Persistently low U.S. labor force participation will likely be a drag on the economy for at least the next decade, continuing a trend that began before the Great Recession and making annual GDP growth of more than 3% on average difficult to achieve. Given this trend, S&P Global Ratings has lowered its estimate for long-term growth in the world's biggest economy to just 1.8%--a full percentage point below the roughly 2.8% we expected for potential growth just 10 years ago.
The historically low percentage of Americans in the workforce will likely also have ramifications for the Federal Reserve, given the central bank's mandate to balance inflation and employment. As policymakers factor in the demographic and cultural shifts affecting the labor force participation rate (LFRP), and perhaps reconsider what constitutes full employment, we expect benchmark interest rates to rise only slowly--and, ultimately, to a lower-than-traditional "neutral rate" (see "U.S. Long-Term Interest Rate Looks To Stay Low For Longer" published Nov. 3, 2016).
Growth in the labor force is one of two key determinants in a country's maximum sustainable, or "potential," economic growth. The other is labor productivity growth (see "The Strange Case Of Shrinking U.S. Productivity Growth: Myth, Mismeasurement, Or Multiyear Phase?", published May 5, 2016). Simply put, the combination of quantity and quality of labor propels output capacity in the long run. By extension, the determinants of labor force growth--changes in the LFPR and the working-age population--are important to economic expansion.
We calculate that declining LFPR has shaved 0.6 percentage points from annual economic growth in the current expansionary period, which is now in its ninth year and the third-longest since World War II. On the (sort of) bright side, S&P Global thinks this drag will moderate--albeit only slightly--in the next decade, as we see some room for a cyclical improvement in the participation rate in the next few quarters.
Much of the decline in the LFPR is due to demographic changes in the U.S.--specifically the aging of the American population and the changing age structure of the working-age population (defined as 16 and older, with no maximum age). In fact, we attribute three-fourths of the drop, from around 66% in 2007 to just 62.8% in the second quarter of this year, to the "pure demographic" effect of the changing age structure of the workforce. The other one-quarter translates into roughly 2 million "missing workers." (For more on this, see Appendix A.)
More specifically, the combination of the shift in the distribution of the population toward retirement age and the fact that older people tend to participate in the labor force at much lower rates than younger people have reduced overall the LFPR substantially in recent years (see chart 1).
To be sure, the retirement rate among workers 55 and older has fallen continuously since 1998--with some recent stabilization--but the increased participation rate is far from enough to carry the overall headline number (see chart 2). Retirement (which, along with disability, is broadly accepted as irreversible) remains the leading cause of nonparticipation since the most recent recession (see table 1).
Meanwhile, the disability rate has reversed somewhat in the prime age group (25-54), and this trend may continue as the current expansion matures. Disability rates among the youngest workers may reverse a bit, as well, as job prospects improve. Still, the overall LFPR has declined for young and prime-aged individuals, as members of successive groups that are replacing baby boomers show less attachment to the labor force.
Another main reason for leaving the labor force is school, and we've seen a structural shift since the late 1990s because of better accessibility of, and higher returns on, education. Students' propensity to work while in school has also declined. Enrollment among the young has stabilized at around 33% in the last few years, up from 25% in the mid-90s, and it's hard to imagine this trend will change--although the "working while in school" rate could improve as the labor market gets tighter. (1)
Nonparticipation by Reason
|As of Q3 2017||Cumulative differences relative to Q4 2017||Negative Outlook|
|(3.1.a)Other: don't want a job because in school||0.3||4.9|
|(3.a.b)Other: don't want a job and not in school||(0.2)||6.9|
|Source: Shigeru Fujita of Philadelphia Federal Reserve, available at www.phil.frb.org/research-and-data/economists/fujita/, updated as of Q3 2017.|
Through The Decades
Birth rates just after the turn of the millennium (combined with assumptions about mortality and net immigration rates) largely determine the working-age population now and in the coming decade. The cohort of Americans 16 and older is set to grow at an annual average of 0.9% from 2016-2026, according to the Bureau of Labor Statistics (BLS)--well below the 1.3% average of 1950-2016.(2) And this doesn't account for recent, or expected, changes in immigration policy.
