This third and final section of the course brings us back to the industrialized world. As before, we consider how demographic and economic concerns are interrelated. We begin with an overview of the fluctuations of fertility that have taken place during the past 60-70 years in the U.S., and some of the important consequences of these fluctuations. We will then move on to consider the possible long-term economic consequences of below-replacement fertility. The concluding section of the course will focus on internal population change in the U.S., largely reflecting internal migration, and on issues surrounding immigration.
A. Fluctuating Fertility: The Baby Boom and the Baby Bust
Booms, busts, and echoes
In examining fertility behavior in the U.S., it is useful to consider both the total fertility rate and the absolute number of births. These two measures will not always move together, since population growth over time and changes in the age composition of the population mean that declines in the total fertility rate can be accompanied by increases in the absolute number of births.
Figure III.A.1 shows these two variables going back to the period just prior to World War II for the TFR and to 1915 for the total number of births. As can be seen clearly from the figure, we have experienced wide swings in fertility in the U.S. over the past 60+ years.
The total fertility rate fell sharply during the Great Depression of the 1930s, and then began rising during World War II. After the War the TFR rose dramatically, averaging almost 3 during the late 1940s and peaking at nearly 3.7 in the late 1950s. Still almost 3.5 in the early 1960s, the TFR began declining sharply -- to below 3.0 in 1965, to about 2.5 (and temporarily holding steady) in the late 1960s, and then down to about 1.8 by the mid-1970s.
Hence, the TFR fell by almost half between the early 1960s and the mid-1970s. After a decade of stability at a level of about 1.8, the total fertility rate rose slowly after 1986, reaching 2.08 in 1990. It presently stands at a little over 2, just slightly below the replacement level of 2.11.
Consider now what has happened to the total number of births. From an annual figure of nearly three million births in the 1920s, Depression-induced postponements resulted in well under 2.5 million births by the mid-1930s. By the end of World War II the level of births was close to its mid-1920s level. After the War there was an extremely sharp and sustained increase in the number of births to a maximum of 4.3 million in 1957.
From this peak, births fell slowly in the early 1960s and then sharply in the mid-1960s, rose slightly at the end of the decade, and then continued their decline to a low of about 3.1 million per year in the mid-1970s. From the mid-1970s to the mid-1980s, the total number of births was again on the rise -- despite a fairly stable total fertility rate -- reflecting the "echo effect" from the large numbers of births following World War II.
By 1989, the number of births surpassed 4 million, the first time this level had been reached since 1964. The number of births peaked in 1990 at almost 4.2 million, and has since declined slightly to about 3.9 million.
Overall, then, five distinct periods can be identified over the past 60+ years with respect to fertility in the U.S.: the Depression era (1930s) with its low TFR and small birth cohorts; the post-War Baby Boom (1946-64), during which the TFR increased by almost 50 percent; the Baby Bust, from the mid-1960s to the mid-1970s, characterized by a sharply falling TFR and declining numbers of births; an "echo phase" from the mid-1970s to the mid-1980s, with a largely stable and below-replacement TFR accompanied by increasing numbers of births due to a greater portion of the population being of childbearing age; and a recent period of below-replacement fertility with some modest variation, with the number of births and the TFR varying together.
These fluctuations in fertility have had and will continue to have important economic consequences. These effects stem largely from life-cycle considerations. That is, to the extent that a particular activity is especially prevalent at a certain life-cycle stage or age range, it should be evident that the frequency or importance of the activity will wax and wane with the passage through life of successive birth cohorts that differ sharply in size (cf., crime and youth).
What's being described here may be called demographic or compositional effects. The source of these effects may be seen most vividly by examining age pyramids showing the progression of the Baby Boom (and later, the Baby Bust) through the life cycle (Figure III.A.2). This is a bit like watching a python that has eaten a piglet -- the bulge moves slowly but steadily through the system.
