Race Differences in Income: Current Income: 1966 and 1967

II. Current Income: 1966 and 1967

When we speak of current incomes, we mean mainly 1966 and 1967, sometimes 1968 and 1969, and never 1970. There is an inevitable long lag between the receipt of income in a given year, a report on it in the sample survey the following March, and the processing of the Survey data and further analysis and publication. Similar data and publication lags in the past partly explain the disparity between what a good many authors have been saying about "current" nonwhite and white income and the greatly changed actual income at the time of saying. It is a hazard of the profession that should inspire a prudent check with more recent indirect measures as well as cycles and trends.

The upshot of an analysis of the 1966, 1967, and 1968 data is that there are very large disparities between nonwhite and white family income, and between the incomes of nonwhite and white persons 14 years old and over who have income. These disparities appear on the whole to be understated by the data, though they are the most complete figures currently available. Moreover, the disparities affect not only the middle and the lower end but the entire income distribution. Drastic reduction in such differences requires work not only on the nonwhite poor, or on reducing "inequalities" in general, or on the "hard core" unemployed but large changes in the occupational distribution of nonwhites, and in the nonwhite to white ratio of earnings within occupations. It will take time, energy, resources, and intelligence.


In looking at current differences in income between whites and nonwhites, we will look first at averages. Median family income for nonwhites in 1966 was $4,628 and for white families $7,722, giving a nonwhite to white median income ratio of 59.9 percent. Median total money income to nonwhite persons (including all persons 14 years old and over who received income) in 1966 was $2,099 and for white persons $3,499, giving a nonwhite to white ratio of 60.0 percent.

The ratios improve slightly for both persons and families in 1967. For families, median nonwhite income was $5,141 and median white income was $8,274, for a ratio of 62.1 percent. For persons, median nonwhite income was $2,322 and for whites $3,725, giving a ratio of 62.3 percent. Data for 1968 indicate that the ratios at the median for both families and persons were about the same as in 1967.


However, we are interested not only in comparing white and nonwhite incomes in the center of the distributions, but at all other parts of the income distribution as well. The statements cited earlier, for example, include some that say that at any given time the ratio of nonwhite to white income is lower in the low-income quantiles than it is at the high end -- that nonwhite relative income increases with increasing quantiles of income; that inequality among nonwhites is greater than among whites and, moreover, is increasing; that ghetto rioters are particularly sensitive to where they stand in relation to other Negroes, not to whites, and believe that inequality among nonwhites is increasing;[1] that better-educated Negroes compare themselves with whites of equal schooling, and therefore feel more deprived than uneducated Negroes. (The latter on this assumption, contrary to the preceding case, are supposed to compare themselves with whites who are not much better off than they.)

Such statements manifest a variety of interests and beliefs about the distributions of nonwhite income. To compare the distribution of income to whites with the distribution of income to nonwhites, we use several methods. One is illustrated in Fig. 1. Here we have white and nonwhite incomes for families in 1967 plotted against proportions of the population. The rightmost vertical scale gives us the proportion of families with income greater than the amount shown, and the leftmost vertical scale gives us the quantiles -- the proportion of families with income equal to or less than the amount shown. Incomes are plotted on the horizontal scale. For example, in Fig. 1 the incomes at the 33rd percentile are about $3,500 and $6,500 for nonwhites and whites respectively. Since the income scale is logarithmic, equal distances on the graph between white and nonwhite incomes indicate incomes in equal ratios.[2]

From Fig. 1, it is clear that income differences between whites and nonwhites involve differences not only in median incomes or in the proportions in poverty, but in the entire distribution. This illustrates the need to distinguish objectives. Reducing the number of poor -- as defined by an arbitrary poverty line -- doesn't help nonwhites above that line, but of course most nonwhites above the line are closer to it than whites and they receive smaller incomes than whites in corresponding positions of the income distribution. Fig. 2 gives us a similar picture of income to persons in 1967. Again, as in the case of families, income differences exist at all points of the income distribution.

