Modernising Access to Social Protection

2. Non-take‑up and the digital transformation of social programmes in OECD countries

Valerie Frey

Raphaela Hyee

Abstract

Across OECD countries, valuable social benefits and services are not captured by people who are eligible for them. Non-take‑up can have serious financial implications for households and limits the effectiveness of social protection systems. This chapter discusses different methods for measuring non-take‑up of social programmes, highlights findings on non-take‑up from OECD countries, and discusses four key barriers to take‑up The chapter concludes by connecting lessons from randomised control trials (RCTs) on programme take‑up to the ongoing digital transformation of social protection.

Key findings

Even when programmes are designed to provide adequate support to those in need, non-take‑up remains a barrier to effective social protection coverage. Non-take‑up refers to people who do not receive a social benefit or service for which they are otherwise eligible according to statutory rules and conditions.

• Take‑up rates illustrate the share of de jure entitled individuals or households who enrol in a specific programme. Estimates of social programme take‑up often apply microsimulation models (of statutory benefit entitlement) to programme enrolment reported in representative survey data. Researchers have also started estimating non-take‑up for specific programmes based on linked administrative data, e.g. from a revenue agency.

• Non-take‑up is a problem across countries. Belgium, for example, estimates non-take‑up to be between 37% and 51% for working-age social assistance programme, while in France, around 34% of households eligible for the minimum income benefit, Revenu de solidarité active (RSA), do not receive it each quarter. Not every country publishes take‑up rates, but the governments of Belgium and France have made intensive, high-profile efforts to study and address the issue of non-take‑up.

• The four primary barriers to programme take‑up are (i) insufficient, unclear or complex information, (ii) “hassle costs” (cumbersome application procedures), (iii) stigma, and (iv) low expected benefits. In randomised control trials (RCTs), treatment interventions testing simple information and support with programme applications – to reduce hassle costs – usually have positive effects on applications and eventual enrolment.

• These findings on information clarity and simplified procedures align well with some aspects of the ongoing digitalisation of service/benefit applications and delivery. For many users, web-based interfaces are likely quicker and easier to use than traditional, in-office, and paper-based approaches, and in some cases have proven to increase the number of applications.

• The successful behavioural interventions that have been found in RCTs are most effective among people who are already in contact with the government for other reasons, e.g. through tax filings or enrolment in other benefits. The groups most in need – such as those with very low income, informal workers, older people, and people speaking a different first language than official government communication – have proven very difficult to reach with simple behavioural interventions. This suggests that efforts to digitalise access to social protection should be accompanied by live client support, to help reach people who may face challenges with electronic applications, renewals and service delivery.

• The findings in this chapter also point to the utility of linked data, detailed further in Chapters 3 and 4. Linked administrative data and social registries can go a long way towards identifying people enrolled in one programme who are likely to be eligible for another, and can support users’ programme take‑up by pre‑filling applications or even automatically enrolling them.

2.1. The challenge of incomplete take‑up of social protection

In the face of major economic, social and climate‑related megatrends in OECD countries, social protection remains a critical tool for reducing poverty, smoothing consumption, and promoting social mobility (OECD, forthcoming[1]). Yet very few – if any – OECD countries reach everyone in need of social benefits and services. Coverage gaps emerge from a variety of factors, including the stringency of eligibility criteria and the adequacy of public funding to cover all eligible potential beneficiaries (see Chapter 1).

Even when programmes are adequately funded and designed to reach those in need, however, non-take‑up remains a barrier to effective social protection coverage. Non-take‑up refers to people who do not receive a benefit or service for which they are otherwise eligible according to statutory rules and conditions.

Potential beneficiaries may not apply for (or re‑enrol) in programmes because they are not aware of them, or because information about the programme is difficult to understand. The claims process may involve high “hassle costs” (cumbersome application procedures, means-tests, or behavioural requirements), stigma around receiving a public benefit or service, or low expected benefit payments or low service value (Section 2.4). The compounding barriers to coverage are illustrated in Chapter 1 in this report (Figure 1.1).

This chapter begins by discussing the measurement of non-take‑up of social programmes (Section 2.2). It then presents a short review of national studies of social programme non-take‑up in OECD countries (Section 2.3) and discusses the main barriers to take‑up identified in the literature (Section 2.4). The chapter concludes by applying lessons from behavioural studies on non-take‑up to the ongoing digital transformation of social protection. This chapter helps to illustrate the sizeable gaps that remain in supporting vulnerable people in OECD countries and motivates continued efforts to improve access to social protection.

2.2. Measuring take‑up of social programmes

Quantitative indicators of social protection coverage often present programme enrolment as a share of some definition of the population potentially in need.1 The denominator measuring the population in need – for example by poverty rates or household size – is often quite broad, and easy to estimate from available survey or administrative data. Such indicators combine information on de jure accessibility (who is eligible based on rules) and de facto take‑up (who actually enrols) in programmes.