The boom, bust, and echo consequences of the fluctuations in fertility have been and are evident in a number of places, but I want to focus on three areas: schooling, labor market entry (and the related issue of cohort size and economic prospects), and social security.
Consequences for the education industry
Beginning in the early/mid-1950s, the Baby Boom began contributing to substantial increases in the demand for primary school education, and hence for primary schools and primary school teachers. New college graduates at that time, however, were from the small birth cohorts of the 1930s, and were unable to meet the growing demand. By the late 1950s and into the 1960s this same pattern was being repeated at the secondary school level.
As a consequence of this rapid growth in demand, then, there were persistent shortages in the market for school teachers in the late 1950s and 1960s. These shortages put upward pressure on teachers' salaries, contributed to a lowering of hiring standards, and resulted in very low unemployment of qualified teachers.
By the end of the 1960s primary school enrollments had leveled off, and they subsequently began declining, reflecting the Baby Bust. This same pattern emerged at the secondary school level by the mid-1970s. As the demand for school teachers was stabilizing, however, the supply -- representing college graduates from the first half of the Baby Boom -- continued growing (reaction to changing market conditions tends to be slow). What had been a booming labor market turned around quite quickly, with predictable consequences.
The dramatic fluctuations in the market for school teachers were repeated at the college level as well, although there is somewhat greater flexibility here due to lower enrollment rates. As the core demographic pool for college enrollments (i.e., those aged 18-22) began to diminish in the early 1980s, colleges responded by more aggressively seeking enrollments from nontraditional students.
At present, we are witnessing a small-scale repeat of what took place 40 years ago. That is, the growth in the number of births, particularly since 1985 (reflecting both the echo effects of the Baby Boom and the uptick in the TFR during the late 1980s), is having an impact on primary school enrollments and hence on the demand for schools and teachers (cf., WSJ articles: "Next population bulge shows its might," "Some communities are caught unprepared," crowding in NYC schools last fall).
In all of these cases, then, the sharp swings in fertility of the past 60+ years have resulted in corresponding fluctuations in the demand for teachers. In the face of institutional rigidities like tenure, these swings in the labor market for educators have resulted in a variety of problems. In addition, there have been dramatic changes in the proportions of college students opting for careers in education, but (as noted above) often these changes tend to lag behind the changes in demand.
Consequences for the labor market
The fluctuations in fertility have also reshaped the labor force on a broader scale. The labor force may be viewed as a population, with "births" (labor force entry or reentry), "deaths" (labor force withdrawal, retirement, or mortality), and migration (mobility between jobs or between employment and unemployment). (Cf., "life expectancy" -- expected lifetime duration of employment.)
Labor force growth will reflect these "births" and "deaths." Among males, for whom labor force participation is fairly continuous from the end of schooling until retirement, labor force growth will consist principally of new entrants minus the sum of retirements plus mortality.
Among females, labor force participation (LFP) is not nearly so continuous over the life cycle as for males, although we saw earlier that periods of withdrawal from the labor force for childbearing and child care appear to have shortened substantially over the past 30-40 years (i.e., women's LFP is looking more like men's).
It's also true that an important (but declining) proportion of women remain more or less permanently out of the labor force. In any case, in considering the effects of fluctuating fertility on the female labor force, we need to keep in mind the secular trend toward higher LFP and more continuous work attachment.
Fertility will have a direct effect on the number of labor force entrants, with a lag of roughly 20 years (18-22, reflecting high school and college completion). The number of retirements will be affected as well, with a lag of 55-65 years. Hence, the fluctuations in fertility over the past 60+ years have influenced the rate of growth of the labor force and -- even more importantly -- the age composition of the working population.
During the 1950s, for example, new entrants to the labor force (apart from adult women entering the work force for the first time) were drawn from the small birth cohorts of the Depression. This exaggerated the increase in the average age of the working population that had been taking place due to the secular decline in fertility. The smaller numbers of young workers resulted in lower rates of job mobility, and this demographic compositional change was misinterpreted by some labor economists to represent a behavioral change (the industrial feudalism hypothesis).
