## 1. Demographic Data and Analysis

## Source details

**Canonical URL:** [1. Demographic Data and Analysis](https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2015/_wp1576.pdf)

## Other formats

- [Markdown version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2015/_wp1576.pdf.md)
- [Structured JSON version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2015/_wp1576.pdf.json)

---

### Key observations and questions
- Current LFPR of 62.8 percent is around the lowest rate since 1978.
- LFPR rose from just below 60 percent in the early 1960s to above 66 percent by 1990, peaking at 67.3 percent in 2000Q3.
- Since the 2001 Recession the LFPR has been largely on a secular decline.
- Central questions:
  - How much of the post-2007 decline is driven by demographics, cyclical, and other structural forces? How much is reversible?
  - What is the baseline forecast for the LFPR over the next few years and the risks around it? What is the current and projected level of labor market slack?
  - What are the macroeconomic and supply-side policy implications?

### Main findings (summary)
- Around ¼–⅓ of the post-2007 decline is reversible.
- Demographics (population aging) explain around 50 percent of the near 3 percentage points LFPR decline during 2007–13.
- State-level panel regression analysis suggests about 30–40 percent of the decline is driven by cyclical factors.
- The remainder is attributable to non-demographic structural factors such as increasing college enrollment, fewer students working, and cohort effects.
- Given continuing population aging, the LFPR will continue to decline in the medium term; the limited reversal from cyclical recovery causes only a near-term flatlining before the secular decline reasserts itself.

### Labor force participation dynamics by gender and age
- Male participation rate trends:
  - Declined by 0.1 percentage points (p.p.) per year between 1995 and 2007.
  - Declined by 0.6 p.p. per year between 2008 and 2014.
- Female participation rates started declining in the late 1990s and have since followed a similar pattern to males.
- Age-group patterns:
  - 16–24 year-olds have been steadily reducing their participation rates since 2000.
  - Prime-aged workers have also reduced participation rates, to a lesser extent.
  - Older workers increased participation:
    - For those aged 65 and above, participation rates increased by almost 50 percent for males since the late 1990s.
    - For those aged 65 and above, participation rates nearly doubled for females since the late 1990s.

### Methodological approaches used
- National-level population models:
  - Model 1: Hold age-group participation rates constant at 2007 levels and let population shares evolve historically to construct an aggregate participation series (demographics-only counterfactual).
  - Model 2: Estimate age-group participation trends (e.g., 2000–2007) and project each age group's participation rate, then combine with population shares (this mixes demographic and structural age-specific trends).
  - Both population models suggest demographics account for around ½ of the LFPR decline since 2007.
- Shift-share analysis:
  - Decomposes total change in aggregate participation into:
    - (a) Changes in population share of each age group weighted by base-year participation rates ("population share shift" or “demographic effect”).
    - (b) Changes in age-group participation rates weighted by base-year population shares ("participation rate shift").
- State-level panel regression analysis:
  - Uses a panel of state-level data to identify cyclical effects.
  - Employs the “Bartik shock”—predicted employment growth based on a state's industry mix—as an instrument to separate labor supply shocks from cyclical conditions.

### Structural versus cyclical decomposition (2007–2013)
- Demographic (population aging) explains around 50 percent of the drop in the aggregate participation rate during 2007–2013.
- State-level panel regressions imply a cyclical effect accounting for 33–43 percent of the near 3 p.p. drop during 2007–13.
- The remainder is non-demographic structural factors (e.g., college enrollment, fewer students working).
- By gender and subperiod (selected figures preserved verbatim):
  - Total LFPR Change 2007-10: -1.3 ppt; Population Shift: -0.6 ppt; Participation Shift: -0.8 ppt.
  - Total LFPR Change 2010-13: -1.5 ppt; Population Shift: -0.8 ppt; Participation Shift: -0.7 ppt.
  - Men 2007-10: Total -2.0 ppt; Population Shift -0.6 ppt; Participation Shift -1.5 ppt.
  - Men 2010-13: Total -1.5 ppt; Population Shift -1.0 ppt; Participation Shift -0.4 ppt.
  - Women 2007-10: Total -0.7 ppt; Population Shift -0.5 ppt; Participation Shift -0.2 ppt.
  - Women 2010-13: Total -1.4 ppt; Population Shift -0.6 ppt; Participation Shift -0.8 ppt.
- Age-group contributions (selected):
  - Pop. Shift 25-54 (Total): -1.4 (2007-10) and -2.1 (2010-13) for total.
  - Part. Shift 16-24: -0.9 (2007-10) and -0.1 (2010-13) for total.

### Cohort effects and age-participation profiles
- Female age-participation profiles shifted outward for cohorts born between the 1940s and mid-1950s; cohorts born after the mid-1950s show profiles that “stopped shifting.”
- For cohorts relevant to post-2007, cohort effects are judged to have had limited impact on aggregate LFPR trends.
- Structural changes specific to particular age groups (e.g., lower teen participation, higher early prime-age participation) have been more important.

### State-level panel regression analysis: methodology and main results
- Model features:
  - Levels model includes state-specific linear and quadratic trends and lagged cyclical terms.
  - First-difference model estimated (equation (3)) to avoid non-stationarity and high persistence.
  - Employment gap measured as difference between payroll or household employment and state-specific HP trend; end-point trend growth from 2002–2005 average imposed from 2006 onward.
- Endogeneity and instrumentation:
  - Unemployment rate change vs participation rate change correlation: -0.16 (weak).
  - Instrument used: industry mix “Bartik shock” (predicted employment growth based on national industry growth weighted by state industry shares).
- Panel regression findings (1976–2012, payroll employment gap IV results summarized):
  - A 1 percent increase in the employment gap leads to a 0.1 percentage point increase in participation rate in the same year, and another 0.1 percentage point increase in each of the subsequent two years (total roughly 0.3 p.p. over three years in pre-crisis estimates).
  - During the Great Recession and recovery, contemporaneous effect reduced by half and adjustment more persistent; total effect of a 1 percent higher employment gap around 0.2 p.p., distributed roughly evenly across 4 years.
  - Recasting regression results: cyclical effect explains 33–43 percent of the near 3 p.p. drop in LFPR during 2007–13.
  - Using the last-column model: cyclical conditions explain about 50 percent of the 1.4 p.p. drop in LFPR during the Great Recession (2007–10); post-2010 cyclical conditions explain 20–35 percent of the LFPR decline.

### Heterogeneity across states and age groups
- State results:
  - Predicted cyclical declines align with states hardest hit by the crisis: Nevada, Arizona, Florida, California.
  - States least affected predicted to have no change or increases: DC, New York, North Dakota.
  - In many states model under-predicts actual LFPR fall, implying demographic and other structural forces also at work.
  - In Nevada and Arizona, model over-predicts decline due to increases in participation among 55+ groups (possible response to housing wealth losses).
- Age-specific cyclicality:
  - Cyclical sensitivity declines with age.
  - Youngest groups (16–24) most sensitive; prime-age (25–54) less sensitive; older groups (55–64, 65+) show volatile coefficients.
  - During crisis and recovery, cyclicality decreased for young and prime-age groups.
  - Pre-crisis, 54–64 group showed counter-cyclical participation (strong economy -> earlier retirement); post-2007 this effect becomes insignificant.

### Youths, SSDI, and older workers
- Youths (ages 16–24):
  - Decline mainly explained by fewer students working rather than higher college enrollment alone.
  - Student worker share: peaked at 46 percent in 2000, declined to less than 40 percent by 2007.
  - Shift-share for 2007–2013: Part. Rate Change -0.8 (annualized); Enrolled Part. Rate Shift -0.5; Enrolled Population Shift 0.0; Unenrolled Participation Rate Shift -0.2; Unenrolled Population Shift -0.1.
  - Much of the student-worker decline predates 2007; a near 5 p.p. plummet in 2008–09 suggests a sizable cyclical component that could be partly reversible.
  - Reversion to pre-Great Recession averages of school enrollment and employment rates for students would increase youth participation by around 7pp.
- SSDI:
  - Applications spiked during the Great Recession, while acceptance rates fell to near historical lows.
  - Change in SSDI beneficiaries/population was 0.6 p.p. during 2007–13.
  - When normalized by population, changes in SSDI recipients did not shift significantly following the Great Recession.
  - Lack of strong correlation between state-level changes in SSDI recipients and LFPRs.
  - Nearly 80 percent of SSDI recipients are above 45 years old; SSDI rises are linked to population aging and are largely irreversible because recipients tend to exit labor force permanently.
- Older workers (55+):
  - LFPR for 55+ increased by around 10 p.p. from mid-1990s up to early 2009; rate of increase slowed and began slight decline in early 2013; stable around 40 percent since late 2013.
  - Factors behind earlier increase: better health and longevity; shift from defined benefit to defined contribution pensions; rising healthcare costs and loss of retiree health benefits; sensitivity to stock market performance increased since 2000.
  - During Great Recession older workers remained in labor force to rebuild net worth; as wealth replenishes, many opted to retire post-2013.

