deltie.net archive

Extracted abstract

Welfare generosity is a multidimensional concept that refers to both the access to benefits and the levels of benefits (in terms of the amounts paid to recipients). However, in analyses of public support for welfare, this distinction has been largely disregarded. To gain a fuller picture of attitudes towards welfare redistribution, the current study explicitly compares the two elements and examines which distributive justice principles-that is, equality, equity, and need-are preferred to govern, on the one hand, the access to benefits and, on the other hand, their levels. The article evaluates this distinction in two different distributive contexts (pensions and unemployment benefits) and contrasts social-structural as well as ideological dividing lines. For this purpose, data from the Belgian National Elections Study 2019 are analysed. The results indicate that the access to and levels of benefits are clearly distinct dimensions in public opinion, as different justice principles are preferred for the two policy dimensions. In addition, structural equation models illustrate that the access dimension is more ideologically structured, whereas preferences regarding the levels of benefits are more strongly stratified along social-structural lines. Overall, the results imply that social justice preferences are clearly different when considering the access to benefits or their level. This distinction should be taken into account in welfare attitude research.

Introduction

Social policy literature widely acknowledges that the concept of welfare generosity is not unidimensional, but has two distinct aspects: it encompasses the access people have to particular social benefits as well as the level of the benefits (Korpi and Palme, 1998; Scruggs, 2006; Jensen et al., 2018; Otto, 2018) . While access refers to the breadth of coverage of welfare provision (which people are entitled to receive a benefit), level relates to the amount (what benefit recipients are actually given) (van Oorschot, 2013) . These two components are crucial ingredients in the resurging question 'who should get what and why?' that guides welfare reform (van Oorschot, 2000) . The who component clearly pertains to the issue of which groups should have access to national circles of solidarity, whereas the what dimension connects to the identification of meaningful thresholds for benefit levels.

Instead of fully engaging with both components, most welfare attitude research tends to focus on people's generic support for redistribution and the role of government in implementing social policies (Svallfors, 1997; Andreß and Heien, 2001; Blekesaune and Quadagno, 2003; Jaeger, 2006a) . However, citizens do not only care about the promotion of redistribution, but are also concerned with the guiding rules behind welfare allocations that structure the balance between burdens and benefits (Bowles and Gintis, 2000; Mau, 2003) . Consequently, what the general public thinks about the access to and level of welfare provision is a crucial factor affecting the legitimacy and feasibility of social policies. Furthermore, as the access and level are generally negatively correlated (in that high coverage is linked to low amounts and vice versa) and the general public can evaluate various policy design dimensions differently (van Oorschot, 2013; Gallego and Marx, 2017) , people could have distinct opinions regarding the access to benefits and the level of them. This warrants disentangling the two dimensions of benefit generosity in welfare attitude research.

In addition to providing much-needed insight into public preferences concerning the 'who ' and 'what' components, the current study also deals with the 'why' factor in the question of who should get what and why. The why refers to the normative criteria that are invoked to establish the access and level dimensions, and hence legitimize the reasons for certain social categories to reach thresholds for public resources. Taking into account this issue, the current study compares how preferred justice principles of equality, equity, and need (i.e., why) determine attitudes towards the access to (i.e., who) and level (i.e., what) of benefits. Instead of merely measuring support for the degree of redistribution in terms of the extent of benefit amounts and coverage, these principles refer to fundamentally distinct types of organizations regarding the access to and level of social benefits (Reeskens and van Oorschot, 2013) .

To investigate which justice beliefs are prevalent with regard to the access and level dimensions of benefits, we study two welfare domains that are institutionalized differently and connected to two types of social risks: pensions and unemployment benefits (Jensen, 2012; Arndt, 2017) . Not only are justice preferences themselves fundamentally reliant on the welfare benefit under consideration (Reeskens and van Oorschot, 2013; Van Hootegem, Abts and Meuleman, 2020) , but the distinction between access and level may also be intensified or weakened depending on the type of social risk concerned. In addition to establishing the distinction between the access to and level of benefits in two distributive contexts, we explore whether preferences for the two dimensions of policy design are linked to ideological and social divides. For example, preferences concerning the access dimension could be more strongly ideologically based, while discussions about levels may be more structurally stratified, as they distinguish between net beneficiaries and contributors (Hedegaard, 2014; Ennser-Jedenastik, 2021) .

In sum, we try to answer two general research questions: (i) To what extent are different justice principles (equality, equity, and need) preferred to govern the access to and level of pensions and unemployment benefits? (ii) To what extent are these justice preferences concerning the access and level socially and ideologically stratified? For this purpose, we analyse data from the Belgian National Elections Study 2019.

Theoretical framework

Public preferences concerning the access to and level of benefits Most existing public opinion studies that assess the legitimacy of the welfare state fail to distinguish between various components of distribution and are hence unable to establish which aspects of social policy design members of the public actually find attractive. What is particularly important is the distinction between two fundamental aspects of benefit generosity: the access to benefits and the level of them (Korpi and Palme, 1998; Scruggs, 2006; van Oorschot, 2013; Jensen et al., 2018; Otto, 2018) . These two dimensions are expected to be assessed differently by the public. Gallego and Marx (2017) , for example, illustrate that labour market reforms that increase the generosity of benefit amounts are the most popular, while simultaneously a restriction to the access by those in need is the most preferred option. Although this indicates that citizens may have different preferences regarding the access to and level of benefits, systematic comparisons of both policy design dimensions are lacking. To remedy this shortcoming, we focus closely on public preferences regarding different distributive mechanisms that regulate the access to and level of welfare benefits.

