Census Data Needs for Housing and Urban Development
In 1992, the Office of Management and Budget (OMB) requested department and agency heads to identify topics to be included in the 2000 census. The Department of Housing and Urban Development's (HUD) response stated that ''HUD cannot fulfill its mission of making communities work for people without the use of census data in developing, administering, and monitoring HUD programs" (Cisneros, 1993:1). HUD uses census data to allocate funds to cities and local jurisdictions most in need of assistance, such as in the assisted housing Community Development Block Grants (CDBG) or Emergency Shelter Grants (ESG). HUD relies heavily on detailed, small-area information only available through a census-like source in monitoring HUD-funded, locally administered programs. Developing a national urban policy and assessing the condition of
The panel was assisted in the preparation of this chapter by Mary Nenno, visiting fellow at the Urban Institute. Ms. Nenno worked with the panel in contacting state and metropolitan housing and urban development officials about their perceptions of census data requirements. She also provided technical advice on the exposition of housing and urban development. The panel also acknowledges the assistance of four housing experts who provided background papers. Roberta F. Garber, a consultant in Westerville, Ohio, prepared a paper on the use of census data in housing policy development. F. Edward Geiger, III, an official with Pennsylvania's Department of Community Affairs, provided a paper on the quality of housing quality statistics. Mr. Geiger received advice and suggestions from Natalie S. Geiger and Dr. Marshall A. Geiger. Franklin F. James, professor of public policy at the University of Colorado at Denver, prepared a paper on the priorities for housing data from future censuses. Thomas B. Cook, deputy director of the California Department of Housing and Community Development, prepared a paper on state and local housing policy development. This appendix also draws on information reported in Cisneros (1993).
urban areas depend largely on measurements made for cities, towns, and counties by the Census Bureau in the decennial enumeration.
Communities are at the heart of HUD's mission, and it needs to continue having information about these integral components of America's cities. Census tracts are the essential building blocks for such community information. HUD uses tract-level data in implementing its own and other federal programs. For example, HUD staff uses tract data to identify qualified census tracts for mortgage revenue bonds and low-income housing tax credits. CDBG program staff monitor the distribution of program assistance using data at the tract-and block-group level. HUD's office of Fair Housing and Equal Opportunity requires census tract-level data to conduct fair housing compliance reviews mandated by Title 6 of the Civil Rights Act of 1964. HUD's field office economists use tract data in performing market analyses required under the Federal Housing Administration (FHA) insurance programs. Location decisions for HUD-assisted housing require tract-level information to avoid undue concentrations of persons living in poverty.
After looking at the array of census 2000 options and reviewing its data needs, HUD (Cisneros, 1993) concluded that the content and design used in 1990 and other recent censuses come closest to meeting the department's needs.
FEDERAL LEGISLATIVE AND CENSUS DATA REQUIREMENTS
Federal legislation specifies census data be used by HUD in several areas. Only a few housing items are strictly "mandated" in the census. Mandated items are census questions that federal legislation specifically requires to be collected in the census. In the narrow legal sense of "mandated" there are several housing items, including number of rooms, tenure, year built, that are required to be collected in the census. Most of the mandated items are listed in federal block-grant formulas.
Because of their reliability and accuracy and for geographic consistency across the nation, census data are especially important for housing and urban development programs. The precision and reliability of population and housing items available from a 1990 census "level of content" seem to be adequate for purposes of HUD (Cisneros, 1993). According to HUD (Cisneros, 1993) reliability and accuracy of the "expanded content" and "continuous measurement" approaches would not be adequate for distributing or allocating funds since sampling errors at low levels of aggregation can result in serious misallocations of resources. And, coverage would not be sufficient for adequate evaluation of urban conditions or community development. Also, the realiability and accuracy of current surveys for income and poverty data would not be adequate for establishing program eligibility and distributing or allocating funds, again because sampling errors at low levels of aggregation can cause serious misallocations of resources. As a result, current surveys would not be adequate for assessing
urban conditions, housing needs of disabled Americans, and CDBG effectiveness.
This section first describes small-area geographic detail needed and then the content of census items that are required by federal legislation.
Small-Area Geographic Detail
The level of aggregation required for allocating funds in the CDBG and HOME programs is generally the unit of government. Allocation of funds using income and poverty counts for the CDBG, ESG, and HOME programs usually also requires aggregation to the unit of government. For many participating jurisdictions, however, data must be available at the census place level to construct "CDBG urban counties." These urban counties represent the balance of counties that contain one or more entitlement cities. In addition, the CDBG program requires that local governments develop community programs to target benefits to low-and moderate-income individuals. This must be done at the block-group level. Fair housing compliance reviews under Title 6 of the Civil Rights Act of 1964, affirmative fair housing marketing reviews under Title 8 of the Civil Rights Act of 1968, and Home Mortgage Disclosure Act compliance reviews require data at the census tract level and block group level. Identification of qualifying census tracts and difficult development areas for low-income housing tax credits and mortgage revenue bonds also requires the use of tract-level data.
