Jump to ContentJump to Main Navigation
Birth QuakeThe Baby Boom and Its Aftershocks$

Diane J. Macunovich

Print publication date: 2002

Print ISBN-13: 9780226500836

Published to Chicago Scholarship Online: February 2013

DOI: 10.7208/chicago/9780226500928.001.0001

Show Summary Details
Page of

PRINTED FROM CHICAGO SCHOLARSHIP ONLINE (www.chicago.universitypressscholarship.com). (c) Copyright University of Chicago Press, 2022. All Rights Reserved. An individual user may print out a PDF of a single chapter of a monograph in CHSO for personal use. Subscriber: null; date: 21 May 2022

Effects of Changing Male Relative Income on Marriage and Divorce

Effects of Changing Male Relative Income on Marriage and Divorce

(p.144) 9 Effects of Changing Male Relative Income on Marriage and Divorce
Birth Quake

Diane J. Macunovich

University of Chicago Press

Abstract and Keywords

This chapter considers the closely related issues of marriage and divorce. Divorce rates appear to have topped out since about 1985 and have even begun to decline, as male relative income has stabilized and begun to improve. For the baby boomers, higher divorce rates appear to leave permanent scars. There is no apparent trend in divorce rates at each age among the baby boomers themselves. Marriage postponement that may have led from relative income effects showed up as simple college enrollment in the Wisconsin sample, which has a strong negative effect on propensity to marry. Young women will be “squeezed out” of the marriage market when there is a relative shortage of unmarried men about two years older than themselves. Rising women's wages appear to have encouraged marriage and discouraged divorce among young working women.

Keywords:   marriage, divorce, male relative income, baby boomers, college enrollment, young women, wages

Our findings suggest that only the male partner's economic resources affect the transition to marriage, with positive economic situations accelerating marriage and deterring separation. Our results imply that despite trends toward egalitarian gender-role attitudes and increasing income provision among women, cohabiting men's economic circumstances carry far more weight than women's in marriage formation.

Smock and Manning, “Cohabiting Partners' Economic Circumstances and Marriage” (1997)

Only a statistician could even imagine that we're through with marriage. The truth is that Americans are nuts about the institution.

Peter Godwin, “Happily Ever After: Americans Will Always Love Marriage and Polls about Other People's Lives”(1999)

Are Americans really “nuts” about marriage?1 Have we simply been living through a temporary aberration caused by the birth quake, or does the much-heralded decline in marriage rates signal the inevitable demise of the institution? Will husbands and wives be replaced by a “posslq”?2 You'd be hard-pressed to answer that question using official statistics on age-specific marriage and divorce rates: the government stopped publishing such statistics back in 1988/89. Ironically, as it turns out, that's just about the time when things began to get interesting. Marriage trends that had followed a seemingly inexorable downward trend since 1970 have begun to turn around, as indicated in figure 9.1. (p.145)

Effects of Changing Male Relative Income on Marriage and Divorce

Figure 9.1 Percentage of men who have been married by age 30, 1965–1998. Official statistics stop in 1989, so they don't show the “turnaround” that appears to have occurred in the 1990s. Source: author's tabulations of data from March Current Population Surveys.

The previous chapters documented dramatic changes in male relative income and female wages and related these to changing relative cohort size. The relative income hypothesis suggests that these declines in ex ante income relative to aspirations should have brought about demographic adjustments aimed at improving the ex post incomes of young adults. There should have been a movement away from marriage, as growing proportions of young men perceived an inability to support a family at a level commensurate with their own material aspirations and as young women responded to the independence afforded them by increased labor force participation. And divorce rates should have risen due to the stresses caused by difficulties in “making ends meet.” The obvious question, then, is whether the turnaround illustrated in figure 9.1 is related to the turnaround in male relative income that began after 1985.

The timing seems right, since these effects should show up most clearly at younger ages, where relative cohort size effects in the labor market are felt most strongly. A turnaround in the economic fortunes of 20-somethings after 1985 would be expected to show up by 1995, in the proportions of men who have chosen to marry by age 30.

You might well question the data presented in figure 9.1: what are they, if not official statistics? They are my own tabulations using (p.146) another set of data provided by the Bureau of Labor Statistics and the Census Bureau—the March Current Population Survey (CPS) for the years 1962–1998. The CPS doesn't provide marriage rates, but does indicate individuals' current marital status, along with a great deal of other demographic information, in a representative sample of the population. Using these data it's possible to calculate not just the proportion currently married or divorced, but also the change in those rates for a given birth cohort from year to year: effectively, cohort marriage and divorce rates.

