Chapter 5. Unemployment-benefit coverage: Recent trends and their drivers

This chapter discusses the scope of unemployment-benefit systems, documents recent trends in the number of benefit recipients, and presents alternative measures of benefit coverage in comparative perspective. A decomposition analysis for selected countries seeks to identify key driving forces behind observed coverage trends, including labour-market and demographic changes, as well as benefit policy reforms. In most countries, only a minority of jobseekers receive unemployment benefits and while benefit receipt has increased substantially during the early post-crisis period, this has failed to arrest the longer-term trend towards falling benefit coverage documented in earlier studies. Although composition effects account for a significant share of the recent decline of benefit coverage, some of it is a result of policy reforms that have reduced unemployment-benefit generosity either in search of budgetary savings or in an effort to articulate job-search incentives for the unemployed.


The statistical data for Israel are supplied by and under the responsibility of the relevant Israeli authorities. The use of such data by the OECD is without prejudice to the status of the Golan Heights, East Jerusalem and Israeli settlements in the West Bank under the terms of international law.

Key findings

Among the range of income-support measures that countries operate, unemployment benefits have a central role in stabilising the incomes of jobseekers and in facilitating access to associated employment-support programmes. Low or declining benefit coverage among jobseekers has been one of the drivers of the long-term rise in income inequality. In addition, with the expansion of new forms of employment and potential risks of higher job displacement through automation or digitalisation there are growing concerns that demand for out-of-work support will increase but that existing benefit systems may not be able to provide effective support for all those in need.

This chapter presents different measures of the effective reach of unemployment benefits and documents how benefit coverage has evolved since before the start of the financial and economic crisis. It then employs a decomposition analysis to identify key driving forces behind the observed coverage trends for selected countries.

The main findings are:

  • While unemployment benefits potentially affect job-seeking behaviours or participation, it is commonly assumed that jobseekers have ready access to such transfers. However, results in this chapter show that, in most countries, only a minority of jobseekers receive unemployment benefits, fewer than one-in-three on average across countries.

  • The scope of unemployment benefit systems differs widely across OECD countries. Reflecting different national policy objectives or constraints, unemployment benefits are received by different labour-market groups, including unemployed people who are actively looking for work, but also groups who do not report active job search or have some employment.

  • While benefit receipt has increased substantially during the early post-crisis period, this has failed to arrest a longer-term trend of falling benefit coverage documented in earlier studies. Coverage rates are now slightly below pre-crisis levels, on average.

  • A decomposition analysis of benefit coverage trends for selected countries shows that the changing composition of the jobseeker population during the early crisis years was a major driver of increasing coverage levels. In particular, soaring job losses led to a large inflow of unemployed with sufficiently long employment histories to qualify for benefits. Policy initiatives to make benefits accessible to a larger group of jobseekers also extended coverage during this period in some countries.

  • However, some or most of these increases were subsequently reversed during the post-recession years. Factors that widened the coverage gap in recent years include the growth of long-term unemployment, migration inflows, and the rising numbers of jobseekers entering the labour force without previous work experience as labour markets tightened during the recovery.

  • Policy reforms also contributed to widening coverage gaps as a number of governments tightened entitlement conditions or reduced benefit durations. In some countries, measures to tackle high or growing numbers of youth who are not in employment, education or training (NEET), accelerated school-to-work transitions. While this ultimately strengthens labour-market participation, it also increased the number of jobseekers without work experience or benefit entitlements.


Income support for jobseekers is a central pillar of social protection and labour market policies. Medium-term social and economic trends, as well as more recent labour market developments, have reinforced interest in the reach and accessibility of unemployment benefits and of related out-of-work transfers. This chapter presents evidence on recent trends in unemployment-benefit coverage, and illustrates their driving factors for selected countries. It focuses on observed coverage, i.e. support that is actually received by jobseekers. This is different from concepts of implicit coverage, such as the share of workers who have built up rights to unemployment insurance, but who may or may not claim or qualify for benefits upon unemployment.

In a rights-and-responsibilities framework, unemployment benefits have a key role in targeting employment-support and activation measures. Declining benefit coverage can erode the effective reach of job-search assistance, training and other social and employment re-integration measures. Out-of-work benefits also serve as a major instrument for countering growing income inequality. For instance, trends towards increasing inequality between the 1990s and mid-2000s have been linked to declining shares of jobseekers receiving benefits (OECD, 2011[1]). Other types of cash support may be available to those not receiving unemployment benefits. But transfers such as last-resort social assistance, disability or early-retirement benefits are less focussed on re-employment and may facilitate temporary or permanent labour-force withdrawal.

Widespread reductions in unemployment-benefit coverage prior to the global financial and economic crisis were documented in earlier OECD work (Immervoll and Richardson, 2011[2]). Since then, concerns about non-coverage have intensified, as demand for out-of-work support escalated during and after the Great Recession (OECD, 2014[3]). More recently, maintaining effective support has been a focus in the Future of Work debate as less predictable career patterns, new forms of employment and a greater risk of job displacement through automation create challenges for traditional forms of social protection (OECD, 2017[4]; forthcoming[5]). One key question in this context is whether the shortening of job tenures that is observed for some countries and groups (OECD, forthcoming[6]), or may be expected for future years, would further erode the accessibility of income support during out-of-work spells.

Aggregate trends in benefit receipt may signal a need for policy responses in order to maintain coverage at desired levels. However, the particular policy levers that are suitable for maintaining effective support for jobseekers cannot be discerned through inspection of headline beneficiary headcounts alone. For instance, the huge inflow of new jobseekers in the aftermath of the Great Recession, the subsequent rise in long-term unemployment, and ongoing demographic changes due to population ageing, have led to sizeable shifts in the composition of jobseeker populations. In turn, these composition effects typically produce changes in observed coverage, independently of any policy changes. In addition, the post-crisis period has seen a high density of policy reforms, including determined measures to extend or restrict access to benefits at different points (OECD, 2014[3]).

Because of major concurrent trends during the post-crisis period, identifying the drivers of changing benefit accessibility is challenging, but also important. Each of the drivers will generally have different sets of policy implications, and understanding them is a necessary input into policy discussions of how to keep social protection effective and accessible. For instance, governments’ policy responses during the post-crisis period, together with successive waves of large flows into and out of unemployment, may have easily swamped other concurrent trends that are of policy interest, such as difficult access to unemployment protection among a rising number of workers engaged in platform work and other new or emerging forms of employment.