U.S. labor force participation hovered from 58%-60% until the early 1970s, before increasing at a relatively rapid pace. It surpassed 66% by the end of the 1980s and continued climbing until 2000, when more than two in every three Americans 16 and older were either working or actively seeking work (see chart 3). That rate slipped only slightly until the end of 2008, when the decline accelerated. While the current rate (mid-year 2017) of about 62.8% is similar to the LFPR in the past three years, it's still more than 3 percentage points lower than just before the Great Recession. The change from the first quarter of 2000 to the second quarter of this year is more than 4 percentage points.
The evolution of the LFPR over the years has been impacted by changing age-cohort-structure of the economy. Movements in larger cohorts tend to propel the overall rates one way or the other, just because their shares in the weighted average calculations are higher. Larger cohorts in age categories with high (low) participation push the overall participation rate upward (downward).
The age-adjusted LFPR using 1960 age structure as a benchmark--when the baby boomers were not yet part of the working age population--shows that when larger age cohorts than in 1960 pass through high (low) participation rate age, this puts upward (downward) pressure on the LFPR (see chart 3). The observed labor force participation rate has been lower than the constructed age-adjusted LFPR since the late 1960s. This was initially due to a large cohort of young baby boomers in the low-participation age tranche, while a small cohort of so-called Silent Generation (born in the mid-1920s through the early 1940s) was simultaneously passing through the high-participation prime working age tranche--both flows pushing down the participation rate. And although the large baby-boomers cohort entered the high-participation prime-age tranche, the gap continued as the boost to the labor force participation rate was still counteracted by a low number of silent generation at the back end of the prime-age working group. More recently, the participation rate has fallen much more dramatically than the age-adjusted LFPR.
Since December 2007, the participation rate has fallen little more than 3 percentage points, while the age-adjusted LFPR has fallen only 1.2 percentage points. This indicates that while the Great Recession did depress labor force participation in the U.S. economy, its impact is overstated in the official LFPR.
Link to Growth
In order to see how labor force participation has affected economic expansion, we decomposed real GDP growth as the sum of growth in real GDP per labor force participant, the LFPR, and the working age population. (For more on our methodology, see Appendix B.)
A long-term view leads to a few important observations. First, since 2000, average annual GDP has increased 2.2%--markedly lower than the prior five decades, when growth averaged 3.6% a year. Second, in the 1970s and 1980s, the supply of potential workers surged as baby boomers (born from 1946-1964) came of age and women's participation in the workforce surged.(3) Third, and in contrast to the preceding decades, the LFPR has increasingly become a drag on growth since the turn of the millennium (see chart 4).
LFPR among the oldest cohort--both men and women--rose from 2000-2009 and has since leveled off. The increase was due, in part, to greater life expectancy (and, thus, the need to save for a longer retirement period), higher educational attainment of the group, and the rapid rise of health care costs.(4)
Among the prime age group (25-54), too, LFPR has been in secular decline since 2000. Within this group, LFPR for men has been on a decline going back to the late 1950s. Since prime-age men still represent almost one-quarter of the working-age population, even relatively small drops in their participation rate can have a meaningful effect on overall LFPR. As in the past, the decrease in participation among men with less education was greater than that of men with more education. LFPR among prime-age women has also trended lower, with the decrease more pronounced for those who didn't attend college.
(Notably, a study this year by Princeton Economics Professor Alan Krueger found that the increase in opioid dependency could account for perhaps a 0.6 percentage point--or 20%--of the decline in male labor force participation from 1999-2015. The study didn't quantify how often opioid addiction contributed to nonparticipation, or the other way around.)
Changes In The LFPR Could Mean Lower Unemployment Rates
Compositional changes in the workforce provide a simple explanation--at least partially--for why the longer-run rate of unemployment, which the Fed currently considers to be around 4.6%, may actually be lower than in previous economic expansions. With the U.S. labor force becoming increasingly made up of groups with historically low unemployment (e.g., those 55 and older, college graduates) and decreasingly comprising groups with high unemployment (e.g., teens), this suggests a lower longer-run unemployment rate. Additionally, to the extent new technology helps in job search and match, the degree of frictional unemployment (caused by people moving from one job to another) may also be lower than in the past, putting downward pressure on unemployment rates at the margin.