The cohort marriage rate is calculated as the increase in the proportion ever married from one year to the next, divided by the proportion never married in the previous year.3 The cohort divorce rate is similarly calculated as the change in the proportion divorced from one year to the next, divided by the proportion who were married or separated in the previous year. This is true of all statistics presented here. It should be noted that the divorce rate is more error-prone than the marriage rate, however, since individuals can move out of the divorced state as well as in, through remarriage. Divorce rates calculated in this way may well be underestimates, especially at younger ages, but both the derived marriage and divorce rates accord well with official statistics up to 1990.

Thus, whereas the official statistics present a fairly bleak picture with no suggestion of a turnaround in marriage rates, the derived rates indicate that there have been tantalizing changes since 1989. Figure 9.2 illustrates some marriage rates estimated in this way, among young men and women by length of time out of school. The annual rates by single year of age derived using the procedure described above have been averaged into three-year age groups in order to minimize “noise” in the data. In addition, data are presented for young men at slightly older ages than for young women, reflecting the traditional age difference between marital partners.

Because of the virtual embargo on age-specific marriage and divorce rates imposed by federal budgetary restrictions since 1990, the data presented here and in Macunovich 2002 contribute significantly to knowledge about recent changes in marital behavior. They confirm the suspicions of many of us who've felt barraged by marriage announcements in recent years: marriage rates are up, particularly among those in their later 20s, and there is a slight trend back toward marriage at earlier ages.

The long-term decline in marriage rates at younger ages that (p.147)

Effects of Changing Male Relative Income on Marriage and Divorce

Figure 9.2 Annual marriage rates among young men and women between 1968/70 and 1998, showing the dramatic decline after 1970 and the tendency toward stabilization and even turnaround after 1985. Patterns here appear to mimic those displayed in the graphs of male relative income in chapter 4. Source: author's tabulations of data from March Current Population Surveys.

proceeded unabated for over fifteen years from the early 1970s appears to have been arrested since the late 1980s, and the break after 1985 appears to be more pronounced among males than among females. In addition, among all of the young women and among male college graduates there has been a marked upward trend in marriage rates at younger ages in the latter part of the 1990s.4

That the post-1985 stabilization was most pronounced among young men suggests that causes of the trend acted more directly on them than on young women: this would be the case with changes in male relative income. That is, 20–22-year-old males responding to increases in relative income may select partners their own age, or in any one of a number of other cohorts. Thus the effect of any change in their relative income on women's marriage rates will be diluted to the extent that males choose partners in various age groups.

If space permitted it would be possible to demonstrate that, as (p.148)

Effects of Changing Male Relative Income on Marriage and Divorce

Figure 9.3 Estimated annual divorce rates between 1968 and 1998 among young men and women who were previously married or separated, showing the dramatic increase after 1970 followed by an apparent stabilization and turnaround after 1985. The patterns here suggest a strong response to changes in male relative income—especially the decline, increase, and renewed decline in the period after 1985. Source: author's tabulations of data from March Current Population Surveys.

with male relative income in chapter 4, the time trends in marriage rates exhibited in figure 9.2 emerge in all conceivable breakdowns of the data on young men and women, for example, by race, region of the country, educational attainment, and part-time/full-time employment status. Just as all young groups experienced staggering declines in male relative income, ranging from one-half to two-thirds, with a hump-shaped recovery after 1985, the proportions of young men entering marriage plunged, as well, by 60–80 percent—with a similar tendency toward resurgence, decline, and resurgence between 1985 and 1995.

Figure 9.3 demonstrates that divorce rates among younger men and women have tended to increase when marriage rates decline and vice versa. This suggests that the same factor(s) might be relevant in explaining both sets of trends. Divorce rates seem to have topped out since about 1985 and have even begun to decline, as male relative income has stabilized and begun to improve.

(p.149) Cohort or Period Effect?

Andrew Cherlin, a respected researcher in the study of marriage and divorce in the United States, has acknowledged the possibility of cohort effects in the trends observed in the post-WWII period, but like many other social scientists he comes down fairly firmly on the side of period effects: “something in the air.”5 His reasoning relies on the fact that many age-specific rates appear to have moved in parallel when major transitions were occurring.

A comparison of trends by age group, however, in percentage ever married suggests that the only reason rates appeared to move in parallel is because the decline at younger ages was so long lasting that downward trends among the first baby boomers as they aged were matched by those of later baby boomers in their early 20s. This is demonstrated in figure 9.4, in which the age groups presented are five years apart—25, 30, 35, 40, and 45. The decline in percentage ever married at each age really began in earnest only after the baby

Effects of Changing Male Relative Income on Marriage and Divorce

Figure 9.4 comparison of time trends in percentage ever married across age groups and time. The vertical lines define five-year periods following the initiation of decline in the percentage married among 25-year-old men. There was a slight drop in that age group after 1967 (that is, among cohorts born 1942–1946), and then the prolonged decline among the baby boom cohorts set in after 1972. The vertical lines indicate that the declines in successively older age groups began only once those same baby boom cohorts had begun entering the age group. Thus, although the trends appear to be parallel, they did not all begin at the same time: this suggests a strong cohort effect, rather than a period effect.

(p.150) boom cohorts (post-1947) had begun entering successively older age groups. As demarcated by the vertical lines, there was an initial drop in percentage married at age 25 among cohorts born 1943–1947 (in 1967–1972), but the real decline occurred among the cohorts born after 1947, from 1972 onward. The decline among 30-year-olds began approximately five years later. This is the case for each successively older age group, suggesting that the decline was a cohort rather than a period effect.

Another point that emerges from figure 9.4 is that this apparent decline was largely a case of delayed marriage, rather than marriages forgone altogether. By the time the boomers reach age 45, their percentage ever married is not much lower than that of their parents. And although I haven't included it here, this is even more the case by age 50: 93.2 percent of 50-year-olds in 1998 were ever married, while twenty-five years earlier that figure was 93.6 percent, and the highest it's been in the last forty years is 96.1 percent, in 1986. But, you might say, very few boomers have yet reached age 50. For this reason it's useful to examine tendencies as they emerge among cohorts rather than in age-group patterns, as in figure 9.5. (And because I've used men to illustrate the points so far, this time I'm showing the patterns among women: there are very few differences between the sexes, in overall marriage patterns!) The figure depicts the life course of cohorts

Effects of Changing Male Relative Income on Marriage and Divorce

Figure 9.5 Cohort marriage patterns by age, showing the strong tendency toward convergence at older ages. Marriage isn't dead—it's just been delayed!

Effects of Changing Male Relative Income on Marriage and Divorce

Figure 9.6 Cohort patterns of divorce among those previously married or separated (top) and their effect on percentages in intact marriages, showing the sharp dichotomy between pre-and post-World War II birth cohorts, as well as the lack of any trend across baby boom cohorts and the tendency toward lower divorce rates among post-boom cohorts.

of women, in terms of the percentage ever married at each age. It shows a wide dispersion at younger ages—up to about 30—as successive baby boom cohorts moved further away from traditional family formation in their 20s, but with a very strong tendency toward convergence after that age. Perhaps Americans really are nuts about marriage!

Despite the apparent tendency toward convergence in percentages ever married, however, a different trend emerges in percentages currently married (i.e., not divorced or separated), as shown in figure 9.6. There, a sharp dichotomy is apparent between pre- and post-WWII birth cohorts: there is a tendency toward convergence, but in two separate groups. For the baby boomers, higher divorce rates appear to leave permanent scars.

It's notable in figure 9.6 that there is no apparent trend in divorce rates at each age among the baby boomers themselves. All cohorts (p.152)

Effects of Changing Male Relative Income on Marriage and Divorce

Figure 9.7 Patterns of divorce at age 30 among men and women who were previously married or separated, showing that the upward trend ended with the baby boomers: rates have been fairly level since the 1945 birth cohort turned 30 in 1975, and now there is evidence of a downward trend in post-boom cohorts.

seem to exhibit a very similar lifecycle pattern. There was simply a break from the pattern established among pre-War cohorts.

There is, however, some indication of a decline in divorce rates among post-boom cohorts, as shown by the position of the curve for the 1975 birth cohort. This is borne out by the pattern of the percentage divorced at age 30, among both men and women, as shown in figure 9.7. Divorce rates ceased their increase with the 1945 birth cohort in 1975, and the trend among these 30-year-olds appears to have been downward since the mid-1990s. Once again this is consistent with the hypothesis that divorce rates were responding to improvements in male relative income that began after 1985 for those in their 20s.

