The 2001 Residential Finance Survey (RFS) was sponsored by the Department of Housing and Urban Development (HUD) and conducted by the Census Bureau. The RFS is a follow-on survey to the 2000 decennial census designed to collect, process, and produce information about the financing of all nonfarm, residential properties. Previous RF surveys have been integral parts of the decennial censuses since 1950. Primary users of RFS data in addition to HUD include the Bureau of Economic Analysis, Fannie Mae and Freddie Mac, and the Congress. Data are collected, tabulated, and presented for properties, the standard unit of reference for financial transactions related to housing. In the RFS, a property is defined as all the buildings and land covered by a single first mortgage. The sample for the RFS is stratified by property size, with large properties overrepresented in the sample. Very large properties are selected with certainty to control their effect on the reliability of the estimates. The RFS is the only standardized single source of detailed information on property, mortgage, and financial characteristics for multiunit properties. Both property owners and mortgage lenders are interviewed, resulting in more accurate information on property and mortgage characteristics. As part of the decennial census, the RFS is mandatory. This is important in collecting information from mortgage lenders.

Facsimiles of the homeowner, rental property and lender questionnaires; a glossary of terms and definitions; a listing of frequently asked questions; and other information about the survey can be found at the following sites:

Residential Finance Survey: 2001 Report Is Now Available.

The main purpose of the survey was to collect information about mortgage financing. The report contains information for all first mortgages including: application method, reasons for refinancing, amounts and uses of cash-outs, year of origination, use of mortgage insurance or guarantees, type of mortgage, origination amounts and current balance, interest rate, interest rate buydowns, original and remaining term of the mortgage, indexes and caps used for ARMs, and items included in and amounts of monthly payments. Similar, but less detailed, information is reported for junior mortgages and home equity lines of credit. The report chapters are organized by tenure – owner-occupied, rental, and vacant units – and by number of units in the property. There are separate chapters for condominiums and for manufactured (mobile) homes. Appendices to the report include facsimiles of the questionnaires, sample design and sampling errors, content descriptions and definitions, and data processing procedures. The report is available for free as a downloadable PDF document from either the HUD Internet site or Census Internet site. In addition to the tables for the nation and the four Census regions included in the report, additional tables for more detailed geographical areas are available on the Census Internet site. Printed (or hard) copies of the report are available from huduser.gov (800-245-2691) or from the Census Bureau Customer Service Center (301-763-4636).

1991 RFS Files and Documentation

Homeowner File, Public Use ASCII Version (5.5 MB)
Rental File, Public Use ASCII Version (7.6 MB)
Micro database File Documentation, Scanned from Hard copy (*.pdf, 4 MB)
Corrections to the 1991 Residential Finance Survey documentation (*.pdf, 754 KB)

The RFS files and documentation are available for downloading:

5^th October, 2005: New versions of Rental Property and Homeowner data and ascii files are available now.

The problem involved 246 cases on the Renter file. These cases had LIVEHERE values of 1 or 2 (owner lived at the property as a primary residence or as a second or vacation home) and NUMUNITS of 1 to 4. According to the rules for classifying cases these would have been considered owner-occupied units and should have been on the Owner file. After review, it was determined that the LIVEHERE values were incorrect, that the cases were rental units and the LIVEHERE values have been corrected in the newly issued file. In addition, not all edit flag variables (those starting with the letter"J") had value labels, and this is corrected in the new files

Download the Rental Property File (*.zip, 4.12 MB)
Download the Homeowner File (*.zip, 2.41 MB)
Residential Finance Survey: 2001 (Order hard copy.)
Comma-delimited ASCII Renters and Vacant File (*.zip, 5.10 MB)
Comma-delimited ASCII Homeowners File (*.zip, 3.17 MB)
Download the Renter codebook (*.txt, 639 KB)
Download the Homeowner codebook (*.txt, 626 KB)
Renter and Vacant Units Top-Coding Values and Cut-offs (*.xls, 20 KB)
Homeowners Units Top-Coding Values and Cut-offs (*.xls, 20 KB)
Join the RFS Electronic Mailing List

Previous Changes:

1^st June, 2005: New codebooks/data dictionaries

New codebooks/data dictionaries have been added. These codebooks have been reformatted in order to be more useful to RFS users.

25^th May, 2005

April 22^nd version of Rental Property and Homeowner data files are available now.
ASCII Versions of the Residential Finance Survey Files and Top Code Information are available now.

Residential Finance Survey — April 22, 2005 Version

The file that is being released is the first reissue since September 2004. After that release several users discovered cases that had suspect values for monetary items:

Very high or very low values of mortgage debt per unit,
High ratios of mortgage debt to value,
Very high or very low ratios of mortgage debt to house value

The Census Bureau has retraced these cases to the original data collection forms and retrieved correct values.

During this time the Census Bureau also was preparing the tables for the hard copy report. This table review also highlighted or uncovered cases that needed correction.

The needed changes and corrections been made and are contained in the new databases.

