Denison and the Contribution of Education to National Income Growth: A Comment

Abstract

There appears to be a hitherto unnoticed error of principle in Denison's analysis of the contribution of education to national income growth (Denison 1962, 1964, 1967). It lies in his index of labor force quality, which incorrectly measures the efficiency units per worker in the economy. The index is only correct if the efficiency unit chosen happens to be a worker earning the national average wage,1 whereas in Denison's practice the choice is arbitrary. The main relevant features of the index may be summarized as follows. Denison assumes, reasonably in my view, that only a fraction (a) of the observed income differential between any pair of educational groups is due to extra education.2 But he then proceeds, wrongly, to assume that, if we take as the efficiency unit the average worker with education e, the number of efficiency units in a worker with education i is 1 + oC x (Wi We)/We, where We is the base-year average wage of workers with education e and Wi is the base-year average wage of those with education i. The index of labor force quality in year t is thus