2004 Survey of Consumer Finances
Summary Results
Three types of summary output are available on this page: a Federal Reserve Bulletin article; two sets of tables, in Excel format, that provide estimates comparable to the tables included in the Bulletin article, one using the internal SCF data and another using the public SCF data; and an Interactive Chart that includes time series charts using triennial SCF data covering the period 1989 to the most recent survey year.
Recent Changes in U.S. Family Finances: Evidence from the 2001 and 2004 Survey of Consumer Finances
Brian K. Bucks, Arthur B. Kennickell, and Kevin B. Moore
Federal Reserve Bulletin, vol. 92 (February 2006), pp. A1-A38.
Article (448 KB PDF) | Article (3.03 MB Postscript)
Erratum
Documentation
Three types of technical documentation for the 2004 SCF are available on this page: the 2004 codebook, the household information packet (HIP), and an executable copy of the computer-assisted personal interviewing (CAPI) program. In addition, technical working papers on the survey methodology are available.
Codebook for the 2004 Survey of Consumer Finances
The codebook contains the text, variable names, and responses for the questions asked in the survey. Also provided are a brief summary of the technical features of the survey design; a copy of the source code for the CAPI program, a concordance of variable names in the final data set and those used in the CAPI program; two programs that calculate standard errors for regression models and another program to calculate standard errors for simple statistics such as weighted medians, all three of which account for sampling error and mutliple imputation; and a list of the variables included in the public data set.
Codebook ( 4.6 MB ASCII)
Last updated 9/4/2014
Showcards
At key points during the administration of the interview, interviewers show the respondents a series of cards containing information relevant to framing or answering a question. Most of this information is displayed as part of the question text or response categories, but some is more general.
Showcards ( 162 KB PDF)
Last updated February 28, 2006
Household Information Packet (HIP)
The 2004 HIP contains forms used for the non-interview reports, the household listing, the first contact with the household, and the record of calls made on the household.
HIP for AP cases ( 46 KB PDF)
HIP for LS cases ( 51 KB PDF)
Last updated May 26, 2004
Executable Computer-Assisted Personal Interviewing (CAPI) Program
The 2004 survey data were collected using CAPI. A copy of the source code for the program is included in the codebook above. To see more directly the process faced by survey interviewers, use the executable version of the CAPI program posted below. The main program files are stored in a self-extracting zip file. A short README file describes the installation and running of the program on a DOS or Windows system. The version of the program offered here has been altered so that no data are saved upon exiting the program. No warranty is made for the performance of these files.
Note: CAPI program may not be compatible with all operating systems.
CAPI installation README ( 2 KB ASCII)
CAPI Program Files ( 3.3 MB ZIP)
Survey Data
Two general types of data set are provided:
Full Public Data: Full data sets of all SCF variables
Summary Extract Public Data: Extract data set of summary variables used in the Federal Reserve Bulletin article – all dollar variables inflation-adjusted to 2022 dollars
All of the versions of the full and summary extract public data sets are provided in compressed form as WINZIP files. (See PKWARE website for software to uncompress WINZIP files.)
Special note to R users: An outside programmer has created scripts for converting and working with SCF data. These scripts are available for download from: https://guilhermejacob.github.io/context/1.6-survey-of-consumer-finances-scf.html#survey-of-consumer-finances-scf
Missing data in the survey have been imputed five times using a multiple imputation technique. The information is stored in five separate and internally coherent imputation replicates (implicates). Thus, for the 4,522 families interviewed for the survey, there are 22,610 records in the data set. Three observations were deleted for the public version of the data set for purposes of disclosure avoidance; thus, there are 22,595 records in the public data set for 4,519 families. The codebook provides more detail on the structure of the data set and the steps taken for disclosure avoidance.
Full Public Data set: SAS format
The full final survey data are available to the public in two SAS transport formats that may be used with the current version of SAS. (Help is available for importing these files as SAS data sets.) One version is created using PROC CPORT; this is the smaller of the two files. The second one is created using PROC COPY with the EXPORT option; this file may be particularly convenient for users who do not have access to SAS, but who do have some means of translating the data set to another form using a software package that does not support CPORT files.
