1. De facto benefit receipt of standard and non-standard workers

Social protection plays a key stabilising role for individuals and societies alike. The recent prominence of social protection in governments’ reform agendas can be seen in the context of unprecedented shocks during the COVID-19 pandemic, heightened uncertainties about the paths of labour-market recoveries and the cost of living, as well as structural transformations driven by digitalisation and other “mega trends”, such as globalisation and climate change. While crises and uncertainties underscore the vital role of social protection, they also highlight the individual and social costs of protection that is ineffective or inaccessible. A future world of work, with less stable career patterns and an emergence of new forms of employment, presents one set of distinct challenges that may erode the prevention, protection or promotion capacities of present-day social protection systems (OECD, 2019[1]; European Commission, 2022[2]; Acemoglu and Restrepo, 2020[3]).

The COVID-19 pandemic has further accentuated structural challenges facing social protection policies (OECD, 2020[4]). Paid sick-leave schemes and unemployment insurance benefits have supported many who have lost their incomes early on during the health crisis. The United States, like many OECD countries quickly expanded benefits and eased access to short-time work schemes. Yet, many emergency measures mostly aided dependent employees. Even in countries with well-developed (or recently reinforced) social protection systems, many workers without standard employment contracts, or with short or unstable work histories, struggled to make ends meet when confronted with a job or earnings loss. Moreover, despite additional support for COVID-related job losses, those who were already out of work before the crisis often faced periods of extended hardship.

To examine the gaps in income support for different groups of workers, this chapter assesses the amount of support that individuals receive when experiencing out-of-work spells, either due to unemployment or to labour-market inactivity. It looks at differences in de facto benefit receipt between wage and salaried- and “non-standard” workers (part-time workers, those on temporary contracts, and self-employed including own-account workers), workers with different racial and ethnic backgrounds, as well as men and women. It compares results for the United States to other OECD countries using a methodology proposed in Immervoll et al. (2022[5]). The approach is based on information on actual (empirical or “de facto”) benefit receipt. It therefore captures the interplay of (i) statutory entitlement rules, (ii) the implementation of these rules across different groups and (iii) the take-up of benefits, which may also vary across groups.

The approach consists of estimating statistical models of benefit receipt while controlling for the most important determinants of social benefit entitlements. The resulting models are then used to “predict” the income support that people receive in specific circumstances (“vignettes”), such as jobless workers with a history of wage and salaried or self-employment.1

This chapter relies on available longitudinal household surveys for 17 OECD countries containing rich information on individual incomes and employment patterns. Administrative records of the universe of workers and sufficiently detailed information on employment history and benefit receipt would have distinct advantages over survey data but are currently not available for most countries.2

In spite of the limitations inherent in available survey data, results are indicative of the approximate patterns of social protection gaps for standard and non-standard workers prior to the COVID-19 crisis. Results therefore point to structural social protection features and challenges that existed already prior to the COVID-19 crisis. The analysis presented here is intended as an illustration using readily available survey data, and as a template for possible future applications with statistically more powerful data sources that may become available for some countries.

This chapter is structured as follows. Section 1.2 provides an overview of the architecture of working-age income support in the United States compared to other OECD countries, to inform the interpretation of the estimates of social protection gaps. To the same end, Section 1.3 provides an overview of statutory access gaps for non-standard workers across the OECD. Section 1.4 provides a short description of the econometric model and the data used for the United States and presents internationally comparable results on social protection gaps between standard and non-standard workers. Section 1.5 presents more granular results on the United States using a model tailored to enable analysing social protection gaps for a larger group of out-of-work individuals, as well as social protection gaps between racial and ethnic groups, as well as between men and women.

Income-support strategies and policy setups differ significantly across countries. Workers in many OECD countries acquire entitlements to earnings-replacement benefits such as unemployment insurance, accident insurance, disability, and parental-leave benefits through contributions. Some groups, e.g. families with children, receive support regardless of income or past employment (universal benefits). In addition, households with limited resources may have access to minimum-income benefits (MIB). This reflects different policy institutions and traditions, but also different strategies for balancing the various objectives of social protection such as risk sharing, income smoothing over time, inequality reduction and poverty alleviation.

Differences in the mix of entitlement criteria across countries are important drivers of social protection gaps between standard and non-standard workers. In 2019, only 11 of 36 OECD countries with available information offered self-employed workers the same unemployment protection as dependent employees, and two of them (Australia and New Zealand) exclusively relied on means-tested income support for jobseekers (OECD, 2022[6]). Several factors make the provision of contribution-based benefits for self-employed workers in particular more complex than for wage and salaried workers (see section 1.3). In contrast, means-tested benefits may be more accessible to non-standard than to standard workers, because they often have lower, and more fluctuating incomes (OECD, 2020[4]). Understanding how the income support architecture in the United States compares to other countries is therefore key for understanding how it compares in terms of social protection gaps between standard and non-standard workers.

Some countries rely very strongly on means-tested benefits for working-age support (e.g. Australia or the United Kingdom, where means-tested or universal child benefits make up the bulk of spending on working-age benefits, see Figure 1.1). Others mainly rely on insurance-based transfers to cushion earnings-losses, with a limited role for means-tested transfers for those who do not have the required contribution history (e.g. Belgium, Italy, Korea or Spain). A third group uses “layered” systems that combine insurance-based out-of-work benefits with universal support for families with children and means-tested benefits as a lower-level safety net (e.g. Austria, France, Germany, the Slovak Republic and to a lesser extent, Hungary). Especially in Austria, Germany and Hungary, universal child benefits account for a significant share of the incomes of working-age households.

This mix of benefits can differ at similar levels of spending. In both France and the United Kingdom, public benefits make up around 8% of the incomes of working-age households (before benefits), but with very different underlying targeting mechanisms (Figure 1.1). The support package in the United Kingdom consists almost entirely of universal and means-tested support, while contributory benefits account for one-third of the support package in France. Similar differences can be seen across Germany and Italy, or Hungary and Spain. In Belgium, where public benefits account for over 10% of the incomes of working-age households, over 80% of payments depend on previous earnings.

In the United States, benefit spending makes up only 2% of the total income of working-age households. Less than one-third of these benefits are contribution based (mainly veterans’ benefits and, to a lesser extent, unemployment compensation).

Looking at the total benefit package for working-age households in the United States, Unemployment Compensation is a relatively minor programme in terms of spending, accounting for only nine percent of overall social spending on working-age households. This is consistent with low coverage and receipt durations of UI (see Chapter 2). With the exception of Veteran’s benefits, which account for 12% of spending for working-age households, all other social spending is means-tested.