Section 5.1, first presents recent changes in the aggregate number of unemployment-benefit recipients drawing on OECD SOCR – the Social Benefit Recipients Database. It then examines benefit coverage among the unemployed in more detail and shows the evolution of benefit receipt patterns for different labour-market groups. Section 5.2 explores different drivers of the observed trends using an empirical approach for separating the role of composition and policy effects. The approach is illustrated using micro-data for six countries: Australia, Denmark, Poland, Spain, Sweden and the United States.

5.1. Access to unemployment benefits: Recent trends

5.1.1. Number of benefit recipients

In most OECD countries, the number of unemployment-benefit recipients rose steeply after 2008 as job losses mounted and unemployment reached historic highs (Figure 5.1). The strong rise also reflects a large inflow of benefit claimants who were at the beginning of their unemployment spell and had sufficiently long employment histories to be entitled to benefits. However, benefit receipt subsequently declined relatively quickly, while unemployment remained high, long-term unemployment increased and many unemployed exhausted their rights to benefits. For the 2007-14 period as a whole, unemployment rose more strongly than the number of benefit recipients, suggesting a decline in coverage.1 This trend was more marked in the European Union (EU) (Panel B of Figure 5.1) than in the OECD area (Panel A).

Before the crisis, about 2.5% of the working-age population received unemployment benefits on average across OECD countries, rising to 3.5% by 2014 (Figure 5.2). By then recipient numbers varied from less than 1% of working-age individuals in Hungary, Japan, Poland, Slovak Republic, and Turkey, to more than 10% in Ireland and Finland. Changes in recipient totals have been very large over this period and the comparison between two years hides much greater swings in several countries. For instance, benefit receipt in the United States soared by 250% between 2007 and 2009, before a gradual decline between 2010 and 2014 brought totals back to their pre-crisis level. Full country details are available in the SOCR Database.2

5.1.2. Benefit coverage among jobseekers

Metrics of benefit coverage relate recipient numbers to a certain population of interest. Different measures are useful for different purposes and each has specific interpretations and data requirements. The simple ratio of total benefit recipients and unemployed – based on the definition of the International Labour Organization (ILO) – is commonly referred to as “pseudo-coverage rate”. Depending on the intended scope of unemployment benefits and the benefit entitlement rules that are in force, not all unemployed qualify for unemployment benefits, while some individuals who are not unemployed – e.g. because they are not actively looking for work or are working a few hours per week – may receive them. Pseudo-coverage can therefore vary from very low rates to more than 100% (see Box 5.1).

Figure 5.1. Benefit receipt rose quickly at the onset of the crisis, then dropped while unemployment remained high
Unemployment benefit recipients and unemployed, headcounts 2007=100

Note: Unweighted country averages. Data for Greece are missing. There is currently no unemployment benefit in Mexico. Unemployed are individuals who are not working, have actively looked for work in the previous four weeks and are available to start work within the next fortnight.

Source: OECD Social Benefit Recipients Database (SOCR) ( and OECD Labour Force Statistics (


Figure 5.2. Unemployment-benefit receipt rose following the financial and economic crisis
Recipient totals from administrative sources, in % of working-age population (ages 16-64)

Note: Includes unemployment insurance and assistance benefits. Benefit recipients data for Greece are missing and there is currently no unemployment benefit in Mexico. 2007 data for Italy and Sweden are omitted for comparability reasons. In some countries, additional forms of income support may be available to some unemployed (e.g. for participants in certain labour-market programmes).

Source: OECD Social Benefit Recipients Database (SOCR) (


Box 5.1. Pseudo-coverage rates derived from benefit recipient totals: Construction and interpretation

The pseudo-coverage rate is a simple ratio of benefit recipients and unemployed. In Figure 5.3, the numerator is the number of beneficiaries of unemployment insurance and assistance benefits. The denominator is the number of ILO unemployed, referred to as unemployed in the remainder of this box, over 15 years of age. The resulting measures are referred to as “pseudo” coverage because the populations in the numerator and denominator typically do not fully overlap.

Figure 5.3. Pseudo-coverage rates across OECD countries
Recipient totals from administrative sources, in % of ILO unemployed

Note: See Figure 5.2. ILO: International Labour Organization.

Source: OECD Social Benefit Recipients Database (SOCR) ( and OECD Employment and Labour Market Statistics (


On the one hand, significant numbers of people who are not unemployed may be able to claim benefits that are categorised under the unemployment heading in SOCR data provided by countries.

On the other hand, some unemployed do not receive benefits, either because they do not meet entitlement requirements (e.g. minimum contribution periods) or because they do not claim benefits to which they are entitled. Very low pseudo-coverage rates signal – again intended or unintended – exclusion of some groups of unemployed from receipt of income support through unemployment benefits.

The above figure shows that pseudo-coverage in four out of five countries was below 70% in 2014 and below 20% in Turkey, Slovak Republic, Poland and Hungary. Rates above 100% in Belgium, Austria, Ireland, Finland and Germany indicate that significant shares of benefit payments go to individuals other than active jobseekers, which may be intended or unintended. On average, pseudo-coverage rates fell from 59% to 57% between 2007 and 2014. But changes were very different across countries: Significant increases in Austria, Chile, Estonia, Finland, Germany, Israel, Korea, New Zealand and Norway contrast with marked drops in Australia, Denmark, Belgium, Canada, Hungary, Latvia, Luxembourg, the Netherlands, Portugal, Spain, Switzerland, and the United States.

Key reasons for differences in scope between the numerator and denominator in the pseudo-coverage rates include the following:

  1. Benefits awarded to groups who are not unemployed according to the ILO definition. Examples of these situations are:

    • Individuals who are not actively looking for work. Recipients may be registered as jobseekers but may still not report active job search in the Labour Force Surveys if job-search and other eligibility conditions are not very demanding or are not strictly enforced (Immervoll and Knotz, forthcoming[7]).

    • Individuals close to retirement age. In some countries, job search requirements are less strict if the beneficiary is approaching retirement age. For instance, in Belgium, unemployment insurance (UI) duration is unlimited and receipt is common among older unemployed. Some categories of older unemployed are exempted from active job-search obligations.