These types of measurements of programme coverage offer broad insights into how well social programmes are reaching a population in need. This can support within-country prioritisation of needs and funding across geographic areas or population groups (Chapter 3), as well as cross-national comparisons and over-time benchmarking of reach.

In the United States, for example, the coverage of the income benefit Temporary Assistance to Needy Families (TANF) is sometimes presented as the “TANF-to-Poverty Ratio.” In 2020, the TANF-to-Poverty Ratio was 21 – meaning that 21 out of every 100 families in poverty received TANF cash assistance nationwide (Shrivastava and Thompson, 2022[2]). In Latvia, an OECD review of affordable housing estimates that only about 23% of households meeting relevant income and household size criteria actually receive housing benefits (OECD, 2020[3])

Estimates focused on take‑up, in contrast, attempt to focus measurement on the share of eligible individuals or households who enrol in a specific programme. While indicators of programme coverage measure recipients as a share of a population potentially in need of a benefit or service, take‑up rates measure recipients of a specific programme as a share of a population ostensibly entitled to the programme or service. For instance, many minimum income benefits (MIBs) in OECD countries provide benefits below the poverty line (OECD, 2023[4]). Estimates of the take‑up rate of such a MIB would therefore zoom in on households with incomes low enough to fulfil the means-test of the benefit, as well as all other possible eligibility criteria, and disregard other households who would still be considered “poor” according to other commonly-used poverty thresholds.

When estimating take‑up for targeted social programmes, determining de jure eligibility is critical. Research on take‑up of social programmes often uses microsimulation models to determine whether a specific household in a representative survey dataset fulfils the statutory requirements of a programme (such as citizenship, age, income or family structure) (Marc et al., 2022[5]). Researchers have also started estimating non-take‑up for specific programmes based on linked administrative data; see, for example, estimates of take‑up of the French minimum pension based on linked tax data (Meinzel, 2022[6]). In Latin American OECD countries like Chile and Costa Rica, social registries are increasingly used to help capture new potential beneficiaries and can be used as a linked data source to measure take‑up.

Few data sources contain all relevant information on benefit eligibility, however, which complicates take-up estimates. For example, data on assets are often incomplete in survey data, and might not be available in administrative or social registry data. Datasets can also rarely be used to identify whether behavioural conditions like job search or school enrolment have been fulfilled, apart from perhaps detailed administrative records on sanctions (e.g. from linked public programmes). These factors can contribute to measurement error in estimates of take‑up.

2.3. Non-take‑up is a problem across OECD countries

Putting aside the empirical challenges in measuring non-take‑up, existing studies of non-take‑up provide useful evidence of the extent of programme participation among people who are likely to be de jure eligible for social programmes. Governments are increasingly investing in producing these estimates. Academic and public research has found high rates of non-take‑up in social programmes across OECD countries.

Belgium in particular is prioritizing research on non-take‑up in social programmes. There are two primary projects on non-take‑up: the BELMOD project, which applies microsimulation to linked administrative data from the labour market and social protection “data warehouse” that includes information on wages and the number of beneficiaries of specific social benefits (CCC, 2023[7]), and the TAKE project. The TAKE project runs a dedicated survey to explore estimates of non-take‑up of public programmes and reasons for it.

TAKE estimates non-take‑up to be between 37% and 51% for the working-age social assistance benefit, between 42% and 59% for the social assistance benefit for the elderly, and 65% and over for the heating allowance (Goedemé , T. et al., 2022[8]).2 (For an elaboration on the BELMOD and TAKE projects vis-à-vis Belgium’s national strategies to identify people living in vulnerable situations, see Chapter 3.)

The government of France has also prioritised the study of benefit non-take‑up. Government researchers have estimated take‑up rates for major social programmes in France and conducted a comparative review of other European countries (Box 2.1).

Around a third (34%) of French households eligible for the minimum income benefit, Revenu de solidarité active (RSA), do not receive it each quarter, and around 20% do not receive it for three consecutive quarters. On average, this amounts to a loss of EUR 330 per month per household. Non-take‑up is highest among households that are not already enrolled in other benefits (such as housing assistance, family benefits or the in-work benefit “prime d’activite”) and among couples without children, young people, homeowners, people living in rural areas or in the Paris metropolitan area. These rates are very similar to estimates from ten years ago, which suggested 36% of potential beneficiaries were not taking up the RSA (Hannafi et al., 2022[9]).

Looking at the minimum retirement pension in France, only 50% of the 646 800 single people aged 65 and older3 actually receive the old-age minimum pension (Meinzel, 2022[6]). Non-recipients would benefit from EUR 205 per month on average. The research finds a significant gender gap in non-take‑up, with a non-take‑up rate of 52% for elderly women and 44% for elderly men. This study used a novel methodological approach, matching microdata from the inter-regime sample of retirees (EIR) of the Direction de la Recherche, des Études, de l’Evaluation et des Statistiques (DREES) with microdata from tax declarations (Meinzel, 2022[6]).

