Profitability and Profit Sharing: Reply to Mukhopadhyay
回应Mukhopadhyay对作者先前研究的批评，澄清了利润分享比例与资本回报率之间关系的误解，并反驳了关于变量遗漏和估计方法的质疑。
IN HIS comment on our paper (FitzRoy and Kraft (F-K) [1986]), Mukhopadhyay (M) is concerned with what we called the surprising lack of from profitability or return on capital (ROC) to the profit share paid per worker (PROSH). Unfortunately his comment is marred by a series of elementary mistakes and confusions. The central result (the effect of PROSH on ROC) is not disputed by M. First, M fails (except in footnote 1) to distinguish between PROSH and the share offraction or profits paid to workers. It is natural for firms to start with a small fraction, say one per cent, and increase to perhaps several per cent when the scheme is successful over time. This increase of several hundred per cent is captured by our AGE (since introduction) variable, and naturally dominates probably comparatively small increases in ROC due to better motivation. This confusion leads M to unjustified claims like their theoretical model dictates that the immediate effect of AGE falls on ROC. Furthermore, profit sharing can be introduced in times of low or high profits. Firms near bankruptcy occasionally start a profit sharing system in order to raise capital. In such a situation PROSH is relatively high while ROC will be low, and with differing motives to introduce profit sharing, a positive correlation does not necessarily follow. Of course, we agree that theoretically there should be some feedback from ROC to PROSH after controlling for AGE, but problems of measuring profits from accounting data and multicollinearity just swamp this relatively small effect. In earlier regressions without AGE, ROC was also insignificant, so there is no detectable feedback. M next claims that the significance of AGE is spurious, and would disappear if we estimated PROSH on the subsample with only positive shares. The inclusion of variables with joint zero values is standard (Heckman [1974]; Wales and Woodland [1980]), and with non-linear Tobit estimation does not produce biased coefficients. Estimates with the sharing subsample lead to very similar results and therefore M's Section II is refuted. Nevertheless, the main purpose of the estimation of the PROSH equation is a test of the frequently stated hypotheses that highly profitable firms share profits and in order to test this the full sample has to be used. The somewhat confusing arguments of M in this section seem to boil down to the point that we have an omitted variable problem, as crucial variables explaining the existence of a profit sharing scheme (what M calls decision process) are missing. We have to admit that, for example, the basic attitude of