Instead of directly assessing the benefit amounts, their coverage, or their eligibility criteria, we study whether intrinsically different norms are involved with regard to governing the access to and level of benefits. Individuals do not behave purely out of self-interest, but are homo reciprocans or sociologicus (instead of economicus) who care deeply about the moral economy of redistribution, that is, the normative assumptions and the principles of social justice behind redistributive mechanisms (Bowles and Gintis, 2000; Mau, 2003) . Instead of adopting a rational choice framework that focuses on hard-to-understand benefit technicalities, we study more fundamental and intuitive normative rules that determine which groups should get how much. This is accomplished by capturing citizens' views on which principles of distributive justiceequality, equity, and need-should govern the access to and level of welfare benefits (Deutsch, 1975) . First, equality means that all citizens are guaranteed equal access to social protection or are entitled to an equal share of benefits and services, without any additional requirements. Second, equity implies that only individuals who have contributed sufficiently to society-in terms of social security contributions, work history, or taxes paid-are covered by social programmes or receive more generous benefits. Last, the principle of need allocates welfare resources only to those who are in need, such as the poor or the disabled, or provides larger benefit amounts to compensate for the perils faced by precarious groups.

The debate on how the access to social benefits and services should be arranged is usually based on the divides between universalism, social insurance, and selective targeting (Korpi and Palme, 1998; Laenen and Gugushvili, 2021) . Although this in part relates to how broad the coverage of benefit is, it also establishes which groups are eligible, based on which criteria. The TWO FACES OF BENEFIT GENEROSITY access dimension hence incorporates both elements of coverage and eligibility. While universal social policies are oriented towards all citizens, social insurance is intended for those who have built up sufficient rights through work contributions, and targeted benefits are only for specific groups, such as those with insufficient personal resources. Although there is an ongoing debate about which types of policies are most popular, benefits that are universal in coverage are generally expected to have greater public support (Hedegaard, 2014) . Not only are they able to align stronger interests from different social groups, but they are also considered superior in terms of procedural and substantive justice (Korpi and Palme, 1998; Rothstein, 1998; Laenen and Gugushvili, 2021) . In terms of distributive justice principles, equality in access comes down to universalism, while equity reflects a social insurance logic and need entails more targeted benefits for those in real need (Clasen and van Oorschot, 2002) .

With regard to the level of benefits, the discussion usually revolves around the distinction between flatrate benefits and earnings-related alternatives (Korpi and Palme, 1998) . Flat-rate benefits give an equal amount to everyone, while earnings-related schemes vary the amount given according to previous earnings or work history. Here, the principle of equality coincides with a flat-rate scheme that gives everyone the same amount, whereas the principle of equity corresponds to an earnings-related benefit that rewards those who have contributed more (Clasen and van Oorschot, 2002) . When the principle of need governs the level of benefits, this translates into higher benefits for lower-income groups, for instance.

Access and level in two distributive contexts: pensions and unemployment benefits

Justice preferences and welfare attitudes depend strongly on the domain under consideration and on the design of a social policy (Reeskens and van Oorschot, 2013; Gallego and Marx, 2017; Van Hootegem, Abts and Meuleman, 2020) . In fact, standards of social justice do not apply universally but vary depending on the context and on what is being distributed. As a result, preferences for equality, equity, and need in governing the access to and level of benefits can also be expected to vary across benefit schemes or welfare domains (Andersen, 2011; Hedegaard, 2014) . To take this domain specificity into account, we examine justice preferences for the access and level dimensions for both pensions and unemployment benefits. After all, these two benefit schemes refer to distinct types of social risks that vary in a number of important characteristics (Jensen, 2012; Arndt, 2017; Green-Pedersen and Jensen, 2019) . However, since the distinction between the access to and level of benefits is central, the goal is to study this division in two distributive contexts rather than focussing on a comparison between the benefits.

In the Belgian welfare state-the context of this study-the institutionalized logics governing the access to and level of pensions and unemployment benefits vary widely. To begin with, pensions are connected to the life course-related risk of retiring or ageing, which is an unavoidable and normal part of the life cycle (Green- Pedersen and Jensen, 2019) . In this sense, this benefit scheme responds to a social risk that is widely prevalent, leading most individuals to have a strong interest in wide access to pensions (Jensen, 2012) . Beyond self-interest, the elderly are regarded as highly deserving of welfare support, as they are not in control of their situation and are perceived to be grateful (van Oorschot, 2006; Meuleman, Roosma and Abts, 2020) . This is also reflected in the Belgian pension system, as entitlement can already be obtained from the moment someone has worked a single day and the access to pensions is hence almost entirely universal.

According to policy feedback theories, citizens can adjust their preferences to be in line with institutional designs and the norms they encapsulate (Rothstein, 1998; Mau, 2004; Kumlin and Stadelmann-Steffen, 2014) . Based on policy feedback and the predominant characteristics of this type of social risk, we would expect the principle of equality to be predominantly supported to govern the access to pensions. However, once this universal access is granted, the justice logic of determining the level of pension benefits is different. Retirement is fully predictable; thus, fostering the belief that individuals should participate in the labour market (Reeskens and van Oorschot, 2013) . In fact, this logic is embedded in the Belgian welfare state, as pensions are earnings-related and based on the number of years worked and the employment history. In line with this, we would expect that the principle of equity is preferred for the level of pensions. In this sense, a clear gap between justice preferences on the access to and level of pensions is expected.

