- Claims of full employment are at present widely off the mark. Even without taking demographic shifts into consideration, current unemployment rates are still higher than they have been at the peaks of recent expansions.
- Once aging of the labor force is accounted for, the difference between current aggregate unemployment rates and those of previous expansions widens by as much as a full percentage point.
- Once declines in labor force participation by prime-age young adults are accounted for, underemployment can be seen to be rampant.
- To assuage this underemployment, US job growth would have to attain a rate of as much as 260,000 jobs per month for the next four years. Current job growth is nowhere near this rate.
- Western Asset’s expectation is that Fed rate hikes will continue to be halting and sporadic, thanks in part to the vast underemployment of the population.
Introduction and Summary
With aggregate unemployment below 5%, it is commonly asserted that the US is at or close to full employment. The analysis behind such assertions concedes that sharp declines in labor force participation rates are the primary factor allowing unemployment rates to decline during this expansion, but the declining participation rates are typically attributed to aging of the adult population, specifically the accession of Baby Boomers into post-55 age groups, where participation rates are much lower than for prime working age groups.
In this paper, we take issue with this assertion. While we concede that the greying of the Baby Boomer cohort has worked to lower aggregate participation rates, it is also the case that labor force participation rates among younger workers have dropped precipitously in recent years. This drop cannot be explained by retirement considerations, but seems to reflect a paucity of viable job opportunities.
The drop in participation rates among younger workers serves to distort aggregate unemployment rate data in two ways. First and most obviously, it has worked to reduce aggregate participation rates as much as has the greying of the Baby Boomers. Second, because unemployment rates are generally higher among younger workers than among older workers, pulling large numbers of younger adults out of the labor force serves to reduce the aggregate unemployment rate in addition to the effects on participation rates.
As we’ll see, while aggregate unemployment is close to the lows of recent cycles, unemployment rates within individual age groups remain quite high. An inordinately high share of the population in older, lower-unemployment-rate age groups is working to pull down reported aggregate unemployment.
Another new wrinkle of the current expansion is an increase in labor force participation among adults 55 and over. Baby Boomers are delaying retirement longer and more commonly than did their parents and older siblings. This almost surely reflects an inability to sustain a satisfactory lifestyle in retirement given historic saving patterns and concerns over the efficacy of Social Security, etc. If so, this phenomenon is unlikely to reverse, and it is a further argument against current claims of full employment. Higher participation rates among adults over 55 serve to further distort participation rates relative to accepted norms that are based on experience under different demographic conditions.
Those asserting the economy to be near full employment prominently cite the demographic factors pushing down aggregate participation rates, namely the greying of the Baby Boomers. However, there is no recognition of demographic factors working to “understate” aggregate unemployment rates, such as a widespread exodus from the labor force by younger adults or longer work-lives for older ones. Whether or not it is conceded that the exodus of younger adults from the workforce reflects slack in the economy, there is no controverting the fact that the exodus is contributing to diminished US living standards. Working-age adults abnormally eliminated from the workforce reduce per-capita GDP and, eventually, per-capita consumption.
Conservatively applying past 2006-peak participation and unemployment rates for individual age groups to present cohort data, we conclude that “under-employment” in the US currently totals about 5 million workers, 1.8% or more of the adult population and equivalent to about 3% of the current labor force. Allowing for permanently elevated participation rates among seniors and/or 2000 cycle-peak experience pushes this number yet higher. Note that we have not even broached the issue of involuntary part-time employment.
Fedspeak on the US economy has been ebullient, and Federal Reserve (Fed) Chair Yellen herself has recently asserted that the economy is at full employment. However, the Fed’s actual policy actions to date suggest an assessment of the economy closer to what we have outlined here. The following analysis supports our expectation that Fed rate hikes will continue to be halting and sporadic, thanks in part to the vast underemployment of the population.
Recent Unemployment Drop Not Driven by Job Gains
While unemployment has fallen sharply during this expansion, employment gains have been more modest. In fact, as seen in Exhibit 1, what we call the “employment rate,” known to Fed officials as the civilian employment-population ratio, has barely increased across the expansion, sitting at 59.7% in July 2016, from 59.4% at the recession trough in June 2009. To repeat, the share of the adult population with a job is barely higher now than it was seven years ago.
It is clear in Exhibit 1 that between 1980 and 2010, unemployment rates and employment rates moved closely (and inversely) together in both good times and bad. In recoveries prior to 2009, unemployment fell primarily because growth in employment vastly exceeded growth in the adult population, thus pushing the employment rate up about as much as the unemployment rate dropped. However, the two lines have diverged markedly throughout this expansion. (As we’ll discuss later, the disparate movement between these lines prior to 1980 reflects the accession of women into full participation in the labor force.)
