Evaluations of employment and training programs often use state unemployment insurance (UI) wage records to measure effects on participants’ employment and earnings. These data offer several benefits for evaluation purposes: They are generally accessible to evaluators (pending agreements with state departments of labor), they cover over 90 percent of employment in a state, and they allow evaluators to track participants over long periods to assess how program impacts evolve over time.

UI wage records also have some constraints. They miss earnings from certain types of work, such as self-employment, informal “off-the-books” jobs, and employment with the federal government. State UI wage records also do not capture out-of-state work—which can limit evaluations, especially those that focus on programs that serve individuals who live near a state border (and therefore may cross the border for work) or individuals who relocate often.

This brief, as part of the Learning from Administrative Data Project, examines the implications of relying only on in-state UI wage records to evaluate programs that are designed to increase employment and earnings. It uses data from the Portland, Oregon, site of the National Evaluation of Welfare-to-Work Strategies (NEWWS), an assessment of a series of programs that were implemented and evaluated in the 1990s. An earlier analysis of the Portland program found that the employment rate impacts differed—primarily five to eight years after study entry—depending on whether national or Oregon-only data were used. This brief builds on that work by presenting differences between the two data sets in employment rate impacts, year by year, through Year 20. It also presents year-by-year differences in earnings impacts, comparing Oregon-only earnings data with data from a broader group of states.