Mature welfare states have long been on their way into an austerity phase marked by growing demands and more insecure revenues. A number of ‘reform pressures’ raise costs and increase the needs addressed by existing social protection. The usual suspects include ageing populations, unemployment, high dependency–employment ratios, immigration, sovereign debt and global competition. We investigate what citizens make of such growing reform pressures, if anything. An experiment implemented in Germany, Norway and Sweden tests if (and how) citizens react to information about reform pressures and, if so, which particular reform pressure frames affect beliefs about the future of the welfare state. We formulate hypotheses based on three distinct bodies of past research that, explicitly or implicitly, raise questions about how influential various types of pressure framing might be. Empirically, we show that while all three literatures contribute to our understanding of the subject matter, none of them provides the full story.
The first body of research grows out of Pierson’s (Reference Pierson1996, Reference Pierson2001) influential ‘new politics of the welfare state’ framework. According to this framework, institutional path dependencies, together with self-interest and risk aversion, are assumed to make citizens strongly supportive of the status quo. If true, a reasonable implication is that many are also hostile to the message that the welfare state may not be affordable. Moreover, strong support for the status quo makes political leaders hesitant to publicly debate pressures and implement reforms. Welfare states remain stable despite mounting pressure, at least unless effective ‘blame avoidance’ strategies can be used to obscure political responsibility. One of many such strategies is to publicly argue that pressure is so ubiquitous and overwhelming that we have ‘no choice’ but to reform the welfare state.
A second and more recent body of work analyses how and why mature welfare states now seem to be changing after all (Hemerijck Reference Hemerijck2013). Several scholars argue that this is not only due to blame avoidance but also to ‘credit claiming’. Reform-minded politicians argue their case by publicly politicizing growing reform pressures (Levy Reference Levy2010; Stiller Reference Stiller2010). Importantly, they mix cognitive arguments about economic pressure with ‘normative’ arguments. The focus here is not on massive pressure that supposedly leaves us with no choice. Rather, serious but ultimately manageable pressures, together with a clear normative cue, are integral to the narratives offered by credit-claiming reformers. Meanwhile, citizens do not seem to punish reforming governments at the polls, as implied by the ‘new politics’ framework. True, citizens still support key welfare institutions, but also worry about the economic consequences (Giger and Nelson Reference Giger and Nelson2013). Studies in this vein put our dependent variable – Welfare State Sustainability Perceptions – centre stage and demonstrate how these perceptions affect preferences and moderate electoral punishment. However, few studies examine how perceptions are formed in the first place. Can they really be regarded as responses to reform pressure, with citizens processing and drawing meaningful conclusions from information about sweeping demographic and economic transformations?
These questions lead to a third group of studies: experimental work on welfare attitudes. Here, and despite surging interest in experiments (Druckman et al. Reference Druckman2011), we largely lack studies about reform pressure framing as a broad phenomenon. A large number of studies, however, shows the importance of cues about how deserving recipients are (Petersen et al. Reference Petersen2010; Slothuus Reference Slothuus2007; Tyler et al. Reference Tyler1997). These studies complement the broader field of welfare state change research by convincingly specifying the ‘normative’ components of credit-claiming narratives. Yet deservingness studies typically overlook the question of economic sustainability. We know little about whether the framing of economic pressures matters in the absence of deservingness cues or – crucially – how the two operate in combination: are citizens more susceptible to economic pressure if they are provided with the right deservingness cue? We address these issues in an experiment that compares the impact of messages about different types of economic pressures, some of which are linked to (un)deserving groups, whereas others are not linked to any groups.
Our results suggest that reform pressure framing is not universally consequential. Several treatments yield no or weak effects, which fits with the ‘new politics’ depiction of welfare institutions and citizens as change resistant despite reform pressure. Other treatments were indeed important. This fits with recent work on welfare state change, by showing that citizens are in principle perfectly capable of processing and reacting to information about pressures. Moreover, such effects indeed depend on the normative cues that are implicit in pressure frames: the framing strategy most clearly supported by our data is to zoom in on pressures linked to groups that are perceived to be undeserving. This is evidenced in particular by immigration-related treatments, but also by treatments related to low employment, and by the fact that a population ageing message had no impact. This hierarchy of effects fits well with deservingness theory. The other possible framing strategy is to zoom out, making messages span a more diverse and broadly threatening set of challenges. This strategy, which is more in line with the ‘new politics’ emphasis on blame avoidance, received some (but clearly weaker) support.
In addition to effect variation across messages, the results also reveal intricate patterns of country variation. The final section incorporates these into an inductive discussion about context-message interactions. This analysis is inductive as our one-shot experiment in ‘only’ three countries cannot (and was not designed to) test contextual hypotheses, but also because the results did not corroborate our initial expectations. We initially expected pressure framing to be especially influential in contexts marked by more real and perceived unsustainability. In this regard, all three countries are among the more economically stable and sustainable European welfare states, at the same time as there is variation among them. Norway, with its exceptional oil revenues and funds, stands out economically and displays more policy stability (with exceptions, for example, in pension and family policy) (Bay et al. Reference Bay, Hatland, Hellevik and Koren2010). Hence, we initially reasoned that pressure framing would be less consequential in Norway than in Sweden or Germany. But the results revealed a much more complicated and – in some key respects – opposite cross-country pattern. This begs new explanations and we identify two broad lessons to be tested in future research. One lesson concerns the importance of immigration. We draw on recent work on the immigration–welfare nexus in opinion formation, raising the possibility that immigration frames not only matter, but perhaps particularly so in Scandinavia. The second lesson is that effects seem to grow in places and policy areas where specific pressures have recently been very salient in the public sphere. We illustrate by juxtaposing key findings with key developments in each country. The salience explanation has an interesting implication: pressure framing can ‘pave the way for its own success’ by contributing to a context in which future messages of the same type work better, at least for a period of time.
