Productivity convergence in OECD manufacturing industries

Abstract

The extent of β- and σ-convergence of average labor productivity across manufacturing industries in 18 OECD countries over the period 1972–1992 shows large inter-industry differences. One reason for these differences is knowledge and capital barriers preventing the occurrence of catch-up. We find the level of average labor productivity, as a proxy for these barriers, is correlated with the extent of convergence.

Introduction

The extent to which economies converge has received abundant attention by economists and politicians. Research has concentrated on the question of convergence of GDP per capita but much less so on the question of convergence of labor productivity at the disaggregated level of industries. The mechanisms behind catch-up and convergence of GDP per capita can never be established unless the developments of its lower level of aggregation equivalent of labor productivity are well understood. The current analysis aims at estimating the extent of convergence in labor productivity at the industry level and relating it to a simple measure of physical capital and knowledge barriers, viz. the average level of labor productivity. Industries with high barriers may also be expected to be industries with relatively low degrees of convergence.

Recently, Bernard and Jones have published a series of papers on productivity convergence at the sector level (Bernard and Jones, 1996a, Bernard and Jones, 1996b, Bernard and Jones, 1996c). They claim that β-convergence at the macro level of GDP per capita has not been caused by productivity convergence in the manufacturing sector but instead by convergence in the service sector (see also Gouyette and Perelman, 1997). Arguments for catch-up resulting from technology transfer in manufacturing between countries and/or globalization of the manufacturing sector therefore seem not to have empirical support. However, the lack of convergence found within the manufacturing sector does not reveal the spread of the extent of convergence across manufacturing industries. Differences in convergence rates across industries help to understand why productivity gaps between countries exist and (dis)appear. Dollar and Wolff (1993) pay some attention to convergence at the industry level in their Chap. 3. The current paper is a more formal approach to this line of research.

The convergence debate has been increasingly shifting into a debate on econometric techniques with claims that the rates of convergence have been overestimated (Lichtenberg, 1994) or underestimated (Islam, 1995, Lee et al., 1998). A first choice a researcher is confronted with is whether to consider β-convergence or σ-convergence. β-convergence implies that less developed countries or industries perform better (catch up) on average when compared to more developed countries or industries. The effect of GDP per capita or productivity in the first period on its relative change in the consecutive period should therefore be negative. The idea behind σ-convergence is that the variance of (log) GDP per capita or productivity decreases as production techniques become more similar. It can easily be shown that σ-convergence is a sufficient condition for β-convergence, but not the other way around (Lichtenberg, 1994, Quah, 1993). In this paper we consider both β-convergence and σ-convergence and compare the results when either of these two measures is chosen.¹

The remainder of this paper is organized as follows. In Section 2 we discuss a simple model of endogenous technological progress to explain differences in speed of convergence across industries. In Section 3 the two measures of convergence are introduced. Next, the measurement of average productivity level at the industry level is discussed. In Section 4 the results are presented for the 28 manufacturing industries in the OECD countries. The measures of convergence are compared and related to the initial level of labor productivity. The last section is used for concluding remarks and questions for future research.