directions for further research.
2 The standard principal-agent model
Labor market theories have different predictions concerning the relationship between the
variability of earnings and that of firm output. In the competitive model with infinitely
elastic labor supply curve, price-taking firms choose employment to equate the marginal
product of labor to the market wage. According to implicit contract models, risk-neutral
firms insure risk-averse workers against fluctuations in levels of activity. As a consequence,
in both models the wage paid is orthogonal to firm-specific shocks.2 In principal-agent
models, by contrast, moral hazard considerations lead firms to link wages to performance.
Under a series of assumptions, the sensitivity of the workers’ pay to performance depends
on many factors, including the noise in measured performance, the marginal cost of effort,
the elasticity of performance to effort and risk aversion (Holmstrom and Milgrom, 1987).
The differing implications that these models have for the compensation scheme allow us to
design an empirical test to discriminate among them.
As noted, empirical studies so far have focused on the consistency of specific compensa-
tion contracts - such as those of CEOs - with the predictions of incentive models. Overall,
the findings broadly support the predictions of the theory: for the groups under study,
compensation responds to performance, implying that less than full insurance is offered
(see Rosen, 1992, for a comprehensive survey of empirical studies of CEOs’ compensation).
Our goal is to assess whether incentive models of wage determination help predict the struc-
ture of compensation for an array of workers that goes beyond the limited groups that have
attracted interest thus far. In addition we want to distinguish between transitory and per-
manent fluctuations in firm performance. As far as we know there has been no attempt to
test the general applicability of the principal-agent theory of compensation and to relate it
to the dynamics of the firm’s performance.
To provide a framework for the subsequent empirical analysis, consider the following
2The two models differ in their predictions concerning lay-offs. For example, following a negative shock,
in the competitive model some workers are fired, while in the implicit contract model they are temporarily
laid-off but receive a payment that equates the utility of those working and those laid-off (Rosen, 1985).