Do local circumstances affect attitudes towards local social policies? Exploring the role of economic, political and demographic conditions

Abstract

Purpose

This article aims to explore whether and how economic, political and demographic municipal conditions shape citizens' attitudes regarding decentralised social policies.

Design/methodology/approach

The authors analysed the 2018 wave of the Dutch Local Election Studies, which includes a novel survey item asking respondents whether they prefer local social policies to be primarily: (1) protection-based, (2) cohesion-building or (3) activation-based. The authors appended context indicators to that survey and performed multilevel logistic regression analyses (1,913 respondents nested in 336 municipalities).

Findings

At the individual level, these preferences are affected by gender, age, income, education and political inclination, as expected. However, preferences towards local social policies are not shaped by local economic, demographic or political conditions. The authors discuss the implications of these findings for future research.

Originality/value

By using unique data, including a newly developed survey item, this study is the first to explore whether and how municipal conditions shape preferences regarding local welfare. Understanding those preferences is increasingly important as many Western European countries have decentralised swathes of social policies from the national to the local level in recent decades.

Citation

Broekema, B., Fenger, M. and van der Waal, J. (2021), "Do local circumstances affect attitudes towards local social policies? Exploring the role of economic, political and demographic conditions", International Journal of Sociology and Social Policy, Vol. 41 No. 9/10, pp. 993-1011. https://doi.org/10.1108/IJSSP-07-2020-0300

Introduction

Many Western European countries have decentralised swathes of social policies from the national to the local level in recent decades (Minas et al., 2012), such as activation-based labour market policies (Fuertes et al., 2020; Jacobsson et al., 2017), long-term care (Dijkhoff, 2014) and youth care (Skamnakis, 2016). Therefore, some of the responsibilities for welfare policy development and implementation have shifted from national to local levels of authority (Ansell and Gingrich, 2003; Kazepov and Barberis, 2013; Van Berkel, 2006). This decentralisation is often justified by pointing to a closer fit between policy measures on the one hand and, on the other, the variation in problems, circumstances and popular preferences that these policies aim to address (Costa-Font and Greer, 2013). The assumption that local conditions affect welfare attitudes towards these new decentralised social policies, however, remains largely understudied.

Existing studies argue that welfare attitudes are shaped by macro-contextual factors, because people are influenced by the cues derived from the environment in which they live (Svallfors, 2012; Van der Waal et al., 2013). As a result, the expectation is that individuals with similar characteristics may have different welfare attitudes depending on their country of residence (Sevä, 2009). This has been corroborated empirically in dozens of cross-national studies focussing on the role of context conditions – including welfare institutions, economic performance and political discourse – in shaping welfare attitudes (e.g. Gelissen, 2000; Larsen, 2008; Sevä, 2009). Consequently, a great deal is known about cross-national variations in welfare attitudes and how these can be explained. However, the same cannot be said about cross-local variations in welfare attitudes towards local welfare.

The limited number of studies that do assess how local context conditions shape welfare attitudes focus on attitudes towards the national welfare system, finding that economic conditions, political discourse, ethnic diversity and urbanisation “clearly do have an effect on the formation of people's attitudes towards aspects of the [national] welfare system” (Blomberg and Kroll, 1999, p. 331; cf. Sevä, 2009). We use those insights as a stepping-stone for exploring whether local circumstances also shape attitudes towards local welfare policies. We do this by focussing on the Dutch case, a pertinent example of large-scale decentralisation, where, in 2015, municipalities became authorised to set their own social welfare policy aims (Vermeulen, 2017).

Specifically, we ask: Do local conditions shape welfare attitudes towards decentralised social policies? Using a unique data set – the Dutch Local Election Studies 2018, which measured citizens' orientations towards local welfare policies across 380 Dutch municipalities – our study is the first to explore whether local circumstances shape attitudes towards local welfare policies.

