Introduction

Should immigrants have the same access to welfare benefits and services as the native population? Emphasized by the populist radical right in particular, this question has received considerable attention in public debates in advanced democracies in recent years. A steady rise in the number of immigrants, heightened by the recent surges associated with the 2015 refugee crisis and the ongoing war in Ukraine, has triggered concerns about the consumption of scarce public resources and the future of the welfare state. Welfare chauvinism – the belief that welfare benefits should be reserved only for the native population (Andersen & Bjorklund, 1990) – has been increasingly evident not only in policy agendas of populist radical right (and, increasingly, mainstream right) parties (e.g. Careja et al., 2016; Afonso & Rennwald, 2018; Ennser-Jedennastik, 2018), but also in public opinion (Eger et al., 2020; Marx & Naumann, 2018).

Research shows that welfare chauvinism at the individual level is associated with both ideological and economic factors. Social identity, values and group belonging are seen as important predictors of welfare chauvinist attitudes (Eger & Breznau, 2017; Ford, 2016). Among economic factors, low income and occupational status, and especially perceived economic insecurity tend to be correlated with welfare chauvinist attitudes (Hjorth, 2016; Kros & Coenders, 2019; Mewes & Mau, 2012). While this scholarship helps us understand why some individuals are more chauvinistic than others, it offers little insight into changes in attitudes, which can occur even without changes in these predisposing factors. We know that political attitudes towards immigration are not fixed and can be shaped by ‘situational triggers’ (Sniderman et al., 2004) and cues, including change in the rhetoric and the tone of media coverage (Brader et al., 2008; Boomgaarden & Vliegenhart, 2009; Hopkins, 2010). Building on this literature, we expect that informational cues can shift attitudes by activating (or decreasing) support for welfare chauvinism. Specifically, we set out to explore if chauvinist attitudes can be generated by negative portrayal of immigrants’ fiscal impact or attenuated by messages about their positive contribution.

Our theory marries insights from two distinct literatures on the effects of information on political attitudes – research on negativity bias and on motivated reasoning. Stemming from prospect theory and evolutionary biology, research on negativity bias has shown that negative information is more powerful in shaping decisions and attitudes than positive information (Kahneman & Tversky, 1979; Tversky & Kahneman, 1981; Norris, 2021). People pay more attention to negative messages, and this information is both more easily noticed and more memorable (Baumeister et al., 2001; Rozin and Royzmanm, 2001). Research on motivated reasoning, on the other hand, suggests that existing values and attitudes moderate any framing effects. For information to be accepted by individuals, it has to be consistent with their prior views (Kunda, 1990). Combining insights from these two lines of research, we hypothesize that negative frames emphasizing fiscal costs of immigration are more consequential than positive frames that emphasize fiscal benefits, but this effect is primarily visible among those whose ideological priors align with the negative information. Since motivated reasoning requires underlying views to be easily accessible, those who occupy a more extreme ideological position should be particularly affected.

We test our argument using a survey experiment on over 4,000 individuals in Germany, where concerns about immigration became particularly amplified following the 2015 refugee crisis. Our analysis demonstrates that negative framing of immigration strengthens welfare chauvinist attitudes, but only among those whose ideological priors are congruent with such information. In line with our hypotheses, positive framing of immigration is considerably weaker and inconsequential for attitudes about immigrants’ access to welfare benefits and services. The results also show that ideology, rather than economic circumstances, is a more important moderator of the framing effects. Our findings have implications for how partisanship influences welfare chauvinist attitudes while also indicating that framing effects are likely to be conditional.

Welfare Chauvinism

The term ‘welfare chauvinism’ was originally coined by Andersen and Bjørklund (1990) to describe the belief that immigrants should be prevented from receiving welfare benefits. The definition has been broadly accepted though some have clarified it to include restriction of welfare benefits and services to the majority population, whether that is defined in terms of ethnicity or nationality (Ennser-Jedenastik, 2018; Reeskens & van Oorschot, 2012).

