The large business-unit effects indicate that there is more intra-industry heterogeneity than has been commonly recognized. Whereas economists are quick to refer to inframarginal rents when this issue arises, the unspoken presumption is that these effects are small, or related to scale. The results are otherwise. The business-unit effects are large and owe only a small fraction of their strength to market share. Some portion of these effects may, of course, be due to measurement biases. But the most obvious sources of bias, differences in industry accounting and differences in corporate policy, should appear as industry or corporate effects.
The presence of strong business-unit effects is consonant with the presumptions in portions of the business strategy literature. Beginning with Caves and Porter's mobility barriers concept, based on the 'hypothesis that sellers within an industry are likely to differ systematically in traits other than size,' (1977: 250) ideas in this area have evolved in the direction of recognizing increasingly disaggregate sources of resource immobility or specificity. 19 According to this view, product-specific reputation, team-specific learning, a variety of first-mover advantages, causal ambiguity that limits effective imitation, and other special conditions permit equilibria in which competitors earn dramatically different rates of return. Although this study cannot discriminate among the various theories regarding the sources of intra-industry heterogeneity, it ncessarily gives broad support to this class of theory and should encourage further work in this vein.
If business-units within industries have large and persistent differences in return, it becomes necessary to ask what the 'industry returns' measures used in many industrial organization studies actually represent. That is, when industries exhibit differing levels of overall return, to what extent are such differences due to systematic industry effects and to what extent are such differences the veiled result of differences in individual business-unit performance? I under take a very simple exploration of this question by examining the variance of Yi = "'ik.r'i· .In; .., the average return observed for industry i over
all four years, so thIat
Yl·=µ+al +
ni·1
·-uIf
ni··
(14)