Given the uncertainty surrounding the LFPR in the near term, it's not outside of the realm of possibility that headline unemployment will decline another 1 percentage point from the current 4.2%. For now, our baseline forecast is 4.0% in the fourth quarter of next year. (For comparison, the low point for headline unemployment in the U.S. was 2.5% in the 1950s; 3.4% in the 1960s; 3.9% in the 1970s; 5.0% in the 1980s; 4.0% in the 1990s; 3.8% in the 2000s; and the current 4.2% in this decade.)
We ran a simulation to see what would happen if labor-force participation were to converge to 62.5%-63.5% by the end of 2018 (with estimated growth in the working-age population of 0.9% for the year and assuming the ratio of establishment survey employment to household survey employment to be constant at the average of the last 12 months, which is 0.955). For each LFPR, we factored in average monthly jobs growth of 75,000-200,000, in increments of 25,000. The results show that unemployment could fall to 3.3%, if the LFPR were to start declining toward 62.5% (starting the current quarter) and the economy continued to add an average of 175,000 jobs per month (more or less the current average) through the end of next year (see table 2). If current conditions were to prevail through the end of 2018--with inflation still below the Fed's target, the LFRP steady at around 62.8%-63%, and average monthly job growth of around the current pace of 175,000--the headline unemployment rate would have room to fall to 3.7%-4%. In this scenario, close to a million workers beyond natural rate of labor supply growth would be potentially pulled into the workforce (presumably the ranks of those currently on the sidelines).
Unemployment Rate By End-2018 (Simulation Table)
|--Payroll employment (monthly change, 000s)--|
|Participation rate (%)||75||100||125||150||175||200|
|Note: Along vertical axis, participation rate converges linearly to between 62.5 and 63.5 by end-2018. For each participation rate, we allow the monthly pace of jobs growth to lie between 75,000 and 200,000 by year-end 2018. Employment growth is designed to linearly reach the target if less than 3Q 2017; if target is higher than 3Q 2017, employment growth is held constant throughout the five quarters at target level.|
While much of the decline in the LFPR since 2007 is irreversible, there could be some cyclical bounce in the next few years. Retirees and those who receive disability aren't likely to return, but we see some room for improvement in older and prime-age workers to make a comeback--and, as job prospects improve, the same goes for those who want a job but don't have one.
However, any cyclical gains may be limited, given that, during the current recovery, the return of Americans to the labor force (after being out of it for whatever reason) hasn't risen. According to Professor Krueger's research, the likelihood of transitioning back into the labor force has edged down during this economic expansion--including in late-2015 and early-2016, when LFPR retraced 0.6 percentage points. This, therefore, poses a risk to our participation rate forecast for the next several quarters, when we expect more people, beyond natural rate of labor supply growth, to re-enter the labor force.(6)
In our latest baseline forecasts (published Sept. 27 2017 in "As the U.S. Economy Chugs Slowly Along, Washington Gridlock Prevails (Recent Deals Aside)"), we see LFPR holding steady around 62.8%-63% next year with a gradual decline in monthly payroll gains to 150,000 on average and unemployment rate at 4% by year end (current rate at 4.2%).
Another Decade-Long Drag
We still see a downward trend in LFPR in the next 10 years--albeit a less steep decline than we've seen in the past six years. There are two reasons for this: The decline in the potential participation rate won't be as sharp given the aging dynamics, and we expect more cyclical gains to soften the drag in the next two years (see chart 5). This means that the drag on the economy in the next 10 years is likely to be smaller than it has been (but it will still be a drag).
Given the 10-year projection of a slower pace of growth in the working-age population and our expectation that the decline in LFPR will slow, we think real GDP growth in the next decade will be closer to that of the past 16 years, on average, than what we saw before the turn of the millennium. Assuming output per labor-force participant remains on par with the historical average, we see the world's biggest economy expanding 2%-2.5% per year in the coming decade. If output per labor force participant matched the higher pace of the 1990s, growth could come closer to 3% a year. On the other hand, if GDP per labor force participant is more like that of the first decade of this millennium, real GDP growth would likely be just 1.5%-2%.
