Demographers have seldom been good prophets. In 1798 Thomas Malthus, the founder of modern demography, based his famous Essay on the principle of population on the fact that American experience showed that “population, when unchecked, goes on doubling itself every twenty-five years.” Malthus was right about America in the eighteenth century, but by 1820 the American birth rate had begun to fall and by 1900 American women were having only half as many children as they had when Malthus wrote. American fertility did not decline because of food shortages, which Malthus had seen as the only effective check on population growth. The decline may have had something to do with “moral restraint,” which was the other factor Malthus thought capable of curtailing population growth, but nothing Malthus wrote explains why Americans should have exercised more restraint in the nineteenth century than they had in the eighteenth. Nor were modern methods of contraception mainly responsible. Most Americans still relied on the same contraceptive techniques at the end of the nineteenth century as at the beginning, namely abstinence, withdrawal, and abortion.

American birth rates continued falling from 1900 to 1940, with the largest decline during the 1920s. By World War II almost all American demographers assumed that fertility would keep falling as long as life expectancy kept rising and urbanization continued. The “baby boom” that began after World War II was therefore a complete surprise. Everyone expected more babies in the first years after the veterans came home, but instead of falling back to its prewar level the birth rate kept rising until 1957. Whereas women who reached childbearing age in the 1920s had had an average of just over two children, those who reached childbearing age in the 1950s averaged three.

Because the rise in American birth rates between 1946 and 1957 was completely unprecedented, most demographers assumed it was temporary. But few anticipated that fertility would fall as low as it did in the 1970s. If we keep having children at our present rate, American women born after 1955 will have an average of only 1.8 children—the lowest average in American history. Casual observers often attribute this decline to the pill, but the decline began before the pill came on the market, and the pill does not seem to have been a major factor since then. Legalized abortions may have played a larger part, but since we don’t know how many abortions took place before legalization, it is hard to be sure. In any event, surveys strongly suggest that the main reason fertility has fallen is that couples want fewer children, though it is also true that they are having fewer children they don’t want. Few demographers have even tried to explain this change in preferred family size.

Richard Easterlin, an economist at the University of Pennsylvania, is the bestknown exception. He has been promoting economic explanations of changes in fertility for nearly twenty years. In Birth and fortune: The Impact of Numbers on Human Welfare he has tried to present his theories in simple, nonquantitative language. Easterlin claims that the birth rate depends partly on how “welloff” young adults feel. Young adults’ feelings of affluence depend, in turn, on how easy it is for them to attain the standard of living they enjoyed when they were growing up. The easier they find it to match their parents’ standard of living, the more children Easterlin thinks they will have. He calls this the “relative income” hypothesis.

Easterlin also claims that fluctuations in the birth rate have strong effects on the quality of young people’s lives a generation later. Americans born into what demographers call small “birth cohorts” can look forward to high starting salaries, favorable promotion prospects, and low rates of unemployment, because the demand for workers of their age will generally exceed the supply. Because of their economic advantages, Easterlin argues, members of small cohorts also marry young and have large numbers of children (the relative income hypothesis). In addition, he suggests that they have low rates of illegitimacy, divorce, crime, and suicide, and have a generally benign view of their society. Members of large birth cohorts, in contrast, have unfavorable economic prospects because the supply of workers their age usually exceeds demand. As a result, these unfortunates tend to marry later, have fewer children, divorce more often, kill themselves more often, commit more crimes, and so on. Social scientists often label such predictions “cohort size” hypotheses, but since this label is not very descriptive I will call them “crowding” hypotheses.

If the “relative income” and “crowding” hypotheses were both correct, they could conceivably interact to produce long-term cycles in which low birth rates in one generation led to high incomes among young adults and hence high birth rates in the next generation, which in turn led to low incomes among young adults and low birth rates in the third generation, ad infinitum. Easterlin claims that this has actually happened and that America is now completing the first such cycle, which ran from about 1940 to 1980. A new cycle of rising relative incomes, rising birth rates, and declining divorce, crime, and suicide should therefore begin soon, as the smaller cohorts born during the 1960s reach maturity. Because this theory is superficially plausible and purports to explain a wide range of otherwise puzzling phenomena, it has exerted considerable influence on scholars during the past decade. Since the publication of Birth and Fortune it has also been taken up by popular prophets like Sylvia Porter.