This study’s primary aim is to address the apparent oversight of local welfare policy in research on welfare attitudes. In doing this, it makes two substantive contributions. First, it transcends a national bias by focussing on attitudes towards local social policies. Second, from a policy perspective, it seeks to shed light on political views at the local level, as Dutch municipal authorities acquired the legal responsibility to consider political preferences as much as local conditions when making important decentralised policy decisions (Ministry of Social Affairs, 2020).

The paper begins by introducing our tailor-made item for measuring attitudes towards decentralised social policies. Building on insights derived from earlier studies on welfare attitudes, we attempt to identify potentially relevant municipal conditions that may shape citizens’ attitudes towards decentralised social policies. Four of these conditions guide our empirical analysis. We conclude with a critical reflection and discussion of the implications of our findings.

Measuring attitudes towards decentralised social policies

To explore the impact of local circumstances on citizens' attitudes towards local social policies, we focus on the basic question: what type of primary aim of such policies do they prefer? The formulation of the response categories was inspired by research that focussed on people’s redistribution preferences and views of government responsibilities (e.g. Svallfors, 2004; Taylor-Gooby, 2001). Broadly speaking, these studies have identified three types of attitudes towards welfare: social-democratic, Christian-democratic and liberal. Those with social-democratic attitudes support extensive state intervention and generous welfare arrangements (Staerkle et al., 2012; Svallfors, 2012; Taylor-Gooby, 2001). This support is weaker among those with Christian-democratic welfare attitudes, who prefer social cohesion and “sources of [welfare] support conceived in a hierarchy of family, firm and state” (Taylor-Gooby, 2001, p. 137). Respondents with liberal welfare attitudes, meanwhile, have the lowest levels of support for redistribution and state intervention, but the strongest for activation-based welfare policies (Staerkle et al., 2012; Svallfors, 2012).

Inspired by the three types of attitude mentioned earlier, we developed a new survey item to measure people's attitudes towards decentralised social policies. To fit this to the local social policy context, we reformulated the type of item often used to measure citizens’ attitudes towards the welfare responsibilities of national governments (for an overview of such items, see Svallfors, 2004, p. 124) [1]. This question asked respondents what they believe should be the primary aim of local social policies, with the response categories discerning the social-democratic, Christian-democratic and liberal takes on the issue. The social-democratic take on the primary aim of local social policy in this study was that citizens ought to be able to appeal to the municipality for problems as the primary (first and foremost) provider of protection for its citizens through policies (Kautto, 2002). Christian-democratic policies support the family and the gendered division of paid and unpaid labour, as well as emphasising the family and surrounding community as the primary social service provider (Seeleib-Kaiser et al., 2005). To enhance this role of the family and community, the Christian-democratic take on the primary aim of local social policies in this study focussed on cohesion-building through connecting members of the society. The liberal stance on welfare policies supports activation-based welfare policy aims (Staerkle et al., 2012; Svallfors, 2012). Therefore, the liberal take on the primary aim of local social policies advocated facilitating citizens to handle their problems themselves by means of activation. We describe the corresponding three types of policy aim as: (1) protection-based local social policies; (2) cohesion-building local social policies; and (3) activation-based local social policies.

In the next section, we theorise how citizens’ attitudes towards decentralised local social policies are possibly shaped by municipal conditions.

How do municipal conditions shape attitudes towards decentralised social policies?

At the individual level, the notion of self-interest implies that various characteristics that pose a greater-than-average risk for facing social or economic problems will influence attitudes towards the role of the state in welfare issues (Kangas, 1997). This needs to be accounted for in our analyses focussed on exploring the role of local conditions in shaping those attitudes.

The general idea is that individuals are risk-averse and thus, when experiencing increasing vulnerability and facing greater risks of becoming dependent of the welfare system, grow more supportive of intensified assistance and less supportive of individual responsibility (Blekesaune and Quadagno, 2003; Hasenfeld and Rafferty, 1989). Such statements can be found in classical readings by Smith (1976) and Downs (1957), and numerous articles presented strong empirical support for these arguments (e.g. Blekesaune, 2007; Taylor-Gooby, 2001). Therefore, it is expected that women (Rantanen et al., 2015), the less educated (Linos and West, 2003), the unemployed and people with low incomes (Bean and Papadakis, 1998) are more supportive of extensive welfare policies.

