How Popular Are Social Investment Policies Really? Evidence from a Survey Experiment in Eight Western European Countries

How Popular Are Social Investment Policies Really? Evidence from a Survey Experiment in Eight... Abstract The concept of the social investment welfare state has received a lot of attention and support both from academics and policymakers. It is therefore commonly assumed that policies such as investing in education or family services would also receive significant support from the mass public. While there are some indications of this, existing comparative surveys of public opinion usually do not take into account how citizens perceive and react to policy trade-offs, i.e. how they respond when forced to prioritize between different types of social policies, which is more realistic given budget constraints. This article presents original data from a representative survey of public opinion in eight Western European countries, studying how support for social investment policies changes when additional spending on these policies would have to be financed with cutbacks in other parts of the welfare state. The central findings are that citizens generally dislike being forced to cut back one type of social spending to expand another, but there is a significant degree of variation across individuals and policy fields. Material self-interest and ideological predispositions as well as their interaction help understanding differences in the acceptance of these trade-offs. The findings have important implications for the political viability of social investment policies. Political parties aiming to expand social investment in a context of fiscal austerity are confronted with different and distinct electoral constraints and challenges given the respective preferences of their electorates. Introduction In recent years, the notion of the ‘social investment’ model of the welfare state has gained much attention both among scholars and policymakers (Morel, Palier and Palme, 2012; Bonoli, 2013; Hemerijck, 2013, 2017). Social investment policies can be defined as policies geared at creating, mobilizing, or preserving human skills (Garritzmann et al., 2017: 36ff.), e.g. education, active labour market, or childcare policies. These ‘new’ types of social policies are often contrasted with more traditional social policies such as pensions, unemployment benefits, or health care (Vandenbroucke and Vleminckx, 2013), which are directed at compensating beneficiaries ex post for socio-economic and labour market losses. Many of the traditional social transfer and insurance policies entail a significant degree of redistribution and are therefore associated with a high degree of income-related conflict on the level of welfare state attitudes (Cusack, Iversen and Rehm, 2006; Busemeyer, Goerres and Weschle, 2009; Rehm, 2009). In contrast, social investments tend to provide relatively more benefits to the affluent middle- and upper-income classes (Van Lancker 2013). Consequently, social investment policies could be widely popular and less contested across the class divide. Existing scholarship using available survey data confirms that large majorities in advanced industrial democracies support higher public spending on education (Ansell, 2010; Busemeyer, 2012; Garritzmann, 2015) and government responsibility for the provision of childcare (Meuleman and Chung, 2012; Bonoli 2013; Busemeyer and Neimanns, 2017). This article takes this literature as point of departure. We probe the robustness of public support for social investment relative to social transfer policies. The motivation for this is the supposition that support for social investments might be vulnerable to the dominance of short-term over long-term concerns. While social investments might have significant benefits in the future, expanding these policies creates short-term costs, particularly if other parts of the welfare state have to be cut back. In an age of ‘permanent austerity’ (Pierson, 2001) policymakers are regularly confronted with policy trade-offs requiring cutbacks in some part of the welfare state to expand other parts such as the social investment pillar. A significant shortcoming of existing surveys on education and other social policies, however, is that the question wording used in these surveys does not force respondents to choose and prioritize between different policy areas. This might contribute to a severe overestimation of the extensiveness of support for social investment policies. To address this shortcoming, this article exploits a novel data set from a representative survey of public opinion in eight Western European countries (Busemeyer et al., 2018). An innovative component in this survey is that it explicitly takes into account policy trade-offs when measuring the extent of public support for social investment policies. Confronting respondents with different possibilities to prioritize among various social policies in a quasi-experimental setting, our analysis reveals that public support for social investment policies drops precipitously once trade-offs are taken into account. Furthermore, we find that material self-interest becomes a stronger determinant of policy preferences when moving from unconstrained to constrained settings. Finally, we also explore interaction effects between self-interest and ideology, showing that material self-interest related to spending trade-offs is a more important determinant of attitudes for those identifying with a right-wing ideology compared to those sympathizing with the left. As we discuss in the conclusion, this is an important and challenging finding for the future of the social investment state. Literature Review and Research Gap The analysis of welfare state attitudes has become a growing field in comparative welfare state research (Kumlin and Stadelmann-Steffen (2014) as well as Svallfors (2012) provide overviews). In general, earlier studies have focused more on identifying the micro-level determinants of attitudes and policy preferences (see, for instance, Hasenfeld and Rafferty, 1989), while more recent work shifts attention towards exploring the feedback effects of macro-level policies and institutions (Jaeger, 2009; Jakobsen, 2010; Jordan, 2013). In the present article, we are mainly concerned with studying the determinants of attitudes on the micro-level, however. Simplifying greatly, the pertinent literature on the micro-level determinants of attitudes can be divided into two categories: First, supporting the rational choice paradigm, materialistic self-interest has consistently been shown to shape preferences: people support those welfare policies from which they or their close relatives (expect to) benefit. Those not benefitting and those paying for these benefits, in turn, tend to be opposed. The literature discusses several indicators of self-interest, for example individuals’ income, educational background, labour market status and skill sets, age, welfare state beneficiary status, and labour market risk (Iversen and Soskice, 2001; Cusack, Iversen and Rehm, 2006; Busemeyer, Goerres and Weschle, 2009; Rehm, 2009; Rehm, Hacker and Schlesinger, 2012; Häusermann, Schwander and Kurer, 2015). More recently, panel data have been used to demonstrate the causal effect of personal experiences of material deprivation on social policy preferences (Jaeger 2006; Brooks and Manza, 2013; Margalit, 2013; Owens and Pedulla, 2014; Naumann, Buss and Bähr, 2015), but results remain inconclusive. Besides material self-interest, secondly, values, norms, and ideological predispositions have been found to influence welfare state attitudes (Kangas, 1997; Lupu and Pontusson, 2011). For example, Lupu and Pontusson (2011) argue that other-oriented attitudes matter for redistribution preferences, i.e. people are not only self-oriented but also altruistic. Moreover, social policy preferences are also driven by non-materialistic factors such as religious beliefs (De la O and Rodden, 2008) and by respondents’ perceptions of ‘deservingness’, i.e. how worthy of support they perceive particular beneficiaries to be (Van Oorschot, 2006). Finally, and maybe most importantly, this literature consistently reports that people’s ideological position is strongly related to their social policy preferences. A newer literature adds the claim that materialistic self-interest can be moderated by ideological positions (Margalit, 2013), an important argument that we also seek to test further. Despite these efforts, a major research gap in the existing literature on welfare state attitudes is that there is hardly any work on the issue of trade-offs between different welfare policies, particularly regarding ‘old’ compensatory vs. ‘new’ social investment policies. A partial exception is Boeri, Börsch-Supan, and Tabellini (2001), who confronted respondents in four countries with policy trade-offs, but only for the case of pension policies. In an earlier study, Hansen (1998) analyses public opinion data for the case of the United States. Both studies provide evidence for the general point, prominently discussed by Pierson (2001) as well as by Brooks and Manza (2006), that citizens dislike cutbacks in welfare state spending and express strong support for the maintenance of the current status quo. Kuklinski et al. (2001) analyse policy trade-offs in healthcare for the case of the United States, but their study deals with the role of information provision and motivation for trade-off choices, not with actual fiscal constraints of policymaking. This gap in research, which is ultimately related to missing survey data on the notion of trade-offs, is significant because studying trade-offs on the level of individual policy preferences has very relevant implications for policymaking and assessments about the political viability and legitimacy of the social investment model of the welfare state. So far, analyses of the transformation of European welfare states towards the social investment model have mostly been confined to the macro-level of policymaking (Morel, Palier and Palme, 2012; Bonoli, 2013; Hemerijck, 2013, but see Häusermann, Schwander, and Kurer, 2015), but it remains unclear to what extent this gradual shift reflects prevailing public attitudes. It might well be the case that public demand for social investment policies—responding to ‘new social risks’ such as single parenthood or skills obsolescence (Esping-Andersen, 2002; Bonoli, 2007)—is robust and significant, and citizens are in fact willing to accept cutbacks in the more traditional pillars of the welfare state. Vice versa, recent policy changes might not be reflected on the level of attitudes at all, which could indicate a certain lack of responsiveness on the part of policymakers, employing social investment reforms to ‘sugarcoat’ unpopular retrenchment. The focus of the present article is to address this gap by adding new empirical information on public opinion about trade-offs for a representative sample of European countries. Our aim is to shed light on political preferences towards policy trade-offs at the individual level, studying particularly the interplay of material self-interest and ideological orientations. Attitudes towards Social Policy Trade-Offs and Their Determinants Before presenting a set of testable hypotheses, we briefly clarify our understanding of a ‘trade-off’ on the level of individual preferences, which centres on the idea that citizens are forced to choose between two different competing policies: expanding one of these policies necessarily implies cutting back the other. Of course, this is a particular and therefore imperfect conception of a policy trade-off because, instead of trading one policy for another, citizens could be willing to accept more taxes or higher levels of public debt to finance additional spending. These are possibilities we explore in different, but related work, using another set of questions from the same survey (Busemeyer and Garritzmann, 2017), showing that citizens in fact prefer to accept higher taxes or higher levels of public debt rather than cutting back spending in other parts of the welfare state. Increasing taxes or debt levels may, however, not always be feasible in practice because of fiscal constraints associated with Europeanization and economic globalization (Genschel, 2004) as well as because of a political discourse dominated by concerns about austerity and debt reduction (Blyth, 2013; Haffert and Mehrtens, 2015). In the present article, we therefore deliberately focus on trade-offs between different policy areas to better understand the dynamics underlying redistribution within the welfare state, which may be different from conflicts about its general size. To measure citizens’ attitudes towards trade-offs, we rely on original public opinion survey data in eight European countries (details below) that forces respondents to choose between the expansion of social investment policies (education and family policies) to the detriment of social insurance and transfer policies (unemployment benefits and pensions). It is a legitimate question to ask to what extent our set-up and the implied notion of a ‘trade-off’ adequately reflects decision-making in the real world. Confronting survey respondents with a fixed set of choices always entails the danger of pre-maturely identifying patterns in public opinion, whereas ‘non-attitudes’ (Berinsky, 2017) might dominate in the real world. Previous research found that citizens might hold conflicting preferences such as a preference for a more generous welfare state and lower taxes (Citrin, 1979; Giger and Nelson, 2013). In spite of these limitations, we believe that our set-up adequately mirrors real-world challenges in the transformation of existing welfare states towards the social investment model. By comparing public support for social investment policies in constrained (i.e. with policy trade-offs) and unconstrained settings (i.e. without trade-offs), our analysis provides robust estimates of the upper and lower bounds for the extensiveness of support. While fiscal constraints in the real world of policymaking are rarely as binding as in our constructed set of limited choices, they are never fully absent either. We acknowledge that our construction of policy trade-offs does not represent the ‘full menu’ of potential policy alternatives. In fact, it is logically impossible to try to model all potential policy options, even when using tools that get closer to capturing multidimensional preferences, like conjoint analyses/factorial surveys, as the number of potential policy options and their combination is infinite. Consequently, we consider our analysis as a complement to existing studies rather than a full-scale replacement. Of course, it might be argued that solving trade-offs between conflicting public demands is the task of policymakers, not citizens. But we believe that studying citizens’ reactions to trade-off scenarios is a worthy and important enterprise because a permanent mismatch between policymakers’ actions and citizens’ preferences could contribute to a decline in the legitimacy of democratic decision-making and declining levels of political trust. Furthermore, the limited evidence from existing studies on fiscal trade-offs (Hansen, 1998; Kuklinski, et al. 2001; Busemeyer and Garritzmann, 2017) shows that citizens are well able to make informed and rational choices in these situations, as trade-offs ‘cause respondents to stop and think’ (Hansen, 1998: 519). In the following, we develop a set of testable hypotheses, which guide the subsequent empirical analysis. Perceptions of Trade-Offs Our first expectation is that citizens—just like politicians—dislike being confronted with trade-offs. Nobody likes making hard choices, particularly if they imply potential negative costs for oneself. Numerous theories in social psychology back this notion, showing that individuals when faced with a difficult decision become more status quo oriented and more resistant to change (Anderson, 2003). Assuming that a considerable share of individuals might prefer both social investment and compensatory social policies (Roosma, Gelissen and Van Oorschot, 2013), individuals should reject policy proposals that conflict with their ideal preferences. This implies that public support for the expansion of a particular social policy should drop significantly when moving from an unconstrained setting with no trade-off to a setting where spending increases would require cutbacks in other welfare areas (Hypothesis 1). Furthermore, we expect variation across policy areas, i.e. different kinds of trade-offs. For example, we expect different dynamics in an ‘education vs. pension’ trade-off than in an ‘education vs. unemployment benefits’ trade-off. This variation might be related to variation in perceived ‘deservingness’ (Van Oorschot, 2006) of different welfare beneficiary groups: the unemployed can be expected to be perceived as less deserving than retired persons or families with children for two reasons: Firstly, they are believed to be able to work themselves out of their predicament (Van Oorschot’s ‘control criterion’), whereas retired persons and young families cannot do the same because of their dependency on the welfare state. Secondly, pensioners are perceived to be more deserving as they have contributed more themselves (‘reciprocity criterion’) and are perceived as being ‘closer’ to the average respondent (‘identity criterion’). Besides the normative implications, the ‘deservingness’ argument also has a materialist, rational component, as individuals are likely to be more supportive of welfare state services and transfers, which they expect to need themselves in the future. Everybody turns old, but the risk of unemployment is more concentrated on certain groups (Busemeyer, Goerres and Weschle, 2009). Furthermore, the duration of personal experiences with the benefits and services of the welfare state differs across policies, as individuals (in Western Europe) have much longer personal experiences with the education system, whereas family policies targeted at families with small children cover a much shorter time span. Hence, cutbacks in pensions are probably less popular than cutbacks in unemployment spending, and spending increases on education can be expected to be more popular than spending increases on family policies for parents with small children (Hypothesis 2). Material Self-Interest Besides variation across policy areas, we expect variation across individuals in citizens’ perceptions of and reactions to trade-offs. We hypothesize that this micro-level variation in attitudes is related to material self-interest and that the explanatory power of indicators of self-interest will increase once we move from an unconstrained to a constrained (trade-off) scenario (Hypothesis 3). The simple reason is that citizens have become used to (and might even be dependent on) the provision of a large set of welfare services and benefits, which they would not want to give up even if other parts of the welfare state were expanded in exchange. Thus, we focus on beneficiary status as an indicator of material self-interest here, but include additional variables related to material self-interest such as household income and age as control variables in the analysis. This implies, for instance, that pupils and students support increases in education spending and are willing to cut compensatory spending on unemployment and pensions to achieve this. Of course, students might well become unemployed in the future, and they will surely turn old at some point, but the immediate short-term benefits of increasing spending on education might outweigh the rather diffuse costs of cutting transfers for the unemployed and pensioners. The same logic applies to parents of young children, whom we expect to support increases in public spending on families and education. In contrast, the unemployed should oppose more spending on social investments when this implies cutting back unemployment benefits. Finally, pensioners should oppose social investment expansion when this necessitates cutbacks in old-age pensions. As hinted at above, we also posit that the explanatory power in terms of statistical significance and effect size for indicators of material self-interest increases when moving from an unconstrained to a constrained (trade-off) scenario. In the constrained scenario, citizens supporting the expansion of a particular policy are also forced to accept cutbacks in other social policies. This reduces the size of the group supporting expansion (see Hypothesis 1 above), but it also changes the composition of this group. It is plausible to assume that those supporting the expansion of a particular policy in spite of cutbacks will be those with the strongest self-interest in that policy. In contrast, those who stop supporting the policy will be respondents whose support might have been more an expression of general support for the welfare state before rather than the particular policy in question (Roosma, Gelissen and Van Oorschot, 2013). Ideological Predispositions In addition to material self-interest, we expect that ideological predispositions matter. An often-discussed problem in the study of the effects of norms and values on attitudes is endogeneity: when assessing the impact of ideology on welfare support it is difficult to disentangle the direction of causality, in particular in cross-sectional