Jump to ContentJump to Main Navigation
Reconciling Work and Poverty ReductionHow Successful Are European Welfare States?$

Bea Cantillon and Frank Vandenbroucke

Print publication date: 2013

Print ISBN-13: 9780199926589

Published to Oxford Scholarship Online: April 2014

DOI: 10.1093/acprof:oso/9780199926589.001.0001

Show Summary Details
Page of

PRINTED FROM OXFORD SCHOLARSHIP ONLINE (www.oxfordscholarship.com). (c) Copyright Oxford University Press, 2018. All Rights Reserved. Under the terms of the licence agreement, an individual user may print out a PDF of a single chapter of a monograph in OSO for personal use (for details see www.oxfordscholarship.com/page/privacy-policy). Subscriber: null; date: 23 January 2019

The Redistributive Capacity of Services in the European Union

The Redistributive Capacity of Services in the European Union

Chapter:
(p.185) 6 The Redistributive Capacity of Services in the European Union
Source:
Reconciling Work and Poverty Reduction
Author(s):

Gerlinde Verbist

Manos Matsaganis

Publisher:
Oxford University Press
DOI:10.1093/acprof:oso/9780199926589.003.0006

Abstract and Keywords

Welfare states provide social benefits in cash and – increasingly – in kind. This chapter analyses the redistributive capacity of services, more specifically health care, education and childcare. First, the methodological issues associated with the quantification of social service delivery and the construction of a counterfactual are explored. The results show that services reduce inequality more than cash benefits do. To correctly interpret this result, it is however important to disentangle the effect of ‘size’ and the effect of ‘design’. Services for non-elderly individuals are, in all countries, much more important in size than cash transfers for the non-elderly, a fortiori if education is included. Using concentration coefficients it is shown that, in most countries, cash transfers are more ‘pro-poor’ than in-kind benefits. Obviously, the larger the share of either spending on cash benefits or spending on services, the larger the equalizing impact on the (extended) income distribution.

Keywords:   social redistribution, income inequality, in-kind benefits, social spending, extended income, cash benefits

6.1 Introduction

Welfare states provide social benefits in cash and in kind. Cash benefits are income transfers, such as retirement pensions, family and unemployment benefits, and social assistance. Benefits in kind are commodities directly transferred to recipients at zero or below-market prices (Barr, 2012).

In Europe, benefits in kind are usually services, such as health, education, child care and care for the elderly. For example, hospital care in most countries is provided either free of charge or at near-zero prices (at the point of use). User fees are even rarer in the case of primary and secondary education: Enrollment is compulsory up to a certain age, while tuition is provided free of charge to all children attending publicly funded schools, irrespective of family income. Moreover, child care is often heavily subsidized; kindergartens are run by the state (most commonly local governments) or government-supervised private organizations, while user fees, where applicable, are usually income-related (in the sense that higher-income families pay higher fees, while lower-income ones pay less or are fully exempted). Elderly care may also be available on similar terms; besides, several countries have developed long-term care insurance schemes, to cater for the future needs of an aging population.

Benefits in kind in the form of goods (rather than services) are rather uncommon in Europe. Housing is a partial exception: In some countries council flats are allocated at subsidized rents (or free of charge) to eligible families. (p.186) Nevertheless, in many countries rent subsidies and the direct provision of social housing have been phased out in favor of means-tested housing allowances in cash, except for emergency accommodation, which remains available for selected groups in acute need (i.e., the homeless, refugees, victims of family abuse, and so on). Furthermore, even though food parcels may be handed out by charities and soup kitchens may be organized by municipalities, these are sporadic, or are limited to emergencies, or cater to the needs of marginal groups such as the homeless.1

The main sources of internationally comparable data on the size and composition of social expenditure are Eurostat and the Organisation for Economic Co-operation and Development (OECD). In the European Union, although cash benefits constitute the lion’s share of expenditure on social protection (which does not include education expenditures), the relative weight of benefits in kind is still significant: In 2007, approximately 38% of all social expenditure involved benefits in kind, corresponding to 8.3% of the combined GDP of the 21 Member States included in this study2 (Figure 6.1). The significance of services varies considerably between national welfare states. Their relative share was highest (around 50%) in Sweden, United Kingdom, and Denmark, near average (33% to 36%) in Spain, France, and Germany, and lowest (around 30%) in Italy and Poland.

Consistent data going back in time are hard to find, but the data in the Appendix of this book clearly indicate that the relative importance of benefits in kind increased (see also Chapter 1 in this volume). Looking at the European Union of older Member States (the EU15), where a longer statistical series is available, spending on benefits in kind has gone up (from 8.1% of GDP in 1998

                   The Redistributive Capacity of Services in the European Union

Figure 6.1. Public expenditure for in-kind and cash transfers, in percentage of GDP, 2007.

Source: OECD (2011), data from OECD Social Expenditure database (www.oecd.org/els/social/expenditure) and OECD Education database (www.oecd.org/education/database).

Note: countries are ranked in increasing order of total expenditure on all social services. 2005 data on education services for Greece and Luxembourg.

(p.187) to 9.2% in 2008), while expenditure on cash benefits has lost ground (from 17.8% to 16.8% of GDP over the same period). As a result, the share of benefits in kind in all social spending clearly has risen, from 31.4% in 1998 to 35.3% in 2008 in the EU15.

Benefits in kind (like most social benefits) tend not to be means-tested, which may have consequences for their distributive impact. In terms of expenditure, a mere 15.2% of benefits in kind were means-tested in 2008 in the EU27 (slightly more as the 8.8% of cash benefits). The relative significance of means-testing for benefits in kind has remained stable since the late 1990s, as data for the EU15 show (15.7% in 2008 compared to 16.2% in 1998), while they slightly increased for cash benefits (9.1% vs. 8.2% over the same period). The share of means-testing in all benefits in kind varies greatly between countries. In the UK, Ireland, and the Netherlands the proportion of benefits in kind that were means-tested was 21%–22%. At the other extreme, this proportion was much lower (around 3% to 4%) in countries like Belgium, Sweden, and Finland, and less than 2% in Romania, the Czech Republic, and Estonia.

Including education expenditures, which are an important in-kind benefit, tips the balance very slightly in favor of benefits in kind: 13.4% of GDP in 2007 in the 21 EU countries included in this study (vs. 13.2% of GDP for cash benefits, Figure 6.1). Spending on in-kind benefits exceeds that on cash benefits in all Nordic countries, as well as in the United Kingdom, Ireland, and the Netherlands. Sweden and Denmark top the league with over 18% of GDP spent on benefits in kind. Education and health (at 5.1% and 6.1% of GDP respectively on average in the EU) make up the bulk of in-kind benefits. Child care, services to the elderly, and other services account between them for another 3.4% of GDP. Compared with this EU average, the United States have similar spending on services, but their cash spending is much lower (Figure 6.1).

Given the increasing importance of services in social spending, it is a natural question to try to gauge their distributional impact. We distinguish three questions. Firstly, do services reduce poverty and inequality, or is it true that they favor the rich more than the poor, as is sometimes asserted in the literature (Le Grand, 1982)? Secondly, how do in-kind benefits from services compare to cash transfers in terms of redistributive impact? And thirdly, has the shift to services eroded the distributive power of the welfare state over time? This chapter attempts to answer these questions by drawing on the latest estimates available and considering the major methodological issues. Its structure is as follows. After this introduction, Section 6.2 reviews the literature on the rationale for in-kind benefits and their possible redistributive impact. Section 6.3 addresses the main methodological questions, presenting an alternative way of taking account of service-related needs in the equivalence scale. Section 6.4 presents the recent findings of analyses on the distributional impact of services, focusing on the first two questions. Section 6.5 concludes, in which we touch upon the third question.

(p.188) 6.2 Redistribution via In-kind Provision

6.2.1 The Rationale for In-kind Benefits

Services are particularly important in the “social investment state” discourse, emphasizing the role of “social protection as a productive factor.” This discourse stresses the contribution of health, education, long-term care, and, crucially, child care to economic efficiency and higher living standards (associated with a healthier, better-educated workforce), as well as to greater equality and lower poverty.3

The choice of in-kind versus cash provision hinges on several considerations. In economics, the standard argument in favor of cash benefits is personal autonomy or “consumer sovereignty, ” while a common justification for benefits in kind is paternalism and interdependent preferences (Curry & Gahvari, 2008).

