The primary data sources for this report are the Current Population Survey (CPS), the Survey of Income and Program Participation (SIPP), the Panel Study on Income Dynamics (PSID), and administrative data for the AFDC/TANF, Food Stamp, and SSI programs. Beginning with the 2001 report on dependence, there was a shift to using CPS rather than SIPP data for several indicators and predictors of welfare recipiency and dependence. This change was necessary because CPS data are updated annually, while SIPP updates are available much less frequently.
If it were not for the lags in data availability, the SIPP would be considered the most useful national survey for measuring welfare dependency. It was used most extensively in the first three annual dependence reports. Its strengths are its longitudinal design, system of monthly accounting, and detail concerning employment, income and participation in federal income-support and related programs. These features make the SIPP particularly effective for capturing the complexities of program dynamics. It continues to be an important source of data in this report, particularly for measures related to AFDC/TANF spell duration and transitions in and out of AFDC/TANF recipiency, dependency, and poverty. This 2003 report provides the first updated analysis of SIPP data beyond 1995, allowing examination of program dynamics under the TANF program.
For measures of receipt, dependency, and poverty at a single point in time, however, the report primarily uses the Annual March Demographic Supplement to the CPS, which measures income and poverty over an annual accounting period. As stated above, the CPS data are available on a more timely basis than the SIPP, and have been widely used to measure trends since the welfare reform legislation of 1996. However, because the CPS does not collect income in the same detail as the SIPP, it has been subject to criticism for underreporting of income, particularly welfare income. To address this concern, some of the indicators in this report are based on CPS data that have been analyzed by the Transfer Income Model (TRIM3), a microsimulation model developed by the Urban Institute under contract to the Office of the Assistant Secretary for Planning and Evaluation. Although its primary purpose is to simulate program eligibility and the impact of policy proposals, the TRIM model has also been used to correct for underreporting of welfare receipt and benefits. Welfare caseloads in TRIM3 are based on CPS data, adjusted upward to ensure that total estimates of recipients equal the total counts from administrative data. Even with these adjustments, some measurement differences between the CPS/TRIM data and SIPP data remain.
As shown in Figure SUM 3, the overall measures of dependency and recipiency have not been greatly affected by the change in data sources. Both data sources show a decline in dependence between 1996 and 1999, from 4.7 to 2.8 percent under the SIPP data, and from 5.2 to 3.3 percent under the TRIM-adjusted CPS data. Still, readers are cautioned against comparing measures for 1987-1995 from the SIPP data in the first three annual reports with the measures for 1996-2000 from the TRIM-adjusted CPS data. In Chapter II, indicators using the CPS data have been analyzed for every year since 1993 (the first year for which TRIM-adjusted CPS data are available), providing a new time series of how the indicators are changing over time from a consistent data source.
Figure SUM 3. Recipiency and Dependency Rates from Two Data Sources: 1987-2000
Note: Recipiency is defined as receipt of any amount of AFDC/TANF, SSI, or food stamps during year. Dependency is defined as having more than 50 percent of annual family income from AFDC/TANF, SSI and/or food stamps. Dependency rates would be lower if adjusted to exclude welfare assistance associated with working.
Source: March CPS data, analyzed using the TRIM3 microsimulation model.
The Panel Study of Income Dynamics (PSID) is another source of data used in this report. Like the SIPP it provides longitudinal data, but over a much longer time period than the three- to four-year time period of the SIPP. The PSID has collected annual income data, including transfer income, since 1968, providing vital data for indicators of long-term welfare receipt, dependence, and deprivation. The PSID measures cover time spans as long as a decade and so are updated less frequently than the CPS-based and SIPP-based measures. The PSID measures in this year's report are unchanged from last year's report, and generally cover the decade ending in 1996.
Finally, the report also draws upon administrative data for the AFDC/TANF, Food Stamp and SSI programs. These data are largely reported in Appendix A. Like the CPS data, administrative data are generally available with little time lags; these data are generally available through fiscal year 2001. To the extent possible, TANF administrative data are reported in a consistent manner with data from the earlier AFDC program, as noted in the footnotes to the tables in Appendix A. The fact remains that assistance under locally designed TANF programs encompasses a diverse set of cash and non-cash benefits designed to support families in making a transition to work, and so direct comparisons between AFDC receipt and TANF receipt must be made with caution. This issue also affects reported data on TANF receipt in national data sets such as the CPS and SIPP.
Most of the data sources allow analysis of the indicators and predictors of welfare dependence across several age and racial/ethnic categories. Where the data are available, statistics are shown for three racial/ethnic groups – non-Hispanic whites, non-Hispanic blacks, and Hispanics.4 In some instances, however, there are not sufficient data on individuals of Hispanic origin, and so the measures are shown for only two racial/ethnic categories. For the primary measure of dependency in this 2003 report, estimates are also provided for individual persons by family structure (see SUM1 and IND1). For these measures, the entire population is subdivided into the following four groups:
- individuals in married-couple families
- individuals in female-headed families, no spouse present
- individuals in male-headed families, no spouse present
- unrelated individuals.
Two technical notes concern the unit of analysis and the difference between annual and monthly measures. The individual, rather than the family or household, is the unit of analysis for most of the statistics in this report. The individual’s dependency status, however, is generally based on total family income, taking into account means-tested assistance, earnings and other sources of income for all individuals in the family.5 This chapter, for example, has reported the percentage of individuals that are dependent (in SUM 1) or poor (in SUM 2) according to annual total family income. Recipiency status is also based on total annual family income in some instances; in SUM 1, for example, recipients are individuals in families receiving assistance at some point in the year. In most other indicators, recipiency is measured as the direct receipt of a benefit by an individual in a month. The difference between an individual and a family measure of recipiency is largest in the SSI program, which provides benefits to individuals and couples, not to families.
Finally, there also are differences between monthly and annual observation of benefit receipt. For example, the measures of annual recipiency (that is, any receipt over the course of a year) shown in Figure and Table SUM 1 are higher than the more traditional measures of recipiency in an average month, as shown in several other indicators.