We revisit measurement of employer-to-employer (EE) transitions in the monthly Current Population Survey. The incidence of missing answers to the question on change of employer sharply increases starting with the introduction of a new software instrument to conduct interviews in January 2007 and of the Respondent Identification Policy in 2008–2009. We document nonrandom nonresponse selection by observable and unobservable worker characteristics that correlate with EE mobility. We propose a selection model and a procedure to impute missing answers. Our imputed EE aggregate series no longer trends down after 2000 and restores a close congruence with the business cycle after 2007.
We analyze sorting in a frictional labor market when workers and jobs have multidimensional characteristics. We say that matching is positive assortative in dimension (j, k) if workers with higher endowment in skill k are matched to a job distribution with higher values of attribute j in the first-order stochastic dominance sense. Crucial for sorting is a single-crossing property of technology. Sorting is positive between worker-job attributes with strong complementarities but negative in other dimensions. Finally, sorting is based on comparative advantage: workers sort into jobs that suit their skill mix rather than their overall skill level.