A BASIC proposition of economic theory is that, under competition, profit rates will tend toward equality. Baumol, on the other hand, has put forward the hypothesis that increased money will not only increase the total profits of the firm, but because it puts the firm in a higher echelon of imperfectly competing groups, it may very well also increase its earnings per dollar of investment [5, p. 33]. The logic of Baumol's proposition is that large firms have all of the options of small firms, and, in addition, they can invest in lines requiring such scale that small firms are excluded. It follows that . . so long as any industries are peculiarly well suited to large investments, and so yield disproportionate returns to sizeable funds, then, provided is prepared to move in response to profit differences, this will tend to be true of all other industries in which large firms operate [5, p. 37]. If this hypothesis is correct, we should find higher rates of return in large enterprises even in the long run and even in the absence of barriers to entry other than those directly associated with availability of capital. An empirical test of this proposition should give a basis for judging the height of the capital requirements barrier [2, p. 156], a potentially important element in the explanation of industry performance. The primary goal of this paper is to test Baumol's hypothesis. It is divided into three parts. In section I we discuss our data and our model, in section II we present and analyze our results, and in section III we draw a few tentative conclusions. I

This paper compares the simple Markov process commonly used in migration studies with an economic model of migration where interregional wage differences are the equilibrating variables. Using the economic model, it appears unlikely that regional exit and entry rates will remain stable as the population is redistributed. As a result, both theory and empirical interstate migration evidence suggest that Markov migration projections will usually understate the population changes required before stochastic equilibrium is reached. IN RECENT YEARS the social sciences, and particularly economics, have experienced

The striking difference between the results reached by Charles Link, using published 1970 Census data, and ours, using the 1967 SEO tape, calls for some explanation. In an effort to find the source of the difference, this comment reports our results using the 1970 Census (1969 income and earnings) individual data from the public use tapes. We shall show that the source of the discrepancies lies with the form of the published Census data rather than in underlying economic relationships. Using individual data, we show that the effect of education on earnings is similar for blacks and whites in 1969, a result which confirms our assertions from the 1966 SEO data. We suspect that similar results would be forthcoming from the 1959 Census data. It seems likely that the low apparent return to black education found bv Lester Thurow was also a statistical artifact. At the very least, we conclude that the use of grouped data for policy analysis of this sort is very suspect and that individual data should be used wherever possible. The published census data is beset with numerous flaws. First, it presents data on income rather than earnings from work or self-employment. Income includes welfare transfers, unemployment insurance, and the like. As a result, the effect of education on income is apt to be weak at low levels of education compared with its effect on earnings. Second, the published data excludes persons with zero income. Since these include the unemployed, their exclusion tends to weaken the education-earnings relationship still further. Third, the published data include all males in the armed forces and in institutions (for example, jail). The education-earnings relationship may be biased downwards even further since income in kind is an enormous component of military pay. Each of these flaws in the published data apply to whites as well as blacks, but the distortion is likely to be more severe for blacks since they have higher dependence on transfers, higher unemployment rates, and higher participation in the armed forces. In addition, the published Census data consist of median incomes for age-education groups, and three problems arise as a result. First, the census groups are quite broad. For example, median incomes are reported for those with 1-4 vears of education, 5-7 years of education, etc., and for ages 18-19, 25-29, . . , 36-44, etc. In contrast, the 1967 SEO tape and the 1970 Census public use tapes permit us to use precise age and education data. Second, the Thurow and Link studies gave each age-education cell equal weight, while our study is based on individual data effectively weighted by the number of individuals in a given cell. Third, the published cell medians do not reflect outliers very well, but they are included in our study. Obviously, a grouped data regression will always yield a higher R2. Bevond that, it is not clear what a priori bias is introduced by grouping. In an attempt to evaluate the effect of these differences we grouped the public use tape data following the census procedures, but utilizing alternative measures of income and alternative sample definitions. These results were then compared with those derived from individual data upon which the grouping was based. The results of these experiments appear in Table 1 for those in the North, Table 2 for those in the South, and Table 3 for the nation as a whole. * The University of Wisconsin, Madison. We gratefully acknowledge the superb research assistance of Nancy Williamson. The research was supported by funds granted to the Institute for Research on Poverty at the University of Wisconsin by the Office of Economic Opportunity. The conclusions are the sole responsibility of the authors.

Bidding theory predicts low er selling winning bids (higher buying winning bids) as numbers of bidders incre ase. Alternative versions predict that winning bids fall (rise) with the second order statistic, the maximum order statistic, or 1/N. The authors test these pre dictions using data on underwriters' spreads on tax exempt bonds, offshore oil t racts, and National Forest timber. They estimate winning bids using general vari ables for the product and dummies for 1, 2, . . .11 bidders. In all cases mor e bidders meansignificantly lower (selling) winning bids. The expected maximum order statistic fits better than either the second order statistic or1/N, in te n out of twelve cases.