surveys. Hence, statistical associations should be interpreted as correlational relations rather than causal statements. Since the dependent variable concerns trade-offs, the theoretical expectations related to ideological predispositions are not straightforward. For instance, people supporting social spending in general are likely to favour compensatory and social investment spending. While this relationship might seem rather trivial, it is difficult to deduce what this would imply for people’s preferences on policy trade-offs, as both supporters and opponents of the welfare state might oppose expanding one spending area at the expense of another, but for different reasons: individuals identifying with a left-wing ideology could object to being forced to choose between different types of social spending, which—in an unconstrained setting—they would both support. Vice versa, respondents subscribing to a right-wing ideology could be opposed to spending increases in general, even if they would partly be financed with cutbacks in other parts of the welfare state. However, as social investment policies tend to be less redistributive compared to many social transfers and therefore tend to benefit the (upper) middle class (Vandenbroucke and Vleminckx, 2013), right-wing supporters could be less opposed to recalibrating the welfare state from classical redistributive policies towards a more investment-oriented and less redistributive model. We further expect the influence of ideology to vary across the different policy trade-offs because each policy area is associated with distinct redistributive consequences. The association between ideology and preferences towards the different policy proposals should be particularly pronounced for policy proposals, which would most strongly reduce the amount of welfare state redistribution. This applies in particular to an expansion of relatively little redistributive education spending at the expense of compensatory spending, and to cuts in highly redistributive spending on unemployment benefits to increase spending on social investment. In contrast, the effect of ideology should cancel out to some extent for policy proposals leading to only minor changes in welfare state redistribution. This includes an expansion of financial support and public services for families with a strong poverty reducing effect (cf. Maldonado and Nieuwenhuis, 2015), which would be financed with cuts in compensatory spending. The same holds for cuts in spending on pensions as a less redistributive type of compensatory spending (Busemeyer, Goerres and Weschle, 2009) to finance additional social investment spending.1 Taken together, the more a recalibration of the welfare state from a compensatory to a social investment model would reduce overall welfare state redistribution, the more individuals with a left-wing ideology should oppose and those with a right-wing ideology support such spending shifts (Hypothesis 4). Furthermore, adapting Margalit’s (2013) argument we hypothesize that ideological predispositions moderate the association between material self-interest and support for social investment in the trade-off scenarios. We expect that when confronted with a policy trade-off, materialistic self-interest plays a less important role for left-wing voters because they are more solidaristic with all disadvantaged social risk groups and therefore oppose sacrificing benefits for one group to expand benefits for another. Thus, we expect that for left-leaning respondents the effect of their beneficiary status on their preferences is attenuated. Right-leaning citizens, in contrast, might be less opposed to recalibrating benefits (as discussed above) and their self-interest might be less attenuated by their ideological position. In sum, Hypothesis 5 posits an interaction effect between material self-interest and ideology. If this reasoning is true, it implies—at first glance counter-intuitively—that if in times of ‘permanent fiscal austerity’ (Pierson, 2001) the expansion of social investment is only possible at the expense of other spending, the ‘turn’ towards the social investment state is in fact more supported (or at least more willingly accepted) by right-leaning individuals. For them, the associated ideological costs of ending up in a potentially less redistributive welfare state model are lower. Research Design The research gap related to individual-level preferences towards trade-offs identified above is at least partly due to the lack of data on these issues as well as on social investment more generally in the existing comparative surveys like the International Social Survey Programme (ISSP), the European Social Survey (ESS), or the Eurobarometer (for details on this discussion and an overview of our survey, see Busemeyer et al., 2018). Thus, we conducted an original survey in eight European countries: Denmark, France, Germany, Ireland, Italy, Spain, Sweden, and the United Kingdom. Unfortunately, due to funding constraints, no Eastern or non-European countries could be included. Hence, the analysis is restricted to Western Europe but covers a wide variety of countries from all ‘worlds of welfare’ in Western Europe here, thereby enhancing representativeness and external validity. In each country, we surveyed a representative sample of 1,000–1,500 adult individuals (aged 18–99 years), overall 8,905 individuals. The survey was conducted by a professional survey-institute via computer-assisted telephone interviewing (CATI) and fulfils all standards of a high-quality survey. Following pre-tests in February 2014, the main fieldwork took place mid-April to early-June 2014.2 Dependent Variables To test our theoretical expectations, we randomly split the sample into four groups in each country (Table C shows that the randomization was successful as the groups are statistically indistinguishable on important covariates). The split in four, equally sized groups and random assignment to these groups are necessary to avoid statistical artefacts (and therefore wrong conclusions): if we had asked everybody these four questions, respondents might tend to reply in a similar fashion to these questions independent of their real preferences and/or they might show effects of survey fatigue or learning, leading to potential biases. Respondents thus see one, and only one, of the following statements (we add emphasis here to highlight the differences for the readers): Q1: ‘Imagine the [COUNTRY] government plans to increase spending on education by 10 per cent and wants to finance this by cutting the benefits for the unemployed’. Q2: ‘Imagine the [COUNTRY] government plans to increase spending on education by 10 per cent and wants to finance this by cutting old age pensions’. Q3: ‘Imagine the government plans to enact reforms involving a 10 per cent increase in the budget for financial support and public services for families with young children; and wants to finance this by cutting the benefits for the unemployed’. Q4: ‘Imagine the government plans to enact reforms involving a 10 per cent increase in the budget for financial support and public services for families with young children; and wants to finance this by cutting old age pensions’. The statements confront respondents with four policy trade-offs: education and family support, on one hand, and unemployment benefits and pensions, on the other. We chose education and financial support and public services for families with young children as prime examples of social investment policies, as they seek to ‘create, mobilize, or preserve skills’ (Garritzmann et al., 2017: 37) of children or their parents. Benefits for the unemployed and old-age pensions, in contrast, are typical examples of compensatory social policies. We can therefore analyse whether and why respondents prefer social investment or compensatory policies as well as differentiate further between two types of ‘old’ and ‘new’ welfare policies. We specified a ‘10% increase’ in the question framing for several reasons. First, this specification makes the policy options more concrete and tangible for respondents. Secondly, it increases the likelihood that respondents think in roughly similar terms about the proposed policy change. Speaking only about spending increases without providing a measuring yardstick, some respondents might think of a 1 per cent increase, while others imagine a 20 per cent change. Of course, the ‘10 percent’ figure is arbitrary to some extent and furthermore implies substantially different amounts of money across policy fields (e.g. pension expenditure is significantly higher than education spending) and countries. However, for our purpose it is not advisable to include different actual spending figures in the questions, since this increases the survey’s administrative complexity and compromises its comparability. Hence, we have to assume that respondents take the 10 per cent figure as a rough proxy signal for a non-trivial but realistic policy change. We used the answers to the four split sample questions to create four dependent variables: those supporting education spending (combining Q1 and Q2), those supporting spending on families with young children (Q3 and Q4), those accepting cutbacks in unemployment spending (Q1 and Q3), and, finally, those accepting cutbacks in pension spending (Q2 and Q4). By partially pooling the split-sample questions we can highlight the specific characteristics of each of the four policy fields regarding cuts and expansions and get an increased sample size for each single model leading to more robust estimates of the coefficients. All models include dummies for sub-group assignment. For each question, we offered five answer categories: ‘strongly agree’, ‘agree’, ‘neither agree nor disagree’, ‘disagree’, and ‘strongly disagree’. Moreover, two residual categories were included: ‘Don’t know’ and ‘No answer’. Thus, we are measuring spending preferences relative to the status quo, not in absolute terms (cf. Kumlin, 2014). A disadvantage of this commonly used five-point Likert scale is that respondents might choose the category ‘neither agree nor disagree’ for different reasons (Goerres and Prinzen, 2012). Admittedly, in our case, it is difficult to pin down exactly why respondents would remain undecided (‘neither nor’) towards a reform proposal: they might indeed reject the proposed spending shift between policy areas or they might simply dislike being forced to choose between different types of spending and reject the implied notion of welfare state reform as a zero-sum game. For this reason, we focus on those respondents who state a clear preference towards agreement of the proposed trade-off scenario against all other respondents. A reassuring indication that we are measuring meaningful responses rather than non-attitudes is that there are only few respondents in the residual categories of ‘dońt know’ and ‘no answer’, which we coded as missings (see Table B for summary statistics). Accordingly, when using regressions to probe individual-level determinants, we dichotomize the five categories into two groups: respondents, who (strongly) agree with the statement (coded ‘1’), and respondents, who (strongly) disagree or are undecided (coded ‘0’).3 That said, we have re-run the analysis using ordered logistic regressions, which produces similar results (see Table H). To test Hypothesis 1, which states that public support for social investment drops considerably once trade-offs are acknowledged, we add two additional dependent variables based on questions from our survey measuring preferences towards spending on ‘education’ and ‘financial support for families’ in an unconstrained scenario (i.e. no trade-offs implied). More specifically, the questions read: Q5 and Q6: ‘In the following, I will name several areas of government activity. Please tell me whether you would like to see more or less government spending in each area. Keep in mind that ‘more’ or ‘much more’ might require a tax increase’.—‘Education (Q5)’—‘Financial support for families (Q6)’. It could be objected that this unconstrained scenario is not ‘truly unconstrained’ because respondents are reminded of the possibility that a spending increase ‘might require a tax increase’. Yet, as we show elsewhere (Busemeyer and Garritzmann, 2017), the reminder about taxes does not have any measurable effect on people’s preference compared to a fully unconditional scenario. We can show this by comparing this question to a ‘truly’ unconditional question in our survey, where we do not include the reminder that more spending might increase taxes, and the support levels between the two questions are indistinguishable. We cannot use the latter question for the present analysis, however, because it is part of another survey experiment and therefore only covers part of the sample. We use the weak reminder about tax increases in the primary question to make it similar to the commonly used spending questions from the ISSP Role of Government module. All in all, we therefore feel confident about treating the question formulation above (Q5 and Q6) as a ‘de facto unconditional’ baseline category. Again, we dichotomize responses and distinguish preferences for ‘more’ or ‘much more’ spending (1) and preferences for the ‘same’, ‘less’, or ‘much less’ (0) spending. We match the split-sample questions from above to the respective items, i.e. unconstrained education spending preferences (Q5) for respondents given Q1 or Q2 and unconstrained family spending preferences (Q6) for those given Q3 or Q4. Of course, these items do not constitute a control group in a strict experimental sense, but nevertheless allow us to compare the difference between constrained and unconstrained settings when measuring spending preferences. Because of the binary nature of the dependent variables, we estimate multivariate single-level logit regressions. We include country fixed-effects and compute country-clustered robust standard errors to reflect the fact that the sample is stratified on countries. In the present article, we focus on associations on the micro-level of attitudes in the pooled sample, while the variation of effects across countries and welfare state regimes will be addressed in future work. We include sampling weights in the descriptive analysis but not in the regression models because weighting can bias the standard errors and is unnecessary when the control variables account for differences in sampling probability (Schmidt-Catran, 2016). Our central explanatory variables related to self-interest capture respondents’ beneficiary status (having small kids at home, being a student, unemployed, or retired). Small kids are here those below the age of 10 years. Because the control variable for kids above the age of 10 years also includes potentially adult children still living in the parental household, we do not have strong expectations in this case. In a further step of the analysis we add self-placement on a left-right scale as a measure for ideological predispositions. In addition, we control for household income, educational attainment, gender, age, perceived risk of unemployment, and household size. In the Online Appendix, we provide additional information on the exact measurement and operationalization of the variables used in the analysis (Table A). Because age correlates heavily with retirement status (r = 0.72), we consider age only for those in full-time employment, with the reference group being those in their 30s. Results First, we examine how spending preferences change once trade-offs are introduced. Figure 1 shows the share of respondents agreeing with spending increases on education (left) and families (right panel). The first important finding is that support for additional spending on social investment policies drops considerably if this would require cutbacks in compensatory spending, providing strong support for Hypothesis 1 and confirming Busemeyer and Garritzmann (2017) for a different set of policy trade-offs. In an unconstrained setting, we see impressive levels of support for spending on social investment: around 77 per cent of respondents support additional spending on education, and 55 per cent are supportive of additional financial support for families. As soon as spending trade-offs are introduced, however, support declines strongly. Across the four spending trade-offs, only 17 per cent of respondents would agree to a reform that increases social investments at the expense of compensatory policies. A majority disagrees with each of the proposed policy changes. This is not totally surprising, but the new contribution of our article is to show that an often-held, empirically largely untested assumption in the literature on the ‘new politics of the welfare state’ (Pierson, 2001) in fact has a solid empirical foundation. There is, of course, some cross-country variation in this respect (see Figure A), but by and large, the general pattern depicted in Figure 1 also holds across countries. Figure 1. View largeDownload slide Shares of respondents supporting increases in social investment spending—unconstrained spending preferences and when implying cuts in compensatory spending Figure 1. View largeDownload slide Shares of respondents supporting increases in social investment spending—unconstrained spending preferences and when implying cuts in compensatory spending Yet, a closer look at Figure 1 reveals that there are considerable differences across the four spending trade-offs, confirming Hypothesis 2: Respondents are much more willing to accept cutbacks in unemployment benefits than in pensions. Every fourth respondent is willing to cut unemployment benefits to increase education spending, but this share drops to 10 per cent when pensions would be reduced. In other words, the difference between these two compensatory policies is 17 percentage points when the trade-off is about education and 6 percentage points when family spending is concerned. This is strong support for our expectation that people are more willing to cut unemployment benefit generosity than pensions, either because of (future-oriented) self-interest or because of differences in perceived deservingness, as discussed in Hypothesis 2. The differences are less pronounced between investments on families with young children compared to educational investment. This might be related to the fact that both policies address groups that are deemed to be deserving of welfare state support or because the redistributive patterns are more similar for education and family policies than for unemployment benefits and pensions. In the next step, we examine to what extent this drop in spending support in the case of trade-offs is related to material self-interest based on individual beneficiary status. Hypothesis 3 postulated that the association between self-interest and support for spending on social investment policies should become stronger once we move from an unconstrained to a constrained (trade-off) scenario. Our logistic regression results in Table 1 provide considerable support for this claim: Models 1 and 2 reveal that in the unconstrained scenarios neither having small kids nor being a student is significantly associated with preferences for more spending. This picture changes when moving to the models for spending trade-offs: respondents with small children are now significantly more likely to support increases in social investment spending (Models 3 and 4), students support increasing education spending (Model 3), and the unemployed (Model 5) and the retired (Model 6) oppose cuts in ‘their’ benefit schemes for the sake of increasing social investment spending. These effects are substantial in size, as indicated by average marginal effect estimates, which allow a comparison of effect sizes across model specifications (see Table D): the likelihood of supporting additional spending increases by 2.7 percentage points for those with small children (Models 3 and 4) and by 6 percentage points for students (Model 3). Spending support drops by 9.3 percentage points for the unemployed in Model 5 and by 4.5 percentage points for the retired in Model 6. In sum, once spending increases for social investment would be financed by cuts in compensatory spending, considerable differences between different welfare state beneficiary groups come to the fore and self-interest becomes a much stronger predictor of preferences, confirming Hypothesis 3. Table 1. Logistic regressions: support for social investment—unconstrained spending preferences and when implying cuts in compensatory spending; maximum likelihood estimates Variables M1 M2 M3 M4 M5 M6 Increase education spending Increase family spending (financial support) Increase education spending, cut compensatory spending Increase family spending, cut compensatory spending Cut unemployment benefits, increase social investment spending Cut pensions, increase social investment spending Unconstrained spending preferences Spending trade-offs between social investment and compensatory spending No post-secondary education (Ref: Higher education) −0.328*** 0.279*** −0.192 0.171 0.075 −0.185** (0.089) (0.045) (0.168) (0.105) (0.166) (0.077) Vocational education −0.084 0.081 −0.119 −0.058 −0.086 −0.158 (0.143) (0.107) (0.164) (0.101) (0.142) (0.132) Household income  (Q1) (Ref: Middle quintile (Q3)) −0.030 0.171 −0.001 −0.347** −0.126 −0.179 (0.118) (0.135) (0.134) (0.166) (0.228) (0.219)  Q2 −0.103 −0.040 0.124 −0.028 0.251** −0.276* (0.142) (0.118) (0.099) (0.126) (0.119) (0.142)  Q4 −0.056 −0.391*** 0.084 −0.037 0.154 −0.184 (0.205) (0.143) (0.183) (0.153) (0.100) (0.202)  Q5 −0.026 −0.500*** 0.359** 0.255*** 0.370*** 0.258 (0.179) (0.087) (0.149) (0.085) (0.095) (0.172) Female −0.041 −0.024 −0.293*** −0.348*** −0.230*** −0.518*** (0.086) (0.077) (0.099) (0.088) (0.089) (0.071) Small child (<10 years) (Ref: No child in HH) 0.103 0.145 0.228** 0.236* 0.201* 0.304* (0.118) (0.150) (0.100) (0.135) (0.118) (0.156)  Older child (≥10 years) 0.358*** 0.092 0.030 −0.128 −0.049 −0.029 (0.107) (0.060) (0.114) (0.168) (0.105) (0.211) Age <30, full-time employed (FTE) (years)  (Ref: Age 30–39, FTE) −0.048 0.386* 0.347 0.510* 0.567*** 0.298 (0.199) (0.221) (0.217) (0.298) (0.173) (0.539)  40–49, FTE −0.027 −0.136 −0.358*** −0.348*** −0.454*** −0.155 (0.159) (0.201) (0.138) (0.077) (0.154) (0.129)  50–59, FTE −0.011 −0.272 −0.966*** −0.546** −0.830*** −0.595*** (0.193) (0.234) (0.109) (0.261) (0.177) (0.209)  >60, FTE −0.179** −0.084 −0.922*** −1.051** −0.904*** −1.248*** (0.079) (0.165) (0.239) (0.473) (0.286) (0.312)  Unemployed 0.412 0.225 −0.577 −0.477* −0.624* −0.318 (0.302) (0.360) (0.358) (0.283) (0.344) (0.434)  Studying 0.693 −0.144 0.377*** 0.551 0.457** 0.544 (0.474) (0.290) (0.133) (0.369) (0.227) (0.378)  Retired 0.122 −0.336 −0.815*** −0.600*** −0.768*** −0.594*** (0.130) (0.225) (0.135) (0.184) (0.141) (0.221)  Other 0.133* −0.063 −0.435*** −0.275*** −0.497** −0.108 (0.080) (0.222) (0.105) (0.096) (0.212) (0.293) Risk of unemployment 0.084 −0.122 −0.362 0.028 −0.262 0.002 (0.172) (0.143) (0.265) (0.095) (0.273) (0.132) Cohabit (Ref: one adult household) −0.022 0.092 0.055 −0.251*** −0.002 −0.254** (0.071) (0.079) (0.138) (0.088) (0.102) (0.108) Country dummies Yes Yes Yes Yes Yes Yes Trade-off dummy Yes Yes Yes Yes Pseudo R2 0.074 0.118 0.094 0.070 0.075 0.043 Observations 3,931 3,849 3,892 3,870 3,929 3,833 Variables M1 M2 M3 M4 M5 M6 Increase education spending Increase family spending (financial support) Increase education spending, cut compensatory spending Increase family spending, cut compensatory spending Cut unemployment benefits, increase social investment spending Cut pensions, increase social investment spending Unconstrained spending preferences Spending trade-offs between social investment and compensatory spending No post-secondary education (Ref: Higher education) −0.328*** 0.279*** −0.192 0.171 0.075 −0.185** (0.089) (0.045) (0.168) (0.105) (0.166) (0.077) Vocational education −0.084 0.081 −0.119 −0.058 −0.086 −0.158 (0.143) (0.107) (0.164) (0.101) (0.142) (0.132) Household income  (Q1) (Ref: Middle quintile (Q3)) −0.030 0.171 −0.001 −0.347** −0.126 −0.179 (0.118) (0.135) (0.134) (0.166) (0.228) (0.219)  Q2 −0.103 −0.040 0.124 −0.028 0.251** −0.276* (0.142) (0.118) (0.099) (0.126) (0.119) (0.142)  Q4 −0.056 −0.391*** 0.084 −0.037 0.154 −0.184 (0.205) (0.143) (0.183) (0.153) (0.100) (0.202)  Q5 −0.026 −0.500*** 0.359** 0.255*** 0.370*** 0.258 (0.179) (0.087) (0.149) (0.085) (0.095) (0.172) Female −0.041 −0.024 −0.293*** −0.348*** −0.230*** −0.518*** (0.086) (0.077) (0.099) (0.088) (0.089) (0.071) Small child (<10 years) (Ref: No child in HH) 0.103 0.145 0.228** 0.236* 0.201* 0.304* (0.118) (0.150) (0.100) (0.135) (0.118) (0.156)  Older child (≥10 years) 0.358*** 0.092 0.030 −0.128 −0.049 −0.029 (0.107) (0.060) (0.114) (0.168) (0.105) (0.211) Age <30, full-time employed (FTE) (years)  (Ref: Age 30–39, FTE) −0.048 0.386* 0.347 0.510* 0.567*** 0.298 (0.199) (0.221) (0.217) (0.298) (0.173) (0.539)  40–49, FTE −0.027 −0.136 −0.358*** −0.348*** −0.454*** −0.155 (0.159) (0.201) (0.138) (0.077) (0.154) (0.129)  50–59, FTE −0.011 −0.272 −0.966*** −0.546** −0.830*** −0.595*** (0.193) (0.234) (0.109) (0.261) (0.177) (0.209)  >60, FTE −0.179** −0.084 −0.922*** −1.051** −0.904*** −1.248*** (0.079) (0.165) (0.239) (0.473) (0.286) (0.312)  Unemployed 0.412 0.225 −0.577 −0.477* −0.624* −0.318 (0.302) (0.360) (0.358) (0.283) (0.344) (0.434)  Studying 0.693 −0.144 0.377*** 0.551 0.457** 0.544 (0.474) (0.290) (0.133) (0.369) (0.227) (0.378)  Retired 0.122 −0.336 −0.815*** −0.600*** −0.768*** −0.594*** (0.130) (0.225) (0.135) (0.184) (0.141) (0.221)  Other 0.133* −0.063 −0.435*** −0.275*** −0.497** −0.108 (0.080) (0.222) (0.105) (0.096) (0.212) (0.293) Risk of unemployment 0.084 −0.122 −0.362 0.028 −0.262 0.002 (0.172) (0.143) (0.265) (0.095) (0.273) (0.132) Cohabit (Ref: one adult household) −0.022 0.092 0.055 −0.251*** −0.002 −0.254** (0.071) (0.079) (0.138) (0.088) (0.102) (0.108) Country dummies Yes Yes Yes Yes Yes Yes Trade-off dummy Yes Yes Yes Yes Pseudo R2 0.074 0.118 0.094 0.070 0.075 0.043 Observations 3,931 3,849 3,892 3,870 3,929 3,833 Note: Country-clustered robust standard errors in parentheses. *** P < 0.01, **P < 0.05, *P < 0.1. Table 1. Logistic regressions: support for social investment—unconstrained spending preferences and when implying cuts in compensatory spending; maximum likelihood estimates Variables M1 M2 M3 M4 M5 M6 Increase education spending Increase family spending (financial support) Increase education spending, cut compensatory spending Increase family spending, cut compensatory spending Cut unemployment benefits, increase social investment spending Cut pensions, increase social investment spending Unconstrained spending preferences Spending trade-offs between social investment and compensatory spending No post-secondary education (Ref: Higher education) −0.328*** 0.279*** −0.192 0.171 0.075 −0.185** (0.089) (0.045) (0.168) (0.105) (0.166) (0.077) Vocational education −0.084 0.081 −0.119 −0.058 −0.086 −0.158 (0.143) (0.107) (0.164) (0.101) (0.142) (0.132) Household income  (Q1) (Ref: Middle quintile (Q3)) −0.030 0.171 −0.001 −0.347** −0.126 −0.179 (0.118) (0.135) (0.134) (0.166) (0.228) (0.219)  Q2 −0.103 −0.040 0.124 −0.028 0.251** −0.276* (0.142) (0.118) (0.099) (0.126) (0.119) (0.142)  Q4 −0.056 −0.391*** 0.084 −0.037 0.154 −0.184 (0.205) (0.143) (0.183) (0.153) (0.100) (0.202)  Q5 −0.026 −0.500*** 0.359** 0.255*** 0.370*** 0.258 (0.179) (0.087) (0.149) (0.085) (0.095) (0.172) Female −0.041 −0.024 −0.293*** −0.348*** −0.230*** −0.518*** (0.086) (0.077) (0.099) (0.088) (0.089) (0.071) Small child (<10 years) (Ref: No child in HH) 0.103 0.145 0.228** 0.236* 0.201* 0.304* (0.118) (0.150) (0.100) (0.135) (0.118) (0.156)  Older child (≥10 years) 0.358*** 0.092 0.030 −0.128 −0.049 −0.029 (0.107) (0.060) (0.114) (0.168) (0.105) (0.211) Age <30, full-time employed (FTE) (years)  (Ref: Age 30–39, FTE) −0.048 0.386* 0.347 0.510* 0.567*** 0.298 (0.199) (0.221) (0.217) (0.298) (0.173) (0.539)  40–49, FTE −0.027 −0.136 −0.358*** −0.348*** −0.454*** −0.155 (0.159) (0.201) (0.138) (0.077) (0.154) (0.129)  50–59, FTE −0.011 −0.272 −0.966*** −0.546** −0.830*** −0.595*** (0.193) (0.234) (0.109) (0.261) (0.177) (0.209)  >60, FTE −0.179** −0.084 −0.922*** −1.051** −0.904*** −1.248*** (0.079) (0.165) (0.239) (0.473) (0.286) (0.312)  Unemployed 0.412 0.225 −0.577 −0.477* −0.624* −0.318 (0.302) (0.360) (0.358) (0.283) (0.344) (0.434)  Studying 0.693 −0.144 0.377*** 0.551 0.457** 0.544 (0.474) (0.290) (0.133) (0.369) (0.227) (0.378)  Retired 0.122 −0.336 −0.815*** −0.600*** −0.768*** −0.594*** (0.130) (0.225) (0.135) (0.184) (0.141) (0.221)  Other 0.133* −0.063 −0.435*** −0.275*** −0.497** −0.108 (0.080) (0.222) (0.105) (0.096) (0.212) (0.293) Risk of unemployment 0.084 −0.122 −0.362 0.028 −0.262 0.002 (0.172) (0.143) (0.265) (0.095) (0.273) (0.132) Cohabit (Ref: one adult household) −0.022 0.092 0.055 −0.251*** −0.002 −0.254** (0.071) (0.079) (0.138) (0.088) (0.102) (0.108) Country dummies Yes Yes Yes Yes Yes Yes Trade-off dummy Yes Yes Yes Yes Pseudo R2 0.074 0.118 0.094 0.070 0.075 0.043 Observations 3,931 3,849 3,892 3,870 3,929 3,833 Variables M1 M2 M3 M4 M5 M6 Increase education spending Increase family spending (financial support) Increase education spending, cut compensatory spending Increase family spending, cut compensatory spending Cut unemployment benefits, increase social investment spending Cut pensions, increase social investment spending Unconstrained spending preferences Spending trade-offs between social investment and compensatory spending No post-secondary education (Ref: Higher education) −0.328*** 0.279*** −0.192 0.171 0.075 −0.185** (0.089) (0.045) (0.168) (0.105) (0.166) (0.077) Vocational education −0.084 0.081 −0.119 −0.058 −0.086 −0.158 (0.143) (0.107) (0.164) (0.101) (0.142) (0.132) Household income  (Q1) (Ref: Middle quintile (Q3)) −0.030 0.171 −0.001 −0.347** −0.126 −0.179 (0.118) (0.135) (0.134) (0.166) (0.228) (0.219)  Q2 −0.103 −0.040 0.124 −0.028 0.251** −0.276* (0.142) (0.118) (0.099) (0.126) (0.119) (0.142)  Q4 −0.056 −0.391*** 0.084 −0.037 0.154 −0.184 (0.205) (0.143) (0.183) (0.153) (0.100) (0.202)  Q5 −0.026 −0.500*** 0.359** 0.255*** 0.370*** 0.258 (0.179) (0.087) (0.149) (0.085) (0.095) (0.172) Female −0.041 −0.024 −0.293*** −0.348*** −0.230*** −0.518*** (0.086) (0.077) (0.099) (0.088) (0.089) (0.071) Small child (<10 years) (Ref: No child in HH) 0.103 0.145 0.228** 0.236* 0.201* 0.304* (0.118) (0.150) (0.100) (0.135) (0.118) (0.156)  Older child (≥10 years) 0.358*** 0.092 0.030 −0.128 −0.049 −0.029 (0.107) (0.060) (0.114) (0.168) (0.105) (0.211) Age <30, full-time employed (FTE) (years)  (Ref: Age 30–39, FTE) −0.048 0.386* 0.347 0.510* 0.567*** 0.298 (0.199) (0.221) (0.217) (0.298) (0.173) (0.539)  40–49, FTE −0.027 −0.136 −0.358*** −0.348*** −0.454*** −0.155 (0.159) (0.201) (0.138) (0.077) (0.154) (0.129)  50–59, FTE −0.011 −0.272 −0.966*** −0.546** −0.830*** −0.595*** (0.193) (0.234) (0.109) (0.261) (0.177) (0.209)  >60, FTE −0.179** −0.084 −0.922*** −1.051** −0.904*** −1.248*** (0.079) (0.165) (0.239) (0.473) (0.286) (0.312)  Unemployed 0.412 0.225 −0.577 −0.477* −0.624* −0.318 (0.302) (0.360) (0.358) (0.283) (0.344) (0.434)  Studying 0.693 −0.144 0.377*** 0.551 0.457** 0.544 (0.474) (0.290) (0.133) (0.369) (0.227) (0.378)  Retired 0.122 −0.336 −0.815*** −0.600*** −0.768*** −0.594*** (0.130) (0.225) (0.135) (0.184) (0.141) (0.221)  Other 0.133* −0.063 −0.435*** −0.275*** −0.497** −0.108 (0.080) (0.222) (0.105) (0.096) (0.212) (0.293) Risk of unemployment 0.084 −0.122 −0.362 0.028 −0.262 0.002 (0.172) (0.143) (0.265) (0.095) (0.273) (0.132) Cohabit (Ref: one adult household) −0.022 0.092 0.055 −0.251*** −0.002 −0.254** (0.071) (0.079) (0.138) (0.088) (0.102) (0.108) Country dummies Yes Yes Yes Yes Yes Yes Trade-off dummy Yes Yes Yes Yes Pseudo R2 0.074 0.118 0.094 0.070 0.075 0.043 Observations 3,931 3,849 3,892 3,870 3,929 3,833 Note: Country-clustered robust standard errors in parentheses. *** P < 0.01, **P < 0.05, *P < 0.1. Of course, some groups of individuals might benefit from both social investment and higher spending on social transfers (e.g. unemployed single parents in need of public childcare). Some might also expect to need certain social transfers in the future, while favouring social investment in the present (e.g. a student expecting high labour market risks once employed). Given the relatively small size of these ‘cross-pressured’ groups, however, we cannot directly measure the preferences of these people with our survey data. In general, the ‘cross-pressure’ effect could contribute to the high levels of opposition to trade-offs. This in fact makes it more likely that we still underestimate the true effects of self-interest. The fact that we find strong and clear effects for the different beneficiary groups indicates that individuals’ current situation is more relevant for their attitudes than their potential future needs. Next, we comment briefly on the control variables. As expected, elderly respondents (those in their 50s and 60s, who are closer to retirement age) oppose expanding social investment spending to the detriment of compensatory spending (Models 3–6). Again, this effect is much more pronounced in the constrained (trade-off) scenarios compared to the unconstrained settings (Models 1 and 2). (Perceived) Unemployment risk is negatively associated with cuts in unemployment spending (Model 5), but this effect is not significant. The significant positive effect of having children above the age of 10 years in Model 1 (but not in Models 3–6) might appear somewhat surprising at first sight. Yet, as we do not have more specific information about the age and status of these children, these children may be grown-ups living with their parents, working or in youth unemployment, which might account for this unexpected finding. Still, this finding needs to be interpreted with caution. Higher-income goes along with stronger support for social investment increases at the expense of compensation. As expected, there are again significant differences between the constrained and unconstrained settings: in the unconstrained settings, higher income is not (in the case of education) or negatively (in the case of family spending, which is likely to be perceived to be more redistributive) related to spending support. In the constrained setting, high-income respondents, in contrast, are more likely to support education and family spending if this would go along with cutbacks in the more redistributive transfer programmes. Higher educational background should be associated with stronger support for education spending increases (Busemeyer, 2015: ch. 5; Garritzmann, 2015). This association tends to be positive in our data as well, but it is not statistically significant. Moreover, women often benefit less than men from social insurance schemes built around the traditional male-breadwinner model and face particular challenges in combining work and family life at the same time (Esping-Andersen, 2002). Hence, there are good reasons to expect that women support higher spending on social investment policies. However, the results show that men are much more likely to accept cutbacks in compensatory spending to increase spending on social investments compared to women. Further research is clearly needed here, but a possible explanation is that women might prefer higher levels of public spending on both policy areas as they are found to be more supportive of the welfare state and redistribution in general (Cusack, Iversen and Rehm, 2006) and might want to achieve the same levels of compensatory safety nets as men while also favouring social investments. This leads us towards examining the role of ideology for spending trade-offs (Hypothesis 4). Table 2 shows the direct association of ideology and trade-off preferences. In additional model steps (Models 5–8) we control for general social spending and redistribution preferences to probe the robustness of this association. Overall, the results in Table 2 reveal a strong direct association between ideology and spending support: left-wing individuals are hostile to spending trade-offs, while those identifying themselves with the right are more supportive. This effect is particularly strong for the two models on expanding education spending and on cutting highly redistributive spending on unemployment (Models 1 and 3). Controlling for general social spending and redistribution preferences (Models 5–8) furthermore shows that these two variables explain a considerable share of the effect of left-right ideology. Individuals with preferences for more social spending reject shifting resources between different areas of welfare spending. Moreover, individuals supporting redistribution are also more likely to oppose spending shifts that would increase education spending or cut back unemployment spending. In contrast, the coefficient of ideology is insignificant for trade-offs that would imply cutbacks in pension spending (Models 4 and 8). This is likely due to the fact that pension spending is less progressive and has a strong appeal for the middle classes. In sum, respondents subscribing to a left-wing ideology are more solidaristic with all social risk groups and are more opposed to expanding benefits for one group at the expense of another, whereas right-leaning citizens are more supportive of shifting spending from more to less redistributive social policies. Table 2. Logistic regressions: effects of ideology on support for social investment when implying cuts in compensatory spending; maximum likelihood estimates Variables Increase education spending, cut compensatory spending Increase family spending, cut compensatory spending Cut unemployment benefits, increase social investment spending Cut pensions, increase social investment spending M1 M2 M3 M4 Right-wing ideology 0.112*** 0.089*** 0.134*** 0.045 (0.026) (0.021) (0.021) (0.029) M5 M6 M7 M8 Right-wing ideology 0.063*** 0.064*** 0.089*** 0.028 (0.021) (0.023) (0.015) (0.030) Social spending preferences −0.395*** −0.346*** −0.442*** −0.250** (0.068) (0.065) (0.101) (0.107) Redistribution preferences −0.094*** 0.005 −0.120*** 0.094 (0.027) (0.026) (0.024) (0.059) Variables Increase education spending, cut compensatory spending Increase family spending, cut compensatory spending Cut unemployment benefits, increase social investment spending Cut pensions, increase social investment spending M1 M2 M3 M4 Right-wing ideology 0.112*** 0.089*** 0.134*** 0.045 (0.026) (0.021) (0.021) (0.029) M5 M6 M7 M8 Right-wing ideology 0.063*** 0.064*** 0.089*** 0.028 (0.021) (0.023) (0.015) (0.030) Social spending preferences −0.395*** −0.346*** −0.442*** −0.250** (0.068) (0.065) (0.101) (0.107) Redistribution preferences −0.094*** 0.005 −0.120*** 0.094 (0.027) (0.026) (0.024) (0.059) Notes: Country-clustered robust standard errors in parentheses; all models include the same independent variables as those models reported in Table 1. See Table E for the full models. *** P < 0.01, **P < 0.05, *P < 0.1. Table 2. Logistic regressions: effects of ideology on support for social investment when implying cuts in compensatory spending; maximum likelihood estimates Variables Increase education spending, cut compensatory spending Increase family spending, cut compensatory spending Cut unemployment benefits, increase social investment spending Cut pensions, increase social investment spending M1 M2 M3 M4 Right-wing ideology 0.112*** 0.089*** 0.134*** 0.045 (0.026) (0.021) (0.021) (0.029) M5 M6 M7 M8 Right-wing ideology 0.063*** 0.064*** 0.089*** 0.028 (0.021) (0.023) (0.015) (0.030) Social spending preferences −0.395*** −0.346*** −0.442*** −0.250** (0.068) (0.065) (0.101) (0.107) Redistribution preferences −0.094*** 0.005 −0.120*** 0.094 (0.027) (0.026) (0.024) (0.059) Variables Increase education spending, cut compensatory spending Increase family spending, cut compensatory spending Cut unemployment benefits, increase social investment spending Cut pensions, increase social investment spending M1 M2 M3 M4 Right-wing ideology 0.112*** 0.089*** 0.134*** 0.045 (0.026) (0.021) (0.021) (0.029) M5 M6 M7 M8 Right-wing ideology 0.063*** 0.064*** 0.089*** 0.028 (0.021) (0.023) (0.015) (0.030) Social spending preferences −0.395*** −0.346*** −0.442*** −0.250** (0.068) (0.065) (0.101) (0.107) Redistribution preferences −0.094*** 0.005 −0.120*** 0.094 (0.027) (0.026) (0.024) (0.059) Notes: Country-clustered robust standard errors in parentheses; all models include the same independent variables as those models reported in Table 1. See Table E for the full models. *** P < 0.01, **P < 0.05, *P < 0.1. Next, we turn to the interaction of ideology and material self-interest (Hypothesis 5). Since we are interested in the interactive effects, we interpret the results based on predicted probability plots (Brambor, Clark and Golder, 2006: 74) and report the full models in Table F. Figure 2 plots predicted probabilities for spending support using models that include an interaction between left-right ideology and beneficiary group status, differentiating between ‘winners’ and ‘losers’ of the proposed reform. ‘Winners’ are students and those with small children in Model 1 and respondents with small children in Model 2; ‘losers’ are the unemployed in Model 3 and pensioners in Model 4. The values shown for the ‘reference groups’ are the average predicted probabilities if the conditioning variables of ‘winner’ (Models 1 and 2) and ‘loser’ status (Models 3 and 4) take the value of 0. The results are the same if we estimate separate models for the two beneficiary groups considered in Model 1 (see Figure B). Figure 2. View largeDownload slide Predicted probabilities of support for social investment when implying cuts in compensatory spending—interactions between self-interest and ideology for affected groups Notes: Predicted probabilities and 95 per cent confidence intervals are based on logistic regressions in Table F. The values shown for the ‘reference groups’ are the average predicted probabilities if the conditioning variables of ‘winner’ (Models 1 and 2) and ‘loser’ status (Models 3 and 4) take the value of 0. Figure 2. View largeDownload slide Predicted probabilities of support for social investment when implying cuts in compensatory spending—interactions between self-interest and ideology for affected groups Notes: Predicted probabilities and 95 per cent confidence intervals are based on logistic regressions in Table F. The values shown for the ‘reference groups’ are the average predicted probabilities if the conditioning variables of ‘winner’ (Models 1 and 2) and ‘loser’ status (Models 3 and 4) take the value of 0. Hypothesis 5 on the interaction effect between ideology and material self-interest is partially supported: we find significant effects in the expected direction in two of the models (Models 1 and 