    • Some recipients may not even be registered as jobseekers. Available recipient headcounts for some countries include sizeable groups of labour-market inactive individuals, e.g. because they are unable to work. For instance, in Germany, about 6 million people lived in households who received unemployment assistance (UA; Arbeitslosengeld II) in December 2016; of those, 1.6 million were not able to work and only 1.7 million unemployed (Bundesagentur für Arbeit, 2017[8]). For many UA recipients, the benefits they receive are, strictly speaking, not an unemployment benefit.

    • Individuals in work. In several countries, it is possible to combine earnings from work with unemployment benefits under certain conditions (e.g. subject to working hours or earnings). For example, in France, about 700 000 recipients of the UI benefit were in work during the second quarter 2015.

  2. Different reporting units: In Germany, UA recipient stocks are reported in number of families rather than individuals. As a result, where two or more unemployed live in the same family, only one payment will be counted in the numerator.

  3. Measurement period: Both recipient stocks and unemployed headcounts are in principle expressed as averages over a 12-month period. If the numerator or the denominator changes significantly during the year, annual pseudo-coverage rates can differ from instantaneous ones.

  4. Double counting of different types of benefits can also push up pseudo-coverage rates in some cases. In a few countries, concurrent receipt of UI and UA is possible (e.g. the United Kingdom and Ireland), resulting in some degree of double counting when summing up recipients of the different benefit programmes.

Pseudo-coverage rates are informative as broad indicators of the scope of unemployment benefits. But they do not show the coverage for specific policy-relevant groups in the labour market, such as the unemployed, and changes over time can be difficult to interpret. A more concise coverage measure can be derived using microdata, such as labour force surveys. While benefit-receipt information is less precise in these surveys than in the administrative SOCR data, it is possible to break down unemployment-benefit recipient totals by employment status.

Although most benefits indeed go to individuals who are unemployed according to the ILO definition (henceforth “unemployed”), other groups also qualify for benefits under certain conditions (Figure 5.4). Among the countries included in Figure 5.4, individuals working and receiving unemployment benefits are relatively numerous in Australia, Austria, Belgium, Denmark, Finland, France, and Latvia. Significant numbers of “working” benefit recipients may simply be due to differences between national definitions of unemployment, which may allow for a few hours of work per week, and the ILO definition, which does not.3 Allowing individuals to combine benefits with some work, perhaps for a limited period of time, may also reflect a policy objective to strengthen work incentives for some jobseeker groups.

In addition, significant shares of benefit recipients do not report active job search (“discouraged” and “other inactive” in Figure 5.4). This group is relatively sizeable in countries with higher benefit caseloads (Australia, Austria, Belgium, Finland, France and Spain). But it also accounts for large proportions of benefit recipients in some countries where benefits are received by small or very small shares of the working-age population (Hungary, Italy, Luxembourg and Slovenia). “Discouraged” jobseekers are those who are available for work, but have temporarily stopped looking, e.g. due to poor job-finding prospects, because participation in active labour market programmes (ALMPs) leaves little time for active job search or formally exempt participants from job-search requirements, or because some groups of benefit recipients (e.g. lone parents or older unemployed) are explicitly or implicitly exempt from some job-search obligations. Depending on national provisions, it can, however, also include some recipients who have already found work but are waiting to start the new job in the medium term.4

Finally, countries may operate exemptions from requirements to be immediately available for work, or their enforcement may be partial. As a result, some individuals who are neither available for work nor actively looking for it (“other inactive” in Figure 5.4) may receive benefits. As part of longer-term activation strategies, some of these recipients may have been intentionally moved onto unemployment benefits from other programmes that do not require availability for work (such as disability or lone-parent benefits).

Figure 5.5 reports coverage rates using individual-level Labour Force Surveys (LFS) data for some of these groups.5 Results are shown both for a narrow definition of unemployed (Panel A) and for a broader definition that comprises both the “unemployed” and “discouraged” (Panel B). Since the number of “discouraged” recipients is substantial in a number of countries, the remainder of this chapter presents results for this extended group of unemployed, and refers to them as “jobseekers”.

Figure 5.4. Different labour-market groups receive unemployment benefits
Recipients in % of the working-age population (ages 16-64), 2016

Note: Some European countries are excluded due to missing information in EU-LFS data. 2015 figures for Australia. LFS data for Sweden do not include a series of benefits that are accessible to jobless individuals who: i) are not in receipt of core unemployment benefits; and who ii) satisfy other conditions such as active participation in employment-support measures. ILO unemployed are out-of-work individuals who want a job, have actively sought work in the previous four weeks, and can start working within the next fortnight. Those who have made arrangements to take up paid employment or self-employment in the near future are also included in the definition of ILO unemployment. Discouraged workers are out-of-work individuals who want a job and are available to start working, but are not actively looking. The Other inactive category refers to out-of-work individuals who are not available to start employment, e.g. because they are students, retired, unable to work, e.g. due to ill-health or care responsibilities, or who prefer not to work for other reasons. ILO: International Labour Organization.

* The breakdown by employment status in the United States is not shown as information on benefit receipt and employment status in the underlying microdata refers to different time periods.

Source: Household, Income and Labour Dynamics in Australia (HILDA) for Australia; European Union Labour Force Survey (EU-LFS) for European countries; and Current Population Survey (CPS) for the United States.


The group of recipients intended by national unemployment-benefit policy may, however, differ significantly from both the broader and the narrow definition of unemployed (see also Figure 5.6 below). It may, for instance, exclude those with short employment histories (including the previously self-employed), those who are judged to have quit their job voluntarily, or those considered to be insufficiently active in their search or preparation for future employment. In addition, benefits may also be limited to an initial period of unemployment, subject to waiting periods before payments start, or limited to jobseekers living in low-income households. The coverage rates presented here reflect these provisions and can serve as metrics for the intended scope of unemployment benefits relative to countries’ population of unemployed.