The bulk of spending is made up of means-tested programs targeting poor households. The nutritional assistance programmes Supplemental Nutrition Assistance Program (SNAP) and Special Supplemental Assistance Program for Women, Infants and Children (WIC) account for 30% of overall spending,3 and the disability benefit Supplemental Security Income (SSI) for disabled, low-income adults for 25%, followed by the employment-related Earned Income Tax Credit (EITC, 20%). Other social assistance programmes (state-level General Assistance, GA, and Temporary Assistance for Needy Families, TANF) make up 4% of spending.

Consistent with the importance of means-tested benefits, more than half of all working-age social spending in the United States is targeted to poorest 20% of households (Figure 1.2). Unemployment Compensation makes up a negligible share of total benefits among the poorest households (3% of benefit spending in the first income quintile). Last resort benefits (SNAP, WIC, GA, and TANF) make up the bulk of benefit spending on the poorest households, implying that many do not have the contribution history to qualify for unemployment compensation. Disability benefits account for 30% of benefit expenditure in the first quintile, indicating that many in the poorest quintile are in fact unable to work because of ill-health. The receipt rate of disability benefits in the working-age population in the United States is, however, not exceptionally high in the international comparison (at 6%, the same as the average across OECD countries, (OECD, 2022[7])).

Unemployment Compensation payments increase with household income (Figure 1.2). This is consistent with the short maximum receipt duration (around five months) – UI recipients are more likely to have labour income during the year. The benefit is also not means-tested and can therefore be received by those living with other income earners.

Coverage and receipt durations for unemployment benefits in the United States are low in comparison to other countries (see Chapter 2). In line with low overall spending on working-age benefits (see section 1.2.1), the United States also provides less generous non-contributory benefits for those who are not or no longer entitled to UI, through unemployment assistance, family benefits, minimum income benefits or through a combination of different measures. Figure 1.3 shows entitlements for a one-adult household, and a couple with two children, in the 13th month of unemployment, to abstract from short unemployment spells, and to show the effects of MIBs that often compensate once unemployment benefit entitlements run out. As many benefits in the United States vary at the state-level, Figure 1.3 shows results for three states with different labour markets and benefit rules: California, Texas and Michigan. It draws on the OECD TaxBEN model, that assesses statutory entitlements in specific policy-relevant, but hypothetical, household circumstances. For an in-depth analysis of work incentives and benefit levels in the United States in a comparative perspective, see the companion paper to this report, (Pearsall, Pacifico and Magalini, forthcoming[8]).

Across the OECD on average, long-term unemployed adults living alone receive benefits amounting to about a third of median household income, significantly below the relative poverty threshold (incomes below 50% of the median household income, following the standard OECD definition). Net incomes are above the poverty threshold in countries where maximum receipt durations of UI benefits exceed one year, including Switzerland, the Netherlands and Portugal, but benefits are close to the relative poverty threshold also in Luxembourg and Japan, where means-tested social assistance takes the place of unemployment benefits after one year of unemployment. In the United States, SNAP is the only benefit that can be received by a long-term jobseeker without children, leading to very low benefit levels in all three states with available data (Figure 1.3, Panel A). Families with children receive family benefits in many OECD countries, which increase household incomes for workless couples with two children, particularly in Australia, Poland and Canada. In the United States, the only family benefit for jobless families is TANF, that has very low coverage rates in practice (see Box 1.1). However, even for households receiving TANF, total benefit packages in the United States are significantly below the OECD average (Figure 1.3, Panel B).

Statutory access to income support varies by employment type and by programme/branch. Temporary and part-time workers are in principle covered in the same way as permanent full-time employees in most countries and for most risks, as long as they meet minimum contribution periods and earnings thresholds.

By contrast, statutory access for self-employed workers is very frequently restricted. Indeed, contributory social protection systems that were mostly set up with a steady employer-employee relationship in mind do not easily accommodate the self-employed (OECD, 2018[14]):

  1. 1. Double contribution issue: Who should be liable for employer contributions in the absence of an employer? Requiring the self-employed to pay the equivalent of both employer and employee contributions brings formal burdens in line with dependent employees. But effective burdens may be higher for the self-employed, especially those with lower earnings, because minimum wages typically do not apply to them and because they may lack the bargaining power to shift any contribution-related costs onto their clients by charging higher prices.

  2. 2. Fluctuating earnings (and margins for avoiding contribution liabilities): The self-employed are often paid at irregular intervals, either because of time lags between work and payment, or because demand for their services is erratic (ISSA, 2012[15]). This complicates the calculation of contributions (as well as the assessment of entitlements). In particular, self-employed workers may be able to avoid or lower contributions by optimising their contribution base, e.g. through timing their work or earnings.

  3. 3. Moral hazard: Demand or price fluctuations affecting self-employed workers are difficult to distinguish from voluntary idleness and this complicates the provision of unemployment insurance. There is no employer to confirm a layoff and efforts to re-establish a business operation are more difficult to monitor than the search for dependent employment. In addition, earnings levels of the self-employed react with more volatility to market developments, e.g. because there are no minimum wages and downward wage rigidity does not apply to them. If entitled to unemployment benefits, those with poor earnings prospects may therefore have relatively strong financial incentives to wind down their business in order to claim benefits.

Box 1.2 describes statutory entitlements to unemployment and other working-age benefits for non-standard workers to aid the interpretation of the empirical support gaps presented in section 1.4.

This section presents new results on empirically observed income support gaps between standard and non-standard workers in the United States. De facto support gaps are the result of a statistical model of benefit entitlements for jobless individuals that controls for the most important determinants of social benefits. Results are intended as shorthand summaries of benefit accessibility and generosity in a comparative perspective. They also allow quantifying the accessibility and generosity of support packages across different population groups, including standard and non-standard workers.

The approach is detailed in a companion paper to this report, (Immervoll et al., 2022[5]), that presents de facto support gaps for 16 OECD countries. This section therefore only provides the main intuition for the model, focussing on the new results for the United States.

The social protection gaps approach aims to estimate receipt probabilities and benefit levels for a specific set of circumstances, and seeks to control for the key characteristics that determine benefit receipt. As benefit access and amounts often depend on past events, the method relies on longitudinal household data that include information on current and past employment, earnings and other relevant individual and family characteristics. The drawback of using easily accessible survey data is comparatively small sample sizes, which can make the analysis of subgroups – such as the out-of-work population with a history of non-standard work – problematic. Data from administrative sources would be more appropriate for this type of analysis, but is currently not readily available for comparative work of the type proposed here.

