transitory shocks and wages cannot be explained by low power of the instruments.
8 Discussion
Our evidence points to three important considerations. First, workers do share the firm’s
fortunes, at least in part. The unexplained earnings variability that can be attributed to
incentive schemes is nearly 10 percent. Second, while transitory shocks to value added do
not affect wages, permanent shocks are partly transferred to earnings. Finally, our search
for heterogeneity in the sensitivity of earnings to shocks shows that insurance increases
with workers’ risk aversion, declines with the amount of responsibility within the firm (as
measured by job position), is less for firms located in an industrial district and greater for
those with high output variability, and is a decreasing and convex function of firm size. Most
of these results are consistent with the predictions of the agency model but very hard to
reconcile with competitive models of wage determination even when some friction is allowed
for.
Consider a competitive model of the labor market, in which price-taking firms choose
employment to equate the marginal product of labor to the market wage. Wages should
not respond at all to idiosyncratic shocks to the firm, which faces an infinitely elastic labor
supply at the prevailing wage. This is the version tested by Blanchflower, Oswald and
Sanfey (1996), although they correlate wages with industry rather than firm profitability.
Our test of the perfect competition model is less stringent, in that we do let wages respond
to industry shocks - as any competitive model of the labor market with limited mobility
across industries would imply. All that we require for the perfectly competitive model to
hold is that wages not respond to firm-specific shocks, as is implied by competition within
the industry. Obviously, this is a much weaker requirement than the null hypothesis that
workers wages do not respond to shocks to the industry. Rejecting perfect competition using
industry data may not be surprising in the presence of, say, some market segmentation or
industry-specific skills. Thus our results can be seen as a more robust rejection of the perfect
competition model.
As Blanchflower, Oswald and Sanfey (1996) note, however, the short-run labor supply
32