We propose that college graduates enter the labor market with less uncertainty regarding which career they are most productive in, and study how this characteristic contributes to the unemployment-education gap. We document several novel facts to support our hypothesis. Notably, college graduates predict their occupation more accurately than those without a college degree. We then develop and calibrate a life cycle search model featuring differences in uncertainty by education and learning about one’s best career fit. Our quantitative analysis suggests large disparities in uncertainty by education, and that such differences can explain a sizeable portion of the unemployment-education gap.

This paper investigates the cyclicality of wage losses from skill mismatch and the underlying mechanisms. Using the NLSY97, I find that wage losses due to skill mismatch are procyclical, with a one percentage point increase in the unemployment rate leading to a 1.3% reduction in wage losses, equivalent to a 14.6% recovery. I document that this effect is more pronounced in industries and occupations with higher training incidence, and that training provision is procyclical. To explain these findings, I develop a directed search model with skill mismatch and endogenous training that encompasses a productivity and training channel. The model demonstrates that declines in labor productivity and reduced training during recessions interact to reduce wage losses from skill mismatch.