By the late 1960s the large baby boom cohorts were spilling out into the labor market, reducing the average age of the work force. The strong labor market of the late 1960s (including a substantial military sector) absorbed these youth, but there was concern about the employment and earnings prospects for youth in the 1970s. Would there be sufficient numbers of new jobs created to accommodate a rapidly expanding labor force swollen by entrants from the huge birth cohorts of the 1950s?
[Note further that the changing demographic composition of the labor force resulted in an increase in the natural rate of unemployment during the 1970s (cf., natural rate concept: absence of cyclical unemployment, and hence presence only of frictional and structural unemployment; demographic aspect pertains to frictional unemployment -- associated with movement between jobs and movement into the labor force -- and hence is but one part of the natural rate).]
An important point to note here in considering whether the labor market would be able to cope with the influx of youth is the role played by wage flexibility. The economist's classical answer to the question above is that if the real wage rate is allowed to decline sufficiently, new jobs will be created and job growth will be able to accommodate the surge in new entrants to the labor force.
The changing age composition of the work force associated with labor market entry of the Baby Boom cohorts can be seen by considering the growth that took place between 1966 and 1976. During this 10-year span the total labor force grew by 23 percent, while the number of 20-24 year olds in the work force grew by 52 percent and the number of 25-34 year old labor force participants grew by 58 percent.
In 1967, the percentage of the male labor force consisting of men in their first five years out of school was 14.5, and by 1976 this figure had jumped to 22.6 percent. Overall, the ratio of workers under age 35 to those 35 and over grew from .46 in 1966 to .67 in 1976.
Unemployment rates of young workers increased during the 1970s, but so did those of older workers (Figure III.A.3). However, there was a clear deterioration in the relative unemployment of young workers. In addition, the relative earnings position of younger workers also worsened. In the words of Jim Smith and Finis Welch, this was "no time to be young."
Smith and Welch were among the first observers to emphasize the role of cohort size as a factor influencing the relative earnings of young workers (i.e., the earnings of younger, inexperienced workers relative to older, experienced workers). In considering the effect of cohort size on earnings, a key aspect is the degree of substitutability among workers.
In particular, if younger workers and older workers were perfect substitutes, then cohort size would be irrelevant to relative earnings, which would simply reflect the rate of substitution between the two groups. While the wage structure (relative earnings) would be unaffected by cohort size, if the two groups were perfect substitutes then large cohorts would result in lower wages for all workers.
However, if younger and older workers are not perfect substitutes, then the marginal productivity and wages of each kind of worker will depend on the relative sizes of the two groups. Smith and Welch discuss this in terms of a simple career-phases model, with apprentices (inexperienced young workers) and professionals (experienced older workers). When there is an increase in the number of apprentices working under each professional, the professionals become more productive; but since each apprentice has less of the time of the professional, the apprentices tend to be less productive than prior to the increase in their number.
In brief, then (reflecting the law of diminishing marginal returns and the need to equalize marginal productivity relative to price for each factor of production in order to minimize the costs of production), the more new workers there are relative to experienced workers, the lower the relative wage of new workers (see Figure III.A.4). Hence, allowing for the fact that new and experienced workers are imperfect substitutes provides a basis for expecting cohort size to influence relative earnings.
Early analyses of Current Population Survey (CPS) earnings data by Welch for white males aged 14-65, covering each year from 1967 through 1975, yielded interesting results on the effects of cohort size. He estimated separate equations for different schooling groups, with the key aspects of each equation (for our purposes) having the following form:
ln Y = a0 + a1 (Ni/NT) + a2 EXPER + a3 [(Ni/NT) * EXPER] + ... ,
where ln Y = natural logarithm of annual earnings, Ni/NT = the proportion of the work force accounted for by the ith cohort of workers (i.e., this is the relative cohort size variable), and EXPER = years of labor market experience.