A great deal of previous research has emphasised the central role of welfare chauvinism in the platforms of radical right parties (RRPs) in advanced and new market economies (e.g. Kitschelt & McGann, 1995; Afonso & Rennwald, 2018; Savage, 2022). Studies focused on the individual level have found that a considerable portion of citizens in advanced democracies are in favour of limited or conditional access to welfare provisions for immigrants (e.g. Ford, 2016; Kros & Coenders, 2019; Mewes and Mau, 2013). Welfare chauvinist attitudes are positively related to ethnic diversity (Quillian, 1995). For example, van der Meer and Reeskens (2021) demonstrate that individuals from more diverse neighbourhoods show more support for restricting immigrants’ access to welfare. More generally, ingroup biases tend to be stronger in more segregated environments with larger outgroups (Enos & Gidron, 2018).

Additional research has focused on the question of how welfare chauvinistic attitudes are formed. Several studies have theorized that welfare chauvinism arises as a result of resource competition (Ford, 2016; Hjorth, 2016; Kros & Coenders, 2019; Reeskens & van Oorschot, 2012). As such, welfare chauvinist attitudes have been shown to be more prevalent among low income groups and those experiencing economic insecurity. Others, however, emphasise the role of values over economic self-interest (Ford, 2016; Hainmueller & Hopkins, 2014; Hjorth, 2016; Marx & Naumann, 2018). These studies indicate that values and social identity trump objective economic indicators when determining who holds welfare chauvinist attitudes. For example, Hjorth (2016) shows that individuals who hold economically right-wing views are more likely to oppose welfare benefits to people from countries that are culturally distant from their own. Similarly, Ford’s (2016) research finds that ethnocentrism shapes perceptions of deservingness and welfare chauvinist attitudes.

The research, to date, has shown that welfare chauvinism can have an impact on public opinion on social policy. We also know that welfare chauvinist attitudes have been linked to both the objective economic status of individuals as well as their subjective ideological beliefs concerning the economic or cultural threat that they believe immigrants pose to society. One question that remains is whether individuals’ welfare chauvinist attitudes can change when presented with new information concerning the impact of immigration in their country.

Information Frames and Welfare Chauvinist Attitudes

Are welfare chauvinist attitudes shaped by frames in the public discourse? A growing literature suggests that frames – which are typically a type of claim or argument to which respondents are exposed to – shape opinion by leading respondents to focus on considerations emphasized in the frame (Chong & Druckmann, 2007a; 2007b; Scheufele, 1999). While framing effects have been demonstrated across a range of policy domains and issues, the existing scholarship also suggests that not all frames are equally effective in shaping attitudes. Our theory builds on two insights of this literature. The first emphasizes the valence of the frames and the presence of negativity bias. The second underlines the importance of motivated reasoning in information processing.

Negativity Bias

Scholars interested in framing effects have started emphasizing the valence of the frames (positive vs. negative) only relatively recently (Boydstun et al., 2019). Instead, much of the research on framing effects focuses on issue frames. Such frames lead respondents to pay attention to particular aspects of an issue or policy (e.g. Nelson et al., 1997; Jacoby, 2000; Sniderman & Theriault, 2004). While this research has demonstrated that public opinion is affected by frames, it is often difficult to separate the effects of the content of a frame from the effects of valence alone. Yet, voters are frequently exposed to valenced frames that do not always emphasize different aspects of an issue, but rather contain information that stresses either positive or negative consequences of a particular policy. Information on the consequences of immigration policy in Germany helps to illustrate this point. Drawing on the same research, but choosing different assumptions, two large media outlets have published very different information about the fiscal impact of immigration. While one reported that on average a foreigner contributed EUR 3,300 more in taxes than they received in terms of state support (Sueddeutsche Zeitung, 27 November 2014), the other argued that each foreigner represents a net cost of EUR 1,800 (Bild, 1 February 2015). What is the effect of such conflicting information on public opinion?

Social psychologists have shown that negative frames tend to be more powerful than positive frames in shaping people’s judgments (Kahneman & Tversky, 1979; Tversky & Kahneman, 1981), that humans are predisposed to pay more attention to negative than positive information (Baumeister et al., 2001), and that negative information is more memorable and seen as more salient (Rozin & Royzman, 2001). This negativity bias is thought to operate automatically because it is an innate trait of our central nervous systems – it “provides an evolutionary advantage, as it is more critical for survival” (Norris, 2021: 68). This presence of negativity bias is also increasingly acknowledged in political science. Several recent studies, focused on issues such as, evaluation of presidential candidates, political campaigns, perceptions of the economy, government approval, and support for the welfare state (Holbrook et al., 2001; Meffert et al., 2006; Soroka, 2014; Avdagic & Savage, 2021) confirm that negative information carries more weight than positive information in democratic politics. These insights help to develop our first hypothesis:

H1: The effects of negative framing of the fiscal impact of immigration on the perceptions of the deservingness of immigrants are stronger than the effects of positive framing.