Under interdependent preferences, if the rich care for the material condition of the poor, a transfer from the former to the latter will leave both better off. However, it could be that the rich are not so much concerned with the welfare of the poor, or with their level of income, but rather with their consumption—and, specifically, with “good” rather than “bad” consumption on the part of the poor, as defined by the rich. Therefore, according to this view, from the point of view of taxpayers, the provision of income transfers may be inferior to the provision of particular goods and services.4 Paternalism is strongly connected to the idea of merit goods and merit wants. Society may be concerned that certain goods should be available to all, or even that all should be forced to consume certain goods. Therefore, school attendance up to a certain age ought to be compulsory, rather than left to the individual preferences of children or their parents. In-kind provision is also supported by the related notion of “specific egalitarianism.” For instance, Tobin (1970) has argued that even those who do not object to income inequality per se may still want to see that all individuals receive adequate food, medical services, or housing.

Although paternalism and merit goods can go some way toward explaining why governments provide health and education directly, rather than paying recipients cash benefits to enable them to buy as much health and education as they like, market failures (and, in particular, information failures such as moral hazard, adverse selection, and so on) remain a more powerful explanation of in-kind provision (Barr, 2012). In this light, the choice between benefits in cash and benefits in kind to achieve equity objectives is constrained by considerations of efficiency. Specifically, when market allocation is efficient (e.g., in the case of food), equity objectives can be pursued via cash benefits that enable recipients to buy what they need at market prices. In contrast, when markets fail (e.g., in the case of health care) cash benefits cannot provide a solution, and equity objectives must be pursued via benefits in kind, such as publicly funded health care and school education (see also Le Grand et al. 2008).

(p.189) The focus on universal provision of a comprehensive range of services (going far beyond health and education to include family services and active labor market policies) has always been a distinct feature of Nordic welfare states (Esping-Andersen, 1990). As a matter of fact, choosing in-kind over cash benefits, even where these may seem interchangeable, has important implications for a number of issues. For instance, it can be argued that paying care allowances in cash (as in much of Continental Europe) favors private provision and reinforces traditional patterns of gender roles within the family, while the direct public provision of child and elderly care (as in Scandinavian countries) limits private sector involvement and promotes gender equality and stimulates labor supply (especially of mothers).

6.2.2 Dimensions of Redistribution

Most of the debate on the distributional impact of services concerns vertical redistribution (e.g., between individuals belonging to different income classes). However, it is worth considering that other dimensions may also be relevant. For instance, we may be simply interested in how resources are distributed between individuals with different needs, quite irrespective of their income. If “the key distributional question is [...] whether what people receive matches their needs” (Hills, 2004: 185), then surely the appropriate dimension is horizontal redistribution. Of course, the difficulty here is that “need” for services is often not observed as such, or cannot easily be disentangled from use of services.

Furthermore, services tend by nature to be mostly used by individuals of particular age groups. For example, health care is more heavily used by the elderly and around birth, education is intended for the young, and child care for the very young (and their parents). In view of that, “a snapshot picture of redistribution may be misleading” (Hills, 2004: 185). The relevant dimension here is life cycle redistribution. This reminds us that the welfare state redistributes resources not just between different individuals, but also between different stages in the life of the same individual.

Again, problems of measurement and data availability abound, making the degree of life cycle redistribution hard to estimate. In spite of such difficulties, it has been estimated that, measured on a lifetime basis, in Britain, “nearly three-quarters of what the welfare state was doing in the late 1980s and early 1990s was more like a ‘savings bank, ’ and only a quarter was ‘Robin Hood’ redistribution between different people” (Falkingham & Hills, 1995; Hills, 2004; Barr, 2001).

6.2.3 Are Services Redistributive?

Benefits in kind are generally considered to be less redistributive than benefits in cash. In particular, their contribution toward reducing poverty and inequality has been questioned, for instance by Le Grand (1982), who famously suggested that “Public expenditure on health care, education, housing and transport (p.190) systematically favors the better off and thereby contributes to inequality in final income” (p. 137). OECD (2008; 2011) evidence shows that net cash transfers reduce overall inequality by one third, whereas services reduce inequality only by one fifth.

Part of the difficulty in assessing whether and to what extent this is true lies in the fact that services actually affect the “primary” distribution of incomes (i.e., before taxes and benefits) in a variety of ways, often subtle. For instance, child care and elderly care arguably promote equality through their effect on female employment—both by freeing up women from family responsibilities so as to pursue careers, and by providing women with job opportunities in the social services sector. In this sense, the “equalizing” effect of services goes beyond what a simple “pre-post comparison” would indicate (Esping-Andersen & Myles, 2009). A similar reasoning, however, also applies to cash benefits, especially if they are linked to activation.

Empirical work on the redistributive role of services has proliferated since the pioneering work of Smeeding (1977). In later work, Smeeding et al. (1993) examined the distributional impact of health, education, and housing in seven European countries, while Evandrou et al. (1993) analyzed the role of services in the British welfare state, and their effect on the distribution of incomes. More recently, Harding et al. (2006) compared the redistributive effect of cash and noncash benefits in Australia and the UK. At about the same time Marical et al. (2008) provided estimates of the distributional impact of a range of publicly provided services in OECD countries, an analysis that was updated and extended in terms of country coverage and categories of services in OECD (2011) and Verbist et al. (2012). Furthermore, Matsaganis & Verbist (2009) estimated the distributional effects of subsidies to publicly funded child care in Belgium, Finland, Germany, Greece, and Sweden, while Paulus et al. (2010) estimated the size and incidence of education, health care, and housing subsidies in Belgium, Germany, Greece, Italy, and the UK.

Institutional design may be crucial in determining distributional impact. As the chapter by Van Lancker and Ghysels in this volume demonstrates, the distributional impact of services may be influenced by a number of factors (see also Van Lancker & Ghysels, 2012). The study shows that in terms of equity the Swedish system of child care outperformed the Flemish one. In Flanders, greater use of child care by high-income groups and the generosity of tax deductions offset the pro-poor design of the tariff (parental fees) structure in public child care centers.

The question of redistributive effect is made complex by the fact that services are typically provided in response to greater need associated with the onset of some life event (from child birth to illness and frailty in old age). Ideally, the horizontal and vertical dimensions of redistribution ought to be identified separately. Controlling for need is one way of estimating the distributional impact of services (i.e., the degree of vertical redistribution) net of horizontal effects. (p.191) Recently, a comprehensive analysis of Norwegian local public services (Aaberge et al., 2010) concluded that while noncash benefits reduced poverty by almost one third and inequality by about 15%, adjusting for differences in need offset a significant part of that impact. The different methodological challenges reflect the complexity of the topic and are the subject of the next section.

6.3 Methodological Issues

Estimating the distributional impact of services (or, indeed, cash benefits) raises the issue of the counterfactual. The European Commission routinely publishes estimates of poverty rates in the EU before social benefits (except pensions) and presents their distance from actual poverty rates (i.e., after social benefits) as a measure of national welfare states’ effectiveness in reducing poverty. Although informative, such an exercise explicitly relies on the ceteris paribus hypothesis. But other things are hardly ever equal: If social benefits had not existed, European societies would be completely different. The abolition of maternity, sickness, and disability benefits, for instance, would oblige individuals to work more even when it was better for them that they did not. Simple pre-post comparisons, far from enabling us to draw safe conclusions on the capacity of welfare states to reduce poverty and inequality, thus need to be interpreted with caution.

For a range of reasons, the counterfactual problem is far more serious with respect to services. To start with, incomes in kind (such as free health and education) are not included in standard definitions of income, and their value has to be computed separately in estimations of “extended income.” On the other hand, as discussed earlier, services are often provided in response to greater need. For instance, over 90% of all health expenditure for an individual occurs in the last year before death. In view of this, claiming that recipients had high “extended incomes” as a result would miss an important part of the story: controlling for needs becomes necessary. Furthermore, the welfare state not only affects net disposable incomes (i.e., after redistribution through income taxes and social benefits), but also shapes market incomes (e.g., gross earnings before taxes and benefits). For example, service-intensive Nordic welfare states have defamiliarized welfare responsibilities with regard to caring for children and the elderly, as a result of which employment rates are virtually identical for men and women. As a consequence, child poverty in Nordic countries is low even before social benefits are taken into account. Ignoring these indirect effects of publicly provided social services on the distribution of market incomes risks seriously underestimating their real distributional impact.

In this section, we discuss the major methodological issues related to including the value of public social services in a distributional analysis (see also, e.g., Aaberge et al., 2010; Garfinkel et al., 2006; Marical et al., 2008; OECD, 2011; Verbist et al., 2012). How should one value the benefits households derive from (p.192) public social services (valuation)? How should we distribute the aggregate value of these services among individuals (allocation)? How should the equivalence scale be adapted to take account of the needs associated with these services (equivalence scales)? Each of these issues has consequences for the counterfactual against which to measure the distributive effect of these services.