In Germany, the working-age minimum income benefit (Arbeitslosengeld II, or Unemployment Benefit II, in contrast to the contribution-based unemployment insurance scheme, Unemployment Benefit I) was estimated to have a non-take‑up rate of 56% in 2014 (Harnisch, 2019[10]). In a more recent analysis, using linked survey and administrative take‑up, Bruckmeier, Riphahn and Wiemers (2020[11]) estimate the benefit to have a non-take‑up rate of 35 – 37%.

In the United Kingdom, the nearly-universal Child Benefit had a take‑up rate of 93% in 2016‑17, a small but significant decrease from previous years. The take‑up rate for the Child Tax Credit was estimated at 83%, and the Working Tax Credit caseload take‑up rate was estimated at 65% (HM Revenue and Customs, 2018[12]). The Working Tax Credits have since been incorporated into the Universal Credit, which is still being rolled out and for which take‑up rates are not yet estimated.

In the United States, the Earned Income Tax Credit (EITC) is the largest poverty alleviation programme for families with children, providing on average nearly 2 500 USD (2018) per family annually through the income tax system. The IRS estimates that only about 80% of eligible families actually received the EITC from 2011‑17, with rates lower among low-income households (Linos et al., 2022[13]). Take‑up of the US Supplemental Nutrition Assistance Program (SNAP) is similar: an estimated 82% of all eligible individuals participated in 2019, with lower rates among the elderly (U.S. Department of Agriculture, 2023[14]).

Box 2.1. National approaches to estimating non-take‑up

Researchers at the Direction de la Recherche, des Études, de l’Evaluation et des Statistiques (DREES) in France conducted a review of non-take‑up of minimum income benefits in the United Kingdom and selected European Union countries (Marc et al., 2022[5]). One of the interesting contributions of this work is an overview of different “data production models” used by different countries to estimate non-take-up. Most countries rely on linked administrative data or survey data plus microsimulation to estimate take‑up rates, but who performs this analysis varies across countries. Prioritisation of non-take‑up studies – and who conducts them – is relevant for national strategies to identify people living in vulnerable situations (Chapter 3).

Centralised “official” estimates in the United Kingdom, and “academic” estimates in Germany

In the United Kingdom, most studies of non-take‑up are carried out by the Department of Work and Pensions (DWP) and His Majesty’s Revenue and Customs (HMRC), the agency responsible for the collection of taxes. With these estimates produced by ministerial statistical offices, the United Kingdom is unusual for having official figures on non-take‑up.

In Germany, estimates of non-take‑up are calculated principally by two research institutes, the Institut für Arbeitsmarkt- und Berufsforschung (IAB, the research centre of the Federal Employment Agency) and the Deutsches Institut für Wirtschaftsforschung (DIW, funded publicly). Estimates often combine IAB’s microsimulation model (which estimates benefit eligibility) with the German Socio-economic Panel (SOEP) managed by DIW. The resulting estimates of non-take‑up are considered more academic, and not “official”.

Decentralised estimates in the Netherlands produced by a private organisation

In the Netherlands, where regional and local authorities are in charge of most social benefits and services, non-take‑up is also measured locally. DREES reports that many municipalities contract analytical work to Kenniscentrum voor Werk en Inkomen en Zorg (KWIZ), a private organisation that applies proprietary software to municipal administrative data to estimate non-take‑up of social programmes.

Increasing focus on non-take‑up in Finland and Belgium

DREES writes that research on non-take‑up in Belgium and Finland is “sporadic and recent, which makes it more difficult to [define their] data production model”. Nevertheless, in both countries, government agencies that collect social benefit data increasingly collaborate with academic researchers, most notably in Belgium as part of the significant BELMOD and TAKE projects.

Source: (Marc et al., 2022[5]); Sécurité sociale Belgium, “Take Project” (https://socialsecurity.belgium.be/fr/sociale-rechten‑toekennen/take-project); BELMOD and TAKE projects elaborated with inputs from Belgium in Chapter 3 in this report.

2.4. Barriers to take‑up among those who are eligible

Four main barriers have been consistently identified in the literature as deterring eligible people from taking up benefits: (i) unclear, complex or insufficient information about the programme; (ii) “hassle costs” (the cost of applying for the benefit or service), (iii) social stigma associated with programme enrolment, and (iv) low expected value of the benefit or service.

Importantly, these barriers can persist after enrolment, restricting clients’ use and renewal of social programmes. Poorer and less-educated people face many barriers to using services for which they may already be inscribed, such as irregular access to the internet, a lack of transportation to visit programme offices, a lack of time to meet conditions for programme maintenance, and weaker communication skills when dealing with providers.