The series plotted in Exhibit 1 are both ratios, but each sports a different denominator: unemployment divided by the labor force and employment divided by adult population. So, they will diverge when there are significant movements in labor force participation: the ratio of the labor force to the adult population. Indeed, again, participation rates have dropped sharply during this expansion, and Exhibit 1 reflects this.
As mentioned earlier, this decline in participation rates has typically been attributed to the accession of Baby Boomers to over-55 age groups, where participation rates are lower. While participation rates do decline as workers pass age 55, the greying of the very large Baby Boomer cohort explains only a fraction of the observed decline in the participation rate. Meanwhile, that greying also serves to pull down aggregate unemployment rates in a way that is not commonly perceived. To make this point, we need to analyze unemployment across age groups.
Unemployment Varies With Age. Always Has, Always Will
Just as participation rates vary with age, so do unemployment rates. Unemployment is very high among teenagers, and then declines gradually as age increases before finally bottoming out at about age 35. Exhibit 2 makes this point for both men and women individually and for both present experience as well as that of the April 2000 cycle-peak (cycle trough in unemployment).
And no, the phenomena displayed here are not specific to these points in time, but are instead typical of experience as far back as these data exist. They reflect the economics of matching prospective employee with prospective employer and the effects of age and experience on workers’ ability to “attract employment.” Since “normal” rates of unemployment unfailingly vary inversely with age and since aggregate unemployment is merely the weighted average of unemployment rates within the various age groups, the implication is that “full employment” unemployment rates will rise when the US adult population gets younger and fall when the population ages.
Baby Boomers are those born between 1947 and 1964. The massive number of births during this period make this cohort the perfect example of the effects of population age on aggregate unemployment. Observe the extremely high unemployment rates among adults 16 to 24, add 18 years onto the 1947 beginning of the Baby Boom, and you immediately understand why unemployment rates began a secular upturn in 1965, as shown in Exhibit 1. Add 18 years onto 1964, the end of the Baby Boom, and you immediately understand why aggregate unemployment began a secular decline in 1982. Observe that the effects of age on unemployment largely peter out by age 35, add 35 years to 1964, and you immediately understand why the secular decline in unemployment hit bottom in 1999.1 Sure, cyclical swings in unemployment occur as well, but the secular swings cited here show up clearly in the historical pattern.
Another demographic phenomenon worth mentioning is the accession of women into the labor force. From 1960 through about 1990, labor force participation rates for women rose steadily. (See Exhibit 5.) Early in that period, jobless rates for women were higher than those for men, but as women’s place in the labor force was fully established, women’s unemployment rates moved to where they were generally lower than those for men. (See Exhibit 3.) This gender-specific demographic shift also worked to raise unemployment rates in the 1960s and 1970s and to lower them in the 1980s and after.
The accession of women into the labor force over 1960–80 is the dominant factor raising aggregate employment rates then, even as aggregate unemployment was rising secularly (because of Baby Boomer effects). Once women were fully assimilated into the labor force by 1980, unemployment and employment rates began to cohere closely (and inversely), up until their divergence during the present expansion. This is why the unemployment rate and the employment rate lines diverge over 1947–80 in Exhibit 1 and why the lines converge over 1980–2010. Neither age- nor sex-demographics provides an obvious explanation for the divergence over 2010–2016.
Recent Demographic Shifts
So, with the youngest Baby Boomers having turned 35 in 1999, with women fully assimilated into the labor force since 1980, and with no other age cohort coming close to the “clout” of the Boomers, what demographic shift might have distorted aggregate unemployment rates recently? The answer is the sharp decline in participation rates by younger workers since 2000, especially since 2007. While young adults’ share of the population has not been affected by this shift, their share in the labor force has, and with younger age groups having higher unemployment rates, falling shares of young adults in the labor force have pulled down aggregate unemployment more than would otherwise have occurred.
One sees this clearly by looking at unemployment rates by age group over time. As stated earlier, aggregate unemployment is close to its cycle lows of 2000 and 2007. However, as can be seen in Exhibit 2, unemployment rates for both sexes and for all age groups under 75 are much higher now than they were in 2000. Exhibit 3 shows unemployment rates for all age groups over time. As seen there, current unemployment rates within the various age groups are generally higher now than at any preceding cycle trough for unemployment. Again, it is the further elevated share of older adults in the labor force that has pulled aggregate unemployment close to the rates seen at previous cycle peaks (unemployment troughs).
