Monday, March 14, 2011

Administrative versus survey data

Low response rates to surveys (with technologies such as Caller ID, etc.) have made the jobs of government agencies such as Census and BLS much harder. The alternative to surveys are administrative data such as wage and unemployment insurance records. However, these records are usually administered at the state level - making it difficult to coordinate. One example of successful coordination has been the LEHD. While the micro-data is not available for public access in the way that the American Community Survey and the CPS is, the aggregate data is available via On The Map.

Even though it is a useful data set for exploring issues such as worker flows, it cannot answer basic questions such as what is the unemployment rate? In order to answer such a question, data linking would have to be made to social security records to identify workers and non-workers (via wages received). Even here, it is unable to distinguish between those who are unemployed versus those who are not in the labor force.

Another useful source of linkage would be from workers to the 4-and 2-year colleges, technical and vocational schools, and graduate and high schools. This education data could provide a wealth of information on the returns to education as well as give a more accurate picture of which colleges have graduates who go on to have higher earnings. One modest project that has been underway for some time is the Texas Schools Project.

While extremely useful, privacy concerns will prevent an expansion and convergence of such two systems.

No comments: