Demographic drivers of long-run labor force participation and unemployment in the US
In this paper,Andreas Hornstein and Marianna Kudlyak build aggregate trends for U.S. labour force participation (LFP) and unemployment from the bottom up, using CPS microdata from 1976–2024. They construct annual LFP and unemployment rates for 44 age-gender-education groups (crossing 2 genders, 5 age brackets, and 4 education levels for adults 25+, plus youth categories). For each group, they estimate a parsimonious state-space model that decomposes each series into three unobserved components:
- Cohort effects: lasting influences specific to a birth cohort (modelled as a random walk for the entering cohort, then fixed as it ages)
- Age effects: influences of an individual’s current age, allowed to vary stochastically over time
- A cyclical component: driven by a common latent factor shared across unemployment and LFP, estimated with an AR(1) and three lags for the LFP response
For the 20-year forecasts, they take age-gender population projections from the U.S. Census Bureau (2025) and forecast future education shares using a separate cohort state-space model of educational attainment.
Findings
- Aggregate unemployment trend declined smoothly from 8% in 1976 to around 5% by 2024. This 3 pp drop is attributed roughly equally to: rising educational attainment (largest single contributor), population ageing, and declining group-level unemployment trends, with cohort effects driving most of the within-group drift.
- Aggregate LFP trend follows a hump shape, rising 4.1 pp from 1976 to the mid-1990s, then falling 4.4 pp through 2024.
The rise was driven mainly by educational upgrading (+2.2 pp) and rising within-group participation trends (+1.2 pp) while the subsequent decline was driven predominantly by population ageing (−4.3 pp), compounded by declining within-group LFP trends (−2.3 pp), partly offset by continued educational upgrading. - Pronounced gender divergence in participation trends:
Men’s trend LFP fell by roughly 10 pp over 1976–2024, driven almost entirely by cohort effects as successive male cohorts participate less. – Women’s trend LFP, on the other hand, rose by 10 pp through 2000, then receded by 2 pp as successive female cohorts participated more, until cohorts entering the late 1990s/early 2000s. The gender gap in LFP change is almost entirely explained by divergent cohort effects, not age or composition shifts. - Twenty-year projections: Trend LFP is forecast to decline a further 1.5 pp from its 2024 level of 62.3%, while trend unemployment is projected to edge down 0.4 pp from 4.7%, remaining historically low. These movements are driven mainly by population ageing (negative for LFP), partially offset by educational upgrading and cohort replacement. Men’s LFP is projected to decline further; women’s to stagnate.
Trends in Labor Force Participation and Unemployment, 1976-2024
Authors: Andreas Hornstein, Marianna Kudlyak
From: Federal Reserve Bank of Richmond, Federal Reserve Bank of San Francisco
What pulls and pushes men out of the labor market generation by generation
In this paper, Leila Bengali, Mary C. Daly, Evgeniya A. Duzhak, and Cindy Zhao use 45 years of CPS monthly microdata (1976–2023) for prime-age men (25–54), organizing the analysis by four generations (Silent, Boomers, Gen X, Millennials) in order to understand why the fraction of men working in the US has declined consistently since the 1950s.
The core method is a linear probability regression, estimated separately per generation, that predicts the probability of being out of the labor force controlling for age, demographics (race, education, marital status, urban/rural residence), the national unemployment rate, and state fixed effects. Push and pull factors are then added to this baseline one group at a time: pull factors (extended schooling, caretaking) and push factors (disability, skills mismatch proxied by declining manufacturing share and the real minimum wage gap), to measure how much of the generational participation gap each explains and whether its influence has grown over time.
Findings
Scale & trend
Prime-age male non-participation tripled from 2.9% in the 1950s to 10.9% in 2023. The rise is structural, not cyclical: generational gaps persist at every age even after controlling for demographics.
Pull factors
- Extended education explains a large share of the higher non-participation of young Millennial men (ages 25–30); without it, Millennials’ profile would look like a simple upward shift from GenX.
- Caretaking is a growing pull across all generations; the effect is strongest when men have young children and an employed partner and is significantly weaker in Canada, where childcare is more subsidized.
Push factors
- Disability is the single largest stated reason for non-participation (38% of cases in 2023), though its policy link via disability insurance is weakening for younger cohorts.
- Skills mismatch hits hardest among the least educated: the Millennial–Boomer non-participation gap at age 25 is 9.4 pp for men with high school or less, versus just 3.2 pp for college graduates.
Pulled Out or Pushed Out? Declining Male Labor Force Participation
Authors: Leila Bengali, Mary C. Daly, Evgeniya A. Duzhak, Cindy Zhao
From: Federal Reserve Bank of San Francisco, Princeton University