WARNING: Some browser versions will download these files in unzipped form; the resulting PROC CPORT transport file will require about 200 MB of disk space and the resulting PROC COPY transport file will require about 500 MB of disk space. When uncompressed and imported as a V9.1.3 engine SAS data set, these files require about 600 MB of disk space.
Main survey data: CPORT version ( 8.2 MB ZIP)
Main survey data: COPY/EXPORT version ( 12.7 MB ZIP)
Last updated 9/4/2014
Full Public Data Set: Stata format
A second version of the full public data set is given as a Stata data file. Because the data set contains approximately 5000 variables, users will need to use Stata SE if they wish to import all the variables. Users of standard Stata will need to specify a subset of variables when using the data set.
WARNING: Some browser versions will download these files in unzipped form; the resulting Stata file will require about 200 MB of disk space.
Main survey data: Stata version ( 8 MB ZIP)
Last updated 9/4/2014
Full Public Data set: ASCII format
A third version of the full public data set is given in ASCII format. The variables included have exactly the same values as those in the SAS and Stata data set. The layout of the variables is given by a format file, which is provided in Excel and tab-delimited formats. Note that under certain operating systems the end of record marker will add one character to the record length listed in the format file.
WARNING: Some browser versions will download these files in unzipped form; the resulting ASCII file will require about 1 GB of disk space.
Main survey data: ASCII ( 31.4 MB ZIP)
Map of variables: Excel ( 825 KB Excel)
Map of variables: Tab-delimited ( 238 KB ASCII)
Last updated 9/4/2014
Summary Extract Data set: SDA On-line analysis tool
The SDA on-line analysis tool 
is a program provided by a third party for the on-line analysis of summary variables such as the statistics found in the Federal Reserve Bulletin article. SDA is developed and maintained by the Computer-assisted Survey Methods Program (CSM) at the University of California, Berkeley. The Help/FAQ file linked to at the top of the SDA page provides instructions on how to perform analysis of the data.
Summary Extract Data set: SAS format
The summary extract data set contains variables used in the Federal Reserve Bulletin article. All dollar variables have been inflation-adjusted to 2022 dollars. Definitions of the variables in the summary extract dataset can be found in the SAS program used to create the data set. The SAS version is created using PROC CPORT, a version using PROC COPY and the EXPORT option is not provided due to the limitation on the length of variable names when using the EXPORT option.
WARNING: Some browser versions will download these files in unzipped form; the resulting PROC CPORT transport file will require about 16 MB of disk space.
SAS extract data (3.2 MB ZIP)
Last updated 4/3/2024
Summary Extract Data set: Stata format
A second version of the extract data set is given as a Stata data file.
WARNING: Some browser versions will download these files in unzipped form; the resulting Stata file will require about 30 MB of disk space.
Stata extract data (3.3 MB ZIP)
Last updated 4/3/2024
Summary Extract Data set: CSV format
A third version of the extract data set is given as a CSV file. Due to column limitiations in versions of Excel prior to 2007, the full file can only be viewed in Excel 2007 and later versions.
WARNING: Some browser versions will download these files in unzipped form; the resulting file will require about 30 MB of disk space.
CSV extract data (3 MB ZIP)
Last updated 4/3/2024
Replicate Weight Files
For many purposes, one must consider the sampling error of SCF estimates. However, because detailed information on the sample design cannot be released, and because of the complexity of the SCF design, users cannot apply some of the standard procedures for variance estimation. A set of sample replicates has been created with bootstrap techniques and analysis weights have been computed independently for each replicate. Analysts may use these weights to make approximate estimates of sampling variance. Replicate weights corresponding to X42001 are available. See the codebook for more details. When uncompressed and restored to a V9.1.3 engine SAS data set, each file requires 49.6 MB of disk space.
WARNING: Some browser versions will download these files in unzipped form; the resulting PROC CPORT transport file will require 41.7 MB of disk space, the resulting PROC COPY transport file will require 47.6 MB of disk space, and the resulting Stata file will require 22 MB of disk space.
Replicate weights--X42001: CPORT version ( 26.5 MB ZIP)
Replicate weights--X42001: COPY/EXPORT version ( 26.5 MB ZIP)
Replicate weights--X42001: Stata version ( 25.9 MB ZIP)
Last updated 9/4/2014