This PDF is a selection from an out-of-print volume from the National Bureau
of Economic Research
Volume Title: Pensions in the U.S. Economy
Volume Author/Editor: Zvi Bodie, John B. Shoven, and David A. Wise, eds.
Volume Publisher: University of Chicago Press
Volume ISBN: 0-226-06285-6
Volume URL: http://www.nber.org/books/bodi88-1
Publication Date: 1988
Chapter Title: Defined Benefit versus Defined Contribution Pension Plans:
What are the Real Trade-offs?
Chapter Author: Zvi Bodie, Alan J. Marcus, Robert C. Merton
Chapter URL: http://www.nber.org/chapters/c6047
Chapter pages in book: (p. 139 - 162)
5
Defined Benefit versus
Defined Contribution
Pension Plans: What are
the Real Trade-offs?
Zvi Bodie, Alan
J.
Marcus, and Robert
C.
Merton
Although employer pension programs vary in design, they are usually
classified into two broad types: defined contribution and defined ben-
efit. These two categories are distinguished in the law under ERISA.
Under
a
defined contribution (DC) plan each employee
has
an account
into which the employer and, if it is a contributory plan, the employee
make regular contributions. Benefit levels depend on the total contri-
butions and investment earnings of the accumulation in the account.
Often the employee
has
some choice regarding the type
of
assets in
which the accumulation is invested and can easily find out what its
value is
at
any time. Defined contribution plans are, in effect, tax-
deferred savings accounts in trust for the employees, and they are by
definition fully funded. They are therefore not
of
much concern to
government regulators and are not covered by Pension Benefit Guar-
antee Corporation (PBGC) insurance.
In
a
defined benefit (DB) plan the employee’s pension benefit entitle-
ment is determined by
a
formula which takes into account years
of
service for the employer and, in most cases, wages or salary. Many
defined benefit formulas also take into account the Social Security
benefits to which an employee is entitled. These are the
so-called
in-
tegrated plans.
See
Merton, Bodie, and Marcus
(1987)
for
a
discussion
of integration.
Zvi Bodie
is
professor of finance and economics at the School
of
Management, Boston
University, and a research associate
of
the National Bureau of Economic Research. Alan
J.
Marcus is associate professor of finance and economics at the School
of
Management,
Boston University, and a faculty research fellow of the National Bureau of Economic
Research. Robert
C.
Merton is
J.
C.
Penney Professor of Management at the Sloan
School
of
Management, Massachusetts Institute
of
Technology, and a research associate
of
the National Bureau
of
Economic Research.
139
140
Zvi
BodieIAlan
J.
MarcusIRobert
C.
Merton
DB and DC plans have significantly different characteristics with
respect to the risks faced by employers and employees, the sensitivity
of benefits to inflation, the flexibility of funding, and the importance
of governmental supervision. Our objective in this paper is to examine
the trade-offs involved in the choice between
DB
and DC plans.
In section
5.1,
we briefly review the mechanics governing the de-
termination and valuation of the benefit streams under DB and DC
pension plans. Section
5.2
contains an informal discussion
of
the rel-
ative advantages of each type of plan. In section
5.3
we develop a
formal model to examine the trade-offs between the two types of plans
in the face of both wage and interest rate uncertainty. Our conclusion
is
that neither plan can be said to wholly dominate the other from the
perspective of employee welfare. Section
5.4
summarizes our results
and concludes the paper.
5.1
Plan Characteristics and Valuation
5.1.1
Defined Contribution Plans
The DC arrangement is the conceptually simpler retirement plan.
The employer, and sometimes also the employee, make regular con-
tributions into the employee's retirement account. The contributions
are usually specified as a predetermined fraction of salary, although
that fraction need not be constant over the course of
a
career.'
Contributions from both parties are tax-deductible,* and investment
income accrues tax-free. Often the employee is given a choice as to
how his account is to be invested. In principle, contributions may be
invested in any security, although in practice most plans limit invest-
ment options to various bond, stock, and money-market funds. At
retirement, the employee either receives a lump sum or an annuity, the
size of which depends upon the accumulated value of the funds in the
retirement account. The employee thus bears all of the investment risk;
the retirement account is by definition fully funded, and the firm has
no obligation beyond making its periodic contribution.
Valuation
of
the DC plan is straightforward: simply measure the
market value of the assets held in the retirement account. However,
as a guide for personal financial planning, the DC plan sponsor often
provides workers with the indicated size of a life annuity starting at
retirement age that could be purchased now with the accumulation in
their account under different scenarios. The actual size of the retire-
ment annuity will, of course, depend upon the realized investment
performance of the retirement fund, the interest rate at retirement, and
the ultimate wage path of the employee.
141
Defined Benefit
versus
Defined Contribution Pension Plans
5.1.2
Defined Benefit Plans
Whereas the
DC
framework focuses on the
value
of the assets cur-
rently endowing a retirement account, the DB plan focuses on theflow
of benefits which the individual will receive upon retirement.
A typical DB plan determines the employee’s benefit as a function
of both years of service and wage history. As a representative plan,
consider one in which the employee receives
1
percent of average salary
(during the last 5 years of service) times the number of years of service.
Normal retirement age is
65,
there are no early retirement options,
death
or
disability benefits, and no Social Security offset provisions.
The actuarially expected life span at retirement
is
80 years.
Assuming the worker is fully vested, at any point in time his claim
is a deferred nominal life annuity, insured up to certain limits by the
Pension Benefit Guarantee Corporation. It is a deferred annuity because
the employee cannot start receiving benefits until he reaches age
65.
It is nominal because the retirement benefit, which the employer is
contractually bound to pay the employee, is fixed in dollar amount at
any point in time up to and including retirement age.
Many people think that under final average pay plans of the sort
described here, retirement benefits are implicitly indexed to inflation,
at least during the employee’s active years with the firm, and therefore
should not be viewed as a purely nominal asset by the employee and
a
purely nominal liability by the firm. We examine this issue in detail
in section
5.2.
For now we focus on the value of the explicit claim only.
Given an interest rate and a wage profile, it is straightforward to
compute the present value of accrued benefits under our prototype DB
plan. Table
5.1
presents such values for workers at different ages as-
suming a constant real annual wage of $15,000. The present value of
accrued liabilities can increase from continued service because of
3
factors:
(1)
as years of service increase,
so
does the defined benefit,
(2)
if the wage increases,
so
will the retirement benefit, and
(3)
as time
passes, less time remains until the retirement benefits begin,
SO
that
their present value increases at the rate of interest.
To illustrate the separate contributions of each of these factors to
the cumulative results reported in table
5.1,
consider the case in which
the benefit formula calls for
1
percent of final year’s salary times years
of service and that the worker lives for
15
years after retiring at age
65. The worker is
35
years old, has worked for the firm
10
years, and
his current salary is $15,000. The nominal interest rate equals a real
rate of
3
percent per year plus the expected rate
of
inflation.
