licensed by King County as of late 2019. Appendix 1 provides more detailed information on the
fielding of the 53-question survey. The survey included questions about driving for a TNC,
vehicle expenses, and demographic characteristics. It asked respondents to enter summary
information from their Uber and/or Lyft app regarding their driving activity during the week of
December 2-8, 2019.
The survey elicited reasonable response rates. About two percent (734) of the emailed surveys
bounced back; 9,379 responses were received. We eliminated returned surveys with no
completed answers and used Qualtrics quality control metrics to screen for possibly fraudulent
responses and likely duplicates. After cleaning the survey sample there were 7,394 surveys. For
the first substantive question, “which company do you drive for?” there were 6,789 responses,
21.5 percent of the universe of driver email addresses to which the survey was sent. By
comparison, the driver surveys that informed Hall and Krueger’s 2018 article on Uber drivers
had response rates of around 10 percent. Predictably, response rates in the Seattle survey varied
across the survey, with response rates tapering off toward the end, depending on the topic and
ease of responding. In the discussion and exhibits that follow, we report the number of responses
for each question or set of questions analyzed.
The key question about “usual weekly working hours” elicited 6,554 responses (20.8 percent).
Questions on demographic characteristics, which appeared toward the end of the survey, elicited
responses from about 12 percent of the driver universe. We trimmed responses to questions
asking about earnings (and related data reported on drivers’ Uber and Lyft apps) and expenses to
eliminate reporting errors and other outliers. Some of the questions asked drivers to consult their
driving record using their Uber or Lyft app for the week of December 2-8, 2019. These questions
elicited a smaller but still substantial response rate: about 2,022 drivers (6.6 percent of the
sampling universe). The drivers who responded to these questions account for a disproportionate
share (13 percent or 92,000) of all Seattle-King County trips provided during the survey week.
Given the racial, ethnic, and linguistic diversity of the driver universe, the survey was translated
into seven languages (Spanish, Simplified and Traditional Chinese, and four East African
languages—Amharic, Oromo, Somali, and Tigrinya), with targeted outreach at community
locations to inform drivers about the survey.
We used a number of methods to check the representativeness of the survey. The first compares
the vehicles used by survey respondents to administrative records compiled by King County to
license vehicles for TNC services. We identified five broad types of vehicles—sedan, sedan
hybrid, SUV, SUV-hybrid, and electric. Exhibit 11 shows the distribution of vehicle types for the
set of TNC vehicles licensed in King County and the vehicles used by survey respondents.
Seventy-six percent of the set of King County licensed TNC drivers have a sedan or hybrid
sedan, compared to 72 percent of Seattle TNC survey respondents. Survey respondents and all
King County licenses reported similar shares of SUVs, 21 and 22 percent, respectively. The
percentage who drive a hybrid SUV or electric vehicle was very low in both groups.
Our second check on the representativeness of our survey respondents compared the
geographical location of driver residences in the King County administrative data and among our