Introduction

In Jensen (1996, 2000), an extended human capital model with imperfect information on the employees' side is estimated as a stochastic earnings frontier with data from the German socioeconomic panel. The costs for the information imperfection are measured by the inefficiency terms of the frontier model. This individual inefficiency in finding suitable jobs is shown to be considerable. The approach leads to a sensible interpretation of the deviations of empirical income from estimated maximum possible income. Distinguishing between ‘potential human capital’ and ‘active human capital’ accounts for the partial failure of human capital models when estimated as average functions.

Keuzenkamp and McAleer (1995, 1997) have revived the discussion about simplicity, a wide-ranging concept in contrast to the well-known but narrow criterion of parsimony. In the present chapter, this concept will be applied in analysing the question of whether the ‘standard OLS approach’ or the frontier approach is to be preferred. The trade-off between simplicity and descriptive accuracy will be examined. The comparison of the models will be conducted verbally because there is no computable simplicity criterion at present.

The simplicity concept helps in the comparison of the OLS and the frontier model, and the comparison of these models demonstrates the rightness of the methodological theses of Keuzenkamp and McAleer (1995, 1997), who hold that simplicity should not be identified with the narrow criterion of parsimony and that Hendry's reductionism is a very narrow methodology.