Ability-to-Repay and Qualified Mortgage Rule Assessment Report

BUREAU OF CONSUMER FINANCIAL PROTECTION | JANUARY 2019

Ability

-to-Repay and

Qualified Mortgage Rule

Assessment Report

1 BUREAU OF CONSUMER FINANCIAL PROTECTION

Message from

Kathleen L. Kraninger

Director

The Bureau of Consumer Financial Protection is pleased to publish this report containing the

results of its assessment of the Ability-to-Repay and Qualified Mortgage Rule that the Bureau

issued in 2013 to implement provisions of the Dodd-Frank Act amending the T ruth in Lending

Act. The provisions were designed to assure that consumers are offered and receive residential

mortgage loans on terms that reasonably reflect their ability to repay the loans.

Separately, section 1022(d) of the Dodd-Frank Act requires the Bureau to conduct an

assessment of each significant rule or order adopted by the Bureau under Federal consumer

financial law. This report has been prepared to satisfy that statutory obligation.

This somewhat unique statutory requirement places a responsibility on the Bureau to take a

hard look at each significant rule it issues and evaluate whether the rule is effective in achieving

its intended objectives, and the purposes and objectives of Title X of the Dodd-Frank Act, or

whether it is having unintended consequences. I see this as a valuable opportunity to assure that

public policy is being pursued in an efficient and effective manner and to facilitate making

evidence-based decisions in the future on whether changes are needed.

The Bureau’s Office of Research took the lead in conducting this assessment. The Bureau’s

researchers began work over two years ago in identifying the questions that needed to be asked

and in exploring the available data sources to answer those questions. The researchers then

developed research plans and solicited public comment on such plans and other information.

The researchers determined that although public and commercially-available data, along with

the National Mortgage Database which the Bureau has developed in collaboration with the

Federal Housing Finance Agency, could be used to examine the effects of the rule on the market

as a whole, those data were insufficient to examine specific market segments where the rule

might have had its largest effect. Accordingly, the Bureau obtained, among other things, a

unique dataset comprised of de-identified, loan-level data from a number of creditors to fill this

gap. The Bureau’s researchers supplemented those data with a survey to which over 175 lenders

2 BUREAU OF CONSUMER FINANCIAL PROTECTION

responded along with data from a survey conducted by the Conference of State Bank

Supervisors.

Through rigorous statistical analyses of the quantitative data and a careful review of the

qualitative data and public comments received in response to the Bureau’s Request For

Information, the Bureau has produced this comprehensive assessment report. I am confident

that this report provides numerous useful findings and insights for stakeholders, policy makers,

and the general public about developments in the origination of mortgages and the effects of the

rule on the availability and cost of credit.

The issuance of this report is not the end of the line for the Bureau. I am committed to assuring

that the Bureau uses lessons drawn from the assessments to inform the Bureau’s approach to

future rulemakings. We are interested in hearing reactions from stakeholders to the report’s

methodology, findings and conclusions. The Bureau anticipates that continued interaction with

and receipt of information from stakeholders about this report will help inform the Bureau’s

future assessments as well as its future policy decisions regarding this rule.

Sincerely,

Kathleen L. Kraninger

3 BUREAU OF CONSUMER FINANCIAL PROTECTION

Table of Contents

Message from Kathleen L. Kraninger ............................................................................1

Table of Contents ..............................................................................................................3

Executive Summary .........................................................................................................6

1. Introduction ..............................................................................................................15

1.1 Purpose and scope of the assessment ..................................................... 19

1.2 Methodology and plan for assessing effectiveness ................................. 29

1.3 Sources of information and data ............................................................. 32

2. The ATR/QM Rule ....................................................................................................36

2.1 Statutory background .............................................................................. 36

2.2 ATR/QM Rule background ..................................................................... 38

2.3 Overview of ATR/QM Rule requirements .............................................. 42

3. Market overview .......................................................................................................49

3.1 The development of the modern mortgage market ................................ 50

3.2 Early 2000s mortgage market expansion............................................... 56

3.3 Financial crisis and Great Recession: 2007-2009 ................................. 63

3.4 Pre-Rule economic recovery: 2009-2013 ............................................... 67

3.5 Mortgage market pre- and post-Rule...................................................... 69

3.6 Compliance with the Rule ....................................................................... 79

4. Assuring the ability to repay..................................................................................82

4 BUREAU OF CONSUMER FINANCIAL PROTECTION

4.1 Ability to repay and loan performance ................................................... 83

4.2 Loans with restricted features ................................................................. 84

4.3 Historical trends in DTI and relationship with loan performance ........ 96

4.4 Effects of the General QM DTI limit on loan performance.................. 106

5. Effects of the Rule on access to mortgage credit and cost of

credit ........................................................................................................................116

5.1 Market trends in origination of loans with DTI greater than 43 percent119

5.2 Evidence from the lender survey........................................................... 123

5.3 Effect of the Rule on access to credit for borrowers with DTI greater

than 43 percent: evidence from the Application Data ..........................131

5.4 Effects of the points and fees requirement on the availability of small

dollar loans and cost of credit ............................................................... 163

5.5 The rebuttable presumption provision ................................................. 180

6. The Temporary GSE QM .......................................................................................188

6.1 Background ............................................................................................ 189

6.2 Conforming originations since the implementation of the Rule ..........191

6.3 Functional features of the Temporary GSE QM requirements ............ 192

6.4 Meeting the goals of the QM requirements ..........................................204

7. Analysis of the small creditor QM category ......................................................207

7.1 Background ............................................................................................ 208

7.2 Analysis using HMDA data ................................................................... 210

7.3 Evidence from CSBS survey data .......................................................... 223

8. Additional effects of the Rule ..............................................................................229

8.1 Effect on closing times........................................................................... 229

8.2 Survey evidence ..................................................................................... 231

Appendix A: The Rule and Bureau purposes and objectives ...........................235

5 BUREAU OF CONSUMER FINANCIAL PROTECTION

Introduction .................................................................................................... 235

Purposes .......................................................................................................... 235

Objectives ........................................................................................................ 238

Appendix B: Comment summaries .......................................................................242

Evidence about ATR/QM Rule effects ........................................................... 243

Recommendations to modify, expand, or eliminate the ATR/QM Rule ...... 254

The assessment plan ....................................................................................... 263

Appendix C: Application data request to nine lenders ....................................267

6 BUREAU OF CONSUMER FINANCIAL PROTECTION

Executive Summary

The mortgage market has been a key to homeownership for an increasing number of American

families since the middle of the 20th century.

T he Dodd-Frank Wall Street Reform and Consumer Protection Act (Dodd-Frank Act) amended

the T ruth in Lending Act (T ILA) in 2010 to place certain new obligations on the origination of

consumer mortgages. T he Dodd-Frank Act also directed the Bureau of Consumer Financial

Protection (Bureau) to issue rules to effectuate c ertain amendments and authorized the Bureau

to prescribe rules as may be necessary or appropriate to enable the Bureau to administer and

carry out the purposes and objectives of T ILA. The Bureau’s initial rule and certain changes to

that rule, which this report refers to collectively as the Ability-to-Repay/Qualified Mortgage

(AT R/QM) Rule or Rule, came into effect in January 2014.

Section 1022(d) of the Dodd-Frank Act requires the Bureau to conduct an assessment of each of

its significant rules and orders and to publish a report of each assessment within five years of

the effective date of the rule or order. The Bureau has determined that the ATR/QM Rule is a

significant rule. The Bureau developed plans for assessments in 2015 and began work on the

ATR/QM Rule assessment in 2016. Pursuant to decisions made at that time, although this

assessment addresses matters relating to the costs and benefits of the Rule, this report does not

include a benefit-cost analysis of the Rule or parts of the Rule. For Section 1022(d) assessments

that the Bureau undertakes going forward, the Bureau in its discretion is reconsidering whether

to include benefit-cost analysis in its assessment and its published report. The Bureau expects

that this report will inform the public about the effects of the Rule and will help inform the

Bureau’s future policy decisions concerning mortgage originations including whether to

commence a rulemaking proceeding to make the Rule more effective in protecting consumers,

less burdensome to industry, or both.

The key requirement of the Rule is that lenders must make a reasonable and good faith

determination, based on verified and documented information, that the consumer has a

For a full definition of t he ATR/QM Rule, see Chapters 1 and 2.

7 BUREAU OF CONSUMER FINANCIAL PROTECTION

reasonable ability to repay (ATR) before issuing a residential mortgage loan. The Rule defines

certain factors that a lender must consider in making such a determination and requires that the

determination must be made using a payment schedule that fully amortizes the loan over the

term of the loan. Lenders who are found to be non-compliant with this requirement can be held

liable for damages under TILA. In addition, non-compliance can be asserted as a matter of

defense by recoupment or setoff in a foreclosure proceeding.

The Rule also defines the category of Qualified Mortgage (QM) loans and provides that QM

loans are presumed to comply with the ATR requirement. In most cases, the presumption is

conclusive (i.e. a safe harbor). However, for “high cost” loans—a term whose definition largely

tracks one developed by the Federal Reserve Board as a proxy for subprime loans—the

presumption is rebuttable, allowing the consumer the opportunity to prove that the lender in

fact failed to make a reasonable determination of the consumer’s repayment ability.

All QM loans must be fully amortizing loans with terms no greater than thirty years and (except

for loans under $100,000) cannot have the sum of points and fees exceed 3 percent of the loan

amount. Additionally, to meet the Rule’s General QM test, the ratio of monthly debt obligations

to income cannot exceed 43 percent (“debt to income ratio,” or “DTI”). For this test, DTI must

be calculated in accordance with the provisions of the Rule’s Appendix Q which incorporates the

FHA underwriting standards from 2013 for calculating debt and income. The Rule creates a

temporary category under which loans eligible for purchase or guarantee by Fannie Mae or

Freddie Mac (the Government Sponsored Enterprises, or GSEs) generally qualify as QM loan.

This exception (the Temporary GSE QM) is scheduled to expire seven years after the effective

date of the Rule (or earlier if the GSEs cease to be in conservatorship). In addition, mortgages

eligible for purchase or guarantee by the Federal Housing Authority (FHA), Veterans

Administration (VA), or Rural Housing Service (RHS) are QMs by virtue of separate regulations

issued by those agencies pursuant to separate Title XIV rulemaking authority under the DFA.

Key Findings

The collapse of the housing market in 2008 sparked the most severe recession in the United

States since the Great Depression. As documented in this report, the years prior to the collapse

were marked by an increased share of lending going to borrowers of lower creditworthiness and

to new loan product types associated with higher risk.

8 BUREAU OF CONSUMER FINANCIAL PROTECTION

A number of different theories have been advanced as to why the housing market collapsed. In

the report that the National Commission on the Causes of the Financial and Economic Crisis in

the United States

issued after the Dodd-Frank Act was enacted the majority pointed to

“dramatic failures of corporate governance and risk management,” “excessive borrowing, risky

investments and lack of transparency,” and “widespread failures in financial regulation and

supervision” as key causes.

A minority report by one commissioner concluded that

“government housing policy” was responsible for “fostering the growth of a bubble of

unprecedented size and an equally unprecedented number of weak and high-risk residential

mortgages.”

A separate minority report by three other commissioners disagreed with both

views and identified ten causes, some global and some domestic, as essential to explaining the

financial and economic crisis.

Since the issuance of the Commission’s report there has been a

vast body of academic literature seeking to explore the contributing causes of the crisis.

It is

beyond the scope of this report to address the question of what caused the housing market to

collapse a decade ago.

The provisions of the Dodd-Frank Act described above and that are the subject of this report

were enacted for the stated purpose of “assur[ing] that consumers are offered and receive

residential mortgage loans on terms that reasonably reflect their ability to repay.”

This report

The Commission was created in May, 2009 by P.L. 111-21 and issued its r eport i n January, 2011. Six m embers of the

Com m ission w ere appointed by the Dem ocratic l eadership of C ongress and four by the Republican leadership. The

m a jority r eported was joined by the six m embers a ppointed by the Dem ocratic l eadership; three m embers

a ppoi nted by the Repub lican l eadership joined on e di ssent and the fourth authored a separate dissent.

T h e Financial C risis In quiry Report at xviii - x x ii (2 011) .

Id. at 444.

Id. a t 445-448.

For a n on-exhaustive list of additional literature on the causes of the crisis, see Adelino, Manuel, Antoinette Schoar,

a n d Felipe S everino. "Loa n originations a nd defaults in the mortgage crisis: The r ole of t he m iddle c lass." T he

Rev iew of Financial Studies 2 9.7 (2016); Amrom in, G ene, et a l. "C om plex m ortgages." Rev iew of Finance 2 2.6

(2 018); Avery, Rob ert B. , and Kenneth P. Br evoort. "The subprim e c risis: Is g ov ernment housing policy to b lame?"

Rev iew of Economics and Statistics 97.2 (2015); Bubb, Ryan, and Alex Kaufman. “Securitization and moral hazard:

Ev idence from credit score cutoff rules.” 63, 1-1 8 (Apr. 2014); Case, Karl E., Robert J. Shiller, and Anne Thom pson.

“ W hat have they been thinking? Hom e buyer b ehavior in hot and col d m arkets.” National Bureau of Econom ic

Research Working Paper #18400 (2012); Chen g, In g-Haw, and Sahil Raina, and Wei Xiong. “Wall Street and the

Hou sing Bu bble.” 104(9): 2797-2829 (2014); Corbae, Dean, and Erwan Quintin, "Leverage and the Foreclosure

Cr i si s, " Journal of Politica l Ec on om y 1 23 , n o. 1 ( Fe b. 2 015 ); Foote, Christ opher L., Kri st opher S. Gerardi, and Paul S.

Willen. “Why Did So Ma ny People Ma ke So Many Ex Post Bad Decisions? The Causes of the Foreclosure Crisis.”

Na tional Bureau of Econom ic Research Working Paper #18082 (2012); Lee, Don ghoon, C hristopher May er, and

Joseph Tracy. “A New Look a t Second Liens.” Federal Reserve Bank of New Y ork S taff Reports (2 012); Mian, Atif,

a n d Amir Sufi. "Fr audulent incom e ov erstatem ent on m ortgage applications during the credit expansion of 2002 to

2 0 05." T he Rev iew of Financial Studies 30.6 (2017); Pa lmer, Christopher, “Why Did S o Ma ny Subprim e Borrowers

Defa u lt During the Crisis: Loose Credit or Plummeting Prices?” (Sept. 24, 2015).

TILA § 1 29B(a)(2), 15 USC § 1 639b(a)(2).

9 BUREAU OF CONSUMER FINANCIAL PROTECTION

does not address the necessity of the new TILA requirements or the merits of possible

alternative ways that Congress might have responded to the housing collapse, but rather, as

required by the Dodd-Frank Act, assesses the effectiveness of the ATR-QM Rule that

implemented those requirements.

Assuring Ability to Repay

A primary purpose of the Rule is to prevent the extension of mortgage credit for which

consumers lack the ability to repay, based on information available at the time of origination.

The report finds that approximately 50 to 60 percent of mortgages originated between 2005 and

2007 that experienced foreclosure in the first two years after origination were mortgage loans

with features that the ATR/QM Rule generally eliminates, restricts, or otherwise excludes from

the definition of a qualified mortgage, such as loans that combined low initial monthly payments

with subsequent payment reset or those made with limited or no documentation of the

consumer’s income or assets. Loans with these features had largely disappeared from the market

prior to the adoption of the Rule, and today they appear to be restricted to a limited market of

highly credit-worthy borrowers.

Further, this report finds that loans with higher debt to income ratios—which is a factor

generally required to be considered in making ATR determinations and is one of the criteria

used to define the General QM category—are historically associated with higher levels of

delinquency, after controlling for other relevant borrower characteristics (even though the

strength of the relationship depends on the economic cycle). In the conventional mortgage

market—which encompasses all mortgages other than those purchased or guaranteed by a

government agency—DTI ratios are constrained from returning to crisis-era levels by a

combination of the ATR requirement, GSE underwriting limits which define the loans which are

eligible for purchase by the GSEs (currently, a DTI limit of 45 percent applies to most loans) and

the Bureau’s General QM DTI threshold which limits the General QM category to loans with

DTIs at or below 43 percent. Even though house prices have largely returned to pre-crisis levels,

currently five to eight percent of conventional loans for home purchase have DTI exceeding 45

percent; in contrast, approximately 24 to 25 percent of loans originated in 2005 – 2007

exceeded that ratio.

Even though it is not possible for the Bureau to directly observe the ability to repay at

origination, an analysis of realized loan performance across a wide pool of loans can be

informative. Among other metrics, this report examines the percentage of loans becoming 60 or

more days delinquent within two years of origination. The analysis finds that the introduction of

the Rule was generally not associated with an improvement in loan performance according to this

metric.

10 BUREAU OF CONSUMER FINANCIAL PROTECTION

In part, this is due to the fact that delinquency rates on mortgages originated in the years

immediately prior to the effective date of the Rule were historically low, as credit was already

tight at that time. The delinquency rate of loans with DTIs exceeding 43 percent made under the

Rule’s ATR underwriting requirements (non-QM loans) remained steady at 0.6 percent; the

delinquency rate of GSE loans with DTIs above 43 percent increased from 0.6 percent for loans

originated in 2012-2013 to 1.0 percent among 2014-2015 originations. Thus, although the

performance of non-QM loans did not improve in absolute terms, it has improved relative to the

performance of comparable QM loans. (Chapter 4)

Access to Credit and Restrictions on Unaffordable Loans

Looking at the market as a whole, there was not a significant break in the volume of mortgage

applications or the average approval rate at the time the Rule became effective. This is

attributable in part to the fact that, as noted, following the financial crisis and before the Rule

took effect credit had tightened substantially and in part to the breadth of the definition of QM

and the safe harbor afforded to most QM loans. The Bureau estimates that 97-99 percent of

loans originated in 2013, the last year prior to the effective date of the Rule, would have satisfied

QM requirements. As explained above, if a loan is a QM loan, it is presumed to meet the Rule’s

ATR requirement, and, therefore, the ATR requirement would not separately decrease access to

credit for borrowers who qualify for a QM loan.

There are, however, certain segments of the market where the Rule is more likely to have

affected access to credit and the Report focuses on those segments.

• Borrowers with high debt to income ratios – For high DTI borrowers—defined

here to mean borrowers with a DTI above 43 percent—who qualify for a loan eligible for

purchase or guarantee by one of the GSEs (or one of the federal agencies), the Rule has

not decreased access to credit since such mortgages meet the standard for QM loans. In

fact, the evidence suggests that the GSEs may have loosened their underwriting

requirements for high DTI borrowers, as evidenced by recent trends (Chapter 6)

There is a segment of high DTI borrowers seeking loans which are not eligible for GSE

purchase (or a government purchase or guarantee), most commonly due to loan size.

Generally, such loans are non-QM loans as they cannot meet the General QM standard

and thus are subject to the ATR provisions. The Bureau’s analysis of detailed application

data from nine larger lenders (further, “Application Data”) indicates that the Rule

displaced between 63 and 70 percent of approved applications for home purchase among

non-QM High DTI borrowers during the period of 2014 – 2016; this translates into a

11 BUREAU OF CONSUMER FINANCIAL PROTECTION

reduction of between 1.5 and 2.0 percent of all loans for home purchase made by these

nine lenders during this period. Evidence from other data sources, including a survey of

mortgage lenders that the Bureau conducted as part of this assessment (further, Lender

Survey) and recent research by the Federal Reserve Board and academic economists

likewise points to sharp reductions in access to credit among this category of borrowers

following the implementation of the Rule. Notably, results in the refinance category are

quite different. For non-QM, High DTI borrowers seeking to refinance their loans, the

Application Data points to an initial reduction in access to credit in 2014, followed by

gradual improvement in the years after. T his is consistent with a notion that consumers

seeking to refinance a mortgage having already demonstrated some ability to repay,

thereby lowering ATR risk and making lenders more likely to extend credit. (Chapter 5)

The Application Data also indicates that among the non-QM High DTI borrowers

seeking to purchase homes, approval rates declined across all credit tiers and income

groupings, with the result that the average credit score and income for declined

applicants increased after the Rule took effect. Further, more broad-based industry data

indicates that despite tightening of credit, delinquency rates for non-QM High DT I

borrowers did not decrease after the Rule took effect. Together, these findings suggest

that the observed decrease in access to credit in this segment was likely driven by

lenders’ desire to avoid the risk of litigation by consumers asserting a violation of the

ATR requirement or other obligations or risks associated with that requirement, rather

than by rejections of borrowers who were unlikely to repay the loan. (Chapter 5)

• Self-employed borrowers – As with high DTI borrowers, the Rule did not impact

access to credit for self-employed borrowers seeking a mortgage which is eligible for

purchase or guarantee by one of the GSEs or federal agencies. In contrast, self-employed

borrowers who do not qualify for a loan eligible for purchase or guarantee by one of the

GSEs or federal agencies generally need to qualify under the General QM standard in

order to obtain a QM loan. Responses to the Lender Survey indicate that specifically for

self-employed borrowers, lenders may find it difficult to comply with Appendix Q

relating to the documentation and calculation of income and debt. However, the

Application Data indicates that the approval rates for non-High DT I, non-GSE eligible

self-employed borrowers have decreased only slightly, by two percentage points.

(Chapter 5)

• Borrowers seeking smaller loan amounts: The points-and-fees cap on QM loans

has potential implications for borrowers seeking smaller loan amounts because, to the

extent there are fixed costs in originating mortgages, those costs will constitute a higher

percentage of the loan amount for smaller loans relative to larger loans. The Bureau’s

12 BUREAU OF CONSUMER FINANCIAL PROTECTION

analysis of HMDA data indicates, however, that the Rule likely had no effect on access to

credit for such loans. This is consistent with responses to Lender Survey which indicate

that applications for which the points and fees limit will be exceeded are sufficiently rare

that lenders handle them on a case-by-case basis. Specifically, lenders typically waive

certain fees, with or without a compensating increase in the interest rate, to avoid

exceeding the cap. Lenders denying an application to avoid exceeding the QM points and

fee cap is rare. (Chapter 5)

Creditor Costs and the Cost of Credit

The Rule introduces certain requirements for documenting income and debt that may differ

from the pre-Rule practices for some lenders. For non-QM loans (as well as high cost QMs), the

Rule also creates potential liability for ATR violations. Furthermore, under a separate rule

administered by other agencies, holders of non-QM loans are required to hold extra capital

against such loans which can add to the cost of funding these loans. The Report examines the

effect of the Rule on lenders’ costs of originating loans and on the prices they charged to

consumers.

At the aggregate market level, the Rule does not appear to have materially increased costs or

prices. A periodic survey conducted by the Mortgage Bankers Association among non-bank

lenders indicates that the costs of originating mortgage loans have increased over the past

decade but that there was not a distinct increase around the time of the implementation of the

Rule. Similarly, the Bureau’s analysis indicates that the spread between the average interest rate

on 30-y ear fixed-rate mortgages over the relevant Treasury rate has remained constant since the

implementation of the Rule. (Chapter 3)

The Bureau would not be able to reasonably obtain evidence that directly measures the extra

cost of originating a loan that may have been created by the Rule. Instead, the Bureau has

obtained qualitative feedback through responses to the Lender Survey, regarding material

changes in credit policy that have occurred. A majority of respondents indicated that their

business model has changed as a result of the Rule. Among those respondents who reported

changed business model, some respondents pointed to increased income documentation or

increased staffing, while others mentioned adopting a policy of not originating non-QM loans.

The Bureau has utilized the Application Data to quantify the cost in the form of foregone profits

from not originating certain non-QM loans; it is found that among the nine lenders that

provided the data, the lost profits amounted to between $20 and 26 million per year.

Focusing specifically on non-QM loans, evidence is mixed as to whether the Rule has increased

the price of such loans. None of the nine lenders that provided Application Data charge extra for

13 BUREAU OF CONSUMER FINANCIAL PROTECTION

non-QM loans specifically and a review of retail rate sheets of approximately 40 lenders revealed

that an extra adjustment for non-QM loans is very infrequent. Nevertheless, 23 out of 204

respondents to the Lender Survey that the Bureau conducted for this assessment indicated

applying such an increase and recent research by the Federal Reserve Board finds that loans

with DT Is above 43 percent are substantially more expensive than similar loans with DTIs at or

below 43 percent. (Chapter 5)

Effects on Market Structure

To a large extent, the current QM category is broad due to the inclusion of loans eligible for

purchase by the GSEs. The inclusion of such loans in the QM category is temporary and is set to

expire by January 2021. Contrary to the Bureau’s expectations at the time of the rulemaking, the

GSEs have maintained a persistently high share of the market in the years following the Rule’s

effective date. The private label mortgage-backed securities market, where investors purchase

loans that are not insured or guaranteed by GSEs or government agencies, remains small

relative to GSE securitizations and primarily provides funding for QM loans made to prime

jumbo borrowers (although recently there has been a number of non-QM securitizations based

on loans made to other types of borrowers). The dominance of GSEs in the conventional loan

segment may be attributable to a range of factors which distinguish GSE loans from those made

under the General QM and ATR criteria, potential advantages in compliance certainty and

flexibility, and robust secondary market liquidity. (Chapter 6)

The Bureau has examined whether the Temporary GSE QM provision of the Rule has caused in

an increased reliance on GSEs’ Automated Underwriting Systems (AUSs) for loans that are not

sold to the GSEs. T he analysis of submissions to AUSs shows no immediate increase in the

aggregate volume of submissions relative to the volume of loans purchased by GSEs. However,

the data do suggest a somewhat higher use of the GSEs’ AUS in recent years, particularly for

loans which do not fit within or are more difficult to document within the General QM

underwriting standards, such as loans made to self-employed borrowers. (Chapter 6)

The Rule contains certain provisions for smaller lenders that allow them to originate High DT I

loans, and in some cases, balloon loans as long as such loans are held on portfolio for at least

two years after the origination by small creditors (Small Creditor QM and Small Creditor

Balloon QM, respectively). Among HMDA reporting depository institutions involved in

mortgage lending in 2016, approximately 90 percent meet the definition of Small Creditor and

these institutions account for about 24 percent of mortgage loans. T he Rule does not appear to

be constraining the activities of these lenders since virtually all fall well below the threshold that

defines Small Creditor. There are systematic differences in the loans made by Small Creditors

14 BUREAU OF CONSUMER FINANCIAL PROTECTION

and non-Small Creditors. The former hold a larger share of their originations on portfolio,

although there was a noticeable decline in the share of portfolio loans made by small creditors in

2016 which coincided with an expansion in the definition of small creditor. Similarly, a larger

share of small creditor mortgages are made in rural counties or to finance manufactured

housing mortgages. Small creditors responding to a survey conducted by the Conference of State

Banking Supervisors (CSBS) in 2015 reported that a larger share of their portfolio loans were

non-QM loans than was true for the larger lenders who responded to the survey, and also

reported declining a smaller percentage of applications than larger creditors. To the extent small

creditors declined applications, these creditors were less likely than larger creditors to attribute

their denial to the requirements of the Rule than larger creditors. (Chapter 7)

15 BUREAU OF CONSUMER FINANCIAL PROTECTION

1. Introduction

The mortgage market is the single largest market for consumer financial products and services

in the United States, with approximately $10.7 trillion in consumer mortgage loans

outstanding.

During the first decade of the 21st century, this market went through an

unprecedented cycle of expansion and contraction. When the housing market collapsed in 2008,

it sparked the most severe recession in the United States since the Great Depression.

An early warning sign of the approaching mortgage crisis was an upswing in early payment

delinquencies and defaults.

For mortgage originations between 2000 and 2004, 1.7 percent

would become 60 or more days delinquent within the first year.

For the 2006 vintage, the

figure was 5.4 percent. Expanding to the first two years of repayment, the growth in

delinquencies was more severe. For mortgages made in 2005, 2006, and 2007, 6.0 percent, 1 3.0

percent, and 14.4 percent became 60 or more days delinquent within the first two years,

respectively. These rates are substantially above the average between 2000 and 2004 of 3.6

percent. As the economy worsened, the share of loans with serious delinquencies (90 or more

days past due or in foreclosure) grew further. For loans with atypical features that became

common during the mid-2000s, the rates of serious delinquency were particularly high. By the

end of 2010, among loans originated from 2005 to 2007, 35.5 percent of short-reset adjustable-

rate mortgages, 29.7 percent of interest only loans, and 27.1 percent of loans with limited or no

documentation were or had been seriously delinquent. Some of those delinquencies may have

resulted from an unanticipated deterioration in the borrowers’ economic situation after the

loans were originated. But the high rate of early delinquencies suggests that for some portion of

Fed. Reserve Sy stem, Mor tgage Debt Outstanding,

h t tps://w ww .federalreserve.gov/data/m ortoutstand/current.htm (last visited Dec. 1 3, 2018).

See T h om as F . S iem s, Branding the Great Recession, at 3, Fin. In sights (Fed. Reserve Bank of Dall., May 2012),

available at https://www.dallasfed.org/~/media/documents/outreach/fi/2012/fi1201.pdf

(stating that the [great recession]

‘‘was the longest and deepest economic contraction, as measured by the drop in

real GDP, since the Great Depression.’’).

Ea r ly payment defaults are g enerally defined a s borrowers being 60 or more days delinquent within the first year.

How ever, where noted, this discussion also uses a more expansive definition of early payment default to include 60

days delinquent within the first two y ears.

All statistics in this paragraph are Bureau calculations using the National Mortgage Database.

16 BUREAU OF CONSUMER FINANCIAL PROTECTION

the borrowers, the loans may have been beyond their ability to repay, either from the start or

shortly thereafter.

The impact of these high rates of delinquency and default was severe on consumers and

communities,

on creditors

who held loans on their books, and on private investors who

purchased loans directly or indirectly through certain types of securitizations.

Because the risk

from these products was spread throughout the financial system,

a severe credit shock

disrupted the American economy. The Federal National Mortgage Association (Fannie Mae) and

the Federal Home Loan Mortgage Corporation (Freddie Mac), which supported the mainstream

mortgage market, experienced heavy losses and were placed in conservatorship by the federal

government in 2008 to support the collapsing mortgage market.

House prices, which had risen

27 percent nationally between 2003 and 2007,

fell an average of 33 percent from their peak in

2006,

and delinquency and foreclosure rates remained elevated

for several years.

See 7 8 Fed. Reg. 6408, 6559–6560 (Jan. 30, 2013).

T h e term “ creditor” and “l ender” are used interchangeably in this report.

“ A larmed by the unexpected delinquencies and defaults that began t o appear in mi d-2 007, i nve stors f l ed t he m ulti-

t r illion dollar market for m ortgage-backed securities (MBS ), dropping MBS v alues—and especially those MBS

ba cked by subprime and other r isky l oans—to fractions of their former prices. Mark-to-m arket accounting then

r equ ired financial institutions t o w rite down t he value of t heir assets—r educing their capital positions and causing

g r eat investor and creditor unease.” U.S. Fin. Crisis In quiry Com m’n, The Financial Cris is Inquiry Report: Final

Report of the National Commission on the Causes of the Financ ial and Economic Cris is in the United States, at

444–445 (Official Gov’t ed. 2011) (FCIC Report), available at https://www.gpo.gov /fdsy s/pkg/GPO-

FCIC/pdf/GPO-FCIC.pdf.

For example, such securities were used as collateral for borrowing. See id. a t 43.

T h e Housing and Econom ic Rec overy Act of 2008 (HERA), which created the Federal H ousing Financ e Agency

(FHFA), granted the Director of FHFA discretionary authority to a ppoint FHFA conservator or receiver of the

En terprises ‘‘for the purpose of r eorganizing, rehabilitating, or winding up the affairs of a regulated entity.’’ Housing

and Economic Recovery Act of 2008, section 1367 (a)(2), amending the Federal Housing Enterprises Financial

Safety a nd Soundness A ct of 1 992, 12 U.S.C. § 4 617(a)(2). On September 6, 2008, FHFA exercised that authority,

pla cing Fannie Ma e a nd Freddie Ma c into conservatorships. Fed. Hous. Fin. Agency, Conservator’s Report on the

Enterpris es’ Financial Performance, at 17 (2 nd Quarter 2012), available at

h t tps://www.fhfa.gov /webfiles/24549/ConservatorsReport2Q2012.pdf

FCIC Report, s upra note 14, at 156.

Fed. Reserve Sy stem, The U. S. Ho using Market: Current Conditions and Po licy Co nsiderations, at 3 ( Fed. Reserve

Bd., White Paper, 2012), available at https://www. federalreserve.gov/publications/other-reports/files/housing-

w h ite-paper-2 012 0104 .pdf.

See Lender Processing Servs., LPS Mortgage Monitor: May 2012 Mo rtgage Performance Observ ations, Data as of

April 2012 Month End, at slide 3, 11 (May 2012), available at

h ttp://www.bkfs.com /CorporateInformation/NewsRoom/MortgageMonitor/201204MortgageMon itor/MortgageM

on i t orApril2 012.pdf.

17 BUREAU OF CONSUMER FINANCIAL PROTECTION

In response to the crisis, Congress passed the Dodd-Frank Wall Street Reform and Consumer

Protection Act (Dodd-Frank Act), which was signed into law on July 21, 2010.

In the Dodd-

Frank Act, Congress included a significant number of new provisions governing the origination

of consumer mortgages. In particular, sections 1411 and 1412 of the Dodd-Frank Act amended

the T ruth In Lending Act (TILA) by adding sections 129C(a) and (b).

These amendments to

T ILA generally provide that no creditor may make a residential mortgage loan unless the

creditor makes a reasonable and good faith determination based on verified and documented

information that, at the time the loan is consummated, the consumer has a reasonable ability to

repay the loan.

The amendments to TILA also establish a presumption of compliance with the

ability-to-repay requirement for creditors originating a category of loan called a “qualified

mortgage” (QM).

Congress directed the Bureau of Consumer Financial Protection (Bureau) to

issue rules to effectuate certain of these amendments

and authorized the Bureau to prescribe

regulations revising, adding to, or subtracting from the criteria that define a qualified

mortgage.

In January 2013, to implement sections 1411 and 1412 of the Dodd-Frank Act, the Bureau

published a final rule titled ‘‘Ability-to-Repay and Qualified Mortgage Standards Under the

Truth in Lending Act (Regulation Z)’’ (January 2013 Rule).

The Bureau amended the January

2013 Rule several times both before it took effect on January 10, 2014 and afterwards. For

purposes of determining whether the January 2013 Rule was significant under section 1022(d)

of the Dodd-Frank Act, the Bureau made its determination based on the January 2013 Rule and

amendments to it that took effect on January 10, 2014.

However, in order to facilitate a clearer

and more meaningful assessment, the assessment and this report take into consideration certain

Pu b. L. No. 111–203, 124 Stat. 1376 (2010).

T ILA section 1 29C(a)–(b) (c odified a s am ended at 1 5 U.S.C. § 1 639c(a)–(b)).

TILA section 129C(a) (codified as amended at 15 U.S.C. § 1639c(a)).

T ILA section 1 29C(b) (c odified a s amended a t 15 U.S.C. § 1 639c(b) ).

Con gress g enerally c onsolidated i n the Bu reau the rulemaking authori ty for Federal consumer financial l aws

pr ev iously vested in certain other federal agencies. Congress a lso provided the Bureau wi th the authority t o, among

oth er things, prescribe rules as may be necessary or appropriate to enable the Bureau to administer and carry out

t h e purposes and objectives of t he Federal consumer financial l aws and to prevent ev asions thereof. 1 2 U. S.C. §

5 5 12(b)(1). The Federal consumer financial laws include TILA.

TILA section 1 29C(b)(3)(B)(i) (codified a s amended a t 15 U.S.C. § 1639c(b)(3)(B)(i)).

7 8 Fed. Reg. 6408 (Jan. 30, 2013).

See Section 1 .1.2, at n. [41].

18 BUREAU OF CONSUMER FINANCIAL PROTECTION

amendments affecting small creditors that took effect in 2016.

Therefore, the term “ATR/QM

Rule” (or “Rule”) generally refers throughout this report to ability-to-repay and qualified

mortgage requirements in effect as of January 2014; except that for certain analyses of small

creditors, the Rule includes requirements on small creditors in effect as of March 2016, as

indicated.

The ATR/QM Rule, among other things, describes certain minimum requirements for creditors

making ability-to-repay determinations, but does not dictate that they follow particular

underwriting standards. Creditors generally must consider certain specified underwriting

factors and use reasonably reliable third-party records to verify the information on which they

rely to determine repayment ability.

The ATR/QM Rule also defines several categories of QM

loans for which, as noted above, compliance with the ATR requirement is presumed.

The

presumption of compliance can be either conclusive (e.g., a safe harbor) for QM loans that are

not ‘‘higher-priced’’, or rebuttable, for QM loans that are ‘‘higher-priced.’’

Section 1022(d) of the Dodd-Frank Act requires the Bureau to conduct an assessment of each

significant rule or order adopted by the Bureau under Federal consumer financial law.

discussed further below, the Bureau has determined that, for purposes of section 1022(d), the

January 2013 Rule and amendments to it that took effect on January 10, 2014 is a significant

rule. Another requirement of section 1022(d) is that the Bureau publish a report of the

assessment within five years of the effective date of the significant rule or order. This document

is the report of the Bureau’s assessment of the ATR/QM Rule in accordance with section

1022(d).

In June 2017, the Bureau published a Request for Information (RFI) requesting public comment

on its plans for assessing the Rule, and requesting certain recommendations and information

See 80 Fed. Reg. 59943 (Oct. 2, 2015) (among other things, this rule increased the mortgage originations threshold

for sm all creditors and expanded t he definition of “ rural area.”) See als o 81 Fed. Reg. 1 6074 (Mar. 25, 2016) ( this

r u le remov ed “predom inantly” as a qualifier of t he “operates in rural or underserved areas” r equirement).

1 2 C.F.R. § 1 026.43(c)(2)–(4). The eight factors that must be considered in an ATR determination are listed in

Section 2.3.2, below.

1 2 C.F.R. § 1 026.43(e)–(f).

1 2 C.F.R. § 1 026.43(e)(1). A “higher-priced covered transaction” is defined at 12 C.F.R. § 1 026.43(b)(4) as “a

cov ered transaction with a n annual percentage rate that exceeds the average prime offer rate for a com parable

transaction as of the date the interest rate is set by 1.5 or more percentage points for a first-lien cov ered transaction,

other than a qualified mortgage under paragraph (e)(5), (e)(6), or (f) of this section; by 3.5 or more percentage

poin ts for a first-lien c ov ered transaction that is a qualified m ortgage under paragraph (e)(5), (e)(6), or (f) of this

section; or by 3.5 or more percentage points for a subordinate-lien c ov ered transaction.”

1 2 U.S.C. § 5512(d).

19 BUREAU OF CONSUMER FINANCIAL PROTECTION

useful in conducting the assessment.

The Bureau received approximately 480 comments in

response to the RFI. The Bureau considered data and other relevant information provided by

commenters, as well as comments on the assessment plan, as it conducted the assessment and

prepared this report.

This report does not generally consider the potential effectiveness of alternative requirements

on the origination of consumer mortgages that might have been or might be adopted, nor does it

include specific proposals by the Bureau to modify any rules. The Bureau expects that the

assessment findings made in this report and the public comments received in response to the

RFI will help inform the Bureau’s future policy decisions concerning consumer mortgages,

including whether to commence a rulemaking proceeding to make the ATR/QM Rule more

effective in protecting consumers, less burdensome to industry, or both. In future policy

development, the Bureau expects to consider other public comments, including comments

received in 2018 in response to a series of requests for information about Bureau activities.

Those comments are not summarized in this report.

Finally, the Bureau’s assessments pursuant to section 1022(d) of the Dodd-Frank Act are not

part of any formal or informal rulemaking proceedings under the Administrative Procedure Act.

This report does not represent legal interpretation, guidance, or advice of the Bureau and does

not itself establish any binding obligations. Only the rules and their official interpretations

(commentary) establish the definitive requirements.

1.1 Purpose and scope of the assessment

1.1.1 Statutory requirement for assessments

Section 1022(d) of the Dodd-Frank Act requires the Bureau to conduct an assessment of each

significant rule or order adopted by the Bureau under Federal consumer financial law.

The

Requ est for In formation Regarding Ability-to-Repay/Qualified Mor tgage Rule Assessm ent, 82 Fed. Reg. 25246

(Ju ne 1, 2017).

Sum maries of t he different types of com ments received in response to the RFI a re included in Appendix B t o t his

r eport. See also Section 1.2 below.

Requ est for In formation Regarding the Bu reau’s A dopted Regulations and New Rulemaking Authorities, 83 Fed.

Reg . 1 2286 (Mar. 21, 2018).

1 2 U.S.C. § 5512(d).

20 BUREAU OF CONSUMER FINANCIAL PROTECTION

Bureau must publish a report of the assessment not later than five years after the effective date

of such rule or order. The assessment must address, among other relevant factors, the rule’s

effectiveness in meeting the purposes and objectives of title X of the Dodd-Frank Act and the

specific goals stated by the Bureau.

The assessment must reflect available evidence and any

data that the Bureau reasonably may collect. Before publishing a report of its assessment, the

Bureau must invite public comment on recommendations for modifying, expanding, or

eliminating the significant rule or order.

The purposes and objectives of title X of the Dodd-Frank Act are set out in section 1021 of the

Dodd-Frank Act. Pursuant to section 1021(a) of the Dodd-Frank Act, the purpose of the Bureau

is to implement and, where applicable, enforce Federal consumer financial law consistently for

the purpose of ensuring that all consumers have access to markets for consumer financial

products and services and that markets for consumer financial products and services are fair,

transparent, and competitive.

The objectives of the Bureau are listed in section 1021(b) of the

Dodd-Frank Act. Specifically, section 1021(b) provides that the Bureau is authorized to exercise

its authorities under Federal consumer financial law for the purposes of ensuring that with

respect to consumer financial products and services:

1. Consumers are provided with timely and understandable information to make

responsible decisions about financial transactions;

2. Consumers are protected from unfair, deceptive, or abusive acts and practices and from

discrimination;

3. Outdated, unnecessary, or unduly burdensome regulations are regularly identified and

addressed in order to reduce unwarranted regulatory burdens;

4. Federal consumer financial law is enforced consistently, without regard to the status of a

person as a depository institution, in order to promote fair competition; and

5. Markets for consumer financial products and services operate transparently and

efficiently to facilitate access and innovation.

The specific goals of the ATR/QM Rule are discussed below in Section 1.1.3.

1 2 U.S.C. § 5511(a).

1 2 U.S.C. § 5511(b).

21 BUREAU OF CONSUMER FINANCIAL PROTECTION

1.1.2 Overview of the ATR/QM Rule

T he Dodd-Frank Act amended T ILA to provide that no creditor may make a residential

mortgage loan unless the creditor makes a reasonable and good faith determination based on

verified and documented information that, at the time the loan is consummated, the consumer

has a reasonable ability to repay the loan. The amendments to TILA also establish a

presumption of compliance with the ability-to-repay requirement for creditors originating a

qualified mortgage.

As noted above, the Bureau first implemented these requirements in the January 2013 Rule.

The Bureau amended the January 2013 Rule several times both before it took effect on January

10, 2014, and afterwards. For purposes of determining whether the January 2013 Rule was

significant under section 1022(d) of the Dodd-Frank Act, the Bureau made its determination

based on the January 2013 Rule and amendments to it that took effect on January 10, 2014.

However, in order to facilitate a clearer and more meaningful assessment, the assessment and

this report take into consideration certain amendments that took effect in 2016. These

amendments revised the definition of a small creditor and preserved the ability of small

creditors to make balloon-payment QMs without regard to whether they operated

predominantly in rural or underserved areas.

Therefore, as stated above, the term “ATR/QM

See 78 Fed. Reg. 6408 (Jan. 30, 2013). The January 2013 Ru le a lso included: (i) special prov isions for creditors

refinancing “non-standard mortgages;” (ii) certain limits on prepayment penalties; (iii) enhanced record retention

requirements; and (iv) anti-evasion provisions. T his Report does n ot discuss these prov isions at length given t heir

m odest im pact on the ov erall effectiveness of t he ATR/Q M rule in m eeting the purposes and obj ectives of t itle x and

t h e goals specified by t he Bureau in the rule. For example, the special prov isions for refinancing “non-standard

m ortgages” prov ide an exception to the ATR r equirement, but a s Chapter 4 points out, these types of loans, (i.e.,

in t erest-only, negative amortization, or A RMs w ith an introductory period of on e year or longer), are quite rare in

the post-Rule period and already made up a sm all share of t he m arket in the years im mediately prior t o the Rule’s

effec tive date. The prepayment penalty and recordkeeping p rovisions are additional standards and requirements

based on other Dodd-Frank Act provisions (sections 1414 and 1416, respectively) not directly r elated to the ATR

det ermination. No c om ments that w ere r eceived on the assessm ent focused on any of t hese prov isions and the

Bu r eau has marshaled its resources to examine prov isions more central to t he ATR determination and the

effec tiveness of t he A TR/QM Ru le.

Wh en the January 2013 Rule was issued, the Bureau concurrently issued a proposal to amend it, and that proposal

w a s finalized on May 29, 2013. See 78 Fed. Reg. 6622 (Jan. 30, 2013) (January 2013 ATR Proposal) ; 78 Fed. Reg.

3 5 430 ( June 12, 2 013) ( May 2013 A TR Rule). The Bureau issued additional corrections and clarifications in the

su mmer and fall of 2 013. See 7 8 Fed. Reg. 44686 (July 24, 2013); 7 8 Fed. Reg. 60382 (Oct. 1 , 2013); 7 8 Fed. Reg.

62993 (Oct. 23, 2 013). Am endments that took effect a fter January 1 0, 2014, are an i nterpretive rule regarding

su ccessors-in-i n ter e st, see 79 Fed. Reg. 41631 (July 17, 2014); a rule related t o nonprofit entities and which also

pr ov ided a cure m echanism for the points and fees limit that a pplies to qualified mortgages, s ee 7 9 Fed. Reg. 65300

(Nov . 3, 2014); r evisions to the definitions of sm all c reditor and rural area, s ee 80 Fed. Reg. 59943 (Oct. 2, 2015); a

procedural rule establishing an application process for designation as a rural area, see 81 Fed. Reg. 11099 (March 3,

2 01 6); and revisions t o the r equirements for QM loans issued by sm all creditors, see 81 Fed. Reg. 16074 (March 25,

2 01 6) .

See 8 0 Fed. Reg. 59943 (Oct. 2, 2 015) (revisions to the definitions of sm all c reditor and rural area); 81 Fed. Reg.

1 6074 (Mar. 25, 2016) (revisions to the requirements for QM loans issued by small creditors).

22 BUREAU OF CONSUMER FINANCIAL PROTECTION

Rule” (or Rule) generally refers throughout this report to ability-to-repay and qualified

mortgage requirements in effect as of January 2014, except for certain analyses of small

creditors, as indicated.

As discussed in greater detail in Chapter 2, the ATR/QM Rule describes certain minimum

requirements for creditors making ability-to-repay determinations. Creditors generally must

consider certain minimum underwriting factors and they generally must use reasonably reliable

third-party records to verify the information they use to determine repayment ability.

The

Dodd-Frank Act attached civil liability to a creditor’s failure to meet the ability-to-repay

requirement.

T he Dodd-Frank Act also established a presumption of compliance with the ability-to-repay

requirement and protection from liability for creditors originating a qualified mortgage. The

Rule defines several categories of qualified mortgages. All categories must meet certain

requirements, which include having terms of 30 years or less, regular periodic payments that are

substantially equal (except in the case of adjustable-rate or step-rate mortgages) that do not

result in the increase of the principal balance, and total points and fees which do not exceed a

certain percentage of the loan amount.

Additional restrictions apply depending on the type of

qualified mortgage.

One category of qualified mortgage is the “General QM.” To fall within this category, the ratio of

the consumer’s total monthly debt payment to total monthly income (DTI) cannot exceed 43

percent and must be calculated using debt and income in accordance with Appendix Q.

The

criteria for General QM further require that creditors calculate mortgage payments based on the

highest payment that will apply in the first five years of the loan.

This category also includes a

restriction on balloon payment features.

A second category of qualified mortgage is the “Temporary GSE QM.” T his is a separate,

temporary category of QM for loans eligible to be purchased or guaranteed by either Fannie Mae

or Freddie Mac (collectively, the Government Sponsored Entities or GSEs) while they operate

under federal conservatorship or receivership. Under the terms of the Rule, the Temporary GSE

1 2 C.F.R. § 1 026.43(c)(2)–(4).

1 2 C .F.R. § 1 026.43(e)(2)(i)–(iii).

Ch apter 2 prov ides a full discussion of t he r equirements for qualified m ortgages. The summary below prov ides

in formation that may be especially useful for understanding the empirical analyses in subsequent chapters.

1 2 C.F.R. § 1 026.43(e)(2)(vi).

1 2 C.F.R. § 1 026.43(e)(2)(iv).

23 BUREAU OF CONSUMER FINANCIAL PROTECTION

QM category will continue to be in effect until the earlier of: (i) the end of conservatorship; or

(ii) January 10, 2021.

The Rule also provided a temporary QM category for loans eligible to be insured by the U.S.

Department of Housing and Urban Development (FHA Loans); guaranteed by the U.S

Department of Veterans Affairs (VA Loans); guaranteed by the U.S. Department of Agriculture

(USDA Loans); or insured by the Rural Housing Service (RHS Loans) (collectively, ‘Temporary

Federal Agency QM’).

The category of Temporary Federal Agency QM no longer exists and has

been replaced by the category of Federal Agency QM because the relevant federal agencies (i.e.,

FHA, VA, and RHS) have all now issued their own qualified mortgage rules.

The Bureau is not

considering these Federal Agency QM rules in the assessment, which is limited to the Bureau’s

own ATR/QM Rule.

A fourth category of qualified mortgages provides more flexible underwriting standards for

small creditor portfolio loans,

and a fifth category allows small creditors that operate in rural

or underserved areas to make balloon-payment portfolio loans that are qualified mortgages.

temporary category that expired in April 2016 allowed any small creditor to make balloon-

payment portfolio loans that are qualified mortgages, even if they did not operate in rural or

underserved areas.

However, amendments prior to the expiration revised the “operate in rural

areas” requirement and preserved the ability of small creditors to make balloon-payment QMs

without regard to whether they operated “predominantly” in rural or underserved areas so long

as such creditors make at least one residential mortgage in a rural or underserved area.

1.1.3 Goals and expected effects of the Rule

The goals of the ATR/QM Rule generally reflect the specific goals set forth by Congress in the

relevant amendments to TILA. Specifically, TILA section 129B, added by section 1402 of the

Dodd-Frank Act, states that Congress created new TILA section 129C upon a finding that

1 2 C .F.R. § 1 026.43(e)(4)(ii)(A).

1 2 C .F.R. § 1 026.43(e)(4)(ii)(B)–(E).

See, e.g., 24 C.F.R. § 2 03.19 (for HUD rules).

1 2 C .F.R. § 1 026.43(e)(5).

1 2 C.F.R. § 1 026.43(f).

1 2 C .F.R. § 1 026.43(e)(5).

1 2 C .F.R. § 1 026.35(b)(2)(iii)(A); s ee also supra note 42.

24 BUREAU OF CONSUMER FINANCIAL PROTECTION

“economic stabilization would be enhanced by the protection, limitation, and regulation of the

terms of residential mortgage credit and the practices related to such credit, while ensuring that

responsible, affordable mortgage credit remains available to consumers.”

T ILA sec tion 129B

further states that the purpose of TILA section 129C is to “assure that consumers are offered and

receive residential mortgage loans on terms that reasonably reflect their ability to repay the

loans and that are understandable and not unfair, deceptive or abusive.”

In its January 2013 Rule implementing these TILA amendments, the Bureau recognized that a

goal of the statute was to prevent a repeat of the deterioration of lending standards which

preceded the financial crisis and which led to various consumer harms.

For example, the

Bureau noted that the ATR requirement of the Rule was intended to prevent consumers from

obtaining mortgages they could not afford.

To the extent that the January 2013 Rule would

reduce credit access, the goal was to reduce lending that ignored or inappropriately discounted a

consumer’s ability to repay.

The Bureau viewed these effects as consistent with congressional

intent and one of the benefits of the Rule.

Similarly, by requiring that creditors determine

ability to repay based on an amortizing payment using the fully indexed rate

(or the maximum

possible rate in five years for certain categories of qualified mortgages), the statute

and the

Rule effectively prohibited underwriting loans based upon low initial monthly payments.

Non-

TILA section 1 29B(a)(1) (codified as amended at 15 U.S.C . § 1639b(a)(1)).

T ILA section 1 29B(a)(2) (codified as am ended at 1 5 U.S.C. § 1639b(a)(2)).

See 78 Fed. Reg. 6408, 6570 (Jan. 3 0, 2013)(“A primary goal of t he statute was to prevent a r epeat of the

deterioration of lending standards that contributed to the financial crisis, which harmed consumers in v arious ways

a n d significantly c urtailed their access t o c redit.”).

“The statutory ability-to-repay standards reflect Congress’s belief that certain lending practices (such as low- or no-

documentation loans or underwriting loans without regard to principal repayment) led to consumers having

m ortgages they could not a fford, r esulting i n high default and foreclosure rates.” Id. a t 6415.

“ The Bureau believes that, t o the extent the final rule r educes c redit access, it will primarily reduce ineffic ient

len ding t hat ignores or inappropriately discounts a consumer’s ability to r epay the loan, thereby preventing

consumer harm, rather than impeding a ccess t o credit for borrowers that do have an ability to r epay.” Id. at 6570.

See also id. at 6558–6560 (Econom ics of A bility T o Repay).

See, supra note, 58.

“ Fu lly indexed rate” means the interest rate calculated u sing the index or formula that will apply after recast, a s

determined at the tim e of consummation, and the maxim um margin that can apply at any time during the loan term.

See 12 C.F.R. § 1 026.43(b)(3).

T ILA section 1 29C(a)(6)(D)(iii) (c odified a s am ended at 15 U.S.C. § 1 639c(a)(6)(D)(iii))(prov iding “the interest rate

ov er the entire t erm of the loan is a fixed rate equal to t he fully indexed rate a t the t ime of t he loan closing, w ithout

considering the introductory rate.”).

For example, low initial payments may occur as the interest-only payments on interest-only loans or negatively

am ortizing option ARMs or result from the introductory rates on hybrid ARMs. The statute required on ly the use of

25 BUREAU OF CONSUMER FINANCIAL PROTECTION

amortizing products were expected likely to persist only in narrow niches for more sophisticated

borrowers who wanted to match their mortgage payment to changes in their expected income

stream, and who had the resources to qualify for the products under the underwriting

assumptions the statute and regulation required.

The Bureau stated a number of other general and particular goals in the January 2013 Rule. The

Bureau stated that it sought to allow for flexible proprietary underwriting standards in ability-

to-repay determinations and to support innovation.

The Bureau also sought to provide

qualified mortgage standards that would allow creditors and the secondary market to readily

determine whether a particular loan is a QM loan. For General QM loans, the ATR/QM Rule

generally requires creditors to use the standards for defining “debt” and “income” in Appendix

Q, which were adapted from FHA guidelines. The Bureau expected that the standards set forth

in Appendix Q, together with the bright-line 43 percent threshold, would provide sufficient

detail and clarity to encourage creditors to provide qualified mortgages to consumers.

The

Bureau also noted, however, that the Rule might have an adverse effect on access to credit for

consumers with atypical financial characteristics, such as income streams that are inconsistent

over time or particularly difficult to document.

The Bureau stated a number of goals for the categories of temporary QM loans.

The Bureau

sought to preserve access to credit for consumers with debt-to-income ratios above 43 percent

during a transition period in which the market was fragile and the mortgage industry was

the fully indexed rate. The Rule requires use of the fully indexed rate or initial rate, whichever i s greater (1 2 C.F.R. §

1 026.43(c)(5)(i)(A)).

See 78 Fed. Reg. 6408, 6562 (Jan. 30, 2013).

Id. a t 6 461 (“The Bureau believes that a variety of underwriting standards c an y ield reasonable, good faith a bility-

to-r epay determinations…. [C]reditors are permitted to dev elop and a pply their own proprietary underwriting

standards and to make changes to those standards over time in response to empirical information and changing

econ om ic and other conditions. T he Bureau believes this fl exibility is necessary given the w ide range of c reditors,

con sumers, and m ortgage products to which this rule appli es.”). Further, “In crafting t he rules t o im plement the

qu a lified mortgage prov ision, the Bureau has sought to b alanc e creating new protecti ons for consumers and n ew

r esponsib ilities for creditors with preserving c onsumers’ a ccess t o credit and allow ing for appropriate lending and

innovation.” Id. at 6505.

“ [T ]he Bureau r ecognizes c oncerns that c reditors should readily be able t o determine w hether indivi dual m ortgage

t r ansacti ons will be deemed qualified mortgages. T he Bureau addresses these c oncerns by adopting a bright-line

debt-to-incom e ratio threshold of 43 percent, as w ell as clear and specific standards, based on FHA guidelines, set

for t h in a ppendix Q for calculating the debt-to-income ratio in i ndividual cases.” Id. at 6525. The 2011 multi-agency

Credit Risk Retention Proposed Rule also r elied on FHA standards for defining “debt” and “ incom e” for purposes of

defining “qualified r esidential mortgage” (QRM) , which would b e exempt from the r isk retention r equirements. Id.

a t 6 5 27; s ee 76 Fed. Reg. 24090 (Apr. 29, 2011).

See 7 8 Fed. Reg. 6408, 6570 (Jan. 30, 2013).

See Chapter 6 for further discussion of the Temporary GSE QM category.

26 BUREAU OF CONSUMER FINANCIAL PROTECTION

adjusting to the final rule.

By providing for most of the conventional market

to continue to

originate higher debt-to-income loans as QM loans, but limiting this to the conforming market

and making the provision temporary, the Bureau sought, over the long term, to encourage

innovation and responsible lending on an individual basis under the ability-to-repay criteria.

The Bureau expected that there would be a “robust and sizable market” for non-QM loans

beyond the 43 percent threshold and structured the Rule to try to ensure that this market would

develop.

The Bureau also stated that because the temporary category of QM loans covers loans

that are eligible to be purchased, guaranteed, or insured regardless of whether the loans are

actually purchased, guaranteed, or insured, private investors could acquire these loans and

secure the same legal protection as the GSEs and Federal agencies.

This would avoid creating a

disincentive for the return of private investors even before the expiration of the temporary

category.

Finally, the Bureau noted that as the market recovered, the GSEs and Federal agencies would be

able to reduce their presence in the market (e.g., by reducing their loan limits). In this scenario,

the percentage of loans granted qualified mortgage status under the temporary category would

7 8 Fed. Reg. 6408, 6533 (Jan. 30, 2013) (“[T]he Bureau acknowledges it may take some time for the non-qualified

m ortgage market t o establish itself in light of the market anxiety regarding litigation risk under the ability-to-repay

r u les, the g eneral slow recovery of the m ortgage market, and the need for lenders t o adjust their operations t o

account for several other major regulatory and c apital r egimes. In light of these factors, the Bureau has concluded

t h at it is appropriate t o prov ide a tem porary a lternative definition of qu alified m ortgage. This will help ensure

a cc ess t o responsible, affordable c redit is available for c onsumers with debt-to-incom e ratios a bov e 43 percent and

fa c ilitate com pliance by lenders by prom oting the use of widely recognized, federally-related underwriting

st a ndards”). On the t ight credit environment at the time of the rulemaking and the general reluctance of lenders

r egarding r isks, see id. a t 6412.

A conventi onal m ortgage l oan is on e that is not insured or guaranteed by the federal gov ernment, including the

Federal Housing Administration (FHA), the U.S. Department of V eterans A ffairs (VA), or the USDA’s Farm Service

A g ency or Rural H ousing Service (FSA/RHS). Conventional loans are either private or guaranteed by one of the two

G ov ernment Sponsored En terprises ( GS Es), the Federal National Mortgage Association (Fannie Ma e) and the

Federal Home Loan Mortgage Corporation (Freddie Mac).

“[The final rule] allows room for a vibrant market for non-qualified m ortgages ov er t ime. The Bureau recognizes

th at there will be many instances in which individual consumers can afford an even higher debt-to-in com e r a tio

based on their particular circumstances, although the Bu reau believes that such loans are better evaluated on an

in dividual basis under t he ab ility-to-repay criteria rather than with a blanket presumption.” Id. at 6506.

“ Ov er the long term, as the market recovers from the mortgage crisis and adjusts to the ability-to-repay rules, the

Bu r eau expects that there will be a r obust and sizable m arket for prudent l oans beyond the 43 percent threshol d

ev en without the b enefit of t he presumption of com plianc e that applies to qu alified m ortgages. In short, the Bureau

does not believe that consumers who do not receive a qualified mortgage because of the 43 percent debt-to-in com e

ratio threshold should be cut off from responsible credit, and has structured the rule to try to ensure that a r obust

a n d affordable a bili ty t o-repay market develops ov er time.” Id. at 6528.