Reporting on the conditions of urban areas requires assessment of concentrations and distributions of activities and needs across neighborhoods, i.e., census tracts. The finest levels of aggregation for these use of data are at the jurisdiction or census place level for comprehensive housing affordability strategies development, CDBG, Fair Housing, and Section 202 elderly and handicapped. And, for calculating income limits and fair market rents, data at the county level are required. Income limits for various assisted housing programs must be defined for separate housing markets—cities, counties, or groups of counties
There are several key items in the census that are needed for the administration of HUD programs: population counts; housing; income and poverty; labor-market information; mobility, health, ethnicity, and language; and homeless population. This section reviews each of these important data items and the federal legislation that requires the information.
Among the most essential census information for HUD are data on persons, households, and family counts by family type, age, disabilities, race, and marital status. These data are used to allocate funds; set income limits; report on housing program accomplishments, needs, and urban conditions; assist local governments in planning; enforce civil rights laws; and identify needy areas and families. For example, this information, along with other data elements, is used to allocate CDBG, ESG, HOME Investment Funds, Section 8 Housing certificates and vouchers, and Section 202 (elderly and handicapped) funds. More effective targeting and more equitable distribution of resources are the goals of such data-driven allocation formulas.
Enforcement and monitoring of the Civil Rights Act of 1964 and the FHA are also carried out through use of these data. Court cases concerning housing discrimination against families with children require such data as evidence. This information is used by local government units to develop timely and well-designed comprehensive housing affordability strategies, which are required to establish eligibility for receiving funds from CDBG, HOME, and other HUD programs. More effective targeting of the low-income housing tax credits and mortgage revenue bonds programs can be achieved by using these data in designating qualified census tracts and difficult to develop areas. These data are under consideration for determining eligibility for new programs such as enterprise zones and community development banks. Congressionally required reports, such as the President's National Urban Policy Report, the State of Fair Housing and Civil Rights Data on HUD Program Applicants and Beneficiaries, and Worst Case Housing Needs, require the use of these data. These data are also used in combination with other data items to set income limits that reflect regional variation in housing costs for several programs, including HUD-assisted housing, Farmers Home Administration, Resolution Trust Corporation, and Treasury Department programs.
One problem, however, is that the 1990 census categories used for race and ethnicity allow an "other race" option. This makes it difficult for HUD and other federal civil rights enforcement agencies to use census data with program data collected and categorized according to OMB guidelines on racial and ethnic information, which do not include an "other race" category.
A central requirement for census data for HUD is information on housing, including data on tenure, cost, rent, value, utilities, quality, size, structure types, amenities, and physical conditions. The 1990 data on housing are used to allocate funds; make grants; establish fair market rents; set income limits; report on housing programs, needs, and urban conditions; assist local governments in planning;
enforce civil rights laws; and identify needy areas and families. For example, this information, along with other data elements, is used to allocate CDBG, ESG, HOME Investment Funds, Section 8 Housing certificates and vouchers, and Section 202 (elderly and handicapped) funds. The goals of the allocations formulas are more effective targeting and more equitable distribution of resources.
Information on rent paid, tenure, number of bedrooms, plumbing and kitchen facilities, utility costs, year moved in, and quality is used to calculate fair market rents used in the Section 8 Housing certificate and voucher program. Enforcement and monitoring of the Civil Rights Acts of 1964 and 1968, the Home Mortgage Disclosure Act, and the Fair Housing Act are also carried out through use of these data. This information is used by local government units to develop timely and well-designed comprehensive housing affordability strategies, which are required to establish eligibility for receiving funds under CDBG, HOME, and other HUD programs. FHA underwriting for single-and multifamily insurance also makes use of these data to undertake market and feasibility analyses. More effective targeting of the low-income housing tax credits and mortgage revenue bonds programs can be achieved by using these data for designating qualified census tracts and difficult development areas.
Congressionally required reports, such as the President's National Urban Policy Report and Worst Case Housing Needs, require use of these data. Like population data, these data are used in combination with other data items in setting income limits that reflect regional variation in housing costs for several programs and agencies, including HUD-assisted housing, Farmers Home Administration, Resolution Trust Corporation, and Treasury Department programs.
Two problems arise, however, with the use of 1990 data. First, collection of rent information using the same set of precoded rent intervals for all areas of the country and all bedroom sizes presents serious accuracy problems in calculating fair market rents when rents exceed $750 per month. Second, the 1990 census categorizations used for race and ethnicity allow an "other race" option, making it difficult for HUD and other federal civil rights enforcement agencies to use census data with program data collected and categorized according to OMB guidelines on racial and ethnic information.
