The Structural Change and the Endogeneity Bias of the College Premium in the United States 1968-2001”

This paper derives and tests a procedure to estimate a general class of paneldata models that display a change of regime, as well as potential correlationbetween specific effects and explanatory variables. This model detects astructural change for all cohorts on the US labor market in the early 80s, andshows that some unobserved skills correlated with schooling have becomerewarded in the second regime. As a result, I quantify the bias in the collegepremium when it is calculated in cross-section: it amounts to circa 30% ofthe cross-sectional estimates after 1980. Interestingly, this bias is carriedalmost entirely by the two most educated cohorts of individuals. A simplemodel where skilled individuals have a comparative advantage to graduatefrom college can account for this fact, since it implies that the correlationbetween unobserved ability and college graduation increases with overallcollege graduation.