“ The t emporary exception has been c arefully structured t o cover loans that are eligible t o be purchased, guaranteed,

or in sured by the G SEs (while in conservatorship) or Federal agencies r egardl ess of whether the loans are actually

so purchased, guaranteed, or insured; this will leave room for private investors to return to the market and secure

t h e same l egal protection as the GS Es and Federal agencie s.” Id. at 6534.

27 BUREAU OF CONSUMER FINANCIAL PROTECTION

also shrink and the market would be able to develop alternative approaches to assessing ability-

to-repay within the General QM requirements.

When the January 2013 Rule was released, the Bureau issued public statements that reiterated

these goals and elaborated on particular aspects of these goals. The Bureau stated that

consumers would be protected from risky lending practices and would not receive loans that

they could not afford.

The Bureau also described “two distinctly different mortgage markets”

over the previous decade, the first in which lending was lax and a more recent one in which

credit was tight. The Bureau stated that its goal with the January 2013 Rule was to address both

of these issues, to make sure borrowers were assured of greater consumer protections and have

reasonable access to credit.

In May 2013, the Bureau amended the January 2013 Rule to exempt certain creditors and

mortgage loans from ability-to-repay requirements; provided an additional definition of a

qualified mortgage for certain loans made and held in portfolio by small creditors, and a

temporary definition of a qualified mortgage for balloon loans; and revised rules on how to

“ A t the same t ime, a s t he m arket r ecov ers and the GSEs and FHA are a ble to r educe their presence in the m arket,

t h e percentage of loans that are granted qualified m ortgage status under the t emporary definition will shrink

t ow ards the long term structure.” Id.

Pr ess Release, Bureau of Consumer Fin. Prot., CFPB Iss ues Rule to Protect Co nsumers from Irresponsible

Mortgage Lending (Jan. 10, 2013), available at https://www.consumerfinance.gov/about-

u s/n ew sr oom /c onsu mer-financial-protection-bureau-issu es-rul e-to-protect-con su mer s-fr om -ir resp onsi ble -

m ortgage-lending/ (“When consumers sit down at the closing table, they shouldn’t be set u p to fail with mortgages

t h ey c an’t afford…. Our Ab ility-to-Repay rule protects borrowers from the kinds of risky lending practices that

r esulted in so m any families losing their hom es. Th is c om mon -sense rule ensures responsible borrowers get

responsible loans.”).

Pr ess Release, Bureau of Consumer Fin. Prot., Prepared Remarks of Richard Cordray at the Ability-to-R epay Rule

Field Hearing (Jan. 1 0, 2013), available at h ttps://www.consumerfinance.gov /about-us/newsroom /prepared-

r e m arks-of-richard-cordray-at-the-abili ty -to-pay-rule-field-hearing/ (“The Ability-to-Repay rule…comes against

t h e backdrop of t wo distinctly different m ortgage m arkets that we have experienced ov er t he past decade. In the

run-up to the financial crisis, we had a housing market that was reckless about lending money. It was driven by

assumptions about property values that turned out to be badly wrong. It had dysfunctional incentives, with lenders

b ein g a bl e t o of f -load v irtually any m ortgage into the secondary market regardless of the quality of the underwriting.

Th ere was broad indi fference to t he ability of m any consumers to repay loans…. Now , in the wake of t he financial

cr ash, we have been experiencing a housing m arket that is tough on people in just the opposi te way—credit is

achingly tight. Since 2008, most mortgages are being priced on very attractive terms. But a ccess to credit has

become so h ighly constrained that many consumers cannot borrow to buy a house ev en with strong credit…. Our

g oa l with the Ability-to-Repay rule is to make sure that people who work hard to buy t heir own hom e can be assured

of n ot on ly greater consumer protections b ut a lso reasonable access to credit so they c an g et a sustainable

m ortgage.”).

28 BUREAU OF CONSUMER FINANCIAL PROTECTION

calculate loan originator compensation for certain purposes.

The Bureau stated that the goals

of these rules were generally to foster access to responsible credit for consumers.

In September 2015, the Bureau issued amendments to further facilitate the origination of

qualified mortgage loans by small creditors, including loans with balloon payments.

The

Bureau stated that the goals of these rules were to help consumers in rural or underserved areas

access mortgage credit.

In March 2016, the Bureau implemented the Helping Expand Lending Practices in Rural

Communities (HELP) Act through an interim final rule.

This rule further expanded the ability

of small creditors to originate qualified mortgage loans with balloon payments.

1.1.4 Determination that the ATR/QM Rule is a significant

rule

As discussed in the June 2017 RFI, the Bureau determined that the ATR/QM Rule—here, the

January 2013 Rule and the amendments that took effect on January 10, 2014—is a significant

7 8 Fed. Reg. 35430 (June 1 2, 2013).

Id. (Regarding sm all creditors, the am endments were “necessary to preserve access to credit for some consumers,

in cluding c onsumer who do n ot qualify for conforming m ortgage credit, and w ill ensure that this credit is prov ided

in a r esponsib le, a ffordable w ay…. [ T]he Bureau understands that sm all creditors are a significant source of

n on conforming mortgage credit”) Id. at 35484; (regarding loan or iginator com pensation, “[T]he Bureau believes

th at there remain some risks of consumer injury from business models in which mortgage brokers attempt to steer

con sumers t o more costly transactions. Including in points and fees compensation paid by creditors to mortgage

br ok ers should h elp reduce those r isks.”) Id. at 35456; see also Press Release, Bureau of Consumer Fin. Pr ot., CFPB

Finaliz es Am endments to Ability-to-Repay Rule (May 29, 2013), available at

h t t p s ://w ww .con su mer fin a n ce.g ov /a bou t-us/newsroom /cfpb-finali zes-amendments-to-ability-to-repay-rule/

(“T oday’s amendments embody our efforts to make reasonable changes t o the rule in order to foster access to

r esponsible credit for c onsumers.”).

80 Fed. Reg. 59943 (Oct. 2, 2015).

Id. (“ The intent of the sm all creditor test is to facilitate lending by those sm all creditors that provide responsible,

a ffor dable credit t o c onsumers, and t o enable c onsumers in rural and underserved areas t o access creditors with a

lending m odel, operations, and pr oducts that may m eet their particular needs.”) Id. at 59 95 0; see also Pr ess Relea se,

Bu r eau of Consumer Fin. Prot., CFPB Finaliz es Rule to Fac ilitate Access to Credit in Rural and Underserv ed Areas

(Sept. 21, 2015), available at https://www.consumerfinance.gov /about-us/newsroom /cfpb-finalizes-rule-to-

fa cilitate-access-to-credit-in-rural-and-underserved-areas/ (“The financial crisis was not caused by community

ba n ks and credit unions, and our mortgage rules reflect the fact that small institutions play a vital role in many

com m uniti es…. These changes will h elp consumers in rural or underserved areas a ccess the mortgage credit they

n eed, w hile still maintaining these im portant new c onsumer protections.”).

81 Fed. Reg. 16074 (Mar. 25, 2016) (“This interim final rule is implementing Congress’s intention to expand the

cohort of small creditors that are eligible for a special prov ision of Regulation Z that permits origination of balloon-

pa y ment qualified m ortgages….”) Id. at 1 6075.

29 BUREAU OF CONSUMER FINANCIAL PROTECTION

rule for purposes of section 1022(d) of the Dodd-Frank Act.

The Bureau stated in the RFI that

it believed that the initial effect of the ATR/QM Rule on costs was muted given market

conditions prevailing in early 2013 and given the Bureau’s decision to create a broad temporary

category of QM loans, particularly the Temporary GSE QM loans. The Bureau recognized that

industry’s strong preference to obtain a presumption of compliance with the ATR/QM Rule by

originating QM loans resulted in meaningful operational changes in originations across the

market. The Bureau also took into consideration the possible effect of the AT R/QM Rule on

access to credit in particular submarkets and possible effects on innovation, overall product

design, and competition. Considering these factors, coupled with the Bureau’s more general

interest to better understand how the Rule’s effects vary under different market conditions, the

Bureau concluded that the ATR/QM Rule was a significant rule for purposes of section 1022(d).

1.2 Methodology and plan for assessing

effectiveness

In general, the Bureau’s methodology for the assessment consisted of three steps:

• First, the Bureau considered at a high level the potential relevant effects of the Rule at a

high level. These effects are the intended and unintended consequences of the Rule that

would potentially be useful in evaluating whether the Rule, or a specific Rule

requirement, furthers the goals of the Rule that were stated at the time of the rulemaking

and, as relevant, the purposes and objectives of the Bureau or other relevant factors. T he

Bureau also considered the broader market context that could influence the effect of the

Rule.

• Second, the Bureau developed specific measures of the potential relevant effects and

market conditions. The Bureau then collected available evidence and data that would

allow the Bureau to compute these measures.

• Third, the Bureau analyzed these measures and considered whether the Rule or specific

Rule requirement furthered the goals of the Rule that were stated at the time of the

rulemaking and, as relevant, the purposes and objectives of the Bureau or other relevant

factors. In doing so, where possible, the Bureau compared the observed measures to

what those measures would be under a counterfactual or “baseline.”

See, supra note, 33.

30 BUREAU OF CONSUMER FINANCIAL PROTECTION

Specifying a baseline against which to evaluate a rule’s effects is necessary for both forecasting

the future effects of proposed regulations and evaluating the historical effects of adopted

regulations.

When a regulation has already taken effect, however, it is often not possible to

find a group of firms or a part of the market that is neither subject to the rule nor indirectly

affected by the rule—but is nevertheless subject to the same other determinants of prices,

quantities and other market outcomes—such that data about those firms or that market provide

a baseline for evaluating the effects of the rule. In particular cases, it may be possible to define a

specific set of outcomes that can serve as the baseline. For example, it may be generally agreed

that the purpose of the rule is to increase (or reduce) particular outcomes relative to some

observed or specified benchmark. In general, however, retrospective analysis requires making a

formal or informal forecast of the market absent a rule, or absent a specific provision of a rule,

to serve as the baseline, and data limitations make this difficult to do in practice.

For purposes of this assessment, the Bureau has generally used a baseline that is the market

absent the Rule as a whole or the specific Rule provision being evaluated.

For certain analyses,

the data is available with which to estimate this baseline. The lender survey that the Bureau

conducted also provides insight into how mortgage origination policies responded to the Rule.

When it is not possible to reliably estimate what a measure would have been absent the Rule or a

specific provision, the analysis uses other baselines, in some cases comparing the relevant

measure to its level before the Rule’s effective date, thus capturing changes since the Rule took

effect. Such changes are an imperfect measure of the effects of the Rule to the extent that market

changes that would have taken place even absent the Rule affect relevant measures.

As noted above, in June 2017, the Bureau published an RFI that, among other things, described

the assessment plan and requested public comment on the plan.

The RFI described the general

focus of the assessment and some of the effects and outcomes that the Bureau would analyze.

See, e.g., Joseph E. Al dy, Learning from Ex perience: An Assessm ent of the Retrospective Rev iews of Agency Rules

a n d the Ev idence for Im prov ing the Desi gn and Im plementation of Regulatory Policy, (Harv., Retrospective Rev.

Rep., 2014), available at https://www.acus.gov/report/retrospective-rev iew -report(prepared for consideration of

t h e Adm inistrative C onferenc e of t he United States) (“In ev aluating the efficacy, b enefits, and costs of any individual

regulation, an analyst must make a determination a bout the counterfactual, i.e., what would have happened i n the

absence of the regulation. In ex ante analysis, this requires constructing an alternative future scenario, or baseline,

fr om which to assess the impacts of the proposed regulation. In ex post analysis, this requires constructing an

alt ernative h istori c scenario for com parison with the im plem ented regulation. T he choi ce of c ounterfactual c an b e

qu it e challenging and subject to c ritic ism .”) Id. at 62–63 . See a lso the extensive list of references c ontained therein.

This r eport also uses other baselines, such as the effects that the Bureau expected would occur at the time of the

r u lem aking. See, for examp le, Chapter 6.

82 Fed. Reg. 25246, 25248–50 (June 1, 2017).

Id. a t 25249. “The Bureau anticipates that the assessm ent wil l primarily focus on t he ATR/QM Rule’s r equirements

in achieving the goal of preserving consumer access to responsible, affordable credit. The Bureau stated with the

31 BUREAU OF CONSUMER FINANCIAL PROTECTION

The major provisions that the RFI said were to be examined were (i) the ATR requirements a

creditor must consider, including the eight underwriting factors; (ii) the QM provisions, with a

focus on the DTI threshold, the points and fees threshold, the small creditor threshold, and the

Appendix Q requirements; and (iii) the applicable verification and third-party documentation

requirements. The outcomes included effects on mortgage costs, origination volumes, approval

rates, and subsequent loan performance; and certain changes in creditors’ underwriting policies

and procedures.

The RFI also described the data that were available to the Bureau at that time

and the data that the Bureau expected to obtain.

Comments on the assessment plan received in response to the RFI generally proposed either

specific analyses for the Bureau to consider or specific data for the Bureau to collect. The

analyses and data collections used in this assessment and discussed in this report are largely

consistent with those proposed by commenters.

It was not possible, however, to consider the

impact of the Rule on every sub-group of creditors or consumers suggested by some

commenters. In particular, a number of commenters recommended that the Bureau assess the

effects on consumers, mortgage brokers and affiliates of including certain payments and

expenses in calculating total points and fees for purposes of meeting the QM threshold. In order

to quantify these effects, however, the Bureau would need data on the frequency with which

total points and fees exceeded the threshold on initial applications—ideally, before and after the

Rule took effect—and then data (post-Rule) on adjustments that took place in order to stay

under the threshold. The Bureau had limited data on points and fees in its possession at the

start of the assessment, there are no publicly-available datasets with the desired information,

and it would have been extremely burdensome to require standardization and reporting of this

information to the Bureau for purposes of the assessment. The Bureau did, however, collect data

from nine lenders and conducted a lender survey in order to acquire certain data on the

Ja n uary 2 013 Rul e its b elief that the ATR/QM Rule ‘will n ot lead to a significant reduction in consumers’ access to

con sumer financial products and services, namely m ortgage credit’ (references om itted). The Bureau t ook into

con si deration, how ever, the potential that the rule ‘m ay have a disproportionate im pact on acc ess to c redit for

consumers with atypical financial characteristics, such as incom e streams that are inconsistent ov er time or

pa r ticularly difficult t o document.’”

Id. In a nalyzing t hese effects, the RFI st ated t hat certain categories of borrowers w ere of special interest, but that

the data for considering any differential im pacts of the Rule on these borrowers were not necessarily availab le.

T h ese c ategories w ere: (i) b orrowers g enerating incom e from self-employ ment (including those w orking a s

“ contract” or “1099” employees); (ii) borrowers anticipated to r ely on incom e from a ssets to r epay the l oan; (iii)

bor r owers who r ely on intermittent, supplemental, part-time, s e a s on al , b on u s, or ov er tim e i n com e; ( iv) b orr ow er s

seek ing sm aller-than-average loan amounts; (v ) borrowers with a debt-to-incom e rati o exceeding 43 percent; (vi)

low a nd m oderate incom e b orrowers; (vii) minority borrowers; and (v iii) rural b orrowers. The assessm ent generally

focu sed on (i), (iv) and (v ), with som e discussion of (v iii) , due to data limitations.

Id. a t 2 5249–25250.

See A ppendix B (T he a ssessment plan) and S ection 1 .3 of t his report.

32 BUREAU OF CONSUMER FINANCIAL PROTECTION

frequency with which applications and originations fail the points and fees threshold and

adjustments that occur. This data, together with the information that the Bureau does have in its

possession, provide some insights into the effects of the points and fees threshold on consumers

and mortgage brokers for different types of loans.

1.3 Sources of information and data

In conducting the assessment the Bureau reviewed available public sources of data, including

both publicly available loan-level data and published studies and reports pertaining to mortgage

originations and performance. The Bureau’s researchers also reviewed information it obtained

through various channels in the normal course of its work and in response to the Request for

Information the Bureau published regarding this assessment. Based on its review, the Bureau

concluded that additional data were needed to conduct this assessment and collected certain

data as described below. Described below are the principal sources of data that the Bureau has

found most probative and on which the findings in this report are primarily based.

• Loan origination and performance data from the National Mortgage Database

(NMDB), Black Knight, CoreLogic, and HMDA. Throughout this assessment, the Bureau

used three sources of de-identified data that combine loan-level performance and

origination data and a fourth source of de-identified origination data. The first is the

National Mortgage Database (NMDB) jointly developed by the FHFA and the Bureau,

which provides loan characteristics and performance for a 5 percent representative

sample of all mortgage originations from 1998 to the present, supplemented by loan and

borrower characteristics from federal administrative sources and credit reporting data

on the additional debts held by these mortgage borrowers. The second is the

commercially available “McDash” data set from Black Knight (McDash Data), which

includes data on approximately 160 million loans serviced from 1989 to 2017. The third

dataset is CoreLogic’s Loan-Level Market Analytics (LLMA) data which contains detailed

loan-level information on originations and performance with market coverage of 76

percent of all residential mortgages in the United States since January 1, 1999. Loan-

level performance information is generally updated on a monthly basis in the McDash

and CoreLogic datasets, and quarterly in the NMDB.

See Section 5 .4 and in particular Section 5.4.6.

The Bureau considered additional available datasets, including publicly available loan-level data from the GSEs, but

fou n d these l ess probative than, or superseded by , the datasets describ ed below .

33 BUREAU OF CONSUMER FINANCIAL PROTECTION

An advantage of these loan performance sources is that they include a large number of

loans from a broad selection of lenders, with information on relevant loan attributes

including debt-to-income ratios, GSE securitization status, loan amounts, interest rates,

Loan-to-value (LT V) ratios, and loan types, as well as borrowers’ credit scores. The

NMDB data are particularly well suited to providing nationally representative results and

insights into loans insured by the GSEs. The McDash and CoreLogic Data supplement

these with additional detail on non-insured loans for large but non-random samples of

the market. However, none of these three datasets can distinguish all non-QM loans

from QM loans, and they do not include data on all loan features affected by the Rule

(e.g., points and fees).

A fourth dataset, coming from required filings under the Home Mortgage Disclosure Act

(HMDA Data), does not contain loan performance data, but does provide loan and

borrower characteristics for over 90 percent of mortgage originations. The dataset used

is the Federal Agency HMDA data, which includes additional fields, notably application

dates and closing dates, which are not contained in the publicly available HMDA data.

The Bureau uses these data to measure market-wide shifts over time in the

characteristics of new mortgage originations.

• Desktop Underwriter and Loan Prospector submissions and acquisitions data provided

by Fannie Mae and Freddie Mac. These data contain disaggregated counts of

submissions to the Desktop Underwriter (DU) and Loan Prospector (LP) Automated

Underwriting Systems operated by the GSEs, and counts of loans acquired by the GSEs

disaggregated by borrower and loan characteristics. The data are used in Chapter 6 to

measure utilization of the Temporary GSE QM provision.

• Application-level data from nine lenders. The Bureau collected de-identified

application-level data from nine mortgage lenders using its authority under section

1022(c)(4) of the Dodd-Frank Act (“Application Data”). The lenders were selected to

represent a range of large nationally operating banks and non-depositories. The data

collection covered all applications received from 2013 to 2016. In total, the Application

Data cover over five million applications. The data include information about each

application’s characteristics and whether the application was approved by the lender.

The Application Data are a unique source of information about the activities that were

directly affected by the Rule. They provide insight into how lenders’ approval rates and

processes may have changed in response to the Rule, as well as into how application

behavior by consumers may have changed. These data are supplemented by lenders'

34 BUREAU OF CONSUMER FINANCIAL PROTECTION

responses to a series of qualitative questions about how they incorporated the

requirements of the Rule into their business practices. Importantly, however, because

they are drawn from the records of only nine lenders, the Application Data may not be

representative of data from all lenders.

• Lender survey. The Bureau conducted a voluntary survey to ask mortgage lenders about

business process changes they have made in response to the Rule. The lenders surveyed

varied both in size and institution type. The survey provides information on lenders’

business practices before and after the Rule, their experience underwriting to Appendix

Q, and how they responded to certain requirements of the Rule. This provides valuable

information on the Points and Fees requirement and Non-QM originations that are

lacking in other data sources. Over 190 lenders responded to the survey. Lenders

responding to the survey had the opportunity to provide more information in the form of

structured interviews as well. Relative to the 1022(c)(4) request associated with the

Application Data, the survey provides more information on how smaller lenders

responded to the Rule. Although informative, the lender survey is not statistically

representative of the market as a whole.

• Supervision Data. The Bureau has utilized data from several fair lending exams to

examine whether the QM points and fees requirements is associated with changes in

closing costs; the data has also been used to examine the impact of the rebuttable

presumption.

• Residential mortgage backed securities (RMBS) data from IMF, Bloomberg, L.P., and

SEC. To analyze possible effects of the Rule on secondary markets for mortgage

securities, the Bureau used aggregate data from Inside Mortgage Finance (IMF). IMF

reports annual aggregate volumes of RMBS issuances by securitizer giving a historical

perspective on the size of the securitization market from 1990 to 2017. This allowed the

Bureau to evaluate how the Rule may have affected the volume and/or the composition

of RMBS issuances.

To assess the post-Rule securitization market, the Bureau used additional loan-level

RMBS data from Bloomberg and the Securities and Exchange Commission (SEC). These

data are used to analyze the market for Non-QM securitizations between 2015 and 2018.

Data for all known Non-QM issuances come from Bloomberg and include fields for loan

and borrower characteristics. The data were then merged to publicly available due

diligence reports from the SEC to determine an individual loan’s QM status to compile

detailed information on the loan and borrower characteristics for Non-QM loans and the

appetite for Non-QM securitizations in the secondary market.

35 BUREAU OF CONSUMER FINANCIAL PROTECTION

• MBA cost data. To understand how the Rule may have affected lender costs, the Bureau

used cost data from the Mortgage Bankers Association’s (MBA) Annual Mortgage

Bankers Performance Reports between 2009 and 2018. The reports contain data on the

revenue and expenses associated with the origination and servicing of one to four unit

residential mortgage loans of independent mortgage companies and other non-

depository institutions. The annual reports also contain information on production and

servicing volume mixes by product type and overall income and balance sheet

summaries. Most lenders included in the data are independent mortgage companies.

While not necessarily representative, the data provide detailed information on the cost

and revenue structure for a large share of independent mortgage companies.

• CSBS Public Survey data. The Bureau used data from the Conference of State Bank

Supervisors’ (CSBS) 2015 National Survey of Community Banks to examine the behavior

of small creditors in relation to the Rule. The 2015 CSBS Public Surv ey inv olved 974

respondents in 39 states, most of which were small creditors. This survey in particular

included many questions related to the Rule. Notably, the CSBS survey data include

information from respondents that do not report to HMDA and consequently may be

underrepresented in available loan-level datasets.

• Evidence from comments received in response to the 2017 RFI concerning the ATR/QM

assessment. The Bureau received approximately 480 comments in response to the RFI.

Approximately 75 percent of the comments came from mortgage brokers or loan

originator organizations. A small number of commenters provided quantitative

information regarding their own experiences with the Rule. A number of commenters

pointed the Bureau toward published research regarding the overall effects of the Rule

and the effects of particular Rule requirements that are within the scope of the

assessment. This information is summarized in Appendix B and incorporated into other

parts of the report as appropriate.

Secondary sources of information. In addition to the primary sources of data discussed above,

the Bureau reviewed a number of secondary sources of information, including reports suggested

by commenters discussed above, the reports of other federal agencies, and published research

on the mortgage market and the Rule. This report discusses and cites these reports in the

relevant sections below. In addition, the Bureau held conversations with industry participants to

understand their experiences with the Rule.

36 BUREAU OF CONSUMER FINANCIAL PROTECTION

2. The ATR/QM Rule

This chapter discusses the statutory basis of the Rule, the development of the Rule, and the

provisions of the ATR/QM Rule.

2.1 Statutory background

The ATR/QM Rule is based on several related provisions enacted in the Dodd-Frank Act.

Section 1411 of the Dodd-Frank Act added a new section 129C(a) to TILA. This new section

generally prohibits a creditor from making a residential mortgage loan unless the creditor has

made a reasonable and good faith determination based on verified and documented information

that, at the time the loan is consummated, the consumer has a reasonable ability to repay the

loan. This requirement does not apply to an open-end credit plan, timeshare plan, reverse

mortgage, or temporary loan with a term of 12 months or less.

New T ILA section 129C(a) also

establishes certain minimum underwriting factors that a creditor generally must consider in

determining the consumer’s repayment ability, including: the consumer's credit history, current

income, expected income the consumer is reasonably assured of receiving, current obligations,

debt-to-income ratio or the residual income the consumer will have after paying non-mortgage

debt and mortgage-related obligations, employment status, and other financial resources not

including the consumer's equity in the dwelling or real property that secures repayment of the

loan.

Creditors that violate the ability-to-repay requirements may be subject to government

enforcement and private actions. As amended by section 1416 of the Dodd-Frank Act, T ILA

provides that a consumer who brings a timely action against a creditor for a violation of the

ability-to-repay requirement may be able to recover special statutory damages equal to the sum

The T ILA sect ion 1 03 (cc)(5 ) definition of “residential m ortgage loan,” added by secti on 1 401 of the Dodd-Fr a n k A ct,

ex cludes open-end credit plans and t imeshare plans. T ILA section 129 C(a)(8) excludes reverse m ortgages and

temporary loans with terms of 1 2 months or less.

TILA section 1 29C(a)(3) (codified a s amended a t 15 U.S.C. § 1639c(a)(3)).

37 BUREAU OF CONSUMER FINANCIAL PROTECTION

of all finance charges and fees paid by the consumer (but not to exceed three years of such

charges and fees), unless the creditor demonstrates that the failure to comply was not

material.

Moreover, TILA section 130(k), added by section 1413 of the Dodd-Frank Act,

provides that if a creditor, an assignee, other holder or their agent initiates a foreclosure action,

a consumer may assert a violation of TILA section 129C(a) (i.e., the ability-to-repay

requirements) ‘‘as a matter of defense by recoupment or setoff” to the initiation of a foreclosure

action, while setting no time limit on consumer use of this defense.

To provide more certainty to creditors that they are in compliance with the ability-to-repay

requirements and not subject to liability while also protecting consumers from loans with terms

that do not reasonably reflect their ability to repay, section 1412 of the Dodd-Frank Act added

TILA section 129C(b).

TILA section 129C(b)(1) states that a creditor or assignee may presume

that a loan has met the repayment ability requirement if the loan is a qualified mortgage.

T ILA

section 129C(b) generally defines a qualified mortgage as a residential mortgage loan for which:

the loan does not contain negative amortization, interest-only payments, or balloon payments;

the term does not exceed 30 years; the points and fees (costs associated with the origination of

the loan) generally do not exceed 3 percent of the loan amount; the consumer’s income or assets

are verified and documented; and the underwriting is based on the maximum interest rate

during the first five years of the loan, uses a payment schedule that fully amortizes the loan over

the loan term, and takes into account all mortgage-related obligations.

A qualified mortgage

must also comply with any guidelines or regulations established by the Bureau relating to total

monthly debt payments to total monthly income ratio (DTI) or alternative measures of ability to

pay regular expenses after payment of total monthly debt taking into account the borrower’s

income.

The Dodd-Frank Act amendments to TILA also authorize the Bureau to prescribe regulations

TILA section 1 30(a)(4), (codified as amended at 15 U.S.C. § 1 640(a)(4)). T his recov ery for a violation of A TR is in

addition to actual damages; statutory damages in an individual action or class action, up to a prescrib ed t hreshold;

an d court costs and attorney fees that would be available for v iolations of other TILA prov isions. See TILA section

1 30(a), (codified as amended at 15 U.S.C. § 1 640(a)). The statute of limitations for an action for a violation of TILA

section 1 29C is three years from the date of the occurrence of the violation, as compared to one year for other TILA

v iolations. See TILA section 130(e), (codified as amended at 15 U.S.C. § 1640(e)).

T ILA section 1 30(k) , (c odified a s amended a t 15 U.S.C. § 1 640(k)).

T ILA section 1 29C(b) (c odified a s amended a t 15 U.S.C. § 1639c(b)).

TILA section 1 29C(b)(1) (codified as amended at 15 U.S.C. § 1639c(b)(1)).

TILA section 1 29C(b)(2)(A), (codified as amended at 15 U.S.C . § 1639c(b)(2)(A)).

T ILA section 1 29C(b)(2)(A)(vi) (codified a s amended a t 15 U.S.C. 1639c(b)(2)(A)(vi)).

38 BUREAU OF CONSUMER FINANCIAL PROTECTION

that would revise, add to, or subtract from criteria that define a qualified mortgage upon a

finding that such regulations are necessary or proper to ensure that responsible, affordable

mortgage credit remains available to consumers in a manner consistent with the purposes of the

ability-to-repay requirements; or are necessary and appropriate to effectuate the purposes of the

ability-to-repay requirements, to prevent circumvention or evasion thereof, or to facilitate

compliance with TILA sections 129B and 129C.

100

The Dodd-Frank Act further provides the

Bureau with certain other specific grants of rulewriting authority with respect to the ability-to-

repay and qualified mortgage provisions, including (for instance) express authority to prescribe

rules adjusting the qualified mortgage points and fees limits to permit creditors that extend

smaller loans to meet the requirements of the qualified mortgage provisions.

101

T ILA section

129C(b)(2)(E), added by the Dodd-Frank Act, grants the Bureau discretion to determine

whether to issue rules providing that the term “qualified mortgage” covers balloon loans that

meet certain minimum criteria, the contours of which the Bureau further has discretion to set

under the statute.

102

As discussed in the next section of this Chapter, the Bureau exercised these

authorities in finalizing rules implementing sections 1411 and 1412 of the Dodd-Frank Act.

2.2 ATR/QM Rule background

This section broadly describes the Bureau’s development of its ATR/QM Rule. Rulemaking

authority for TILA was originally vested in the Board of Governors of the Federal Reserve

System (Board). General rulemaking authority for TILA, including the ATR/QM Rule,

transferred from the Board to the Bureau on July 21, 2011, pursuant to the Dodd-Frank Act. In

May of 2011, before the transfer of rulemaking authority to the Bureau went into effect, the

Board published for public comment a proposed rule (May 2011 Proposed Rule) proposing to

amend Regulation Z to implement the ability-to-repay and qualified mortgage amendments to

TILA made by the Dodd-Frank Act.

103

The Bureau reopened the comment period on June 5,

2012 to solicit comment on new data and information submitted during or obtained after the

close of the original comment period.

104

The Bureau’s January 2013 Rule implemented the

100

TILA section 1 29C(b)(3)(B)(i), (codified a s amended a t 15 U.S.C. § 1639c(b)(3)(B)(i)).

101

TILA section 129C(b)(2)(D), (codified as amended at 1 5 U.S.C. § 1 639c(b)(2)(D)).

102

T ILA section 129C(b)(2)(E), (codified as amended at 1 5 U.S.C. § 1 639c(b)(2)(E)). S ecti on 1 0 1 of th e Ec onom i c

Gr owth, Regulatory Relief, and Consumer Protection Act, Pub. L. 115–1 74, enacted May 24, 2018, established an

a dditional category of qu alified mortgages for loans held in portfolio by certain lenders.

103

7 6 Fed. Reg. 27390 (May 11, 2011).

104

7 7 Fed. Reg. 33120 (June 5, 2012).

39 BUREAU OF CONSUMER FINANCIAL PROTECTION

statutory ability-to-repay provisions after reviewing and considering the comments submitted in

response to the Board’s May 2011 Proposed Rule and to the additional comment request by the

Bureau.

The January 2013 Rule, among other things, generally requires creditors to make a reasonable,

good faith determination of a consumer’s ability to repay any consumer credit transaction

secured by a dwelling, other than an open-end credit plan, timeshare plan, reverse mortgage, or

temporary loan with a term of 12 months or less.

105

The January 2013 Rule describes certain

minimum requirements for creditors making ability-to-repay determinations, but does not

dictate that they follow particular underwriting standards. The January 2013 Rule also provided

four categories of “qualified mortgage” loans, for which compliance with the ability-to-repay

requirement is presumed.

106

The Bureau amended the January 2013 Rule several times prior to its effective date to address

important questions raised by industry, consumer advocacy groups, and other stakeholders. The

Bureau determined that these amendments were necessary to protect consumers better, avoid

potentially significant disruption in mortgage markets, and clarify standards by making

technical corrections and conforming changes.

As discussed in the Introduction, the Bureau has determined that the January 2013 Rule and

related amendments that took effect on the Rule’s effective date collectively make up a

significant rule, the ATR/QM Rule, for purposes of this assessment. The amendments that the

Bureau did and did not consider as part of the assessment are described below.

2.2.1 Amendments to the Rule considered in the

assessment

May 2013 final rule. To avoid impairing access to credit for consumers on terms that reasonably

reflect their ability to repay, the May 2013 final rule

107

provided exemptions from the ability-to-

repay requirements for loans made by creditors pursuant to identified development and Federal

emergency economic stabilization programs and for loans made by certain nonprofit

105

1 2 C .F.R. § 1 026.43(c)(1).

106

1 2 C .F.R. § 1 02 6.43 (e) and ( f) established t he G eneral, T emporary GSE, T emporary Agency, and

Ru r al/Undeserved Sm all Creditor Ba lloon Payment QM c ategories.

107

7 8 Fed. Reg. 35430 (June 12, 2013). The rule w as finalized and issued on May 29, 2013, but was not published

u n til the later date.

40 BUREAU OF CONSUMER FINANCIAL PROTECTION

creditors.

108

Further, the May 2013 final rule added two new qualified mortgage categories to

the four categories provided in the January 2013 Rule. One of the new QM categories was for

loans held in portfolio by small creditors

109

and the other new QM category was a temporary

category that allowed all small creditors to make balloon-payment qualified mortgages.

110

July 2013 final rule. The final rule published on July 24, 2013,

111

included clarifications to the

Temporary GSE QM and Temporary Federal Agency QM categories and to Appendix Q, which

prescribes the income and debt a creditor uses to determine a consumer’s debt-to-income ratio

for purposes of the General QM category.

October 1, 2013 final rule. The October 1, 2013, final rule

112

expanded the small creditor balloon-

payment QM category to include certain high-cost mortgages and to cover additional creditors,

those that met the “rural or underserved” definition in any of the three preceding years rather

than only in the preceding year.

October 23, 2013, interim final rule. The interim final rule published on October 21, 2013,

113

included a minor technical correction to the Federal Agency QM loan category.

2.2.2 Other substantive rules affecting the ATR/QM Rule

In addition to the above rules amending the January 2013 Rule before its effective date that the

Bureau is considering as part of the assessment, the Bureau has issued other substantive rules

that affect the ATR/QM Rule. Although these other rules technically fall outside the five-y ear

assessment period, they are considered to the extent they are reflected in the data and are

relevant to the analysis of the ATR/QM Rule’s effectiveness in meeting its purposes, objectives,

and goals.

November 2014 final rule. A final rule published on November 3, 2014,

114

excluded certain

subordinate loans originated by nonprofit creditors from the number counted for purposes of

108

1 2 C.F.R. § 1 026.43(a)(3)(iv)—(vi) .

109

1 2 C .F.R. § 1 026.43(e)(5).

110

1 2 C .F.R. § 1 026.43(e)(6).

111

7 8 Fed. Reg. 44686 (July 24, 2013).

112

7 8 Fed. Reg. 60382 (Oct. 1, 2013).

113

7 8 Fed. Reg. 62993 (Oct. 23, 2013).

114

7 9 Fed. Reg. 65300 (Nov. 3, 2014).

41 BUREAU OF CONSUMER FINANCIAL PROTECTION

the nonprofit exemption from the ability-to-repay requirements.

115

This final rule also

implemented a temporary points and fees cure provision.

116

October 201 5 final rule. Among other changes, the October 2015 final rule

117

increased the

number of creditors that could meet the definition of “small creditor” by raising the originations

limit from 500 first-lien mortgage loans to 2,000 and excluding loans held in portfolio.

118

T he

October 2015 final rule also substantially expanded the definition of “rural” by adding census

blocks

119

that are not in an “urban area,” as defined by the Census Bureau, to the definition of

rural areas.

1 20

March 2016 interim final rule. The Bureau published an interim final rule

1 21

on March 25, 2016

amending Regulation Z to implement the HELP Rural Communities Act provision

1 22

that

removed “predominantly” from the TILA requirement that small creditors operate

“predominantly in rural or underserved areas”

1 23

to qualify for certain special provisions,

including eligibility to make balloon-payment qualified mortgages. The Bureau implemented the

removal of “predominantly” by replacing the “extended more than 50 percent of their total

covered transactions in rural or underserved counties” requirement with “extended a first-lien

covered transaction on a property that is located in an area that is designated either ‘rural’ or

‘underserved.’”

1 24

115

1 2 C .F.R. § 1 026.43(a)(3)(vii).

116

12 C.F.R. § 1 026.43(e)(3)(iii)—(iv).

117

80 Fed. Reg. 59943 (Oct. 2, 2015).

118

1 2 C .F.R. § 1 026.35(b)(2)(iii)(B); c om men ts 12 C.F.R. § 1 026.43(e)(5)–4; 12 C.F.R. § 1 026.43(f)(2)(ii)–1 . The

changes were made to the exemption provisions of an escrow rule, which are cross-referenced in the sm all creditor

qu a lified mortgage prov isions. See 12 C.F.R. § 1 026.43(e)(5)(i)(D); (e)(6)(i)(B); (f)(1)(vi).

119

A census block is the smallest geographic area for which the U.S. C ensus Bu reau collects and t abulates decennial

census data. See 80 Fed. Reg. 5 9943, 59956 (Oct. 2, 2015).

1 20

1 2 C .F.R. § 1 026.35(b)(2)(iv)(A)(2); com ment 12 C.F.R. § 1 026.43(f)(1)(vi)–1.

1 21

81 Fed. Reg. 1 6074 (Mar. 25, 2016).

1 22

Pu b. L. 114–94, section 89003 (2015).

1 23

TILA section 1 29C(b)(2)(E)(iv)(I), (codified as amended at 15 U.S.C. § 1 639c(b)(2)(E)(iv)(I)).

1 24

1 2 C .F.R. § 1 026.35(b)(2)(ii)(A); comments 12 C.F.R. § 1 026.43(f)(1)(vi)–1 ; 12 C.F.R. § 1 026.43(f)(2)(ii)–1.

42 BUREAU OF CONSUMER FINANCIAL PROTECTION

2.3 Overview of ATR/QM Rule requirements

2.3.1 Scope of the ATR/QM Rule

This section describes the scope and major substantive provisions of the ATR/QM Rule. With

certain exceptions, the ATR/QM Rule applies to any consumer credit transaction that is secured

by a dwelling.

1 25

The Rule does not apply to an extension of credit primarily for a business,

commercial, or agricultural purpose, even if it is secured by a dwelling.

1 26

As noted above, TILA

excludes from coverage open-end home equity lines of credit, timeshare plans, reverse

mortgages, and temporary loans with terms of 12 months or less.

1 27

In addition, the May 2013 final rule provided exemptions from the ability-to-repay requirements

for programs administered by housing finance agencies; creditors designated as Community

Development Financial Institutions, Downpayment Assistance through Secondary Financing

Providers, or Community Housing Development Organizations; certain nonprofit creditors;

certain homeownership stabilization and foreclosure prevention programs; and certain Federal

agency and GSE refinancing programs.

1 28

These exemptions addressed concerns that the

ATR/QM Rule’s ability-to-repay requirements were substantially different from the

underwriting requirements employed by these creditors or required under these programs.

Without an exemption, creditors might have been may be discouraged from participating in

these programs and significantly impair access to credit for consumers under these programs.

1 29

2.3.2 Major provisions of the ATR/QM Rule

This section describes the major topics addressed in the ATR/QM Rule. As indicated, many of

the requirements in the Rule, which was promulgated to implement Dodd-Frank Act

amendments to TILA, mirror the statute.

1 25

1 2 C.F.R. § 1 026.43(a)(1)–(3).

1 26

See Comment 1 2 C.F.R. § 1 026.43(a)–1.

1 27

See, supra note, 92.

1 28

1 2 C.F.R. § 1 026.43(a)(3)( iv ) –(vi).

1 29

See 7 8 Fed. Reg. 35430, 35440 (June 12, 2013).

43 BUREAU OF CONSUMER FINANCIAL PROTECTION

Ability-to-Repay provisions (§ 1026.43(c))

To implement TILA section 129C(a), 12 C.F.R. § 1026.43(c)(1) provides that a creditor shall not

make a loan that is a covered transaction unless the creditor makes a reasonable and good faith

determination at or before consummation that the consumer will have a reasonable ability to

repay the loan according to its terms.

Eight factors. In making the repayment ability determination, creditors generally must

consider, at a minimum, eight underwriting factors:

130

(i) current or reasonably expected income or assets, other than the value of the dwelling,

including any real property attached to the dwelling, that secures the loan;

(ii) current employment status, if the creditor relies on income from employment in

determining repayment ability;

(iii) the monthly payment on the covered transaction;

(iv) the monthly payment on any simultaneous loan(s) that the creditor knows or has

reason to know will be made;

(v) the monthly payment for mortgage-related obligations;

(vi) current debt obligations, alimony, and child support;

(vii) the monthly debt-to-income ratio or residual income; and

(viii) credit history.

Verification. Creditors generally must verify the information that they will rely upon in

determining a consumer's repayment ability, using reasonably reliable third-party records

specific to the individual consumer.

131

For example, a creditor must verify the amounts of

income or assets relied on to determine a consumer’s ability to repay the loan using third-party

records that provide reasonably reliable evidence of a consumer’s income or assets.

132

Payment calculation. Monthly payments on the loan must generally be calculated by assuming

that the loan is repaid in substantially equal monthly payments during its term.

133

For

adjustable-rate mortgages, the monthly payment must be calculated using the fully indexed rate

130

1 2 C .F.R. § 1 026.43(c)(2).

131

12 C.F.R. § 1 026.43(c)(3); com ment 12 C.F.R. § 1 026. 43(c)(3)–1.

131

1 2 C.F.R. § 1 026.43(c)(3); comment 12 C.F.R. § 1 026. 43(c)(3)–1.

132

1 2 C .F.R. § 1 026.43(c)(4).

133

1 2 C .F.R. § 1 026.43(c)(5)(i)(B).

44 BUREAU OF CONSUMER FINANCIAL PROTECTION

or an introductory rate, whichever is higher.

134

Special payment calculation rules apply for loans

with balloon payments, interest-only payments, or negative amortization.

135

Loans with such

features are not prohibited under the ability-to-repay standards, which were intended to provide

flexibility in underwriting standards so that creditors could adapt their underwriting processes

to a consumer’s particular circumstances.

136

Qualified Mortgage provisions (§ 1026.43(e)(1) through (3))

To implement TILA section 129C(b), 12 C.F.R. § 1026.43(e) and (f) provide for a class of

“qualified mortgage” loans, for which compliance with the ability-to-repay requirement is

presumed.

137

Presumption of compliance. T he Dodd-Frank Act provides that ‘‘qualified mortgages’’ are

entitled to a presumption that the creditor making the loan satisfied the ability-to-repay

requirements, but it does not specify whether the presumption of compliance is conclusive (i.e.,

creates a safe harbor) or is rebuttable. Under the ATR/QM Rule, the presumption of compliance

can be either conclusive, i.e., a safe harbor, for QM loans that are not “higher-priced,” or

rebuttable, for most QM loans that are “higher-priced.”

138

Generally, if the annual percentage

rate (APR) of a qualified mortgage exceeds the average prime offer rate (APOR)

139

for a

comparable loan product by 1.5 or more percentage points for a first-lien covered transaction,

the loan is a higher-priced covered transaction (HPCT)

1 40

and a rebuttable presumption

qualified mortgage.

1 41

The 1.5 percent limit is raised to 2.5 percent in the case of a subordinate-

134

1 2 C .F.R. § 1 026.43(c)(5)(i)(A).

135

1 2 C .F.R. § 1 026.43(c)(3).

136

See 7 8 Fed. Reg. 6408, 6460 (Jan. 30, 2013).

137

TILA section 129C(b) (codified as amended at 15 U.S.C. § 1639c(b)); 12 C.F.R. § 1 026.43(e).

138

1 2 C.F.R. § 1 026.43(e)(1).

139

Th e ATR/QM Ru le relies upon the definition of “APOR.” See 1 2 C.F.R. § 1026.35(a)(2). “Average prime offer rate”

m eans an annual percentage rate that is derived from average i nterest rates, points, and other loan pricing t erms

cu rrently offered to consumers by a representative sample of creditors for mortgage transactions that have low-risk

pr icing characteristics. The Bu reau publishes average prime offer rates for a b road range of ty pes of transactions in

a ta ble updated a t least w eekly a s well a s the m ethodology the Bureau uses t o derive these rates.

1 40

1 2 C .F.R. § 1 026.43(b)(4).

1 41

1 2 C .F.R. § 1 026.43(e)(1)(ii)(A).

45 BUREAU OF CONSUMER FINANCIAL PROTECTION

lien qualified mortgage. If a qualified mortgage is not an HPCT, it is a safe harbor qualified

mortgage.

1 42

Product and cost restrictions. The ATR/QM Rule defines QM loans in part by establishing

restrictions on product features and costs. Specifically, restrictions on product features generally

include prohibitions against negative amortization, balloon payments, interest-only

payments,

1 43

and terms greater than 30 years.

1 44

In addition, the total points and fees (certain

charges in connection with the loan’s origination) payable in connection with a QM Loan must

not exceed a certain percentage of the loan amount. The ATR/QM Rule establishes five tiers of

points and fees limits, based on loan size, with higher points and fees permitted for smaller

loans. These tiers range from three percent for loans of $100,000 or more to eight percent for

loans under $12,500.

1 45

Categories of Qualified Mortgages (§ 1026.43(e)(4) through (6))

General QM loans. One category of qualified mortgages is referred to as “General QM” loans. In

addition to complying with the product and cost restrictions noted above, for a loan to be a

General QM loan, a creditor must:

Underwrite the loan taking into account the monthly payment on the loan calculated by

using the maximum rate during the first five years after the date on which the first

regular periodic payment will be due and a payment schedule that will repay either

(i) the outstanding principal balance over the remaining term of the loan as of the date

that the interest rate adjusts to the maximum rate (and assuming the consumer will have

made all required payments as due prior to that date); or (ii) the loan amount over the

loan term;

1 46

Consider and verify at or before consummation the consumer’s current or reasonably

expected income or assets other than the value of the dwelling (including any real

property attached to the dwelling) that secures the loan, current debt obligations,

alimony, and child-support obligations, in accordance with Appendix Q, which sets

1 42

1 2 C .F.R. § 1 026.43(e)(1)(i).

1 43

1 2 C .F.R. § 1 02 6.43 (e)(2)(i). H owever, sm all creditors that operate in rural and underserved areas m ay make QM

loan s with balloon payment features. See 12 C.F.R. § 1 026.43(f).

1 44

1 2 C .F.R. § 1 026.43(e)(2)(ii).

1 45

1 2 C .F.R. § 1 026.43(e)(2)(iii) and (3). The threshold amounts are adjusted annually.

1 46

1 2 C .F.R. § 1 026.43(e)(2)(iv).

46 BUREAU OF CONSUMER FINANCIAL PROTECTION

standards for determining the “debt” and “income” that may be used for General QM

loan purposes;

1 47

Ensure that the ratio of the consumer’s total monthly debt to total monthly income at the

time of consummation, as determined in accordance with appendix Q, does not exceed

43 percent (DTI ceiling).

1 48

1 49

Temporary GSE QM loans. The ATR/QM Rule provides a separate, temporary, qualified

mortgage category for loans eligible to be purchased or guaranteed by either the Federal

National Mortgage Association or the Federal Home Loan Mortgage Corporation (collectively,

the GSEs) while they operate under Federal conservatorship or receivership or until January 10,

2021, whichever is earlier (‘‘Temporary GSE QM’’ loans).

150

The product and points and fees cost

restrictions that generally apply to qualified mortgages must be followed, but the GSE

underwriting standards generally are used instead of the General QM standards, which include

Appendix Q and the 43 percent DTI ceiling.

Temporary Federal Agency QM loans. The ATR/QM Rule also provided a temporary category

of QM loans for loans eligible to be insured or guaranteed by the U.S. Department of Housing

and Urban Development (FHA Loans); guaranteed by the U.S. Department of Veterans Affairs

(VA Loans); guaranteed by the U.S. Department of Agriculture (USDA Loans); or insured by the

Rural Housing Service (RHS Loans) (collectively, ‘‘Temporary Federal Agency QM” loans).

151

The category of Temporary Federal Agency QM loans no longer exists and has been replaced by

the category of Federal Agency QM loans because since 2014 the relevant Federal agencies (i.e.,

FHA, VA, and USDA/RHS) have all issued their own qualified mortgage rules

152

as permitted by

T ILA.

153

Because these Federal Agency QM rules are neither rules nor orders adopted by the

1 47

1 2 C .F.R. § 1 026.43(e)(2)(v).

1 48

1 2 C.F.R. § 1 026.43(e)(2)(vi). The monthly debt obligation must include the monthly payment for mortgage-

related obligations and any simultaneous loan the creditor knows or has reason to know will be made.

1 49

In establishing the DTI ceiling, the Bureau stated it “believes, based upon its r eview of the data it has obtained and

t h e c om ments r eceived, that the use of t otal debt-to-incom e as a qu alified m ortgage criterion prov ides a w idespread

a n d useful measure of a consumer’s abi lity to r epay, and that the Bureau should exercise its authority to adopt a

specific debt-to-income ratio that must be met in order for a loan to meet the requirements of a qualified mortgage.

T h e Bureau believes that the qualified mortgage criteria should include a standard for ev aluating w hether

con sumers have the ability t o repay their m ortgage loans, in addition t o the product feature requirem ents specified

in the statute.” See 78 Fed. Reg. 6408, 6526 (Jan. 30, 2013).

150

1 2 C .F.R. § 1 026.43(e)(4)(ii)(A).

151

1 2 C .F.R. § 1 026.43(e)(4)(ii)(B)–(E).

152

See, e .g., 24 C.F.R. § 2 03.19 (HUD rules).

153

TILA section 1 29C(b)(3)(B)(ii), added by section 1412 of the Dodd-Frank Act.

47 BUREAU OF CONSUMER FINANCIAL PROTECTION

Bureau under Federal consumer financial law, their effectiveness is beyond this assessment’s

scope.

Small Creditor Portfolio QM loans. The ATR/QM Rule permits small creditors, defined as

creditors that fall below certain assets and originations thresholds,

154

to make “Small Creditor

Portfolio QM” loans. Such loans must generally conform to all of the requirements of General

QM loans but do not have to follow Appendix Q and are not subject to the 43 percent DTI

ceiling. These loans must be held in portfolio, generally for a minimum of three years, to

maintain their qualified mortgage status. The APR over APOR safe harbor limit is increased for

these loans from 1.5 percentage points to 3.5 percentage points, making it easier for loans made

by small creditors to qualify for the safe harbor.

155

Rural/Underserved Small Creditor Balloon Payment QM loans. Although generally excluded

from being qualified mortgages, balloon payment loans can be qualified mortgages if made by

small creditors that fall below certain asset and origination thresholds and that operate in rural

and underserved areas.

156

These “Small Creditor Balloon Payment QM” loans are only eligible

for qualified mortgage status if certain product and cost restrictions that generally apply to

qualified mortgages are followed and if they have a term of at least five years and a fixed interest

rate. Income and debt must be considered and verified, and the consumer's monthly debt-to-

income ratio or residual income must be considered, but the standards in Appendix Q and the

43 percent DTI ceiling do not apply.

157

Except in limited circumstances, a Small Creditor Balloon

Payment QM will lose its QM status if, post consummation, it is sold, assigned, or otherwise

transferred to another person within three years of consummation.

158

As with Small Creditor

Portfolio QM loans, the APR over APOR safe harbor limit is increased for these loans from 1.5

percentage points to 3.5 percentage points.

159

Temporary Small Creditor Balloon Payment QM loans. The ATR/QM Rule also included a

temporary qualified mortgage category for small creditors that fall below certain asset and

origination thresholds. The “Temporary Small Creditor Balloon Payment QM” loan category was

temporary, providing a two-year transition period through April 1, 2016, during which small

154

Th e assets and or iginations thresholds are in 12 C.F.R. § 1 026.35(b)(2)(iii)(B) and (C), respectively.

155

1 2 C.F.R. § 1 026.43(e)(5).

156

1 2 C .F.R. § 1 026.35(b)(2)(iii)(A) through C, cross-referenced in 1 2 C.F.R. § 1 026.43(f)(1)(vi).

157

1 2 C.F.R. § 1 02 6.43 (f) .

158

1 2 C.F.R. § 1 026.43(f)(2).

159

T h ese Sm all C reditor Ba lloon Payment QM prov isions are prov ided in 1 2 C.F.R. § 1 02 6.43 (f).

48 BUREAU OF CONSUMER FINANCIAL PROTECTION

creditors that did not operate predominantly in rural or underserved areas could make balloon-

payment qualified mortgages if they held the loans in portfolio and otherwise followed the

balloon-payment qualified mortgage requirements applicable to creditors operating in rural or

underserved areas.

1 60

12 C.F.R. § 1 026.43 (e)(6). The requirem ent to “ operate predom inantly in rural or underserved areas” did not affect

th e ability of small creditors to make balloon payment QMs because the Temporary Small Creditor Balloon Payment

QM loan category that allowed all sm all creditors to make balloon-payment QMs was in effect until the

“ pr edom inantly” requirem ent was dropped in the Ma rch 2016 interim final rule.

49 BUREAU OF CONSUMER FINANCIAL PROTECTION

3. Market overview

This chapter provides background on the mortgage market and the economy as relevant to the

Bureau’s assessment of the ATR/QM Rule. It starts by providing an overview of the

development of the modern mortgage market starting around the Great Depression. The chapter

next focuses on the expansion in the mortgage market that started in the early 2000s. Next

comes a brief discussion of the subsequent mortgage market contraction, financial crisis, and

Great Recession.

1 61

The chapter then turns to the moderate economic recovery that took place

leading up to the implementation of the Rule. The chapter finally describes relevant dimensions

of the mortgage market shortly before and after the implementation of the Rule and compliance

with the Rule. The measures based on aggregate market data described in this section give a first

take on any effects on the market the Rule may have had. These complement and anticipate the

analyses in Chapters 4 through 8 which analyze effects of the Rule or specific provisions of the

Rule (e.g. QM provisions) on narrower segments of the mortgage market. At all points in the

chapter, the dynamics of relevant variables are presented in figures often covering a long time

range. These figures are discussed in multiple steps in the various sections covering different

time periods.

The main themes emerging from this chapter are as follows:

• The roots of the modern mortgage market can be traced to the Great Depression, after

which housing finance innovations made mortgages more available and affordable, and

World War II, after which the housing market went through two decades of expansion.

• Another robust expansion of the mortgage market started around 2000. This more

recent expansion saw an increased share of lending going to borrowers of lower

creditworthiness and to newer loan product types associated with higher risk.

1 61

T h e National Bu reau of Econom ic Research considers the most r ecent recession t o have lasted from December

2 007 to June 2009. See Nat’l Bureau of Econ. Research, US Business Cycle Expansions and Contractions,

h t tp://www.nber.org/cycles.html (last visited Dec. 17, 2018).

50 BUREAU OF CONSUMER FINANCIAL PROTECTION

• In 2006 and 2007, the performance of loans originated became worse and worse. House

prices started a significant correction in 2007 and the US economy experienced its most

severe recession since the Great Depression. The ensuing reduction in originations was

especially stark amongst certain loan products associated with higher risk and among

lower credit score borrowers.

• The path to economic recovery after the recession ended in June 2009 was slow and the

recov ery in the housing market was particularly slow. Between 2011 and the

implementation of the ATR/QM Rule, the volume of mortgage lending gradually

increased but credit remained tight in 2013, the last year prior to the Rule.

• Many trends in the mortgage market evolved smoothly around the time of the Rule’s

implementation. This includes the volume of mortgage applications, the approval rate of

these applications, the spread of the average interest rate on fixed-rate mortgages over

the relevant Treasury rate, and the revenues and expenses associated with originating a

mortgage loan reported by non-depository lenders. There was an increase in the share of

purchase originations sold to the GSEs before the Rule took effect, although this share

did not shift appreciably in the years following the Rule’s implementation. There was an

increase in the share of jumbo loans and a reduction in the spread between the cost of

jumbo and conforming loans following the effective date of the Rule, although both of

those effects are likely attributable to market forces rather than the Rule.

3.1 The development of the modern

mortgage market

162

As highlighted already in Chapter 1, the mortgage market is the single largest market for

consumer financial products and services in the United States with approximately $10.7 trillion

in consumer mortgage loans outstanding as of mid-2018. Figure 1 plots mortgage debt

1 62

The discussion in this section r elies on several background sources, including: Nat’l Bureau of Econ. Research,

Housing and Mortgage Markets in Historical Perspective, (Eugene N. White, et al., eds., Univ. Of Chi. Press 2014);

Da n i el K. Fetter, How Do Mortgage Subsidies Affect Ho me Ow nership? Ev idence from the Mid-Century GI Bills, 5

A merican Econ. J. 111 (2013); Richard K. Green & Susan M. Wachter, The American Mortgage in His toric al and

International Context, 1 9 J. of Econ. Persp. 93 (2005);Edward M. Gramlich, Subprime M o rtgages: Am erica’s

Lates t Boom and Bust, (Urban Inst. Press 2007); N. Er ic W eiss & Ka tie Jones, Ove rview of the Ho using Finance

System in the United States (Jan. 2017) (CRS Report) (report on the housing finance sy stem in the United States

pr epared for m embers and c ommittees of Congress).

51 BUREAU OF CONSUMER FINANCIAL PROTECTION

outstanding as a share of personal income since 1949.

1 63

The significant growth in mortgage

holdings as a share of personal income up to mid-1960s can be attributed to New Deal policies

that promoted homeownership and the post-war housing boom that occurred between 1945 and

1960.

FIGURE 1: MORTGA GE DEBT OUTSTA NDING AS SHA RE OF PERSO NA L INCOME, 1949-2017

Between the 1940 and 1960 censuses, the homeownership rate in the United States increased

from 43.5 to 61.9 percent. To show which households were affected the most by the expansion of

mortgage credit, Figure 2 shows the homeownership rate by family income quartile since 1940

1 63

See FRED Ec on om ic Da ta, Mortgage Debt Outstanding by Type of Property: One- to Four-Family Residences,

Fed. Reserve Bank of St. Louis, https://fred.stlouisfed.org/series/MDOTP1 T4FR (last v isited Dec. 17, 2018); FRED

Econ om ic Data, Mortgage Debt Outstanding by Type of Property: Multifamily Residences,

h ttps://fred.stlouisfed.org/series/MDO T P1T4 FR (last visited Dec . 17, 2 018); FRED Econ om ic Da ta, Personal

Inco m e, Fed Reserve Bank of St. Louis, https://fred.stlouisfed.org/series/A065RC1A027NBEA ( last visited Dec . 17,

2 018) (Personal incom e is the incom e that persons r eceive in return for their prov ision of labor, land, and capital

a n d the n et current transfer payments that they receive from business and from gov ernment.).

52 BUREAU OF CONSUMER FINANCIAL PROTECTION

using publicly available Census Bureau data.

1 64

Between 1940 and 1960, there is substantial

growth in homeownership that is most pronounced for the top two income quartiles. The top

income quartile’s homeownership rate grows from 49.6 percent to 77.4 percent over this period,

while the growth for the second income quartile is from 42.1 percent to 66.4 percent. Growth for

both of these groups continues up until 1980, at which point the homeownership rate of the two

groups is 87.3 percent and 72.7 percent, respectively. Subsequently, the homeownership rate

stays relatively stable for these two groups.

1 64

W e c a lculate the hom eow nership rate a s the share of h eads of h ousehol d at least 1 8 y ears of a ge who report t hat

th ey own their housing unit outright or are in the process of buying it. Da ta com e from the In tegrated Public Use

Mic r odata Series ( IPUMS), w hich provides harmonized C ensus Bureau microdata. For 1 940 to 2 000, m icrodata

fr om the Dec ennial Census are used. H om eownership data are n ot available for t he 1 950 Census. Du e t o data

limitations, microdata for 2010 come from the American Community Survey (ACS). T he AC S is a survey m anaged

by the Census Bureau, w hich uses a representative sample of the US population. See Steven Ruggles, Sarah Flood,

Ron a ld Goeken, Josiah Grov er, Erin Mey er, Jose Pa cas, and Matthew S obek. IPUMS USA, U.S. Census Data for

So c ial Eco nomic, and Health Research, https://doi.org/10.18128/D010.V8.0 (last v isited Dec . 17, 2018). Family

in com e is n ot reported in the 1940 c ensus. Family incom e for that year i s proxied by the sum of in dividual wage

income in the family and is imputed for families with no reported wage income based on educational attainment,

soc ioec onom ic index, occupational incom e score, and the presence of chil dren in the household.

