T HE proposition has been advanced by Dobb that . . where wage-earners are strongly organized in trade unions, one might expect labour to succeed in obtaining a larger share of the product than elsewhere. ' While there are several alternative forms in which this hypothesis may be stated, only one such will be examined in this paper. Accordingly, the purpose of this study is to test the hypothesis that labor's relative share of the income produced by manufacturing industries in the United States is, in some significant sense, positively correlated with the degree of union organization or, alternatively, with changes in the degree of union organization. The principal finding is that no significant correlation can be established and, therefore, it is concluded that the hypothesis must be rejected. While the results of this study, as it turns out, do not differ substantially from those of other research more or less closely related to it,2 the methods and procedures used are believed to be different and to possess at least some general validity. It is contended, therefore, that more credibility can be attached to the results contained in this paper than to those of related studies. In particular, the research design employed here incorporates two principal innovations. First, the data are those of individual industries rather than of sectors of the economy or the nation as a whole. Since it is neither the nation nor arbitrary sectors of it that have been unionized, but rather concrete, identifiable, individual industries, the use of data flowing from the latter would seem to be appropriate if one is seeking to isolate the impact of unionism on distributive shares. Second, the method is essentially one of employing a simple trend analysis to determine the intra-industry behavior of labor's share over the long-run. This is in opposition to the currently popular, but potentially misleading, terminial-years approach. There is no reason to expect a priori that any set of results obtained from a mere comparison of two more or less widely separated end years will not be vitiated by analysis of the data from another set of end years. Only in the case where the deviations from the trend are consistently small does a terminal-years approach appear to serve the purpose as well as a trend analysis. But since the trend must first be ascertained in order to determine the extent of such deviations, there would not seem to be any advantage in working only with terminal years. Because almost any result, depending only on the choice of years to serve as bench-marks, can be obtained by using this method, it would appear that no general validity can be associated with any particular finding.3 A trend

where W is the annual rate of change of money wages in manufacturing, P is the annual rate of change of consumer prices, U is the average annual civilian unemployment rate, and R is the average annual profit rate of manufacturing corporations.' On the assumption that P -(W), specifically, P = W r, where r is the trend rate of increase of output per man-hour in the private non-farm economy the steady-state solution for W is a a-blr + b3R + b2 t. (1.2) 1-b, 1-b,