### LFPR forecasts and labor market slack
- Baseline forecast decomposition:
  - (i) Pure demographic effect holding age-group participation rates constant at 2007 levels and using census baseline population forecast.
  - (ii) Cyclical bounce-back benchmarked off state-level analysis.
  - (iii) Judgment on non-demographic structural forces (college enrollment, student work share, retirement patterns).
- State-level panel regression projection:
  - Predicts a cyclical bounceback of around ¼–⅓ of the LFPR decline during 2007–13 over the next 5 years.
  - Despite cyclical bounceback, LFPR continues to decline due to structural forces.
- Baseline scenario specifics:
  - Youth LFPR expected to bounce-back by around 2 p.p. as school enrollment declines closer to 2007 levels and more students work; older workers forecast to have no bounce-back due to resumed retirements as wealth is re-accumulated.
  - Aggregate participation roughly flat for 2014–16, then secular decline resumes from 2017 as aging dominates.
  - Baseline LFPR at 2019 around 62.3 percent (text compares to CBO projection of 62 percent by end-2019).
- Employment gap and broader slack:
  - Broader employment gap constructed as weighted sum of unemployment and participation gaps, with an adjustment for part-time workers due to slack.
  - Participation gap grew to almost 1.0 ppt by end-2014 and became the main component of overall employment gap.
  - Despite participation gap emergence, unemployment gap accounted for around three-fifths on average of overall employment gap since 2008.
  - Broader employment gap peaked in 2010 at 4½ percent; fell to around 2¾ percent in 2013; around ¾ of a percentage point by end-2014.
  - Expectation: overall labor market slack to close in 2016, but uncertainty over NAIRU, participation trend and part-time gap could push closure to 2017.

### Key numeric findings and estimates (preserved verbatim)
- Structural component estimates (selected):
  - Aaronson et. al. 2007-14: 54%
  - CEA (2014) 2007-14: 52%
  - CBO (2014) 2007-13: 50%
  - Mishel, Bivens, Gould, and Shierholz (2012) 2007-11: 33%
  - Fujita (2013) 2000-13: 65%
- Aggregate LFPR changes:
  - 2000-2013 Total Change in LFPR: -3.8 ppt.
  - 2000-2007 Total Change in LFPR: -1.0 ppt.
  - 2007-2013 Total Change in LFPR: -2.9 ppt.
- Cyclical contributions (from Table 5 summary):
  - Using payroll employment model (whole sample): Cyclical Contribution around -0.9 to -1.2 ppt.
  - Closing labor demand gap in medium term would increase LFPR by (in ppt): whole sample 0.8 (99 pct confidence interval 0.6-1.0) using payroll emp; other variants: 0.9 (0.7-1.2), 0.7 (0.5-0.9), 0.7 (0.4-1.0) depending on specification and sample.
- Youth-specific numbers:
  - Student worker share: peaked at 46 percent in 2000, declined to less than 40 percent by 2007.
  - Reversion to pre-Great Recession averages of school enrollment and employment rates for students would increase youth participation by around 7pp.
- SSDI:
  - Change in SSDI beneficiaries/population was 0.6 p.p. during 2007–13.
  - Nearly 80 percent of SSDI recipients are above 45 years old.
- Older workers:
  - LFPR for 55+ around 40 percent since late 2013; increased by around 10 p.p. from mid-1990s to early 2009.

### Conclusions and policy recommendations
- Conclusions:
  - Around ¼–⅓ of the post-2007 LFPR decline is reversible (cyclical), but LFPR will continue to fall over medium term because population aging explains around 50 percent of the near 3 p.p. LFPR decline during 2007–13; state-level regressions attribute 33–43 percent to cyclical factors; remaining decline due to non-demographic structural factors.
- Risks and uncertainties:
  - Forecasting LFPR for youths and older workers is challenging; alternative demographic, enrollment, and behavioral scenarios can produce differences up to 1 full percentage point (Aaronson et al.) or up to 2 percentage points by mid-2020s (CEA) depending on recovery of non-aging components.
  - Potential for further hysteresis effects from persistent long-term unemployment.
- Policy priorities to enhance labor supply and offset aging headwinds:
  - Enhance training and job search assistance programs (such as sectoral training), particularly those that engage industry and higher education institutions.
  - Better family benefits (including childcare assistance) to reverse downward trend in female LFPR.
  - Modify disability program to allow for part-time work by recipients; reduce penalties for working during application process; re-examine eligibility rules to prevent misuse (especially for disability related to mental illness).
  - Provide greater visa opportunities for high-skilled immigrants.
  - Expand the EITC to childless workers and lower the age threshold from 25.