Income and Poverty
Income information and numbers of persons in poverty are used in several HUD programs. For example, income information for local areas collected in the census is used to define low, very low, and moderate incomes for eligibility determinations in conventional public housing, Section 8 Housing voucher and certificate programs (existing, new construction, and moderate and substantial rehabilitation); rent supplements, CDBG targeting (low-and moderate-income benefit requirement); allocation of funds in the CDBG, Fair Share, Emergency
Shelter Grants, Section 202 elderly housing and HOME programs; developing comprehensive housing affordability strategies by local governments; reporting on worst case housing needs; and identifying qualified census tracts and difficult development areas for low-income housing tax credit and mortgage revenue bond programs. HUD income limits and estimates are used by numerous other agencies and organizations: the Federal Reserve Board, Treasury, Farmers Home Administration, Resolution Trust Corporation, Federal Deposit Insurance Corporation, Federal Housing Finance Board, Federal National Mortgage Corporation, and Federal Home Loan Mortgage Corporation. The 1990 census data on income and poverty are used to define income eligibility limits for rent supplements and conventional public housing. The income limits are also being used by the Federal Reserve Board, U.S. Treasury Department, Farmers Home Administration, Resolution Trust Corporation, Federal Deposit Insurance Corporation, Federal Housing Finance Board, Federal National Mortgage Corporation, and Federal Home Loan Mortgage Corporation. In addition, census data are being used to report on conditions and needs for housing and urban development, assist local governments in planning, and enforce civil rights laws.
Labor Market Information
These data—education, school attendance, employment and unemployment, work location, transportation, industry, occupation, type of firm, and number of weeks worked—are used in assessing the condition of urban areas in the President's Urban Policy Report and in assessing and evaluating the effectiveness of the CDBG program. Secretary Cisneros has initiated a new process of monitoring the condition of cities that uses information from the 1990 census.
Mobility, Health, Ethnicity, and Language
The 1990 census data on mobility, health, ethnicity, and language have been used in the National Urban Policy Report, CHAS development, Section 202 grants, the CDBG evaluation, and in fair housing compliance and enforcement. Information on health conditions and health limitations is used in the Section 202 Elderly Housing Grant Program (now Supportive Housing for Persons with Disabilities Program). The President's National Urban Policy Report has used of information on ethnicity, 5-year mobility, language, and health conditions as indicators of urban conditions. Comprehensive housing affordability strategies developed by local governments are required to use information on health difficulties in assessing local needs. Evaluation of the CDBG program has made use of the information on mobility and language.
The Emergency Shelter Grants, Supportive Housing Demonstration, Supplemental Assistance for Facilities to Assist the Homeless, and Section 8 Moderate Rehabilitation Single Room Occupancy—programs created by the McKinney Act—require allocation of funds, determination of eligibility, and identification of areas with severe homelessness problems. At present these funds are awarded using competitive procedures or proxy measures.
Current procedures for census data on the homeless lack adequate coverage, reasonable reliability, and accuracy. While the shelter and street night enumeration (S-night) count is reasonable in the areas where it was done, there was no universal cooperation at the local jurisdiction level.1 The street count suffers because of numerous methodological and enumeration problems.
The shelter count on the homeless from the S-night operation was used by HUD to assess alternative approaches for allocating homeless assistance by formula. At present, funds for homeless populations are generally allocated and programs are administered at the jurisdiction level. The street count, however, was not considered to be sufficiently accurate for allocating funds.
PROGRAM USES OF CENSUS DATA
Comprehensive Housing Affordability Strategies
As mentioned above, there are only a few housing items specifically mandated by legislation. There are other housing items in the census, however, that are required for effective program and policy purposes. Data on telephones, kitchen, rents, number of bedrooms, plumbing, and number of rooms are routinely used in HUD program work. At the moment, these data are updated annually using the American Housing Survey (AHS) and the consumer price index.
One of the most important current uses of census data by HUD is for the CHAS legislation. CHAS data are assembled to help give states and areas submitting their CHAS applications equal access to data. CHAS requirements include calculating the annual median family income for the area and adjusting it for family size. HUD looked at other possible tabulations for local area housing strategies, but after examining a number of options and after comprehensive discussion with state and local housing officials, it was decided that the CHAS data tabulations would offer the most accurate, equitable information for making housing decisions.
For the CHAS data, HUD obtained decennial census data tapes from the Census Bureau, processed them, and then published a databook. At the same time, the Census Bureau produced equivalent tabulations for many areas of the nation and released them on CD-ROMs. HUD has offered many workshops on
the use of the CHAS data, usually conducted with private consultants. Consultants to HUD work with state and local governments so that local agencies can prepare census data for HUD applications and program monitoring.