We expect that, compared to the control group, respondents exposed to the negative frame will be less supportive of granting immigrants the same access to benefits and services as the natives. The mention of the net costs of immigration is likely to trigger concerns about fairness, reduce solidarity towards immigrants, or even raise concerns about a possible aggregate reduction in welfare provision. The positive frame, at the same time, is unlikely to have the opposite effect. As outlined above, negative information tends to be more powerful, and thus more likely to affect the opinion than positive information. In addition, given the salience of immigration, respondents have likely had some previous exposure to public discourse on immigration. Recent research in social psychology tells us that negative information stays longer in people’s minds and that shifting the opinion of those previously exposed to negative information is more difficult (Boydstun et al., 2019; Sparks & Ledgerwood, 2017). Given that a portion of respondents in the positive frame group would have likely been exposed to previous negative information about immigrants, this would likely make it more difficult for the positive frame to have a significant effect.

Motivated Reasoning

We argued above that the valence of a frame matters, but it is also important to recognise that frames do not always shape opinion because individuals typically do not process new information in a rational and unbiased way (Druckman & Bolsen, 2011). Existing values and attitudes tend to moderate framing effects. As Lakoff argues, “People think in frames… To be accepted, the truth must fit people’s frames” (2004:17). In other words, when faced with new information, individuals tend to interpret it considering their existing values and attitudes. Typically, individuals engage in motivated reasoning, which means that they are more likely to accept or seek out information that is consistent with their prior views, irrespective of whether that information is accurate (Kunda, 1990). While this is mostly done automatically and subconsciously, it is sometimes consciously recognised (Lodge & Taber, 2013). Studies have demonstrated the presence of partisan motivated reasoning in areas such as evaluations of politicians, perceptions of public support for particular policies, evaluations of policy outcomes, and redistribution preferences (e.g. Goren, 2002; Nir, 2011; McCabe, 2016; Savage, 2020).

While a range of pre-existing values and attitudes affects information processing, political ideology is a relevant proxy of political attitudes in general, and views about immigration in particular. Political ideology has been associated with the psychological need to manage uncertainty and threat, with right-wing individuals typically reporting less tolerance for uncertainty and more concern with threat (Jost et al., 2003). Behavioural research shows that individuals of right-wing orientation display significantly higher physiological sensitivity to threat than those of left-wing orientation (Oxley et al., 2008). These proclivities towards threat and uncertainty have been linked to preferences about inequality, with right-wing individuals being less critical of inequality and more likely to embrace the idea that not all groups are equally deserving (Jost & Amodio, 2012). As immigration is often portrayed as a threat, this is likely to trigger considerations about deservingness of immigrants particularly among right-wing individuals.

A growing body of evidence shows that exposure to ideologically congenial information reduces perceptions that the information is biased (Kelly, 2019). Similarly, Lodge and Taber (2006) show that ideology leads to selective information processing, which is driven largely by automatic affective processes. For most people, salient socio-political concepts – such as immigration – act as “hot cognitions”, automatically motivating ideological or partisan “goals that drive normatively suspect selectivity in information processing” (Lodge & Taber, 2006: 756). Thus, frames that are in line with one’s political ideology or partisanship are likely to be more effective in activating and strengthening the already existing views. Hence, we expect Hypothesis 1, which emphasizes the negativity bias effect, to be conditional on respondents’ ideological leaning. In other words:

H2: While generally negative frames about the fiscal impact of immigration are likely to be more consequential than positive frames, their effect should be visible primarily for those whose ideological priors are congruent with such frames.

Although the congruence of frames and attitudes is important, it is reasonable to expect that the extremity of attitudes also matters. For motivated reasoning to occur, the underlying views and traits must be easily accessible and retrievable from memory (Aldrich et al., 1989). Research in political cognition suggests that the accessibility of attitudes depends on the extremity of the individual’s position. More extreme attitudes are associated with increased selective judgement (Lavine et al., 2000; Pomerantz et al., 1995) as such individuals are more susceptible to disregarding information that clashes with their position. Hence, we also expect an attitude strength effect, such that those citizens who are more extreme ideologically will be more prone to motivated reasoning. Combining insights from the discussion about negativity bias, motivated reasoning and attitude extremity, we arrive at our final hypothesis:

H3: The effect of negative frames should be most evident among more extreme ideologues who hold frame-congruent attitudes.

