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, 2018. All Rights Reserved. Under the terms of the licence agreement, an individual user may print out a PDF of a single chapter of a monograph in CHSO for personal use (for details see www.chicago.universitypressscholarship.com/page/privacy-policy). Subscriber: null; date: 19 January 2019

Effects of Relative Cohort Size on Inequality and the Overall Structure of Wages

Effects of Relative Cohort Size on Inequality and the Overall Structure of Wages

Chapter:
(p.103) 6 Effects of Relative Cohort Size on Inequality and the Overall Structure of Wages
Source:
Birth Quake
Author(s):

Diane J. Macunovich

Publisher:
University of Chicago Press
DOI:10.7208/chicago/9780226500928.003.0008

Abstract and Keywords

This chapter analyzes the wages at all age levels throughout the workforce. It tries to identify whether the age structure of the population has had a significant effect on the primary dimensions of inequality in the United States over the last thirty-five years: on the return to experience, the return to skill, hours and weeks worked, and on the general structure of wages and level of inequality. Changing relative cohort size can explain a significant proportion of the variation over time in measures of work intensity, such as hours and weeks worked and the proportion working full time, at all levels of experience. Cohort size effects—on wages, unemployment, and hours and weeks worked—have occurred not just among younger workers, but throughout the labor force. Furthermore, changing relative cohort size has been a major factor increasing all types of inequality in the society over the past few decades.

Keywords:   wages, inequality, United States, relative cohort size, work intensity, unemployment, labor force

The income of low-wage workers finally is rising and wage inequality has slightly declined. This surprising turn of events is forcing economists and policymakers to reopen the debate about why income distribution in America, which once prided itself on being a middle-class society, has evolved over the last twenty years into the most unequal in the industrialized world. … [E]vents of the past eighteen months have raised new doubts about the conventional wisdom. Workers on the bottom half of the wage ladder, including those with minimal skills, have seen real gains for the first time in a generation. If a lack of education and skills held them back before, what's causing their wages to rise now?

Merrill Goozner, “Gains by Low-Wage Workers Raise Eyebrows” (1998)

The previous chapter looked at the RCS:RY (relative cohort size:relative income) relationship among just the youngest workers, using summary measures of relative income and earnings at the aggregate level. It examined the average wage of all entry-level workers relative to that of all prime-age workers, given the size of the entry-level cohort relative to the size of the prime-age cohort for whom it is assumed to be least “substitutable.” This is the element of relative cohort size most relevant to the “second-order effects” that will be discussed in subsequent chapters: behavioral changes in things like marriage, divorce, fertility, juvenile crime rates, and the labor force participation of younger women.

It was assumed in the analysis in chapter 5 that if the RCS:RY relationship didn't show up at the aggregate (national) level, then even (p.104) if it exists at more disaggregated levels it must not be very important. But there's a downside to the more intuitive aggregate analysis in chapter 5: social scientists are often not satisfied with findings based on such aggregated data—especially when the data consist of annual observations just since 1960. For various technical reasons relationships that appear at this aggregate level might be “spurious” rather than real or causal. Findings are much more believable if they can also be demonstrated to exist among subgroups and individuals in more disaggregated data, where there is a great deal more variation in behavior. Thus a more stringent test of the RCS:RY relationship demonstrated in chapter 5 would be based on such disaggregated “microlevel” data.

In addition, the findings reported in chapter 5 have much wider implications. They raise questions about potential cohort size effects that might account for wage differences throughout the labor force, among workers at all levels of experience. Do cohort size effects follow workers throughout their career: do large cohorts never escape their fate? Or can that longer-term fate be affected by what else is happening in the age structure of the population? In other words, for example, is it better to be a large prime-age cohort when entry-level cohorts are small, or when they're large? And, given the huge variations in the size of entry-level cohorts that we've experienced over the past thirty-five years, is it possible that cohort size might account for the disturbing changes in income inequality that have also occurred?

It's ironic that if these cohort size effects do reverberate throughout the labor force, affecting workers at all levels of experience, then individuals actually influence their own fate in the labor market through their fertility decisions early in life: the child you have at age 25 becomes the new entry-level cohort when you're 45. If there are positive effects of the size of entry-level cohorts, an attempt to improve one's economic status when young by reducing or postponing fertility might backfire later in life.