Welch found a1 < 0, while a2 and a3 were > 0. The effects on earnings of relative cohort size and work experience are given by the partial derivatives:
partial ln Y with respect to (Ni/NT) = a1 + a3 EXPER; and
partial ln Y with respect to EXPER = a2 + a3 (Ni/NT).
Hence, when a cohort enters the labor market (EXPER=0), its size adversely affects earnings (since a1 < 0). However, over time the negative impact is eroded, in that the positive payoff to work experience (reflecting a2 > 0) tends to be greater for larger cohorts (since a3 > 0). Welch's results implied that the adverse consequences of large cohort size would wash out after about 10 years (cf., career-phases model).
A related analysis by Richard Freeman ("The Effect of Demographic Factors on Age-Earnings Profiles") concludes: "The principal finding is that the age-earnings profile of male workers, which has traditionally been viewed as a stable economic relation determined by human capital investment decisions, appears to be significantly influenced by the age composition of the work force.
"Apparently because younger and older workers are imperfect substitutes in production, changes in the number of young male workers relative to older male workers substantially influence the ratio of earnings of younger men to the earnings of older men. The effect of changes in the relative numbers of workers of different ages on age-earnings profiles is especially marked among college graduates. (Cf., Murphy et al.'s Figure 2.2 and related data.)
"By contrast, the age-earnings profile of female workers, which tends to be quite flat, appears to be little influenced by the age composition of the female workforce, possibly because the intermittent work experience of women makes younger women and older women closer substitutes in production."
In effect, then, these studies suggest that the labor market suffered from a short-run "indigestion" in the 1970s, associated with the entry of large Baby Boom cohorts. Presumably, this problem should have been relieved as we moved through the 1980s and 1990s.
Data from Murphy et al. (to be presented in class) provide a hint of the beginning of this phenomenon with respect to earnings. Further, as can be seen in Figure III.A.3, the relative unemployment of young workers since 1980 has improved, and -- in contrast to earlier experience -- improved rather than deteriorated during the most recent recession in the early 1990s.
Indeed, at present, with labor market entry coming from the small birth cohorts of the mid/late 1970s, the arguments above regarding the impact of cohort size would suggest that this is a very good time to be young (having a strong national economy with a low unemployment rate is extremely helpful in this regard; cf., demographic contribution to recent lowering of the natural rate of unemployment). Further, the substantial growth in female labor force attachment in recent decades suggests that the cohort-size arguments may well be more relevant to women's earnings now than they were in the past.
Consequences for social security
The fluctuations in fertility will soon begin to have a profound impact on the population aged 60 and over. Due to the historical trend of declining fertility coupled with falling mortality, the proportion of the population that is older had been growing anyway. However, the survivors of the increasing birth cohorts beginning even prior to the Baby Boom will soon reach age 60, and in less than a decade the leading edge of the Boom will cross this threshold.
If we focus on age 65 (the current age for receipt of full social security benefits), it is clear that the impact of the Baby Boom on the population this age and older will begin in 2011 and continue until almost 2030. At present, there are nearly 34 million people in the U.S. aged 65 and over, and by 2010 this number is projected (middle series of the Census Bureau's population projections) to increase to more than 39 million. Then will come two decades of rapid growth, with increases to 53+ million in 2020 and 69+ million in 2030, followed by a slowdown of growth of the elderly population to 75 million in 2040 and nearly 79 million in 2050.
Consider the aged dependency ratio, defined here as the number of individuals aged 65 and over per hundred people aged 18-64. This ratio is presently about 21, and will remain at that level until about 2010. Reflecting the entry of the Baby Boom into the aged category, however, Census Bureau projections indicate that the ratio will rise to 28 by 2020 and to almost 36 by 2030. It will remain at about 36 through 2040 and 2050.
