For years, policymakers and program managers within the U.S. Department of Housing and Urban Development (HUD), as well as housing advocates and analysts both inside and outside the Government, have needed better information on the characteristics of the households that live in assisted housing. Until recently, reliable answers to basic questions, such as the average income of assisted households and the racial composition of households in assisted units, were hard to come by even though various HUD data systems collected relevant information. In 1992, HUD began releasing comprehensive data on assisted housing from the American Housing Survey.¹ Although these data filled the gap, they had two serious limitations: incomplete samples introduced bias into the reported measures, and local data were not available.²

Thanks to the work of many individuals throughout the Department, HUD's program data systems were greatly improved during the early 1990s. HUD has now begun to make detailed information available on the characteristics of assisted households. This article describes the newly available data and explains how to access them.

Making Program Data Work

Recipients of HUD-assisted housing must report their incomes and other pertinent information at the time of admission and annually thereafter to local housing agencies or private owners who manage the housing. These managers use the information to determine initial eligibility for assistance and to set the amount of contribution toward rent that will be expected from the household.

For years, managers recorded these data on HUD-approved forms and stored them in offices throughout the country. For some programs, copies were sent to local HUD offices and filed there as well. Some of these data were then entered into various program data systems. In the late 1980s and early 1990s, HUD set in motion a number of initiatives to collect this information more systematically. Managers were encouraged to submit data in electronic formats to centralized locations. HUD hired a contractor to enter hard-copy data into electronic form. Meanwhile, HUD staff began assembling these data into two major data systems: the Multifamily Tenant Characteristics System (MTCS) and the Tenant Rental Assistance Certification System (TRACS).

After MTCS and TRACS became operational and had acquired data representing more than 60 percent of assisted households, staff in the Office of Policy Development and Research aggregated these data into files that could easily answer program or policy questions without compromising the confidentiality of information on individual households. One product of these efforts that is available to the general public is A Picture of Subsidized Households.

A Picture of Subsidized Households

Paul Burke, a HUD analyst, has compiled electronic files and hard-copy reports under the name A Picture of Subsidized Households. Within confidentiality constraints, Picture presents summary statistics by program type for the United States as a whole as well as for each State, public or Indian housing agency, project, and census tract. Using the files, one can create city, county, or metropolitan totals. The following programs are identified separately: public housing, Indian housing, Section 8 certificates and vouchers, Section 8 moderate rehabilitation, Section 8 new construction and substantial rehabilitation, and Section 236. All other HUD subsidy programs are reported in an "Other Subsidy" category. Picture also uses HUD's low-income- housing tax credit database to show where housing units receiving tax credits are located, but no tenant characteristics data are available for this program.³

Table 1 presents a number of the demographic variables that are available on the file from the national level all the way down to the project or census tract level. Among the four program types reported, public housing and Section 8 new construction and substantial rehabilitation serve the lowest income households, with Section 8 certificates and vouchers close behind. The majority of households in Section 8 new construction and substantial rehabilitation are elderly. Section 8 certificates and vouchers has the lowest percentage of elderly households and the highest percentage of households with children. Picture also contains more detailed information that allows one to see the place-to-place variation in whom these programs serve and how they operate.

Users of either the hard-copy or electronic versions of Picture need to be aware of a number of cautions:

• The most recent file (1997) covers all projects and units, but occupancy characteristics are based on only 80 percent of the assisted units; program managers did not report on the other 20 percent. In addition, reporting rates vary widely by location.

• Tenant characteristics are not shown for projects or census tracts if the number of assisted units is 10 or fewer or if the reported data represent less than 40 percent of all assisted units.

• Missing data on individual items are identified, as described in the report.

• Percent-occupied data are available only for public housing. In other programs, plausible assumptions are made about occupancy rates to calculate reporting levels.

Before using any variable, users should carefully read the definition of the variable and associated notes that may indicate particular problems with the variable. For example, the variable measuring the percentage of households that have moved into their units within the past year occasionally reads 0 percent or 99 percent -- the former indicates no new admissions, and the latter indicates that the entire project contains new admissions or that the census tract contains only new assisted households. Although 0 percent and 99 percent may be legitimate values, the definition notes indicate that these values probably indicate data quality problems for that project or census tract.