The ‘New Politics’ Framework and Welfare State Change
Paul Pierson’s (Reference Pierson1996, Reference Pierson2001) work on ‘the new politics of the welfare state’ framework is the natural starting point for our analysis. This influential framework contains assumptions that imply welfare state stability in the face of considerable reform pressures. Importantly, the institutions of mature welfare states are thought to generate their own support. This is partly due to large, self-interested and ‘risk-averse’ constituencies defending their benefits. However, broader mechanisms of institutional inertia are invoked to explain why non-beneficiaries also see the status quo as more rational, natural and even desirable.
Several implications seem to follow. For example, fear of electoral punishment will make office-seeking actors think twice before they propagate, let alone implement, visible welfare reform. Moreover, if it is true that a risk-averse public strongly supports an institutionalized status quo, then a reasonable implication is that citizens are also predisposed against information that suggests the welfare state is unsustainable. Granted, Pierson’s framework does not directly address the psychological mechanisms of information processing. But considerable research in political psychology, for example recent work on ‘motivated reasoning’, shows how strong and stable political predispositions serve as information filters that make people more prone to accept messages that are consistent with their predispositions. Conversely, people resist, or even argue against, messages that challenge their predispositions (Taber and Lodge Reference Taber and Lodge2006; Zaller Reference Zaller1992). Thus a plausible implication of the ‘new politics’ framework, and its depiction of welfare attitudes, is that pressure-framing effects are attenuated by strong welfare state support, and are not very consequential overall.
The ‘new politics’ framework has been amended and partly questioned in recent research on welfare state change (Jacobs Reference Jacobs2016). A key finding is that significant reforms have occurred in the last decade or so (Beramendi et al. Reference Beramendi2015; Hemerijck Reference Hemerijck2013). While radical and quick retrenchment is unusual, there is increasing focus on cost containment (Palier and Martin Reference Palier and Martin2007; Taylor-Gooby Reference Taylor-Gooby2001), together with non-negligible cutbacks (Korpi and Palme Reference Korpi and Palme2003; Scruggs Reference Scruggs2008). These processes were accelerated in several countries by the great recession and the eurozone crisis. However, austerity-driven retrenchment is not the only reform trajectory (van Kersbergen, Vis and Hemerijck Reference van Kersbergen, Vis and Hemerijck2014). Welfare states are also being recalibrated, with ‘new social risks’ receiving more attention, resulting for instance in expansive reform in family- and active labour market policy (Bonoli Reference Bonoli2005; Ferrera Reference Ferrera2008). Relatedly, scholars see a shift from traditional goals like income security and equality to a social investment-oriented welfare state that is more focused on human capital and equality of opportunity than on condition (Morel, Palier and Palme Reference Morel, Palier and Palme2012).
To explain these changes, scholars have reassessed the strategies that actors employ to make change politically and electorally feasible. Of course, under the ‘new politics’ framework reform is very difficult but occasionally possible through blame avoidance. Increasingly, however, there is also evidence of more offensive ‘credit-claiming’ strategies. From this vantage point even unpopular reform may not be ‘politically suicidal’ – as Pierson (Reference Pierson2001, 416) puts it – if it is visibly and publicly legitimized. In an overview, Levy argues that welfare reforms have not only been more frequent than ‘new politics’ implies, but have also materialized via a more open and communicative route (Levy Reference Levy2010, 561). Another overview by Bonoli draws similar conclusions but notes that research on credit claiming is less developed than the extensive work on blame avoidance (Bonoli Reference Bonoli2012; Hood Reference Hood2007; Lindbom Reference Lindbom2007; Vis Reference Vis2016; Wenzelburger and Hörisch Reference Wenzelburger and Hörisch2016).
A small but growing group of studies, however, does find empirical evidence of active credit-claiming strategies. Although scholars rely on somewhat different concepts, including ‘ideational leadership’ or ‘strategic reframing’, they demonstrate a number of reoccurring features (Elmelund-Præstekær and Emmenegger Reference Elmelund-Præstekær and Emmenegger2013; Stiller Reference Stiller2010). One is proactive agenda setting of reform pressures. This goes beyond blame avoidance (where problems are kept off the agenda, unless somebody else can be blamed, or where an economic crisis or otherwise massive pressure can be used as a blame avoidance device). Instead, actors put reform pressure firmly on the agenda, explaining how it makes the search for alternatives sensible. Among other things, this allows politicians to make a positive point out of exercising daring leadership and accusing opponents of sweeping problems under the rug. A further feature is the legitimization of visible reform using both cognitive and normative arguments. Cognitive arguments explain the economic logic behind proposed solutions, such as why reform alleviates resource scarcity. Normative arguments link solutions to values of some sort, such as cultural legacies, fairness or deservingness.Footnote 1 This stance, too, is different from the archetypal blame avoidance storyline that somebody else (or, say, a massive economic crisis) forces us to do what is all bad. In sum, then, actors are not necessarily afraid to put pressure on the agenda and publicly argue that choosing seemingly unpopular reform is both economically sensible and can have a normative payoff. This implies a wider applicability of reform pressure framing in political communication than suggested by the ‘new politics’ framework.
How do citizens fit into this picture of accelerating welfare state change and public argumentation? One set of findings suggests that reforming and retrenching governments are rarely punished at the polls (Armingeon and Giger Reference Armingeon and Giger2008; Arndt Reference Arndt2014; Giger Reference Giger2011; Kumlin Reference Kumlin2007; Lindbom Reference Lindbom2014). This may seem puzzling, given the oft-reported high levels of normative welfare support (Svallfors Reference Svallfors2012). However, newer findings show that perceptions of economic unsustainability are typically more negative than, and partly orthogonal to, normative support (Roosma, Gelissen and van Oorschot Reference Roosma, Gelissen and van Oorschot2013). Sustainability perceptions, moreover, appear to have behavioural and attitudinal consequences. Analysing voting behaviour with European Social Survey data, Giger and Nelson (Reference Giger and Nelson2013, 1091) show that ‘the perceived economic strain of social spending tempers generalized support for redistribution and makes many voters relatively more tolerant of a retrenchment agenda’. They also suggested to future research that ‘[…] fiscal austerity and severe economic conditions together with the demographic challenges might make people more responsive to claims of endangered economic sustainability of the current system of social welfare’.