Under the
7
percent inflation scenario, the sources of the change in
the value of the pension benefit from the passage of an additional year
are as follows. Prior to this year, the worker had accrued a life annuity
Table
5.1
Present Value
of
Accrued Benefits and Marginal Change
in
Benefits for Hypothetical Worker,
No
Early Retirement
Present Value
of
Accrued Benefits
in Constant
Dollars
0%
Inflation
7%
Inflation
3%
Discount
10%
Discount
Rate Rate
Starting Age
25
Current Age
Marginal Change in
Present Value
of
Accrued Benefits
from
an Additional
Year’s Work
0%
Inflation
7%
Inflation
3%
Discount
10%
Discount
Rate Rate
Constant
%
of
Constant
%
of
Dollars Salary Dollars Salary
30
35
40
45
50
55
60
65
$2,274
$5,271
$9,167
$14,169
$20,532
$28,563
$38,63 1
$51,181
$144
$463
$1,120
$2,404
$4,840
$9,354
$17,575
$32,329
$455
$527
$61 1
$708
$82
1
$952
$1,104
$1,242*
3.03
3.51
4.07
4.72
5.47
6.35
7.36
8.28
$4
1
$82
$158
$297
$546
$938
$1,768
$2,794*
.27
.55
1.05
1.98
3.64
6.25
11.79
18.63
NOTES:
Worker currently paid
$15,000
per year with no real wage growth.
Worker will retire at age
65.
Pension plan pays
1
percent
of
average salary in last
5
years times years
of
service.
Pension plan contains no early retirement provisions
or
makes correct actuarial adjustment
for
early retirees.
Benefits are vested after
5
years.
Real
interest rate is
3
percent, nominal rate increases one
for
one with inflation.
*Value calculated for age
64
rather than age
65.
SOURCE:
Adapted from Ellwood
(1985)
143
Defined Benefit versus Defined Contribution Pension Plans
of
$1,500
per year
(1
percent
x
10
years
x
$15,000)
beginning at age
65.
With a nominal interest rate of
10%
per year, the present value
(PV) of this deferred annuity at age
35
is
$654.
The increase in pension
benefits as
a
result of working an additional year can be broken into
three parts:
Factor
1:
One additional year of service at
a
salary of
$16,050 ($15,000
x
1.07) entitles him to an additional deferred annuity of
$160.50
per year, and
Factor
2:
The salary increase of
$1050
entitles him to an additional
deferred annuity of
$105
per year
(1
percent
x
10
years
x
$1,050).
The PV of these additional accrued benefits from factors
1
and
2
at the
end of the year is
$127.
This represents the nominal value of the newly
earned pension benefits, which is an annuity of
$265.50
per year starting
at
retirement.
Factor
3:
The
PV
of his previously accrued benefits increases by
10
percent from
$654
to
$719.40
because the date of their
eventual receipt has drawn one year closer.
As
a
result of
all
three factors, the nominal value of his pension wealth
increases from
$654
to
$846
and
its
real value to
$791.
Now let
us
refer to table
5.1
to see how these factors manifest them-
selves in the time pattern of benefit accrual in the no-inflation and in
the
7
percent inflation scenarios. The right-hand panel shows the con-
stant dollar present value of benefits attributable to continued work
with the same employer; these benefits are represented by factors
I
and
2
only. In the no-inflation case, there is no salary growth and hence
only factor
1
is at work. For each additional year of service an additional
deferred annuity of
$150
per year is earned. Note, however, that the
value of the incremental benefits earned at each age increases with age,
from
$455
(3.03
percent of salary)
at
age
30
to
$1,242 (8.28
percent of
salary) at age
64.
This
is
a reflection of the fact that the additional
$150
per year deferred life annuity has
a
higher PV the closer the employee
is to age
65.
The accrual of benefits under a
DB
plan is thus inherently
‘
‘backloaded
.’
’
For a fixed real interest rate, this backloading effect is much more
pronounced in the
7
percent inflation scenario because of the impact
of inflation on the nominal interest rate. In this case the constant-dollar
value of additional pension benefits earned increases from
$41 (.27
percent of salary) at age
30
to
$2,794 (18.63
percent of salary)
at
age
64.
In contrast, backloading or frontloading in
DC
plans is independent
of inflation as well as interest rates3 This is because employers can
achieve any backloading pattern by simply choosing an appropriate
pattern of contribution rates over the course of the employee’s career.
The left-hand panel of table
5.1
illustrates the effect of inflation on the
144
Zvi
BodielAlan
J.
MarcudRobert
C.
Merton
PV
of total accumulated pension benefits under the DB plan assuming
no real salary growth.
5.1.3
Funding
As
mentioned before, DC plans are by their nature fully funded, that
is, the market value of the plan’s assets equals the liability of the
sponsor to the plan’s beneficiaries. In sharp contrast, the calculation
of the funding status of DB plans is complex and controversial. If the
plan’s assets are invested in traded securities, their market value is
relatively easy to ascertain. The source of difficulty is in measuring
the sponsor’s liability.
From a strictly legal point of view the sponsor’s liability is the present
value of the accrued vested benefits which would be payable if the plan
were immediately terminated. But many pension experts contend that
sponsors have an implicit semicontractual obligation which makes it
more appropriate to take account of projected future salary growth in
the computation of the firm’s pension liability. The contention of a
further obligation beyond the legal one makes it unclear whether a real
or nominal interest rate should be used in discounting future benefits
(either with or without salary growth projections) to compute their
present value.
To
evaluate the strict obligation of the sponsor, the DB
liabilities could be determined by deriving the cost of an immunized
or dedicated bond portfolio using current market prices. While clearly
superior to a simple interest rate assumption, this valuation procedure
is itself only an approximation because the payment dates of pension
liabilities typically extend far beyond the maturity range that is rich
enough to extract our discount bond prices from traded coupon bonds.
Hence, an exact bond-dedication scheme is not feasible. Immunization
techniques that rely on duration measures are not wholly reliable be-
cause duration measures are sensitive to the specification of term struc-
ture dynamics. (See Bierwag 1977, Bierwag and Kaufman 1977, and
Cox, Ingersoll, and
Ross
1979). Beyond the term structure, the default
risk associated with partially funded pension obligations adds the fur-
ther problem of choosing equivalent-risk bonds from the securities
market.
For the past several years the Financial Accounting Standard Board
(FASB) has been grappling with these issues, trying to establish a uni-
form set of valuation standards for firms to use in their financial
statements.
The government guarantees, up to a limit, employer pension benefits
through the PBGC. The valuation of guaranteed benefits therefore should
utilize the riskless-in-terms-of-default interest rate. However, in prac-
tice, only
80
percent of accrued benefits is vested while only 90-95
percent of vested benefits is guaranteed
so
that roughly one-quarter of
145
Defined Benefit versus Defined Contribution Pension Plans
accrued benefits is not guaranteed (Amoroso
1982).