A Survey Experiment of Welfare Chauvinist Attitudes

We test our hypotheses using a survey experiment in Germany, a country that represents a ‘hard’ case for our theory. Prior research has shown that welfare chauvinist attitudes are more likely to be found in states where welfare benefits are distributed on the basis of need rather than equity or equality. Social insurance-based systems, such as Germany’s, are less likely to elicit welfare chauvinist attitudes (Ennser-Jedenastik, 2018; Reeskens & Van Oorschot, 2013). The experiment was carried out by YouGov and was inserted into their regular political omnibus survey.Footnote 1 The sample of 4,158 respondents was drawn from a panel of 320,000 individuals and is weighted by age, region, gender, education, political interest, and voting behaviour at the last election to ensure that it is nationally representative.Footnote 2

All individuals in the survey (including the control group) were given the following initial prompt to get them thinking about the way the welfare state is funded and the relationship between taxation and spending.

(Priming information): The government provides a range of social benefits and services to address the needs associated with unemployment, sickness, education, housing, family circumstances, and retirement. Such benefits and services are financed through taxation and national insurance and all legal residents in Germany are entitled to receive them. To spend more on social benefits and services, the government may need to increase taxes and national insurance contributions.

Respondents were then randomized into three groups: the control group, which received no additional information; the negative frame group, and the positive frame group. The negative frame group received information suggesting that immigrants have deleterious effect on public finances:

(Negative frame): Because immigrants are also entitled to receive social benefits and use public services, the economic implications of immigration are an increasing concern. Recent research shows that immigration is a drain on government finances – on average, immigrants take out significantly more from the welfare state in social benefits and services than they contribute in taxes and national insurance.

Respondents in the positive frame group received information suggesting that immigrants are net contributors to public finances:

(Positive frame): Because immigrants are also entitled to receive social benefits and use public services, the economic implications of immigration are an increasing concern. However, recent research shows that immigration is in fact a boost to government finances – on average, immigrants contribute significantly more to the welfare state in taxes and national insurance than they take out in social benefits and services.

Respondents were then asked the question regarding their preferences for immigrants’ access to welfare benefits. Our focus on immigrants’ access to gauge welfare chauvinism is in line with most individual-level studies on the subject. There are, however, studies (particularly at the party-level) that adopt a broader operationalisation of welfare chauvinism, encompassing not only attitudes towards immigrants’ access to welfare provision, but also support for welfare for the ingroup (see Careja & Harris, 2022).

Using a 5-point scale, with responses ranging from ‘strongly disagree’ to ‘strongly agree’, respondents were asked to what extent they agree or disagree with the following statement:

The government should guarantee that immigrants have the same rights to social benefits as German citizens.

Randomization of the treatment groups ensures that they are almost identicalFootnote 3 in all respects in terms of observable and unobservable variables that may confound cross-group comparison. The groups only differ with respect to the information that they received. One limitation is the absence of information about respondents’ pre-existing preferences about welfare in general. While support for welfare in Germany tends to be high, it is possible that some respondents oppose welfare entirely as a matter of principle. It is potentially ambiguous how those individuals would respond to the question about immigrant access.

We also hypothesize that the negative treatment will be more likely to elicit welfare chauvinist attitudes among those who are likely to be predisposed to such views. To test this, we use two indicators of respondents’ ideological disposition. The first is an indicator of which party the respondent voted for at the previous parliamentary election. The second variable asks individuals to place themselves on an ideological scale ranging from 1 to 7, where 1 indicates ‘very left-wing’ and 7 indicates ‘very right-wing’.

Results

To aid interpretation, we use Ordinary Least Squares regression to estimate the hypothesized framing effects. The online appendix shows that the results remain substantively the same when using ordered logit models (Table A4). As our sample is weighted to be representative and individuals are randomized into treatment groups, spurious correlation is unlikely to be a problem in our models. We therefore follow the recommendations of Mutz (2011) and do not include socio-demographic controls in order to keep our models as simple as possible. Table 1 shows the distribution of responses to the dependent variable question (see also Table A1 in the appendix).