3). Substantively, these models show that left-wing ideology attenuates differences in preferences between winner and loser groups. That is, materialistic self-interest plays less of a role for left-wing voters, while it matters significantly for right-leaning citizens. For example, left-wing students or left-wing parents of small children are no more likely than the remaining population to demand spending increases on education at the expense of compensatory spending (Model 1). Also, while left-wing unemployed respondents most strongly reject cutting unemployment spending to finance additional social investment, this opposition is nearly as strong among left-wing individuals without unemployment spells (Model 3). In both models, the predicted probability for left-wing individuals to support a spending trade-off ranges around the value of 10 per cent, irrespective of beneficiary status. Put differently, ideology trumps materialistic self-interest. In contrast and as theorized, right-wing individuals who derive a material benefit from expanding social investment are even more supportive of shifting resources from compensatory to social investment spending. The predicted probability of supporting education spending expansion is 36 per cent for right-wing students or parents of small children, but only 20 per cent for rightist respondents who would not immediately benefit from such a reform. Similarly, the predicted probability of a right-wing respondent to support cuts in unemployment spending to finance additional social investment is 32 per cent, but only 20 per cent if this person is unemployed. This difference in terms of predicted probabilities reaches statistical significance at the 95 per cent level for those with moderate right-leaning positions (7 on the 0–10 scale). The widening confidence intervals for those at the far-right are to some extent related to the low number of unemployed with a right-wing ideology (see Table G for a cross-tabulation of ideology and beneficiary status). The differences in support between affected and unaffected groups are insignificant for those with right-leaning positions with regard to trade-offs involving an expansion of family spending and cutbacks in pensions (Models 2 and 4). This finding indicates that redistributive implications of programme characteristics matter beyond the distinction between social investment and compensatory policies. The distributive effects of cash transfers and services for families with small children (Maldonado and Nieuwenhuis, 2015) and the cross-class relevance of public pensions (Busemeyer, Goerres and Weschle, 2009) appear to cancel out the effects of ideology in these cases. Taken together, the potential of conflict about redistribution between different welfare beneficiary groups is muted among left-wing individuals who jointly reject policy reforms, which would reduce the overall amount of redistribution. In contrast, right-wing individuals are more supportive of policy proposals reducing the overall amount of redistribution, and the effects of material self-interest based on individual beneficiary status are particularly pronounced for those individuals. Conclusion This article contributes to recent debates about welfare state reform in ‘hard times’ by analysing novel data collected in an original survey on social investment and associated trade-offs. The innovation of this survey is to link respondents’ support for additional spending on presumably popular policies to cutbacks in other parts of the welfare state. This new way of measuring spending preferences is in certain ways superior to common approaches in international comparative surveys, but it also has limitations, and we therefore recommend it as a complement to existing research rather than a replacement. Its limitations are that we only look at trade-offs of a particular kind (social investment vs. social transfers), neglecting other policy fields, which citizens might be more inclined to sacrifice (such as spending on defence, for instance). Future work could use alternative methodological tools, such as vignette studies or conjoint analyses, to test these claims further in a multidimensional setting. Moreover, in this article we focused on policy trade-offs and do not take into account the possibility that citizens might in fact be willing to pay for higher spending with tax increases, although we explore this possibility elsewhere (Busemeyer and Garritzmann, 2017). Finally, we only have data for one time point for a limited number of countries, whereas using panel data on trade-offs would help understanding temporal dynamics. Obviously, results might be different when these parameters are changed. In general, our analysis demonstrated that social investments are popular, but this popularity has limits. For example, citizens are more willing to accept cutbacks in unemployment benefits rather than pensions to finance additional investments in education and family policies. Not surprisingly, we also find that those who benefit from social investment spending, i.e. students and parents with small children, are more willing to accept cutbacks in pensions and unemployment benefits to increase spending on social investment, whereas the potential losers of this deal, i.e. retired persons and to a lesser extent the unemployed, are opposed. Ideological predisposition also matter in the sense that traditional supporters of the welfare state are more inclined to reject cutbacks in unemployment benefits and pensions, even if this would lead to additional spending on social investment policies. We also found evidence for interaction effects between ideology and material self-interest in line with recent work (Brooks and Manza, 2013; Margalit, 2013) on social policy preferences. This finding has crucial implications for the future political viability of the social investment state. Our findings imply that—while it is true that left-leaning citizens are more in favour of social investment—it might actually be easier for right-leaning politicians to ‘recalibrate’ welfare states towards social investment because their voters are more willing to accept accompanying cutbacks in other (‘old’) social policies. Left-wing politicians could become stuck between a rock and a hard place: while their voters demand the expansion of social investments, they are not willing to accept policy trade-offs. One way out of this trap might be to increase taxes or public debt, i.e. to expand the size of the cake rather than aim at a different distribution. Yet, in a time of ‘permanent fiscal austerity’ (Pierson, 2001) and an omnipresent political discourse around debt tax reductions (Blyth, 2013; Haffert and Mehrtens, 2015), the associated political costs of such a strategy might be prohibitively high. In any case, our results imply that welfare state modernization implies different political dynamics for left- and right-wing parties. Erik Neimanns is postdoctoral researcher at the University of Konstanz. He received his PhD from the University of Konstanz in 2017 with a dissertation on ‘Public opinion and social investment: How political-institutional context shapes support and opposition towards expanding childcare’. He works in the ERC-funded project ‘Investing in education in Europe: Attitudes, Politics, Policies’. His current research interest is in the linkages between public opinion and policymaking in the area of education policies, with a particular focus on early childhood education and care. His work has been recently published in the Journal of European Social Policy. Marius R. Busemeyer is a professor of political science at the University of Konstanz. His current research focuses on public attitudes towards education policies and the welfare state, comparative political economy, and social inequality. His work has been published, among others, in the British Journal of Political Science, the Socio-Economic Review, the Journal of European Social Policy, and the European Journal of Political Research. Julian L. Garritzmann is senior researcher at the Universities of Konstanz and Zurich. He received his PhD from Konstanz in 2014 with a dissertation on the ‘Political Economy of Higher Education Finance’, published with Palgrave Macmillan. He served as a visiting fellow at Harvard University, at Duke University, and at Rutgers. His work has appeared in the Journal of European Social Policy, the Journal of European Public Policy, West European Politics, and PS: Political Science and Politics. In 2014, he was awarded the ESPAnet/JESP Doctoral Researcher Prize. His dissertation has received the Doctoral Researcher Prize of the German Political Science Association and is shortlisted for the Deutscher Studienpreis. Footnotes 1 Note that with regard to family policies we not only consider spending in kind, such as childcare but also cash transfers targeted to families with young children (see operationalization below). The latter can be expected to be more redistributive than the former, as it has strong poverty reducing effects, whereas access to childcare is often socially stratified (Van Lancker 2013; Maldonado and Nieuwenhuis, 2015). 2 More information on the survey can be found in a detailed background report (https://www.polver.uni-konstanz.de/en/busemeyer/research/current-projects/erc-project-inveduc/). 3 The results largely stay the same if the middle category is excluded. Supplementary Data Supplementary data are available at ESR online. Acknowledgements We presented previous versions of this article at the workshops ‘Welfare State Reform, Public Opinion and the Media’ at the University of Lund, ‘Citizens in Changing Welfare States’ at the University of Gothenburg, at the ESPAnet conference in Odense, and at various occasions at the University of Konstanz. We thank the participants, the editors of European Sociological Review, and the three anonymous reviewers for very helpful comments and suggestions. Funding Research for this project was funded by the Starting Grant ‘Investing in Education from Europe: Attitudes, Politics and Policies’ (INVEDUC) from the European Research Council (ERC), Grant No. 311769. References Anderson C. J. ( 2003 ). The psychology of doing nothing: forms of decision avoidance result from reason and emotion . Psychological Bulletin , 129 , 139 – 167 . Google Scholar CrossRef Search ADS PubMed Ansell B. W. ( 2010 ). From the Ballot to the Blackboard—The Redistributive Political Economy of Education . Cambridge : Cambridge University Press . Google Scholar CrossRef Search ADS Berinsky A. J. ( 2017 ). Measuring public opinion with surveys . Annual Review of Political Science , 20 , 309 – 329 . Google Scholar CrossRef Search ADS Blyth M. ( 2013 ). Austerity: The History of a Dangerous Idea . Oxford : Oxford University Press . Boeri T. , Börsch-Supan A. , Tabellini G. ( 2001 ). Would you like to shrink the welfare state? A survey of European citizens . Welfare State Reform , 16 , 8 – 50 . Bonoli G. ( 2007 ). Time matters: postindustrialization, new social risks, and welfare state adaptation in advanced industrial democracies . Comparative Political Studies , 40 , 495 – 520 . Google Scholar CrossRef Search ADS Bonoli G. ( 2013 ). The Origins of Active Social Policy: Labour Market and Childcare Policies in a Comparative Perspective . Oxford : Oxford University Press . Google Scholar CrossRef Search ADS Brambor T. , Clark W. R. , Golder M. ( 2006 ). Understanding interaction models: improving empirical analyses . Political Analysis , 14 , 63 – 82 . Google Scholar CrossRef Search ADS Brooks C. , Manza J. ( 2006 ). Why do welfare states persist? . Journal of Politics , 68 , 816 – 827 . Google Scholar CrossRef Search ADS Brooks C. , Manza J. ( 2013 ). A broken public? Americans’ responses to the great recession . American Sociological Review , 78 , 727 – 748 . Google Scholar CrossRef Search ADS Busemeyer M. R. ( 2012 ). Inequality and the political economy of education: an analysis of individual preferences in OECD countries . Journal of European Social Policy , 22 , 219 – 240 . Google Scholar CrossRef Search ADS Busemeyer M. R. ( 2015 ). Skills and Inequality: Partisan Politics and the Political Economy of Education Reforms in Western Welfare States . Cambridge : Cambridge University Press . Busemeyer M. R. , Goerres A. , Weschle S. ( 2009 ). Attitudes towards redistributive spending in an era of demographic ageing: the rival pressures from age and income in 14 OECD countries . Journal of European Social Policy , 19 , 195 – 212 . Google Scholar CrossRef Search ADS Busemeyer M. R. , Garritzmann J. L. ( 2017 ). Public opinion on policy and budgetary trade-offs in European welfare states: evidence from a new comparative survey . Journal of European Public Policy , 24 , 871 – 889 . Google Scholar CrossRef Search ADS Busemeyer M. R. , Neimanns E. ( 2017 ). Conflictive preferences towards social investments and transfers in mature welfare states: the cases of unemployment benefits and childcare provision . Journal of European Social Policy , 27 , 229 – 246 . Google Scholar CrossRef Search ADS Busemeyer M. R. et al. ( 2018 ). Investing in education in Europe: evidence from a new survey of public opinion . Journal of European Social Policy , 28 , 34 – 54 . Google Scholar CrossRef Search ADS Citrin J. ( 1979 ). Do people want something for nothing: public opinion on taxes and government spending . National Tax Journal , 32 , 113 – 129 . Cusack T. , Iversen T. , Rehm P. ( 2006 ). Risks at work: the demand and supply sides of government redistribution . Oxford Review of Economic Policy , 22 , 365 – 389 . Google Scholar CrossRef Search ADS De la O A. , Rodden J. A. ( 2008 ). Does religion distract the poor? Income and issue voting around the world . Comparative Political Studies , 41 , 437 – 476 . Google Scholar CrossRef Search ADS Esping-Andersen G. ( 2002 ). Why We Need a New Welfare State . Oxford : Oxford University Press . Google Scholar CrossRef Search ADS Garritzmann J. L. ( 2015 ). Attitudes towards student support: how positive feedback-effects prevent change in the four worlds of student finance . Journal of European Social Policy , 25 , 139 – 158 . Google Scholar CrossRef Search ADS Garritzmann J. L. , Häusermann S. , Palier B. , Zollinger C. ( 2017 ). WoPSI: The World Politics of Social Investment . LIEPP Working Paper No 64 , March. Sciences Po: Paris. Genschel P. ( 2004 ). Globalization and the welfare state: a retrospective . Journal of European Public Policy , 11 , 613 – 636 . Google Scholar CrossRef Search ADS Giger N. , Nelson M. ( 2013 ). The welfare state or the economy? Preferences, constituencies, and strategies for retrenchment . European Sociological Review , 29 , 1083 – 1094 . Google Scholar CrossRef Search ADS Goerres A. , Prinzen K. ( 2012 ). Can we improve the measurement of attitudes towards the welfare state? A constructive critique of survey instruments with evidence from focus groups . Social Indicators Research , 105 , 515 – 534 . Google Scholar CrossRef Search ADS Haffert L. , Mehrtens P. ( 2015 ). From austerity to expansion? Consolidation, budget surpluses, and the decline of fiscal capacity . Politics and Society , 43 , 119 – 148 . Google Scholar CrossRef Search ADS Hansen J. M. ( 1998 ). Individuals, institutions, and public preferences over public finance . American Political Science Review , 92 , 513 – 531 . Google Scholar CrossRef Search ADS Hasenfeld Y. , Rafferty A. J. ( 1989 ). The determinants of public attitudes toward the welfare state . Social Force , 67 , 1027 – 1048 . Google Scholar CrossRef Search ADS Häusermann S. , Schwander H. , Kurer T. ( 2015 ). High-skilled outsiders? Labor market vulnerability, education and welfare state preferences . Socio-Economic Review , 13 , 235 – 258 . Google Scholar CrossRef Search ADS Hemerijck A. ( 2013 ). Changing Welfare States . Oxford : Oxford University Press . Hemerijck A. ( 2017 ). The Uses of Social Investment . Oxford : Oxford University Press . Google Scholar CrossRef Search ADS Iversen T. , Soskice D. ( 2001 ). An asset theory of social policy preferences . American Political Science Review , 95 , 875 – 894 . Jakobsen T. G. ( 2010 ). Public versus private: the conditional effect of state policy and institutional trust on mass opinion . European Sociological Review , 26 , 307 – 318 . Google Scholar CrossRef Search ADS Jaeger M. M. ( 2006 ). What makes people support public responsibility for welfare provision: self-interest or political ideology? A longitudinal approach . Acta Sociologica , 49 , 321 – 338 . Google Scholar CrossRef Search ADS Jaeger M. M. ( 2009 ). United but divided: welfare regimes and the level and variance in public support for redistribution . European Sociological Review , 25 , 723 – 737 . Google Scholar CrossRef Search ADS Jordan J. ( 2013 ). Policy feedback and support for the welfare state . Journal of European Social Policy , 23 , 134 – 148 . Google Scholar CrossRef Search ADS Kangas O. E. ( 1997 ). Self-interest and the common good: the impact of norms, selfishness and context in social policy opinions . The Journal of Socio-Economics , 26 , 475 – 494 . Google Scholar CrossRef Search ADS Kuklinski J. H. et al. ( 2001 ). The political environment and citizen competence . American Journal of Political Science , 45 , 410 – 424 . Google Scholar CrossRef Search ADS Kumlin S. ( 2014 ). Policy feedback in political context: unemployment benefits, election campaigns, and democratic satisfaction. In: Kumlin S. , Stadelmann-Steffen I. (Eds.), How Welfare States Shape the Democratic Public. Policy Feedback, Participation, Voting, and Attitudes . Cheltenham; Northhamption : Edward Elgar , pp. 181 – 197 . Kumlin S. , Stadelmann-Steffen I. ( 2014 ). How Welfare States Shape the Democratic Public. Policy Feedback, Participation, Voting, and Attitudes . Cheltenham; Northhamption : Edward Elgar . Lupu N. , Pontusson J. ( 2011 ). The structure of inequality and the politics of redistribution . American Political Science Review , 105 , 316 – 336 . Google Scholar CrossRef Search ADS Maldonado L. C. , Nieuwenhuis R. ( 2015 ). Family policies and single parent poverty in 18 OECD countries, 1978–2008 . Community, Work & Family , 18 , 395 – 415 . Google Scholar CrossRef Search ADS Margalit Y. ( 2013 ). Explaining social policy preferences: evidence from the great recession . American Political Science Review , 107 , 80 – 103 . Google Scholar CrossRef Search ADS Meuleman B. , Chung H. ( 2012 ). Who should care for the children? Support for government intervention in childcare. In: Ervasti H. , Andersen J. G. , Fridberg T. , Ringdal K. (Eds.), The Future of the Welfare State: Social Policy Attitudes and Social Capital in Europe . Cheltenham : Edward Elgar , pp. 107 – 133 . Morel N. , Palier B. , Palme J. ( 2012 ). Towards a Social Investment Welfare State: Ideas, Policies and Challenges . Bristol : The Policy Press . Naumann E. , Buss C. , Bähr J. ( 2015 ). How unemployment experience affects support for the welfare state: a real panel approach . European Sociological Review , 32 , 81 – 92 . Google Scholar CrossRef Search ADS Owens L. A. , Pedulla D. S. ( 2014 ). Material welfare and changing political preferences: the case of support for redistributive social policies . Social Forces , 92 , 1087 – 1113 . Google Scholar CrossRef Search ADS Pierson P. ( 2001 ). The New Politics of the Welfare State . Oxford : Oxford University Press . Google Scholar CrossRef Search ADS Rehm P. ( 2009 ). Risks and redistribution an individual-level analysis . Comparative Political Studies , 42 , 855 – 881 . Google Scholar CrossRef Search ADS Rehm P. , Hacker J. S. , Schlesinger M. ( 2012 ). Insecure alliances: risk, inequality, and support for the welfare state . American Political Science Review , 106 , 386 – 406 . Google Scholar CrossRef Search ADS Roosma F. , Gelissen J. , Van Oorschot W. ( 2013 ). The multidimensionality of welfare state attitudes: a European cross-national study . Social Indicators Research , 113 , 235 – 255 . Google Scholar CrossRef Search ADS PubMed Schmidt-Catran A. W. ( 2016 ). Economic inequality and public demand for redistribution: combining cross-sectional and longitudinal evidence . Socio-Economic Review , 14 , 119 – 140 . Google Scholar CrossRef Search ADS Svallfors S. ( 2012 ). Contested Welfare States: Welfare Attitudes in Europe and beyond . Stanford : Stanford University Press . Google Scholar CrossRef Search ADS Vandenbroucke F. , Vleminckx K. ( 2013 ). Disappointing poverty trends: is the social investment state to blame? Journal of European Social Policy , 21 , 450 – 471 . Google Scholar CrossRef Search ADS Van Lancker W. ( 2013 ). Putting the child-centred investment strategy to the test: evidence for the EU27 . European Journal of Social Security , 15 , 4 – 27 . Google Scholar CrossRef Search ADS Van Oorschot W. ( 2006 ). Making the difference in social Europe: deservingness perceptions among citizens of European welfare states . Journal of European Social Policy , 16 , 23 – 42 . Google Scholar CrossRef Search ADS © The Author(s) 2018. Published by Oxford University Press. All rights reserved. For permissions, please e-mail: journals.permissions@oup.com This article is published and distributed under the terms of the Oxford University Press, Standard Journals Publication Model (https://academic.oup.com/journals/pages/about_us/legal/notices) http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png European Sociological Review Oxford University Press