This suspicion has been most thoroughly tested with respect to immigration. Several studies use experimental designs to examine how attitudes react to different types of exposure to immigration. A reoccurring finding and/or interpretation is that effects exist and are not only due to perceived ‘cultural threat’, racism, and the like, but also to the notion that immigration is an economic reason to worry about welfare state sustainability (Aalberg, Shanto and Messing Reference Aalberg, Shanto and Messing2012; Bay, Finseraas and Pedersen Reference Bay, Finseraas and Pedersen2016; Cappelen and Midtbø Reference Cappelen and Midtbø2016; Fietkau and Hansen Reference Fietkau and Hansen2018; Hjorth Reference Hjorth2016; Naumann and Stoetzer Reference Naumann and Stoetzer2018).Footnote 2
Only a handful of experimentalFootnote 3 studies have considered the impact of reform pressure beyond immigration. Naumann (Reference Naumann2017) found that experimental subjects in Germany who were exposed to detailed quantitative facts about demographic pressures become more accepting of a raised retirement age. By contrast, Brooks reported that information emphasizing the global economic crisis did not sway welfare support in the USA (Brooks Reference Brooks2011).Footnote 4 Kangas, Niemelä and Varjonen operationalized both ‘moral sentiments’ and ‘factual viewpoints’ on the costs of administrative social assistance reform in Finland and found that the former in particular built reform acceptance (Kangas, Niemelä and Varjonen Reference Kangas, Niemelä and Varjonen2014). These pioneering studies are valuable, and we continue down their path. Still, they are limited in two ways. First, most are single-country studies that cannot reveal whether the results are highly context specific or signal a more universal (in)ability to process sustainability problems. Of course, our three-country design is also limited in this regard but offers improvement. A second limitation is more crucial to our endeavour: these studies generally examine only one category of reform pressure (for example, population aging or immigration). Thus we know little about whether the effects vary across conceivable ways of framing the message that the welfare state is pressured.
Framing generally means that certain aspects of reality are highlighted, and others ignored, when defining a problem (Entman Reference Entman1993, 52). Moreover, while framing can generate entirely new beliefs, it is likely to be more effective when it triggers and strengthens existing considerations. However, this cannot happen so frequently that it eliminates competing considerations, or crystallizes attitudes completely. Thus people are seen as ambivalent and malleable, and potentially open to competing considerations (Zaller Reference Zaller1992), with attitudes varying depending on which ones are accessible in short-term memory. ‘Framing effects’, then, occur when changes in emphasis produce opinion change (Chong and Druckman Reference Chong and Druckman2007, 104). The reform pressure framing studied here concerns variation in emphasis on how pressured the welfare state is, and what it is pressured by. The next section presents hypotheses about the nature of and conditions for this phenomenon, drawing especially on the ‘new politics’ framework and ‘deservingness theory’. These two frameworks have useful and partly contrasting implications for what the ingredients of successful pressure framing might be.
Which Types of Reform Pressure Framing Matter – and Among Whom?
Past research has demonstrated that citizens from most social and political groups are generally and ‘sociotropically’ concerned with the health of the overall economy. This is seen most clearly in research on economic retrospective and prospective voting, which shows that the country’s economy matters more than personal pocketbook concerns.Footnote 5 Macroeconomic concerns are also widely seen as the ultimate ‘valence issue’; while there is disagreement over policy, the underlying goal is valued in all social and ideological camps (Stokes Reference Stokes1963). Thus one might reasonably formulate the baseline prediction that references to cost-inducing reform pressures generally make citizens worried about welfare state sustainability.
Hypothesis 1: Citizens will be more concerned about welfare state sustainability when exposed to reform pressure framing.
This hypothesis is a useful starting point, and it certainly fits with the emphasis on leadership and generally pressure-aware citizens found in recent work on welfare state change. But there are reasons to believe the situation is more complex. The next section uses the ‘new politics’ framework and research on ‘deservingness theory’ to develop more curtailed expectations.
Insights from the New Politics framework
The ‘new politics’ framework assumes that welfare state institutions have built strong support for the status quo. This is partly due to the rise of self-interested constituencies that defend their own benefits and services. The framework also relies on institutional theory to explain why the status quo often seems both more rational and normatively desirable also beyond immediate beneficiaries. These assumptions suggest there is limited room for reform pressure framing, particularly among self-interested and status quo-supporting citizens, but also among the many who have more generalized support for social protection and redistribution. Such groups are assumed to punish incumbents for unpopular policy changes prompted by reform pressures. By implication, we expect them to resist messages about pressure and sustainability problems, and formulate two hypotheses:
Hypothesis 2: Generalized welfare state support diminishes reform pressure framing effects.
Hypothesis 3: Support for a specific welfare state policy area diminishes the framing effects concerning an associated reform pressure.
It may not show at first sight, but ‘new politics’ assumptions do allow for more effective reform pressure framing of a certain kind. As noted previously, strong vested interests and normative support for the status quo make blame avoidance strategies necessary for reforming politicians. Work on blame avoidance has conceptualized and examined a long menu of possible strategies (Hood Reference Hood2007). Many of these are not communicative in nature but rather concern the design of political institutions and public policy. However, some strategies are ‘presentational’ and relate to communication with the public. Of special interest here is the notion that politicians can argue that reform pressure is so massive and ubiquitous that we have no choice but to implement necessary changes. Probably the best-documented version of this storyline is that a major economic crisis, with severe unemployment-related budgetary strains, necessitate policies that neither citizens nor decision makers prefer (Kuipers Reference Kuipers2006; Starke Reference Starke2008; Jakobsson and Kumlin Reference Jakobsson and Kumlin2017). But one can also imagine that politicians simultaneously bring up several pressures in a single argument, implying exceedingly strong pressure from multiple angles. In these situations, self-interested and risk-averse citizens may feel it is no longer possible to postpone painful reforms. Put differently, the distinction between immediate self-interest and vague long-term collective interests (not prioritized by citizens under the ‘new politics’ framework) becomes smaller. Additionally, broader references to encompassing pressure make it harder to discern which policy/group will suffer. Overall, we hypothesize:
Hypothesis 4: Reform pressure frames that invoke a larger number of reform pressures, or are otherwise encompassing and broadly threatening, matter more for perceptions of welfare state sustainability.
