Existing studies address several different purposes. The priority assigned to these purposes affects the design of the study.
- Measure the outcomes of reforms or particular policies. Surveys and administrative data can provide good information about whether former recipients are working, what kinds of jobs they are getting, how much they are earning, and whether they are facing hardships. Because specific outcomes are the focus, the study could use a sample of former recipients. Sampling enables a study to concentrate its resources to ask in-depth questions and achieve the high response rates needed to have confidence in the findings.
- Identify barriers to recipients finding work so policies can be changed. Personal interviews or telephone surveys are the best ways to collect information about recipients' barriers. For this purpose, a sample is a valid means of collecting information. A survey could focus on former recipients who are not working or who are unable to support their families. Queries about barriers to employment could be an additional set of questions on a survey directed only to recipients who are unemployed or underemployed. (Alternatively, questions about employment barriers could be added to a survey directed to recipients identified as not working through unemployment insurance or new-hire registry records.) The survey might ask questions about access to child care, transportation, health benefits, and training.
- Identify families at risk of extreme hardship or neglect and abuse. If this is the primary purpose, the study should include all families that leave welfare or, at the very least, all families that are sanctioned. Sanctioned families are those most likely to suffer hardships. Home visits are the best approach to identify these families because they permit first-hand observation of the families' circumstances.
- Evaluate welfare reforms to determine their impact. A scientific evaluation of policy impacts is the most comprehensive and expensive form of study. The best research design involves randomly assigning recipients to treatment and control groups. The control group would participate in the existing program and the treatment group would come under the new requirements and services. Only through this kind of design can valid conclusions be drawn about the effects of the welfare program. Otherwise, analysts cannot determine whether changes in employment or caseload levels were the result of the reforms or other factors, such as a strong economy or the particular characteristics of the recipients in the program.