The main variable of interest is the value of the total benefit package, rather than any individual category of social transfer, reflecting the fact that countries provide support through different channels and programmes. The policy scope comprises the most important social transfers to working-age individuals and their families: unemployment and disability benefits, (employer as well as publicly provided) sick pay,4 family (including maternity) benefits, any benefits tied to education (such as public student aid), in-work and minimum income benefits (means-tested transfers aimed at reducing poverty, most importantly social assistance and housing benefits).5 For the United States, the analysis furthermore includes Supplemental Nutrition Assistance Program (SNAP) and Special Supplemental Nutrition Program for Women, Infants and Children (WIC).6 For details on the benefits included for the United States, see Annex 1.A.

The variable of interest also encompasses support provided through the tax system that is akin to cash benefits (such as refundable child or in-work tax credits) when these are reported in the data.7 For the United States, EITC amounts are imputed for individuals who report EITC receipt in the data; however EITC coverage is underestimated because of survey underreporting, as well as inconsistencies between reported EITC receipt and previous year’s earnings, see Annex 1.B for details. Receipt of the Child Tax Credit is unfortunately not reported in the US data, and the analysis therefore does not account for CTC receipt. Estimated benefit receipt incidence and amounts will therefore underestimate the true extent of income support for households with children in the United States (see Annex 1.B).

Benefit receipt is measured over an entire year and therefore accounts for both generosity in a given month, any benefit reductions for longer out-of-work spells, and the effective duration of entitlements (including any waiting periods or other possible gaps between benefit entitlement and pay-out). For benefits that are observed/reported at the household rather than the individual level (family benefits, minimum income benefits), amounts are divided equally across all adult household members on a per-capita basis.

The empirical assessment of income support for different labour market groups proceeds in two steps. A first step estimates the relationship between individual benefit receipt and a number of key structural drivers of support. The model specification includes the following independent variables, along with relevant interactions and higher-order terms: Main employment status and pre-transfer household income during the reference period (year 0), main employment status and earnings during the two years preceding the reference period (years -1 and -2), household composition in year 0, including the presence of dependent children (plus children under the age of six to capture maternity/paternity benefits), as well as health status, housing tenure and housing costs, sex and age (all year 0). See Figure 1.5 and Figure 1.6.

Separate models are estimated for benefit receipt (yes/no) and benefit levels (benefit amounts), see (Immervoll et al., 2022[5]) for details. A second step uses the estimated relationships for inference on the benefit gaps between standard and non-standard workers in specific concrete circumstances (“vignettes”) that are defined in a consistent way across countries. The use of a vignette-based analysis facilitates the communication of complex statistical results in a comparative setting, and the identification of possible policy mechanisms driving entitlement gaps.

Benefit “gaps” for non-standard workers are calculated relative to a baseline standard worker, who is likely to require out-of-work support. This baseline standard worker is an individual who was out of work (either unemployed or labour-market “inactive”) for at least six months during the reference year, lives in a low-income household in the reference period (bottom 20% of the national distribution), and has neither significant health problems, nor young children under the age of six.8 In the two years prior to the reference period, the baseline standard worker was a dependent full-time employee with earnings at or above the 40th percentile of the national distribution.9 The comparator vignette is an otherwise similar individual, with a history of “non-standard” work, detailed in the notes to Figure 1.6, as well as the figures showing model results.

It is important to note that the analysis looks at a subset of out-of-work individuals who are both persistently out of work (for at least six months in year 0) and may not be actively seeking work, which is a requirement for receiving unemployment benefits in all countries barring special circumstances.10 It also focuses on individuals who are in clear need of support (in the bottom 20% of the income distribution before benefits), and therefore likely to satisfy means-tests.

Model estimates for the United States11 are based on the first three waves of the 2014 Panel of the Survey of Income and Programme Participation (SIPP), covering labour market histories and incomes in the years 2013, 2014 and 2015.12 While more recent waves of SIPP data are available, a redesign of the survey means that it is difficult to obtain three consecutive waves of data.

While in principle it is possible to link the three waves of the 2018 SIPP survey (corresponding to the years 2018, 2019, and 2020), low response rates in 2019 and 2020 led the Census Bureau to issue a warning for researchers to use the data with caution. In 2019, a lapse in survey funding disrupted the data collection process, while the 2020 survey relies exclusively on telephone-based interviews due to the COVID-19 pandemic, making it difficult for survey administrators to follow up on non-responders. In addition to data collection issues, the COVID-19 emergency measures implemented in 2020 represent an extraordinary situation that was not the focus of the original cross-country analysis. This means that the data from the 2014 SIPP survey remains the most appropriate for use in the analysis.

The sub-sample of interest comprises all working-age individuals aged 18-64 who are potentially in need of working-age support: individuals who (i) have not worked for pay or profit during the majority (six months or more) of the observation/reference period (year 0: 201613 for the United States, 2018 for all European countries and Korea, 2019 for Australia), (ii) are not already retired,14 and (iii) were not in education or military service in year -1 (and thus had the opportunity to accumulate entitlements to any insurance-based benefits). This is typically a small share of the working age population, with effective sample sizes for the 1st stage coverage model ranging from fewer than 2000 observations in Austria, Korea and the Baltic countries, to more than 6 000 observations in Australia, Greece, Italy, Spain. For the United States, the estimation sample comprises 3 430 unique individuals, representing a population of about 28 Million, see Annex Table 1.C.1.

The descriptive statistics in Annex Table 1.C.1 show that women are strongly over-represented in the out-of-work estimation sample, accounting for two-thirds of out-of-work individuals. This tendency is slightly more pronounced in the United States compared to the cross-country average. Furthermore, compared to other OECD countries, out-of-work individuals in the United States are less likely to be in the bottom 20% of the income distribution, and more likely to be in the top 20%. Only 14% of the sample population are unemployed, with the remainder being inactive, compared to 34% in the cross-country average (not shown). The share of unemployed individuals in the sample is only significantly lower in Korea (4%). A high share of economically inactive women living in households in the middle and upper end of the income distribution is in line with the male-breadwinner model.

In line with the categorisation of the (partial) simulations of unemployment benefits in Chapter 2, part-time workers are wage and salaried workers working below 35 hours per week. Self-employed workers include workers in the residual category of “other work arrangement” according to guidance from the Census Bureau that these workers are mostly independent contractors or consultants. Workers who are wage and salaried workers as well as self-employed are categorised as self-employed if their earnings from self-employment in a given month are higher than their wage or salaried income (and vice-versa). For individuals with missing earnings the simulations use the number of hours worked as either a wage or salaried worker or as self-employed to determine the primary status (see also Box 2.1 in Chapter 2).

Raw receipt rates of social benefits are significantly lower than in other countries – 49% in the United States compared to 68% across countries on average. Previous standard workers – who worked mostly full-time as wage- and salaried employees during the two years before the income reference period – were less likely to receive benefits than the overall out-of-work sample, consistent with the prevalence of means-tested benefits in the United States (see section 1.2).