Thus, the funding
status of a plan is important to employees as well as to the PBGC.
In
effect, adequate funding protects accrued-but-not-yet-vested benefits.
See Marcus
(1987)
for an analysis of PBGC insurance and corporate
funding policy.
5.2
Trade-offs
Our
original belief was that defined contribution plans would nec-
essarily dominate defined benefit plans because of the flexibility of DC
plan design. We would have guessed that anything that could be ac-
complished with
a
DB plan could be replicated
in
a
cleverly constructed
DC plan. However, this belief is not borne out. DB plans create implicit
securities that can be welfare improving and that are neither currently
available in capital markets, nor likely to be created in capital markets
in the future. Some examples of these “securities” are factor-share
claims, price-indexed claims, and perhaps deferred life annuities at fair
interest rates.
Moreover, some of the “real-world’
’
complications in plan design,
such as incentive effects, tend to favor DB over DC plans. Thus, the
optimal plan design is likely to be firm specific. At this point, all we
can do is enumerate the relative advantages
of
each plan type and
describe the circumstances in which one plan might dominate.
5.2.1
Investment Performance and Choice
The most obvious source of risk to an employee in the DC plan is
the investment performance of the fund. However, this source of un-
certainty can be controlled.
For
example, the periodic contributions
of the DC plan could,
in
principle, be used to purchase deferred an-
nuities which would generate retirement income streams similar to
those provided by DB plans. Alternatively,
it
is feasible for the plan
to
select an investment strategy with low variance rates of
real
returns.
Bodie
(1980)
has shown that commodity futures can be added to port-
folios to successfully provide an effective hedge against inflation.
Therefore,
in
either nominal
or
real terms, DC plans do not necessarily
impose substantial risk on participants, given the availability of low-
variance investment strategies.
There are, however, no strong
a
priori
reasons to believe that most
individuals would choose to invest accumulated DC funds in the lowest
risk asset. DC plans typically offer sufficient flexibility to select
a
risk-
return strategy suited to the employee’s individual preferences and
circumstances. In contrast, DB plans force individuals to accumulate
the pension portion of retirement saving in the form of deferred life
annuities and thus limit the risk-return choice.
146
Zvi BodieIAlan
J.
MarcudRobert
C.
Merton
5.2.2
Accrual Patterns
As
noted and illustrated in table
5.
I,
DB plans are inherently back-
loaded. DC plans can be backloaded too by choosing
a
contribution
rate that rises with a worker’s age and ten~re.~ Therefore, the salient
inherent difference in accrual patterns between the two plan designs
is that DB backloading is stochastic in the sense that real benefit ac-
cruals depend upon the rate of wage inflation. This seems to
us
an
avoidable source of uncertainty which both parties (employer and em-
ployee) might benefit by shedding. On this score, DC plans would
appear to be superior, although implicit contracting to provide em-
ployees with aprotective “wage floor” (cf. Diamond and Mirrlees
1985)
can be implemented more effectively with DB-type plans.
5.2.3
Termination and Portability
It is commonly asserted that considerations of portability favor DC
plans. The typical justification is that the worker in a DB plan who
leaves his job for reasons beyond his control forfeits future indexation
of benefits already accrued. It is further asserted that there are implicit
contracts between employees and firms which require larger total com-
pensation (wage
plus
pension accrual) for more highly tenured workers.
Hence, termination of employment causes
a
forfeiture of the ability to
work for advantageous total compensation rates (and, in particular,
indexation of total pension accruals). Under this line of reasoning, DC
plans are more portable.
It should be realized, however, that the portability issue is intimately
tied to the accrual pattern. For DC plans with contribution rates tied
to tenure as well as age, the penalty to early termination can be as
great as for any DB plan. In practice, however, contribution rates for
DC plans are rarely tied to tenure and are usually not as heavily back-
loaded as DB plans. Therefore, in practice it would appear that port-
ability considerations do favor DC plans over DB plans.
5.2.4 Incentives
Pension benefits in DC plans depend upon the wage trajectory over
the worker’s entire career. In contrast, benefits in most DB plans de-
pend on final average salary. For this reason, workers in DB plans
should have a greater incentive to sustain a high level of effort over
the entire career in order to achieve a high career-end salary. Final
salary has greater leverage
in
DB plans because of its greater effect on
pension benefits.
In conclusion, it seems that there is
a
trade-off between the goals
of
portability and incentives. Portability dictates low backloading, while
incentives require high backloading. While DC plans opt in practice
147
Defined Benefit
versus
Defined Contribution Pension Plans
for lower backloading than DB plans, this pattern is not an inherent
property of the two plans.
5.2.5
Informational Economies in Plan Design and Implementation
Retirement income planning is one
of
the most complex areas of
personal finance. Many employees would consider it
a
service to have
their employer define and provide an adequate level of savings for them.
Since retirement-income goals are typically defined as percentage re-
placement rates of salary, the benefits of DB plans which are defined
in exactly those terms are easier to interpret.
One could in principle achieve the goal of a specific replacement rate
with a
DC
plan of the so-called target benefit type. Under these plans,
the contribution rate is adjusted periodically to achieve the target re-
placement rate, taking into account the discrepancy between actual
and assumed investment return. However, such plans are rare.
5.2.6 Wage-Path Risk
The pegging of benefits in DB plans to final average wage would
appear to provide employees with a type of income-maintenance in-
surance not available in
DC
plans. This observation has been used to
support the selection of these plans over DC plans. This conclusion
is, however, not robust. If wage paths are unpredictable at the start of
a career, then individuals may view it as very risky to have their re-
tirement benefits depend
so
heavily on final salary. Indeed, employees
might prefer
a
retirement benefit tied to (inflation-adjusted) career-
average earnings
so
as to eliminate excessive dependence on the re-
alized wage in the final years of employment. This time-averaging fea-
ture is achieved by a DC plan because benefits will depend on the
contribution in each year of service, rather than on a final wage formula.
Although inflation-adjusted career-average DB plans would achieve the
same goal, in practice these plans are quite rare. In fact, the only major
DB plan that pays a benefit computed in such a fashion is the Social
Security system. We pursue this issue further in the analysis in
section
5.3.
5.2.7 Interest-Rate Risk
As
noted earlier, one major source of uncertainty in
DC
plans con-
cerns the terms under which the stock of retirement wealth can be
transformed into a flow of retirement income. DB plans, by offering
life annuities, effectively guarantee the interest rate at retirement. It
should be noted, however, that without indexation of benefits, this is
a
guarantee of the nominal rather than the real interest rate. The value
to the employee of a nominal-rate guarantee is questionable when in-
flation over a
10-
or 20-year period can be highly unpredictable.
148
Zvi BodieIAlan
J.
MarcudRobert
C.
Merton
In principle, DC plans can offer at retirement the same nominal
interest rate guarantee through the purchase of deferred life annuities
as a DB plan. However, in practice, with the notable exception of the
Teacher’s Insurance and Annuity Association (TIAA), the capitaliza-
tion rates used to compute benefits in the private annuity market are
far below the interest rates available in competitive financial markets.