This "graying" of the population has important implications for social security because of the fact that the social security system is one in which today's benefits paid out to retired recipients come from today's social security tax receipts, which in turn are derived from taxes levied on those currently working and their employers. This arrangement, described as a "pay-as-you-go" system, contrasts with the type of individual retirement accounts that some critics of the current system have recently been advocating, in which individuals invest money in personal accounts and collect returns in the future (cf., defined benefit vs. defined contribution pension plans).
The arithmetic of a pay-as-you-go system is straightforward. Using Gramlich's notation, let B = the average level of social security benefits, S = the total number of social security recipients, t = the social security tax rate (on both workers and employers), W = the average taxable wage (income), and N = the number of workers. The total in taxes taken in by the system each period equals (t * W * N), and the total in benefits paid out equals (B * S).
For the system to work on a pay-as-you-go basis, taxes must be enough to cover benefits: t * W * N = B * S. As Gramlich notes, a slight rearrangement makes it clear that with such a system, the balance between tax receipts and benefit expenditures is inherently linked to changes in the age composition of the population. That is, t = (B/W) * (S/N) -- the tax rate required to keep the system afloat must equal the product of the aggregate replacement rate (B/W) times the effective dependency ratio (S/N).
Hence, fluctuations in fertility like those we've seen in the U.S. render the system inherently unstable. When the Baby Boomers begin retiring, their social security benefits will (under the current system) be paid out of taxes levied on those from the Baby Bust and subsequent birth cohorts. The effective dependency ratio will rise, so maintaining the current benefit level (replacement rate) will require increases in social security taxes.
There are short-term influences that have had adverse impacts on the social security system. For example, legislated automatic cost-of-living adjustments (COLAs) were first passed in 1972, and the subsequent high inflation in the 1970s dramatically increased benefit payments. At the same time, the stagnant economy of the 1970s meant that social security taxes on earnings grew relatively slowly.
In 1983 several modifications designed to shore up the system were introduced, based on recommendations of a commission headed by Alan Greenspan, currently Chairman of the Board of Governors of the Fed. These included increasing tax rates and the ceiling on social security taxes (which has happened again several times), raising the social security retirement age from 65 to 67 beginning after 2000, taxing a portion of social security benefits, expanding coverage to include some government employees, and delaying cost-of-living adjustments. At present there is fairly widespread agreement among economists that the Consumer Price Index (on which social security COLAs are based) overstates inflation, but the political will to make downward adjustments in increases does not yet exist.
Gramlich reports current estimates that the payroll tax will have to be raised from its present level of 12 percent to 17 percent by 2030 and to 19 percent by 2070 in order to retain the pay-as-you-go system. Even without such increases, the "money's worth" ratio -- the ratio of the present discounted value of future benefits to the discounted taxes paid -- is expected to fall. That is, the rate of return to individuals will decline in the future, even with no changes. If taxes are raised to keep the system afloat, or if benefits are lowered, the money's worth ratio will decline even faster.
This is the delicate political economy issue that no politician has been willing to touch, then: keeping the system financially sound requires either raising taxes or reducing benefits, but doing so makes the system highly unpopular. The system would be unpopular not only for the younger cohorts emphasized by Gramlich, but also for current recipients and their advocates like AARP (cf., Americans for Intergenerational Equity).
As Gramlich notes, different approaches for addressing this political economy issue were considered by the social security advisory council that he recently chaired and that issued its report earlier this year. Much was made in the press of the recommendation that savings for old age be invested in the stock market.
More noteworthy is the need for raising national saving. This can be done, as in the past, by some combination of raising taxes, reducing benefits, taxing benefits, or gradually raising the retirement age. Alternatively, it can be accomplished by mandating private saving and creating individual retirement accounts.
In any case, it is clear that changes are required. Despite the problems, Gramlich ends on a fairly optimistic note: "It is already too late to stem the fall in money's worth ratios for workers born in the 1950s, but it is not too late to stem the fall for workers born in the 1960s or later."