RESEARCH DOSSIER
Disability Insurance Labor Supply and Claimant Reintegration Outcomes
Abstract. Disability insurance allocates earnings risk across time, individuals, employers, insurers, and public institutions. Its labor-market consequences arise jointly from the incidence and duration of illness or impairment, eligibility assessments, benefit replacement and offsets, employers’ incentives to retain workers, available accommodations, medical and vocational services, and the legal protection of work choices. This monograph examines how disability insurance participation changes labor supply and how claimants regain, maintain, or reshape paid employment. It separates work disability from functional limitation, medical need from a contractual or statutory benefit entitlement, and benefit exit from durable reintegration. It synthesizes causal evidence from U.S. disability examiner and appeals-judge assignment, Dutch screening reforms, administrative and survey statistics, international labor-market comparisons, and actual work-incentive rules. Maestas, Mullen, and Strand (2013) estimated that, among U.S. applicants at the margin of initial allowance, employment would have been 28 percentage points higher without benefit receipt; French and Song (2014) estimated a 26-percentage-point reduction in participation three years after an appeals decision. These are different local causal effects, and neither describes every disabled person or every insurance product. Parallel evidence demonstrates severe baseline employment constraints among denied applicants, substantial demand-side exclusion, and wide differences in national reintegration architecture. The manuscript develops an original, source-bound analytical synthesis connecting eligibility margins, replacement income, timing of assistance, employer retention, employment quality, health, and the durability of earnings without proposing a new proprietary theory. The resulting scientific conclusion is that an evaluation of disability insurance must examine joint distributions of income security, health, work quality, and sustained employment, including unsuccessful attempts and transitions back into support; a benefit-termination rate by itself is an insufficient outcome measure.
Keywords: Disability insurance; labor supply; Social Security Disability Insurance; employment; earnings; work incentives; benefit eligibility; causal inference; claimant reintegration; return to work; vocational rehabilitation; partial disability; employment retention; workplace accommodation; duration analysis; disability rights; social protection; international comparative policy.
Introduction
Disability insurance protects a person whose health or functional circumstances interrupt or reduce the ability to earn. Its economic effect is therefore inseparable from the event that creates the claim. The central research problem is to determine how benefit rules, entitlement decisions, and reintegration services change behavior beyond what would have occurred under the same illness, labor market, and household circumstances without that particular intervention. A worker who remains unemployed after a claim can reflect severe incapacity, absence of accessible work, weak employer retention, unfavorable benefit incentives, delayed rehabilitation, or several of these together. Observed nonemployment does not identify the contribution of any single mechanism.
The WHO classification of functioning and disability treats functioning as an interaction between health conditions and contextual factors, including the environment. That measurement principle is essential when comparing people who have the same diagnosis but different jobs, commuting arrangements, assistive technology, caregiving obligations, and employer accommodations. Income insurance contracts address a legally defined earnings-loss contingency; the functioning classification describes domains of health and lived activity. The two can be linked analytically while retaining their different evidentiary purposes.
Disability and employment statistics also require population discipline. The OECD's comparative disability and work report describes disability across a wider working-age population than disability-insurance beneficiaries. The International Labour Organization's disability social protection assessment measures benefit coverage among people with severe disabilities. The U.S. Social Security Administration's annual disability insurance statistical report concerns particular insured-program beneficiaries, applications, and termination processes. These denominators cannot be substituted for one another. A population disability employment gap, an insurance take-up rate, an allowance probability, and an exit-to-work rate are four distinct objects.
This dossier uses an international documentary and critical empirical method. Its evidentiary material comprises published causal studies, original program statutes and administrative guidance, international statistical and legal instruments, and institutional evaluations. It does not create a synthetic individual-level dataset or estimate a new causal coefficient. Numerical examples below are explicitly hypothetical, with their assumptions and arithmetic disclosed. Where full journal articles or datasets were not inspected, claims are confined to official abstracts, authors' research summaries, public statutes, or published institutional tables. The research was structured around three questions: which sources identify behavior and entitlement; how causal effects can be separated from selection; and which changes in employment, earnings, health, and security are observed for different populations.
The analytical method has five stages. It first fixes the insured event and the eligible population. It then records the contractual or public-policy treatment whose consequences are under examination, including waiting periods and benefit interactions. Third, it chooses an estimand and considers whether assignment, a legal discontinuity, or a credible comparison identifies it. Fourth, it maps reported outcomes and time horizons without equating participation, earnings, employment, or benefit exit. Finally, it assesses what the finding can support outside the original jurisdiction and claimant group. This order prevents a country-specific empirical estimate from silently becoming a universal feature of disability insurance.
Historical and Conceptual Foundations
The insured risk and the relevant populations
Disability insurance is an institutional response to the stochastic loss of earned income following a health-related impairment. For a privately insured claimant, the object is normally a contractual promise defined by occupation, income, duration, elimination period, and exclusions. For a public social insurance claimant, entitlement depends on legislative definitions and administrative criteria, including insured status, relevant work and earnings tests, and disability determination. Employer sick pay and workers' compensation form adjacent systems with different causal mechanisms and legal purposes. A program of personal care or long-term services protects needs that may coincide with work disability without being an income-replacement contract.
The relevant risk sequence includes health shock; persistence or recovery; job retention or separation; claim application; screening; allowance or denial; benefit payment; available care and supports; any attempted return to work; changes in hours or earnings; and either stable employment, intermittent work, continued support, or renewed claim. The timing is substantively important. A person can retain a job while receiving partial benefits, lose a job while awaiting adjudication, work while benefits continue, or leave benefits without obtaining employment. The ICF environmental-factor approach explains why an identical medical limitation can produce different employment outcomes depending on the tasks, access arrangements, and social environment.
A defensible evaluation defines at least the following populations in advance: all working-age people with a relevant health condition; people satisfying a functional survey definition; insured workers; claimants who file; claimants initially approved; appellants; long-term recipients; people offered reintegration services; and people actually receiving those services. For example, the causal effect of initial SSDI award on marginal applicants pertains to individuals whose allowance can be shifted by examiner assignment. It does not identify the effect of rehabilitation among current long-term recipients, because eligibility margins and later service selection differ. Likewise, self-reported disability measured in the EU labor force survey encompasses people who never made an insurance claim.
Historical development of disability income protection
Industrial wage employment exposed households to concentrated losses when a breadwinner became unable to work. Mutual aid, employer provisions, state sickness benefits, accident compensation, occupational pensions, private disability income contracts, and public disability insurance gradually produced overlapping forms of protection. Their incentives reflect different insured risks. Occupational accident systems attach liability or coverage to work causation; sickness insurance may begin shortly after an episode; long-term disability income protection often follows prolonged incapacity and evidence assessment; and general means-tested income support addresses low household resources.
The historical expansion of U.S. disability income programs illustrates the importance of institutions and labor demand. Autor and Duggan's analysis of rising disability rolls links disability program participation in the late twentieth century to screening stringency, changes in the relative value of benefits, and declining employment prospects for less-educated workers. The work cautions against treating the number of beneficiaries as a mechanical measure of impairment prevalence. An increase in applications can reflect layoffs and wage declines alongside medical need. A fall in unemployment can partly coexist with a rise in labor-force withdrawal to disability programs, which changes measured labor-market categories without resolving underlying employment difficulties.
