Insurance by BogdanovaThe Encyclopedia of Insurance and Risk Science

RESEARCH DOSSIER

Social Insurance Gender Pension Gaps and Unpaid Care

Abstract. Gender pension gaps are the cumulative retirement-income consequences of unequal employment histories, pay, contributory coverage, care responsibilities, pension rules, and survival patterns. This monograph examines the role of unpaid childcare, care for disabled relatives and older adults, and domestic work in the formation of pension entitlements. It distinguishes the measured gap among pension recipients from differences in access to contributory pensions, the distribution of pensions across institutional pillars, and old-age poverty. The research combines OECD and Eurostat benefit statistics, ILO social-protection coverage data, United Nations time-use definitions, peer-reviewed evidence on child-related earnings penalties, and original public documentation of pension credit rules in Germany, Canada, the United Kingdom, Ireland, Sweden, Australia, and the United States. It reads European equal-treatment and work-life-balance legislation alongside international social security standards. Explicit arithmetic illustrates the separate effects of interrupted careers, post-care wage losses, credit formulas, and minimum benefits. The comparative synthesis finds that care recognition is instrument-specific: credited years, credited pensionable earnings, excluded low-earning months, pension contributions on paid leave, and residence-based guarantees address different gaps. Their outcomes depend on eligibility, take-up, indexation, funding, and labor-market institutions. The study specifies empirically testable implications without claiming an original causal estimate, new theoretical framework, or comprehensive global statute census.

Keywords: Social insurance; gender pension gap; unpaid care; pension coverage; pension adequacy; childcare; eldercare; career interruption; pension credits; child-related earnings penalties; defined benefit pensions; defined contribution pensions; social protection; old-age poverty; survivor pensions; family policy; contributory pension; time-use statistics; comparative pension law; retirement inequality.

Introduction: The Life-Course Production of Pension Inequality

A pension is a claim on resources during retirement, created under rules that distribute rights across time, people, and contingencies. Pension outcomes observed at age sixty-five or seventy often reflect decisions and institutions operating decades earlier. Labor-market entry, hours, occupations, wages, unemployment, illness, childbirth, family caregiving, formal contribution records, divorce, and the death of a partner can each affect retirement income. The gender pension gap is therefore a life-course outcome whose explanation requires reconstructing the institutions through which paid and unpaid activities become pensionable rights.

The most recent comprehensive OECD comparison, Pensions at a Glance 2025, reports an average gap of 23 percent across the OECD in 2024, down from 28 percent in 2007. The underlying gender pension gap chapter defines the measure among pension recipients aged sixty-five and above, taking the difference between the mean pensions of men and women relative to men's mean. On February 25, 2026, Eurostat separately reported that women's average pensions in the EU in 2024 were 24.5 percent lower than men's and that the corresponding median gap was 24.9 percent, using its own statistical universe and 2024 pension data. Similar magnitudes do not make the two aggregates identical or interchangeable.

The ILO's global old-age protection assessment supplies a second, distinct perspective. It estimates that 79.6 percent of people above statutory retirement age receive an old-age pension of some kind, yet contributory pension coverage is 49.2 percent among older women and 63.2 percent among older men. These statistics show why analysis must address entitlement and access as well as average payment levels. A person with no pension is missing from a pension-recipient mean even though that exclusion is central to the social-protection problem.

Unpaid care connects these outcomes to the organization of family and public services. Caring for children or dependent adults may displace work, reduce hours, redirect occupational choices, and limit the ability to accumulate retirement contributions. Policies governing childcare and long-term care provision affect these decisions, while pension formulas convert them into deferred claims. The ILO's Care at Work study documents legal variations in care leave and services across 185 countries. Labor-market studies identify mechanisms by which the birth of a child can lead to persistent earnings differences. These literatures are complementary: care-policy entitlement, realized labor supply, and pension receipt must be examined as separate steps.

The research question is how social insurance systems transmit unequal care histories into unequal retirement rights, which mechanisms mitigate that transmission, and how reliably observed data identify these effects. The unit of comparison is a legal pension arrangement in a specified country and effective period. The unit of statistical observation may instead be a person, household, contribution year, claim, or pension recipient. Every substantive comparison identifies its denominator. The study concerns public and mandatory retirement protection primarily, with occupational and voluntary arrangements included to explain boundaries and interactions. Country comparisons are deliberately selective and sourced to official agency descriptions; they are mechanisms, not a ranking of national systems.

Conceptual and Historical Foundations

Social Insurance and the Boundaries of the Pension System

Social insurance organizes protection against legally designated contingencies through compulsory or collectively organized membership, contributions or public financing, and enforceable benefit rules. Its old-age branch may combine current-worker financing, accumulated assets, general revenue, and transfers from other branches. The ILO Minimum Standards Convention, adopted in 1952, treats old age as a distinct social security contingency and specifies minimum rules for protected persons and benefits in the Parts accepted by ratifying states. Its terms have to be read alongside national ratifications and implementing law; an international standard does not establish identical individual pension claims in every country.

An earnings-related pension is generally sensitive to earnings and covered service, although the precise mapping differs between defined-benefit, points, notional defined-contribution, and funded defined-contribution systems. Flat-rate basic pensions, residence-based benefits, and means-tested old-age assistance can establish a first tier independent of some aspects of employment history. Survivor benefits derive wholly or partly from another person's insurance history. Occupational pensions and individual savings products can add income and amplify differences in wages, employer access, tenure, and savings capacity. These sources require separate measurement before combined retirement-income analysis.

The [Social Protection Floors Recommendation](https://normlex.ilo.org/dyn/nrmlx_en/f?p=NORMLEXPUB%3A12100%3A0%3A%3ANO%3A%3AP12100_INSTRUMENT_ID%3A3065524 articulates nationally defined basic guarantees, including income security in old age. That policy orientation places adequacy and effective coverage beside contribution-related entitlements. It does not eliminate the accounting distinction between a pension earned on a person's contribution record and an allowance financed through taxation. Conflating those mechanisms conceals both redistribution and gaps in protection.

