In the early 1990s the Census Bureau began the Small Area Income and Poverty Estimates (SAIPE) Program to meet growing demands for regularly updated estimates of key income and poverty measures for subnational areas, such as states, counties, and school districts. SAIPE is not the first program for providing more frequent estimates than are provided by the decennial census of population, but it is the most ambitious effort of its type to date (see below, “Background”).
The first SAIPE estimates were issued in early 1997 for states and counties for income year 1993. Estimates for states and counties for income year 1995 were issued in early 1999; also issued at that time were estimates of the numbers of poor school-age children for school districts in 1995. The SAIPE estimates were developed by using a variety of survey, census, and administrative records data sources with statistical modeling techniques.
An important application of intercensal small-area estimates of poverty is for the allocation of over $7 billion of funds annually for programs for educationally disadvantaged children under Title I of the Elementary and Secondary Education Act. Reauthorization of that act in 1994 provided that updated estimates of poor school-age children for counties and school districts, produced every 2 years by the Census Bureau, should be used for Title I allocations in place of estimates from the most recent census, unless the Secretaries of Commerce and Education found that they were “inappropriate or unreliable” on the basis of a review by a panel of the Committee on National Statistics at the National Research Council.
Below are the first 10 and last 10 pages of uncorrected machine-read text (when available) of this chapter, followed by the top 30 algorithmically extracted key phrases from the chapter as a whole.
Intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text on the opening pages of each chapter. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.
Do not use for reproduction, copying, pasting, or reading; exclusively for search engines.
OCR for page 11
Small-Area Income and Poverty Estimates: Priorities for 2000 and Beyond 1 Introduction In the early 1990s the Census Bureau began the Small Area Income and Poverty Estimates (SAIPE) Program to meet growing demands for regularly updated estimates of key income and poverty measures for subnational areas, such as states, counties, and school districts. SAIPE is not the first program for providing more frequent estimates than are provided by the decennial census of population, but it is the most ambitious effort of its type to date (see below, “Background”). The first SAIPE estimates were issued in early 1997 for states and counties for income year 1993. Estimates for states and counties for income year 1995 were issued in early 1999; also issued at that time were estimates of the numbers of poor school-age children for school districts in 1995. The SAIPE estimates were developed by using a variety of survey, census, and administrative records data sources with statistical modeling techniques. An important application of intercensal small-area estimates of poverty is for the allocation of over $7 billion of funds annually for programs for educationally disadvantaged children under Title I of the Elementary and Secondary Education Act. Reauthorization of that act in 1994 provided that updated estimates of poor school-age children for counties and school districts, produced every 2 years by the Census Bureau, should be used for Title I allocations in place of estimates from the most recent census, unless the Secretaries of Commerce and Education found that they were “inappropriate or unreliable” on the basis of a review by a panel of the Committee on National Statistics at the National Research Council.
OCR for page 11
Small-Area Income and Poverty Estimates: Priorities for 2000 and Beyond The 1994 act also provided for the Panel on Estimates of Poverty for Small Geographic Areas to carry out that task. The panel's findings, which supported use of the estimates, were published in three interim reports (National Research Council, 1997, 1998, 1999). The panel subsequently combined the three interim reports into a single technical volume (National Research Council, 2000c), which documents the current methods for producing SAIPE estimates of poor school-age children for states, counties, and school districts, and the evaluations of them that have been conducted to date. The technical volume is designed to complement this report. In this, its final report, the panel addresses its broader charge to review the SAIPE Program as a whole. The report offers recommendations to the Census Bureau for future research and development that can lead to improved SAIPE estimates for use in programs, such as Title I and others, that require updated small-area income or poverty estimates for such purposes as fund allocation. The report also identifies issues that user agencies need to consider in deciding to adopt small-area income and poverty estimates from SAIPE or other sources for program purposes. BACKGROUND A large and growing number of federal programs use small-area income and poverty estimates for allocating funds to states and localities, providing matching funds for state expenditures, performance monitoring, and program evaluation. This trend reflects a shift in policy away from individual entitlement programs to formula block grant programs. An example is the 1996 Personal Responsibility and Work Opportunity Reconciliation Act, which replaced the Aid to Families with Dependent Children entitlement with the Temporary Assistance to Needy Families (TANF) block grant. Data sources used for fund allocation and related federal program purposes include the decennial census, SAIPE estimates, and small-area personal income estimates from the Bureau of Economic Analysis (BEA). States also allocate both federal and state funds to localities under formulas that target less well-off areas or populations. Data sources used by the states include the decennial census, the new SAIPE estimates, and, for many programs, state and local administrative records that relate to poverty or income, including TANF caseloads, numbers of children approved for free or free and reduced-price meals under the National School Lunch Program, and, in a few instances, income estimates from state tax records. Historically, decennial census estimates have been the most frequently