At the contextual level, informed by the wider welfare attitudes literature, we focus on four types of context to explore how citizen’s attitudes towards decentralised social policies can be shaped by municipal conditions: (1) economic; (2) political; (3) ethnic; and (4) urbanisation.

Economic

One contextual factor frequently linked to welfare attitudes is economic circumstances and in particular the unemployment rate (e.g. Blekesaune, 2007; Blomberg and Kroll, 1999). These conditions can affect attitudes towards local welfare via two mechanisms: sociotropic concerns or self-interest.

According to the sociotropic mechanism, public concern for those in economically precarious positions is greater in areas that are doing worse economically (Blekesaune and Quadagno, 2003). The underlying notion is that being confronted with unemployed friends, relatives or fellow local citizens on a regular basis breeds solidarity with them (Kiewiet and Lewis-Back, 2011). It is also believed that such sociotropic concerns breed higher levels of solidarity with the unemployed among the wider public (Staerkle et al., 2012; Van Oorschot and Meuleman, 2012). If this is correct, the first expectation is that the support for protection-based social policies is higher in municipalities with higher rates of unemployment.

According to the self-interest mechanism, welfare attitudes are influenced differently by unemployment rates (Blekesaune and Quadagno, 2003), as not everyone is affected equally by high unemployment and unemployment-related risks. As these risks vary between societal groups, already existing conflicting welfare interests are further reinforced where the unemployment rate is high. In other words, those in a poorer labour-market position have an even greater risk of becoming unemployed, while those who are unemployed face even worse job prospects (Blekesaune and Quadagno, 2003; Edlund, 2006). As a result, one would expect these people to favour protection-based social policies more in municipalities with high unemployment rates.

Those in a stronger labour-market position are likely to bear a higher financial burden in places that fare worse economically, as they are the people who, in the main, must contribute more to the welfare system (Hasenfeld and Rafferty, 1989). Consequently, this is likely to further reduce their already low support for protection-based social policies (Linos and West, 2003). In line with this argument, it might be expected that the negative effect of income on support for protection-based social policies is stronger in municipalities with higher unemployment rates and that the positive effect of unemployment on support for protection-based social policies is stronger in municipalities with higher unemployment rates.

Political

It is argued, and has been found, that attitudes and preferences are influenced by our interactions with those around us (Blomberg and Kroll, 1999; Miller, 1978). These interactions can both weaken and strengthen ideological differences, depending on how we respond to dissimilar others (Miller, 1978). Consequently, two mechanisms are discussed here: consensual and reactive (Miller, 1978).

The consensualism mechanism is derived from contact theory (Campbell, 2006), which suggests that contact between ideological and political adversaries produces dialogues. These dialogues, in turn, lead to information elaboration, mutual understanding, tolerance and, by extension, consensual tendencies (Miller, 1978). In other words, “people are influenced towards agreement with their contacts, whether or not those contacts have similar characteristics to themselves” (Miller, 1978, p. 265). Previous studies do indeed show that contact with political opponents breeds more agreement in terms of political ideology (e.g. Campbell, 2006). Using this insight as a stepping-stone for theorising about the role of local political climates in shaping welfare attitudes, we expect views on decentralised social policies to be influenced by the attitudes of the majority on these issues and that people are moved into agreement with each other. The first scenario, therefore, states that: (1) non-social democrats are more in favour of protection-based local social policies in social-democratic municipalities; (2) non-Christian democrats are more in favour of cohesion-building local social policies in Christian-democratic municipalities; and (3) non-liberals are more in favour of activation-based local social policies in liberal municipalities.