Pension-Gap Measures and Denominators

For a particular scheme, calendar year, age group, and population of recipients, define mean benefit amounts for men and women as M and W. The conventional recipient gender pension gap is G = (M − W) / M. Thus M = 1,000 and W = 750 imply G = 25 percent. A gap measured this way can change because one group receives different pensions or because the mix of pension recipients changes. It is a conditional descriptive statistic; by itself it identifies neither a cause nor average welfare.

A second indicator is the coverage gap, measured as the difference or ratio between the share of older men and older women receiving a specified benefit. A third is the gap in expected lifetime benefits, which requires payment duration, mortality, indexation, survivor status, discounting assumptions, and possibly household pooling. A fourth is the household-equivalized poverty gap, measuring income available for living expenses. OECD explicitly cautions that the recipient pension-gap indicator excludes people without a pension and does not describe living standards directly.

The distinction is substantively important. In a purely hypothetical group of ten older women, eight each receive a pension of 80 units and two receive none; ten men each receive 100. Among recipients the female–male gap equals 20 percent. Counting zero benefits for all older residents produces a female per-person average of 64 and a gap of 36 percent. Neither number is an error: they answer different questions. They must not be combined with pension-income surveys without reconciling the definition of pension, recipient, age, and reference period.

How Formal Equality Coexists with Unequal Pension Histories

European law makes the distinction between equal legal formulas and unequal labor-market histories especially clear. Council Directive 79/7/EEC establishes progressive equal treatment in specified statutory social security fields, including old age, while its text contains defined exclusions and qualifications concerning retirement ages, childrearing advantages, and derived rights. Legal assessment therefore requires close examination of its scope and of national law at the relevant time, rather than treating the directive as a universal pension design.

The later Work-Life Balance Directive provides EU minimum rights concerning paternity, parental, and carers' leave and flexible working arrangements for workers covered by the instrument. Those rights can affect later paid-work histories; the directive itself does not prescribe a single cross-country method of assigning retirement pension credits for every care spell. The legal channels of employment protection and pension accrual are analytically distinct and often operate through different statutes.

Unpaid Care as Work and as an Insurable Policy Concern

The United Nations' SDG 5.4.1 statistical guidance defines unpaid domestic and care services by activity and time spent, drawing on time-use surveys. The category includes care of household or non-household family members, as well as domestic activities such as food preparation, cleaning, maintenance, shopping, and management of household services. Its indicator measures the proportion of a day devoted to specified activities. The concept is wider than the set of activities recognized under a particular pension-credit rule, which may recognize only primary childcare, care for a person with an assessed disability, or prescribed weekly hours.

The EIGE care findings, released in 2026, document gender differences in care intensity and external support in the European Union. Such surveys illuminate the exposure to caring activity but cannot directly reveal pensionable years or statutory eligibility. The institutional research task is to trace which recorded forms of care change contributions and which entitlements compensate for that effect.

The history of pension provision thus combines three developments: expansion of individual social insurance rights; growing awareness of employment-history inequality; and targeted recognition of nonmarket caregiving. These developments proceed at different speeds. The existence of a gender-neutral pension rule can coexist with unequal career earnings and unequal entitlement amounts, while a care credit that raises a claimant's recorded earnings can still leave substantial gaps arising from occupational progression and pay after return to work.

Career Histories

Entry, Earnings, Hours, and Career Duration

The pension consequences of gendered labor-market participation begin with the transition from education to covered employment and continue through job mobility, unemployment, caregiving, disability, migration, and retirement timing. Annual pensionable pay is the product of paid hours, hourly compensation, covered status, and the scheme's definitions of assessable income. A lower annual contribution record can arise from fewer hours even where hourly wages are identical. Conversely, equal full-time participation does not ensure equal pension accrual if occupational sorting, wage progression, bonuses, or ceilings differ. Any decomposition that treats earnings as an exogenous explanation of the pension gap thereby risks concealing the effects of care on earnings themselves.

Goldin's analysis of the gender earnings gap explains how occupations that reward long, continuous, or inflexible working hours can generate disproportionate earnings penalties for workers requiring temporal flexibility. This mechanism matters for pensions with long assessment periods, since a temporary caring obligation can alter subsequent promotions, occupational tenure, and the earnings base for many later years. A pension credit that replaces the year of leave may therefore repair the immediate missing contribution without restoring the entire subsequent trajectory. The magnitude of this cumulative effect differs between an individual with a short leave and full wage recovery and an individual who moves into persistent part-time employment.

A standard individual contribution panel should distinguish years employed, annual hours, gross and assessable earnings, employer contributions, employee contributions, state-financed credits, periods excluded from averaging, and periods of eligibility without payment. It should identify whether the pension record derives from employees, self-employed workers, informal activity, or multiple schemes. Migrants may have contribution rights in more than one jurisdiction, and occupational pension participation can vary even among otherwise observationally similar wage earners. A count of employed years alone loses these dimensions.

A life-course analysis must also distinguish cohorts. Women retiring in 2024 worked under earlier labor-market conditions and earlier pension rules, whereas an employment policy enacted in 2024 principally affects future retired cohorts. The OECD's 2025 pension comparison combines observed outcomes of retired cohorts with forward-looking pension model cases; these are different empirical objects. Treating projections as measured pension outcomes would overstate what is known.

Pension Accrual in Different System Designs

Consider a simplified pension account indexed by working age t. Let e(t) denote pensionable earnings, a(t) the annual accrual rate or contribution factor, and c(t) a separately recognized care credit. An abstract earnings-related accrual ledger can be expressed as A = sum over t of [a(t) × e(t) + c(t)]. This is a bookkeeping illustration, not a universal statutory pension formula. A real defined-benefit scheme may index historical wages, apply the average of selected high-earning years, limit creditable service, impose benefit ceilings, and supplement with minimum entitlements. Pension points schemes may divide earnings by an annually defined reference wage. Defined-contribution schemes accumulate contributions plus investment returns minus costs, and retirement income depends on conversion and decumulation assumptions.

In a funded account, a missing contribution early in a career may also lose decades of compound returns. In a wage-indexed benefit formula, the missing accrual depends on how the system indexes or revalues past earnings and whether minimum pensions or service ceilings bind. In a highest-years formula, a low-earning childcare year may have little effect if excluded by sufficiently many higher-earning years, but a permanently shortened career can introduce zero years. These institutional contrasts explain why a single percentage of interrupted employment cannot be mapped mechanically to a single percentage pension loss.