Table 1 The government should guarantee that immigrants have the same rights to social benefits as German citizens
Full size table

Figure 1 shows the direct effect of our experimental treatments (see also Table A3 in the appendix). Those exposed to the negative framing information are less likely to support granting the same welfare rights to immigrants as German citizens when compared to the control group. Similarly, individuals in the positive frame group are more likely to support giving immigrants equal access to welfare. However, the effects of both frames are not statistically significant at conventional levels. The effect of the negative frame (p = 0.08) is greater, as we hypothesize, shifting a respondent’s likelihood of supporting equal access to welfare benefits by almost one-tenth of a point on the response scale; an effect that is more than four times greater than that of the positive frame group (p = 0.7). Although the lack of statistical significance does not allow us to confirm H1, the direction of this result is in line with prior research which finds a similarly stronger effect of negative frames compared to positive frames (Avdagic & Savage, 2021; Soroka, 2014). One caveat must be noted here: both frames share the same first sentence, which suggests that “the economic implications of immigration are an increasing concern”. Although this information is rebutted in the next sentence of the positive treatment, this could have contributed to the weaker results of this frame. To explore this possibility, we re-ran the experiment using a more neutral wording including “increasingly debated” instead of “increasing concern”. We find no significant difference between the two wordings of the positive frame (see Appendix p.18). However, we cannot fully exclude the possibility that even this more neutral version may not have raised doubt about immigrants’ entitlement among respondents who received the positive frame.

Fig. 1
figure 1

Marginal effect of framing on support for equal access to welfare for immigrants

Full size image

Our second hypothesis is that framing is more likely to be consequential for individuals of certain ideological profiles. Specifically, we expect supporters of right-wing parties to be more susceptible to negative framing than supporters of left-wing parties. A more restrictive stance towards immigration is a staple of right-wing party platforms across Europe, particularly among parties of the radical right (Mudde, 2007). Over the past decade, we have also seen mainstream right-wing parties shift their positions on immigration towards those of radical right parties in an attempt to accommodate such views among the electorate (Meyer & Wagner, 2013). As Marx and Naumann (2018) have shown, the 2015 refugee crisis in Germany saw the CDU/CSU shift towards a more restrictive position on immigration.

Figure 2 shows that both the positive and negative framing function as expected in almost every case, but these effects are not significant for supporters of the SPD, Greens, or FDP. As expected, the negative framing is most potent among CDU/CSU supporters, reducing support for greater welfare access for immigrants by around a fifth of a point (p = 0.04). This is not surprising as supporters of centre-right parties are likely to be ideologically moderate compared to radical right supporters. CDU/CSU supporters may be initially less committed to a more hard-line stance on immigrant access, which makes them more open to persuasion. By contrast, AfD supporters are likely to already be committed to the idea that immigrants should not have equal access.

Fig. 2
figure 2

The marginal effect of framing on support for equal access to welfare for immigrants, conditional on partisanship

Full size image

There are two other notable, and curious, significant findings. The negative framing shifts support for a less restrictive immigration policy by almost half a point (p = 0.01) among Die Linke supporters, while the positive framing shifts support for a more restrictive policy, again, by half a point (p = 0.04) among AfD supporters.Footnote 4 One explanation for these counterintuitive results is that they are examples of a ‘boomerang effect’ of partisan motivated reasoning. A boomerang effect occurs when a message produces the opposite effect to that intended due to the way in which the message is processed by individuals (Bayes & Druckman, 2021; Hart & Nisbet, 2012). The AfD is renowned for promoting a restrictive approach to immigration, while Die Linke has traditionally favoured an open borders policy. It could therefore be the case that exposing supporters of these parties to information that challenges their policy preferences triggers a partisan-motivated inverse response as they seek to maintain their partisan identities (Leeper & Slothuus, 2014). For Die Linke supporters, who instinctively favour a liberal immigration policy, receiving information which shows that that policy has negative effects may trigger a defensive increase in support for their prior position.