Past Trends in Income Inequality

Although a search of the economics literature over the past fifteen years would turn up little in terms of “male relative income,” it would produce volumes on the more general issue of “rising income inequality.” Between 1968 and 1994, income inequality in the United States increased 22.4 percent—more than wiping out the 7.4 percent (p.105) improvement that had occurred in the 1950s and 1960s.1 Analysts tend to trace a large proportion of this change in family and household income inequality to increasing inequality in male earnings.2 While the average wage in the bottom quintile3 increased 20 to 30 percent relative to the top quintile in each of the three decades prior to 1970, it grew only 5 percent in relative terms in the 1970s and actually declined 8 percent in the 1980s. It is true that the rich have been getting richer and the poor, poorer.

This change is apparent in table 6.1, which shows that those in the eleventh to twentieth percentiles experienced a real wage gain of 31.5 percent between 1940 and 1950, far outstripping the gain of those in the eighty-first to ninetieth percentiles, who saw only a 9.1 percent increase. But that situation had more than reversed itself by the 1980–1990 period, when the lower wage earners lost 16.9 percent of their earnings while the wealthiest experienced a 1.1 percent gain. As a result, the United States enjoys the unenviable distinction of having the largest standard deviation4 of earnings in any developed country: in the late 1980s this was 0.774, as compared with a non-U.S. weighted average standard deviation of 0.480.5

There is little consensus among analysts regarding the cause of this dramatic deterioration in the relative (and absolute) position of lower-paid workers over the past twenty-five years. Early researchers identified the labor market entry of the post-WWII baby boom as an important factor: a dramatic increase in the supply of younger, less experienced workers that depressed their wages and generally reduced their employment prospects. In recent years researchers have focused on other factors such as sectoral shifts (“deindustrialization” and the growth of the service sector), technological change such as computerization, and globalization of the economy reflected in immigration and the trade deficit. They point out that what had

Table 6.1 Real Wage Growth Rate, by Wage Percentile, 1940–1990

Percentile

1940–1950

1950–1960

1960–1970

1970–1980

1980–1990

11–20

.278

.192

−.015

−.169

21–40

.277

.292

.207

.015

−.116

41–60

.301

.232

.073

−.072

61–80

.127

.302

.252

.096

−.024

81–90

. 091

.300

.284

.089

.011

1–100

.194

.297

.241

.050

−.078

(p.106) been an “increased return to experience” (i.e., the wages of older, more experienced workers rising relative to those of younger, less experienced workers) became in the 1980s an “increased return to skill” (a sharp rise in the wages of college graduates relative to those with less than a college education).

The purpose of the analysis reported in this chapter was to determine whether the age structure of the population—as measured by relative cohort size—has had a significant effect on the primary dimensions of inequality in the United States over the last thirty-five years: on the return to experience, the return to skill, hours and weeks worked, and on the general structure of wages and level of inequality. That analysis is reported with full technical detail elsewhere.6 Here we'll just look at the implications of its findings.

Comparing the Experiences of Lucky, Big Bob, and Little Larry

The data available to researchers indicate that there are significant differences among cohorts not just in their entry-level work experience, but throughout their working lifetimes. My analyses using those data have led me to believe that these lifelong differences have been largely the result of variations in cohort size and position, and that they depend not just on the size of one's own cohort, but also on the pattern of births (that is, changes in cohort size) throughout one's lifetime. In order to highlight the very different lifetime experiences apparent in the data, I'll personify the characteristic patterns exhibited in three cohorts using the stories of Lucky, Big Bob, and Little Larry.

“Lucky” is perhaps a misnomer for members of the first of these cohorts—born between 1935 and 1940—since they were born during the Depression and then spent their childhood in the throes of World War II. But they certainly lucked out in the labor market. Fertility rates were very low when Lucky was born, so his cohort was small relative to his parents', and he thus entered the labor market when starting wages were soaring relative to the average wage. Some estimates put his relative income at an extraordinary 1.0 or higher: that is, his starting salary in his early twenties matched his parents' total annual income just a few years earlier, when he was still living at home (and his parents were in their 40s or 50s).

Lucky benefited not just from an optimal relative cohort size: he was followed in the labor market by ever-larger cohorts—the product (p.107) of the 1950s baby boom. His cohort position was optimal. These larger cohorts in his wake generated an increasing demand for goods and services that contributed to growth in an economy already buoyed up by returning soldiers forming households. Thus Lucky enjoyed a rapid rate of increase in his already generous starting wage. Then, to top it off, in his prime years he found himself once again benefiting from small relative cohort size—but this time small relative to the size of entry-level cohorts, rather than relative to the size of his parents' cohort. As the wages of younger workers fell, his continued to rise. He decided to cash in on his favorable career and switch from his full-time job to a part-time one as he moved closer to retirement. And here again he benefited since entry-level cohorts were now very small, thus generating little competition for the less-demanding jobs that now attracted him.