53 BUREAU OF CONSUMER FINANCIAL PROTECTION

FIGURE 2: HOMEOW NE RS HI P RA TE BY INCOME QUA RTIL E, 1940-2010

The third income quartile also experiences increases in the homeownership rate, but at a lesser

pace than the top two groups. Finally, homeownership rates have decreased at the bottom of the

income distribution. The homeownership rate of this group was 50.8 percent in 1960 and

dropped to 42.8 percent by 2000.

This growth in homeownership was facilitated by innovations in housing finance. In response to

the wave of foreclosures that accompanied the Great Depression, in 1934 the government

established the Federal Housing Administration (FHA) to provide mortgage insurance nec essary

for investors to purchase mortgages with confidence. By creating the standards that loans had to

meet to be insured by the FHA, the modern American mortgage was created with minimum

quality standards, full amortization, a long (eventually 30 year) term which substantially

reduced monthly payments and the risk of default for borrowers as compared to earlier loan

products that were non-amortizing, had shorter maturity periods, and most often had balloon

payments. During the 1930s, similar programs were established by the United States

Department of Agriculture (USDA). Following the war, the Veterans Administration (VA) also

created a mortgage insurance program similar to that of the FHA in order to serve the needs of

returning soldiers. Compared to products before the Great Depression which often limited loan-

54 BUREAU OF CONSUMER FINANCIAL PROTECTION

to-value ratios to at most 50 percent, the new products also allowed for higher loan-to-value

ratios, thereby making mortgages affordable for more households. Over time, 20 percent arose

as the typical downpayment for conventional mortgages;

1 65

programs through the FHA and VA

sometimes allowed even smaller downpayments.

Most mortgages prior to the Great Depression were funded using lenders’ funds, known as

portfolio lending. The Federal National Mortgage Association (Fannie Mae) was created in 1938

to purchase FHA-insured loans, pool them, and sell them as securities to investors on financial

markets as residential mortgage backed securities (RMBS). This created the secondary mortgage

market and gave lenders a new source of capital. In the process Fannie Mae mandated certain

lending practices. If lenders didn’t meet Fannie Mae’s guidelines about underwriting practices

or other loan terms and lending practices, then their loans would not be packaged as securities.

As part of the Housing and Urban Development Act of 1968, Fannie Mae was split into two

entities: Ginnie Mae and the “new” Fannie Mae. Ginnie Mae was established as a government-

owned entity that provides an explicit government guarantee of timely payment for RMBS

backed by federally insured or guaranteed loans—loans insured or guaranteed by the FHA, the

VA, and the USDA.

1 66

Fannie Mae in contrast became a publicly-traded company.

In 1970, Congress created Federal Home Loan Mortgage Corporation (Freddie Mac), which

operated similarly to Fannie Mae. In 1972, Fannie Mae and Freddie Mac both began to purchase

conventional mortgages that were not guaranteed or insured by the FHA, VA, or USDA; high

leverage conventional loans could be insured instead by Private Mortgage Insurance (PMI)

companies. By the mid-1980s funds provided through the securitization of mortgages in the

secondary market had overtaken depository portfolio funding as the primary source of mortgage

capital.

Fannie Mae and Freddie Mac are jointly known as the Government Sponsored Enterprises

(GSEs). Although their securities are not explicitly backed by the government, most investors

have long believed that the government would not allow them to default on their obligations.

1 67

1 65

A conventional mortgage loan is on e that is not insured or guaranteed by the federal gov ernment, including the

Federal Housing Administration (FHA), the U.S. Department of Veterans Affairs (VA), or the USDA’s Farm Service

Agency or Rural Housing Service (FSA/RHS). Conventional loans are either private or guaranteed by one of the two

G ov ernment Sponsored En terprises ( GS Es), the Federal National Mortgage Association (Fannie Ma e) and the

Federal Home Loan Mortgage Corporation (Freddie Mac).

1 66

Gi nnie Ma e, 50 Years of Ginnie Mae: We Make Affordable Ho using a Reality, (Oct. 2018), available at

h t tps://w ww .ginniem ae.gov /new sroom /m ediaresources/Documents/about_ginniem ae.pdf.

1 67

Con gressional Bu dget O ff., Fa nn ie Mae, Freddie Mac, and the Federal Role in the Secondary Mortgage Market,

(Dec. 2010), available at https://ww w.cbo.gov /publication/21992.

55 BUREAU OF CONSUMER FINANCIAL PROTECTION

This belief proved to be true during the financial crisis and is probably felt even more strongly

today with the GSEs in government conservatorship.

FIGURE 3: SHA RE OF MORTGAGE DEBT OUTSTA NDI NG FUNDED BY PRIV A TE SECUR I T IZ A T ION, 1980-

2017

Finally, private companies also issue mortgage backed securities.

1 68

In contrast to Agency

(Ginnie Mae or GSE) RMBS, these private label securities (PLS) have no government guarantee.

As a result, investors see these securities as riskier than Agency RMBS. Despite this, the share of

outstanding mortgage debt accounted for by PLS grew gradually over the next two decades as

can be seen in Figure 3, reaching 7.5 percent by 2000.

These developments then set the stage in the mortgage market for the long expansion of the

1990s and beyond.

1 68

The first private label mortgage backed security was issued by Ba nk of America in 1 977.

56 BUREAU OF CONSUMER FINANCIAL PROTECTION

3.2 Early 2000s mortgage market expansion

The recession of 2001 ended a decade of economic growth, but it was brief and shallow.

1 69

Its

effect in the mortgage market was limited. The share of personal income accounted for by

outstanding mortgage debt hovered around 55 percent in the decade before 1998 and then

began a climb that was rapid and was not slowed down by the recession. By 2007, total

mortgage debt outstanding as a share of personal income stood at 93.9 percent. No previous

period has experienced the same rapid growth in household housing leverage and the only

period that came close was that of the post-war expansion. T he expansion of the 2000s was

markedly different than the post-war expansion, however, as demonstrated by Figure 2, since

the more recent expansion was much less concentrated among high-income households.

Another new development of this era was the significant growth in funding accounted for by

private securitization (see Figure 3), the share of which reached 20.9 percent by 2006. All in all,

the types of mortgage products available, the types of borrowers participating in the market, and

the type of funding available during the post-war period were all much more limited compared

to those observed since 2000.

Figure 4 shows the number of purchase and refinance mortgage originations for each year from

1998, the first year of full NMDB coverage, to 2016, the last year of full NMDB coverage.

170

The

number of originations declined in 2000 compared to the late 1990s, especially among

refinances, but the market quickly recovered and went through an unprecedented refinance

boom in 2003. This was partly attributable to the large drop in interest rates that took place at

the time. Figure 5 shows the average 30-year fixed mortgage rate together with the benchmark

of the 10-year Treasury rate and the spread between the two. Following 2003, the number of

refinance originations dropped, while the number of purchase originations continued growing

through 2005.

1 69

T h e recession officially lasted from Ma rch 2001 t o Nov ember 2001. See Kevin L. Kliesen, The 2001 Recession:

How was it different and what developments may have caused it?, ( Federal Reserve Ba nk of St. Louis, 2 003)

available at https://files.stlouisfed.org/files/htdocs/publications/review/03/09/Kliesen.pdf.

170

In Figure 4, the sam ple is restricted to first-lien originations since only those appear in the NMDB.

57 BUREAU OF CONSUMER FINANCIAL PROTECTION

FIGURE 4: NUMB ER OF PURCHA S E AND REFINA NCE ORIGINA TIONS, 1998-2016

58 BUREAU OF CONSUMER FINANCIAL PROTECTION

FIGURE 5: 30-YEAR AVERAGE FIXED MORTGA GE RA TE, 10-Y EAR CONSTA NT MA TURITY TREA SURY

RA TE, A ND THEIR SPREA D, 1998-2017

The early 2000s saw some changes in the mix of lending by creditworthiness as measured by a

borrower’s credit score. Figure 6 shows the share of total amount of mortgage lending (in

dollars) accounted for by the various credit score groups.

171

The share of lending accounted for

by lower credit score groups expanded in the last two years of the 1990s. The 2001 recession led

to a temporary reversal in this trend, but by 2004 the share of lending accounted for by

borrowers with a credit score below 680 was 36.4 percent and this share peaked in 2006 at 41.3

percent.

171

T h e credit score used is the VantageScore 3.0, which is what i s available in the NMDB. FICO sc ores have been

m or e c om monly used for underwriting during the period of study. The distribution of Vantage scores in the NMDB

h a s a t hicker left tail than the distribution of FICO sc ores in the CoreLogic data, so t here are relatively m ore low

sc or e borrowers using t he Vantage score.

59 BUREAU OF CONSUMER FINANCIAL PROTECTION

FIGURE 6: DISTRIB UT IO N OF ORIGINA TED MORTGA GE DEBT BY CREDIT SCORE GROUP, 1998-2016

There was also a shift in the type of loan products originated. Figure 7 shows the distribution of

originations by loan product type. While there is a multitude of products and product features

that existed and were introduced into the market at the time, for sake of exposition, this figure

distinguishes between three broad groups of loan products. Traditionally, fixed-rate mortgages

with a loan-to-value (LTV) ratio of 80 percent or less have been the most common and have

been considered the least risky. These are referred to as “Low-leverage fixed-rate loans.” On the

other end are products that have features that turned out to be so highly correlated with default

risk that the Rule generally does not provide them with a presumption of compliance with the

ability-to-repay requirement or otherwise limits them. These features are interest-only, negative

amortization, term over 30 years, and balloon loans (restricted to non-QM loans by the Rule)

172

and ARMs with reset periods under five years (limited by the Rule given its payment calculation

172

Ba lloon loans are allowed as a QM for sm all, rural lenders.

60 BUREAU OF CONSUMER FINANCIAL PROTECTION

provisions). These are referred to as “Restricted feature loans.”

173

T he remaining loans are

categorized as “Other loans” and consist of higher leverage fixed-rate loans, ARMs with longer

reset periods, and so on. As can be seen from the figure, there was a rise in loans with restricted

features and in other loans during the late 1990s, but this trend reversed temporarily with the

2001 recession. From 2000 to 2003 the share of low-leverage fixed-rate loans rose from 51.8

percent to 63.7 percent. Then the prevalence of restricted feature loans picked up, reaching a

peak of 30.3 percent in 2006. This followed the relaxation of underwriting standards, which

allowed borrowers to be approved in a short amount of time and with less documentation of

their ability to repay the loan.

173

No documentation loans are also restricted by the Rule, but the NMDB data do not distinguish these loans from

low documentation loans.

61 BUREAU OF CONSUMER FINANCIAL PROTECTION

FIGURE 7: DISTRIB UT IO N OF ORIGINA TIO NS BY LOA N PRODUCT TYPE, 1998-2016

Figure 8 shows the “early delinquency rate” by the loan product types used in Figure 7 and by

year of origination.

174

The early delinquency rate is measured as the percent of loans that

become 60 days or more past due within two years of origination. As expected, restricted feature

loans had the highest early delinquency rate while low-leverage fixed-rate loans had the lowest

early delinquency rate during the expansion. The performance of loans originated in the two

years subsequent to the 2001 recession was better than in 2001. Delinquency rates started

increasing in 2004, but it was not until 2005 that delinquency rates surpassed the levels of

2001. By the 2007 cohort of loans, the delinquency rate for each loan product type was more

than two and one half times its level in 2001.

174

Figure 8 starts in 2001, the first year that loan performance information is available in the NMDB.

62 BUREAU OF CONSUMER FINANCIAL PROTECTION

FIGURE 8: EARLY DELINQUENCY RATE BY LOA N PRODUCT TY PE, 2001-2016

Overall, the expansion of mortgage credit coincided with both an increased share of lending to

borrowers of lower creditworthiness and an increase in the share of loan product types

associated with higher risk (e.g. interest-only loans). While the shift toward less creditworthy

borrowers and riskier product types explain part of the rise in the overall early delinquency rate,

early delinquency rates within loan product type groups reported in Figure 8 (and even more

finely within loan product type and borrower credit score groups) also increased.

63 BUREAU OF CONSUMER FINANCIAL PROTECTION

3.3 Financial crisis and Great Recession:

2007-2009

In 2007, a significant correction in the housing market began. Between 2000 and the beginning

of 2007, prices of single-family homes rose 69.5 percent nationally as shown in Figure 9. House

prices peaked in March of 2007 and then started falling and ultimately fell 23.4 percent by the

time they hit their trough in January of 2012.

175

The declining value of borrowers’ collateral

partly contributed to the surge of delinquencies documented in Figure 8.

FIGURE 9: HOUSE PRICE INDEX ( JAN 2000=100) , JAN 2000 – DEC 2016

175

For a com parison of growth rates in m edian hom e prices and median rents, see the C ensus Bureau’s Quarterly

Residential Vacancies and Homeownership data series, available at

h ttps://www.census.gov/housing/hvs/files/currenthvspress.pdf.

64 BUREAU OF CONSUMER FINANCIAL PROTECTION

Purchase originations quickly declined from their previous levels starting in 2007 and the

contraction lasted through 2011. Refinance originations were also muted. The drop in refinances

between 2005 and 2008 was comparable to that experienced by purchases, but the refinance

market experienced a weak recovery by 2009. The contraction was especially stark amongst

loans with restricted features (see Figure 7)—almost no loans with such features were made by

2009—and among the lower credit score groups (see Figure 6).

FIGURE 10: TRA NSIT ION RA TE INTO 90+ DA Y S DELINQ UE NCY A ND FORECL O S U R E START RA TE, 2002 –

2016

Concurrently, the transition rate into serious delinquency (90 or more days past due or in

foreclosure) and the foreclosure start rate remained at elevated levels for several years. Figure

10 shows for each year the annual rate at which borrowers with existing mortgages transitioned

65 BUREAU OF CONSUMER FINANCIAL PROTECTION

into serious delinquency or had a foreclosure started.

176, 177

In 2007, the transition rate into

serious delinquency surpassed its previous high of 1.93 percent reached in 2003 and stayed

above that level through 2015, peaking at 6.54 percent in 2009. At the same time, the

foreclosure start rate reached 2.71 percent, almost triple its previous peak of 0.98 percent

reached in 2003. Simultaneously, as the new issuance of PLS all but faded after 2007,

178

the

share of outstanding mortgage debt funded by PLS gradually declined reaching less than half its

2006 share by 2012.

176

T h e quarterly transition rate into serious (90+ days) delinquency m easures the percent of a ll m ortgages not i n

ser ious delinquency at the end of a quarter in which the loans are reported to be seriously delinquent at the end of

t h e subsequent quarter. The quarterly foreclosure start rate m easures the percent of a ll m ortgages not i n foreclosure

h eld at the end of a quarter that are r eported t o b e in foreclosure at the end of t he subsequent quarter. T he annual

ra tes r eported are the sums of the quarterly rates during a year. Note that the foreclosure start rate is different from

t h e c om monly used forec losure (inventory) rate ( used, for example, by the Mor tgage Bankers of Am erica National

Deli n quency Survey), w hich is the share of m ortgages a t a point in time that are in foreclosure. An advantage of t he

for eclosure start rate is that, unlike the foreclosure inventory rate, it is not influenced by the length of time that a

m ortgage is in foreclosure which can v ary across st ates and ov er time due t o differences in state regulations and

ch anges in the speed of processing foreclosures ov er t ime. See Timothy Dunne & Guhan V enkatu, Foreclosure

Metrics , Ec on. C om mentary, Fed. Reserve Bank of Cleveland (Apr. 2009), available at

h t tps://w ww .clevelandfed.org/newsroom -and-events/publications/economic-com men tary /e con omi c-

commentary-arch iv es/2 009 -ec onom i c-com mentaries/ec-20090409-foreclosure-m etr ics.a spx (for a discussion of

v arious foreclosure metrics).

177

Note that the horizontal axis in Figure 10 is the year when the loan entered serious delinquency or foreclosure,

w h ile in Figure 8 it was the year the loan was or iginated.

178

La uri e G oodman, A Pro gress Report o n the Private-Label Securities Market, Hous. Fin. Pol’y Ctr. Commentary,

Ur b. Inst. (2016), available at https://www.urban.org/research/publication/progress-report-priv ate -label-

sec urities-m arket.

66 BUREAU OF CONSUMER FINANCIAL PROTECTION

FIGURE 11: CIV ILIAN UNEMPL OY MENT RATE, JAN 1998 – DEC 2017

The housing crisis soon was followed by a full financial crisis as the value of mortgage backed

securities and the derivative securities tied to them dwindled.

179

Ultimately, the US economy

experienced its most severe postwar recession, commonly known as the Great Recession.

1 80, 1 81

The economic effects were widespread and severe. The fall in housing prices is estimated to have

resulted in 7.4 trillion dollars of household wealth lost.

1 82

As shown in Figure 11, the

179

Ja m es Bu llard, et a l., Systemic Risk and the Financial Cris is : A Prim er, at 403–4 17, Fed. Reserve Bank of St .

Lou is Rev . (Sept./Oct., Part 1, 2009), available at

h ttps://files.stlouisfed.org/files/htdocs/publications/review/09/09/part1/Bullard.pdf.

1 80

See Federal Reserve Bank of Minneapolis Special Study: Recession in Pe rspectiv e, available at

h t tps://www.minneapolisfed.org/publications/special-studies/recession-in-perspective.

1 81

See Siems, s upra note, 9.

1 82

See FRED Ec on om ic Da ta, Households; Owners’ Equity in Real Es tate, Level,

h ttps://fred.stlouisfed.org/series/OEHRENWBSHNO (last visited Dec. 1 8, 2018). The loss was calculated as the

difference between 2006 Q1 peak and 2009 Q1 trough in households’ equity in real estate.

67 BUREAU OF CONSUMER FINANCIAL PROTECTION

unemployment rate reached 10 percent in October of 2009, levels that the US economy had not

experienced since 1983.

1 83

This was also a time of substantial household deleveraging as the

share of personal income held as mortgage debt fell 17 percentage points to reach 77.2 percent

by 2011 as shown in Figure 1.

3.4 Pre-Rule economic recovery: 2009-2013

The path to economic recovery after the recession ended in June 2009 was slow and uneven.

The unemployment rate started to slowly decline in November of 2009. Consumer spending

started to recover as early as the middle of 2009, while nonmortgage lending to consumers

began recovering in 2010.

1 84

In contrast to consumer spending and nonmortgage lending, recovery in the housing market

was much slower. The housing market remained depressed throughout 2009 and 2010. Over

this period, the market saw decline and then stagnation in new home construction and sales in

combination with an increase in real estate owned by lender (REO) and short sales.

1 85

As shown

in Figure 9, house prices continued falling through 2011 reaching their trough in December of

that year.

The housing market began showing signs of recovery starting in 2012. The recovery was

attributed, at least in part, to continued improvements in the labor market documented in

Figure 11, to historically low interest rates as shown in Figure 5, and to pent-up demand from

the post-recessionary period.

1 86

Loan modification programs also became available to aid

distressed borrowers.

1 87

House prices began to slowly increase in 2012 and experienced

1 83

See FRED Ec on om ic Da ta, Civilian Unemployment Rate, Fed. Reserve Ba nk of St. Louis,

h ttps://fred.stlouisfed.org/series/UNRATE/ (last v isited Dec . 18, 2018); FRED Econ om ic Research, 30-Y ear Fixe d

Rate Mortgage Average in the United States, Fed. Reserve Ba nk of St. Louis,

h t tps://fred.stlouisfed.org/series/MORTGAGE3 0US/ (last v isited Dec. 1 8, 2018).

1 84

See Fin. Stability Oversight Council, 2011 Annual Report, U.S. Dep’t of the Treasury (2011), available at

h t tps://www.treasury.gov /initiatives/fsoc/studies-reports/Pages/2011-Annu al-Report.aspx.

1 85

See Fin. Stability Oversight Council, 2014 Annual Report, U.S. Dep’t of the Treasury (2014), available at

h t tps://www.treasury.gov /initiatives/fsoc/studies-reports/Pages/2014-Annual-Report.aspx.

1 86

Id.

1 87

Ea r ly in the h ousing crisis, t he availability and terms of m ortgage m odification programs varied widely and oft en

failed to lower monthly payments for the borrower. As these early m ortgage modifications rarely improv ed

a ffor dability, the m odified loans were highly likely t o re-default. In response t o the fi nancial c risis, the federal

g ov ernment established programs a imed at assisting distressed hom eow ners. H om e Affordable Modification

Pr og ram (HAMP) was introduced i n March 2009, p rov iding i ncentive payments t o mortgage lenders, servicers,

borr owers, and investors for m odifying loans to conform to the HAMP guidelines. See Maximilian D. Schmeiser &

68 BUREAU OF CONSUMER FINANCIAL PROTECTION

substantial growth starting in 2013. T he number of purchase mortgage originations started

increasing again in 2012 as shown in Figure 4. In the same year refinance loans experienced an

upturn, which was largely attributed to low interest rates. The market for these loans cooled off

in the second half of 2013 as interest rates rose slightly.

1 88

Although the housing market was recovering, the experience of the financial crisis led to tighter

underwriting standards in mortgage lending compared to the standards used by some during

the preceding expansion.

1 89

By 2013, the number of purchase originations was well below the

levels observed in the late 1990s (Figure 4). High credit score group borrowers experienc ed a

much more robust recovery than those in lower credit score groups, for whom mortgage credit

availability was significantly lower than in the late 1990s (Figure 6). Borrowers with credit

scores above 760 accounted for close to 58 percent of originated dollars in 2013, compared to

just 25 percent in 2006 and 39 percent in 2003. Loans with restricted features all but

disappeared from the market, at 1.05 percent their share in 2013 was lower than their 1998

share at 1.63 percent. Note that the disappearance of these loans, which started in 2009, took

place prior to the effective date of the Rule’s requirements regarding restricted features.

All in all, between 2011 and the implementation of the ATR/QM Rule, mortgage lending

recovered somewhat from its trough but remained tight, especially for borrowers of lower

creditworthiness and those using riskier loan product types. Correspondingly, early delinquency

rates fell to below 2 percent for loans originated in 2011 and stayed below that level through

2016. The tightness of the mortgage market was also reflected in further household

deleveraging, with the share of personal income accounted for by mortgage debt falling to below

67 percent by 2014.

Ma t thew G r oss, Th e Determinants of Subprime Mortgage Pe rform ance Fo llow ing a Loan Mo dific ation, 52 J. of

Real Est . Fin. & Econ . 1 (2016).

1 88

Cer tain com menters suggested that the housing m arket recov ery has been weaker than the data examined by t he

Bu r eau suggest. See Appendix B.

1 89

See Laurie Goodman et a l., Where Have All the Loans Gone? The Impact of Credit Availability on Mortgage

Volume, Hous. Fin. Pol’y Ctr. Commentary (2014), available at

h t tps://www.urban.org/sites/default/files/publication/22386/413052-Where-Have-All-the-Loans-Gone-The-

Im pact-of-Credit-Availability-on-Mortgage-Volume.PDF; Laurie G oodm an et al ., The Impact of Tight Credit

Standards on 2009–13 Lending, Hous. Fin. Pol’y Ctr. Commentary, Urb. Inst. (2015), available at

h t tps://www.urban.org/sites/default/files/publication/48731/2000165-The-Im pact-of-Tight-Credit-Standards-on-

2 009-13-Lending.pdf.

69 BUREAU OF CONSUMER FINANCIAL PROTECTION

3.5 Mortgage market pre- and post-Rule

This section provides additional observations on the mortgage market shortly before and after

the implementation of the Rule using more refined aggregate market data in order to highlight

possible effects of the Rule. It considers applications and approval rates, breaks down mortgage

originations by loan size and purchaser type, shows interest rate trends by loan size, and

discusses lenders’ costs of and revenues from mortgage origination over time. The purpose of

the present section is to assess whether the implementation of the Rule had a large, discrete

effect on the market that would be apparent in these aggregate trends. Later chapters included

in this assessment analyze narrower segments of the mortgage market that may have been most

directly affected by the Rule or particular requirements of the Rule.

FIGURE 12: PURCHA S E MORTGA G E A PPLICA T I O NS A ND SHA RE OF A PPROVALS, JAN 2010 – DEC 2016

70 BUREAU OF CONSUMER FINANCIAL PROTECTION

An imperfect but measurable correlate of mortgage loan demand is the number of applications

made for a mortgage loan.

1 90

The share of applications that are approved as opposed to denied

(or withdrawn), in turn, reflects the considerations of lenders in the market and of investors who

purchase loans from lenders. Examining these two drivers of mortgage market outcomes can

signify how borrower and lender behavior are changing in the market. Figure 12 reports the

number of purchase mortgage applications on the left axis and the percent of such applications

approved on the right axis as reported under HMDA between 2010 and 2016, both seasonally

adjusted.

1 91

There is no significant break in either applications or the approval rate around the

effective date of the Rule, implying that, at this aggregate level, neither demand nor supply were

significantly disrupted.

To examine any shifts in the distribution of loan size, Figure 13 shows the share of purchase

mortgage originations for loan size categories above the standard conforming limit over the

period 2010 to 2016 as reported under HMDA.

1 92

Super conforming loans are defined as loans

with a size above the standard conforming loan limit and up to the county-specific maximum

that are permitted in designated high-cost areas.

1 93

Jumbo loans are defined as loans that are

originated with values above either the standard conforming loan limit or the high-cost county

maximum, whichever is greater. Not shown in the figure are conforming loans, which account

for over 90 percent of mortgage originations through early 2013. The share of jumbo loans grew

by 64 percent between early 2013 and the end of 2016 while the growth in the share of super

conforming loans was more muted at 35 percent.

1 94

Despite the significant growth in the jumbo

1 90

It is im portant to note that applications cannot be taken to be a direct measure of demand. To the extent that

bor rowers anticipate variation in the approval rate, they may turn their latent demand for mortgage loans into

actual applications with different propensity depending on the approv al rate they anticipate. Effects of this type are

a n alyzed in Chapter 5.

1 91

In Figure 12, the measure of a pplications includes those applications that are ultimately originated, a pprov ed but

n ot a ccepted, denied, w ithdrawn by the applicant, c losed for incom pleteness, and purchased by an institution. The

sa m ple excludes pre-approval requests and is restricted to purchase applications for first-lien loans on single-family

r esidences. See U.S. Census Bureau, The X13 Arima-Seats Seasonal Adjustment Program, ( 2 014 –2017), available

at h t tps://www.census.gov/srd/www/x13as/ (seasona l a d ju stmen t i s p er forme d u sin g th e Census Bureau’s X13

seasonal adjustment program).

1 92

In Fig ure 13, the sample is r estricted to first-lien c onventional purchase originations.

1 93

Th e national conforming loan limit for mortgages for single-family one-unit properties was $417,000 for 2006-

2 0 08, with limits 50 percent higher for four statutorily-designated high cost areas: Alaska, Hawaii, Guam, and the

U.S. Virgin Islands. Since 2008, various l egislative a cts, including the Housing and Ec onom ic Recov ery Act of 2 008,

in creased the loan lim its in certain high-cost areas in the United States. See Fed. Hous. Fin. Agency, Conforming

Lo an Lim its , h ttps://ww w.fhfa.gov /DataTools/Downloads/Pages/Conforming-Loan-Lim its.aspx (last visited Dec .

1 8 , 2 018) (to determine the applicable limits, H MDA data are m atched at the year and county level t o the high-cost

cou nty limits).

1 94

Not e t hat t he sharp drop in the share of su per conforming loans and the equivalent increase in the share of jumbo

loans between 2010 and 2011 was due to the c ou nty-specific m axima being significantly lowered in sev eral c ounties

in 2 011. At all other times, county-specific maxima stayed the same or increased.

71 BUREAU OF CONSUMER FINANCIAL PROTECTION

share, later chapters using more refined analysis will examine whether the growth of jumbo

originations would have been even higher absent the Rule.

1 95

FIGURE 13: SHA RE OF ORIGINA T IONS BY LOAN SIZE, JAN 2010 – DEC 2016

At the other end of the loan size distribution, the cap on points and fees for qualified mortgages

introduced by the Rule could be binding. Figures 14 and 15 show the share of small loans among

purchase mortgage originations using loan size thresholds defining the Rule’s points and fees

cap, for site-built and manufactured home loans, respectively.

1 96

1 95

A s discussed further in C hapters 5 and 6, with regards to t he treatment of j umbo l oans under the A TR/QM Ru le,

th e primary difference is that they cannot qualify for Temporary GSE QM status.

1 96

In th ese figures, t he sample is restricted t o first-lien conventional purchase originations that are valued under

$170,000, the median loan size in 2011.

72 BUREAU OF CONSUMER FINANCIAL PROTECTION

FIGURE 14: SHA RE OF ORIGINA T IONS BY LOAN SIZE, SMALL SITE-BUILT HOME LOA NS, JA N 2010 – DEC

2016

Among site-built home loans, the share of loans under $170,000 was 47.1 percent at the

beginning of 2010 and declined to 34.4 percent by the end of 2016. This was largely due to a

combination of the price increases documented in Figure 9, borrowers purchasing larger homes,

and borrowers taking out loans with a higher loan-to-value ratio.

1 97

The Rule’s points and fees

cap may also have contributed to this trend, an issue further examined in Section 5.4.5.

1 97

The median loan-to-v alue ratio for site-built hom e loans increased from 76 percent to 8 0 percent ov er t he same

per iod.

73 BUREAU OF CONSUMER FINANCIAL PROTECTION

FIGURE 15: SHA RE OF ORIGINA T IONS BY LOAN SIZE, SMALL MA NUFA CT UR ED HOME LOANS, 2010 Q1 –

2016 Q4

Among manufactured home loans, there was a distinct shift from loans under $60,000 to loans

over this value. The share of loans under $60,000 decreased from 68.6 percent to 46.8 percent

over the seven years studied, while the share of loans between $60,000 and $170,000 in size

increased from 29.9 percent to 47.4 percent. The shift towards larger sizes was more

pronounced among manufactured home loans as compared to site-built home loans. This was

also reflected in the growth in the median loan amount from $44,000 in the first quarter of

2010 to $63,000 in the last quarter of 2016, a 43.2 percent increase.

1 98

The changes observed

among small site-built and manufactured home loans were gradual and there were no sharp

discontinuities observed around the effective date of the Rule. Section 5.4 .5 further ex plores the

effect of the Rule on small balance manufactured home loans.

1 98

Ov er the same period, the median size of site-built home loans grew by 24.4 percent, from $180,000 to $224,000.

A lso, the loan-to-v alue ratio for manufactured h om e loans grew from 6 5 percent in 2010 to 68 percent in 2016.

74 BUREAU OF CONSUMER FINANCIAL PROTECTION

Figures 16 and 17 show the distribution of home purchase and refinance mortgage originations,

respectively, by loan purchaser type over the period of 2010 to 2016 using data from the NMDB.

GSE loans represented 35.5 percent of purchase originations in 2010, but their share grew to

44.3 percent by 2013 and stayed around that level thereafter. Private originations constituted

19.1 percent of purchase originations in 2010; their share grew to 23.7 percent by 2014, but then

declined back to 21.4 percent by 2016.

1 99

Ov erall, the share of conventional (GSE plus Private)

originations grew from 55 percent in 2010 to around 67 percent in the years from 2013 to 2016,

while the composition of conventional originations did not shift appreciably during this time.

This composition is discussed in further detail in Chapters 5 and 6.

FIGURE 16: DISTRIB UTIO N OF PURCHA S E ORIGINA TIO NS BY LOA N PURCHA S ER TY PE, 2010-2016

1 99

Pr iv ate originations com prise of loans securitized by PLS and loans financed by portfolio lending by com mercial

ba nks, credit unions, savings banks, savings associations, mortgage banks, life insurance companies, finance

com panies, t heir affiliate institutions, and other private purchasers.

75 BUREAU OF CONSUMER FINANCIAL PROTECTION

Conventional loans play a more dominant role among refinance originations throughout the

period. There is a slight shift in the composition of conventional loans as the origination share of

GSEs declines from 73.7 percent in 2010 to 67.9 percent in 2013. The private origination share

grows from 13.0 percent in 2010 to 17.3 percent by 2013. After that point, the share of GSE

originations experiences further decline to 59.9 percent by 2016 while private originations show

a small increase to 18.4 percent by 2016. This shift in composition may reflect possible effects of

the Rule.

FIGURE 17: DISTRIB UTIO N OF REFINA NCE ORIGINA TIO NS BY LOAN PURCHASER TYPE, 2010-2016

To capture changes in pricing, Figure 18 shows average interest rates on jumbo and conforming

mortgage loans among fixed rate originations.

200

The conforming loan category contains both

standard conforming and super-conforming loans. Both conforming and jumbo interest rates

200

Fig ure 1 8 pool s purchase and refinance or iginations sinc e the trends are very similar.

76 BUREAU OF CONSUMER FINANCIAL PROTECTION

trend downward from 2010 to early 2013, falling from around 5 percent to roughly 3.5 percent.

A small spread exists between the two categories with jumbo loans having a slightly higher

interest rate than conforming loans. Following the rise of the benchmark 10-year Treasury rate

(Figure 5), mortgage interest rates increase in the second half of 2013 back to around 4.5

percent for conforming loans. Around the same time, interest rates on conforming loans become

higher than those for jumbo loans. Both rates trend downward through the end of 2016 again

following the benchmark 10-year Treasury rate, and the positive spread between conforming

and jumbo loans is sustained over that period. While the inversion of the rates roughly coincides

with the implementation of the Rule, these data are inconsistent with the proposition that the

Rule caused a significant increase in the price of jumbo loans relative to those of conforming

loans.

77 BUREAU OF CONSUMER FINANCIAL PROTECTION

FIGURE 18: A V ERA GE FIXED INTEREST RA TE BY CONFORMING LIMIT, JAN 2010 – DEC 2016

This section closes by considering the revenues and expenses associated with originating a

mortgage loan over time. Since 2008, the Mortgage Bankers Association has been publishing

the Annual Mortgage Bankers Performance Report that provides data on the revenues and

expenses associated with the origination of one-to-four unit residential loans.

201

Most providers

of the data are independent mortgage companies. In 2017, 280 respondents provided data.

These lenders originated 8,822 residential mortgages on average with an average loan size of

$240,191 and with an average origination volume of $2.13 billion. Respondents represented

around 74.2 percent of the mortgage origination volume of independent mortgage companies

and 34.4 percent of the volume originated market-wide.

202

While a large share of independent

mortgage companies are represented in the data, it is not possible to know exactly how

representative the reported numbers are among all independent mortgage companies.

201

The data also cov er the costs of servicing mortgage loans. Those costs are not considered here.

202

These share calculations are based on the 2016 volume of m ortgage or iginations for respondents, a ll i ndependent

m ortgage com panies, and for the m arket as a whole and the latter two volume figures are derived from HMDA .

78 BUREAU OF CONSUMER FINANCIAL PROTECTION

Furthermore, because the data are limited to independent mortgage companies, they do not

provide insight into the expenses of depository institutions. However, to the Bureau’s

knowledge, these data give the most detailed information on the expense and revenue structure

of mortgage origination.

203

FIGURE 19: AVERAGE REVENUES AND EXPENSES A SSOCIA T ED WITH ORIGINATING A LOA N FOR

INDEPE NDE NT MORTGA G E COMPA N I ES, 2008-2017

Figure 19 plots average revenues and expenses associated with originating a loan and their

respective components. Both revenues and expenses have been rising substantially from a little

over $4,000 in 2008 to around $8,000 in 2017. Revenue growth has somewhat outpaced the

growth in expenses, with revenues growing 75.1 percent and expenses growing 68 percent ov er

this time. In terms of revenues, net secondary marketing income (which includes the gain or loss

on the sale of loans in the secondary market, pricing subsidies and overages, as well as

capitalized servicing and servicing released premiums, together with a small amount of interest

income) has a large and growing share. Its share has grown from 51.7 percent of revenues in

2008 to 81.9 percent of revenues in 2017. Correspondingly, loan origination fee income’s share

has shrunk, from 33.7 percent in 2008 to 12.1 percent in 2017. Finally, the share of other

203

Sev eral c om menters noted the discussion of these data in U.S. Dep’t of t he T reasury, A Financial System that

Creates Economic Opportunities: Bank s and Credit Unions, June 2017. In addition, several c om menters referenced

su rv eys of r egulatory burden for c redit unions conducted by the Credit Union National A ssoc iation, available at

h t tps://www.cuna.org/regburden/. See also Appendix B.

79 BUREAU OF CONSUMER FINANCIAL PROTECTION

originations income (such us underwriting and processing income, administration and other

fees, and fee income earned on loans acquired from correspondents and brokers) has remained

small and relatively stable.

In terms of expenses, non-personnel expenses (occupancy and equipment expenses and other

direct expenses, including technology-related expenses, outsourcing and professional fees, and

other operating expenses) have grown moderately over the period covered from $1,570 per loan

in 2008 to $2,174 in 2017. Personnel expenses, in contrast, have grown rapidly both in absolute

amount and as a share of overall expenses, from $2,905 per loan in 2008 to $5,346 in 2017,

reaching 71.1 percent of all expenses by 2017. This increase can be attributed to increased

compensation per employee as the growth in the average number of employees (at 230 percent

over the period covered) has not outpaced the growth in the number of originations (241

percent).

While the above reported trends clearly establish that the revenues and expenses associated with

originating mortgage loans have increased over the past decade, it is uncertain whether the

increase or some part of it was caused by the ATR/QM Rule. First, the increase that took place

was gradual and there was no distinct increase around the time of the implementation of the

Rule. Second, multiple changes in the mortgage market have affected the cost of doing business

in this market over the period examined. On the regulatory side, the Secure and Fair

Enforcement for Mortgage Licensing (SAFE) Act was enacted into law on July 30, 2008 and the

Bureau’s TILA-RESPA Integrated Disclosure Rule also came into effect on October 3, 2015. On

the non-regulatory side, there was pressure to keep up with consumer expectations for a more

streamlined process with investments in better technology

204

, ongoing uncertainty about GSE

reform, and reduced volume of lending in part because of historically low refinance activity.

For

these reasons, it is not possible to determine from these aggregate trends alone if the ATR/QM

Rule contributed, in part or at all, to the observed increase in mortgage origination expenses.

3.6 Compliance with the Rule

Section 1025 of the Dodd-Frank Act grants the Bureau exclusive authority to examine insured

depository institutions and insured credit unions with total assets of more than $10 billion and

their affiliates to (among other things) assess these entities’ compliance with the requirements

of Federal consumer financial laws. Section 1024 of the Dodd-Frank Act separately

204

See Da ily Dose, Keeping Pace with Digitiz ation in th e Mortgage M ark ets, MReport (Aug. 27, 2018), available at

h t tps://themreport. com /daily-dose/08-27-2 018 /ke ep in g-pace-with-digitization-in-mortgage-lending.

80 BUREAU OF CONSUMER FINANCIAL PROTECTION

authorizes the Bureau to examine depositories

205

and certain non-bank depositories engaged in

residential mortgage lending,

206

among other things, and assess these entities’ compliance with

the requirements of Federal consumer financial laws.

207

The Bureau created its non-

depository supervision program in January 2012.

208

After the effective date of the Rule, the Bureau allowed four months to pass in order for

financial institutions to address compliance and technical issues that may be impacted by major

system changes.

209

Supervisory examinations of mortgage originators since 2014 have generally

focused on reviewing for compliance with the Rule. The Bureau discusses in its Supervisory

Highlights patterns and trends found during exams.

21 0

This section focuses on ATR-related

findings from mortgage origination exams.

Supervision has observed that most entities, depository or non-depository, examined by the

Bureau are generally complying with the ATR/QM Rule. However, as first described in the Fall

2016 Supervisory Highlights

21 1

and further discussed in the Spring 2017 Supervisory

Highlights

21 2

, with respect to certain ability-to-repay provisions

21 3

, the Bureau’s examinations

identified the following violation:

205

1 2 U.S.C. § 5515.

206

1 2 U.S.C. § 5514. (326).

207

1 2 U.S.C. § 5514(a)(1)(A): this provision applies to any covered person who “offers or prov ides origination,

brokerage, or servicing of loans secured by real estate for use by consumers primarily for personal, family, or

h ou sehold purposes, or loan modification or foreclosure relief services in connection with such loans.”

208

See S t eve Antonakes & Peggy Twohig, The CFPB Launches its Nonbank Supervision Program, CFPB Blog (Jan. 5,

2 01 2), available at https://www.consumerfinance.gov/about-us/blog/the-c fp b-launches-its-nonbank-su p erv isi on -

pr ogram/.

209

See Bu reau Consumer Fin. Prot., Supervis ory High lights, Issue 7 (Winter 2015), available at

h t tps://files.consumerfinance.gov/f/201503_cfpb_supervisory-highl ights-winter-2015.pdf.

21 0

See generally Bu reau Consumer Fin. Pr ot., Supervisory Highlights, https://www.consumerfinance.gov/policy-

com pliance/guidance/supervisory-highlights/ (last visited Dec. 3 1, 2018) (for a list of a ll published Supervisory

Highlights); s ee also Bu reau Consumer Fin. Prot., Superviso ry Highlights, Is sue 15 (Spring 2017), available at

h ttps://s3.amazonaws.com /f ile s.c onsume r fina nc e.gov /f /d ocu m en t s/2 0 170 4 _cfpb_Su p erv i s or y -Highlights_Issue-

1 5 .pdf (for Supervision’s observations and a pproach to compliance with the ATR/QM Rule provisions).

21 1

See Bu reau Consumer Fin. Prot., Supervis ory High lights, Issue 13 ( Fall 2 016), available at

h t tps://files.consumerfinance.gov/f/documents/Supervisory_Highlights_Issue_1 3__Final_1 0.31.16.pdf.

21 2

See Bu reau C onsumer Fin. Prot., Supervisory Highlights, Issue 15 (Spring 2017), available at

https ://s 3.amazonaws.c om/files.consumerfinance.gov/f/documents/201704_cfpb_Supervis ory-

Highlights _Is sue-15.pdf.

21 3

1 2 C .F.R. § 1 026.43(c)(2)(vii), .43(c)(4), and .43(c)(7).

81 BUREAU OF CONSUMER FINANCIAL PROTECTION

Income Verification

• A creditor violated the ATR requirements by failing to properly verify income relied upon

when considering the consumer’s monthly debt-to-income ratio and determining the

consumer’s ability to repay.

21 4

The Bureau also has enforcement authority with respect to non-depository mortgage

originators

21 5

and depositories with assets over $10 billion,

21 6

and the prudential regulators

have enforcement authority with respect to smaller depositories.

Since the effective date of the

Rule, the Bureau has not brought enforcement actions against any entities, depository or non-

depository, for violating the Rule.

21 4

Supra n ote 211, at 14.

21 5

For en forcement authority of n on-depositories, see 12 U.S.C. § 5514(c).

21 6

For en forcement authority of depositories, s ee 12 U.S.C. § 5515(c).

82 BUREAU OF CONSUMER FINANCIAL PROTECTION

4. Assuring the ability to repay

This chapter assesses the effectiveness of the 2013 ATR/QM Rule in assuring that mortgages

consumers received are on terms that reasonably reflected their ability to repay the loans. The

Rule’s Ability-to-Repay provisions require that lenders consider and verify specific underwriting

factors, while the Qualified Mortgage provisions provide a legal presumption of compliance

(that is either conclusive or rebuttable) for loans which satisfy certain underwriting

requirements and restrictions, including those on interest-only payments, negative

amortization, balloon payments, terms exceeding 30 years, and debt-to-income (DT I) ratios.

These provisions apply to covered loans applied for on: 1) the relationships between some of the

key restricted loan characteristics and loan performance; 2) changes in loan characteristics

when the Rule became effective; and 3) measures of loan performance for those segments of the

market where loan characteristics changed.

The main findings in this chapter include:

• Loans with risky features, including interest-only pay ments, low documentation,

negative amortization, balloon payments, adjustable-rate mortgages (ARMs) for which

the interest rate can reset in under five years, and terms exceeding 30 years, had largely

disappeared from the market prior to the effective date of the Rule and today appear to

be restricted to a limited market of highly credit-worthy borrowers. Such loans had

particularly high rates of default among 2005 to 2007 originations. By subjecting the

origination of loans with risky features to the ATR requirement, and limiting the ability

of such loans to obtain QM status, the Rule is likely to mitigate the reemergence of risky

loans should a similar overexpansion of the mortgage market take place.

• In the current market, DTI ratios are likely constrained from returning to crisis-era

levels by a combination of the ATR requirement, GSE underwriting limits which define

the loans which are eligible for purchase by the GSEs (currently, a DTI limit of 45

percent applies to most loans) and the Bureau’s General QM DTI threshold which limits

the General QM category to loans with a DTI at or below 43 percent. Even though house

prices have largely returned to pre-crisis levels, currently 5 to 8 percent of conventional

loans for home purchase have DTI exceeding 45 percent; in contrast, approximately 24

to 25 percent of loans originated in 2005 – 2007 exceeded that ratio. Given the negative

relationship between higher DTIs and loan performance, this restraint likely contributes

83 BUREAU OF CONSUMER FINANCIAL PROTECTION

to ensuring that borrowers receive loans they are able to repay, in addition to potentially

mitigating systemic risks.

• Early delinquency rates (measured as the percentage of loans becoming 60 or more days

past due over the first two years since origination) remain historically low in the post-

crisis era. T he early delinquency rate of loans with DTI exceeding 43 percent made under

the Rule’s ATR underwriting requirements (non-QM loans) has remained steady at 0.6

percent. In contrast, the early delinquency rate of GSE loans with DTIs above 43 percent

rose from 0.6 percent in 2012-2013 to 1 percent among 2014-2015 originations. Thus,

the performance of non-QM loans with DTI greater 43 percent has improved relative to

the performance of comparable loans purchased by the GSEs following the

implementation of the Rule.

The first section describes the loan performance statistics used to measure borrower distress,

and how such measures relate to the idea of assuring the ability to repay. T he second section

provides evidence on several restricted features which prevent loans from satisfying the General

QM requirements, including interest-only payments, balloon payments, negative amortization,

terms exceeding 30 years, and loans made with limited income or asset documentation. Loans

with restricted features are quite rare in the post-Rule period, but where the data allow, their

performance is analyzed and compared to that of loans without such features. Effects on

adjustable-rate mortgages, which the QM provisions require to be underwritten to the

maximum payment within the first five years of the loan, are also examined. T he third section

provides historical evidence on the trends in DTI ratios and their relationship to loan

performance. The final section documents how DTIs changed for some covered loans originated

after the Rule became effective and compares the performance of these loans to those which

were not directly affected by the Rule’s General QM DTI threshold.

4.1 Ability to repay and loan performance

Because the affordability of a given mortgage will vary from consumer to consumer based upon

a range of factors, there is no recognized metric that can directly measure whether the terms of

mortgage loans made after the Rule’s effective date reasonably reflect consumers’ ability to

repay. This analysis instead measures a proxy for the lack of ability to repay across a wide pool

of loans by considering the frequency of early borrower distress, measured as whether a

borrower was ever 60 or more days past due within the first two years after origination.

21 7

This

21 7

Day s past due is defined using the Mor tgage Ba nkers Association (MBA) c alculation m ethod.

84 BUREAU OF CONSUMER FINANCIAL PROTECTION

measure is referred to as the “early delinquency rate” in the analyses in this chapter. The focus

on early delinquencies is intended to capture borrowers’ difficulties in making payments soon

after the origination of the loan, even if these delinquencies do not lead to a borrower potentially

losing their home. To evaluate more serious borrower distress, some analyses use a measure of

whether a borrower was ever in foreclosure within the first two years after origination, referred

to as the “early foreclosure rate.” For purposes of this assessment, the Bureau assumes that the

average “early delinquency rate” and “early foreclosure rate” across a wide pool of Qualified

Mortgages (QM) are probative of whether QM loans reasonably assure repayment ability, and

that the dependence of these rates on the defining characteristics of QM loans is probative of

how those characteristics may influence repayment ability. Likewise, the av erage “early

delinquency rate” and “early foreclosure rate” among a wide pool of non-QM loans are probative

of whether such loans reasonably assure repayment ability.

To be clear, this analysis does not define or otherwise identify any acceptable limits of

delinquencies and defaults for QM and non-QM loans. Delinquencies are measured but are not

assessed against any assumed benchmark. Defining or otherwise identifying benchmarks for

acceptable levels of delinquencies for new loans is beyond the scope of this report and, in any

event, is difficult in part because the level of delinquencies at a given time (and thus for vintages

of loans made around that time) will depend not only on the characteristics and underwriting of

the loans themselves but also on the subsequent health of the economy as a whole. The primary

goal of this chapter is to present relevant evidence over time and across products.

4.2 Loans with restricted features

The Rule imposed specific documentation, verification and underwriting requirements for loans

to meet the General QM criteria, generally eliminating or restricting no-documentation and

certain low-documentation loans; furthermore, the General QM category excludes loans with

particular features that are viewed as higher risk such as interest-only payments, balloon

payments, negative amortization, and terms over 30 years.

21 8

The Rule also imposes

requirements on how creditors determine the monthly payment obligations used in

underwriting. In particular, in order for a loan to be a General QM loan, it must be underwritten

based on the maximum interest rate permitted during the first five years of repayment, whereas

21 8

Ba lloon loans are particularly rare in both McDa sh and CoreLogic. One possible explanation is that coverage i n

bot h data sources i s sk ew ed t owards larger lenders and therefore m ay not fully capture loans originated by sm all

creditors. Chapter 8 further discusses a prov ision in the Rule that allows small creditors to originate balloon

pa y ment QMs, subject to similar restrictions as General QM loans. See 12 C.F.R § 1 026.43(f) for more information.

85 BUREAU OF CONSUMER FINANCIAL PROTECTION

for non-QM loans the underwriting must be based on the maximum interest rate permitted

under the mortgage.

21 9

Figure 20 shows the share of conventional purchase loans with each restricted loan feature

based on NMDB data.

220

The prevalence of restricted feature loans in the market was already

quite limited prior to the Rule’s implementation, in contrast to their more widespread use in the

years preceding the financial crisis.

221

While it is beyond the scope of this assessment to model

how the U.S. economy and the housing market would have progressed had the ATR/QM Rule

been in place at the beginning of the 21

century or how it would progress in the future absent

the Rule, some simple calculations can shed light on some of the changes that the Rule would

have likely brought about had it been in place at the time.

21 9

1 2 C .F.R. § 1 026.43(e)(2)(vi)(A).

220

Loa n s may have multipl e restricted features, and thus a ppear in multiple groups i n Figure 20.

221

Com parable patterns have been found in other studies of these restricted feature loans. See Bing Ba i et al., Has the

QM Rule Made it Harder to Get a Mortgage?, Hous. Fin. Pol’y Ctr. Com mentary, Urb. Inst. (2016), available at

h t tps://www.urban.org/sites/default/files/publication/7 8266/2000640-Has-the-QM-Rule-Made-It-Harder-to-

Get-a-Mor tgage.pdf.

86 BUREAU OF CONSUMER FINANCIAL PROTECTION

FIGURE 20: SHA RE OF CONV ENT IO NA L PURCHA S E LOA NS WITH RESTRI CT ED FEATURES, 1998-2016

Figure 21 shows the share of loans with at least one restricted feature

222

among early foreclosure

loans (defined here as loans that foreclosed within two years of origination) and among

performing loans (defined here as those loans that did not foreclose within two years of

origination) by origination year. The figure shows that 50 to 60 percent of early foreclosed loans

from the 2005 to 2007 originations that preceded the crisis had features that the Rule generally

subsequently restricted or eliminated in some manner. The Rule would likely have prevented at

least some of the early foreclosed loans that had these features from being originated in the first

place, potentially eliminating a majority of early foreclosed loans if the Rule had been in place at

222

Gi v en that loans with n o documentation are not distinguished from loans with low documentation in the NMDB,

t h is figure does not c lassify loans a s having restricted features based on documentation alone. Forec losure shares

w h en classifying these loans as restricted are similar.

87 BUREAU OF CONSUMER FINANCIAL PROTECTION

the time.

223

On the other hand, it is not possible to assess to what extent performing loans with

these features would have been originated under terms allowed by the Rule and to what extent

they would have been eliminated. Further, while it is not possible to assess the likelihood that

risky lending with these features would occur again in the future absent the Rule, an important

benefit of the Rule is that it limits such an outcome and any consequent consumer harm or

macroeconomic disruption.

FIGURE 21: SHA RE OF LOA NS WITH RESTRICTE D FEA TURES AMONG EA RL Y FORECL OS U RE LOA NS A ND

A MONG PERFO R MING LOANS BY ORIGINA TION Y EA R

The remainder of this section focuses on the more narrow use of products with these restricted

features in the post-crisis era. To assess borrowers’ ability to repay within this space, the

223

An analysis u sing CoreLogic data finds that the national foreclosure inventory, as measured by the number of

m ortgaged r esidences t hat have been placed into the foreclosure process by the servi cer, peaked i n January 2 011.

See Unite d States Residential Fo re closure Cris is : Te n Years Late r ( Ma rch 2 017), available here

h t tps://www.corelogic.com /research/foreclosure-report/national-foreclosure-report-10-y ear.pdf.

88 BUREAU OF CONSUMER FINANCIAL PROTECTION

following analyses compare the characteristics and performance of such loans to loans in the

overall market, taking advantage of the larger sample sizes available in the McDash and

CoreLogic datasets.

224

Figure 22 shows the limited prevalence of restricted feature loans in these

datasets since 2012. Further, in qualitative responses to the Bureau’s survey of lenders

concerning mortgage applications from 2013 through 2016, multiple lenders reported

discontinuing products with balloon and interest-only payments, as well as changing the

structure or income requirements of ARM products.

225

224

The variable that indicates documentation status in the McDash data is reported as missing or unknown for over

6 0 percent of loan observations beginning in 2014. For this reason, this subsection uses CoreLogic LLMA data in

order to measure loan performance by documentation status and the McDash data to analyze the other loan

fea tures. The CoreLogic data indicate whether a loan is “full documentation”, “low or minimal documentation”, or

“no a sset /inc om e v erification”. A “full docum entation” l oan is described a s one in which the borrower’s em pl oym ent,

in com e and assets have b een verified. In contrast, loans are c ategorized a s “no documentation” if the provi der of the

loan data clearly indicates that the loan was originated with no documentation. The third category of loans is “low or

m inimal documentation”, which includes any loan that does not fit in the previous two categories and is n ot missing

th is information. Low documentation loans may include phrasing from the data provider such as streamlined,

r educed, or lim ited verification.

225

See Section 8 .2 in this report for additional details of r esponses t o the survey of lenders. In a ddition, responses to

th e Bureau’s 1022(c)(4) information request to nine anonymous mortgage lenders also indicate that sev eral l enders

preemptively discontinued some of these restricted product features altogether prior to the Rule’s effective date.

Ot h er lenders r eport that they continue t o offer interest-only loan products a s non-QM loans or that they still

orig inate loans with documentation exceptions in limited circumstances.

89 BUREAU OF CONSUMER FINANCIAL PROTECTION

FIGURE 22: SHA RE OF CONV ENT IO NA L PURCHA S E LOA NS WITH RESTRI CT ED FEATURES: 2012-2017

4.2.1 Post-crisis characteristics and performance of loans

with restricted features

Given their rarity and the low overall delinquency levels in the current market, the available data

do not allow for informative comparisons of loan performance for loans with balloon payments,

negative amortization, and/or terms over 30 years.

226,

227

Further, loans with no asset or income

226

Balloon loans are particularly rare in both the McDash and CoreLogic datasets. One possible explanation is that

cov erage in both data sources is sk ew ed t owards larger lenders and therefore m ay not fully c apture loans originated

by sm all creditors. Chapter 7 further discusses a prov ision in the Ru le that allows sm all creditors to or iginate

ba lloon payment QMs, subject to similar restrictions as General QM loans. See 12 C.F.R. § 1 026.43(f).

227

The McDash data are reported on a monthly basis and as such, this analysis does not consider loans with terms of

3 6 1 or 362 m onths t o exceed 3 0 years. C onsidering such loans a s 30-year term also accounts for the possibility that

90 BUREAU OF CONSUMER FINANCIAL PROTECTION

verification are non-existent in the sample used in this chapter, even before the ATR/QM Rule

took effect. For those reasons, performance is only estimated for loans with interest-only

payments, loans with limited documentation of borrower assets or income, and (in the next

subsection) loans with ARM resets under five years.

Figure 23 shows that from 2012 through 2015, loans with interest-only payments had

considerably lower early delinquency rates than the market as a whole. T his likely reflects the

more limited use of such products after the crisis era. There was a small rise in the use of such

loans at the time the Rule went into effect (see Figure 22), which was accompanied by an uptick

in their early delinquency rate while still staying significantly below the early delinquency rate of

non-interest only loans (see Figure 23). Figure 24 shows that the limited number of loans

reported as having minimal documentation of either assets or income performed comparably to

the broader population of mortgages from 2012 through 2015.

228

a loa n origination could occur a month or more before the borrower’s first payment is due. For example, if a

borr ower closes on a loan on January 15

, the first payment may not be due until March 1

and th i s m ay r esu lt i n a

r eported loan term that is on e to two months longer than 360 months. See 12 C.F.R. § 1 026.17(c)(4)(iii) for more

in formation about disclosures relating to the calculation of payment schedules.

228

Du e to limitations in data availability, the delinquency rates show n in Figure 24 are m easured for loans that w ere

or ig inated from January 2012 through September 2015.

91 BUREAU OF CONSUMER FINANCIAL PROTECTION

FIGURE 23: EARLY DELINQ UE NCY RA TES BY INTEREST-ONLY PAYMENT STA TUS, CONV ENT IO NA L

PURCHA S E LOA NS, 2012 THROUGH 2015

92 BUREAU OF CONSUMER FINANCIAL PROTECTION

FIGURE 24: EARLY DELINQ UE NCY RA TES BY A SSET A ND INCO ME DOCUMENTA TION STATUS,

CONV ENTIONA L PURCHA S E LOANS, 2012 THROUGH 2015

To further examine the underwriting of these loans, T able 1 shows average borrower and loan

characteristics for 2014 originations of the two loan products analyzed in this subsection, as well

as the adjustable rate products analyzed in the next subsection. On average, the small subset of

borrowers who took out low documentation loans in 2014 tended to have similar characteristics

to the general population of borrowers, consistent with their comparable loan performance

shown in Figure 24. On the other hand, borrowers who took out interest-only loans tended to

have higher credit scores and markedly lower LTV ratios and introductory rates on average.

These characteristics suggest that loans with these restricted features may be largely confined to

highly creditworthy borrowers.

93 BUREAU OF CONSUMER FINANCIAL PROTECTION

TABL E 1 : AV ERA GE BORROW ER A ND LOA N CHA RA CTERISTICS OF CONV ENT IO NA L PURCHA S E

LOA NS WITH RESTRICTE D FEATURES, 2014 ORIGINA T IONS

Loan sample Credit score DTI ratio LTV ratio Interest rate Observations

All loans (McDash) 755.48 29.69 81.93 4.23 875,044

Interest-only (McDash) 770.79 32.57 69.46 2.99 8,108

ARMs that reset in under 5

years (McDash)

772.63 31.71 74.4 2.88 3,635

All loans (CoreLogic) 755.47 33.45 80.48 4.24 579,931

Low documentation

(CoreLogic)

756.44 32.62 80.54 4.04 9,700

4.2.2 Effects on adjustable rate mortgage characteristics

In addition to prohibiting certain features on QM loans, the QM provisions of the Rule require

that creditors underwrite based on the maximum interest rate permitted during the first five

years of repayment.

229

These provisions operate in part to prevent the widespread return of

loans underwritten based on a “teaser” rate payment used for the first two or three years of the

loan, which would then reset to a much higher level.

230

In qualitative responses to the Bureau,

several lenders noted that they had changed the structure of some adjustable rate mortgages

(ARMs) in response to this requirement, increasing the time until first payment reset to five

years or longer.

To assess whether such a shift occurred across the market more broadly, Figure 25 examines the

share of ARMs with initial reset timing below five years.

231

The sample is restricted to the

conventional, non-GSE market where General QM provisions, rather than Temporary GSE QM

or Federal Agency QM provisions, are likely to bind. The data show that while ARMs with initial

229

1 2 C .F.R. § 1 026.43(e)(2)(iv)(A).

230

“ [T ]he abi lity-to-repa y prov is i on s of t h e Dod d -Frank Act w ere codified in response t o lax l ending t erms and

pr actices in the mid-2000's, which led to increased foreclosures, particularly for subprime borrowers. The statutory

u n derwriting r equirements for a qu alified m ortgage—for exampl e, the requirem ent that loans be underwritten on a

fully amortized b asis using the m aximum interest rate during the fi rst fiv e years and n ot a t easer rate, and the

r equ irement t o c onsider and v erify a consumer's incom e or a ssets—wi ll help prevent a return t o such lax lending.”