The presentation of results starts out by discussing receipt patterns for the standard worker “baseline vignette”: an able-bodied individual who has been out-of-work for six months or more, and who has a record of stable full-time wage or salaried employment, and whose income before benefits put them into the bottom 20% of the income distribution. The results therefore show benefit entitlements for individuals who are both in clear need of support as well as “deserving” in having made prior contributions. The support available for standard workers is indicative of cross-country differences in income support architectures, and useful for building intuition for the resulting drivers of support patterns.

In Belgium and France, the “baseline” standard worker’s chance of receiving support is 95% or more, compared to less than 60% in Greece and Italy, and below 50% in Korea and the United States (“baseline: past standard work” in Panel A, Figure 1.7). In most other countries, the share is about 70%-80%.

The countries with the lowest estimated receipt probabilities, the United States, Korea and Greece, also have the lowest levels of spending on working-age transfers in the analysed countries (under 4%, see Figure 1.1). Working-age transfer spending is somewhat higher in Italy where the benefit receipt probability was also under 60%, but benefits are largely insurance based and thus less targeted to low-income groups (for instance, in Italy in 2018, 43% of all working age benefits went to the top income quintile according to the OECD Income Distribution Database). In contrast, in the countries with the highest receipt probabilities, Belgium and France, government transfers represent over 10% of the incomes of working-age households. In Belgium, unemployment benefit receipt durations are in principle not limited (see section 1.3), while France and Germany combine insurance-based unemployment benefits with means-tested unemployment and social assistance programmes.

For the United States, the fact that the vignette is defined to have no significant health problems may also contribute to the low receipt probability, given the importance of (means-tested) disability benefits for low-income households (see Figure 1.2).

Yet, the low benefit receipt probability for a long-term out-of-work individual with a recent history of employment is likely connected to the overall low coverage rates for unemployment benefits in the United States (see Chapter 2). Unlike in many other OECD countries, standard workers who are “voluntarily” unemployed, are not entitled to unemployment benefits. Statutory benefit receipt durations are also comparatively short (up to 26 weeks, depending on the state of residence, see Chapter 2). Last-resort benefits, in the case of the United States mainly SNAP, General Assistance and TANF, are therefore the only benefit that many jobless workers may claim. These benefits can be associated with social stigma, and may have strict means- and asset tests, leading to low take-up (see Box 1.1 and (Hyee et al., 2022[20])).

The predicted average sizes of overall benefit packages also vary enormously across countries. They range from under 20% of the national median household income in Korea, the United States and Greece, to around 30% or less in parts of Central and Eastern Europe (Lithuania, Estonia, Latvia, and Hungary), Germany and Australia, 35 to 40% in Poland and the United Kingdom, and 40 to 50% in Spain, Portugal, Italy, and France. At 60% of median income, estimated benefit levels are highest in Belgium (Panel B of Figure 1.7).

In terms of country rankings, these benefit levels are broadly in line with “theoretical” benefit entitlements, as calculated with policy simulation models (e.g. http://oe.cd/TaxBEN). For a number of reasons, however, actual values as estimated here differ from – and are generally lower than – theoretical levels implied by headline indicators for “typical workers”, such as replacement rates at the beginning of an unemployment spell:

  1. 1. De facto estimates are based on actually observed spells of joblessness. Unlike “typical worker” replacement rates, the resulting entitlements reflect the characteristics of those experiencing job loss, such as past earnings histories. Since those with lower earnings or shorter career histories tend to be over-represented among job losers, the resulting entitlements to any earnings-related insurance benefits can be noticeably lower than those of an “average” worker.

  2. 2. Results refer to support received over an entire calendar year. They therefore capture differences in benefit amounts in a given month, in benefit duration limits, and in the average duration of out-of-work spells. The latter varies across countries, even among the selected sample with jobless spells of six months or longer. For instance, the average spell duration for previous standard workers who have been jobless for six months or longer in the United States is almost two months shorter than in Italy (see Annex Table 1.C.1 for the United States and Table B-1 in the online Annex15 of (Immervoll et al., 2022[5]) for other OECD countries with available information).

Estimates of de facto benefit levels refer to recent job losers receiving any type of cash support. This can include people who do not receive out-of-work benefits, but transfers of a lower value, such as housing benefits or nutritional support in the case of the United States or “universal” benefits (e.g. jobless people with children may receive universal child benefits in Austria and Germany). In the case of the United States, unemployment compensation plays a minor role for the selected vignette (who is in the bottom income quintile), as the majority of payments are received by households further up the income distribution (Figure 1.2).

With that in mind, the (comparatively small) group of out-of-work Italians with past standard employment who do qualify for benefits receive significantly more generous support on average (over 40% of median household income) than, for example, an equivalent individual in Australia, where (flat-rate and means-tested) benefits amount to about 20% of median household income. In both cases (and in most other countries), those relying on benefit income alone would typically have income below commonly used relative poverty cut-offs, but poverty gaps would be significantly bigger for benefit recipients in Australia.

Across countries, there is no obvious general link between accessibility and generosity. As noted, benefit access in Italy is comparatively difficult, but benefit levels for recipients are higher than in the majority of other countries. Hungary, Germany and the Baltic countries follow the opposite pattern, with implied coverage above 80%, but with comparatively low benefit levels around 30% of median household incomes. Accessibility and generosity scores are both high in Belgium.

In the United States, Korea and Greece, accessibility and generosity scores are both low. In the United States and Korea, low annual support levels are partly driven by short durations of (unemployment) benefits as noted above. In addition, effective minimum-income entitlements also tend to be lower than in many other countries (http://oe.cd/TaxBEN).

In the United States, as well as four other countries (Austria, Germany, Hungary and the United Kingdom) both coverage and generosity gaps between standard and non-standard workers are statistically insignificant. In Australia and Belgium, access gaps are statistically insignificant, and receipt probabilities are at around 70% or above for both the standard and non-standard vignettes. While results for France and Spain point to statistically significant gaps, with somewhat lower point estimates for the implied coverage for non-standard workers, receipt probabilities for both types of worker are also above 70%. As these eight countries follow very different social protection strategies, these results suggest that accessible support for non-standard workers is achievable with different targeting mechanisms.

Out-of-work support in the United States, Australia and the United Kingdom is largely means-tested (and therefore unrelated to previous employment and earnings, see section 1.2); in the case of the United States, support is almost entirely means-tested for low-income households (Figure 1.2). By contrast, Hungary and Belgium offer earnings-related unemployment protection to both standard and non-standard workers.

A finding of small or insignificant gaps in the protection afforded to standard and non-standard workers in such a diverse set of countries is notable. For instance, it raises questions about recent prominent calls for a strong reliance on means-tested safety-net benefits, or for a universal basic income, that are sometimes motivated by concerns that insurance-based systems cannot provide effective protection for non-standard workers (World Bank, 2018[21]; Gentilini et al., 2019[22]; Browne and Immervoll, 2017[23]).