Big Bob was born in 1955—went through school with Lucky's son, in fact. Schools were having trouble accommodating the growing numbers of students during that period, so most of his education occurred in overcrowded classrooms and porta-cabins, and often on half-day schedules. Predictably, then, he found himself nearly crowded out when he tried to enter the labor market. But he figured college wouldn't help him much, since his older friends who'd gone to college were having a tough time finding jobs. He was forced into part-time work, even as his father's wage was rising—so his relative income, only about 0.4, was very low in comparison with Lucky's: he earned only 40 percent of his parents' income. He decided he'd have to put off any hopes of starting his own family, and for a time he even continued living with his parents, to make ends meet. But he at least benefited from the fact that for a time the size of entry-level cohorts following him was still growing and generating new demand for goods and services in the economy, so he was eventually able to settle into a full-time job.

Then, just as Big Bob was beginning to feel more confident about his career, the bottom seemed to fall out of the market. The economy stumbled as the size of entry-level cohorts began to decline and the demand for goods and services no longer grew at expected rates: producers cut back on production as inventories grew. They began downsizing and Big Bob was laid off. His fortunes finally turned when he entered his 40s, however: the size of entry-level cohorts began to grow and the economy recovered its old dynamism.

Little Larry entered the world in 1970. He was part of a much (p.108) smaller cohort than Big Bob's, relative to that of his parents, but that didn't seem to help when he tried to strike out on his own twenty years later. The economy was still stumbling along as entry-level cohort size continued to decline and demand failed to increase at historic rates. Nearly 20 percent of his older friends were unemployed, so he figured he wouldn't lose much by taking time out for college. Things still looked pretty gloomy when he graduated, so he had to take a job at lower pay than he'd hoped for and work as a “temp.” His relative income was just 0.25: that is, he was earning only 25 percent of his parents' income. He decided to share a large apartment with several friends. But then his fortunes began to improve dramatically after 1995 as the size of entry-level cohorts began to increase once more, and producers cashed in on the spurt in demand.

Where will Little Larry's fortunes take him over the next two decades? My analyses suggest that at any given point in his lifetime the trajectory will be strongly affected by the pattern of births that occurred twenty-some years earlier. The point of these tales has been to emphasize that one's own cohort size and position are important not only at the point of labor market entry, but also later in life, and that the size of the cohorts following one into the labor market play a significant role, as well. It is too simplistic to expect one's lifetime wage profile to reflect only one's own cohort size and position. Both Lucky and Little Larry were members of smaller birth cohorts, but their lifetime experiences diverge because Lucky was followed by a surge in births while Little Larry was followed (initially) by a continued decline.

In order to control for this dynamic effect of changing cohort size, I used an augmented version of the model described in chapter 5 in the analysis underlying the results presented in this chapter. That is, while the augmented model is based on the same three factors used in chapter 5—relative cohort size and position (the GFR twenty-one years before an individual's labor market entry, and its rate of change); the level and rate of change in young men's share of the total military; and the level of durable goods imports and exports—it controls not just for each cohort's own size and position, but in any given year it also controls for the size of the current entrylevel cohort. I used this augmented model to try to explain movements over the last thirty-five years not just in the wages of the youngest workers, but in the entire wage structure: wages at all levels of education and experience.

(p.109) The data I analyzed with this augmented model focus specifically on the hourly wages of unenrolled white male, full-time, full-year workers, rather than on the relative income of all entry-level workers—a change aimed at making its results comparable with those of standard labor models. But related analyses I have conducted show that the results I found using the more restricted data set are very similar to those found for the total labor force, including non-white workers and those working less than full year, full time.

A Pictorial Summary of Model Results

The model's results can be examined in many different ways. It's possible to look at the absolute wages of various groups as well as their levels relative to each other—measures such as male relative wages and the college wage premium that were discussed in chapter 5. In addition, one can examine hours and weeks worked, together with what is termed “within cell variance”: the degree of variation that's observed among the wages of workers with the same level of experience and skill. Researchers have found that this within-cell variance accounts for the lion's share of the growth in inequality since 1980. In all of these dimensions, relative cohort size appears to account for most of the changes we've observed in the wage structure over the past thirty-five years.