In sum, then, the ‘new politics’ framework implies limited effectiveness for pressure framing. It is likely to be present and effective mainly in areas with weak initial support or when multiple reform pressures are portrayed as so strong and immediate that obstacles related to self-interest and strong support for the status quo are overcome.
Insights from deservingness theory
Deservingness theory implies a somewhat different answer to what increases sensitivity to economic messages about reform pressures. Here, the key to stronger effects is to zoom in on specific pressures that make citizens think about particular undeserving groups (rather than zooming out on general or multiple pressures). Much past research shows that welfare state-related attitudes are sensitive to perceived deservingness. Deservingness-oriented reasoning is regarded as a deep-seated, perhaps genetic, tendency that is emotionally based, automated and chronically salient. It is thought to matter as soon as there is any information about deservingness (Petersen et al. Reference Petersen2010; Petersen et al. Reference Petersen2012; Slothuus Reference Slothuus2007; Van Oorschot et al. Reference van Oorschot2017). Deservingness theory is related to a broader class of models suggesting that people evaluate the political world in terms of social justice just as much as in terms of personal gain (Tyler et al. Reference Tyler1997). Theorists have referred to this notion as dual utility function (Rothstein Reference Rothstein1998), contingent consent (Levi Reference Levi1997) and moral economy (Mau Reference Mau2003).
There are at least two general ways of judging deservingness. The more direct one involves specific deservingness criteria. Van Oorschot distilled five such criteria from a large number of previous studies (Van Oorschot Reference van Oorschot2000). First, people are thought to assess ‘control over neediness’, where those who cannot help their predicament are more deserving. Secondly, the greater the ‘level of need’, the greater the perceived deservingness. Thirdly, ‘identity’ is thought to matter: that is, mainly needy people who ‘belong to us’ are deserving. A fourth criterion taps the attitudes of beneficiaries, while a fifth one concerns the reciprocal relationship between benefactors and beneficiaries; the solidarity of the former is thought to be contingent on whether the latter adhere to behavioural norms.
If they lack the necessary facts to apply these criteria, citizens can instead use general stereotypes about groups. Van Oorschot finds that Europeans share a common deservingness culture: the old are perceived as the most deserving of public welfare, followed by the sick and disabled, and then the unemployed (Van Oorschot Reference van Oorschot2006). ‘Immigrants’ are more or less universally seen as the least deserving of all. Our experiment draws on this rather universal pecking order. It has implications for reform pressure framing, as some (but not all) reform pressures are implicitly linked to specific groups. This is the case for three of the reform pressures included as treatments in the experiment. The aging population, for example, concerns a group universally regarded as deserving, whereas the opposite is true for immigration. Pressure emanating from high unemployment/low employment concerns groups that are more mixed and where deservingness is evaluated somewhere in between.
Importantly, we extend deservingness assumptions to also cover economic messages about the welfare state. We test the idea that pressure framing is more effective if information about costs is mixed with cues about which groups are involved, allowing respondents to see mounting economic pressure through a normative deservingness lens. The suspicion is that deservingness provides the key normative ingredient in broader ‘credit-claiming’ narratives that also incorporates ‘cognitive’ arguments about economic pressure (Esmark and Schoop Reference Esmark and Schoop2017; Slothus 2007. Two psychological mechanisms may be at play here. First, deservingness-related processing is widely seen as affective and automated, and so the evaluation about reform pressures themselves may be shaped by a quick ‘transfer of affect’ preceding, and subsequently shaping, more elaborate thinking about the plausibility of costs. Second, a further mechanism arises from the fact that deservingness criteria themselves are linked to issues of reciprocity, contributions and even belongingness. Thus a pressure linked to a less deserving group would seem a more plausible generator of costs, perhaps due to welfare abuse, or lack of taxpaying contributions. We formulate the following general hypothesis:
Hypothesis 5: Reform pressure frames associated with groups perceived as undeserving (immigration/immigrants) elicit stronger effects than pressure associated with deserving groups (ageing population/the old).
It is useful to note that Hypothesis 5 partly overlaps with Hypothesis 3, which predicts stronger effects with weaker concrete policy support. The overlap arises as policy support correlates with the deservingness perceptions of associated groups. A crucial nuance, however, is that the deservingness perspective predicts the strongest impact for very concrete frames that single out specific undeserving groups. ‘new politics’ assumptions imply at least as strong effects when pressures are big and scary (and hence useful in blame-avoiding ‘no choice’ narratives, see Hypothesis 4). From a deservingness viewpoint, by contrast, such encompassing frames should be less effective as they are by definition less clear about groups. Hence, frames that mix many pressures should impact less as it becomes harder to extract deservingness information.
Research Design, Data and Methods
We designed three identical survey experiments in Germany, Norway and Sweden, carried out in 2014–15. The German experiment was embedded in an online survey of 1,860 respondents fielded in April 2015 by Yougov. The sample is a quota sample based on census data about region, age, gender and education from its pool of volunteers.Footnote 6 The Norwegian experiment was embedded in a survey of 2,836 respondents fielded in late March 2014 by TNS Gallup. The sample is drawn from TNS Gallup’s pre-recruited panel of 50,000 individuals. The Swedish experiment was embedded in the citizen panel (3,729) of the University of Gothenburg’s Laboratory of Opinion Research, and fielded in early spring 2015. It uses a non-random quota design as in Germany.Footnote 7 All three samples are equally good in terms of internal validity. In terms of external validity, the Norwegian data fare better as the sample comes from a group of volunteers who originally were recruited by a random selection from the Norwegian resident population.