Thus if we wish to explain marriage trends we need to explain the forces that caused baby boomers to delay marriage at earlier ages: this appears to be the source of the overall trends that have emerged over the past thirty years. Why did such a large proportion of young people in the baby boom cohorts put off marriage—and why did so many of their marriages fail?

(p.153) College Enrollment as a Deterrent to Marriage

Studies have confirmed that current enrollment in school has a marked effect on marriage rates. One influential study found that current enrollment was in fact one of the strongest deterrents: “[Y]oung men's earnings and the time spent in schooling to increase them were found to be important influences on marriage timing. Additional schooling had little effect net of the time it absorbed.”6 That is, the researchers found that being in school had a significant deterrent effect on the probability of marriage, but that additional schooling on its own, after controlling for the time taken to acquire it, had no significant impact on marriage probabilities.

That one study has tended to reinforce, in the minds of most researchers, Cherlin's dismissal of cohort effects. It “provide[d] no support for Easterlin's hypothesis that marriage will occur earlier when young men judge their economic prospects favorably with respect to their parents' income.”7 But this study, based on an analysis of Wisconsin males who graduated high school in 1957, bears closer examination. The database was unusual in its depth of coverage: it contained Social Security earnings records for the young males, together with income tax reports for their parents and follow-up information through 1975. Researchers conducted a year-by-year analysis, measuring the probability of marriage among these young males in the years 1958–1965. The results show a strong positive effect of the young males' own income on their propensity to marry, but little or no effect of their parents' income. The researchers concluded that the relative income concept is not only erroneous, but “obfuscates important effects of young men's current earnings.”

This finding is extremely significant in that it appears to ignore the potential connection between relative income and enrollment rates. If enrollment rates as well as marriage rates are a function of relative income, then the analysis produced only a partial estimate of the effect of relative income on marriage, ignoring the secondary effect operating through enrollment rates. As described in the previous chapter, the relative income model suggests that young males, in making a decision regarding higher education, will compare their potential earnings as high school graduates to those of their fathers and will assess the potential returns to education relative to the size of the “gap” between these two. In this context, the males in the Wisconsin 1957 sample who enrolled in college would have been composed largely of those who (p.154) had judged their potential high school earnings to be deficient relative to their aspirations. In this sense, then, the sample of wage earners was actually truncated, and truncated in a way that would have removed just those young males whose parental earnings were “high” and thus indirectly caused postponement of marriage.

Thus marriage postponement that may have resulted from relative income effects showed up as simple college enrollment in the Wisconsin sample—which did, indeed, have a strong negative effect on propensity to marry. It seems possible that a re-estimation of the Wisconsin model with college enrollment as a function of relative income would have shown support for the relative income hypothesis, and hence for cohort size effects on male marriage rates.

Changing Values

Probably the most common theme running through various analyses of delayed and declining marriage (and fertility) rates is that of “changing values”: attitudinal changes that have, in the words of one group of researchers, “undermined the social and economic forces that maintained the institution.”8 Some feel that these changes were essentially independent of economic forces, instead arising from the continuing growth of individualism that began with the Enlightenment.9 But others maintain that such change is not autonomous: it derives from—or at least is intensified by—a host of factors, including the growing financial independence of women; the decline of religious values; improved contraceptive methods, which made sex outside of marriage less risky; and lower fertility, which reduced some of the “function” of marriage.10 Also stressed, however, is the increasing social acceptance of such factors: not only are the associated behaviors more attractive, they are also less “costly” in terms of potential social sanction. This leads one to envision a snowball effect in which one set of attitudinal changes brings about behavioral change, which then reinforces the original attitudinal shift. My own position is that declining male relative income was a significant factor in accelerating the snowball.

The Marriage Squeeze Hypothesis: Too Many Women?

There is another theory, in addition to the relative income hypothesis, that links cohort size and marriage rates: the “marriage squeeze” (p.155) hypothesis.11 Because women tend to look for marriage partners among males approximately two years older than themselves, periods of mismatch between any given cohort and the cohort two years older—such as those that occurred during the baby boom—will affect marriage rates, due to imbalances between the demand for and the supply of mates. This suggests that cohorts born during the first half of a baby boom will exhibit declining marriage rates for females, while cohorts born during the latter half of such a boom will exhibit increasing marriage rates for females, because of this tendency for women to marry older men. In other words, young women will be “squeezed out” of the marriage market when there is a relative shortage of unmarried men about two years older than themselves—and this type of shortage would occur on the leading edge of any baby boom.