Advertisement

Unfortunately, Easterlin is better at inventing theories than at evaluating them. Birth and Fortune contains a number of charts, tables, and footnotes that seem to support his claims, but both the evidence he presents and the reasoning he adduces to link this evidence to his theories are full of problems.

Consider the “relative income” hypothesis. According to Easterlin, recent fluctuations in the birth rate derive from a pattern of individual behavior that has remained the same for decades, namely that couples with high incomes relative to those of their parents (i.e., high “relative incomes”) have more children than couples with low incomes relative to their parents’. Since couples raised during the Depression reached childbearing age in the affluent years after World War II, an unusually large proportion had high incomes relative to their parents’, and the birth rate rose. When couples raised during the affluent postwar years reached childbearing age in the 1960s, a smaller proportion were able to match their parents’ incomes, not because times were hard in the 1960s but because their parents had done so well in the 1940s and 1950s. As a result, birth rates fell. Those who reached childbearing age during the stagnant 1970s were even less likely to feel affluent compared to the standards they had absorbed in the 1950s and 1960s, so the birth rate fell even further.

While Easterlin is right in claiming that the birth rate has risen and fallen since 1940 in tandem with the proportion of young couples enjoying high incomes relative to their parents’, he offers no evidence whatever that couples with high incomes relative to the income of their parents actually have unusually large families, much less that this relationship is strong enough to account for the dramatic changes in fertility between 1940 and 1980. Instead, he claims that “the requisite income history data [for testing this claim] do not exist.” This assertion is grossly misleading. Much relevant data exists, and none of it supports Easterlin’s theory.

Sociologists have been studying the relationship between fathers’ and sons’ economic positions for half a century, and in recent years demographers have looked quite carefully at the relationship of economic mobility to fertility. Because it is hard to gather retrospective data on parental income, demographers have usually measured economic mobility by comparing a man’s current occupation to his father’s occupation. Such investigations always find the same thing. Men working in highly paid occupations have fewer children than men working in poorly paid occupations. This remains true even when one compares men whose fathers were both in the same occupation. 1 This is precisely the opposite of what Easterlin’s relative income hypothesis would lead one to expect. Of course, incomes vary a lot within occupations, and Easterlin may believe that if demographers measured relative income instead of relative occupational status the results would be different. But he does not make this argument. He simply brushes the issue aside with his sweeping claim that “the requisite income history data do not exist.”

In fact, recent surveys also provide direct evidence on the effects of relative income. Siblings raised together, for example, have roughly the same standard of living while they are growing up. Easterlin’s theory therefore predicts that if one sibling ends up with more income than another, the more affluent sibling should have more children. Three years ago two sociologists at the University of Wisconsin tried to test Easterlin’s theory by looking at the relationship of income differences between brothers to fertility differences. Their findings supported Easterlin in the limited sense that the more affluent of two brothers did tend to have more children than the less affluent. But a 25 percent income differential between two brothers was associated with an average fertility difference of only 1 percent.2

Since Easterlin’s calculations suggest that the average American couple’s relative income declined less than 25 percent between 1957 and 1977, the evidence on brothers implies that this decline in relative income should not have reduced fertility by more than I percent. Since the overall decline in fertility was at least 40 and perhaps more than 100 percent, depending on the measure one uses, the decline in relative income was not, on this evidence, an important explanation of the decline in fertility. It is true that this study of brothers covers only one city, Kalamazoo, Michigan, so Easterlin might argue that it is not representative of the nation as a whole. But he makes no such argument. Instead he claims that income data on individuals suitable for testing his theory does not exist at all.

