YOUTH LABOR MARKETS IN THE U.S: Discussion and Conclusions 3

One way to assess explanations regarding unobservables, such as this one, is to look at evidence on observables that should behave in the same way as the unobservable. In this particular case, it seems reasonable to suppose that education and the unobserved wage or productivity component should have similar relationships with early job stability. However, Tables 2-5 show that education is positively correlated with early job stability (as is AFQT, although these results are not reported in the tables). It is possible that the relationship is different for the unobservable fixed effect, if the unobservable matters mainly within education and ability levels. But it is probably best to be skeptical about an hypothesized unobservable related to wages or productivity that is inversely related to education and measured ability.

Alternatively, however, suppose that there is heterogeneity in the returns to search. With such heterogeneity, those with relatively higher returns to search exhibit less early job stability, do more searching and more successful searching, and find better matches as adults, generating a negative correlation between early job stability and adult wages. But when early labor market conditions provide the identifying variation in early job stability, the negative correlation between early job stability and the quality of the adult job match is broken, and the IV estimate of the effect of early job stability on adult wages should rise relative to the OLS estimate. For example, those with predicted high early job stability are less likely to be those with low returns to search-who end up with worse matches and lower adult wages-than those who endogenously choose high early job stability.28,29 In addition to explaining the differences between the IV and OLS estimates, the explanation based on heterogeneous returns to search is easier to reconcile with the positive estimated relationship between education and early job stability. In particular, those with more education may have lower returns to search because they enter the labor market with much better information about their skills and abilities, and with human capital investments that are more occupation specific.