Table 4 compares the variance partition for sample A with that reported by Schmalensee(1985). Schmalensee estimated that 19.59 percent of the total variance was due to industry effects. In this study, I find that somewhat less, 16.12 percent, is due to all industry effects (stable plus year-to-year fluctuations). The difference between the estimates arises mainly because 1975 was an abnormal year-repeating Schmalensee's one-year analysis in 1976 and 1977 yields smaller industry components. More importantly, I find that only one-half of this variance is due to stable effects. Long-term industry effects account for only 8.28 percent of the observed variance among sample A business-unit returns.
Turning to the intra-industry variance, Schmalensee reported that 80.41 percent of the variance was unexplained by industry; the comparable figure in this study is 83.08 percent. However, my partition of this intra-industry variance into stable and year-to-year components reveals that over one-half is due to stable business-unit effects. Indeed, the variance among stable business-unit effects is six times as large as the variance among stable industry effects-business-units differ from one another within industries much more than industries differ from one another.
Despite the fact that this is a descriptive study, some strong general results can be reported:
1. There are significant business-unit effects in U.S. manufacturing activities that strongly outweigh industry and corporate membership as predictors of profitability. The variance among business-unit effects is much larger than the variance among industry effects (six times larger in sample A and eleven times larger in sample B).
2. Corporate effects, although present in sample B, are not important in explaining the dispersion in observed rates of return among business-units.