Years of Poverty, Years of Plenty, by Greg J. Duncan et al., Ann Arbor: Institute for Social Research, University of Michigan, 184 pp., $24.00
Issues revolving around poverty and "income distribution" are among the most controversial of our time, so a book that offers new facts on these subjects—some light to go with the heat—has at least the promise of being valuable. Years of Poverty, Years of Plenty is such a book. It exhibits both the best and the worst of modern "social science."
First, the best: It summarizes and analyzes for a more general public the results of a massive, ongoing study at the University of Michigan on income and poverty in America. This is done with a clarity that belies the impression too often created that social scientists are incapable of writing English. Tables are organized well and explained with care, so that the meaning of the numbers becomes readily apparent. The actual findings of the study are also very important and challenging to many widely held beliefs.
Then the worst: Interspersed with this careful analysis, especially in the second half of the book, is freewheeling policy prescription on issues that would each require a book twice this size. Indeed, the selection and presentation of factual evidence suffers from the constant oscillation between analysis and policy prescription. An otherwise excellent book thus degenerates repeatedly into tendentious politics. The authors clearly want to "do good," and they insist on doing it between the covers of the same slim volume in which they are analyzing complex social phenomena.
In contrast to a vast literature that uses annual income statistics to determine inequality or poverty among Americans, this study sees this annual data as only an instantaneous snapshot of a complex ongoing process that must be understood as a process, rather than as a collection of momentary details. For example, nearly half the families who were in the bottom 20 percent in 1971 were not there in 1978, and 6 percent of these "poor" families were later in the top 20 percent. Likewise, less than half of those families in the top 20 percent remained there for the entire period. This fluidity was common throughout the income distribution. Less than half of all American families remained in the same income stratum—the same quintile—both years.
What does this do for our popular concepts of "the poor," or "the affluent," who tend to be discussed as if they were people persistently in the same relative positions in the income hierarchy? Those who are persistently poor—in the bottom 20 percent at least eight out of 10 years—are less than 3 percent of the US population.
The very concept of a family also becomes less set in stone from a multiyear perspective. At the end of 11 years, nearly half of all families were headed by a different person. The growing up of children, the remarriage of widows, the divorce of parents, all create a very dynamic process that confounds static definitions.
This study reveals that the family, among other things, redistributes income—far more income than the government redistributes. The formation and dissolution of families is one of the major factors in getting women into and out of poverty. More than half the difference between black and white women in their rate of change in income was due to the lower marriage rates of black women.
A dynamic study of welfare recipients likewise presents a different picture from that derived from one-year statistics. Just as there is a considerable turnover among people at different income levels, there is also a considerable turnover of people on welfare. This means that there is a much larger proportion of the population on welfare over a decade than at any given time. About one-fourth of the total population received at least one of the many forms of welfare, at some time or other during the decade beginning in 1969. But less than 1 percent of the population received half or more of their income from welfare in as many as eight years of that decade.
Employment was also found to be quite different when seen in a one-year snapshot, as compared to a multiyear motion picture. The hourly rates of pay for men in their prime working ages changed by an average of 25 percent from one year to the next. Moreover, while both black and white male household heads averaged more than 2,000 hours worked in 1969, most of them also had hours fluctuations over the next decade of at least 250 hours per year—the equivalent of more than six weeks' work. A substantial majority had hours variations of at least 500 hours for at least one year of the decade.
Years of Poverty, Years of Plenty is at its best in its descriptions of dynamic income changes. When it proceeds to the next step of determining causation, it ventures onto shaky ground, and when it tries to prescribe policy it sinks into a swamp. While this book's virtues are its own, its faults are common to a large body of "social science" literature and so are well worth exploring for that reason.
Statistical "explanations" of causation are essentially statements that holding constant some variable (education, urbanization, etc.) will reduce the income differences between groups by some fraction. But, like others who use specialized terminology with highly restricted meaning, the authors of this study drift into using it as if its more general meaning also applied. Thus, after various factors used to "explain" male-female income differences still leave a substantial unexplained residual, the authors ask: "If skills do not 'explain' most of the wage gap, then what does explain it?" After a momentary pause to note the difficulty of testing alternative hypotheses, the authors proceed to ignore this caveat, come down heavily on the side of discrimination, and launch into advocacy of various government policy "solutions," including affirmative action, equal pay laws, and projecting a different image of women in textbooks. They even assert that sex discrimination has been "proven" before the Equal Employment Opportunity Commission—omitting the crucial point that such "proof" almost invariably consists of nothing more than statistics used in precisely the same way that this study uses them to "explain" economic differences. The authors add that discrimination is "confirmed by psychologists and sociologists," a statement that is undoubtedly correct as to the general state of political opinion in the so-called social sciences but says nothing about the actual situation in the real world.
