Introduction

As early as 1936, T. P. Wright observed that the number of manhours required to produce an airframe is a decreasing function of the number of airframes of the same type previously produced. Indeed, the connection is very precise; the amount of labor required to produce the nth airframe is proportional to n−⅓. This relation has subsequently become basic in the Rand Corporation studies of production and cost planning for the U.S. Air Force.

Other empirical studies have revealed similar results. Hirsch showed the existence of ‘learning curves’ or ‘progress ratios’ in the production of various types of machines. Lundberg described the ‘Horndal Effect’. The Horndal Iron Works in Sweden had no new investment, and presumably no important change in its methods of production, for a period of fifteen years; yet output per man increased on average by two per cent per year. More recent studies by Sturmey and Rapping have also emphasized the crucial point. Specifically, in observing manufacturing processes, one finds steadily improving performances that can only be attributed to learning from experience on the job.

This conclusion is, indeed, quite commonplace; it is confirmed by casual empiricism as well as statistical studies; it is a technological fact of life that hardly requires explanation. Yet until Arrow's seminal article, this basic relation had not been incorporated into theories of production, distribution, and growth.

Arrow's model

The fact that a person may learn from experience is, as noted above, commonplace knowledge. How to introduce this fact into an economic model is not. The examples cited above would seem to suggest that cumulative gross output is a suitable index of experience.