A Structural Estimation of the CES Preferences and Linear Labor Supply: The Case of Prime-Age Males in Japan

The tax simulation studies in Japan have necessarily relied on arbitrary sets of preference parameters due to the paucity of the empirical estimates. Motivated by this state of the art, we estimate the labor supply function and preference parameters for Japanese prime-age males, allowing for the complex Japanese income tax system and taking advantage of a large microdata set we obtained for this study. We employ two versions of the method of maximum likelihood. One is the celebrated Hausman method which assumes a linear labor supply function (Hausman 1979). The other is the method proposed by Zabalza (1983), which takes advantage of an explicit specification of the CES preferences. We have examined several estimation patterns and calculated elasticities over the whole or sub-samples of observations. While the sample averages of uncompensated elasticities are estimated between 0.06 -- 0.21, those of compensated counterparts result in higher ranges (0.08 -- 1.39) due to rather large negative estimates for the income effects. It may be interesting to point out that the compensated elasticities tend to be higher in the CES case (0.41 -- 1.39) than the Hausman case (0.08 -- 1.12) despite the fact that the former is immune from the MaCurdy critique (MaCurdy et al. 1990).