A care credit can modify one or more of four objects: covered service, imputed earnings, the divisor or years used to compute average earnings, or actual funded contributions. The mechanisms are not actuarially equivalent. Imputing an average national wage in a defined-benefit account may disproportionately help workers with previously low wages, depending on formula rules. Excluding low-earning care months from a denominator can strongly improve a pension calculation for a shorter career but may have limited effect if other exclusions already apply. A funded contribution financed by government creates a balance that changes with investment performance. Eligibility and recordkeeping determine whether the nominal right becomes an actual benefit.

Selection into Pension Receipt and Survivor Entitlements

Pension-recipient statistics condition on receipt. Contributory qualifying periods, minimum contribution thresholds, migration histories, vesting rules, and informal employment can select people out of the recipient population. The ILO's coverage comparison documents distinct access patterns for older women and men worldwide. Where women excluded from contributory pensions rely on noncontributory support, a recipient-based measure restricted to the contributory scheme can appear narrower even as aggregate protection is weaker. Adding a tax-financed minimum might increase measured inclusion and simultaneously lower the overall average of newly counted recipients. Such an increase in the measured gender pension gap can occur alongside a real improvement in protection.

Derived and survivor rights also complicate interpretation. A survivor pension may protect individuals with limited pension contributions but depends on relationship status, prior insured partner earnings, statutory eligibility, and the duration of the benefit. An individual can have a low pension based on her own record and substantial survivor income; aggregating these without separation obscures individual entitlement autonomy. The OECD chapter on pension gaps discusses survivor pensions and the distinction between individual pensions and living standards. Neither individual gender differences nor household welfare is captured fully by a single all-pension mean.

The Long Delay Between Policy Change and Pension Outcomes

The observed pension gap is a stock of claims held by living retirees, while a newly created care credit is initially a flow into working-age pension accounts. Effects on the observed stock emerge slowly as credited contributors reach retirement, as pension rules mature, and as the composition of survivor cohorts changes. Retrospective credits may operate differently: a reform granting recognition for past care can affect present claimants, but its cost and distribution depend on eligibility, evidence, and statutory retroactivity.

Analytical evaluation should therefore report three horizons: current income effects on existing pensioners; medium-term changes in accounts of workers approaching retirement; and long-run effects on entering cohorts. Policy studies that report only one horizon can understate or overstate the total distributional consequences. The same temporal separation is essential when examining increases in pension ages or changing contribution ceilings, since such reforms interact with caregiving, healthy life expectancy, and labor-market participation.

Unpaid Care Effects

Mechanisms and Causal Interpretation

Unpaid care affects pension outcomes through at least six empirically distinct channels: time out of work; reduced current hours; lower subsequent wages or occupational mobility; altered access to workplace pensions; reduced ability to save; and direct statutory offset through credits, contributions, or service recognition. Childbirth can trigger several channels at once. Long-term caregiving for older parents may begin later, involve intermittent or unpredictable intensity, and interact with a caregiver's own health. Even where care is unpaid, the pension consequences arise from institutional choices about which activities generate individual social security rights.

Kleven, Landais, and Søgaard's Danish child-penalty study uses an event-study design to trace changes in earnings, participation, and hours around the birth of a first child. Their evidence identifies substantial long-run divergence in women's and men's labor-market trajectories after parenthood in Denmark. The published study is Children and Gender Inequality. These results support a mechanism by which child-related labor-market penalties can affect eventual pension accrual, but the paper does not itself estimate cross-country pension gaps or the effect of every national care-credit rule.

The distinction matters for causal inference. A correlation between being a caregiver and having a smaller pension can reflect preexisting wage differences, educational choices, labor-market institutions, and health conditions that influence both care and pension outcomes. Observing a pension penalty after care is informative but is not automatically an estimate of the causal contribution of caregiving. An event study or policy discontinuity may identify a more local parameter, conditional on valid design assumptions. The aggregate national gap additionally depends on the frequency of care spells, benefit formulas, selection, and cohort composition.

Childcare, Long-Term Care, and Heterogeneous Care Spells

Childcare and long-term care share the feature of time displacement but differ in timing, predictability, eligibility, and documentation. Care for a small child tends to be identified through age and parenthood records; eligible care for an adult with disability often requires an assessment of dependency and a threshold number of weekly care hours. A pension scheme may give an automatic child-related credit while requiring a caregiver of an adult to apply for a different benefit. Equating the two categories would conceal major differences in take-up and legal design.

Care intensity interacts with employment choices. Ten hours of unpaid care can coexist with full-time work in some occupations, while the same schedule can require part-time employment or job exit elsewhere. Multiple children, disability, single parenting, and overlapping responsibilities for children and older relatives can concentrate gaps. Carers may face competing incentives when pension credit rules impose hours thresholds that differ from paid employment rules. The ILO comparative care-policy report details diversity in care-related leave and services, emphasizing why formal rights must be measured against access, adequacy, and actual use.

Care also has a social allocation dimension. If pension credits attach to one designated caregiver within a household, a formally gender-neutral rule can produce highly gender-skewed outcomes where women perform most qualifying care. Individual pension autonomy may improve when credits are paid to actual caregivers and when both parents can share paid leave and employment protection. Whether a particular policy changes the division of care requires observed behavior rather than inference from statutory language alone.

Comparative National Pension-Credit Instruments

German statutory pension insurance treats eligible periods of childrearing as pension-relevant contribution periods. Deutsche Rentenversicherung's official childrearing guidance states that, at the research cutoff, up to 36 months per child born from 1992 and up to 30 months for children born earlier can be recognized; child-consideration periods may extend to ten years. The pension authority's legal glossary explains the credited earnings basis and public contributions. These are distinct durations and effects. Recognition generally belongs to only one parent for a given month, and rules for concurrent employment and other credit types need individual account verification. This case illustrates positive pensionable earnings being constructed through law for qualifying care.