Unemployment benefits are associated with a labour market-related social risk, as unemployment does not occur as an intrinsic part of everyone's life cycle but is fundamentally tied to a person's position in the labour force. As unemployment is especially prevalent among low-income groups, the skewed social stratification weakens broad interests in encompassing social protection against this risk (Jensen, 2012) . In addition, there is a strong sense of internal control regarding unemployment, as the belief is widespread that it is self-inflicted and that the unemployed could find a job if desired (Furåker and Blomsterberg, 2003; Meuleman, Roosma and Abts, 2020) . Institutionally, access to unemployment benefits is not universal in Belgium, but contingent on having worked or contributed for a minimum period to become covered and be eligible for unemployment protection (Van Lancker et al., 2015) . These elements are related to both the institutionalization and the characteristics of the underlying social risk and may stimulate public preferences for organizing access to unemployment benefits around the principle of equity. In terms of the level, the Belgian system of unemployment benefits is based on the principles of equity and need. Benefit levels are determined in line with the previous earnings of a benefit claimant and also differentiate with regard to need (for example, family status). As a result, one could expect that a considerable proportion of citizens would prefer equity or need to govern the levels of unemployment benefitsalthough previous research has shown that equality for unemployment benefit levels is also widely supported (Reeskens and van Oorschot, 2013; Van Hootegem, Abts and Meuleman, 2020) . This pattern also implies that we would expect smaller differences in preferences for justice principles governing the access to versus the level of unemployment benefits, compared with pensions.

Comparing social and ideological divides

As a last step in the comparison, we analyse how justice preferences regarding the access to and level of benefits are driven by self-interest and ideology (Arts and Gelissen, 2001; Jaeger, 2006b; Reeskens and van Oorschot, 2013) . First, the self-interest framework assumes that individuals make cost-benefit calculations and support policies of which they are likely to be (or to become) beneficiaries (Kangas, 1997; Jaeger, 2006b) . Groups with a higher socio-economic status are generally found to be more in favour of equity, while deprived groups gravitate more towards equality and need (D'Anjou, Steijn and Van Aarsen, 1995; Arts and Gelissen, 2001; Reeskens and van Oorschot, 2013) . Second, the ideology framework assumes that social policy preferences are a function of broader coherent systems of political norms and values (Jaeger, 2006b; van Oorschot, Reeskens and Meuleman, 2012) . Right-wing individuals, for instance, tend to adopt more restricted notions of solidarity, potentially leading to lower support for egalitarian distribution and greater popularity of allocation according to the principle of equity or need (van Oorschot, 2006; Reeskens and van Oorschot, 2013) .

According to the self-interest framework, the core conflict is between redistributive winners and losers, who benefit to varying degrees from the allocation of welfare. However, it is difficult to identify clear beneficiaries and losers in terms of access, as 'the welfare state provides something for everyone' (Mau, 2003: p. 90 ). Even when distribution is restricted to particular groups in terms of access, the possibility of future exposure to risk remains. This blurring of interests could lower social-structural conflicts in terms of access and instead increase interest in universalism for various social groups (Laenen and Gugushvili, 2021) . We thus expect that preferences concerning the access to benefits are most clearly structured along ideological dividing lines, whereby the more conditional interpretations of solidarity among right-wing and authoritarian individuals translate into greater support for targeting rather than universalism (cf. van Oorschot, 2006; Van Hootegem, Abts and Meuleman, 2020) . Ideological contestation resolves strongly around defining the 'similar others' who should be part of the collective insurance schemes and hence will be entitled to support from the state. For example, this is clear when radical right-wing parties define the logic of welfare deservingness in terms of ethnic identity (cf. De Koster, Achterberg and Van der Waal, 2013). Such conflicts regarding the deservingness of groups and the debate between targeting versus universalism are strongly ideologically driven and introduce a substantial degree of partisan conflict between left and right (Ennser-Jedenastik, 2021) .

Once the fundamental debate on who should get access to benefits has been settled, the question then becomes what these beneficiaries should get. This creates a more pronounced divide between those who receive higher levels of benefits (the recipients) and those who mainly contribute (the payers) (Mau, 2003) . In fact, research indicates that people's degree of proximity or direct interest in welfare provision shapes their attitudes about the level of benefit schemes (Hedegaard, 2014) . However, while various social groups may still benefit from relatively broad access, we expect stronger socio-economic divisions in the criteria governing the levels, where earnings-related versus flat-related benefits clearly have distinct implications concerning how much different groups actually receive. Although disputes about levels are certainly still ideologically driven (Jensen and Kevins, 2019; Pedersen, 2019) , this may constitute less of a polarized issue than the access to welfare. There is a much more blurred ideological gap in debates on the level or degree of redistribution and taxation (Esarey, Salmon and Barrilleaux, 2012; Ennser-Jedenastik, 2021) . In sum, for the level of benefits, we expect social stratification in particular to occur but that there will be a weaker ideological differentiation than for the access dimension.

To assess these social and ideological cleavages for both the access and the level, they are again studied in the specific contexts of pensions and unemployment benefits. This enables us to establish whether the differences in the determining factors of justice preferences for the access as well as the level hold across TWO FACES OF BENEFIT GENEROSITY distributive contexts or whether they are specific to the benefit under consideration.

Data and methods

Data

The current investigation uses data from the Belgian National Elections Study 2019, a large probability-based survey. The National Register functioned as the sampling frame to select individuals of 18 years and older who were entitled to vote in the federal elections of May 2019. A two-stage random sampling design was used, in which a response rate of 32.81 per cent was achieved. Due to the COVID-19 pandemic during the fieldwork period, respondents were interviewed using different survey methods, namely Computer-Assisted Personal Interviewing (a mix of face-to-face and video calls) and Computer-Assisted Web Interviewing. All the respondents who participated in the interview also received a follow-up drop-off questionnaire with additional survey questions (including items on the justice preferences). In total, 1,659 respondents participated in the main interview, of whom 1,129 completed the drop-off survey. Post-stratification weights (for age, gender, and education) were applied to correct for selective non-response.