How Popular Are Social Investment Policies Really? Evidence from a Survey Experiment in Eight Western European Countries

Loading next page...
 
/lp/ou_press/how-popular-are-social-investment-policies-really-evidence-from-a-9dZC5EA13l
Publisher
Oxford University Press
Copyright
© The Author(s) 2018. Published by Oxford University Press. All rights reserved. For permissions, please e-mail: journals.permissions@oup.com
ISSN
0266-7215
eISSN
1468-2672
D.O.I.
10.1093/esr/jcy008
Publisher site
See Article on Publisher Site

Abstract

Abstract The concept of the social investment welfare state has received a lot of attention and support both from academics and policymakers. It is therefore commonly assumed that policies such as investing in education or family services would also receive significant support from the mass public. While there are some indications of this, existing comparative surveys of public opinion usually do not take into account how citizens perceive and react to policy trade-offs, i.e. how they respond when forced to prioritize between different types of social policies, which is more realistic given budget constraints. This article presents original data from a representative survey of public opinion in eight Western European countries, studying how support for social investment policies changes when additional spending on these policies would have to be financed with cutbacks in other parts of the welfare state. The central findings are that citizens generally dislike being forced to cut back one type of social spending to expand another, but there is a significant degree of variation across individuals and policy fields. Material self-interest and ideological predispositions as well as their interaction help understanding differences in the acceptance of these trade-offs. The findings have important implications for the political viability of social investment policies. Political parties aiming to expand social investment in a context of fiscal austerity are confronted with different and distinct electoral constraints and challenges given the respective preferences of their electorates. Introduction In recent years, the notion of the ‘social investment’ model of the welfare state has gained much attention both among scholars and policymakers (Morel, Palier and Palme, 2012; Bonoli, 2013; Hemerijck, 2013, 2017). Social investment policies can be defined as policies geared at creating, mobilizing, or preserving human skills (Garritzmann et al., 2017: 36ff.), e.g. education, active labour market, or childcare policies. These ‘new’ types of social policies are often contrasted with more traditional social policies such as pensions, unemployment benefits, or health care (Vandenbroucke and Vleminckx, 2013), which are directed at compensating beneficiaries ex post for socio-economic and labour market losses. Many of the traditional social transfer and insurance policies entail a significant degree of redistribution and are therefore associated with a high degree of income-related conflict on the level of welfare state attitudes (Cusack, Iversen and Rehm, 2006; Busemeyer, Goerres and Weschle, 2009; Rehm, 2009). In contrast, social investments tend to provide relatively more benefits to the affluent middle- and upper-income classes (Van Lancker 2013). Consequently, social investment policies could be widely popular and less contested across the class divide. Existing scholarship using available survey data confirms that large majorities in advanced industrial democracies support higher public spending on education (Ansell, 2010; Busemeyer, 2012; Garritzmann, 2015) and government responsibility for the provision of childcare (Meuleman and Chung, 2012; Bonoli 2013; Busemeyer and Neimanns, 2017). This article takes this literature as point of departure. We probe the robustness of public support for social investment relative to social transfer policies. The motivation for this is the supposition that support for social investments might be vulnerable to the dominance of short-term over long-term concerns. While social investments might have significant benefits in the future, expanding these policies creates short-term costs, particularly if other parts of the welfare state have to be cut back. In an age of ‘permanent austerity’ (Pierson, 2001) policymakers are regularly confronted with policy trade-offs requiring cutbacks in some part of the welfare state to expand other parts such as the social investment pillar. A significant shortcoming of existing surveys on education and other social policies, however, is that the question wording used in these surveys does not force respondents to choose and prioritize between different policy areas. This might contribute to a severe overestimation of the extensiveness of support for social investment policies. To address this shortcoming, this article exploits a novel data set from a representative survey of public opinion in eight Western European countries (Busemeyer et al., 2018). An innovative component in this survey is that it explicitly takes into account policy trade-offs when measuring the extent of public support for social investment policies. Confronting respondents with different possibilities to prioritize among various social policies in a quasi-experimental setting, our analysis reveals that public support for social investment policies drops precipitously once trade-offs are taken into account. Furthermore, we find that material self-interest becomes a stronger determinant of policy preferences when moving from unconstrained to constrained settings. Finally, we also explore interaction effects between self-interest and ideology, showing that material self-interest related to spending trade-offs is a more important determinant of attitudes for those identifying with a right-wing ideology compared to those sympathizing with the left. As we discuss in the conclusion, this is an important and challenging finding for the future of the social investment state. Literature Review and Research Gap The analysis of welfare state attitudes has become a growing field in comparative welfare state research (Kumlin and Stadelmann-Steffen (2014) as well as Svallfors (2012) provide overviews). In general, earlier studies have focused more on identifying the micro-level determinants of attitudes and policy preferences (see, for instance, Hasenfeld and Rafferty, 1989), while more recent work shifts attention towards exploring the feedback effects of macro-level policies and institutions (Jaeger, 2009; Jakobsen, 2010; Jordan, 2013). In the present article, we are mainly concerned with studying the determinants of attitudes on the micro-level, however. Simplifying greatly, the pertinent literature on the micro-level determinants of attitudes can be divided into two categories: First, supporting the rational choice paradigm, materialistic self-interest has consistently been shown to shape preferences: people support those welfare policies from which they or their close relatives (expect to) benefit. Those not benefitting and those paying for these benefits, in turn, tend to be opposed. The literature discusses several indicators of self-interest, for example individuals’ income, educational background, labour market status and skill sets, age, welfare state beneficiary status, and labour market risk (Iversen and Soskice, 2001; Cusack, Iversen and Rehm, 2006; Busemeyer, Goerres and Weschle, 2009; Rehm, 2009; Rehm, Hacker and Schlesinger, 2012; Häusermann, Schwander and Kurer, 2015). More recently, panel data have been used to demonstrate the causal effect of personal experiences of material deprivation on social policy preferences (Jaeger 2006; Brooks and Manza, 2013; Margalit, 2013; Owens and Pedulla, 2014; Naumann, Buss and Bähr, 2015), but results remain inconclusive. Besides material self-interest, secondly, values, norms, and ideological predispositions have been found to influence welfare state attitudes (Kangas, 1997; Lupu and Pontusson, 2011). For example, Lupu and Pontusson (2011) argue that other-oriented attitudes matter for redistribution preferences, i.e. people are not only self-oriented but also altruistic. Moreover, social policy preferences are also driven by non-materialistic factors such as religious beliefs (De la O and Rodden, 2008) and by respondents’ perceptions of ‘deservingness’, i.e. how worthy of support they perceive particular beneficiaries to be (Van Oorschot, 2006). Finally, and maybe most importantly, this literature consistently reports that people’s ideological position is strongly related to their social policy preferences. A newer literature adds the claim that materialistic self-interest can be moderated by ideological positions (Margalit, 2013), an important argument that we also seek to test further. Despite these efforts, a major research gap in the existing literature on welfare state attitudes is that there is hardly any work on the issue of trade-offs between different welfare policies, particularly regarding ‘old’ compensatory vs. ‘new’ social investment policies. A partial exception is Boeri, Börsch-Supan, and Tabellini (2001), who confronted respondents in four countries with policy trade-offs, but only for the case of pension policies. In an earlier study, Hansen (1998) analyses public opinion data for the case of the United States. Both studies provide evidence for the general point, prominently discussed by Pierson (2001) as well as by Brooks and Manza (2006), that citizens dislike cutbacks in welfare state spending and express strong support for the maintenance of the current status quo. Kuklinski et al. (2001) analyse policy trade-offs in healthcare for the case of the United States, but their study deals with the role of information provision and motivation for trade-off choices, not with actual fiscal constraints of policymaking. This gap in research, which is ultimately related to missing survey data on the notion of trade-offs, is significant because studying trade-offs on the level of individual policy preferences has very relevant implications for policymaking and assessments about the political viability and legitimacy of the social investment model of the welfare state. So far, analyses of the transformation of European welfare states towards the social investment model have mostly been confined to the macro-level of policymaking (Morel, Palier and Palme, 2012; Bonoli, 2013; Hemerijck, 2013, but see Häusermann, Schwander, and Kurer, 2015), but it remains unclear to what extent this gradual shift reflects prevailing public attitudes. It might well be the case that public demand for social investment policies—responding to ‘new social risks’ such as single parenthood or skills obsolescence (Esping-Andersen, 2002; Bonoli, 2007)—is robust and significant, and citizens are in fact willing to accept cutbacks in the more traditional pillars of the welfare state. Vice versa, recent policy changes might not be reflected on the level of attitudes at all, which could indicate a certain lack of responsiveness on the part of policymakers, employing social investment reforms to ‘sugarcoat’ unpopular retrenchment. The focus of the present article is to address this gap by adding new empirical information on public opinion about trade-offs for a representative sample of European countries. Our aim is to shed light on political preferences towards policy trade-offs at the individual level, studying particularly the interplay of material self-interest and ideological orientations. Attitudes towards Social Policy Trade-Offs and Their Determinants Before presenting a set of testable hypotheses, we briefly clarify our understanding of a ‘trade-off’ on the level of individual preferences, which centres on the idea that citizens are forced to choose between two different competing policies: expanding one of these policies necessarily implies cutting back the other. Of course, this is a particular and therefore imperfect conception of a policy trade-off because, instead of trading one policy for another, citizens could be willing to accept more taxes or higher levels of public debt to finance additional spending. These are possibilities we explore in different, but related work, using another set of questions from the same survey (Busemeyer and Garritzmann, 2017), showing that citizens in fact prefer to accept higher taxes or higher levels of public debt rather than cutting back spending in other parts of the welfare state. Increasing taxes or debt levels may, however, not always be feasible in practice because of fiscal constraints associated with Europeanization and economic globalization (Genschel, 2004) as well as because of a political discourse dominated by concerns about austerity and debt reduction (Blyth, 2013; Haffert and Mehrtens, 2015). In the present article, we therefore deliberately focus on trade-offs between different policy areas to better understand the dynamics underlying redistribution within the welfare state, which may be different from conflicts about its general size. To measure citizens’ attitudes towards trade-offs, we rely on original public opinion survey data in eight European countries (details below) that forces respondents to choose between the expansion of social investment policies (education and family policies) to the detriment of social insurance and transfer policies (unemployment benefits and pensions). It is a legitimate question to ask to what extent our set-up and the implied notion of a ‘trade-off’ adequately reflects decision-making in the real world. Confronting survey respondents with a fixed set of choices always entails the danger of pre-maturely identifying patterns in public opinion, whereas ‘non-attitudes’ (Berinsky, 2017) might dominate in the real world. Previous research found that citizens might hold conflicting preferences such as a preference for a more generous welfare state and lower taxes (Citrin, 1979; Giger and Nelson, 2013). In spite of these limitations, we believe that our set-up adequately mirrors real-world challenges in the transformation of existing welfare states towards the social investment model. By comparing public support for social investment policies in constrained (i.e. with policy trade-offs) and unconstrained settings (i.e. without trade-offs), our analysis provides robust estimates of the upper and lower bounds for the extensiveness of support. While fiscal constraints in the real world of policymaking are rarely as binding as in our constructed set of limited choices, they are never fully absent either. We acknowledge that our construction of policy trade-offs does not represent the ‘full menu’ of potential policy alternatives. In fact, it is logically impossible to try to model all potential policy options, even when using tools that get closer to capturing multidimensional preferences, like conjoint analyses/factorial surveys, as the number of potential policy options and their combination is infinite. Consequently, we consider our analysis as a complement to existing studies rather than a full-scale replacement. Of course, it might be argued that solving trade-offs between conflicting public demands is the task of policymakers, not citizens. But we believe that studying citizens’ reactions to trade-off scenarios is a worthy and important enterprise because a permanent mismatch between policymakers’ actions and citizens’ preferences could contribute to a decline in the legitimacy of democratic decision-making and declining levels of political trust. Furthermore, the limited evidence from existing studies on fiscal trade-offs (Hansen, 1998; Kuklinski, et al. 2001; Busemeyer and Garritzmann, 2017) shows that citizens are well able to make informed and rational choices in these situations, as trade-offs ‘cause respondents to stop and think’ (Hansen, 1998: 519). In the following, we develop a set of testable hypotheses, which guide the subsequent empirical analysis. Perceptions of Trade-Offs Our first expectation is that citizens—just like politicians—dislike being confronted with trade-offs. Nobody likes making hard choices, particularly if they imply potential negative costs for oneself. Numerous theories in social psychology back this notion, showing that individuals when faced with a difficult decision become more status quo oriented and more resistant to change (Anderson, 2003). Assuming that a considerable share of individuals might prefer both social investment and compensatory social policies (Roosma, Gelissen and Van Oorschot, 2013), individuals should reject policy proposals that conflict with their ideal preferences. This implies that public support for the expansion of a particular social policy should drop significantly when moving from an unconstrained setting with no trade-off to a setting where spending increases would require cutbacks in other welfare areas (Hypothesis 1). Furthermore, we expect variation across policy areas, i.e. different kinds of trade-offs. For example, we expect different dynamics in an ‘education vs. pension’ trade-off than in an ‘education vs. unemployment benefits’ trade-off. This variation might be related to variation in perceived ‘deservingness’ (Van Oorschot, 2006) of different welfare beneficiary groups: the unemployed can be expected to be perceived as less deserving than retired persons or families with children for two reasons: Firstly, they are believed to be able to work themselves out of their predicament (Van Oorschot’s ‘control criterion’), whereas retired persons and young families cannot do the same because of their dependency on the welfare state. Secondly, pensioners are perceived to be more deserving as they have contributed more themselves (‘reciprocity criterion’) and are perceived as being ‘closer’ to the average respondent (‘identity criterion’). Besides the normative implications, the ‘deservingness’ argument also has a materialist, rational component, as individuals are likely to be more supportive of welfare state services and transfers, which they expect to need themselves in the future. Everybody turns old, but the risk of unemployment is more concentrated on certain groups (Busemeyer, Goerres and Weschle, 2009). Furthermore, the duration of personal experiences with the benefits and services of the welfare state differs across policies, as individuals (in Western Europe) have much longer personal experiences with the education system, whereas family policies targeted at families with small children cover a much shorter time span. Hence, cutbacks in pensions are probably less popular than cutbacks in unemployment spending, and spending increases on education can be expected to be more popular than spending increases on family policies for parents with small children (Hypothesis 2). Material Self-Interest Besides variation across policy areas, we expect variation across individuals in citizens’ perceptions of and reactions to trade-offs. We hypothesize that this micro-level variation in attitudes is related to material self-interest and that the explanatory power of indicators of self-interest will increase once we move from an unconstrained to a constrained (trade-off) scenario (Hypothesis 3). The simple reason is that citizens have become used to (and might even be dependent on) the provision of a large set of welfare services and benefits, which they would not want to give up even if other parts of the welfare state were expanded in exchange. Thus, we focus on beneficiary status as an indicator of material self-interest here, but include additional variables related to material self-interest such as household income and age as control variables in the analysis. This implies, for instance, that pupils and students support increases in education spending and are willing to cut compensatory spending on unemployment and pensions to achieve this. Of course, students might well become unemployed in the future, and they will surely turn old at some point, but the immediate short-term benefits of increasing spending on education might outweigh the rather diffuse costs of cutting transfers for the unemployed and pensioners. The same logic applies to parents of young children, whom we expect to support increases in public spending on families and education. In contrast, the unemployed should oppose more spending on social investments when this implies cutting back unemployment benefits. Finally, pensioners should oppose social investment expansion when this necessitates cutbacks in old-age pensions. As hinted at above, we also posit that the explanatory power in terms of statistical significance and effect size for indicators of material self-interest increases when moving from an unconstrained to a constrained (trade-off) scenario. In the constrained scenario, citizens supporting the expansion of a particular policy are also forced to accept cutbacks in other social policies. This reduces the size of the group supporting expansion (see Hypothesis 1 above), but it also changes the composition of this group. It is plausible to assume that those supporting the expansion of a particular policy in spite of cutbacks will be those with the strongest self-interest in that policy. In contrast, those who stop supporting the policy will be respondents whose support might have been more an expression of general support for the welfare state before rather than the particular policy in question (Roosma, Gelissen and Van Oorschot, 2013). Ideological Predispositions In addition to material self-interest, we expect that ideological predispositions matter. An often-discussed problem in the study of the effects of norms and values on attitudes is endogeneity: when assessing the impact of ideology on welfare support it is difficult to disentangle the direction of causality, in particular in cross-sectional surveys. Hence, statistical associations should be interpreted as correlational