6.3.1 Valuation

The valuation of public services is a particularly difficult issue, given that these services are provided outside market settings, and hence there is no market price valuation. In the literature, the standard practice is to value the benefit deriving from public services at the production cost, which means that its measurement is based on the inputs used to provide these services rather than on the actual outputs produced (see, e.g., Aaberge et al., 2006; Marical et al., 2008; Smeeding et al., 1993). An initial drawback of this approach is that it does not take account of the quality and efficiency in the provision of these services. Both total and public health care spending is, for instance, much higher in the United States than in any EU country. This corresponds to very high-quality care in some areas (e.g., cancer care), but not necessarily in others, such as primary care, in which many other countries (e.g., United Kingdom) perform better (Pearson, 2009). Moreover, U.S. standard health indicators in general are not always better than in many European countries (Anderson et al., 2003; Garfinkel et al., 2006). Within the national accounts framework, attempts have been undertaken to develop output-based measures, which precisely try to capture (changes in) quality. Deveci et al. (2008) find, for instance, that output-based production of health services grew more rapidly than input-based production. Another problem with using the production cost is that it does not necessarily reflect the user’s value of the service, given that the public service cannot (easily) be exchanged for other goods. Therefore, economists often assume that in-kind benefits are worth less to recipients than their equivalent in cash (Smeeding, 1977; Nolan & Russell, 2001; Garfinkel et al., 2006; Barr, 2012).

6.3.2 Allocation

The second question relates to the allocation of these benefits across individuals: Who are the beneficiaries to whom the value of public services is attributed? The literature distinguishes two approaches, namely the “actual consumption approach” and the “insurance value approach” (see, e.g., Marical et al., 2008). The actual consumption approach allocates the value of public services to the individuals who are actually using the service; it can, of course, only be applied if actual beneficiaries can be identified. This approach is typically used in the case of education services (Antoninis & Tsakloglou, 2001; Callan et al., 2008), childcare services (Matsaganis & Verbist, 2009; Vaalavuo, 2011; Chapter 7 in this volume) and social housing (OECD, 2011; Verbist et al., 2012).

(p.193) The actual consumption approach has also been used for public healthcare services, based on detailed data on the effective use of healthcare services by individuals (see, e.g., for the UK Evandrou et al., 1993; Sefton, 2002). Several authors, however, point out that this approach ignores the greater needs that are associated with being ill: It implies that, ceteris paribus, sick people are better off than healthy people because they receive more health care services (see, e.g., Aaberge et al., 2006). Therefore, many studies use an insurance value approach, which means that one imputes the “insurance value” of coverage to each person based on specific characteristics (e.g., age, sex, socioeconomic position). The insurance value is the amount that an insured person would have to pay in each category (e.g., age group) so that the third party provider (i.c. the government) would have just enough revenue to cover all claims for such persons (Smeeding, 1982). It is based on the notion that what the government provides is equivalent to funding an insurance policy where the value of the premium is the same for everybody sharing the same characteristics, such as age (Marical et al., 2008). The insurance value approach also incorporates the value of access to this type of service.5 Both approaches can lead to quite different results. Marical et al. (2008) have applied the insurance value and the actual consumption approach for healthcare services in eight European countries. On average, the reduction in inequality when including healthcare expenditures in the income concept turned out to be considerably lower on the basis of the actual consumption approach than with the insurance-value approach.6

6.3.3 Correction for Needs: Equivalence Scales

As the needs of a household grow with each additional member in a nonproportional way, equivalence scales are commonly used in distribution analyses to take account of such economies of scale (OECD, 2005). In this book (as in many recent publications) the equivalence scale used for adjusting household disposable income is the so-called OECD modified equivalence scale.7 But as some types of noncash income may have needs associated with them that are unmeasured in usual equivalence scales, using a cash income equivalence scale when noncash income components are included in the income concept, may give rise to what Radner (1997) has called the “consistency” problem. Consider two single-person households with each EUR 1, 000 cash income. Person A is ill and receives public health care worth EUR 200, whereas person B is healthy and needs no health care. Consequently, person A could be said to have 20% more needs than B because of differences in health care needs, and his equivalence scale should be 1.2 compared with 1 for B.

Despite recognition of this issue in the literature, most empirical studies still apply the same (cash income) equivalence scale for both cash and extended income. Garfinkel et al. (2006) defend this approach because “on the one hand, in-kind benefits do not exhibit economies of scale, which implies they should be divided by household size rather than the square of the household size. On the (p.194) other hand, in-kind benefits are not shared equally by all family members, which suggests that they should be added to equivalent cash income on an individual basis. (...) Thus our use of the same equivalence scale for both cash and in-kind expenditures is a reasonable middle-of-the-road solution.” However, this reasoning neglects the fact that health care or education-related needs do not only depend on economies of scales as captured by a standard cash income equivalence scale. This issue is tackled in Paulus et al. (2010), whose basic point of departure is that the equivalence scale used to measure inequality of disposable income is conditional on the existence of free public services such as education and health care. They propose a fixed cost approach, “assuming that the needs of the recipients of these services are equal to a specific sum of money. For example, we could assume that the per capita amounts spent by the state for age-specific population groups on public education and public health care depict accurately the corresponding needs of these groups. Then the recalculation of equivalence scales is straightforward.” They propose the following formula, which should be valid for a household to remain at the same welfare level before and after including public services in the income concept:

(1)
y / e = ( y + k ) / e
with y being cash disposable income, e the modified OECD equivalence scale, k the value of public services and e' the new equivalence scale that incorporates the extra needs of the household members for public services. Hence, (y+k)/e' can be considered as the income concept that incorporates both the in-kind benefit from services, as well as the corresponding needs for these services.

This formula can be rewritten as:

(2)
e = ( y + k ) e / y
meaning that for all households (with y different from zero) the new equivalence scale can be derived. Note that this scale is income-dependent, because its value decreases with income level. The value of k differs across countries, and can reflect differences in social priorities. Paulus et al. (2010) calculate this adjusted equivalence scale using EU average values to calculate k (see further), and then calculate how this impacts on inequality measures. They actually do not really calculate the redistributive effect of services, but they basically test the sensitivity of inequality outcomes for differences in relative spending levels on services across countries.

In order to measure the distributive impact of services, an extra step needs to be introduced, which is the track followed in this chapter. We start from formula (1) and decompose it into two steps, thus developing a service-needs-adjusted counterfactual for measuring the redistributive effect of public services. Firstly, it shows the effect of including the needs for the services (e.g., health care) in the (p.195) equivalence scale by moving from y/e to y/e'[. As e' is an equivalence scale that incorporates a measure for health care needs, y/e' can be considered as an indicator of what the living standard would be if there were no publicly provided healthcare services. Because cash equivalent income is conditional upon the existence of publicly provided services, y/e' is a way of removing this conditionality. Consequently, y/e' can be used as a counterfactual against which to measure the redistributive impact of services, which is represented by the transition from y/e' to (y+k)/e'.

The demarcation of needs (of the target groups; see Aaberge et al., 2010) requires careful consideration, and might differ according to type of service. In the case of health care and compulsory education, it can be argued that all individuals have a need for health care, and that all pupils of compulsory school age have a need for education (which is one of the reasons it is compulsory; Callan & Keane, 2009). In the case of other services, this issue is more debatable: Does the use of child care correspond to a need (and thus require an adaption of the equivalence scale), or is it more a reflection of preferences (thus not requiring a modification of the equivalence scale)? We come back to this issue in section 6.3.4.

The level of needs is calculated as the average spending per individual in target group i for the respective services per individual in this target group (i.c. age group).8 The value of k is calculated for each age group using a correction in spending levels (as a share of spending per age group per service in GDP per capita) toward the EU-level based on the formula presented in Paulus et al. (2010):

(3)
k = i = 1 n ( k E N i * S E E U i S E N i + k H N i * S H E U i S H N i + k E C N i * S C E U i S C N i )
with k ENi, k HNi, k ECN being the country’s spending for, respectively, public education, health care, and early childhood education and child care (ECEC) for persons with characteristics i; S ENi, S HNi, S CNi, being the country’s spending figures for the different types of services expressed as a share of national GDP per capita and S EEUi S HEUi S CEUi being the corresponding EU averages. The new equivalence scale e' of formula (2) is then recalculated for all households using the new value of k, which reflects EU averages of spending.

