INTRODUCTION

One of the most widely estimated regression equations in economics is that of the earnings or wage function, which relates a measure of market remuneration (e.g., hourly wage rates or annual earnings) to measures of human capital stocks such as completed schooling and market work experience. In the first half of this paper, I address two related questions: (1) What interpretation should be given to the wage equation? and (2) Why should we care about estimating it?

Twenty-five years ago, Zvi Griliches, in his 1975 Presidential Address to the World Congress of the Econometric Society, raised the first question of interpretation, together with several others that were concerned with the appropriate specification and estimation of the wage function. In that address, Griliches chose to adopt the competitive market–human capital production interpretation of Ben-Porath (1967), in which the wage an individual is offered is the product of the competitively determined skill rental price and the amount of human capital or skill units cumulatively produced by and thus embodied in the individual. About a decade later, Heckman and Sedlacek (1985) and Willis (1986) extended the competitive market–human capital production interpretation of the wage function to a multidimensional skill setting, formally incorporating the self-selection model of Roy (1951) to the choice of employment sector in the former case and to the choice of occupation cum schooling in the latter. More recently, Keane and Wolpin (1997) extended and generalized the Heckman and Sedlacek and Willis papers to a dynamic setting of schooling, employment, and occupational choice, adopting as well the competitive market–human capital production paradigm.