FR3059_20180829_omb_B

Supporting Statement Part B for the
Survey of Consumer Finances
(FR 3059; OMB No. 7100-0287)
Summary
For all information collections that involve surveys or require a statistical methodology,
the Board of Governors of the Federal Reserve System (Board) is required to provide a complete
justification and explanation of the use of such a methodology. For collections that employ
surveys without such a methodology, the Board should be prepared to justify its decision not to
use statistical methods in any case where such methods might reduce burden or improve
accuracy of results.
Background
For many years, the Board has sponsored consumer surveys to obtain information on the
financial behavior of households. The 2019 Survey of Consumer Finances (SCF) would be the
latest in a triennial series, which began in 1983, that provides comprehensive data for U.S.
households on the distribution of assets and debts, along with related information and other data
items necessary for analyzing financial behavior. The SCF is the only survey conducted in the
United States that provides such financial data for a representative sample of all households.
In addition to providing baseline information for current analysis, data from earlier SCFs
have proved directly useful in policy work at the Board. For example, these surveys have been
used in Board briefings and numerous memoranda to examine wealth and income inequality,
changes in the distribution of debt burdens in the population, coverage of household deposits by
federal deposit insurance, ownership of mutual funds and stocks, automobile leasing, and many
other areas. The surveys have also been used extensively for longer-term research within the
Federal Reserve System, in the Office of Tax Analysis and other parts of the Treasury, in other
government agencies, in academia, in other research institutions, and in businesses.
Universe and Respondent Selection
For the survey, the contractor and the Board would draw samples for the survey designed
to obtain at most 7,000 completed interviews. The sample selected by the contractor would be
an area-probability design which is a type of geographically structured random sample that
provides good national coverage of widely-distributed behavior. The sample selected by the
Board would be drawn from statistical records derived from individual tax returns; this “list”
sample would be designed to oversample wealthy households. Both samples would be selected
in a way that is consistent with procedures used in earlier SCFs. When the two samples are
combined and sample weights applied, the survey is representative of the number of households
in the United States in the survey year. The 2016 SCF represented 126 million households.
The response rate in the area-probability sample is more than double that in the list
sample. In 2016, about 65 percent of households selected for the area-probability sample
actually completed interviews. The overall response rate in the list sample was about one-third;

in the part of the list sample likely containing the wealthiest households, the response rate was
only about one-half that level.
Procedures for Collecting Information
The survey is expected to provide a core set of data on family income, assets, and
liabilities. The major aspects of the sample design that address this requirement have been
largely constant since 1989. The SCF combines two techniques for random sampling. First, a
standard multistage area-probability sample (a geographically based random sample) is selected
to provide good coverage of characteristics, such as homeownership, that are broadly distributed
in the population.
Second, a supplemental sample is selected to disproportionately include wealthy
households, which hold a relatively large share of such thinly held assets as noncorporate
businesses and tax-exempt bonds. Called the “list sample,” this group is drawn from a list of
statistical records derived from tax returns. These records are used under strict rules governing
confidentiality, the rights of potential respondents to refuse participation in the survey, and the
types of information that can be made available. Persons listed by Forbes magazine as being
among the wealthiest 400 people in the United States are excluded from sampling.1
Of the 6,254 interviews completed for the most recent survey, the 2016 SCF, 4,754 were
from the area-probability sample, and 1,500 were from the list sample.
The SCF data are collected solely for statistical purposes under a pledge of
confidentiality. That is, no use is made of the data that would directly affect any survey
participant outside the administration of the survey. As required by the Confidential Information
Protection Act (when agencies other than the major statistical agencies use a contractor for data
collection), the survey information is processed within an officially designated statistical unit of
the government, which includes the Microeconomic Surveys section of the Division of Research
and Statistics.
Methods to Maximize Response
Various methods are used in the SCF to maximize response rates and deal with issues of
non-response bias and survey error. To provide a measure of the frequency with which
households similar to the sample households could be expected to be found in the population of
all households, an analysis weight is computed for each case, accounting both for the systematic
properties of the sample design and for differential patterns of nonresponse. The SCF response
rates are low by the standards of some other major government surveys, and analysis of the data
confirms that the tendency to refuse participation is highly correlated with net worth. However,
unlike other surveys, which almost certainly also have differential nonresponse by wealthy
households, the SCF has the means to adjust for such nonresponse. A major part of SCF

1

For more details about the SCF sample design, see Bricker, Jesse, Alice Henriques, and Kevin Moore (2017).
“Updates to the Sampling of Wealthy Households in the Survey of Consumer Finances,” Finance and Economics
Discussion Series 2017-114. Board of Governors of the Federal Reserve System (U.S.).

2

research is devoted to the evaluation of nonresponse and adjustments for nonresponse in the
analysis weights of the survey.2
Errors may be introduced into survey results at many stages. Sampling error—the
variability expected in estimates based on a sample instead of a census—is a particularly
important source of error. Such error can be reduced either by increasing the size of a sample or,
as is done in the SCF, by designing the sample to reduce important sources of variability.
Sampling error can be estimated, and the SCF uses replication methods to do so. Replication
methods draw samples, called replicates, from the set of actual respondents in a way that
incorporates the important dimensions of the original sample design. In the SCF, weights were
computed for all of the cases in each of the replicates.3 To estimate the overall standard error, a
measure of the variability of these estimates is combined with a measure of the variability
because of imputation for missing data.
Other errors include those that interviewers may introduce by failing to follow the survey
protocol or misunderstanding a respondent’s answers. SCF interviewers are given lengthy,
project-specific training and ongoing coaching to minimize such problems. Respondents may
introduce error by interpreting a question in a sense different from that intended by the survey.
For the SCF, extensive pretesting of questions and thorough review of the data tend to reduce
this source of error.
Item nonresponse, which is nonresponse to selected items within the survey, may be
another important source of error. To address missing information on individual questions
within the interview, the SCF uses statistical methods to impute missing data; the technique
makes multiple estimates of missing data to allow for an estimate of the uncertainty attributable
to this type of nonresponse.
Testing of Procedures
The major aspects of sample design for the SCF have remained unchanged since 1989,
but research on improving the sample design is ongoing. Prior the field period for the SCF,
which runs roughly from May to December of the survey year, there is extensive testing of the
survey computer-assisted personal interview (CAPI) questionnaire and all of the systems related
to data capture and processing. This testing culminates in a pretest of the survey instrument and
systems prior to the main data collection. For this pretest, the contractor would interview no
more than 150 households based on samples they would select. After the pretest, the contractor
would forward the data collected to the Board and would conduct a formal debriefing of
interviewers and others involved in the administration of the surveys. In response to issues that
emerge, the contractor would revise procedures for the main survey in consultation with the
Board.
2

The weights used in this article are adjusted for differential rates of nonresponse across groups. See Arthur B.
Kennickell (1999), “Revisions to the SCF Weighting Methodology: Accounting for Race/Ethnicity and
Homeownership” (Washington: Board of Governors of the Federal Reserve System, January),
www.federalreserve.gov/econres/scf_workingpapers.htm.
3
See Arthur B. Kennickell (2000), “Revisions to the Variance Estimation Procedure for the SCF” (Washington:
Board of Governors of the Federal Reserve System, October),
www.federalreserve.gov/econres/scf_workingpapers.htm.

3