The U.S. statutory design is not a generic template for every jurisdiction. Section 223 of the Social Security Act's disability insurance provisions establishes entitlement and work-related legal rules in a specific federal insurance system. By contrast, in the Netherlands, employer reintegration obligations require engagement with the return-to-work process during a period of sick-pay responsibility before the long-term work-capacity assessment. Norway's sick-leave follow-up requirements provide a further example of employer planning obligations. National institutional sequences determine which actor has responsibility while employment attachment remains intact and which event triggers long-term payment.
Functional disability, work capacity, and contractual eligibility
The word disability refers to different variables in medicine, rights law, labor statistics, and insurance. An activity limitation may be measured by self-report. An insurer may evaluate capacity for the claimant's own occupation or any suitable occupation, according to precise contractual wording. A statutory authority may assess ability to perform substantial gainful activity or a locally defined work-capacity classification. The decision to allow a claim is an administrative outcome, not an unbiased laboratory measurement of health, and it may depend on procedural standards and evidence.
Consider three people with a comparable mobility limitation. One has remote professional work with accessible software and flexible hours; another has a physically demanding job with no feasible modified duties; a third has intermittent employment and a period without insured earnings. Their functional classification may overlap while their disability-income entitlement and employment paths differ sharply. This example is analytical, not evidence about any observed sample. It shows why comparisons must separate medical information, occupational demands, contracted definitions, and employment opportunity.
The composition of claims matters as well. Physical impairments, mental health conditions, episodic illness, multiple chronic conditions, and progressive disease can produce different patterns of limitation and service needs. Severity should not be proxied only by a diagnostic label. Employer wage offers, education, local labor demand, and household responsibilities can interact with functional limitations. The OECD's evidence on disability inclusion documents large differences in work participation and points to early intervention, skills, and broad labor-market policy rather than a single screening rule.
Disability rights and the normative context of reintegration
Claimant reintegration is a process of securing, retaining, resuming, or adjusting work consistent with a person's capacities, choices, and access needs. It includes job retention and changed hours, occupation, or accommodations, as well as return to a former employer. The United Nations Convention on the Rights of Persons with Disabilities establishes in Article 27 on work and employment a right to opportunity in freely chosen or accepted work, with attention to nondiscrimination, accommodation, vocational support, job retention, and return to employment. This source supplies a normative boundary for policy analysis. A program that removes financial benefits through administrative termination without improving access to suitable employment cannot be counted as an inclusive labor-market success simply because caseload expenditure fell.
The International Labour Organization's Vocational Rehabilitation Convention No. 159 describes vocational rehabilitation in terms of enabling people to obtain, retain, and advance in suitable employment. The particular legal force of the convention depends on ratification and national implementation. Its conceptual contribution is broader: reintegration includes retention and progression, which a first-day return indicator misses. Work can benefit income, identity, social relations, and participation when it is sustainable and accessible, while a premature return to unsuitable work can worsen health or produce repeated absence. The scientific task is to measure these outcomes rather than assuming them.
Labor Supply and Benefit Incentives
Budget constraints, replacement rates, and benefit loss
Economic labor supply depends on the relationship between earnings, program payments, taxes, required work expenses, effort, health constraints, and the value of time outside paid employment. Disability insurance changes the opportunity set by replacing some income when covered incapacity occurs. Let wage earnings be w(h), where h denotes feasible work hours, benefit receipt be B(h,s,t), taxes be T, and work-related disability costs be C(h,s), with s representing health state and t time since entitlement. A simple accounting identity is disposable resources R = w(h) + B(h,s,t) - T(w,B) - C(h,s). This is a descriptive framework, not an estimated structural labor-supply model; it reveals that the effects of a benefit rule cannot be evaluated from the nominal replacement percentage alone.
An abrupt earnings threshold can produce a discontinuity: increasing earnings just beyond that limit may reduce the benefit sufficiently to lower disposable resources. A gradual offset changes the effective return to an additional dollar of earnings. A protected trial-work interval changes the timing and perceived reversibility of risk. Yet the labor-supply effect still depends on whether suitable employment is offered, the person's health permits the extra work, and the claimant knows and trusts the rules. Income security may also finance treatment, transport, assistive equipment, or housing stability that increases the feasibility of subsequent work. Accordingly, a positive benefits effect on near-term consumption can coexist with either direction of longer-term employment response.
The U.S. trial work period regulation permits qualifying beneficiaries to test work in nine service months without those services being used to terminate disability during the trial. SSA's 2026 return-to-work guidance reports a trial-work monthly earnings trigger of $1,210 and a post-trial substantial-gainful-activity threshold of $1,690 for a nonblind beneficiary, with a different higher threshold for statutory blindness. These annual amounts are indexed program rules, not fixed structural parameters of disability insurance. The Social Security Administration's work-incentive explanation also describes access to employment services, protection of medical coverage in specified circumstances, and mechanisms intended to reduce the financial risk of experimenting with work.
Threshold effects should be interpreted at the relevant margin. A claimant offered flexible, well-compensated work may be able to cross a threshold and remain employed. A claimant offered irregular low-hours work may encounter fluctuating earnings and unpredictable requalification. A fixed threshold applied to gross earnings can produce different net-resource changes for the two. Evaluations should record deductions, partial benefits, health coverage, household transfers, and work expenses, not solely wages and cash benefits. Every benefit rule also has a time dimension: an early work attempt under the trial period is a different treatment from resumed work after a long spell of benefit receipt.
Selection, timing, expectations, and work attachment
The choice to apply for disability insurance often follows a sequence of illness, reduction in working hours, job separation, exhausted sick pay, medical treatment, and reduced income. At that point, application behavior is endogenous to health severity, anticipated recovery, and labor-market opportunity. Selection into the claimant pool must therefore be modeled before estimating the effect of benefit receipt. A simple comparison of recipients with workers who never applied typically confounds program participation with pre-existing functional and employment differences.
Bound's foundational study of rejected DI applicants examined survey evidence showing that fewer than half of rejected male applicants worked, with low earnings among those employed. The result undermines a counterfactual in which all denied claimants can immediately obtain ordinary employment. The finding does not establish that benefits have no behavioral effect. It establishes that the observed disability applicant population faces major employment limitations independent of ultimate allowance. Modern assignment designs examine a narrower, causally interpretable question by comparing applicants whose decisions were shifted by examiner or judge variation.
Expectations may matter before decisions are made. A claimant who anticipates an adversarial assessment can delay work resumption for fear that earnings will be taken as evidence against an application. Another may preserve a job through graded duties because temporary benefits and employer accommodations are jointly available. Waiting periods and lengthy appeals can lead to skill depreciation, loss of employer attachment, reduced confidence, and economic insecurity. The French and Song appeals study documents substantial subsequent reapplications and awards following an initial administrative-law-judge denial, illustrating that appeal-stage treatment is embedded within a continuing claim process.