For a considerable part of the distribution of family income in the modal range, the distance between the two curves narrows with increasing levels of income -- that is, the nonwhite to white ratio tends to increase. All of this behavior is more clearly visible in the income-ratio-at-quantiles curve shown in Fig. 3 which is derived by taking the nonwhite-to-white income ratio at selected percentiles and plotting the ratio against the percentiles. However, these changes are not monotonic and their behavior at the tails in particular is quite different. For 1967, for example, in the lowest and highest fifths, the ratios decline with increasing income. Moreover, for the case of income to persons, for the distribution on the whole, there is no tendency toward increased ratios with increasing income such as is shown in the modal range of family income; and the relative decline in the bottom and top percentiles is plain. See Fig. 4. (We shall make extensive use of similar curves.)

The Ends of the Distribution

There are good reasons to expect the two ends of these relative distributions to behave differently from the middle. At the bottom, income maintenance programs, in spite of their many familiar limitations, tend to put at least a low floor under nonwhite as well as white income and so to raise low level nonwhite relative to low level white income. Moreover, the lower the position in the nonwhite income distribution the more one would expect these income supports to decrease nonwhite economic disadvantage relative to whites at the corresponding position in the white distribution. This effect would be even stronger with more generous income maintenance programs, but they are quire visible with current ones.

At the other end, nonwhites run into a variety of obstacles lost from view in averages and totals. Even though young whites and nonwhites have come close together in median years of schooling, they are still very far apart in the advanced education associated with high income. In the older groups the schooling differences are larger. Moreover, even allowing for differences in years of schooling and in occupational structure, many social hurdles apparently make it particularly hard for nonwhites to earn very high incomes; examples include the remoteness of nonwhites from top-level communication and decision networks, a particular white prejudice against nonwhites in supervisory and other authoritative roles.[3] Such expectations about the two ends of the distribution are confirmed by the analyses (summarized below, see Sections III and IV) of the changes over time in money income distributions and their components and by the analysis of the effects of schooling and occupational distributions.

Patterns of the Income Ratio Curves

In fact, the behavior of the income-ratio-at-quantiles curves exhibit quite consistent patterns in recent years. The 1967 curve (Fig. 3) is quite typical for families. It is neither linear nor monotonic. It does appear that there are three distinct segments of the curve, each of which may be approximately linear. Fitting straight lines to each of the three segments (that is, 0th to 21st percentile, 21st to 83rd percentile, and 83rd to 100th percentile) gives strong evidence that these three parts of the distribution follow quite different patterns. The slopes for the nonwhite-to-white income ratios against quantiles of the distribution are negative for the two end segments, and positive for the middle one. (The positive slope for families in the modal range appears to be associated with a greater tendency in nonwhite middle income families for wives to work as well as husbands; and perhaps with the positive association of wives' job qualifications with husbands'.) The data points are also very close to linear over the middle segment.[4] In Fig. 4, the segments of the relative distribution for persons (using 0th to 38th percentiles, 38th to 79th percentiles, and 79th to 100th percentiles) are not as clearly linear as for families, but the change in behavior of the curve is quite pronounced over these three parts of the distribution. And the evidence is particularly strong for a negative slope in the last segment (the upper part of the distribution).[5]

The Meaning of "Equal Opportunity"

The relevance of the low end of the distribution is clear, but why should anyone bother with higher percentiles, especially the top ones? There are good reasons for wanting to put the same floor under the income of nonwhites and whites. But to focus only on the lowest quantiles is to neglect the fact that there is a lower statistical ceiling over nonwhite income, one that is hard to penetrate. It may suggest a kind of settlement-house attitude, as distinct from the wider range of concerns indicated by the goal of equal opportunity. First, it should be observed that even the top 1 percent of nonwhite income receivers are hardly rich beyond the wildest dreams of avarice. The top 1 percent of nonwhite males who had income in 1966 received at least $12,000 (see Fig. 5). However, the top 10 percent of white males received at least $12,000,[6] and the top 1 percent of white males received at least $26,000. This seems a natural interpretation of the phrase, "equality of opportunity." The phrase clearly covers an equal chance to exceed the poverty line, but getting an even chance with whites at middle and higher incomes seems to be implied as well. More than just the symbolism of top nonwhite earners (like top athletes or top politicians), it may provide useful incentives for those in lower ordinal groups to aim that high: even if they don't reach top income levels, they may get more than if they hadn't had that incentive. Finally, the growth of the nonwhite middle class and of a class of high-level managers, professionals, or entrepreneurs who make, say, $26,000 or more might be directly associated with the economic improvement of other nonwhites -- through savings and investment, by helping to build information networks, and through key positions of influence that affect entry, promotion, and profit in higher-paying occupations.