Exhibit 4 is an attempt to estimate the size of these effects. It shows unemployment rates by age and sex presently and at cycle peaks in 1969, 1979, 1989, 2000 and 2007. The aggregate unemployment rates in the second-to-last row are merely the weighted averages of these cohort rates, using the labor force weights in place at the time. The last column in the table shows current labor force weights for each cohort. When these current weights are used in conjunction with past cycle-peak unemployment rates, we get the last row of the table: full employment unemployment rates adjusted for the demographics shifts experienced since these points.
Adjusting for demographic differences increases the difference between current aggregate unemployment and those of 2000 and 2007 by 0.1 to 0.2 percentage points (pp). For earlier cycle peaks, the effects are larger, nearly a full pp for 1979. If we are going to judge current unemployment rates by the standards of the past, then we should adjust those “standards” downward by as much as a full pp because of changes in labor force age demographics. Again, current commentary frequently cites the importance of demographic shifts in reducing labor force participation rates. The equally important effects of demographic shifts on aggregate unemployment discussed here receive no such popular attention.
Participation Rate Changes
Furthermore, even adjusting full employment unemployment rates for demographic shifts fails to take account of the sharp declines in participation rates among younger adults. The charts in Exhibit 5 detail these declines. As seen there, 2000 was the peak year for labor markets and for participation rates. Participation rates for cohorts under 55 declined some over the course of the 2000–07 business cycle, and they have declined more markedly since the onset of recession in 2007.
By inspecting the scales of the various charts in Exhibit 5, you can see that participation rates for age groups over 55 are markedly lower than for those under 55, so, yes, the accession of Baby Boomers toward seniority has worked to reduce the aggregate participation rate. However, it is clear from the specifics of Exhibit 5 that the meaningful declines in age-specific participation rates have been among younger cohorts, and the effects of these declines on the aggregate participation rate (and aggregate unemployment rate) should not be ignored.
Employment Rates and Demographic Effects
We would argue that employment rates are the most meaningful measure of the job market. They encompass shifts in both unemployment rates and in participation rates. Also, they reflect the contribution of the population to GDP and, thus, to living standards. (Idleness is not necessarily bad, but it doesn’t put any food on the plate.)
Of course, as the amalgam of participation rates and unemployment rates,2 employment rates are just as affected by demographic shifts as are unemployment and participation rates. Earlier, we pointed out that employment rates are little changed now from their levels at the bottom of the recession, but we should examine the influence of demographic shifts on this comparison.
The charts in Exhibit 6 show employment rates for the same age/sex cohorts detailed earlier. As seen there, most cohort employment rates do show some improvement from June 2009 cycle lows (right border of grey area). (The exceptions to this are women aged 45–54 and 55–64.) So, we concede that the greying of Baby Boomers has worked to reduce aggregate employment rates and that labor market conditions have improved more since June 2009 more than is reflected in the aggregate employment rate.
Still, it is also clear from these charts—and the earlier ones—that within cohorts, employment statuses have deteriorated markedly on net from their positions at the cycle peak in 2007 and even further from those of the cycle peak of 2000. The sharp drops in employment rates for cohorts under 55 are quite at odds with assertions of full employment.
Current data suggest that younger adults have the “luxury” of dropping out of the labor force now and presumably waiting for better days to come. Those younger adults without such “luxury” are having more trouble finding work, as indicated by higher cohort unemployment rates for younger adults. Older workers, however, do not have such luxury and, faced with inadequate retirement savings, they have postponed retirement and extended their work-lives en masse. As much as employment rates have risen among over-55ers, unemployment rates there are also higher than in 2006/2000 (see Exhibit 3), suggesting that even for these cohorts, labor market recovery is less than complete.
So, let’s combine unemployment and participation rate demographic effects to get an estimate of where we stand presently relative to “full employment.” Exhibit 7 compares current cohort employment rates to those in place at cycle peaks (for employment rates) in April 2000 and December 2006.3 The second-to-last column shows current weights for each cohort in the total adult population.
The second-to-last row shows aggregate employment rates currently and at the 2000 and 2006 reference points. As seen there, the present aggregate employment rate is 5 pp lower than at cycle peaks 10 and 16 years ago (equivalent to 8% of the current labor force). Now, it is inappropriate to compare the current employment rate directly to those of 2006 or 2000. The population has aged too much. However, adjusting those past employment rates for observed demographic shifts still results in employment rates markedly higher than presently.