This discrepancy is often attributed to an adverse selection problem
and discourages participation in the annuity market by unhealthy in-
dividual~.~ The adverse selection issue is largely avoided in DB plans
because workers are precommitted to participation regardless of health
status.
5.3
A
Model
of
Wage and Interest Rate Uncertainty
In this section we develop a model to focus on the twin issues of
wage and interest rate uncertainty using stylized versions of DB and
DC plans. We find that the putative replacement rate advantages of DB
plans are not supported by our model, and that the interest rate guar-
antee is only partially supported: specifically, DB plans do offer welfare-
improving opportunities with respect to postretirement interest rate
uncertainty, but
not
with respect to preretirement uncertainty.
For the most part, we will concentrate on individual welfare in a
model in which all wage uncertainty is employee-specific and, from
the firm’s perspective, is perfectly diversifiable. This framework is at
a
polar extreme from Merton’s (1983) model of Social Security, in which
all uncertainty regarding marginal product derives from uncertainty in
the aggregate production function, with no individual-specific effects.
In Merton’s framework, labor-income uncertainty is perfectly corre-
lated across individuals, and in such an environment, DB plans may
offer superior risk-sharing properties that are not captured in our model.
Although our model focuses exclusively on uncertainty at the individual
worker’s level and interest rate risk, we will discuss further the im-
plications of Merton’s model for
our
results.
As
indicated earlier, in-
terest rate uncertainty emerges as
a
central determinant of the relative
advantages of DB versus DC plans.
5.3.1 Pension Plan Design
We consider a 3-period model in which the individual works in pe-
riods
0
and
l
and is retired in period
2.
Current wage,
W,
is known,
while period-1 wage,
W,,
is uncertain until
t
=
1.
For simplicity, we
will assume that the time
0
expectation of
W1
is
W,,.
Trends in wage
paths could easily be incorporated into the analysis, but would simply
clutter the algebra; hence we ignore such trends. Wages are measured
in real dollars
as
of time
0.
149
Defined Benefit versus Defined Contribution Pension Plans
Consumption occurs at three points:
t
=
0,
1,
2.
A
pension benefit,
P,
is paid at
t
=
2.
The real interest rate prevailing between
t
=
(0,
1)
is denoted
r,,
and is known at time
0.
The real rate between
t
=
(1,
2)
is
rl
and is not known until time
1.
Finally, we assume that individuals
have initial nonhuman wealth of
Ao.
The timing assumptions of the
model are presented in figure
5.1.
If financial markets were complete, then, of course, the choice of
pension plan would be irrelevant because the employee could use se-
curities to trade to an optimal position. There are two important de-
viations from complete markets that make pension design crucial from
the employee’s perspective. First, there are neither markets in which
wage uncertainty can be insured, nor ones in which claims to future
wages can be sold. This feature of our model precludes employee-
initiated risk pooling. Second, because of adverse selection problems,
the market for deferred life annuities is assumed to be closed. Although
such markets do in fact exist, as discussed, the rates of return typically
offered are
so
low as to discourage widespread participation. In our
model, the absence of such annuities will be captured by not allowing
individuals to invest at
t
=
0
in two-period bonds which pay specified
returns during the retirement period,
t
=
2.
The goal of the firm is to offer
a
pension plan that maximizes the
utility of
a
“typical” worker, subject to the constraint that all pension
plans considered have equal present value of costs to the firm. Subject
to the firm’s indifference condition, we compare the utility value of
DB versus DC plans.
In DB plans, firms typically promise workers a prespecified fraction
of career-end wages, possibly averaged over the last several years of
working life, and this is the type of plan we model. We will assume
that the pension benefit at
t
=
2
equals
W,,
so
that expected income
in each period of life is equal. We assume further that pension benefits
are explicitly linked to the price level. While this practice is uncommon
in the private sector in the United States, it is true of Social Security
and it serves as a useful base case from which to analyze the potential
efficacy of competing pension designs.
Time
Income
Consumption
Financial
Wealth
Interest Rate
Fig.
5.1
Timing Assumptions
150
Zvi BodieIAlan
J.
MarcudRobert
C.
Merton
The present value at
t
=
0
of the firm’s time-2 pension obligations
is
(1)
PVDB
=
E(Wl)B(O,
2),
where
B(0,
2) is the present value at
t
=
0
of a claim to an expected
payoff of
$1
at
t
=
2, with an uncertainty equivalent to that of the
wage distribution. If wage uncertainty were completely diversifiable,
then
B(0,
2) would equal the present value of
a
certain dollar to be
received in two periods;
B(t,
T)
would be the discount function at
t
for
payments at
T.
However, for the moment, we will not restrict the nature
of wage uncertainty.
In contrast to
DB
plans, DC plans require firms to contribute
a
prespecified fraction of wages into the worker’s retirement saving ac-
count each period. For simplicity we will assume that explicit wages
paid in each period are the same for each type
of
pension plan provided.
Hence, the indifference condition for the firm is that the present value
of periodic contributions into the DC plan equals the present value of
the DB commitment. The prespecified (at
t
=
0)
DC contribution sched-
ule
is
set at time
0
and therefore can depend only on observed variables
at
t
=
0.
While the contribution rates may depend on expectations of
future interest rates, they cannot be updated
ex
post
facto
to reflect
realizations of interest rates or any other factor.
There is an infinite number of DC contribution schedules which have
the same PV. Among these, we will select the one which has the same
timing pattern as the PV
of
accruing benefits under the DB plan.
The contribution schedule,
k,,
as a fraction of wages is given by:
1
k
-
-B(O,
2);
O-2
(2)
t
=
0,
t
=
I.
1
2
kl
=
-B(O,
2)/B(O,
1);
The present value at
I
=
0
of the
DC
plan contribution equals
PVDC
=
koWo
+
klEo(Wl)B(O,
1)
1
1
2
=
-WOB(O,
2)
+
;E.(WI)B(O,
2).
Since
Eo(Wl)
=
Wo, the present values of the firm’s contributions in
the DB and DC plans are equal.
Notice that since
B(0,
I)
is less than
I,
the DC plan as specified above
embodies some backloading. In fact, any degree of prespecified back-
loading may be built into the DC plan simply by changing the coeffi-
cients in equation (2) from their values of 1/2.
Any
coefficient pair for
151
Defined Benefit versus Defined Contribution Pension Plans
kl
and
k2
that sums to
1
will ensure that the present value of the DC
plan equals the present value of the DB plan.
The pension benefit in the DC plan will accumulate at
t
=
2
to
a
value that depends on the investment experience of the plan. Call the
rate of return on the pension portfolio in each period
zf
and let
2,
=
1
+
zr.
Then the pension benefit paid at
t
=
2
in the DC plan will be
(3)
whereas in the DB plan,
(4)
PDB
=
w,.
