YOUTH LABOR MARKETS IN THE U.S: The Data The Sample 7

The sensitivity of the results to alternative sources of variation in unemployment is considered by constructing (Uijt – U jt), the deviation of the individual-level unemployment rate from the cohort-year average, and using both “components” of the unemployment rate Uijt as instruments. In the latter specification, using Newey’s (1985) overidentification test is particularly informative, because there are, conceptually, two different instrumental variables (even though they each represent sets of instrumental variables)-the cohort-average unemployment rates, for which the identifying assumptions guaranteed payday loan
w6581-4
w6581-5
That is, is the average unemployment rate faced by individual i in cohort j over the five-year postschooling period, and Uj. is the average unemployment rate faced by individuals in cohort j over the five-year post-schooling period. As in the specification in which the unemployment rates from each of the five years are used, U j. is the most plausibly exogenous variable, because it removes individual cross-sectional variation that may be related to migration and residence decisions. One advantage of using these averages is that they make the estimated effects of early unemployment rates in the first-stage regressions easier to interpret, because there is one coefficient rather than a set of coefficients for the unemployment rates from each of the first five years. On the other hand, the pattern of unemployment rates over the initial post-schooling period may be important, and the five-year averages discard this information are more plausible a priori, and the deviations, for which the overidentification test may be most relevant.