In the next few weeks we expect to issue a revised codebook or data dictionary and a listing of the top-coding values for variables that were top-coded.

Within the next six weeks we expect to issue a Source and Accuracy write-up that would include a section on estimating standard errors.

ASCII Versions of the Residential Finance Survey Files and Top-Code Information

Comma-delimited ASCII versions of the two Residential Finance Survey databases – Renters and Vacant Units and Homeowners – are now available.

Top-coding information for the two files is now available in an Excel spreadsheet. The information consists of the variable name, description of the variable, the top-code chosen (the highest respondent-reported value on the file), the weighted mean of the cases above the top-code value (i.e., the value that replaces the respondent-reported value if it is above the top-code amount,) and the number of cases that have been top-coded.

Note: 21st September 2004
New RFS File Expected in Early October

A new file was just released (the September 14 vintage) that contained corrected information for multifamily (5+) properties originated in 2000 and 2001 that had ORIGAMT1/NUMUNITS over $200,000. As noted at that time, we discovered about 200 other multifamily cases (i.e., those with origination years other than 2000 or 2001) that had ORIGAMT1/NUMUNITS over $200,000. A list has been sent to Census and they have found numerous cases that need to be corrected. The Census Bureau notes that the vast majority of these problems were caused by the same data input problems as found with the 2000 and 2001 originations — missing trailing zeroes (decimal slides) or spurious leading digits.

Since these cases have been identified, other possible problems with multifamily cases have been identified (A case may appear in more than one problem category):

About 700 cases with excessive debt-to-value ratios, i.e., ORIGAMT1/ORIGVAL1 over 200%
About 45 cases with low per-unit debt, i.e., ORIGAMT1/NUMUNITS less than $500.
About 52 cases with low per-unit values, i.e., ORIGVAL1/NUMUNITS less than $500.

The Census Bureau is now checking the original input data sources for these cases and expects to make corrections.

Once these additional cases are reviewed and corrected, we will issue a new database. It is expected that this will occur in about two weeks, early in October. We will issue an email at that time.

We expect to continue updating the files as users report issues to us and as we uncover possible problems as the final report tables are programmed and reviewed. So users should be sure that they are using the most recent file.

14th September, 2004

Editing of the RFS microdata base continues as the Census Bureau is preparing the software and programs for generating the multivariate tables that will form the body of the final hardcopy (and Internet) versions of the 2001 Residential Finance Survey report. We had released the database as early as possible and expected to be uncovering problems as the Census Bureau reviews draft tables, and as users report to us possible problems they have encountered. We have already issued a revised version in response to users concerns. The Census Bureau will continue to respond to users concerns until the final tables are prepared. At that point, we would expect that almost all errors will have been identified through table review or user input.

Last week, a user alerted us to some unusual values for ORIGAMT1 for multifamily ( MF 5+) properties that were originated in 2000 and 2001. The> focus was on MF cases with ORIGAMT1/NUMUNITS over $200,000. There were 22 cases for 2001 originations and 40 cases for 2000 originations. The Census Bureau traced these values back to the original questionnaires or electronic data submissions and found that many of the problems were caused by decimal slides or by spurious leading digits. These cases are now being corrected and a new data file should be posted in the next few days.

As a follow-up to these problems, the Census Bureau is now looking at all MF originations with high origination values per units, i.e., ORIGAMT1/NUMUNITS > $200,000. If needed, a new database will be issued if these values need to be changed. Furthermore, editing will continue as part of the process of generating the final report so it is possible that other versions of the file will be released in the future.

26th July, 2004 - A new version of Rental Property and Homeowner data files are available now.

The new versions of the files correct the following problems or concerns:

All the dollar variables are top-coded using a weighted mean.

The Census Bureau has responded to user concerns and has revised the files accordingly. They will continue to look at issues that surface as the files are used, so changes might be implemented in the future.

21st July, 2004 - Changes on Rental Property and Homeowner data files.

The new versions of the files correct the following problems or concerns:

For Renter units with 5 or more units and origination years of 2001 there has been a change to assigning values for mortgage balances (MGTBAL1, MGTBAL2 and MGTBAL3) when the balances were not reported. In earlier versions of the files, values were allocated or assigned to these cases using a random assignment method, i.e., hot deck. In this version, the mortgage values were calculated based on other information on the file (interest rate, term, and payment). This will, hopefully, restore some balance to the relationship between original mortgage amount and current mortgage balance for 2001 originations for properties with five or more units.

In this revision, the values for year acquired will be in single-year increments for properties acquired in 1990 and later years. This was supposed to have been done in the last revision, but was not.

16th July, 2004 - Changes on Rental Property, Homeowner data and documentation files.

Data file changes: The variable YRACQ on both datasets is being rounded to the nearest 5 for years earlier than 1990. The years 1990 and above are left as they came in, i.e., single years. There were three rental property records that had bad ORIGAMT1 or MTGBAL1. They have been fixed.

Documentation file changes: The formatting has been revised to allow more spacing between variables, and some typos have been corrected.