Canada uses a different method. The government's CPP child-rearing provisions distinguish a base-component child-rearing drop-out, excluding qualifying low/no earnings months from the pension calculation where advantageous, and an enhanced-component child-rearing drop-in, providing qualifying credits for some periods since January 2019. Eligibility concerns primary care of a child younger than seven, benefit history, and qualifying earnings conditions. The government's description cautions against treating all years of care as automatically credited at a uniform wage. This approach chiefly protects how an earnings history enters a pension formula.

In the United Kingdom, National Insurance credits eligibility links qualifying childcare and other circumstances to contribution records for specified state pension purposes. Claiming Child Benefit for a child under twelve can protect a parent's record, and distinct Carer's Credit rules address qualifying adult or disability-related care for at least twenty hours per week. Those credits do not automatically replace lost employer pension contributions or every component of career earnings. They operate through a qualifying-record system and are subject to claimant status and administrative conditions.

Ireland's HomeCaring Periods arrangement recognizes specified periods outside the workforce caring for children under twelve or other people needing care when assessing entitlement under the Total Contributions Approach. The Homemaker's Scheme separately addresses certain qualifying periods through the older yearly-average assessment. These instruments should be compared only after identifying the applicable assessment route and birth dates or claim conditions; a legal title shared across cohorts does not establish equal effective value.

Sweden's public pension system recognizes pensionable credit for years with small children as part of a broader scheme that combines income-related accrual and a guarantee pension. The Swedish Pensions Agency's English pension-system explanation distinguishes public pension components and child-year adjustments. Parental leave compensation, child-related pension amounts, and a means-tested or guarantee component perform different functions; their effects cannot be summarized as a single year-for-year credit without calculating the individual record.

Australia demonstrates another institutional channel. The Australian Taxation Office's parental leave superannuation guidance, updated July 2026, states that eligible government-funded Parental Leave Pay for children born or adopted from July 1, 2025, receives a superannuation contribution, with payments commencing in financial year 2026–27. Services Australia's September 2026 scheme update states a twelve-percent contribution on qualifying payments. This is a funded pension-contribution instrument rather than a recalculation of statutory pensionable years; its long-run impact depends on the number of qualifying weeks, contributions, investment returns, and retirement balance.

For contrast, the United States Social Security retirement benefit generally uses the highest 35 years of indexed earnings. An interrupted earnings career may therefore create low or zero years in the calculation, although excluded low years may be replaced by other higher years when sufficient earnings years exist. U.S. Social Security also has spousal and survivor benefits and other protective provisions; this formula comparison concerns the direct retirement-worker calculation, not a claim that the whole system offers no household protection.

These examples identify four observable designs: imputed pensionable earnings; an altered assessment divisor or excluded low-earning periods; credited contribution years; and government-funded contributions into a savings account. They represent the legal mechanics of selected schemes as described by official sources available at the cutoff. The illustrations do not establish a uniform welfare ranking, since net benefits depend on pension formulas, taxes, minima, effective claims, and financing.

Illustrative Benefit Arithmetic and Sensitivity

The following examples are fully hypothetical and test the logic of pension mechanisms. Their amounts are not calibrated to a country's benefit levels, actuarial assumptions, or policy costs. All pensions are expressed in illustrative constant-value units; taxes, ceilings, minimum guarantees, inflation, and mortality are initially omitted so the comparison exposes one feature at a time.

Begin with a simple career-average formula that pays 1.5 percent of the arithmetic sum of forty annual pensionable earnings records, scaled to an illustrative unit convention. A person earning 40 units in each of forty years has an accrual index equal to 0.015 × 40 × 40 = 24 units. A person with five zero-earning years and thirty-five years at the same earnings obtains 0.015 × 35 × 40 = 21 units, a 12.5 percent reduction relative to the full-career index. If five missing years are credited at 40 units, the index returns to 24. If the credit instead grants only qualifying service while pensionable earnings are imputed at 20 units, the index becomes 22.5. The arithmetic shows that the label care credit does not determine its value: the credited earnings base does.

Now introduce the post-care labor-market channel. Suppose the worker's thirty-five paid years include five years at 32 units after return, instead of 40. Without a care credit, accrual becomes 0.015 × [(30 × 40) + (5 × 32)] = 20.4. Credit the five years of absence at 40 and the index becomes 23.4. Even a complete offset for the immediate interruption leaves a 2.5 percent shortfall against the full-career 24 because lower earnings after care remain in the record. If lower wages instead persist for twenty paid years, the uncredited index is 0.015 × [(15 × 40) + (20 × 32)] = 18.6, and with the same five years credited at 40 it becomes 21.6. A full care-year credit cannot by itself equalize a twenty-year wage trajectory.

A dropout works differently. Consider a purely hypothetical average of thirty-five recorded years at 40 and five years at zero, calculated over forty years. Its average is 35 units. Excluding the five zero years from both the numerator and denominator raises the resulting average to 40. If actual benefit entitlement also requires a minimum number of years of service, exclusion may repair the average but not the qualifying-period shortfall; a complete statutory calculation must check both. A funded contribution behaves differently again: a deposit of 1 unit made thirty years before retirement has an illustrative future value of approximately 2.43 units at a constant three-percent annual net return, or 4.32 at five percent. These are compound-interest examples, not forecasts; volatile returns, fees, taxes, and inflation change real purchasing power.

The examples establish a crucial measurement rule. A scenario comparison should hold constant the baseline earnings profile, care timing, credit eligibility, wage recovery, pension indexation, retirement age, contribution rates, and terminal benefit convention except for the policy treatment being tested. Otherwise a reported care-credit effect can absorb changes in wages or contribution rates unrelated to the intervention.

Pension Disparities

OECD, European Union, and Global Coverage Evidence

The OECD and Eurostat report related yet nonidentical retirement benefit indicators. The OECD's 2025 gender pension gap chapter reports a 23 percent average gap in 2024 among OECD countries, measuring the percentage difference in average pensions of men and women and noting that country samples and coverage vary. Eurostat's 2024 EU figures report 24.5 percent for average pensions and 24.9 percent for median pensions in the EU, for people aged sixty-five or older. They also show large cross-country variation: a mean gap of 5.6 percent in Estonia contrasts with more than 35 percent in the Netherlands, Austria, and Malta. The median and mean can differ substantially, confirming that upper tails and distributional shape affect inference.