Our final hypothesis states that negative framing has a greater impact on individuals who are more ideologically extreme and who hold congruent positions on immigrant access. We test this hypothesis using a seven-point scale of ideological self-placement, which is commonly used to measure ideological extremity (Devine, 2015; Mason, 2018).Footnote 5 Figure 3 shows that the positive frame has no significant effect on attitudes about immigrant access. Among those on the left of the spectrum there is a positive effect, meaning that individuals are more likely to support equal access for immigrants. But this effect does not meet conventional levels of statistical significance. Conversely, the negative framing does significantly shift the views of individuals towards a welfare chauvinistic stance for respondents on the right of the political spectrum. This effect increases the further to the right an individual lies on the scale. The marginal effect for those on the furthest right of the ideological spectrum is shift of a third of a point (p = 0.01) towards more welfare chauvinistic policy preferences. To test the robustness of the shape of this relationship, we split respondents into three groups (Left, Centre, and Right). Figure A1 of the online appendix shows that, as expected, the negative framing has a greater effect on those that are placed into the right-wing group compared to both the center and left-wing groups.

Fig. 3
figure 3

The marginal effect of framing on support for equal access to welfare for immigrants, conditional on an individual’s ideological self-placement

Full size image

Taken together, these results provide some support for the argument that the framing of immigrants’ impact on public finances can influence attitudes towards welfare policy. However, it is not straightforward. On their own, frames do not have a significant effect on welfare attitudes. But when combined with an individual’s ideological and partisan orientation, negative framing in particular can engender chauvinistic welfare attitudes. As we have argued, this should be expected as partisan motivated reasoning makes some individuals more receptive to information that is congruent with their prior beliefs, especially those further to the right on the political spectrum (Lavine et al., 2000).

One may question, however, if motivated reasoning is the only information processing framework that can explain our results. Bayesian updating, as shown by the recent literature (Druckman & McGrath, 2019; Little et al., 2022; Coppock, 2023), is compatible with many of the findings attributed to motivated reasoning, where there is heterogeneity of prior attitudes and of beliefs about what information is credible. Bayesian learning assumes that posterior attitudes reflect a combination of new information and the prior attitudes, but new information is effective only if it is perceived as credible. Because “the very sources people find credible are the ones with whom they share common beliefs” (Druckman & McGrath, 2019: 114), Bayesian updating is in principle not incompatible with our general findings. However, we find no clear support for Coppock’s (2023) Bayesian-inspired argument that information pushes all respondents in the same direction. The difference between left and right-leaning individuals in our analysis is not only in the magnitude of the effects, but their responses seem qualitatively different. Although Fig. 2 suggests that the negative frame pushes both CDU and SPD voters in the same direction (albeit the difference is not significant for SPD voters), we do find differences among parties and some evidence of backlash or boomerang effects. Moreover, looking at the ideological scale, left- and right-leaning individuals do not respond in the same way to the negative information (see Figures A1 and A2 in the appendix). While those further to the right grow more supportive of welfare chauvinist attitudes, those on the left move in the opposite direction (although the difference from the control group in this case is not statistically significant).

On the whole, therefore, our findings are in line with recent experimental research that emphasises the role of ideology as a moderator of framing effects in other policy areas. Partisan bias in citizens’ response to frames has been found in research ranging from the elderly care issues and trade policy in Denmark (Slothuus & de Vreese (2010), to Obama’s healthcare reform (McCabe, 2016) and the attribution of responsibility for the economic crisis in the UK (Bisgaard, 2015). However, other research on the subject does not find ideology to be a moderator of framing effects. Negative framing of immigration has been found to reduce support for welfare spending irrespective of ideological priors (Avdagic & Savage, 2021). Similarly, recent research shows that most individuals, regardless of their partisanship, believe that natives are more deserving of welfare benefits than immigrants (Magni, 2022). These differences are likely attributable to variation in substantive focus and methodological approach. For example, Avdagic and Savage (2021) focus on general support for welfare spending, rather than welfare chauvinism. Magni (2022), who focuses on deservingness of migrants, relies on a conjoint experiment requiring choices between different profiles of individuals rather than a framing experiment. In addition, it is possible that that framing effects may be inconsistent across issue domains for parties of differing ideological persuasion.