Additionally, because the Fed has a mandate for maximum employment as well as price stability, the degree of labor market slack (labor underutilization of working age population) is a key factor when the central bank determines the course of monetary policy. How fast (or slowly) the Federal Open Market committee lifts interest rates depends heavily on how many workers could re-enter the labor force.
If current conditions were to prevail through the end of 2018--with inflation still below the Fed's target, the LFRP steady at around 62.8%-63%, and average monthly job growth of around the current pace of 175,000--the headline unemployment rate would have room to fall to 3.7%-4%, as per the aforementioned scenario. In such a scenario, for the Fed, communications would be challenging given their latest median projection for the unemployment rate in the fourth quarter of next year at 4.1%. The uncertainties surrounding LFPR--both on a stand-alone basis and in terms of their interaction with wages, inflation, and monetary policy--gives the data-dependent Fed justification to keep in place what has been historically gradual (and cautious) monetary policy normalization.
Given the varying degree of changes of LFPR among age and gender groups, we can isolate the "demographic effect" of the decline in the rate by holding participation among each cohort constant at the level of a particular year (2007, in this case) and letting the shares of each group evolve as the populations did. Doing so allows us to construct the aggregate participation rate that would have been obtained if the only changes through time stemmed from changes in the population share of each group. In this scenario, the LFPR in August of this year would have been 63.6%, versus the actual 62.8% (see chart 6). (5) This translates into a "missing" 2.05 million workers. Thus, three-fourths of the decline in LFPR in the past decade is due to the "pure demographic" effect of the changing age structure of the U.S. workforce.
Changes in the LFPR within age cohorts over the years suggest that nonaging components also have meaningful role in the decline in the overall LFPR. This effect is embedded in the "other" component of the decline in participation rate (see chart 5), which represents the net effect of changes in participation rates since 2007 within specific age-gender group. We quantify the relative importance of variation in the participation rate within age group using shift share analysis broken into eight age-gender groups (see table 3). (The shift share analytical framework breaks down the total change in the participation rate with respect to a base year equals the sum of (a) changes in the population share of each group weighted by their base-year participation rate; (b) changes in the participation rate of each group weighted by their base-year population share; and (c) an interaction term that is typically small for years not too far from the base year (IMF, 2015).)
Contribution Of The Nonaging Component Of The Decline In The U.S. LFPR
The breakdown of contribution of nonaging components indicates that within-group declines in participation among younger workers (ages 16 to 24) and prime-age men (25 to 54) have been important in explaining the decline in age-adjusted participation rates. Increasing rates of participation by the 55+ group has helped stem the degree of decline. Generally, the pace of changes in participation within most demographic groups has slowed during this expansionary period, relative to past expansionary periods. The notable exception is prime-age men, who have seen an even faster decline in age-adjusted labor force participation rates. According to our calculations, 1.4 million of them are "missing."
A Modified Growth Accounting Framework
We use a modified version of a simple growth accounting framework to assess the impact of falling labor force participation and slowing working age population growth on the economy's long-run economic performance.
We delineate the growth identity of real GDP into its two primary contributions: GDP per labor force participant and labor inputs.
GDP = (GDP/hrs worked)*(hrs worked/labor force)*(labor force/population 16+)*population 16+
Or, simplifying the above equation gives:
GDP = (GDP/labor force)*(labor force/population 16+)*population 16+
GDP = GDP per labor force * labor force participation rate * working age population
The multiplicative factor is transformed into additive relationship by using the natural log of both sides, thus giving us the growth rate contribution of each component (in percentage points) to GDP growth (see chart 4).
Note that the measure of GDP per labor force has its denominator reflecting the labor force (which includes not only employed but also unemployed folks), as opposed to reflecting only the employed population of the labor force. Number of employed-only is generally used to calculate labor productivity in the literature. Accounting reduction similar to the one shown above can be done based on the employed-only segment of the labor force.
