Ruben Hernandez-Murillo and colleagues report on "Patterns of Interstate Migration in the U.S. from the Survey of Income and Program Participation." Panel data (which the SIPP is) that report on individuals before and after some event (in this case migrating) are the best. And I am drawn to these types of studies because macro-level labor market analyses can mislead. Unemployment remains high in many places and many sectors because of bad matches between jobs and workers. Migration can help.
The authors are able to consider before-and-after wage and employment characteristics of the movers. They can do this for by gender, age, race, and level of education. An odd result is that they find that the employment rate falls for all groups. Then why move? For reasons that are unclear the authors only look at a three-month window at either end of the move. Do people eventually become better off? Or are the movers making a huge mistake?
The authors can also compare origin vs. destination unemployment rates, incomes, foreclcosure rates school rankings and homicide rates. That's the good news. The bad news is that even their rich data source tells them little. "Surprisingly, there seems to be little difference between between a majority of the characteristics." I think that it's a problem of aggregation. The authors also had metropolitan-level data available from the SIPP, but only for two of the panels, 1996 and 2001. I expect that a five-year window and smaller areas would have been more useful. So I wish they would not have done a state-level study.
There is good news and bad news. Panel data are best. Maximum spatial disaggregation is best and wider time-windows are a good idea. There are dissertations to be written.