Data and Methodological Issues for Tracking Former Welfare Recipients: A Workshop Summary (1999)

leave welfare when the economy is strong or who returns to welfare when the economy is weak can help predict future program needs.

Finally, this part of the workshop discussion focused on how the grants could be used to make evaluation and causal inferences about the new welfare policies. Researchers discussed potential problems in making causal inferences with these studies. Comparing the two cohorts (pre and post-PRWORA cohorts) for making causal policy inferences is problematic because the composition of the caseload during these two times may be different and because the socioeconomic environment may also differ. Serious efforts to control for these factors will require that quality data on local labor market and economic conditions and historical data on characteristics of cohort members is collected. The short-term nature of the study periods (most studies follow recipients for 1 year) also limits the ability to make causal conclusions about the long-term effects of the policies.

A major topic of discussion at the workshop centered on what outcomes of former recipients should be measured and monitored. The state plans propose to examine many outcomes of those who have left welfare. The following is a generalized list of policy questions of interest to states:

• What are the employment situations of those who leave welfare?

• Are those who leave welfare properly trained for gaining employment?

• Do those who leave welfare return to cash assistance and/or do they use other public assistance programs?

• How are the children of former recipients doing?

• Are the child care needs of former recipients being met?

• Are those who have left welfare living in material hardship?

• What informal supports and privately provided supports are former recipients using?

• What barriers to employment do former recipients face?

Beyond these general questions of interest, there is variation in the specific outcomes grantees are measuring, and much of the discussion focused on which outcomes should be measured. There are tradeoffs between the depth at which surveys and administrative data can provide information and the breadth of the information provided. While collecting information on a wide range of outcomes will give a good overall picture of the circumstances of those who have left welfare, it may also limit the depth at which information can be collected because resources to add questions to surveys are limited and because administrative data are limited in the outcome measures available.

One area in which the tradeoff between the depth and breadth of outcomes was particularly apparent is child outcomes. Most grantees agreed that the effects of policy changes on children of former recipients is a major policy interest. However, some said that with the small surveys that each state was conducting, there is little time and funding to develop good scales and measures of

child outcomes. Many argued that the short length of time for which leavers will be followed is not enough time to observe effects of policies on child outcomes. Consequently, many researchers from states and counties have decided to include only a few child outcomes.

Other participants said that the child well-being outcomes are very important because collecting such data now can at least provide a baseline measure of child well-being against which future comparisons can be made. Furthermore, there has been some research on developing survey modules and measures of child outcomes that are of varying lengths and that can be used to assess child development, health, and well being (Child Trends, 1999). Child Trends also developed the self-administered qestionnaire for adolescents in the Survey of Program Dynamics. Modules like these can be used as models for survey questions. A survey developed by DHHS's National Center for Health Statistics, State and Local Area Integrated Telephone Survey (SLAITS), incorporated some of these questions and was also suggested as a model.

Most participants agreed that future work needs to be devoted to examining how administrative data sources can be used to map outcomes normally collected in surveys. Using existing administrative data sets is generally much less expensive than funding new surveys, and with limited research budgets, they are a possible alternative to large surveys. For child outcomes, an example is the use of public school administrative records to measure student attendance and, possibly, even performance at school.

Another set of outcomes that are potentially important for assessing the circumstances of those who have left welfare include family structure, family formation, and fertility. These outcomes encompass marriage and marital dissolution, household living arrangements, family functioning, and the fertility behaviors of recipients and former recipients. For women and their children, marrying a spouse who is employed is one path to greater economic security. Sharing household quarters with other working adults (relatives, cohabitating partners, friends) is another means by which former recipients can find support or cut expenses. One aspect of family functioning is the degree of conflict within a family. The extent to which recipients are victims of domestic abuse and the extent to which such violence is a barrier to employment are also important. These outcomes were highlighted as important outcomes to assess the well-being of adults and children and should be included in data collection efforts.

To obtain data on household living arrangements of recipients, some participants encouraged the use of surveys. Because there are strong incentives not to report all household members and their earnings to a caseworker (in order to obtain more benefits), administrative data can often underestimate the number of people living in the household of a former recipient.

Another issue for assessing the status of households of former welfare recipients is how to collect information on other household members other than the former recipient and her children. The emphasis on collecting income data on other household members reflects the need to measure the amount of resources available to a household. Most grantees will rely on survey data for this information. It was suggested that the Social Security numbers of household members could be obtained through a survey and would make it possible to link their earnings data from UI records to those of the recipients with whom they live. One problem with this approach is that it may impose an extra burden to obtain human subjects permissions if the Social Security numbers and records of