Indicators: dependent variables

Our dependent variables are support for the justice principles in governing (i) the access to benefits and (ii) the level of benefits. For these dimensions, three justice principles were presented (equality, equity, and need) to the respondents, who were asked to rate how fair they deemed this situation to be (on a five-point answer scale, from 'very unjust' to 'very just'). Every respondent was thus asked six questions referring to access and level for each of the three justice principles applied to a single benefit context. This within-subject design-which presented all the respondents with questions about the preferred principles for the access to and level of benefits-was intended to maximize power and hence facilitate the comparison between the two components. A rating approach in which all individuals expressed their support for each of the principles was adopted rather than a ranking approach that forces individuals to choose a single principle as their favourite. This rating approach reduces order effects by allowing individuals to express support for multiple ideas simultaneously (Vriens, 2015; Van Hootegem, Abts and Meuleman, 2020) . The specific benefit referred to (pension vs. unemployment benefit) was randomized between respondents in a split-ballot experiment. This between-subject variation of the domain under consideration was intended to avoid overburdening respondents and to assess absolute support in a single domain rather than relative to another social policy programme. The question wordings and percentages of respondents perceiving the different principles as equitable are presented in Table 1 . In addition, the correlations between support for the principles governing the access and level are shown in Supplementary Table A1 .

Table 1 Question wordings and percentage of respondents considering distributive principles as just To what extent do you think that the following situations are just or unjust? Pensions (% just) Unemployment benefits (% just) Equality access-that everyone gets equal access to pensions/unemployment benefits 58.3 56.1 Equality level-that the level of pensions/ unemployment benefits is equal for everyone 38.5 49.6 P-value McNemar test for difference between access and level 0.000 0.000 Equity access-that only people who have worked hard enough get access to pensions/unemployment benefits 32.0 32.9 Equity level-that people who have worked harder receive higher pensions/unemployment benefits 56.4 41.3 P-value McNemar test for difference between access and level 0.000 0.000 Need access-that only the poor and people in need get access to pensions/unemployment benefits 4.6 14.2 Need level-that people who are poor and in need receive higher pensions/unemployment benefits 15.5 22.3 P-value McNemar test for difference between access and level 0.000 0.000 Downloaded from https://academic.oup.com/esr/advance-article/doi/10.1093/esr/jcad053/7251287 by guest on 02 January 2024

Indicators: independent variables

Two ideological dimensions are included as independent variables in our regression models. Left-right placement, as a broad indicator of political ideology, is measured by a single item on an 11-point scale (0 = left; 10 = right). As a cultural dimension of ideology, we additionally include authoritarianism to predict justice preferences. This concept is measured by three, five-point items (disagree/agree) to examine whether respondents thought that most problems would be solved if we could get rid of immoral and dishonest people, that obedience and respect for authority are important virtues, and that laws should become stricter. To test the measurement validity of this ideological dimension, we conducted a multigroup confirmatory factor analysis for respondents who were given questions on pensions and unemployment benefit. A scalar invariant model that restricts loadings and intercepts to be equal across group shows a good fit with the data (χ 2 = 3.809; df = 4; root-mean-square error of approximation = 0.000; comparative fit index = 1.000; Tucker-Lewis index = 1.000; standardized root-mean-square residual = 0.037). Supplementary Table A2 shows that the first item of the authoritarianism scale has a somewhat weak loading (<0.40). As all other loadings are sufficiently high and using only two items for a latent concept leads to problems of model identification, we proceeded with this model and saved the factor scores (to reduce the complexity of the final model) to be included in the regression models.

The social structure is operationalized by education, employment situation, and subjective income. For education, we created the following three categories: none to lower-secondary, upper-secondary (reference category), and tertiary education. Current employment situation is divided into three categories and hence operationalized by two dummy variables: the employed (reference category), pensioners, and the inactive (including students, the unemployed, disabled people, etc.). Subjective income is used instead of objective income because the latter variable suffers from a high degree of item non-response. The subjective income variable is measured by using four answer categories that respondents could choose from to describe how they felt about their income: (i) 'have more than enough and can easily save', (ii) 'have sufficient to get by without difficulties', (iii) 'have just sufficient to get by', and (iv) 'regularly have difficulties getting by'. To create groups sufficiently large to compare, a dummy was created to combine the first two categories and the last two categories, respectively (high income = reference category). The models also control for gender (female = reference category), age, and region-Flanders (reference category) and Francophone Belgium (Wallonia and Brussels).

Statistical modelling

Two distinct approaches were taken in order to answer our research questions. To start with, we provide a descriptive overview of preferences for equality, equity, and need in governing access and levels and compare them for both pensions and unemployment benefits (Table 1 ). To assess whether the percentage of respondents that rated a particular situation as equitable differs across the access and level dimensions, a series of McNemar tests was conducted. The McNemar test was chosen because of the paired nature of the comparison (a single respondent rates the level as well as the access) (McNemar, 1947) . In this case, the test was used to assess whether the marginal distribution of a particular item differs depending on whether the access or the level dimension is referred to. If this null hypothesis is rejected, there is a significant difference in justice preferences between access and level. As a second step in the analysis, a multigroup structural equation model (SEM) was estimated to compare the impact of the social structure and ideology between respondents who answered questions on the access and levels of pensions versus on unemployment benefits. These coefficients were evaluated for all six dependent variables simultaneously. Descriptive statistics were generated using SPSS version 27, and all other analyses were conducted in Mplus version 8.4 (Muthén and Muthén, 2017) .