Long-duration dependence also changes the meaning of an employment coefficient. An effect measured three years after an appeal is a combined result of initial decision, subsequent applications, employment attempts, and later eligibility changes. If researchers classify all denied applicants as permanently untreated, estimated contrasts can be attenuated or otherwise distorted by crossover. A rigorous design therefore distinguishes initial assignment, actual receipt over time, administrative outcomes, and earned income in each follow-up interval.
Partial capacity, episodic impairment, and benefit design
Many impairments produce partial or fluctuating work capacity. A full-versus-zero binary benefit creates a mismatch if the person's realistic option is fewer hours, less physically demanding work, modified tasks, or a different occupation. Residual earnings benefits in some private products and partial-capacity public arrangements can help preserve job attachment. Their impact depends on formulas, reporting burdens, offsets with other benefits, and the predictability of reassessment. The existence of a partial-benefit provision is not proof of its effectiveness; evaluations require actual eligibility and claims records.
A useful distinction is between capacity to perform a specific duty on a particular day and sustainable work over months. A claimant may complete a work trial and then return to medical leave. That sequence represents a meaningful attempt, but it does not satisfy a one-year employment-retention criterion. Episodic conditions make spells, interruptions, fatigue, symptom variability, and time-to-recurrence central outcome variables. Counting any observed earnings in a calendar year as a permanent return exaggerates reintegration when work episodes are short or recurrent.
The insurer faces a design problem because both severe disability risk and labor-market behavior affect claims durations. Encouraging supported partial work can reduce some insured wage loss, but an overly rapid assessment may increase appeals, health deterioration, or repeated claims. Private insurer records generally observe covered claimants and insured-contract measures rather than the full universe of disabled workers, and an insurer's claim closure may occur because of policy expiration, a definition change, death, retirement, income offsets, or occupational requalification. Such closure must be classified by cause before interpreting it as return to work. The Society of Actuaries' group long-term disability incidence study is an example of actuarial experience research whose incidence denominators and product scope should remain distinct from longitudinal causal measures of claimant employment.
Household and employer responses
Income protection changes household risks as well as claimant incentives. A disability benefit can stabilize housing expenditure, reduce emergency borrowing, and alter a partner's labor supply; the change in household welfare is consequently broader than the recipient's own earnings. The relevant outcome can be net household resources, consumption volatility, hours of unpaid care, or financial distress. Household effects cannot be inferred from beneficiary head counts alone. If an additional family member starts paid work in response to a claim, aggregate labor supply can rise while the claimant's hours fall; the individual and household signs need not match.
Employers affect the feasible set of work options through task design, schedule flexibility, remote work, occupational rehabilitation, job redesign, and the decision to terminate or preserve a contract. In the Netherlands, the official reintegration obligations for employers describe wage-continuation duties and a process lasting up to 104 weeks before assessment for certain longer-term incapacity benefits. This changes the timing of employer incentives compared with systems where disability determination chiefly follows job loss. In Norway, official workplace adaptation guidance requires early follow-up planning and describes changes to hours, work organization, equipment, and duties.
Employer retention should be measured separately from rehiring after separation. The former preserves firm-specific knowledge and often the worker's original pay scale; the latter may involve search, retraining, lower wages, and risk of mismatch. A labor-supply response that changes hours at the same employer cannot simply be compared with an employment-transition rate among people who already lost jobs. Claims adjudicators, vocational counselors, workers, employers, and healthcare providers occupy different positions in that sequence; treatment attribution requires recording who delivered which service and when.
Empirical Identification Strategy
Potential outcomes and the target estimand
The empirical question must specify a treatment and a population. Let D indicate actual award of a specified disability benefit, and let Y_t(d) be the employment outcome at follow-up time t that would occur under treatment state d. The individual causal effect Y_t(1) - Y_t(0) cannot ordinarily be observed because the same claimant is not simultaneously awarded and denied under identical circumstances. The average treatment effect across applicants is conceptually different from the effect for applicants whose decisions were changed by variation in decision-maker stringency. The latter is a local average treatment effect, or LATE, for a set of compliers defined by the research instrument.
A useful analysis begins with the treatment definition. "Disability receipt" may denote initial legal allowance, at least one payment, any payment in a year, continuing receipt, a statutory entitlement, or an income replacement of given size. These should not be treated as equivalent interventions. Moreover, an initial allowance changes expected resources, potential health-care coverage, decisions about appeals, and future labor-market choices, so the estimated effect is generally a composite policy response rather than an isolated income effect. The treatment should also be located on a timeline: allowance at an examiner stage differs from allowance by an appeals judge after protracted unsuccessful applications.
For a simple binary employment outcome, a causal estimate of minus 0.28 indicates a 28-percentage-point fall in the probability of employment within the studied compliers and the specified follow-up window. It does not say every recipient reduces employment by 28 percent. The percentage-point versus percent distinction matters, especially if the baseline probability of employment is low. If the denied group's hypothetical employment probability were 0.45, an absolute 0.28 reduction would imply 0.17 for the treated group, while a 28 percent relative reduction would imply 0.324. Those figures are explanatory arithmetic, not reported subgroup outcomes from the studies.
Potential-outcome language also requires a careful choice of event time. Follow-up at 12, 36, or 60 months can include different employment pathways and policy changes. Researchers may estimate any paid work, earnings above a statutory substantial-gainful-activity threshold, hours worked, average annual earnings, or cumulative earnings. These outcomes capture different behavioral and economic margins. A participant returning for several hours a week might count as employed while earning less than the substantial gainful activity level. A participant with no paid work but substantial unpaid caregiving could be classified as out of the labor force without being inactive.
Examiner assignment and appeals-judge assignment
Maestas, Mullen, and Strand's examiner-assignment causal study uses variation in disability examiners' allowance propensities as an instrument for benefit receipt among U.S. SSDI applicants. The published AER abstract reports that about 23 percent of applicants were at the margin of program entry and that employment among this group would have been 28 percentage points higher in the absence of receipt. It further describes strong heterogeneity: the estimated employment effect is negligible for people with relatively more severe impairments and can be much larger for the less severely impaired within the affected margin. These claims refer to a specified historical administrative sample and local variation in decisions; they should not be applied as direct estimates for all existing SSDI beneficiaries.
The design is informative because the actual allowance decision is correlated with impairment severity and other factors that are hard to measure. Examiner assignment can create plausibly exogenous differences in allowance probability after conditioning on the administrative assignment process. The key assumptions are that examiner stringency is relevant to receipt, that assignment is as-good-as-random within the relevant offices and periods, that a more lenient examiner does not systematically decrease awards for otherwise identical claimants, and that examiner identity affects later work through the receipt pathway rather than through unmodeled direct channels. These restrictions require institutional evidence and empirical tests. For instance, variation in processing speed or referral to other services could threaten an exclusion restriction if it affects employment independently of awards.
French and Song's appeals-judge assignment analysis estimates the effect of benefit receipt using effectively random assignment of disability appeals to judges. The American Economic Association abstract reports a 26-percentage-point reduction in labor-force participation three years after determination, with smaller estimates for older people, college graduates, and those with mental illness. It also reports that more than 60 percent of individuals denied by an administrative law judge obtained benefits within ten subsequent years. The long-run crossover matters: the contrast of people initially denied and initially approved does not preserve untreated status indefinitely.