Experimental design
The treatments were randomly assigned at the individual level. In Germany, the random assignment was stratified by region of residence (East/West), with 20 per cent of all individuals in a treatment group living in the East. We operationalized five types of pressures: ‘the aging population’, ‘the high number of people at an employable age not working’, two types of immigration (‘EU’ and ‘non-Western’)Footnote 8 and the ‘financial crisis’. We also developed a frame that combined these pressure types (see Table 1). In addition, as all three countries are relatively affluent welfare states, we designed a frame that highlighted their (relatively) favourable economic situation with positive welfare state implications. This leaves us with seven treatment groups.Footnote 9 These are compared to a control group that, like the other groups, received an opening sentence telling them there is ‘some debate’ about welfare state costs. This is in itself a weak reminder of reform pressures. The control group, however, did not receive any additional information telling them what specifically pressures the welfare state or why ‘many people’ think so. Since all respondents learned there is some debate about pressures, the experiment becomes a conservative but well-controlled test largely capturing the impact of receiving additional information about specific pressure in various combinations. Specifically, the experiment was embedded in the following question:Footnote 10
There is some debate about (insert A Table 1) the costs associated with social security systems and public services in [country]. (Insert B Table 1). Thinking ahead 10 years from now, for each of the following social security and public services, where would you place yourself on a scale from 1-7, where 1 means that [country] will not be able to afford the present level of social security and public services, and 7 means that [country] will be able to afford to increase the level?
Treatment (insertions) and experimental groups
Dependent variable
The dependent variable comes from a subsequent battery on sustainability in seven policy areas: public health care, old age pensions, sickness benefits, unemployment benefits, social welfare benefits, care for the elderly and child care. These items are combined into an additive ‘sustainability index’ comprising all these items. Principal components analysis shows that all items load strongly on one underlying factor. Cronbach’s Alpha for the index is 0.96, and is almost the same across all countries.Footnote 11
Results
Table 2 gives an impression of the seven sustainability items and the index. As might be expected, Norwegians clearly worry the least about welfare state sustainability. By contrast, Germans are the most worried, with Swedes falling exactly in between. The order of countries is the same across policy areas.
Means for all respondents on the seven sustainability items, and the sustainability index
Q: ‘Thinking ahead 10 years from now, for each of the following social security and public services, where would you place yourself on a scale from 1–7, where 1 means that [country] will not be able to afford the present level of social security and public services, and 7 means that [country] will be able to afford to increase the level?’. Explicit ‘don’t know’ answers recoded to the neutral category. Only for respondents who answered all items.
We now proceed in two steps to test the six hypotheses. We first investigate the main effects of reform pressure framing, comparing these across stimuli, thus testing Hypotheses 1, 4 and 5. We analyse the experimental effects using a simple ordinary least squares (OLS) regression model with dummies for countries.Footnote 12 In a second step we examine the moderation effects arising from both general welfare state support (Hypothesis 2) and support for government responsibility in specific policy areas that are associated with a given frame (Hypothesis 3).
Main experimental effects
Table 3 reports two series of regressions.Footnote 13 One shows the impact of receiving any kind of information about a negative pressure, alongside that of the one positive frame. On average, receiving any of the negative frames reduced belief in sustainability by about 0.83 scale points compared to the control group (p=0.03).Footnote 14 This effect across the three countries is about 7 per cent of the standard deviation. Interestingly, there is no corresponding positive effect of the largely positive storyline that the welfare state is in good shape despite some talk about costs. Breaking down these results by country yields a further observation: effects are only significant in the two Scandinavian countries; indeed, the p-value of an F-test comparing models with and without experimental variables is highly insignificant at 0.76 in Germany. So overall, hearing about these pressures does so little in Germany that the variation may be due to chance alone.
OLS regression models of the perceived welfare state sustainability index
a p-value of the F-test that all coefficients of broad experimental groups equal to zero.
b p-value of the F-test that all coefficients of single experimental groups equal to zero.
Note: b-coefficients with p-values in brackets. All regressions include a control for whether the value of the dependent variable has been imputed (see Methods section). Models with German data include a control variable for region of residence (West, East) to account for residual regime socialization effects (Goerres and Tepe Reference Goerres and Tepe2012). Additional models with extra control variables for gender, income, age and education do not differ much in the estimated experimental effects. Models that are fitted separately for East and West Germany do not yield statistically significant differences.
Models 5–8 further unpack the negative frames. Already the pooled Model 5 suggests considerable variation. The strongest effects, relative to the control group, come from the two immigration-related stimuli. Individuals who are informed about EU migration pressure are, on average, about 1.36 scale points lower on the sustainability index compared to the control group (p=0.005). Pressure emanating from non-Western (Scandinavian formulation)/non-European (German formulation) countries has a somewhat stronger estimate effect of about −1.86 (p=0.000). This corresponds to about 16 per cent of the standard deviation. The third-largest effect stems from the ‘all pressures’ treatment containing all the negative pressures used elsewhere (b=−0.85; p=0.08). All other treatments have non-significant effects with population ageing performing the worst, and ‘too few working’ only approaching significance in the pooled analysis (b=−0.72; p=0.14). Consistent with Hypothesis 5, then, the uncovered hierarchy of effects reflects what we know about Europe’s deservingness culture (van Oorschot Reference van Oorschot2006). The effects indeed appear to depend on whether information about reform pressure ‘zooms in’ on specific costs associated with a particular group, and how (un)deserving that group is regarded to be.