Advertisement

Several other major surveys have measured relative income even more directly. The best records come from a long-term study of men who finished high school in Wisconsin in 1957. After taking elaborate precautions to ensure confidentiality the University of Wisconsin obtained access to these men’s Social Security records from 1957 through 1971 and to their parents’ state income tax returns for 1957 through 1960. Easterlin may not have known about these records’ existence. But they are among the richest and best-known sources of evidence on matters of this kind in all of social science, so even casual inquiries among demographers would have sufficed to reveal their existence. Are we to believe Easterlin made no such inquiries? If he didn’t how could he assert with such confidence that the data needed to test his theory did not exist? If he did, how could he have failed to discover this data?

A little more inquiry would also have told Easterlin that two Wisconsin demographers were using the Wisconsin records to test his own relative income hypothesis. Their work is now complete, and it provides even less support for his theory than the data on Kalamazoo brothers. In Wisconsin, a 25 percent difference in relative income was associated with a difference of only 0.5 percent in fertility.3 On this evidence, then, declining relative income accounts for about 1 percent of the decline in fertility over the past generation.

These are not isolated instances. So far as I can discover, no study of individual fertility has ever suggested that any measure related to relative income appreciably affects fertility. Partly for this reason, very few demographers or economists accept Easterlin’s theory. Economists of the “Chicago school” do, however, continue to claim that children constitute a costly “investment” and that decisions about family size must therefore involve the same kinds of considerations as decisions about how big a house to buy or how many cars to maintain.

This view does not bear scrutiny. If the costs of having children were mainly financial, we would expect affluent couples to “consume” more children than poor couples, just as they consume more of almost everything else. Since affluent couples have fewer children than poor couples, at least in America, the nonmonetary costs of parenthood must greatly outweigh the monetary costs for most Americans. The most obvious costs of having children are time (“freedom”) and energy. Affluent couples without children have more appealing ways to use their time than poor couples do, so it is not surprising that they marry later and have fewer children.

But why did couples at all income levels start having more children in the 1950s, reversing the 150-year decline in family size? We don’t know the answer to this question, but we do know that fertility increased throughout American society, among rich and poor, highly educated and poorly educated, urban and rural, black and white. This strongly suggests that the change was in some sense a collective decision, one that grew out of the mood of the times. Couples find it very hard to judge in advance how much they will enjoy having children. They therefore depend heavily on their friends and relatives to reach a judgment about whether big families are more or less satisfying than small ones. Like other predictions based on inadequate evidence, these judgments are inherently volatile. Since they depend partly on ideas and images transmitted by television, movies, magazines, and even books, it is tempting to credit producers and editors with exerting a heavy influence on them. But most producers and editors try to tell their audiences what they think they want to hear. I therefore suspect that the “mass media” usually end up accentuating swings in opinion about family size rather than creating them.

The argument that decisions about family size are to a great extent collective rather than individual in character does not in itself explain birth rate fluctuations. But it does suggest that if economists want to explain such fluctuations they might try comparing the “demand for children” to the demand for fads and luxuries such as hula hoops, designer jeans, or cocaine instead of comparing it to the demand for “consumer durables” like housing and automobiles.

Whatever the causes, birth rate fluctuations have important consequences. For Malthus, of course, the birth rate was important because it influenced total population and hence the ratio of population to natural resources. Recently, however, a number of observers have become interested in the effects of short-term fluctuations in the birth rate on social and economic life. In 1973 Daniel Patrick Moynihan even went so far as to blame the social upheavals of the late 1960s and early 1970s on the disproportionate number of teenagers and young adults at the time. More recently, the financial troubles of the Social Security system, which derive partly from the small number of men born in the 1920s and 1930s, have again underlined the importance of “lumpy” age distributions.