6.3.4 Data and Implementation

In our empirical analysis we build on the work presented in OECD (2011) and Verbist et al. (2012), focusing on three major categories of services, namely health care, education (with separate results for compulsory and tertiary education), and ECEC. The underlying database is EU-SILC 2007. We present results for 21 EU countries.9

For allocating public education expenditures over the population, we use the actual consumption approach. EU-SILC provides current participation in education for individuals 16 years and older, distinguishing six ISCED (International (p.196) Standard Classification of Education) levels (pre-primary, primary, lower secondary, upper secondary, postsecondary, nontertiary, tertiary).10 As this information is not available for individuals younger than 16 years we have imputed education levels for this group using enrollment rates per education level and age reported in the OECD Education Database. This data source also provides us the average amount of public spending on education per year per pupil/student for the different education levels. These amounts are allocated to pupils/students participating in the corresponding education level.

For health care we have applied the insurance value approach11 using the health care age profiles as published by the European Commission (2009) to derive public healthcare spending per age group. Note that these age profiles only consider differences in age and gender. This approach might underestimate the equalizing effect of public healthcare services in countries where elements of the system are targeted toward low-income groups (e.g., in the form of reduced out-of-pocket payments). On the other hand, given that research has indicated that poorer people have in general worse health conditions, and consequently greater needs for health care (see, e.g., Hernandez-Quevedo et al., 2006), the results may overestimate the distributive impact if these needs remain unmet.

Beneficiaries of early childhood education and child care are identified in EU-SILC on the basis of their participation (number of hours) in either pre-school education or day-care centers. The amounts for the imputation come on the one hand from the OECD Education Database for pre-primary education and on the other from various national sources for child care facilities (see Verbist et al., 2012, for an overview). The imputations are based on number of hours of reported use, thus incorporating intensity of use. A limitation is that EU-SILC does not differentiate between the use of private and public child care. By treating all child care as “public, ” our results will overstate the number of recipients of public subsidies and understate the value of such subsidies per user. However, the resulting bias may in practice be rather limited, given that most ECEC is in fact school pre-primary education, which tends to be overwhelmingly publicly funded.12

With respect to the choice of equivalence scale, the cash income equivalence scale is the modified OECD scale, which is the starting point for constructing the services needs adjusted equivalence scale. For this adjustment for needs related to services, we first have to specify the target groups for which we assume that there are corresponding needs (see also Aaberge et al., 2010). For health care, all individuals have needs, the extent of which differs with age and gender (in line with the health care age profiles used for the imputation).13 For education, we assume that all individuals in the age bracket of 6–16 years have education needs (this corresponds for most European countries with compulsory education),14 which is in line with the approach in Paulus et al. (2010). However, given increased participation in higher education and increased demand for a better educated workforce, it can also be argued that education needs extend to a older age group. Therefore, we use the age group 6–22 years as corresponding (p.197) to education needs. For ECEC, a similar reasoning can be applied: Given the importance attached to child care (see also Lisbon targets), one can assume that child care use is increasingly recognized as a need. Therefore, we also include in the equivalence scale needs for ECEC. For each target group (based on age), we calculate a value for k, which is then averaged over the EU.

6.4 The Distributive Impact of Services

In order to show the distributive impact of public services we first present the size and the incidence of these services measured against cash disposable income equivalized with the modified OECD scale. This shows the relationship between the in-kind benefit of services and the indicator of living standards commonly used in distribution analyses. In a next step, we adjust this living standard concept by incorporating the needs for services in the equivalence scale, as well as the value of these services in the income concept. We then use this adjusted measure of living standards to test the inequality and poverty effect of incorporating services both in the income concept and in the equivalence scale. In order to present an indicator of the distributive characteristics of the various policy instruments that is independent of their size, we finally calculate concentration coefficients. Because the social investment strategy is oriented toward the working-age population, we focus here on benefits targeted at non-elderly individuals. This means that healthcare expenditures allocated to individuals age 65+ are not included and that cash transfers do not include pensions.

6.4.1 Size and Incidence of Services

Figure 6.2a presents the value of total in-kind benefits for working-age individuals as a share of disposable income, as well as the distribution over cash income quintiles. With on average 23% of disposable income, these services are important for living standards, and even more important than cash transfers (excluding pensions), which account for 10% of disposable income (see Figure 6.2b). For services, the size ranges from 16% of disposable income in Greece to 30% in Sweden. For cash transfers, Greece reports again the lowest size (3%) and Hungary has the highest score with 15%.

The distributive pattern of services over the cash income quintile distribution is remarkably equal. In most countries, the share of the bottom quintile in total services is around 20%, with slightly more pro-poorness in Poland and Luxembourg (around 25%). The distributive pattern of cash transfers (excluding pensions) is in general somewhat more oriented toward lower incomes than that of services: On average 26% of the total mass of cash transfers goes to the bottom quintile, whereas the top quintile has a share of only 15%. This is most pronounced in the Netherlands (with 35% going to the bottom quintile) and absent in Spain and Italy (15% for Q1).

(p.198)

                   The Redistributive Capacity of Services in the European Union

Figure 6.2a. Size (percentage of dpi, rh axis) and distribution of in-kind benefits (non-elderly only) over dpi quintiles.

Notes: Countries are ranked in decreasing order by share of benefits in disposable income. Dpi = disposable cash income (equivalence scale = modified OECD-scale).

Source: Calculations based on OECD/EU database on the distributional impact of in-kind services.

Let us now look into the size and distribution of four types of services, namely health care, compulsory education, tertiary education, and ECEC.15 Health care (excluding expenditures going to the elderly) is the most important type, with on average a share in disposable income of almost 10% (Figure 6.2c). The size is lowest in Greece (7%) and highest in France (12%). On average, the share going to
                   The Redistributive Capacity of Services in the European Union

Figure 6.2b. Size (percentage of dpi, rh axis) and distribution of cash benefits (excluding pensions) over dpi quintiles.

Source: Calculations based on OECD/EU database on the distributional impact of in-kind services.

Notes: Countries are ranked in decreasing order by share of benefits in disposable income.

(p.199)
                   The Redistributive Capacity of Services in the European Union

Figure 6.2c. Size (percentage of dpi, rh axis) and distribution of healthcare in-kind benefits (non-elderly only) over dpi quintiles.

Source: Calculations based on OECD/EU database on the distributional impact of in-kind services.

Notes: Countries are ranked in decreasing order by share of benefits in disposable income.

the bottom quintile is slightly below 20%, and this is the case in almost all countries (with Poland and Luxembourg as the only exceptions with a share of 20%).

Figure 6.2d gives the distribution over income quintiles of public education expenditures targeted at 6–16-year-old pupils, which corresponds in most countries to compulsory education. After health care this is the second most

                   The Redistributive Capacity of Services in the European Union

Figure 6.2d. Size (percentage of dpi, rh axis) and distribution of compulsory education in-kind benefits over dpi quintiles.

Source: Calculations based on OECD/EU database on the distributional impact of in-kind services.

Notes: Countries are ranked in decreasing order by share of benefits in disposable income.

(p.200) important component of services (on average 7% of disposable income, with low values in Germany and the Slovak Republic [almost 5%] and a high value of 11% in Hungary). Compulsory education is more oriented toward low incomes: On EU average, the bottom quintile receives 24% of expenditures of this education category, whereas it is only 14% for the top quintile. This pattern is strongest in the Czech Republic and Poland (with a Q1 share of almost 30%), and then decreases gradually over the countries toward 20%. In only three countries the bottom quintile share is below 20% (in Finland, Denmark, and Germany). The general progressive pattern elsewhere is driven by the fact that children in compulsory education tend to be situated more in the lower parts of the income distribution, which is far less the case in these three countries.

The pattern of tertiary education is quite different from that of compulsory education. With on average 2% of disposable income, its size is much lower (ranging from 1% [UK and Italy] to 3% [Slovak Republic and Slovenia], Figure 6.2e). The EU-average appears to indicate a rather even spread of tertiary education expenditures over the entire income distribution. This, however, hides considerable cross-country variation. In Estonia, Portugal, and Slovenia around 10% of tertiary education expenditures are going to the bottom quintile, and, not surprisingly, the share of the top quintile in these countries is, with 30% to 40%, considerable. This is the most common pattern, namely an underrepresentation of the bottom quintile and an overrepresentation of the top quintile. There are, however, some notable exceptions, namely the Nordic countries and Germany. In these countries expenditures to the bottom quintile account for 27% (Iceland) up to

                   The Redistributive Capacity of Services in the European Union

Figure 6.2e. Size (percentage of dpi, rh axis) and distribution of tertiary education in-kind benefits over dpi quintiles.

Source: Calculations based on OECD/EU database on the distributional impact of in-kind services.