Notice that there is no assurance,
or
even likelihood, that the expected
pension benefits will be equal across the two plans, despite the fact
that the ex ante present values are equal.
5.3.2
Welfare Analysis
Pension benefits are subject to uncertainty from both stochastic wage
paths and stochastic investment returns. Rather than consider these
effects jointly, we will examine polar cases in which one
or
the other
source of uncertainty dominates.
1
1
2
PDC
=
-B(O,
2)W,Z,Z,
+
po,
2)/B(O, 1)W121,
Wage
Uncertainty
Consider first the case in which all investment returns can be made
certain by investing pension assets in default-free bonds. Therefore
both
r,,
and
r1
are known at
t
=
0.
Moreover, suppose for the moment,
that all wage uncertainty is perfectly diversifiable to the firm,
so
that
B(t,
Z‘)
is simply the discount function for riskless future cash flows.
Under these hypotheses,
2,
=
1
+
r,
=
R,
and
z1
=
1
+
rI
=
RI.
Further, with no uncertainty regarding the evolution of future interest
rates,
B(0,
1)
=
1/R,
and
B(0,
2)
=
l/RoRl.
Thus, equation
(3)
reduces
to
(3‘)
1
2
PDC
=
-(W,,
+
WJ.
In this simple case, it is clear that the DC plan must dominate the DB
plan for any risk-averse utility function. With
E,(W,)
=
W,,
both plans
have equal expected benefits, while the DC plan imposes less uncer-
tainty on participants because
of
the “wage averaging” embodied in
equation
(3’).
Essentially the only uncertainty in this case derives from
W1.
The DC plan allows
for
limited risk pooling of wage uncertainty
152
Zvi
BodieIAlan
J.
MarcudRobert
C.
Merton
through the firm (and ultimately the stock market), while the DB plan
allows for none. This advantage of DC plans may be thought of as a
pure efficiency gain.
The advantage of DC plans in the wage-uncertainty-only scenario
does not hinge solely on the diversifiability of wage risk. Suppose that
final wage is highly correlated with some marketable security, such as
the value of the stock of the firm or the value of a broad market index.
In this case, the DB plan implicitly forces the participant to invest
a
large fraction of wealth in this asset, since the pension benefit essen-
tially duplicates the payoff to the asset. In contrast, the DC plan allows
the participant to take the pension contribution each period and invest
it in any security. In essence, the DC plan allows participants to get
their money out of the (0ver)investment in
W,
and achieve superior
portfolio diversification. This advantage of DC plans is incremental to
the pure efficiency gain from the risk pooling opportunity that was
noted above.
Interest Rate Uncertainty
In this section, we will assume that wage paths are either given
or
uncorrelated with the interest rate, and that the only investment ve-
hicles are bonds. However, the future path of interest rates is not known
at the time the pension contract is established. Because wages pose
no systematic risk,
B(t,
T)
is simply the riskless discount function, and
B(0,
1)
=
l/Ro.
As in Merton
(1983),
we will assume that the lifetime utility function
for the individual at time
0
is
(5)
At time
1,
all uncertainty
is
resolved since both
WI
and
Rl
(and hence
P)
are known. Lifetime utility at
t
=
1
is thus
uo
=
log
(Co)
+
E,[log(C,)
+
log(C2)I.
UI
=
log(C,)
+
log(C2),
u2
=
log(C2).
and at
t
=
2
is
Upon arriving at
t
=
2,
the individual will consume all of his financial
wealth, plus all pension benefits:
(6)
C2
=
(A,
+
Wl
-
CI)Rl
+
Z?
Thus, at
f
=
1, the optimization problem is
max[log(C1)
+
log(C,)l,
which results in the first-order condition
(7)
C1
-
CJRI
=
0.
153
Defined Benefit versus Defined Contribution Pension
Plans
Using equation
(6),
equation
(7)
can be solved to yield
C,*
=
(Al
+
W1
+
P/R1)/2,
C,*
=
RlCI*.
Using the expressions for
P
from equations
(3)
and
(4),
we find that
(8-DB)
CPB
=
(A,
+
W,
+
WI/R1)/2,
while
(8-DC)
Cpc
=
[A,
+
W1
+
(1/2)(WO
+
Wj)B(O,
2)RO]/2.
As expected, the difference between equations (8-DC) and (8-DB)
reflects the “wage diversification” attribute of DC plans, in that con-
sumption depends upon
a
weighted sum of earnings over the entire
career. A perhaps surprising feature of equations
(8)
is that consumption
for individuals in DC plans is not a function of the realized interest
rate, R1, although it is for individuals in DB plans. This is true despite
the fact that retirement
wealth
is subject to interest rate risk for DC
plans, but not for DB plans.
This feature of the model turns out to be an artifact of the log utility
function, but nevertheless highlights an important feature of DB versus
DC plan design. Recall the first-order condition
(7)
for optimal con-
sumption allocation across times 1 and 2, which requires that time-2
consumption be R, times time-1 consumption. For an individual in
a
DC plan, all wealth already is held and can be invested
at
rate
RI
at
t
=
1.
Thus, the simple rule is to consume one-half of wealth at
t
=
1,
invest the remainder, and thus consume RI times one-half of wealth at
t
=
2. Consumption at
t
=
1
is thus
independent
of R,. In contrast,
in a DB plan, the pension benefit to be received at
t
=
2 already is
fixed at
t
=
1. Thus, a large value of R1 requires
a
decrease in
t
=
1
consumption in order to satisfy the first-order condition for an optimum.
Another way of seeing this is to note that, for the log utility function,
consumption at
t
=
1 depends
only
on wealth, not on the interest rate.
For DC plans, wealth at
t
=
1
is
independent of R,, since all assets
are already in hand. For DB plans, pension benefits are still deferred
at
t
=
1,
and wealth depends on R,.
For more general utility functions, consumption at
t
=
1
depends
on both wealth
and
R,. However, DC plans still offer a type of con-
sumption smoothing that is not offered by DB plans. Specifically, the
generalized first-order condition at
t
=
1
requires that the ratio of the
marginal utility of consumption at
t
=
1
to that at
t
=
2 equals RI. A
larger R, thus induces more time-2 consumption. This can be attained
with less (or no) sacrifice of current consumption when assets are
already in hand since assets currently invested can earn the higher rate
of interest. In DB plans, in contrast, there is no offset between income
154
Zvi Bodie/Alan
J.
MarcudRobert
C.
Merton
and substitution effects. A larger
R,
decreases
pension wealth and,
simultaneously, requires a reallocation of consumption to the retire-
ment period,
t
=
2. Thus, the consumption stream in
DC
plans is less
sensitive to the interest rate during the accumulation phase, and indeed,
in the log utility case, is actually independent of the realization of the
interest rate.