More generally, our results reflect changes in the nature of party competition in many European countries. Recent studies have found that centre-right parties have adjusted their policy positions on immigration and the economy in the direction of the platforms espoused by the radical right (Abou-Chadi & Krause, 2018; Chueri, 2023; Fischer & Giuliani, 2023). Although this is not always electorally successful at winning back radical right voters (Abou-Chadi et al., 2022), supporters of mainstream center-right parties, like the CDU, may find themselves committed to these more extreme policies due to partisan motivated reasoning. In Germany specifically, there was a general movement towards more restrictive immigration policy among parties following the migration crisis, but this was more notable among right-wing parties such as the CDU/CSU (Marx & Naumann, 2018). Our results therefore have implications for how far welfare chauvinism may spread among the electorate. If partisan motivated reasoning is as effective as we have shown in this paper, then parties have considerable potential to influence voters via their programmatic shifts. The mainstreaming of more restrictive immigration policies by the CDU/CSU in the aftermath of the migration crisis did appear to make their voters more receptive to negative messages about migrants, though in recent years, several democratic countries have experienced rising support for immigration (Ford & Morris, 2022; Mutz, 2018). It may be that a shift by mainstream parties back towards a more liberal migration policy could weaken the effects of negative frames that we find. Our results therefore suggest that rising welfare chauvinism is, to a significant degree, a response to parties rather than a more fundamental bottom-up change in individual attitudes.

Alternative Explanations

Our hypotheses suggest that negative frames are more effective, but they need to be ideologically congruent to have an effect. In line with the literature on partisan motivated reasoning (Bisgaard, 2015; Bolsen et al., 2014; Druckman et al., 2013; Petersen et al., 2013), we argue that individuals are more receptive of information that reinforces their political identity. Yet, research on issue motivated reasoning (Lodge & Taber, 2013) suggests that political ideology is not the only factor that shapes preferences. Self-interest is another consideration in the formation of preferences that may affect how information is received (Bolsen & Leeper, 2013). Specifically, personal economic circumstances may influence respondents’ attitudes on immigrant access, and thus moderate the effect of frames. Consider a left-leaning individual who is facing dire economic circumstances. Would they support equal access for immigrants in line with their political identity, or oppose it for fear of welfare competition? As Mullinix (2016) argues, it is likely that both types of processes – political identity and issue motivated reasoning – are often simultaneously at work. If our argument is correct, including self-interest considerations will not affect the results about the importance of one’s ideological orientation.

We consider several indicators of an individual’s economic circumstances that should capture self-interest considerations. These include standard indicators, such as the household income, as well as social grade, which captures the occupational make up that may affect immigration attitudes. In addition, we consider respondents’ perceptions about their own job security and their expectations about their household’s financial position in the next twelve months. These tap into any potential concerns that immigration may heighten competition over jobs, benefits and public services.

The full results of these models are presented in Table A5a in the appendix. In all the models the interaction between the negative treatment and ideology remains significant, confirming our hypotheses. Simultaneously, as Fig. 4 shows, there is no evidence that self-interest considerations moderate the effect of the frames as none of the indicators we consider influence the effectiveness of the frames in a consistent manner. While low-income individuals seem susceptible to the negative framing, this effect if not statistically significant. Social grade also does not seem to moderate the effects of the frames. Beliefs about job prospects do not moderate the framing effects in a consistent manner. While the positive framing has no influence, the negative framing affects those who are only a little sure about their ability to keep their jobs, as well as those who are not currently employed. However, this effect is not evident among those who have no confidence in being able to keep their job. Expectations about the financial situation of the respondents’ households offer no clear predictions. Paradoxically, at a first glance it appears that the negative treatment reduces support for equal access among those who expect their situation to remain the same or improve, but this effect is not statistically significant. As a robustness check and to reduce the number of interactions, we ran separate models in which job and household financial prospects are treated as continuous rather than categorical variables, and social grade is operationalised as a variable with only four categories (Table A5b). The interaction between the negative treatment and job prospects now appears statistically significant, suggesting that the negative treatment has a strong effect among those facing worse job prospects or presently not having a job. However, our main results remain unaffected as the interaction between the negative treatment and ideology remains significant in all models. On the whole, our results suggest that for the most part ideology is a more important moderator of the framing effects than one’s economic circumstances.