It is not entirely clear, however, how this variable is expected to affect male marriage rates: arguments of symmetry would imply that the marriage rates of young men should move inversely to those of young women in the same birth cohort, since a period of abundant mates for women will coincide with a period of scarcity for young men, and vice versa. But here again, symmetry may not hold. The “too many women” hypothesis suggests that young men will not feel compelled to marry when there is an abundance of potential mates: they need not commit themselves to any one woman.12 In this case male and female marriage rates would be expected to move together, declining (increasing) in periods of increasing (declining) relative cohort size.

Testing the Various Theories on Marriage and Divorce

How can competing explanations of marriage and divorce trends be tested? To what extent, if any, are marriage and divorce rates a function of relative cohort size (whether through male relative income or through the “marriage squeeze”) or of absolute incomes, both male and female?13 To what extent are economic variables important? Of these potential factors, the relative income hypothesis is the most difficult to test because of the need to compare the earnings of young adults with the income of their parents, a task hampered by the lack of appropriate data.

I have made two attempts to get around this data limitation, one using aggregate national-level data and the other using more disaggregated (p.156) regional data.14 In both cases the March Current Population Survey data have been invaluable. These data on hundreds of thousands of individuals over a forty-year period can be aggregated up to any level—whether a neighborhood, county, city, state, region, or nation—and used to “simulate” parent-child groupings by comparing the average earnings of young men in an area with the average income of older families in the same area. Using this approximation makes it possible to test for a relationship at the national level between economic variables (young men's earnings, women's wages, and older families' income), as well as the “marriage squeeze” variable, and annual observations of age-specific marriage rates provided by Vital Statistics for the years through 1990 in a simple time-series analysis.

In addition, by using CPS data on proportions married and divorced, the analysis can be extended down to the regional level for a much larger and richer time series of data cross-sections.15 The overall trends across regions are similar, as was demonstrated in figure 4.2, but there are significant regional differences as well, which permit testing to determine whether the apparent relationship between male relative income and marriage/divorce patterns is statistically significant. Specifically, I have identified subgroups of young adults by year, race, region, educational attainment, and years out of school16 and tested to see if there is a relationship between proportions married and divorced in each subgroup and the following economic and non-economic variables:17

  1. 1. the average annual earnings of young men in each subgroup by year, race, region, educational attainment, and years out of school;

  2. 2. the average starting hourly wage of young women (that is, the average wage in the first full year of work experience) in each subgroup by year, race, region, and educational attainment;18

  3. 3. the average income of families with children with head of either sex aged 45–54, by year, race, and region;

  4. 4. the “marriage squeeze” variable; and

  5. 5. a simple time trend.

In both sets of analyses I have found a strong and highly significant effect of male relative income on marriage, as hypothesized. That is, young men's absolute earnings exert a strong positive effect on marriage for both young men and women, as hypothesized by (p.157) other researchers, but an additional and much stronger effect—of the opposite sign—is exerted by older families' income. Young men's earnings were not found to be significant for divorce, but once again the effect of parental income was estimated to be very large and significant (positive). This finding in my analyses is consistent with recent survey data. In 1999 the New York Times and CBS News conducted a survey in which “1,038 young people aged 13–17 were asked to compare their lives with their parents' experience when growing up”. There was a definite effect of parental income, with the more affluent teenagers “whatever their race or gender, substantially more likely than those from more modest homes to report that their lives were harder and subject to more stress.” The comparative percentages were 50(38) reporting their lives as harder, 28(41) easier and 21(18) about the same among those with household incomes of $75,000+(〈$30,000).19

In disaggregated analyses I have found that for both men and women, a 10 percent change in older families' income results in a 5 percent change in both divorce and marriage rates among younger adults. The effect on divorce for men and women six to ten years out of school is even stronger: a 10 percent change in parental income would produce an 11 to 12 percent increase in divorce.20 These disaggregated results support my additional finding that at the national level a relative income model explains 99 percent of variation over time in marriage rates among young men 20–24 years old and even explains 75 percent of the year-to-year change at that level.

I have found in the disaggregated analysis that consistently, in regions, periods, and racial subgroups in which older families' incomes are high, marriage rates tend to be lower and divorce rates tend to be higher among young adults—even after controlling for other factors such as education and race—suggesting that these young people's material aspirations are indeed a function of older families' income. And, consistent with the hypothesis, this effect of older families' income tends to be strongest among the youngest adults (those less than seven years out of school), gradually weakening with age.