While the coverage levels in Figure 5.5 are naturally different from pseudo-coverage rates in Box 5.1, changes since 2007 are broadly similar. Across 24 OECD countries, fewer than one-in-three unemployed, and fewer than one-in-four jobseekers, receive unemployment benefits on average. Coverage rates for jobseekers are below 15% in Greece, Italy, Poland, Slovak Republic, Slovenia and the United States. Austria, Belgium and Finland show the highest coverage rates in 2016, ranging between approximately 45% and 60%: In countries with the highest coverage in the OECD, at least four-in-ten jobseekers still report not receiving an unemployment benefit.

Figure 5.5. Only a minority of jobseekers receive unemployment benefits

Note: Some European countries are excluded due to missing information in EU-LFS data. OECD-24 corresponds to the unweighted average of the countries shown. 2015 figures for Australia. LFS data for Sweden do not include a series of benefits that are accessible to jobless individuals who: i) are not in receipt of core unemployment benefits; and who ii) satisfy other conditions such as active participation in employment-support measures. ILO: International Labour Organization.

Source: Household, Income and Labour Dynamics in Australia (HILDA) for Australia; European Union Labour Force Survey (EU-LFS) for European countries; and Current Population Survey (CPS) for the United States.


5.2. Access to unemployment benefits: Driving factors

Coverage trends are affected by a number of policy and non-policy factors interacting with one another (Figure 5.6). Non-policy factors include demographics (e.g. ageing, migration) and labour-market conditions, although each of them may, in turn, shape policies regarding the accessibility of benefits.

Figure 5.6. Drivers of unemployment benefit coverage

5.2.1. Policy levers

The conditionality built into unemployment-benefit programmes, such as employment conditions, means-tests or activation-related behavioural requirements, are the most direct policy lever for making support more or less accessible initially, while limited benefit durations exclude longer-term unemployed from support provisions. Those who qualify for benefits may decide not to take them up if benefit levels are seen as low relative to the cost of claiming, or if other types of transfers are more generous or easier to obtain. In addition, the perceived accessibility and generosity of benefits can affect the job-search and (re-)employment decisions of unemployed individuals.6

Figure 5.7 presents information on three important aspects of benefit policy across OECD countries.7 As for coverage rates, country differences in the policy parameters governing benefit accessibility are very substantial. Claimants in Slovak Republic, Turkey and Lithuania need to be employed for at least one year and a half before qualifying for unemployment benefits, while employment requirements can be less than six months in several other countries, including in Australia and New Zealand, where no previous employment is needed to qualify for means-tested assistance benefits (Panel A). Behavioural eligibility conditions, such as formal requirements to report active job search and be available for taking up employment, also vary greatly. One indicator of overall strictness suggests that requirements are tightest in Portugal, Luxembourg and Estonia, and comparatively lenient in Hungary, Czech Republic and Turkey (Panel B). For those claiming benefits successfully, maximum benefit durations are half a year or less in Hungary, the United States and Czech Republic but unlimited in Belgium and in several countries operating (means-tested) assistance benefits either as the main form of unemployment support or as a follow-up to first-tier insurance benefits (Panel C).

Figure 5.7. Benefit access provisions vary widely across countries

Note: Information for 2014 or as specified. Data are not shown for Mexico as there is no unemployment benefit. Employment requirements: For individuals with full-time open-ended contracts prior to employment loss. Minimum earnings/contributions requirements in the United Kingdom, the United States and Norway and are assumed to be met. Behavioural eligibility criteria: Scores from 1 (least strict) to 5 (most strict). See Langenbucher (2015[9]) and Immervoll and Knotz (forthcoming[7]) for content and scope of the strictness indicator. Benefit durations: For a 40 year-old with a “long” employment record. Unemployment-assistance durations are for individuals who have exhausted unemployment-insurance rights. Unlimited durations are shown as 48 months. Benefit durations in the United States vary by State and unemployment rate. The 20-week benefit duration in the United States refers to Michigan as at 1 July 2015. Unemployed individuals in Chile can draw unemployment insurance pay-outs provided there are sufficient assets in their individual savings account. In Sweden and in some other countries, additional unemployment support can be available for unemployed individuals participating in activation and employment support programmes.

Source: OECD Tax-Benefit Policy Database (; Langenbucher (2015[9]), “How demanding are eligibility criteria for unemployment benefits, quantitative indicators for OECD and EU countries”,; Immervoll and Knotz (forthcoming[7]), “How demanding are activation requirements for jobseekers? New evidence on activity-related eligibility criteria for unemployment and social assistance benefits”.


In addition to unemployment-benefit policy, a number of indirect policy factors also have an impact on unemployment-benefit coverage, e.g. if reforms make it easier or harder to substitute other types of benefits for unemployment support. In combination, the different benefit policy parameters determine the likelihood of receiving benefits for a specific individual with a given set of characteristics and preferences.

5.2.2. Composition effects

Demographics and labour-market conditions, in turn, determine the number of jobseekers with each specific combination of characteristics and preferences. Some groups are significantly more likely to receive benefits than others (Figure 5.8) and changing sizes of different groups of jobseekers alter observed coverage rates through composition effects. For instance those with less stable temporary employment whose contracts have ended are less likely to meet relevant entitlement conditions for benefits that require a minimum duration of past employment (such as contribution-based insurance benefits). A growing share of jobseekers with less stable employment patterns will therefore tend to drive down coverage rates. By contrast, an increase in the number of older jobseekers with long previous job tenure can have the opposite effect and drive up coverage rates. The same applies to workers who were made redundant, and who typically have longer job tenure than the average job seeker (see Chapter 4).

Figure 5.9 illustrates possible magnitudes of composition effects using data for two countries. In the United States (Panel A), the share of jobseekers who were dismissed from their previous job rose sharply from 23% in 2007 to 46% in 2010 before falling back to 30% by 2016. Overall benefit coverage moved in the same direction, consistent with a positive composition effect as jobseekers dismissed in a steep labour-market downturn include large shares with sufficiently long employment histories to qualify for benefits. In Denmark, the share of young jobseekers increased between 2005 and 2008, fell between 2008 and 2010, and rose again between 2010 and 2016. Overall coverage moved in the opposite direction, consistent with a negative composition effect as youth are less likely to receive benefits than other jobseekers.