With the available CD-ROMs, published databook, and software, local and state agencies are able to define specific geographic areas. They examine different household types and define the low-income and poverty universe for the population in the geographic areas. Next, they analyze housing costs for various income groups. Finally, they calculate the composition of the population for affordability groups. Such a procedure helps to ensure that all parts of the nation—in different regions, as well as urban, suburban, and rural areas—are treated equally by HUD in assisting groups who lack affordable housing.
The strengths of 1990 census data for a local agency are access to the data (with a microcomputer and a CD-ROM reader) and software that enables them to make local decisions about housing areas. The current system for handling CHAS data has one major attraction: there is remarkable trust at the local level in the quality of the data. All the data were collected nationwide at the same time, with comparable questions asked of respondents in different states. No other local housing data provide such comparability. Local administrative data on housing, for example, vary markedly in the definition of a household or family. For instance, different metropolitan housing agencies use different definitions of "dependents." Administrative data have variations in the coverage of the lower income groups, which make the data worrisome for decisions about housing affordability. Finally, few local or state agencies have made quality checks on their housing data. With the CHAS data, housing officials are able to examine neighborhoods, which are re-emerging as a central focus for urban policy. A key use of the small-area data in the census is the ability to define neighborhoods locally.
Housing Policy Development
One example of the use of census data at the local level is that of the Mid-Ohio Regional Planning Commission, which in 1990 contracted with Roberta F. Garber Consulting and Ohio Capital Corporation for Housing to prepare a five-year CHAS for the city of Columbus and Franklin County (see Garber, 1994). This joint city-county planning process was intended not only to meet HUD's requirements for the CHAS, but, more importantly, to reach consensus among a broad array of public, private, and nonprofit actors regarding affordable housing goals and strategies.
One of the most challenging parts of this process was the collection, analysis, and presentation of data on affordable housing needs and the affordable housing market to the 35-member CHAS planning committee. Data were gathered from surveys, locally generated reports, focus groups, the 1990 census, and the 1991 AHS. Data analysis and presentation were key for successfully reaching
consensus on policy directions. There were, however, a number of obstacles to overcome.
Census data have inherent limitations in their use for housing policy development. First, they provide little direct data on physical housing conditions. Data on the year a structure was built and the number of vacant, boarded-up units, overlaid with data on household income or poverty, can help target areas in an urban community where there is a high likelihood of deteriorated housing. Data on units lacking complete plumbing and kitchen facilities—though useful in rural areas—have little relationship to housing conditions in urban areas. The other indicator of housing problems—extent of overcrowding—may or may not relate to housing condition and is a fairly minor housing problem in Columbus and Franklin County.
Second, until HUD requested special tabulations from the 1990 census in a CHAS databook, there was no way to use census data to understand the characteristics of households with housing problems. The special HUD tabulations available in 1993 were valuable in developing the most recent 5-year CHAS for Columbus and Franklin County. For the first time, data that described the number and characteristics of persons with housing cost burden, severe housing cost burden, overcrowding, and units lacking complete plumbing and kitchens were available. In addition, data on the affordability of the local housing stock and of vacant housing units were provided. For Columbus and Franklin County, it was possible to use the CHAS data book to determine, for example, that nearly 80 percent of renter households with severe cost burden (paying more than 50 percent of their income for rent and utilities) also had incomes below 30 percent of area median income. This indicated that most of the programs subsidizing the production of affordable rental housing, such as the low-income housing tax credits program, are still not serving the housing needs of the community. Typically those programs are designed to serve households at 50 to 60 percent of area median income. These data resulted in strategies to make significant changes in the city of Columbus and Franklin County housing programs in order to target the use of CDBG and HOME program funds to projects that serve households with the greatest need.
Although the CHAS data book was a great improvement in the availability of census data for affordable housing policy development, a major area of data was still missing, specifically, data that cross-tabulated housing condition with housing cost burden. The availability of these data is important because of the common perceptions local community leaders have about housing problems—that most housing problems involve poor housing conditions or lack of available affordable units. As a result of this misperception, strategies generally involve housing rehabilitation or new construction.
Fortunately, larger cities like Columbus have another data resource, the AHS, that provides a great deal of data on housing conditions and numerous cross-tabulations of data on housing condition with data on other household
characteristics. The AHS indicated that 80 percent of households with "worst case housing needs" in Columbus and Franklin County had only severe cost burden as a housing problem; only 17 percent lived in substandard housing. The availability of these data has significant implications for housing policy: for many households, what is commonly thought of as a "housing problem" is in reality as much or more so an "income problem." This is particularly true for Columbus and Franklin County, where there are also comfortable vacancy rates. This suggests a need to focus on economic self-sufficiency and rental assistance as integral parts of an affordable housing strategy. However, it should be noted that responding to the income support needs for families with high rent burdens involves more than assistance from HUD, which has a priority mission to improve the condition/supply of housing and condition of neighborhoods; it also involves other federal departments whose missions are focused on income and family support. In addition, 17 percent of households living in substandard housing is not insignificant, especially if condition of structure is related to neighborhood condition. Decennial census data are a critical resource for identifying local housing needs that go beyond one federal department of the local housing agency.