In Hungary, non-standard workers, including the self-employed, are entitled to unemployment benefits (Albert, Gáspár and Gal, 2017[24]). In Belgium, non-standard workers can qualify for unemployment insurance support though benefit amounts are much more generous than in Hungary, and benefits for self-employed workers in Belgium account also for household needs (De Wispelaere and Pacolet, 2017[25]). In both countries, means-tested support provides further layers of protection for those not entitled to insurance benefits. Austria and Germany also combine a first-tier unemployment insurance system with a second layer of means-tested support. In France, a key explanation for the insignificant coverage gaps is the very short qualification period for unemployment benefits, paired with the possibility to retain unused benefit entitlements for future out-of-work periods, and to cumulate benefit rights across successive out-of-work spells for the (large and growing number) of workers with short-duration employment contracts.16 Like Austria, Belgium and Germany, France also provides multi-layered income support that benefits workers across different types of non-standard employment (as well as others who may not qualify for first-tier insurance benefits).

Implied access gaps are largest in Korea, Portugal and Italy, where standard workers were between 50% (Italy, Portugal) and 100% (Korea) more likely than non-standard workers to receive income support following a job loss (Figure 1.7, “past non-standard work”). Gaps are also large in Latvia, Lithuania and Estonia.

The example of Portugal, in particular, illustrates the need to consider effective access in addition to statutory entitlements: Portugal has one of the biggest access gaps of all considered countries, mostly driven by the low coverage of self-employed workers (see Annex 1.D). This is despite unemployment benefits being open to owners of businesses and independent contractors with only one client (Perista and Baptista, 2017[26]). But not all self-employed workers have access (e.g. unincorporated self-employed workers, or those working for more than one client), and the required contribution period for self-employed workers is twice as long as for employees. Self-employed workers also have legal access to cash sickness benefits, but the maximum entitlement period is one-third of the duration for employees. These factors result in limited effective access to cash support for self-employed workers in Portugal, even though they do have better statutory protection than in other countries.

Statutory entitlement rules vary significantly across different types of non-standard work (section 1.3). Self-employed workers for example may not be covered at all for certain types of risks, while part-time workers and those with interrupted/unstable work histories may suffer reduced effective access, because they fail to meet the required earnings or contribution histories. Results on the gaps between standard workers and a heterogeneous group of all non-standard workers may therefore mask significant differences between different types of non-standard employment. Understanding these differences is necessary for designing tailored policy strategies for tackling unintended gaps.

The analysis for granular employment types is only possible for a sub-group of countries where sample sizes allow such disaggregation (see Annex Table 1.C.1 for the United States and Table B-1 in the online Annex17 of (Immervoll et al., 2022[5]) for other OECD countries with available information). For the United States, sample sizes are sufficient to estimate gaps for all granular employment types.

Results highlight that self-employed workers (including independent contractors) are typically least likely to receive support, while gaps are less common, and tend to be smaller, for part-time and unstable workers (see Annex 1.D). Accessibility gaps for those with past self-employment are sizeable in three of the six countries considered: In Portugal, Spain and Italy, implied coverage gaps (the difference of estimated receipt probabilities between previous standard and self-employed workers) are around 50 percent. In Italy and Spain, self-employed workers do not have access to unemployment benefits, and in Portugal, access is incomplete. Self-employed workers thus have to rely on lower-tier income support such as social assistance and housing benefits, which typically feature strict eligibility requirements including income and asset tests, and are subject to significant non-take-up, lowering their effective reach. For instance, receipt of means-tested support is particularly low in a number of southern European countries. Minimum-income benefits also tend to be less generous than social insurance transfers.

In Poland, self-employed workers can receive unemployment benefits, but only after a 90-day waiting period (compared to seven days for dependent employees). The benefit is not linked to previous earnings, which explains the small and insignificant gap between self-employed and standard workers in access, and the comparatively small gap in generosity.

Access gaps for part-time workers are less common, in line with the statutory entitlement results, affecting only three out of 11 countries (see Annex Figure 1.D.2, Panel A). Benefit levels for part-time workers were significantly lower than for (full-time) standard workers in six of the 11 countries considered, in line with the strong previous earnings link that shapes entitlements in many unemployment benefit programmes (Annex Figure 1.D.2, Panel B). Gaps in levels were largest in Southern European countries where insurance-related benefits dominate. In Australia, somewhat higher benefit payments to those with previous part-time work likely reflect the importance of means-testing in combination with lower household incomes of households with (past) part-time work.

In Italy and Poland, those with interrupted work histories are less likely to receive out-of-work support than standard employees (Annex Figure 1.D.3, Panel A). In some countries, workers can qualify for unemployment insurance support after comparatively short periods in work, e.g. three months in France. In Austria, qualification periods are shorter for workers with repeated unemployment spells (such as seasonal workers). And in some countries, jobseekers are able to keep unused unemployment benefit entitlements for future claims if they found work prior to benefit expiration, among them Australia, Austria, France, Spain and the United Kingdom. This is, however not the case in Latvia and Poland. In France, a recent reform in 2021 has reduced entitlements of workers with short contracts and repeated unemployment spells by taking out-of-work spells into account when assessing the earnings base for benefit entitlements.

For the United States, effective protection gaps are insignificant for all granular employment types. This is connected to the specific mix of benefits in the United States and the choice of vignette – a low-income individual who has been persistently out of work (at least six months over the reference year) and is therefore in clear need of support:

  • Means-tested benefits make up the bulk of payments to working-age households in the United States, in particular at the bottom end of the income distribution (see section 1.2).There are typically no access gaps to means-tested benefits for non-standard workers; in fact, means-tested benefits may be more accessible and generous for non-standard workers (see Box 1.2).

  • The only benefit that depends on past employment – unemployment compensation – has a maximum receipt duration of 26 weeks depending on the state (see Chapter 2).

    Because of data limitations, it is not possible to perform the above analysis for jobless individuals with shorter out-of-work durations for European countries.18 Looking at shorter unemployment durations is, however, possible with the SIPP. Section 1.5 presents an adaption of the econometric model that allows also looking at shorter episodes of joblessness.

Because benefit receipt is reported in the SIPP on a monthly basis, it is possible to adjust the model described in section 1.4 to include shorter out-of-work spells (below six months). Since the maximum duration of unemployment compensation is 26 weeks (depending on the state, see section 2.3), this approach is more suitable to detecting differences in unemployment compensation receipt. Including shorter out-of-work spells also increases the size of the sample of out-of-work individuals, which enables a more granular look at de facto access gaps for racial and ethnic groups as well as women and men.