The problems inherent in this shifting use of words such as explain and proof become apparent if we consider height as an explanation of wage differences. If the wages earned by each person in the population are tabulated along with his or her height, we will undoubtedly find a significant correlation between height and earnings, because most people under five feet tall are too young to be earning anything. Statistically, holding height constant will therefore reduce income differences and so "explain" a certain amount of variance. But in fact it has explained nothing in any meaningful cause-and-effect sense. Growing up simply happens to be related to acquiring schooling, skills, and experience, all of which affect earnings.
The authors of this study understand this principle. They simply proceed as if they don't—like many "social scientists" trying to do good. They acknowledge more than once that the variables they specify may not be defined and measured with the precision required for conclusive evidence. Ultimately that is true of all variables in all studies. However, what is crucial here is not that these authors have approached no closer to perfection than anyone else, but that large and well-known qualitative differences in the education of men and women are utterly ignored in their statistical "explanation."
What the authors treat in their statistics as the "same" education is not close to being the same between males and females, beginning in high school and especially in higher education. The mathematics component—crucial in the social sciences and engineering, and increasingly important in other fields—is not even remotely the same. Young women enter college with far less mathematics background than young men—so much so as to put large areas of study off-limits to women from the beginning. Ignoring this is not a question of general measurement problems but rather of ignoring things known in advance to be very different in the groups being compared. It is as if the height "explanation" of earnings were to be used in "explaining" economic differences between the Swedes and the Japanese, knowing in advance that these two groups differ in average height at every age. By dichotomizing all possible reasons for sex differences in earnings into discrimination and socialization, the authors also utterly ignore physical differences that affect the likelihood that women will enter physically demanding and well-paid fields, such as construction, mining, and the like.
Most important of all, discrimination is not treated like other possible explanations and subjected to empirical tests and analytical critiques. It is the residual beneficiary of anything not explained by other specified variables. If it isn't discrimination, then what is it? That is the basic challenge. Others have asked: If it isn't genetics, then what is it? Still others might ask: If it isn't culture, then what is it? If it isn't luck, etc., etc.?
This approach is in complete contrast to the way the same authors proceed when testing alternative theories of the labor market in general. Here they compare alternative "human capital" and "dual labor market" theories, and test them against each other and the facts. They do not say that any residual unexplained by one must be due to the other.
Given the huge unexplained residuals commonplace in the social sciences at this stage of their development, the automatic attribution of the unexplained to some favored theory is especially indefensible. At its worst, it borders on the Marxist tactic of declaring that something is "no accident"—the only alternative to chance being their special theory.
Other parts of the liberal-left vision are "proven" in equally cavalier fashion. People's own attitudes and values have little effect on their economic fate, the authors conclude, because attitude survey answers have little or no correlation with subsequent economic levels. But this ignores the fundamental point that values are ultimately questions of trade-offs, of how much one is willing to sacrifice of one thing to achieve other things. Attitude surveys are tests of lip service, not sacrifices. It is hardly surprising that such surveys often show low-performance groups with the same "values" as high-performance groups when values are defined in ways that have no relevance to the issue.
Finally, when the authors turn directly to policy prescription, the arbitrary and the unsubstantiated come into their own. After having demonstrated, at the beginning of the book, the slippery and ever-changing human content of such categories as "the family" or "the poor," the authors nevertheless end by implicitly accepting, as the touchstone of social policy, statistical parity of results among such categories. Why it is morally imperative that statistical categories look equal on paper is a question never addressed. Worse, the authors falsely attribute to others the notion that labor markets are morally "fair," that women "deserve" less pay than men. This is quite a degeneration for a book whose special qualities might represent a major contribution to understanding, if it did not so readily succumb to the fashions, visions, and moral self-indulgence that make a mockery of much contemporary "social science."
Thomas Sowell, an economist, is a senior fellow at the Hoover Institution. He is the author of numerous books, including Knowledge and Decisions, The Economics and Politics of Race, Civil Rights: Rhetoric or Reality?, and, most recently, Marxism: Philosophy and Economics.