Classification-Error Models and Labor-Market Dynamics

Measurement error is often a severe problem in longitudinal data such as labor-market flows. Two correction methods have been used for labor-market flows, but the methods tend to adjust the data in opposite directions. One method shows a rigid labor market with long unemployment spells, whereas the other shows the opposite. Because the methods appear theoretically similar, one might interpret the differing results as an indictment of both procedures. This article documents and explains the differences between the two methods. More important, it argues that the differences result from the reliance of one method on assumptions that are theoretically implausible and rejected by the data.