Results

Comparing opinions on the access to and level of benefits in two welfare domains Table 1 displays the percentage of respondents considering equality, equity, and need as (very) just in governing the access to and level of pensions and unemployment benefits. It should be noted that individuals were asked to report support for each principle separately, which means that the percentages do not necessarily sum up to 100. For the access to pensions, a majority of 58 per cent considered it to be just to guarantee equal coverage spanning everyone. This is in line with the strong interest across groups in broad access to pensions due to the high prevalence, as well as with the institutional design of relatively universal access to pensions in Belgium. When the level of unemployment benefits is considered, however, only 39 per cent consider equality to be just. The majority prefer to put in place earnings-related benefits in line with the principle of equity (56 per cent). This preference can be understood from the strong predictability of this social risk and the institutional logic of earnings-related pensions in Belgium, as this takes into account previous income as well as work trajectories in calculating the levels of pensions. Making the access to (rather than the level TWO FACES OF BENEFIT GENEROSITY of) pensions dependent on this reciprocal logic is, however, only supported by about a third of the respondents. The proportion considering the principle of need as being just is marginal for both the access and levels of pensions, but there is still a stronger degree of support for need in the level (16 per cent) than for access (5 per cent). According to the McNemar tests, all differences in justice preferences between the access and level dimensions of pensions are strongly significant.

For unemployment benefits, we also see notable differences between the popularity of the principles governing the coverage and amounts, although they are less pronounced. With regard to pensions, equality in access is most broadly supported (56 per cent). This is contrary to what was expected and is not aligned with the socially stratified nature of this labour-specific risk and the institutional organization of unemployment benefits in Belgium. In addition, equality is the most popular justice principle concerning the benefit levels (about half of the respondents consider this principle as being just). The dominance of support for flat-rate unemployment benefits is somewhat surprising, given the strong perceived internal control in facing unemployment as well as the predominance of equity and need to set the levels of this benefit in the Belgian welfare state. Nevertheless, the pronounced preferences for equality could also be related to the order of the survey questions, as equality came first and could hence have been given broader support because the alternatives had not yet been presented. Furthermore, we observe that distribution based on equity and need is more popular when the level of benefits is concerned rather than the access. For both justice principles, the difference between access and level equals almost 10 percentage points (for equity: 33 versus 41 per cent; for need: 14 versus 22 per cent). These differences are significant for both the access and the level dimensions. It should be noted that need-based distribution is more popular when unemployment benefits are concerned instead of pensions.

Although not the primary interest of this article, it is also interesting to examine the correlations between support for the principles in the access to and level of benefits (see Supplementary Table A1 ). The correlations show that support for equality in access is mostly combined with support for equality and need in terms of levels, while equity in access is only strongly associated with earnings-related benefits (i.e., equity). Support for benefits that are only granted to people in need goes hand in hand with support for higher levels of benefits for those in need or flat-rate benefits. Overall, the correlations between access and level are moderate, indicating their separate conceptual nature. In sum, we can conclude that the support for the principles governing the access and level differs in both distributive contexts and that it is highly relevant to distinguish the two aspects of benefit generosity.

The social-structural and ideological conflicts surrounding opinions about access and level

In the next step, we analyse whether justice preferences for the access and level dimensions are differently related to social-structural and ideological predictors. Table 2 shows the unstandardized regression effects of social-structural and ideological variables on preferences for equality, equity, and need in governing the access to and level of benefits. Respondents in the condition referring to pensions and unemployment benefits are treated as two groups in a multigroup SEM in order to facilitate comparison between the two experimental conditions. The significance of the difference between the parameters for the access and level of each justice principle is also presented. However, this should be interpreted with caution, as power is limited and it is hence more interesting to look at differences in which predictors work significantly.

We start by looking at preferences for the equality principle in the case of pensions. Only three determinants predict support for equal access and equal levels similarly. Those with a tertiary education as well as younger respondents and right-wing individuals are less inclined to consider equality as a just basis for both the policy dimensions. This can be understood in terms of, respectively, the lower interest these groups have in broad redistribution and their more conditional notions of solidarity (Jaeger, 2006b; van Oorschot, 2006) . For the equality in access and levels, pensioners are less likely to consider this equitable than those in employment, which is relatively surprising and counter to their own interests (although this relationship is not significant for the levels of pensions). However, the effects of income and authoritarianism differ substantially between the access and level dimensions, illustrating that different forces are at play. Authoritarianism has a strong negative impact on support for equality in the access dimension, but is not significantly related to preferences for equality in the level dimension. This supports the theory that discussions about the coverage of pensions are more strongly rooted in ideological factors than debates on pension levels. For the level dimension, we observe stronger social-structural gradients: men and higher-income individuals are less inclined to support equal benefit levels, which aligns with their own interests. Once the ideologically loaded access debate has been settled, material interests and conflicts between net recipients and contributors apparently play a role in determining justice preferences regarding the level of pensions (cf. Mau, 2003; Hedegaard, 2014) . Yet despite substantial differences in terms of the significance of several predictors, we