As for the econometric model described in section 1.4 and in more detail in (Immervoll et al., 2022[5]), the sample consists of working-age, non-retired individuals with a complete employment history over the years 2014 – 2016.19 To guarantee that the model captures all relevant work and earnings history that determines eligibility to unemployment compensation, the model contains two groups of jobless individuals:

  • Individuals who have been out-of-work for the entire duration of the panel – 2014-16, either unemployed or labour-market inactive. These individuals are persistently out-of-work.

  • Individuals who have worked for the first 12 months of the panel and became jobless – either unemployed or labour-market inactive – at some point between month 13 and month 36 of the panel. The 12-month observation period is chosen to coincide with the assessment period for unemployment compensation (the period that is used to calculate unemployment compensation amounts, see section 2.3). The model categorises these workers as full-time, part-time, or self-employed if, out of the 12 months directly preceding the out-of-work spell, they were full-time or part-time wage and salaried workers, or self-employed, for at least nine months. Workers who combined full-time/part-time/self-employed work and were in neither status for more than eight months are classified as hybrid workers.

The left-hand-side variable of interest is the monthly average of the total package of working-age benefits as in the model described in section 1.4, calculated over the entire spell.20 The regression models determining benefit receipt (yes/no) and benefit amounts are equally similar to the comparative model, but only control for earnings over the 12 months preceding the start of the out-of-work spell, in-line with the short assessment period for unemployment compensation in the United States compared to other countries (see section 2.3). Deciles of current household incomes are similarly calculated on a monthly basis over the entire spell, see Annex Table 1.D.1 for descriptive statistics, and the table notes for details on the calculation of explanatory variables of the econometric model.

In line with the comparatively short duration of UI benefits in the United States, differences between standard and non-standard workers appear in the US-specific model that includes individuals with shorter (one to six months) spells of joblessness. Previous part-time, self-employed or hybrid workers are about 10 percentage points, or around 20%, less likely to receive any benefits during an out-of-work spell than previous full-time, wage or salaried workers Figure 1.8.21 The gap for self-employed workers is not significant, likely because point estimates are less precise because of the lower number of self-employed workers in the sample, see Annex Table 1.D.1. Estimated generosity gaps are significant for part-time workers, in line with lower earnings, and therefore lower entitlements to UI.22

Larger sample sizes in the US-specific model also enable an analysis of social protection gaps for non-Latino whites,23 African Americans (including Latino African Americans) and Latinos. Sample sizes are insufficient to analyse social protection gaps for Asian Americans and other racial or ethnic minority groups (including mixed-race individuals).

For a given employment history (at least 12 months of standard or non-standard work immediately preceding the start of the out-of-work spell with earnings at or above the 40th percentile of the earnings distribution), there are no statistically significant differences in the de facto probability of receiving any benefits between non-Latino white and African American or Latino previous standard and non-standard workers (Figure 1.9, Panel A and B, left hand side). African American previous standard workers have significantly lower estimated benefit entitlements than white standard workers (10% compared to 15% of median household income, Figure 1.9, Panel A, right hand side) whereas Latino previous non-standard workers have significantly lower entitlements than white previous non-standard workers (6% compared to 11% of median household income, Figure 1.9, Panel B, right hand side). This is most likely due to lower previous earnings of African American and Latino previous standard and non-standard Latino workers (see descriptive statistics in Annex Table 1.D.1).

These results however only indicate that, for workers with similar employment histories, benefit receipt patterns are comparable among racial and ethnic groups. They do not inform about differences in employment patterns between groups: for instance, Latino jobseekers are more likely than non-Latino whites to have worked part-time before becoming unemployed, and African American jobseekers are more likely to have been self-employed (see Figure 2.11). Thus, as previously non-standard workers, they are less likely to be covered by benefits in the first place (Figure 1.8 and Figure 1.9). African Americans are also over-represented in the group of long-term unemployed workers, who do not have access to more generous unemployment compensation (see Chapter 2).

Holding previous work status constant may therefore mask larger labour market inequalities, with knock-on effects on benefit access and poverty. Chapter 2 looks in more detail at racial and ethnic differences in previous employment patterns for current jobseekers. There are no significant gender differences in benefit receipt patterns (not shown).


[3] Acemoglu, D. and P. Restrepo (2020), “Robots and Jobs: Evidence from US Labor Markets”, Journal of Political Economy, Vol. 128/6, pp. 2188-2244, https://doi.org/10.1086/705716.

[9] Aizer, A., H. Hoynes and A. Lleras-Muney (2022), “Children and the US Social Safety Net: Balancing Disincentives for Adults and Benefits for Children”, Journal of Economic Perspectives, Vol. 36/2, pp. 149-174, https://doi.org/10.1257/jep.36.2.149.

[23] Browne, J. and H. Immervoll (2017), “Mechanics of replacing benefit systems with a basic income: comparative results from a microsimulation approach”, The Journal of Economic Inequality, Vol. 15/4, pp. 325-344, https://doi.org/10.1007/s10888-017-9366-6.

[19] ESPN (ed.) (2017), ESPN Thematic Report on Access to social protection of people working as self-employed or on non-standard contracts - Italy, European Commission.

[25] ESPN (ed.) (2017), ESPN Thematic Report on Access to social protection of people working as self-employed or on non-standard contracts: Belgium, European Comission.

[24] ESPN (ed.) (2017), ESPN Thematic Report on Access to social protection of people working as self-employed or on non-standard contracts: Hungary, European Commission.

[26] ESPN (ed.) (2017), ESPN Thematic Report on Access to social protection of people working as self-employed or on non-standard contracts: Portugal, European Commission.

[2] European Commission (2022), Access to social protection for workers and the self-employed. (Partial) Update of the monitoring framework - 2021, https://ec.europa.eu/social/main.jsp?catId=738&langId=en&pubId=8358&furtherPubs=yes.

[30] European Commission (2022), Open data maturity report 2021, https://doi.org/10.2830/394148.

[22] Gentilini, U. et al. (2019), Exploring Universal Basic Income: A Guide to Navigating Concepts, Evidence, and Practices, Washington, DC: World Bank, https://doi.org/10.1596/978-1-4648-1458-7.

[12] Hoynes, H. and D. Schanzenbach (2018), “Safety Net Investments in Children”, Brookings Papers on Economic Activity, Vol. 2018/1, pp. 89-150, https://doi.org/10.1353/eca.2018.0001.

[20] Hyee, R. et al. (2022), “How reliable are social safety nets?: Value and accessibility in situations of acute economic need”, IZA Discussion Papers 15232, https://docs.iza.org/dp15232.pdf.

