The American Housing Survey (AHS) provides the only detailed information available between the decennial censuses on the Nation's housing stock and how American families are sheltered. HUD and the Bureau of the Census have conducted this survey program since 1973. The program consists of the national survey, conducted every 2 years, and 47 metropolitan area surveys, conducted every 4 to 6 years in the largest metropolitan areas of the country. Both surveys collect information on the entire housing stock including characteristics of the structure such as size and number of rooms, plumbing, amenities, and household equipment (including heating); the neighborhood; financing and housing costs; demographics such as incomes, migration, and commutes of the family and occupants; and occupancy and tenure of the housing unit. The national survey involves interviews or visits to about 60,000 housing units every 2 years, whereas each metropolitan survey involves interviews or visits to 2,500 to 5,000 housing units every 4 to 6 years. A unique and important aspect of the AHS is that the same housing units are revisited during each survey, providing a history of American housing. In addition, newly constructed units are added each survey year so the survey results represent all housing.

Results from the 1997 national survey are now available¹ and form the focus of this article. This article discusses the special nature of the 1997 survey, presents selected results, and explains the many ways to access this rich data source.

The 1997 AHS National Survey: Change and Challenge

The 1997 AHS national survey marks a dramatic evolution in data collection and processing for the AHS program. For the first time, interviewing and data collection were done with computer-assisted personal interviewing (CAPI), which involved the complete transformation of the questionnaire from a pencil-and-paper format to a series of computer screens on a laptop computer. CAPI is expected to improve data quality, reduce respondent burden, speed up reporting, and reduce costs.

Experience has shown that any change in questionnaire format or interview technique can produce different answers to similar questions. Every attempt was made to minimize the change to wording and phrasing of questions while taking advantage of the capabilities of the personal computer. The change to CAPI was accompanied by a change in the data processing system.

Still, given these major changes, HUD and the Census Bureau were very concerned that the new survey data collection methods and data processing and analysis systems be thoroughly reviewed for errors and that survey changes be carefully documented. This was accomplished with an extraordinary quality control and data review process conducted during the last year.

One aspect of the process involved independent development of key components of the processing and estimation system. In this way we were able to ensure that the new software was correct. Estimates from a wide array of AHS publication tables from the 1993 and 1995 surveys were entered into an automated system that compared those results with intermediate estimates developed from the 1997 survey. The system automatically identified large and anomalous differences across the three sets of survey estimates. This "comparer" system was run on each new and revised version of the 1997 data.

After each run, staff from HUD, the Census Bureau, and the independent contractor (ICF Consulting) met to determine possible reasons for the large differences. This review involved looking at the logic of the CAPI program, specifications and coding of the editing and allocation software, specifications and computer programs used to recode questionnaire data to variables used in the AHS public use file and publication tables, and table specifications and software. After each review, the data file was revised and corrected.

This review-revise cycle was done numerous times until we felt that there were no more errors and that the remaining differences could not be eliminated. A much-improved data file and documentation of remaining differences are the outcomes of the quality control and review process.

Following the discussion of survey results, the dissemination of results to the AHS user community will be discussed.

Selective Housing and Family Information From the 1997 National AHS

The composition of the 112 million housing units of the American housing stock is presented in Table 1. Of these housing units about 3 percent are seasonal units and nearly 9 percent are vacant. Of the nearly 100 million occupied units in 1997, nearly 66 percent are owner occupied.

Table 1. Composition of the 1997 Housing Stock

The AHS found that three-quarters of American housing units are in single-family structures. Approximately 60 percent of America's housing is single-family, detached homes. Townhouses and other single-family, attached units account for 6 percent of the housing stock. Manufactured or mobile homes account for 7.4 percent of the housing stock, and the remaining 26 percent are in multiunit buildings with about one-third of those being in small (two to four units) buildings (see Table 2).

Table 2. Units in Structure

In 1997, the AHS determined that American housing is fairly new, with more than one-fourth of the units less than 20 years old (having been built since 1980). The median age is 32 years. Yet housing units are very durable and long-lived, with 1 in 11 units more than 80 years old (having been built in 1919 or earlier) and 1 in 4 units more than 50 years old (having been built before 1950) (see Table 3).

Table 3. Year Structure Built

Most American housing is located in urban areas. More than three-quarters of American housing is located in metropolitan areas, and about 24 percent is located outside metropolitan areas, according to the AHS (See Table 4). Of those in metropolitan areas, 60 percent are in suburbs and 40 percent are in central cities. Regionally, more than one-third of American housing is in the South, one-fourth is in the Midwest, about one-fifth is in the West, and about one-fifth is in the Northeast.

Table 4. Location of U.S. Housing

The AHS reveals that American housing is made up of fairly roomy houses. The median number of rooms is 5.3, with nearly half of the homes having six or more rooms. Only about 2 percent of the housing units have only one or two rooms (see Table 5).

