Do higher wages reflect higher productivity?

Norway has invested heavily in education in recent decades. The average educational attainment in Norway was below the OECD average in 1960, while in 1995 Norway was among the OECD countries with the highest average educational attainment ¹ . However, growth accounting calculations as documented in Hægeland (1997) show that these investments have contributed relatively little to economic growth. The low estimate for the growth contribution from education is driven by the low wage premiums for education in Norway, and a crucial assumption for such estimates is that relative wages correspond to relative productivity.

A number of studies have documented small educational wage differences in Norway relative to other countries, see e.g. Asplund et al. (1996) and Kahn (1996). Does the compressed wage distribution reflect small differences in productivity between various categories of labor, or are relative wages poor proxies for the productivity differenences between educational categories? Several authors, e.g. Freeman (1996) and Kahn (1996), have pointed out that Norway represents an exception to the trend in most Western countries in the eighties and nineties, towards a more decentralized wage determination and increasing wage differences. The wage distribution in Norway is very compressed compared to other countries ² , especially at the bottom. Both Freeman and Kahn attribute this to a high degree of centralized bargaining. The importance of centralized bargaining in the wage determination suggests that the growth accounting assumption of wage differences reflecting productivity differences may be questionable.

In this empirical analysis we compare differences in wages and productivity across workers with different educational levels. We also examine the wage and productivity effects of experience. Various theories of compensation and human capital investment predict that wage profiles over the working career do not follow the productivity profiles and we test such predictions. A third issue we consider is whether wage differences between male and female workers reflect differences in productivity. Our results indicate that the higher wages earned by workers with higher education largely correspond to their higher productivity. Experienced workers are more productive than unexperienced workers and they also earn higher wages; the wage premium is lower than the productivity premium for workers with 8-15 years of potential experience, while the opposite is the case for workers with more than 15 years of experience. Women are found to be less productive than men, with wages corresponding to their lower productivity.

Our analysis is based on a new data set for plants, combining information from the annual Norwegian census of manufacturing plants with register files for individual workers. This new data set makes it possible to disaggregate the work force of each plant according to different characteristics. Workers are classified according to education, gender and experience, and we estimate relative wages and relative productivity of these labor categories for the period 1986-93.

There exist few studies comparing wages and productivity at the plant level. Most earlier studies at the micro level have used various proxies for productivity, such as performance pay (Foster and Rosenzweig, 1993) or superiors' evaluations (Medoff and Abraham, 1980, 1981). Griliches and Regev (1995), in their study of Israeli manufacturing firms, found that an occupation-based, wage-weighted labor quality index contributed significantly to explaining labor productivity. We employ a more direct measure by estimating how plant productivity vary with work force composition. More specifically, we estimate plant level production functions where workers with different characteristics are allowed to have potentially different productivity levels. The estimated productivity differences are compared to corresponding estimated wage differences. Our methodology draws on the work by Hellerstein, Neumark, and Troske (1996) who estimated relative wages and relative productivity for various worker characteristics on U.S. plant level data. Their work in turn builds on Hellerstein and Neumark (1994, 1995), who analyzed relative wages and productivity with respect to gender and experience on Israeli data. Due to a better match of workers and plants, we are able to relax some of the restrictive assumptions made by Hellerstein and her coauthors.

The rest of the paper is organized as follows. Section 2 presents the econometric framework for estimating relative wages and productivity, and discusses some econometric issues. Section 3 offers details on sample and variable construction, while estimation results are presented and discussed in sections 4 and 5. Section 6 concludes and points out directions for future research.

'See Wolff (1994) and OECD (1997).

² See ch. 3 in OECD (1995), especially fig. 16.