The detailed results of the analysis are presented in the full study.7 As in the simple model in chapter 5, the results demonstrate a strong and highly significant negative effect of own RCS (birth cohort size), which is stronger for those born on the trailing edge (Little Larry) than on the leading edge (Big Bob) of the baby boom. In addition, there is a strong positive effect of the size of the current entrylevel cohort each year (because of its members' demand for goods and services as they set up their own households), which varies for an individual worker depending on his own level of experience as suggested in the tale of Lucky, Big Bob, and Little Larry. A large entry-level cohort competes with me for jobs when I'm a younger, inexperienced worker or an older worker approaching retirement, but complements me (effectively making me more productive) when I'm in my prime years. In addition, any demographically induced economic slowdown will hurt me more if I'm a more “expendable” worker with little experience.

The model described here—the model from chapter 5, augmented (p.110)

Effects of Relative Cohort Size on Inequality and the Overall Structure of Wages

Figure 6.1. Observed, simulated, and predicted wages of young white males working full time relative to males with 26–35 years of work experience. The high degree of correspondence between the observed and the simulated values suggests that the model explains the time trend of the data extremely well, although it uses just individuals' own cohort size and the size of the current entry-level cohort at each stage in those individuals' careers.

to control for current as well as birth cohort size—was fitted on data describing workers at all levels of education and experience: nearly 150,000 observations of individuals and their wages over a thirty-five year period.8 It's not possible to present here the many graphs I've prepared for all of these education and experience groups: instead I've just presented graphs showing the model's “fit” for the youngest age group, which are very representative of the model results as a whole.9

Figures 6.1 and 6.2 address the two basic concepts from chapter 5: male relative earnings and the college wage premium. These two measures trace the patterns of what statisticians term “between-cell” variance: variations among individuals' wages that can be explained by differences in experience and education.

Figure 6.1 presents the results for male relative hourly wages: the wages of younger relative to prime-age workers. Included in the figure are three different sets of values: those actually observed in the data; fitted, or “predicted,” values produced by the full model (including controls for own and current cohort size, the military, and trade); and what are termed “simulated” values, produced by the model when all variables other than own and current cohort size (p.111) measures are held constant. This model explains the time trend of the data extremely well, although it uses just individuals' own cohort size and the size of the current entry-level cohort at each stage in those individuals' careers (i.e., there is a high degree of correspondence between the observed and the “simulated” values).

Figure 6.2 presents the same type of information for the college wage premium. This graph shows that the military and trade variables add more explanatory power (i.e., the “fitted” curve matches the observed data better than the “simulated,” in which military and trade are held constant). The difference between the simulated and actual values of the college premium show that trade and the military depressed the college wage premium in the 1960s (because the actual value is lower than the simulated) and boosted it in the 1980s. But even here, changing relative cohort size explains most of the marked decline and then rise in the return to a college education over the last thirty years that's been the focus of so much concern among analysts (because the actual and simulated values still follow very similar patterns over time).

Figure 6.3 addresses “within-cell” variance—differences among

Effects of Relative Cohort Size on Inequality and the Overall Structure of Wages

Figure 6.2. Observed, simulated, and predicted college wage premium (ratio of average hourly wage of college grads relative to that of high school grads) for young white males working full time. The difference between the simulated and actual values of the college premium show that trade and the military depressed the college wage premium in the 1960s (because the actual value is lower than the simulated) and boosted it in the 1980s. But even here, changing relative cohort size explains most of the marked decline and then rise in the return to a college education over the last thirty years.

(p.112)
Effects of Relative Cohort Size on Inequality and the Overall Structure of Wages

Figure 6.3. Observed, simulated, and predicted variance of logged hourly wages within year-state-education-experience cells for white males working full time. Similarities between the simulated and predicted values indicate that the longer-term trends in variance were the result of changing age structure in the population.

wages earned by individuals with the same levels of education and experience that have proven very difficult to explain. The observed values in figure 6.3 show the substantial increase in within-cell variance observed in the labor force since about 1975: corresponding graphs in Macunovich 2002 demonstrate that this was a common trend across experience levels.10 This variance has been one of the major factors underlying the growing inequality in U.S. society in recent decades. Similarities between the simulated and predicted values in figure 6.3 indicate that the longer-term trends in variance were the result of changing age structure in the population.

But why—or how—would changing age structure increase within-cell wage variance? As indicated in chapter 5, between-cell variance is thought to be the result of “imperfect substitutability” between different categories of worker. Recent work by other researchers suggests that increased within-cell variance may be accounted for by variation in (1) the proportion working full time; (2) hours worked per week; and (3) weeks worked per year.11 When there is a supply glut, employers can choose among workers and will tend to favor those within any given category who demonstrate higher levels of ability and motivation—qualities not necessarily captured by simple measures like years of experience and education. My (p.113) analyses, reported elsewhere, demonstrate that changing relative cohort size can explain a significant proportion of the variation over time in measures of work intensity, such as hours and weeks worked and the proportion working full time, at all levels of experience.12

The model appears to explain well the long-term trends in wages—both relative and absolute—at all levels of education and experience: the strong increases of the 1960s and early 1970s as well as the marked declines experienced by younger and less skilled workers since 1973, with stable and sometimes even increasing wages among other groups.