The comparison between Estonia and Malta cannot by itself identify differences in care participation or pension-credit success. Country gaps depend on statutory pension design, historical employment opportunities, household patterns, occupational pensions, population composition, survey definitions, and cohort aging. Indeed, a low recipient gap can coexist with a high level of poverty among both sexes, or with uneven pension access. No single country rank is an index of gender equality across all old-age outcomes.

The ILO's 2024–26 global pension chapter reports that 79.6 percent of people beyond statutory retirement age receive some pension, while among older people contributory pensions reach 49.2 percent of women and 63.2 percent of men. It also finds that 29.1 percent of working-age women contribute to pensions, compared with 40.9 percent of working-age men. Global old-age coverage combines contributory and tax-financed schemes, and countries report different administrative and survey periods. Its distinction between contributor coverage and beneficiary coverage is essential when examining informal employment and care work. A generous formula on paper can benefit few carers if qualifying contributors are rare.

Three noninterchangeable gaps must be reported jointly: observed recipient amounts, access to a pension, and economic security among all older persons. A cross-country difference in the first measure may arise partly from transitions in the second. The third additionally reflects household composition, asset income, costs, and services. The OECD Pensions at a Glance 2025 volume reports higher old-age relative poverty among women than among men across the comparison group, underscoring why pension benefits must be related to overall resources.

Labor-Market Institutions and Household Arrangements

Gender-related earnings differences reflect both pay received for comparable work and the composition of occupations and work histories. The labor-market transmission mechanism is sensitive to employment protection, childcare availability, norms governing parental care, labor taxation, collective bargaining, the incidence of informal work, and access to flexible schedules. A system that gives full pension credit during parental leave may preserve the immediate pension record while leaving later promotions and employer pension access differentiated. An employer-linked occupational pension may be especially sensitive to part-time thresholds, vesting periods, and job changes.

Goldin's research on workplace flexibility explains why the earnings premium for uninterrupted, inflexible hours matters for the trajectory of lifetime compensation. Kleven and colleagues' Danish parental event study isolates a child-related earnings discontinuity under specific identification assumptions. These findings justify tracing wage effects beyond the leave period but do not authorize assuming a fixed global child-penalty percentage or a universal pension effect of the Danish estimate.

Household relationships can alter both care allocations and pension rights. Marriage and registered partnerships may entitle partners to survivor benefits, pension splitting, or rights during divorce, with scope varying by jurisdiction. Sole parents may face more restricted alternatives for sharing care and paid work. Informal unions may generate caring obligations without the same derived benefit rights as formally recognized partnerships. Pension policy should be evaluated at the level of individual rights and at the level of household resources, because a high household pension does not always imply equal individual ownership or access.

Income Adequacy, Longevity, and Survivor Protection

A retirement pension can be adequate for a single person, inadequate for another with high disability-related costs, and a different share of household income when partners live together. Benefit adequacy should therefore be assessed against a transparent benchmark, such as income relative to median equivalized income, rather than inferred from a pension-gap ratio alone. Pension inclusion, replacement rate, and minimum income eligibility remain different quantities.

Women generally experience different mortality and survival profiles from men in many populations, so monthly pension amounts and lifetime expected payment duration need separate treatment. A smaller monthly pension could be paid over more expected years; this does not demonstrate adequate income during each year. In schemes that use sex-differentiated annuity factors, institutional and legal rules affect benefit conversion, while survivors' benefits condition on marital or partnership status. A full actuarial assessment requires current mortality tables, country-specific conversion methods, interest and indexation assumptions, and treatment of retirement timing. No universal actuarial adjustment is estimated here.

Public social protection also influences household behavior before retirement. Means-tested benefits can offset deprivation but create withdrawal rates as other income rises. Universal or residence-based floors can secure coverage where contribution records are fragmented. Survivor benefits improve financial security following the death of a spouse but may induce distributional differences between household types. These mechanisms warrant explicit disclosure rather than collapsing all transfers into a generic social pension.

Care-Policy Financing, Distribution, and Tradeoffs

Care-recognition benefits require financing. In a pay-as-you-go social insurance system, crediting pensionable earnings creates future entitlements financed through contributions, public transfers, benefit adjustments, or some combination. Where a government pays explicit contributions into a funded account, cost is incurred contemporaneously and investment risk subsequently affects benefits. A minimum pension guarantee raises expenditure selectively among beneficiaries who would otherwise fall below its conditions. Design alternatives therefore vary in the timing, certainty, and distribution of public costs.

Credits calculated at a national reference wage may produce larger absolute additions for some beneficiaries than flat minimum protection, especially where pension accrual remains linked to the wage base. Low-income carers with no attachment to the insured system may be excluded from a contributory credit even though their protection needs are substantial. A minimum benefit can help them if its residence, age, income, and claim requirements are satisfied. Conversely, a means-tested minimum can weaken the visible incremental return to an earnings credit among those whose benefits are already topped up. These interactions are consequences of formulas, not value judgments implied by an aggregate gap.

A policy appraisal should report gross expenditure, changes in public revenue, take-up, benefit distribution by earnings and family status, and administrative compliance costs. Fiscal comparisons must use the same time horizon and discounting assumptions. The legal statement that care is recognized must be supplemented by a realistic assessment of claim rates, data availability, appeals, and information barriers. The ILO's care leave and services survey demonstrates variation in care policy environments and provides context for considering the complementarity of services, leave, and pension rules.

Evidence Verification and Sensitivity

Source Hierarchy, Time, and Jurisdiction

The evidence base combines several source categories whose inferential roles must remain distinct. Official pension-authority descriptions identify current administrative rules; original legislation determines the legally operative entitlement and the hierarchy of national norms; OECD, Eurostat, and ILO statistical releases define populations and indicators; academic research evaluates mechanisms and local causal effects. Agency guidance is strong evidence of the administration's published interpretation at the date checked, but a claimant-specific legal question can turn on statutory amendments, transitional regulations, court decisions, or facts omitted from a public summary. The monograph therefore describes institution-level mechanics as of the November 2026 research cutoff and does not purport to adjudicate individual rights.