*Italic Source: U.S. Bureau of Labor Statistics, Haver Analytics and IMF staff calculations as presented in _wp1576 - Annex I for more details).*

### 1. Demographic Data and Analysis .......................................................................................

### 1. Demographic Data and Analysis

### Key observations and questions
- Current LFPR of 62.8 percent is around the lowest rate since 1978.
- LFPR rose from just below 60 percent in the early 1960s to above 66 percent by 1990, peaking at 67.3 percent in 2000Q3.
- Since the 2001 Recession the LFPR has been largely on a secular decline.
- Central questions addressed:
  - How much of the post-2007 decline is driven by demographics, cyclical, and other structural forces? How much is reversible?
  - What is the baseline forecast for the LFPR over the next few years and the risks around it? What is the current and projected level of labor market slack?
  - What are the macroeconomic and supply-side policy implications?

### Main findings (summary)
- Around ¼–⅓ of the post-2007 decline is reversible.
- Demographics (population aging) explain around 50 percent of the near 3 percentage points LFPR decline during 2007–13.
- State-level panel regression analysis suggests about 30–40 percent of the decline is driven by cyclical factors.
- The remainder is attributable to non-demographic structural factors such as increasing college enrollment, fewer students working, and cohort effects.
- Given continuing population aging, the LFPR will continue to decline in the medium term; the limited reversal from cyclical recovery causes only a near-term flatlining before the secular decline reasserts itself.

### Labor force participation dynamics by gender and age
- Male participation rate trends:
  - Declined by 0.1 percentage points (p.p.) per year between 1995 and 2007.
  - Declined by 0.6 p.p. per year between 2008 and 2014.
- Female participation rates started declining in the late 1990s and have since followed a similar pattern to males.
- Age-group patterns:
  - 16–24 year-olds have been steadily reducing their participation rates since 2000.
  - Prime-aged workers have also reduced participation rates, to a lesser extent.
  - Older workers increased participation:
    - For those aged 65 and above, participation rates increased by almost 50 percent for males since the late 1990s.
    - For those aged 65 and above, participation rates nearly doubled for females since the late 1990s.

### Methodological approaches used
- National-level population models:
  - Model 1: Hold age-group participation rates constant at 2007 levels and let population shares evolve historically to construct an aggregate participation series (demographics-only counterfactual).
  - Model 2: Estimate age-group participation trends (e.g., 2000–2007) and project each age group's participation rate, then combine with population shares (this mixes demographic and structural age-specific trends).
  - Both population models suggest demographics account for around ½ of the LFPR decline since 2007.
- Shift-share analysis:
  - Decomposes total change in aggregate participation into:
    - (a) Changes in population share of each age group weighted by base-year participation rates ("population share shift" or “demographic effect”).
    - (b) Changes in age-group participation rates weighted by base-year population shares ("participation rate shift").
- State-level panel regression analysis:
  - Uses a panel of state-level data to identify cyclical effects.
  - Employs the “Bartik shock”—predicted employment growth based on a state's industry mix—as an instrument to separate labor supply shocks from cyclical conditions.

### Comparative references to other studies (as reported in text)
- Fujita (2013): Using CPS micro data on reported reasons for non-participation, finds retirement and disability account for two-thirds of the decline in participation between 2000 and 2013, with decline due to retirement occurring after 2010.
- CBO (2014) and CEA (2014): Use approaches similar to the demographic models here and find structural/demographic forces account for around 50–60 percent of the participation rate decline during 2007–13.
- Mishel et al. (2012): Find the structural component explains only one-third of the fall between 2007 and 2011; the difference reflects use of a longer-term trend (1989–2007) in that study.