The reactive mechanism is derived from the group threat theory. In contrast to the consensualism mechanism, it assumes that people will be antagonised by confrontation with dissimilar others (Andrews, 2011; Miller, 1978; Scala and Johnson, 2017) and that frequent contact with political opponents will reinforce our ideological and political views (Andrews, 2011). In the Big Sort, for example, Bishop (2009) demonstrates that bringing different groups together does not instantly foster mutual respect. On the contrary, political discussions can often lead to group polarisation. Glaser (1994) also found support for this threat theory, revealing that the larger the size of the “group” threatening our views and resources, the stronger our beliefs become. With this in mind, the opposite can be expected with regard to the consensualism mechanism, hence: (1) social democrats are more in favour of protection-based local social policies in non-social democratic municipalities; (2) Christian democrats are more in favour of cohesion-building local social policies in non-Christian democratic municipalities; and (3) liberals are more in favour of activation-based local social policies in non-liberal municipalities.

Ethnic

The third contextual indicator that might affect citizens' attitudes towards local social policies is ethnic diversity. In his article E pluribus Unum, Putnam proposes the constrict theory, arguing that such diversity in a residential setting reduces both in- and out-group solidarity (2007). As a result, those in ethnically diverse contexts tend to turn inwards, distrust people regardless of the colour of their skin and give less to charity, a phenomenon that Putnam calls “hunkering down” (Gijsberts et al., 2012; Putnam, 2007; Sturgis et al., 2011). Given this “hunkering down” thesis, it might be expected that views towards protection-based welfare policies are less positive when the share of ethnic minority residents is higher, as this makes people less willing to pay for social security in general. In other words, we expect the support for protection-based local social policies to be lower in municipalities with a larger share of ethnic minority citizens.

Urbanisation

The final contextual condition that may be relevant for the local welfare attitudes discussed here concerns the level of urbanisation (Kelly and Lobao, 2019; Sevä, 2009). Researchers argued more than 30 years ago that socialisation in rural areas affects attitudes in a different way to socialisation in urban areas (e.g. Davis, 1988; Scala and Johnson, 2017). Previous studies largely expected people living in rural areas to have more negative views towards welfare recipients (Camasso and Moore, 1985). Generally, rural socialisation: (1) teaches people to have a high regard for self-reliance and a strong work ethic, instilling the idea that a precarious position in life is the result of personal shortcomings (Osgood, 1977); and (2) emphasises “a sense of local responsibility for and control of the problem” (Davis, 1988, p. 70). Socialisation in urban environments, however, leads to “a greater tendency toward unconventional lifestyles, individualisation, and greater acceptance of stigmatised behaviour, less conservative attitudes, and more variance in family structure” (Rank and Hirschl, 1993, p. 608). Consequently, those in urban areas are more inclined to support welfare policies that enable individual self-actualisation, irrespective of labour-market participation and family obligations (Blomberg and Kroll, 1999). In line with this reasoning, we expect that: higher levels of urbanisation will reduce support for activation-based local social policies.

Data and methods

Data set

We explore aforementioned scenarios using multilevel modelling on a unique data set that combined individual-level data from the Dutch Local Election Studies, 2018 (DLES2018), and municipal-level data retrieved from the various sources outlined further. The DLES2018 data set was collected by the Longitudinal Internet Studies for the Social Sciences (LISS) panel and administered by CentERdata (Tilburg University, the Netherlands). The LISS is a long-running panel which accesses high-quality, random samples drawn from the Dutch population by Statistics Netherlands (CBS). In our study, respondents aged 18 years and above were randomly selected in 2018. Of the 3,380 individuals chosen, 2,704 (80%) replied and 2,652 completed the survey. The DLES2018 includes respondents from all 380 Dutch municipalities (for more information on the survey, see: Jansen and Denters, 2018, pp. 60–61).

Outcome measure

A three-category dependent variable – local social policy preference – indicates a respondent’s preferred primary aim of local social policies. This question was introduced and posed as follows:

Now we turn to a number of questions about municipal social policies, for example health care, tackling unemployment, and youth policy (…) We would like to know what you think about that. Social policy in my municipality must above all ensure that people….