2.4.1. Programme awareness and information complexity

To apply for a social programme, potential claimants have to be aware of its existence, its basic entitlement rules, and how to put in a claim. These can be high barrier for people with complex needs, limited time, and/or limited educational and economic resources. The OECD Risks that Matter (RTM) Survey – a representative survey conducted in 27 OECD countries – finds that 36% of respondents, on average, are uncertain whether they would qualify for benefits, and 33% are not sure how to apply (OECD, 2023[15]).

Among those who are aware, information complexity in enrolment can present challenges. All individuals have a finite amount of cognitive resources when making decisions, and high amounts of information can impair understanding (Datta and Mullainathan, 2014[16]). These cognitive limits can be particularly harmful when potential programme clients must participate in detailed processes and applications in order to receive welfare‑improving benefits or services.

As (Datta and Mullainathan, 2014[16]) write, “without realizing [it], we often design programmes assuming that people have unbounded cognitive capacity. We assume that they can think through complex problems effortlessly and quickly arrive at the correct choice. We often assume unbounded self-control, which leads us to expect people will always […] do what they intend to do. These assumptions are often unstated, implicit, or even unconscious, but they show up” in programme design.

To participate in social programmes, individuals need to pay attention to various rules and processes. This focus exacts a mental cost, and experimental evidence has found that poverty actually impedes cognitive function (Mani et al., 2013[17]). An individual’s preoccupation with budgets and financial decision-making – which poor individuals do daily – consumes mental resources. Decision fatigue, in turn, leaves less energy for other tasks and leads to poorer decision-making.

All individuals who are eligible for programmes face costs in learning about a programme, but these costs are often the highest for those individuals with the greatest need (Finkelstein and Notowidigdo, 2019[18]). In thinking about intersecting disadvantage, this means not only people living in situations of poverty, but those who may be in frail health, speak a different first language from official government communication (often only in one language), or have physical or mental disability – among other potential conditions.

The importance of clear information for social programme take‑up is by now well documented (Heckman and Smith, 2003[19]; Bhargava and Manoli, 2015[20]). Insufficient information about a programme’s benefits, its application process, its interaction with other benefits, and a client’s likely eligibility can discourage programme participation.

For vulnerable individuals, the cost of learning about a programme often falls to an agent, such as a family member, a caregiver, or a social worker. Since an agent bears the costs of learning about (and perhaps applying to) a programme from which another person benefits, the agent may be less incentivised to apply time and energy to the process.

Intricate policy designs can also lead to non-take‑up as recipients can be unsure about their entitlement or miscalculate amounts (Hyee and Immervoll, 2022[21]). Uncertainty about benefit levels can also lead to a fear of back payments, which can be problematic especially for poor and liquidity-constrained individuals. For instance, for the Earned Income Tax Credit (EITC) in the United States, it has been shown that few recipients opt for an early pay-out of their tax credit because they (possibly mistakenly) fear overpayments (Nichols and Rothstein, 2015[22]). Back-payments can also result from insufficient responsiveness of the benefit, e.g. if the benefit is not automatically adjusted to a changing income situation (Eurofund, forthcoming[23]).

Lessons from the Earned Income Tax Credit

A randomised field experiment carried out by Bhargava and Manoli (2015[20]) and the US Internal Revenue Service illustrates the important role of clear information in take‑up of the US Earned Income Tax Credit (EITC). To test whether different information treatments influence EITC take‑up among those who are eligible, researchers randomly assigned different mailings to over 35 000 individuals in California who filed their taxes but failed to claim their EITC. The researchers found that receiving the text-dense, standard reminder letter (control group) encouraged 14% of contacted non-respondents to take up the EITC. However, a simpler layout, with less repetitive information, improved take‑up to 23%. Providing benefit information also significantly improved take‑up, relative to the standard reminder notice, while language intended to lower programme stigma and lower perceptions of time costs had no effect (Bhargava and Manoli, 2015[20]).

To note, (Bhargava and Manoli, 2015[20]) ran these randomised treatments within a sample of people who had already filed taxes. This is presumably a group with at least basic skills in using an online tax-filing system or liaising with an accountant. Simple information treatments also show positive but much smaller effects on take‑up of refundable tax credits in RCTs with a larger sample, including people who had not filed a tax return before (Guyton et al., 2017[24]; Goldin et al., 2022[25]). Linos et al (2022[13]) finds no effect of randomised information treatments on take‑up of the EITC among people who had never filed before, though they did find the information treatments improved people’s engagement with government via click-through rates to government websites.

These RCTs offer evidence that clear information can help improve participation and offer good examples of how to test different reminders of enrolment. Yet they also illustrate the difficulty of reaching people who are not already in “the system,” such people who have not already filed taxes or who are not receiving other benefits or services.