Using equations
(S),
we may now compute the derived or indirect
utility function at
t
=
1:
(9-DB)
JDB(A1,
Wl,
t
=
1)
=
lOg(C,*)
+
lOg(C,*)
=
2
log[l/2(A1
+
W1
+
Wl/Rl)]
+
log(R1);
(9-DC)
JDc(A1,
W1,
t
=
1)
=
2 10g{l/2[A,
+
W1
+
W’B(0,
2)RJ)
+
log(R,);
where
W‘
=
(W,
+
WJ2, that is, career-average earnings.
As
a base case to compare equations
(9),
consider the situation in
which the expectations hypothesis for the term structure of interest
rates holds. ThenB(0, 2)
=
(l/Z?,)Eo(l/Rl). In this instance, with
Eo(W’)
=
W,
and
W,
uncorrelated with
R1,
the expectations of the arguments
of the log terms in equations
(9)
are equal. However, the argument of
the log term in
(9-DC)
is subject to less uncertainty (as oft
=
0)
than
in
(9-DB).
This is due to both the wage diversification embodied in the
DC
plan and the interest rate risk that appears only in the
DB
plan.
Using equations
(9),
we may obtain the derived utility function at
r
=
0:
(10)
J(A,, W,,
t
=
0)
=
max{log(C,)
+
EO
[J(Al,
WI,
t
=
I)]}.
From equation (lo), it is easy to show that time-0 utility is higher in
the
DC
plan (still assuming that the expectations hypothesis holds).
Consider the optimizing value of time-0 consumption under the
DB
plan. This consumption choice is also feasible in the
DC
plan and will
result in an identical value for
Al
.
However, for any given A
,,
E,[J(A
,,
W1,
t
=
l)]
is greater in the
DC
plan. This last point follows from the
equal expected values of the arguments
of
the log function in equations
(9),
the greater dispersion
of
the argument in the
DB
plan, and the
concavity of the log function. Because
DC
plans offer greater welfare
than
DB
plans at consumption levels that are optimal for
DB
plans,
they must do
so
afortiori
when
C,
is chosen to be optimal for the
DC
environment.
For the
DB
plan to dominate the
DC
plan, it would be necessary for
it to offer
a
greater
expected pension benefit at
r
=
2.
This would
155
Defined Benefit versus Defined
Contribution
Pension Plans
require that
B(0,
2)
be less than
E,,(l/Z?&,),
that is, that there be
a
positive liquidity
or
risk premium for investing in long-term bonds
rather than rolling over shorts.
At this point,
it
is worth reconsidering the assumptions of our model.
It should be apparent that the zero expected growth rate of real wages
is not essential to the argument. Our analysis would have been similar
even with
a
positive trend in real wages. The only major modification
would involve an adjustment for the fact that a DB plan with a
100
percent replacement rate of final salary would promise retirement-period
income greater than career-average wages. The per period contribu-
tions to the retirement fund in the equal-present-value DC plan would
thus need to be correspondingly increased. In the nomenclature of
equation
(2),
the sum of
k,
and
k2
would need to exceed
1
.O.
However,
aside from this adjustment, the analysis would be similar.
The issue of interest rate uncertainty during the retirement period is
more difficult and poses issues not easily treated in the above model.
In
our
3-period model, the individual simply consumes total retirement
wealth in the last period. If, however, retirement itself is viewed as
a
many-period interval, then real retirement
income
and not
wealth
may
be the significant determinant of welfare. Given a stock of wealth at
retirement, the real consumption stream that is feasible for the retiree
depends on the real long-term interest rate at the time of retirement,
when the purchase of a (real) life annuity is contemplated. Even if
retirement wealth can be predicted fairly precisely with
a
low-
investment-risk DC retirement fund, the real income stream that can
be generated by that wealth
is
subject to considerable uncertainty.6 In
contrast, by guaranteeing a specified income (and hence, consumption)
stream upon retirement, the (price-level-indexed)
DB
plan eliminates
the risk associated with the conversion, at retirement, of a stock
of
retirement wealth into a flow of equivalent-present-value consumption.
DC plans cannot offer a guaranteed capitalization rate at retirement
because of our assumption that life annuities and bonds
of
long- enough
maturity do not exist.
In order to examine some potential effects of uncertainty in the
interest rate at retirement, we will consider
a
simple adjustment to our
model. Suppose that at
t
=
2,
the financial assets of individuals are
multiplied by some increasing function of
R2,
f(R2),
where
R2
equals
one plus the postretirement rate of interest. The multiplication byf(R2)
reflects the increased retirement-income stream that is available to DC
participants when interest rates at retirement turn out to be high. In
contrast, for DB plans, the retirement-income stream is guaranteed by
the firm
so
that interest-rate risk is not borne by plan participants.
Reconsider now the optimal consumption program for DC plan par-
ticipants. At
t
=
2,
156
Zvi
BodieIAlan
J.
MarcudRobert
C.
Merton
(1
1)
which now is stochastic at
t
=
1
because of the dependence on
RZ.
Thus, at
t
=
1,
the maximization problem becomes
C2
=
[(A,
+
Wi
-
CI)R,
+
f'~clf(Rd7
max log(C0
+
E,tlog(C2)1,
which has first-order condition
But examination of equations (11) and (12) shows that
AR,)
drops
out of the first-order conditions
so
that (12) results in exactly the same
consumption level at
t
=
1
as in the nonstochastic
R2
model. Lifetime
utility, however, may change. For example, for an actuarially fair
AR,)
adjustment, such as
AR,
=
R2/E,(R2),7
consumption
at
t
=
1 is un-
changed, while consumption at
t
=
2 has the same expected value as
in the previous model, but greater uncertainty. In this case, expected
time-2 utility falls. If time-2 interest rate uncertainty is sufficiently great
relative to wage and time-1 interest rate uncertainty, DC plans could
become inferior to DB plans from the viewpoint of plan participants.
Thus, retirement-period interest rate uncertainty emerges as a potential
advantage
of
DB relative to DC plans.
5.3.3 Factor-Share Uncertainty
Merton (1983) has examined a model in which labor-income uncer-
tainty derives entirely from an aggregate production function in which
income shares accruing to capital and labor are stochastically deter-
mined. In contrast to the model above, in which labor-income uncer-
tainty is diversifiable, in Merton's model labor income is perfectly
correlated across individuals. Given the nontradeability of human cap-
ital, economic inefficiencies arise in this economy, since early in life,
individuals hold too much of their wealth in human capital relative to
physical capital, while at retirement all wealth is invested in physical
capital. These portfolio imbalances preclude optimal sharing of factor-
share risk. Merton suggests that a Social Security system which pays
retirees a share of current wage income implicitly provides diversifi-
cation across factor shares and can increase welfare by improving the
efficiency of risk bearing in the economy.
A
similar argument can be made with regard to DB versus DC plans.