relations rather than causal statements. Since the dependent variable concerns trade-offs, the theoretical expectations related to ideological predispositions are not straightforward. For instance, people supporting social spending in general are likely to favour compensatory and social investment spending. While this relationship might seem rather trivial, it is difficult to deduce what this would imply for people’s preferences on policy trade-offs, as both supporters and opponents of the welfare state might oppose expanding one spending area at the expense of another, but for different reasons: individuals identifying with a left-wing ideology could object to being forced to choose between different types of social spending, which—in an unconstrained setting—they would both support. Vice versa, respondents subscribing to a right-wing ideology could be opposed to spending increases in general, even if they would partly be financed with cutbacks in other parts of the welfare state. However, as social investment policies tend to be less redistributive compared to many social transfers and therefore tend to benefit the (upper) middle class (Vandenbroucke and Vleminckx, 2013), right-wing supporters could be less opposed to recalibrating the welfare state from classical redistributive policies towards a more investment-oriented and less redistributive model. We further expect the influence of ideology to vary across the different policy trade-offs because each policy area is associated with distinct redistributive consequences. The association between ideology and preferences towards the different policy proposals should be particularly pronounced for policy proposals, which would most strongly reduce the amount of welfare state redistribution. This applies in particular to an expansion of relatively little redistributive education spending at the expense of compensatory spending, and to cuts in highly redistributive spending on unemployment benefits to increase spending on social investment. In contrast, the effect of ideology should cancel out to some extent for policy proposals leading to only minor changes in welfare state redistribution. This includes an expansion of financial support and public services for families with a strong poverty reducing effect (cf. Maldonado and Nieuwenhuis, 2015), which would be financed with cuts in compensatory spending. The same holds for cuts in spending on pensions as a less redistributive type of compensatory spending (Busemeyer, Goerres and Weschle, 2009) to finance additional social investment spending.1 Taken together, the more a recalibration of the welfare state from a compensatory to a social investment model would reduce overall welfare state redistribution, the more individuals with a left-wing ideology should oppose and those with a right-wing ideology support such spending shifts (Hypothesis 4). Furthermore, adapting Margalit’s (2013) argument we hypothesize that ideological predispositions moderate the association between material self-interest and support for social investment in the trade-off scenarios. We expect that when confronted with a policy trade-off, materialistic self-interest plays a less important role for left-wing voters because they are more solidaristic with all disadvantaged social risk groups and therefore oppose sacrificing benefits for one group to expand benefits for another. Thus, we expect that for left-leaning respondents the effect of their beneficiary status on their preferences is attenuated. Right-leaning citizens, in contrast, might be less opposed to recalibrating benefits (as discussed above) and their self-interest might be less attenuated by their ideological position. In sum, Hypothesis 5 posits an interaction effect between material self-interest and ideology. If this reasoning is true, it implies—at first glance counter-intuitively—that if in times of ‘permanent fiscal austerity’ (Pierson, 2001) the expansion of social investment is only possible at the expense of other spending, the ‘turn’ towards the social investment state is in fact more supported (or at least more willingly accepted) by right-leaning individuals. For them, the associated ideological costs of ending up in a potentially less redistributive welfare state model are lower. Research Design The research gap related to individual-level preferences towards trade-offs identified above is at least partly due to the lack of data on these issues as well as on social investment more generally in the existing comparative surveys like the International Social Survey Programme (ISSP), the European Social Survey (ESS), or the Eurobarometer (for details on this discussion and an overview of our survey, see Busemeyer et al., 2018). Thus, we conducted an original survey in eight European countries: Denmark, France, Germany, Ireland, Italy, Spain, Sweden, and the United Kingdom. Unfortunately, due to funding constraints, no Eastern or non-European countries could be included. Hence, the analysis is restricted to Western Europe but covers a wide variety of countries from all ‘worlds of welfare’ in Western Europe here, thereby enhancing representativeness and external validity. In each country, we surveyed a representative sample of 1,000–1,500 adult individuals (aged 18–99 years), overall 8,905 individuals. The survey was conducted by a professional survey-institute via computer-assisted telephone interviewing (CATI) and fulfils all standards of a high-quality survey. Following pre-tests in February 2014, the main fieldwork took place mid-April to early-June 2014.2 Dependent Variables To test our theoretical expectations, we randomly split the sample into four groups in each country (Table C shows that the randomization was successful as the groups are statistically indistinguishable on important covariates). The split in four, equally sized groups and random assignment to these groups are necessary to avoid statistical artefacts (and therefore wrong conclusions): if we had asked everybody these four questions, respondents might tend to reply in a similar fashion to these questions independent of their real preferences and/or they might show effects of survey fatigue or learning, leading to potential biases. Respondents thus see one, and only one, of the following statements (we add emphasis here to highlight the differences for the readers): Q1: ‘Imagine the [COUNTRY] government plans to increase spending on education by 10 per cent and wants to finance this by cutting the benefits for the unemployed’. Q2: ‘Imagine the [COUNTRY] government plans to increase spending on education by 10 per cent and wants to finance this by cutting old age pensions’. Q3: ‘Imagine the government plans to enact reforms involving a 10 per cent increase in the budget for financial support and public services for families with young children; and wants to finance this by cutting the benefits for the unemployed’. Q4: ‘Imagine the government plans to enact reforms involving a 10 per cent increase in the budget for financial support and public services for families with young children; and wants to finance this by cutting old age pensions’. The statements confront respondents with four policy trade-offs: education and family support, on one hand, and unemployment benefits and pensions, on the other. We chose education and financial support and public services for families with young children as prime examples of social investment policies, as they seek to ‘create, mobilize, or preserve skills’ (Garritzmann et al., 2017: 37) of children or their parents. Benefits for the unemployed and old-age pensions, in contrast, are typical examples of compensatory social policies. We can therefore analyse whether and why respondents prefer social investment or compensatory policies as well as differentiate further between two types of ‘old’ and ‘new’ welfare policies. We specified a ‘10% increase’ in the question framing for several reasons. First, this specification makes the policy options more concrete and tangible for respondents. Secondly, it increases the likelihood that respondents think in roughly similar terms about the proposed policy change. Speaking only about spending increases without providing a measuring yardstick, some respondents might think of a 1 per cent increase, while others imagine a 20 per cent change. Of course, the ‘10 percent’ figure is arbitrary to some extent and furthermore implies substantially different amounts of money across policy fields (e.g. pension expenditure is significantly higher than education spending) and countries. However, for our purpose it is not advisable to include different actual spending figures in the questions, since this increases the survey’s administrative complexity and compromises its comparability. Hence, we have to assume that respondents take the 10 per cent figure as a rough proxy signal for a non-trivial but realistic policy change. We used the answers to the four split sample questions to create four dependent variables: those supporting education spending (combining Q1 and Q2), those supporting spending on families with young children (Q3 and Q4), those accepting cutbacks in unemployment spending (Q1 and Q3), and, finally, those accepting cutbacks in pension spending (Q2 and Q4). By partially pooling the split-sample questions we can highlight the specific characteristics of each of the four policy fields regarding cuts and expansions and get an increased sample size for each single model leading to more robust estimates of the coefficients. All models include dummies for sub-group assignment. For each question, we offered five answer categories: ‘strongly agree’, ‘agree’, ‘neither agree nor disagree’, ‘disagree’, and ‘strongly disagree’. Moreover, two residual categories were included: ‘Don’t know’ and ‘No answer’. Thus, we are measuring spending preferences relative to the status quo, not in absolute terms (cf. Kumlin, 2014). A disadvantage of this commonly used five-point Likert scale is that respondents might choose the category ‘neither agree nor disagree’ for different reasons (Goerres and Prinzen, 2012). Admittedly, in our case, it is difficult to pin down exactly why respondents would remain undecided (‘neither nor’) towards a reform proposal: they might indeed reject the proposed spending shift between policy areas or they might simply dislike being forced to choose between different types of spending and reject the implied notion of welfare state reform as a zero-sum game. For this reason, we focus on those respondents who state a clear preference towards agreement of the proposed trade-off scenario against all other respondents. A reassuring indication that we are measuring meaningful responses rather than non-attitudes is that there are only few respondents in the residual categories of ‘dońt know’ and ‘no answer’, which we coded as missings (see Table B for summary statistics). Accordingly, when using regressions to probe individual-level determinants, we dichotomize the five categories into two groups: respondents, who (strongly) agree with the statement (coded ‘1’), and respondents, who (strongly) disagree or are undecided (coded ‘0’).3 That said, we have re-run the analysis using ordered logistic regressions, which produces similar results (see Table H). To test Hypothesis 1, which states that public support for social investment drops considerably once trade-offs are acknowledged, we add two additional dependent variables based on questions from our survey measuring preferences towards spending on ‘education’ and ‘financial support for families’ in an unconstrained scenario (i.e. no trade-offs implied). More specifically, the questions read: Q5 and Q6: ‘In the following, I will name several areas of government activity. Please tell me whether you would like to see more or less government spending in each area. Keep in mind that ‘more’ or ‘much more’ might require a tax increase’.—‘Education (Q5)’—‘Financial support for families (Q6)’. It could be objected that this unconstrained scenario is not ‘truly unconstrained’ because respondents are reminded of the possibility that a spending increase ‘might require a tax increase’. Yet, as we show elsewhere (Busemeyer and Garritzmann, 2017), the reminder about taxes does not have any measurable effect on people’s preference compared to a fully unconditional scenario. We can show this by comparing this question to a ‘truly’ unconditional question in our survey, where we do not include the reminder that more spending might increase taxes, and the support levels between the two questions are indistinguishable. We cannot use the latter question for the present analysis, however, because it is part of another survey experiment and therefore only covers part of the sample. We use the weak reminder about tax increases in the primary question to make it similar to the commonly used spending questions from the ISSP Role of Government module. All in all, we therefore feel confident about treating the question formulation above (Q5 and Q6) as a ‘de facto unconditional’ baseline category. Again, we dichotomize responses and distinguish preferences for ‘more’ or ‘much more’ spending (1) and preferences for the ‘same’, ‘less’, or ‘much less’ (0) spending. We match the split-sample questions from above to the respective items, i.e. unconstrained education spending preferences (Q5) for respondents given Q1 or Q2 and unconstrained family spending preferences (Q6) for those given Q3 or Q4. Of course, these items do not constitute a control group in a strict experimental sense, but nevertheless allow us to compare the difference between constrained and unconstrained settings when measuring spending preferences. Because of the binary nature of the dependent variables, we estimate multivariate single-level logit regressions. We include country fixed-effects and compute country-clustered robust standard errors to reflect the fact that the sample is stratified on countries. In the present article, we focus on associations on the micro-level of attitudes in the pooled sample, while the variation of effects across countries and welfare state regimes will be addressed in future work. We include sampling weights in the descriptive analysis but not in the regression models because weighting can bias the standard errors and is unnecessary when the control variables account for differences in sampling probability (Schmidt-Catran, 2016). Our central explanatory variables related to self-interest capture respondents’ beneficiary status (having small kids at home, being a student, unemployed, or retired). Small kids are here those below the age of 10 years. Because the control variable for kids above the age of 10 years also includes potentially adult children still living in the parental household, we do not have strong expectations in this case. In a further step of the analysis we add self-placement on a left-right scale as a measure for ideological predispositions. In addition, we control for household income, educational attainment, gender, age, perceived risk of unemployment, and household size. In the Online Appendix, we provide additional information on the exact measurement and operationalization of the variables used in the analysis (Table A). Because age correlates heavily with retirement status (r = 0.72), we consider age only for those in full-time employment, with the reference group being those in their 30s. Results First, we examine how spending preferences change once trade-offs are introduced. Figure 1 shows the share of respondents agreeing with spending increases on education (left) and families (right panel). The first important finding is that support for additional spending on social investment policies drops considerably if this would require cutbacks in compensatory spending, providing strong support for Hypothesis 1 and confirming Busemeyer and Garritzmann (2017) for a different set of policy trade-offs. In an unconstrained setting, we see impressive levels of support for spending on social investment: around 77 per cent of respondents support additional spending on education, and 55 per cent are supportive of additional financial support for families. As soon as spending trade-offs are introduced, however, support declines strongly. Across the four spending trade-offs, only 17 per cent of respondents would agree to a reform that increases social investments at the expense of compensatory policies. A majority disagrees with each of the proposed policy changes. This is not totally surprising, but the new contribution of our article is to show that an often-held, empirically largely untested assumption in the literature on the ‘new politics of the welfare state’ (Pierson, 2001) in fact has a solid empirical foundation. There is, of course, some cross-country variation in this respect (see Figure A), but by and large, the general pattern depicted in Figure 1 also holds across countries. Figure 1. View largeDownload slide Shares of respondents supporting increases in social investment spending—unconstrained spending preferences and when implying cuts in compensatory spending Figure 1. View largeDownload slide Shares of respondents supporting increases in social investment spending—unconstrained spending preferences and when implying cuts in compensatory spending Yet, a closer look at Figure 1 reveals that there are considerable differences across the four spending trade-offs, confirming Hypothesis 2: Respondents are much more willing to accept cutbacks in unemployment benefits than in pensions. Every fourth respondent is willing to cut unemployment benefits to increase education spending, but this share drops to 10 per cent when pensions would be reduced. In other words, the difference between these two compensatory policies is 17 percentage points when the trade-off is about education and 6 percentage points when family spending is concerned. This is strong support for our expectation that people are more willing to cut unemployment benefit generosity than pensions, either because of (future-oriented) self-interest or because of differences in perceived deservingness, as discussed in Hypothesis 2. The differences are less pronounced between investments on families with young children compared to educational investment. This might be related to the fact that both policies address groups that are deemed to be deserving of welfare state support or because the redistributive patterns are more similar for education and family policies than for unemployment benefits and pensions. In the next step, we examine to what extent this drop in spending support in the case of trade-offs is related to material self-interest based on individual beneficiary status. Hypothesis 3 postulated that the association between self-interest and support for spending on social investment policies should become stronger once we move from an unconstrained to a constrained (trade-off) scenario. Our logistic regression results in Table 1 provide considerable support for this claim: Models 1 and 2 reveal that in the unconstrained scenarios neither having small kids nor being a student is significantly associated with preferences for more spending. This picture changes when moving to the models for spending trade-offs: respondents with small children are now significantly more likely to support increases in social investment spending (Models 3 and 4), students support increasing education spending (Model 3), and the unemployed (Model 5) and the retired (Model 6) oppose cuts in ‘their’ benefit schemes for the sake of increasing social investment spending. These effects are substantial in size, as indicated by average marginal effect estimates, which allow a comparison of effect sizes across model specifications (see Table D): the likelihood of supporting additional spending increases by 2.7 percentage points for those with small children (Models 3 and 4) and by 6 percentage points for students (Model 3). Spending support drops by 9.3 percentage points for the unemployed in Model 5 and by 4.5 percentage points for the retired in Model 6. In sum, once spending increases for social investment would be financed by cuts in compensatory spending, considerable differences between different welfare state beneficiary groups come to the fore and self-interest becomes a much stronger predictor of preferences, confirming Hypothesis 3. Table 1. Logistic regressions: support for social investment—unconstrained spending preferences and when implying cuts in compensatory spending; maximum likelihood estimates Variables M1 M2 M3 M4 M5 M6 Increase education spending Increase family spending (financial support) Increase education spending, cut compensatory spending Increase family spending, cut compensatory spending Cut unemployment benefits, increase social investment spending Cut pensions, increase social investment spending Unconstrained spending preferences Spending trade-offs between social investment and compensatory spending No post-secondary education (Ref: Higher education) −0.328*** 0.279*** −0.192 0.171 0.075 −0.185** (0.089) (0.045) (0.168) (0.105) (0.166) (0.077) Vocational education −0.084 0.081 −0.119 −0.058 −0.086 −0.158 (0.143) (0.107) (0.164) (0.101) (0.142) (0.132) Household income  (Q1) (Ref: Middle quintile (Q3)) −0.030 0.171 −0.001 −0.347** −0.126 −0.179 (0.118) (0.135) (0.134) (0.166) (0.228) (0.219)  Q2 −0.103 −0.040 0.124 −0.028 0.251** −0.276* (0.142) (0.118) (0.099) (0.126) (0.119) (0.142)  Q4 −0.056 −0.391*** 0.084 −0.037 0.154 −0.184 (0.205) (0.143) (0.183) (0.153) (0.100) (0.202)  Q5 −0.026 −0.500*** 0.359** 0.255*** 0.370*** 0.258 (0.179) (0.087) (0.149) (0.085) (0.095) (0.172) Female −0.041 −0.024 −0.293*** −0.348*** −0.230*** −0.518*** (0.086) (0.077) (0.099) (0.088) (0.089) (0.071) Small child (<10 years) (Ref: No child in HH) 0.103 0.145 0.228** 0.236* 0.201* 