Empirical Calibration of a Least-Cost Conservation Reserve Program

agent. By its nature, this line of research is descriptive rather than prescriptive. It allows
the econometrician to infer the principal’s beliefs, but provides no guidance to the principal
regarding how to develop them.

A second branch deals with how to estimate parameters of a contracting model
without assuming a sophisticated principal. Work in this area has used the contracting
framework of Laffont and Tirole (1986), focusing on heavily regulated industries such as
urban transport where existing regulations create moral hazard problems in addition to
adverse selection. In an initial attempt along this line, Dalen and Gomez-Lobo (1997)
interpreted regression residuals as an indication of the type of an individual agent. That
model was rather restrictive since it did not allow for any random error. In their study
of French urban transport firms, Gagnepain and Ivaldi (2002) employ a similar approach,
assuming that labor efficiency (type in their model) is an unobserved time-invariant random
variable with a beta distribution. Importantly, they also append a normally-distributed
random error. With panel data, they are able to use maximum likelihood techniques to
estimate the parameters of firms’ cost functions as well as the parameters of the distribution
of agent types.

Like Gagnepain and Ivaldi, the estimation strategy employed here allows for a ran-
dom distribution of agent types as well as a stochastic noise. Rather than analyzing a
regulated monopoly, it analyzes a sector in which all production decisions are chosen by the
agents themselves.¹ As a result, differences in regulatory framework (e.g., cost-plus versus
fixed price) do not affect agent actions. In addition, output cannot be considered exogenous.

The key difference, however, lies in econometric methodology. In the spirit of the
stochastic frontier models pioneered by Aigner et al. (1977) and Meeusen and Broeck (1977),
the specification employed here allows use of a two-part additive composite error structure.

<<   1   2   3   4   5   6   >>

More intriguing information