«ECONOMIC GROWTH AND INCOME INEQUALITY* By SIMONKUZNETS The central theme of this paper is the character and causes of longterm changes in the ...»
The American Economic Review
VOLUMEXLV NUMBER ONE
ECONOMIC GROWTH AND INCOME INEQUALITY*
The central theme of this paper is the character and causes of longterm changes in the personal distribution of income. Does inequality
in the distribution of income increase or decrease in the course of a country's economic growth? What factors determine the secular level and trends of income inequalities?
These are broad questions in a field of study that has been plagued by looseness in definitions, unusual scarcity of data, and pressures of strongly held opinions. While we cannot completely avoid the resulting difficulties, it may help to specify the characteristics of the size-ofincome distributions that we want to examine and the movements of which we want to explain.
Five specifications may be listed. First, the units for which incomes are recorded and grouped should be family-expenditureunits, properly adjusted for the number of persons in each-rather than income recipients for whom the relations between receipt and use of income can be widely diverse. Second, the distribution should be complete, i.e., should cover all units in a country rather than a segment either at the upper or lower tail. Third, if possible we should segregate the units whose main income earners are either still in the learning or already in the retired stages of their life cycle-to avoid complicating the picture by includingincomes not associated with full-time, full-fledgedparticipation in economic activity. Fourth, income should be defined as it is now for national income in this country, i.e., received by individuals, including income in kind, before and after direct taxes, excluding capital gains. Fifth, the units should be grouped by secular levels of income, free of cyclical and other transient disturbances.
For such a distribution of mature expenditure units by secular levels * Presidentialaddressdelivered at the Sixty-seventhAnnual Meeting of the American EconomicAssociation,Detroit, Michigan,December29, 1954.
Numiber 56 of a scrics of pbotographlIs of past pircsidecits of thte Associattioni.
2 THE AMERICAN ECONOMIC REVIEWof income per capita, we should measure shares of some fixed ordinal groups-percentiles, deciles, quintiles, etc. In the underlying array the units should be classified by average income levels for a sufficientlylong span so that they form income-status groups-say a generation or about 25 years. Within such a period, even when classified by secular income levels, units may shift from one ordinal group to another. It would, therefore, be necessary and useful to study separately the relative share of units that, throughout the generation period of reference, were continuously within a specific ordinal group, and the share of the units that moved into that specific group; and this should be done for the shares of "residents" and "migrants" within all ordinal groups.
Without such a long period of reference and the resulting separation between "resident" and "migrant" units at different relative income levels, the very distinction between "low" and "high" income classes loses its meaning, particularly in a study of long-term changes in shares and in inequalities in the distribution. To say, for example, that the "lower" income classes gained or lost during the last twenty years in that their share of total income increased or decreased has meaning only if the units have been classified as members of the "lower" classes throughout those 20 years-and for those who have moved into or out of those classes recently such a statement has no significance.
Furthermore, if one may add a final touch to what is beginning to look like a statistical economist's pipe dream, we should be able to trace secular income levels not only through a single generation but at least through two-connecting the incomes of a given generation with those of its immediate descendants. We could then distinguish units that, throughout a given generation, remain within one ordinal group and whose children-through their generation-are also within that group, from units that remain within a group through their generation but whose children move up or down on the relative economic scale in their time. The number of possible combinations and permutations becomes large; but it should not obscure the main design of the income structure called for-the classification by long-term income status of a given generation and of its immediate descendants. If living members of society-as producers, consumers, savers, decision-makers on secular problems-react to long-term changes in income levels and shares, data on such an income structure are essential. An economic society can then be judged by the secular level of the income share that it provides for a given generation and for its children. The important corollary is that the study of long-term changes in the income distribution must distinguish between changes in the shares of resident groups-resident within either one or two generations-and changes in the income shares of
KUZNETS: ECONOMICGROWTHAND INCOME INEQUALITY 3groups that, judged by their secular levels, migrate upward or downward on the income scale.
Even if we had data to approximate the income structure just outlined, the broad question posed at the start-how income inequality changes in the process of a country's economic growth-could be answeredonly for growth under defined economic and social conditions.
And, in fact, we shall deal with this question in terms of the experience of the now developed countries which grew under the aegis of the business enterprise. But even with this limitation, there are no statistics that can be used directly for the purpose of measuring the secular income structure. Indeed, I have difficulty in visualizing how such information could practicably be collected-a difficulty that may be due to lack of familiarity with the studies of our colleagues in demography and sociology who have concerned themselves with problems of generation or intergenerationmobility and status. But although we now lack data directly relevant to the secular income structure, the setting up of reasonably clear and yet difficult specifications is not merely an exercise in perfectionism. For if these specifications do approximate,and I trust that they do, the real core of our interest when we talk about shares of economic classes or long-term changes in these shares, then proper disclosure of our meaning and intentions is vitally useful. It forces us to examine and evaluate critically the data that are available; it prevents us from jumping to conclusions based on these inadequate data; it reduces the loss and waste of time involved in mechanical manipulations of the type represented by Pareto-curvefitting to groups of data whose meaning, in terms of income concept, unit of observation, and proportion of the total universe covered, remains distressingly vague; and most important of all, it propels us toward a deliberate construction of testable bridges between the available data and the income structure that is the real focus of our interest.
