As always with instrumental variables approaches, caution should be exercised in drawing overly-strong conclusions until a consensus emerges from complementary evidence using other data sources and alternative identifying assumptions. In this particular case, however, past work by Beaudry and DiNardo (1991), and replication of that work with the data set used in this paper (which differs from the data sets they analyzed), suggest that the identifying assumptions are plausible.
Third, the policy evaluation that this paper attempts does not consider the effects of actual policies-such as school-to-work programs-implemented to encourage early job stability. Study of real-world programs is likely to lead to additional insights about alternative means by which early job stability can be encouraged, and the effects of doing so.
Finally, the evidence is silent on the issue of whether there is any need for policy intervention. In particular, individuals might choose to forego early job stability even if it would ultimately result in higher wages. Such behavior may maximize utility, and there is no obvious market failure that causes individuals to experience less job stability when young than is optimal, although their own shortsightedness may be the culprit. In addition to estimating the effects of alternative policies, the need for intervention is an important area of inquiry for those interested in policies that might transform the workings of youth labor markets in the U.S.