A few states, including California, Kentucky, Maryland, and Tennessee, are moving toward the creation of data warehouses that will store information extracted from administrative records that can be readily used for research purposes. The term "data warehouse" refers to the archiving and integration of data from a variety of sources in a single commonly formatted database that provides a "snapshot" of comparable information at different points in time. Creating data warehouses involves developing compatible data structures, definitions, and coding conventions. It also involves archiving files, such as the unemployment insurance wage records, that would otherwise be "dumped" so they can be easily accessed over time.
California is trying to create files on the total universe of clients rather than just samples of clients extracted from administrative records. People frequently move in and out of households, making it difficult to track them with less than a universal approach. Despite the added expense, this approach may prove more useful over the long term.
Data-sharing often is carried out through partnerships between state agencies and state universities. California's longitudinal database of Medicaid recipients, which serves as the core for its data-integration efforts, is being developed through the UCDATA project at the University of California at Berkeley. Maryland is building a data warehouse at Towson State University that links TANF, food stamps, child support enforcement, child welfare, and unemployment insurance records. Georgia State University, the University of Missouri, the University of Baltimore, and the University of Texas all have unemployment insurance data archives dating back about ten years. In working with universities and contractors, state agencies must make it clear that they retain control over what can be done with the data.
It is also important to have "computer people" talk to "computer people" so they are clear on the content of each data field and can make sure they have the technical capacity to manipulate the relevant databases. For example, policymakers in Kentucky have committed one person from each of the relevant cabinet agencies to work together full time for up to one month so they can resolve all of the data issues related to linking key databases. In addition to the technical people, program staff need to be involved so they can help explain the data. For example, they could explain a rise in the number of reported child abuse cases during a certain period because they know the increase was associated with publicity about the death of a child in the child welfare system.
Increased demand for data matching and information on the outcomes of welfare reform is placing even greater demands on states' already-overburdened information-processing capabilities. Many of the states' information systems staff are overwhelmed by the workload required to make massive federally mandated changes in welfare, child support, and other state data systems and address Year 2000 conversion issues. Changes in information systems are complicated by state agency reorganizations and devolution, decentralization, and privatization initiatives that create new organizational relationships and bring in new players who must be trained. (For example, California's system is based only on county-level administrative data because counties are responsible for administering welfare in California.)