Unfortunately, the AHS also has a major limitation. It only aggregates data on a jurisdictional basis. There are no census tract-level or neighborhood-level data. Neighborhood-level planning and policy development are required for the CHAS, for the new HUD consolidated plan, and for the strategic planning process for the empowerment zone and enterprise community program. Lack of subarea data on housing conditions creates a significant void in planning at this scale. It is increasingly evident that housing, economic, and social data on a neighborhood basis—blocks and census tracts—are essential in designing and carrying out effective local programs. Recent studies have emphasized that it is important to measure neighborhood quality, as well as the quality of housing structures (Newman and Schnare, 1994). For smaller communities it may be possible to conduct windshield surveys to determine neighborhood housing conditions. For larger urban areas, this process can take years. In addition, lack of AHS subarea data makes it difficult to clearly describe to suburban officials the housing problems in their communities that need to be addressed.
Another example of the use of decennial census data for housing policy development is that of the California Department of Housing and Community Development. The deputy director of the department, Thomas B. Cook (1994), cites the use of these data in state housing planning, including the state review of required local housing planning documents, evaluation of financial assistance applications for both state and federal funds, and technical assistance both in the preparation of local planning documents and local loan and grant applications. This includes the use of small-area data from the census. The state has acquired and uses 1980 and 1990 cross-tabulations of population and housing items at
state and substate levels. It also utilizes decennial data as a benchmark for surveys that update information between census periods.
In conclusion, the need for the housing data provided in the decennial census certainly has not decreased in recent years, and in fact will continue to grow as the federal government adds increasing planning requirements to its programs. The focus on comprehensive and strategic planning means that housing data are important not only for housing planning, but also for planning related to economic development, welfare reform, crime reduction, and transportation. A minimal level of data is included in the census. Without special HUD data runs and the AHS, it would be very difficult to accurately link investments of resources with real needs. Unfortunately for rural areas, these other data resources are not available at low levels of geography, leaving the decennial census as the primary database. Fewer census questions pertaining to housing will only make it more difficult to effectively target scarce public and private resources.
Analysis, Development, and Evaluation of Urban Policy Initiatives
A loss of small-area data from the census would forestall a variety of important urban research and would impede the effective implementation of a number of urban programs that currently rely on such data. Many urban processes can be meaningfully examined and documented only using small-area census data.
Urban Poverty, Need, and Economic Trends
Poverty has become concentrated in massive and growing neighborhoods within cities. Data describing socioeconomic and demographic characteristics of census tracts or blocks are used to track the dimensions and causes of this type of poverty. Census small-area data can also permit at least crude analysis of mobility patterns in and out of poverty-stricken areas. The data are also used to analyze the effects that living in such areas has on the opportunities and achievements of residents. There is no substitute for census data as a basis for such analyses.
Over the past 20 years, urban experts have documented the importance of small-area geographic specialization of urban economies. Researchers have shown that office and retail trade trends in the downtowns of American cities have differed significantly from those elsewhere in big cities. Thus, place-of-work data describing trends in central business districts are helpful for documenting this trend. Such data have been provided from the decennial census.
Indeed, there is a general shortage of up-to-date employment information for even the municipalities. The Census Bureau's County Business Patterns provides annual data for counties, however, most cities are not coterminous with counties. The Census Bureau has the ability to compile a ''city business patterns" data series, though such an effort would require financial support. Such a
report would he invaluable for local, state, and federal economic development planning.
Research has shown that geographic access to jobs is a major determinant of unemployment and employment in urban areas. Thus, information on the spatial locations of workers and jobs is important for understanding unemployment. Those data also help in the analysis of urban workers' transportation needs. Federal laws such as the Intermodal Surface Transportation Efficiency Act and the Clean Air Act impose complex planning requirements on states and local governments that can only be accomplished using small-area data on housing, population, employment, and other characteristics of small areas that determine transportation demands. Local governments have a panoply of planning needs that also require such data.
The geographic level of detail required for such research and planning is difficult or impossible to specify without detailed knowledge of the particular area. Therefore, small-area data that enable the researcher or policy analyst to combine small areas flexibly into meaningful areas for analysis are needed.
Housing and Neighborhoods
Small-area data help document the characteristics and trends—physical and social—in urban neighborhoods. They also help identify and document the effects of abandonment, revitalization, social transition, and so forth.
Neighborhood patterns of income, race, and ethnicity have been shown again and again to affect the development of a neighborhood or community. Overall patterns of segregation are important indicators and determinants of housing, neighborhood, school, and job opportunities facing racial and ethnic minorities in urban areas. Studies of neighborhood segregation obviously require small-area data. Indeed, it is clear that the smaller the area the better the data. Census blocks play a central role in research on segregation.