In a DC plan, the income of a retired individual depends solely on
investment performance and is independent of retirement-period un-
certainty in factor shares. Retirees thus have no stake in labor income
during their retirement period. In a DB plan, retirement income is also
determined upon retirement. However, if factor-share uncertainty is
157
Defined Benefit versus Defined Contribution Pension Plans
primarily attributable to unforeseeable long-term secular trends (rather
than to transitory business-cycle effects) then a final-salary DB plan
may provide risk-sharing benefits similar to Merton’s Social Security
scheme. Such secular uncertainty could arise, for example, from un-
anticipated changes in labor-augmenting technical progress.
Since the pension benefit under the DB plan is tied to final salary,
individuals participating in such a scheme are invested in an implicit
security that is tied to the wage share in the neighborhood of the
retirement period.
To
the extent that firms offer ad hoc increases in
pension benefits when wages of current employees increase, the re-
tiree’s stake in aggregate labor income
is
further enhanced. Of course
DC plan benefits also depend to some extent on end-of-career earnings.
However, the career averaging properties
of
DC plans greatly reduce
the magnitude of this dependence. Thus, if labor-income uncertainty
is predominantly dependent on economy-wide factors, then this source
of risk would favor DB over DC plans.
5.3.4
Inflation
In the preceding model, we assumed that wages and pension benefits
were all contracted in real terms. It is clear that the vast majority of
DB plans as currently implemented are not contractually indexed dur-
ing the retirement period. This weakens the case for viewing DB plans
as offering income-maintenance or interest-rate insurance.
Moreover, there is controversy surrounding the degree
of
indexation
during the worker’s active life. Bulow
(1982)
has argued that wages in
firms administering DB plans should
not
be expected to keep pace with
the price level. His argument is based on the notion that labor markets
clear as spot markets (with respect to pension issues) and that any
implicit contracts between firms and workers are independent of pen-
sion issues. In this case, the market-clearing employee compensation
will determine the
sum
of wages plus accruing pension benefits. The
level of either wages or pension accruals alone, however, is
indeterminate.
To
illustrate Bulow’s point, consider the effects of an unanticipated
increase in the price level. The increase imposes a real loss on workers,
since their pension benefits are defined in nominal terms. Of course,
the worker’s
loss
is the firm’s gain.
If,
however, the employees were
to receive a pay raise in the subsequent period which would keep
their real wage constant, then the earnings base upon which pension
benefits are calculated also would rise
at
the inflation rate, and the
worker’s pension
loss
would be eliminated. Real compensation
in
the
second period would in effect be higher than in the first: real wages
are constant, but pension transfers have increased in order to com-
pensate for the effects of the unanticipated inflation. The firm has, in
158
Zvi BodieIAlan
J.
MarcudRobert
C.
Merton
effect, issued insurance against the effect of inflation on the value of
pension benefits.
Bulow argues that firms neither behave in this way nor should they
be expected to. His competing model holds total real compensation
exogenous. Because pension benefits in DB plans increase with the
wage level, the wage component of compensation will
not
rise at the
inflation rate in the subsequent period. Instead, the sum of the partially
indexed wage increase and partially indexed recovery of real pension
benefits
together
will provide
an
increase in nominal compensation
which matches the inflation rate. However, the initial loss of pension
value due to inflation is borne entirely by the worker.
Under the Bulow model, DB plans pose significant risk to partici-
pants. The nominal nature of the pension contract is to be taken quite
seriously; workers bear the entire brunt of inflation risk. Thus, while
DB plans provide a less variable final-salary replacement rate to work-
ers than do DC plans, the final real salary itself becomes more sensitive
to inflation. Whether DB
or
DC plans are riskier in
a
utility sense is
therefore an open question. Bulow’s model is far from universally
accepted. Several observers (e.g., Cohn and Modigliani
1983)
believe
that firms do in fact offer implicit indexation to workers. In this view,
the wage decision is made separately from the pension decision, and
the effects of wage increases on pension benefits are ignored in the
determination of worker compensation.
5.4
Concluding Comments:
Is
There
a
Better
Way?
The major advantage of DB plans is the potential they offer to provide
a stable replacement rate of final income to workers.
If
the replacement
rate is the relevant variable for worker retirement
utility,
then DB plans
offer some degree of insurance against real wage risk. Of course, pro-
tection offered to workers
is
risk borne by the firm.
As
real wages
change, funding rates must correspondingly adjust. However, to the
extent that real wage risk is largely diversifiable to employers, and
nondiversifiable to employees, the replacement rate stability should be
viewed as an advantage of DB plans.
The advantages of DC plans are most apparent during periods of
inflation uncertainty. These are: the predictability of the value of pen-
sion wealth, the ability to invest in inflation-hedged portfolios rather
than nominal DB annuities, and the fully-funded nature of the DC plan.
Finally, the DC plan has the advantage that workers can more easily
determine the true present value of the pension benefit they earn in
any year, although they may have more uncertainty about future pen-
sion benefit flows
at
retirement. Measuring the present value of accruing
159
Defined Benefit versus Defined Contribution Pension Plans
defined benefits
is
difficult
at
best and imposes severe informational
requirements on workers. Such difficulties could lead workers to mis-
value their total compensation and result in misinformed behavior.8
Of
interest for future research is the possibility
of
pension plan de-
signs that combine the best attributes of DB and DC plans. Many firms
already offer
DB
plans supplemented by DC plans. An interesting al-
ternative
is
the so-called floor plan, which is in essence
a
DC plan
together with a guarantee
of
a minimum retirement income based on
a DB-type formula. Employers and employees can trade
off
the level
of guaranteed floor against the size of the expected DC benefit. These
plans offer the downside protection of DB plans, yet still allow em-
ployees to take positions in high-expected-return assets. Floor plans
already are offered by some firms and allow for
a
great deal of flex-
ibility and creativity.
Notes
1.
It is important to distinguish here between several subcategories
of
DC
plans: money purchase, profit sharing, and thrift plans.
For
money purchase
plans, like TIAA-CREF, contributions are usually based on the employee’s
compensation, as stated in the text. But in profit-sharing plans employer con-
tributions are based
on
the sponsor’s profitability, and
in
thrift plans contri-
bution levels are usually determined voluntarily by employees, with employer
matching contributions at some prespecified rate. Thrift plans are usually
of-
fered as a supplement to a DB
or
other DC plan.
2.
Until the late 1970s, employee contributions to many DC plans were not
tax-deductible, the main exception being employees of certain nonprofit
or-
ganizations (403[b] plans).
But
recently the government has expanded tax-
deductibility of employee contributions to the private for-profit sector through
401(k) plans.
3.
There is a separate question
of
whether the difference in backloading
patterns is of importance to workers. Consider
a
scenario in which the inflation
rate is fixed and only the interest rate varies.
In
this case, the impact of interest
rates
on
accrual patterns would be irrelevant to workers from a welfare stand-
point. The real stream of benefits to be paid starting at retirement is independent
of the trajectory
of
the present value
of
accrued benefits. When inflation rates
are stochastic, however, backloading patterns can have important effects on
welfare. The real benefit stream during retirement moves inversely with the
stochastic price level.