The two instrumental-variable results address different decision stages and different complier groups. Even when the point estimates are numerically close, their agreement does not establish a common universal response. Examiner-stage marginal applicants and appeals-stage marginal appellants differ in their health trajectories, duration without work, legal representation, prior denial history, and possible selection into appeal. A robust interpretation identifies the two as evidence that awards can reduce paid employment for specific marginal groups under these U.S. program conditions while distinguishing the employment gains hypothetically possible among denials from the actual quality and accessibility of jobs available to those individuals.
Instrument strength also matters. A comparison based on a small difference in average allowance rates may generate imprecise estimates and sensitivity to how examiner or judge effects are computed. Best practice uses leave-one-out or similarly protected stringency measures when appropriate, geographic and time controls consistent with assignment, specification checks for balance on pre-treatment characteristics, and interval estimates. A published abstract cannot substitute for re-estimating original microdata; this dossier accordingly attributes the reported results to the authors and restricts its claims to their published descriptions.
Screening experiments, policy changes, and external validity
De Jong, Lindeboom, and van der Klaauw's Dutch screening experiment studies intensified screening of disability insurance applications in two of 26 Dutch regions. The published article reports reductions in long-term sickness absence and disability applications, evidence of earlier work resumption and self-screening, and no detected spillover into unemployment insurance inflow in the studied design. The intervention differs fundamentally from an award-versus-denial instrument: stricter screening can change expectations and behavior before an application is made, while administrative assignment designs identify behavior among people whose receipt was changed after application.
The Dutch study invites two linked but distinct causal questions. First, do screening obligations change employer and employee actions during sick leave? Second, does the screening rule change which people enter long-term disability programs? A decrease in approved claims can be produced by faster recovery or work accommodation, deterrence of potentially eligible applicants, shifts to other income programs, or administrative reclassification. The article's experimental setting supports more credible inference for its particular intervention than an unadjusted before-and-after caseload trend, but its institutional specifics still limit transfer to countries without comparable employer duties or health-service arrangements.
Policy discontinuities can supply additional research opportunities. Statutory ages, earnings thresholds, benefit duration limits, employer-size rules, and eligibility contribution requirements sometimes create regression-discontinuity or difference-in-differences designs. Each requires careful assessment of sorting, anticipation, simultaneous changes, and whether treated and comparison groups had parallel trends before the reform. A statutory discontinuity is credible only if other outcome-relevant determinants do not jump at the same threshold. An income test often encourages bunching or strategic timing, so manipulation around a threshold must be investigated rather than assumed absent.
External validity should be assessed along several axes: health severity; occupation and skill; claim stage; demographic composition; calendar period; local wage offers; alternative social protection; public or private coverage; and availability of graded work. A cross-national reading of the OECD disability work evidence makes clear that employment gaps and activation institutions differ substantially. An IV estimate obtained for marginal U.S. awards does not identify the effect of a Dutch employer reintegration obligation or of a private short-term income product. Comparative synthesis can identify mechanisms likely to recur, but their coefficients must be re-estimated or appropriately transported with transparent assumptions.
Dynamic selection, duration, and competing events
Employment reintegration is a multi-state process rather than a single transition. Define states such as employed with full duties, employed with modified duties, sick leave while attached to an employer, unemployed and seeking work, out of labor force, receiving full disability benefits, receiving partial benefits, vocational rehabilitation, renewed medical leave, and death or retirement. Some states can coexist; benefit receipt and employment are often parallel dimensions. A person may move repeatedly between work and benefit receipt as health and job circumstances change. A credible dataset therefore needs time-indexed employment, earnings, entitlement, program services, and health histories.
A first return-to-work hazard h(t) summarizes the probability of an initial return in a time interval conditional on having remained without that return until t. A survival curve for time to first work can help compare timing, but its denominator contracts as people return, die, retire, or otherwise leave observation. Competing-risk methods are necessary when alternative absorbing outcomes prevent a work return; recurrent-event methods are needed when relapse and re-entry matter. Censoring at policy expiration or administrative closure may be informative because it can depend on health, income, or work capacity.
Dynamic selection creates a particularly serious challenge for comparisons of program participants. People who engage with employment services may be healthier, more motivated, more informed, or face a different set of jobs than those who do not. A raw difference in employment after services cannot be described as their causal impact. O'Leary and Roessel's Ticket to Work evaluation explicitly considers participation bias while assessing employment outcomes, disability payment savings, and program costs. The authors find improved outcomes and net payment savings for many clients, while aggregate savings did not completely cover program costs in their evaluation. Their stated selection corrections do not convert the evaluation into a randomized trial.
An appropriate longitudinal evaluation pre-specifies administrative start date, episodes of treatment, and censoring rules. A claimant's eligibility for vocational services may begin well after the initial insurance award; starting follow-up at service entry for participants while starting at claim entry for controls would create immortal-time bias. The study should establish a comparable index date for both groups or use a design that handles time-varying service allocation. Time since health shock and time since application are separate clocks. Their conflation can conceal whether early intervention prevents job separation or whether late intervention improves transitions after separation.
Employment and Health Outcomes
U.S. causal findings and their boundaries
The U.S. empirical literature supplies strong evidence that disability insurance receipt can change observed work among applicants with marginal eligibility decisions. In the 2013 examiner study, the estimated counterfactual employment effect among compliers is substantial. In the 2014 appeals study, an effect is visible three years after a decision. These causal results should be read alongside the descriptive evidence in Bound's analysis of denied applicants' earnings, which documents low employment and earnings even without award. Together they yield an important interpretation: among some applicants a benefit changes labor supply materially, while a large part of nonemployment reflects underlying health and labor-market barriers.
The results also distinguish extensive from intensive labor supply. Participation captures whether someone enters or remains in the labor force; employment captures work status; hours and earnings capture intensity and remuneration. An IV finding for employment cannot be automatically converted into a wage elasticity, fiscal multiplier, or welfare effect. A claimant with a part-time accommodated job may have higher health-related spending and lower gross earnings than before impairment even while employment participation increases. Likewise, a return to work during the first year that is followed by repeated absences may have a different welfare value from sustained work over several years.
Administrative program data place the causal findings in an institutional context. The 2025 SSA disability report was released in October 2026 and records more than 8.5 million disabled beneficiaries, including worker beneficiaries and other eligible categories, receiving benefits in December 2025. The agency's disabled-worker statistics show 7,125,880 disabled-worker beneficiaries in its reported 2025 series. These figures describe different population aggregations and should not be presented as contradictory. Program stock depends on incidence, awards, mortality, recovery, retirement transitions, and other exits; declining stock is not itself evidence of claimant reintegration.
The agency's statistical tables about trial-work participation and earnings-related terminations are administratively valuable but require reading of processing lag and revisions. In the 2025 program statistical report, work-related continuing disability review counts are subject to later revision, and some recent-year data are incomplete at publication. For an evaluation, a work-incentive count should be divided by a defined eligible population and linked to follow-up records. Comparing annual participation counts across years without adjusting for revised records, changes in the beneficiary population, and economic conditions could produce misleading trends.