OCR for page 11
Small-Area Income and Poverty Estimates: Priorities for 2000 and Beyond used data source for federal programs that require small-area estimates of income and poverty, and they have been used by many state programs as well. However, for most areas, census estimates decline in accuracy over the years between census enumerations and, consequently, increasingly misrepresent the distribution of income or poverty as the period since the last census lengthens. For example, from 1989 to 1993, not only did the United States as a whole experience a 21 percent increase in the number of poor people because of economic recession, but the increase in the poor population was not uniform across the nation. Some states experienced greater than average increases in their poor populations (e.g., 52% in Florida, 44% in California), while other states experienced smaller than average increases (e.g., 4% in Texas, 7% in Illinois).1 The Census Bureau originally began the SAIPE Program in response to inquiries about updated estimates of per capita income for local governmental jurisdictions. Such estimates had been regularly produced by the Census Bureau every 2 years from 1971 to 1987, using changes in income reported on tax returns and BEA personal income estimates to update census income estimates (see National Research Council, 1980, for a review of the methodology). The estimates were used in allocating funds to 39,000 local governments—states, counties, cities, towns, and other units—under the General Revenue Sharing Program; but when that program expired, the Bureau discontinued its per capita income series. Efforts to build a consortium of federal agencies to fund a broader set of SAIPE estimates got under way in early 1993. By August, five federal agencies had agreed to provide sufficient funding to initiate the project: the U.S. Department of Agriculture, Food and Nutrition Service; the U.S. Department of Education, National Center for Education Statistics; the U.S. Department of Health and Human Services, Head Start Program; the U.S. Department of Housing and Urban Development, Office of Policy Development and Research; and the U.S. Department of Labor, Employment and Training Administration. The Statistics of Income Division (SOI) of the Internal Revenue Service (IRS) agreed to become a partner in the project, and its data were essential for several anticipated estimation methodologies. About the same time, Congressman Thomas C. Sawyer (D-Ohio), who then chaired the House of Representatives Subcommittee on Census, Statistics, and Postal Personnel, organized a hearing on the need for 1 From tabulations of the March Current Population Survey (CPS) (see the Census Bureau's web site: http://www.census.gov/hhes/www/saipe.html). The United States subsequently experienced a 12 percent decrease in the number of poor people from 1993 to 1998; however, state differences in the rate of decline were not large enough to be reliably distinguished by the March CPS.
OCR for page 11
Small-Area Income and Poverty Estimates: Priorities for 2000 and Beyond more current measures of poverty for small areas and how those measures might be developed. Following this hearing, Congressman Sawyer introduced authorizing legislation requiring the Secretary of Commerce to develop methods to produce “intercensal data relating to the incidence of poverty for each State, county, and local jurisdiction.” The legislation further called for estimates of the numbers of poor children aged 5-17 for school districts and of the numbers of poor people aged 65 and over for states and counties. The legislation was passed by the House of Representatives in November 1993, but the Senate did not act on it. One year later, in September 1994, Congress passed the “Improving America's Schools Act,” which called for the use of updated Census Bureau estimates of poor school-age children for allocation of Title I funds, if they were found sufficiently reliable by a panel of the National Research Council. SAIPE IN BRIEF The main objective of the SAIPE program is to produce updated income and poverty estimates for the administration of federal programs, including the allocation of federal funds to local jurisdictions. At present, SAIPE provides the following estimates: For states, biennially beginning in early 1997 (for income year 1993) and annually beginning in 1999 (for income year 1996): median household income; number of poor people; number of poor children under age 5; number of poor related children aged 5-17;2 and number of poor people under age 18. For counties, biennially beginning in early 1997 (for income year 1993):3 median household income; number of poor people; number of poor related children aged 5-17; and number of poor people under age 18. 2 Related children include family members in a household under age 18, except married sons, daughters, or spouse of the householder and foster children (see the Census Bureau's web site for a detailed definition and the reason for using it: http://www.census.gov/hhes/www/saipe.html). 3 The county estimates (for income year 1993) were revised and reissued in early 1998.
