The Size of Individual Incomes: Socio-Economic Variables and Chance Variation

W HILE many economists have focused their descriptive and analytical energies on the size distribution of income, no wholly useful or empirically validable explanation for the inequality of individual income receipts has been developed. This article describes some empirical research intended to throw some new light on the factors which determine the size of individual incomes. The study involves a crosssection analysis bringing together data on the income and economic characteristics of a large number of individuals. We hypothesize that the amount of an individual's income is a consequence of the action of a number of discernible independent variables and of chance variation. In other words, we suggest that size of income can be explained on the basis of the following regression model: Y = ao + aIX1 + a2X2 . * + ctnXn + U where Y is the size of an individual's income; X1 . . . X,. are the independent variables; and u represents random variation. Such a statistical approach is frequently employed with regard to other social and economic phenomena, but it has never been explicitly applied in the study of the income distribution. While this methodological approach does not limit the nature and number of independent variables which may be introduced into the equation, in practice it becomes necessary to restrict the scope of the problem. We have postulated a particular hypothesis. Casual observation what may facetiously be termed "armchair empiricism" suggests a connection between the diverse nature of the productive services offered by individuals and the inequality of their incomes. We hypothesize that such heterogeneity, as illustrated by the differing socio-economic characteristics of the workerincome-recipients, plays a part in determining the size of incomes. In other words, we suggest that personal socio-economic characteristics are important independent variables in our regression model. We aim to ascertain whether such a notion has a place in a theory of income size distribution. In addition, we have investigated the variation of income which remains after the socioeconomic heterogeneity of its recipients has been allowed for. An exhaustive analysis of all other factors which may have a relationship to size of income was not possible, and no effort in this direction was made. However, our regression model does postulate the presence of random residual variation, and so the second part of the study was focused on this aspect of the distribution process. The analysis was limited to income from wages and salaries. In each case, we have dealt only with the earnings in one year of one individual employed full-time (as nearly as can be determined) for wages and salaries only. It is one of the complicating aspects of statistical work dealing with the size distribution that so much of the data concern total income receipts of spending units regardless of source, number of recipients, length of time worked, etc. Interesting as these data may be from a welfare point of view, they throw together a number of separate problems which cannot be handled properly when we are concerned with explaining the size of earnings. The statistical information upon which this research is based comes from the I950 to I953 Surveys of Consumer Finances of the Survey Research Center at the University of Michigan. These sample surveys, carried on annually under the sponsorship of the Federal Reserve Board, gather information on the financial status, spending and saving habits, etc. of a representative sample of approximately 3,000 American spending units. Among the varied information gathered are extensive data on in* This article embodies material from my unpublished doctoral dissertation, "Some Aspects of the Income Size Distribution," University of Michigan, I956. I am particularly indebted to Dr. L. R. Klein and Professors James N. Morgan and Daniel B. Suits. Appreciation is expressed to the Survey Research Center at the University of Michigan who supplied the empirical data on which the study is based.