2.4.2. “Hassle costs”

So-called “hassle costs” (Bertrand, Mullainathan and Shafir, 2006[26]) are an important deterrent to programme take‑up (Currie, 2006[27]; Ko and Moffitt, 2022[28]). Barriers to take‑up are necessarily higher in targeted programmes than in universal ones. Because targeted programmes are intended to reach a select group of beneficiaries, service and benefit providers need to confirm that statutory eligibility requirements are fulfilled. This can, however, imply significant time and energy costs as potential beneficiaries attempt to comply with enrolment (and re‑enrolment) procedures and conditions.

The factors complicating enrolment and persistence can include the “hassle” of finding transportation to apply for benefits; the length and complexity of the application form(s); providing the necessary supplemental documents; the operating hours of the benefit office; communicating with benefit providers; and the processes for maintaining eligibility (Ko and Moffitt, 2022[28]; Currie, 2006[27]; Bertrand, Mullainathan and Shafir, 2006[26]). The claims process requires time, money, and energy on the part of applicants. Potential recipients who do not want to or cannot fulfil behavioural requirements, such as job search requirements, may also not apply for a programme or service.

In social programmes where benefits can be applied for and managed online – as is the case in many OECD countries (see Chapter 4) – the hassle of take‑up can include many of the annoyances of modern life. Complex password requirements, changing user interfaces, unreliable internet access, multi-step application and renewal processes, scanning papers into electronic versions, and difficulty reaching a human service provider, among others, may accumulate to depress applications, enrolment and re‑enrolment.

When asked about the ease of applying for public benefits, respondents to the OECD’s 27‑country RTM Survey identify hassle costs as the biggest challenge: 51% of 27 000 respondents say they believe the application process for public benefits would be difficult and lengthy (OECD, 2023[15]).

Lessons from nutritional assistance, health insurance, and housing and income benefits

A key federal welfare programme in the United States, the Supplemental Nutrition Assistance Program (SNAP), has been evaluated for take‑up outcomes among elderly people (a group with low take‑up) in a randomised control trial in the state of Pennsylvania. The sample was drawn from elderly individuals who were not enrolled in SNAP but were enrolled in Medicaid (means-tested public health insurance). Treatment groups were exposed to information about their eligibility, or to eligibility information plus phone‑based application assistance from “Benefits Data Plus,” a non-profit community organisation. The “status quo” control group – receiving no additional information about SNAP – had a take‑up rate of 6% over nine months. 11% of the information-only treatment group, and 18% of the information plus assistance treatment group, enrolled in the same time (Finkelstein and Notowidigdo, 2019[18]). Those who applied in response to the SNAP treatment interventions had higher net income and better health status than the average enrolee in the control group (Finkelstein and Notowidigdo, 2019[18]). These outcomes reinforce the idea that a lack of information and administrative costs hinder programme enrolment, particularly for individuals in greater need.

An RCT in France found that in-person visits – which could be viewed as time‑intensive – actually helped to improve take‑up because the meeting reduced transaction costs in applications. Job seekers who were randomly selected to attend a meeting with a social worker to learn about a range of potential benefits were 31% more likely to take up any new benefit, relative to those in the control group. Treated job seekers were particularly likely to take up housing and income benefits because the social worker could directly help them with their application. In contrast, in a related RCT, job seekers were exposed to a treatment of an online simulator which provided information; this treatment had no significant effect on take‑up (Castell et al., 2022[29]).

Using observational data, (Bitler, Currie and Scholz, 2003[30]) find that the United States’ Special Supplemental Nutrition Program for Women and Children (WIC) also appears to suffer from high transaction costs. WIC provides supplemental food and nutrition education to low-income mothers and children up to age five. States with stricter eligibility rules have lower participation, and requirements of more frequent visits to WIC offices have reduced participation. (Currie, 2006[27]) also attributes a large gap in take‑up rates between children’s low enrolment in the means-tested US State Children’s Health Insurance Program (“SCHIP”, now “CHIP”) and mothers’ higher enrolment in the federal Medicaid programme to the difficulty of the application process. Hospitals have an incentive to get eligible pregnant women signed up for Medicaid, because they are legally required to provide service to women in childbirth even if they cannot pay. Consequently, most U.S. hospitals have set up Medicaid enrolment offices on-site in order to help patients complete applications and obtain the necessary documents, while CHIP counts on parents living in poverty to apply directly for their children.

2.4.3. Social stigma

One longstanding explanation for incomplete programme take‑up is social stigma. Claiming social benefits may be associated with feelings of shame, which can be exacerbated by the labelling of the benefit itself (as a last-resort benefit for those worst off), or the claims procedure or benefit delivery (especially in small communities). Early literature on take‑up suggested stigma was a major cost of participation in means-tested programmes (see, for example, Moffit (1983[31])).

