Retirement Age Stickiness

Definition

Retirement age stickiness is the empirical phenomenon in which workers' actual labor force exit behavior remains anchored at a historically significant age — particularly the old Full Retirement Age (FRA) of 65 — even after that age is no longer the financial or institutional anchor in the Social Security system. Deshpande, Fadlon, and Gray (2021) document this divergence using Social Security Administration (SSA) administrative data for $\approx 2.76$ million workers: when the 1983 Amendments raised the FRA from $65$ to $66$ , benefit claiming spikes shifted one-for-one to the new FRA, but retirement spikes remained at age 65 at essentially unchanged magnitude across a decade of post-reform cohorts. The stickiness is not explained by information frictions, spousal coordination, or Medicare — the most supported mechanism is employer-level norms and incentives.

Key Ideas

The Social Security literature had generally assumed that claiming and retirement move together. Stickiness shows they do not — they are governed by different institutions and respond to different incentives.
For post-reform workers (FRA=66) who exit the labor force at 65, there is a gap year with no income from either source — neither working nor collecting Social Security (SS) benefits. This income gap is a direct consequence of stickiness.
The stickiness persists across 10 birth-year cohorts of post-reform workers, ruling out transitional explanations (learning, adjustment frictions, preparation time).
Employer norms — retirement customs, demand-side age preferences, pension incentive structures, and related workplace conventions — appear to mediate retirement more than Social Security's financial incentives do.
The finding challenges a core rationale of FRA increases: that raising the retirement age will extend working years. It does not, for a substantial fraction of workers.

The Claiming-Retirement Divergence

The 1983 Social Security Act Amendments raised the FRA from $65$ to $67$ over $22$ years. For cohorts born $1943$ – $54$ , the FRA settled at exactly $66$ . The paper compares:

Pre-reform cohorts (born $1935$ – $37$ , FRA= $65$ )
Post-reform cohorts (born $1943$ – $45$ , FRA= $66$ )

Claiming (when SS payments begin):

Age	Pre-reform	Post-reform	Change
Claim at 62	$36.4\%$	$28.9\%$	$-7.5$ percentage points (pp)
Claim at 65	$22.0\%$	$8.5\%$	$-13.5$ pp
Claim at 66	$2.5\%$	$18.5\%$	$+16.0$ pp

The spike migrates almost entirely from 65 to 66. Claiming is flexible and responds to the FRA.

Retirement (labor force exit):

Age	Pre-reform	Post-reform	Change
Retire at 62	$5.9\%$	$6.0\%$	$+0.1$ pp
Retire at 65	$4.9\%$	$6.1\%$	$+1.2$ pp
Retire at 66	$4.0\%$	$5.7\%$	$+1.7$ pp

No spike migrates to 66. The 65 spike persists. Retirement is sticky.

The underlying retirement distribution does shift modestly toward later ages (consistent with income and substitution effects from the changed benefit schedule), but the discrete spike at 65 does not disappear. Even after controlling for cohort and time effects, the post-reform 65 spike is as large as the pre-reform 65 spike.

Regression Discontinuity Evidence

The cohort analysis uses large FRA changes (full year). The regression discontinuity (RD) complements it using 2-month incremental FRA changes at the January 2nd birthday cutoffs for phase-in cohorts (born $1938$ – $42$ ). Multiple cutoffs are stacked for statistical power; the sample is $\approx 507{,}000$ .

Claiming (RD): $2$ -month FRA increase $\to$ $0.84$ -month average increase in claiming age; propensity to claim within 2 months of the new FRA $+12$ pp. Strong, precise response.

Retirement/earnings (RD — high-powered null results):

Can rule out earnings increases above $\$292$ at age $65$ and $\$312$ at age $66$
Expected earnings increase if retirement responded as flexibly as claiming: $\approx \$1{,}154$ at $65$ , $\approx \$955$ at $66$
Ratio of upper bound to expected response: $25\%$ at $65$ , $33\%$ at $66$

The RD provides causal identification: the two-month variation is quasi-random conditional on birth cohort, and the Continuous Work History Sample (CWHS) sample is large enough to detect even modest retirement responses if they existed.

Mechanisms

Ruled Out

1. Information and learning The FRA increases were announced in 1983, decades in advance. SSA mailed annual SS statements to all workers from $1999$ – $2011$ . Even after $2011$ , mailed statements to workers over 60. Post-reform workers were informed. Moreover, the stickiness does not shrink across the 10 birth-year cohorts in the post-reform sample ( $1938$ – $1948$ ) — the earliest cohorts were already past their FRA when the paper was written, meaning they had years to observe the new FRA in practice. Learning cannot explain a stickiness that does not diminish across cohorts.

2. Joint spousal retirement Female spouses are on average younger than male spouses. If women retired at 65 to coincide with older husbands already at the old FRA, stickiness should be stronger for women. The spike at 65 is equally large for men and women in post-reform cohorts.

3. Medicare eligibility at 65 Even though the FRA increased, Medicare eligibility remained at 65. Workers who continue working for employer health insurance might exit at 65 upon Medicare access. Using Form 5500 employer benefit-plan data linked by employer identification number (EIN) to CWHS earnings, regressions of workplace-level retirement behavior on whether the employer's plan covers retirees (health continuation coverage) find no relationship. Firms with and without retiree health coverage show the same 65 stickiness.