7 8 Fed. Reg. 6511 (Jan. 30, 2013).

231

The initial reset of an ARM is also referred to by the term “recast.”

94 BUREAU OF CONSUMER FINANCIAL PROTECTION

reset timing under five years already made up less than 20 percent of ARMs prior to the Rule’s

effective date, their share fell further after the effective date of the Rule.

FIGURE 25: SHA RE OF CONV ENTIONA L, NON-GSE ADJUSTABLE RATE MORTGA GES SPLIT BY INITIA L

RESET TIMING, 2012 TO 2015

Like the non-QM loan features discussed in the previous subsection, short timing reset ARMs

already made up a much smaller share of the market in the years immediately prior to the Rule’s

effective date than during the financial crisis. Column 3 of Table 1 shows that short timing reset

ARMs appear to be restricted to highly creditworthy borrowers. This is also reflected in the very

low early delinquency rates for such loans, shown in Figure 26, though the strong initial

performance for all ARMs is due in part to loans for which initial payments have yet to reset.

Together, the ATR and General QM underwriting requirements of the Rule will likely prevent

loans with these characteristics from re-emerging as a means of enabling borrowers to gain

approval for a mortgage.

95 BUREAU OF CONSUMER FINANCIAL PROTECTION

FIGURE 26: EARLY DELINQ UE NCY RA TES OF CONV ENTIO NA L, NON-GSE ADJUSTABL E RA TE

MORTGAGES, SPLIT BY INITIA L RESET TIMING, 2012 TO 2015

96 BUREAU OF CONSUMER FINANCIAL PROTECTION

4.3 Historical trends in DTI and relationship

with loan performance

4.3.1 Historical trends in DTI

While the primary focus of this chapter is on the years surrounding the Rule’s implementation,

this subsection provides context by examining how DT Is have evolved since 2000. As was

shown in Figure 8 of Chapter 3, recent vintages of mortgage originations have had very low early

delinquency rates, on the order of 1 to 3 percent depending on the product type, relative to peak

vintage early delinquencies of 7 to 25 percent in 2007. These low early delinquency rates are

seen for vintages both before and after the Rule’s 2014 effective date, likely reflecting both

steady economic growth and changes in lender practices following the collapse of the mortgage

market and the 2007 to 2009 recession as discussed in Chapter 3.

97 BUREAU OF CONSUMER FINANCIAL PROTECTION

FIGURE 27: CONV ENTIONA L PURCHA S E MORTGA G E DTI DISTRIBUTIO NS BY ORIGINA TIO N YEAR

Shifts over time in both housing costs and underwriting practices can be seen in the changing

distribution of DTI for loans originated in the years prior to, during, and after the financial

crisis. Figure 27 plots these distributions for conventional purchase mortgages and in each case

indicates the Rule’s General QM DTI threshold of 43 percent.

232

The distribution of DTIs shifted

substantially higher from the pre-crisis era (2000 to 2003) to the years surrounding the crisis-

era (2004 to 2007) as a result of rising home prices and loosening underwriting requirements,

and included many loans above both the General QM DTI threshold and recent GSE DT I limits

(45 percent without what the GSEs consider compensating factors like required cash reserves or

232

DTI data are only available for 33 percent of loans in this sample of the McDash data. Where possible, analyses are

r eplicated using NMDB da ta on GSE and FHA loans, for which close to full DTI cov erage is available.

98 BUREAU OF CONSUMER FINANCIAL PROTECTION

LTV restrictions, 50 percent with such compensating factors).

233

Following the crisis, by 2012

nearly all conventional loans had DTIs below these GSE limits, and substantially fewer loans

were made with DTIs above 43 percent. While DT Is for conventional loans are trending higher

in the current market, they are likely being constrained from returning to crisis-era levels by

these GSE limits, combined with limited appetite from lenders to originate non-QM loans above

the General QM DTI threshold.

234

In 2017, 5 percent of conventional purchase loans had DT Is

over 45 percent, compared to 24 percent in the years surrounding the crisis.

235

The role of the GSE DTI limits is highlighted in Figure 28, which shows the comparable DTI

distributions for GSE and FHA loans over this time period, using NMDB data. DTI levels for

GSE loans have been held below the GSE-imposed 45 percent and 50 percent limits in current

years, driving the results seen for all conventional loans in Figure 27. In contrast, current DTI

levels for FHA loans exceed their crisis-era levels, with numerous loans originated up to an

apparent DTI limit of 57 percent.

236

233

Typical required compensating factors for GSE loans with a DTI above 45 percent include twelve months of cash

reserves for the borrower and a maximum LTV ratio of 80 percent. See Steve Holden & Walt Scott, Desktop

Underwriter Version 10.1 – Updates to the Debt-to-Income (DTI) Ratio Assessment, Credit Risk Sharing

Commentary, Fannie Ma e (July 1 0, 2 017), available at http://www.fanniem ae.com /portal/funding-the-

m arket/credit-risk/news/desktop-underwriter-debt-to-in com e -ratios-071017.html.

234

Ch apters 5 , 6 , and 8 provide further evidence and discussion on lenders’ a pproaches to n on-QM lending.

235

In th e C oreLogic data, 8 percent of 2 017 c onventional purchase or iginations had DT Is ov er 45 percent, c om pared

t o 2 5 percent in the y ears surrounding the crisis.

236

FHA underwriting allows DTI ratios above those seen in the conventional space. Section 4.1.2 provides a brief

analysis of DTI and delinquency for recent FHA originations.

99 BUREAU OF CONSUMER FINANCIAL PROTECTION

FIGURE 28: GSE A ND FHA PURCHA S E MORTGA G E DTI DISTRIB UTIONS BY ORIGINA T ION YEAR

While the remainder of this section focuses on the relationship between DTI and loan

performance, the Rule’s focus on DTI was intended to have additional benefits. The Rule’s

underwriting requirements were not only meant to improve assessments of individual

consumers’ ability to repay, but were also intended to limit potential systemic effects of

overextended credit.

237

Underwriting limits on maximum allowable DTIs can provide a

237

For example, regarding misstated incomes used in underwriting, the Rule stated that “ [t]he systemic effects were

ev ident: the extension of credit against inflated incomes expanded the su pply of credit, which in turn continued the

r a pid r ise of h ouse prices in the later years of t he housing boom and exacerbated the ev entual crash.” 78 Fed. Reg.

6 4 08, 6561 ( Jan. 3 0, 2013) . For the prevalence of t easer rate products w hich did n ot r eflect true debt payment

lia bilities, the Rule stated that “. . . the widespread use of t he product put many borrowers in precarious financial

posit ions and m ay also have fueled the sy stemic rise in h om e prices. The elimination of t hese products should lim it

both the individual and sy stemic harms which ultimately translate, in the largest part, into harms to the individual

consumers.” Id. The Rule’s requirements to accurately document and use income and debt payment information

100 BUREAU OF CONSUMER FINANCIAL PROT ECTION

meaningful constraint on borrowing levels. In turn, a DTI limit which binds for the most highly

leveraged borrowers can potentially benefit the broader population of consumers, by

constraining excessive house price growth and subsequent resulting price declines in a

downturn. Such a limit effectively imposes a link between borrowing and household incomes.

Recent research, notably Greenwald (2018), has studied this mechanism in depth, finding that

in a market with low downpayment requirements and large numbers of borrowers at or near

DTI limits (as exists in the post-Rule period), small changes in DTI limits can lead to substantial

house price and borrowing changes.

238

In simulations conducted in that paper, the existence of a

DTI limit significantly reduces the magnitude of house price fluctuations and the resulting

borrower distress from pricing corrections. This report does not attempt to estimate these

systemic effects, but they represent a potentially substantial benefit of DTI thresholds for overall

market stability and loan performance, in addition to the relationships described in the next

subsections.

4.3.2 Relationship between DTI and loan performance

The following figures examine the relationship between DTI and early delinquency rates, across

different time periods and mortgage loan types, and find that relationship to be generally

positive. The relationships are shown through both observed mean early delinquency rates for

loans with different DTIs, as well as expected mean early delinquency rates which have been

adjusted to control for differences between loans in other characteristics.

The included control variables reflect underwriting information used directly to assess mortgage

riskiness (credit scores, LTV ratios), characteristics which may indirectly signal the risk of a loan

(documentation type, interest rate, loan amounts), and the month and year of origination to

account for changes in the economy over time which influence market-wide performance.

239

The

en sure that DT I lim its cannot be ev aded by m isrepresentation, which in turn allow s the DTI lim its to im pose a

m eaningful constraint on borrowing levels.

238

Da n iel G reenw ald, The Mortgage Credit Channel of Macroeconomic Transmission, (MIT Sloan Research Paper

No. 51 84–16, 2 016 ). See also Dean Corbae and Erwan Quintin (2015), Leverage and the Foreclosure Cris is, Jou r n al

of Political Economy, v ol. 123(1), pg. 1-65.

239

Th ough not necessarily indicators of repayment ability, these control variables reflect standard risk factors used

both in practice to set mortgage pricing and by researchers to study loan-level risk based on characteristics at

or ig ination. See, e.g., Fannie Mae, Lo a n-Le vel Pric e A d justmen ts ( LLPA ) M atrix, (June 5, 2018), available at

h t tps://www.fanniem ae.com /content/pricing/llpa-m atrix.p df; Freddie Ma c, Credit Fees in Pric e, at E1 9–1 2 ( Dec. 5 ,

2 01 8), available at h ttp://ww w.freddi emac.com /singlefamily/pdf/ex19.pdf (for mortgage pricing); s ee Robe r t B.

A v ery et al., Credit Ris k, Credit Scoring, and the Performance of Home Mortgages, Fed. Reserve Bu ll. (July 1996);

Ha milton Fout, Grace Li, & Mark Palim, Credit Risk of Low Income Mortgages, (Fannie Mae, Econ. & Strategic

Resea rch White Pa per, 2 017), available at

h t tp://www.fanniemae.com /resources/file/research/datanotes/pdf/credit-risk-of-low-incom e -m ortg a ge s-white-

101 BUREAU OF CONSUMER FINANCIAL PROT ECTION

inclusion of these variables helps assess the extent to which the relationships between DTI and

mean early delinquency in the data are driven by correlation with these other characteristics (for

example if DTI is positively correlated with LTV), and whether they persist after accounting for

such correlations.

240

When the full set of control variables is included, the estimated

relationship between DTI and early delinquency reflects the expected early delinquency for two

otherwise similar loans originated in the same month and year with the same credit score, LTV

ratio, loan amount, documentation type, and interest rate.

241

Figure 29 shows the relationship between DTI and early delinquency rates for conventional

single-family purchase loans originated from 2006 to 2008 during the latter part of the financial

crisis. The green data points reflect the mean early delinquency rate for loans originated within

each equally sized (by count of loans) DTI bin. The black line shows the linear best fit line for the

underlying data. The right panel shows that for two loans which are otherwise identical

according to the characteristics listed above, the expected early delinquency rate for a loan with

a DTI of 20 percent was approximately 8 percent, while a loan with a DTI of 40 percent had an

expected early delinquency rate near 13 percent.

paper.pdf; Christopher Mayer, Karen Pence, & Shane M. S herlund, The Rise in Mortgage Defaults, 23 J. of Econ.

Persp. 27 (2009) (studies of loan-level risk based on characteristics at origination).

240

Note that this analysis does not attempt to estimate the relative explanatory or predictive pow er of different

v ariables that could be used in underwriting, but rather seeks to establish the relationship between DTI and

per formance with and without controlling for these other underwriting factors. See Diana Farrell, Kanav Bhagat, &

Ch en Zhao, Falling Behind: Bank Data on the Role of Income and Savings in Mortgage Default, (JP Morgan Chase

In st., 2018), available at https://www.jpmorganchase.com /corporate/institute/insight-in com e-shocks-mortgage-

default.htm; Mark Zandi & Cristian DeRitis, Special Report: The Skinny on Skin in the Game, (Moody ’s Analytics,

Ec on . & C on sumer Credit Analytics, 2011), available at https://www.economy.com /m ark-

za n di/documents/QRM_030911.pdf.

241

A sh ift in reporting of documentation type occurs in 2014 in the McDash data, with a substantially h igher share of

loans reporting “unknown” documentation type. The specifications used in this analysis includes an interaction

term for documentation type and dates after 2014 to account for potentially differential categorizations used in the

la t ter part of t he sample. For a dditional details on the m ethodol ogy used for these figures, see Michael Stepner,

2 01 3. "Binscatter: Stata module to generate binned scatterplots," Statistical Software Components S457709, Boston

College Department of Economics, available at https://ideas.repec.org/c/boc/bocode/s457709.html.

102 BUREAU OF CONSUMER FINANCIAL PROT ECTION

FIGURE 29: RELATIONSHIP BETWEEN EARLY DELINQUE NCIES A ND DTI, 2006 TO 2008 CONV ENTIONA L

PURCHA S E ORIGINA T IONS

Figure 30 shows the same relationship for conventional single-family purchase loans originated

in the years surrounding the implementation of the Rule from 2012 to 2015. As discussed

earlier, the overall early delinquency rate during this period is approximately one tenth of that

for the 2006 to 2008 vintages (see different scale on y-axis). A similar positive correlation is

seen, though with a decline in delinquencies at the highest levels of DTI, particularly those

above the General QM threshold (dashed gray line) or the GSE limit without compensating

factors (purple short-dashed line). The sharp decrease in originations above these levels (shown

in Figures 27 and 28), captures the increased underwriting requirements for loans with DTI

above these levels on characteristics that are observable in the data (e.g., credit scores, LTV) and

unobservable in the data (e.g., asset and savings requirements, lender accommodation).

Examples of such unobservable underwriting criteria required by one or more of the lenders

from the Application Data include reserve asset requirements of 10, 25, or 50 percent of the

original loan amount depending on the extent to which DTIs exceed 43 percent. Controlling for

the observable underwriting dimensions reduces the decline in delinquencies for those loans

above the GSE DTI threshold of 45, and the remaining gap is likely due to the unobserved

underwriting factors.

103 BUREAU OF CONSUMER FINANCIAL PROT ECTION

FIGURE 30: RELATIONSHIP BETWEEN EARLY DELINQUENCIES AND DTI, 2012 TO 2015 CONV ENTIONA L

PURCHA S E ORIGINA T IONS

The previous two figures use McDash servicing data, which has missing DTI values for a

substantial share of loans. Figures 31 and 32 show the relationship between DTI and early

delinquencies for the period 2012 through 2016 for nationally representative samples of

purchase originations of GSE and FHA insured loans, respectively, for which there exist

complete DTI data coverage in the NMDB dataset. For context, GSE and FHA loans represented

44.6 and 18.5 percent of purchase loans in 2014, respectively (Figure 16 in Chapter 3). In these

samples, the same strong positive relationship exists both unconditionally and with controls. As

highlighted earlier, Figure 31 shows that GSE early delinquencies decline for loans with DT Is

above 45 percent, likely due to unobservable underwriting criteria.

242

However, because FHA

242

T h is decrease i n GSE delinquencies abov e DT Is of 45 percent, t hough with a positive relationship below this level.

See also Karan Kaul & Laurie Goodman, Updated: What, If Anything, Should Replace the QM GSE Patch, Hous.

Fin . Pol’y Ctr. Commentary (2018), available at https://www.urban.org/research/publication/updated-what-if-

anything-sh ou ld -replace-qm -gse-patch.

104 BUREAU OF CONSUMER FINANCIAL PROT ECTION

originations do not decrease as substantially at DTI thresholds over 43 percent, the positive

relationship in Figure 32 continues through these higher levels.

243

FIGURE 31: RELATIONSHIP BETWEEN EARLY DELINQUE NCIES A ND DTI, GSE HOMEBUY ERS, 2012 TO

2016 PURCHA S E ORIGINATIONS

These figures document post-crisis loan performance relative to historical levels, and the basic

relationship between DTI and performance across large segments of the mortgage market. For

all periods and samples studied, the positive relationship between DTI and early delinquency is

present and economically meaningful. In all cases, the slope of the relationship is stronger for

the unconditional early delinquency rate, providing evidence that higher DTI is correlated with

other higher risk loan characteristics. However, because the positive relationship still exists after

243

Cu rrent FHA manual underwriting guidelines loans do a pply a dditional credit sc ore requirem ents or

com pensating factors (e.g., documented cash reserves, residual income tests, minimal increases in housing

pa y ments, si gnificant a dditional incom e l ike bonuses or ov ertim e) a t DT I thresholds of 4 0, 43, 4 7, and 50, but these

a r e generally less stringent than GSE r equirem ents, and thus have a more lim ited effect on origination patterns. See

U.S. Dep’t of Hous. & Ur ban Dev ., FHA Single Fam ily Ho using Po licy Handbook, at 25 (Dec. 30, 2016), available

at h t tps://www.hud.gov /sites/documents/40001H SGH.pdf.

105 BUREAU OF CONSUMER FINANCIAL PROT ECTION

controlling for other underwriting criteria, these figures suggest that higher DTI does

independently increase expected early delinquency, regardless of the other factors.

FIGURE 32: RELA TIO N S HI P BETWEEN EARLY DELINQUE NCIES A ND DTI, FHA HOMEB UY ERS, 2012 TO

2016 PURCHA S E ORIGINATIONS

The slope of the relationship in Figures 29 and 30 also appears to scale with the overall level of

early delinquencies between time periods, suggesting that adverse changes to the housing

market as a whole may lead to proportionately higher delinquencies for loans with higher DTIs.

Thus, while higher DTI loans have low overall early delinquencies in recent vintages, the

potential for higher DTI loans to default at higher rates in a weaker housing market persists. To

the extent that underwriting responses to the combination of GSE requirements and the Rule

limit such loans to a narrower set of consumers with strong borrowing characteristics, the Rule’s

General QM DTI threshold and Temporary GSE QM provision contribute to ensuring borrowers

receive loans they are able to repay.

106 BUREAU OF CONSUMER FINANCIAL PROT ECTION

4.4 Effects of the General QM DTI limit on

loan performance

To further assess whether the implementation of the Rule’s General QM DTI limit may have had

immediate effects on the early delinquency rate—which, as previously discussed, serves as a

proxy for measuring the effect of the Rule on ability to repay—this section first identifies the

market segments in which the Rule meaningfully changed loan origination behavior.

Specifically, loans covered by the Rule’s General QM DTI threshold likely saw a reduction in

originations with a DTI above the limit of 43 percent and may have increased originations just

below the limit. The latter effect would occur, for example, if borrowers started choosing to buy

homes of a somewhat lower value or putting down larger downpayments. The full set of

responses to the threshold may also have affected loan performance.

107 BUREAU OF CONSUMER FINANCIAL PROT ECTION

This section primarily examines first-lien, conventional, single-family purchase mortgages

originated in the year preceding (2013) and the year following (2014) the Rule’s effective date of

January 10, 2014. The focus is on comparing the origination and performance trends, before

and after the Rule, of a segment of loans not purchased by the GSEs (and therefore, unless

eligible for GSE or government agency purchase, guarantee, or insurance, or else made by and

held on the portfolios of Small Creditors, must comply with the General QM DTI limit to obtain

QM status) with a segment of loans that are purchased by GSEs (and therefore not subject to the

General QM DTI limit due to the Temporary GSE QM).

244

The trends in originations for these

segments are presented first in Section 4.4.1, while their performance is measured in Section

4.4.2.

245

These approaches draw in part on academic research into the Rule’s effects, notably

DeFusco, Johnson, and Mondragon (2017).

246

4.4.1 Effects on DTI distributions

To first demonstrate the starkest potential origination changes due to the General QM DTI

threshold, Figure 33 below shows DTI distributions of jumbo single-family purchase loans in the

McDash data, which are ineligible for GSE purchase due to their size, in 2013 (prior to the Rule’s

effective date) and 2014 (after the Rule became effective). For context, total jumbo purchase

originations increased from an estimated 108,700 to 1 30,200 between 2013 and 2014, based on

nationally representative NMDB data. While jumbo loans are not representative of the market

as a whole, their ineligibility for GSE purchase allows a clean look at changes for a market

segment where essentially all loans are subject to the General QM DTI threshold. Each point on

a given line shows the percentage of loans originated in that year which had DTI—rounded up to

the nearest whole number—equal to the level shown on the horizontal axis. Vertical lines

244

Not e t hat loans not p urchased by the G SEs could nonetheless b e QMs under the T emporary G SE Q M if su ch loans

were eligible for GSE purchase. The available loan performance data does not identify which non-purchased loans

w ere eligible for purchase, nor does it provide a reliable means to estimate purchase eligibility. In the data, the

pr esence of eligible ( and therefore QM) loans not purchased by the GSEs i s likely to lessen any performance

differences ob served between loans purchased and those not purchased by the GS Es. While they do n ot prov ide

per formance data, the Application Data used in Chapter 5 allows for the comparison of GSE-eligible and ineligible

a pplications and originations. Da ta on submissions to GSE Autom ated Underw riting Sy stem s are used in Chapter 6

to examine possible lender u tilization of the Temporary GSE QM for loans not sold to the GSEs.

245

T h ese c om parisons b efore and after the Ru le of l oans purchased and not p urchased by the G SEs are in the style of

so-ca lled “ differences-in-differences” analyses, with discussion and ev idence on the caveats and assumptions

r equ ired to interpret them as such developed i n the sections b elow. The further comparisons of average

per formance when splitting the loans within these segments above and below the 43 percent DT I threshold are in

t h e style of “triple difference” analyses, though the pot ential substitution of borrowers a cross the DT I threshold

r equ ires caution in interpreting them as such.

246

See A nthony A. Defusco, Stephanie Johnson & Joh n Mondragon, Regulating Ho usehold Leverage, (NW. Univ.

Kellogg Sch. of Mgm t., 2017). Available at SSRN: https://ssrn.com/abstract=3046564.

108 BUREAU OF CONSUMER FINANCIAL PROT ECTION

separate loans above and below the General QM DTI limit of 43 percent (gray dashed) and the

GSE limit without compensating factors of 45 percent effective during this period (purple short-

dashed). In 2014, the share of loans originated above a DTI of 43 percent fell, while the share of

loans originated at and just below a DTI of 43 percent increased. This change likely reflects

some general market trends from one year to the next, but may also reflect the ability of

borrowers to adjust DT I, with some borrowers who would have obtained a loan with a DTI

above 43 percent absent the Rule instead obtaining a loan with a DTI just below 43 percent.

109 BUREAU OF CONSUMER FINANCIAL PROT ECTION

FIGURE 33: CHA NGE IN DTI DISTRIB UTION FROM 2013 TO 2014, CONVENTIONAL JUMBO PURCHA S E

ORIGINA TIO NS

A pattern of “bunching” at DTIs below (but not exactly at) the limit of 43 percent in 201 4 may be

due to the difficulty of precisely measuring DTI as well as the fact that while some methods of

lowering DTI are continuous, such as that of increasing downpayments, other methods are

discrete in nature, such as purchasing a less costly home or eliminating other installment debt

payments by paying off such loans.

The General QM DT I threshold’s apparent effect of reducing jumbo originations at DTIs over 43

percent while potentially increasing them at DTIs under 43 percent is consistent with the

findings of DeFusco, Johnson, and Mondragon (2017). Using the 2013 to 2014 shift in DTIs for

loans of $41 7,000 or less to model the counterfactual shift for loans above that amount, the

paper estimates that the DTI limit caused 15 percent of originations abov e $417,000 with DT Is

over 43 percent that would have been made in 2014 absent the Rule to no longer be made with

the Rule in effect, and caused an additional 20 percent of these originations to be made at lower

DT Is. Chapter 5 explores these patterns and the implications for credit access in depth, using

110 BUREAU OF CONSUMER FINANCIAL PROT ECTION

new Application Data from nine lenders. In particular, the Application Data can distinguish

loans based on GSE eligibility rather than only GSE purchase, allowing for a more precise

accounting of borrower and lender responses to the Rule.

While the DTI distributions for jumbo loans are suggestive of the Rule’s effects, the following

two figures examine the broader and more representative comparison samples used for the loan

performance analysis in Section 4.4.2. Figure 34 shows the DTI distributions for a “treated”

sample including not only jumbo loans, but all conventional loans not purchased by the GSEs

within two years of origination. It is important to note that not all loans within this sample will

be subject to the General QM DTI threshold, for example those that qualify as Small Creditor

QM or are GSE eligible—and thus covered by the T emporary GSE QM—but not sold to the

GSEs.

247

Figure 35 shows the comparable DTI distributions for those loans that were purchased

by the GSEs within two years of origination, assuring that they were not subject to the General

QM DTI limit (the “control” sample). One potential concern for this comparison is that the

assignment of GSE eligible loans to either the treatment or control samples is not random, but

rather will reflect potential changes in the insurance and securitization choices made by lenders

after the Rule’s effective date, as well as changes in consumers’ loan choices. While lenders

could, in response to the Rule, choose to sell to the GSEs eligible loans with high DTIs that the

lender would have kept on portfolio absent the Rule, analysis in Chapter 6 suggests that such

substitution was not prevalent at the time the Rule was implemented. Looking at estimated total

conventional purchase originations in the NMDB data, GSE purchased loans decreased slightly

from 1,403,200 in 2013 to 1,397,500 in 2014, while non-GSE conventional purchase

originations increased from 732,700 in 2013 to 74 1,300 in 2014.

248

The pattern of a decreased shared of DTIs over 43 percent and an increased share of DT Is at or

below 43 percent is also present in the larger sample of loans not purchased by the GSEs (Figure

34), though less pronounced than for the subsample of only jumbo loans (Figure 33). This

suggests that any borrower or lender response was less pronounced for non-jumbo loans than

for jumbo loans.

247

Com parable to the limitations discussed in Foot note 244 regarding GSE eligible loans not purchased by the GSEs,

t h e available loan performance data does n ot identify which loans w ere originated by lenders eligibl e for the Sm all

Cr editor QM, nor does it prov ide a reliable m eans to estimate su ch eligibility. In the data, the presence of Sm all

Cr editor QM l oans is likely to l essen any performance differences observed betw een loans purchased by the GSEs

a n d those n ot purchased. Chapter 7 specifically analyzes the effects of t he Sm all Creditor QM c ategory.

248

Du e to differences in data availability, loans in the NMDB data are categorized as GSE or non-GS E b ase d on

w h ether they have been r eported in credit r ecords as purchased by a GSE as of September 2018, rather than within

tw o years of or igination.

111 BUREAU OF CONSUMER FINANCIAL PROT ECTION

FIGURE 34: CHA NGE IN DTI DISTRIBUTION FROM 2013 TO 2014, CONVENTIONAL NON-GSE PURCHA S E

LOA NS

In comparison, for the control group of loans purchased by the GSEs, Figure 35 shows that there

was an increase from 2013 to 2014 in the share of loans with DTI of 44 or 45 percent, while the

shares with DTI above 45 percent stayed comparable. This shift is consistent with the general

market trend towards higher DTI that can be observed throughout the DTI distribution in

Figures 28 and 35.

249

T h e increased share of high DT I loans i s similar just b elow t he 4 3 perc ent t hreshold and just abov e. G iven that

su bstitution into GSE securitization due to the Rule’s DTI threshold would be expected to occur only abov e the 43

per cent threshold, this pattern suggests limited substitution of this type.

112 BUREAU OF CONSUMER FINANCIAL PROT ECTION

FIGURE 35: CHA NGE IN DTI DISTRIBUTION FROM 2013 TO 2014, GSE PURCHA S E LOA NS

The next section evaluates how loan performance changed for loans not purchased by the GSEs

just above and below the General QM DTI threshold, relative to loans purchased by the GSEs

(and hence not subject to the General QM DTI threshold) above and below the threshold.

4.4.2 Effects on loan performance

Figure 36 compares early delinquency rates for GSE and non-GSE conventional purchase loans,

dividing the sample into two segments: 1) those with DTIs ranging from 30 to 43; and 2) those

with DT Is ranging from 44 to 50. In addition to originations from 2013 and 2014, the figure

113 BUREAU OF CONSUMER FINANCIAL PROT ECTION

adds originations from 2012 and 2015 to provide more statistical power for the comparisons.

250

The gray bars represent 95 percent confidence intervals.

Prior to the Rule’s implementation, non-GSE loans had higher early delinquency rates than GSE

loans at DTIs up to 43, and comparable early delinquency at DTIs above 43. Looking at changes

in delinquency from 2012-2013 to 2014-2015, GSE early delinquency rates increased for both

DT I bins (0.5 percent to 0.8 percent at DTIs below 43, 0.6 percent to 1.0 percent at DTIs above

43), while the non-GSE early delinquency decreased at DTIs up to 43 (1.0 percent to 0.7

percent) and remained steady at 0.6 percent for DTIs above 43.

251

Under an assumption that

GSE and non-GSE loans would have followed parallel trends absent the Rule, this would suggest

that lenders were more cautious in making non-GSE loans to borrowers with DTIs near the

General QM threshold as a result of the Rule but that the Rule did not similarly affect

underwriting for GSE loans with similar DTIs. Notably, the relative improvement of non-GSE

loans is seen both immediately above and immediately below the threshold, suggesting that

these differences may result from either more general responses to the Rule (beyond the DTI

threshold) or from compositional changes in the set of loans taken out by borrowers as part of

their shift from above the threshold to below.

250

Th e delinquency results using only 2013 and 2014 are qualitatively sim ilar, but less precise. Similarly, the sh ifts in

t h e DTI distributions w hen including the additional years of 2012 and 2015 are sim ilar to those shown in Figures 34

an d 35. Note also that the QRM risk-retention rule became effective in February 2015, and b ecause t hat rule

provides that a QM equals a QRM, the QM DTI thresholds are applicable to securitized residential mortgage loans

th at are QRMs.

251

Fu rther, G SE delinquency rates at DT Is ov er 43 p ercent exceeded the delinquency rates of n on-GSE (and therefore

g en erally non-QM) loans at these DT I lev els in 2014 to 2015.

114 BUREAU OF CONSUMER FINANCIAL PROT ECTION

FIGURE 36: EARLY DELINQ UE NCY RA TES BY DTI FOR GSE V ERSUS NON-GSE PURC HASE LOANS, 2012

THROUG H 2015

For non-GSE loans above the General QM DTI limit, their very low, albeit unchanged, early

delinquency rates combined with their reduced origination volume following the Rule’s

implementation (as shown in Figure 34) suggests that lenders continued to provide such loans

to only a limited segment of borrowers with strong creditworthiness along other underwriting

dimensions. T o further highlight the role additional underwriting criteria may play, Figure 37

shows performance for conventional purchase loans with DTIs above 43, split between those up

to the effective GSE limit of 45 without compensating factors, and those above 45 requiring such

factors. Consistent with tighter underwriting above these thresholds, the higher GSE early

delinquency rates after the Rule’s implementation are concentrated in originations with DTIs of

44 or 45, which exceed the early delinquency rates of the less common originations with DTI

exceeding 45. It is noteworthy that for both groups of loans the early delinquency rate for GSE

115 BUREAU OF CONSUMER FINANCIAL PROT ECTION

loans originated post-Rule increased whereas the early delinquency rate for non-GSE loans

remained relatively flat.

FIGURE 37: EARLY DELINQ UE NCY RA TES BY DTI (OV ER 43) FOR GSE V ERSUS NON-GSE PURCHA S E

LOA NS, 2012 THROUG H 2015

Overall, the Rule appears to have reduced the share of mortgages originated with DTI over 43

percent, while potentially increasing the share originated with DTIs at or just below 43 percent.

These patterns are studied in more detail in Chapter 5. Further, both above and below the DTI

threshold of 43 percent, the improvement in performance of non-GSE loans relative to GSE

loans provides some evidence that those loans that continue to be made under the General QM,

other non-T emporary GSE QM, or non-QM AT R guidelines are underwritten in a way that

reflects consumers’ ability to repay.

116 BUREAU OF CONSUMER FINANCIAL PROT ECTION

5. Effects of the Rule on

access to mortgage credit

and cost of credit

This chapter presents evidence regarding the impact of the ability-to-repay (ATR) requirement

on access to mortgage credit and cost of credit among borrowers who do not qualify for a QM

loan. The Bureau estimates that the segment of non-QM loans primarily consists of loans that

are not eligible for purchase by GSE’s, with debt to income ratios exceeding 43 percent. In the

home purchase category, such loans constituted approximately 1-3 percent of all loans in 2013.

Although the most common reason for such loans being not eligible for GSE purchase is loan

size (e.g. “jumbo loans”), the available data indicates that there may be a substantial number of

borrowers with DTI exceeding 43% that do not qualify for a GSE loan for other reasons. Such

borrowers may include those with irregular income, certain self-employed borrowers, and those

with little or no credit history. Although such borrowers may not fit into a standard GSE (or

FHA) product, or otherwise qualify for a QM loan, they may nevertheless have the ability to

repay. Unfortunately, the available data does not always distinguish all types of non-QM

borrowers.

The impact of the Rule on access to credit for non-QM borrowers derives primarily from the fact

that, relative to the pre-Rule period, such originations carry an extra risk (actual or perceived)

and impose extra costs for the lender, collectively referred to as “ATR risk.” It is a combined

result of a host of various risks and/or cost factors, such as: a) risk of litigation by private parties

asserting that the lender failed to assess ATR; b) cost of complying with documentation and

verification requirements of the Rule (if different from the pre-Rule practice); c) additional cost

of funds, due to a separate requirement, adopted by other federal agencies, that lenders retain

extra capital to cover the risk associated with non-QM loans; d) and, additional cost of funds due

to the cost of originating less liquid assets (non-QM loans are not easily sold on the secondary

market).

The impact is separately considered for two types of non-QM loans: a) loans with DTI>43

percent (further, “High DTI” loans) that are not eligible for purchase by GSEs; these are

primarily jumbo loans, with some presence of conforming size loans that aren’t eligible for GSE

117 BUREAU OF CONSUMER FINANCIAL PROT ECTION

purchase for other reasons; and b) loans where the sum of applicable points and fees exceeds the

QM limit (particularly, small balance loans). The available data do not allow for the study of the

impact of the Rule among other types of non-QM borrowers.

The chapter then goes on to examine the impact of the rebuttable presumption provision that

applies to first-lien mortgages with annual percentage rates (APRs) that are 1.5 or more

percentage points over the benchmark Average Prime Offer Rate (APOR) for a comparable

transaction, and second-lien mortgages with APRs that are 3.5 percentage points over the

comparable APOR.

The main findings are:

• Application level data obtained from nine large lenders (further, “Application Data”)

indicates that among these lenders, the Rule eliminated between 63 and 70 percent of

non-GSE eligible, High DTI loans for home purchase over the period of 2014 to 2016. In

absolute terms, this represents a loss of between 9,000 and 12,000 approved

applications for such loans among these lenders, over the period of three years. For

context, these lenders have approved approximately 615,000 applications for home

purchase during the same period. Thus, the number of displaced loans represents

between 1.5 to 2 percent of loans approved over three years. Notably, the impact of the

Rule in the refinance category is much more muted than in the purchase category.

252

This is consistent with a notion that consumers seeking to refinance a mortgage having

already demonstrated some ability to repay, thereby lowering ATR risk and making

lenders more likely to extend credit.

• The findings from the Application Data are corroborated by a lender survey conducted

by the Bureau for this assessment. Among 89 lenders who responded to the appropriate

survey question, 30 indicated introducing a 43 percent DTI limit or not originating non-

QM loans that the lenders intend to hold rather than selling (“portfolio loans”) and loans

intended for sale to investors other than the GSEs or a government agency. Recent

research by the Federal Reserve Board and academic economists also suggests

significant reductions in lending among non-QM High DTI borrowers following the

Rule.

252

Som e of t he r efinanced m ortgages may have been loans g overned by 1 2 C.F.R. § 1 026.43(d). Unfortunately, the

data on loans being refinanced is not available.

118 BUREAU OF CONSUMER FINANCIAL PROT ECTION

• The analysis of characteristics of rejected applications suggests that the Rule did not

have a differential impact on access to credit among particular categories of borrowers,

along dimensions such as credit score, income and downpayment amount. Thus, the

observed effect on access to credit was likely driven by lenders’ avoidance of litigation or

other risks associated with the ATR requirement, rather than by rejections of borrowers

who were unlikely to repay the loan.

• There is significant heterogeneity in the extent to which lenders have tightened credit for

non-GSE eligible High DTI borrowers after the Rule. This heterogeneity in lender’s

responses to the Rule, and its persistence during the years following the Rule, is

consistent with a notion that the industry has not developed a common approach to

measuring and predicting ATR risk, as it has accomplished for other types of risk, such

as prepayment and default.

• The Application Data indicates that, notwithstanding concerns that have been expressed

about the challenge of documenting and verifying income for self-employed borrowers

under the General QM standard and the documentation requirements contained in

Appendix Q to the Rule, approval rates for non-High DT I, non-GSE eligible self-

employed borrowers have decreased only slightly, by two percentage points.

• One of the criteria for a QM is that the total points and fees charged at the time of

origination cannot exceed a set limit, which is 3 percent of the loan amount for loans

above 100,000 dollars; higher limits apply to mortgages with smaller balances. This

research finds that non-QM loans where the sum of applicable points and fees exceeds

the QM limit are generally not originated. According to conversations with lenders,

instances when an application indicates that the points and fees limit will be exceeded

are sufficiently rare that lenders handle them on a case by case basis. The lender survey

indicates that the violation is typically remedied by waiving certain fees, with or without

a compensating increase in the interest rate; denying an application is rarely done. The

analysis of data reported by lenders under the Home Mortgage Disclosure Act on

approval rates of small balance loans similarly indicates that the Rule likely had no effect

on access to credit for such loans.

• Research using HMDA data indicates that the rebuttable presumption status applicable

to HPML loans did not reduce access to such loans by consumers, both in the site-built

and in the manufactured housing segments.

When interpreting these results, one must keep in mind that the credit standards were already

relatively tight by the time the Rule took effect; it is possible that the impacts would be different

during times when credit is more abundant.

119 BUREAU OF CONSUMER FINANCIAL PROT ECTION

5.1 Market trends in origination of loans

with DTI greater than 43 percent

The Bureau has utilized two servicing datasets: McDash and CoreLogic, for measuring the

originations of non-GSE eligible loans with DTI>43 percent. Although neither of these datasets

is statistically representative of the market, both are large datasets, with millions of loan level

observations, covering 30 to 40 percent of conventional originations for home purchase. In the

GSE segment, the National Mortgage Database (NMDB) provides a representative share of High

DTI originations. The sample of GSE loans included in NMDB is representative of the

population of GSE loans, and the DTI information is provided directly by the GSEs (whereas it is

often missing in the servicing datasets).

FIGURE 38: SHA RE OF GSE HOME PURCHA S E LOA NS WITH DTI A BOV E 43 PERC E NT, 2012-2017 BY DATA

SOURCE

120 BUREAU OF CONSUMER FINANCIAL PROT ECTION

FIGURE 39: SHA RE OF NON-GSE PURCHASE LOANS OVER $417,000 WITH DTI ABOVE 43 PERC E NT, 2012-

2017 BY DATA SOURCE

Figures 38 and 39 plot shares of High DTI originations over time, separately for GSE and non-

GSE loans, among first lien loans for home purchase. The loan is labeled as “GSE” if the data

indicates it was sold to the GSEs within 2 years of origination; otherwise, it is labeled as Non-

GSE. Because some GSE eligible loans are never sold to the GSEs, the set of loans in the Non-

GSE category is further restricted to loans with amount above $417,000. In the GSE segment,

both CoreLogic and NMDB show the share of High DTI loans at approximately 15 percent before

the Rule, and growing after the Rule. Notably, McDash shows a much lower value,

approximately 10 percent before the Rule. It is not clear what drives the difference between

these two datasets. In the Non-GSE segment, NMDB data is sparse and not shown in Figure 39;

both servicing datasets indicate that in the post-Rule period, the share of High DTI loans among

non-GSE loans is fluctuating between 4 and 8 percent.

The General QM requirements of the Rule apply to loans in the Non-GSE segment, but not to

loans in the GSE segment, which serves as a control group. Although the share of High DTI

loans in the Non-GSE segment has clearly dropped after the Rule, the visual inspection of Figure

121 BUREAU OF CONSUMER FINANCIAL PROT ECTION

39 is not a reliable method of identifying the impact of the Rule on credit access for High DT I

borrowers in the Non-GSE segment. The general concern is that due to reasons unrelated to the

Rule, such as house price growth, the number of High DT I borrowers seeking to purchase a

home may be increasing over time.

253

Because GSE lending and non-GSE lending generally have

different geographic footprint (with non-GSE borrowers, primarily jumbo borrowers, being

concentrated in metropolitan areas), the impact of the house price growth on the proportion of

High DTI borrowers is likely different between two segments. If house price growth did not

occur after the Rule was introduced, the observed declines in the share of High DTI loans in the

Non-GSE segment would have been deeper than what is currently observed. Similar concerns

apply to other relevant characteristics of applicants, such as credit score, income,

downpayment—all of which may be affected by changes in economic conditions unrelated to the

Rule. To properly control for these changes, Section 5.3 provides an econometric analysis using

application level data.

254

In terms of pricing, Figure 40 shows the interest rate on loans for High-DTI and non-High DT I

loans over time among the non-GSE loans over $417,000. The figure compares loans with DTIs

between 40 and 43 percent (i.e. just below the General QM threshold) and those with 44 to 45

percent (i.e. just above the threshold).

255

There does not appear to be a marked change in the

relative price of High DTI loans in the year following the Rule. The difference between the two

interest rates (shown on the left vertical axis) becomes positive, albeit fairly small, in late 2015

and early 2016.

253

Com m enters noted c om parable trends in data and r eports produced by the Urban In stitute’s H ousing Finance

Poli cy Center, available at h ttps://www.urban.org/policy-centers/housing-finance-pol icy -center, and by the

Am erican Enterprise In stitute’s Center on H ousing Market s and Finance, available at h t t p : //w ww . a ei.or g /h ou sin g /.

254

Ca l culations using C oreLogic data (which includes or iginated loans, n ot a pplications) suggest a 3 5 percent decline

in t he origination of Hi gh DT I loans ov er $417,000 after the Rule, w ith 15 p ercent n ot originated a t all and 20

per cent shifting t o l ower DT Is. See “Regulating Household Leverage,” by Anthony DeFusco, Stephanie Johnson

a n d John Mondragon, available a t https://ssrn.com/abstract=3046564.

255

T h e tighter the b and around 43 percent, the m ore likely that the loans are c om parabl e, but the sm aller are the

sa m ple sizes. For the b ands chosen, t he average m onthly sam ple size is 732 for DT I b etween 40 and 43 percent and

2 1 9 for DT I between 44 and 45 percent.

122 BUREAU OF CONSUMER FINANCIAL PROT ECTION

FIGURE 40: INTERES T RA TES ON NON-GSE LOANS OVER $417K BY DTI, 2012-2016

This finding is in contrast with research recently done by the Federal Reserve Board

256

, which

utilizes application data from Optimal Blue (a platform that provides rate locks for lenders), and

has shown that non-QM High DTI loans are more expensive than comparable non-High DT I

loans by approximately 25 basis points (2013 to 2018 average). This research also finds that

there was not an immediate increase in the relative cost of High DTI loans after the Rule, but

rather a gradual increase during 2015 through 2018. The difference in results may be

attributable to the fact that different lenders contribute their data to McDash and Optimal Blue

datasets. T he Bureau’s own research suggests that not all lenders charge extra for a non-QM

loan (more details in the next section).

256

A urel Hizm o & Shane Sherlund, The Effects of the Ability-to-Repay Rule/Qualified Mortgage Rule on Mortgage

Lending, FEDS Notes (Nov . 1 6, 2018), available at https://www.federalreserve.gov/econres/notes/feds-

notes/effects-of-the-ability-to-repay-qualifi ed-m ortga ge -rule-on-mortgage-lending-20181116.htm.

123 BUREAU OF CONSUMER FINANCIAL PROT ECTION

5.2 Evidence from the lender survey

In summer 2018, the Bureau conducted a survey of mortgage lenders in order to gain insight

into policy responses to the Rule. An email with a survey link was sent to almost 2,000 lenders

using email addresses from the Bureau’s HMDA operations. In total, 195 responses were

received. Twenty five respondents did not answer most of the questions in the survey and two

respondents were Community Development Financial Institutions (CDFIs) and are therefore

not covered by the Rule. The survey was not sent to the nine lenders that provided the

Application Data; instead, these lenders provided more detailed written responses regarding

their policy responses to the Rule. This information was sufficient to impute answers to a subset

of questions on the survey for these lenders. Thus, the total number of respondents in the results

presented below is 177; however, the actual number of respondents depends on the specific

question. Although the sample of respondents is not statistically representative of the overall

population of mortgage lenders and, like any such survey, is subject to non-response bias, it

includes a diverse group of lenders.

The purpose of this section is to summarize results from the survey that are relevant to the issue

of the impact of the Rule on access to credit and the cost of credit; it is not meant to be a

complete summary of the survey. Other parts of this report make use of data provided by this

survey where relevant.

TABL E 2 : WHICH OF THE FOLLOWING OPTIONS BEST DESCR I B ES THE TY PE OF Y OUR INSTITUT ION?

CHOOS E ONE.

Institution Type Count of Respondents Percent of Respondents

Bank with <$2 billion in total ass et s 45 25%

Bank with $2-10 billion in total assets

Bank with >$10 billion in total assets

13%

Credit Union

13%

Non-DI

40%

Total responses

177

100%

124 BUREAU OF CONSUMER FINANCIAL PROT ECTION

TABL E 3 : HOW MA NY MORTGA G ES DID Y OUR INSTITUT ION DIRECTLY ORIGINA T E IN 2017? PLEASE

USE COUNT OF LOA NS.

Origination volume in 2017 Count of respondents Percent of respondents

0-299

17.8%

300-499 17 9.8%

500-999

9.8%

1000-1999 29 16.7%

2000-4999 26 14.9%

5000-9999 26 14.9%

10000-19999

5.2%

>=20000 19 10.9%

Total responses

174

100.0%

Table 2 and Table 3 provide the breakdown of respondents by institution type and by the

volume of originations in 2017. Lenders of every category have provided a meaningful number

of responses to the survey.

TABL E 4 : AMONG MORTGAGES YOUR INSTIT UT IO N ORIGINA TED IN 2017, WHAT WA S THE COMBINED

SHARE ELIGIBLE TO BE PURCHA S ED, GUA RA NTEE D OR INSURE D BY A GSE, FHA, VA, OR

USDA /RHS?

Percentage of originations

(count of loans)

Count of respondents Percent of respondents

Less than 80% 94 55%

Approximately 80% 13 8%

Approximately 90%

21%

100% (all loans)

16%

Total responses

171

100%

The degree to which a mortgage lender’s business is potentially affected by the requirements of

the Rule is represented by the share of originations that are not eligible for purchase or

guarantee by the GSEs, FHA or VA, because such loans generally need to satisfy the General QM

requirements in order to obtain QM status. According to Table 4, for 28 respondents all or

almost all loans were eligible for purchase or guarantee by GSE/FHA/VA. This suggests the

General QM requirements of the Rule currently do not affect those lenders; however, it is

possible that some of them may have decided to originate only T emporary and Agency QM loans

in response to the Rule.

125 BUREAU OF CONSUMER FINANCIAL PROT ECTION

TABL E 5 : AMONG MORTGAGES Y OUR INSTITUTION ORIGINA T ED IN 2017, WHA T WA S THE COMBINED

SHARE ELIGIBLE TO BE PURCHA S ED, GUA RA NTEE D OR INSURE D BY A GSE, FHA, VA, OR

USDA /RHS? BREAKDOW N BY INSTIT UT I O N TY PE.

Share of originations Depository institution Non-DI (Independent mortgage banker)

Less than 80% 74 20

Approximately 80% 12 1

Approximately 90% 15 21

100% (all loans) 3 25

Total responses 104 67

Note: only includes observations where response to both questions was provided.

According to Table 5, out of 28 respondents who originate only T emporary QM loans or Agency

QM loans, the majority (25 out of 28) are non-depository lenders. The inv erse is not true,

however: there are 42 non-depository lenders who reported originating loans that do not meet

the Temporary QM or Agency QM standards. Likely, these are jumbo loans that are

subsequently sold to private investors.

TABL E 6 : CONSIDER Y OUR 2013 BUSINESS MODEL. WHAT SPECIFI CA LLY CHA NGED A S A DIRECT

RESULT OF THE A TR/QM RULE?

Response Count of respondents Percent of respondents

Business model changed 100 62.50%

Does not apply/No change 60 37.50%

Total responses 160 100.00%

Among the 160 lenders who responded to question in Table 6, approximately 63 percent

indicated that the Rule had an impact on their business operations. For the remaining 37

percent, it may be inferred that the Rule did not produce a material impact; this may occur

either because these lender’s business only focused on originating QM loans, or because their

lending standards already were in compliance with the requirements of the Rule.

126 BUREAU OF CONSUMER FINANCIAL PROT ECTION

TABL E 7 : IMPA CT OF THE A TR/QM RULE ON THE BUSINES S MODEL BY THE SHA RE OF 2017

ORIGINATIONS ELIGIBLE TO BE PURCHA SED, GUA RANTEED OR INSURED BY A GSE, FHA, VA,

OR USDA /RHS.

Share of originations Business model changed Does not apply/No change

Less than 80% 55 32

Approximately 80% 10 2

Approximately 90% 19 12

100% (all loans) 15 13

Total responses 99 59

Note: only includes observations where response to both questions was provided.

Table 7 suggests that the set of lenders who reported that their business model was not changed

by the Rule is not identical to the set of lenders who originate only Temporary or Agency QM

loans. In sum, the reported impact of the Rule is not restricted to institutions of particular type.

Among the 100 lenders who responded that the business model has changed, 87 provided write-

in responses detailing what specifically has changed.

TABL E 8 : ANALYSIS OF WRITE-IN RESPONSES TO THE QUESTIO N "CONSIDE R Y OUR 2013 BUSINESS

MODEL. WHA T SPECIFICA LLY CHA NGED A S A DIRECT RESUL T OF THE ATR/QM RULE?"

Issue mentioned

Count of

respondents

Percent of

respondents

Increased income documentation

36%

DTI cap of 43% was introduced

32%

Products with balloon feature discontinued

20%

Increased staffing / compliance costs

15%

Determined not to originate non-QM

13%

Changes to cap structure or income requirements for ARM

products

13 15%

Products with interest only feature discontinued

18%

Difficulties with meeting points and fees test

13%

Longer closing times

Asset depletion no longer allowed

Total responses

100%

Table 8 lists the issues mentioned by respondents, sorted in the order of declining frequency.

The percentages in the right most column of do not sum up to 100 percent because some

respondents indicated multiple issues. Two findings are notable. First, “Increased income

127 BUREAU OF CONSUMER FINANCIAL PROT ECTION

documentation” is the most frequently mentioned change that was prompted by the Rule. This

finding is somewhat surprising given the general notion that income documentation standards

already had been fairly strict at the time of the introduction of the Rule.

257

Some respondents

explicitly link the increased documentation to Appendix Q requirements, while others mention

general ability to repay requirement as the reason. Second, the third most popular issue is

“Products with balloon feature discontinued”, mentioned by 19 percent of respondents.

Some respondents have indicated that the business model change was to stay away from non-

QM originations, either through a DTI cap of 43 percent on portfolio or investor loans, or as a

more general policy of not originating non-QM loans regardless of the reason. Nevertheless, a

number of lenders do originate non-QM loans, according to Table 9.

TABL E 9 : WHA T SHA RE OF Y OUR 2017 ORIGINA TIO NS IS REPRESENTED BY NON-QM LOA NS? CHOOSE

ONE OPTIO N.

Non-QM sha re

Count of respondents

Percent of respondents

None

30%

<5%

44%

>5%

21%

Do not know

Total responses

167

100%

Among lenders who provided responses to this question, 30 percent mentioned not originating

any non-QM loans. Among those who report originating non-QM loans, the majority indicated

that the share of such loans among their originations was low, less than 5 percent. The Bureau

has obtained more detailed data from several large lenders, including those who provided the

Application Data; generally, the share of non-QM loans was found to be less than 1 percent.

257

7 8 Fed. Reg. 6408, 6564 (Jan. 30, 2013).

128 BUREAU OF CONSUMER FINANCIAL PROT ECTION

TABL E 10 : ORIGINA TIO N OF NON-QM LOANS BY INSTITUTIO N TY PE

Non-QM

sha re

Bank with <$2

billion in total

a sse ts

Bank with $2-

10 billion in

total assets

Bank with

>$10 billion in

total assets

Credit Union Non-DI

None

38%

36%

29%

37%

<5%

38%

18%

64%

38%

54%

>5%

25%

45%

32%

33%

Total

responses

40 11 22 21 65

Note: only includes observations where response to both questions was provided.

Table 10 examines which institution type is more likely to originate non-QM loans. The last row

in the table indicates the number of responses in the corresponding column. Non-depository

lenders (“Non-DI”) are significantly less likely to originate a substantial amount (“>5%”) of non-

QM loans than any other lender type. At the same time, the percentages on the row labeled

“None” indicate that non-depository lenders do originate some non-QM loans at a rate that is

comparable to banks. Originations of non-QM loans by non-depository lenders, who generally

do not hold loans on balance sheets, suggest there exists a secondary market for this type of

loans. Large banks, with more than $10 billion in assets, almost all originate some non-QM

loans; this is in contrast to other institution types, where about a third do not originate any non-

QM loans. Chapter 7 provides additional information on non-QM originations by small and

medium banks, utilizing a separate survey conducted by the Conference of State Bank

Supervisors.

TABL E 11 : DO Y OU SELL A NY OF Y OUR NON-QM LOA NS TO THIRD PA RTIES ? BREA KDOW N BY

INSTIT U T IO N TY PE.

Response

Depositary

institution

Non-DI (independent mortgage

banker)

No, we keep all or almost all such loans on

portfolio

61 3

Yes, we sell most or all of our non-QM loans 1 32

Yes, we sell some of our non-QM loans 3 1

Total responses 65 36

Almost all depository institutions hold non-QM loans they originate on portfolio, as Table 11

suggests. Conversations with lenders suggest that one possible explanation is that depository

institutions originate non-QM loans on an occasional basis through their general portfolio

129 BUREAU OF CONSUMER FINANCIAL PROT ECTION

products, while some non-depository lenders maintain specialized non-QM mortgage products,

financed by investors.

TABL E 12 : A MONG THE NON-QM LOA NS THA T Y OU ORIGINA TE, DO AT LEAST SOME OF THEM HA V E

THE FOLLOWING FEA TURES?

A jumbo loan with DTI>43%

Count

Percent

Rarely or never

46%

Sometimes

49%

Often

Total responses

101

100%

A non-jumbo loan with DTI>43% (Only consider mortgages not

eligible to be purchased, guaranteed or insured by a GSE, FHA,

VA, or USDA/RHS)

Count Percent

Rarely or never

32%

Sometimes

56%

Often

12%

Total responses

100

100%

Borrower did not (could not) provide documentation required by

Appendix Q (Only consider mortgages not eligible to be

purchased, guaranteed or insured by a GSE, FHA, VA, or

USDA/RHS)

Count Percent

Rarely or never

60%

Sometimes

31%

Often

Total responses

100%

As Table 12 indicates, the phenomenon of non-QM High DTI loans is not restricted to the jumbo

segment. This finding suggests that the Rule may have had an impact on originations of loans

that are conforming in loan size but do not fit into the GSE guidelines on other parameters. The

analysis in the following section provides further insight into this issue.

TABL E 13 : OV ER THE NEXT Y EA R, DO Y OU EXPECT Y OUR INSTITUT ION’ S NON-QM LENDING WILL:

Response

Count of respondents

Percent of responses

Decrease

Increase

26%

Stay about the same

69%

Total responses

100%

130 BUREAU OF CONSUMER FINANCIAL PROT ECTION

When asked about expectations regarding the future growth of non-QM originations, 70 percent

of respondents who replied to the question indicated that it would “Stay about the same”

according to Table 13 above.

Finally, with regards to pricing, the Bureau inquired whether lenders apply extra pricing

adjustment for non-QM loans when the DTI exceeds 43 percent. T his policy option is of

particular interest as it represents an alternative to rejecting a non-QM application. Table 14

presents the count and share of respondents who responded “yes” to the question whether they

applied a pricing adjustment in situations where the DTI on a mortgage loan exceeded 43

percent. Responses are restricted to lenders who did not qualify for a Small Creditor QM status.

The respondents were asked to only consider mortgages not eligible to be purchased, guaranteed

or insured by a GSE, FHA, VA, or USDA/RHS.

TABL E 14 : COUNT OF RESPO ND ENTS A PPLY ING PRICING A DJUSTME NT FOR LOA NS OV ER 43 PERCENT

DTI (ONLY LENDERS THA T DO NOT QUA LIFY FOR A SMA LL CREDITO R QM)

Institution type Count of respondents

Apply adjustment

Bank with <$2 billion in total assets 12 3

Bank with $2-10 billion in total assets 7 4

Bank with >$10 billion in total assets 20 5

Credit Union 10 0

Non-DI 32 11

Total 81 23 (28%)

Note: only includes observations where response to both questions was provided.

Overall, about 28 percent of respondents indicated applying a pricing adjustment. The Bureau

has investigated this issue further by examining retail ratesheets from a number of lenders

(approximately, 40). Only a few lenders from the examined set have a pricing adjustment that

applies specifically to High DTI loans. None of the nine lenders who provided Application Data

apply such an adjustment. Overall, it appears that using extra pricing adjustment to compensate

for ATR risk is a less popular policy response to the Rule among lenders, as compared to

tightening of underwriting standards (introducing a 43 percent DTI cap), particularly when

institution size is taken into account.

131 BUREAU OF CONSUMER FINANCIAL PROT ECTION

5.3 Effect of the Rule on access to credit for

borrowers with DTI greater than 43

percent: evidence from the Application

Data

5.3.1 Description of the data

To analyze the impact of the ATR requirement on access to credit, the Bureau acquired de-

identified, application level data from nine large lenders, including depository and non-

depository institutions, spanning the four years from 2013 to 2016.

258

None of the lenders is a

small creditor under the Rule and none are credit unions. Although these lenders account for a

significant percentage of mortgage originations (over 15 percent of jumbo originations in 2016),

they are not representative of the entire market and thus the analyses that follow must be read

with that limitation in mind.

For each lender, the data contain information on each application received by the lender, its

affiliates, and correspondent lenders and brokers for a closed-end, first-lien consumer mortgage

to purchase or refinance an owner-occupied one to four family residential property. If several

applications were submitted by an applicant with respect to a single transaction, then

respondents were instructed to include information on only the final application. Respondents

were instructed not to include pre-approval requests.

For each application, the available data fields include the outcome of the loan application—

approved, denied, or withdrawn—as well as a broad set of characteristics of the borrower and of

the mortgage. See Appendix C for details on the available data fields and their values. To

minimize the risk of re-identifying individual borrowers, the numeric data fields, such as loan

amount, loan-to-value (LTV) ratio, income reported on the application, etc., were reported in

bins with each bin identified by the range of included values. Furthermore, the date of

application was coarsened to the year-month level, and the location of the property was

recorded at the county level.

The universe of applications is classified into loan types, according to the mortgage product

applied for: GSE, FHA, VA, USDA/RHS, and Private. The latter is a catch-all category that

258

See Appendix C for details, including the data dictionary.

132 BUREAU OF CONSUMER FINANCIAL PROT ECTION

represents privately funded loans (either held in portfolio or by investors). Such classification is

based on the type of product the borrower applied for. For instance, the “GSE” category consists

of applications for GSE mortgage products, or products where the lender sells most or all loans

to the GSE’s. Applications in the “Private” category are those which the lender either retains on

portfolio or sells to private investors. Lender-level statistics reported in this chapter are de-

identified and randomized across tables to reduce re-identification risk.

The Temporary GSE QM provision of the Rule maintains that loans eligible for purchase by

GSE’s generally are QM loans. Such loans constitute a control group for the purposes of this

analysis.

259

This group consists of GSE applications (for loans that would have been

subsequently sold to GSEs) and of GSE eligible Private applications (for loans that generally

would not have been sold to GSEs). Indeed, not all GSE eligible loans are sold to GSEs by the

nine lenders that contributed the Application Data. The data includes an appropriate indicator

that distinguishes between GSE eligible and non-GSE eligible Private applications. The Bureau

assumes for purposes of analysis that Private non-GSE eligible applications must satisfy General

QM provisions of the Rule in order to obtain the QM status.

260

Applications from other segments (FHA, VA, and USDA/RHS) are not used in this analysis

because they are subject to these agencies’ own QM rules.