Notes: Countries are ranked in decreasing order by share of benefits in disposable income.

(p.201)
                   The Redistributive Capacity of Services in the European Union

Figure 6.2f. Size (percentage of dpi, rh axis) and distribution of ECEC in-kind benefits over dpi quintiles.

Source: Calculations based on OECD/EU database on the distributional impact of in-kind services.

Notes: Countries are ranked in decreasing order by share of benefits in disposable income.

52% (Denmark) of total tertiary education expenditures. This is largely the effect of the compositional factors, as in these countries, large proportions of students live outside the parental home and are thus classified as a separate household. With their low incomes they are often concentrated in the lowest quintile of the distribution. In a study that focuses on the distributive effect of tertiary education spending, Callan et al. 2008 compare for a selection of European countries the distributive effect of excluding students who live independently from the analysis.16 They find that attributing the public transfers to students living with their parents only, rather than to all students, alters the distribution of higher education expenditures considerably, in the sense that results become less pro-poor.

Figure 6.2f shows the size and distribution of the value of ECEC subsidies over income quintiles. On average it represents around 2% of disposable income, with high levels in Sweden, Hungary, and Denmark and a much lower level in Ireland. In countries like Hungary, Luxembourg, Austria, the Netherlands, the Czech Republic, and Iceland, ECEC expenditures tend to go more to lower incomes than to the top groups: In these countries, the first quintile receives between 23% and 28%. The opposite is the case in the Nordic countries, Greece, and Ireland, where the lowest quintile receives less than 20%.

6.4.2 Incorporating Services in the Income Concept and Equivalence Scale

We now incorporate the needs associated with services in the income concept, in order to construct a counterfactual that can be used to measure the impact of (p.202) services on inequality and poverty. This counterfactual is cash income corrected with an equivalence scale that accounts for needs of health care (all individuals), education (individuals aged 6 years to 22 years) and ECEC (children aged 0 to 5 years), as explained in Section 3.3.17 This income concept can be interpreted as an indicator for the living standard under the assumption that these services would not be publicly provided: It indicates in a hypothetical way how much worse off people would be without these services given their needs for health care, education, and child care. A comparison of columns (1) and (3) in Table 6.1 shows the inequality effect of moving from a cash income equivalence scale to one that includes needs for services. Column (3) gives inequality under the assumption that the needs for services are not met, and is hence our counterfactual against which we measure the effect of services on income inequality. This equivalence scale adjustment leads to a considerable increase in measured inequality of disposable income, indicating that these needs are relatively more concentrated at the bottom of the income distribution: On average the Gini increases from 0.2856 to 0.3461. This increase is rather similar across countries.

One may be surprised that the Gini coefficients in columns (1) and (4) are rather similar, suggesting that the redistributive effect of services is rather limited. This outcome follows, of course, from our framework discussed in Section 3.3: The income concept used in column (1) is an indicator of living standards conditional on the existence of free public services, whereas the one in column (4) basically makes this conditionality explicit by incorporating both the needs for and the value of these services—as expressed in formula (1). As we have used an EU-level corrected value of k for each target group to calculate the services-needs adjusted equivalence scale (see formula [3]‌), the comparison of columns (1) and (4) (as is done in Paulus et al., 2010) is in fact a sensitivity test for differences in relative spending across EU countries.

The difference between columns (3) and (4) in Table 6.1 results in the Reynolds-Smolensky index, which gives the reduction in inequality following from including all services (so also including healthcare expenditures going to the elderly) in the income concept. On average for the 21 EU countries, inequality drops substantially, from 0.3461 to 0.2842, which corresponds to around 18% in the counterfactual scenario. Relative reductions in inequality are strongest in Denmark and Sweden (around 23%) and lowest in Greece (13.4%). In relative terms, these reductions are somewhat smaller than the ones calculated on the basis of the modified OECD scale, which is shown in the first panel of Table 6.1 (RE as percent of (1)). The fact that reductions are stronger in the Nordic countries compared with Southern Europe follows from the fact that the services-needs adjusted equivalence scale is based on the average values of k for the EU. As relative spending levels on services are above average in the Nordics, these countries provide a better coverage of these needs (compared with the EU average) than countries with below average spending levels (such as in Southern Europe). (p.203)

Table 6.1. Effect of all services on inequality (Gini coefficient; RE = redistributive effect), a comparison of the modified OECD and a services-needs adjusted1 equivalence scale.

Equival. Scale Income concept

Cash disposable (1)

Modified OECD scale Extended (all services) (2)

RE (1)–(2)

RE as % of (1)

Cash disposable (3)

Adjusted for services needs Extended (all services) (4)

RE (3)–(4)

RE as % of (3)

AT

0.2615

0.2091

0.0524

20.0%

0.3172

0.2591

0.0581

18.3%

BE

0.2622

0.2014

0.0608

23.2%

0.3189

0.2536

0.0653

20.5%

CZ

0.2524

0.1949

0.0575

22.8%

0.3154

0.2535

0.0619

19.6%

DE

0.2995

0.2413

0.0581

19.4%

0.3508

0.2912

0.0596

17.0%

DK

0.2451

0.1894

0.0557

22.7%

0.2973

0.2299

0.0673

22.7%

EE

0.3344

0.2714

0.0630

18.8%

0.4067

0.3457

0.0610

15.0%

ES

0.3125

0.2481

0.0644

20.6%

0.3718

0.3128

0.0589

15.9%

FI

0.2616

0.2137

0.0479

18.3%

0.3174

0.2598

0.0576

18.1%

FR

0.2638

0.2031

0.0607

23.0%

0.3194

0.2529

0.0665

20.8%

GR

0.3427

0.2840

0.0587

17.1%

0.4025

0.3486

0.0539

13.4%

HU

0.2571

0.1961

0.0610

23.7%

0.3273

0.2586

0.0687

21.0%

IE

0.3121

0.2391

0.0731

23.4%

0.3842

0.3138

0.0704

18.3%

IT

0.3216

0.2591

0.0625

19.4%

0.3774

0.3168

0.0607

16.1%

LU

0.2736

0.2147

0.0589

21.5%

0.3460

0.2819

0.0641

18.5%

NL

0.2731

0.2185

0.0546

20.0%

0.3356

0.2751

0.0605

18.0%

PL

0.3217

0.2597

0.0620

19.3%

0.3873

0.3283

0.0590

15.2%

PT

0.3682

0.2888

0.0794

21.6%

0.4319

0.3597

0.0723

16.7%

SE

0.2342

0.1798

0.0544

23.2%

0.2934

0.2255

0.0679

23.1%

SI

0.2278

0.1868

0.0410

18.0%

0.2727

0.2236

0.0491

18.0%

SK

0.2446

0.1921

0.0525

21.5%

0.3093

0.2553

0.0539

17.4%

UK

0.3283

0.2609

0.0674

20.5%

0.3863

0.3222

0.0641

16.6%

EU-21

0.2856

0.2263

0.0593

20.8%

0.3461

0.2842

0.0619

17.9%

(1) Equivalence scale is constructed on the assumption that all individuals have health care needs; 6–22-year-olds have education needs; and 0–5-year-olds have ECEC needs.

Source: Calculations based on OECD/EU database on the distributional impact of in-kind services.

6.4.3 The Impact on Inequality and Poverty

In Table 6.2 we look at the inequality and poverty impact of the various services going to the non-elderly population. Panel A gives the effect on inequality of the different income components. We start from extended income (including all services and using the services-needs adjusted equivalence scale) as the baseline (thus corresponding to Column (4) in Table 6.1). Baseline income inequality is lowest in Sweden, Slovenia, and Denmark (a Gini of around 0.22), and (p.204)

Table 6.2. Effect of services on inequality (Gini) and poverty, equivalence scale adjusted for needs for services1,2

A: Effect on inequality

B: Effect on poverty rate

Gini

Reynolds-Smolensky (change in Gini when excluding income component)

Poverty rate

Change in poverty rate when excluding income component

Extended

All services

Health care Compulsory

Tertiary

Cash (excl pensions)