0.304* (0.118) (0.150) (0.100) (0.135) (0.118) (0.156)  Older child (≥10 years) 0.358*** 0.092 0.030 −0.128 −0.049 −0.029 (0.107) (0.060) (0.114) (0.168) (0.105) (0.211) Age <30, full-time employed (FTE) (years)  (Ref: Age 30–39, FTE) −0.048 0.386* 0.347 0.510* 0.567*** 0.298 (0.199) (0.221) (0.217) (0.298) (0.173) (0.539)  40–49, FTE −0.027 −0.136 −0.358*** −0.348*** −0.454*** −0.155 (0.159) (0.201) (0.138) (0.077) (0.154) (0.129)  50–59, FTE −0.011 −0.272 −0.966*** −0.546** −0.830*** −0.595*** (0.193) (0.234) (0.109) (0.261) (0.177) (0.209)  >60, FTE −0.179** −0.084 −0.922*** −1.051** −0.904*** −1.248*** (0.079) (0.165) (0.239) (0.473) (0.286) (0.312)  Unemployed 0.412 0.225 −0.577 −0.477* −0.624* −0.318 (0.302) (0.360) (0.358) (0.283) (0.344) (0.434)  Studying 0.693 −0.144 0.377*** 0.551 0.457** 0.544 (0.474) (0.290) (0.133) (0.369) (0.227) (0.378)  Retired 0.122 −0.336 −0.815*** −0.600*** −0.768*** −0.594*** (0.130) (0.225) (0.135) (0.184) (0.141) (0.221)  Other 0.133* −0.063 −0.435*** −0.275*** −0.497** −0.108 (0.080) (0.222) (0.105) (0.096) (0.212) (0.293) Risk of unemployment 0.084 −0.122 −0.362 0.028 −0.262 0.002 (0.172) (0.143) (0.265) (0.095) (0.273) (0.132) Cohabit (Ref: one adult household) −0.022 0.092 0.055 −0.251*** −0.002 −0.254** (0.071) (0.079) (0.138) (0.088) (0.102) (0.108) Country dummies Yes Yes Yes Yes Yes Yes Trade-off dummy Yes Yes Yes Yes Pseudo R2 0.074 0.118 0.094 0.070 0.075 0.043 Observations 3,931 3,849 3,892 3,870 3,929 3,833 Variables M1 M2 M3 M4 M5 M6 Increase education spending Increase family spending (financial support) Increase education spending, cut compensatory spending Increase family spending, cut compensatory spending Cut unemployment benefits, increase social investment spending Cut pensions, increase social investment spending Unconstrained spending preferences Spending trade-offs between social investment and compensatory spending No post-secondary education (Ref: Higher education) −0.328*** 0.279*** −0.192 0.171 0.075 −0.185** (0.089) (0.045) (0.168) (0.105) (0.166) (0.077) Vocational education −0.084 0.081 −0.119 −0.058 −0.086 −0.158 (0.143) (0.107) (0.164) (0.101) (0.142) (0.132) Household income  (Q1) (Ref: Middle quintile (Q3)) −0.030 0.171 −0.001 −0.347** −0.126 −0.179 (0.118) (0.135) (0.134) (0.166) (0.228) (0.219)  Q2 −0.103 −0.040 0.124 −0.028 0.251** −0.276* (0.142) (0.118) (0.099) (0.126) (0.119) (0.142)  Q4 −0.056 −0.391*** 0.084 −0.037 0.154 −0.184 (0.205) (0.143) (0.183) (0.153) (0.100) (0.202)  Q5 −0.026 −0.500*** 0.359** 0.255*** 0.370*** 0.258 (0.179) (0.087) (0.149) (0.085) (0.095) (0.172) Female −0.041 −0.024 −0.293*** −0.348*** −0.230*** −0.518*** (0.086) (0.077) (0.099) (0.088) (0.089) (0.071) Small child (<10 years) (Ref: No child in HH) 0.103 0.145 0.228** 0.236* 0.201* 0.304* (0.118) (0.150) (0.100) (0.135) (0.118) (0.156)  Older child (≥10 years) 0.358*** 0.092 0.030 −0.128 −0.049 −0.029 (0.107) (0.060) (0.114) (0.168) (0.105) (0.211) Age <30, full-time employed (FTE) (years)  (Ref: Age 30–39, FTE) −0.048 0.386* 0.347 0.510* 0.567*** 0.298 (0.199) (0.221) (0.217) (0.298) (0.173) (0.539)  40–49, FTE −0.027 −0.136 −0.358*** −0.348*** −0.454*** −0.155 (0.159) (0.201) (0.138) (0.077) (0.154) (0.129)  50–59, FTE −0.011 −0.272 −0.966*** −0.546** −0.830*** −0.595*** (0.193) (0.234) (0.109) (0.261) (0.177) (0.209)  >60, FTE −0.179** −0.084 −0.922*** −1.051** −0.904*** −1.248*** (0.079) (0.165) (0.239) (0.473) (0.286) (0.312)  Unemployed 0.412 0.225 −0.577 −0.477* −0.624* −0.318 (0.302) (0.360) (0.358) (0.283) (0.344) (0.434)  Studying 0.693 −0.144 0.377*** 0.551 0.457** 0.544 (0.474) (0.290) (0.133) (0.369) (0.227) (0.378)  Retired 0.122 −0.336 −0.815*** −0.600*** −0.768*** −0.594*** (0.130) (0.225) (0.135) (0.184) (0.141) (0.221)  Other 0.133* −0.063 −0.435*** −0.275*** −0.497** −0.108 (0.080) (0.222) (0.105) (0.096) (0.212) (0.293) Risk of unemployment 0.084 −0.122 −0.362 0.028 −0.262 0.002 (0.172) (0.143) (0.265) (0.095) (0.273) (0.132) Cohabit (Ref: one adult household) −0.022 0.092 0.055 −0.251*** −0.002 −0.254** (0.071) (0.079) (0.138) (0.088) (0.102) (0.108) Country dummies Yes Yes Yes Yes Yes Yes Trade-off dummy Yes Yes Yes Yes Pseudo R2 0.074 0.118 0.094 0.070 0.075 0.043 Observations 3,931 3,849 3,892 3,870 3,929 3,833 Note: Country-clustered robust standard errors in parentheses. *** P < 0.01, **P < 0.05, *P < 0.1. Table 1. Logistic regressions: support for social investment—unconstrained spending preferences and when implying cuts in compensatory spending; maximum likelihood estimates Variables M1 M2 M3 M4 M5 M6 Increase education spending Increase family spending (financial support) Increase education spending, cut compensatory spending Increase family spending, cut compensatory spending Cut unemployment benefits, increase social investment spending Cut pensions, increase social investment spending Unconstrained spending preferences Spending trade-offs between social investment and compensatory spending No post-secondary education (Ref: Higher education) −0.328*** 0.279*** −0.192 0.171 0.075 −0.185** (0.089) (0.045) (0.168) (0.105) (0.166) (0.077) Vocational education −0.084 0.081 −0.119 −0.058 −0.086 −0.158 (0.143) (0.107) (0.164) (0.101) (0.142) (0.132) Household income  (Q1) (Ref: Middle quintile (Q3)) −0.030 0.171 −0.001 −0.347** −0.126 −0.179 (0.118) (0.135) (0.134) (0.166) (0.228) (0.219)  Q2 −0.103 −0.040 0.124 −0.028 0.251** −0.276* (0.142) (0.118) (0.099) (0.126) (0.119) (0.142)  Q4 −0.056 −0.391*** 0.084 −0.037 0.154 −0.184 (0.205) (0.143) (0.183) (0.153) (0.100) (0.202)  Q5 −0.026 −0.500*** 0.359** 0.255*** 0.370*** 0.258 (0.179) (0.087) (0.149) (0.085) (0.095) (0.172) Female −0.041 −0.024 −0.293*** −0.348*** −0.230*** −0.518*** (0.086) (0.077) (0.099) (0.088) (0.089) (0.071) Small child (<10 years) (Ref: No child in HH) 0.103 0.145 0.228** 0.236* 0.201* 0.304* (0.118) (0.150) (0.100) (0.135) (0.118) (0.156)  Older child (≥10 years) 0.358*** 0.092 0.030 −0.128 −0.049 −0.029 (0.107) (0.060) (0.114) (0.168) (0.105) (0.211) Age <30, full-time employed (FTE) (years)  (Ref: Age 30–39, FTE) −0.048 0.386* 0.347 0.510* 0.567*** 0.298 (0.199) (0.221) (0.217) (0.298) (0.173) (0.539)  40–49, FTE −0.027 −0.136 −0.358*** −0.348*** −0.454*** −0.155 (0.159) (0.201) (0.138) (0.077) (0.154) (0.129)  50–59, FTE −0.011 −0.272 −0.966*** −0.546** −0.830*** −0.595*** (0.193) (0.234) (0.109) (0.261) (0.177) (0.209)  >60, FTE −0.179** −0.084 −0.922*** −1.051** −0.904*** −1.248*** (0.079) (0.165) (0.239) (0.473) (0.286) (0.312)  Unemployed 0.412 0.225 −0.577 −0.477* −0.624* −0.318 (0.302) (0.360) (0.358) (0.283) (0.344) (0.434)  Studying 0.693 −0.144 0.377*** 0.551 0.457** 0.544 (0.474) (0.290) (0.133) (0.369) (0.227) (0.378)  Retired 0.122 −0.336 −0.815*** −0.600*** −0.768*** −0.594*** (0.130) (0.225) (0.135) (0.184) (0.141) (0.221)  Other 0.133* −0.063 −0.435*** −0.275*** −0.497** −0.108 (0.080) (0.222) (0.105) (0.096) (0.212) (0.293) Risk of unemployment 0.084 −0.122 −0.362 0.028 −0.262 0.002 (0.172) (0.143) (0.265) (0.095) (0.273) (0.132) Cohabit (Ref: one adult household) −0.022 0.092 0.055 −0.251*** −0.002 −0.254** (0.071) (0.079) (0.138) (0.088) (0.102) (0.108) Country dummies Yes Yes Yes Yes Yes Yes Trade-off dummy Yes Yes Yes Yes Pseudo R2 0.074 0.118 0.094 0.070 0.075 0.043 Observations 3,931 3,849 3,892 3,870 3,929 3,833 Variables M1 M2 M3 M4 M5 M6 Increase education spending Increase family spending (financial support) Increase education spending, cut compensatory spending Increase family spending, cut compensatory spending Cut unemployment benefits, increase social investment spending Cut pensions, increase social investment spending Unconstrained spending preferences Spending trade-offs between social investment and compensatory spending No post-secondary education (Ref: Higher education) −0.328*** 0.279*** −0.192 0.171 0.075 −0.185** (0.089) (0.045) (0.168) (0.105) (0.166) (0.077) Vocational education −0.084 0.081 −0.119 −0.058 −0.086 −0.158 (0.143) (0.107) (0.164) (0.101) (0.142) (0.132) Household income  (Q1) (Ref: Middle quintile (Q3)) −0.030 0.171 −0.001 −0.347** −0.126 −0.179 (0.118) (0.135) (0.134) (0.166) (0.228) (0.219)  Q2 −0.103 −0.040 0.124 −0.028 0.251** −0.276* (0.142) (0.118) (0.099) (0.126) (0.119) (0.142)  Q4 −0.056 −0.391*** 0.084 −0.037 0.154 −0.184 (0.205) (0.143) (0.183) (0.153) (0.100) (0.202)  Q5 −0.026 −0.500*** 0.359** 0.255*** 0.370*** 0.258 (0.179) (0.087) (0.149) (0.085) (0.095) (0.172) Female −0.041 −0.024 −0.293*** −0.348*** −0.230*** −0.518*** (0.086) (0.077) (0.099) (0.088) (0.089) (0.071) Small child (<10 years) (Ref: No child in HH) 0.103 0.145 0.228** 0.236* 0.201* 0.304* (0.118) (0.150) (0.100) (0.135) (0.118) (0.156)  Older child (≥10 years) 0.358*** 0.092 0.030 −0.128 −0.049 −0.029 (0.107) (0.060) (0.114) (0.168) (0.105) (0.211) Age <30, full-time employed (FTE) (years)  (Ref: Age 30–39, FTE) −0.048 0.386* 0.347 0.510* 0.567*** 0.298 (0.199) (0.221) (0.217) (0.298) (0.173) (0.539)  40–49, FTE −0.027 −0.136 −0.358*** −0.348*** −0.454*** −0.155 (0.159) (0.201) (0.138) (0.077) (0.154) (0.129)  50–59, FTE −0.011 −0.272 −0.966*** −0.546** −0.830*** −0.595*** (0.193) (0.234) (0.109) (0.261) (0.177) (0.209)  >60, FTE −0.179** −0.084 −0.922*** −1.051** −0.904*** −1.248*** (0.079) (0.165) (0.239) (0.473) (0.286) (0.312)  Unemployed 0.412 0.225 −0.577 −0.477* −0.624* −0.318 (0.302) (0.360) (0.358) (0.283) (0.344) (0.434)  Studying 0.693 −0.144 0.377*** 0.551 0.457** 0.544 (0.474) (0.290) (0.133) (0.369) (0.227) (0.378)  Retired 0.122 −0.336 −0.815*** −0.600*** −0.768*** −0.594*** (0.130) (0.225) (0.135) (0.184) (0.141) (0.221)  Other 0.133* −0.063 −0.435*** −0.275*** −0.497** −0.108 (0.080) (0.222) (0.105) (0.096) (0.212) (0.293) Risk of unemployment 0.084 −0.122 −0.362 0.028 −0.262 0.002 (0.172) (0.143) (0.265) (0.095) (0.273) (0.132) Cohabit (Ref: one adult household) −0.022 0.092 0.055 −0.251*** −0.002 −0.254** (0.071) (0.079) (0.138) (0.088) (0.102) (0.108) Country dummies Yes Yes Yes Yes Yes Yes Trade-off dummy Yes Yes Yes Yes Pseudo R2 0.074 0.118 0.094 0.070 0.075 0.043 Observations 3,931 3,849 3,892 3,870 3,929 3,833 Note: Country-clustered robust standard errors in parentheses. *** P < 0.01, **P < 0.05, *P < 0.1. Of course, some groups of individuals might benefit from both social investment and higher spending on social transfers (e.g. unemployed single parents in need of public childcare). Some might also expect to need certain social transfers in the future, while favouring social investment in the present (e.g. a student expecting high labour market risks once employed). Given the relatively small size of these ‘cross-pressured’ groups, however, we cannot directly measure the preferences of these people with our survey data. In general, the ‘cross-pressure’ effect could contribute to the high levels of opposition to trade-offs. This in fact makes it more likely that we still underestimate the true effects of self-interest. The fact that we find strong and clear effects for the different beneficiary groups indicates that individuals’ current situation is more relevant for their attitudes than their potential future needs. Next, we comment briefly on the control variables. As expected, elderly respondents (those in their 50s and 60s, who are closer to retirement age) oppose expanding social investment spending to the detriment of compensatory spending (Models 3–6). Again, this effect is much more pronounced in the constrained (trade-off) scenarios compared to the unconstrained settings (Models 1 and 2). (Perceived) Unemployment risk is negatively associated with cuts in unemployment spending (Model 5), but this effect is not significant. The significant positive effect of having children above the age of 10 years in Model 1 (but not in Models 3–6) might appear somewhat surprising at first sight. Yet, as we do not have more specific information about the age and status of these children, these children may be grown-ups living with their parents, working or in youth unemployment, which might account for this unexpected finding. Still, this finding needs to be interpreted with caution. Higher-income goes along with stronger support for social investment increases at the expense of compensation. As expected, there are again significant differences between the constrained and unconstrained settings: in the unconstrained settings, higher income is not (in the case of education) or negatively (in the case of family spending, which is likely to be perceived to be more redistributive) related to spending support. In the constrained setting, high-income respondents, in contrast, are more likely to support education and family spending if this would go along with cutbacks in the more redistributive transfer programmes. Higher educational background should be associated with stronger support for education spending increases (Busemeyer, 2015: ch. 5; Garritzmann, 2015). This association tends to be positive in our data as well, but it is not statistically significant. Moreover, women often benefit less than men from social insurance schemes built around the traditional male-breadwinner model and face particular challenges in combining work and family life at the same time (Esping-Andersen, 2002). Hence, there are good reasons to expect that women support higher spending on social investment policies. However, the results show that men are much more likely to accept cutbacks in compensatory spending to increase spending on social investments compared to women. Further research is clearly needed here, but a possible explanation is that women might prefer higher levels of public spending on both policy areas as they are found to be more supportive of the welfare state and redistribution in general (Cusack, Iversen and Rehm, 2006) and might want to achieve the same levels of compensatory safety nets as men while also favouring social investments. This leads us towards examining the role of ideology for spending trade-offs (Hypothesis 4). Table 2 shows the direct association of ideology and trade-off preferences. In additional model steps (Models 5–8) we control for general social spending and redistribution preferences to probe the robustness of this association. Overall, the results in Table 2 reveal a strong direct association between ideology and spending support: left-wing individuals are hostile to spending trade-offs, while those identifying themselves with the right are more supportive. This effect is particularly strong for the two models on expanding education spending and on cutting highly redistributive spending on unemployment (Models 1 and 3). Controlling for general social spending and redistribution preferences (Models 5–8) furthermore shows that these two variables explain a considerable share of the effect of left-right ideology. Individuals with preferences for more social spending reject shifting resources between different areas of welfare spending. Moreover, individuals supporting redistribution are also more likely to oppose spending shifts that would increase education spending or cut back unemployment spending. In contrast, the coefficient of ideology is insignificant for trade-offs that would imply cutbacks in pension spending (Models 4 and 8). This is likely due to the fact that pension spending is less progressive and has a strong appeal for the middle classes. In sum, respondents subscribing to a left-wing ideology are more solidaristic with all social risk groups and are more opposed to expanding benefits for one group at the expense of another, whereas right-leaning citizens are more supportive of shifting spending from more to less redistributive social policies. Table 2. Logistic regressions: effects of ideology on support for social investment when implying cuts in compensatory spending; maximum likelihood estimates Variables Increase education spending, cut compensatory spending Increase family spending, cut compensatory spending Cut unemployment benefits, increase social investment spending Cut pensions, increase social investment spending M1 M2 M3 M4 Right-wing ideology 0.112*** 0.089*** 0.134*** 0.045 (0.026) (0.021) (0.021) (0.029) M5 M6 M7 M8 Right-wing ideology 0.063*** 0.064*** 0.089*** 0.028 (0.021) (0.023) (0.015) (0.030) Social spending preferences −0.395*** −0.346*** −0.442*** −0.250** (0.068) (0.065) (0.101) (0.107) Redistribution preferences −0.094*** 0.005 −0.120*** 0.094 (0.027) (0.026) (0.024) (0.059) Variables Increase education spending, cut compensatory spending Increase family spending, cut compensatory spending Cut unemployment benefits, increase social investment spending Cut pensions, increase social investment spending M1 M2 M3 M4 Right-wing ideology 0.112*** 0.089*** 0.134*** 0.045 (0.026) (0.021) (0.021) (0.029) M5 M6 M7 M8 Right-wing ideology 0.063*** 0.064*** 0.089*** 0.028 (0.021) (0.023) (0.015) (0.030) Social spending preferences −0.395*** −0.346*** −0.442*** −0.250** (0.068) (0.065) (0.101) (0.107) Redistribution preferences −0.094*** 0.005 −0.120*** 0.094 (0.027) (0.026) (0.024) (0.059) Notes: Country-clustered robust standard errors in parentheses; all models include the same independent variables as those models reported in Table 1. See Table E for the full models. *** P < 0.01, **P < 0.05, *P < 0.1. Table 2. Logistic regressions: effects of ideology on support for social investment when implying cuts in compensatory spending; maximum likelihood estimates Variables Increase education spending, cut compensatory spending Increase family spending, cut compensatory spending Cut unemployment benefits, increase social investment spending Cut pensions, increase social investment spending M1 M2 M3 M4 Right-wing ideology 0.112*** 0.089*** 0.134*** 0.045 (0.026) (0.021) (0.021) (0.029) M5 M6 M7 M8 Right-wing ideology 0.063*** 0.064*** 0.089*** 0.028 (0.021) (0.023) (0.015) (0.030) Social spending preferences −0.395*** −0.346*** −0.442*** −0.250** (0.068) (0.065) (0.101) (0.107) Redistribution preferences −0.094*** 0.005 −0.120*** 0.094 (0.027) (0.026) (0.024) (0.059) Variables Increase education spending, cut compensatory spending Increase family spending, cut compensatory spending Cut unemployment benefits, increase social investment spending Cut pensions, increase social investment spending M1 M2 M3 M4 Right-wing ideology 0.112*** 0.089*** 0.134*** 0.045 (0.026) (0.021) (0.021) (0.029) M5 M6 M7 M8 Right-wing ideology 0.063*** 0.064*** 0.089*** 0.028 (0.021) (0.023) (0.015) (0.030) Social spending preferences −0.395*** −0.346*** −0.442*** −0.250** (0.068) (0.065) (0.101) (0.107) Redistribution preferences −0.094*** 0.005 −0.120*** 0.094 (0.027) (0.026) (0.024) (0.059) Notes: Country-clustered robust standard errors in parentheses; all models include the same independent variables as those models reported in Table 1. See Table E for the full models. *** P < 0.01, **P < 0.05, *P < 0.1. Next, we turn to the interaction of ideology and material self-interest (Hypothesis 5). Since we are interested in the interactive effects, we interpret the results based on predicted probability plots (Brambor, Clark and Golder, 2006: 74) and report the full models in Table F. Figure 2 plots predicted probabilities for spending support using models that include an interaction between left-right ideology and beneficiary group status, differentiating between ‘winners’ and ‘losers’ of the proposed reform. ‘Winners’ are students and those with small children in Model 1 and respondents with small children in Model 2; ‘losers’ are the unemployed in Model 3 and pensioners in Model 4. The values shown for the ‘reference groups’ are the average predicted probabilities if the conditioning variables of ‘winner’ (Models 1 and 2) and ‘loser’ status (Models 3 and 4) take the value of 0. The results are the same if we estimate separate models for the two beneficiary groups considered in Model 1 (see Figure B). Figure 2. View largeDownload slide Predicted probabilities of support for social investment when implying cuts in compensatory spending—interactions between self-interest and ideology for affected groups Notes: Predicted probabilities and 95 per cent confidence intervals are based on logistic regressions in Table F. The values shown for the ‘reference groups’ are the average predicted probabilities if the conditioning variables of ‘winner’ (Models 1 and 2) and ‘loser’ status (Models 3 and 4) take the value of 0. Figure 2. View largeDownload slide Predicted probabilities of support for social investment when implying cuts in compensatory spending—interactions between self-interest and ideology for affected groups Notes: Predicted probabilities and 95 per cent confidence intervals are based on logistic regressions in Table F. The values shown for the ‘reference groups’ are the average predicted probabilities if the conditioning variables of ‘winner’ (Models 1 and 2) and ‘loser’ status (Models 3 and 4) take the value of 0. Hypothesis 5 on the interaction effect between ideology and material self-interest is partially supported: we find significant effects in the expected direction in two of the models (Models 1 and 3). Substantively, these models show that left-wing ideology attenuates differences in preferences between winner and loser groups. That is, materialistic self-interest plays less of a role for left-wing voters, while it matters significantly for right-leaning citizens. For example, left-wing students or left-wing parents of small children are no more likely than the remaining population to demand spending increases on education at the expense of compensatory spending (Model 1). Also, while left-wing unemployed respondents most strongly reject cutting unemployment spending to finance additional social investment, this opposition is nearly as strong among left-wing individuals without unemployment spells (Model 3). In both models, the predicted probability for left-wing individuals to support a spending trade-off ranges around the value of 10 per cent, irrespective of