The female wage appears to play a less consistent role in these disaggregated analyses. It has a less certain effect than that of the relative income variables: negative and significant in most cases on young women's propensity to divorce, and positive and significant in most cases on their propensity to marry.21 But its effect on male (p.158) marriage and divorce propensities is significant in only one case, for young men's divorce propensity in the first five years out of school. There it exerts a significant negative effect.

Interestingly, however, these estimated positive effects of women's earnings on their own propensity to marry contradict a popular belief that women's increasing financial independence has been the cause of declining marriage rates. One researcher, however, has argued strenuously that women's financial independence is not the cause, but that the real culprit has been declining male earnings and employment prospects.22 These results support her contention, but add to young men's own characteristics the very significant negative effect of parental income on their children's marital propensities.

There remains, however, even after controlling for these variables, a very significant time trend: negative in all cases for men's and women's propensity to marry. This might be the result of a reinforcing “snowball” effect mentioned earlier, but it might also reflect changes in societal attitudes not captured by the economic variables. The models also estimate a significant negative time trend with respect to the propensity of younger men and women to divorce (that is, men and women less than ten years out of school). This could be a feedback effect of declining marriage rates: a self-selection process weeds out those who might have had a higher propensity to divorce.

The estimated effect of the “marriage squeeze” variable in the disaggregated analysis is paradoxical, however. It is highly significant, but its estimated effect on marriage rates is negative, rather than positive: an abundance of marital partners for women appears to decrease women's chances of marrying, rather than increase them. This result contradicts the expectations of the marriage squeeze hypothesis and seems somewhat nonsensical until one realizes that it simply means marriage rates continued to decline even for baby boom cohorts born after the peak. If the variable isn't measuring a marriage squeeze effect, it might be acting as a proxy for some type of “snowball” or “cascade” effect of relative cohort size on marriage propensities, as younger cohorts respond to the negative marital experiences of cohorts preceding them. This hypothesis is supported by the fact that the marriage squeeze variable loses its significance in analyses at the national level, after controlling either for a time trend or for male relative income. Many researchers have begun to suggest that a negative time trend of this type could well be attributable to the rising acceptability and incidence of cohabitation, (p.159) a natural outcome of a “snowball” effect in terms of changing social attitudes and mores. Young adults still value many of the benefits associated with marriage, it would appear, but given the economic uncertainties associated with low male relative income, they haven't been prepared to make as firm a commitment as in the past. One researcher has concluded, “The increase in the proportion of unmarried young people should not be interpreted as an increase in “singlehood” as traditionally regarded: young people are setting up housekeeping with partners of the opposite sex at almost as early an age as they did before marriage rates declined…. The picture that is emerging is that cohabitation is very much a family status, but one in which levels of certainty about the relationship are lower than in marriage.”23


Statistics presented in this chapter suggest that the media's marriage obituaries may have been premature, if not altogether off track. Those obituaries were encouraged by the virtual embargo on official age-specific marriage and divorce statistics that has existed since the late 1980s because of federal budgetary restrictions. Since 1990 we have seen only “crude” marriage and divorce rates (rates per 1,000 total population) from official sources, which are severely distorted by changing age composition in the population. Because the baby boom during the 1990s was moving out of marriageable ages and into divorce-prone ones, it seemed like marriage rates were continuing their inexorable decline, and divorce rates their worrying increases.

Age-specific rates presented in this chapter demonstrate that trends appear to have turned around among young adults: they're beginning to marry at earlier ages and at higher rates, and their divorce rates—which in any case hadn't increased in this group since about 1975—are on the decline. And the time trends in this behavior are remarkably similar to time trends in male relative income, suggesting that young adults were significantly affected by declines in this measure and scrambling to maintain their desired living standards by delaying family formation.

Statistical analyses confirm this relationship that seemed so apparent visually. As suspected by many researchers, declining male absolute earnings did play a role in the marriage decline—but this (p.160) was a contingent role. The prime mover was parental income: young people were significantly affected by the rising standard of living in their parents' homes and felt compelled to establish a standard in their own households at commensurate levels. This tendency was so strong that for every 10 percent increase in the income of older families, marriage rates declined by 5 percent among younger adults—and divorce rates increased by about 5 percent. But since 1985 young people's earnings relative to those of their parents have begun to rise for the first time since the 1960s, and marriage rates appear to be following. It seems that marriage rates declined not so much because marriage was becoming unpopular as an institution, but rather because it began to seem increasingly out of reach for young adults.