International participation and employment evidence
Population evidence situates claims behavior within broader employment exclusion. The OECD's 2022 comparative report describes an employment gap of approximately 27 percentage points in 2019 between people with and without disability across its included set of 32 countries. The same publication reports that roughly one in four people with high support requirements and one in two with moderate support requirements had a job. These statements characterize the OECD's sampled working-age populations and functional categories. They do not constitute causal estimates of the effect of insurance receipt.
Eurostat's 2025 social scoreboard reports a disability employment gap of 24.2 percentage points for the EU in 2025. The European Commission's Employment and Social Developments analysis reports an approximately 24-percentage-point gap for 2024 and relates that gap to work participation differences and severity. The years and disability measures must remain attached to the reported estimates. Their broadly similar magnitude to selected U.S. treatment-effect coefficients must not be interpreted as the same causal quantity. Population employment gaps combine severity, barriers, discrimination, education, institutional incentives, job availability, and all other differences correlated with disability status.
Coverage determines the geographic meaning of insurance research. The ILO's global disability-benefit coverage findings estimate that 38.9 percent of people with severe disabilities globally receive a disability benefit, with very large regional variation. Because this figure measures benefit coverage in a specified severity population rather than work capacity among insured claimants, it identifies a first-order access problem: many people are outside benefit systems altogether. Studying the incentives faced by the covered group alone can therefore overlook the workers with the least income protection and limited ability to invest in rehabilitation.
Cross-country rankings depend on whether disability is measured by survey activity limitations, clinical diagnosis, benefit receipt, or a national register. Differences in disability prevalence, labor-force denominators, employment age bands, and survey nonresponse alter cross-country estimates. EU data on self-reported activity limitation should not be merged directly with U.S. SSA eligibility data. Even within the same dataset, working-age definitions of 15–64 versus 20–64 can affect reported gaps. Comparative tables should retain each source's reference year, territory, population definition, and exact indicator.
Reintegration interventions and work quality
Reintegration involves the organization of work as well as incentives in the benefit system. The Dutch official employer reintegration framework places a structured responsibility on the employer during prolonged sickness absence. Norwegian follow-up and adaptation requirements explicitly discuss planning, dialogue, and modifications to duties or equipment. These instruments demonstrate how early workplace processes can precede a final disability insurance determination. Whether they raise sustainable employment depends on compliance, labor demand, appropriateness of tasks, and the claimant's health status.
In the United States, SSA's Ticket to Work and work incentives emphasize support services and protection during experimentation with work. The 2023 Ticket program assessment finds improvements for many service recipients but a mixed fiscal balance. Its fiscal measure and employment measure are different outputs. A program can improve participants' work prospects without paying for its total administrative and service costs from benefit reductions; whether it is socially worthwhile depends additionally on participant earnings, health, autonomy, tax revenue, and the counterfactual program costs.
Sustainable work requires more than a positive earnings entry. Evaluations should report job retention at 6, 12, and 24 months; weekly paid hours; earnings levels and volatility; access to benefits; workplace accommodations; perceived safety; occupational match; health changes; repeated sickness absence; and claimant-defined work goals. Work quality includes adequate pay, control over hours, discrimination exposure, and the practical availability of modifications. A return to unsafe or involuntary work can generate negative welfare even when it reduces insured claim duration. Conversely, reduced hours in a medically suitable role can be a successful reintegration outcome if it preserves health and stable resources.
The UN disability employment guarantee and ILO vocational rehabilitation principles provide a clear interpretation of suitable work as accessible, appropriately supported, and consistent with human rights. They do not imply that all people are medically able to return or that each claimant must accept any available job. For research, autonomy should be observable through stated preferences and voluntary participation where feasible; legal protections and practical access barriers should be explicitly recorded rather than inferred from a claimant's final labor-market classification.
Health outcomes and welfare measurement
Employment and health are jointly determined. Better health can make work possible, while suitable employment can improve material security and social participation. The direction of causality is difficult to isolate because the work transition occurs along with treatment, rehabilitation, symptom change, and family circumstances. Observational correlations between return to work and improved health may reflect the fact that healthier claimants returned. Conversely, a rise in sick leave after return could indicate worsening health, successful access to healthcare, or better administrative reporting rather than a simple deterioration caused by employment.
The WHO's Rehabilitation 2030 agenda emphasizes integration of rehabilitation into health systems and equitable access to assistive products and services. This supports a study design that records the timing, intensity, and availability of health and rehabilitation services while following employment. The clinical treatment effect should not be assumed identical to the labor-supply effect of benefit entitlement. Health outcomes might include functional change, symptom burden, self-rated health, clinically relevant assessments, and the incidence of hospitalization or relapse, depending on ethically accessible data.
A welfare analysis asks whose outcomes matter and how they are valued. The claimant values security, health, feasible work, and life plans. The household experiences consumption and informal care changes. The employer experiences productivity, replacement and accommodation costs, and retention of skills. The insurer or public program incurs benefits and administrative costs; taxpayers bear or receive changes in financing. Because these elements are distributed across actors and time, a reduction in benefit payments is only one component of the social balance. The ILO's disability social protection discussion highlights income security, disability-related extra costs, healthcare, and rehabilitation as connected dimensions of inclusion.
A scientifically credible evaluation does not assign a positive value to every observed work transition or a negative value to every benefit continuation. If impairment is severe and durable, income protection can be a successful insurance outcome without labor-market return. If a person with partial capacity secures a stable accommodated role, continuation of a residual benefit can be consistent with successful work and income protection together. The proper outcome is a joint profile of employment, health, resources, and autonomy, measured relative to a justified counterfactual and followed long enough to identify relapses or durable transitions.
Primary Source and Model Validation
Original data, measurement, and source lineage
The primary unit of observation must be defined before statistical interpretation. A claimant record can represent an application, a person, a benefit episode, a contract, an employer absence spell, or a monthly payment. The same person can have multiple applications, concurrent programs, and repeated spells. Without unique identifiers and a documented linkage procedure, the apparent count of reintegrated claimants may include duplicate events. Administrative insurance data rarely identify the entire working-age population; surveys may identify disability broadly while lacking verified entitlement details. Linking the two can enrich analysis but creates selection and privacy concerns.
The most useful design records the date and source of health information, job status before the shock, occupation, covered earnings, contract or statutory eligibility, application and decision dates, trial work, rehabilitation services, benefit amounts, other public transfers, reported earnings, and later employment status. The data dictionary must distinguish scheduled benefit from actually paid benefit; gross earnings from take-home resources; a job contract from paid hours; employer retention from employment at a new firm; and temporary program suspension from final termination. Ideally, outcome observation continues after the last insurance payment to distinguish durable reintegration from administrative closure.
Official administrative source lineage illustrates the point. The Social Security Administration's 2025 insurance statistics give program-level counts with particular administrative dates and definitions. Its disabled-worker statistical series uses a different presentation of specific worker beneficiaries. The OECD's comparative disability measurements arise from harmonized but still country-specific population surveys. Eurostat's EU disability employment dashboard records a disability employment gap with a defined survey population. These are source-specific datasets, not interchangeable measures of the same unit.