It is interesting that the encompassing pressure frames matter less than the immigration frames. Specifically, the international crisis message is wholly inconsequential, while the ‘all pressures at once’ effect is just under half and two-thirds, respectively, of the immigration estimates. These observations are not very consistent with Hypothesis 4, which expects multiple and encompassing reform pressure framing to be especially effective. Note also that the immigration information is part of the ‘all pressures’ information. Thus only hearing about immigration produces stronger effects than hearing about immigration as well as aging, low employment and a financial crisis. These observations fit well with deservingness theory in that citizens react more strongly to information that singles out supposedly undeserving groups. They are less consistent with the blame avoidance-inspired notion that citizens are especially sensitive to ubiquitous and overwhelming pressure.
At the same time, the ‘all pressures’ frame is not wholly inconsequential. Thus while Hypothesis 4 is not strongly supported, encompassing pressure framing adds something over and above deservingness. Apparently reform pressure framing can play some role even though it is confusing about deservingness (that is, mixing pressures linked to deserving and undeserving groups, and with pressures unconnected to groups), at the same time as it is clear and dramatic on the multiple sources of rising welfare state costs.Footnote 15
Breaking down the results by country (Models 6–8) again reveals no effects in Germany, stronger effects with significant coefficients in Sweden and the strongest single effects in Norway.Footnote 16 The non-Western immigrant treatment has a significant impact in both Scandinavian countries, but is more influential in Norway (b=−3.11; p=0.000, whereas b=−1.98; p=0.02 in Sweden;). Only in Norway, moreover, does the EU immigration treatment matter (b=3.0; p=0.000). Further, the estimates for ‘all pressures’ in the two Scandinavian countries are −1.51 and −1.34, respectively, with p-values around 0.07. Finally, only in Sweden do we see an effect of the ‘too few working’ frame (−1.62; p=0.05). Overall, we can say that the Swedish effects are slightly stronger but also more dispersed across frames. By contrast, Norwegians react most strongly to immigration, with the non-Western frame causing a decline in the belief in long-term sustainability of about 30 per cent of the standard deviation.
To summarize, we have unearthed considerable heterogeneity and limitations in reform pressure framing effects, across both frames and contexts. Contrary to Hypothesis 1, there is little in the way of generally present effects. Instead, Hypothesis 5 receives a good amount of support in that deservingness cues matter. Also, and somewhat consistent with Hypothesis 4, a frame that ‘zooms out’ and encompasses many pressures at once (but is unclear on deservingness) made some impression in two of the three countries.
Finally, the results suggest variation not just across frames. There also seem to be interactions between country context and specific stimuli. The obvious example is the strong immigration effects in Norway, but one can mention the importance of low employment in Sweden, which is underscored by further analyses of specific components of the sustainability index (not shown, see Table E.1). While in general the experimental treatments have relatively similar effects on all components of the sustainability index, items related to the labour market behave differently in Sweden. To begin with, the ‘too few working’ treatment affects sustainability perceptions in most policy areas in Sweden, but these effects are above average in two areas most directly linked to this pressure: unemployment benefits and sick pay. In Sweden, moreover, sick pay perceptions are affected by reform pressure framing of all kinds. Said differently, all conditions significantly impact Swedes’ sustainability perceptions in this particular policy area. The concluding section will discuss possible explanations and avenues for further research on these discoveries about country-policy area interactions.
Moderated effects?
Hypotheses 2 and 3 predict that people with high levels of welfare state support are less affected. However, we find no support for Hypothesis 2, which states that general welfare state support hinders pressure framing. We ran a series of regressions involving interactions between seven treatments and general welfare state support, on the seven policy items and the sustainability index, in two countries.Footnote 17 We found no significant interactions.Footnote 18 Further, Hypothesis 3 states that those with stronger support for a specific related policy area are less affected by an associated reform pressure. Here, we ran four regressions each for Germany and Norway to check whether support for government responsibility conditions the effects (detailed results in Table F.1). As moderators, we used well-known items on ‘government responsibility’ to provide a reasonable standard of living for the old, for the unemployed and for migrants (0 to 10, with higher values indicating more support).Footnote 19 Specifically, three items are relevant for four corresponding frames: ensure a reasonable standard of living for the unemployed (‘too few working’), ensure a reasonable standard of living for the old (‘grey wave’) and ensure a reasonable standard of living for immigrants (the two immigration treatments).
We find only sporadic support for Hypothesis 3. As seen in Appendix Table F.1, F-tests comparing models with and without ‘treatment X specific support’ interactions suggest no model fit improvement for the population ageing and low employment treatments. However, for the strongest effects in our analysis – that is, those of immigration – we find interactions with support for government responsibility in this area. This provides a modicum of support for Hypothesis 3 in the area in which pressure framing matters the most.
Figure 1 illustrates the marginal effects of the two immigration frames, relative to the control group, in Germany and Norway. The x-axis represents the moderator variable, that is, support for the government providing a reasonable standard of living for immigrants.
Predicted marginal effects of the immigration frames across levels of support for government responsibility for providing reasonable standard of living for immigrants. Note: the graphs show the point estimates of the marginal effect of the immigration frames compared to the control group at all 11 values of the contextual variables (government should provide a reasonable standard of living for immigrants). The band stands for the 95 per cent confidence interval.
The EU immigration frame is not significantly moderated here, as illustrated by a rather flat gradient in the first row for both Germany and Norway. When individuals are confronted with the non-Western immigration frame, however, their reaction is clearly contingent. In Norway, those at scale values 8 to 10 (about one-third of the sample) are not estimated to be significantly affected by the non-Western frame compared to the control group. Those who indicate a value of 0 to 7, however, react significantly. Germany exhibits a similar pattern, although the effect estimates are positive for values of the moderator variable of 6 and higher (about 45 per cent of the sample). The estimated negative effect is close to significance at the 0.05 level only for those with very low support for government responsibility for immigrants.
Conclusion: One or Two Paths to Effective Pressure Framing?