The European directives analyzed here establish legally binding frameworks for Member States within their scope, subject to implementation requirements and legal interpretation. The 1979 statutory security directive addresses progressive equality of treatment in specified statutory schemes. The 2019 work-life balance directive sets leave and flexible-working minimum standards. Each instrument concerns a defined regulatory subject and effective legal horizon. Neither can be converted into a present-day universal pension amount; the pension effect depends on national pension rules and the affected worker's record.

International labor standards serve a different function. The Social Security Convention No. 102 establishes agreed minimum standards for specified social security branches subject to ratification and acceptance conditions. The Social Protection Floors Recommendation No. 202 states principles for nationally defined basic social guarantees. The latter is a recommendation rather than a directly applicable national statute. Both are relevant to evaluating the architecture of old-age protection, yet neither supplies a quantitative estimate of unpaid care's causal effect on a national pension gap.

The traceability strategy in this study links each substantive external assertion to a contextual source in the prose and a full bibliographic record in References. The accompanying EVIDENCE.json identifies claim scope, source type, and access limits. Historical enactments are not silently treated as 2026 rules. Government information pages with unspecified publication dates are marked as current web guidance accessed at the research cutoff; their underlying legislation may have been amended. A complete national entitlement audit would require consolidated statute texts, official gazettes, regulations, and judicial records for each effective claim year.

Statistical Comparability and Missing Populations

National and cross-country pension statistics differ in reporting ages, benefit types, tax treatment, treatment of survivor benefits, private and occupational income inclusion, and survey versus administrative collection. The OECD's recipient-based gender pension gap definition is directly suited to comparing average pension amounts where definitions have been reconciled. Eurostat's mean and median pension series adds a distributional perspective but has its own reference population. In small populations, compositional and sampling differences can materially affect estimated gaps. A sample-weighted mean and an unweighted mean answer different questions if subgroup structures differ.

The ILO's old-age coverage statistics reflect different underlying program definitions and population denominators. Receiving any pension, receiving an earnings-related contributory pension, and being formally insured as a working-age contributor are not nested in a way that allows their quoted percentages to be simply subtracted from one another. Regional gaps in administrative capacity and informality further limit the portability of a single national dataset.

Time-use studies create an additional challenge: diary-based estimates vary by reference day, season, simultaneous care, unpaid domestic work definitions, and whether the survey records care within or beyond one's household. The UN indicator 5.4.1 handbook recommends activity-based classification; EIGE's 2026 EU care survey factsheet provides a different empirical window into participation and intensity. Combining these measures with pension administration demands explicit person-level matching, time alignment, and eligibility coding. A correlation of aggregate national care hours and pension gaps risks an ecological fallacy.

Missingness deserves substantive treatment. Informal caregivers may lack an administrative record. Pension beneficiaries may not identify prior childcare or adult care in retrospective surveys. Persons with no pension are omitted from recipient means; deceased members of a cohort are missing from current survivor samples, so mortality selection can change the composition of older pension populations. Pension-account data may lack occupational plans or personal savings. A credible statistical appendix would identify coverage fractions for each source and quantify plausible effects of missing groups.

Causal Identification and Decomposition

An accounting decomposition can isolate differences in observed employment years, earnings levels, and accrued rights. Such a decomposition does not establish whether the employment differences were themselves caused by care. For instance, conditioning on post-childbirth earnings can remove from a regression part of the pathway by which childcare affects pensions. Analyses must specify whether the estimand is total association, direct effect holding employment constant, or the policy effect of a defined care-credit amendment.

Event-study research such as Kleven, Landais, and Søgaard's Danish analysis helps identify post-childbirth changes in earnings, using trajectories around a birth and comparison assumptions. That design examines a labor-market outcome and particular population; pension transmission is a further step to be modeled. Difference-in-differences designs around pension-credit reform can estimate local benefit changes where treated and comparison cohorts exhibit appropriate pre-reform trends and reform timing is not confounded by other changes. Regression discontinuity at eligibility cutoffs can identify local effects, but its estimates may not generalize to carers far from the threshold.

Policy analysis should disaggregate at least the following components: predicted baseline pension under the actual earnings history; statutory care credits mechanically applied to that history; changes in working behavior plausibly induced by the policy; and redistribution through floors, survivor benefits, taxes, and means tests. The mechanical component is deterministically computable with complete administrative rules. The behavioral component requires a credible research design and evidence. Treating both as a single observed difference exaggerates the certainty of policy impact.

An assessment of care recognition should further distinguish an intention-to-treat population of persons eligible under law from actual beneficiaries who successfully claim credits. Failure to record a past care period is a take-up and administration issue, not evidence that the underlying pension formula lacks a provision. Selection into application and eligibility can bias comparisons of recipients with and without credits. Administrative process data and controlled outreach designs can help estimate this component.

Model Sensitivity and Replication Protocol

The first source of sensitivity is the assumed length and timing of care. A five-year spell in one's twenties affects a funded account differently from a five-year spell immediately before retirement; in a points system, year-specific national wage averages matter. The second is subsequent earnings recovery. The hypothetical examples above show that credits for the direct absence cannot eliminate wage scars after reentry. The third is eligibility. A parent outside formal pension membership may be unable to benefit from a credit; residence-based and tax-financed schemes may reach that person. The fourth is indexing. Revaluation of earnings, pension adjustments, and inflation determine the real value of a credited unit.

A reproducible comparative simulation would process: (i) an explicit annual earnings and care calendar; (ii) the pension rules and effective dates; (iii) eligibility and claim status; (iv) actual and counterfactual credit assignment; (v) resulting accrued annual benefit under the statute; (vi) public cost and financing conventions; and (vii) poverty and coverage outcomes under household income assumptions. It would use identical simulated histories for policy and comparator cases and publish sensitivity scenarios, errors, and versioned legal source references.

As an elementary sensitivity calculation, take the initial full-career index of 24 and the five-year interruption index of 21. Crediting zero, one-half, and all of the five absence years at 40 produces indices of 21, 22.5, and 24. These imply remaining reductions from the baseline of 12.5 percent, 6.25 percent, and zero, respectively. Those outcomes follow from one invented linear accrual formula. A real system's minimum benefit, nonlinear earnings ceiling, qualifying threshold, or survivor supplement could make the marginal benefit of the credit smaller, larger, or zero. A sensitivity table must therefore label every input and not recast hypothetical arithmetic as estimated national effects.