A reproducible source register should record for each publication or administrative instrument: issuing authority or authors, title, date and version, stable URL or DOI, country, population, reference period, measurement definitions, observed variables, exact claims verified, access scope, and possible revisions. The accompanying EVIDENCE.json performs that source-to-claim function for this documentary monograph. It should not be confused with access to underlying confidential claimant microdata. The empirical results summarized here come from the cited authors, not from reanalysis of their datasets.
Jurisdictional and contractual verification
Law must be checked in its appropriate temporal and territorial scope. A U.S. SSDI trial-work rule is created within Social Security law and implementing regulations. The statutory disability insurance provisions, the trial work federal regulation, and SSA's current program guidance provide distinct layers of authority and explanation. A legal exposition should distinguish statute, regulation, and public-facing summary, particularly when annual monetary thresholds change. For example, a threshold stated for 2026 must not be silently generalized to earlier or later years.
The British Universal Credit system provides another instructive boundary. The government's health condition and disability guide explains Work Capability Assessment and the relationship between work requirements, health information, and claimant commitments in its 2026 updated version. Universal Credit is a means-tested income system, so its labor-market incentives cannot simply be equated with U.S. insured SSDI or a private disability income policy. The policy's eligibility categories, household assessment, and work allowances matter for interpretation.
The Dutch official reintegration rules describe employer wage continuation and actions during the first two years of illness. Norwegian return-to-work planning duties similarly define institutional responsibilities. The two countries' employer requirements demonstrate different routes by which job attachment can be maintained before long-term benefit entitlement. They do not automatically establish that a particular rehabilitation service caused a particular claimant to resume work. That causal question requires design-specific data.
Private disability coverage further demands a contract-by-contract evidentiary method. Relevant clauses include definition of covered disability, own-occupation and any-occupation transitions, pre-existing-condition exclusions, partial or residual disability provisions, rehabilitation participation requirements, income offsets, elimination periods, proof-of-loss rules, benefit termination, and review or appeal mechanisms. A group experience study such as the SOA's long-term disability incidence research can help identify actuarial population and coverage definitions, but its aggregate experience is not a ready-made experiment in employment behavior. Contract versions and years must be recorded before combining private claims.
Confounding, sensitivity, and missingness
The principal confounders include baseline health severity, functional limitations, anticipated disease trajectory, education, occupation, local wages, employer accommodations, previous employment, household wealth, other transfers, age, and prior benefit applications. Several are measured imperfectly. Prior earnings provide information about labor-market attachment but are also a determinant of benefit calculation. Diagnosis can proxy medical risk without capturing work capacity. Employer accommodation can be both a cause of work retention and a response to the same health condition that triggers claims. These relationships require a causal graph or explicit temporal argument in any econometric analysis.
Random examiner or judge assignment is valuable when its institutional premise is supported, yet a publication-level identification claim must still be examined for balance and validity. An instrument based on administrative stringency assumes that otherwise comparable cases face different award probabilities because of assignment rather than unmeasured severity. Exclusion can fail if decision makers also affect delay, claimant information, or other channels relevant to work. Monotonicity may fail if case types receive different treatment patterns across examiners. Estimates should report first-stage relevance, reduced-form effects, instrument diagnostics, and confidence intervals rather than a single point estimate.
The Dutch screening-study results are informative about effects in a particular regional experiment. A comparison of 2 treated and 24 other regions still requires attention to region-level characteristics, spillovers, common shocks, employer compliance, and the limited number of treated clusters. Publicly accessible abstracts describe the principal results but do not permit this dossier to replicate every statistical test. As a result, the conclusions here concern the published evidence and theoretical interpretation rather than an independently reproduced estimate.
Attrition and administrative missingness can also generate selection. If earnings data omit informal work or cross-border employment, observed work can be understated. If self-reported health is collected only from responders, people with severe conditions or unstable housing may be underrepresented. If deceased or retired beneficiaries are counted as never returning to work without a competing-risk treatment, a longer follow-up can make reintegration appear less frequent. If disability reassessment is incomplete, recent earnings-triggered exits can be undercounted, as the SSA's statistical reporting documentation warns for some series.
Sensitivity analysis should vary employment definitions, observation windows, earnings thresholds, functional severity strata, baseline labor attachment, geographic controls, and treatment timing. The researcher should also investigate omitted outcomes that change the policy interpretation, particularly health deterioration, failed work attempts, informal care, and financial hardship. Whenever an estimate cannot support all these outcomes, the conclusion should identify the narrower estimand rather than extrapolating. A favorable employment coefficient cannot by itself certify improved health, equity, or social welfare.
Transparent worked calculations and evaluation protocol
The following computations illustrate how program design changes observed resources. All amounts and rates in the next two examples are invented for explanation. They describe no current national statute or commercial product.
In a stylized partial-benefit arrangement, suppose a claimant receives $1,800 per month while not working. Suppose eligible work brings $500 in monthly earnings and reduces the benefit by 50 cents for each dollar earned, before taxes and work-related expenses. The new benefit is $1,800 - 0.5 × $500 = $1,550. Gross resources are $1,550 + $500 = $2,050, which is $250 higher than under nonemployment. If work-related costs and taxation exceed that incremental amount, net resources can still fall. The gross gain is therefore an incomplete measure of the individual's actual financial incentive.
Consider an otherwise similar stylized benefit that terminates entirely above an earnings threshold of $400. Just below the threshold, earnings of $399 plus the $1,800 benefit imply $2,199 gross monthly resources. At $401, earnings alone would imply $401. The sharp drop is a mathematical property of this fictional cliff design; real systems often have trial periods, grace periods, partial payments, or reinstatement rules that change the outcome. The illustration shows why the complete statutory or contractual schedule must be reconstructed rather than imputing incentives from one nominal threshold.
A separate hypothetical difference-in-differences calculation uses employment rates. Assume employment in a treated group rises from 40% to 47%, while employment in a comparison group rises from 42% to 45% over the same two dates. The first change is 7 percentage points, the second is 3, and the difference-in-differences estimate is 4 percentage points. Interpreting the 4-point difference as causal would require a credible parallel-trends assumption in the absence of treatment, no relevant anticipation, stable population composition, and attention to treatment spillovers. The example is an arithmetic demonstration, not an observed program result.
Likewise, assume a reported local effect of 28 percentage points among compliers representing 23% of applicants. Multiplying 0.28 × 0.23 yields 0.0644, or 6.44 percentage points; this would equal a population average only under an additional assumption that every noncomplier has zero effect and that the study's local identification maps consistently to the desired full applicant population. Neither assumption follows from an IV design. The arithmetic illustrates why the Maestas–Mullen–Strand local estimate cannot be mechanically rebranded as a universal applicant treatment effect.
A defensible empirical protocol would specify a target trial or quasi-experimental emulation with eligibility determined from pre-treatment records, a common starting date, the treatment and alternative treatment precisely described, and a declared follow-up horizon. It would use monthly work, benefit, and health outcomes; record employer retention, service use, and failed attempts; stratify by severity, age, sex, occupation, and coverage; allow recurrent work episodes; and publish sensitivity analyses for missingness and alternative definitions. Researchers should report absolute levels and percentage-point effects, confidence intervals, event-time trajectories, and relevant denominators together. Open scientific code and nonidentifying data dictionaries can improve reproducibility without disclosing private health information.
