might be upward sloping, due to temporary friction. For instance, in the presence of workers’
mobility costs of the type considered by Bertola (1999), firms face an upward-sloping labor
supply curve. In these models, the responsiveness of wages to idiosyncratic shocks to firm
performance is determined by the slope of the labor supply curve: the further away from
infinite elasticity, the stronger the implied correlation between performance and earnings.
Thus, one could argue that while the correlation that we find between wages and firm-
specific shocks is inconsistent with the frictionless, perfectly competitive model, it could
be made consistent with an extended version that allows for some temporary friction. Yet
our evidence is difficult to reconcile with these models as well. Competitive models with
either firing or mobility costs carry direct implications for the response of wages to shocks of
different duration. Specifically, since adjusting employment rather than wages is relatively
more advantageous vis-à-vis permanent shocks, wages should respond more to temporary
than to permanent shocks, a prediction that is strongly at variance with our empirical
findings.
Apart from this direct evidence, the competitive model fails to fit with some other
features of our results. First, whereas in principal-agent models the response of wages to
performance depends in predictable ways on well identified firm and worker characteristics,
competitive models (with or without friction) have no clear implication of how the correla-
tion between wages and firm performance varies with those attributes. Second, one could
argue that an environment corresponding more closely to the competitive paradigm is that
of industrial district firms. In particular, firms are highly similar and employ workers with
similar characteristics; workers’ mobility and search costs are negligible (firms being located
very close to one another); firms are small, implying price-taking behavior. Thus, if the
competitive model were valid one should expect lesser sensitivity of wages to performance
among district firms, since they are likely to have a more elastic short run labor supply.
Again, however, this is the opposite of what we actually find.
Overall, we take our results on wage determination to be remarkably consistent with the
predictions of the agency model and inconsistent with competitive theories, not only in the
extreme version characterized by continuously clearing markets but also in more realistic
33