TABL E 15 : COMPOS ITIO N OF THE APPLICATION DATA BY LEND ER

GSE Eligibility Min Mean Max

GSE

0.48

0.82

100

Private GSE eligible

0.00

0.03

0.24

Private Non-GSE eligible

0.00

0.15

0.33

Table 15 provides the breakdown of applications by: GSE applications, Private GSE eligible

applications and Private non-GSE eligible applications. The sample includes all conventional

applications for home purchase or refinance, 2013 to 2016 (all years of data). On average across

259

Im portantly, the DTI requirements for GSE eligibility have r emained unchanged during the study period (2013 to

2 01 6) .

260

Sec tion 1 026.43(4)(ii)(A) provides t hat a QM mortgage must be eligible for purchase or guaranty by the GSE

“except with regard to matters wholly unrelated to ability to repay . . . .” It is therefore conceivable that som e

per centage of n on-GS E eligible l oans nevertheless c ould m eet the r equirement for the Tem porary G SE Q M

pr ovision if the ineligibility was attributable to a matter wholly unrelated to the ability to r epay. The Application

Da t a does n ot a llow such differentiation, hence the assumption m entioned ab ove.

133 BUREAU OF CONSUMER FINANCIAL PROT ECTION

nine lenders, the share of GSE applications was 82 percent, the share of Private GSE eligible

applications was 3 percent, and the share of Private non-GSE eligible applications was 15

percent. The share of the Private non-GSE eligible applications (treated group) determines the

degree to which the lender’s business is potentially affected by the General QM DTI

requirement; that share varies significantly by lender, between 0 and 33 percent. One lender

only originates GSE loans, which means that this lender’s data is entirely in the control group.

The heterogeneity in the share of non-GSE eligible loans holds more broadly across mortgage

lenders, as seen from responses to the Lender Survey.

In total, the sample includes close to 3.5 million applications for GSE products, and close to half

a million applications for Private products. The large number of observations is important as it

provides sufficient power to identify effects in small segments. The large size of the dataset also

implies that the estimated impacts of the Rule, albeit obtained for a non-representative sample

of lenders, affect large number of borrowers.

Table 16 and Table 17 compare GSE eligible and non-GSE eligible applications along a number

of loan and borrower characteristics. Because each characteristic is recorded as a categorical

variable, these tables show percentages separately for each sub-sample. For example, Table 16

indicates that 23.45 percent of applications in the GSE eligible category were made for home

purchase; in the non-GSE eligible category, the share of home purchase applications is 52.43

percent. T he primary difference between the two c ategories is loan size: over 70 perc ent of non-

GSE eligible applications exceed the general $417,000 conforming limit effective at the time,

whereas only about 4 percent of GSE eligible applications exceed this limit, all located in high-

cost counties. In other words, 30 percent of non-GSE eligible applications are within conforming

limits, which suggests that loan size alone (e.g., “jumbo loan”) is a rather imperfect proxy for

non-GSE eligibility. On dimensions other than loan size, it is often possible to find comparable

GSE eligible borrowers for a given non-GSE eligible borrower who may serve as a control group.

134 BUREAU OF CONSUMER FINANCIAL PROT ECTION

TABL E 16 : DISTRIBUTION OF A PPLICA T ION A ND B O RROW ER CHA RA CT ERIS TICS BY A PPL ICATION TY PE,

2013-2016

Variable

Percent of GSE eligible

applications

Percent of Non-GSE eligible

applications

Decision

Approved

70.03

67.26

Denied

17.25

16.04

Withdrawn

12.72

16.70

Loan Purpose

Purchase

23.45

52.34

Refinance

76.55

47.66

Loan Amount

< 60,001

4.54

3.52

60,001-100,000

14.42

5.00

100,001-150,000

21.28

5.72

150,001-250,000

30.87

7.44

250,001-417,000

25.06

6.95

417,001-625,000

3.38

30.79

> 625,000

0.46

40.59

FICO Score

<620

4.46

8.23

620-659

7.19

2.02

660-679

6.05

2.25

680-699

8.35

4.50

700-719

9.66

7.39

720-739

10.20

10.54

>=740

54.08

65.07

Back-end DTI

< 21%

13.12

14.71

21-30%

22.06

22.65

31-40%

31.24

36.76

41-43%

10.36

13.69

44-45%

8.40

3.61

46-50%

5.93

3.68

> 50%

8.89

4.91

135 BUREAU OF CONSUMER FINANCIAL PROT ECTION

TABL E 17 : DISTRIBUTION OF A PPLICA T ION A ND B O RROW ER CHA RA CT ERIS TICS BY A PPL ICATION TY PE,

2013-2016 (CONTINUED)

Variable

Percent of GSE eligible

app

lications

Percent of Non-GSE eligible

applications

Application income, dollars per

month

< 2

,501

9.05 11.21

2,5

01-5,000

24.62 7.83

5,0

01-7,500

23.99 4.58

7,5

01-10

,000

17.22 6.57

10,001

-12,500

10.45 9.11

12,501-15,000 5.86 9.75

> 15,000

8.81 50.94

LTV

< 5

17.9 15.92

51-80% 54.86 61.4

81-90% 11.63 6.58

91-95% 9.18 4.38

> 95%

6.43 11.72

Number of borrowers

1 51.69 44.78

2 48.31 55.22

Self-employed

borrower

No 86.13 77.02

Yes 13.87 22.98

Fixed rate mortgage

No 3.61 27.19

Yes 96.39 72.81

Delin

quency on other loans

No 99.6 99.67

Yes 0.4 0.33

Bankrupt

No 98.19 99.53

Yes 1.81 0.47

136 BUREAU OF CONSUMER FINANCIAL PROT ECTION

FIGURE 41: SHA RES OF HIGH DTI A PPLICATIONS A MONG GSE EL IGIB L E A ND NON-GS E EL IGIBL E

CATEGORIES, BY LENDER, 2013

Figure 41 provides, separately for each lender, the share of High DTI applications in the GSE

eligible and non-GSE eligible segments in 2013. The pre-Rule data is used to eliminate the

influence of the Rule on the data. Almost all points lie above the 45-degree line, meaning that for

each lender, the share of High DTI applications was higher in the GSE eligible segment than in

the non-GSE eligible segment. Before the implementation of the Rule, there was wide variation

across lenders in the proportion of High DTI applications, in both segments, ranging from 10 to

30 percent.

In addition, shares of High DTI applications in both segments are positively related. This could

result from common factors, such as the geographic footprint of a lender, but also from the

lender’s credit policy towards High DTI borrowers. Note that not all borrowers who contact a

lender end up submitting an application. From the borrower’s perspective, filing an application

137 BUREAU OF CONSUMER FINANCIAL PROT ECTION

requires effort and often a fee, which makes sense only if there is a reasonable expectation of

approval, which itself is a function of the lenders’ underwriting approach. Therefore, this

analysis considers the share of High DTI applications as an outcome that may be affected by the

Rule, along with the more traditional outcome, the approval rate of High DTI borrowers.

FIGURE 42: APPROVAL RATES BY DTI A ND BY LOA N TY PE, 2013

The approval rate is defined as the ratio of approved applications to all applications. Figure 42

plots approval rates by DTI bin, separately for GSE eligible and non-GSE eligible applications,

using 2013 data to capture the state of the market before the Rule took effect. There is no

significant difference in how lenders approach applications with DTI in the [41-43%] bin, and

applications in the [44-45%] bin. Also, there is almost no difference in approval rates between

GSE eligible and non-GSE eligible applications in the [44-45%] bin. In other words, the 43

percent cutoff was immaterial in the pre-Rule environment from the point of view of credit

policy. This finding helps us identify the impact of the Rule, because it alleviates a concern that

applicants just above 43 percent were different from those just below 43 percent on dimensions

not observed in the data.

138 BUREAU OF CONSUMER FINANCIAL PROT ECTION

FIGURE 43: SHARE VS APPROVAL RATE OF HIGH-DT I NON-GSE ELIGI BL E A PPL ICA T IONS, 2013

Figure 43 plots two key indicators—the share of High DTI applications and the approval rate of

High DTI applications—against each other, among non-GSE eligible applications, using 2013

data. There seem to be two clusters of lenders, but within each cluster the relationship between

the two indicators is positive, suggesting that both are likely influenced by a given lender’s

underwriting approach towards High DTI borrowers.

139 BUREAU OF CONSUMER FINANCIAL PROT ECTION

FIGURE 44: CHA RA CTERIS T ICS OF HIGH-D T I A PPL ICA T IO NS, 2013

Figure 44 compares characteristics of High DTI applications between two segments: GSE

eligible applications and non-GSE eligible applications. The height of a bar corresponds to the

share of High DTI applications among all applications that belong to that category. For instance,

the “LTV” graph indicates that among GSE eligible applications with LT V>95 percent,

approximately 35 percent were High DTI. For LTV, FICO score and number of borrowers, the

“GSE eligible” bars are all higher than the “Non-GSE eligible,” reflecting the fact that there are

universally more High DTI applicants in the GSE segment. However, with respect to income, the

two segments differ substantially. Among GSE eligible applications, as the income reported on

the application (measured in thousands of dollars per month) rises, the share of applications

that are High DTI falls. In contrast, among non-GSE eligible applications, there is an inverse U-

shape relationship: applicants with the lowest incomes (less than $5,000 per month) and

highest incomes (more than $15,000 per month) are less likely to be High DTI than applicants

in the middle. This evidence is consistent with the notion that the share of High DTI

applications is reflective of lender’s underwriting approach. It must be noted, however, that this

observation belongs to 2013 data and does not suggest that lower income non-GSE eligible High

DTI applicants have been particularly affected by the Rule—this specific hypothesis is explored

later.

140 BUREAU OF CONSUMER FINANCIAL PROT ECTION

To conclude the description of the data, Figure 45 compares Application Data to two servicing

datasets along the key metric that can be computed in all three datasets: the share of loans with

DT I>4 3 percent among non-GSE eligible loans. It appears that the Application Data tracks to

the CoreLogic dataset quite closely, which is about twice as large by count of loans, and more

importantly includes many more lenders.

FIGURE 45: COMPA R I S O N OF A PPL ICATION DATA TO SERVICING DATA SETS (TRENDS IN THE SHA RE OF

HIGH DTI LOA NS FOR HOME PURCHA S E)

5.3.2 Estimation approach

The goal of this analysis is to isolate the effect of the ATR requirement on two key metrics of

interest: the share of High DTI applications and the approval rate of High DTI applications,

among non-GSE eligible applications. The estimation approach utilizes the DTI threshold

141 BUREAU OF CONSUMER FINANCIAL PROT ECTION

established by the General QM standard. This analysis assumes that non-GSE eligible

applications with DTI less or equal to 43 percent will comply with the ATR provision by

complying with the General QM requirements, whereas non-GSE eligible applications with DTI

greater than 43 percent (High DTI) will comply with the ATR requirement directly.

To estimate the effect of the Rule on the share of High DT I applications, it is necessary to control

for the influence of confounding factors that may affect the share of High DTI applications for

reasons not related to the Rule. For instance, the income of applicants or the amount of debt

they apply for are influenced by economic conditions and house price growth. The analysis

controls for application income directly, as this field is available in the data. Changes in house

prices are controlled for by comparing the share of High DTI borrowers in the non-GSE eligible

segment to the contemporaneous share of similar High DTI borrowers in the GSE eligible

segment, within the same lender. It was not possible to include house price indices directly into

the regression because geographic data is not available for all applications.

To estimate the effect of the Rule on the approval rate of High DTI applications, a triple-

differences estimation approach is adopted. Whereas the previous approach (for the share of

High DTI applications) performed two comparisons (before vs. after, and treatment vs control

group), the approach for approval rates performs a three-way comparison. This helps further

eliminate the potential effects of confounding factors on the outcome. First, the approval rate of

High DT I applications is compared to the approval rate of otherwise similar non-High DT I

applications in the GSE eligible segment; the same comparison is performed within the non-

GSE eligible segment as well. This helps isolate the influence of confounding factors that affect

the approval rate of all borrowers, regardless of DTI. Second, the approval rate of High DTI

applications in the non-GSE eligible segment is compared to the approval rate of High DTI

applications in the GSE segment, revealing the effect of differences in lenders’ underwriting

approaches to such applicants in these two segments. Third, the analysis examines how the

above mentioned differences have changed after the implementation of the Rule. See Section

5.3.8 for details of both specifications.

The plan for the rest of Section 5.3 is as follows. Sections 5.3.3 and 5.3.4 present estimation

results for home purchase loans and refinance loans, respectively. Sections 5.3.5 and 5.3.6

examine the impact of the Rule within specific groups of borrowers. Section 5.3.7 calculates the

combined effect of the Rule on the number of approved non-GSE eligible High DTI applications.

Finally, Section 5.3.8 is the technical appendix that contains the details of econometric

specifications and certain regression tables not included in the main text.

142 BUREAU OF CONSUMER FINANCIAL PROT ECTION

5.3.3 Results for home purchases

Table 18 provides information on aggregate changes in the outcomes of interest—the share of

High DTI applications and the approval rate of High DTI applications—over time, separately for

the GSE eligible and non-GSE eligible segments in the home purchase category. The share of

High DTI applications in the GSE eligible segment grew each year, whereas in the non-GSE

eligible segment this share declined substantially in 2014 and remained at approximately that

level afterwards. The approval rate of High DTI applications in the GSE eligible category

segment stayed relatively constant during 2013-2016, whereas in the non-GSE eligible segment

the approval rate declined substantially in 2014 and remained at approximately that level

afterwards.

TABL E 18 : SHA RE A ND A PPROV A L RA TE OF HIGH DTI A PPLICATIONS BY GSE ELIGIBILITY, HOME

PURCHA S E, 2013-2016

Year

Share of High DTI

GSE Eligible

applications

Approval rate of

High DTI GSE

Eligible

applications

Share of High DTI

Non-GSE eligible

applications

Approval rate of

High DTI Non-

GSE eligible

applications

2013

0.16

0.68

0.15

0.53

2014

0.17

0.68

0.09

0.33

2015

0.17

0.68

0.09

0.35

2016

0.19

0.69

0.09

0.39

The statistics in Table 18 are simple averages that do not control for any underlying changes in

loan and borrower characteristics. Figures 46 and 47 show model estimates of the dynamics of

these outcomes at a monthly level that would have been observed if loan and borrower

characteristics stayed constant. See Section 5.3.8 for the specification of the model that

produced these estimates.

143 BUREAU OF CONSUMER FINANCIAL PROT ECTION

FIGURE 46: ESTIMA TED EFFECT ON THE SHA RE OF HIGH DTI LOA NS A MONG PURCHA S E A PPLICA TIO NS,

2013-2016

144 BUREAU OF CONSUMER FINANCIAL PROT ECTION

FIGURE 47: ESTIMA TED EFFECT ON THE APPROVAL RATE ON HIGH DTI PURCA HS E A PPL ICA T IO N S, 2013-

2016

Consider first Figure 46, which plots the share of High DTI applications. The vertical dotted line

divides the pre-Rule and post-Rule periods (it is placed between the December 2013 and

January 2014 data points). The line labeled “Data Average” plots the difference in the share of

High DTI applications between non-GSE eligible and GSE eligible segments, for each month of

data. This difference is normalized to zero for January 2014. As an example, the Data Average in

January 2013 is graphed at 0.02. T his means that the difference in the share of High DTI

applications between non-GSE eligible and GSE eligible segments was approximately 2

percentage points higher than the same difference in January 2014. The negative values after

January 2014 indicate a negative impact of the Rule on the relative share of High DTI applicants

in the non-GSE eligible category. The line “Model Estimate” plots the predicted difference in the

shares of High DTI loans between the two segments that would have been observed if the mix of

applicants—on dimensions other than DTI—stayed constant throughout the entire period. Both

lines are fairly close to each other because non-DTI characteristics of the borrower have low

predictive power of the High DTI status. Finally, the shaded area around the Model Estimate

145 BUREAU OF CONSUMER FINANCIAL PROT ECTION

line represents the 95% confidence interval. Figure 47 is interpreted in a similar fashion. On that

figure, the “Data Average” and “Model Estimate” lines diverge, implying that changes in average

approval rates understate the actual impact of the Rule.

For the share of High DTI applications in the non-GSE eligible category, the regression model

estimates show an average decline of 10 percentage points after the implementation of the Rule

(see Section 5.3.8). In other words, in the absence of the Rule, there would have been 10

percentage points more High DTI applications in the non-GSE eligible segment. In 2013, the

share of High DTI applications in the non-GSE eligible segment was 14.7 percent. Therefore, in

relative terms the change represents a 68 percent reduction in the number of High DTI

applications over 2014 to 2016. In absolute terms, e.g. relative to the total number of

applications made by these lenders over 2014 to 2016, this change is small, only several

percentage points.

For the approval rate of High DTI applications in the non-GSE eligible category, the regression

model estimates show an average decline of 21 percentage points after the implementation of

the Rule (see Section 5.3.8). This decline in the approval rate is very large: it is larger than the

difference in approval rates between a borrower with a FICO score of 620 and a borrower with a

FICO score of 740 or above (see Table 28 for reference). Relative to the 2013 baseline, this

change represents a 40 percent decline.

Beyond the average effects, the dynamics of these outcomes over time are also important.

Figures 46 and 47 show that the introduction of the Rule was associated with a sharp drop in

both the share and approval rate of High DTI, non-GSE eligible applications, relative to High

DTI GSE eligible applications. After this initial decline, the outcomes gradually declined further.

While the average approval rate difference seems to have returned to the level of January 2014

by the end of 2016, the model estimates suggest that this reversal is due to changes in the mix of

High DTI applicants rather than due to changes in lenders’ credit policies. For both outcomes,

the model estimates suggest no convergence of outcomes to pre-Rule levels, implying that

lenders tightened underwriting approaches toward non-GSE High DTI purchase applicants at

the time the Rule became effective, and had not relaxed these approaches by 2016.

146 BUREAU OF CONSUMER FINANCIAL PROT ECTION

FIGURE 48: ESTIMA TED EFFECT OF THE GENERA L QM DTI PROV IS ION, BY LENDER. A PPLICA TIO N S FOR

HOME PURCHA S E, 2013-2016

Figure 48 plots the estimated effect of the General QM DTI provision separately for each lender.

The figure also includes 95% confidence intervals for each estimate, indicated as dashed lines.

There is substantial heterogeneity in the extent to which lenders have changed their

underwriting approaches toward non-GSE eligible High DTI applicants after the Rule.

For one lender in the upper right corner of the figure, the approach to High DTI applicants in

the non-GSE eligible segment has not substantially changed after the Rule. Another lender, in

the bottom middle part of the figure, has reduced its approval rate of High DTI applications by

close to 45 percentage points. The remaining six lenders are located between these two

extremes.

147 BUREAU OF CONSUMER FINANCIAL PROT ECTION

The fact that lenders have reacted so differently to the Rule is important for several reasons.

261

First, the heterogeneity of responses leaves open the possibility that lenders not included in the

sample may have reacted to the QM DTI requirement differently from those that are included.

For this reason, the average results presented here are valid only for this specific set of lenders.

The significance of the results presented in this section stems from the large combined size of

lenders included in the Application Data: based on HMDA 2016 data, these nine lenders

processed close to 20 percent of all applications for jumbo loans (a crude approximation of the

non-GSE eligible segment). Further, Figure 45 shows that according to a key metric—the share

of High DTI loans in the non-GSE eligible category—the Application Data is close to the

CoreLogic servicing dataset, which represents data from many more lenders.

Second, the differences in lenders’ reactions to the Rule, and the persistence of these differences

across time, suggests that lenders have not yet developed a common approach to measure and

model ATR risk in the same way as they approach other types of risk, such as the risk of

delinquency and default. For instance, cross-lender differences in both the level and the change

in approval rates of High DTI applications are much larger than, for example, differences in

approval rates by FICO category.

5.3.4 Results for refinances

This subsection briefly discusses results for High DTI refinance applications. Figures 49 and 50

present the dynamics of outcomes of applications for refinance. Similar to the patterns found

among applications for home purchase, there was a sharp drop in the relative share and the

relative approval rate of High DTI non-GSE eligible applications immediately after the

introduction of the Rule. However, in contrast to the home purchase category, the outcomes

show a trend toward convergence back to pre-Rule levels by the end of 2016, as seen from Table

19, and from Figures 49 and 50. This finding is consistent with a notion that lenders have

developed a common approach to ATR risk on refinance loans; for example, there may be a

consensus that on such loans the ability to repay has already been demonstrated.

261

Bey ond the issues discussed b elow, the differenc es in l ender’s a pproaches t o High DT I borrow ers em phasize t he

im portance of sh opping for a loan. Existing r esearch points to limited amount of shopping. See A lexei Alexandrov &

Ser gei Koulayev, No Shopping in the U.S. Mortgage Market: Direct and Strategic Effects of Prov iding In formation,

(Bu r eau Consumer Fin. Prot., O ffic e of Research, Working Paper No. 2 017–01, 2017).

148 BUREAU OF CONSUMER FINANCIAL PROT ECTION

TABL E 19 : SHA RE A ND A PPROV A L RA TE A ND SHA RE OF HIGH DTI A PPLICA TIO NS BY GSE ELIGIB IL ITY ,

REFINANCE, 2013-2016

Year

Share of High DTI

GSE Eligible

applications

Approval rate of

High DTI GSE

Eligible

applications

Share of High DTI

Non-GSE eligible

applications

Approval rate of

High DTI Non-

GSE eligible

applications

2013

0.27

0.67

0.19

0.46

2014

0.26

0.58

0.14

0.22

2015

0.23

0.57

0.13

0.28

2016

0.23

0.56

0.13

0.35

FIGURE 49: ESTIMA TED EFFECT ON THE SHA RE OF HIGH DTI LOA NS A MONG A PPLICA T I O N S FOR

REFINANCE, 2013-2016

149 BUREAU OF CONSUMER FINANCIAL PROT ECTION

FIGURE 50: ESTIMA TED EFFECT ON THE APPROVAL RATE ON HIGH DTI REFI NA N C E A PPLICA T IO NS ,

2013-2016

5.3.5 Effects on specific groups of borrowers: FICO, LTV,

Income

The previous results have shown that the Rule likely caused some non-GSE eligible High DT I

applications not to be submitted and if they were submitted then to be denied. This finding on

its own does not speak for or against the effectiveness of the DTI restriction in achieving the

purposes of the Rule or the Act. If the denied applicants in fact lacked the ability to repay, then

the reduction in approval rates is an intended consequence of the Rule. If the opposite were the

case, and the rejected applicants did have the ability to repay, then an unintended result is

observed.

150 BUREAU OF CONSUMER FINANCIAL PROT ECTION

Even though the ability to repay is not directly observed in the Application Data used in this

analysis, it is nevertheless informative to examine whether the Rule has affected home buy ers in

a way that is consistent with the expected loan performance.

262

Specifically, this analysis

differentiates borrowers by FICO score and application income, as these variables are traditional

predictors of delinquency and default. Figures 51 plots, separately for each FICO score bin

(FICO<680, 680-720, and FICO>720), the difference between the share and approval rate of

High DTI applicants in the GSE eligible and non-GSE eligible categories, normalized to zero in

January 2014; Figure 52 plots similar trends by income. Large decreases in the shares and

approval rates of High DTI, non-GSE eligible borrowers are observed for applicants with high

credit scores (>720) and high income, as well as for those with low credit score and low income.

Econometric analysis confirms that the Rule did not have differential impact on any specific

category of non-GSE eligible High DTI borrowers (see Section 5.3.8).

Table 20 illustrates how the pool of denied non-GSE eligible High DTI applicants has changed

between 2013 and 2014. After the introduction of the Rule, the pool of denied applicants

contains more of borrowers with higher income, higher FICO score and higher downpayment.

Together, these findings suggest that the observed decrease in access to credit in this segment

was likely driven by lenders’ desire to avoid the risk of litigation by consumers asserting a

violation of the ATR requirement or other risks associated with that requirement, rather than by

rejections of borrowers who were unlikely to repay the loan.

262

Sev eral c om menters noted the possibility that the Rule c ould have differential access to c redit effects on different

se g m en ts of th e b orr ow in g p opulation. For background, see Neil Bhutta and Glenn B. Canner, Mortgage Market

Co nditions and Borrow er Outcomes: Ev idence from the 2012 HMDA Data and Matched HMDA-Credit Record

Data, Fed. Res. Bull., Nov . 2013.

151 BUREAU OF CONSUMER FINANCIAL PROT ECTION

FIGURE 51: DIFFERENCE IN GSE EL IGI BL E A ND NON-GS E ELIGIBL E APPROVAL RATE AND SHA RE OF

HIGH DTI A PPLICA T IO NS FOR HOME PURCHA S E BY FICO BIN, 2013-2016

152 BUREAU OF CONSUMER FINANCIAL PROT ECTION

FIGURE 52: DIFFER EN C E IN GSE EL IGIBL E A ND NON-GS E ELIGIBL E APPROVAL RATE AND SHA RE OF

HIGH DTI A PPLICA TIO NS FOR HOME PURCHA S E BY A PPLICA T I O N INCOME, 2013-2016

153 BUREAU OF CONSUMER FINANCIAL PROT ECTION

TABL E 20 : AVERAGE CHARACTERISTICS OF DENIED NON-GSE ELIGIBL E HIGH DTI A PPLICA T I O NS FOR

HOME PURCHA S E, 2013-2016

Variable

Percent frequencies among

High DTI applications denied

in 2013

Denied in 2014

Application income, dollars per

month

< 2,501

46.86

38.45

2,501-5,000

10.39

11.93

5,001-7,500

6.91

7.34

7,501-10,000

7.29

7.67

10,001-12,500

6.36

6.88

12,501-15,000

5.42

7.01

> 15,000

16.77

20.73

FICO Score

<620

40.54

30.74

620-659

3.04

4.5

660-679

3.34

3.6

680-699

5.19

5.72

700-719

6.05

6.28

720-739

7.27

8.36

>=740

34.57

40.79

LTV

< 50%

5.44

5.36

51-80%

40.07

42.55

81-90%

5.53

8.63

91-95%

5.47

6.18

> 95%

43.49

37.29

5.3.6 Effects on self-employed borrowers

Another borrower characteristic of interest is self-employment.

263

The Bureau has examined

whether the Rule has had a disproportionate impact on self-employed borrowers who do not

263

Sev eral c om menters noted the im portance of self-em pl oyment i ncom e for m ortgage borrowers. For survey and

other ev idence on the prevalence of self-em ploym ent incom e, see Board of G ov ernor s of the Federal Reserv e Sy stem ,

Report on the Economic Well-Being o f U.S. Ho useholds in 2016, May 2017, available at

h t tps://w ww .federalreserve.gov/publications/files/2017-report-e conom i c-well-being-us-househol ds-2 018 05.pd f;

El k a T orpey and Andrew Hogan, Bureau of La bor Statistics, Career O utlook: “Working in a gig economy,” May

2 01 6, available at https://www.bls.gov /careeroutlook/2016/article/pdf/what-is-the-gig-e con om y .pd f .

154 BUREAU OF CONSUMER FINANCIAL PROT ECTION

qualify for a GSE loan. This impact may occur through two different channels: first, through the

ATR risk, where lenders may perceive self-employed borrowers as presenting higher ATR risk;

second, through compliance with Appendix Q requirements on documenting income and debt,

where self-employed borrowers may have a harder time presenting the required documentation,

or lenders may perceive a greater uncertainty of compliance among self-employed borrowers.

The Application Data contain a flag identifying self-employed applicants and therefore can be

used to examine the effect of the Rule on this group of borrowers. Table 21 reports the share of

home purchase applications submitted by self-employed borrowers between 2013 and 2016 for

the nine lenders in the Application Data. There is no discernible drop in the share of

applications from self-employed borrowers after the Rule was introduced; instances where the

share has declined in 2014 over 2013 are often reversed in later years.

TABL E 21 : PERCE N TA G E OF A PPL ICA TIONS FOR HOME PURCHA S E SUBMITTE D BY SELF-EMPLOY ED

BORROW ERS, 2013-2016.

Lender 2013 2014 2015 2016

26%

28%

25%

17%

15%

13%

17%

12%

17%

26%

22%

21%

22%

38%

37%

38%

13%

10%

11%

12%

13%

11%

14%

15%

13%

16%

18%

Total

17%

16%

17%

More definitive conclusions can be obtained using a regression analysis of the approval rates for

self-employed borrowers, similar to the one employed above for High DTI borrowers. See

Section 5.3.8 for specification and detailed results. The effect of the Rule on non-GSE eligible

self-employed borrowers is considered separately for High DT I and non-High DT I borrowers. In

the High DTI segment, there is no differential impact on self-employed borrowers (in other

words, their approval rate has declined in the same fashion as it did for all High DTI, non-GSE

eligible borrowers). Among non-High DT I, non-GSE eligible borrowers, the analysis finds that

approval rates are reduced by 2 percentage points for self-employed borrowers, compared to

similar non-self-employed borrowers. This effect is statistically significant at the 95 percent

confidence level, but is relatively small in magnitude compared to the overall approval rates.

This implies that the above described channels through which self-employed borrowers are

disproportionately affected, such as the Appendix Q requirements, are present but not

prohibitive.

155 BUREAU OF CONSUMER FINANCIAL PROT ECTION

The lender survey also sheds some light on the effect of Appendix Q, which, in all likelihood, is

most relevant for self-employed borrowers. T able 22 reports survey respondents’ responses to

how often they originate non-QM loans (therefore not eligible to be purchased, guaranteed or

insured by a GSE, FHA, VA, or USDA/RHS) where the borrower did not (could not) provide

documentation required by Appendix Q.

TABL E 22 : LENDER SURV EY: LA CK OF DOCUME NTA TION REQUIRED BY APPENDIX Q

Frequency of originating non-QM loans where borrower did not (could

not) provide documentation required by Appendix Q

Count Percent

Often

Sometimes

31%

Rarely or never

60%

Total responses

100%

Among 87 lenders who responded to this question (all of them originating non-QM loans), the

majority indicated “Rarely or never”. This suggests that in most cases, borrowers, including the

self-employed ones, are able to provide the documentation required by Appendix Q.

Nevertheless, a non-trivial portion of respondents indicated that such difficulties occur

“Sometimes” or “Often”, leaving open the possibility that Appendix Q requirements may have

had an impact on access to credit.

5.3.7 Combined effect of the Rule on High DTI borrowers

The two previously estimated effects on application counts and approval rates can be combined

into an overall estimated effect of the Rule on the number of approved High DTI non-GSE

eligible applications. Importantly, this analysis does not speak to the alternatives that were

chosen by borrowers who did not appear among non-GSE eligible High DTI applications after

the Rule. The possibilities include, but are not limited to: a) documenting more income or

paying off other debt, in order to fit under 43 percent DTI; b) choosing a lower loan amount

either by increasing the downpayment or purchasing a home at a lower purchase price

264

; c)

postponing home purchase; or (d) obtaining a non-GSE eligible High DTI loan from a lender

outside of the sample.

264

Lowering the loan amount may result in a conforming size loan, which may be GSE eligible. The Bureau has also

examined submissions to GSE platforms that received approve/eligible status, and did not find any sharp increase

in the number of High DTI submissions that would be consistent with substitution away from the non-QM segment

(see Chapter 6 below for more details).

156 BUREAU OF CONSUMER FINANCIAL PROT ECTION

Some of these alternatives, if chosen, may manifest themselves in an increased number of

applications just under the regulatory threshold of 43 percent (the so-called “bunching”

behavior). Evidence for such bunching has been presented in Chapter 4. It is possible to account

for “bunching” by considering changes in the share of applications with DTI above 40 percent

rather than 43 percent; in this way, the redistribution of applications from above 43 percent to

just below 43 percent will not affect the estimated change in the total number of High DT I

applications. Table 23 and Table 24 present lower bound and upper bound estimates by utilizing

DT I thresholds of 40 percent and 43 percent, respectively. It is estimated that the General QM

DTI provision has eliminated between 63 and 70 percent of approved non-GSE High DTI

applications for home purchase among the nine lenders that contributed the data, over the

period of 2014 – 2016; this change translates into a reduction of between 1.5 and 2 percent of all

loans for home purchase made by these lenders during this period.

265

Taking the average profit

margin of 48 basis points for 2014 through 2016 from the Annual Mortgage Bankers

Performance Report and considering the average non-GSE eligible loan size, the impact of the

Rule amounts to a cost of between $20 million and $26 million per year in lost profits.

265

Recent research by the Federal Reserve Board, using Optimal Blue data, fi nds an ap proximate ly 5 0 percent

r eduction in originations of High DTI, non-GSE eligible loans. Available a t

h t tps://w ww .federalreserve.gov/econres/notes/feds-notes/effects-of-the-ability-to-repay-qual ifi ed-m or tg a ge -rule-

on -m ortgage-lending-20181116.htm.

157 BUREAU OF CONSUMER FINANCIAL PROT ECTION

TABL E 23 : LOWER BOUND ESTIMA T E OF THE COMBINED EFFECT OF THE RULE (A CCOUNTI NG FOR

BUNCHING UNDER 43 PERC E NT), OVER 2014 – 2016, HOME PURCHA S E.

Step Re sul t

Pre-Rule share of DTI>40% among Non-GSE eligible .408

Pre-Rule share of DTI>40% among GSE eligible .435

Post-Rule share of DTI>40% among Non-GSE eligible .375

Post-Rule share of DTI>40% among GSE eligible .456

Post-Rule change in relative share of DTI>40%

(.375-.456) - (.408-.435)

= -.053

Counterfactual post-Rule share of DTI>40% among Non-GSE eligible .375 - (-.053) = .429

Post-Rule count of NOT DTI>40% among Non-GSE eligible 70112

Counterfactual post-Rule count of DTI>40% among Non-GSE eligible

.429 = X/(70112+X)

=> X = 52744

Actual post-Rule count of DTI>40% among Non-GSE eligible 42215

Lost post-Rule count of High DTI among Non-GSE eligible 52744 - 42215 = 10529

Actual post-Rule count of High DTI applications among Non-GSE eligible 15135

Counterfactual post-Rule count of High DTI applications 15135 + 10529 = 25664

Post-Rule approval rate of High DTI applications among Non-GSE eligible .359

Causal estimate of the impact on approval rate -.212

Counterfactual post-Rule approval rate of High DTI .359 - -.212 = .571

Actual count of post-Rule approved High DTI applications 5439

Counterfactual count of approved High DTI applications post Rule .571*25664 = 14672

Estimate of the count of approved High DTI applications lost due to Rule 14672-5439 = 9233

Estimate of the share of approved High DTI applications lost due to Rule (14672-5439)/14672 = .629

Estimate of lost profits assuming 48 bps profit rate, millions of dollars 19.281

158 BUREAU OF CONSUMER FINANCIAL PROT ECTION

TABL E 24 : UPPER BOUND ESTIMA T E OF THE COMBINE D EFFECT OF THE RULE (ACCOUNT ING FOR

BUNCHING UNDER 43 PERC E NT), OVER 2014 – 2016, HOME PURCHA S E.

Step Re sul t

Pre-Rule share of DTI>43% among Non-GSE eligible .232

Pre-Rule share of DTI>43% among GSE eligible .260

Post-Rule share of DTI>43% among Non-GSE eligible .134

Post-Rule share of DTI>43% among GSE eligible .271

Post-Rule change in relative share of DTI>43% (.134-.271) - (.232-.260) = -.109

Counterfactual post-Rule share of DTI>43% among Non-GSE eligible .134 - (-.109) = .244

Post-Rule count of NOT DTI>43% among Non-GSE eligible 97192

Counterfactual post-Rule count of DTI>43% among Non-GSE eligible

.244 = X/(97192+X) => X =

31392

Actual post-Rule count of DTI>43% among Non-GSE eligible 15135

Lost post-Rule count of High DTI among Non-GSE eligible 31392 - 15135 = 16257

Actual post-Rule count of High DTI applications among Non-GSE

eligible

15135

Counterfactual post-Rule count of High DTI applications 15135 + 16257 = 31392

Post-Rule approval rate of High DTI applications among Non-GSE

eligible

.359

Causal estimate of the impact on approval rate -.212

Counterfactual post-Rule approval rate of High DTI .359 - -.212 = .571

Actual count of post-Rule approved High DTI applications 5439

Counterfactual count of approved High DTI applications post Rule .571*31392 = 17946

Estimate of the count of approved High DTI applications lost due to

Rule

17946-5439 = 12507

Estimate of the share of approved High DTI applications lost due to

Rule

(17946-5439)/17946 = .696

Estimate of lost profits assuming 48 bps profit rate, millions of dollars 26.118

159 BUREAU OF CONSUMER FINANCIAL PROT ECTION

5.3.8 Technical Appendix

To analyze changes in the share of High DTI applicants, the following linear probability model is

estimated:

ℎ



= 

+ 





+ 

2013



+ 





× 2013



+ 



′

 + 



Here ℎ



is a variable that takes on the value of 1 if application i has DTI>43 percent and 0

otherwise (such variables are known as indicators); 



is an indicator for application i being

non-GSE eligible; 



is a set of control variables including a full set of dummies for calendar

month (January – December), dummies for each bin of loan amount, FICO, LTV, CLTV, income,

number of borrowers, self-employment status, number of units, payment type (fixed or

adjustable rate mortgage), past foreclosure or current delinquency. See Appendix C for data

dictionary. The parameter of interest in this model is 

which applies to the interaction term





× 2013



. This parameter measures the change in the share of High-DTI applications

in the non-GSE eligible segment relative to the GSE-eligible control group that occurred after

2013. Results are found in the first and the second column of Table 25 (the second column

includes borrower controls 



, while the first does not).

To analyze changes in the approval rate of High-DTI applications, the following linear

probability model is estimated:





= 

+ 





+ 

2013



+ 

ℎ



+ 





× 2013



+ 





ℎ



+ 

2013



× ℎ



+ 





× 2013



× ℎ



+ 



′

 + 



The set of borrower controls 



is the same as in the regression for the High DT I status above.

Given that the outcome is a binary indicator of approval, all coefficient estimates are interpreted

as changes in the approval probability, expressed in percentage points. The parameter of

interest in this model is 

, which applies to the triple interaction term 



× 2013



ℎ



. This parameter measure the change in the approval rate of High-DT I, non-GSE

eligible applications relative to the High-DT I, GSE-eligible control group in each year after 2013.

Results are found in the third and the fourth column of Table 25 (the fourth column includes

borrower controls 



, while the third does not).

Table 25 shows the average impact across lenders for the entire post-Rule period, for

applications for home purchase. Table 26 shows results for refinances. Among applications for

160 BUREAU OF CONSUMER FINANCIAL PROT ECTION

home purchase (Table 25), the likelihood that a given application is High DTI declined by 10

percentage points, and the likelihood that a given High DTI application is approved declined by

more than 21 percentage points

266

. This decline in the approval rate is very large: it is larger than

the difference in approval rates between a borrower with a FICO score of 620 and a borrower

with a FICO score of 740 or above (see Table 28 for reference).

TABL E 25 : ESTIMA TED EFFECT OF THE RULE ON HIGH DTI A PPLICA TIO NS FOR HOME PURCHA S E, 2013-

2016

Share High DTI

Share High

DTI

Approval

rate

Approval rate

NGSE -0.028*** -0.02*** -0.086*** 0.038***

Post2013

0.011***

0.014***

0.011***

-0.005***

Post2013 X NGSE

-0.109***

-0.102***

0.062***

0.012***

High DTI

-0.098***

-0.106***

NGSE X High DTI

-0.067***

-0.031***

Post2013 X High DTI

-0.005

0.007**

Post2013 X NGSE X High DTI

0.234***

0.212***

Borrower controls

Yes

Observations

686,334

686,304

686,334

686,304

0.013

0.053

0.03

0.184

266

Lev el of significance is reported in the tables as * p<0.10, ** p<0.05, and *** p<0.01.

161 BUREAU OF CONSUMER FINANCIAL PROT ECTION

TABL E 26 : ESTIMA TED EFFECT OF THE RULE ON HIGH DTI A PPLICA TIO NS FOR REFINA NCE, 2013-2016

Share High DTI

Approval rate

NGSE

-0.114***

-0.048***

-0.048

-0.041***

Post2013

0.082***

0.045***

-0.007***

-0.017***

Post2013 X NGSE

-0.032***

-0.046***

0.067***

0.056***

High DTI

-0.050***

-0.003**

NGSE X High DTI

-0.158***

-0.192***

Post2013 X High DTI

-0.086***

-0.094***

Post2013 X NGSE X

High DTI

-0.139*** -0.106***

Borrower controls

Yes

Observations

1,876,221

1,876,165

1,876,221

1,876,165

0.012

0.118

0.023

0.068

TABL E 27 : ESTIMA TED EFFECT OF THE QM DTI PROV IS ION WITHIN FICO A ND INCOME CATEGORIES:

HIGH DTI A PPLICA T IO NS FOR HOME PURCHA S E, 2013-2016

Share of High DTI Approval rate among High DTI

Post

-0.00211

-0.00236

FICO 1-699

-0.0524***

-0.193***

P o s t X FICO 1 -699

0.0583***

0.0454***

Income < 5K

0.170***

-0.311***

Income 5-10K

0.0872***

-0.0423***

Post X Income < 5K

0.0544***

0.0342**

Post X Income 5-10K

0.0123***

0.00453

SelfEmp

0.127***

-0.0803***

Post X SelfEmp

-0.0028

0.00159

P ri vat e

-0.0501***

0.00904

Private X FICO 1-699

0.0978***

-0.125***

Private X Income < 5K

-0.0390***

-0.0635***

Private X Income 5-10K

0.0700***

-0.0437**

Private X SelfEmp

-0.0284***

-0.0131

Private X Post

-0.0970***

-0.251***

Private X Post X FICO 1-699

-0.0416***

0.117***

Private X Post X Income < 5K

-0.0158

0.0167

Private X Post X Income 5-10K

-0.00996

0.0507**

Private X Post X SelfEmp

-0.00433

-0.00301

Constant

0.281***

0.997***

Observations

357,638

79,625

0.061

0.175

162 BUREAU OF CONSUMER FINANCIAL PROT ECTION

TABL E 28 : ESTIMA TED EFFECT OF THE RULE ON A PPROV A L RA TES FOR SELF-EMPLOY ED

BORROW ERS FOR HOME PURCHASE, 2014 – 2016

High DTI

Not High DTI

NGSE

-0.001

0.037***

SelfEmp

-0.090***

-0.033***

NGSE & SelfEmp

-0.004

-0.010*

NGSE & Post2013

-0.185***

0.015***

SelfEmp & Post2013

0.002

0.013***

NGSE & SelEmp & Post2013

-0.040**

-0.021***

Post2013

-0.006

-0.008***

FICO < 620

omitted

FICO 620-659

0.302***

0.556***

FICO 660-679

0.354***

0.591***

FICO 680-699

0.383***

0.621***

FICO 700-719

0.413***

0.632***

FICO 720-739

0.435***

0.659***

FICO >= 740

0.472***

0.680***

LTV controls

YES

Loan size controls

YES

More than one borrower

0.017***

0.028***

Fixed rate loan

0.029***

0.017***

CLTV>LTV

-0.070***

-0.013***

One unit

0.064***

0.088***

Quarter of application controls

YES

Constant

-0.025*

-0.097***

Observations

168496

887670

R-squared

0.143

0.188

163 BUREAU OF CONSUMER FINANCIAL PROT ECTION

5.4 Effects of the points and fees

requirement on the availability of small

dollar loans and cost of credit

As amended by the Dodd-Frank Act, TILA requires that a “qualified mortgage” at or above

$100,000 have total “points and fees” that do not exceed three percent of the total loan amount,

except for “smaller loans” for which Congress directed the Bureau to adopt points and fee

caps.

267

Points and fees are charges paid for the loan to the creditor, loan originator, or an

affiliate. In addition to the general three percent cap, the ATR/QM Rule provides for

proportionally higher points and fees limits for smaller loan amounts: eight percent of the loan

amount for loans less than $12,500; $1,000 for loans that are at least $12,500 but less than

$20,000; five percent of the loan amount for loans that are at least $20,000 but less than

$60,000; $3,000 dollars for loans between $60,000 and $100,000, which are all indexed for

inflation, and a 3 percent cap for loans above $100,000.

268

From conversations with lenders, the Bureau has learned that borrowing scenarios where the

QM points and fees threshold (further, “QM PF threshold” or simply “PF threshold”) is exceeded

are relatively rare and are typically dealt with through an “exception process”. Loan origination

software calculates the PF status on each application and produces a message if the status is

negative (i.e., PF threshold is exceeded); on some systems, such an alert puts a stop on the

application process until a loan officer with sufficient authority creates an exception. If the

lender wishes to avoid originating a loan that exceeds the QM PF threshold, the options include:

a) reduce fees to adhere to the threshold, or b) deny the application.

There are several major difficulties with quantifying the direct impact of the QM PF threshold

on access to credit and the cost of credit. First, before the Rule lenders did not perform the

points and fees calculation in a manner prescribed by the Rule, and therefore such data does not

exist. Second, the Bureau does not have data on individual loan charges with sufficient detail to

reconstruct the results of the points and fees calculation. As a result, there does not exist a

longitudinal dataset of loan originations that would allow a comprehensive examination of the

impact of the QM PF provision, of the kind that was utilized to study the impact of the General

QM DTI provision. Third, even if such data did exist, it would reflect the results of PF calculation

267

Dodd-Fr ank Act section 1 412; TILA section 1 29C (b)(2 )(A)(vii). The lim its on points and fees for qualified

m ortgages are im plemented in 12 C.F.R. § 1 026.43(e)(3).

268

1 2 C.F.R. § 1 026.43(e)(3).

164 BUREAU OF CONSUMER FINANCIAL PROT ECTION

on the originated loans would reflect the final status of the application, e.g., after lenders have

made necessary fee adjustments to stay under the PF threshold. Such data would say little about

the frequency at which PF violations occur at the initial status of the application, or about the

magnitude of fee adjustments. For instance, if broker compensation was adjusted (lowered) to

fit within the PF threshold, the existing data on individual loan charges is not informative on

what the broker compensation would have been absent the Rule.

Although it is not possible to produce a causal estimate of the impact of the QM PF provision on

access to credit and cost of credit, the available data (some collected specifically for the purposes

of this assessment) allows to answer partial questions that are indicative of that impact.

Specifically, the questions explored this section are the following: 1) How often is the QM PF

threshold initially exceeded on an application (e.g., before adjustments are made)?; 2) Which

borrowing scenarios are most affected?; and 3) What are the lender policies in the situation

where an application indicates that the PF threshold would be exceeded?

5.4.1 Summary of the points and fees requirement

TILA defines points and fees to include: 1) all items included in the finance charge

269

except

interest or the time price differential; 2) all compensation paid directly by either a consumer or a

creditor to a mortgage originator from any source; 3) certain real estate related settlement

charges that are generally excluded from the finance charge (such as title insurance,

27 0

document preparation, and appraisal fees) unless the charge is paid to an unaffiliated third

party and meets other conditions; 4) certain charges such as credit insurance and prepayment

penalties; and 5) such other charges as the Bureau determines to be appropriate.

27 1

The Bureau

did not determine it appropriate to add any charges beyond those listed in the statutory

269

1 2 C.F.R. § 1 026.4.

27 0

For r esearch relevant t o the title insurance m arket, the r ole of a ffiliated service providers, and m ore g eneral

r esearch on vertical i ntegration, see Lawrence J. White, The Title Ins urance Industry, Reverse Competition, and

Co ntrolled Bus iness - A Diffe rent View , The Journal of Risk and In surance, V ol. 51, No. 2 (1984); Analysis Group,

In c., Co mpetition and Title Ins urance Rate s in California, January 2006, available at

h ttps://www.analysisgroup.com/link/4b5d321ac3c1459cb5d09ca007e4dbca.aspx; Birny Birnbaum, Report to the

Califo rnia Ins urance Co mmissio ner, “An Analysis of Competition in t he California Title Insurance and Escrow

In d u stry,” De c em ber 2005, available at http://www.insurance.ca.gov /0400-news/0200-studies-

reports/upload/CATitleCompetitionReport0512Public.pdf; Harris/Nielsen, O ne -Stop Shopping Consumer

Preferences, October 2015, available at h ttp://narfocus.com/billdatabase/clientfiles/1 72/25/2950.pdf; Mi chael H.

Rioda n, Competitive Effects of Vertical Integration, Columbia University Department of Econom ics Discussion

Paper Series, 2005; Timothy Bresnahan and Jonathan Levin, Vertical Integration and Market Structure, S tanford

Institute for Econom ic Policy Research March 2012.

27 1

TILA section 1 03(bb)(4).

165 BUREAU OF CONSUMER FINANCIAL PROT ECTION

definition of points and fees when it implemented this list in the ATR/QM Rule.

27 2

In a loan

originated directly by a creditor, points and fees will generally be limited to direct charges by the

creditor and charges by any affiliates it chooses to use for settlement services.

27 3

As amended by the Dodd-Frank Act, TILA provides certain exclusions from the points and fees

definition.

27 4

TILA excludes any mortgage insurance premium charged by a government agency,

such as FHA.

27 5

TILA may also exclude up-front premiums for private mortgage insurance, but

certain conditions must be met. Typically, private mortgage insurance is paid monthly after

settlement, usually as part of the loan’s escrow payment. TILA excludes from points and fees any

mortgage insurance premium paid after closing.

27 6

However, there can also be a sizable up-front

premium for private mortgage insurance that is paid at settlement. TILA excludes any up-front

private mortgage insurance charge that is not in excess of the typical up-front amount charged

by the FHA, as long as the excluded amount is automatically refundable pro rata when the loan

is paid off.

27 7

In order to avoid the points and fees cap interfering with creditors offering discount points to

consumers, TILA excludes from points and fees either one or two “bona fide” discount points,

depending on the difference between the interest rate without any discount being purchased and

the average prime offer rate (APOR) for the transaction.

27 8

For example, if a creditor originates a

loan that would have an interest rate of 5 percent with no discount purchased, and the APOR for

the transaction is 4 or higher (1 percent difference or less), the creditor may exclude up to two

bona fide discount points from the points and fees for the transaction. If a creditor originates a

loan that would have an interest rate of 6 percent with no discount, and the APOR is 4 percent

or between 4 and 5 percent (2 percent difference or less, but more than 1 percent), the creditor

27 2

1 2 C.F.R. § 1 026.32(b)(1).

27 3

1 2 C.F.R. § 1 026.32(b)(1).

27 4

The exclusions described here do not include exclusions provided in the list of points and fees items above from

T ILA section 103(bb)(4), such as reasonable unaffiliated third-party real estate charges from w hich t he creditor

receives no compensation and credit insurance calculated and paid monthly.

27 5

TILA section 1 03(bb)(1)(C)(i).

27 6

TILA section 1 03(bb)(1)(C)(ii), (iii).

27 7

These provisions are implemented in the ATR/QM Rule at 1 2 C.F.R. § 1 026.32(b)(1)(i)(B)–(C).

27 8

T ILA section 1 03(dd). T his provision is im plemented in the A TR/Q M Ru le at § 1 026.32(b)(1)(i)(E)–(F) and

§ 1 026.32(b)(3)(i). The provision has a separate test for exclusion of discount points on loans for n on-real estate

m anufactured housing. The average prime offer rate (APOR) is an annual percentage rate that is derived from

a v erage interest rates, points, and other loan pricing t erms c urrently offered to c onsumers by a representative

sample of creditors for mortgage transactions that have low-risk pricing characteristics. 12 C.F.R. § 1026.35(a)(2).

166 BUREAU OF CONSUMER FINANCIAL PROT ECTION

may only exclude one bona fide discount point. TILA requires that bona fide discount points

excluded from points and fees result in a real discount and that the amount of the rate reduction

purchased be reasonably consistent with established industry norms and practices.

27 9

5.4.2 Evidence from the lender survey

In the lender survey, the Bureau asked: “How often does a loan application initially exceed the

QM cap for points and fees? Please only consider applications for loans of less than $100,000.”

This question focuses on the initial status of the application, as opposed to the final status which

reflects fee adjustments that may have been made in order to accommodate the QM cap. T he

question also focuses on applications for smaller loan amounts as those generally have higher

PF ratios. A total of 159 responses were received on this question.

TABL E 29 : HOW OFTEN DOES A LOAN A PPLICA T IO N INITIA LLY EXCEED THE QM CA P FOR POINTS A ND

FEES? PLEA SE ONLY CONSI D E R A PPLICA TIO NS FOR LOA NS OF LESS THA N $100,000.

Percentage of applications Count of respondents Percent of responses

<1% of applications 74 47%

1 - 3% of applications

14%

>3% of applications 24 15%

Do not know 39 25%

Total responses 159 100%

Table 29 provides the breakdown of responses to the question stated above. Notably, a

substantial portion of respondents, 39 out of 159, indicated that they “Did not know” the

incidence in which the points and fees on an application initially exceed QM PF threshold. As

some of them indicated in the write-in response, this is because the final status of the

application as it is recorded in the loan originations software already reflects fee adjustments

and thus is uninformative regarding the initial PF violation. In the majority of cases, 96 out of

159, the incidence of PF violations is infrequent, less than 3 percent of applications.

The write-in responses provide additional detail regarding the situations where the PF cap

might be initially exceeded. For example, one bank indicated that over 15 percent of their

27 9

TILA section 1 03(dd).

167 BUREAU OF CONSUMER FINANCIAL PROT ECTION

wholesale portfolio applications had points and fees that initially exceeded the cap, due to the

inclusion of broker compensation in the calculation of PF. Other respondents mentioned

circumstances leading to QM PF violations including: lower loan amounts, particularly in rural

areas; second homes; private mortgage insurance (PMI); discount points paid to buy down the

rate; fees on special loan programs (HFA, FHA); and fees due to an affiliated appraisal or title

company. Respondents also mentioned low credit scores, high LTVs, and manufactured homes

as additional circumstances although it is less clear in these cases what the underlying causes

may be. In addition to the lender survey responses, seven out of 31 lenders that provided

comments to the initial Federal Register notice on this assessment also provided detail on the

circumstances leading to QM PF violations. These comments pointed to: smaller loans in rural

areas; generally loans below $50,000 and; broker compensation as contributing causes.

TABL E 30 : DO Y OU TA KE THE FOLLOWING A CTIONS WHEN A LOA N A PPLICA TIO N IS BEING PROC ES S ED

A ND THERE IS A N INDICA TIO N THA T THE QM CA P FOR POINTS A ND FEES COULD BE

EXCEEDED?

Policy for PF violation

Number of

respondents who

mentioned this

policy

Percent of

respondents who

mentioned this

policy

Waive certain fees to keep points and fees ratio under

the limit, and increase interest rate

51 39%

Waive certain fees to keep points and fees ratio under

the limit, without increasing interest rate

112 76%

Deny the loan application 29 22%

Proceed without making any changes 22 17%

Total responses 137

The next question on the survey asks: “Do you take the following actions when a loan

application is being processed and there is an indication that the QM cap for points and fees

could be exceeded?” T able 30 shows the breakdown of responses to this question. A total of 137

lenders have responded to this question. Because some respondents chose multiple options

(e.g., more than one policy) the sum of values in the first column exceeds 137. By far the most

168 BUREAU OF CONSUMER FINANCIAL PROT ECTION

popular option was “Waive certain fees … without increasing the interest rate,” followed by

“Waive certain fees … and increase the interest rate”.

280

TABL E 31 : PF V IOLA TION POLICES BY INSTIT UT IO N TY PE

Institution type

Waive certain

fees to keep

points and

fees ratio

under the

limit, and

increase

interest rate

Waive certain

fees to keep

points and

fees ratio

under the

limit, without

increasing

interest rate

Deny the loan

application

Proceed

without

making any

changes

Bank with <$2 billion in total assets

(45)

17% 68% 17% 27%

Bank with $2-10 billion in total

assets (16)

63% 73% 22% 14%

Bank with >$10 billion in total

assets (23)

37% 70% 11% 16%

Credit Union (23) 29% 62% 17% 18%

Non-DI (70) 51% 90% 30% 11%

All (177) 39% 77% 22% 17%

Table 31 presents the breakdown of policy options by institution type (the number in brackets

indicates the number of respondents among each type). The percentage values show the

percentage of respondents of a given type (indicated by the row) mentioning a given policy

(indicated by the column). Notably, 70 percent of banks with >$10 billion in total assets

mentioned “Waive certain fees to keep points and fees ratio under the limit, without increasing

interest rate” as applicable policy. Thus, this policy option is not only the most popular among

lenders, but is also most likely to be applied across borrowers as it is often chosen by larger

institutions.

280

This finding is consistent with other surveys on this topic. See National Association of Realtor’s Survey of

Mortgage Providers, April 2014, available at https://www.nar.realtor/research-and-statistics/research-

r eports/m ortgage-originators-survey/june-2014-m ortg a ge -originators-su rvey .

169 BUREAU OF CONSUMER FINANCIAL PROT ECTION

5.4.3 Evidence from the Application Data

As part of the data collection that resulted in the Application Data from nine lenders, the Bureau

requested, that for each application, a data field indicating whether the application has passed

the QM PF test. Seven out of nine lenders were able to provide these data (the other two do not

routinely collect this information). Because the PF calculation is different for FHA applications,

the foregoing analysis below only focuses on applications for conventional loans. The analysis

further focuses on applications for loan purchase, as these are more likely to exceed the QM PF

threshold due to extra origination-related charges that apply to purchase transactions.

TABL E 32 : PERCE N T OF PF V IOLA TIONS A MONG CONV ENT IO NA L A PPLICA TIO NS FOR HOME

PURCHA S E IN THE RETA IL CHA NNEL

Lender

Approved

in 2013

Approved

in 2014

Approved

in 2015

Approved

in 2016

Denied

in 2013

Denied

in 2014

Denied

in 2015

Denied in

2016

0.0%

1.5%

2.0%

2.5%

(0.0%)

. 0.0% 0.0% 0.0% . . . 0.0%

(100%)

(14.6%)

(15%)

(18%)

(100%)

0.0%

2.9%

(100%)

(1.4%)

(0.0%)

(100%)

(1.8%)

(0.0%)

32.5%

1.5%

0.0%

40.0%

5.0%

0.0%

1.1%

(36.3%)

(0.1%)

(0.0%)

(75.2%)

(0.6%)

(0.0%)

0.0%

1.1%

(100%)

(1.7%)

(0.0%)

(100%)

(1.9%)

(0.0%)

0.0%

0.2%

0.0%

0.5%

0.1%

(0.0%)

44.7%

1.6%

0.9%

53.4%

8.1%

4.2%

5.5%

(97%)

(4.5%)

(1.1%)

(1.2%)

(98.0%)

(6.1%)

(2.1%)

(2.0%)

All

9.7%

0.5%

0.2%

0.1%

2.4%

2.8%

1.9%

(64%)

(4.8%)

(4%)

(4.3%)

(79.5%)

(29.9%)

(26.7%)

(21.0%)

Table 32 present the percentage of applications where the QM PF threshold was exceeded, for

the retail channel (results for the corresponding channel are almost identical). The broker

channel is examined below in Section 5.4.6. For each lender–year combination, the first row

indicates the incidence of PF violations; the values in brackets indicate the percentage of records

where information on PF status is missing. Unfortunately, for the pre-Rule period (2013

applications), the information is missing in majority of cases. For this reason, the 2013

percentages are considered to be unreliable.

170 BUREAU OF CONSUMER FINANCIAL PROT ECTION

In the post-Rule period, 2014 to 2016, almost all approved applications indicate passing the QM

PF test. Importantly, the Application Data indicates the final status of the application, i.e., after

fee adjustments were made. For this reason, it is not possible to examine the impact of the initial

PF violation on the eventual approval rate of an application.

Presumably, denied applications did not go through the same fee adjustment process and thus

may provide some indication, albeit imprecise, of how often the PF threshold is initially

exceeded. Because denied applications are systematically different from approved applications

(lower FICO score, higher LTV, etc.), for this analysis only denied applications with at least a 50

percent probability of approval were selected. The probability of approval on denied

applications was calculated using estimates from the approval regression model (see Section

5.3.8 for detail). Among this selected group of denied applications, between 2 and 3 percent of

applications are found to exceed the QM PF threshold. This finding indicates that PF violations

do occur, albeit infrequently.

TABL E 33 : PERCE N T OF PF V IOLA TIONS AMONG APPLICA TIONS FOR LOA N A MOUNTS BEL OW $100,000

(RETA IL A ND CORRES P O NDENT APPLICA TIONS FOR HOME PURCHA S E)

Lender

Approved

in 2014

Approved

in 2015

Approved

in 2016

Denied in

2014

Denied in

2015

Denied in

2016

0.1%

0.0%

3.2%

4.4%

3.8%

0.0%

5.2%

2.6%

3.3%

4.2%

0.0%

0.1%

10.1%

0.0%

4.4%

0.0%

2.6%

2.7%

2.2%

0.0%

0.1%

0.0%

0.9%

0.8%

1.4%

0.6%

6.8%

2.6%

5.0%

Total

0.7%

0.2%

3.9%

2.8%

2.9%

Table 33 examines the incidence of PF violations among applications for loan amounts less than

$100,000. Among approved applications, there is almost no difference from applications for

larger loan amounts (e.g., almost all applications pass the QM PF test). However, among denied

applications, an application for a loan under $100,000 is 1-2 percentage points more likely to

fail the PF test than an application for a larger loan. Nevertheless, overall incidence is low, under

3 percent in most cases. This finding corroborates the results of the lender survey where most

respondents indicated that fewer than 3 percent of applications initially exceeded the PF

threshold.