But, as in so many other cases, male relative income was probably not the only factor operational during this period. A significant negative time trend in the statistical analyses suggests either there are other—presumably non-economic—factors at work here, or else declining marriage rates have created over time a “cascade” or “snowball” effect, with younger baby boom cohorts responding not only to their own relative economic circumstances but also to the lifestyle example set by older baby boomers. If this latter is the case, however, it demonstrates even more the power of male relative income: it implies that recent economic changes have been influential even in the face of strong antimarital cultural biases that have developed against marriage as a result of baby boomer experience.

One last significant finding of the analyses reported in this chapter is that young women's increasing financial independence has in fact played a positive rather than the often expected negative role in its effect on marriage rates, after accounting for the effects of male relative income. Rising women's wages appear to have encouraged marriage and discouraged divorce among young working women, but their effect was swamped by the strong discouragement of declining male relative income (and, presumably, other non-economic factors).


(1.) The work presented here is described in more detail in Macunovich 2002.

(2.) No, that's not a typo! “Posslq” (pronounced “possle-q”) is a term coined by Statistics Canada in its 1981 census to describe unmarried “Persons of the Opposite Sex Sharing Living Quarters.”

(3.) The proportion ever married is used here, rather than the proportion currently married, to allow for the fact that individuals can move out of the married state, as well as in—through divorce, separation, or death of a partner.

(4.) Some of these age-specific rates calculated from the March CPS are presented in Macunovich 2002.

(5.) Cherlin 1992.

(6.) MacDonald and Rindfuss 1981, 123.

(7.) Ibid., 131.

(8.) Schoen, Urton, Woodrow, and Baj 1985.

(9.) Lestaeghe and Surkyn 1988. Their position is presented primarily in terms of fertility, but is extended to the full range of family change.

(10.) These positions are nicely summarized in Preston 1986 and Westoff 1986.

(11.) Glick, Beresford, and Heer 1963.

(12.) Guttentag and Secord 1983.

(13.) For theories regarding the impact of absolute male earnings and the female wage, see Oppenheimer, Kalmijn, and Lew 1993; Oppenheimer, Blossfield, and Wackerow 1995; Oppenheimer, Kalmijn, and Lim 1997.

(14.) The results described here are presented in more detail in Macunovich 2002.

(15.) The regions used in the analyses described here are twenty-one state groupings that can be identified consistently across the entire period covered by the CPS.

(16.) The data set used in the analyses contained more than 23,000 (13,000) observations of young men (women) one to five years out of school and more than 76,000 (48,000) in their first fifteen years out.

(17.) These econometric analyses also contained full sets of controls for education, race, years out of school, and region, as well as a time trend.

(18.) In “matching up” young men's and women's characteristics, a two-year age difference was assumed between men and women. That is, the characteristics of young women x years out of school were paired with those of young men x + 2 years out of school.

(19.) T. Lewin, “Next to Mom and Dad: It's a Hard Life (or Not),” New York Times, November 17, 1999.

(20.) The disaggregated marriage model (dependent variable: proportion ever married) was first estimated for unenrolled full-time male workers one to fifteen years out of school, where the t-statistic on relative income (F-statistic for the full regression) was 27.3 (515.14 [23, 113]) for young men in the first five years out of school, rising to 35.4 (748.9 [34, 648]) for those one to seven years out. The corresponding t (F) statistics for “parental” income when the numerator and denominator of relative income were entered separately were −5.4 (485.7) and −5.4 (699.52). These results are presented and described more fully in Macunovich 2002.

(21.) This finding of a positive effect of the female wage on marriage propensities is consistent with findings in a large number of microlevel analyses, such as Cherlin 1980; Goldscheider and Waite 1986; Mare and Winship 1991; McLoughlin and Lichter 1993; Oppenheimer, Blossfield and Wackerow 1995; and Teachman, Polonko, and Leigh 1987.

(22.) Oppenheimer, Kalmijn, and Lew 1993; Oppenheimer, Blossfield, and Wackerow 1995; Oppenheimer, Kalmijn, and Lim 1997.

(23.) Bumpass and Sweet 1991, 913.