Easterlin’s theories on “crowding” imply that because fewer young people will be entering the labor market during the 1980s and 1990s, virtually all the problems now endemic among teenagers and young adults—problems that derive from unemployment and a sense that their chances of success are narrowing—will diminish. Those who want to abandon political efforts to deal with these problems are bound to find such a theory useful, though I doubt that that was Easterlin’s intent. It is therefore important to emphasize that most of Easterlin’s “crowding” theories are speculative and that some are at odds with the evidence he himself cites.

Central to Easterlin’s “crowding” theories is his assumption that when employers hire young workers they usually have to hire older ones as well both to supervise the young and to do other jobs that novices cannot handle. To the extent this is true, demand for old and young workers will fluctuate together. When the supply of young workers rises faster than the supply of older workers, as happened from 1960 to 1980, the wage gap between young and old should therefore widen. When the supply of young workers rises less rapidly than the supply of older workers, as happened from 1940 to 1960 and will happen again from 1980 until the end of the century, the wage gap between young and old should narrow. Therefore the prospects for young people will grow brighter.

Easterlin offers no direct evidence on wage changes between 1940 and 1955. But he does show that whereas a regularly employed twenty- to twenty-four-year-old man earned 73 percent of what a regularly employed forty-five- to fifty-four-year-old man earned in 1955, this ratio had fallen to 56 percent in 1977. This is consistent with his argument that young people born during the baby boom are having a harder time in the job market. Furthermore, young men’s wages have risen a little faster than mature men’s wages since 1977, suggesting that the decline in births after 1957 may once again be narrowing the wage gap.4 Other economists, using far more precise methods than Easterlin uses, have also found that the relative wages of the young depend to some extent on the size of the birth cohort from which they are drawn. Thus I think Easterlin is probably right in predicting that young men’s wages will rise faster than mature men’s wages during the 1980s and 1990s.

There is also some evidence that high birth rates lead to higher unemployment a generation later, though the effect is not large. Easterlin, for example, indicates that unemployment among males aged forty-five to fifty-four fell from 4.0 percent to 2.8 percent between 1950 and 1978, while unemployment among males aged twenty to twenty-four rose from 8.1 percent to 9.1 percent. Unemployment rose much more dramatically among young blacks, but this was partly because able young blacks were much more likely to be in school in 1978, so those looking for work were those employers least wanted to hire.

Easterlin’s argument does not seem to apply to women, however. In 1955 regularly employed women aged twenty to twenty-four earned 99 percent of what those aged forty-five to fifty-four earned. This reflected the fact that women worked in occupations where young, inexperienced workers did virtually the same thing as older workers. An employer looking for a school-teacher, cashier, waitress, sales clerk, or typist would usually favor applicants with a year or two of experience over applicants with no experience whatever, but he would not pay the same premium for twenty years of experience in these occupations that he would pay for a doctor, corporate manager, or plumber with twenty years of experience. If employers had still regarded old and young women as interchangeable in 1977, the glut of young women seeking work should have depressed the wages of everyone in traditionally female occupations, regardless of age. Yet the fact is, as Easterlin notes, that regularly employed women aged twenty to twenty-four earned only 82 percent of what forty-five to fifty-four-year-old women earned in 1977, compared to 99 percent in 1955.

This suggests that by 1977 a significant number of older women must have held jobs traditionally reserved for men, where experience paid off. If women are in fact moving into nontraditional jobs where experience matters, the wage differential between old and young women is likely to widen in the 1980s and 1990s, instead of narrowing as Easterlin predicts.

Easterlin also argues that when young men’s incomes are low relative to their fathers, as they were in the 1960s and 1970s, women are more likely to work, since this is the only way they can attain the standard of living to which they aspire. When young men’s incomes are high relative to their fathers, as they were in the 1940s and 1950s and should be again in the 1980s, more women feel they can afford to become full-time mothers. Employment patterns are consistent with this theory.Employment opportunities for women have expanded steadily since 1940. But the rates of participation of women under thirty-five in the labor force were no higher in 1960 than in 1940. According to Easterlin this was because young men were doing so well economically during the 1940s and 1950s that young women felt they could afford to get married, stay home, and have children. As a result, employment rates grew only among women over thirty-five. After 1960, when young men’s relative incomes began to fall, the proportion of younger women in the labor force grew dramatically. The mothers of the children born during the baby boom, in contrast, remained at home even after their children reached adolescence, living on their husbands’ still ample incomes.