The loss of small-area data would hinder the implementation of many of the most important urban programs in the nation. The loss would be particularly great for recently implemented urban programs.
If federal resources are to be used most efficiently, they must be accurately targeted to the places and people who need them most. Many federal programs use measures of local need in allocation formulas or eligibility standards. Some programs, such as the Community Development Block Grants, attempt to measure community need or distress using formulas that reflect conditions in entire municipalities or counties.
The most exciting new urban programs are being implemented at the neighborhood scale (e.g., community development programs). Community development
efforts can sometimes effect fundamental change in a community by instituting a comprehensive effort that has the support of residents, property owners, local governments, and others affected by the programs. The first new federal economic development initiatives in almost 20 years—enterprise and empowerment zones—also are targeted to either areas smaller than a municipality or overlapping areas of more than one locality. Small-area data are needed to evaluate the effects of these programs on the socioeconomic, demographic, and housing conditions in the community. Such data are also needed to target assistance and support to community development organizations.
Census data are also being used to target federal, state, and local efforts to ameliorate large concentrations of poverty.
ALTERNATIVE SOURCES OF CENSUS DATA
Many areas of the country (counties, cities, and towns) are not included or identified in the sample-based surveys. Sample sizes in small areas are too small to provide any confidence in the data.
Alternative sources of data on population; housing; income and poverty; labor markets; and mobility, health, ethnicity, and language include the Current Population Survey (for income and poverty data, the March supplement), AHS (conducted nationally every other year), and the AHS large metropolitan area data (collected for selected areas every four years). These surveys are sample-based and have fairly small sample sizes and very low levels of precision and accuracy. This is especially true for population data, income and poverty data, and labor market information, which have statutory requirements that specify the level of aggregation needed for meeting program and funding requirements.
A second alternative source is HUD administrative records, which capture information from the part of the population already receiving benefits from HUD programs. These records are limited, however, because they only cover a part of the population. Areas or families needing assistance cannot be identified with these records. Also, because the records contain little information on labor market characteristics, they not be very useful for obtaining labor market information. Model-based estimates cannot be used because they artificially reduce the amount of variance, which leads to incorrect allocations. Such estimates also rely on strong assumptions that are difficult to defend. These estimates could not be used to obtain labor market information because they are not independent assessments of program outcomes and conditions.
A final source for housing data is the use of random digit dialing telephone surveys. Such surveys have been used to determine fair market rents; however, they would be extremely costly to use in the 2,700 market areas for which data are needed.
As could be expected, there are no other consistent estimates on the homeless that cover all areas and places on a national basis. Prior survey estimates of
the homeless are dated (a national survey was conducted in 1984), did not measure the number of homeless (i.e., measured shelter bed capacity), or did not provide reliable estimates at the local level.
PRIORITIES FOR HOUSING DATA
Data From Future Censuses
There is the possibility that future censuses may not continue to provide data for small geographic areas such as census tracts, census blocks, and small cities or towns. The loss of such data would have profound, adverse impacts on federal urban policy and research and would impede the effective targeting and administration of a number of important urban programs. It would also harm state and local policy and programs. Vitally important urban research and urban programs would be crippled if high-quality data for small areas are not provided by the Census Bureau in the future.
Franklin J. James, professor of public policy at the University of Colorado at Denver, previous director of the U.S. Department of Housing and Urban Development's legislative and urban policy staff and still active in urban policy organizations, notes that high-quality data for small areas are necessary for accuracy in defining urban conditions (James, 1994):
I have a variety of experience which enables me to judge the value of such small area data. As the director of HUD's Legislative and Urban Policy Staff, I was involved in the urban policy development during President Carter's administration. I remain active in urban policy research. Recently, HUD asked me to assist in preparing a background paper on urban economic development issues for President Clinton's urban policy. Moreover, I am on research advisory committees which keep me up to date on research and policy issues. Examples are the advisory committee of the Community Development Research Center of the New School for Social Research, and the research advisory panel of the Federal National Mortgage Association.
Improving Data on the Condition of Housing Structures
Research and analysis of housing quality has been hampered by the lack of useful data. While independent researchers could undertake housing surveys to fill this gap, such surveys would be too costly or lack the credibility and reliability of data produced by the Census Bureau. Unfortunately, the decennial census is not an adequate source for housing quality statistics, and other Census Bureau data have significant limitations too. With only minimal additions to the decennial census, data on housing quality could be expanded tremendously, thus making existing Census Bureau data considerably more useful.
Outdated and Too Little Data
The only real indicators of housing quality in the decennial census are the questions regarding lack of a full bath and lack of a complete kitchen. Through the middle of this century, indoor plumbing was a reasonable gauge of the quality of a home. However, these two housing features have outlived their usefulness since only about 1 percent of all housing units in the United States lack complete plumbing facilities (Bureau of the Census, 1993).