Summary

The results presented in this chapter support and extend those in chapter 5, and in the two chapters together I have tried to demonstrate the “first-order” effects mentioned in the overview. Cohort size effects—on wages, unemployment, and hours and weeks worked—have occurred not just among younger workers, but throughout the labor force, providing a plausible explanation for a significant portion of the disturbing rise in income inequality observed in our society since 1980.

Younger workers were hurt relative to older workers, and the “return” to investing in a college education dipped and then soared as the baby boom began entering the labor force. Marginal workers (those with less education, experience, ability, or training) were initially carried on a rising tide of growing aggregate demand for goods and services as the baby boomers set up households, but those marginal workers were then the “first fired” when the wave ebbed and the economy stumbled. Those with less skill, motivation, or ability were often forced into part-time work, creating differences between their wages and those of other workers with similar levels of education and experience. But this was a cyclical effect related to the life cycle of the baby boom, rather than a gloomy indicator of what was to come in the twenty-first century. This argument is perhaps more easily accepted now after the economy's recovery in the mid-1990s, but in the early 1990s it was very difficult to convince many analysts of its merits.13

The analyses I have described in this chapter indicate that these effects are significant not just in aggregate data at the national level, where we have relatively few years of observation, but also using (p.114) observations on the wages of hundreds of thousands of individuals over a thirty-five-year period. Changing relative cohort size has been a major factor increasing all types of inequality in our society over the past few decades. My findings accord with recent work by other researchers using international data. Their results suggest that cohort size also accounts for a significant proportion of changes in inequality in countries around the world, at all stages of development.14

Not least among the sources of inequality here in the United States has been the growing gap between the wages of entry-level and prime-age workers that accounts for declining male relative income. Thus the first link has been forged in the chain of effects leading from cohort size to many of the societal changes that have perplexed us in the post-WWII period. The next chapters in part 2 will carry on the analysis, looking at connections between male relative income and many of those societal changes. In this chapter and in chapter 5, the “prime mover” in explanatory models—the most significant independent variable—has been relative cohort size, approximated using the general fertility rate twenty-one years earlier. In part 2 the focus shifts to the indirect, or “secondary” effects of relative cohort size. There, relative income becomes the primary independent variable.

Notes:

(1.) As measured by the Gini coefficient for family income, a statistic that indicates how far a society is from perfect equality. This figure is taken from Ryscavage 1995, 51.

(2.) About one-third, according to Karoly and Burtless 1995.

(3.) Income or earnings quintiles are groupings of individuals based on the level of their (annual) income or earnings. Those in the bottom quintile earn wages that put them among the 20 percent of the population with the lowest earnings. Table 6.1 presents data for such quintiles, except that it reports figures for those in the eleventh to twentieth and eighty-first to ninetieth percentiles—rather than the first to twentieth and eighty-first to one hundredth—because it is often felt that there are distortions in the estimates for those in the very top or bottom of the income distribution.

(4.) Standard deviation is an indicator of the amount of variation in a measure like earnings, observed among members of a group—the spread between those at the bottom and those at the top.

(5.) Blau and Kahn 1996, 806.

(6.) Macunovich 1999a.

(7.) Macunovich 1999a focuses on younger workers, and Macunovich 2002 extends the analysis to the full labor force.

(8.) These 150,000 observations represent over 650,000 individual observations, grouped by education, race, state, and experience. When the model was estimated on the fully disaggregated data, it explained over 27 percent of the variance in the data.

(9.) Similar graphs for other age groups, and by education level, are presented in Macunovich 1999a and 2002. These demonstrate the model's ability to explain wage movements throughout the labor force.

(10.) Although figures 6.1, 6.2, 6.3 illustrate the situation only for younger workers, Macunovich 2002 contains similar results for workers at all education and experience levels.

(11.) McCall (2000, 426) finds that “high rates of joblessness, immigration and casualization (part-time work, temporary work, and unincorporated self-employment) exert significant positive effects on the level of residual wage inequality within labor markets.”

(12.) These results are reported in Macunovich 1999a and 2002.

(13.) I say this from firsthand experience after trying—and failing—to convince some fellow members of the 1994–1995 Social Security Technical Panel that the economy could be expected to rebound again, despite its performance since the mid-1970s.

(14.) Higgins and Williamson 1999.