171 BUREAU OF CONSUMER FINANCIAL PROT ECTION

5.4.4 Evidence from the fair lending data

From the above analyses, it appears that lenders waive fees and sometimes increase interest

rates in response to an initial PF violation. However, the ultimate impact of these adjustments

on the cost of credit remains unclear. To study this issue, the Bureau has utilized data it received

from lenders for the purposes of certain fair lending exams. Data from seven exams contained a

sufficient number of observations for the pre-Rule and post-Rule period.

Unfortunately, only two exams included loans that were closed before 2014. For the remaining

five exams, the pre-Rule data includes applications that were submitted before 2014, but closed

in 2014. This may introduce trends in the data due to potential selection of applications that

may be correlated with points and fees.

172 BUREAU OF CONSUMER FINANCIAL PROT ECTION

FIGURE 53: QUANTIL ES OF NET TOTA L POINT S A ND FEES FOR LOA NS BELOW $150,000, BY LENDER

2013-2015

Figure 53 presents the main result of this analysis. Separately for each lender, it plots the first

and fifth quantiles of the distribution of “Net Total Points and Fees” paid on that lender’s loans,

along with the median. The calculation of this cost metric is different from the one involved in

the QM PF definition, so the data should be interpreted as only a proxy. And, the Bureau has

been unable to compare this proxy to points and fees calculated according to the QM PF

definition, because the fair lending data does not have sufficient detail to perform this

calculation. Nevertheless, changes in this measure of cost of credit should be informative of the

impact of the Rule.

173 BUREAU OF CONSUMER FINANCIAL PROT ECTION

For some lenders, the net total points and fees at the first, fifth and fiftieth (median) percentile

declined in 2014 compared to 2013. For other lenders, this cost metric has stayed generally

constant; and for one lender substantial increases are observed. This is somewhat consistent

with the evidence from the prior evidence that lenders may be, on the margin, taking steps to

ensure that loans are under the threshold. However, the fact that declines in mortgage costs are

observed at all quantiles, while the impact of the PF provision is expected to be limited to

borrowers with the highest values of points and fees, limits any conclusion that the dynamics

here are related to the Rule.

5.4.5 Approval rates and originations of small balance

loans: evidence from HMDA

Several trade groups and individual commenters on the RFI stated that the current points and

fees tiers make smaller size mortgages less attractive to lenders given the relatively high cost of

originating such loans. As detailed in the next section, qualitative responses to the lender survey

indicate that smaller size loans may be more likely to exceed the points and fees threshold than

larger size loans, rendering them ineligible for QM status. The survey also indicates that such

situations may be uncommon.

This section summarizes analysis studying how approval rates for small loans changed in

response to the implementation of the points and fees cap. The analysis utilizes HMDA data and

focuses on conventional loans under $170,000 between 2012 and 2016, covering two years

before and three years after the Rule’s effective date.

281

The HMDA data do not allow for a direct measurement of points and fees for a given loan. In

order to estimate the impact of the cap, the analysis instead compares approval rates before and

after the implementation of the Rule across loan size thresholds established under the Rule.

Specifically, the analysis relies on the observation that, given the structure of the cap, it is likely

to be more or less restrictive at different loan sizes. For instance, given a fixed points and fees

cap of $3,000 for loan sizes between $60,000 and $100,000, to the extent that some of the cost

of originating a mortgage varies positively with loan size, the cap is expected to be more

restrictive for loans at or just below $100,000 in size as compared to loans at or just above

$60,000 in size. Therefore, if the points and fees cap has a negative effect on the rate at which

281

T h e value of $170,000 is the m edian loan size in 2011 before the beginning of the data under study. Restricting to

loa n sizes below this median results in a set of loans that are similar to each other in important market

characteristics, such as geography. For example, hom es with h igher prices—and therefore higher loan sizes—are

m or e likely to be located in metropolitan areas. In com parison, the smaller size loans t hat are the focus of t his

a n alysis are m ore likely t o b e found in n on -m etropolitan, micropolitan, or smaller metropolitan areas.

174 BUREAU OF CONSUMER FINANCIAL PROT ECTION

loan applications are approved, among loans between $60,000 and $100,000 in size, this effect

is expected to be more pronounced for larger loans.

To test this hypothesis, the analysis defines loan size bins of width $10,000, with the first bin

covering $60,000 to $69,999 and the last bin covering $90,000 to $99,999.

282

A statistical

model is constructed where approval is modeled as a function of occupancy status, the month

and year of application, the county where the property is located, whether the property is located

inside a metropolitan statistical area, the loan purchaser type, and loan size bin. Then an

additional term is introduced to allow the approval rate by loan size bin to be different starting

in 2014. Again, under the hypothesis outlined above regarding the effect of the points and fees

cap, the estimated effects after 2014 would show a lower approval rate for the higher size bins

than for the lower size bins relative to the earlier years.

Site-built home loans

Figure 54 shows the estimated effects for site-built home loan applications (classified in HMDA

as applications for one to four family home loans other than manufactured housing).

282

The analysis is restricted to first-lien conventional purchase loans.

175 BUREAU OF CONSUMER FINANCIAL PROT ECTION

FIGURE 54: A PPROV A L RA TE MODEL ESTIMA TES FOR SITE-BUILT HOME LOA NS, 2012-2016

The green bars show for each loan size bin the estimated approval rate for a loan with average

occupancy status in the average county and the average month and year of origination over the

2012-2013 period. Higher size loans experienced slightly higher approval rates between 2012

and 2016. When allowing for differential approval rates starting in 2014, there is no discernible

difference at higher loan size bins. The black bars represent the 95 percent confidence intervals,

which are tightly estimated. The model shows no statistically significant changes in the relative

approval rates after 2014. This implies that the points and fees cap was not binding for these

loans.

Manufactured home loans

Figure 55 shows the estimated effects on the approval rate for originations for properties

classified as manufactured homes in HMDA.

176 BUREAU OF CONSUMER FINANCIAL PROT ECTION

FIGURE 55: A PPROV A L RA TE MODEL ESTIMA TES FOR MA NUFA CT UR ED HOME LOANS, 2012-2016

Manufactured home loan applications have a substantially lower approval rate than those for

site-built homes. Unlike with site-built home loans, higher loan size applications have a lower

approval rate than lower loan size applications. After 2014, the relative approval rate of higher

size loan applications increases, and this increase is statistically significant. This does not lend

support to the hypothesis that the points and fees cap suppressed the approval rate of higher

size loans relative to lower size loans in this loan size range.

283,284

283

Not ice that these results are r eflective of a l arger t rend among m anufactured housing l oan a pplications during this

period, that of the approval rate of larger loans becoming higher relative to that of smaller loans. In particular, the

a v erage a pprov al rate of a pplications with a size between $2 0,000 and $5 9,99 9 dropped from 48.3 percent in 2 012

t o 4 0 .8 percent in 2016, the average a pprov al rate of a pplications with a size b etween $6 0,000 and $99 ,999

dr opped from 41.7 percent in 2012 t o 3 7.5 percent in 2016, w hile the average approv al rate of a pplications with a

size b etween $1 00,000 and $1 69,999 increased from 3 3.6 percent in 2012 to 41.3 percent in 2016.

284

In terms of interpreting these results, it is important to note the changes to the Home Ownership and Equ ity

Pr ot ection Act (HOEPA) pursuant to the Dodd-Frank Act w ere im plemented at t he same time as the ATR/QM Ru le.

These changes likely increased the share of manufactured home loans that are classified as HOEPA loans

su bstantially. See Bureau of Consumer Fin. Prot., Manufactured-housing Consumer Finance in the United States,

(Sept. 2014), available at https://files.consumerfinance.gov /f/201409_cfpb_report_m anufactured-h ou sing .pd f.

177 BUREAU OF CONSUMER FINANCIAL PROT ECTION

While the potential for an effect from the points and fees caps is most clear from a fixed cap, as

between loan size of $60,000 and $100,000, a percentage cap can also have an effect on the rate

of approvals. If some of the cost of originating a loan is fixed as opposed to changing with the

size of the loan, then a percentage cap becomes less restrictive as loan size increases. This

hypothesis is tested for loans between $20,000 and $60,000 (where the points and fees cap is

five percent of the loan size) and for loans between $100,000 and $170,000 (where the points

and fees cap is three percent of the loan size). Among manufactured housing loan applications,

the relative approval rate increased with loan size among these ranges, too. While this does not

rule out the above hypothesis, it may also be a result of the larger trend of a shift towards larger

loans in this market documented in Chapter 3.

Using a similar methodology, the Bureau also analyzed year-on-year growth rates of originations

at the state level taking into account home price changes and allowing for variation with the

quarter and year of origination, the state, and by loan size bin. No statistically significant effects

of the points and fees cap were found. The lack of statistical significance is partly due to the

small sample size (unlike the previous analysis that uses individual data, this analysis relies on

state-level observations), but the point estimates do not indicate an economically significant

effect either.

5.4.6 The effect of the QM points and fees provision in the

broker segment

The Rule specifies that the loan originator compensation paid by a creditor to a non-employee

(e.g., a mortgage broker) must be included in points and fees, even if the creditor is paying this

fee on the consumer’s behalf.

285

In contrast, the points and fees formula does not include

payments that the creditor makes to its own employees. As a result, brokered transactions will

generally have a higher sum of points and fees than retail transactions, and thus are more likely

to exceed the QM PF threshold. The Bureau has examined the available data to investigate

whether this is indeed the case, and whether brokered transaction have declined after the

Rule.

286

285

The ATR/QM Rule implements the loan originator compensation part of points and fees. See C.F.R.

§ 1 02 6.32(b )(1)(ii). T he discussion of this prov ision summarized in the t ext above is from the preamble t o the

ATR/QM Rule. See 78 Fed. Reg. 6408, 6432–6438 (Jan. 30, 2013).

286

For a dditional research on the role of m ortgage brokers in lending com petition and pricing, see M. Cary Collins

and Keith D. Harvey (2010), Mo rtgage Bro k ers and Mortgage Rate Spreads: Their Pricing Influence Depends on

178 BUREAU OF CONSUMER FINANCIAL PROT ECTION

TABL E 34 : PERCE N T OF PF V IOLA TIONS IN THE A PPL ICA T ION DA TA FROM NINE LENDERS: BROKER

CHA NNEL

Lender

Approved

in 2013

Approved

in 2014

Approved

in 2015

Approved

in 2016

Denied

in 2014

Denied

in 2015

Denied

in 2016

43.7%

41.1%

43.9%

46%

35.8%

100%

1.9%

0.0%

1.5%

0.0%

# 0.0% 0.0% 0.0% 7.6% 0.027 5.5%

100%

1.6%

0.0%

2.0%

0.0%

33.3%

0.3%

0.0%

0.3%

0.0%

100%

99.8%

4.4%

1.2%

0.8%

99.3%

100%

Total

33.3%

4.8%

5.2%

7.3%

9.4%

4.5%

7.8%

99.9%

2.0%

0.3%

0.2%

11.9%

16.9%

11%

Table 34 presents the percentage of brokered applications where the QM PF threshold was

exceeded, separately for three lenders (the other six lenders did not utilize brokers). For each

lender–year combination, the first row indicates the incidence of PF violations; the second row

indicates the percentage of records where information on PF status is missing. Only one lender

shows a significant percentage of approved brokered transactions where PF is exceeded. The

other two lenders are curing all or almost all PF violations on the approved applications. Among

denied applications, for two lenders reported in Table 34, the percentage of PF violations in the

broker segment is higher than in the retail segment. For the remaining lender, almost all data

on denied brokered applications is missing. To conclude, this limited examination suggests that

brokered applications are initially more likely to result in a PF violation.

Neighborhood Type, Journal of Housing Research 19(2); Amany El Anshasy, Gregory Elliehausen, and Yoshiaki

Shimazaki, The Pricing of Subprime Mortgages by Mortgage Brokers and Lenders, July 2005, available at

h t tp://www.chicagofed.org/digital_assets/others/events/2 005/prom ises_and_pitfalls/paper_pricing.pdf.

179 BUREAU OF CONSUMER FINANCIAL PROT ECTION

TABL E 35 : SHA RE OF BROKERE D A PPLICA T IO NS IN THE A PPLICA TIO N DA TA FROM THREE LENDERS

THA T EMPL OY BROKER S

Lender

2013 (all

loans)

2014 (all

loans)

2015 (all

loans)

2016 (all

loans)

2013

(<150k)

2014

(<150k)

2015

(<150k)

2016

(<150k)

# 2.1% 3.8% 5.4% 10.7% 6.3% 9.3% 12% 17.7%

# 21.6% 19.1% 15.5% 11.4% 14.7% 11.3% 13.1% 8.6%

# 3.5% 3.5% 6.1% 5.5% 2% 2% 3.9% 3.5%

Total 7.6% 9% 9.7% 9.2% 5.8% 6.2% 8.4% 7.1%

Table 35 displays the share of brokered applications for three lenders. While there are some

fluctuations, there is no evidence that these lenders have systematically reduced their reliance

on brokered loans after the introduction of the Rule. A similar result is obtained if one examines

the share of brokered loans in the CoreLogic data, as seen in Figure 56.

180 BUREAU OF CONSUMER FINANCIAL PROT ECTION

FIGURE 56: SHA RE OF BROKERE D ORIGINATIONS AMONG MORTGA G E LOA NS FOR HOME PURCHA S E,

2010-2017.

The Bureau has received numerous comments from mortgage brokers with examples that

suggest that the QM PF provision may have created difficulties both for brokers and for

consumers involved in these transactions. Unfortunately, the Bureau could not reasonably

obtain data that would allow it to evaluate these examples on a market-by-market lev el. T he

impact of the QM PF provision in the mortgage broker segment is an area that requires further

research.

5.5 The rebuttable presumption provision

The Rule’s safe harbor and rebuttable presumption provisions provide different liability

protection for QM loans depending on whether they are higher-priced covered transactions

(HPCTs). HPCT s are generally defined as first-lien mortgages with annual percentage rates

181 BUREAU OF CONSUMER FINANCIAL PROT ECTION

(APRs) that are 1.5 or more percentage points over the benchmark Average Prime Offer Rate

(APOR) for a comparable transaction, and second-lien mortgages with APRs that are 3.5

percentage points over the comparable APOR.

287

For Small Creditor Portfolio and Small

Creditor Balloon Payment QMs, the first-lien HPCT threshold is an APR that is 3.5 or more

percentage points over APOR.

288

QM loans that are not HPCTs, referred to here as “Safe Harbor QMs,” receive a complete safe

harbor from civil liability—i.e., the Rule conclusively presumes creditors originating these loans

complied with the Ability-to-Repay (ATR) requirements.

289

By contrast, the Rule establishes

only a rebuttable presumption that creditors originating QM loans that are HPCTs complied

with the ATR requirements—i.e., a consumer who purchased a HPCT qualified mortgage can

provide evidence to attempt to rebut that presumption.

290

For example, the Rule provides that a

consumer may rebut the presumption with evidence demonstrating that the consumer’s residual

income was insufficient to meet living expenses. With the potential legal risk associated with

HPCTs, this section analyzes how the Rule’s rebuttable presumption provision may have

impacted HPCT originations.

Lenders had requirements to monitor their origination of higher-priced loans prior to the

adoption of the ATR/QM Rule. The Board began tracking loan pricing data for higher-priced

loans through HMDA in 2004, so that government agencies would be able to “identify more

easily price disparities that require investigation.”

291

In 2008, the Board adjusted the reporting

thresholds for higher-priced loans, and issued amendments to Regulation Z which defined these

loans as higher-priced mortgage loans (HPMLs).

292

The thresholds for HPMLs are generally the

same as the first-lien (other than for small creditor QMs) and second-lien HPCT APR/APOR

thresholds (other than for small creditor QMs).

293

The Board’s rule also required that a creditor

287

1 2 C.F.R. § 1 026.35(a)(1).

288

1 2 C .F.R. § 1 026.43(b)(4).

289

1 2 C .F.R. § 1 026.43(e)(1)(i).

290

1 2 C .F.R. § 1 026.43(e)(1)(ii).

291

Fed. Reserve Boa rd, Frequently Asked Questions About the New HMDA Data, (Mar. 31, 2005), available at

h t tps://w ww .federalreserve.gov/b oarddocs/press/bcreg/2005/2 0050331/attachment.pdf.

292

Fed. Fin. In sts. Ex amination Council, History of HMDA, https://www.ffiec.gov /hmda/history2.htm (last

m odified S ept. 6, 2018).

293

1 2 C.F.R. § 1 026.35(a)(1). The threshold is 2.5 or more percentage points over APOR for “ jumbo” loans. 12 C.F.R. §

1 026.35(a)(1)(ii).

182 BUREAU OF CONSUMER FINANCIAL PROT ECTION

make an ATR determination, and only applied this requirement to HPMLs.

294

In 2010, the

Dodd-Frank Act extended the ATR requirement to all mortgage loans, beginning in January

2014 when the ATR/QM Rule took effect.

Using HMDA data, this section first analyzes whether the Rule’s rebuttable presumption

provision had an immediate impact on HPML origination volume. HPMLs, because of their

nearly identical definition, may serve as a proxy for HPCTs. Since the Board’s 2008 rule

required lenders to make an ATR determination for HPMLs, and therefore already increased the

potential legal risk associated with these loans, the impact in this category is likely muted

(although cannot be ruled out a priori). The analysis focuses on conventional first-lien

mortgages originated for the purchase of owner-occupied homes from 2012 to 2016. Site-built

homes and manufactured homes are evaluated separately. Loans insured and guaranteed by the

Federal Housing Administration (FHA), the U.S. Department of Veterans Affairs (VA), and the

U.S. Department of Agriculture/Rural Housing Service (USDA/RHS) are not subject to the

Bureau’s ATR/QM Rule. For such loans, QM status is determined using each Agency’s own

metrics. For this reason, the analysis excludes these loans.

The second part of this section analyzes loan rate spreads, the difference between the APR of a

loan and APOR, to assess if the rebuttable presumption rate spread threshold was a binding

constraint for lenders after the effective date of the Rule. An increase in loan originations

directly under the threshold would suggest an immediate impact of the rebuttable presumption

provision of the Rule, as it might indicate that some lenders have responded to the Rule by

originating loans that are just within 1.5 percentage points of APOR to maintain their Safe

Harbor status. This section makes use of data obtained through several fair lending

examinations. For the reasons discussed in the preceding paragraph, these data are similarly

restricted to conventional first-lien owner-occupied purchase loans but only include site-built

originations and not manufactured housing loans. The data are further restricted to only include

lenders whose exams cover a pre- and post-Rule period.

5.5.1 Site-built home loans

The level of HPML lending as reflected in the HMDA data for a period before and after the

effective date of the Rule as shown in Figure 57. From 2012 to 2016, HPMLs were a small share

of mortgage originations; they represented fewer than 4 percent of first-lien conventional home

purchase loan originations for owner-occupied site-built homes. HPML lending slightly

294

In 2008, the Board also revised the definition of HPML.

This definition matches the one later adopted for HPCTs

in the ATR/QM Ru le. See Truth in Lending, 73 Fed. Reg. 44522 (July 30, 2008).

183 BUREAU OF CONSUMER FINANCIAL PROT ECTION

increased after the adoption of the ATR/QM Rule (2014 to 2016), but overall, the share of

HPMLs originated remained relatively constant and did not vary beyond one percentage point

over the course of five years.

Figure 57 also shows the share of HPMLs originated for three loan size groups based on some of

the points and fees thresholds—loans less than $60,000, loans greater than or equal to $60,000

up to $100,000, and loans greater than or equal to $100,000. Under the ATR/QM Rule, small-

balance mortgages have higher limits on points and fees to qualify as a QM than larger loans. As

small-balance mortgages are often more expensive to originate, as they have the same fixed

costs as larger loans, yet bring in less revenue, it is important to analyze the potential impact on

these different loan size groups. Additionally, the definition of an HPML depends on the APR for

a loan which can be affected by the fees relative to the size of the loan.

FIGURE 57: HPML ORIGINA TION SHA RE BY Y EA R A ND LOA N SIZ E, SITE-BUIL T HOMES 2012-2016

In Figure 57, it is clear that HPML lending is more common among the loans in the smaller loan

size buckets. However, over time, each loan size group has a similar pattern. There were no large

changes from year to year, and since the adoption of the ATR/QM Rule, the share of HPMLs

184 BUREAU OF CONSUMER FINANCIAL PROT ECTION

originated slightly increased. Loans under $60,000 experienced a small decrease in the share of

HPMLs originated directly following and leading up to the ATR/QM Rule effective date in

January 2014; however, HPML lending in that group promptly rebounded in 2015.

5.5.2 Manufactured home loans

The majority of mortgages originated for the purchase of manufactured homes are HPMLs.

From 2012 to 2016, over three quarters of first-lien conventional mortgage originations for

owner-occupied manufactured homes were HPMLs, as seen in the following figure. Similarly to

site-built homes, the change was quite small, and the share of HPMLs originated remained

relatively constant over the five year period. Additionally, an analysis of the loans by loan

balance suggests that this small decrease is driven by a change in the size of the loans originated

rather than a direct change in lender pricing.

FIGURE 58: HPML ORIGINA TIO N SHA RE BY YEAR A ND LOA N SIZE, MA NUFA CT URE D HOMES 2012-2016

Manufactured housing loans are divided into the same buckets as in Figure 58. When broken

out into these loan size groups, in contrast with the aggregate picture, the share of HPMLs

185 BUREAU OF CONSUMER FINANCIAL PROT ECTION

originated in each size group increases in the years following the Rule’s effective date (2014 to

2016). Again, this increase is small. However, the overall decrease in the share of HPMLs

originated for manufactured homes is primarily driven by an increase in the origination of loans

with larger balances. In fact, the share of manufactured housing loans greater than or equal to

$100,000 increased from 12 percent of originations in 2013 to 20 percent of originations in

2016. As with site-built homes, HPML lending is more common for manufactured housing loans

with smaller balances. So an increase in the share of larger loans originated drove down the

overall share of HPMLs originated.

5.5.3 Analysis of rate spreads

Pricing data from a sample of seven fair lending exams that cover the pre- and post-Rule periods

provided the APR for approximately 60,000 conventional loans. Rate spreads were then

calculated for each loan using the respective APOR rate effective the same week as the loan rate

lock date. Figure 59 shows the distribution of the computed rate spreads for 2013 and 2014.

Post-Rule bunching directly below the 1.5 percentage point threshold, as indicated with a

vertical grey line, is not apparent when comparing the 2013 and 2014 distributions. Bunching is

typically associated with a binding constraint, as lenders change parameters of the loan (in this

case, rate spread) to stay under a regulatory threshold.

186 BUREAU OF CONSUMER FINANCIAL PROT ECTION

FIGURE 59: RA TE SPREA D DISTRIB UTION BY YEAR, 2013 AND 2014

To further analyze potential bunching effects not directly observed in the Figure 59

distributions, loans are divided into two rate spread groups around the threshold—rate spreads

above the 150 basis point threshold and rate spreads between 100 and 150 basis points above

APOR. Figure 60 reports the share of originations for each rate spread group between the

second quarter of 2013 and the fourth quarter of 2015

295

. Consistent with the patterns seen in

Figure 59, no substantial shift in the shares of originations with differing rate spreads is

observed in Figure 60 around the Rule’s effective date. Loans with rate spreads below 100 basis

points are not shown on the figure but commanded the majority of the share of total loans

(above 90 percent) in the data compared to the between and above rate spread groups. The

share of loans above the threshold experienced a slight increase in the beginning of 2015 but still

remained below 2 percent of total loans.

295

Du e t o t he a forementioned r estrictions, data from the Fa ir Lending exams is on ly available b eginning in the third

qu a rter of 2013 until the end of 2 015.

187 BUREAU OF CONSUMER FINANCIAL PROT ECTION

FIGURE 60: ORIGINA TIO N SHA RE BY RA TE SPREA D GROUP, 2013Q3-2015Q4

Based on the findings above, the Rule’s rebuttable presumption provision does not appear to

have had a significant impact on HPML lending, nor is there systematic evidence that the rate

spread threshold is binding. The share of mortgage origination volumes accounted for by site-

built and manufactured housing HPMLs were relatively steady from 2012 to 2016, suggesting

that lenders were not driven away from originating HPMLs after the implementation of the Rule.

The Bureau’s findings from the Fair Lending data also suggest that the rebuttable presumption

threshold of 150 basis points above APOR was not a binding constraint for lenders as the share

of originations above the threshold remained steady after the implementation of the Rule.

These results are likely explained by the fact that the Board’s 2008 rule applied an ATR

requirement to HPMLs. These results do not rule out the possibility that the Board’s rule

significantly impacted HPML lending however, analysis of that question is beyond the scope of

this Report.

188 BUREAU OF CONSUMER FINANCIAL PROT ECTION

6. The Temporary GSE QM

This chapter considers trends in the volume, characteristics and sale into the secondary market

of conforming loans that are originated under the Temporary GSE QM provision of the

AT R/QM Rule. Previous chapters considered the performance of these loans and the role that

they have played in preserving access to credit. This chapter considers potential explanations for

the large and persistent market share of Temporary GSE QM originations in the conforming

segment of the market.

296

The main findings include the following:

• The GSE’s share of the conventional purchase market was large prior to the Rule’s

introduction and has seen a further small increase in the years following. The large and

persistent market share may be attributable to a range of factors which distinguish GSE

loans from those made under the General QM and ATR criteria, potential advantages in

compliance certainty and flexibility, and robust secondary market liquidity. As a result of

these and other factors, at least with respect to loans originated for sale on the secondary

market, given the option to extend mortgage credit to a particular borrower through a

GSE loan, originators have generally done so.

• The market for private label mortgage backed securities remains quite small relative to

its pre-crisis level. This limits the funding available for loans that are not eligible for

purchase or guarantee by the GSEs or government agencies, a category of loans that

includes non-QM loans. Although there have been some issuances containing non-QM

loans, the majority of new private label securities consist of prime jumbo loans made to

borrowers with strong credit characteristics.

• The evidence does not suggest that there was an immediate shift to increased use of the

GSEs’ Automated Underwriting Systems (AUSs) for loans not intended to be sold to the

GSEs, as a preferable method of establishing a loan’s QM status compared to meeting the

296

The jumbo segment of t he m arket is analyzed in Chapters 4 and 5 and is not directly affected by the T emporary

G SE Q M pr ov isions. Thus, t his Chapter considers on ly t he conforming segment of t he m arket.

189 BUREAU OF CONSUMER FINANCIAL PROT ECTION

General QM underwriting requirements. However, the data do suggest a somewhat

greater use of the GSEs’ AUS in recent years, particularly for loans which do not fit

within or are more difficult to document within the General QM underwriting standards,

such as loans made to self-employed borrowers.

Section 6.1 briefly reviews the Temporary GSE QM criteria and the Bureau’s expectations, stated

at the time of the rulemaking, of how QM and non-QM lending would evolve over time. Section

6.2 presents trends in the share of GSE originations in the years before and after the Rule.

Section 6.3 describes certain functional features of the T emporary GSE QM requirements and

considers how these features may have contributed to the large and persistent market share of

Temporary GSE QM loans in the conforming segment. These fundamental features are

compliance certainty and flexibility, the ability to accommodate high debt-to-income mortgage

demand, and access to liquidity through the secondary market. This section also presents

empirical results on the use of GSE eligibility to secure QM status. Section 6.4 then briefly

considers four goals of the QM requirements and draws on the analysis in the previous section

to inform why these goals have or have not been met.

6.1 Background

As discussed in detail in Chapter 2, the Temporary GSE QM, sometimes referred to as the Patch,

is a temporary qualified mortgage category that under the terms of the Rule will be in effect until

the earlier of: (i) the end of GSE conservatorships; or (ii) January 10, 2021.

297

The Temporary

GSE QM category includes the product and cost restrictions that generally apply to qualified

mortgages. However, the Temporary GSE QM category generally uses the GSE underwriting

standards instead of the General QM standards and does not establish a DTI threshold. The

General QM underwriting standards include Appendix Q of the Bureau’s Regulation Z and the

43 percent threshold on DTI.

298

As with other types of QM loans, the presumption of compliance

for Temporary GSE QM loans can be either conclusive, i.e., a safe harbor, for QM loans that are

not “higher-priced”; or rebuttable, for QM loans that are “higher-priced.”

299

297

At the time the Rule took effect, the temporary category of qu alified mortgages also included loans eligible to be

guaranteed or insured (as appropriate) by the U.S. Department of Housing and Urban Development, U.S. Dept. of

Veterans Affairs, or the U.S. Dept. of Agriculture or Rural Hous. Serv. These provisions of the temporary c ategory

ph a sed ou t as these federal agencies issue t heir own qualified m ortgage rules and w ould have expired after seven

y ears. See 78 Fed. Reg. 6408, 6409 (Jan. 30, 2013).

298

1 2 C .F.R. § 1 026.43(e)(2)(v)–(vi).

299

1 2 C.F.R. § 1 026.43(e)(1).

190 BUREAU OF CONSUMER FINANCIAL PROT ECTION

In establishing the categories of temporary QM loans, the Bureau stated that it sought to

preserve access to credit for consumers with debt-to-income ratios above 43 percent during a

transition period in which the market was fragile and the mortgage industry was adjusting to the

final rule.

300

By providing for most of the conventional market to continue to originate higher

debt-to-income loans as QM loans, but limiting this to the conforming market and making the

provision temporary, the Bureau sought, over the long term, to encourage innovation and

responsible lending on an individual basis under the ability-to-repay criteria. The Bureau

expected that there would be a robust and sizable market for non-QM loans beyond the 43

percent threshold and structured the Rule to try to ensure that this market would develop.

The Bureau also stated that because the temporary category of QM loans covers loans that are

eligible to be purchased, guaranteed, or insured regardless of whether the loans are actually

purchased, guaranteed, or insured, private investors could acquire these loans and secure the

same legal protection as the GSEs and Federal agencies. This would avoid creating a disincentive

for the return of private investors even before the expiration of the temporary category.

Finally, the Bureau noted that as the market recovered, the GSEs and federal agencies would be

able to reduce their presence in the market (e.g., by reducing their loan limits). In this scenario,

the percentage of loans granted qualified mortgage status under the temporary category would

also shrink and the market would be able to develop alternative approaches to assessing ability-

to-repay within the General QM requirements.

The continued prominence of Temporary GSE QM originations is contrary to the Bureau’s

expectations at the time of the rulemaking, and certain goals of the Rule have therefore not been

met. In accounting for the continued prominence of Temporary GSE QM originations, two

factors can be distinguished. First, the scope of GSE-eligible loans is broad, and it grew even

broader for a period of time after the Rule became effective as the GSEs loosened their credit

eligibility in various respects. Second, for a number of reasons, investors in mortgage-backed

securities favor funding GSE-guaranteed loans over other loans, including GSE-eligible, General

QM and non-QM loans. Thus, at least with respect to loans originated for sale on the secondary

market, given the option to extend mortgage credit to a particular borrower through a GSE-

loan, originators will generally do so; and any expansion of the scope of GSE eligible loans will

grow the share of Temporary GSE QM originations. To the extent there is a preference for GSE-

eligible but not guaranteed loans over General QM or non-QM loans either among investors or

among creditors originating loans to hold in portfolio, this too will contribute to the prominence

of Temporary GSE QM originations. This chapter addresses both of these factors in considering

300

See Chapter 1 at Section 1.1.3, for r eferences.

191 BUREAU OF CONSUMER FINANCIAL PROT ECTION

potential explanations for the large and persistent market share of Temporary GSE QM

originations in the conforming segment of the market.

6.2 Conforming originations since the

implementation of the Rule

Chapter 4 examined changes in the DTI distribution of GSE loans and found that for loans

originated in 2014, there was an upward shift in DTIs for GSE loans, which was most

pronounced among loans with DTIs approaching 45 percent. Chapter 5 examined changes in the

share of high DTI loans among GSE and non-GSE loans and, for the nine lenders in the

Application Data, among GSE-eligible and non-GSE eligible originations. Those data show a

decline in high DTI lending in the non-GSE space relative to the GSE space and thus the

continued prominence of the Temporary GSE Exemption among high DTI borrowers. Thus,

although the Bureau expected that loans with DTI above the 43 percent threshold would

increasingly be originated outside the Temporary QM category, i.e., as non-QM loans, the

available data suggests that the opposite is happening.

Figure 61 broadens the analysis and presents the share of conventional purchase-mortgage

originations insured by the GSEs since 2000, for loans at or below $417,000, which was the

conforming loan limit in most counties at the time the Rule became effective.

301

The share of

GSE insured loans was large prior to the Rule’s introduction and has seen a further small

increase since the Rule’s 2014 effective date. The GSE share of conventional purchase loans

under $417,000 rose from 69 percent in 2013 to 70 percent in 2014, and remained at 71 percent

by 2017.

302

Counter to the Bureau’s expectations, the percentage of GSE insured loans has not

shrunk since the finalization of the Rule.

303

The next sections of this chapter discuss potential

reasons for the sustained GSE share of conventional purchase loans, and analyze data from GSE

301

As discussed in Section 3.5, some “high-cost” counties had higher conforming loan lim its. Given that patterns in

conforming loan originations in these typically large urban markets may reflect more local trends, the analysis of

conforming loans is restricted to loans at or below $417,000. However, the broad shifts in the GSE share over time

sh own in Figure 61 are robust to the inclusion of these larger loans in these high-cost c ounties. See Figure 1 6 in

Chapter 3 for a breakdown of purchase originations for a ll (conventional a nd nonconventional) loan types.

302

Originations of conventional refinance loans fell in aggregate from 2013 to 2014. The GSE share for such loans fell

fr om 83 percent in 2013 to 7 8 percent in 2014, and remained at 76 percent by 2017.

303

Com m enters cited industry survey r esults c onsistent w ith t hese findings, see 2 3

Annual ABA Residential Real

Est a t e Survey Report, April 2016 , available at

h ttps://www.aba.com/Tools/Function/Mortgage/Documents/2016ABARealEstateLendingSurveyReport.pdf.

192 BUREAU OF CONSUMER FINANCIAL PROT ECTION

Automated Underwriting Systems (AUS) to further assess the role of the Temporary GSE QM in

observed market trends.

FIGURE 61: LOAN TYPE COMPOSITION OF CONV ENTIONA L PURCHA S E ORIGINA T IONS UNDER $417,000,

2000 TO 2017

6.3 Functional features of the Temporary

GSE QM requirements

6.3.1 Compliance certainty and flexibility

As noted above, given the option to extend mortgage credit to a particular borrower through

either a Temporary GSE QM or a General QM, originators generally offer a T emporary GSE QM,

at least with respect to loans intended to be sold in the secondary market. While the existence of

a secondary market is certainly one factor favoring the Temporary GSE QM, there are other

193 BUREAU OF CONSUMER FINANCIAL PROT ECTION

factors that may help account for the large and persistent market share of Temporary GSE QM

originations in the conforming market.

First, Fannie Mae and Freddie Mac provide a high degree of specific detail for the method to be

used to calculate income and debt. Although admittedly a crude measure of detail, Fannie Mae

and Freddie Mac guidelines for creditors originating loans for sale to them each provide 108 and

125 pages,

304

respectively on these topics. In contrast, the regulatory text of Appendix Q is

contained within only 11 pages.

305

Second, there is a perceived lack of clarity in Appendix Q. The Bureau viewed the use of FHA

guidelines as providing clear, well-established standards for determining whether a loan is a

qualified mortgage.

306

However, some respondents to the RFI disagree. For example,

respondents to the RFI stated that it “is ambiguous and leads to uncertainty,” “confusing and

unworkable” and that “additional guidance . . . is needed.”

307

These concerns with Appendix Q

may have contributed to investors’—and at least derivatively, creditors’—preference for

Temporary GSE QM lending and interfered with the achievement of policy goals for the

Temporary GSE QM category.

308

Third, Appendix Q has been static since the adoption of January 2013 Rule. In contrast, the

T emporary GSE QM provides flexibility and has changed over time. Flexibility potentially allows

for clarification and refinement in the face of ever-changing market conditions as well as for

innovation as discussed later in this Chapter. The GSEs regularly adjust and update their

underwriting guidelines, often monthly and sometimes more frequently. These changes affect

allowable DTI calculation methods and can address emerging issues with respect to the

304

In th e O ctob er 2, 2 018, the PDF v ersion of Fannie Ma e’s S elling Guide, Chapter B3-3, Income Assessment is 86

pa g es ( pages 313–398) and Chapter B3-6, Liability Assessment is 22 pages (pages 501–522). In the October 1 8,

2 01 8, the PDF v ersion of Fr eddie Ma c’s S ingle-Family Seller/S ervi cer Guide, T opic 5 300, Stable Mon thly In com e

an d Asset Qualification is 106 pages (pages 5301-1–5 301-6, 5302-1–5 302-9, 53 03 -1 –5 303-33, 5304-1 –5 304-1 4,

5 305-1–5 305-1 6 , 53 06 -1–5 306-1 8, 5 307-1–5 307-1 0) and T opic 5 400, Ev aluation of Mon thly Obligations is 19 pages

(pa ges 5 401-1–5401-1 9).

305

1 2 C.F.R. § 1 026, appendix Q, at 446–456.

306

7 8 Fed. Reg. 6408, 6527 (Jan. 30, 2013).

307

See A ppendix B.

308

Perhaps supporting an assessment of ambiguity, the Federal Housing Administration, the agency responsible for

t h e source m aterial, “prov ide[ed] more definitive underwriting standards . . . to overcome lender uncertainty,” by

r ev ising the source m aterial in December 2013 . See Truth in Lending, 78 Fed. Reg. 7 5238, 75243 (Dec. 11, 2013).

194 BUREAU OF CONSUMER FINANCIAL PROT ECTION

treatment of certain types of debt or income categories.

309

In contrast, as discussed in Chapter 5,

of the 87 respondents to the lender survey who responded to a question regarding Appendix Q,

27 percent said that sometimes borrowers who were approved for loans could not provide

documentation required by Appendix Q and 8 percent said this was often true.

310

Finally, although technically the Temporary GSE QM applies to loans that are eligible for

purchase or guarantee by one of the GSEs, market participants believe that extra compliance

certainty is assured for loans actually sold to the GSEs.

6.3.2 Accommodating high debt-to-income mortgage

demand

A further reason for the continued use of Temporary GSE QM is that the GSEs were able to

accommodate demand for mortgages above the 43 percent DT I ceiling as the DTI distribution

shifted up in recent years due to house price appreciation, increases in debt load (especially for

those with student loans) and other factors. At the time of the rulemaking, the Bureau

understood that FHA had been using the 43 percent DTI threshold for many years as a general

boundary for defining affordability. The Bureau found the threshold a relatively liberal one

relative to the GSE guidelines with a benchmark of 36 percent, before consideration of

compensating factors.

311

However, while the Bureau was aware at the time of the rulemaking

that 18 percent of GSE and federal agency loans had a DTI over 43 percent,

312

the Bureau did not

attempt to predict how readily the GSEs—and thus the Temporary GSE QM category—would

accommodate loans with higher DTI as house prices and the interest rate recovered. Indeed, the

Bureau expected that over time the GSEs role in the housing market would shrink.

In fact, the opposite has occurred, especially within the segment of high DTI borrowers.

Evidence presented in prior chapters shows that high-DTI loans have recently been an

increasing share of Temporary GSE QM originations. Figure 35 of Chapter 4 demonstrates the

rising DTIs of GSE originations in the year following the effective date of the Rule, while Figure

309

T h ere can, howev er, be a t radeoff between flexibility and com pliance c ertainty. T o support t he pace of t hese

u pdates, both Fannie and Freddie prov ide robust im plementation su pport to lenders and ot her stakeholders on

their websites with videos, fact sheets, searchable FAQs, training schedules and various job aids.

310

See T ab le 2 2 in this report.

311

Id. a t 6 505 (“[T]he 4 3 percent threshold has been utilized by the Federal Housing Adm inistration (FHA) for m any

y ears as its g eneral boundary for defining affordab ility. Relative t o other b enchmarks that are used in the m arket

(su ch a s GSE guidelines) that have a benchmark of 36 percent, before c onsideration of c om pensating factors, this

th reshold is a relatively liberal one which a llows am ple room for consumers to qu alify for an a ffordable m ortgage”).

312

Id. a t 6 569 (“Based on t he data a s of year-end 2011, such loans are approximately 18 percent of the market.”).

195 BUREAU OF CONSUMER FINANCIAL PROT ECTION

38 of Chapter 5 specifically notes the increased share of originations with a DTI over 43 percent

through 2017. Some of this growth is likely a product of rising house prices as well as rising

interest rates, which directly increase borrowers’ required monthly payments for any given loan

size, but more recent growth also is attributable to actions by the GSEs.

313

Each of the GSEs uses a proprietary automated underwriting system (AUS) to determine

eligibility for most of its business and with each new release of the AUS the GSEs can adjust

their criteria. In particular, in May 2017, Fannie Mae announced that its July 2017 release of its

Desktop Underwriter (DU) would include an expansion of high-DTI eligibility by removing the

preexisting requirement that borrowers with DTIs abov e 4 5 percent have at least 12 months of

reserves and a loan-to-value of at least 80 percent. Fannie explained that, “A higher DTI

presents a higher degree of risk and therefore, the updated risk assessment (DU version 10.1)

will require compensating risk factors to address this additional risk. However, for loans with up

to 50% DTI, the assessment will now be made entirely within the DU risk assessment and

without the use of a model overlay.”

314

This policy change resulted in a dramatic increase in high-DTI originations by the GSEs. For

example, Fannie Mae reported that its purchases with DTIs over 45 percent increasing from 6

percent in June 2017 to 19 percent in December 2017.

315

For the first five months of 2018, 29

percent of Fannie Mae’s loans and 21 percent of Freddie Mac’s loans had DTI ratios above 43

percent, up from 13 and 14 percent respectively in 2013.

316

These increases were larger than

anticipated, and, after evaluating the profile of loans Fannie Mae responded by tightening their

DU underwriting criteria for such loans in March 2018 to limit “risk layering.

317

Over the same

period, Freddie Mac made no significant announced changes to their compensating factors

313

A v erage 3 0-year fixed rate mortgage interest rates increased from a recent weekly low of 3 .41 perc ent in July 2016

t o a s h igh as 4.94 percent in Nov emb er 2018, b ased on Freddie Ma c Prim ary Mor tgage Market Survey data,

available at http://www.freddiemac.com /pm ms.

314

See Steve Holden & Walt Scott, Desktop Underwriter Version 10.1 – Updates to the Debt-to-Income (DTI) Ratio

Ass essment, C redit Risk S haring C om mentary, Fannie Ma e (July 1 0, 2017), available at

h t tp://www.fanni emae.com /p ortal/fund ing-the-m arket/credit-risk/news/desktop-underwriter-deb t-to-in com e-

r a tios-071017.html.

315

See “Fannie Mae’s Efforts to Ease Mortgage Access Show How Hard it is to Balance Risk and Access,” April 5, 2018

available at https://www.urban.org/urban-wire/fannie-m aes-efforts-ease-m ort g a ge -access-sh ow -h ow -hard -it-

balance-risk-and-access.

316

Karan Kaul & Laurie Goodman, Updated: What, If Anything, Should Replace the QM GSE Patch, Hous. Fin. Pol’y

Ctr . Com mentary (2018), available at https://www.urban.org/research/publication/updated-what-if-anything-

sh ou ld-replace-qm -gse-patch.

317

See Fannie Mae, Desktop Underwriter/Desktop Originator Release Notes DU Version 10.2, (Jan. 30, 2018),

available at https://w ww.fanniemae.c om /content/release_notes/du-do-release-notes-03172018.pdf.

196 BUREAU OF CONSUMER FINANCIAL PROT ECTION

required for loans with high DTIs, and generally saw the high DTI share of their overall portfolio

increase gradually.

In contrast, the underwriting guidelines and DTI limits for General QM loans have remained

static since they were issued. As noted above, these calculation methods under the General QM,

which are provided in Appendix Q, are a subject of concern for a number of commenters on the

RFI.

6.3.3 Liquidity through the secondary market

A final reason for the continued use of the Temporary GSE QM relative to the General QM is the

immediate liquidity available to creditors through the robust secondary market available for

loans originated to the GSE standards.

When lenders adhere to the GSEs’ guidelines—guidelines that are standardized and that are

provided with robust implementation and client management support—they also gain access to

a highly liquid secondary market. In contrast, while private market securitizations have grown

somewhat in recent years, their volume is extremely small compared to their pre-crisis level. (In

2017, there were less than $20 billion in new origination PLS issuances, while the same number

was over $1 trillion in 2005.)

Figure 62 depicts the level and composition of new origination PLS issuances.

318,31 9

To the

extent that there have been private securitizations since 2014, the majority of new origination

PLS issuances consisted of prime jumbo loans made to borrowers with strong credit

characteristics.

320

These securities have a low share of non-QM loans and their non-QM loans

have better credit characteristics than non-QM loans found in securities with a high percentage

of non-QM loans.

321

The adoption of the ATR/QM rule in 2014 does not seem to have led to the

318

Sin ce the financial crisis in 2008, the majority of PLS issuances have consisted of pools of loans originated prior to

th e crisis, som etimes referred t o as “seasoned deals.” These are g enerally securities of r epackaged loans from

existing RMBS and securities of seasoned re-performing or non-performing loans.

319

Un less otherwise n oted, statistics regarding PLS are from In side Mortgage Finance.

320

For ex ample, in 2017, the average FICO score of a loan in a prim e jumbo issuance was m ore than 2 0 points greater

t h an the average score of a loan in an Agency RMBS and the average DT I was 2 percentage points lower. See In side

Mort g . Fin., Prime Jumbo MBS Characteristics: 2013 through 2018; Inside Mor tg. Fin., Agency/Channel Purpose

Loan Characteristics, https://w ww.insidemortgagefinanc e.com /data/gse_m bs_characteristics.html (last visited

Dec . 3 1 , 2018).

321

As an example, in 2017, only 3 percent of loans in p rime jumbo securities w ere interest-only, a nd therefore n on-

QM, a n d these were generally loans t o high-income borrowers.

197 BUREAU OF CONSUMER FINANCIAL PROT ECTION

development of a private market for non-QM loans, as 94 percent of the loans securitized during

this time were QM.

FIGURE 62: PLS ISSUANCES BACKED BY NEWLY ORIGINA TED LOANS BY TYPE, 2008-2017

Following the implementation of the Rule in 2014, non-QM issuances appeared first in 2015 and

continued to grow over the past three years, though they still made up only 21 percent of new

origination PLS issuances in 2017.

322

From 2015 until the beginning of the fourth quarter of

2018, 52 non-QM securities were issued.

323

On average, about 66 percent of the loans held in

these securities were non-QM loans. When observing the loan-level detail, compared to the QM

loans in the same issuances, non-QM loans were often low documentation (most often in case of

self-employed borrowers), but had similar LTVs and DTIs as the QM loans in the same

322

Da ta on non-QM securities are only available post-Rule.

323

Ra t ings agency reports w ere used to identify n on-QM issuances and their aggregate statistics.

198 BUREAU OF CONSUMER FINANCIAL PROT ECTION

issuance.

324

In terms of pricing, controlling for observable loan characteristics (such as

documentation status, whether the interest rate is fixed or variable, the purpose of the loan, the

occupancy status, the size of the loan and the loan-to-value ratio at origination) and the

borrower’s credit score and year of origination reveals that non-QM loans carried an estimated

premium of 119 basis points over safe harbor QM loans.

325,

In sum, the percentage of loans that are granted Qualified Mortgage status under the Temporary

GSE and Federal Agency QM categories has not shrunk and there appears to be limited

momentum toward a long-term structure with a more pronounced role for private market

securitization.

A review of potential explanations unrelated to the issuance of the ATR/QM Rule for the

absence of private market securitizations is outside the scope of this assessment.

326

6.3.4 The use of GSE eligibility to secure QM status

As previously noted, in defining the Temporary GSE QM to include GSE-eligible loans

regardless of whether the loans were actually guaranteed by one of the GSEs, the Rule sought to

avoid creating a disincentive that would inhibit the growth of a private securitization market.

The prior section shows that such a market has not emerged. To further assess use of the

Temporary GSE QM, the Bureau analyzed data provided by Fannie Mae and Freddie Mac on the

utilization of their respective Desktop Underwriter and Loan Prospector AUSs from 2013 Q1 to

2017 Q1. The data include counts of applications submitted and determined eligible as well as

the number of loans actually purchased by the GSEs, broken down by various loan and borrower

characteristics.

324

Loa n -level data were found for 43 of the 52 non-QM securities on the Bloom berg Terminal. T o determine the

loan -level QM status for these 43 securities, the data were m atched t o due diligence reports from the EDGAR

database at S EC .g ov of which 1 1 securities matched fully c ontaining 5,378 loans (Not c overed/Exem pt loans w ere

removed).

325

T h ere were 1 65 l oans that were QM w ith a rebuttable presumption. The estimated premium was 100 basis points

ov er safe harbor QM loans.

326

See, e.g., Laurie G oodm an, The Rebirth o f Securitization: Where is the Priv ate-Label Mortgage Mark et, (Hous.

Fin . Pol’y Ctr., Urb. Inst., Research Paper, 2015), available at

h t tps://www.urban.org/research/publication/rebirth-securitization-w here-private-label-mortgage-

m arket/view/full_report; see also Azar Abramov et al., Private-Label Mortgage Securitization Market Challenges

and the Implications for Insurers and Insurance Regulation, (Nat’l Ass’n of Insurance Comm’rs, CIPR S tudy Series

2 016–2, 2016), available at http://www.naic.org/documents/cipr_study_1 61208_private-

la bel_m ortgage_securitization.pdf.

199 BUREAU OF CONSUMER FINANCIAL PROT ECTION

Given that any loans eligible to be purchased by the GSEs are QM under the Temporary GSE

QM, the data on eligible submissions to the GSEs’ AUSs are used to assess whether lenders

responded to the rule by submitting additional loans to the AUSs beyond those intended to be

sold to the GSEs. Such a response could occur if lenders perceived the GSEs’ AUSs as a

preferable method of establishing a loan’s QM status, compared to General QM underwriting

requirements, either for loans originated for sale or for loans originate to be held on portfolio.

Such a response also could occur if lenders perceived the safe harbor or presumption of the

Temporary GSE exemption as preferable to underwriting under the ATR requirements. If

lenders used the Temporary GSE QM in either or both of these ways, it would be reflected in

increased submissions to the GSEs’ AUSs relative to measures of total loan applications or total

GSE purchases (under an assumption these are unaffected by the Rule). Further, any such

increases should be strongest for loans which may be more difficult to underwrite under the

General QM or ATR requirements.

Aggregating the submission and purchase data, Figure 63 shows the ratio of loans purchased by

the GSEs to eligible submissions to the GSEs from 2013 Q1 through 2016 Q4 by loan amount

bin, where purchases are shifted two months earlier (i.e., 2013 acquisitions for June, July, and

August are plotted in line with 2013 Q2 submissions to account for the lag between submissions

and purchases).

327

If lenders responded to the Rule by submitting additional loans to the AUSs

without increasing their sales to the GSEs, this ratio would be expected to fall. In contrast, the

figure shows that while the ratio of purchases to eligible submissions varies with loan amount

and fluctuates from quarter-to-quarter, the aggregate level remained fairly stable over the

period observed.

327

Indiv idual submissions and purchases cannot be observed or linked in the data, and it is common for loans to be

pu r chased several months after they were submitted. For this reason, the figures in this chapter shift acquisitions

tw o months earlier, to better align the submissions to their eventual purchases. Since only the submission dates are

available for submissions, and only acquisition dates are available for purchased loans, the ratio of submissions to

pu r chases w ithin a given qu arter will reflect som e spillovers of su bmissions and purchases from the prior and

follow ing quarters.

200 BUREAU OF CONSUMER FINANCIAL PROT ECTION

FIGURE 63: RA TIO OF A PPROV ED/ EL IGIBLE GSE SUBMISSIONS TO GSE PURCHASES, BY LOA N A MOUNT,

2013 TO 2016

While Figure 63 does not provide evidence of any aggregate shifts in the use of the GSEs’ AUSs,

lenders’ incentive to respond may be strongest for loans which are either definitively not

General QM (e.g., DTI over 43 percent) or those loans for which Appendix Q underwriting

requirements may be most difficult (e.g., self-employed borrowers). In qualitative responses to

the Bureau’s Lender Survey, underwriting for self-employed borrowers was one of the most

frequently reported sources of difficulty in originating mortgages using Appendix Q, and 61

respondents reported making changes to income documentation requirements for self-

employed borrowers. The Bureau also received numerous comments stating that Appendix Q

was ambiguous regarding how to account for particular sources of income and recurring

expenses in calculating DTI. These examples typically involve borrowers who are self-employed,

have irregular income, or wanted to use asset depletion as income.

328

When considering these

328

Com m enters also described the documentation and certain other requirem ents a s ambiguous or ov erly restrictive.

See, e.g., Com ment l etter from Structured Fin. In dus. Grp. (July 3 1, 2017); Com ment letter from T eacher’s In s. and

201 BUREAU OF CONSUMER FINANCIAL PROT ECTION

effects, it is important to keep in mind the results of Section 5.3.6 that imply that an adverse

differential effect on the approval rate of self-employed applicants in response to the Rule is

present but limited.

FIGURE 64: RA TIO OF A PPROV ED/ EL IGIBLE GSE SUBMISSIONS TO GSE PURCHASES, SPLIT BY DTI BINS,

LOANS UNDER $417,000

Figure 64 shows the ratio of eligible submissions to GSE purchases for loans split between those

with DTIs above 43 percent versus those at or below 43 percent, while Figure 65 shows the same

ratio for self-employed borrowers relative to those who are not self-employed. In Figure 64, the

ratios for DTIs exceeding 43 percent increase relative to those below 43 percent over time.

Figure 65 shows more limited differences between ratios for self-employed and not self-

employed borrowers, though with a relative increase in the ratio for self-employed borrowers

potentially emerging by the end of 2016. With the caveat that these patterns likely reflect a mix

Annuity Ass’n, (July 31, 2017); Comment letter from JPMorgan Chase Bank, (July 31, 2017) (See Appendix B for

fu rther details.).

202 BUREAU OF CONSUMER FINANCIAL PROT ECTION

of market trends, the findings are consistent with somewhat higher use of the GSEs’ AUSs for

loans which do not fit within (or are more difficult to document within) the General QM

underwriting standards.

FIGURE 65: RA TIO OF A PPROV ED/ EL IGIBLE GSE SUBMISSIONS TO GSE PURCHASES, SPLIT BY SELF-

EMPL OY ME NT INCOM E, LOANS UNDER $417,000

Finally, given the possibility of changes in the propensity of lenders to sell loans to the GSEs at

the time the Rule was implemented (which would affect the denominator of the ratios in the

previous figures), Figure 66 assesses the trend in eligible GSE AUS submissions relative to the

number of approved conventional, conforming applications in HMDA, with both samples

restricted to purchase loans at or below $417,000.

329

By comparing eligible submission to all

HMDA-reported, conventional conforming loans, this measure captures both changes in the

propensity to submit loans to the GSEs’ AUS, as well as any market shifts towards (or away

329

A comparable com parison for submissions with DTI over 43 percent or self-em ployed b orrowers similar t o

Fi g ures 64 and 65 is not possible for these groups, a s they are not distinguished in the HMDA data.

203 BUREAU OF CONSUMER FINANCIAL PROT ECTION

from) loan products typically sold to the GSEs. As with the preceding figures, these ratios can

exceed one, as borrowers may shop between multiple loans before (or without) originating a

loan.

330

The data show a relatively steady pattern of submissions to approved applications from

2013 to 2014, followed by an increase in submissions to approved applications in the years that

follow. This potential longer-run market pattern of increased overall submissions to the GSEs’

AUSs, coupled with the persistent share of eligible loans actually sold to the GSEs, suggests that

lenders generally have not decreased their use of the Temporary GSE QM in the non-jumbo

conventional market segment in the years following the implementation of the Rule. Rather, the

evidence as a whole suggests that lenders may be increasingly taking advantage of the provisions

in certain market segments even with respect to loans that are not sold to the GSEs.

FIGURE 66: RATIO OF GSE APPOVED/ELIGIBLE AUS SUBMISSIONS TO A PPROV ED HMDA CONV ENTIO NA L

APPLICATIONS, BY LOAN AMOUNT, 2013 TO 2016

330

The submissions data have been de-duplicated, m eaning that if m ultiple submissions were received for a g iven

borrower attempting to take out a single loan, only the last submission is kept in the data.

204 BUREAU OF CONSUMER FINANCIAL PROT ECTION

6.4 Meeting the goals of the QM

requirements

6.4.1 Access to responsible credit that consumers have the

ability to repay

Although a robust market outside of the qualified mortgage space has not emerged, the

mortgage market has successfully maintained fairly broad credit access, including maintaining

or exceeding the preexisting 18 to 22 percent range of originations above 43 percent DT I after

the implementation of the Rule.

331

Other QM provisions, including those that allowed more

flexibility for portfolio originations by small creditors, may also have supported this market

stability. The market as a whole has experienced minimal disruption, as evidenced in Chapter

3’s market overview. Chapters 4, 5, and 7 present evidence on potential access to credit issues

affecting narrower segments of the market.

6.4.2 Providing a clear QM framework

The Bureau adopted a specific debt-to-income ratio threshold in General QM because it viewed

the approach as providing a clear, bright line criterion for a qualified mortgage that ensured that

lenders in fact evaluate consumers’ ability to repay qualified mortgages while also providing

certainty for lenders, assignees, and investors in the secondary market.

332

However, as

documented in the multiple comment letters and survey responses received by the Bureau citing

specific challenges and seeking additional clarity regarding Appendix Q’s requirements, there is

sometimes no bright line criterion a lender can use to assess whether the amounts used for

monthly income and for monthly debt are in compliance with Appendix Q.

333

To be sure, lenders may also experience difficulties when they attempt to interpret the method

of calculating income and debt in compliance with GSE standards. However, the industry has

had more experience with these GSE standards and more tools available for the resolution of

interpretive uncertainty. Thus the use of the GSEs adds compliance certainty for loans that

could also satisfy the General QM test, and for high DTI loans the Temporary GSE QM provides

331

7 8 Fed. Reg. 6408, 6527, 6569 (Jan. 30, 2013).

332

Id. a t 6527.

333

See A ppendix B.

205 BUREAU OF CONSUMER FINANCIAL PROT ECTION

the only means of compliance certainty. These factors may have contributed to investors’

persistent preference for GSE-guaranteed loans as well as to creditors increased use of GSE

underwriting for certain categories of loans, their reluctance to originate non-QM loans, and

their shift away from high-DTI loans in the non-GSE eligible space.

6.4.3 Supporting the emergence of a non-QM market

Chapter 5.1 discusses the origination of high-DT I (non-GSE eligible) loans. Given available data,

this is the only sizeable segment of non-QM loans that it is possible to identify with any

certainty, and even this segment represents only 1 to 2 percent of the market. The analysis finds

that among the lenders supplying the Application Data, around two thirds of the purchase loans

in this specific segment were eliminated following the implementation of the Rule. Further, as

discussed above, a vibrant primary and secondary market for non-QM loans was a goal of the

Rule, but does not yet exist.

Overall, it is possible that the breadth of the T emporary GSE QM category in itself is inhibiting

the growth of the non-QM market. However it is also possible that this market might not exist

even with a narrower Temporary GSE QM category and narrower Federal Agency QM category,

if borrowers were not willing to pay the price required for the potential litigation risk associated

with non-QM loans, or lenders were unwilling or lacked the funding to make such loans.

Commenters on the RFI stated that creditors and investors are uncertain as to how individual

judges might interpret the standards that exist in the ATR regulations, and that there is little

litigation experience with which to guide the identification of legal risks. As a result, they claim

that it is not possible to measure this risk and consider whether it leads to pricing above the

amount that prospective borrowers for whom a non-QM loan is the only option would pay.

6.4.4 Innovation

As discussed previously, when establishing the General QM presumption of ATR compliance

standard, the Bureau sought to strike a balance between appropriate lending and innovation.

The Bureau expected that private mortgage market participants would innovate at least in the

non-QM space. Innovation could also occur in the General QM space with respect to

underwriting approaches that would be consistent with the General QM criteria.

The original proposal of the Rule contained a comment that indicated that lenders could look to

widely accepted governmental or nongovernmental underwriting standards to evaluate a

206 BUREAU OF CONSUMER FINANCIAL PROT ECTION

consumers ability to repay.