### Labor market slack and policy implications
- The paper’s “employment gap” measure incorporates participation and part-time worker gaps as well as the unemployment gap and indicates persistent slack that will decline only gradually under the baseline.
- Policy implications highlighted:
  - Stimulative macroeconomic policies remain important to help reach full employment.
  - Labor supply measures are essential to boost potential growth given continued downward pressure on the LFPR.
  - Stimulative macroeconomic and labor supply policies can help reduce the scope for further hysteresis effects (e.g., loss of skills, discouragement).

### Forecast and outlook (baseline)
- Only around ¼–⅓ of the post-2007 LFPR decline is forecast to be reversed over the next few years as job prospects improve.
- Population aging continues to exert downward pressure, so any cyclical rebound leads primarily to a near-term flatlining of the LFPR; the secular decline resumes as cyclical improvements wane.

*Source: _wp1576 - 1. Demographic Data and Analysis (excerpts) from the provided IMF content*

### Annex I for more details).

### Appendix 1. Demographic Data and Analysis

### Appendix 1. Demographic Data and Analysis

### Methodological approach
- Two-pronged strategy to disentangle the effect of population dynamics on the participation rate:
  - Demographic approach: relies on disaggregated population and participation data by age group (10 groups) and gender to estimate the demographic component of the decline in participation rates.
  - Shift-share analysis: investigates the behavior of specific age groups by decomposing total change in the participation rate with respect to a base year into three components:
    - (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;
    - (c) an interaction term that is typically small for years not too far from the base year.
- The shift-share identity is presented as equation (A.1) where the aggregate participation rate and group-specific participation rates and population shares are denoted by the corresponding symbols in the text.

### Data sources and coverage
- Labor force data by gender and age groups from the Household Employment Survey of the Bureau of Labor Statistics (BLS).
- Age groups used: 16-19, 20-24, 25-34, 35-44, 45-54, 55-59, 60-64, 65-69, 70-74, 75+.
- Time coverage: 1981 to present.
- Population data, including forecasts for 2014–2019, from the BLS.
- Immigration data used in simulations (described in section II of the Annex) from the US Census Bureau.

### Appendix 2 — State-Level Regression Model (household employment)
- Purpose: summarize regression results of estimating equation (3) using household employment to construct the employment gap as a measure of the state business cycle.
- Trend-cycle decomposition and end-point adjustment follow the same procedure as for payroll employment.

Key points from regression analysis:
- Instrument validity:
  - The industry mix variable remains a very strong instrument for the employment gap when household employment is used, evidenced by positive, statistically significant first stage coefficients and large F-statistics.
- OLS bias:
  - The bias of OLS is positive and substantially larger than when using payroll employment.
  - Rationale: household employment encompasses a wider definition of employment (including self-employment) and is more prone to an endogenous response to labor supply; an increase in labor supply raises both the LFPR and self-employment, biasing OLS upward.
  - Household employment is derived from the same survey as the LFPR, introducing possible mechanical correlation that could render OLS regression spurious.
  - Instrumentation is therefore even more important when using household employment; this is reflected in the Hausman test results reported in the table notes.
- Magnitude and persistence of cyclical response:
  - When instrumenting using the industry mix variable and accounting for sufficient lags, the total effect of a 1 percent increase in the employment gap leads to a total of 0.2 p.p. increase in LFPR within 2 years, the same estimate obtained using payroll employment (columns 3 to 5).
  - The cyclical response has been more sluggish and persistent following the Great Recession, though it still adds up to the same total effect as estimated with the whole sample.
- Estimation details (as reported in the table notes):
  - Column 1: OLS, no lags in the employment gap variable.
  - Column 2: instruments the contemporaneous employment gap with the industry mix based employment growth in equation (4).
  - Columns 3 to 6: introduce further lags in the employment gap variable.
  - Columns 4 to 6: weight the data by the average working-age population in each state.
  - Columns 5 and 6: split the sample to sub-samples before and following the Great Recession.
  - The Hausman test result reports the p-value of the null hypothesis that the contemporaneous employment gap is exogenous.
  - The 1st stage panel reports the first stage coefficient for the contemporaneous employment gap and the first stage F-statistics.
  - All specifications include state-specific intercepts and trends (not shown).
  - Standard errors are robust to heteroskedasticity and auto-correlation (using Newey-West kernel).
  - ***, ** denote 1 and 5 percent statistical significance respectively.

*Source: Appendix 1 and Appendix 2, _wp1576 - Appendix 1. Demographic Data and Analysis.*

---


_Source: https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2015/_wp1576.pdf_