I. Trends in Income Inequality Forewarned of the difficulties, we turn now to the available data.
These data, even when relating to complete populations, invariably classify units by income for a given year. From our standpoint, this is their major limitation. Because the data often do not permit many size-groupings, and because the difference between annual income incidence and longer-termincome status has less effect if the number of classes is small and the limits of each class are wide, we use a few wide classes. This does not resolve the difficulty; and there are others due to the scantiness of data for long periods, inadequacy of the unit usedwhich is, at best, a family and very often a reportingunit-errors in the
4 THE AMERICAN ECONOMICREVIEWdata, and so on through a long list. Consequently, the trends in the income structure can be discerned but dimly, and the results considered as preliminary informed guesses.
The data are for the United States, England, and Germany-a scant sample, but at least a starting point for some inferences concerning long-term changes in the presently developed countries. The general conclusion suggested is that the relative distribution of income, as measured by annual income incidence in rather broad classes, has been moving toward equality-with these trends particularly noticeable since the 1920's but beginning perhaps in the period before the first world war.
Let me cite some figures, all for income before direct taxes, in support of this impression. In the United States, in the distribution of income among families (excluding single individuals), the shares of the two lowest quintiles rise from 13? per cent in 1929 to 18 per cent in the years after the second world war (average of 1944, 1946, 1947, and 1950); whereas the share of the top quintile declines from 55 to 44 per cent, and that of the top 5 per cent from 31 to 20 per cent. In the United Kingdom, the share of the top 5 per cent of units declines from 46 per cent in 1880 to 43 per cent in 1910 or 1913, to 33 per cent in 1929, to 31 per cent in 1938, and to 24 per cent in 1947; the share of the lower 85 per cent remains fairly constant between 1880 and 1913, between 41 and 43 per cent, but then rises to 46 per cent in 1929 and 55 per cent in 1947. In Prussia income inequality increases slightly between 1875 and 1913-the shares of the top quintile rising from 48 to 50 per cent, of the top 5 per cent from 26 to 30 per cent; the share of the lower 60 per cent, however, remains about the same. In Saxony, the change between 1880 and 1913 is minor: the share of the two lowest quintiles declines from 15 to 14Y2 per cent; that of the third quintile rises from 12 to 13 per cent, of-the fourth quintile from 16?2 to about 18 per cent; that of the top quintile declines from 562 to 54j/2 per cent, and of the top 5 per cent from 34 to 33 per cent. In Germany as a whole, relative income inequality drops fairly sharply from 1913 to the 1920's, apparently due to decimation of large fortunes and property incomes during the war and inflation; but then begins to return to prewar levels during the depression of the 1930's.'
The following sources were used in calculatingthe figurescited:
United States. For recent years we used Income Distributionby Size, 1944-1950(Washington, 1953) and Selma Goldsmith and others, "Size Distributionof Income Since the Mid-Thirties," Rev. Econ. Stat., Feb. 1954,XXXVI, 1-32; for 1929, the BrookingsInstitution data as adjustedin Simon Kuznets, Shares of Upper Groupsin Income and Savings (New York, 1953), p. 220.
UnitedKingdom.For 1938 and 1947, Dudley Seers,The Levellingof Income Since 1938
KUZNETS: ECONOMICGROWTHAND INCOME INEQUALITYEven for what they are assumed to represent, let alone as approximations to shares in distributions by secular income levels, the data are such that differences of two or three percentage points cannot be assigned significance. One must judge by the general weight and consensus of the evidence-which unfortunately is limited to a few countries.
It justifies a tentative impression of constancy in the relative distribution of income before taxes, followed by some narrowing of relative income inequality after the first world war-or earlier.
Three aspects of this finding should be stressed. First, the data are for income before direct taxes and exclude contributions by government (e.g., relief and free assistance). It is fair to argue that both the proportionand progressivity of direct taxes and the proportion of total income of individuals accounted for by government assistance to the less privileged economic groups have grown during recent decades. This is certainly true of the United States and the United Kingdom, but in the case of Germany is subject to further examination. It follows that the distribution of income after direct taxes and including free contributions by government would show an even greater narrowing of inequality in developed countries with size distributions of pretax, exgovernment-benefitsincome similar to those for the United States and the United Kingdom.
Second, such stability or reduction in the inequality of the percentage shares was accompanied by significant rises in real income per capita.