4. The contribution pattern for a DC plan required to match the accrual
pattern of
a
DB plan could
run
into IRS limits on annual contributions at older
ages, particularly for higher paid employees.
5.
An alternative explanation is that insurance companies view their an-
nuitants as members of
a
captive market and try to recoup past losses by
offering them below-market rates.
160
Zvi BodieIAlan
J.
MarcusIRobert
C.
Merton
6. As always, it is impossible to tell from first principles of welfare analysis
whether an individual would necessarily choose to convert wealth into
a
risk-
less stream of retirement benefits.
7. This is an
actuariafly
fair adjustment
in
the sense that the expected value
of period-2 income would be unaffected.
8. While workers are more likely to be informationally disadvantaged than
employers, the
level
of complication is such that employers also may make
significant mistakes. All of this is perhaps an issue in evaluating the Bulow
argument, since that argument turns
on
accurate perceptions of the “true”
pension benefits and costs.
9. Among the companies offering floor plans are Xerox, Hewlett-Packard,
and Georgia-Pacific.
References
Amoroso, Vincent. 1983. Termination insurance for single-employer pension
plans:
Costs and benefits.
Transactions, Society
oj
Actuaries
35:71-83.
Bierwag,
G.
0.
1977. Immunization, duration and the term structure of interest
rates.
Journul of Financial and Quantitative Analysis
(December): 725-42.
Bierwag,
G.
O.,
and George
G.
Kaufman. 1977. Coping with the risk of interest
rate fluctuations.
Journal
of
Business
(July): 364-70.
Bodie, Zvi. 1980. An innovation for stable real retirement income.
Journal of
Portfolio Management
7 (Fall): 5-13.
Bulow, Jeremy
I.
1982. What are corporate pension liabilities?
Quarterly Jour-
nal
of
Economics
97:435-52.
Cohn, Richard A., and Franco Modigliani. 1983. Inflation and corporate
fi-
nancial management. Mimeo.
Cox, John C., Jonathan E. Ingersoll, and Stephan A. Ross. Duration and the
measurement
of
basis risk.
Journal of Business
52
(January):
51
-61.
Diamond, Peter, and James Mirrlees. 1985. Insurance aspects
of
pensions.
In
Pensions, labor, and individual choice,
edited by David Wise. Chicago: Uni-
versity
of
Chicago Press.
Ellwood, David T. 1985. Pensions and the labor market: A starting point.
In
Pensions, labor, and individual choice,
edited by David Wise. Chicago: Uni-
versity of Chicago Press.
Marcus, Alan J. 1987. Corporate pension policy and the value
of
PBGC in-
surance.
In
Issues in Pension Economics,
edited by Zvi Bodie, John Shoven,
and David Wise. Chicago: University of Chicago Press.
Merton, Robert C. 1983.
On
the role
of
Social Security as
a
means for efficient
risk sharing in an economy where human capital is not tradeable.
Financial
aspects
of
the United States pension system,
edited by Zvi Bodie and John
Shoven. Chicago: University of Chicago Press.
Merton, Robert C., Zvi Bodie, and Alan J. Marcus. 1987. Pension plan inte-
gration as insurance against Social Security risk.
In
Issues in Pension Eco-
nomics,
edited by Z. Bodie, J. B. Shoven, and D. A. Wise. Chicago: Uni-
versity of Chicago Press.
161
Defined Benefit versus
Defined
Contribution Pension Plans
Comment
Laurence
J.
Kotlikoff
This paper provides
a
very insightful comparison of defined contri-
bution and defined benefit pension plans. While the authors are cau-
tious, one
is
left with the impression that the defined contribution form
of pension plans is superior in many, if not all, respects to the defined
benefit form.
I
certainly concur with that conclusion. Many defined
benefit plans appear to subject workers and employers to unnecessary
earnings risk by tying the pension payment to the average of earnings
at the end of workers’ careers; while hedging inflation, such provisions
mean that workers’ pensions are very sensitive to earnings late in their
careers. Such earnings may be unusually low for reasons including
poor health, changes in market conditions, and
so
forth.
Other defined benefit plans relate the pension to longer averages of
earnings, which make the initial real pension benefit potentially quite
sensitive to inflation. Still other defined benefit pensions are indepen-
dent of earnings, positing a nominal benefit that depends only on ser-
vice. The real values of these latter pensions are also very sensitive to
inflation.
In contrast to the defined benefit plans, defined contribution plans
appear to be riskier with respect to the real rate of return. However,
as the authors point out, one can devise close to riskless portfolios
that get around this objection. In addition, they make the important
point that defined benefit plans also are sensitive to the real rate of
return because changes in real rates alter the present value of future
defined benefits.
Once workers retire the defined benefit pensions are subject to con-
siderable inflation risk. While many firms do provide cost-of-living
increases on an ad hoc basis, these increases do not keep pace with
inflation, as Robert Clark has shown in his study of cost-of-living in-
creases in the
1970s.
In contrast, defined contribution plans give work-
ers the option of withdrawing their funds and investing them them-
selves.
As
mentioned, one can safely hedge inflation and secure a real,
if minuscule, rate of return; the problem is, however, that many retirees
may not know how to devise such riskless portfolios that involve using
future commodity markets. While financial markets could provide such
safe assets, they do not appear readily available at the current time.
A
problem with defined contribution plans not discussed in detail by
the authors is that defined contribution plans that pay off at retirement
do not provide retirees with an annuity, and therefore do not provide
Laurence
J.
Kotlikoff
is a professor
of
economics at Boston University and a research
associate
of
the National Bureau
of
Economic Research.
162
Zvi
BodieIAlan
J.
MarcudRobert
C.
Merton
retirees with insurance against life-span uncertainty. Those that do pay
off in the form of an annuity provide a stream of nominal retirement
benefits that are subject to inflation risk just like the benefits of defined
benefit plans.
Another important issue that the authors do not consider is whether
defined benefit plans are too complicated for workers-and, indeed,
even employers-to understand and properly evaluate. It is not atypical
to find a defined benefit plan that has
(1)
an age- and service-related
benefit formula,
(2)
an average earnings base,
(3)
age- and service-
dependent early retirement reduction formulas,
(4)
special early re-
tirement supplemental benefits, and
(5)
actuarial reductions for workers
terminating prior to early retirement.
To
calculate correctly one’s ac-
crual of pension benefits in such plans requires actuarial skills which
typical workers do not possess. In addition, in many cases even if the
workers possessed such skills, the booklets describing the pension
plans are
so
poorly written, if not intentionally misleading, that it is
very difficult to figure out what one is actually receiving. Hence, an
important advantage of defined contribution plans is that they provide
workers with better information about their retirement finances.
My guess is that defined benefit plans emerged because they were
attractive to older union members and to employers who thought they
could generate strong retirement incentives without being explicit about
those incentives. In the process the country has been straddled with
a
very risky private pension system that provides insufficient information
to both workers and employers about the benefits and costs of financing
retirement.