1. Become more connected

2. Are better able to handle their problems themselves

3. Can appeal to the municipality for their problems [2].

Variables measuring local conditions

The unemployment rate was included as a continuous variable and ranged in 2017 from 2.9% (municipality: Kapelle) to 9.2% (municipality: Zevenaar) (Vereniging van Nederlandse Gemeenten, 2018). This shows the unemployed labour force as a share of the labour force overall (unemployed and employed). This is the standard indicator for measuring unemployment in the Netherlands, as applied by Statistics Netherlands (CBS, 2019d).

In order to measure political climate, we identified three types of municipality: social-democratic (n = 93); liberal (n = 235); and Christian-democratic (n = 52). This was determined by the share of votes for parties with a social democratic [(1) = Socialist Party (SP); GreenLeft (GL); Labour Party (PvdA)], conservative [(2) = Party for Freedom and Democracy (VVD); Democrats' 66 (D66)] or Christian-democratic [(3) = Christian Democratic Appeal (CDA); Christian Union (CU); Political Reformed Party (SGP)] economic agenda, respectively. These shares were calculated using the average of the votes in the 2012 and 2017 national elections (Kiesraad, 2012, 2017). For our analyses, we needed a categorical variable. Therefore, a municipality was coded as social-democratic (1), liberal (2) or Christian-democratic (3), when parties of one of those types cumulatively received a higher share of the votes than parties from the other two types. To optimise the validity, we chose to use vote shares during national, rather than local, elections, as turnout in the former is much higher than in the latter. Moreover, a substantial number of votes go to local parties during local elections (Boogers and Voerman, 2010). As the economic agendas of those local parties vary greatly across municipalities (Boogers and Voerman, 2010), ranging from very social-democratic to very liberal, votes cast for those parties are difficult to classify accurately.

The share of non-Western ethnic minorities in a municipality was used as a proxy for the degree of ethnic diversity, as there was no municipal-level indicator available for this. It should be noted that previous research has demonstrated that ethnic diversity and the share of non-Western ethnic minorities in an area affect social cohesion in a roughly similar way in the Netherlands (Gijsberts, van der Meer and Dagevos, 2012). In other words, they can largely be used interchangeably for the research problem at hand. The share of non-Western ethnic minorities is included in the Dutch Local Election Studies as an ordinal variable with three categories: (1) <3%; (2) 3–10%; and (3) >10%. According to Statistics Netherlands, a citizen is considered a non-Western ethnic minority member if at least one parent was born in an African, Latin American or Asian (excluding Indonesia and Japan) country (CBS, 2019b, 2019c).

Urbanisation measures the number of addresses within a radius of 1 km, that is, “address density” (CBS, 2019a), recalculated by Statistics Netherlands to a range of 1–5. The urbanisation scores of the municipalities in our data set in 2018 (CBS, 2019e) ranged from 1.0 (municipalities: Alphen-Chaam, Loppersum and Westerveld) to 4.8 (municipality: Amsterdam). Consequently, we cover almost the entire range and included the five types of municipality according to their urbanisation level per km², as discerned by Statistics Netherlands: (1) non-urban: fewer than 500 addresses; (2) small urban: 500–1,000 addresses; (3) moderate urban: 1,000–1,500 addresses; (4) strong urban: 1,500–2,500 addresses; (5) very strong urban: 2,500 or more addresses (CBS, 2019a).