This research does not use confidential administrative microdata, conduct a preregistered systematic evidence search, or estimate a new cross-country causal effect. Its numerical contributions are transparent demonstrations and reconciled source-reported statistics. Claims about specific legal rules use official current descriptions, with statutory audit limits disclosed. Those limits confine generalization while retaining the explanatory value of the mechanisms.

Scientific Synthesis and Research Agenda

Integrated Findings

The reviewed evidence supports a connected explanation of gender differences in pension outcomes. Gendered distributions of caregiving affect hours of employment and earnings trajectories. Employment and earnings become pensionable through country-specific legal formulas and administrative records. Pension systems may credit certain care histories, protect benefit assessment periods, or provide minimum or survivor benefits. Coverage of these protections varies with the type of care, eligibility thresholds, application procedures, financing, and employment structure. The pension gap observed among older recipients is therefore the cumulative result of both unequal life histories and institutional transformations of those histories into rights.

A key comparative finding is that distinct credit mechanisms target distinct points in this chain. Germany's childrearing periods add pension-relevant contribution records; Canada's dropout and drop-in rules adjust the treatment of low-earning periods in different CPP components; the United Kingdom's National Insurance credits help preserve qualifying records; Ireland's HomeCaring Periods influence the Total Contributions Approach; Sweden's child-year treatment creates additional pension-relevant rights; Australia contributes to funded superannuation accounts based on eligible paid parental leave. These are grounded institutional distinctions that clarify the limits of comparing policies by their headline duration alone.

A second finding is that pension payment gaps and pension coverage gaps require separate remedies. Redistributing pension accrual within an existing contributory scheme does little for people wholly outside its qualifying population unless the credit itself provides an entry route. Broader basic-income support may improve coverage while leaving employment-related pension differentials intact. The ILO's worldwide pension access data make this separation central to international comparison. The design question becomes which group, account, and outcome a policy actually reaches.

A third finding is that care-related pension inequality can persist after a complete credit for the interruption itself. Evidence on long-run child-related labor-market effects indicates that care may influence wages, occupations, and hours over many years. Credits address a subset of the resulting accumulation channels. Childcare services, leave design, workplace flexibility, and equitable sharing of care can therefore affect pension outcomes through future employment histories. The pension and care-service institutions should be compared jointly while retaining separate outcome measurements.

Comparative Policy Appraisal

The selection of a pension-care instrument should start with the targeted impairment. If the principal problem is a missing qualifying year, a credited service year can be decisive. If it is a low average of assessable earnings, a dropout or imputed earnings credit may be more effective. If employer-linked funded pension balances are lost during a paid leave period, state-funded contributions can address a portion of the loss. If the intended beneficiary has never contributed to the qualifying scheme, a floor or noncontributory old-age entitlement may be the relevant instrument. A full assessment includes administrative feasibility, treatment of nonresident caregivers, and interactions with means tests.

Distribution also depends on which caregiver receives the credit. Child-related pension recognition can be allocated by default, by application, or by transfer between eligible parents. Default attribution may reflect typical division of care, but individual claims should be examined in relation to actual responsibilities, administrative evidence, and legal recognition of family arrangements. Eligibility thresholds for adult care should be compared against real care patterns and measured take-up. A seemingly generous benefit with burdensome documentation can yield weaker realized protection than its nominal formula suggests.

Intertemporal budget incidence is central. Credits to today's workforce may create benefits decades later in pay-as-you-go schemes, while direct funded contributions generate immediate public expenditure and investment exposure. Minimum guarantees may entail more immediate costs among retired beneficiaries. No option's fiscal merits can be evaluated only by its future headline pension uplift. Adequate assessment combines projected cohort costs, distributional consequences, and funding arrangements, with explicit discount-rate and longevity assumptions.

Researchable Hypotheses and Data Requirements

The following propositions are researchable implications from the comparative synthesis, not newly founded theories or claims of established universal causal effects.

First, in an earnings-related system with a childrearing earnings credit, the direct pension loss associated with an eligible care spell should be smaller than for a legally equivalent worker without the credit, holding earnings history and other pension rules constant. The strength of this predicted mechanical attenuation depends on the credit's imputed earnings rate, overlap rules, and benefit ceiling. A linked administrative dataset should permit direct calculation.

Second, systems that exclude care months from average earnings may show larger gains among claimants whose pension formula includes otherwise low-earning months than among claimants already eligible for enough higher-earning reference years. The relevant data are complete annual earnings records, care documentation, legal calculator rules, and claimant-specific benefits.

Third, credits attached to contributory membership may have smaller protective effects on pension coverage among informal or intermittently insured caregivers than tax-financed minimum guarantees. Evaluating this comparison requires observing nonclaimants and older persons without a pension, not merely benefits among pensioners.

Fourth, the policy's impact on the gender pension gap may depend on whether the credit changes labor-market behavior. A pension credit that repairs an account after care and a childcare service that allows continued work affect different stages of the life course. Data linking time use, employment, pension records, and childcare access are necessary to separate these effects. Suitable natural experiments could exploit carefully documented eligibility reforms while checking concurrent policy changes.

Fifth, the measured recipient pension gap can rise after a reform that newly includes low-income people with small pensions. The hypothetical inclusion example shows the accounting possibility. Empirical testing would compare coverage rates, conditional recipient means, unconditional population means, poverty, and benefit distributions before and after implementation, using consistent sampling and valuation.

A sustained research program should prioritize linked administrative histories, harmonized surveys and time-use diaries, historical reform dates, migration and pension portability records, retirement mortality, and complete benefit regulations. Access restrictions and privacy concerns require anonymization and lawful governance. Cross-national causal studies must be designed around comparable estimands and explicit country-specific law rather than averages of heterogeneous treatment effects.

Institutional and Editorial Implications

A scientific pension analysis has two obligations: measure the observed distributions accurately and identify the legal production of entitlements. Statistical releases provide the first dimension; statutory and administrative materials supply the second. Labor-market causal research describes behavioral channels but does not substitute for calculating benefit formulas. Keeping these evidentiary roles distinct makes the research reproducible and avoids turning pension inequalities into an untestable aggregate narrative.