F. Edward Geiger, III, senior policy analyst of the Pennsylvania Department of Community Development, emphasizes the need for improved data in the decennial census on condition of housing structures for individual communities and by census tracts and makes recommendations for additional condition items (see Table H.1; see also Geiger, 1994). This need is critical in Pennsylvania because 72 percent of Pennsylvania's housing stock is over 30 years old and only 42 percent of the state's municipalities have enacted building codes (Bureau of the Census, 1992). Yet, a scant 1-2 percent of the state's housing would be considered substandard using lack of kitchen or bath as the measure of decent quality housing (Bureau of the Census, 1992). Common sense tells us that this decennial census information does not measure up to the task.
The AHS is the most comprehensive source of housing quality data in the nation. The survey covers roughly 150,000 housing units and contains a wealth of housing quality data. The AHS examines condition, structural features and equipment, financial characteristics, and even neighborhood quality (Bureau of the Census, 1991). However, even the AHS has its limitations. It does not provide data for individual communities (much less census tracts) and does not provide detailed reports for areas other than the 44 metropolitan areas covered by the survey.
Additional Decennial Census Questions
Assuming it is not feasible to expand the AHS to cover larger portions of the nation, a few minor additions to the decennial census's long form should be considered. Adding additional questions to the decennial census would improve the data available for all areas (including metropolitan and nonmetropolitan areas) and all levels of geography (state, county, county subdivisions, and tracts/block numbering areas). The additions would also provide the opportunity to generate even more detailed estimates of housing quality by extrapolating with data from the AHS.
The additional questions should cover the operation and breakdowns of major housing systems: plumbing, heating, electrical, and roofing. One or two questions about each of these housing features (about 8 questions in total) would vastly expand the housing quality information gathered at a relatively minimal cost compared to expanding the AHS to more homes in other areas of the country.
(The existing questions about plumbing and kitchens might be eliminated to help limit the length of the census form.) Although example questions are suggested in Table H.1., the questions should be selected based on their relationship to other housing quality features covered in the AHS. With additional research, questions could be developed that would facilitate the extrapolation of housing quality data. For example, additional research might determine that x percent of homes having broken toilets also have serious leaks inside the structure from pipes or plumbing fixtures. If the decennial census gathered statistics on broken toilets, a good estimate of homes with leaking pipes would be available. Using this methodology, many housing quality features found in the AHS could be extrapolated from the slightly expanded decennial census. Both the actual census data and the extrapolated data could provide detailed information about the quality of housing in specific communities and other small geographic areas.
In addition to the research needed for extrapolation purposes, the added questions should be tested to determine how accurately this information can be self-reported in the long-form questionnaire. In order to truly improve housing quality data, the responses from the housing quality questions must be reliable.
One major issue neglected to this point is why the additional data on housing quality are needed. These additional data could be used to determine the level of funding for housing programs and how those funds should be used. Government agencies that fund housing programs could use this information to determine where additional monies are needed for housing rehabilitation rather than new construction. The expanded housing quality data also could be used to determine whether housing programs are effective. Over the long-term, it is unclear whether the quality of the housing stock is improving or declining. These additional data may help answer the question.
The private sector could also use the expanded housing quality data. Home repair and maintenance firms could use this information to target markets and determine the need for their products and services. The manufactured housing industry (commonly referred to as mobile home builders) could verify whether its housing is sound and durable. The industry could use the improved census data to compare the quality of manufactured homes to traditional, site-built housing.
Data for Small Areas
High-quality data for small areas are necessary for the accurate analysis and diagnosis of urban problems, and for the development, implementation, and evaluation of urban programs. Survey data such as the Current Population Survey or the AHS can frequently substitute for census data when the focus is on large geographic areas such as the nation, populous states, or major metropolitan areas. Indeed, survey data are sometimes superior to census data for such purposes because survey questionnaires can be focused and fine tuned for particular
tasks. However, it is prohibitively expensive to use surveys to describe situations and trends in small areas. For technical reasons, much the same sample size is required for accurate survey data for a small area as is required for a survey of a large area. Sample size is among the most important determinants of the costs of a data collection effort.
Small-area data from the Census Bureau have wide-ranging usefulness because they provide data of fundamental importance to a wide range of research and policy analysis, such as demographic information, labor-force and employment information, and income and housing descriptions. From the vantage point of the federal government, it is much more efficient to provide such basic data from the census than to finance or require others to finance the large numbers of special surveys that would be needed to replace census data for small areas. The marginal cost of adding additional data items to an ongoing survey or census is small. Similarly, the potential savings from dropping individual questions is also small if other survey parameters are unchanged. Moreover, the loss of quality census data describing small areas would impede special surveys that might be done to replace the lost data. Currently, census data increase the efficiency of special surveys. Generic data from the census can help target special survey efforts to the places where the survey is most appropriate and facilitate the design of efficient lists for more detailed sample surveys. The data provided by the census are used to delineate areas for special studies. This targeting function increases the efficiency of a large number of special surveys in small areas.