Supported (Suggestive)

4. Employer norms and incentives

The movers analysis leverages workers who switched employers between ages $50$ – $60$ . The specification compares movers from the same origin employer who land in destinations with different retirement-behavior norms:

Destination rate: Moving to a workplace where other employees are $1$ pp more likely to retire at 65 is associated with an $8.4$ pp higher propensity for the mover (post-reform cohort, FRA= $66$ ) to retire at the old FRA of 65.
Destination change in rates: Moving to a workplace that did not shift its retirement rate from 65 to 66 after the FRA increase is associated with a $7.4$ pp higher propensity to exhibit stickiness.
No analogous relationship holds for moves across counties — geography does not predict retirement behavior.

The identifying assumption (no nonrandom selection across destinations conditional on origin fixed effects) is supported by the finding that destination intensity is nearly uncorrelated with movers' baseline lifetime earnings — the strongest observable predictor of retirement behavior. This is consistent with but does not prove causality.

Candidate employer-level sub-mechanisms: workplace retirement norms and conventions, demand-side preferences for workforce age composition, employer-sponsored pension plan structures and withdrawal norms.

The Gap Year Problem

For post-reform workers who remain stuck at 65, the claiming-retirement divergence creates a gap year: they exit the labor force at 65 (no earnings) but do not claim SS benefits until 66 (no SS income). During this gap, their income sources are limited to savings, spousal earnings, or other transfer programs.

Workers with qualifying health impairments who face this gap year are potential disability insurance (DI) applicants — the income loss from the gap year may push them toward disability program entry if they also face the health and work-capacity criteria for DI. The stickiness thus creates a partial pathway from retirement anchor effects to disability program enrollment, particularly for workers in declining health who anticipated retiring at 65 but can no longer rely on immediate SS income. See Early Retirement Safety Net Gap and Conditional DI Applicants.

International Comparison

A notable contrast: studies of FRA increases in Austria, Switzerland, Germany, and Italy find strong shifts in labor supply in response to FRA changes, with claiming and retirement moving in lockstep. The U.S. finding of stickiness appears to be institutionally distinctive. The authors attribute the difference to the stronger role of U.S. employer institutions — defined-benefit pension plan norms, informal retirement conventions, and workplace demand-side factors — relative to the more centralized retirement systems in continental Europe.

Why It Matters

For Social Security Reform Policy

FRA increases have been proposed as a fiscally neutral way to reduce SS program costs while extending working years and improving labor force participation among older Americans. The stickiness finding splits these two rationales: FRA increases do reduce costs (by shifting claiming later and thus paying benefits for fewer years), but they largely do not extend working years for a substantial fraction of workers. Reform proposals premised on both rationales simultaneously will achieve only the first.

For Employer Policy as a Complement

The employer mechanism implies that extending working years requires engaging employers, not just adjusting Social Security incentives. Employer-based pension restructuring, elimination of mandatory retirement conventions, and demand-side policies for older workers are suggested as necessary complements to FRA increases.

For the Early Retirement Safety Net Gap

The stickiness creates a new version of the safety net gap: not the $62$ – $64$ structural gap (workers too impaired for DI/Supplemental Security Income (SSI), relying on early Old-Age and Survivors Insurance (OASI)) but a $65$ – $66$ behavioral gap (post-reform workers who retire at 65 per employment norms but receive no income from any source for up to a year). Workers with impairments in this gap are likely DI applicants. See Early Retirement Safety Net Gap.

For DI Caseload

If employers push workers out at 65 through norms and pension structures, and those workers cannot claim SS for another year, the path from employment to DI application shortens for those with health impairments. This is an indirect channel from retirement stickiness to DI enrollment growth.

Open Questions

What specific employer-level features drive the stickiness? The movers design identifies employer-level aggregate norms but cannot separate pension incentives, informal customs, demand-side age preferences, or health insurance linkages within the employer channel.
Would voluntary employer-based policies (e.g., phased retirement, flexible scheduling) reduce stickiness, or is the norm-setting so entrenched that structural incentives are required?
Does the stickiness interact with health: are workers who retire sticky at 65 systematically less healthy than those who follow the new FRA? If so, stickiness may be partly a health-driven phenomenon mislabeled as a norm effect.
How does stickiness interact with the earnings test? The earnings test penalizes earnings above a threshold for early claimers. Its elimination above the FRA in 2000 may affect the gap-year calculus for post-reform workers.
Will stickiness persist as the FRA increases further toward 67? The paper's phase-in cohorts do not yet include the second tranche of FRA increases (67 for those born $1960$ +), leaving open whether a second anchor forms at 66 or retirement follows the new FRA to 67.
How does employment lock interact with stickiness? Singleton (2024) shows that workers who receive new health diagnoses near age 62 extend employment to retain employer health insurance — an insurance-driven employment extension distinct from the employer-norm anchor at 65. Both mechanisms delay labor force exit, but via different channels and at different ages; their joint distribution is unknown.