Figure 5.8. Some groups of jobseekers are more likely to receive benefits than others
Coverage rates by selected group relative to overall coverage, average across 24 OECD countries

Note: “Jobseekers” include both the unemployed and discouraged workers. Country averages are for the same 24 countries as in Figure 5.5 but are not fully balanced across years due to missing data. They exclude the United Kingdom in 2010 and 2011, Denmark in 2015 and Australia in 2016.

Source: Household, Income and Labour Dynamics in Australia (HILDA) for Australia; European Union Labour Force Survey (EU-LFS) for European countries; and Current Population Survey (CPS) for the United States.


Figure 5.9. Positive and negative composition effects: An illustration
Overall coverage rate and relative size of selected jobseeker groups, % of jobseekers

Note: “Jobseekers” include both the unemployed and discouraged workers.

Source: European Union Labour Force Survey (EU-LFS) for Denmark and Current Population Survey (CPS) for the United States.


5.2.3. Disentangling different drivers of benefit coverage

In practice, different composition and policy effects occur in parallel and may interact. The mechanics behind observed change in overall coverage determines whether or not it may be a policy concern, and which policy responses may be suitable. For instance, stable overall coverage can hide a need for policy action if it is the result of offsetting composition effects that go in different directions (e.g. higher coverage due to ageing, and lower coverage due to increasing own-account work or shortening job tenure). Different drivers may also indicate which policy levers would be effective or appropriate for maintaining coverage at desired levels. For instance, a modest drop in coverage following tighter job-search requirements and associated sanctions might be intended or acceptable whereas a similar drop due to a surge in long-term unemployment may motivate a review of the balance between benefit adequacy and activation provisions.

This section aims to shed light on the concomitant forces that drove changes in coverage since the onset of the global financial and economic crises for six countries: Australia, Denmark, Poland, Spain, Sweden and the United States. The countries were chosen based on data availability and quality, notably the match between benefit-receipt information recorded in labour-force surveys, and the recipient totals available from administrative sources as documented in the OECD SOCR Database. The selected countries also represent different benefit-policy regimes, crisis exposures and broader labour-market contexts.

The method is adapted from common statistical decomposition techniques – see Blinder (1973[10]), Oaxaca (1973[11]) and Fairlie (2005[12]). The decomposition separates changes in observed coverage into those that can be attributed to changing characteristics of the unemployed population (“explanatory variables”), and those that are due to “structural” shifts, notably including benefit reforms, but also changes in behaviours and other factors that are not directly observed in micro-data.8 Coverage changes and their respective drivers are assessed relative to a reference year prior to the onset of the financial and economic crisis: 2005 for Sweden and 2006 for the remaining five countries. The earlier reference year for Sweden is intended to facilitate capturing the impact of an important benefit reform that was enacted in 2006-07 (see below). OECD (2018[13]) provides details on the decomposition method, data and the criteria that were used to select explanatory variables.

The share of jobseekers receiving benefits grew in all six countries during the immediate aftermath of the crisis (Figure 5.10). But coverage trends for the core unemployment benefits over a more extended post-crisis period up until 2016 were either decreasing (Sweden, Denmark, and the United States) or stable (Australia, Poland and Spain).9

Changes in the composition of the unemployed population were important drivers of observed coverage trends in most of the countries (Figure 5.11). Composition effects (blue line) explain almost the entire observed trend in Australia, Denmark and Poland. But the part of the trends explained by compositional changes is smaller in Spain, the United States and Sweden, indicating that other factors, such as benefit policy changes, have shaped coverage trends in important ways.10

Figure 5.10. Coverage trends in selected OECD countries
Individuals who receive unemployment benefit as % of jobseekers

Note: The population of jobseekers includes both the unemployed and discouraged workers. Data for Sweden do not include a series of benefits for jobless individuals who: i) are not in receipt of core unemployment benefits; and who ii) satisfy other conditions such as active participation in employment-support measures. Due to data availability, results for the United States refer only to those who have been unemployed for 52 weeks or less. Records with missing values were excluded from the samples. OECD (2018[13]) provides full details,

Source: Household, Income and Labour Dynamics in Australia (HILDA) for Australia; European Union Labour Force Survey (EU-LFS) for European countries; and Current Population Survey (CPS) for the United States.


The composition effect can be examined more closely by isolating the role of different characteristics, shown as stacked bars in the chart (see notes in Figure 5.11 for details). For instance, the net effect of changes in out-of-work durations on measured coverage is shown by the light-blue bars. In Spain and the United States, and to a lesser extent in Denmark and Poland, large inflows to unemployment in the aftermath of the crisis tended to push up coverage rates. Increasing shares of those experiencing long-term unemployment during the later stages of the labour-market crisis had the opposite effect.11 The effect of “reasons for entering unemployment” (light-grey bars in Figure 5.11) is similar to the case of out-of-work durations. Increasing proportions of jobseekers that were recently dismissed or reached the end of a temporary employment contract pushed up measured coverage in Spain and the United States during the recession and in the first years of the recovery.12

Figure 5.11. Observable changes driving coverage in six OECD countries
Contribution of observable characteristics to changes in coverage among jobseekers, relative to the baseline year

Note: The reference year is 2005 in Sweden and 2006 in Australia, Denmark, Poland, Spain and the United States. Composition effects for different groups (e.g. changes in the size of each different age group) are aggregated into broader domains (e.g. “age”) to facilitate visual inspection. The additional dotted line for Sweden is an estimate of the total change in coverage after accounting for two other important activity-related benefit programmes (Activation Grant [AG] and Development Allowance [DA]) for which aggregate data were available. The estimates assume that all recipients are “jobseekers” as per definition in this chapter. Decomposition results for Sweden however are only for the core unemployment benefits reported in the European Union Labour Force Survey. p.p: percentage points. UI: Unemployment insurance.

Source: Household, Income and Labour Dynamics in Australia (HILDA) for Australia; European Union Labour Force Survey (EU-LFS) and administrative data for Sweden; European Union Labour Force Survey (EU-LFS) for other European countries; and Current Population Survey (CPS) for the United States.