Of course, none of the above disparages the importance of lifting the low end of the nonwhite distribution. Rather, it stresses that the problems of discrimination apply to all of the distribution; that some useful aids to solving the troubles at the low end are not independent of progress at the middle and at the top; and finally, that slogans of "equal economic opportunity," if they mean what they say, imply for nonwhites an even chance with whites -- not simply at scraping by above some arbitrary minimum, but at doing moderately well, and a fair chance at making it big.

Income Comparisons for Men and Women

The greater disparity between whites and nonwhites for men (Fig. 5) than for aggregate persons (Fig. 2) implies that the disparity may be smaller between white and nonwhite women. The income distributions for white and nonwhite women are illustrated in Fig. 6, along with the distributions for men. Here we see that income to both white and nonwhite women is lower than income to either white or nonwhite men. Note also that the two curves for women are closer to each other than are the two curves for men, which means nonwhite to white income ratios are higher for women than for men. Thus, among nonwhites, men have higher incomes than women, but the difference is smaller than that between white men and women.

Men are, of course, the principal breadwinners and for that reason the object of more study; and the "status" of women on the other hand is often thought to be primarily determined by the income, occupation, and education of their husbands. It is also true that more data are published about men than women and this tends to further focus research on male income. This in turn is reinforced by several traditions of research. For example, S. M. Lipset has suggested to us that the familiar sociologists' methods of studying occupational mobility by relating the occupational status of the son to occupation of the father is of clear interest for men, but has no ready correlative for women. Be that as it may, many studies focus exclusively on comparisons of white and nonwhite men, yet sometimes draw conclusions about nonwhites in general. However, from Fig. 6 we can see that men and women have quite different incomes, and any inference from one to the sum of the two (or in the reverse direction) requires caution. This caution has not always been observed, and it has misled inferences about changes over time.

Nor is it justified, in spite of the principal importance for family welfare of male income, to focus only on male income and to neglect the incomes of white and nonwhite women. This might be done on the premise that, in the case of white women, low incomes are almost entirely a matter of unconstrained choice, that white women who participate less or receive smaller rewards in the labor force do so only because they are married to rich men and are not interested in middle or high level jobs. However, that premise may be doubted even by some who are not charter members of Women's Lib. First, for nonwhite compared with white women, the ratio of the medians of year-round full-time earnings is even larger than that for all earnings, part and full-time (in 1967, .761 compared with .688; see Table 2). Second, even when the husband's income is high, nonwhite wives participate more than whites. "Among families where the husband's yearly income was $10,000 or more, about half of the Negro wives compared with a third of the white wives were working or looking for work"[7] in March 1969. Since the interval is not bounded above and nonwhite husbands are undoubtedly clustered nearer the lower limit, this second point is suggestive rather than conclusive. Third, however, the income contribution of wives, in any case, bears no simple inverse relation to the income of their husbands. For nonwhites in the middle ranges, as we suggested earlier (p. 18), wives' incomes are positively correlated with their husbands' income. And, finally the willingness to work of white as well as nonwhite women has something to do not only with pay, but also with the non-monetary rewards of the work available -- its interest, the status it confers, and so on. An abundance of evidence suggests that (in parallel as we shall observe with nonwhite minorities) women find it especially hard to get authoritative, high-paying jobs that call for high levels of education. See Archibald, 1970.