But the relationship between relative income and birth rates illustrates the danger of drawing conclusions about individual behavior from such aggregate statistics. And just as with birth rates, Easterlin does not offer a shred of evidence that wives whose husbands have low relative incomes are more likely to work than wives whose husbands have high relative incomes. In the absence of such evidence I would attribute changes in the age distribution of working women to changes in collective attitudes toward motherhood and work, and I would bet against a reversal of these attitudes in the 1980s.

Easterlin’s theories become even more speculative when he considers issues that are not economic. He claims, for example, that relative income affects marital stability, and that young people who find themselves markedly worse off than their parents were will tend to have more divorces. To support this claim he shows that while the divorce rate has climbed steadily throughout the twentieth century, it climbed faster when young men’s earnings lagged behind their fathers’, as happened during the 1930s, 1960s, and 1970s. But he does not show that a couple’s relative income is actually associated with their chances of getting divorced. We know that, contrary to folklore, the rich get divorced slightly less than the poor, which is consistent with the relative income hypothesis. But whether rich couples with poor parents stick together more often than rich couples with rich parents nobody seems to know, and Easterlin has not tried to find out.

Easterlin even argues that the illegitimacy rate depends on relative income. This is not because he thinks relative income determines the frequency of premarital sex or the efficiency of premarital contraception. Rather, he argues that after a baby is conceived out of wedlock, the parents are more likely to marry before the baby is born if the father’s relative income is high. Easterlin shows that the percentage of illegitimately conceived babies whose parents married before the baby’s birth fell steadily from the late 1950s to the late 1970s, as his theory says it should. But this percentage was as low during the late 1940s and early 1950s as it is today, which is contrary to his theory.

Easterlin also uses cohort size to explain changes in crime, suicide, alienation, and unhappiness. But the evidence is so thin and the emphasis on aggregate rather than individual statistics so pervasive that it is hard to call these arguments much more than speculations.

I began reading Birth and Fortune just as David Stockman began trying to eliminate federal support for social science on the grounds that social science is of little social value. I must confess that I felt a certain sympathy for Stockman’s views as I read. Despite fifty years of research, almost none of the behavior that Easterlin tries to explain is at all well understood. Demographers cannot even predict birth or divorce rates as accurately as, say, meteorologists can predict the weather. Unlike physical systems, social systems are conscious of their history and allow this history to influence them. This makes it impossible to abstract a few common elements from different historical moments and assume that these few elements act independently of the others to produce the same consequences time after time. They don’t.

On reflection, however, Easterlin’s work also provides an excellent example of how social science can, in fact, be extremely valuable. The absence of rigorous, well-validated social theories does not lead most people to give up theorizing and learn to live with uncertainty. When we don’t have good theories we usually invent bad ones. The greatest accomplishment of social science in the twentieth century has been to develop methods for determining just how bad most such theories are. A generation ago, for example, it would have been virtually impossible to say whether Easterlin’s relative income hypothesis was right or wrong. We simply would not have had enough evidence to pass judgment. Even if we had had the relevant evidence, we would not have had the tools for analyzing it. Since the relative income hypothesis is superficially plausible, it would probably have exerted enormous influence for many years, just as Malthus’s theory did. Today we can test such theories quite easily, at least if we want to.

In most cases, our theories prove deficient; social reality resists simplification. The major value of modern social science is thus to promote skepticism about bad social theories. Such achievements will not win social scientists much thanks from their fellow citizens, especially those who, like Stockman and the “supply-side” economists, have a bad theory of their own to peddle. Nonetheless, skepticism remains more honorable than shamanism.

This Issue

October 8, 1981