334

The proposed comment was not adopted because the Bureau

concluded that an emphasis on widely accepted underwriting standards could distract lenders

from ability-to-repay determinations that are reasonable and in good faith, hinder lenders’

ability to respond to changing market and economic conditions, and stifle market growth and

positive innovation.

335

In the final rule, the Bureau emphasized that lenders were permitted to

develop and apply their own proprietary underwriting standards and to make changes to those

standards over time in response to empirical information and changing economic and other

conditions.

336

Nevertheless, the vast majority of loans are originated as QMs: as discussed in

Section 6.1, three quarters of current originations are GSE and Federal Agency loans.

Innovation is occurring in the mortgage market under the umbrella of the Temporary GSE QM.

For example, the GSEs are providing pre-closing assurances of purchase that rely upon

automated verifications and validations.

337

Fannie Mae has a program that “allows lenders to

validate a borrower’s income, assets, and employment with a single report from a single

approved vendor that the lender chooses.”

338

The Temporary GSE QM does not require that

these new methods of income verification and calculation be compliant with Appendix Q, and it

would be difficult for a creditor to determine if they were, as much of the underlying

requirements and technical specifications are maintained under proprietary confidentiality

between the vendors and the GSEs. Similarly, while a private investor or lender could seek to

originate and privately securitize mortgage loans using these same innovations, the complexity

of the GSE-approved methods, at least in some cases, and the fact that these methods are

private, would make it difficult for an entity to know if the loan was in fact eligible for purchase

by the GSEs. These constraints may explain, at least in part, why innovation in one segment of

the market does not appear to have spurred growth and innovation in others.

334

Pr oposed Supplement I to Pa rt 1026—Official Interpretations, Paragraph 43(c)–1, at 76 Fed. Reg. 27390, 27492

(May 11, 2011).

335

7 8 Fed. Reg. 6408, 6461 (Jan. 30. 2013).

336

Id.

337

Mic h al Tucker, Fannie Mae, Freddie Mac Tout New Programs to Boost Access to Credit, Newslink (Oct. 25,

2 01 6) , available at https://www.mba.org/servicing-newslink/2016 /october/servicing-newslink-10-25-16/news-

and-trends/fannie-m ae-freddie-mac-tout-new-programs-to-boost-access-to-credit.

338

Kel sey Ramirez, Fannie Ma e Rev eals Ma jor Upgrade t o its Day 1 Certainty Pr oduct: Here are the C om panies

In v olved in the Pilot Program, HousingWire (Oct. 23, 2017), available at

h t tps://www.housingwire.com /articles/41638-fannie -m ae-reveals-m ajor-upgrade-to-its-day-1 -cer ta in ty -product.

207 BUREAU OF CONSUMER FINANCIAL PROT ECTION

7. Analysis of the small creditor

QM category

This chapter considers the Rule’s Small Creditor QM category and the associated asset and

origination requirements and explores whether these thresholds are influencing lender

behavior. This section also analyzes to what extent, if any, small creditors moved in and out of

the Small Creditor definitions and also looks at their lending activity in rural and underserved

counties.

Main findings in this chapter include the following:

• There was no bunching of small creditors just below the loan thresholds defining a small

creditor as most small creditors fell well below the 500-loan threshold that was in effect in

2014 and 2015 and the amended 2,000 loan threshold that took effect in March 2016.

• The geographic market coverage of small creditors increased substantially with the new

2,000 loan threshold in the March 2016 amendment to the Rule. The number of counties

served and the market share held by small creditors within individual counties increased

in 2016 compared to 2014 allowing for more lenders to qualify as small creditors and

increasing access to credit for borrowers in rural and underserved areas who have DTIs

above 43 percent.

• From 2012 to 2015, the share of depository institutions that met the definition of small

creditor ranged from 81 percent to 86 percent although the share of loans made by these

creditors ranged from 12 percent to 16 percent. In 2016, when a broader definition of

small creditor took effect, the share of depositories that were small creditors increased

to 89 percent and their share of loans increased to 24 percent.

• There are systematic differences in the loans made by small and non-small depository

institutions. Small creditors hold a larger share of their originations in portfolio, although

there was a noticeable decline in the share of portfolio loans made by small creditors in

2016 which coincided with the change in the definition of small creditor.

208 BUREAU OF CONSUMER FINANCIAL PROT ECTION

Similarly, a larger share of small creditor mortgages are made in rural counties or for

manufactured housing.

• Results from the CSBS survey show that small creditors declined a smaller percentage of

applications than larger creditors. To the extent small creditors declined applications,

these creditors were less likely to attribute their denial to the requirements of the Rule

than larger creditors.

7.1 Background

The Rule contains provisions directed at smaller lending institutions, including provisions that

are meant to preserve access to mortgage credit in rural and underserved areas. Lenders who

meet certain asset and origination criteria are considered to be “small creditors” and can

originate loans that are classified as Qualified Mortgages (QM) even if they contain

characteristics or are underwritten in a manner that would otherwise render them non-QM

loans.

339

In addition to loans that meet standard QM definitions, small creditors can originate Small

Creditor QM loans and small creditors who operate in rural and underserved areas can originate

Small Creditor QM Balloon loans. While these loans must meet many of the standard QM

criteria, they have a higher threshold to be considered higher priced for purposes of determining

whether they qualify for the QM safe harbor.

340

They are also not subject to the 43 percent DTI

limit nor are they required to use Appendix Q to calculate debt and income.

341

Small creditors

who operate in rural or undeserved areas can originate certain loans with a balloon payment

that are still considered to be QM provided they meet other QM criteria.

342

Finally, Small

339

Th e focus of this chapter is on “small creditors” as defined under 1026.35(b)(2)(iii)(B) and (C) and 12 C.F.R. §

1 026.43(e)(5).This chapter also discusses “rural small creditors,” which are small creditors that operate in a rural or

underserved area and can make Small Creditor Ba lloon QMs. See 12 C.F.R. § 1 026.35(b)(2)(iii)(A)–(C) and

1 026.43(f).

340

QM m ortgages are g enerally c onsidered to be higher priced if they have an A PR that exceeds the applicable APO R

by at least 1.5 percentage points for first-lien loans and at least 3.5 percentage points for subordinate-lien loans. In

con trast, Sm all Creditor QM loans, including balloons, are only c onsidered higher priced i f the APR exceeds A POR

by at least 3.5 percentage points for either a first- or subordinate-lien loan. 12 C.F.R. § 1 026.43(b)(4). QMs which

a r e higher priced enjoy on ly a rebuttable presumption of c om pliance with the ATR requirem ents, w hereas Q Ms

w h ich are n ot higher pric ed enjoy a safe harbor.

341

1 2 C .F.R. § 1 026.43(e)(5)(i)(A).

342

For example, Small Creditor Balloon QM loans may n ot have negative-amortization or interest-only features and

m ust comply with the points and fees limits to which other QM loans are subject. In a ddition, Small Creditor

Balloon QM loans m ust carry a fixed interest rate, payments other than the balloon m ust fully amortize the loan

209 BUREAU OF CONSUMER FINANCIAL PROT ECTION

Creditor QM loans and Small Creditor QM Balloon loans must be held in portfolio for three

years.

343

This section does not analyze the Small Creditor Balloon QM category specifically but

does look into rural lending by small creditor status.

344

As noted above, lenders must be within certain asset and origination thresholds to be considered

small creditors.

345

These thresholds have been modified over time as shown in Table 36. For

example, when the Rule was first implemented in 2014, small creditors were defined as lenders

that originated 500 or fewer loans (including loans originated by any affiliates) and had assets of

no more than $2 billion (not including the assets of any affiliates) in the previous calendar

year.

346

The asset threshold is adjusted annually for inflation.

The small creditor criteria were amended along three dimensions with the changes becoming

effective in 2016.

347

First, lenders who originated up to 2,000 loans could be considered small

creditors. Second, any loans held in the lender’s portfolio were exempt from the 2,000 loan

limit. Third, the asset threshold was required to include the assets of any affiliates. Again, these

changes became effective in 2016 with a grace period for small creditors who may move out of

small creditor status due to the change.

TABL E 36 : SMA LL CREDITO R QM CA TEGO RY REQUIREMENTS

Year

Origination

Threshold

Include

Affiliate

Originations?

Exclude

Portfolio

Loans?

Asse t

Threshold

Include

Affiliate

Asse ts?

2014

500

Yes

$2 billion

2015

500

Yes

$2.06 billion

2016

2,000

Yes

$2.052 billion

Yes

This chapter makes use of both administrative and survey data to report trends in lending

among institutions classified as small creditors. The first section of the analysis uses data

ov er 30 years or less, and the loan term must be at least five years. The lender m ust determine the borrower’s a bility

t o m ake periodic payments other t han the balloon and verify income and assets.

343

1 2 C .F.R. § 1 026.43(e)(5)(ii); 1 2 C.F.R. § 1 026.43(f)(2).

344

See Section 8.2 for a discussion on balloon loans based on the results of the Bureau’s lender survey.

345

1 2 C .F.R. § 1 026.43(e)(5)(i)(D) (cross-referencing 12 C.F.R. § 1 026.35(b)(2)(iii)(B)–(C)).

346

1 2 C.F.R. § 1 026.35(b)(2)(iii)(B) and (C), January 1, 2015 edition.

347

1 2 C.F.R. § 1 026.35(b)(2)(iii)(B)–(C) (as amended at 81 Fed. Reg. 16074 (Mar. 25, 2016)).

210 BUREAU OF CONSUMER FINANCIAL PROT ECTION

reported under HMDA to estimate the number of small creditors among HMDA reporting

institutions. The HMDA data are also used to examine lending behavior before and after the

Rule. The second section makes use of the CSBS survey data as described in Chapter 1 to

understand how small creditors are engaging in the origination of qualified mortgages.

7.2 Analysis using HMDA data

The first section of the HMDA analysis describes the data and methods used to estimate the

number of small creditors who report HMDA data and to analyze the impact, if any, the Rule

had on these lenders and the borrowers they serve. The second section focuses on estimating the

share of mortgage lenders that may meet the small creditor criteria before and after the Rule’s

implementation in 2014 using administrative data.

348

The prevalence and type of mortgage

lenders that met the small creditor requirements over this time period and the extent to which

mortgage lenders moved between size groups are also reported.

The next section examines whether the distribution of lenders by the number of covered loans

they originate changed over time in response to the Rule and provides a summary of findings on

mortgage origination activity for lenders below and above the origination threshold.

349

This

section also shows portfolio lending activity and, for loans sold in the secondary market, the

typical type of purchaser broken down by small creditor status. The last section provides an

analysis on the role that small creditors play in rural counties and in other housing markets,

such as in manufactured home lending.

Overall, small creditors account for a large portion of mortgage lenders and a small but growing

share of loans. T he data provide evidence that the share of small lenders has been growing over

the period of analysis used in this section. The analysis shows that most lenders in the data that

may meet the small creditor criteria are well within the origination and asset thresholds.

348

Resu lts r eported for sm all creditors in y ears b efore t he ATR/QM Rule took effect use the small creditor thresholds

in effect in 2014 to classify lenders as sm all creditors. A s noted in the next section, assets and originations are used

to determine which lenders in the data would qualify as “small creditors.” However, “small creditors” as defined in

this analysis may differ from the lender’s a ctual status.

349

Th e analyses in this chapter focus on HMDA data for first-lien purchase m ortgages on 1 -4 unit single-f am i ly

properties when observing origination trends.

211 BUREAU OF CONSUMER FINANCIAL PROT ECTION

7.2.1 Data and methods

This analysis uses the non-public HMDA data described in Chapter 1.

350

In the context of this

chapter, any reference to “lender(s)” or “creditor(s)” only refers to HMDA reporting lenders.

Estimates of small lenders in this analysis are not the complete universe of mortgage

originators. The Bureau estimates that there are over 4,000 depository institutions which

originate mortgages but are not HMDA reporters. Most, if not all, of HMDA non-reporters

would qualify as small creditors due to their small size.

351

To estimate asset levels, data from the Federal Financial Institutions Examinations Council

(FFIEC)

352

and National Credit Union Administration (NCUA)

353

are used. Asset data on banks

and credit unions are matched to loan counts for HMDA reporting institutions. Using the

matched data, small creditor status is defined for each year between 2012 and 2016 based on the

requirements in Table 36.

354

Any affiliates of a lender are identified as such in HMDA and are

included in the calculation of a lender’s prior year originations that go into the small creditor

determination for all years between 2012 and 2016. The origination count is determined based

on the institution’s prior year covered mortgage transactions in HMDA data that are subject to

the ATR/QM Rule. Covered transactions in HMDA are identified as first-lien purchase and

refinance originations on owner-occupied site-built and manufactured housing properties.

Current year asset holdings are determined based on total assets from the last quarter of the

350

HMDA r equires cov e r ed d e posi tor y a n d n on -depository institutions to collect and publicly disclose information

about applications and originations of mortgage loans used to purchase a home, refinance an existing mortgage loan,

or for h om e im prov ement purposes. For m ore i nformation on HMDA data and reporting. See Fed. Fin. Insts.

Ex a m ination Council, A Guide to HMDA Reporting: Getting it Right!, available at

h t tps://www.ffiec.gov/hmda/guide.htm (effective Jan. 1 , 2018).

351

In a separate analysis, the Bureau estimated the universe of mortgage lenders using HMDA and Call Report data.

Th e analysis estimates that the universe of mortgage lenders in 2016 was 11,656 that includes HMDA reporters and

len ders that did not meet the reporting requirem ents for H MDA due t o their size b ut still originated mortgage loans.

Of t h is estimate, t here were roughly 4 ,892 m ortgage lenders who did not report H MDA data and 9,106 lenders

con si dered to be sm all based on the ATR/QM definitions, suggesting that the analysis of HMDA data in this report

is limited to about 56 percent of a ll estimated small creditors.

352

Ev ery national bank, st ate m ember bank, and i nsured nonmember b ank is r equired by its primary federal

r egulator to file a Call Report a s of the c lose of business on the last day of ea ch calendar quarter. The specific

r eporting requirements depend upon the size of t he b ank a nd w hether it has any foreign offices.

353

The NCUA Call Report includes data on all federally insured credit unions. These credit unions make up 98

percent of all credit unions and 99 percent of all insured deposits in credit unions. See U.S. Gov ’t Accountability

Office, GAO–17–259, Priv ate Deposit Insurance: Credit Unions Largely Co mplied with Dis closure Rules But Rules

Should b e Clarified, at 5, Figure 1 (2017), available at https://www.gao.gov /assets/690/683779.pdf.

354

A lthough the Rul e was not im plem ented until January 1 0, 2014, lenders are r etroactively identified as sm all

cr editors in 2012 and 2013 if they m et the requirem ents in the respective previous years to analyze h ow the n ew

Sm all C reditor QM status may have c hanged lenders’ behaviors after implementation.

212 BUREAU OF CONSUMER FINANCIAL PROT ECTION

preceding year using the matched FFIEC and NCUA data. Assets of a lender’s affiliates in 2016

are combined when determining small creditor status in that year. All prior year’s assets from

affiliates are not included in a lender’s asset holdings.

355

Finally, the primary sample consists of

depository institutions (i.e., banks and credit unions).

356

For purpose of this analysis, banks and

credit unions are not broken out separately.

357 ,

358

Estimates of small creditors within this analysis do not represent the universe of small creditors

due to data limitations associated with how HMDA data are collected and reported. Creditors

are required to report under HMDA only if they have assets above a specified threshold and a

home or branch office within a metropolitan area. These non-reporters are excluded from the

analysis that follows. Other limitations may lead to underestimating

359

or overestimating

360

when determining a lender’s small creditor status in HMDA although the aforementioned

limitation results in an overall underestimation of small creditors.

7.2.2 Distribution of mortgage lenders by size over time

Within the HMDA data, most mortgage lenders are small institutions, but the share of

origination volume accounted for by these institutions is small. The majority of mortgage loans

are originated by the relatively few large lenders.

Table 37 reports estimates of small creditor status. The table indicates that at least a large

majority of lenders in the sample likely met the small creditor criteria during the 2012-2016

355

Sm all creditors in 2016 are estimated in two ways–(1) using the am ended 2 016 thresholds, denoted as 2016b in

t h is chapter and; (2) u sing the 2015 t hresholds to estimate t he number of sm all c reditors in 2016 in the absence of

t h e 2016 threshold am endments denoted as 2016a. If 2 016a or 2016b is n ot specified in a table or figure, the 2016

a m endments were u sed.

356

Non -depositories (e.g., independent brokers and a ffiliated lenders) are excluded from the analysis because they

are not expected to benefit from small creditor status. Small Creditor QM loans require the lender to h old t he l oan

on por tfolio for a three year period. Non-depositories do not h old loans on portfolio and, therefore, are not expected

to or iginate Small Creditor QM loans.

357

For summary statistics on market share, average costs and profits ov er t ime for credit unions, see

https ://www.ncua.gov/analysis/Pages/industry/fact-sheets.aspx.

358

For a ddi tional analysis on t he im pacts the Ru le m ay have had on credit unions spec ifically, see

https://www.nafcu.org/research/reportoncreditunions-archive; see also A ppendix B.

359

Sm all creditor status may be underestim ated is due t o dw elling si ze classifications. Single-family housing units are

defined as one to four unit properties in the HMDA data. The small creditor exemption only applies to single-unit

h om es. This difference w ould increa se the num ber of lender s that fall abov e the origination threshold and, therefore,

r educes the number of sm all credi tors ob served in the data.

360

Ov erestimation may oc cur in 2016 a s som e affiliates of HMDA reporters m ay not sh ow up in the data and thus

cannot be included in the overall origination amounts in that year.

213 BUREAU OF CONSUMER FINANCIAL PROT ECTION

time period. In 2016, 89 percent of HMDA depository institutions were small creditors but

made up only 24 percent of the total count of mortgage originations in that year. The share of

depository institutions in the data that met the small creditor criteria ranged from 81 percent

to 86 percent between 2012 and 2015, although most of the change in share was attributable to a

decrease in the total number of lenders reporting HMDA data. Notably, there was an increase to

89 percent in 2016 due to the origination threshold change that took effect in 2016 (see Table

36). Before 2016, the share of loans originated by small creditors ranged from 12 to 14 percent

but then rose to 24 percent in 2016. As described earlier, the origination threshold increased

from 500 to 2,000 in 2016, with loans held in portfolio no longer counting towards this limit.

On the other hand, all affiliates’ assets were now taken into account for the asset threshold,

which stood at just over $2 billion in both 2015 and 2016. Without the 2016 amendment

increasing the origination threshold, the number of small creditors would have decreased by 267

and the number of non-small creditors would have risen by 83 lenders. The absolute number of

small creditors also increased from 2014 to 2016 while the overall number of mortgage lenders

in the data declined.

361

TABLE 37: ESTIMA TED NUMBER OF HMDA-REPORTING MORTGAGE LENDERS WHO MET THE SMALL

CREDITO R CRITERIA , 2012-2016

2012

2013

2014

2015

2016a

2016b

Total Non-Small Lenders

(number)

976 1,152 1,067 806 889 619

Non-Small Lenders (share of

loans)

86% 88% 86% 84% 86% 76%

Total Small Lenders (number) 5,408 5,066 5,047 5,128 4,860 5,130

Small Lenders (share of loans) 14% 12% 14% 16% 14% 24%

Total Lenders

6,384

6,218

6,114

5,934

5,749

Note: 2016a uses the 2015 amendments to while 2016b uses the 2016 amendments.

7.2.3 Small creditor originations

Next, the analysis reports the distribution of the number of covered originations across lenders.

Shifts in the distribution of loan originations -after the Rule may provide evidence that lending

361

Com m enters prov ided CUNA Mutual Group’s Credit Union Trends Report (2017) for ov erall trends in credit

u n ions, available at h ttps://w ww .cunamutual.com /resource-library/insights/industry/credit-union-trends-

report?shortURL=https://www.cunamutual.com/resource-library/publications/credit-u ni on -trends-report. See

also A ppendix B.

214 BUREAU OF CONSUMER FINANCIAL PROT ECTION

institutions are sensitive to the originations threshold that determines small creditor status. For

example, changes in origination behavior after the Rule might be observed through bunching in

the distribution of originations suggesting that lenders changed their lending activity to fall

below the origination threshold. Alternatively, the data may show evidence of a concentration of

institutions just below the covered origination threshold.

362

Figures 67 and 68 are overlaid origination distributions that show that the majority of lenders in

the data originated fewer than 500 loans annually during the 2013-2016 time periods. A smaller

number of lenders in the data originated 500-1,000 loans, and originating over 1,000 loans

annually is far less common.

363

There is a dip in the number of lenders in the data who

originated fewer than 100 loans in 2015 compared to 2016 (Figure 64). Overall, when comparing

2013 to 2014 and 2015 to 2016, the distributions have little variation over time.

362

In a sim ilar analysis (not shown here), no ev idence was found t o support lender sensitivity to b oth the originations

a n d asset threshold. Few lenders were within ± 2 0 percent of t he or igination t hreshold and ± 25 percent of t he a sset

t h reshold.

363

The data for Figures 67 and 68 are t op-coded at a value of 1 ,000 loans originated.

215 BUREAU OF CONSUMER FINANCIAL PROT ECTION

FIGURE 67: DISTRIB UTIO N OF THE NUMB ER OF ORIGINA T IONS PER MORTGA G E LENDER, 2013 AND 2014

216 BUREAU OF CONSUMER FINANCIAL PROT ECTION

FIGURE 68: DISTRIB UTIO N OF THE NUMB ER OF ORIGINA T IONS PER MORTGA G E LENDER, 2015 AND 2016

Next the analysis provides time series evidence on originations growth for depository

institutions that are just below and above the origination threshold. If lenders are reducing

originations in order to stay or become a small creditor, there would be evidence of bunching

just around the thresholds. To analysis this, lenders are placed into loan groups based on their

originations that were used to determine small creditor status according to the Rule (see Section

7.2.1). The loan groups are as follows for covered loans: 1) up to 250 covered loans; 2) over 250

and up to 500; 3) over 500 but up to 750; and 4) over 750 covered loans. Figure 69 reports the

growth rates in mortgage originations for lenders by these loan groups. Overall, lenders in all

four groups exhibit a similar trend in origination growth. Origination growth decreases for all

groups between the 2013 and 2014 periods. Growth picked up after 2014 but then declined

slightly in 2016. As there are no clear differences in the growth rates for the threshold groups,

this implies that the thresholds did not impact the lending behavior of depository institutions.

217 BUREAU OF CONSUMER FINANCIAL PROT ECTION

FIGURE 69: TIME SERI ES GROWT H IN ORIGINA T IO NS OF LENDERS A BOV E A ND BELOW THE

ORIGINA TIO N THRES HOL D BY COV ERED LOA N GROUP

Finally, the analysis on small creditor originations considers the relationship between small

creditor status and loan purchaser type.

364

Figures 70 and 71 show the share of originations in

each year by purchaser type for non-small and small creditors respectively. Compared to non-

small creditors, small creditors held a higher share of their originations in portfolio between

2012 and 2016.

365

The share in portfolio for small creditors inc reased up to 2014 to about 62

percent but then experienced a decline to roughly 44 percent by 2016. A potential reason for this

364

Because portfolio loans are recorded in the HMDA data only if the loans are originated and sold in the same

ca lendar year, l oans originated tow ard t he end of t he y ear are less likely t o be r eported as sol d. For that r eason,

statistics on portfolio loan are computed using only loans originated during the first three quarters of the year.

How ev er, when determining small creditor status in 2016, portfolio loans are used for the entirety of the year as this

w ou ld be the number lenders use w hen determining their sm all creditor st atus.

365

Du e to data limitations, portfolio lending can only be observed at or igination. The length of time the or iginated

loa n is h eld in portfolio c annot be ob served. The numbers provide an u pper b ound on the number of S m all Creditor

QM loa ns.

218 BUREAU OF CONSUMER FINANCIAL PROT ECTION

decline may be that the higher origination threshold of 2016 brought in lenders who did not

hold a substantial share of originations in portfolio thus decreasing the overall share for small

creditors in 2016. This is confirmed if the 500 origination threshold is applied in 2016 as the

composition of loan purchaser type for small creditors remains largely unchanged compared to

2015. Non-small creditors sell a higher share of originations to affiliate lenders compared to

small creditors in the data while small creditors sell a higher share of loans to non-affiliates.

FIGURE 70: SHA RE OF NON-SMALL ORIGINA TIO NS BY PURCHA S ER TY PE AND YEAR

219 BUREAU OF CONSUMER FINANCIAL PROT ECTION

FIGURE 71: SHA RE OF SMA LL ORIGINATIONS BY PURCHA S ER TY PE AND YEAR

The differences across small and non-small institutions show that portfolio originations are a

much larger proportion of mortgage originations for small institutions compared to non-small

institutions. As the Rule generally requires Small Creditor QMs to be held on portfolio for three

years after consummation, this may suggest that small creditors are utilizing this category of

QM.

7.2.4 Small creditors in rural and manufactured housing

markets

Small creditors are more likely to operate in rural areas compared to larger creditors. This

section provides insight into the geographic distribution of small creditors, along with their role

in providing access to credit in rural and manufactured housing markets. As previously

discussed, this analysis is missing data from lenders who do not have a branch or office in a

metropolitan area, and is therefore likely missing a large number of rural lenders.

Figures 72 and 73 show the geographic distribution of small lenders and their market share for

2014 and 2016, respectively. Market share is calculated by looking at the total originations of

220 BUREAU OF CONSUMER FINANCIAL PROT ECTION

depository institutions in a given county and then identifying the share that are accounted for by

small creditors. In 2014, small lenders originated loans throughout much of the United States.

In rural areas, small lenders often carried a large market share. This was especially true in much

of the Southern, Midwestern, and Mountain states.

Comparing Figures 72 and 73, market coverage among small creditors increased substantially

between 2014 and 2016. The figures show both increases in the number of counties that small

creditors serve and the market share held by small creditors within individual counties. T he

number of small creditors increasing between this time (see Table 37) and the 2016 amendment

may explain this increase of coverage.

FIGURE 72: COUNTY -L EV EL MA RKET SHA RE OF SMA LL CREDITORS IN 2014

221 BUREAU OF CONSUMER FINANCIAL PROT ECTION

FIGURE 73: COUNTY -L EV EL MA RKET SHA RE OF SMA LL CREDITORS IN 2016

Table 38 reports mortgage originations by small and non-small creditors in rural counties.

366

Among small creditors, the share of total originations occurring in rural areas is much larger

than for non-small creditors. This appears to be consistent with the higher likelihood that small

creditors operate only or predominantly in rural or underserved areas compared to non-small

creditors. The 2016a columns suggests that without the 2016 threshold amendments, the rural

share of small creditor originations would have stayed the same instead of decreasing to 21

366

T h e Bureau publishes a yearly list of rural and underserved counties that are exempt from certain regulatory

requ irements of the Truth in Lending A ct. Bureau of Con sumer Fin. Prot., Rural and Underserved Co unties Lis t,

available at https://www.consumerfinance.gov /policy-com pliance/guidance/im plem entation-guidance/rural-and-

underserved-coun ti es-li st/ (last v isited Dec. 31, 2018) (for the current and previous y ear’s lists).

222 BUREAU OF CONSUMER FINANCIAL PROT ECTION

percent in 2016 with the amendments. The 2016 amendments did however increase the share of

small creditors operating in rural areas due to being more inclusive of larger lenders. The

amendments in 20

TABL E 38 : ESTIMA TED NUMBER OF HMDA -REPORTING MORTGA G E LENDERS WITH ORIGINA TIONS IN

RURA L COUNTIES

2012 2013 2014 2015 2016a 2016b

Total Non-Small Lenders 976 1,152 1,067 806 889 619

Share of Non-small Originations 26% 20% 11% 11% 11% 10%

Total Small Lenders 5,408 5,066 5,047 5,128 4,860 5,130

Share of Small Originations 44% 34% 24% 24% 24% 21%

Total 6,384 6,218 6,114 5,934 5,749 5,749

Table 39 reports manufactured housing mortgage originations by small and non-small creditors.

Manufactured housing loans make up a larger share of small lenders’ originations compared to

non-small lenders. Similar to rural loan originations, these patterns are consistent with small

creditors being more likely to provide access to mortgage credit for manufactured housing

compared to larger creditors although the share or manufactured originations that make up a

small creditor’s lending has been declining since 2012.

TABL E 39 : ESTIMA TED NUMBER OF HMDA -REPORTING MANUFA CTURED HOUSING MORTGAGE

LENDERS

2012 2013 2014 2015 2016a 2016b

Total Non-Small Lenders 976 1,152 1,067 806 889 619

Share of Non-small Originations 4% 4% 3% 2% 2% 2%

Total Small Lenders 5,408 5,066 5,047 5,128 4,860 5,130

Share of Small Originations 13% 12% 10% 9% 9% 7%

Total 6,384 6,218 6,114 5,934 5,749 5,749

223 BUREAU OF CONSUMER FINANCIAL PROT ECTION

7.3 Evidence from CSBS survey data

7.3.1 Overview of survey data

The section utilizes survey evidence from the Conference of State Bank Supervisors’ (CSBS) 2015

National Survey of Community Banks, an annual survey of community banks.

367

T he survey

provides additional insight into the small creditor exemption implemented under the Rule

because many of the survey respondents are small banks that are not required to report under

HMDA.

368

In total, 974 community banks responded to the CSBS survey.

369

Survey respondents

are also disproportionately rural institutions: over 65 percent of respondents reported that the

majority of their lending was in rural areas or did an equal amount of lending in urban and rural

settings.

The focus of the 2015 survey is mortgage originations that occurred in 2014, although some

survey questions ask about future lending expectations. The survey evidence provides a

snapshot of lending activity in the year after the Rule was implemented. One limitation of the

data is that no comparison to the pre-Rule period is available, as no information directly related

to the Rule was collected before 2015.

The CSBS survey provides information related to lender characteristics, including lenders’

primary lines of business, ownership structure, asset size, and market areas. The survey asks

several questions related to qualified mortgage lending, including questions concerning

qualified mortgage portfolio lending, mortgage application denial behavior related to the Rule,

and future plans for qualified mortgage lending. The reported asset size is used to determine a

lender’s small creditor status. Lenders who reported assets below $2 billion are considered to be

367

Mor e in formation and findings, such a s compliance costs on the annual survey c onducted by CSBS, see

h t tps://www.com munitybanking.org/~/m edia/files/c b21pub_2017_book_web.pdf.

368

A bout 30 percent of survey respondents that r esponded to a questi on a bout t heir H MDA r eporting status

in dicated that they were not HMDA reporters. Respondents t o this qu estion represented 16 percent of a ll survey

respondents in 2015.

369

Com m enters prov ided additional information on community banks and lending, see Appendix B. Data on the

profitability of community banks, available at h ttps://www.fdic.gov /bank/analytical/quarterly/2016-vol10-

4 /a rticle1 .pdf and

h t tps://www.fdic.gov /bank/analytical/quarterly/2016_v ol10_4/fdic_v10n4_3q1 6_quarterly.pdf; the decl ine in

community banks, available at

h t tps://www.com munitybanking.org/~/m edia/files/c b21pub_2017_book_web.pdf); and the health of community

b a n ks p ost-crisis relative to l arger b a nks, available at

h t tps://www.com munitybanking.org/~/m edia/files/com munitybanking/2 015/session3_paper4_bassett.pdf).

224 BUREAU OF CONSUMER FINANCIAL PROT ECTION

small in this dataset.

37 0

Based on this, number of small lenders were determined to be 677 and

the number of non-small was 30. The remaining 267 lenders were not included in the analysis

since as their asset size was unknown and therefore could not be identified as either small or

non-small.

There are a few limitations to the CSBS survey data. A limitation of these results is that they

cannot be compared to pre-Rule mortgage lending among small creditors and all other

creditors, as the survey was conducted only once, in 2017. Also, the CSBS survey is not

nationally representative and mostly includes smaller FDIC-insured institutions from an

unequal geographic distribution.

7.3.2 Analysis of CSBS survey data

This section discusses mortgage lending among small creditors responding to the CSBS survey,

comparing their behavior to that of non-small creditors. The analysis examines how denial rates,

portfolio lending, and non-QM lending vary across the two groups based on survey responses

from the CSBS data.

Figures 74 and 75 report differences in denial rates for small and non-small creditors in 2014.

Percentages on the vertical axis of the figures represent the share of lenders responding to each

possible response to a question by size. Creditors may have different rates at which mortgage

applications are denied based on their lending strategy and the pool of applications they receive,

among other factors.

Figure 74 examines the distribution of surveyed institutions by application denial rates. In total,

649 lenders responded to this question provided the share of denied applications in 2014. Of

this total, 623 were small creditors and 26 were non-small. T he figure shows that a about 27

percent of small creditor survey respondents denied between 0 and 10 percent of loans, whereas

application denial rates from non-small institutions peak at the 10 to 20 percent and 20 to 30

percent marks. There were some respondents who indicated that they did not deny any loans, all

of these were small creditors (roughly 5 percent of small creditor respondents). Among survey

respondents, then, small creditors generally denied a smaller share of applications relative to

non-small respondents.

37 0

The Bureau c onducted a m atch t o HMDA data and for the lenders in the C SBS surv ey data who matched t o H MDA,

a v ast majority a lso met the origination threshold to be considered small creditors.

225 BUREAU OF CONSUMER FINANCIAL PROT ECTION

FIGURE 74: SHA RE OF ALL A PPLICA T IO N S DENIED IN 2014, BY SIZE OF INSTITUT IONS

Figure 75 reports the share of denied applications that respondents stated would have been

approved in absence of the ability-to-repay standard under the Rule.

37 1

The total number of

lenders responding to this question was 530 with 21 being non-small and 509 small. The figure

shows that over 30 percent of small creditor respondents stated that they did not deny any

applications due to the ATR standard, while most of the non-small creditors responded that

between 0 and 10 percent of denied applications were denied because of the ATR rule. This may

indicate that small creditors were taking advantage of the small creditor exemption and

rejecting fewer applications due to the ATR standard than they would have otherwise.

37 1

The data reflect responses to the question, “Of the 1-4 family mortgage loan applications that you denied in 2014,

w h at percentage of t hem would you have a pprov ed if the Ability-to-Repay underwriting standard had not been in

pla ce?” For the purposes of t his qu estion, the Bu reau a ssumes that sm all c reditors interpreted the question to

sig n ify that the loans in qu estion would have violated som e portion of the ab ility-to-repay requirem ents. The

qu estion itself l inks to t he Bureau’s ov erview of the ATR r equirements for qualified mortgages.

226 BUREAU OF CONSUMER FINANCIAL PROT ECTION

FIGURE 75: SHA RE OF DENIED A PPLICA T IO NS THA T WOULD HA V E BEEN A PPROV ED WITHOUT THE A TR

STA NDA RD IN 2014, BY SIZ E OF INSTITUTIO NS

Finally, the CSBS survey included questions that illuminate small creditor non-QM lending

behavior. This analysis makes use of two CSBS survey questions that provide information on

differences in non-QM lending among small and not small creditors. T he first is a question

inquiring about the non-QM share of lenders’ 2014 portfolio originations and the second being a

question on lenders’ expected future plans for non-QM lending.

Figure 76 reports the share of non-QM loans held in portfolio among survey respondents. For

this question, 609 lenders responded with 584 of this total being small and 25 non-small. About

35 percent of small creditor survey respondents indicated that between 0 and 10 percent of their

portfolios consisted of non-QM loans. Many small creditors also held no non-QM loans in

portfolio (thereby indicating that they did not make non-QM loans in 2014). Notably, there is a

small spike in the distribution for small creditors at 90 to 100 percent non-QM loans being held

in portfolio. Such spike is not present for non-small creditors. This may be due to the small

creditor exemption, which, as previously discussed, required that small creditors hold Small

Creditor QM loans on their portfolios for three years to ensure QM status. This may indicate

that small creditors were more willing to extend credit to borrowers whose loans may not have

been approved by larger institutions due to non-QM loan features. It should also be noted that

227 BUREAU OF CONSUMER FINANCIAL PROT ECTION

about 16 percent of non-small respondents stated that between 70 and 90 percent of their

portfolios were non-QM in 2014 however, this amounts to four non-small lenders.

FIGURE 76: SHA RE OF LENDERS’ 2014 PORTFOL IOS THAT WERE NON-QM, BY SIZE OF INSTIT UT IO N S

Figure 77 reports institutions’ expectations of non-QM lending in 2015 with 664 lenders

responding to this question in the survey. Respondents were made up of 637 small creditors and

27 non-small. Specifically, this figure examines survey respondents’ plans for non-QM lending

in 2015 at the beginning of the year. Most small creditor respondents indicated that they would

offer non-QM loans in 2015, either as a part of their normal lending process or on an exception

basis. Though non-small institutions seemed proportionally more inclined to non-QM lending

going forward, this information indicates that small respondents generally expected to engage in

non-QM lending in 2015.

228 BUREAU OF CONSUMER FINANCIAL PROT ECTION

FIGURE 77: PLANS FOR OFFERING NON-QM LOANS IN 2015, BY SIZ E OF INSTITUTIONS

229 BUREAU OF CONSUMER FINANCIAL PROT ECTION

8. Additional effects of the Rule

The preceding chapters have outlined several effects the Rule has had on the mortgage market

in general, and on specific segments and types of lenders. This chapter discusses some

additional potential effects of the Rule. The chapter starts by presenting a study of how closing

times changed in response to the Rule. The chapter then presents additional results from the

lender survey the Bureau conducted that have not been discussed in previous chapters.

Main findings include the following:

• Closing times immediately after the effective date of the Rule increased by about three

and a half days for refinance loans and little over a day for purchase loans. It is likely that

these short term effects attenuated over time due to learning and adaptation to the new

requirements.

• The survey of lenders conducted by the Bureau establishes that a substantial share of

respondents changed their business model and/or their product offerings in response to

the Rule. The survey did not attempt to quantify the cost of these changes. About two

thirds of respondents report originating non-QM loans.

8.1 Effect on closing times

The Bureau used loan-level data reported under HMDA between 2011 and 2016 to study the

effect of the ATR/QM Rule on mortgage closing times—the number of days between a borrower

applying for a loan and the eventual closing of that loan. The Rule applied to all loans with an

application date on or after the Rule’s effective date. Therefore, the analysis estimates the Rule’s

effect by comparing the distributions of closing times on loans with application dates shortly

before and after the effective date. As such, the analysis is only equipped to estimate short-run

effects.

230 BUREAU OF CONSUMER FINANCIAL PROT ECTION

Simply comparing applications received either seven days or 21 days before and after the

effective date of the Rule, average closing times lengthened by less than a day, both for purchase

and refinance loans.

37 2

This comparison does not control for seasonal variation or fluctuations

in the volume or average loan size of applications or in borrower characteristics. To do so,

closing times are modelled as flexible functions of these variables.

37 3

With the model’s

predictions in hand, the Rule’s effect is estimated by examining how the difference between each

loan’s observed closing time (which reflects the Rule’s effect) and the predicted closing time

(which only accounts for the modelled characteristics and not the Rule) varied 30 days after the

effective date compared to the 30 days before.

37 4

Examining loans with application dates close to

the effective date allows focusing on the Rule’s effect rather than potential factors that vary

further from the effective date but are not captured in the model.

According to the model, average closing times lengthened by 1.2 days for purchase loans and 3.6

days for refinance loans after the ATR/QM Rule took effect.

37 5

This is larger than the simple

difference reported above, especially for refinance loans. This is because the model predicts that,

given seasonal factors and other covariates, the closing time of refinance loans would have

decreased following the Rule’s effective date. Instead, it increased slightly, implying the larger

effect estimated for such loans. The larger estimated effect on refinance loans compared with

purchase loans might be seen as suggestive evidence that the Rule added relatively more

documentation requirements for refinance loans, which were often more streamlined and less

costly to originate, than for purchase loans.

In sum, after controlling for confounding factors, average closing times increased by about three

and one half days for refinance loans, but the estimated effect on purchase loan closing times

was much smaller, at a little over one day.

37 6

Again, these are short term effects applicable to the

month immediately after the effective date of the Rule. It is possible that these short term effects

37 2

T o pu t that into c ontext, around the t ime t he Rule was im plemented, average c losing time was around 55 days for

pu r chases and around 45 days for r efinances.

37 3

Ex p l anatory v ariabl es inc lude 14 trigonometric terms to capture seasonality; purchase-, refinance-, a n d h om e -

im prov ement a pplication volumes; t he property’s state; the borrower’s incom e; whether there w as a c o-borrower;

bor r ower and, i f a ppl icable, c o-borrower race and ethnicity; loan amount; lien status; application day of the week;

w h ether the loan w as higher-priced or classified as a high-cost l oan; indicators for loan type and property type; and

sev eral interaction t erms.

37 4

Resu lts are sim ilar u sing a 60-day window and excluding 14 days before and after the effective date. This window

a llows for t he possibility that borrowers or l enders tried t o have c ertain types of a pplications submitted before the

Rule took effect.

37 5

Standard errors calculated using a methodology akin to bootstrapping establish that these changes were

st a tistically significant at t he 95 percent level of c onfidence.

37 6

Com m enters also r eported experiencing an increase in closing t imes, see Appendix B.

231 BUREAU OF CONSUMER FINANCIAL PROT ECTION

reflect delays due to one-time changes in software and systems, or staff needing to learn new

policies, practices, and systems, and that the effect on closing times may have attenuated over

time and with experience.

8.2 Survey evidence

Some of the responses to the Bureau’s survey of lenders described in Chapter 1 were

summarized in Chapter 5. Here we summarize responses to additional questions that address

the effects of the Rule on lenders. When considering these results, it is important to keep in

mind that the survey respondents are likely not representative of the market as a whole. As

mentioned in Section 5.2, the sample used from the lender survey includes 177 lenders which

excludes CDFIs and lenders that did not provide a substantive number of responses to

questions.

Table 40 reports responses by institution type to whether the lender’s business model changed

as a direct result of the ATR/QM Rule.

TABL E 40 : RESPO NS ES TO WHETH E R BUSINES S MODEL CHA NGED A S A DIRECT RESUL T OF THE

A TR/QM RULE

Bank with <$2

billion in

a sse ts

Bank with $2-

10 billion in

a sse ts

Bank with

>$10 billion in

a sse ts

Credit union

Independent

mortgage

lender

Business

model changed

27 5 22 17 29

No change 13 7 1 4 35

Total

responses

40 12 23 21 64

Overall, 100 respondents reported changing their business model, while 60 reported no

changes.

37 7

Generally, depository lenders were somewhat more likely to report changing their

business model compared to independent mortgage lenders.

37 7

Rec a ll that there were 1 96 r espondents t o the lender survey and an additional eight respondents t o the survey

am ong lenders providing the Application Data. Twenty-five of t he r espondents prov ided n o responses to the survey

a n d tw o respondents were CDFIs a nd thus not c overed by the Ru le, leaving 177 cov ered respondents with responses

232 BUREAU OF CONSUMER FINANCIAL PROT ECTION

TABL E 41 : WA Y S IN WHICH BUSINES S MODEL S WERE REPO R TE D TO HA V E CHA NGED

Changes mentioned

Depository

institution

Independent

mortgage

lender

Increased income documentation 26 5

DTI cap of 43 percent was introduced 25 3

Balloon loans discontinued 15 2

Structure of or income requirements for ARM loans

changed

12 1

Interest only loans discontinued 10 1

Decided not to originate non-QM loans 8 5

Experienced higher staffing and/or compliance costs 7 9

Had difficulties meeting points and fees test 5 6

Experienced longer closing times 3 0

Asset depletion no longer allowed 2 0

Total responses 62 25

Table 41 reports the specific ways in which respondents reported changing their business

models. The number of responses reported are not the same as the number of respondents as a

respondent could provide multiple, one, or no responses regarding the ways in which their

business model changed. Depository lenders were more likely to report increasing income

documentation requirements and introducing a 43 percent monthly DTI ratio cap while

independent mortgage lenders were more likely to report increased staffing and/or compliance

costs. The discontinuation of balloon and interest-only loans and the restructuring of ARMs

point towards effects on loans with restricted features studied in the first part of Chapter 4.

Table 42 reports responses by institution type to whether the lender discontinued or materially

modified mortgage products for reasons related to the requirements of the ATR/QM Rule.

t o a t least som e of t he questions. S eventeen respondents did not prov ide an answer t o the question regarding

bu si ness model change, t herefore the r esponses of 1 60 respondents are reported in Table 40.

233 BUREAU OF CONSUMER FINANCIAL PROT ECTION

TABL E 42 : RESPO NS ES TO WHETH E R PRODUCT WA S CHA NG ED OR DISCO NT I NUED

Bank with

<$2 billion

i n a sse ts

Bank with

$2-10

billion in

a sse ts

Bank with

>$10

billion in

a sse ts

Credit

union

Inde pendent

mortgage

lender

Made changes or

discontinued

products

15 4 18 10 16

Made no material

changes

25 9 5 10 47

Total responses 40 13 23 20 63

63 respondents out of 159 reported making changes to or discontinuing products.

37 8

Depository

lenders—especially large ones—were more likely to report making changes to or discontinuing

products compared to independent mortgage lenders. Two specific discontinued products were

mentioned by more than five respondents: balloon loans (15 responses) and interest-only loans

(6 responses). Eleven respondents stated that the discontinuation affected more than 10 percent

of their loans, 13 respondents said 5 to 10 percent of their loans were affected, while the

remaining 21 respondents giving a quantitative response said that less than 5 percent of their

loans were affected.

Table 43 shows lenders’ reported share of 2017 originations represented by non-QM loans by

institution type. For each institution type, except for the largest banks, about one third of

respondents do not originate non-QM loans.

37 9

There is one bank with assets over $10 billion

that does not originate non-QM loans, but this bank originates few mortgages.

37 8

Eig h teen respondents did not answer this question.

37 9

Con sistent w ith the findings here, c om menters noted industry survey ev idence on the lim ited share of non-QM

len ding by independent m ortgage lenders, see National Association of Realtors’ Survey of Mor tgage Originators,

Fir st Q u arter 2017: The Future of t he CFPB, QM, and Sm all Lender Rule, available at

https://www.nar.realtor/research-and-statistics/research-reports/mortgage-originators-survey /survey-of-

m ortgage-originators-fir st-quarter-2017.

234 BUREAU OF CONSUMER FINANCIAL PROT ECTION

TABL E 43 : SHA RE OF NON-QM LOANS

Non-QM Share

Bank with

<$2 billion

i n a sse ts

Bank with

$2-10 billion

i n a sse ts

Bank with

>$10 billion

i n a sse ts

Credit union

Independent

mortgage

lender

None

<5%

6 - 10%

11 - 15%

21 - 30%

>30%

Do not know 4 2 1 0 1

Total responses

Among the 98 lenders making non-QM loans and responding to the question regarding their

expectation regarding the change in their non-QM lending over the coming year, five expected

decreasing their non-QM lending, 25 expected increasing their non-QM lending, and 68 said

that their non-QM lending would stay about the same.

Finally, 67 respondents indicated originating loans in rural areas in 2013 and one indicated

discontinuing doing so in 2014. All others continued doing so through 2017.

Balloon loans were originated by 49 respondents in 2013 and 14 respondents indicated

discontinuing doing so in 2014 with the propensity to discontinue being larger for independent

mortgage lenders, credit unions and large banks as opposed to smaller banks. Five additional

respondents discontinued offering balloon loans after 2014, leaving 33 respondents offering

them in 2017.

235 BUREAU OF CONSUMER FINANCIAL PROT ECTION

Appendix A: THE RULE AND BUREAU PURPOSES AND

OBJECTIVES

Introduction

As discussed in Chapter 1, section 1022(d) of the Dodd-Frank Act requires the Bureau to

conduct an assessment of each significant rule or order adopted by the Bureau under Federal

consumer financial law. Section 1022(d) requires that the assessment address, among other

relevant factors, the rule’s effectiveness in meeting the specific goals stated by the Bureau, as

well as the Bureau’s purposes and objectives specified in section 1021 of title X of the Dodd-

Frank Act. Whereas the body of the report addresses the specific goals stated by the Bureau, this

appendix highlights certain core findings in the body of the report with respect to the latter

requirement.

380

Purposes

Under section 1021(a) of the Dodd-Frank Act, “[t]he Bureau shall seek to implement and, where

applicable, enforce federal consumer financial law consistently for the purpose of ensuring that

all consumers have access to markets for consumer financial products and services and that

markets for consumer financial products and services are fair, transparent, and competitive.”

381

All consumers have access to markets for consumer financial products and

services.

In issuing the Rule, the Bureau stated that it “sought to balance creating new protections for

consumers and new responsibilities for creditors with preserving consumers’ access to credit

380

As evidenced below, the degree to which the ATR/QM Rule implicates each of the purposes and objectives of title

X v aries, and the Bureau has endeavored t o include in this a ppendix i nformation that m ay be r elevant to those

pu r poses and objectives directly and indirectly im plicated. The Bureau further a cknow ledges that som e of the title X

pu rposes and objectives may overlap and some of the findings discussed below may be relevant for multiple

pu rposes and objectives. Thus, while this appendix distinguishes between purposes and objectives in order to

h ighlight key findings in t he body of t he report, the appendix is n ot m eant as a comprehensive summary of a ll

fi n dings relevant to each purpose and ob jective.

381

1 2 U.S.C. § 5511(a).

236 BUREAU OF CONSUMER FINANCIAL PROT ECTION

and allowing for appropriate lending and innovation.”

382

This concern recognizes that

establishing the ability to repay requirement for making residential mortgage loans created

litigation risk that could disrupt markets and thus consumers’ access to credit. Overall,

consumer access to residential mortgage loans after the Rule’s effective date has been preserved

at levels comparable to those before the effective date. Applications and approval rates analyzed

for the market overview discussed in Chapter 3 indicate that there were no significant breaks in

either applications or approval rates around the time of the Rule became effective. At this

aggregate level, neither supply nor demand were significantly disrupted, although applications

and approval rates might have been higher if not for the Rule. To the extent that the Rule may

have impacted mortgage pricing, In terms of mortgage pricing, Figure 5 of Chapter 3 shows that

the spread of the mortgage interest rate over the 10-year Treasury rate did not change

significantly around the time the Rule came into effect, so mortgage pricing did not appreciably

change at the market level after the Rule took effect. Mortgage pricing might, however, have

been lower if not for the Rule.

The Rule may have decreased consumer access to credit in certain sub-markets for which data

were available and reasonably obtainable. Indicating a decreased access to credit for a specific

group of consumers, Figure 45 in Chapter 5 shows that the introduction of the Rule was

associated with a sharp drop in both the share and approval rate of High DTI, non-GSE eligible

applications. After this initial drop, these outcomes continued to decline further. While the

average approval rate seems to have returned to the January 2014 level by the end of 2016, the

model estimates suggest that this reversal is due to changes in the mix of High DTI applicants

rather than due to changes in lenders’ propensity to approve applicants of a given set of

characteristics. Section 5.3.7 estimates that the QM DTI provision eliminated approximately

10,000 loans among this group of consumers over three years, for the lenders considered,

thereby decreasing access to credit. Section 5.3.6 also considers the effect of the Rule on self-

employed borrowers’ access to credit and finds the effects of the Rule to be relatively neutral.

Application Data presented in Table 21 indicates that approval rates for non-High DT I, non-GSE

eligible self-employed borrowers have decreased only slightly, by two percentage points, after

the Rule. Section 7.3 also considers loans that were denied due to the ATR requirements of the

Rule using CSBS survey data. Figure 71 in this chapter shows that 44 lenders, out of 693 who

responded to this question, denied at least half of the loans that would have been approved

absent the Rule in 2014, a decrease in access to credit.

382

7 8 Fed. Reg. 6408, 6505 (Jan. 30, 2013).

237 BUREAU OF CONSUMER FINANCIAL PROT ECTION

Chapter 7 indicates that amendments to the Rule in 2015 appear to have increased access to

credit for markets and consumers served by small creditors. The geographic market coverage of

small creditors increased substantially with the October 2015 amendments to the Rule, which

increased the loan originations threshold from 500 to 2,000 first-lien covered transactions and

excluded loans held in portfolio in determining whether this threshold had been crossed.

383

T he

number of counties served and the market share held by small creditors within individual

counties increased in 2016 compared to 2014. Without the amendment, the number of small

creditors, and markets served, would have decreased according to HMDA data.

Markets for consumer financial products and services are fair, transparent,

and competitive.

The Rule applies to all creditors that make residential mortgage loans. This broad coverage

promotes fairness in the sense of establishing a level playing field among creditors in this

market.

The Rule also prohibits practices (i.e. making loans without assessing the consumer’s ability to

repay) that the Board of Governors of the Federal Reserve System had found to be “unfair” at

least with respect to subprime borrowers. The Rule also helps ensure the markets for consumer

financial products are fair. Chapter 4 finds that loans with potentially risky non-QM

characteristics, including interest-only payments, low documentation, negative amortization,

ARM resets under five years, and terms exceeding 30 years, appear to be almost nonexistent or

restricted to a limited market of highly credit-worthy borrowers. These types of products largely

disappeared from the market prior to the adoption of the Rule, and it is not possible to assess

whether absent the Rule the incentives for such practices would return, but the Rule would

constrain the origination of loans for which consumers lack a reasonable ability to repay and the

resulting harms.

At the same time, the Rule does not appear to have inhibited competition among creditors, as

indicated by analyses in Chapter 5. Figure 45 in Chapter 5, analyzing the effect of the QM DTI

provision on non-GSE eligible High DTI applicants for purchase loans, indicates substantial

heterogeneity in response to the Rule across lenders. From a consumer’s point of view,

differences in competing lenders’ approaches to High DTI applicants evidence robust

competition and varied strategies for meeting that competition.

383

8 0 Fed. Reg. 59943 (O ctober 2, 2015).

238 BUREAU OF CONSUMER FINANCIAL PROT ECTION

Some aspects of the rule may also have resulted in limiting competition in the secondary market

for residential mortgage loans. Chapter 6 notes the market share of the GSEs has not decreased

in the years after the Rule went into effect, contrary to the Bureau’s expectation at the time Rule

was written. The persistent prominence of Temporary GSE QM lending likely reflects the GSEs’

advantages in compliance certainty, underwriting flexibility, accommodation of high-DTI

lending, and the availability of a robust secondary market. An entity that attempted to compete

with the GSE’s by selling bonds from securitizing loans eligible for purchase by the GSEs may

find it more difficult to ensure compliance with GSE requirements (and hence ensure

Temporary GSE QM status) and would not have the potential pricing advantages that come with

conservatorship. Also in Chapter 6, the data suggests a somewhat higher use of the GSEs’ AUS

in recent years, particularly for loans which do not fit within or are more difficult to document

within the General QM underwriting standards, such as loans made to self-employed

borrowers.

Objectives

The objectives of the Bureau are listed in section 1021(b) of the Dodd-Frank Act.

384

Consumers are provided with timely and understandable information to

make responsible decisions about financial transactions.

Although the Rule, and particularly the QM requirements, encourage creditors to originate loans

with understandable loan features, the Rule does not include requirements for information

disclosures that creditors must provide to consumers.

Consumers are protected from unfair, deceptive, or abusive acts and

practices and from discrimination.

In the years leading up to the passage of the Dodd-Frank Act, the Federal Reserve Board

imposed certain restrictions on high-cost loans based on a conclusion that such restrictions were

necessary to prevent unfair and deceptive acts or practices in connection with mortgage loans.

385

The ATR/QM Rule is not strictly a rule designed to address unfair deceptive or abusive acts and

practices, or to protect consumers from discrimination, and the Bureau has not determined that

384

1 2 U.S.C. § 5511(b).

385

7 3 Fed. Reg. 44522 (July 30, 2008)

239 BUREAU OF CONSUMER FINANCIAL PROT ECTION

the absence of repayment ability or the presence of any particular loan feature would render a

mortgage transaction unfair, deceptive, abusive, or discriminatory. The Dodd-Frank Act

nevertheless states that one purpose of the ATR requirement is to assure that consumers are

offered and receive residential mortgage loans on terms that reasonably reflect their ability to

repay the loans and that are understandable and not unfair, deceptive, or abusive. T he ATR/QM

Rule in turn has the potential to reduce the likelihood of unfair, deceptive, abusive, or

discriminatory acts or practices by generally restricting or eliminating the origination of loans

with riskier loan characteristics. Chapter 4 notes several consumer protection outcomes,

particularly those associated with the QM and DTI provisions of the Rule. Loans with potentially

risky non-QM characteristics, including interest-only payments, low documentation, negativ e

amortization, ARM resets under five years, and terms exceeding 30 years, appear to be almost

nonexistent or restricted to a limited market of highly credit-worthy borrowers, in contrast to

their more widespread use during the housing crisis. Chapter 4 provides evidence on the very

high foreclosure rates of loans with these features in the years leading up to the housing crisis.

Such products now appear to be restricted to credit-worthy borrowers, and likely will be

prevented from re-emerging on a large scale by the QM underwriting requirements.

For the most highly-leveraged conventional loan borrowers, DTI ratios are likely constrained

from returning to crisis-era levels by a combination of GSE underwriting limits and the Bureau’s

General QM DTI ceiling. Given the negative relationship between higher DTIs and loan

performance, demonstrated across loan types and over time, these limits contribute to ensuring

borrowers receive loans they are able to repay, in addition to potentially mitigating systemic

risks. In addition, non-QMs originated under the Rule’s ATR requirements also demonstrate

strong performance. As noted in section 4.4.2, both above and below the DTI limit of 43, the

improvement in performance of non-GSE loans relative to GSE loans provides some evidence

that those loans that continue to be made under the General QM, other non-T emporary GSE

QM, or non-QM ATR guidelines are underwritten in a way that reflects consumers’ ability to

repay.

Outdated, unnecessary, or unduly burdensome regulations are regularly

identified and addressed in order to reduce unwarranted regulatory

burdens.

The Bureau amended the January 2013 Rule several times before and after its effective date to

address important questions raised by industry, consumer advocacy groups, and other

stakeholders. For example, the May 2013 final rule added two new qualified mortgage categories

to the four categories provided in the January 2013 Rule. One of the new QM categories was for

loans held in portfolio by small creditors and the other was a temporary category that allowed all

small creditors to make balloon-payment qualified mortgages.

240 BUREAU OF CONSUMER FINANCIAL PROT ECTION

More changes were made by additional rules, including amendments that clarified provisions of

Appendix Q,

386

implemented a temporary points and fees cure provision,

387

and expanded the

definition of “rural” by adding census blocks

388

that are not in an “urban area,” as defined by the

Census Bureau, to the definition of rural areas.

389

The Bureau determined that these amendments were necessary to protect consumers better,

avoid potentially significant disruption in mortgage markets, and clarify standards by making

technical corrections and conforming changes. In these ways, the Bureau reduced the regulatory

burden imposed by the Rule several times.

Federal consumer financial law is enforced consistently, without regard to

the status of a person as a depository institution, in order to promote fair

competition.

The specific goals of the Rule, which are noted in Chapter 1, do not include consistent

enforcement of Federal consumer financial law without regard to status as a depository or non-

depository institution.

As noted in section 3.6, the Bureau has enforcement authority with respect to non-depository

mortgage originators

390

and depositories with assets over $10 billion,

391

and the prudential

regulators have enforcement authority with respect to smaller depositories.

Since the effective

date of the ATR/QM Rule, the Bureau has not brought enforcement actions against any entities,

depository or non-depository, for violating the Rule.

The Bureau has supervisory authority with respect to depositories with assets over $10 billion

392

and non-depositories engaged in residential mortgage lending.

393

As discussed in Chapter 3, the

Bureau has conducted examinations among large depositories and non-depository mortgage

386

7 8 Fed. Reg. 44686 (July 24, 2013).

387

1 2 C .F.R. § 1 026.43(e)(3)(iii) and (iv).

388

A census block is the smallest geographic area for which the Census Bureau collects and tabulates decennial

census data. See 80 Fed. Reg. 5 9943, 59956 (Oct. 2, 2015).

389

12 C.F.R. § 1 026.35(b)(2)(iv)(A)(2) and Supplement I t o Pa rt 1026—Official Interpretations, Paragraph

4 3 (f)(1)(vi)-1.

390

For en forcement authority of n on-depositories, see 12 U.S.C. § 5514(c).

391

For en forcement authority of depositories, s ee 12 U.S.C. § 5515(c).

392

For su pervisory authority of depositories, s ee 12 U.S.C. § 5515(a)-(b).

393

For su pervisory authority of n on-depositories, see 12 U.S.C. § 5514(a)(1)(A).

241 BUREAU OF CONSUMER FINANCIAL PROT ECTION

originators. Most mortgage originators examined by the Bureau, depository or non-depository,

have generally been complying with the ATR rule.

Finally, although it is not directly related to the consistent enforcement of the law, the Bureau

observes that the Rule applies to all creditors that make residential mortgage loans, promoting

fair competition by establishing a more level playing field among creditors in this market.

Markets for consumer financial products and services operate transparently

and efficiently to facilitate access and innovation.