Fig. 4
figure 4

(A) The marginal effect of framing on support for equal access to welfare for immigrants – alternative explanations; (B) Social grade: 1 = Higher technical, 2 = Manager/senior administrator, 3 = Junior management, 4 = Sales/service sector, 5 = Foreman/supervisor, 6 = Skilled manual, 7 = Semi-skilled/unskilled, 8 = Miscellaneous, 9 = Never been employed; (C) “How confident are you that you will keep your current job for the next 12 months?”; (D) “How do you think the financial situation of your household will change in the next 12 months? It will…”

Full size image

Conclusion

In this paper, we have brought together two distinct literatures on the effects of information on political behavior – negativity bias and motivated reasoning – to examine how the interplay between valenced frames and ideological bias affect public opinion on immigrant access to welfare provision. We know from existing research that negative messages are more influential than positive messages (Kahneman & Tversky, 1979; Norris, 2021), and we also know that citizens are more receptive of information that is in line with their partisan views (e.g. Slothuus & de Vreese, 2010). Yet, prior research has not fully illuminated how ideological priors influence the receptivity of valenced frames. In doing this, we contribute to the literature on framing in general and on welfare chauvinism in particular.

Our findings demonstrate that negative frames are indeed more powerful, but the effects of these frames are visible primarily among those whose ideological priors align with this information, and particularly among those who occupy a more extreme ideological position. This contrasts with the literature that finds more general effects of negative information on policy attitudes (e.g. Avdagic & Savage, 2021; Soroka, 2014), suggesting that the effects of frames may vary from issue to issue and even within specific aspects of policy domains. For example, limiting immigrants’ access to welfare provision may be seen as a policy that concerns primarily those on the right. By contrast, support for general welfare spending is an issue that affects all individuals, so responses to negative information may be less influenced by motivated reasoning. Some recent contributions to the framing literature have identified personal salience of issues, the type of issues and the role of political awareness (Mullinix, 2016; Slothuus & de Vreese, 2010) as factors that determine the extent to which partisan motivated reasoning shapes framing effects. However, further research is required to identify systematically the general conditions under which the receptivity to valenced frames is affected by motivated reasoning and partisan bias.

Our findings also have implications for research on welfare chauvinism. First, while much of this literature focuses on individual determinants of welfare chauvinist attitudes, such as socio-economic characteristics (e.g. Mewes & Mau, 2012) or cultural identity (e.g. Hjorth, 2016), we contribute by examining whether messaging can generate or attenuate these attitudes. We show that negative messaging about immigration does not seem to induce welfare chauvinism for all individuals or even for most of those experiencing economic insecurity. Instead, it is primarily those further to the right on the political spectrum who are receptive of such information. Second, our findings contrast with recent research on attitudes towards deservingness of immigrants vis-à-vis natives, which suggests that ideology has little influence and that conservatives and liberals alike penalise immigrants (Magni, 2022). We suspect that this contrast reflects the fact that our framing captures the general political competition over immigration, while Magni’s reliance on a conjoint experiment with forced choices between particular types of immigrants and natives offers more granular information that may not trigger equally strong partisan considerations. A fruitful avenue for further research, therefore, would be to explore more systematically if and why ideology matters more in the assessment of general policies than of specific characteristics of individual immigrant profiles.

Overall, our results demonstrate that framing can be influential under some circumstances, but there are still questions that are beyond the scope of this current project. The first is that we do not know how persistent the effects of negative framing are over time. Lack of funding meant that we were unable to follow-up our respondents to see if any attitudinal changes had persisted. However, as prior research has shown, negative information is likely to be more memorable and more powerful so we have reason to believe that our results would be persistent. The second question is how respondents react to immigrants from different ethnic groups – do their perceptions of whether benefits should be restricted depend on the ethnicity of the potential recipients? This is a pertinent question as European countries are currently receiving a large number of refugees as a result of Russia’s invasion of Ukraine, a predominantly white, Christian country. Are migrants from Ukraine seen as more entitled to welfare benefits than those who arrived primarily from North Africa and the Middle East during the 2015 migration crisis? How negative or positive information is interpreted by individuals could possibly be affected by their conceptions of what a typical migrant looks like at any given time.

Those questions aside, it is evident that negative information about immigration can harden attitudes about immigrant access to welfare. Given that the tone of media coverage of immigration is largely negative across European states, this presents a challenge to those on the pro-immigration side of the debate. Our results suggest that one strategy may be to focus less on converting those who are already hostile to immigration, and instead concentrate on solidifying the support of individuals who are less susceptible to anti-immigration messaging. Emphasizing specific social and economic problems that immigration can help withFootnote 6, rather than general fiscal gains from immigration, may be a way to try and counter the effects of the negativity bias.