The interpretation is internationally applicable as an investigative method because it asks comparable questions of different pension schemes: who qualifies, what care is recognized, where accrual is credited, how claims are made, how benefits are financed, and which people remain outside the protective mechanism. The legal answers remain jurisdiction- and period-specific. Further work should integrate funded occupational pensions and migration-linked portability, which this monograph treats as boundary conditions rather than exhaustive country surveys.

Conclusion

Gender differences in old-age pensions are accumulated through the interaction of employment, earnings, unpaid care, individual and derived rights, and pension-system design. The OECD and Eurostat measurements document substantial inequalities in benefits among recipients, while the ILO records unequal effective access to contributory protection. These outcomes reflect different denominators and dimensions of retirement security. Their interpretation requires separating contribution coverage, actual pension payment, benefit adequacy, and household resources.

Unpaid care has an identifiable institutional pathway into pension outcomes: it can reduce paid hours, interrupt contributions, depress subsequent pay, and shift workers away from occupational pension opportunities. Comparative labor-market evidence shows why these mechanisms can extend beyond a discrete absence from employment. Pension systems respond through credited service, imputed earnings, exclusions from benefit assessments, transfers to funded accounts, and minimum or survivor protection. The national examples demonstrate the diversity and specificity of those instruments.

The original analytical synthesis establishes a practical comparative conclusion: the effectiveness of recognizing unpaid care depends on the precise pension-account object protected, the lifetime trajectory of the caregiver, the availability of basic coverage outside contributory schemes, and the operation of claims administration. The same nominal entitlement may have markedly different consequences across earnings histories and countries. A rigorous policy evaluation consequently requires both source-specific legal interpretation and life-course data, accompanied by distributional and fiscal sensitivity analysis.

The remaining empirical work concerns causal estimation from linked records, the treatment of nonclaimants, adult-care histories, cohort dynamics, and long-run behavioral responses to pension-credit reforms. These are research priorities derived from the evidence assembled here. The monograph offers a source-grounded comparative analysis and transparent illustrative calculations; it advances no independent new doctrine or universal causal estimate.

References

  • OECD. (2025). Pensions at a Glance 2025: OECD and G20 Indicators. OECD Publishing, Paris. Source document.

  • OECD. (2025). Gender pension gap. In Pensions at a Glance 2025: OECD and G20 Indicators. OECD Publishing. Source document.

  • Eurostat. (2026, February 25). Women’s pension 25% lower than men’s in 2024. Eurostat News, European Commission. Source document.

  • International Labour Organization. (2024). Old-age pensions. Chapter 4, World Social Protection Report 2024–26: Universal Social Protection for Climate Action and a Just Transition. ILO. Source document.

  • Addati, L., Cattaneo, U., and Pozzan, E. (2022). Care at Work: Investing in Care Leave and Services for a More Gender Equal World of Work. International Labour Office, Geneva. Source document.

  • International Labour Organization. (1952). Social Security (Minimum Standards) Convention, 1952 (No. 102). Adopted June 28, 1952; official ILO text edition. Source document.

  • Council of the European Communities. (1978). Council Directive 79/7/EEC of 19 December 1978 on the progressive implementation of the principle of equal treatment for men and women in matters of social security. Official Journal of the European Communities, L 6, 10 January 1979. Source document.

  • European Parliament and Council. (2019). Directive (EU) 2019/1158 on work-life balance for parents and carers and repealing Council Directive 2010/18/EU. Official Journal of the European Union, L 188, 12 July 2019. Source document.

  • United Nations Statistics Division. (2018; accessed 2026). Indicator 5.4.1: Proportion of time spent on unpaid domestic and care work, by sex, age and location. E-Handbook on SDG Indicators. Source document.

  • European Institute for Gender Equality. (2026, March 30). Factsheet: Gender Gaps in Care—EU. Publications Office of the European Union. DOI: 10.2839/5919324. Source document.

  • Goldin, C. (2014). A grand gender convergence: Its last chapter. American Economic Review, 104(4), 1091–1119. DOI: 10.1257/aer.104.4.1091. Source document.

  • Kleven, H., Landais, C., and Søgaard, J. E. (2018). Children and Gender Inequality: Evidence from Denmark. NBER Working Paper No. 24219. National Bureau of Economic Research. DOI: 10.3386/w24219. Source document.

  • Kleven, H., Landais, C., and Søgaard, J. E. (2019). Children and gender inequality: Evidence from Denmark. American Economic Journal: Applied Economics, 11(4), 181–209. DOI: 10.1257/app.20180010. Source document.

  • Deutsche Rentenversicherung. (n.d.; accessed November 17, 2026). Kindererziehung: Ihr Plus für die Rente. German statutory pension authority. Source document.

  • Deutsche Rentenversicherung. (n.d.; accessed November 17, 2026). Kindererziehungszeiten. Rentenlexikon, German statutory pension authority. Source document.

  • Government of Canada. (n.d.; accessed November 17, 2026). Child-rearing Provisions. Canada Pension Plan. Source document.

  • UK Government. (n.d.; accessed November 17, 2026). National Insurance Credits: Eligibility. GOV.UK. Source document.

  • UK Government. (n.d.; accessed November 17, 2026). Carer’s Credit: Eligibility. GOV.UK. Source document.

  • Government of Ireland, Department of Social Protection. (2019; updated January 2, 2026). HomeCaring Periods. Source document.

  • Government of Ireland, Department of Social Protection. (2019; updated January 2, 2026). Homemaker’s Scheme. Source document.

  • Swedish Pensions Agency (Pensionsmyndigheten). (n.d.; accessed November 17, 2026). The Swedish Pension System. Source document.

  • Australian Taxation Office. (2026; updated July 24, 2026). Paid Parental Leave Superannuation Contribution. Source document.

  • Services Australia. (2026; updated September 14, 2026). Paid Parental Leave Scheme Changes. Source document.

  • U.S. Social Security Administration. (n.d.; accessed November 17, 2026). The Age You Start Receiving Benefits and the Age You Stop Working. Benefits Planner. Source document.

  • International Labour Organization. (2012). Social Protection Floors Recommendation, 2012 (No. 202). Adopted June 14, 2012. NORMLEX instrument text. Source document.