Extended

All services

Health care Compulsory

Tertiary

Cash (excl pensions)

income

non-elderly

non-elderly

education

education

ECEC

income

non-elderly

non-elderly

education

education

ECEC

AT

0.2591

0.0479

0.0139

0.0221

0.0014

0.0033

0.0399

0.1175

0.1553

0.0501

0.0608

0.0062

0.0097

0.0959

BE

0.2536

0.0465

0.0140

0.0160

0.0027

0.0033

0.0501

0.1410

0.1545

0.0614

0.0415

0.0110

0.0103

0.0942

CZ

0.2535

0.0467

0.0147

0.0163

0.0001

0.0038

0.0363

0.1001

0.1560

0.0550

0.0397

0.0065

0.0090

0.0776

DE

0.2912

0.0460

0.0158

0.0132

0.0036

0.0037

0.0427

0.1451

0.1354

0.0505

0.0347

0.0111

0.0083

0.0778

DK

0.2299

0.0502

0.0062

0.0207

0.0084

0.0046

0.0619

0.1062

0.1730

0.0380

0.0455

0.0203

0.0120

0.1240

EE

0.3457

0.0437

0.0149

0.0130

-0.0006

0.0023

0.0134

0.2106

0.1340

0.0483

0.0392

0.0060

0.0087

0.0395

ES

0.3128

0.0414

0.0146

0.0141

0.0013

0.0029

0.0127

0.2021

0.1223

0.0530

0.0347

0.0097

0.0120

0.0323

FI

0.2598

0.0420

0.0091

0.0170

0.0028

0.0038

0.0590

0.1352

0.1500

0.0372

0.0527

0.0108

0.0148

0.1194

FR

0.2529

0.0563

0.0171

0.0187

0.0028

0.0045

0.0369

0.1198

0.1808

0.0703

0.0500

0.0079

0.0140

0.0944

GR

0.3486

0.0387

0.0145

0.0140

0.0028

0.0018

0.0113

0.2127

0.1005

0.0443

0.0334

0.0091

0.0056

0.0270

HU

0.2586

0.0611

0.0141

0.0305

0.0002

0.0074

0.0458

0.1246

0.1924

0.0602

0.0771

0.0076

0.0201

0.1113

IE

0.3138

0.0576

0.0178

0.0207

0.0013

0.0003

0.0608

0.1608

0.1897

0.0713

0.0573

0.0125

0.0035

0.1403

IT

0.3168

0.0469

0.0136

0.0163

0.0015

0.0031

0.0088

0.1940

0.1232

0.0457

0.0405

0.0060

0.0106

0.0315

LU

0.2819

0.0559

0.0154

0.0230

0.0065

0.0301

0.1449

0.1569

0.0646

0.0685

0.0144

0.0709

NL

0.2751

0.0509

0.0116

0.0232

0.0024

0.0044

0.0384

0.1108

0.1584

0.0521

0.0648

0.0072

0.0127

0.0822

PL

0.3283

0.0532

0.0157

0.0215

0.0011

0.0032

0.0319

0.1792

0.1400

0.0506

0.0520

0.0077

0.0107

0.0668

PT

0.3597

0.0536

0.0209

0.0212

0.0002

0.0021

0.0205

0.1793

0.1485

0.0559

0.0462

0.0042

0.0058

0.0406

SE

0.2255

0.0517

0.0068

0.0211

0.0037

0.0063

0.0465

0.1028

0.1874

0.0478

0.0505

0.0084

0.0160

0.1164

SI

0.2236

0.0403

0.0119

0.0162

-0.0004

0.0022

0.0383

0.1124

0.1589

0.0439

0.0436

0.0080

0.0075

0.0945

SK

0.2553

0.0407

0.0123

0.0139

0.0023

0.0013

0.0224

0.1201

0.1282

0.0522

0.0348

0.0138

0.0028

0.0571

UK

0.3222

0.0443

0.0161

0.0158

0.0012

0.0013

0.0462

0.1756

0.1155

0.0496

0.0389

0.0044

0.0052

0.0880

EU-21

0.2842

0.0484

0.0138

0.0185

0.0019

0.0034

0.0359

0.1474

0.1505

0.0525

0.0479

0.0089

0.0102

0.0801

(1) Equivalence scale is constructed on the assumption that all individuals have health care needs; 6–22-year-olds have education needs; and 0–5-year-olds have ECEC needs.

(2) Poverty rate is calculated on the basis of 60% of median equivalized extended income.

Source: Calculations based on OECD/EU database on the distributional impact of in-kind services.

(p.205) highest in Estonia, Greece, and Portugal (more than 0.34). We then calculate the Reynolds-Smolensky index for different income components going to the non-elderly population (total services non-elderly; health care non-elderly; compulsory education; tertiary education; ECEC; cash transfer excluding pensions). This index is calculated as the difference between the baseline Gini coefficient and the counterfactual Gini of income without the income component. On average, the Reynolds-Smolensky index for the total of non-elderly services is with 0.0484 much more important than the one for cash benefits (excluding pensions: 0.0359). Exceptions to this pattern are Belgium, Denmark, Finland, Ireland, and the United Kingdom, where the Reynolds-Smolensky index is higher for cash than for in-kind. In Southern Europe and Estonia, on the contrary, the inequality impact of cash benefits is relatively small (below 0.02), and this is combined with a below average inequality effect of services. Turning to the different categories of services, compulsory education has on average and in most countries the strongest effect on inequality, followed by health care. For tertiary education, the effect is much smaller, and in some countries close to zero or even slightly anti-equalizing (Slovenia and Estonia). The effect of ECEC is relatively small, but positive in all countries.

Panel B of Table 6.2 presents the poverty reducing effect, starting from a similar baseline as for our inequality analysis, namely a hypothetical income concept that incorporates both the needs associated with services as well as the value of these services. The baseline is the at-risk-of-poverty rate calculated on the basis of extended income, with the poverty line set at 60% of median equivalent extended income. The poverty reducing effect is presented as a point change from moving from income without the component to extended income. For example, on average the hypothetical at-risk-of-poverty rate without services (non-elderly) would be 29.8%, implying that incorporating these services in the income concept reduces the poverty rate by 15.1 percentage points to 14.7. It shows that in the absence of these services, and given that individuals have needs for these services, poverty would be much higher than it is currently.

It is striking that the poverty reducing effect of services is much larger than the one of cash transfers (excluding pensions), which is 8 percentage points. This result is found in all countries, even those where inequality reduction due to cash transfers was stronger. The poverty reducing effect of services ranges from 10 percentage points (Greece) to almost 20 percentage points (Hungary), whereas for cash transfers it ranges from around 3 percentage points (Southern Europe) to 14 percentage points (Ireland). The main drivers for services are health care and compulsory education (which have a similar poverty reducing effect of around 5 percentage points), while the effect of both tertiary education and ECEC is rather limited (no more than 2 percentage points on average and in all countries).

(p.206) 6.4.4 The Weak and Strong Pro-poorness of Cash and In-kind Benefits

The stronger poverty and inequality effect of services may come as a surprise, as a comparison of Figure 6.1 showed that cash transfers are on average across EU countries more directed toward lower incomes than services. The distributive impact of polices, however, depends both on size and design (see also, Chapter 5 in this volume). Calculating concentration coefficients indicates how income components are distributed, irrespective of their size.18 To calculate these concentration coefficients, we rank individuals according to their extended income. When the concentration coefficient has a value that is lower than the Gini coefficient of extended income (which is given in Column [1]‌ of Table 6.1), then lower incomes benefit relatively more: Individuals receive a higher share of the income component than their share of extended income. Thus, these concentration coefficients provide insight into the pro-poorness of the various income components, independent of their size. We can make a distinction here between weak and strong pro-poorness. Strong pro-poorness corresponds to a negative concentration coefficient, whereas weak pro-poorness is captured by a concentration coefficient between zero and the value of the Gini coefficient of extended income.

The concentration coefficients of the various income components (Table 6.3) show that for the EU on average the structure of cash benefits is more pro-poor than that of in-kind benefits (−0.0344 resp. 0.0834). For almost all countries the concentration coefficient for cash transfers is negative, pointing to strong pro-poorness, and much smaller than that of services. Exceptions to this pattern are Spain and Italy, where the concentration coefficients indicate that cash benefits are only weakly pro-poor and less pro-poor than services.

When looking at the different types of services, we found that compulsory education has the strongest inequality reducing effect, even though its size is smaller than that of non-elderly healthcare expenditures. The negative concentration coefficient on average and in most countries indicates that poorer income groups receive a higher share of these services than their share of extended income and points to strong pro-poorness. This result is driven by the distribution of compulsory-aged individuals, which are in most countries slightly more concentrated in the bottom quintile. In almost all countries, the concentration coefficient for tertiary education is the highest of all income components, indicating that this is the least pro-poor of all. Exceptions are Denmark and Sweden, which have a strong presence of students in the bottom quintile, because of students living separately (cf. supra). In some countries the concentration coefficient for tertiary education is even higher than the Gini of extended income, pointing to a pro-rich distribution. This is not only the case in countries where the inclusion of tertiary education expenditures was slightly anti-equalizing (Estonia and Slovenia) but also in, for example, the Czech Republic, Hungary, and Poland.