Individual-level variables

The analysis included the individual-level variables needed to account for the notion of self-interest when exploring our scenarios, as well as those conventionally used as controls in studies on welfare attitudes. Gender was coded as male (0) and female (1). Age was measured in years. Education level indicated the highest level attained by the respondent according to the following three categories (based on the International Standard Classification of Education 2011; ISCED): (1) low = “lower than lower secondary” and “lower secondary”; (2) medium = “lower tier secondary”, “upper tier secondary” and “advanced vocational”; and (3) = “lower tertiary” and “higher tertiary” [3]. Employed was coded as (1) and unemployed as (0), while income was measured as the combined net monthly incomes of all household members in Euros, ranging from 0 to 12,475 euros per month. Vote choice was categorised into four groups reflecting how the respondents voted during the 2017 national elections: (0) vote for other (i.e. non-social democratic, non-liberal and non-Christian democratic) party = Party for Freedom (PVV), Forum for Democracy (FvD), Senior party (50+), Party for the Animals (PvD) and Think (Denk); (1) vote for liberal party = Party for Freedom and Democracy (VVD) and Democrats' 66 (D66); (2) vote for social-democratic party = Socialist Party (SP), GreenLeft (GL) and Labour Party (PvdA); or (3) vote for a Christian-democratic party = Christian Democratic Appeal (CDA), Christian Union (CU) and Political Reformed Party (SGP).

Table 1 depicts the descriptive statistics of all the variables.

Modelling approach

As our respondents were nested in municipalities, we estimated a set of multilevel models to explore our ideas (Hox, 1998). Because all the expectations enabled us to simplify the three-category dependent variable into a dichotomous version, we can report our analyses as clearly as possible (i.e. a logistic, instead of a multinomial logistic, regression), improving the interpretability of our results.

Results

Our independent variable – local social policy preferences – clearly indicates that the preferences regarding the type of local social policy varied among the respondents. Option 3 had the most support among our respondents (i.e. 44% favoured protection-based local social policies), followed by option 2 (31% preferred activation-based local social policies) and option 1 (21% favoured cohesion-building local social policies).

In advance of assessing whether municipal factors shaped the local social policy preferences of local citizens, we first explored whether conventional individual-level factors in welfare-attitude research affected those preferences as expected. Model 1 in Table 2 therefore includes the individual-level control variables. The results were largely in accordance with earlier research (e.g. Van Oorschot, 2000, 2006): Model 1 indicates a preference for protection-based local social policies over activation-based and cohesion-building versions, and this was stronger among females (cf. Gelissen, 2000), the elderly (cf. Van Oorschot, 2000), those on lower incomes (cf. Bean and Papadakis, 1998) and the less educated (cf. Linos and West, 2003). Surprisingly, the unemployed did not favour protection-based local social policies over activation-based or cohesion-building ones.

We then explored if the scenarios on the context conditions found support in the data. Model 2 indicates that there was municipal-level variation over and above individual-level variation in local social policy preferences. In other words, preferences pertaining to local social policies, as measured by our novel survey item, are not only determined by individual-level characteristics such as age and gender, but also by municipal conditions. In what follows, we are the first to scrutinise whether the types of conditions described in the literature – which have generally been found to shape attitudes towards national welfare – are also relevant in determining attitudes towards local welfare.

Models 2 and 3 address our scenarios about the impact of local economic conditions on attitudes towards local welfare, namely that high unemployment rates shape local social policy preferences according to the sociotropic or self-interest mechanisms. Model 2 clearly demonstrates that the former finds no evidence: support for protection-based local social policies was not higher in municipalities with high unemployment rates. To ease interpretation, and as recommended (cf. Brambor et al., 2006), we provide a graphic representation of the interaction effects of the unemployment rate with income (Figure 1) and employment status (Figure 2). Figures 1 and 2 show that the self-interest mechanism was also not at play: a high rate of unemployment did not exacerbate the difference between the economically well off (high incomes and employed) and the economically weak (low incomes and unemployed) in terms of support for protection-based local social policies over activation-based or cohesion-building versions [4]. This means that we did not find support for the scenario that local economic conditions might be relevant for attitudes towards local welfare.

Models 4, 5 and 6 explored the scenarios of the influence of the local political climate, which assumed that the local political climate might shape inhabitants’ local social policy preferences via the consensualism or reactive mechanisms. Surprisingly, all three models indicated that neither of the scenarios found support in the data: the interaction effects of political climate with vote choice on local social policy preferences were not significant. To aid interpretation, we depict the central findings of models 5–7 in Figures 3–5. The error bars in all these cases convincingly demonstrate that the local political climate did not affect the relationship between vote choice and preferences towards local social policies.