The age composition of the jobseeker population was a significant driver of coverage trends in Denmark, Sweden and Spain (grey bars). However, the underlying reasons driving these shifts differ across these countries. Spain saw an increasing proportion of young jobseekers who have become inactive NEET and are therefore no longer counted as jobseekers. By contrast, in Denmark and Sweden the negative age effect is consistent with a series of policy changes that led to sizeable increase in the number of young jobseekers with no or limited previous work experience. In Sweden, the Adult Education Initiative in mid-2003, which had previously taken up to 4% of the jobless out of the labour force while they acquired educational qualifications (OECD, 2004[14]) was terminated from mid-2003.13 In Denmark the 2003 More People in Work reform package, the 2006 Welfare Agreement, and a series of reforms that took place in 2007 and 2008 (e.g. the reforms of the voluntary 10th form and of the public study grants) all included measures to accelerate school-to-work transitions and incentivise youth participation in training and other active programmes that, however, frequently do not provide rights to unemployment benefits (OECD, (2005[15]; 2006[16]; 2008[17]).14

Migration flows can affect coverage as recent immigrants are less likely to have built up the employment records that may be needed to qualify for benefits. Increasing shares of non-native unemployed has reduced coverage especially in Denmark and Sweden since 2013 – see also OECD (2012[18]). In Australia, a larger inflow of recent immigrants into unemployment between 2011 and 2013 is associated with increasing proportions of jobseekers not meeting applicable residence-duration conditions for benefit receipt.

In addition to composition effects, coverage rates were impacted by reforms of unemployment benefit systems that were enacted over recent years. These policy changes have either offset or added to the composition effects. OECD (2018[13]) provides an overview of key characteristics of these unemployment-benefit systems in 2016 and summarises major policy changes since 2005.15

  • In Spain, a number of reforms during the earlier phase of the labour-market downturn have made benefits more accessible to long-term unemployed, but subsequent policy changes have tended to reduce coverage.16 Results in Figure 5.11 show a declining, and ultimately disappearing, “structural” effect, suggesting that these reforms have partly offset each other over the observation period.

  • In the United States, sizeable extensions of unemployment benefit durations explain the growing “structural” effect in 2009 and 2010.17 After 2011, as unemployment began to decline, several states gradually became ineligible for parts of the federal extensions and a number of them further restricted standard benefit durations. The “structural” effect consequently declined and turned negative in 2014 as access to benefits became more restrictive on average than it had been in the 2006 baseline year.

  • In Denmark, composition effects explain almost all of the coverage changes until 2013, but negative structural effects start playing a role from 2014. This is consistent with a substantial shortening of the maximum duration of unemployment insurance benefits that was approved in 2010 but only came fully into effect several years later.18

  • In Sweden, composition effects explain only around one-third of the overall change in coverage levels since 2005 and unemployment-benefit reforms are therefore likely to have had a major role. Important benefit reforms implemented in Sweden during this time period include measures to shorten benefit durations or make membership in the voluntary unemployment insurance funds more costly to workers.19 Results in Figure 5.11 suggest that these policy changes were indeed associated with a significant decline in coverage of the core unemployment benefits.20 The full effects of some of these reforms appear to have materialised only gradually as the negative “structural” component grew significantly for several years. This was followed by a slight narrowing since 2011, consistent with a roll-back of the earlier reforms.

  • Benefit coverage in Poland increased in the aftermath of the crisis but was back to 2006 levels by 2016, with a negative “structural” effect offsetting the positive composition effects from 2010. Maximum benefit durations were shortened in 2008/2009.21 From 2013 onwards, unemployment rates in several regions fell below the threshold that triggered the 12-month benefit extension, contributing to the widening negative “structural” effect after 2013.

5.3. Concluding remarks

This chapter has shown evidence of a declining trend in unemployment benefit coverage since the end of the recession. This decline could be one factor making jobseekers less selective as regards job offers and, therefore, contribute to explaining the increase in lower-quality jobs that has been observed in recent years (see Chapter 1). However, the trend has been far from uniform, both across countries and over time, with sometimes temporary extensions of potential benefit durations increasing coverage in the early recovery years. Although composition effects account for a significant part of the evolution of benefit coverage in many countries, some change can also clearly be ascribed to policy reforms aimed at reducing unemployment benefit generosity either to contain public spending or in order to dampen job-search disincentives for the unemployed. In countries with generous systems and high coverage levels, a reduction in the share of unemployed receiving benefits may reflect temporary changes in the jobseeker population or more effective activation provisions without necessarily involving the weakening of its protective role. But in countries with very low coverage further reductions may cast doubts on the capacity of the unemployment benefit system to effectively contribute to labour market inclusiveness by helping people to weather negative labour-market shocks. Low coverage also fuels apprehensions about new forms of employment and a risk of job displacement through automation. Together, these labour-market developments create additional demands for out-of-work support but also challenges for maintaining effective support for all those in need (OECD, 2017[4]) (OECD, 2018[19]).

Low and declining unemployment benefit coverage rates can also be a concern for other reasons. For example, unemployment benefits provide the principal instrument for linking jobless people to employment services and active labour market programmes to improve their job prospects. In the absence of accessible unemployment benefits, it can be difficult to reach out to those facing multiple barriers to employment, who therefore risk being left behind (Immervoll, 2012[20]). In these cases, achieving good benefit coverage can be essential to make an activation strategy effective and sustainable. For this reason the new OECD Jobs Strategy calls for clear policy action to extend access to unemployment benefit within a rigorously-enforced “mutual obligation” framework, in which governments have the duty to provide jobseekers with benefits and effective services to enable them to find work and, in turn, beneficiaries have to take active steps to find work or improve their employability (OECD, forthcoming[5]).


[35] Arni, P., R. Lalive and J. Van Ours (2013), “How effective are unemployment benefit sanctions? Looking beyond unemployment exit”, Journal of Applied Econometrics, Vol. 28/7, pp. 1153-1178,

[10] Blinder, A. (1973), “Wage Discrimination: Reduced Form and Structural Estimates”, The Journal of Human Resources, Vol. 8/4, p. 436,

[8] Bundesagentur für Arbeit (2017), Grundsicherung für Arbeitsuchende (SGB II) -,

[12] Fairlie, R. (2005), “An extension of the Blinder-Oaxaca decomposition technique to logit and probit models”, Journal of economic and social measurement., Vol. 30/4, pp. 305-316,

[25] Fernández, R. (2017), Faces of Joblessness in Spain: Main Results and Policy Inventory, OECD, Paris, (accessed on 15 February 2018).