Table 2 Unstandardized regression coefficients of the social structure and ideology on preferences for equality, equity, and need in governing the access to and level of pensions and unemployment benefits (N pensions = 465; N unemployment = 500) Equality access Equality level Diff. Equity access Equity level Diff. Need access Need level Diff. Pension Gender Female (ref.) Male -0.165 -0.416*** * 0.190 0.155 0.006 -0.075 Age 0.013** 0.015** -0.009* -0.008* 0.006 0.012*** Education Lower -0.055 0.194 0.142 -0.272* * 0.234* 0.135 Upper secondary (ref.) Tertiary -0.378** -0.316* 0.102 -0.028 -0.141 -0.015 Employment Employed (ref.) Pensioner -0.314* -0.313 0.185 0.194 0.049 0.126 Inactive 0.010 -0.151 -0.069 0.127 0.167 0.421** Income Low income -0.011 0.284* * 0.199 -0.126 0.069 0.134 High income (ref.) Region Flanders (ref.) Francophone Belgium 0.144 -0.165 * -0.124 0.165 -0.193* -0.323** Left-right -0.078*** -0.058* 0.125*** 0.054* -0.002 -0.021 Authoritarianism -0.403* -0.285 0.659*** 0.508** -0.164 -0.166 R² 0.103 0.131 0.132 0.075 0.095 0.140 Unemployment benefits Gender Female (ref.) Male 0.112 -0.105 -0.035 0.002 0.052 -0.006 Age -0.002 0.006 0.000 0.001 -0.001 0.005 Education Lower 0.387** 0.103 0.128 -0.067 0.155 -0.011 Upper secondary (ref.) Tertiary 0.188 -0.399*** *** -0.153 0.146 * -0.175 -0.052 Employment Employed (ref.) Pensioner 0.036 0.009 -0.013 0.038 -0.055 -0.007 Inactive 0.218 0.189 -0.067 0.036 0.157 0.026 Income Low income 0.008 -0.061 0.068 0.303* 0.052 0.212* High income (ref.) Region Flanders (ref.) Francophone Belgium 0.220* -0.134 ** -0.121 -0.134 -0.283** -0.195* Left-right -0.070** -0.040 0.138*** 0.089*** * 0.006 -0.034 Authoritarianism -0.614** -0.526* 0.005 0.070 0.115 -0.224 R² 0.113 0.073 0.112 0.063 0.063 0.044 Note: Diff. = significance of difference between parameters for the access and the level. *P ≤ 0.05; **P ≤ 0.01; ***P ≤ 0.001. Downloaded from https://academic.oup.com/esr/advance-article/doi/10.1093/esr/jcad053/7251287 by guest on 02 January 2024 TWO FACES OF BENEFIT GENEROSITY

only observe significant differences in the size of the parameters for gender, income, and regional gradient.

For the principle of equity for pensions, the social-structural and ideological effects are more similar in terms of their statistical significance. Older respondents and those who are more left-wing and less authoritarian show a lower likelihood to support reciprocity both in access and levels. This broadly supports both the ideology and self-interest theories. However, we again observe that the ideological effects are considerably stronger for the access dimension, suggesting that there seems to be stronger ideological debate concerning access than levels (Ennser-Jedenastik, 2021) . Despite the considerably larger coefficients of the ideological parameters with regard to access, only the left-right factor has a statistically significant different regression coefficient for access and level. The main difference regarding equity is that lower education decreases support for equity in the levels of pensions, but does not affect the preference for equity in the access to benefits. The gap in the size of the parameters is also statistically significant for education. This demonstrates that social stratification appears to be slightly stronger for levels than for access. With regard to the need principle, we observe notable differences between the access and level factors, as only political ideology has the same impact on both policy dimensions (in that right-wing respondents are less supportive of need-based distribution). In contrast to equity and equality, the ideological stratification is stronger for the levels of pensions than for access. In addition, lower-educated individuals are more inclined to support the principle of need, with regard to access, while older respondents and the inactive are more in favour of higher benefits for those in need. However, this does not translate into statistically significant differences in the size of the parameters.

In sum, the patterns for pensions are largely in line with the expectations of the self-interest and ideology frameworks. The results also clarify that especially for equality and equity, the access dimension is more strongly ideologically debated, while the level aspect follows more social-structural dividing lines. This provides further evidence that the discussion on whether to choose universalism, social insurance, or need-based targeting-as well as which groups are deserving of welfare support-is heavily disputed across ideological lines, while discussions on how to establish benefit levels show a less clear ideological gap. In addition, this illustrates that personal stakes might be much clearer for benefit amounts when the more fundamental debate on who are members of redistributive circles has been decided on. The need principle provides an exception to this pattern. This could be related to the more ambiguous conceptualization of this principle, which can be interpreted in various ways (Kittel, 2020; Van Hootegem, Meuleman and Abts, 2021) .

In the case of unemployment benefits, there are also relevant differences between the access and level dimensions. Authoritarianism significantly predicts lower support for both equality in access to and levels of unemployment benefits, but the parameter is larger for the access than for the levels (yet not significant). In addition, lower-educated individuals, those from Francophone Belgium and left-wingers, are more supportive of equality in access, thus supporting the self-interest and ideology theories. For the level, we observe that individuals with tertiary education are less in favour of equality than those with upper-secondary education, and this also differs significantly between the access and level aspects. The high number of significant ideological predictors for the access illustrates stronger ideological contestation, but the social stratification in justice preferences is not necessarily more apparent for the level dimension. In terms of support for the equity principle, fewer differences between access and level are observed, as only political ideology has a significant influence on equity preferences. With regard to both the access and level aspects, right-wing individuals are more conditional and hence are more supportive of distribution types that involve a quidpro-quo welfare model. However, the coefficient is considerably and significantly larger for the access than for the level. For the need principle, only the region significantly shapes support for both dimensions, in that individuals living in Francophone Belgium are less likely to prefer need-based access and levels for unemployment benefits. In addition, those with a low income are significantly more supportive of higher benefit amounts for those who are more in need (without prioritizing need in access), thus also conforming to the self-interest theory.

In the distributive context of unemployment benefits, it thus becomes clear that the access dimension is more strongly ideological polarized. This is once more in line with the prediction that there is a broader ideological gap surrounding the fundamental debate on who is entitled to welfare support (Ennser-Jedenastik, 2021). However, in contrast to pensions, the stronger social stratification for the level dimensions was not observed in the context of unemployment benefits. This corresponds to the smaller overall differences between the access and level dimensions for unemployment benefits than that for pensions reported in the previous section.