We conducted an experiment in three countries testing whether exposure to welfare state reform pressure frames triggers concerns about the long-term financial sustainability of the welfare state. We formulated hypotheses that draw on three distinct bodies of research. While all three have implications that receive some support, none of them provides the full story of the empirical results.
The results reveal that pressure framing has a clear causal potential that is not always realized. In fact, exposure to several major cost-inducing welfare state challenges – most notably population aging – did not make people more worried about sustainability. This contradicts the literature on the politics of aging that assumes objective pressure is dominant also in citizens’ heads (Goerres and Vanhuysse Reference Goerres and Vanhuysse2012). Instead this observation, together with the fact that most significant coefficients are moderately sized, fits the ‘new politics’ depiction of welfare states and citizens as change resistant despite heavy reform pressure. Other observations, however, are less consistent with the implications we teased out of the ‘new politics’ framework. For example, we found that whenever framing matters, it usually does so across broad attitudinal groups. Here the migration frames worked differently, with clearer interactions with immigration support. But overall, and mostly inconsistent with Hypotheses 2 and 3, successful pressure framers may not risk indifference, and certainly not backlash, among groups strongly predisposed against their message. Thus whatever obstacles there are to pressure framing, strong welfare state support may for the most part not be one of them. Nor did we find that a positive message saying the welfare state is in good shape has any positive effects. So even though pressure framers may struggle to sway the public, they may not need to worry about pressure deniers (as could be expected based on the assumption that people are biased towards the status quo).
Some treatments indeed yielded substantively important effects, which generally echoes recent work on welfare state change. Some scholars have observed that welfare reforms have come with open ‘credit claiming’ by leaders. Others report that citizens have economic perceptions of the welfare state that are partly orthogonal to normative support, showing how perceptions moderate electoral punishment for potentially unpopular reforms. We add evidence suggesting that citizens are at least in principle perfectly able to draw key conclusions about welfare sustainability from brief reminders about reform pressures. This is consistent with the view that successful pressure framing helps explain why significant reforms have not triggered electoral punishment despite strong support for the status quo. Of course, our experimental data cannot determine whether this inference to real-world politics is valid. But overall, evidence is mounting that ‘new politics’-inspired blame avoidance does not provide the full story behind accelerating welfare state change.
Crucially however, effects vary greatly across the negative pressure frames. Consistent with ‘deservingness theory’, they elicit stronger effects if pressure is linked to groups generally seen as undeserving. Linking immigration and, to a lesser extent, low employment rates with costs appears to boost effects compared to population ageing, which elicits no impact. This hierarchy of effects indeed suggests deservingness theory is key to understanding also economic messages about the future of the welfare state. Specifically, it suggests deservingness theory is useful for specifying the normative arguments emphasized in work on credit claiming, ideational leadership and similar phenomena that have been argued to drive welfare state change.
There may, however, be another path to effective pressure framing. Consistent with ideas taken from the ‘new politics’ framework, we found some impact of a broader stimulus that was dramatic and clear about reform pressure, but confusing about deservingness. Although its effect was clearly weaker than the immigration frames, this ‘all pressures’ treatment played some role in two of the three countries studied.
Overall, these results suggest that reform-minded politicians who want to take their case to the public are left with a choice. Either they zoom in on pressures linked to undeserving groups, or they zoom out, making messages span a broader mix of multiple challenges. The choice is not easy. Especially when it comes to stressing immigration, there are obvious ideological and political obstacles for most party families. In some countries, only populist right parties are comfortable with it. Immigration treatments were also the most divisive ones in our experiments. While interactions with attitudes were generally absent, antipathy towards immigrants did increase the impact of hearing that immigrants pressure our welfare state.
Politicians who do not want to play the immigration card, or who are otherwise unable to formulate a deservingness-based narrative, can still zoom out and formulate encompassing messages about multiple economic pressures. This may be palatable for more party families in most democracies. However, judging from our results, this message is likely to make a weaker, if any, imprint. Future research is clearly needed here, however. One issue has to do with the precise content of our ‘all pressures’ frame. We only tested one of the many possible permutations of the message that the welfare state is under multiple pressures. Perhaps citizens are more susceptible to particular combinations of, and narratives about, broad reform pressure? Another limitation of the study relates to the experimental design. Our experiment made progress by combining economic pressures and deservingness in a single design, using differences across types of economic pressures to measure variation in deservingness. But one can certainly imagine more factorial designs that independently vary these and other relevant aspects, including pressure type, pressure severity and implied deservingness of related groups, such that the different possible combinations can be disentangled.
Outlook: Two Lessons About Country Variation
We found clear country differences in pressure framing effects. This section identifies two possible explanations that may be examined in future research. They grow directly out of the empirics, but were largely unanticipated. We treat them as inductive products of our undertaking and emphasize that they require future deductive testing.
One lesson concerns the importance of immigration. Our data indicate that especially Norwegian, but also Swedish respondents, are more prone than their German counterparts to react to immigration as a cost-inducing reform pressure. This is consistent with several single-country studies showing that exposure to various immigration-related stimuli affects welfare state-related attitudes in Scandinavia (Aalberg, Shanto and Messing Reference Aalberg, Shanto and Messing2012; Bay, Finseraas and Pedersen Reference Bay, Finseraas and Pedersen2016; Cappelen and Midtbø Reference Cappelen and Midtbø2016; Hjorth Reference Hjorth2016). But what is more, our results fit with a recent two-country experimental comparison in which Fietkau and Hansen (Reference Fietkau and Hansen2018, 136) found that:
Danes react more strongly to immigrants’ educational and qualification background than do Germans. We argue that because of Denmark’s larger welfare state and significantly larger social benefit spending, Danes are more afraid that immigration will pose a threat to their universal welfare system. Danes may perceive immigrants as exploiting welfare benefits more than natives and thus as bearing a high economic cost.