(p.207) Health care is rather evenly distributed and is hence only weakly pro-poor. On average the concentration coefficient has a value similar to that of ECEC (both around 0.11), though variation across countries is limited for health care. ECEC services exhibit strong pro-poorness in Austria, the Czech Republic, and Luxembourg, where it is (one of) the most pro-poor policy instruments. In the Nordic countries the pattern is far less pro-poor. A more detailed discussion of this category of services is the subject of Chapter 7 in this volume, which focuses on the distribution of different family care policies across households with children (instead of all households, as is done in this chapter).

Summarizing, these results suggest that the stronger redistributive effect and poverty reduction of in-kind benefits should be attributed mainly to their

Table 6.3. Concentration coefficients of cash benefits and services, equivalence scale adjusted for needs for services.

All services non-elderly

Health care

Compulsory education

Tertiary education

ECEC

Cash (excl pensions)

AT

0.0402

0.0913

−0.0637

0.2696

−0.1422

−0.0769

BE

0.0811

0.1154

0.0148

0.1702

0.0874

−0.0837

CZ

0.0677

0.1226

−0.0973

0.3525

−0.0546

−0.0961

DE

0.0583

0.0944

−0.0363

0.1554

−0.0005

−0.0900

DK

0.1055

0.1563

0.0273

0.0825

0.1598

−0.1261

EE

0.1597

0.1620

0.0898

0.4331

0.1854

0.1382

ES

0.1235

0.1224

−0.0008

0.3264

0.1936

0.1425

FI

0.1164

0.1311

0.0368

0.2028

0.2021

−0.1255

FR

0.0592

0.0996

−0.0241

0.1463

0.0308

−0.0147

GR

0.0970

0.1182

0.0152

0.2279

0.2018

−0.0077

HU

0.0615

0.1233

−0.0445

0.3207

0.0630

−0.0146

IE

0.0771

0.1193

−0.0361

0.3002

0.2318

−0.1172

IT

0.0910

0.1157

0.0316

0.1460

0.2157

0.1862

LU

0.0293

0.1054

−0.0187

−0.0516

−0.0148

NL

0.0460

0.1123

−0.0919

0.2065

0.0143

−0.1579

PL

0.0693

0.0955

0.0006

0.2984

0.1623

−0.0685

PT

0.1162

0.1305

0.0245

0.3877

0.1702

0.0373

SE

0.1088

0.1707

−0.0018

−0.0063

0.2348

−0.0236

SI

0.1044

0.0929

0.0387

0.2883

0.2225

−0.0155

SK

0.0770

0.1390

−0.1237

0.2413

0.1113

−0.0055

UK

0.0614

0.0944

−0.0434

0.1435

0.2278

−0.1891

EU-21

0.0834

0.1196

−0.0144

0.2347

0.1174

−0.0344

Note: Equivalence scale is constructed on the assumption that all individuals have health care needs; 6–22-year-olds have education needs; and 0–5-year-olds have ECEC needs.

Source: Calculations based on OECD/EU database on the distributional impact of in-kind services.

(p.208) size, rather than to the way they are distributed over the population. The design of cash transfers is apparently more oriented toward lower incomes and is the instrument with the strongest pro-poorness.

6.5 Conclusion

Over the last 25 years, expenditure on services, especially health and child care, has increased significantly in many European countries. At the same time, spending on cash transfers other than pensions has declined as a proportion of GDP (Vandenbroucke & Vleminckx, 2011). Given that in-kind benefits are generally considered to be less pro-poor than cash benefits, this trend has been identified as a key reason European welfare states proved unable to reduce relative poverty in spite of favorable conditions in terms of economic and employment growth (Cantillon, 2011).

In this chapter we analyzed empirically the impact of services on inequality and poverty. In view of the conceptual and methodological issues, this task is fraught with difficulties. We have discussed the issues of valuation, allocation, and the use of an equivalence scale adjusted for needs associated with these services. In this chapter we have chosen to build further on a discussion of alternative approaches in Verbist et al. (2012) and the methodology proposed in Paulus et al. (2010) and to construct a hypothetical counterfactual that incorporates service-related needs. We thus compare an estimate of the current distribution of in-kind benefits with a counterfactual that depicts inequality and poverty in a hypothetical situation in which no publicly provided services exist.

In a first instance, we have tried to answer the question whether services targeted at non-elderly individuals (who are the focus of the social investment strategy) are redistributive. When looking at the total of health care, education, and ECEC, the answer is clearly affirmative for the 21 EU countries considered in this chapter. Compulsory education and health care, especially, both reduce inequality and poverty when compared with a hypothetical situation without these publicly provided services. The answer to the question whether these in-kind benefits are less redistributive than cash transfers requires more consideration. When comparing the Reynolds-Smolensky index for cash transfers with that of in-kind benefits, one is inclined to conclude that services reduce inequality more than cash benefits do. However, it is important to disentangle a size and a design effect, as services going to non-elderly individuals are in all countries much more important in size than cash transfers (excluding pensions). Therefore, we have also calculated concentration coefficients of the various instruments, thus allowing us to focus on the distributive structure independent of size. These coefficients tell us that cash transfers turn out to be more pro-poor than in-kind benefits in most countries. In this perspective, cash transfers are on average in the EU the most pro-poor of the policy instruments considered here, closely (p.209) followed by the in-kind benefit from compulsory education. The in-kind benefits derived from tertiary education expenditures is the least pro-poor (and in some countries even pro-rich). These results are to an important extent driven by the distribution of beneficiaries: In general, compulsory education pupils tend to be situated more in the middle and the bottom of the income distribution, whereas tertiary education students are overrepresented in the higher regions of the income distribution.

For answering the question of evolution, that is, whether the relative shift to services has eroded the redistributive power of the welfare state, there is little evidence on comparisons over time. The only exception is OECD (2011), which compares the inequality reduction through services for 2000 and 2007 for 17 OECD countries (including 14 “old” EU Member States), indicating that, on average across countries, inequality reduction of services has remained remarkably stable over this period. However, countries that improved inequality reduction through services were also those countries that experienced an increase in terms of size (expressed as a share of cash disposable income) (and vice versa). Whether these changes went hand in glove with corresponding (or opposite) changes in the redistributive impact of cash transfers requires further investigation.

Acknowledgments

This chapter uses data prepared in a joint co-funded project of the OECD and the European Commission (see also, OECD, 2011; Verbist et al., 2012). The authors are grateful to Bea Cantillon, Michael Förster, John Hills, and Frank Vandenbroucke for their constructive remarks and valuable suggestions.

Notes

(1) Outside Europe, the direct provision of food to the poor as a matter of course (i.e., not only in the case of famine relief and other emergencies) is still quite common in the United States and some Latin American countries. Examples are Programa Apoyo Alimentario (PAL) in Mexico or food stamp programs in the United States.

(2) The selection of countries is driven by data availability, see section 6.3.4.

(3) The “Samaritan’s dilemma, ” proposed by Buchanan (1975), may be thought of as the libertarian case for social investment. The argument goes along the lines of the Chinese proverb that it is better to teach someone how to fish than simply give them a fish. In Buchanan’s formulation, recipients have an incentive to remain poor if they are entitled to benefits as long as they are poor. Hence benefits should primarily be designed to discourage benefit dependency and eliminate moral hazard. The latter may arise when the availability (p.210) of benefits (when poor) undermines the willingness of individuals to invest in human capital (so they avoid poverty). While the argument can be evoked to support cuts in social provision, it can and has been used to support the public provision of in-kind transfers, whether in the form of job training or social insurance (Coates, 1995).

(4) Sen’s approach (1993), redefining well-being in terms of capabilities like being able to read, write, remain healthy, etc., can be seen as a more enlightened form of paternalism (cf. Deneulin, 2002), justifying social investment in public services such as education and health.

(5) A more pragmatic reason for using the insurance value approach is that most data sets used in distributional analysis (e.g., EU-SILC) do not contain information on effective use of healthcare services.

(6) This rather surprising outcome is largely due to the effect of re-ranking. Because part of the expenditures (notably those on in-hospital care) are concentrated among a very small group, this may lead more easily to re-ranking of individual beneficiaries, which dampens the equalizing effects of healthcare services (5% of the population in the survey data accounted for more than 90% of the nights spent in hospital, whereas out-of-hospital care was more widespread over the population; see Marical et al., 2008 for more details).