While stigma likely does induce a cost to benefit receipt, reviews of experimental and observational programme evaluations suggest that transaction costs and complex or insufficient information seem to deter participation much more than stigma does (see, for instance, (Bhargava and Manoli, 2015[20]; Castell et al., 2022[29]; Currie, 2006[27]; Ewoudou, Tsimpo and Wodon, 2009[32]; Remler and Glied, 2003[33]). Even universal (presumably less stigmatised) social programmes have sizeable non-participation challenges, suggesting that stigma is not the main deterrent to means-tested programme participation (Currie, 2006[27]).

At the same time, stigma is difficult to study in academic research. Most surveys use proxies for stigma that are difficult to interpret, which may help explain the weak results (Remler and Glied, 2003[33]). RCT treatments that simulate stigma cannot truly replicate real-world feelings of shame or disrespect.

Another potential associated barrier to take‑up is distrust, though this is perhaps even more difficult to measure. (Linos et al., 2022[13]) write “[distrust…] may be a particular challenge for EITC outreach. Outreach messages often include promises of free cash that can be hard to distinguish from scams to which families are frequently exposed.”

2.4.4. Low expected benefits

Finally, potential programme users are unlikely to initiate enrolment procedures if they believe programme benefits (level and expected duration) will not outweigh the costs – the sum of psychological frictions, transaction costs and stigma noted above. Take‑up of social benefits has been consistently shown to increase with benefit amounts and durations (Janssens and Van Mechelen, 2017[34]; Ko and Moffitt, 2022[28]).

For example, benefit levels have been positively associated with unemployment benefit take‑up rates in the United States (Anderson and Meyer, 1997[35]; OECD, 2023[36]), along with other factors. Rozema and Ziebarth (2017[37]) exploit cross-state and time variation in cigarette taxes to examine the importance of benefit levels for non-take up of SNAP. While the price of cigarettes does not affect benefit eligibility, it does affect smoking households’ budget constraints because demand for tobacco is inelastic, at least in the short term. They find that a one dollar (USD) increase in cigarette taxes raises SNAP take‑up by eligible smoking households by 3.2 percentage points while leaving take‑up of non-smoking households unaffected.

Looking at take‑up rates for the working-age social assistance benefit in Germany, Bruckmeier and Wiemers (2011[38]) find that a marginal EUR 100 increase in the benefit level increases take‑up by 5.8 – 7.6%. They also find that households whose income satisfies the means-test over the course of an entire year have an around 10 percentage point higher take‑up rate than those who have a shorter low-income spell, underlining the importance of the duration-adjusted value of benefits. Similarly, Whelan (2009[39]) shows that an increase in calculated benefits of CAN 100 increases the propensity of eligible households to claim social assistance in Canada by between 5.2 and 6.8 percentage points. In France, the take‑up of unemployment benefits is higher for jobseekers with higher benefit entitlements and longer maximum benefit durations: those with long working histories who are dismissed from open-ended contracts (Eurofund, forthcoming[23]).

The fact that households with higher expected benefit entitlements are more likely to take up the benefits is also indicated by the fact that the share of overall expenditure taken up (the share of the sum of all statutory entitlements) is usually higher than claimant take‑up (the share of eligible households that claim the benefit (Ko and Moffitt, 2022[28]; Fuchs et al., 2020[40]).

2.5. Connecting lessons on take‑up to the digital transformation of social protection

The remainder of this report presents a stocktaking of OECD governments’ strategies to identify groups in vulnerable situations in need of social protection (Chapter 3); collect, curate and link data across different sources; and apply technology to facilitate beneficiaries’ applications, renewals and providers’ outreach to beneficiaries (Chapter 4). The findings of this chapter offer some lessons for social protection systems attempting to modernise for the challenges ahead.

Throughout the literature, the people who are least connected to state institutions – such as people living in situations of homelessness, non-citizens, or workers who do not file income taxes – are also the least likely to take up social programmes, even when prompted. This holds if they are reached through communication tools like postcards or weblinks sent through text messages. The information, psychological, “hassle” and other barriers are high. This suggests the need for governments to continue to produce probabilistic, survey-based estimates of groups or regions in need (Chapter 3) to support targeted information campaigns.

At the same time, clear information is probably not a sufficient intervention for individuals with the greatest need. There is often also a need for individualised, personal outreach, for example by community groups who can help with applications (as in (Finkelstein and Notowidigdo, 2019[18]) and (Castell et al., 2022[29])).

Lessons from behavioural research illustrate that programme application and renewal processes must be simplified, and not only by “going online.” Modern communication technology presents hassles even for those well-versed in it, and even higher barriers for those without regular access to mobile phones or computers. The digital transformation of social service/benefit enrolment and delivery must be accompanied by handrails, including live human resources to support people who may face challenges with electronic applications and service delivery (see more in Chapter 5).