The last row of Exhibit 7 show aggregate employment rates when 2000 and 2006 cohort employment rates are weighted by current population demographics (second-to-last column). This calculation attempts to express 2000 and 2006 labor market conditions in a manner consistent with current demographics. Even after this adjustment, the aggregate employment rates of 2006 and 2000 remain 1.6% to 2.1% higher than present experience, equivalent to 2.6% to 3.3% of the current labor force.
In other words, the current employment rate of 59.7% reflects a substantially underemployed labor force. Raising the employment rate today to 61.3%—equivalent to market conditions in 2006—would require average job growth of 240,000 per month for the next four years. Raising it to 61.8%—equivalent to market conditions of 2000—would require average job growth of 265,000 per month.
Finally, note that these calculations assume ALL cohort employment rates return to previous cycle-peak levels. If participation rates and employment rates among over-55 groups are indeed permanently higher due to the need to postpone retirement, then the recent increases in employment rates among these cohorts are likely to be sustained. In that case, previous cycle-peak employment rates should be utilized only for under-55 cohorts, with recent employment rates utilized for cohorts aged 55-and-over.4 The last column posits this combination of cohort employment rates as a possible full employment norm. Given current population weights, that would imply an aggregate employment rate of 61.7%, 2 pp above current levels and requiring 260,000 jobs per month for the next four years in order to be attained.
Assertions that the US is at full employment typically make mention of demographic changes that buttress that argument, but they completely ignore demographic shifts that have the opposite impact. In this analysis, we have incorporated ALL observable demographic changes within the data. We have analyzed participation rates and unemployment rates across all age groups, both for men and women.
Allowing for the observable effects of age-demographics on labor force participation rates, we found that the differences between current unemployment rates and those in place at previous expansion peaks are understated by as much as 0.9 pp by the aging of the labor force that has occurred in recent years. Similarly, we found that while some of the decline in participation rates since 2007 is indeed due to an aging population, much of the observed decline is due to an exodus of below-35 adults from the labor force, an exodus that would seem to reflect a lack of job opportunities rather than demographics. Finally, we also detailed a sustained increase in recent years in participation rates among 55-and-older adults, a change that points to further possible downward potential for aggregate unemployment rates and upward potential for aggregate employment rates.
Together, these effects suggest that underemployment in the US economy at present amounts to between 4.1 million and 5.2 million workers, equivalent to 2.6% to 3.3% of the current labor force. To assuage this underemployment, US job growth would have to attain a rate of as much as 260,000 jobs per months for the next four years. (The peak one-year growth rate for the expansion to date is 251,000 per month in 2014.)
Now, only a small fraction of this shortfall can be attributed to higher rates of unemployment within cohorts. The bulk is due to the cessation of labor force participation by millions of adults aged 20 to 44. It is debatable whether this exodus indeed reflects “slack” in the labor market, as these individuals have voluntarily left the market to return to school, to subsist off government benefits or to live with parents and family. It is not debatable that the absence of these individuals from the labor force and from productive employment is a drag on domestic growth and living standards. And it is plausible that many or most of these disaffected workers would re-enter the labor force were job prospects to improve. It is certainly the case that the vast bulk of these potential workers are nowhere near retirement age.
So, there is ample reason to believe that claims of full employment are at present widely off the mark. Even without allowing for demographic shifts, current unemployment rates are a bit higher now than at the peaks of recent expansions. Upon allowing for the effects of observed demographic shifts, current unemployment rates are even further above previous experience. Once the observed declines in labor force participation by prime-age young adults are accounted for, underemployment can be seen to be rampant. Despite the rhetoric to the contrary, we believe Fed personnel are aware of this underemployment and that it will continue to be a factor retarding their rate-hike actions.
- Our 1947-64 definition of the Baby Boom coincides with the period when birth rates were actually extraordinarily high. Other time spans for the Boom period coincide with one writer’s definition or another, but they don’t jibe with the actual demographic experience.
- Take the unemployment rate, add one and then multiply that quantity by the participation rate, and you get the employment rate.
- And, no, neither of these past peaks would qualify as overheated job markets. The Employment Cost Index for private-sector labor costs rose a mere 3.1% in 2006 and 4.4% in 2000, while PCE deflator inflation was 2.1% in 2006 and 1.8% in 2000.
- We might even use yet higher employment rates for over-55 cohorts given that unemployment rates across these cohorts are generally still higher than what was observed in 2000 and 2006, as detailed in Exhibit 2 and Table 1.