Can Administrative Data Kept by States be Used to Study the Duration of Program Participation?

John Marcotte

December, 1994

The Urban Institute

Executive Summary

I. Purpose of Study

Recent efforts to reform welfare programs have focused on limiting the duration of a spell on welfare to 24 months. After 24 months, clients would no longer be eligible for benefits. A central question that emerges from this policy is: what proportion of program participants have not left welfare by 24 months? Furthermore, how does that proportion vary over clients' social and demographic characteristics? Data to examine such questions must be longitudinal. Longitudinal data are usually collected in panel studies, and panel studies are expensive to implement. A potential ready source of longitudinal data is the administrative records of program participants that the states maintain to operate programs and manage their caseloads. These administrative records could serve as an alternative to panel surveys. Besides determining the proportion of cases still on welfare after 24 months, longitudinal data can ascertain how the chances of leaving welfare change with the duration of time spent on welfare (i.e. does prolonged program participation foster welfare dependency?) Administrative data also could provide a way to study the interrelationships between various welfare programs. For example, does participation in multiple programs indicate a pattern of dependency?

State administrative records might be a rich source of longitudinal data to study program participation. The purpose of this study is to assess whether administrative data kept by states can be used to study the duration of program participation. Programs under consideration are:

public assistance (AFDC), food stamps, medical assistance and child support (IV-D).

II. Types of Potential Data

Two types of data may be obtainable:

(1) retrospective data: information that is currently available including prior case histories; this type of data collection requires that states have kept case histories;

(2) prospective data: information that is assembled as events occur; this type of data collection requires states to store all information as it happens.

These two types of data are not mutually exclusive; either type may be assembled or both. In general, retrospective data are easier and less expensive to collect but often less detailed; however, some policy questions cannot be addressed with these type of data. Prospective data require that information be collected over a period of time which is more time-consuming and expensive.

In the course of this study, 15 states were contacted by telephone to obtain preliminary information about their respective state data systems. Four of the 15 states (California, Georgia, New York and Texas) were visited to collected more detail material.

Given current circumstances, California and Texas present the best opportunities for using pre- existing or retrospective data. Data for both states, however, do not include links with Child Support (IV-D) records.

For prospective data, New York and California offer the best possibilities. California has an on-going project called the California Longitudinal Data Base (LDB). These data, however, do not include grant amounts or child support information. New York's welfare management system has the potential for linking records across all programs (including child support) but existing computer resources make the production of data files impractical at this time.

III. Options for ASPE

To assemble longitudinal data about program participation from administrative records (under the assumption that the administrative data have the necessary data elements), ASPE has two basic options:

(1) ask states to produce these files themselves by providing them with funds for additional computational and human resources (this option would require states to build research capacity into their current operating systems);

(2) ask states to provide administrative data in whatever form it exists to a third party who will assemble the longitudinal files.

These two basic strategies have both positive and negative aspects.

A. Option 1: Positive Aspects

For the first strategy of inducing states to produce these files themselves, the positive aspects are:

• the states develop capacity to produce these data on a regular basis;
• state personnel who are most familiar with state-specific participation regulations would have active role in documenting and annotating the data;
• state personnel who are familiar with the nuances of their administrative system would be able to assemble the data more easily.

B. Option 1: Negative Aspects

Negative aspects of asking states to produce these data themselves are:

• most states have neither the computational nor the human resources to produce these data;
• data would not have a standard format which would make comparisons among states more difficult;
• each state would have to develop the same capacity with respect to personnel and computation which is expensive and redundant.

C. Option 2: Positive Aspects

For the second strategy of requesting states to provide data to a third party data archivist organization, the positive aspects are:

• data for all states would be put in the same format which facilitate state comparisons;
• data for a variety of states would be centrally available; public use files for policy analysts could easily be made available over electronic medium such as the Internet;
• the same personnel could assemble files for all states which would avoid having similar personnel in all the states;
• the same computational resources could be used for each state;
• policy analysts would be more likely to have an active role in assembling the data and would be able to specify the essential data fields needed to address policy relevant questions.

D. Option 2: Negative Aspects

Negative points to using a third party archivist are:

• because of different administrative systems and program regulations, non-state personnel would have difficulty assembling the data;
• data with individual identifiers are confidential; before the data are given to a third party archivists, states might insist that individual identifiers be stripped or scrambled; this scrambling of identifiers may make linking records over time and across programs impossible;
• the states would not build up any capacity to produce these files on a regular basis;
• when files are not assembled at the state government, itself, they are less likely to be used by state policy analysts;
• the data archivist contract would need to be fairly long term (at least 3 years).

IV. Eagerness of States to Participate

California and New York are the most enthusiastic about trying to assemble administrative data for research. At the present time, other states have little incentive to supply or assemble data. Most states have limited research departments. In most states, the research department is also responsible for producing the already federally mandated reports. Most states would probably be willing to participate if federal resources were made available to the state.

Assembling data of this type with federal seed money could provide a way for states to develop research capacity that they do not currently have. This new research capacity could provide states with the ability to make changes in their respective welfare programs based on empirical research rather than on opinions or anecdotes.

V. Conclusion

On balance from a research standpoint, option 2 is the better choice. Under option 2, data are more likely to be usable to address relevant policy questions and to make state comparisons. Data produced under option 2 are more likely to be used by a greater variety of analysts in both the public and private sectors (including universities). Economies of scale will probably make option 2 more cost effective in the long run. After assembling data for the first few states, the option 2 contractor will have developed capacities which can then be used to compile data for other states.

At the moment, Option 1 is most viable in New York State. The New York welfare management system contains the necessary information to produce longitudinal files, and the system includes child support records.

If neither option is considered viable, data for California and Texas are available now, but these data lack information about all programs.

Whatever option ASPE undertakes, new longitudinal data will not be available for at least 9 months from the date of funding. Therefore, ASPE must act quickly if it wants to use longitudinal data from a variety of states to examine how program changes have affected the use of welfare.