Both nonwhite and white women are discriminated against, though discrimination by reason of sex is even harder to disentangle and quantify than color discrimination. One needs to separate not only unequal pay for equal work from non-market discrimination in the opportunities to acquire skills, but also inequalities in opportunity available to women who want to and can perform the same job as well as men from educational choices and income differences due to differing social functions, capabilities, and preferences. The annals of prosecution under the Fair Labor Standards Acts, as well as common observation, make clear that some substantial part of the difference is discriminatory.

Regional and Farm-Nonfarm Comparisons

There are various ways of decomposing income distributions other than by sex. For example, for families we can look at income differences over different regions of the country or by farm-nonfarm residence. Fig. 7 compares the ratio-at-quantiles curves for the South[8] and for the Nonsouth for 1967. For both whites and nonwhites incomes are lower in the South than in the rest of the country. However, notice that the difference is greater for nonwhites than for whites. Thus, Southern nonwhites have lower incomes relative to Southern whites than have nonwhites living outside the South compared with whites outside the South. This greater income disparity in the South is very important for the overall picture, because at present about half of all nonwhite families live in the South, whereas only about one-quarter of white families do.

We find even greater disparities between white and nonwhite incomes for families who live on farms. However, the effect on the overall income distribution is small, since only a small proportion of families live on farms, and today proportionately fewer nonwhites than whites. In 1967 about 5.5 percent of white families lived on farms, while about 4.4 percent of nonwhite families did.

The intersection of some of these unfavored categories show even larger contrasts in nonwhite to white ratios.

[1]Caplan and Paige, 1968.

[2]For some of the reasons for our extensive use of logarithms, percentage differences and the like, see pages 48 ff.

[3]Blumer, 1961, p. 219; Siegel, 1965, pp. 41-57.

[4]Let X = quantiles of the distribution, and Y = the nonwhite-to-white income ratio for families. Then the straight line fits are Y = .599-.376X for 0 < X .209 with R2 = .826, Y = .441 + .336X for .209 X .826 with R2 = .991, Y = 1.182 - .565X for .826 X < 1 with R2 = .829. The slopes for all three fits are significantly different at the 5 percent level from the slope of .1567 which resluts from fitting a line to the whole distribution. All three fits took account of the sampling errors for each point; i.e., each point had a weight inversely proportional to the variance in the estimate at that point.

[5]Let X = quantiles of the distribution, and Y = the nonwhite-to-white income ratio for persons. Then the straight line fits are Y = .832 - .530X for 0 < X .382 with R2 = .971, Y = .580 + .085X for .382 X .787 with R2 = .503, and Y = 1.116 - .560X for .787 X < 1 with R2 = .808. All three fits took account of the sampling errors for each point; i.e., each point had a weight inversely proportional to the variance in the estimate at that point.

[6]Forming the nonwhite to white ratio of these proportions with income above $12,000, that is, 1 percent/10 percent, as a measure of nonwhite disadvantage at that income level has an obvious intuitive appeal. The variation of such ratios with increasing income levels at a given date and the variation over time of such ratios above a given dollar income level have both been suggested to us as useful measures of changing nonwhite disadvantage. In fact, in 1967 nonwhite relative proportions with income above various dollar amounts declined linearly with income from .963 at $500 to .233 at $10,000, suggesting a steady and drastic worsening of nonwhite relative disadvantage with higher income levels. The regression, moreover, was statistically significant at the .001 level. But its substantive significance is slender. The ratios below various dollar levels suggest the opposite: a steady relative improvement with increasing income. Something like this measure is rather popular for indicating changes over time and is equally misleading for this purpose. We comment on it in Section III.

[7]Waldman, 1970, p. 12.

[8]The South as used here includes, besides the District of Columbia, 16 states: Maryland, Delaware, Virginia, West Virginia, North Carolina, South Carolina, Georgia, Florida, Kentucky, Tennessee, Alabama, Mississippi, Louisiana, Arkansas, Oklahoma, and Texas.

Previous Chapter
Next Chapter