Table 5. Number of Rooms

The number of bedrooms also indicates the size of the American home in 1997. The median number of bedrooms is 2.6, with 39 percent of the housing units having 3 bedrooms. Few housing units (less than 1 percent) have no bedrooms (see Table 6).

Table 6. Number of Bedrooms

The survey found that, in 1997, almost all American homes (98.5 percent) had one or more complete bathrooms and 38 percent had 2 or more (see Table 7).

Table 7. Number of Bathrooms

When AHS interviewers asked about problems, respondents noted that deficiencies are rare. Few units (2 percent or less) have holes in the floor, lack or have exposed electrical wiring, or have rooms without electrical outlets. About 6 percent of survey respondents reported less serious problems such as open cracks or holes in the interiors of the units, while 3 percent reported broken plaster or peeling paint (see Table 8).

Table 8. Select Deficiencies

The 1997 AHS found that Americans had median monthly housing costs of $542. Slightly more than 18 percent of the households reported housing costs of $1,000 or more, while 24 percent reported costs of less than $300 (see Table 9). In terms of income, the typical household spent about 21 percent of its income for housing. This figure was different for owners versus renters: Owners paid a median ratio of 17 percent of their income for housing, while renters paid 29 percent of their income for housing.

Table 9. Monthly Housing Costs

The preceding discussion only touches on the wealth of information about American families and their housing. The report (available in hard copy printed format or as a downloadable PDF document) contains 453 pages of tables that cover general housing characteristics, building and housing unit size, plumbing, equipment, fuels, problems with equipment, neighborhood conditions, occupant family types and composition, recent movers, income and housing costs, and value and financing. There are separate chapters for all housing units, occupied units, owner occupants, renter families, African-American households, Hispanic households, and elderly households. The next section discusses the many ways to tap into this rich and extensive source of information.

Accessing Housing Information

HUD and the Census Bureau have given high priority to making the AHS accessible. The AHS information is made available in print, electronic, tabular, and microdata formats. A full array of media is used: hard copy, PDF versions, CD-ROM, and downloadable Internet files. Copies of the printed versions are available through the Census Bureau (301-457-4100) and from HUD USER (1-800-245-2691). Both the Bureau and HUD maintain Internet sites, and HUD operates an AHS listserv for the AHS user community.

The HUD Web site for the AHS is at http://www.huduser.gov/datasets/ahs.html. At this site you will find:

• The 1997 microdata files for downloading in both SAS and ASCII format.

• Summary statistics for the 1997 survey.

• Line-by-line comparisons of 1993, 1995, and 1997 estimates for a wide array of housing and occupant characteristics.

• The codebook for the survey.

• Links to the Census Bureau Web site.

• Microdata in downloadable format for the 1995 national survey and the 1996 and 1995 metropolitan surveys.

• Information for ordering reports and documents from HUD USER. (P.O. Box 23268, Washington, D.C. 20850.)

• PDF versions of the AHS reports, including the 1997, 1995, and 1993 national survey reports, and the 1994, 1995, 1996, and 1998 (forthcoming) metropolitan survey reports. Other AHS and housing reports are also available there.

• Internet access to the microdata files for extracts and analyses using automated systems. The 1997 national survey is available through the FERRET system, which allows the user to download ASCII extracts or to create user-specified tables. The 1993 and 1995 national surveys are available for extracting and downloading thorough the Census Bureau's Data Extraction System.

• A description of the surveys, historical changes in the surveys, definitions of concepts and variables, sample design and sizes, estimation weights, and survey results in brief formats.

• Information for ordering reports and documents from Customer Services. (Bureau of the Census, Washington, DC 20233-8500.)

Stay Tuned: More To Come

Further documentation of the impact of the change in survey data collection procedures on the 1997 national survey is being prepared. This documentation will provide users with information about changes and comparability with earlier national surveys, and is expected to be completed within the next 6 months.

As this issue of U.S. Housing Market Conditions goes to press, interviewing for the 1999 AHS national sample will be nearing completion. HUD and the Census Bureau are planning to have the data file available early next year and to have the published report available by next summer. The questionnaire and report format will be nearly identical to the 1997 survey, and because the same housing units from the 1997 survey will be revisited, the data will be even richer because survey-to-survey comparisons can be done on the same housing units.²

The database or microdata files for 1998 metropolitan surveys are nearly completed and should be released in the next month. Metropolitan areas surveyed in 1998 include Baltimore, Maryland; Birmingham, Alabama; Boston, Massachusetts-New Hampshire; Cincinnati, Ohio-Kentucky-Indiana; Houston, Texas; Minneapolis-St. Paul, Minnesota-Wisconsin; Norfolk-Virginia Beach-Newport News, Virginia; Oakland, California; Providence-Pawtucket-Warwick, Rhode Island-Massachusetts; Rochester, New York; Salt Lake City, Utah; San Francisco, California; San Jose, California; Tampa-St. Petersburg, Florida; and Washington, D.C.-Maryland-Virginia. Printed reports for these areas should be available early next year.