Discussion Paper

No. 2009-2 | January 05, 2009
Distribution of Labour Productivity in Japan over the Period 1996–2006


The distribution of labour productivity is investigated by analyzing the longitudinal micro-level data set which contains detailed financial condition of large numbers of Japanese companies over the period 1996–2006. The generalized beta function of the second kind is applied to explain the distribution. We calculate marginal labour productivity by using the fitting parameters, and show that the economy in the labour market is not in equilibrium. By comparing parameters characterizing high productivity range and low productivity range, we show that inequality of low productivity range is larger than that of high productivity range. In addition, it is shown that the change of inequality in low productivity has strong correlation with GDP.

Paper submitted to the special issue “Reconstructing Macroeconomics”

JEL Classification:

C16, E23, L60


  • Downloads: 1807


Cite As

Wataru Souma, Yuichi Ikeda, Hiroshi Iyetomi, and Yoshi Fujiwara (2009). Distribution of Labour Productivity in Japan over the Period 1996–2006. Economics Discussion Papers, No 2009-2, Kiel Institute for the World Economy. http://www.economics-ejournal.org/economics/discussionpapers/2009-2

Comments and Questions

Anonymous - Referee Report
March 03, 2009 - 08:28

- The paper seemed unfinished to me, although the topic is interesting and important.

- The motivation for the study and the significance of the results were not properly communicated. For example, one main conclusion is that "the labor market is not in equilibrium" since "labor productivity varies between ...[more]

... plants and companies".

But the authors implicitly define equilibrium as zero variance of labor productivity (i.e. deterministic equilibrium). If the authors had discovered that the labor market was actually in this kind of state of deterministic equilibrium -- well that would be very surprising! Finding the opposite is much less surprising, given that the labor market has a huge number of degrees of freedom and weak micro-level coordination.

The authors do not mention that the labor market might in fact be close to another kind of equilibrium -- namely statistical equilibrium. In this kind of equilibrium we would not expect zero variance of labor productivity, although we should expect a relatively constant distribution of labor productivity over time. Perhaps the authors could consider testing this proposition? This would be better than knocking down a straw man.

- The authors assume a Cobb-Douglas production function and find that the data seems to support this assumption. The authors might be interested in the point that such empirical corroboration in fact simply reflects an accounting identity. E.g., see Laws of Production and Laws of Algebra: The Humbug Production Function
(1974) The Review of Economics and Statistics, Volume 56(1), February 1974, p. 115-120.

- Perhaps this is an unfair observation, but I found the discussion of the fitting of the Generalized Beta distribution a little repetitive and overlong. I also got the feeling that the authors might mention what other distributions they tried fitting to the data, and why they selected this particular one.

- The authors did not explain why they thought it important to split the data into manufacturing from non-manufacturing parts. What is different about these sectors that would justify this split?

- The paper would benefit from the removal of typos and some better sentence construction. But this doesn't get in the way of comprehension, apart from a misspelling of "two" as "tow".

- The authors note that the "inequality of labor productivity of the non-manufacturing industry is higher than that of the manufacturing industry". But they don't attempt to explain why this might be so.

In fact, the analysis and conclusions I think are quite limited, which is why the paper feels unfinished and not fully polished.

Wataru Souma - Reply to Referee Report
March 11, 2009 - 08:28

see attached file