- Aaronson, Daniel, L. Hu, A. Seifoddini, and D. Sullivan. 2015. "Changing Labor Force Composition and the Natural Rate of Unemployment," in Chicago Fed Letter.
- Balkrishan, Ravi, M. Dao, J. Sole and J. Zook, 2015. "Recent U.S. Labor Force Dynamics: Reversible or Not?" in IMF's Working Paper Series.
- Congressional Budget Office (CBO). 2017. "The Budget and Economic Outlook: 2017 to 2027."
- Council of Economic Advisers. 2016. "The Long-Term Decline in Prime-Age Male Labor Force Participation."
- Erceg, Chris and A. Levin. 2013. "Labor Force Participation and Monetary Policy in the Wake of the Great Recession" in the Federal Reserve Board of Governor's Research Publication.
- Fujita, Shigeru. 2014. "On the Causes of Declines in the Labor Force Participation Rate" in Philadelphia Fed's Working Paper Series.
- Goldin, Claudia and J. Mitchell, 2017. The New Life Cycle of Women's Employment: Disappearing Humps, Sagging Middles, Expanding Tops. The Journal of Economic Perspectives, 31(1), pp.161-182.
- Hipple, Steven. 2016. "Labor Force Participation: What Has Happened Since the Peak?" in Bureau of Labor Statistics Monthly Labor Review
- Juhn, Chinhui, and S. Potter. 2006. "Changes in Labor Force Participation in the United States." Journal of Economic Perspectives 20(3): 27-46.
- Kruger, Alan, J.Cramer, and D. Cho. 2014. "Are the Long-Term Unemployed on the Margins of the Labor Market?" in Brooking's Papers on Economic Activity.
- Kruger, Alan. 2017. "Where Have all the Workers Gone? An Inquiry into the Decline of the U.S. Labor Force participation Rate" in Brookings Papers on Economic Activity.
- Szafran Robert, F. 2002. Age-Adjusted Labor Force Participation Rates, 1960–2045. Bureau of Labor Statistics, Monthly Labor Review.
- Zandweghe, Willem. 2012. "Interpreting the Recent Decline in Labor Force Participation" in the Federal Reserve Bank of Kansas City's Research Publication.
(1) Between 2000 and 2015, the school enrollment rate of teenagers 16-19 increased to 82.3% from 77.2%. The rising school enrollment rate among teenagers could have contributed to their falling labor force participation rate, because those enrolled in school are much less likely to participate in the labor force. From 2000 to 2015, the labor force participation rate of teenagers enrolled in school fell to 25.5% from 41.8%.
(2) The annual growth rate has declined to 1.0% over the 2006–2016 decade, from 1.3% between 1996 and 2006. The shares of the youth (16 to 24) and the prime age (25 to 54) groups in the civilian noninstitutional population are projected to decline over 2016–2026. In contrast, the share of the 55-years-and-older age group in the civilian noninstitutional population is projected to increase considerably.
(3) LFPR among women surged to roughly 60% by the end of the 1990s, from 34% at the beginning of the 1950s (LFPR among men was gradually declining. The U.S. women age-participation profile experienced a continuous outward shift throughout 1950s to 1980s, reflecting evolving cultural trends and behavioral changes that affect lifetime participation (such as education, marriage, etc.). That said, for cohorts born in the 60s and thereafter, the age participation profiles stopped shifting as much, and their lifetime participation profiles have basically been constant. This suggests no much cohort effect coming from this channel on the overall participation rate since the mid-90s.
(4) Changes over time to retirement programs--such as Social Security reforms and the rising use of defined-contribution rather than defined-benefit retirement plans--have increased the incentive to work longer. The decline in equity and residential asset values during the last recession may have created additional financial incentives to keep working.
(5) Because of data availability, nonseasonally adjusted (NSA) data were used for calculation. Numbers may differ a bit from BLS publications since we use 12-month-moving averages of NSA data. Most granular age-gender group data are of five-year spans available only NSA.
(6) If a cyclical recovery in labor force participation is unlikely, then a reversal of secular trends toward declining labor force is the only way to achieve an increase in labor force participation.
Writer: Joe Maguire