The findings in this chapter also illustrate the value of data linked across agencies or ministries to identify beneficiaries in one programme who likely have eligibility in another (explored further in Chapters 3 and 4). People who are already “in the system” are the ones most likely to apply for and eventually take up other programmes for which they are eligible.

Automatic enrolment in social programmes – using personally-identified linked data – is another promising tool for solving the take‑up challenge, at least among people known to the state. While data governance structures and technical capacity are not in place for this yet in many OECD countries, it is an area of emerging interest.

Finally, improving enrolment (and re-enrolment) processes, including by targeted outreach, is unlikely to be successful if benefit amounts and the services provided are not adequate to meet the needs of the target population. Though connected to broader questions of financing, benefit levels and service quality and quantity cannot be discounted when considering how to improve take‑up.

References

[35] Anderson, P. and B. Meyer (1997), “Unemployment Insurance Takeup Rates and the After-Tax Value of Benefits”, The Quarterly Journal of Economics, Vol. 112/3, pp. 913-937, https://doi.org/10.1162/003355397555389.

[27] Auerbach, A., D. Card and J. Quigley (eds.) (2006), The take-up of social benefits, Russell Sage, https://www.russellsage.org/publications/public-policy-and-income-distribution.

[26] Bertrand, M., S. Mullainathan and E. Shafir (2006), “Behavioral Economics and Marketing in Aid of Decision Making among the Poor”, Journal of Public Policy & Marketing, Vol. 25/1, pp. 8-23, https://doi.org/10.1509/jppm.25.1.8.

[20] Bhargava, S. and D. Manoli (2015), “Psychological Frictions and the Incomplete Take-Up of Social Benefits: Evidence from an IRS Field Experiment”, American Economic Review, Vol. 105/11, pp. 3489-3529, https://doi.org/10.1257/aer.20121493.

[30] Bitler, M., J. Currie and J. Scholz (2003), WIC Eligibility and Participation, The Journal of Human Resources.

[11] Bruckmeier, K., R. Riphahn and J. Wiemers (2020), “Misreporting of program take-up in survey data and its consequences for measuring non-take-up: new evidence from linked administrative and survey data”, Empirical Economics, Vol. 61/3, pp. 1567-1616, https://doi.org/10.1007/s00181-020-01921-4.

[38] Bruckmeier, K. and J. Wiemers (2011), “A new targeting: a new take-up?”, Empirical Economics, Vol. 43/2, pp. 565-580, https://doi.org/10.1007/s00181-011-0505-9.

[29] Castell, L. et al. (2022), Take-up of social benefits: Experimental evidence from France.

[7] CCC (2023), Crossroads Bank for Social Security - Datawarehouse Labour Market and Social Protection, https://www.ccc-ggc.brussels/en/observatbru/data-sources/crossroads-bank-social-security-datawarehouse-labour-market-and-social.

[16] Datta, S. and S. Mullainathan (2014), “Behavioral Design: A New Approach to Development Policy”, Review of Income and Wealth, Vol. 60/1, pp. 7-35, https://doi.org/10.1111/roiw.12093.

[23] Eurofund (forthcoming), Coverage Gaps, Eurofund, https://www.eurofound.europa.eu/.

[32] Ewoudou, J., C. Tsimpo and Q. Wodon (2009), Stigma And The Take-Up Of Social Programs, The World Bank, https://doi.org/10.1596/1813-9450-4962.

[18] Finkelstein, A. and M. Notowidigdo (2019), “Take-Up and Targeting: Experimental Evidence from SNAP”, The Quarterly Journal of Economics, Vol. 134/3, pp. 1505-1556, https://doi.org/10.1093/qje/qjz013.

[40] Fuchs, M. et al. (2020), “Falling through the social safety net? Analysing non‐take‐up of minimum income benefit and monetary social assistance in Austria”, Social Policy & Administration, Vol. 54/5, pp. 827-843, https://doi.org/10.1111/spol.12581.

[8] Goedemé , T. et al. (2022), TAKE Reducing poverty through improving take up of social policies. Final Report, Belgian Science Policy Office, https://takeproject.files.wordpress.com/2023/01/221207_brain-be-final-report_take.pdf.

[25] Goldin, J. et al. (2022), “Tax filing and take-up: Experimental evidence on tax preparation outreach and benefit claiming”, Journal of Public Economics, Vol. 206, p. 104550, https://doi.org/10.1016/j.jpubeco.2021.104550.

[24] Guyton, J. et al. (2017), “Reminders and Recidivism: Using Administrative Data to Characterize Nonfilers and Conduct EITC Outreach”, American Economic Review, Vol. 107/5, pp. 471-475, https://doi.org/10.1257/aer.p20171062.

[9] Hannafi, C. et al. (2022), “Mesurer régulièrement le non-recours au RSA et à la prime d’activité : méthode et résultats”, HAL Working Papers, https://hal.science/hal-03618416.