[22] Fredriksson, P. and B. Holmlund (2006), “Improving Incentives in Unemployment Insurance: A Review of Recent Research”, Journal of Economic Surveys, Vol. 20/3, pp. 357-386,

[20] Immervoll, H. (2012), “Reforming the Benefit System to 'Make Work Pay': Options and Priorities in a Weak Labour Market”, IZA Policy Papers, No. 50, IZA, Bonn,

[7] Immervoll, H. and C. Knotz (forthcoming), “How demanding are activation requirements for jobseekers? New evidence on activity-related eligibility criteria for unemployment and social assistance benefits”, OECD Social, Employment and Migration Working Papers, OECD, Paris,

[31] Immervoll, H. and L. Richardson (2013), “Redistribution Policy in Europe and the United States: Is the Great Recession a 'Game Changer' for Working-age Families?”, OECD Social, Employment and Migration Working Papers, No. 150, OECD Publishing, Paris,

[2] Immervoll, H. and L. Richardson (2011), “Redistribution Policy and Inequality Reduction in OECD Countries: What Has Changed in Two Decades?”, OECD Social, Employment and Migration Working Papers, No. 122, OECD Publishing, Paris,

[33] Knittler, K. (2017), Die Definition macht die Zahl, Statistik Austria, Vienna, pp. 180-191.

[36] Kolsrud, J. (2018), “Voluntary unemployment insurance as an option for non-standard work: The case of Sweden”, in OECD (ed.), The Future of Social Protection: What works for nonstandard workers?, OECD publishing, Paris.

[9] Langenbucher, K. (2015), “How demanding are eligibility criteria for unemployment benefits, quantitative indicators for OECD and EU countries”, OECD Social, Employment and Migration Working Papers, No. 166, OECD Publishing, Paris,

[29] Le Barbanchon, T. (2012), The effect of the potential duration of unemployment benefits on unemployment exits to work and match quality in france, (accessed on 15 February 2018).

[21] Nekoei, A. and A. Weber (2017), “Does extending unemployment benefits improve job quality?”, American Economic Review,

[11] Oaxaca, R. (1973), “Male-Female Wage Differentials in Urban Labor Markets”, International Economic Review, Vol. 14/3, p. 693,

[13] OECD (2018), “Supplementary material for Chapter 5”, in OECD Employment Outlook 2018, OECD Publishing, Paris,

[19] OECD (2018), The Future of Social Protection: What works for non-standard workers?, OECD, Paris,

[4] OECD (2017), “Basic Income as a policy option: Can it add up?”, Policy Brief on the Future of Work,

[26] OECD (2017), OECD Economic Surveys: Spain 2017, OECD Publishing, Paris,

[27] OECD (2016), Investing in Youth: Sweden, Investing in Youth, OECD Publishing, Paris,

[30] OECD (2015), Back to Work: Sweden: Improving the Re-employment Prospects of Displaced Workers, Back to Work, OECD Publishing, Paris,

[3] OECD (2014), “The crisis and its aftermath: A stress test for societies and for social policies”, in Society at a Glance 2014: OECD Social Indicators, OECD Publishing, Paris,

[18] OECD (2012), OECD Economic Surveys: Sweden 2012, OECD Publishing, Paris,

[1] OECD (2011), Divided We Stand: Why Inequality Keeps Rising, OECD Publishing, Paris,

[17] OECD (2008), OECD Economic Surveys: Denmark 2008, OECD Publishing, Paris,

[16] OECD (2006), OECD Economic Surveys: Denmark 2006, OECD Publishing, Paris,

[15] OECD (2005), OECD Economic Surveys: Denmark 2005, OECD Publishing, Paris,

[14] OECD (2004), OECD Economic Surveys: Sweden 2004, OECD Publishing, Paris,

[6] OECD (forthcoming), “Are jobs becoming less stable?”, Policy Brief on the Future of Work, OECD Publishing, Paris.

[5] OECD (forthcoming), Good Jobs for All in a Changing World of Work: The OECD Jobs Strategy, OECD Publishing, Paris,

[32] Rothstein, J. (2011), “Unemployment Insurance and Job Search in the Great Recession”, Brookings Papers on Economic Activity, Vol. 2011/2, pp. 143-213,

[28] Swedish Social Insurance Agency (2017), Social Insurance in Figures 2017,

[34] Tatsiramos, K. and J. van Ours (2014), “Labor market effects of unemployment insurance design”, Journal of Economic Surveys, Vol. 28/2, pp. 284-311,

[23] van den Berg, G., B. van der Klaauw and J. van Ours (2004), “Punitive Sanctions and the Transition Rate from Welfare to Work”, Journal of Labor Economics, Vol. 22/1, pp. 211-241,

[24] van den Berg, G. and J. Vikström (2014), “Monitoring Job Offer Decisions, Punishments, Exit to Work, and Job Quality”, The Scandinavian Journal of Economics, Vol. 116/2, pp. 284-334,

Supplementary material for Chapter 5

Supplementary material for Chapter 5 is available online only in English at the following DOI:


← 1. 2014 is currently the latest data available in the OECD SOCR Database.

← 2. The Social benefit recipients Database -

← 3. See, e.g. Knittler (2017[33]) for Austria.

← 4. Those waiting to start a job in the short term are classified as ILO unemployed in Labour Force Surveys data.

← 5. OECD (2018[13]) provides details on data sources and limitations.

← 6. For instance, once a sanction is imposed, the cost of unemployment increases for jobseekers, creating additional incentives for finding and accepting work more quickly than they may otherwise have done – ex-post effect of sanctions Fredriksson and Holmlund (2006[22]), Arni, Lalive and van Ours (2013[35]), van den Berg, van der Klaauw and van Ours (2004[23]). However, positive effects of higher exit rates from unemployment, and higher re-employment rates, can also come at the cost of poorer job matches, a lower quality of post-unemployment outcomes with respect to job stability and earnings (Arni, Lalive and Van Ours, 2013[35]), or a higher probability of working in part-time jobs (van den Berg and Vikström, 2014[24]). Evidence of the effect of unemployment support on post-unemployment job quality is mixed (Tatsiramos and van Ours, 2014[34]) with some newer research finding no effect (Le Barbanchon, 2012[29]) while others document a statistically and economically significant positive effect of benefit durations on post-unemployment wages (Nekoei and Weber, 2017[21]). The latter study also reconciles disparate results by carefully accounting for the two countervailing forces of: i) reduced wages due to longer unemployment; and ii) higher wages due to a better bargaining position and more careful job search.