Special Descriptive Measures

Picture also contains some special statistics that were calculated from the reporting forms, other program records, or 1990 census data. (See Table 2.)

Among the statistics that housing policy analysts would typically want most are the monthly costs of providing assistance under the various programs. Picture cannot provide these data, but it does provide an interesting substitute, namely, the monthly current costs of providing assistance. From the annual forms filled out on each Section 8 tenant, Picture calculates the difference between the total monthly rent of the unit and the rent paid by the tenant.⁴ (For Section 8, this is the total cost to the Government.) For public and Indian housing, Picture estimates the operating and modernization funds per unit per month. (This calculation ignores the initial costs of constructing public and Indian housing.) For Section 236 and other subsidy programs, Picture reports only the Section 8 costs, if any, of Section 236 tenants who also receive Section 8 assistance. (This estimate does not include the monthly Section 236 interest subsidy.) Although these deficiencies make it inappropriate to compare spending across programs, the spending estimates give a sense of the ongoing costs to the Government and, more importantly, how these costs vary across places and across local housing agencies or projects.

The average income numbers (see Table 1) show a simple summary of whom the program serves; but it is also possible to measure, place-by-place, the variation around this mean. For this purpose, Picture uses a standard statistic -- the coefficient of variation, which is the ratio of the standard deviation to the mean. The larger the ratio, the greater the range of variation around the mean. This measure is reported at various levels, from project or census tract to Nation. At the national level, public housing has the greatest relative variation in income, followed by Section 8 certificates and vouchers.

Housing policy analysts would also like to know the extent to which assisted housing is provided in a racially diverse manner. Because Picture reports at the project level, the percentage minority for each project -- as well as the percentage African American, Hispanic, and so forth -- can be determined directly from the data.⁵ Because local housing agencies provide public housing and the Section 8 certificate and voucher opportunities in many different locations, Picture also provides a useful summary measure for the particular program administered by a local housing agency.

This measure of racial diversity averages the absolute value of the difference between the percentage minority in a project (or percentage minority of the Section 8 certificate and voucher holders living in a census tract) and the percentage minority among all the clients served by that program at a local housing agency. For example, if a local housing agency has only two projects, each having 100 units, and one project is all minority and the other is all white, then the agency has a client base that is 50 percent minority. The measure of racial diversity is 50 because the absolute differences are 100 - 50 = 50 and 50 - 0 = 50, which averages to 50. Using another example, if the two projects were both 50 percent minority, then the measure of racial diversity is 0 because the absolute differences are 50 - 50 = 0 and 50 - 50 = 0, which averages to 0. These examples show the two extremes of the scale, with 0 representing project (or tract) diversity that exactly matches the diversity in the client group and 50 representing no diversity at the project (or tract) level.

This racial difference statistic is reported only for public housing and Section 8 certificates and vouchers. The averages of these measures across the Nation are 13 and 14, respectively. These readings suggest a high level of diversity on average.

Picture also incorporates 1990 census data to describe the environment in which assisted housing is provided. Table 2 shows that the Section 8 certificates and vouchers, Section 8 new construction, and Section 236 operate in census tracts where an average of about 20 percent of the populations have incomes below the poverty level. This suggests that these programs typically operate in poor neighborhoods. In 1990, 71.3 percent of all metropolitan census tracts had a poverty level lower than 15 percent, and 78.5 percent had a poverty level lower than 20 percent. The average poverty rate for the neighborhoods in which public housing projects are located was 37 percent. In 1990, only 8.8 percent of metropolitan census tracts had poverty rates higher than 35 percent.

Public housing projects also tend to be located in more heavily minority neighborhoods. Based on 1990 census data, the average proportion of minorities in tracts with public housing projects was 59 percent, compared with 40 percent for Section 236 and 36 percent for Section 8 new construction and substantial rehabilitation. Section 8 certificate and voucher holders lived in census tracts that were, on average, 39 percent minority in 1990.