[10] Harnisch, M. (2019), “Non-Take-Up of Means-Tested Social Benefits in Germany”, SSRN Electronic Journal, https://doi.org/10.2139/ssrn.3352378.

[19] Heckman, J. and J. Smith (2003), The Determinants of Participation in a Social Program: Evidence from a Prototypical Job Training Program, National Bureau of Economic Research, Cambridge, MA, https://doi.org/10.3386/w9818.

[12] HM Revenue and Customs (2018), Child Benefit, Child Tax Credit and Working Tax Credit, https://assets.publishing.service.gov.uk/government/uploads/system/uploads/attachment_data/file/851148/Child_Benefit__Child_Tax_Credit_and_Working_Tax_Credit_take-up_rates_2016_to_2017_restated.pdf.

[21] Hyee, R. and H. Immervoll (2022), The new work incentive for Spain’s national Minimum Income Benefit. Policy issues and incentives in the international comparison, https://www.oecd.org/social/benefits-and-wages/Note-on-the-new-work-incentive-Spain.pdf.

[34] Janssens, J. and N. Van Mechelen (2017), “Who is to Blame? An Overview of the Factors Contributing to the Non-Take-Up of Social Rights”, Working Paper No. 1708, Herman Deleeck Centre for Social Policy, University of Antwerp, https://EconPapers.repec.org/RePEc:hdl:wpaper:1708.

[28] Ko, W. and R. Moffitt (2022), “Take-up of social benefits”, IZA Discussion Paper 15351, https://www.nber.org/papers/w30148.

[13] Linos, E. et al. (2022), “Can Nudges Increase Take-Up of the EITC? Evidence from Multiple Field Experiments”, American Economic Journal: Economic Policy, Vol. 14/4, pp. 432-452, https://doi.org/10.1257/pol.20200603.

[17] Mani, A. et al. (2013), “Poverty Impedes Cognitive Function”, Science, Vol. 341/6149, pp. 976-980, https://doi.org/10.1126/science.1238041.

[5] Marc, C. et al. (2022), Non-take-up of minimum social benefits: Quantification in Europe.

[6] Meinzel, P. (2022), Le non-recours au minimum vieillesse des personnes seules (Non-take-up of minimum retirement pensions by single people).

[31] Moffit, R. (1983), “An Economic Model of Welfare Stigma”, The American Economic Review, Vol. 5/73, pp. 1023-1035, https://www.jstor.org/stable/1814669.

[22] Nichols, A. and J. Rothstein (2015), The Earned Income Tax Credit (EITC), National Bureau of Economic Research, Cambridge, MA, https://doi.org/10.3386/w21211.

[4] OECD (2023), Belgium’s Labour Market and Social Policy Response, OECD Publishing, Paris, https://doi.org/10.1787/990b14aa-en.

[36] OECD (2023), Benefit Reforms for Inclusive Societies in the United States: Income Security During Joblessness, OECD Publishing, Paris, https://doi.org/10.1787/32d8f005-en.

[15] OECD (2023), Main Findings from the 2022 OECD Risks that Matter Survey, OECD Publishing, Paris, https://doi.org/10.1787/70aea928-en.

[3] OECD (2020), Policy Actions for Affordable Housing in Latvia, OECD Publishing, Paris, https://read.oecd-ilibrary.org/view/?ref=137_137572-i6cxds8act&title=Policy-Actions-for-Affordable-Housing-in-Latvia.

[1] OECD (forthcoming), Megatrends and the Future of Social Protection, OECD Publishing, Paris.

[33] Remler, D. and S. Glied (2003), What other programs can teach us: Increasing participation in health insurance programs, American Journal of Public Health.

[37] Rozema, K. and N. Ziebarth (2017), “Taxing Consumption and the Take-up of Public Assistance: The Case of Cigarette Taxes and Food Stamps”, The Journal of Law and Economics, Vol. 60/1, pp. 1-27, https://doi.org/10.1086/692072.

[2] Shrivastava, A. and G. Thompson (2022), TANF cash assistance should reach millions more families to lessen hardship.

[14] U.S. Department of Agriculture (2023), Trends in USDA SNAP participation rates: FY 2016-2019.

[39] Whelan, S. (2009), “The take-up of means-tested income support”, Empirical Economics, Vol. 39/3, pp. 847-875, https://doi.org/10.1007/s00181-009-0329-z.

Notes

← 1.  For a discussion of government efforts to identify populations at need, see Chapter 3.

← 2.  The report presents a range which illustrates findings of the more “strict” model (which attempts to include the income of cohabitating ascendants and descendants in the means test) and a more “lenient” model (which does not consider additional income sources).

← 3.  As well as other people meeting other sufficient conditions, e.g. related to disability.