Conclusion and discussion

The two aspects of benefit generosity, namely the access to and level of benefits, are recognized extensively in social policy literature (Korpi and Palme, Downloaded from https://academic.oup.com/esr/advance-article/doi/10.1093/esr/jcad053/7251287 by guest on 02 January 2024 1998; Scruggs, 2006; van Oorschot, 2013; Jensen et al., 2018; Otto, 2018) . Nevertheless, public opinion research analysing people's support for redistribution and government responsibility in implementing social policies does not explicitly recognize the distinction between the two dimensions. Accordingly, previous studies fail to separate the question of who should acquire resources from that of what amount should be provided. These two modalities concerning the balance between benefits and the burdens of welfare distribution are, however, at the forefront of contemporary welfare debates and public concerns (Bowles and Gintis, 2000; van Oorschot, 2000) . To fill this gap, the current article provides insights into the distinction between justice preferences regarding the access to and level of benefits. Concretely, we study public support in the two distributive contexts of pensions and unemployment benefits and compare the social-structural and ideological dividing lines.

A descriptive overview reveals that public opinions regarding the access and the level are clearly distinct, as different redistributive principles were considered fair for both dimensions of benefit generosity. For pensions, equality is most broadly supported in governing the access, while equity proves most popular to govern how benefit levels are established. This is in line with the strong prevalence and simultaneous predictability of the associated life course-related social risk of retiring, as well as with the institutional logic in accordance with which the access and level are actually determined in the Belgian welfare state. However, policy feedback effects could not be tested explicitly, as there is no counterfactual case with a different institutional setup. For unemployment benefits, equality is found to be considered the fairest principle for both the access and the level, but there are still considerable differences: when the level rather than the access is the subject, the popularity of the equality principle decreases, while equity and need become more popular. The importance of the distinction between access and levels is further demonstrated by the finding that the access dimension is more strongly rooted in ideological predictors (especially authoritarianism), while the level dimension is structured more strongly along a social gradient. The social stratification of the level dimension is, however, especially apparent in the domain of pensions and less so for unemployment benefits. All in all, these findings illustrate the relevance of distinguishing opinions about the coverage of social schemes and the amount of benefits.

The findings support the multidimensional nature of welfare attitudes and demonstrate that merely focussing on aggregate support for redistribution or government involvement misses important internal differentiation in public preferences (Roosma, Gelissen and van Oorschot, 2013) . For example, when finding high support for the role of government in organising redistribution, this does not yet indicate which allocation mechanisms are actually liked (or disliked). In this sense, we go a step further by indicating that policy design and distinct aspects of benefit generosity matter in understanding what types of distribution the public actually prefers (cf. Gallego and Marx, 2017) . This is especially essential in an era in which the question is no longer so much whether redistribution should be institutionalized, but in which debates especially revolve around how to organize this in a way that guarantees that acceptable amounts are given to rightfully deserving social groups (van Oorschot, 2000) . However, as this article focuses on a broad comparison between the access to and level of benefits, it does not explicitly assess how these aspects are combined and balanced against each other in the design of specific welfare benefits. As a result, this could be a fruitful trajectory for future research efforts.

The distinction between the access and level is nevertheless important not only to provide a clearer perspective on welfare state legitimacy, but also to further our understanding of how policies and institutions impact on public ideas. As mentioned, the policy feedback literature assumes that institutional norms and ideas can be taken over by citizens, leading to public preferences conforming to the predominant institutional logics (Rothstein, 1998; Mau, 2004; Kumlin and Stadelmann-Steffen, 2014) . In unravelling policy feedback effects, it can be crucial to distinguish the access to and level of social policies, as a match (or mismatch) between preferences and institutions depends on the dimension under consideration. This could also partly explain why it is often difficult to establish a clear link between typologies of welfare regimes, as these include both access and level dimensions to categorize social security systems, as well as public opinion (Esping- Andersen, 1990; Svallfors, 2012; Kumlin and Stadelmann-Steffen, 2014) . Instead, the institutionpreference link could be situated more strongly on a meso-level that considers the design characteristics and multidimensionality of benefits and services (Laenen, 2018) .