OCR for page 11
Small-Area Income and Poverty Estimates: Priorities for 2000 and Beyond For school districts, biennially beginning in early 1999 (for income year 1995): number of poor related children aged 5-17. In addition, the Census Bureau produces small-area population estimates of the numbers of persons in relevant age groups, which can be used to construct estimates of poverty rates. Population estimates by age are developed annually for states and counties from a demographic estimates program that has been active for many years. For school districts, the Census Bureau is now producing biennial estimates of the numbers of school-age children and the total population; the first estimates were issued in 1999 for July 1996. The SAIPE estimates are developed by using a variety of data sources with statistical estimation methods. There is at present no single data source—either a sample survey, such as the March CPS, or an administrative records system—that can be used to produce reliable direct estimates between decennial censuses.4 Instead, multiple sources must be combined in statistical models to produce reliable indirect estimates. Model-dependent indirect estimators use data from other areas and, possibly, other time periods that are obtained from several sources to “borrow strength” and improve the precision of estimates for small areas.5 The basic methodology was first developed several decades ago, and the Census Bureau has used this strategy for several types of estimates. Specifically, it used model-dependent methods in the 1970s to improve 1970 census small-area income estimates for use in developing updated per capita income estimates for governmental jurisdictions (Fay and Herriot, 1979) and, in part, to develop population estimates for states and counties (see National Research Council, 1980:App. A). More recently, it used model-dependent methods to estimate median family income for states (Fay, Nelson, and Litow, 1993). The SAIPE estimates for states incorporate predicted values from statistical regression models that predict state poverty or income in the March 4 Census income and poverty estimates, which are based on the long-form sample, themselves exhibit a considerable degree of error due to sampling variability for many small areas, in addition to other kinds of error, such as misreporting. 5 By “model-dependent” we mean that the accuracy of the estimates depends on the validity of the assumptions of the estimation model: see Marker (1999) and Rao (1999) for overviews of small-area estimation methods; see U.S. Office of Management and Budget (1993) for definitions of direct and indirect estimators and other terms used in the research literature.
OCR for page 11
Small-Area Income and Poverty Estimates: Priorities for 2000 and Beyond CPS on the basis of variables derived from such sources as IRS tax returns, food stamp records, the 1990 census, and population estimates. The regression predictions are combined with the corresponding direct state estimates from the March CPS by using a procedure in which the weights given to the predicted values and the direct estimates depend on their relative precision. Similar procedures are used for county estimates, although the state and county models differ in several respects. For poverty measures (but not median household income), each set of county estimates is adjusted by state to sum to the applicable SAIPE state estimates. For school districts, a different procedure is used to develop school-age poverty estimates because of the lack of district-level data from administrative records that could be used to form predictor variables in a school district model similar to the SAIPE county model. Instead, school district estimates are produced by using 1990 census data to calculate the percentage shares or proportions of poor school-age children for the school districts (or parts of school districts) in each county, using more recent district boundaries from a special survey that is conducted every 2 years. These shares are then applied to updated SAIPE county estimates of poor school-age children. PLAN OF THE REPORT This report has two goals: to identify and discuss key issues for federal agencies and other users when applying small-area income and poverty estimates for such purposes as fund allocation; and to outline priority areas for research and development for the Census Bureau that can lead to improved small-area income and poverty estimates from the SAIPE Program. Chapter 2 focuses on uses and users. It describes the growing needs for small-area income and poverty estimates for federal and state program purposes, important criteria for such estimates (e.g., timeliness, appropriate concept of poverty or income), and the extent to which available data sources satisfy the criteria. Chapter 3-Chapter 5 focus on the SAIPE estimates, which the panel believes will be more widely used in the future. Chapter 3 provides a technical discussion of the SAIPE estimation models, focusing on the models for poor school-age children for states, counties, and school districts. It summarizes earlier evaluations of the models and lists priorities for model research and development, previously identified by the panel, that the Census Bureau should pursue in the short to medium term (see also National Research Council, 2000c). Chapter 4 and Chapter 5 discuss, respectively, new survey data sources, such as the 2000 census and the planned Ameri-
OCR for page 11
Small-Area Income and Poverty Estimates: Priorities for 2000 and Beyond can Community Survey, and improvements in administrative records that, if implemented, could support major improvements in the SAIPE estimates in the longer term, particularly for such subcounty areas as school districts. Chapter 6 returns to the user perspective, considering interactions between the properties of small-area income and poverty estimates and the properties of allocation formulas. Estimates will always have some degree of error or uncertainty, and users must be aware of the unintended effects that errors in estimates, in conjunction with formula features, may have on allocations. Chapter 7 provides general recommendations to users and producers of small-area income and poverty estimates from SAIPE (and other sources) in the following areas: practices that are important to follow in the production, evaluation, and documentation of estimates; assessments that users should conduct of estimates; and the need for policy makers to consider carefully the use of estimates for fund allocation and other program purposes in light of their uncertainty. The appendix presents some results of simulating fund allocations with varying levels of uncertainty of estimates and different rules for allocating funds. Chapter 2, Chapter 6, and Chapter 7 are addressed primarily to administrators, analysts, and policy makers in federal agencies who are considering the use of estimates of income and poverty for programs. Chapter 3, Chapter 4, and Chapter 5, which are more technical, are addressed primarily to researchers in small-area estimation, including the staff at the Census Bureau. However, each of the technical chapters includes an overview of the key points for nontechnical readers.