Integrated Scientific Results and Further Questions
Joint interpretation of income security and work
The literature supports an integrated interpretation with several distinct evidence levels. Random assignment of claim evaluators and judges offers causal evidence that particular eligibility awards reduce paid work for some marginal U.S. applicants. Research on denied applicants shows that labor-market capacity is already substantially constrained in the applicant pool. Dutch screening evidence shows that administrative expectations and employer or claimant behavior can change before long-term receipt. Population statistics from the OECD and Eurostat document employment exclusion far beyond the insured population. Administrative rules describe practical work incentives and reintegration responsibilities but are not themselves evaluations of causal effectiveness. Each level contributes a different part of the explanation.
The main analytical conclusion is that disability insurance changes the timing, risk, and returns of work attempts. The direct income effect of a benefit interacts with the prospects of a suitable job, administrative review risk, service access, and the cost of repeated failure. A program with protected trial work and available accommodations gives a claimant a different opportunity set from a program with immediate benefit loss and no job-retention pathway. Yet that difference alone does not determine effectiveness. The relevant counterfactual includes the claimant's health and realistic employment opportunities, not merely a budget constraint drawn for a fully able worker.
The economic value of insurance can be realized even where no employment return follows. A person with severe permanent impairment may receive essential income protection and maintain access to health services; employment-focused performance measures should not erase that benefit. Where feasible work exists, a good reintegration outcome combines earnings and acceptable working conditions with income continuity and health stability. The ILO's global social protection assessment grounds this wider interpretation in coverage, adequacy, assistive services, and protection against disability-related extra costs.
An actuarial balance is also affected by claim duration, incidence, and the transition between full and partial benefits. If a supported work attempt reduces paid claim months without increasing relapses, an insurer may see lower future expenditure. If it produces repeated reopenings and clinical deterioration, an initially favorable closure statistic could conceal higher long-term costs. An employer may save training costs by retaining a worker on modified duties. The claimant may prioritize income stability or health even when a narrow fiscal measure favors a different action. The appropriate evaluation records the distribution of consequences across these actors and time.
Heterogeneity, institutional design, and distribution
Severity is central but does not exhaust heterogeneity. The examiner-assignment evidence reports different effects by unobserved severity; the appeals-judge evidence reports smaller effects for certain demographic and impairment groups. These are study-specific subgroup results, and overlapping classifications may complicate their causal interpretation. Educational attainment, occupation, local hiring, and benefit levels can modify the work incentives faced by applicants. Gender and household care duties can alter feasible hours and access to rehabilitation services. A pooled estimate hides such differences, while an excessive number of post-hoc subgroup comparisons risks mistaking noise for real variation.
Institutional design determines which margin can respond. Early sick-leave planning can preserve employment before detachment; late vocational services primarily assist people already separated from work. Partial disability payments support graded work differently from full-income replacement rules. A secure reinstatement pathway reduces the cost of a failed attempt; uncertainty over eligibility or healthcare coverage increases perceived risk. The SSA work-incentive system provides examples of such safeguards, while Dutch employer obligations and Norwegian sick-leave planning place responsibilities at an earlier stage. Their comparisons identify relevant institutional features, not universal rankings of effectiveness.
A comparative empirical result should therefore describe policy treatment, claimant population, baseline employment, expected health trajectories, and labor-demand context in one evidentiary statement. Where data are incomplete, the document should say which dimension remains unobserved. The OECD's comparative inclusion study identifies persistent gaps and insufficient early intervention, while the CRPD employment provisions establish obligations around accessible work and reasonable accommodation. Jointly they support evaluating the availability and quality of opportunities rather than focusing solely on administrative benefit counts.
Distribution also means recognizing people outside the insured population. In countries with limited contributory coverage, many people with severe disabilities have no disability income benefit to respond to. The ILO's 38.9% global coverage statistic pertains to those with severe disabilities and a particular reporting framework; it underscores the need to analyze eligibility access and social protection before debating marginal benefit disincentives. Empirical results from high-income formal employment systems may not generalize to informal labor markets, migrant workers, agricultural households, or people without continuous contribution histories. A comparative science of disability insurance must retain these population boundaries explicitly.
A practical scientific research agenda
Future primary research should distinguish three intervention windows: prevention of job loss during early illness, the effects of benefit eligibility during adjudication, and support for reintegration after an established claim. For early intervention, credible employer- or workplace-level policy changes could estimate sustained retention and health outcomes. For adjudication, assignment designs can estimate effects of award and delay on earnings, health, and household resources. For later reintegration, randomized service offers or credible implementation variation can estimate employment, benefit, health, and cost outcomes. Each study must assess whether service take-up and follow-up observation are affected by unmeasured readiness for work.
A priority is the linked analysis of health and employment. Claims datasets often contain functional restrictions and administrative status but limited clinical outcomes; healthcare records contain diagnoses and treatment but little job information; tax records contain earnings but not the suitability of work. Secure linkage, strong data governance, and participant involvement can allow researchers to examine whether an intervention changes income protection, treatment continuity, relapse, employment, and job quality jointly. Disability itself should be operationalized according to a documented purpose, following the WHO functioning framework.
A second priority is systematic attention to claimants who attempt work but do not sustain it. Most reports emphasize first returns and benefit exits; a sequence of failed trials, repeated sickness leave, partial capacity, and renewed benefits may reveal whether incentives and workplace services genuinely match health conditions. Researchers should pre-register repeat-work and repeat-claim outcomes and document how they handle competing risks. Return-to-work surveys should not exclude people whose work experiment was unsuccessful, because that can generate survivor bias and exaggerate the apparent durability of program success.
A third priority is a genuinely comparative institutional study with aligned microdata measures. The U.S., the Netherlands, Norway, and the United Kingdom exemplify different mixes of income insurance, employer obligations, medical evaluation, and reintegration service delivery. The legal texts cited here establish selected features but do not supply harmonized causal estimates. A comparative design should begin from common event definitions and then preserve jurisdiction-specific program rules instead of forcing all systems into one undifferentiated benefit variable. Particular attention is warranted for partial-capacity awards, changes in earnings thresholds, job retention in smaller firms, and populations with episodic impairments.
A fourth priority is the relationship between claimant agency and program outcomes. Studies can examine whether information about trial work, the opportunity to choose services, accessible administrative communication, and procedural safeguards alter perceived risk and actual behavior. However, changes in satisfaction or confidence should not be substituted for employment outcomes, and work transitions should not be presumed to demonstrate voluntary choice. The ILO's vocational rehabilitation standard and the UN work and employment provisions supply the relevant rights and suitability context.
The present synthesis produces a practical research interpretation: design, evidence, and welfare criteria should follow the full claimant pathway rather than a single administratively convenient endpoint. This is an analytical conclusion derived from the reviewed studies and institutional documents. It is not presented as a newly coined theory or doctrine. Its testable implication is straightforward: studies with linked employment, benefit, health, and workplace data should find that measured reintegration success differs materially depending on whether the endpoint is first work, sustained suitable work, net household resources, or a reduction in program expenditure.