[5] Immervoll, H. et al. (2022), “De-facto gaps in social protection for standard and non-standard workers: An approach for monitoring the accessibility and levels of income support”, OECD Social, Employment and Migration Working Papers, No. 271, OECD Publishing, Paris, https://doi.org/10.1787/48e282e7-en.

[15] ISSA (2012), Handbook on the extention of social security coverage to the self-employed, ISSA, https://ww1.issa.int/sites/default/files/documents/publications/2_handbook-extension-selfemployed-26571.pdf.

[7] OECD (2022), Disability, Work and Inclusion in Slovenia: Towards Early Intervention for Sick Workers, OECD Publishing, Paris, https://doi.org/10.1787/50e655b3-en.

[18] OECD (2022), Disability, Work and Inclusion: Mainstreaming in All Policies and Practices, OECD Publishing, Paris, https://doi.org/10.1787/1eaa5e9c-en.

[6] OECD (2022), OECD Employment Outlook 2022: Building Back More Inclusive Labour Markets, OECD Publishing, Paris, https://doi.org/10.1787/1bb305a6-en.

[28] OECD (2021), After the pandemic: what model for social protection? Project Proposal, https://one.oecd.org/document/DELSA/ELSA(2021)11/en/pdf.

[4] OECD (2020), “Supporting livelihoods during the COVID-19 crisis: closing the gaps in safety nets”, ELS Policy Brief on the Policy Response to the COVID-19 Crisis, http://oe.cd/il/covid19briefsupport.

[1] OECD (2019), “Left on your own? Social protection when labour markets are in flux”, in OECD Employment Outlook 2019: The Future of Work, OECD Publishing, Paris, https://doi.org/10.1787/bfb2fb55-en.

[29] OECD (2018), Open Government Data Report: Enhancing Policy Maturity for Sustainable Impact, OECD Digital Government Studies, OECD Publishing, Paris, https://doi.org/10.1787/9789264305847-en.

[14] OECD (2018), The Future of Social Protection: What Works for Non-standard Workers?, OECD Publishing, Paris, https://doi.org/10.1787/9789264306943-en.

[8] Pearsall, E., D. Pacifico and E. Magalini (forthcoming), “Unemployment benefit reforms to support employment and inclusiveness in the United States”, Social, Employment and Migration Working Papers, https://www.oecd.org/employment/listofsocialemploymentandmigrationworkingpapers.htm.

[17] Raitano, M. (2018), “Italy: Para-subordinate workers and their social protection”, in The Future of Social Protection: What Works for Non-standard Workers?, OECD Publishing, Paris, https://doi.org/10.1787/9789264306943-9-en.

[13] Shrivastava, A. and G. Thompson (2022), TANF Cash Assistance Should Reach Millions More Families to Lessen Hardship, https://www.cbpp.org/research/income-security/tanf-cash-assistance-should-reach-millions-more-families-to-lessen (accessed on 1 February 2023).

[16] SSA and ISSA (2017), “United States of America”, in Social Security Programs Throughout the World: The Americas, Social Security Administration, https://www.ssa.gov/policy (accessed on 2 December 2018).

[11] The Urban Institute (2021), Welfare Rules Databook: State TANF Policies as of July 2019, https://www.urban.org/research/publication/welfare-rules-databook-state-tanf-policies-july-2019.

[27] U.S. Department of Commerce, Economics and Statistics Administration, U.S. Census Bureau (2019), Survey of Inocme and Program Participation. 2014 Panel Users’ Guide. 2019 Edition., https://www2.census.gov/programs-surveys/sipp/tech-documentation/methodology/2014-SIPP-Panel-Users-Guide.pdf.

[21] World Bank (2018), World Development Report 2019: The Changing Nature of Work, Washington, DC: World Bank, https://doi.org/10.1596/978-1-4648-1328-3.

[10] Ziliak, J. (ed.) (2016), “Temporary Assistance for Needy Families”, in Economics of Means-Tested Transfer Programs in the United States, Volume I, University of Chicago Press, https://doi.org/10.7208/chicago/9780226370507.001.0001.

The analysis is restricted to the working age population, and considers the following government transfers:

  • Unemployment Compensation benefits,

  • Sickness benefits,

  • Disability benefits (Supplemental Security Income SSI and veterans’ benefits),

  • Earned Income Tax Credit (EITC), partially imputed for the analysis (see below),

  • Supplemental Nutrition Assistance Program (SNAP),

  • State-level General Assistance (GA) benefits,

  • Temporary Assistance to needy Families (TANF), and

  • Special Supplemental Nutrition Program for Women, Infants and Children (WIC).

With the exception of the EITC (see Annex 1.B), all benefit receipt information is taken directly from (self-reported) benefit receipt information in the SIPP (see Annex 1.C).

SNAP, GA, TANF and WIC are household-level benefits. In the SIPP, they can be received by an individual person, subset of the household or the entire household. As such, it is possible for multiple programme units (and received benefit amounts) to exist within one household (U.S. Department of Commerce, Economics and Statistics Administration, U.S. Census Bureau, 2019[27]).

The SPgaps approach assumes that the entire household benefits from these payments. Household level benefits are therefore summed up over the entire household and assigned to the adult members of the household in equal shares. This leads to higher benefit receipt rates compared to the United States Census Bureau Statistics (see Annex Table 1.A.1). Differences are especially large for TANF and WIC, family benefits that are likely to be received by those living in larger households. In contrast, receipt rates for individual benefits such as Unemployment Compensation or disability benefits are virtually identical to Census Bureau calculation based on the entire SIPP. It is reassuring that the sample exclusions necessary for the analysis do not lead to a distortion of overall benefit receipt rates.

A particularity of the income support architecture in the United States is its comparatively heavy reliance on tax credits as a tool for working age income support. As the Social protection gaps analysis aims to compare the effective reach of income support across countries, including tax credits in the total package of income support as far as possible was a priority.

As a tax credit, the EITC amounts are not directly reported in the SIPP. However, the SIPP does include a question on whether an individual received an Earned Income Tax Credit (EITC) on their tax form. Because de facto receipt of the EITC is recorded in the SIPP, and amounts are deterministic, the Secretariat decided to impute EITC amounts based on self-reported recipients’ previous year’s income and the number of dependent children in a given household. As the income reference year for this analysis covers receipt during the calendar year 2016, the EITC amounts were imputed using policy rules as of the 1 January 2015.

While the imputation was straightforward, there were some discrepancies between self-reported incidence of EITC receipt and entitlement based on past income and household composition. About a third of all individuals who reported receiving the EITC had imputed amounts of zero because their previous year’s earnings were either too low or too high to be entitled to the tax credit. Self-reported EITC receipt was also lower than according to tax records (Annex Table 1.B.1).