← 7. Full details on unemployment-benefit policy, and changes since the early 2000s, is available from the OECD Tax-Benefit Policy Database (

← 8. Decomposition analysis, in general, does not rely on the identification of structural or “causal” relationships from the data and does not explicitly reveal which policy mechanisms, if any, are driving the observed changes in coverage levels. Results therefore require careful interpretation. For instance, composition effects can themselves be a result of policy changes (e.g. the indirect policy levers depicted in Figure 5.6). For instance, a new active labour market policy for young jobseekers might increase the share of youth who are registered with the Public Employment Service and actively searching for jobs. Since young jobseekers are comparatively less likely to receive unemployment benefits, this would tend to reduce coverage.

← 9. Sweden operates a series of benefits that are available to jobless individuals who: i) are not in receipt of core unemployment benefits; and who ii) satisfy other conditions such as active participation in employment-support measures. These benefits are included in the SOCR data presented in Figure 5.2 and Figure 5.3. But they are unlikely to be reported in EU-LFS survey data and thus not included in the resulting coverage measures and in the decomposition analysis. Some of these programmes existed since the 1990s, including the Activity Grant and the Development Allowance, which both provide a mix of employment and income support measures. After a decline in the 1990s and early 2000s, programme participation increased between 2008 and 2010 and fell again until 2016. In 2016, about 90 000 individuals (about 8% more than in 2005) participated in a given month, accounting for about 19% of the jobseeker population. The Introduction Benefit (not included in this study) became available in December 2010 and is intended for those who have recently arrived in Sweden. There were about 48 700 recipients in 2016 (Swedish Social Insurance Agency, 2017[28]; OECD, 2016[27]).

← 10. Composition effects for different groups (e.g. changes in the size of each different age group) are aggregated into broader domains (e.g. “age”) to facilitate visual inspection. Fuller granularity, as well as group definitions for each country, are available in OECD (2018[13]).

← 11. For instance, the fraction of long-term unemployed (12+ months) among jobseekers in Spain increased from 34% in 2009 to 50% in 2015. Details for each country are in OECD (2018[13]).

← 12. In Spain the increasing number of jobseekers from 2012 onwards who entered the labour force without any recent work experience partially reversed this positive effect on coverage. One likely explanation for this is women starting to look for work in an attempt to offset other household members’ loss of earnings (the so-called “added worker effect”), a phenomenon that was common in Spain in the aftermath of the crisis (OECD, 2017[26]; Fernández, 2017[25]).

← 13. The inflow of young jobseekers without previous work experience can also explain why in Sweden the overall net effect of other observable characteristics such as the “Reason for entering unemployment” is negative. For instance, the increasing number of jobseekers without previous work experience who entered unemployment for reasons “other than dismissals or terminations of temporary contract” increased significantly from 2005 onwards. As this group is less likely to qualify for unemployment insurance the overall net effect on coverage is negative.

← 14. These reforms can also explain why in Denmark changes in the numbers of jobseekers reporting “active job search” (grid-pattern bars) is among the drivers of coverage trends. Between 2006 and 2008, there was a rising share of jobseekers who had not actively sought employment due to participation in training and other active labour market programmes. Such “lock-in” effects can, in part, be related to the 2006 reform that strengthened active programmes for people under 29 years old (OECD, 2008[17]).

← 15. No major benefit reform occurred in Australia in the period of observation. As a result fluctuations in coverage are fully explained by composition effects (see Figure 5.11).

← 16. A new temporary unemployment benefit was introduced in 2009 (Programa Temporal de Protección por Desempleo e Inserción, PRODI), extending benefit durations by six months for those who had exhausted entitlements to contributory unemployment benefits and were not eligible for other support. In 2011, PRODI was replaced by a programme providing up to six months of benefit support to jobseekers undertaking professional qualification (Programa de recualificación personal de las personas que agoten su protección por desempleo, PREPARA). In 2012, the so-called “pre-retirement” age (the age at which it is possible to receive unemployment assistance until retirement) was increased from 52 to 55 years.

← 17. In addition to the 26 weeks standard benefit duration prior to the crisis, the Extended Benefits programme provides up to 20 weeks of additional entitlement during periods when a state experiences high unemployment. In addition, the federal Emergency Unemployment Compensation (EUC) enacted in 2008 extended benefit durations by 13 weeks, increasing to up to 53 weeks of federally financed additional benefits (a useful summary by state is provided by Rothstein (2011[32])). Last-resort benefits (SNAP, formerly known as the “Food Stamp” programme) also became significantly more accessible from 2007 (Immervoll and Richardson, 2013[31]).

← 18. Denmark approved a reform that reduced the maximum duration of unemployment insurance from four to two years, with a clause that durations could be extended temporarily for six months during periods of economic downturns. This exception was applied twice, in 2011 and 2012. The impact of the shorter benefit duration was further dampened by the introduction of a series of other temporary programmes in 2012/13, such as a new education allowance and access to social assistance with more lenient means-testing following expiry of unemployment insurance entitlements.

← 19. In 2007 Sweden abolished the possibility to maintain unemployment insurance benefits beyond the standard duration. In addition, two reforms in 2007 and 2008 made contributions to the voluntary unemployment insurance dependent on the rate of unemployed workers covered by each fund, raising contribution payments by 300% on average and reducing the share of workers who are fund members. The reforms were rolled back in 2014 but fund density has yet to recover to its pre-reform levels (Kolsrud, 2018[36]). In addition, a number of job-search and other activity-related eligibility conditions became more strict, see Immervoll and Knotz (forthcoming[7]) and OECD (2015[30]).

← 20. See Endnote 9 for related benefit programmes extended or brought in during this period.

← 21. Standard benefit duration limits were six months before the reform, with a 12-month special extension for jobseekers living in regions where the unemployment rate was more than 1.25 times the national average. In special cases, the maximum duration could be extended by up to 18 months instead of 12. The reform abolished the possibility for the 18-month extension and increased the unemployment-rate threshold for the 12-month extension.