Conclusion
Disability insurance changes labor supply through its influence on income security, claim eligibility, the expected financial consequences of work, and the institutions that govern employment retention and return. The causal literature establishes meaningful work responses for some marginal U.S. disability applicants: Maestas, Mullen, and Strand (2013) report a 28-percentage-point employment counterfactual among the affected initial-award margin, and French and Song (2014) report a 26-percentage-point labor-force participation response three years after an appeals determination. These estimates apply to different groups, methods, and time horizons. Evidence on rejected applicants, screening reforms, international employment gaps, and reintegration programs supplies complementary information rather than interchangeable causal coefficients.
The documentary record also establishes that most serious questions about reintegration cannot be answered by insurance payments or administrative claim closure alone. Suitable employment depends on health trajectories, occupations, employer accommodations, accessible services, local labor demand, household resources, and protection against unsuccessful work attempts. Rights instruments require attention to freely chosen work and reasonable accommodation. International statistics describe substantial gaps in employment and benefit coverage, including among people who never reach the claimant stage.
The scientific implication is to evaluate disability insurance using multiple explicitly defined outcomes: income and benefit adequacy, employment and earnings, job retention and job quality, health, duration and recurrence, and the distribution of economic consequences among households, employers, insurers, and public programs. Causal interpretations must specify the identified population and policy margin. This evidence-linked evaluation permits rigorous comparison of benefits, screening, and reintegration measures while preserving the distinct purposes of insurance security and inclusive employment. The results provide an original synthesis of existing science and an operational agenda for further causal research, without asserting a new authorial theory.
References
Autor, David H., and Mark G. Duggan. 2003. “The Rise in the Disability Rolls and the Decline in Unemployment.” The Quarterly Journal of Economics 118 (1): 157–206. https://doi.org/10.1162/00335530360535171.
Bound, John. 1989. “The Health and Earnings of Rejected Disability Insurance Applicants.” The American Economic Review 79 (3): 482–503. Public research version: NBER Working Paper 2816, January 1989. https://www.nber.org/papers/w2816.
de Jong, Philip, Maarten Lindeboom, and Bas van der Klaauw. 2011. “Screening Disability Insurance Applications.” Journal of the European Economic Association 9 (1): 106–129. https://doi.org/10.1111/j.1542-4774.2010.01002.x.
European Commission. 2025. Employment and Social Developments in Europe 2025, chapter 2 (disability employment participation and employment gaps). https://ec.europa.eu/employment_social/employment_analysis/esde/2025/Chapter%202.html.
Eurostat. 2026 access. Social Scoreboard: Disability Employment Gap, EU 2025 reported value. European Commission. https://ec.europa.eu/eurostat/cache/dashboard/social-scoreboard/.
French, Eric, and Jae Song. 2014. “The Effect of Disability Insurance Receipt on Labor Supply.” American Economic Journal: Economic Policy 6 (2): 291–337. https://doi.org/10.1257/pol.6.2.291.
Government of the Netherlands. 2026 access. Reintegration Obligations for Employers. Business.gov.nl. https://business.gov.nl/regulations/reintegration-obligations/.
Government of the United Kingdom, Department for Work and Pensions. 2026, updated August 4. Universal Credit: Health Conditions and Disability Guide. https://www.gov.uk/government/publications/universal-credit-if-you-have-a-disability-or-health-condition-quick-guide/universal-credit-if-you-have-a-disability-or-health-condition.
International Labour Organization. 1983. Vocational Rehabilitation and Employment (Disabled Persons) Convention, 1983 (No. 159), adopted June 20, entered into force June 20, 1985. NORMLEX instrument text. https://normlex.ilo.org/dyn/nrmlx_en/f?p=NORMLEXPUB%3A12100%3A0%3A%3ANO%3A%3AP12100_ILO_CODE%3AC159.
International Labour Organization. 2024. World Social Protection Report 2024–26, chapter 4, section 4.2.5 (disability benefits and inclusion). https://webapps.ilo.org/static/english/reports/flagship/world_social_protection_report_2024-26/Chapter%2004.html.
Maestas, Nicole, Kathleen J. Mullen, and Alexander Strand. 2013. “Does Disability Insurance Receipt Discourage Work? Using Examiner Assignment to Estimate Causal Effects of SSDI Receipt.” American Economic Review 103 (5): 1797–1829. https://doi.org/10.1257/aer.103.5.1797.
Norwegian Labour Inspection Authority. 2026 access. Follow-Up of Employees on Sick Leave. Official guidance, Arbeidstilsynet. https://www.arbeidstilsynet.no/en/working-hours-and-organisation-of-work/workplace-adaptations/follow-up-of-employees-on-sick-leave/.
O'Leary, Paul, and Emily Roessel. 2023. “Effects of the Ticket to Work Program: Return on Investment and Overall Assessment of Outcomes Versus Design.” Social Security Bulletin 83 (1), online edition. https://www.ssa.gov/policy/docs/ssb/v83n1/v83n1p1.html.
Organisation for Economic Co-operation and Development (OECD). 2022. Disability, Work and Inclusion: Mainstreaming in All Policies and Practices. Paris: OECD Publishing. https://doi.org/10.1787/1eaa5e9c-en.
Society of Actuaries (SOA). 2026 access. 2015–2022 Group Long-Term Disability Incidence Study. SOA Research Institute. https://www.soa.org/resources/experience-studies/15-22-grp-ltd-inc/.
Social Security Administration (SSA). 2026. Annual Statistical Report on the Social Security Disability Insurance Program, 2025. Released October 2026. https://www.ssa.gov/policy/docs/statcomps/di_asr/2025/index.html.
Social Security Administration (SSA), Office of the Chief Actuary. 2026 access. Disabled-Worker Statistics. https://www.ssa.gov/OACT/STATS/dibStat.html.
Social Security Administration (SSA). 2026 access. Try Returning to Work Without Losing Disability. Published 2026 thresholds and trial-work guidance. https://www.ssa.gov/disability/work.
Social Security Administration (SSA). 2026 access. Work Incentives: Ticket to Work. https://www.choosework.ssa.gov/about/work-incentives.
United States Social Security Administration. 2026 access. Social Security Act, Title II, Section 223: Disability Insurance Benefit Payments. Statutory text. https://www.ssa.gov/OP_Home/ssact/title02/0223.htm.
United States Social Security Administration. 2026 access. 20 CFR § 404.1592: The Trial Work Period. Official federal regulation text. https://www.ssa.gov/OP_Home/cfr20/404/404-1592.htm.
United Nations. 2006. Convention on the Rights of Persons with Disabilities, Article 27: Work and Employment. https://www.un.org/development/desa/disabilities/convention-on-the-rights-of-persons-with-disabilities/article-27-work-and-employment.html.
World Health Organization (WHO). 2001, updated online. International Classification of Functioning, Disability and Health (ICF). Geneva: WHO. https://www.who.int/standards/classifications/international-classification-of-functioning-disability-and-health.
World Health Organization (WHO). 2017 onward. Rehabilitation 2030: A Call for Action. Geneva: WHO. https://www.who.int/initiatives/rehabilitation-2030.