The volume of EITC payments was somewhat lower in the imputation than in administrative statistics, with the average imputed value 13% lower than according to IRS tax returns data (Annex Table 1.B.1). Underreporting of the EITC is therefore likely to contribute to the overall low receipt rate of income support in the United States as compared to other countries.

Similarly, the Child Tax Credit (CTC) is an important programme for working parents: in 2017, 28.2 million recipients received USD 998 on average.24 CTC receipt is not recorded in in the SIPP either, but unlike the EITC, the SIPP does not contain information on whether the CTC was claimed. While it would have been possible to impute the CTC from statutory entitlement rules, this would be against the purpose of the social protection gaps approach, as the goal is to measure and compare de facto benefit receipt, taking into account differences in actual programme implementation as well as non-take-up. The analysis therefore does not account for CTC receipt, and estimated benefit receipt incidence and amounts will therefore underestimate the true extent of income support for households with children in the United States.


← 1. Self-employed workers are all non-wage or salaried workers including independent contractors, see section 1.4.2 for details on the data used.

← 2. National and international open-data and digital government initiatives seek to facilitate the preparation and accessibility of such data sources for research purposes, including in the social policy domain (OECD, 2021[28]; OECD, 2018[29]; European Commission, 2022[30]). Yet, no comparative cross-country database of individual-level administrative data is currently available.

← 3. The Supplemental Nutrition Assistance Program SNAP makes up the bulk of spending on nutritional assistance; the Special Supplemental Nutrition Program for Women, Infants and Children (WIC) for pregnant, breastfeeding, or post-partum mothers and children under the age of five is much smaller.

← 4. Information on sickness benefits is not available for Germany, while data for other European countries as well as Australia include employer provided sick pay whenever available. Receipt information on employer-paid sickness benefits is not available in the KLIPS.

← 5. The unemployment benefit variable includes severance payments for all countries except the United States; they are quantitatively important in Korea and in some Southern and Eastern European countries. The SIPP does not contain any information on severance payments.

← 6. The Social Protection Gaps analysis generally excludes in-kind benefits (such as free school-meals or subsidised housing) because cross-country comparisons between benefits in-kind are difficult to impossible. However, it considers SNAP and WIC to be close enough to cash to include it in the analysis.

← 7. Note that the package of working-age benefits in the United Kingdom and Korea includes refundable, income-related child and in-work tax credits, whereas related programmes are not recorded as social transfers in some other countries. Receipt of means-tested tax credits in Korea may be somewhat under-reported (both the Earned Income Tax Credit, and also the Childcare Tax Credit, though the latter is not relevant for adults living alone).

← 8. The statistical model controls for age (including a higher order term), gender, education, household composition such as household size, presence of a partner and dependent children (under the age of 18) and young children (under the age of 6), as well as housing tenure and rent paid. The “vignette” only specifies the presence of children under the age of six, previous work status and earnings, income, health status, and that the worker worked for 12 months in the year before the reference period. This reflects the trade-off between the sample size and good comparisons across countries.

← 9. In year -1, the baseline standard worker worked for the entire 12 months (this is to ensure recent contribution periods for contributory unemployment benefits), with at least six of them in full-time dependent employment, or five months in full-time dependent employment and three months in part-time dependent employment. In year -2, the baseline standard worker was in full-time dependent employment for at least six months, and out of work during at most two months.

← 10. E.g. in Austria, pregnant women are sometimes not required to actively seek work.

← 11. For details on the data sources for the other countries, see (Immervoll et al., 2022[5]).

← 12. The activity calendar and income in the SIPP always refers to activities and income in the year prior to the interview. Thus, the SIPP 2014 interviewed households in the years 2014 – 2017, and therefore holds labour market and income history for 2013-16. The first three waves therefore indeed cover the period 2013 – 2015, with interviews one year later. This report however follows the convention of using the interview, not the income reference year used by other publications to facilitate comparisons.

← 13. Income and labour-market information pertain to 2015, but interviews were conducted in 2016, see endnote 12.

← 14. Individuals are defined as retired if they receive an old-age pension during the reference year. This can lead to imprecisions in countries where pensions are provided as a lump sum and therefore a pension income stream cannot be observed. In the country sample for the present study, this is only the case for Australia. Any means-tested old-age pension payments, veteran’s pensions etc. can however be identified in the Australian data source (HILDA).

← 15. https://www.oecd.org/els/soc/SEM271_Online Annexe Table B.xlsx

← 16. Responding to a steep increase in the number of “micro contracts” in France over the past 10 years, a 2021 reform reduced benefit generosity for those alternating repeatedly between short-duration employment and unemployment.

← 17. https://www.oecd.org/els/soc/SEM271_Online Annexe Table B.xlsx

← 18. The EU-SILC records benefit receipt amounts only at the annual basis, it is therefore not possible to link benefit receipt to short periods of joblessness.

← 19. Restricting the sample to individuals with complete observations for the entire three-year panel duration does not lead to a change in the distribution of individuals over the categories race and income. Put differently, restricting the sample in this way does not lead to a disproportionate “loss” of disadvantaged labour market groups (as captured by the dimensions race and income).

← 20. A different specification looking exclusively at unemployment compensation receipt yields very similar results.

← 21. Alternative model specifications for vignettes with household disposable incomes in the middle of the distribution (between the 41st and the 70th percentile), as well as specifying that the vignette be unemployed at the beginning of the spell and not labour-market inactive, yield very similar gaps. Of course, the overall coverage rate is lower for the higher-income vignette (20% for previous standard workers, compared to around 50% for the low-income vignette).

← 22. Note that the vignette fixes previous earnings at or above the 40th percentile of the earnings distribution – part-time workers are expected to have lower earnings within this range.

← 23. There are two variables that identify race and ethnicity in the SIPP: race, that groups individuals into the categories “White alone”, “Black alone”, “Asian alone” and “Residual”, with the residual containing those considering themselves to be mixed race as well as other racial or ethnic minorities; and origin, that identifies “Spanish, Hispanic or Latino” individuals. In the 2016 SIPP, only 5.5% of African Americans and 3.1% of Asian Americans are also of Spanish/Hispanic or Latino origin. They are grouped with African Americans and Asian Americans, respectively.

← 24. Tax foundation (2020): “The Child Tax Credit: Primer”, published online under https://files.taxfoundation.org/20200413132740/Child-Tax-Credit-A-Primer.pdf

Metadata, Legal and Rights

This document, as well as any data and map included herein, are without prejudice to the status of or sovereignty over any territory, to the delimitation of international frontiers and boundaries and to the name of any territory, city or area. Extracts from publications may be subject to additional disclaimers, which are set out in the complete version of the publication, available at the link provided.

© OECD 2023

The use of this work, whether digital or print, is governed by the Terms and Conditions to be found at https://www.oecd.org/termsandconditions.