beneficiary status. Put differently, ideology trumps materialistic self-interest. In contrast and as theorized, right-wing individuals who derive a material benefit from expanding social investment are even more supportive of shifting resources from compensatory to social investment spending. The predicted probability of supporting education spending expansion is 36 per cent for right-wing students or parents of small children, but only 20 per cent for rightist respondents who would not immediately benefit from such a reform. Similarly, the predicted probability of a right-wing respondent to support cuts in unemployment spending to finance additional social investment is 32 per cent, but only 20 per cent if this person is unemployed. This difference in terms of predicted probabilities reaches statistical significance at the 95 per cent level for those with moderate right-leaning positions (7 on the 0–10 scale). The widening confidence intervals for those at the far-right are to some extent related to the low number of unemployed with a right-wing ideology (see Table G for a cross-tabulation of ideology and beneficiary status). The differences in support between affected and unaffected groups are insignificant for those with right-leaning positions with regard to trade-offs involving an expansion of family spending and cutbacks in pensions (Models 2 and 4). This finding indicates that redistributive implications of programme characteristics matter beyond the distinction between social investment and compensatory policies. The distributive effects of cash transfers and services for families with small children (Maldonado and Nieuwenhuis, 2015) and the cross-class relevance of public pensions (Busemeyer, Goerres and Weschle, 2009) appear to cancel out the effects of ideology in these cases. Taken together, the potential of conflict about redistribution between different welfare beneficiary groups is muted among left-wing individuals who jointly reject policy reforms, which would reduce the overall amount of redistribution. In contrast, right-wing individuals are more supportive of policy proposals reducing the overall amount of redistribution, and the effects of material self-interest based on individual beneficiary status are particularly pronounced for those individuals. Conclusion This article contributes to recent debates about welfare state reform in ‘hard times’ by analysing novel data collected in an original survey on social investment and associated trade-offs. The innovation of this survey is to link respondents’ support for additional spending on presumably popular policies to cutbacks in other parts of the welfare state. This new way of measuring spending preferences is in certain ways superior to common approaches in international comparative surveys, but it also has limitations, and we therefore recommend it as a complement to existing research rather than a replacement. Its limitations are that we only look at trade-offs of a particular kind (social investment vs. social transfers), neglecting other policy fields, which citizens might be more inclined to sacrifice (such as spending on defence, for instance). Future work could use alternative methodological tools, such as vignette studies or conjoint analyses, to test these claims further in a multidimensional setting. Moreover, in this article we focused on policy trade-offs and do not take into account the possibility that citizens might in fact be willing to pay for higher spending with tax increases, although we explore this possibility elsewhere (Busemeyer and Garritzmann, 2017). Finally, we only have data for one time point for a limited number of countries, whereas using panel data on trade-offs would help understanding temporal dynamics. Obviously, results might be different when these parameters are changed. In general, our analysis demonstrated that social investments are popular, but this popularity has limits. For example, citizens are more willing to accept cutbacks in unemployment benefits rather than pensions to finance additional investments in education and family policies. Not surprisingly, we also find that those who benefit from social investment spending, i.e. students and parents with small children, are more willing to accept cutbacks in pensions and unemployment benefits to increase spending on social investment, whereas the potential losers of this deal, i.e. retired persons and to a lesser extent the unemployed, are opposed. Ideological predisposition also matter in the sense that traditional supporters of the welfare state are more inclined to reject cutbacks in unemployment benefits and pensions, even if this would lead to additional spending on social investment policies. We also found evidence for interaction effects between ideology and material self-interest in line with recent work (Brooks and Manza, 2013; Margalit, 2013) on social policy preferences. This finding has crucial implications for the future political viability of the social investment state. Our findings imply that—while it is true that left-leaning citizens are more in favour of social investment—it might actually be easier for right-leaning politicians to ‘recalibrate’ welfare states towards social investment because their voters are more willing to accept accompanying cutbacks in other (‘old’) social policies. Left-wing politicians could become stuck between a rock and a hard place: while their voters demand the expansion of social investments, they are not willing to accept policy trade-offs. One way out of this trap might be to increase taxes or public debt, i.e. to expand the size of the cake rather than aim at a different distribution. Yet, in a time of ‘permanent fiscal austerity’ (Pierson, 2001) and an omnipresent political discourse around debt tax reductions (Blyth, 2013; Haffert and Mehrtens, 2015), the associated political costs of such a strategy might be prohibitively high. In any case, our results imply that welfare state modernization implies different political dynamics for left- and right-wing parties. Erik Neimanns is postdoctoral researcher at the University of Konstanz. He received his PhD from the University of Konstanz in 2017 with a dissertation on ‘Public opinion and social investment: How political-institutional context shapes support and opposition towards expanding childcare’. He works in the ERC-funded project ‘Investing in education in Europe: Attitudes, Politics, Policies’. His current research interest is in the linkages between public opinion and policymaking in the area of education policies, with a particular focus on early childhood education and care. His work has been recently published in the Journal of European Social Policy. Marius R. Busemeyer is a professor of political science at the University of Konstanz. His current research focuses on public attitudes towards education policies and the welfare state, comparative political economy, and social inequality. His work has been published, among others, in the British Journal of Political Science, the Socio-Economic Review, the Journal of European Social Policy, and the European Journal of Political Research. Julian L. Garritzmann is senior researcher at the Universities of Konstanz and Zurich. He received his PhD from Konstanz in 2014 with a dissertation on the ‘Political Economy of Higher Education Finance’, published with Palgrave Macmillan. He served as a visiting fellow at Harvard University, at Duke University, and at Rutgers. His work has appeared in the Journal of European Social Policy, the Journal of European Public Policy, West European Politics, and PS: Political Science and Politics. In 2014, he was awarded the ESPAnet/JESP Doctoral Researcher Prize. His dissertation has received the Doctoral Researcher Prize of the German Political Science Association and is shortlisted for the Deutscher Studienpreis. Footnotes 1 Note that with regard to family policies we not only consider spending in kind, such as childcare but also cash transfers targeted to families with young children (see operationalization below). The latter can be expected to be more redistributive than the former, as it has strong poverty reducing effects, whereas access to childcare is often socially stratified (Van Lancker 2013; Maldonado and Nieuwenhuis, 2015). 2 More information on the survey can be found in a detailed background report (https://www.polver.uni-konstanz.de/en/busemeyer/research/current-projects/erc-project-inveduc/). 3 The results largely stay the same if the middle category is excluded. Supplementary Data Supplementary data are available at ESR online. Acknowledgements We presented previous versions of this article at the workshops ‘Welfare State Reform, Public Opinion and the Media’ at the University of Lund, ‘Citizens in Changing Welfare States’ at the University of Gothenburg, at the ESPAnet conference in Odense, and at various occasions at the University of Konstanz. We thank the participants, the editors of European Sociological Review, and the three anonymous reviewers for very helpful comments and suggestions. Funding Research for this project was funded by the Starting Grant ‘Investing in Education from Europe: Attitudes, Politics and Policies’ (INVEDUC) from the European Research Council (ERC), Grant No. 311769. References Anderson C. J. ( 2003 ). The psychology of doing nothing: forms of decision avoidance result from reason and emotion . Psychological Bulletin , 129 , 139 – 167 . Google Scholar CrossRef Search ADS PubMed Ansell B. W. ( 2010 ). From the Ballot to the Blackboard—The Redistributive Political Economy of Education . Cambridge : Cambridge University Press . Google Scholar CrossRef Search ADS Berinsky A. J. ( 2017 ). Measuring public opinion with surveys . Annual Review of Political Science , 20 , 309 – 329 . Google Scholar CrossRef Search ADS Blyth M. ( 2013 ). Austerity: The History of a Dangerous Idea . Oxford : Oxford University Press . Boeri T. , Börsch-Supan A. , Tabellini G. ( 2001 ). Would you like to shrink the welfare state? A survey of European citizens . Welfare State Reform , 16 , 8 – 50 . Bonoli G. ( 2007 ). Time matters: postindustrialization, new social risks, and welfare state adaptation in advanced industrial democracies . Comparative Political Studies , 40 , 495 – 520 . Google Scholar CrossRef Search ADS Bonoli G. ( 2013 ). The Origins of Active Social Policy: Labour Market and Childcare Policies in a Comparative Perspective . Oxford : Oxford University Press . Google Scholar CrossRef Search ADS Brambor T. , Clark W. R. , Golder M. ( 2006 ). Understanding interaction models: improving empirical analyses . Political Analysis , 14 , 63 – 82 . Google Scholar CrossRef Search ADS Brooks C. , Manza J. ( 2006 ). Why do welfare states persist? . Journal of Politics , 68 , 816 – 827 . Google Scholar CrossRef Search ADS Brooks C. , Manza J. ( 2013 ). A broken public? Americans’ responses to the great recession . American Sociological Review , 78 , 727 – 748 . Google Scholar CrossRef Search ADS Busemeyer M. R. ( 2012 ). Inequality and the political economy of education: an analysis of individual preferences in OECD countries . Journal of European Social Policy , 22 , 219 – 240 . Google Scholar CrossRef Search ADS Busemeyer M. R. ( 2015 ). Skills and Inequality: Partisan Politics and the Political Economy of Education Reforms in Western Welfare States . Cambridge : Cambridge University Press . Busemeyer M. R. , Goerres A. , Weschle S. ( 2009 ). Attitudes towards redistributive spending in an era of demographic ageing: the rival pressures from age and income in 14 OECD countries . Journal of European Social Policy , 19 , 195 – 212 . Google Scholar CrossRef Search ADS Busemeyer M. R. , Garritzmann J. L. ( 2017 ). Public opinion on policy and budgetary trade-offs in European welfare states: evidence from a new comparative survey . Journal of European Public Policy , 24 , 871 – 889 . Google Scholar CrossRef Search ADS Busemeyer M. R. , Neimanns E. ( 2017 ). Conflictive preferences towards social investments and transfers in mature welfare states: the cases of unemployment benefits and childcare provision . Journal of European Social Policy , 27 , 229 – 246 . Google Scholar CrossRef Search ADS Busemeyer M. R. et al. ( 2018 ). Investing in education in Europe: evidence from a new survey of public opinion . Journal of European Social Policy , 28 , 34 – 54 . Google Scholar CrossRef Search ADS Citrin J. ( 1979 ). Do people want something for nothing: public opinion on taxes and government spending . National Tax Journal , 32 , 113 – 129 . Cusack T. , Iversen T. , Rehm P. ( 2006 ). Risks at work: the demand and supply sides of government redistribution . Oxford Review of Economic Policy , 22 , 365 – 389 . Google Scholar CrossRef Search ADS De la O A. , Rodden J. A. ( 2008 ). Does religion distract the poor? Income and issue voting around the world . Comparative Political Studies , 41 , 437 – 476 . Google Scholar CrossRef Search ADS Esping-Andersen G. ( 2002 ). Why We Need a New Welfare State . Oxford : Oxford University Press . Google Scholar CrossRef Search ADS Garritzmann J. L. ( 2015 ). Attitudes towards student support: how positive feedback-effects prevent change in the four worlds of student finance . Journal of European Social Policy , 25 , 139 – 158 . Google Scholar CrossRef Search ADS Garritzmann J. L. , Häusermann S. , Palier B. , Zollinger C. ( 2017 ). WoPSI: The World Politics of Social Investment . LIEPP Working Paper No 64 , March. Sciences Po: Paris. Genschel P. ( 2004 ). Globalization and the welfare state: a retrospective . Journal of European Public Policy , 11 , 613 – 636 . Google Scholar CrossRef Search ADS Giger N. , Nelson M. ( 2013 ). The welfare state or the economy? Preferences, constituencies, and strategies for retrenchment . European Sociological Review , 29 , 1083 – 1094 . Google Scholar CrossRef Search ADS Goerres A. , Prinzen K. ( 2012 ). Can we improve the measurement of attitudes towards the welfare state? A constructive critique of survey instruments with evidence from focus groups . Social Indicators Research , 105 , 515 – 534 . Google Scholar CrossRef Search ADS Haffert L. , Mehrtens P. ( 2015 ). From austerity to expansion? Consolidation, budget surpluses, and the decline of fiscal capacity . Politics and Society , 43 , 119 – 148 . Google Scholar CrossRef Search ADS Hansen J. M. ( 1998 ). Individuals, institutions, and public preferences over public finance . American Political Science Review , 92 , 513 – 531 . Google Scholar CrossRef Search ADS Hasenfeld Y. , Rafferty A. J. ( 1989 ). The determinants of public attitudes toward the welfare state . Social Force , 67 , 1027 – 1048 . Google Scholar CrossRef Search ADS Häusermann S. , Schwander H. , Kurer T. ( 2015 ). High-skilled outsiders? Labor market vulnerability, education and welfare state preferences . Socio-Economic Review , 13 , 235 – 258 . Google Scholar CrossRef Search ADS Hemerijck A. ( 2013 ). Changing Welfare States . Oxford : Oxford University Press . Hemerijck A. ( 2017 ). The Uses of Social Investment . Oxford : Oxford University Press . Google Scholar CrossRef Search ADS Iversen T. , Soskice D. ( 2001 ). An asset theory of social policy preferences . American Political Science Review , 95 , 875 – 894 . Jakobsen T. G. ( 2010 ). Public versus private: the conditional effect of state policy and institutional trust on mass opinion . European Sociological Review , 26 , 307 – 318 . Google Scholar CrossRef Search ADS Jaeger M. M. ( 2006 ). What makes people support public responsibility for welfare provision: self-interest or political ideology? A longitudinal approach . Acta Sociologica , 49 , 321 – 338 . Google Scholar CrossRef Search ADS Jaeger M. M. ( 2009 ). United but divided: welfare regimes and the level and variance in public support for redistribution . European Sociological Review , 25 , 723 – 737 . Google Scholar CrossRef Search ADS Jordan J. ( 2013 ). Policy feedback and support for the welfare state . Journal of European Social Policy , 23 , 134 – 148 . Google Scholar CrossRef Search ADS Kangas O. E. ( 1997 ). Self-interest and the common good: the impact of norms, selfishness and context in social policy opinions . The Journal of Socio-Economics , 26 , 475 – 494 . Google Scholar CrossRef Search ADS Kuklinski J. H. et al. ( 2001 ). The political environment and citizen competence . American Journal of Political Science , 45 , 410 – 424 . Google Scholar CrossRef Search ADS Kumlin S. ( 2014 ). Policy feedback in political context: unemployment benefits, election campaigns, and democratic satisfaction. In: Kumlin S. , Stadelmann-Steffen I. (Eds.), How Welfare States Shape the Democratic Public. Policy Feedback, Participation, Voting, and Attitudes . Cheltenham; Northhamption : Edward Elgar , pp. 181 – 197 . Kumlin S. , Stadelmann-Steffen I. ( 2014 ). How Welfare States Shape the Democratic Public. Policy Feedback, Participation, Voting, and Attitudes . Cheltenham; Northhamption : Edward Elgar . Lupu N. , Pontusson J. ( 2011 ). The structure of inequality and the politics of redistribution . American Political Science Review , 105 , 316 – 336 . Google Scholar CrossRef Search ADS Maldonado L. C. , Nieuwenhuis R. ( 2015 ). Family policies and single parent poverty in 18 OECD countries, 1978–2008 . Community, Work & Family , 18 , 395 – 415 . Google Scholar CrossRef Search ADS Margalit Y. ( 2013 ). Explaining social policy preferences: evidence from the great recession . American Political Science Review , 107 , 80 – 103 . Google Scholar CrossRef Search ADS Meuleman B. , Chung H. ( 2012 ). Who should care for the children? Support for government intervention in childcare. In: Ervasti H. , Andersen J. G. , Fridberg T. , Ringdal K. (Eds.), The Future of the Welfare State: Social Policy Attitudes and Social Capital in Europe . Cheltenham : Edward Elgar , pp. 107 – 133 . Morel N. , Palier B. , Palme J. ( 2012 ). Towards a Social Investment Welfare State: Ideas, Policies and Challenges . Bristol : The Policy Press . Naumann E. , Buss C. , Bähr J. ( 2015 ). How unemployment experience affects support for the welfare state: a real panel approach . European Sociological Review , 32 , 81 – 92 . Google Scholar CrossRef Search ADS Owens L. A. , Pedulla D. S. ( 2014 ). Material welfare and changing political preferences: the case of support for redistributive social policies . Social Forces , 92 , 1087 – 1113 . Google Scholar CrossRef Search ADS Pierson P. ( 2001 ). The New Politics of the Welfare State . Oxford : Oxford University Press . Google Scholar CrossRef Search ADS Rehm P. ( 2009 ). Risks and redistribution an individual-level analysis . Comparative Political Studies , 42 , 855 – 881 . Google Scholar CrossRef Search ADS Rehm P. , Hacker J. S. , Schlesinger M. ( 2012 ). Insecure alliances: risk, inequality, and support for the welfare state . American Political Science Review , 106 , 386 – 406 . Google Scholar CrossRef Search ADS Roosma F. , Gelissen J. , Van Oorschot W. ( 2013 ). The multidimensionality of welfare state attitudes: a European cross-national study . Social Indicators Research , 113 , 235 – 255 . Google Scholar CrossRef Search ADS PubMed Schmidt-Catran A. W. ( 2016 ). Economic inequality and public demand for redistribution: combining cross-sectional and longitudinal evidence . Socio-Economic Review , 14 , 119 – 140 . Google Scholar CrossRef Search ADS Svallfors S. ( 2012 ). Contested Welfare States: Welfare Attitudes in Europe and beyond . Stanford : Stanford University Press . Google Scholar CrossRef Search ADS Vandenbroucke F. , Vleminckx K. ( 2013 ). Disappointing poverty trends: is the social investment state to blame? Journal of European Social Policy , 21 , 450 – 471 . Google Scholar CrossRef Search ADS Van Lancker W. ( 2013 ). Putting the child-centred investment strategy to the test: evidence for the EU27 . European Journal of Social Security , 15 , 4 – 27 . Google Scholar CrossRef Search ADS Van Oorschot W. ( 2006 ). Making the difference in social Europe: deservingness perceptions among citizens of European welfare states . Journal of European Social Policy , 16 , 23 – 42 . Google Scholar CrossRef Search ADS © The Author(s) 2018. Published by Oxford University Press. All rights reserved. For permissions, please e-mail: journals.permissions@oup.com This article is published and distributed under the terms of the Oxford University Press, Standard Journals Publication Model (https://academic.oup.com/journals/pages/about_us/legal/notices)

Journal

European Sociological ReviewOxford University Press

Published: Mar 23, 2018

There are no references for this article.

You’re reading a free preview. Subscribe to read the entire article.


DeepDyve is your
personal research library

It’s your single place to instantly
discover and read the research
that matters to you.

Enjoy affordable access to
over 18 million articles from more than
15,000 peer-reviewed journals.

All for just $49/month

Explore the DeepDyve Library

Search

Query the DeepDyve database, plus search all of PubMed and Google Scholar seamlessly

Organize

Save any article or search result from DeepDyve, PubMed, and Google Scholar... all in one place.

Access

Get unlimited, online access to over 18 million full-text articles from more than 15,000 scientific journals.

Your journals are on DeepDyve

Read from thousands of the leading scholarly journals from SpringerNature, Elsevier, Wiley-Blackwell, Oxford University Press and more.

All the latest content is available, no embargo periods.

See the journals in your area

DeepDyve

Freelancer

DeepDyve

Pro

Price

FREE

$49/month
$360/year

Save searches from
Google Scholar,
PubMed

Create lists to
organize your research

Export lists, citations

Read DeepDyve articles

Abstract access only

Unlimited access to over
18 million full-text articles

Print

20 pages / month

PDF Discount

20% off