In sum, then, current evidence is consistent with an institutional explanation of country variation: citizens in larger and more universal welfare states, which have more generous (and possibly easier to access) services and benefits, are more sensitive to the notion that immigration constitutes an economic reason to worry about welfare state sustainability.Footnote 20
Our three-country study of multiple reform pressures also uncovered a more complex contextual pattern indicative of a second lesson. It seems that effects grow where reform pressures have been recently and massively debated in the public sphere. This ‘salience explanation’ is consistent with standard assumptions in political psychology in which recently activated ‘cognitive schemas’ are more likely to be applied again than those rarely used. Zaller drew on this idea in his ‘accessibility axiom’, stating that ‘the more recently a consideration has been called to mind or thought about, the less time it takes to retrieve that consideration or related considerations from memory and bring them to the top of the head for use’ (Zaller Reference Zaller1992, 48). Similarly, scholars working specifically on framing often assume, or find, that people ‘embrace the frame they hear most often and that most easily comes to mind’ (Chong and Druckman Reference Chong and Druckman2007, 104).
The relevance of salience is evident when juxtaposing key results with recent developments in each country. Beginning with Norway, we find that the strongest effects in our data concern both kinds of immigration. This is logical from a salience perspective, given that the populist right had in 2014 recently entered the government, and given that an influential government task force had for several years forcefully drawn the attention of policy-making elites, and the wider public, to the welfare–immigration nexus (NOU 2011, 7; Brochman and Grødem Reference Brochmann and Grødem2013). Importantly, this debate concerned immigration from the EU/European Economic Area as much as refugee immigration from outside Europe.
In Sweden, the 2006–14 centre-right government’s main winning narrative was arbetslinjen. The idea was that employment levels were too low to sustain the welfare state and the economy, and that incentive-oriented tax reforms, activation, tightened conditionality and some retrenchment were needed. Employment issues were highly salient among voters in the 2006 and 2010 elections. The centre-right coalition took over issue ownership in this area (Oscarsson and Holmberg Reference Oscarsson and Holmberg2013). Because arbetslinjen dominated Swedish politics for several years, and because some of the most controversial reforms involved sick pay and unemployment benefits, it is logical that we found pressure framing effects in Sweden for both treatments and dependent variables related to these topics.
The non-Western immigration frame (but not EU immigration) also matters in Sweden. From a salience perspective this makes sense. In summer 2014, the conservative Prime Minister Fredrik Reinfeldt held a massively debated speech, and a press conference, to kick off the election campaign. He forcefully put the increase in refugees after the Syria crisis on the campaign agenda. Although he famously asked his fellow citizens to ‘open their hearts’, he also used the occasion to be frank about budgetary implications, arguing that there could now be no expansive election promises. In the subsequent election, the populist radical right Sweden Democrats received an eye-popping 13 per cent, while the prime minister was widely interpreted as, and often blamed for, pitting immigration against the welfare state and making this a defining campaign feature. Overall, it makes sense that roughly six months later we find that Swedes are sensitive to pressure framing concerning non-Western immigration in particular.
The German data revealed little impact and hence fewer clues. At the same time, the largely absent effects in Germany could be explained by declining overall attention to welfare sustainability in recent years. This drop might in turn have happened partly because major reforms addressing reform pressure were implemented in an earlier phase. Of course, Germany’s Bismarckian social insurance systems, with their emphasis on status maintenance, were long seen as impervious to reform. This perception, however, has changed, with public agenda attention at its peak during the Reformstau debate of the late 1990s and early 2000s (Palier Reference Palier2010). This debate eventually produced reforms of pensions, unemployment benefits and family policy. After that, Zohlnhöfer notes, debates about pressures and reforms declined markedly; subsequent governments were ‘unwilling to pursue further some of the reform projects they themselves had embraced […] and talk about a ‘German employment miracle’ began […] This meant that reforms appeared less urgent for the time being, while it would have become particularly difficult to communicate the reforms as necessary and appropriate to the voters’ (Zohlnhöfer Reference Zohlnhöfer2015, 13–14).
The salience perspective has an important implication: reform pressure framers can ‘pave the way for their own success’. That is, if pressure frames accumulate in a somewhat concentrated period, they contribute to an overall context in which further opinion formation on the same topic works smoother. A momentum may be created in which pressure awareness builds at the same time that the conditions for future framing effects of the same kind improve. Of course, such momentum is unlikely to last forever. Eventually even the most salient issues will wane, in part due to ever-changing issue attention cycles (Downs Reference Downs1972), or to more sudden and dramatic ‘punctuated equilibria’ (Baumgartner et al. Reference Baumgartner2009). Moreover, agenda shifts may be especially likely whenever politicians change policy in the process (Soss and Schram Reference Soss and Schram2007). In such cases, ‘thermostatic’ reactions to reforms may quickly return citizens to a more skeptical stance (Naumann Reference Naumann2014; Soroka and Wlezien Reference Soroka and Wlezien2010), thus further contributing to a closing of the reform window that pressure framing initially helped open.
Supplementary Material
Replication data sets are available in Havard Dataverse at: https://dx.doi.org/10.7910/DVN/F05CTI and online appendices at: https://doi.org/10.1017/S0007123418000224
Acknowledgements
We thank our excellent research assistants: Robin Arens, Hanna Bugge, Hedda Haakestad, Hayfat Hamidou, and Atle Haugsgjerd. Earlier versions were presented at conferences organized by the Council for European Studies (Paris), the European Political Science Association (Edinburgh), the Nordic Political Science Association (Gothenburg), Policy Network (Oxford), in workshops/seminars at Lund University (organised by Moira Nelson and Georg Wenzelberger) and the University of Berne (organized by Isabelle Stadelmann-Steffen), and in various talks in our respective home institutions. A collective round of thanks to the many participants in these events. Finally, we thank the three anonymous reviewers for their helpful comments.
Financial support
This research was carried out under the auspices of the Support for the Affluent Welfare State project, directed by Staffan Kumlin and funded by the Research Council of Norway (no. 217122). Achim Goerres received further funding from the University of Duisburg-Essen.