(7) This scale assigns a value of 1 to the household head, of 0.5 to each additional adult member, and of 0.3 to each child. See OECD (2005) (http://www.oecd.org/dataoecd/61/52/35411111.pdf) for further explanations and specifications. This is a pragmatic equivalence scale, which takes into account only differences in household size.

(8) This means that higher spending levels (either due to high use of the services or high expenditures per user) result in higher corresponding needs (and vice versa). Thus a higher level of spending corresponds to a higher recognition of needs. Remember that, as already indicated in section 6.3.1, differences in efficiency or quality of services cannot be accounted for in this analysis.

(9) Bulgaria, Malta, and Romania are not included in EU-SILC 2007. Cyprus, Lithuania, and Latvia are not included in the analysis because for these three countries we do not have estimates for the value of ECEC. Moreover, their amounts for tertiary education include expenditures on research and development, reducing comparability with other countries.

(10) Note that EU-SILC does not allow us to distinguish between participation in publicly or privately funded educational institutions, nor between general and technical secondary education, nor between Type A and Type B tertiary education. For tertiary education, amounts exclude direct expenditures for research and development activities.

(11) This approach is defendable for EU-countries, where public healthcare coverage is quasiuniversal in most countries.

(p.211) (12) In some of these countries (Denmark, Germany, and Sweden), the distinction between day care and pre-primary education in EU-SILC is probably erroneous: The number of children in the former is probably overestimated, while the number in the latter is underestimated (when compared with enrollment rates in pre-primary education in the OECD Education Database). In most countries where child care is more frequently used, day-care centers tend to be heavily subsidized.

(13) Unfortunately, there is no information available that would allow us to differentiate according to other parameters such as socioeconomic background.

(14) The starting age of compulsory education is six years in most countries, whereas some countries start earlier at age five years (Hungary, the Netherlands, and the United Kingdom), and others at age seven years (Denmark, Estonia, Finland, Poland, and Sweden). (See OECD, Education at a Glance 2010.) The ending age of compulsory education varies between 14 years and 18 years in Europe. Ninety percent of the population are enrolled in education for at least 13 years, ranging from 11 (e.g., Greece) to 15 years (Belgium, France, Norway, and Sweden).

(15) These four categories do not sum up to the total presented in Figure 6.2a, which includes also non-tertiary education outside the age category 6-16.

(16) This was a relevant exercise for five of the seven countries in their study, namely Germany, Greece, Ireland, the Netherlands, and the UK. In Belgium and Italy students in the survey are included together with their household of origin, thus making this kind of correction unnecessary.

(17) The calculations have also been done for other specifications of needs (e.g., only children aged 6 years to 16 years have education needs). Overall, the results are similar to those with the broader definition of needs as used here.

(18) The concentration coefficient of an income component is calculated in a similar way as the Gini coefficient (see e.g., Kakwani, 1977; Lambert, 2002; OECD, 2008). The difference between the two lies in the variable according to which income units are ranked. With a concentration coefficient of an income component, income units are ranked according to extended income (and not by the income component itself), while for a Gini coefficient the variable of interest and the ranking income variable are the same (namely, extended income). Because extended income is used for all income components as the ranking variable, concentration coefficients can be used to compare the distributive structure across income components. They can be considered as a summary indicator of the information provided by quintile distributions. Note that in Figure 6.2, quintiles are constructed on the basis of cash incomes (with modified OECD scale), whereas here units are ranked on the basis of extended income equivalized with the services-needs adjusted equivalence scale.

Notes:

(1) Outside Europe, the direct provision of food to the poor as a matter of course (i.e., not only in the case of famine relief and other emergencies) is still quite common in the United States and some Latin American countries. Examples are Programa Apoyo Alimentario (PAL) in Mexico or food stamp programs in the United States.

(2) The selection of countries is driven by data availability, see section 6.3.4.

(3) The “Samaritan’s dilemma, ” proposed by Buchanan (1975), may be thought of as the libertarian case for social investment. The argument goes along the lines of the Chinese proverb that it is better to teach someone how to fish than simply give them a fish. In Buchanan’s formulation, recipients have an incentive to remain poor if they are entitled to benefits as long as they are poor. Hence benefits should primarily be designed to discourage benefit dependency and eliminate moral hazard. The latter may arise when the availability (p.210) of benefits (when poor) undermines the willingness of individuals to invest in human capital (so they avoid poverty). While the argument can be evoked to support cuts in social provision, it can and has been used to support the public provision of in-kind transfers, whether in the form of job training or social insurance (Coates, 1995).

(4) Sen’s approach (1993), redefining well-being in terms of capabilities like being able to read, write, remain healthy, etc., can be seen as a more enlightened form of paternalism (cf. Deneulin, 2002), justifying social investment in public services such as education and health.

(5) A more pragmatic reason for using the insurance value approach is that most data sets used in distributional analysis (e.g., EU-SILC) do not contain information on effective use of healthcare services.

(6) This rather surprising outcome is largely due to the effect of re-ranking. Because part of the expenditures (notably those on in-hospital care) are concentrated among a very small group, this may lead more easily to re-ranking of individual beneficiaries, which dampens the equalizing effects of healthcare services (5% of the population in the survey data accounted for more than 90% of the nights spent in hospital, whereas out-of-hospital care was more widespread over the population; see Marical et al., 2008 for more details).

(7) This scale assigns a value of 1 to the household head, of 0.5 to each additional adult member, and of 0.3 to each child. See OECD (2005) (http://www.oecd.org/dataoecd/61/52/35411111.pdf) for further explanations and specifications. This is a pragmatic equivalence scale, which takes into account only differences in household size.

(8) This means that higher spending levels (either due to high use of the services or high expenditures per user) result in higher corresponding needs (and vice versa). Thus a higher level of spending corresponds to a higher recognition of needs. Remember that, as already indicated in section 6.3.1, differences in efficiency or quality of services cannot be accounted for in this analysis.

(9) Bulgaria, Malta, and Romania are not included in EU-SILC 2007. Cyprus, Lithuania, and Latvia are not included in the analysis because for these three countries we do not have estimates for the value of ECEC. Moreover, their amounts for tertiary education include expenditures on research and development, reducing comparability with other countries.

(10) Note that EU-SILC does not allow us to distinguish between participation in publicly or privately funded educational institutions, nor between general and technical secondary education, nor between Type A and Type B tertiary education. For tertiary education, amounts exclude direct expenditures for research and development activities.

(11) This approach is defendable for EU-countries, where public healthcare coverage is quasiuniversal in most countries.

(p.211) (12) In some of these countries (Denmark, Germany, and Sweden), the distinction between day care and pre-primary education in EU-SILC is probably erroneous: The number of children in the former is probably overestimated, while the number in the latter is underestimated (when compared with enrollment rates in pre-primary education in the OECD Education Database). In most countries where child care is more frequently used, day-care centers tend to be heavily subsidized.

(13) Unfortunately, there is no information available that would allow us to differentiate according to other parameters such as socioeconomic background.

(14) The starting age of compulsory education is six years in most countries, whereas some countries start earlier at age five years (Hungary, the Netherlands, and the United Kingdom), and others at age seven years (Denmark, Estonia, Finland, Poland, and Sweden). (See OECD, Education at a Glance 2010.) The ending age of compulsory education varies between 14 years and 18 years in Europe. Ninety percent of the population are enrolled in education for at least 13 years, ranging from 11 (e.g., Greece) to 15 years (Belgium, France, Norway, and Sweden).

(15) These four categories do not sum up to the total presented in Figure 6.2a, which includes also non-tertiary education outside the age category 6-16.

(16) This was a relevant exercise for five of the seven countries in their study, namely Germany, Greece, Ireland, the Netherlands, and the UK. In Belgium and Italy students in the survey are included together with their household of origin, thus making this kind of correction unnecessary.

(17) The calculations have also been done for other specifications of needs (e.g., only children aged 6 years to 16 years have education needs). Overall, the results are similar to those with the broader definition of needs as used here.

(18) The concentration coefficient of an income component is calculated in a similar way as the Gini coefficient (see e.g., Kakwani, 1977; Lambert, 2002; OECD, 2008). The difference between the two lies in the variable according to which income units are ranked. With a concentration coefficient of an income component, income units are ranked according to extended income (and not by the income component itself), while for a Gini coefficient the variable of interest and the ranking income variable are the same (namely, extended income). Because extended income is used for all income components as the ranking variable, concentration coefficients can be used to compare the distributive structure across income components. They can be considered as a summary indicator of the information provided by quintile distributions. Note that in Figure 6.2, quintiles are constructed on the basis of cash incomes (with modified OECD scale), whereas here units are ranked on the basis of extended income equivalized with the services-needs adjusted equivalence scale.