CONSEQUENCES OF NOT COLLECTING CENSUS DATA
There would be severe consequences for HUD programs if census data were not available for population and housing characteristics. First, HUD would be unable to meet legislatively mandated requirements for several programs—community development block grants, emergency shelter grants, low-income housing tax credits, and mortgage revenue bonds. Second, lower levels of accuracy and precision associated with other data sources would lead to poorly targeted programs, inefficient use of limited federal resources, and seriously inequitable distributions of program benefits. Third, HUD would not be able to effectively carry out its civil rights and fair housing responsibilities.
There would be other consequences if census data were lacking for special characteristics. Without census data on income and poverty, HUD would be unable to meet legislatively mandated requirements for several programs—community development block grants, emergency shelter grants, low-income housing tax credits, and mortgage revenue bonds. Second, lower levels of accuracy and precision associated with other data sources would lead to poorly targeted programs, inefficient use of limited federal resources, and seriously inequitable distributions of program benefits. Without census data on labor markets, the
President's Urban Policy Report would not be able to report on the required conditions or would have to rely on other unsatisfactory sources. CDBG evaluation would be difficult or impossible. Lack of information on disabilities would lessen the ability of HUD to effectively target Section 202 grants, for local governments to include this special needs group in their CHAS, and for the Office of Fair Housing and Equal Opportunity to determine need for accessible housing. Without census data on the homeless, decisions concerning funding allocations, program design, and national priorities will continue to be made in an information vacuum.
Barrett, D.F., I. Anolik, and F.H. Abramson 1992 The 1990 Census Shelter and Street Night Enumeration. Paper presented at the annual meeting of the American Statistical Association, Boston, Mass. Bureau of the Census, Washington, D.C.
Bureau of the Census 1991 What is the American Housing Survey? Washington, D.C.: U.S. Department of Commerce. 1992 Age of Pennsylvania's Housing Stock: 1990 Census of Population and Housing, Summary Tape File 3A, prepared by Pennsylvania State Data Center, May. 1993 Statistical Abstract of the United States, 1993. Washington, D.C.: U.S. Department of Commerce.
Cisneros, H.G. 1993 Response of the Department of Housing and Urban Development to Office of Management and Budget Request to Identify Topics to Be Included in the 2000 Census. Letter and Attachment to Philip Lader, Office of Management and Budget, June 18.
Cook, T.B. 1994 Letter dated March 3 to B. Edmonston. Division of Housing Policy Development, Department of Housing and Community Development, Sacramento, Calif.
Garber, R.F. 1994 Use of Decennial Census Data in Housing Policy Development. Paper prepared for the Panel on Census Requirements in the Year 2000 and Beyond, Committee on National Statistics, National Research Council, Washington, D.C.
Geiger, F.E. III 1994 Improving the Quality of Housing Quality Statistics. April. Paper prepared for the Panel on Census Requirements in the Year 2000 and Beyond, Committee on National Statistics, National Research Council, Washington, D.C.
James, F.J. 1994 Priorities for Data from Future Censuses. Memorandum dated April 20 to B. Edmonston. Panel on Census Requirements in the Year 2000 and Beyond. Graduate School of Public Affairs, University of Colorado, Denver, Colo.
Newman, S.J., and A.B. Schnare 1994 Back to the Future: Housing Policy for the Next Century. Baltimore, Md.: Johns Hopkins University.
Pennsylvania Department of Community Affairs 1994 Municipalities without Building Codes: Local Land Use Controls in Pennsylvania, 2nd ed. Harrisburg, Pa.: Pennsylvania Department of Community Affairs.
TABLE H.1 Suggested Additional Decennial Census Questions on Housing Condition
Data Need Addressed
During the past year, have all the toilets in this housing unit been broken for 6 hours or more?
Availability of adequate toilet and wastewater disposal
During the past year have plumbing problems prevented you from using a sink, bath, or shower for longer than 6 hours?
Presence of adequate plumbing system
Were you uncomfortably cold last winter for more than 24 hours because of problems affording or operating heating equipment in your home?
Affordability and operation of heating system
Did your heating equipment last winter break down at least 3 times for 6 hours or more?
Adequate maintenance and repair of heating equipment
Does your home have any exposed wiring?
Presence of serious electrical hazards
Does your home have any room where all the electrical outlets do not work?
Adequate maintenance and operation of electrical system
Did you have water leaks in your home because the roof remained unrepaired for longer than 3 months?
Adequate roof construction and maintenance
Did you have water leaks in your home because the windows, walls, or doors remained unrepaired for longer than 3 months?
Soundness of structure and maintenance (other than roof)
Source: Geiger (1994).