Section 8 certificate and voucher holders live in census tracts in which, on average, 40 percent of the households in 1990 were owners of single-family detached homes, compared with 36 percent in Section 8 new construction and substantial rehabilitation, 33 percent in Section 236, and 26 percent in public housing.

All statistics in Table 2 are available in the data file at the national, State, agency, and project levels and for most or all programs. In an earlier publication of the data, a short commentary included several tables illustrating analyses possible by using the file. Two of these tabulations deserve mention. To assess the extent to which subsidized units are spatially concentrated, one tabulation looked at subsidized households as a percentage of all households in census tracts.⁶ The median tract with subsidized housing contained 12 percent subsidized units. Fewer than one out of seven subsidized units was located in a tract where 50 percent or more of the households lived in subsidized units. At the other extreme, 20 percent of the population lived in census tracts where there were no subsidized households. The second tabulation looked at the size of subsidized housing projects.⁷ Approximately 12 percent of public housing households lived in projects with fewer than 50 units; another 12 percent lived in projects with 1,000 or more units. Among residents of Section 8 new construction and substantial rehabilitation units, 22 percent lived in projects with fewer than 50 units while less than 1 percent lived in projects with 1,000 or more units.

How To Get the Data

A Picture of Subsidized Households in 1997: United States: Totals & Agencies With Over 500 Units was published on October 23, 1997, and contains data reported by local housing agencies and project managers through the middle of 1997. It can be obtained from HUD USER at 1-800-245-2691 (or 202-708-3178 in the Washington, D.C. metropolitan area). The order number for this publication is HUD-8562.

An earlier published version contained information from local housing agencies through September 1996 and from housing projects through October 1995. This version was published in 11 volumes, 1 for each of the 10 HUD regions that report on local housing conditions in U.S. Housing Market Conditions and a national summary volume. Much greater local detail is provided. Volume 11, the national summary, also contains 44 maps showing the location of assisted housing in various parts of the country. Each of the 11 volumes has the general title A Picture of Subsidized Households, and individual volumes can be ordered from HUD USER. The order numbers run from HUD-7611 through HUD-7621.

Additionally, all of the data are available at www.huduser.gov/datasets/assthsg.html. The files are more complete than the publications and contain data on all sizes of agencies, although they omit the maps. The 1993 file and a special historical file with partial data for the 1970s contain only public and Indian housing data.

Notes

1. Characteristics of HUD-Assisted Renters and Their Units in 1989, U.S. Department of Housing and Urban Development, March 1992. This report matched addresses of the more than 15,000 renters in the American Housing Survey with the addresses of HUD- assisted households. Followup reports using the 1991 and 1993 American Housing Surveys were published in 1997. Separately, the American Housing Survey has been publishing data on tenants who identify themselves as living in public housing and "other federally assisted housing" since 1973. Unfortunately, neither address matching nor respondents give complete data. Address matching finds 50 percent as many sample units as it should in public housing, 37 percent in other project-based assistance, and 137 percent in certificates and vouchers. (Over coverage means the lists included households that were unsubsidized or were in other programs.) On the other hand, self-identification by the respondents finds 196 percent as many sample units as it should in public housing and 76 percent in other Federal assistance. The total assisted sample from address matching is 70 percent of the size it should be. The total sample from self-identification is 110 percent of the size it should be. (These are 1993 figures.)

2. Using the American Housing Survey as the source also provided a valuable benefit, namely, information on the characteristics and quality of the unit occupied by the assisted household. Program data cannot provide comparable information.

3. See the August 1996 edition of U.S. Housing Market Conditions for a discussion of the low-income-housing tax credit database.

4. For Section 8 certificates and vouchers, spending also includes the administrative fee; that is, 8 percent of the two-bedroom fair market rent.

5. Note that the national aggregates presented in Table 1 do not provide this information. Two 100-unit projects, one all minority and one all white, would average out as 50 percent minority. Diversity must be measured at the local level.

6. A Picture of Subsidized Households Volume 11, United States: Large Projects & Agencies, December 1996, page 13.

7. Ibid., page 14.