References

1. J G Andersen Ambivalent Values: Universalism or Targeting? Welfare State Attitudes in Denmark. CCWS Working Paper No. 2011-73 2011. June 2021
2. Public policy-making and risk profiles: Scandinavian centre-right governments after the turn of the millennium H Andreß T Heien European Political Science Review 17 2001. 2017 Arndt, C Four worlds of welfare state attitudes? A comparison of Germany, Norway, and the United States European Sociological Review
3. Welfare states, solidarity and justice principles: does the type really matter? W Arts J Gelissen 2001 Acta Sociologica 44
4. Public attitudes toward welfare state policies: a comparative analysis of 24 nations M Blekesaune J Quadagno European Sociological Review 19 2003
5. Reciprocity, self-interest, and the welfare state S Bowles H Gintis Nordic Journal of Political Economy 26 2000
6. Changing Principles in European Social Security Jochen Clasen Wim Van Oorschot 10.1023/a:1020520321533 European Journal of Social Security European Journal of Social Security 1388-2627 2399-2948 4 2 2002 SAGE Publications
7. Social position, ideology, and distributive justice L D'anjou A Steijn D Van Aarsen Social Justice Research 8 1995
8. The new right and the welfare state: the electoral relevance of welfare chauvinism and welfare populism in the Netherlands W De Koster P Achterberg J Van Der Waal International Political Science Review 34 2013
9. Equity, equality, and need: what determines which value will be used as the basis of distributive justice M Deutsch Journal of Social Issues 31 1975
10. What drives partisan conflict and consensus on welfare state issues? Laurenz Ennser-Jedenastik 0000-0002-0107-5093 10.1017/s0143814x20000240 Journal of Public Policy J. Pub. Pol. 0143-814X 1469-7815 41 4 2021 Cambridge University Press (CUP)
11. What motivates political preferences? Self-interest, ideology, and fairness in a laboratory democracy J Esarey T C Salmon C Barrilleaux Economic Inquiry 50 2012
12. The Three Worlds of Welfare Capitalism G Esping-Andersen 1990 Polity Press Cambridge
13. Attitudes towards the unemployed. An analysis of Swedish survey data B Furåker M Blomsterberg International Journal of Social Welfare 12 2003
14. Multi-dimensional preferences for labour market reforms: a conjoint experiment A Gallego P Marx Journal of European Public Policy 24 2017
15. Electoral competition and the welfare state C Green-Pedersen C Jensen West European Politics 42 2019
16. The policy design effect: proximity as a micro-level explanation of the effect of policy designs on social benefit attitudes T F Hedegaard Scandinavian Political Studies 37 2014
17. Welfare regimes and attitudes towards redistribution: the regime hypothesis revisited M M Jaeger European Sociological Review 22 2006
18. What makes people support public responsibility for welfare provision: self-interest or political ideology? A longitudinal approach M M Jaeger Acta Sociologica 49 2006b
19. Labour market-versus life course-related social policies: understanding cross-programme differences C Jensen Journal of European Public Policy 19 2012
20. Policy instruments and welfare state reform C Jensen Journal of European Social Policy 28 2018
21. Self-interest and the common good: the impact of norms, selfishness and context in social policy opinions C Jensen A Kevins The Journal of Socio-Economics O E Kangas 67 2019. 1997 Numbers and attitudes towards welfare state generosity Political Studies
22. Need-Based Distributive Justice: An Interdisciplinary Perspective B Kittel Traub S. and Kittel B. 2020 Springer Cham Need-based justice: a sociological perspective
23. The paradox of redistribution and strategies of equality: welfare state institutions, inequality, and poverty in the Western countries W Korpi J Palme American Sociological Review 63 1998
24. S Kumlin I Stadelmann-Steffen How Welfare States Shape the Democratic Public: Policy Feedback, Participation, Voting, and Attitudes Cheltenham Edward Elgar Publishing 2014
25. Do institutions matter? The interplay between income benefit design, popular perceptions, and the social legitimacy of targeted welfare T Laenen Journal of European Social Policy 28 2018
26. Are universal welfare policies really more popular than selective ones? A critical discussion of empirical research T Laenen D Gugushvili International Journal of Sociology and Social Policy 41 2021
27. The Moral Economy of Welfare States: Britain and Germany Compared S Mau 2003 Routledge London
28. Welfare Regimes and the Norms of Social Exchange Steffen Mau 10.1177/0011392104039314 Current Sociology Current Sociology 0011-3921 1461-7064 52 1 2004 SAGE Publications
29. Note on the sampling error of the difference between correlated proportions or percentages Q Mcnemar Psychometrika 12 1947
30. Welfare deservingness opinions from heuristic to measurable concept: the CARIN deservingness principles scale B Meuleman F Roosma K Abts Social Science Research 85 102352 2020
31. Mplus User's Guide. 8th edn L K Muthén B Muthén 2017 Muthén and Muthén Los Angeles
32. Social expenditure, social rights, and benefit receipt as indicators of welfare state generosity: three peas in a pod, or a different kettle of fish altogether? A Otto International Journal of Sociology and Social Policy 38 2018
33. Attitudes on the size of unemployment benefits: Ideology and numbers R T Pedersen Scandinavian Political Studies 42 2019
34. Equity, equality, or need? A study of popular preferences for welfare redistribution principles across 24 European countries Tim Reeskens Wim Van Oorschot 10.1080/13501763.2012.752064 Journal of European Public Policy Journal of European Public Policy 1350-1763 1466-4429 20 8 2013 Informa UK Limited
35. 10.1093/esr/jcad053/7251287 January 2024
36. The Multidimensionality of Welfare State Attitudes: A European Cross-National Study Femke Roosma John Gelissen Wim Van Oorschot 10.1007/s11205-012-0099-4 Social Indicators Research Soc Indic Res 0303-8300 1573-0921 113 1 2013. 1998 Springer Science and Business Media LLC Rothstein, B; Cambridge The multidimensionality of welfare state attitudes: a European cross-national study
37. The generosity of social insurance, 1971-2002 L Scruggs Oxford Review of Economic Policy 22 2006
38. Worlds of welfare and attitudes to redistribution: a comparison of eight Western nations S Svallfors European Sociological Review 13 1997
39. S Svallfors Contested Welfare States: Welfare Attitudes in Europe and Beyond Stanford Stanford University Press 2012
40. Differentiated distributive justice preferences? Configurations of preferences for equality, equity and need in three welfare domains A Van Hootegem K Abts B Meuleman Social Justice Research 33 2020
41. Measuring public support for distributive justice principles: assessing the measurement quality of the Basic Social Justice Orientations scale A Van Hootegem B Meuleman K Abts International Journal of Public Opinion Research 33 2021
42. Leven aan de dop: De Impact van Werkloosheid op het inkomen in België in Vergelijkend Perspectief W Van Lancker 2015 Antwerp Universiteit Antwerpen
43. W Van Oorschot Who should get what, and why? On deservingness criteria and the conditionality of solidarity among the public. Policy and Politics 2000 28
44. Making the difference in social Europe: deservingness perceptions among citizens of European welfare states W Van Oorschot Journal of European Social Policy 16 2006
45. Comparative welfare state analysis with survey-based benefit recipiency data: the 'dependent variable problem' revisited W Van Oorschot European Journal of Social Security 15 2013
46. Popular perceptions of welfare state consequences: a multilevel, cross-national analysis of 25 European countries W Van Oorschot T Reeskens B Meuleman Journal of European Social Policy 22 2012
47. I Vriens Two of a Kind? Comparing Ratings and Rankings for Measuring Work Values Using Latent Class Modeling Tilburg BOXPress BV 2015