Aspects of the Rule have facilitated access to credit, but in doing so may have had a negative

effect on innovation. As discussed in Chapter 6, the Temporary GSE QM was likely crucial in

maintaining short term access to responsible credit, in part due to its compliance certainty and

flexibility advantages, relative to the newly adopted Appendix Q, and in part due to robust

secondary market liquidity. However, given the large share of originations able to meet the

Temporary GSE QM criteria at the time the Rule became effective, there has been limited

momentum toward the emergence of a robust non-QM market, likely limiting innovation in the

non-GSE market.

As noted in Chapter 6, though, innovation is occurring in the mortgage market under the

umbrella of the Temporary GSE QM. However, the ability of the private sector to leverage this

innovation outside the GSE space has been limited. This limited ability may explain, at least in

part, why innovation in one segment of the market does not appear to have spurred growth and

innovation in others.

242 BUREAU OF CONSUMER FINANCIAL PROT ECTION

Appendix B: COMMENT SUMMARIES

On June 1, 2017, the Bureau published a request for information on the ATR/QM Rule (“Rule”)

assessment and invited the public to submit comments and information on a variety of topics.

394

The public comment period closed on July 31, 2017. The Bureau received approximately 480

comments in response to the RFI. The Bureau provides a description of the comments and

summarizes the information received on certain topics below and the full comments are

available on www.regulations.gov.

395

Generally, commenters reported on their own experiences, and provided information from

surveys and other types of research, regarding the overall effect of the Rule and the effects of

particular requirements that are within the scope of the assessment report. This information is

summarized here and incorporated into other parts of the report as appropriate. See Chapter 1,

“Sources of information and data,” for a summary of the data and information used in the

assessment.

396

The Bureau inventoried over 80 studies, surveys, and other types of research cited by

commenters regarding effects of the Rule on the market as a whole; on loan originators; on

consumers or particular subgroups of consumers; on credit unions; and on affiliated settlement

service providers. These research items were reviewed and their content and potential

394

82 Fed. Reg. 25246 (June 1, 2017). Under section 1022(d)(3), before publishing an assessment report, the Bureau

is r equ ired t o seek com ment on recommendations for m odifying, expanding, or elim inating the n ewly a dopted

significant rule or order. In the RFI, the Bureau invited the public to submit: (1) c om ments on the feasibility and

effectiveness of the assessment plan, the objectives of the ATR/QM Rule that the Bureau intends to emphasize in

the assessment, and the outcomes, metrics, baselines and analytical methods for assessing the effectiveness of the

Rule; (2) data and other factual information that may be useful for executing the Bureau’s assessment plan;

(3) recommendations to improve the assessment plan, as well as data, other factual information, and s ou r ces of da ta

t h at would b e useful a nd available to execute any recom mended im prov ements t o the a ssessm ent plan including

da t a on certain exceptions and prov isions; (4) data and other factual information about the benefits and costs of the

A T R/QM Ru le for consumers, creditors, and other stakeholders in the mortgage industry; and about the impacts of

th e Rule on transparency, efficiency, access, and innovation in the mortgage market; (5) data and other factual

in formation about the Ru le’s effectiveness in m eeting the purposes and ob jectives of t itl e X of the Dodd-Frank Act

(section 1021); and (6) recom mendations for modifying, expanding, or elim inating the A TR/QM Ru le. Id. at 25250.

395

A s st ated in the RFI, the Bureau is n ot generally responding to each comment received pursuant to the RFI. “The

Bu r eau plans to c onsider r elevant com ments and other information r eceived a s it conducts the a ssessm ent and

pr epares a n assessm ent report. T he Bureau does n ot, however, expect that i t will r espond in the assessm ent report

to each comment received pursuant to this document. Furthermore, the Bureau does not anticipate that the

assessment report will include specific proposals by the Bureau to modify any rules, although the findings made in

t h e assessm ent wil l help t o inform the Bureau’s thinking as to whether to consider commencing a rulemaking

proceeding in the future.” See 82 Fed. Reg. 25246, 25247 (June 1, 2017).

396

Sec tion 1 022(d)(1) prov ides that the a ssessm ent report shall reflect available ev idence a nd any data that the

Bu r eau reasonably may collect. Som e c om menters a lso directed the Bureau tow ard published research, which the

Bu reau reviewed and incorporated into other parts of the report as appropriate.

243 BUREAU OF CONSUMER FINANCIAL PROT ECTION

relationship to the Bureau’s assessment taken into account in developing this report. Some are

cited within the body of the assessment and all are listed, using the citation provided by the

respective commenter, below in this appendix, grouped under the individual subject headings

they address.

This appendix also contains a summary of recommendations for modifying, expanding or

eliminating the Rule.

397

Finally, section IV of the RFI described the assessment plan, and the

Bureau also invited comments on the plan. These comments are summarized below. The Bureau

continued to develop the assessment plan after publishing the RFI, taking into account the

comments received.

Evidence about ATR/QM Rule effects

398

General Comments about the Rule’s Effectiveness, Costs,

and Effects on the Mortgage Market

A number of commenters supported the Rule and noted what they considered to be the Rule’s

positive effects. A trade group and consumer advocacy organizations stated that the Rule

strengthened underwriting standards and eliminated higher risk products and features. A trade

group and a law professor stated that, while further improvements to the Rule are needed, the

Rule has restored common-sense principles to the mortgage origination market and has done so

without restricting access to credit. A creditor and some individual commenters also provided

general support without additional arguments.

397

Sec tion 1 022(d)(3 ) provides that before publishing a r eport of its assessment, the Bureau shall invite public

comm ent on r ec omm endations for m odify ing, expanding, or elim inating the newly adopted significant rule or order.

The Bureau invited these recommendations in the RFI.

398

Cer tain commenters offered evi dence on the ov erall effects of t he Rule. Chapter 3 presents an ov ervi ew of t rends

in the mortgage market, including trends in originations by loan size and consumer characteristics, and considers

w h ether the Rule affected these trends. Other effects for w hich com menters presented ev idence are prim arily

discussed as follows. Effects on small creditors are discussed in Section 7.2; credit availability, Sections 3.5, 5.3 and

5 .4.5; costs of or igination, Section 3.5; time to cl osing, Section 8.1; liability concerns general ly, C h apter 5 ( non-GSE

eligible loans) and Chapter 6 (GSE eligible loans); jumbo loans, Sections 3.5 , 4.4.1 and 6 .3.3; Appendix Q, Section

5 .3 .6 (bu t a l so Ch a pt er 6 ); m or tg a g e br oker s, Se ct i on 5 .4 .6 ; s peci fi c g r ou ps of c on su m er s, Sect i on s 5 .3 .5 ( FICO, LT V ,

in come), 5.3.6 (self-em ploy ed), 5 .4.5 (sm all l oans), 7 .2.4 ( rural c onsumers). C om menters generally presented

ev idence of a dverse changes and attributed t hese changes t o the Rul e. In som e cases Bu reau findings were

qu a ntitatively or directionally consistent w ith t his ev idence and i n other cases opposed; see each subsection for the

specific results.

244 BUREAU OF CONSUMER FINANCIAL PROT ECTION

Commenters that recommended that the Rule be relaxed or eliminated generally noted what

they considered to be negative effects of the rule. A trade group stated that the affordability and

availability of mortgage credit and the complete recovery of the housing market continue to be

adversely affected by a steady rise in regulatory compliance costs, loan origination entities’ fear

of enforcement action by the Bureau, and a lack of clear and reliable regulatory guidance. The

same trade group, along with another trade group and one individual commenter stated that the

Rule restricts access to credit.

As discussed under separate headings below, some commenters made general statements

concerning the Rule’s effects on the mortgage market, and some commenters made statements

on the Rule’s effect on specific market participants, consumers, and specific consumer sub-

groups.

Comments on the effects on the market as a whole

Commenters made several general statements about the effect on the mortgage market as a

whole. Trade groups and a number of small creditors and loan originators stated that the Rule

has reduced competition in the mortgage market and favors large creditors. A trade group stated

its survey of members suggests that the Rule is having a downward impact in lending, with 72

percent of survey participants responding that the ATR/QM Rule is affecting credit availability

and 7 percent responding that the impact is “severe.” The commenter also noted that the current

state of homeownership has remained between 62.9 and 63.7 percent, a "plateau" that

constitutes the lowest rate in more than 50 years, according to the Census Bureau.

A creditor and a trade group stated that the Rule has increased compliance costs and risks to

creditors, with the trade group stating that a member creditor has increased the number of

employees to comply with the Rule at a cost in additional salaries of $238,000 in addition to

third-party costs, such as compliance support and audits of roughly $52,000 annually. Another

trade group stated that market studies indicate that over the past decade the cost of originating a

mortgage has increased by 72 percent, from approximately $4,376 in the third quarter of 2009

to approximately $7,562 by the fourth quarter of 2016.

A trade group stated that the Rule, along with other regulatory changes, has made it difficult for

creditors to stay profitable or continue operations in some markets in response to increased

costs and legal risks, leaving consumers with fewer options. A creditor stated that a $12,000

home equity loan now requires the same processing time as an $800,000 mortgage loan.

Another creditor stated that the Rule is not working as desired, since time to closing and

creditor costs have increased while credit has become less available and that small loans,

especially for low to middle income consumers, are less likely to be originated, as a result.

245 BUREAU OF CONSUMER FINANCIAL PROT ECTION

Studies, surveys, and research cited by commenters regarding effects of the Rule on the market

as a whole:

• a creditor cited Report on the Economic Well-Being of U.S. Households in 2016,

(Washington DC, 2017); a trade group cited Changes in U.S. Family Finances from 2007

to 2010: Evidence from the Survey of Consumer Finances, June 2012;

• a creditor cited Neil Bhutta & Glenn B. Canner, Mortgage Market Conditions and

Borrower Outcomes: Evidence from the 2012 HMDA Data and Matched HMDA-Credit

Record Data, Fed. Res. Bull., Nov. 2013;

• a creditor, mortgage insurer, and a trade group of credit unions cited A Financial System

that Creates Economic Opportunities, (June 2017).

• a trade group cited the Real Estate Service Providers Council Survey dated 2013;

• another trade group cited Ken Fears, Reach of New Risky Loans Still Modest,

Economist’s Outlook Blog, National Association of Realtors, June 15,2017

• a consumer advocacy organization cited the a report for the Conference of State Banking

Supervisors by William F. Basset and John C. Driscoll, Post Crisis Residential Mortgage

Lending by Community Banks (2015);

• a trade group, a creditor and consumer advocacy organizations cited Laurie Goodman,

Jun Zhu, Bing Bai, Tight credit standards prevented 5.2 million mortgages between 2009

and 2014, Urban Wire, January 28, 2016;

• a trade group cited Laurie Goodman, Jun Zhu, Bing Bai, Overly tight credit killed 1.1

million mortgages in 2015, Urban Wire, November 21, 2016 ;

• that same trade group, two groups of consumer advocacy organizations, and a creditor

cited Laurie Goodman, Bing Bai, Ellen Seidman, Has the QM Rule Made It Harder to Get

a Mortgage?, February 2016;

• the same trade group cited Edward Golding, Laurie Goodman, Jun Zhu, Fannie Mae

Raises the DTI Limit, July 2017;

• a law professor cited Laurie Goodman, Quantifying the Tightness of Mortgage Credit and

Assessing Policy Actions;

• a group of consumer advocacy organizations cited Jim Parrot and Mark Zandi, Opening

up the Credit Box (2013)

246 BUREAU OF CONSUMER FINANCIAL PROT ECTION

• a creditor cited Edward Golding, Laurie Goodman and Jun Zhu, Fannie Mae Raises the

DTI Limit, July 2017

• a trade group cited Richard Green, The Trouble with DTI as an Underwriting Variable-

and as an Overlay” Richard’s Real Estate and Urban Economics Blog, December 7, 2016;

• a trade group cited Diane Katz, Heritage Foundation, A Better Path for Mortgage

Regulation, Feb. 28, 2017;

• the same trade group also cited a statement of Todd Zywicki, Professor of Law and

Executive Director of the Law and Economics Center, George Mason University School

of Law, Assessing the Effects of Consumer Financial Regulations: Hearing Before the S.

Comm. on Banking, Hous. & Urban Affairs, 115th Cong. (2016);

• the same consumer advocacy organizations cited Michael Stone, What is Housing

Affordability? The case for the Residual Income Approach, Housing Policy Debate, Vol.

17, Issue 1 (Aug. 31, 2006);

• a trade group cited Steve Holden and Walt Scott, Desktop Underwriter Version 10.1 –

Updates to DTI Ratio Assessment, Fannie Mae, July 10, 2017

• a creditor cited Matthew Jozoff, The Cost of Post-Crisis Regulation on Mortgage

Lending, J.P. Morgan Research (March 31, 2017);

• consumer advocacy organizations cited CoreLogic, United States Residential Foreclosure

Crisis, Ten Years Later (2017);

• the same consumer advocacy organizations cited CoreLogic, Mortgage Performance

Continues Steady Improvement in April 2017;

• a different group of consumer advocacy organizations cited CoreLogic, Home Equity

Lending Landscape (Feb. 2016)

• a creditor cited a report from Moody Analytics by Mark Zandi & Cristian DeRitis, The

Skinny on Skin in the Game, March 11, 2011;

Comments on the effect of liability concerns

One trade group stated that the mortgage market remains dominated by loans covered by the

qualified mortgage safe harbor. The commenter further stated that this concentration is due

primarily to liability concerns and uncertainties around what can happen outside of the QM

“safety zones.” The commenter stated that Bureau did a commendable job in eliminating

specious class action probabilities, but nonetheless creditors and investors remain uncertain as

247 BUREAU OF CONSUMER FINANCIAL PROT ECTION

to how the courts will interpret the numerous standards that exist in the Rule and apply them to

specific circumstances. The trade group also stated that its members anticipate that foreclosure

challenges asserting non-compliance with the Rule will emerge based on whether creditors

complied with general statutory requirements, for which there is also assignee liability under

T ILA.

Comments on the effect of the Temporary GSE QM definition

Several commenters specifically provided statements concerning the impact of the Temporary

GSE QM provision of the Rule. A few trade groups stated that the Temporary GSE QM provision

has worked to prevent significant disruption in the mortgage market and enable lenders to

continue to originate loans seamlessly. One trade group stated that the fact that all loans sold to

GSEs automatically are qualified mortgages provides a great benefit to both consumers and

creditors by significantly reducing the amount of burdensome and oftentimes confusing

paperwork that goes in the mortgage application, allowing creditors to lend more efficiently and

to more consumers. Another trade group stated that the temporary GSE qualified mortgage

provision has: combined a bright-line definition with underwriting flexibility, which has allowed

creditors to reach deeper into the population of credit-worthy consumers and permitted

responsible lending above a 43 percent debt-to-income ratio. However, the commenter also

stated that not including jumbo mortgages into the Temporary GSE Qualified Mortgage

definition contributed to the retarded recovery of private label securities market because the

investment community has rejected these mortgages.

Comments on the effects of Appendix Q of Regulation Z

399

A trade group and a creditor stated that Appendix Q is ambiguous and leads to uncertainty,

inappropriate results, and restricts appropriate access to credit. Another trade group stated that

the current definition of income causes documentation problems, litigation and liability risk,

and harms consumers with less than meticulous credit records. A consumer advocacy

organization stated that the documentation standards for self-employment income add time and

expense to the mortgage application process and can discourage creditors and borrowers from

pursuing loans when such income is present. A trade group stated that for income from part-

time employment, the amount of time it takes to properly document and assess a two-year

history of income, consumers essentially need to have been working up to three years for it to be

used to determine income, which is extremely burdensome on those consumers. A trade group

399

See also recom mendations to m odify, expand or eliminate Appendix Q below .

248 BUREAU OF CONSUMER FINANCIAL PROT ECTION

and a creditor stated that the treatment of work history gaps, as well as documentation of a new

job, interferes with appropriate access to credit.

Comments on effects to specific market participants

Comments concerning the Rule’s effectiveness, costs, and effect on the mortgage market often

focused on effects to specific market participants and ranged from effects on loan originators

and credit unions to those on consumers generally, first-time homebuyers, retired consumers,

self-employed consumers, rural consumers, minority consumers, credit-challenged consumers,

special program consumers (e.g., doctors), and consumers seeking loans with a low loan

amount.

LOAN ORIGINATORS

Trade groups and individual commenters, especially mortgage brokers, stated that the Rule had

negative effects on non-employee loan originators in particular, including: reducing competition

that limits consumer options; unequal treatment of loan originators that work for mortgage

brokers compared to those employed by depository institutions, forcing small loan origination

organizations out of business; and reducing compensation to mortgage brokers while increasing

compliance costs. Some of these commenters also stated that the Rule favors banks, but should

create an even playing field for brokers, wholesale lending, small independent mortgage

originators, and banks. These commenters stated no other business has regulations of caps on

income and expenses like mortgage brokers and since the loan origination compensation

requirements under other rules protects consumers, there is no need to limit loan originator

compensation further through the Rule’s points and fees provisions. A creditor stated that

mortgage broker lending is less costly and better for consumers and that there are fewer

complaints concerning the conduct of mortgage brokers in the Bureau’s consumer complaint

database than complains about banks, but that the Rule has resulted in reduced lending by

mortgage brokers.

Studies, surveys, and research cited by commenters regarding effects of the Rule on Loan

Originators:

• A trade group cited to a paper from the Chicago Federal Reserve Bank by Amany El

Anshasy, Gregory Elliehausen & Yoshiaki Shimazaki, The Pricing of Subprime

Mortgages by Mortgage Brokers and Lenders (July 2005) (unpublished manuscript);

• A trade group cited the National Association of Realtors’ Research Division, 2017

Member Profile, (May 2017);

249 BUREAU OF CONSUMER FINANCIAL PROT ECTION

• a trade group cited M. Cary Collins & Keith D. Harvey, Mortgage Brokers and Mortgage

Rate Spreads: Their Pricing Influence Depends on Neighborhood Type, 19 J. HOUSING

RES. 153, 168 (2010)

CONSUMERS

Some commenters referenced effects of the Rule on consumers. A number of individual

commenters stated that the Rule made it difficult for many consumers to get a loan. A number

of individual commenters, creditors, and a trade group stated that consumers are being harmed

by the Rule since it has resulted in a lack of competition between providers of mortgages. A few

individuals and a creditor stated that the Rule has resulted in increased costs to consumers.

Some individuals and a creditor stated that the Rule has harmed specific groups of consumers,

including retirees, self-employed consumers (due to a lack of consideration of their income in

Appendix Q), consumers residing in rural areas, consumers with challenging credit (also

referred to as subprime borrowers), consumers seeking low loan amounts, and consumers who

seek special programs, such as doctors. One commenter stated that the Rule has resulted in

reducing supply of consumers attempting to move up into another home which leads to an

oversupply of higher priced homes. A trade group stated that the Rule has harmed minority

borrowers. A trade group stated that the vast amounts of paperwork required to meet the Rule’s

requirements often leads consumers to become frustrated with the mortgage process and back

out. The commenter further stated that consumers are frustrated when they apply for mortgages

since the requirements are the same for a small home equity loan or for a large purchase loan.

Studies, surveys, and research cited by commenters regarding effects of the Rule on consumers

or particular subgroups of consumers:

• A title company cited the economic analysis conducted in connection with HUD’s Final

Rule Amendments to Regulation X, the Real Estate Settlement Procedures Act:

Withdrawal of Employer-Employee and Computer Loan Origination Systems (CLOs)

Exemptions, 61 Fed. Reg. 29238 (Jun. 7, 1996);

• the same commenter cited a study commissioned by HUD conducted by Peat, Marwick,

Mitchell & Co., Real Estate Closing Costs, RESPA, Section 14a, prepared for HUD under

Contract H-2910, Project Code 4 .3.01.103 (Oct. 1980);

• a creditor cited Report on the Economic Well-Being of U.S. Households in 2016,

(Washington DC, 2017); a trade group cited Changes in U.S. Family Finances from

2007 to 2010: Evidence from the Survey of Consumer Finances, June 2012;

250 BUREAU OF CONSUMER FINANCIAL PROT ECTION

• a creditor cited Neil Bhutta & Glenn B. Canner, Mortgage Market Conditions and

Borrower Outcomes: Evidence from the 2012 HMDA Data and Matched HMDA-Credit

Record Data, Fed. Res. Bull., Nov. 2013;

• a joint letter from consumer advocacy groups cited Underwriting Standards Ease for

Fourth Consecutive Year, OCC Survey Shows, Office of the Comptroller of the Currency

(2016)

• a joint letter from consumer advocacy groups cited Title Insurance: Actions Needed to

Improve Oversight of the Title Industry And Better Protect Consumers, Government

Accountability Office (2007);

• a creditor cited Elka Torpey and Andrew Hogan, Working in a gig economy, Career

Outlook, Bureau of Labor Statistics, May 2016;

• a trade group cited an examination of closed loans in Duval County, Florida, apparently

conducted by the commenter or at least without a reference to its source, concerning

consumer costs at closing for a creditor transaction averaging $6,222 and for a

mortgage broker transaction, after credits applied, averaging $3,479

• a trade group cited a survey conducted on behalf of the National Association of Realtors

by Harris/Nielson, One-Stop Shopping Consumer Preferences, October 6, 2015;

• a trade group, a creditor and consumer advocacy organizations cited Laurie Goodman,

Jun Zhu, Bing Bai, Tight credit standards prevented 5.2 million mortgages between

2009 and 2014, Urban Wire, January 28, 2016;

• a trade group cited Laurie Goodman, Jun Zhu, Bing Bai, Overly tight credit killed 1.1

million mortgages in 2015, Urban Wire, November 21, 2016;

• that same trade group, two groups of consumer advocacy organizations, and a creditor

cited Laurie Goodman, Bing Bai, Ellen Seidman, Has the QM Rule Made It Harder to

Get a Mortgage?, February 2016;

• the same trade group cited Edward Golding, Laurie Goodman, Jun Zhu, Fannie Mae

Raises the DTI Limit, July 201 7;

• a law professor cited Laurie Goodman, Quantifying the Tightness of Mortgage Credit

and Assessing Policy Actions;

• a group of consumer advocacy organizations cited Jim Parrot and Mark Zandi, Opening

up the Credit Box (2013)

251 BUREAU OF CONSUMER FINANCIAL PROT ECTION

• a trade group cited M. Cary Collins & Keith D. Harvey, Mortgage Brokers and Mortgage

Rate Spreads: Their Pricing Influence Depends on Neighborhood Type, 19 J. HOUSING

RES. 153, 168 (2010)

• a creditor cited Sean M. Hoskins, The Ability to Repay Rule: Possible Effects of the

Qualified Mortgage Definition on Credit Availability and Other Selected Issues, January

9, 2014;

• the same creditor cited Patrick T. O'Keefe, Qualified Mortgages & Government Reverse

Redlining: How the CFPB's Qualified Mortgage Regulations Will Handicap the

Availability of Credit to Minority Borrowers, Fordham Law Review;

• a group of consumer advocacy organizations cited Michael Calhoun, The Federal

Housing Administration can do more with more, Brookings (2017);

• a title insurance company cited Michael H. Riodan, Competitive Effects of Vertical

Integration, Columbia University Department of Economics Discussion Paper Series,

2005;

• the same title insurance company cited T imothy Bresnahan and Jonathan Levin,

Vertical Integration and Market Structure, Stanford Institute for Economic Policy

Research Mar. 2012;

• the same title insurance company also cited Lawrence J. White, The Title Insurance

Industry, Reverse Competition, and Controlled Business - A Different View, The

Journal of Risk and Insurance, Vol. 51 , No. 2 (1 984);

• a group of consumer advocacy organizations cited Heather K. Way & Lucy Wood,

Contracts for Deed: Charting Risks and New Paths for Advocacy, 23 J. Affordable Hous.

& Cmty. Dev. L. 37 (2014);

• the same consumer advocacy organizations cited Michael Stone, What is Housing

Affordability? The case for the Residual Income Approach, Housing Policy Debate, Vol.

17, Issue 1 (Aug. 31, 2006);

• the Center for Responsible Lending, by Ellen Schloemer, et al., Losing Ground:

Foreclosures in the Subprime Market and Their Cost to Homeowners (2006);

• the Center for Responsible Lending, by Sarah Wolff, Analysis of HMDA Data 2012-

2015;

252 BUREAU OF CONSUMER FINANCIAL PROT ECTION

• a creditor cited Center for Responsible Lending, by Debbie Gruenstein Bocian, Wei Li,

and Keith S. Ernst, Foreclosures by Race and Ethnicity—The Demographics of a Crisis,

June 18, 2010 ; and a law professor also cited the Center for Responsible Lending,

Roberto Quercia, et al., Balancing Risk and Access (2012);

• a creditor cited Rachel Witowski, Blacks and Hispanics Likely to Be Hurt by Qualified

Mortgage Rule, Nat'l Mortgage News, October 22, 2013;

• a trade group of credit unions cited a member credit union as stating that they Rule,

together with another rulemaking, has required the credit union to pass long costs to

the consumer of $54 per transaction at consummation and $247 in increased annual

mortgage costs

CREDIT UNIONS

A number of trade groups and credit unions described effects to their segment of the mortgage

industry. A few trade groups stated that the Rule makes it unnecessarily difficult for credit

unions to provide high quality, consumer-friendly products to their members. One trade group

stated that credit unions are good-faith partners helping their members buy a home, and the

Rule should reflect this. A trade group stated that a study conducted by the Credit Union

National Association indicated the Rule had an impact of $7.2 billion on credit unions in 2014. A

credit union stated that the Rule resulted in higher mortgage costs, more restrictive portfolio

lending, and reduced efficiencies in its lending process. The credit union further stated that the

Rule has required additional staffing and oversight, upgrades in technology systems, and

additional time to evaluate loans under the new underwriting guidelines and verification

processes.

Studies, surveys, and research cited by commenters regarding effects of the Rule on Credit

Unions:

• a survey conducted on behalf of the Credit Union National Association by Haller, Jon;

Ledin, Paul; and Malla, Bandana, Credit Union National Association Impact of CFPB

Rules Survey (Feb. 2017);

• a trade group of credit unions cited a study for the Credit Union National Association by

Hui, Vincent; Myers , Ryan; and Symour, Kaleb, Regulatory Financial Impact Study;

• a consumer advocacy group cited the Credit Union National Association Mutual

Group’s Credit Union Trends Report (2017);

253 BUREAU OF CONSUMER FINANCIAL PROT ECTION

• a trade group of real state service providers cited a series of surveys, the most recent

dated 2015 conducted by Harris/Nielsen concerning affiliated business relationships in

real estate transactions;

• a trade group and consumer advocacy groups cited surveys by the National Association

of Insurance Commissioners Of State Insurance Laws Regarding Title Data And Title

Matters, from 2010 and 2015;

• a title insurance company cited Michael H. Riordan, Competitive Effects of Vertical

Integration, Columbia University Department of Economics Discussion Paper Series,

2005;

• the same title insurance company cited T imothy Bresnahan and Jonathan Levin,

Vertical Integration and Market Structure, Stanford Institute for Economic Policy

Research Mar. 2012;

• the same title insurance company also cited Lawrence J. White, The Title Insurance

Industry, Reverse Competition, and Controlled Business - A Different View, The

Journal of Risk and Insurance, Vol. 51 , No. 2 (1 984 );

• a title company cited Analysis Group, Inc., Competition and Title Insurance Rates in

California, Jan. 2006;

• a trade group of credit unions cited a member credit union as stating that they Rule has

required the credit union to add additional positions costing $238,000 annually along

with an additional $70,000 for a different rulemaking, $52,000 in extra annual costs in

third party contracts for the two rules.

AFFILIATED SETTLEMENT SERVICE PROVIDERS

Some commenters referenced effects of the Rule on affiliated settlement service providers,

stating that the Rule has had an adverse effect on the ability of affiliated settlement service

providers to compete with other market participants.

Studies, surveys, and research cited by commenters regarding effects of the Rule on affiliated

settlement service providers:

• A title company cited the economic analysis conducted in connection with HUD’s Final

Rule Amendments to Regulation X, the Real Estate Settlement Procedures Act:

Withdrawal of Employer-Employee and Computer Loan Origination Systems (CLOs)

Exemptions, 61 Fed. Reg. 29238 (Jun. 7, 1996);

254 BUREAU OF CONSUMER FINANCIAL PROT ECTION

• the same commenter cited a study commissioned by HUD conducted by Peat, Marwick,

Mitchell & Co., Real Estate Closing Costs, RESPA, Section 14a, prepared for HUD under

Contract H-2910, Project Code 4 .3.01.103 (Oct. 1980);

Recommendations to modify, expand, or

eliminate the ATR/QM Rule

This section discusses the comments received that recommend modification, expansion, or

elimination of the Rule. As noted in the Request for Information, the findings made in this

assessment, and these comments, will help inform the Bureau as to whether to consider

commencing rulemaking in the future in relation to the Rule.

Commenters provided numerous suggestions for specific changes to components of the Rule.

The areas discussed included requirements associated with definition of a qualified mortgage;

the definition of points and fees and the maximum cap of 3 percent of the loan amount on points

and fees; changes to, or elimination of, the maximum debt-to-income ratio for qualified

mortgages; elimination or extension of the Temporary GSE QM definition; establishing

permissible calculations related to a residual income calculation that can be used in lieu of the

debt-to-income ratio; changes to Appendix Q of Regulation Z, which can be used to determine

the amount of income and debt of the consumer used to calculate the debt-to-income ratio;

changes to permit asset-based lending; changes to post-consummation cures to ensure qualified

mortgage status; exemption of credit unions from the Rule; and miscellaneous other changes.

Comments on general principals and goals for Rule modifications

A few commenters discussed general principles and goals of modifications to the Rule. A trade

group stated that the Dodd-Frank Act gave the Bureau tremendous latitude and discretion, so

the assessment should determine revisions to the Rule, and that it is crucially important to

preserve the fundamental intent of law, which the commenter identifies as restricting risky

practices and encouraging traditional prime lending. A creditor stated that the Rule should

provide simple guidelines, rather than complex underwriting rules. A trade group stated that the

Rule should clarify bank statement lending programs. A creditor stated that the Rule should

change ability-to-pay reliance on historical performance to allow for new products with no

performance history. A trade groups stated that the Rule must advance towards a uniform and

transparent set of guidelines, criteria, and compensating factors that are objective and policy-

based and independent of any institutional market participant. Two trade groups stated that any

changes to the Rule should be applied holistically and not vary based on size of the institution or

business model. One of these trade groups also stated that the Rule should responsibly widen

the credit box so that many more consumers can benefit from safe, sustainable mortgages. A

255 BUREAU OF CONSUMER FINANCIAL PROT ECTION

trade group of State regulators stated that the Rule should provide more flexibility for

community banks that rely on relationship lending and use more qualitative data in their

lending. A group of consumer advocates stated that property assessed clean energy loans and

home equity lines of credit should be covered by the Rule.

Commenters stated that various changes to the definition of a qualified mortgage should be

made to the Rule. A trade group supported the statutory definition for qualified mortgage, but

recommended removing the debt-to-income ratio requirement. Several trade groups, creditors

and individual commenters stated that the qualified mortgage safe harbor should be expanded

to mortgages held in portfolio by creditors, or at least those held by credit unions or community

banks. These commenters stated such a change would facilitate worthwhile lending since the

Rule discourages lending outside of the safe harbor. Consumer advocacy groups stated that the

qualified mortgage safe harbor should not be expanded to mortgages held in portfolio by larger

institutions, as the risk of foreclosure is not a constraint on lending if a consumer has enough

equity in the property that secures the mortgage, which could lead to equity stripping by

creditors originating mortgages to hold in portfolio. These consumer advocacy groups also

stated that high-cost mortgages should not receive the qualified mortgage safe harbor since

high-cost mortgages are inherently dangerous and have a long track record of consumer harm.

A number of trade groups stated that the scope of the safe harbor should be expanded by

adjusting the threshold of high-cost mortgages to a higher threshold of 250 basis points above

the average prime offer rate. A trade group stated that non-qualified mortgages should be

considered to be qualified mortgages if the consumer consistently makes periodic payments on

the mortgage for a certain period of time and the mortgage is considered “seasoned”. Two

individuals stated that streamline refinances should meet the definition of qualified mortgages.

One individual stated that land installment contracts should not be considered qualified

mortgages, while two consumer advocacy organizations stated that the Bureau should make

clear that land installment contracts are mortgages and are covered by the Rule.

Comments related to the inclusion of loan originator compensation in the

definition of points and fees

Approximately three quarters of the comments consisted of comments from loan originators or

loan originator organizations. These commenters were consistently critical of how loan

originator compensation is treated under the points and fees requirement for qualified

mortgages and requested various changes for regulatory relief. Several individual commenters

stated that the points and fees definition in the statute was a drafting error. These commenters

stated that the points and fees definition double counts mortgage broker compensation.

Commenters suggested changes to the definition of “mortgage broker” to recognize the

difference between a loan originator and a loan originator organization. Commenters suggested

that the Rule should exclude compensation paid by the creditor to the mortgage broker from the

256 BUREAU OF CONSUMER FINANCIAL PROT ECTION

definition of points and fees or only count compensation received directly from the consumer.

The commenters stated the compensation is merely the gain on sale of the promissory note into

the secondary mortgage market, and that the consumer does not directly pay these funds to the

mortgage broker. Some individual commenters also stated that the Bureau has recognized the

error in the points and fees provision, because the integrated mortgage disclosure forms do not

include a requirement to disclose the split in compensation between a loan originator

organization and the individual loan originator.

Some creditors and a large number of individual commenters stated that the Rule should

remove the statutory 3 percent cap on points and fees from the qualified mortgage definition

altogether because of the deleterious impact that the provision has had on small-business

mortgage brokers and low- and moderate-income consumers, as well as because the operation

of the loan originator compensation rule sufficiently constrains the actions of mortgage brokers

and protects consumers.

Comments concerning the limit on points and fees in relation to loans for

smaller amounts

A few trade groups and individual commenters stated that the Bureau should modify the tiers

for smaller loan amounts from their existing thresholds because the high costs associated with

origination for creditors make the current levels less economically feasible, reducing the

attractiveness of low balance loans to creditors. A group of consumer advocates stated that the

levels for smaller loans should not be changed unless there is clear evidence that the Rule is

artificially reducing access.

Comments recommending raising the points and fees limit

A few individuals stated that the points and fees limit should be increased from the 3 percent

amount to 5 percent, or 5.5 percent. A group of consumer advocates and a trade group stated

that the definition of points and fees should not change since there is no possibility of a refund if

there is an overpayment of points and fees, while a consumer can at least try to refinance a

mortgage if they happen to have a higher interest rate.

Comments on the inclusion of various charges in the points and fees

definition

AFFILIATE FEES

A few trade groups and individuals stated that fees paid to affiliates should be excluded from the

definition of points and fees to encourage the development of ‘one-stop shopping’ for mortgages.

257 BUREAU OF CONSUMER FINANCIAL PROT ECTION

Comments from groups of consumer advocate organizations stated that affiliate charges should

remain included in the points and fees definition. These commenters also stated that the

removal of affiliate charges from the definition of points and fees would significantly increase

prices for those services.

PRIVATE MORTGAGE INSURANCE

Two trade groups stated that upfront private mortgage insurance payments should be excluded

from the definition of points and fees since equivalent government program upfront fees are not

included in the definition and the determination of whether a loan can be repaid is not

determined by a mortgage insurance provider’s identity.

BONA FIDE DISCOUNT POINTS

An individual commenter stated that the Bureau should provide more guidance on the exclusion

of bona fide discount points from the definition of points and fees.

Comments recommending changes to the maximum permissible debt-to-

income ratio

A number of commenters stated that the maximum permissible debt-to-income ratio for a

qualified mortgage should be eliminated or modified. A number of trade groups, a creditor, and

some individual commenters stated that the maximum debt-to-income ratio for a qualified

mortgage should be eliminated because it makes no sense, has hurt consumers with difficult to

document income, and that other measurements, such as cash flow, would be a more inclusive

indicator of the ability to repay the loan. One trade group stated that the maximum debt-to-

income amount should be eliminated or other methods for satisfying the qualified mortgage

definition should be seriously considered if there is no significant change to mortgage

performance when the debt-to-income ratio exceeds 43 percent.

One trade group commenter stated that the varying levels of permissible debt-to-income ratios

between the GSEs and loans guaranteed or insured by a government agency versus a General

QM’s 43 percent debt-to-income ratio creates opportunities for regulatory arbitrage, leading

creditors to direct consumers to a mortgage product based on regulatory provisions rather than

the needs of the consumer. This commenter stated that the results of the assessment should be

used to examine debt-to-income ratios above 43 percent and its interplay with compensating

risk factors with a view towards creating a more expansive, uniform, and transparent standard.

This commenter further stated that if harmonizing the various debt-to-income standards

amongst qualified mortgages is not possible, the various government agencies that guarantee or

258 BUREAU OF CONSUMER FINANCIAL PROT ECTION

insure mortgages should be required to justify higher debt-to-income ratios and any permissive

lending standards should be accompanied by periodic monitoring and reporting requirements.

Several trade groups, creditors, and individual commenters stated that the debt-to-income ratio

should be changed to an amount higher than 43 percent, some without specifying the amount

and others stating that the amount should be increased by various levels ranging from 45 to 60

percent and either outright increases, under extenuating circumstance, or when the Temporary

GSE QM provision expires, in order to provide consumers with more access to credit. A creditor

and a couple of individual commenters stated the maximum debt-to-income ratio should be

eliminated for jumbo mortgages since the Rule does not effectively measure the ability to repay

for them and that high-income consumers do not need the same protections as other

consumers. One individual commenter stated that the maximum debt-to-income ratio should be

a sliding scale based on gross income. A trade group stated that a ‘one size fits all’ regulation

does not work, and an individual commenter sated that investors should be permitted to

establish their own debt-to-income ratio. In contrast to the suggested increases to the maximum

debt-to-income ratio amount, two individual commenters stated that there should be no

increases.

Comments in relation to the elimination or extension of the Temporary GSE

QM provision

Two trade groups and an individual commenter stated that the Temporary GSE QM provision

should be removed and the Bureau should rely on core statutory requirements to define a

qualified mortgage. One trade group stated that although it believes that the Temporary GSE

QM provision is essential for mortgage market support at the present, the Temporary GSE QM

provision must eventually sunset. Another trade group stated that it supported the Temporary

GSE QM provision unless and until the Bureau develops a standard that more effectively

balances the need for a bright line with the reasonable credit underwriting that balances

multiple factors. A few trade groups stated that the Temporary GSE QM provision has combined

a regulatory bright line with underwriting flexibility for creditors by permitting the creditor to

use GSE underwriting standards in order to comply with the qualified mortgage requirements,

allowing creditors to reach deeper into a population of credit-worthy consumers. Two groups of

consumer advocates and two trade groups stated that the Temporary GSE QM provision has

worked and should be maintained for various reasons, including the effect of an expiration on

the availability of credit to consumers, the necessity of doing responsible lending above a 43

percent debt-to-income ratio, and maintaining underwriting flexibility that is incorporated in

the GSE standards, which is not possible in a regulation. An individual commenter suggested

extending the Temporary GSE QM provision for seven years, while three trade groups and a

consumer advocacy organization suggested an indefinite extension until an alternative is in

place. Three trade groups, two creditors and a consumer advocacy organization stated that the

259 BUREAU OF CONSUMER FINANCIAL PROT ECTION

Temporary GSE QM provision should be permanent. A creditor and two trade groups also

supported extending the Temporary GSE QM provision to the jumbo mortgage market since the

discrepancy in treatment of jumbo mortgages interferes with the securitization process. One

trade group stated that the Bureau should clarify that the documentation requirements for the

Temporary GSE QM provision does not require that the mortgage actually be purchased or

guaranteed by a GSE. A provider of credit scoring models stated the implicit endorsement of the

FICO credit model in GSE underwriting standards should be expanded to include other credit

scoring models.0

Comments concerning Appendix Q of Regulation Z

GENERAL

Many commenters suggested modifications, modernization, simplification, and alternatives to

Appendix Q of Regulation Z, which describes how to determine income and debt for use in the

maximum debt-to-income ratio of 43 percent under the Rule. A few trade groups, a creditor, and

a consumer advocacy organization argued that Appendix Q should be completely eliminated.

Two of these trade groups with the creditor and consumer advocacy organization stated that the

Bureau should develop a transparent set of criteria, including compensating factors, to define a

qualified mortgage to replace Appendix Q. A trade group stated that Appendix Q was borrowed

from static, vague, and outdated guidelines that do not reflect today’s employment and income

trends and documentation standards, let alone technological norms for complying and verifying

information and a consumer’s ability to repay. Two trade groups and two individual

commenters stated that the Bureau should approve alternatives to Appendix Q, including

commonly accepted underwriting standards such as those of the GSEs, FHA, VA and RHS.

CALCULATING INCOME AND DEBT

Several commenters specifically discussed modifications to the method used to determine

income set forth in Appendix Q, especially in relation to stated difficulties for specific groups of

consumers. These groups included consumers that receive income from self-employment, part-

time employment, renting real property, social security, and nontraditional income.

As noted above, a trade group stated that for income from part-time employment, the amount of

time it takes to properly document and assess a two-year history of income, consumers

essentially need to have been working up to three years for it to be used to determine income,

which is extremely burdensome on those consumers. This trade group further stated that

Appendix Q is not clear if the income can come from any part-time job held during the two year

period, or if it must be from the same job, and that the Bureau should clarify this requirement.

In addition this trade group states that in determining social security income under Appendix Q,

260 BUREAU OF CONSUMER FINANCIAL PROT ECTION

creditors should be permitted to gross up to a standard 125 percent instead of to the consumer’s

tax rate, as it would make the process easier and less burdensome on the consumer, without any

impact on underwriting.

An individual commenter stated that the 25 percent cap on variable income should be increased

to something more reasonable like 45 to 50 percent. A trade group and a creditor stated that the

Bureau should clarify that income from vacation rentals received by a consumer should be

considered a valid income for Appendix Q. A creditor stated that income received from a line of

credit to manage cash flow from accounts receivables of self-employed consumers is not

considered income under Appendix Q, but should be. Another trade group stated that Appendix

Q is confusing on how to use asset depletion as income.

A credit reporting agency stated that the Rule should allow income estimation models, such as

the Income Insight Score, to be used to determine income and only require third party

verification if the Income Insight Score does not match the consumer’s stated income. An

individual commenter stated that Appendix Q contains a dichotomy in treatment between

income and debt related to student loans, as the payments due on student loans are included in

determining debt while anticipated increases in income from the consumer’s education are not

considered in determining income. A creditor stated that Appendix Q counted certain debts

twice in determining debts and should be changed to avoid such a result.

An industry trade group stated that Appendix Q should permit the use of compensating factors

and residual income in determining income. This trade group, along with a few other trade

groups and a creditor, stated that the Bureau should consider the VA’s residual income test as an

option to use in the definition of a qualified mortgage. Another two trade groups and a creditor

stated a residual income threshold would allow more flexibility in the Rule. Another trade group

stated that the Bureau should provide a clear definition of residual income, since the current

definition causes documentation problems, litigation and liability risk, and harms consumers

with less than meticulous credit records.

DOCUMENTING INCOME AND DEBT

Several commenters stated there are problematic issues related to documentation of income

under Appendix Q. A number of commenters, including several trade groups, creditors and

individual commenters, stated that additional guidance from the Bureau in relation to Appendix

Q is needed in relation to verification of employment (especially for foreign nationals), lending

to non-permanent resident aliens, and clarity on the effect of a decline in income for self-

employed consumers on the determination of income.

261 BUREAU OF CONSUMER FINANCIAL PROT ECTION

A trade group and a creditor stated that the treatment of work history gaps, as well as

documentation of a new job, interferes with appropriate access to credit. Another creditor stated

that provisions requiring explanations of 30-day gaps in employment is unnecessarily short and

should be expanded, and that alternative employment documentation to what is currently

accepted should be sufficient for Appendix Q. A trade group and a creditor stated that Appendix

Q is confusing and unworkable on how to use a consumer’s tax return for documentation. This

trade group, and another creditor, stated that tax transcripts should be used to document

income instead of copies of the complete tax return forms completed by consumers, since the

majority of consumers file and sign tax returns electronically. This trade group also stated that

requirements for a balance sheet and profit and loss statements for sole proprietors and

partnerships to document income should be eliminated in Appendix Q.

A trade group stated that documentation requirements that include current lease information

for rental income is duplicative as creditors are already required to collect tax information for at

least two years, creating another burden on consumers. This trade group also stated that social

security income should be able to be documented by a direct deposit or checking statement

showing that the social security funds were deposited in the consumer’s account, instead of

requiring a benefit verification letter. In addition, this trade group stated that the written

verification of continuance requirements for military add-on income (such as basic allowance

for housing or subsistence) should be eliminated because requiring consumers to provide

written verification of this additional income is not required for civilian consumers and is almost

impossible to obtain by the consumer or creditor.

A creditor stated that Appendix Q should permit the annuitization of assets to substitute for

income verification. An individual commenter stated that reduced documentation should be

available for self-employed consumers that demonstrate a 5 year history of being in business.

Comments recommending extension or adoption of a post-consummation

cure provision

A few trade groups and a consumer advocacy organization stated that the post-consummation

cure for points and fees that will expire in January 2021 should be made permanent. An

industry trade group and two creditors stated that the post-consummation cure should be

expanded to include instances of missing documentation to establish compliance with the Rule.

The same trade group, along with two other trade groups stated that the post-consummation

cure should also be extended to include debt-to-income ratio issues. A creditor and the same two

trade groups stated that creditors should be able to use a post-consummation cure for the points

and fees threshold by providing refunds to consumers.

262 BUREAU OF CONSUMER FINANCIAL PROT ECTION

Comments recommending creation of an exemption from the Rule for credit

unions

A few credit unions and a trade group stated that an exemption from the Rule for credit unions

was appropriate. These commenters stated that credit unions did not cause the mortgage crisis,

do provide a financial benefit to their members, and do not profiteer off of them. These

commenters suggested the Bureau exercise its authority under Dodd-Frank Act section

1022(b)(3)(A) and exempt credit unions from the Rule. Another trade group stated that the

asset size limit for credit unions to originate small creditor qualified mortgages should be

increased.

Other comments recommending modification, expansion, or elimination of

the Rule

Some commenters included proposed modifications to the Rule that are not tied to specific

provisions nor a general comment on the Rule itself. One trade group stated that creditors

should be permitted to make decisions based on a consumer’s overall credit profile, including

looking at the consumer’s credit score and the loan-to-value ratio for consumers relying on

income from assets to repay the loan. Further, this trade group stated that that creditors should

be permitted to make decisions on where a consumer falls within a higher credit score bracket,

relaxing guidance for loan-to-value ratios if the credit score analysis supports the transaction for

consumers seeking smaller-than-average loan amounts. A trade group stated that no-

documentation and low-documentation mortgages are no longer possible, so the Rule should be

modified since the risks associated with these categories is gone. A creditor stated that profit on

the sale of a mortgage is a good indication of risk and should be considered by the Rule, and that

a consumer’s cash reserves and loan-to-value ratio should be offsetting factors. Another creditor

stated that the Rule should permit third-party verification of consumer’s records. One group of

consumer advocates stated that the Bureau should examine whether it is appropriate to have

different standards for the Rule based on the source of credit insurance for the mortgage (e.g.,

GSEs, FHA, or VA). One trade group of State banking regulators stated that asset size is a bad

way to define community banks, and that the Bureau should adopt the definition used by the

FDIC.

263 BUREAU OF CONSUMER FINANCIAL PROT ECTION

The assessment plan

400

Comments concerning the baseline measurements

An industry trade group commenter stated that, because the market retraction underway at the

time of the Rule’s effective date would suggest that using the law’s inception point as a baseline

to measure impact could lead to erroneous conclusions, the selection of a baseline for

comparison be carefully weighed and considered by the Bureau. T his commenter urged the

adoption of a multidimensional perspective that at a minimum adopts multiple baselines that

can be compared with current lending activity. The commenter indicated that the use of

multiple baselines allows for broader comparisons of potential policy courses that should be

considered in determining optimal solutions and regulatory restructuring of the Rule.

A trade group stated that the assessment should include an analysis of the products and

structures that caused the 2008 mortgage crisis, as well as the role that securitization of non-

document and low-document loans played in the crisis including inaccurate ratings of mortgage-

back securities. This commenter also stated that the assessment should focus on cost,

origination volume, approval rates, subsequent loan performance, millennial and immigrant

markets, self-employed borrowers, and the interaction between the dramatic increase in closing

costs since 2008 and the inclusion of affiliates in the definition of points and fees in the Rule.

Comments concerning the sufficiency of the data

Some commenters stated that the data and other factual information to be used was insufficient,

specifically that the use of daily rate-sheets would not be useful in determining trends in the cost

of credit.

Comments recommending specific data to be reviewed

Two commenters suggested that the assessment plan should also focus on the number of safe

harbor qualified mortgages, rebuttable presumption qualified mortgages, and non-qualified

mortgages, and the accompanying income, credit score, and demographic data for each

category. Another suggested the use of HMDA reports and information from the GSEs, FHA,

VA, and industry databases, such as Mortgage Banker Association surveys.

400

A s n oted a bov e, the Bureau c ontinued t o develop the assessment plan after publishing the RFI, t aking into

a cc ount the com ments received. See also Chapter 1 , section 1.2.

264 BUREAU OF CONSUMER FINANCIAL PROT ECTION

Comments recommending the use of qualitative methods

Some commenters suggested the use of qualitative methods would help the assessment,

specifically interviews of creditors and surveys of various mortgage market participants and

subsets of participants. One trade group stated that the Bureau should interview a broad-based

sample of community banks that lend in both rural and non-rural markets, including those that

qualify for the small creditor and rural exemptions and those that do not. Consumer advocacy

groups stated that the assessment should include outreach to consumers and consumer

advocates.

Comments concerning data on impacts on creditors

A commenter stated that the assessment plan only focuses on consumer outcomes, and does not

consider effects on the mortgage industry and marketplace, and that the conclusions will be

unnecessarily constrained and not fulfill Congressional intent. Some commenters stated that the

assessment should include reviews of information and impacts on creditors and the market. A

trade group stated that the assessment should include the perspective of all the purposes and

objectives laid out in Dodd-Frank Act sections 1021 and 1022, and ensure the Bureau identifies

and addresses outdated, unnecessary or unduly burdensome rules within the scope of its

rulemaking authority. Another trade group stated that the assessment should carefully consider

the effect the Rule has had on credit unions and their members.

Comments concerning review of regulatory costs

A creditor stated that the assessment’s scope should be expanded to include market outcomes,

measured by elapsed time from application to consummation, costs, credit availability, and

regulatory burden. A trade group also stated that the assessment should analyze regulatory costs

in terms of the overall compliance environment, taking into account the interrelation of all

mortgage reforms that currently impact lending operations. This commenter also stated that the

assessment should analyze the Rule and other rules, e.g. loan originator compensation.

Comments concerning review of access to sustainable credit

An academic commenter stated that metrics will need to be developed to evaluate whether

mortgage regulation, including the Rule, increases access to sustainable credit. A trade group

stated that the assessment should determine which consumers that have been shut out of the

mortgage market, and whether enhancements to qualified mortgage standards, or further

harmonization of the various qualified mortgage standards, can improve access to credit while

still protecting consumers.

265 BUREAU OF CONSUMER FINANCIAL PROT ECTION

Comments recommending a focus on specific metrics

A number of commenters stated that the assessment should focus on particular areas

encompassed by the Rule and specific metrics related to mortgage origination.

A trade group stated the assessment should look at the volume of small balance versus higher-

balance qualified mortgages, demand for loans at various loan amounts, and the impact of

points and fees on smaller loan balances. This commenter also stated the assessment should

look at the performance of mortgages at various debt-to-income ratios, since before the current

ratio was selected, the Bureau considered and provided FHFA loan performance data for public

comment and that data on current loan performance should again be obtained and offered for

public comment. The commenter further stated that observable compensating factors that may

affect the performance of mortgages at various debt-to-income ratios should be better

understood for the assessment, as well as the comparative debt-to-income ratios between

qualified mortgages and the performance of GSE, VA, FHA, and RHS mortgages. The same

commenter also stated the assessment should analyze the effect of the Rule on access to credit

by estimating the number of mortgages in each category of qualified mortgages and non-

qualified mortgages, such as those that meet the Temporary GSE QM provision, those that meet

the General QM definition, and those that have a safe harbor or rebuttable presumption of

compliance.

This commenter, as well as a group of consumer advocates and another trade group, stated that

the assessment should focus on the Rule’s effect on access to credit by consumers. The group of

consumer advocates also stated that the assessment should focus on preventing unaffordable

lending. A trade group stated that the assessment should prioritize analysis of the market impact

of the Temporary GSE Qualified Mortgage provision and begin the process of identifying

appropriate uniform standards that can eventually replace this provision without disrupting

markets. A trade group stated that if the assessment considers only the Rule’s QM standards, it

will have a significant gap because the assessment will achieve an understanding of only some,

but not all, products being offered to consumers. A trade group stated that costs of court

litigation and eventual settlements must form a part of the assessment, as unknown litigation

risks associated with non-qualified mortgages has been a primary factor in the failure of

investors to support a reemergence of private label security markets.

266 BUREAU OF CONSUMER FINANCIAL PROT ECTION

Comments on effect of a lack of clarity in the Rule

A trade group stated that the assessment should consider whether difficulties with originating

non-qualified mortgages are based on a lack of clarity on how to comply with the ability-to-pay

requirements. This commenter, an industry participant, and a consumer advocacy organization,

state that the assessment should review whether underwriting guides of the GSE and

governmental programs should serve as alternatives to Appendix Q.

Comments recommending a focus on effects to affiliates

A number of trade groups and other commenters, stated that the assessment should measure

the Rule’s effect on affiliates, sometimes generally and sometimes in the context of the Rule’s

definition of points and fees.

267 BUREAU OF CONSUMER FINANCIAL PROT ECTION

Appendix C: APPLICATION DATA REQUEST TO NINE

LENDERS

As mentioned in Sections 1.3 and 5.3.1, the Bureau collected de-ide ntified application-level data

from nine lenders

using its authority under section 1022(c)(4) of the Dodd-Frank Act. The

lenders represent a range of national banks and non-depositories and includes applications

received from 2013 to 2016. This amounts to five million applications in total with information

about each application’s characteristics and whether the application was approved, denied or

withdrawn by the lender.

The nine lenders provided data on applications received by their institution, affiliates,

correspondent lenders, and mortgage brokers. For applications received from the correspondent

or broker channel, only applications where the lender made the final credit decision were

included. Applications data were requested for first-lien home purchase or refinance closed-end,

owner-occupied, one to four family residential consumer mortgages. Certain restrictions were

placed such as excluding pre-approval requests, incomplete applications, any personally

identifiable information, and information on the race or ethnicity of the applicant or co-applicant.

Below is a comple te list of the requested fields contained in the Application Data.

Application Fields

# Field name Field description

Labels and value

ranges

Note s

[1]

fico

FICO score of the

applicant at

origination. If more

than one applicant,

report the lowest

score.

1 = <620

2 = 620 – 659

3 = 660 – 679

4 = 680 – 699

5 = 700 – 719

6 = 720 – 739

7 = 740+

[2]

numborr

Number of

applicants on the

application.

1 = One

2 = More than one

[3]

inc_src

Primary source of

income used to

qualify for a loan. If

more than one

applicant, this

applies to the

applicant with the

highest income.

1 = Full time

employment

2 = Part time

employment

3 = Monthly

income from

retirement/pension

plan

4 = Monthly

withdrawals from

This field will be

adjusted to reflect

detail recorded in

your data systems.

268 BUREAU OF CONSUMER FINANCIAL PROT ECTION

# Field name Field description

Labels and value

ranges

Note s

assets

5 = Other

U = Unknown

[4]

self_emp

Indicator for a self-

employed

applicant whether

it was a primary or

secondary source

of income. If more

than one applicant,

this is an indicator

for whether one of

the applicants is

self-employed.

1 = Self-employed

0 = Not self-

employed

U = Unknown

[5]

delinq

Delinquent at the

time of application

or in default on

any Federal debt

or any other loan,

mortgage, financial

obligation, bond,

or loan guarantee?

If more than one

applicant, then

indicate whether

any applicant is

delinquent.

1 = Delinquent

0 = Not delinquent

N/A = Not

applicable

U = Unknown

[6]

bankruptcy

An indicator for a

bankruptcy in the

last 7 years, by

any of the

applicants.

1 = Bankruptcy

0 = No bankruptcy

U = Unknown

[7]

dti Back-end DTI.

1 = <21.01%

2 = 21.01 –

30.00%

3 = 30.01 –

40.00%

4 = 40.01 –

43.00%

5 = 43.01 –

45.00%

6 = 45.01 –

50.00%

7 = >50.00%

[8]

amount

Loan amount, in

dollars

1 = <60,001

2 = 60,0001 –

100,000

3 = 100,001 –

150,000

4 = 150,001 –

269 BUREAU OF CONSUMER FINANCIAL PROT ECTION

# Field name Field description

Labels and value

ranges

Note s

250,000

5 = 250,001 –

417,000

6 = 417,001 –

625,000

7 = >625,000

[9]

inc

Application

income, per

month, in dollars.

1 = <2,501

2 = 2,501 – 5,000

3 = 5,001 – 7,500

4 = 7,501 – 10,000

5 = 10,001 –

12,500

6 = 12,501 –

15,000

7 = >15,000

[10]

purpose Loan purpose.

1 = Purchase

2 = Refinance

Do not include

applications for

other purposes

[11]

app_date

Application date,

MM/YYYY.

MM/YYYY

Include

applications

received between

01/01/2013 –

12/31/2016

[12]

gse_eligib

An indicator of an

application for a

loan that is eligible

to be purchased,

insured, or

guaranteed by a

GSE (regardless

of whether you

sold, or intended

to sell, the loan to

GSE's).

1 = GSE-eligible

0 = Not GSE-

eligible

U = Unknown

[13]

loantype

The type of

mortgage product

applied for.

Response options:

"GSE" - you

typically sell loans

originated under

this product to

GSE's

"FHA" - an

application for an

FHA insured loan

"VA" - an

application for VA

guaranteed loan

1 = GSE

2 = FHA

3 = VA

4 = USDA/RHS

5 = JUMBO

6 = Other

270 BUREAU OF CONSUMER FINANCIAL PROT ECTION

# Field name Field description

Labels and value

ranges

Note s

"USDA/RHS" - an

application for a

guaranteed

USDA/RHS loan

"Jumbo" - a loan

above the

applicable GSE

conforming limit.

[14]

portfolio

An indicator for an

application for a

portfolio mortgage

product. You

typically keep

loans originated

under this

mortgage product

in portfolio, for at

least 1 year after

origination.

1 = Portfolio

product

0 = Not a portfolio

product

U = Unknown

[15]

ltv LTV

1 = <50.01%

2 = 50.01 –

80.00%

3 = 80.01 –

90.00%

4 = 90.01 –

95.00%

5 = >95

Discuss cases

where appraisal

information is not

available on an

application

[16]

cltv

An indicator of

whether the

CLTV>LTV on the

loan application.

This usually

occurs if there is a

contemporaneous

application for a

second mortgage.

1 = CLTV>LTV

0 = CLTV=LTV

U = Unknown

[17]

paytype

Payment type of

the mortgage

product.

1 = Fixed

2 = ARM

3 = Balloon

4 = Other

[18]

term

Amortization term,

in years.

Numeric values

(e.g., 5, 10, 15, 30

etc.)

[19]

arm_t

Length of initial

term before reset

Numeric values,

such as 1,3,5,7,10

271 BUREAU OF CONSUMER FINANCIAL PROT ECTION

# Field name Field description

Labels and value

ranges

Note s

for an ARM loan,

in years, for ARM

loans.

N/A = non-ARM

loans

[20]

balloon_t

Length of the term

before the balloon

payment, for a

balloon loan.

Numeric values

N/A = Not a

balloon loan

[21]

points_test

An indicator for

whether the

application has

passed the QM

points and fees

test (e.g., the sum

of applicable

points and fees

does not exceed

the relevant

threshold).

1 = Application has

passed the QM

points and fees

test

0 = Application has

not passed the QM

points and fees

test

U = Unknown

[22]

hoepa

An indicator for a

HOEPA loan.

1 = HOEPA

0 = Not HOEPA

U = Unknown

[23]

channel

Origination

channel where the

application was

acquired.

1 = Retail

2 = Correspondent

3 = Broker

U = Unknown

[24]

units

Number of units in

the property.

1, 2, 3, 4

U = Unknown

Only applications

for 1-4 unit single

family homes. If

the exact number

of units is not

captured,

alternatively

indicate whether

this is a 1 unit or

>1 unit property

[25]

Manufactured

housing indicator.

1 = Manufactured

home

0 = Not a

manufactured

home

U = Unknown