Figure 3 shows that non-social democrats are not more likely to support protection-based local social policies when they live in a municipality with a social-democratic political climate; nor do social democrats favour these policies more in municipalities with a non-social-democratic political climate. Similarly, Figure 4 shows that those who do not vote for Christian-democratic parties are also no more likely to support cohesion-building local social policies when they live in a municipality with a Christian-democratic political climate, nor are those who vote for Christian-democratic parties in municipalities with a non-Christian-democratic political climate. Figure 5 demonstrates that both non-liberals and liberals are likewise no more in favour of activation-based local social policies in municipalities with a liberal, or non-liberal, political climate, respectively. Overall, neither the consensualism nor the reactive mechanism seem to play a role in shaping attitudes towards local social policies in the Dutch context.

Model 7 estimated the effect of a municipality's share of non-Western ethnic minority residents on local social policy preferences. Although this effect was in the expected direction, it was too weak to achieve statistical significance, meaning that we found no evidence to support the scenario that the share of non-Western ethnic minorities might affect attitudes towards local welfare. Finally, model 8 estimated whether inhabitants of rural municipalities were more supportive of activation-based local social policies, as we had expected based on ideas suggesting that urban socialisation leads to more individualistic explanations for economic hardship. Yet, again, this proved not to be the case. The effect of urbanity was far from significant. Our sixth scenario, therefore, also finds no support in the data. By no means can we therefore say that local attitudes towards municipal social policies vary according to the economic, demographic, political or urban local context conditions [5]. What we can say, however, is that these attitudes do vary according to individual-level characteristics, presenting new support for the self-interest theory also being applicable to the local level of governance.

Conclusions

The aim of this study was to explore whether and how attitudes towards decentralised local social policies are shaped by municipal conditions. This is an important issue, given the trend in many Western countries of decentralising social policies from national to local authorities. The existing literature focusses on national welfare attitudes, and so we are the first to assess attitudes towards local welfare. To do this, we developed a novel survey item through which we assessed whether people prefer (1) protection-based, (2) cohesion-building or (3) activation-based local social policies. This was included in the 2018 wave of the Dutch Local Election Studies. By combining that survey with municipal-level data from various sources, we were able to answer the question: How do local conditions shape welfare attitudes towards decentralised social policies? Informed by a large body of welfare-state literature, we focussed on the following local conditions: (1) the unemployment rate; (2) the political climate; (3) the share of non-Western ethnic minority citizens; and (4) the level of urbanisation.

Surprisingly, our study suggests that none of these conditions shape preferences towards local social policies in addition to the standard individual-level characteristics of gender, age, income and level of education. Unlike the “self-interest” and “sociotropic” notions, unemployment rates do not have an impact on those preferences. Likewise, living in a municipality where one's political adversaries are dominant does not shape one's preferences towards such policies according to the consensualism (informed by contact theory) and reactive mechanisms (informed by group threat theory). Contrary to Putnam’s theory of “hunkering down” (2007), the share of non-Western ethnic minorities does not reduce support for protection-based local social policies. Finally, we also found no evidence to support the idea that attitudes may be influenced by socialisation in rural or urban regions: inhabitants of rural areas are not more supportive of activation-based local social policies.

In summary, while preferences towards local social policies clearly vary across individuals and municipalities, it has been shown that they are not shaped by the “usual suspects” among the context conditions.

Debate

Given the aforementioned, from a citizen's perspective, there does not seem to be a strong case for decentralising social policies, or at least not for tailoring those policies to local conditions. That said, by taking our study's characteristics into account, we suggest that there are at least four reasons for this unexpected finding.

The first and most obvious reason is our focus on the Dutch case, as determined by the availability of unique survey data across all municipalities on citizens’ preferences towards local social policies. This provided a strict test of our hypotheses: local conditions are less likely to vary in the Netherlands and are thus not expected to have as strong an effect as in larger countries such as the United States, Germany or the United Kingdom. This could, for instance, be the reason why Blomberg and Kroll (1999) and Sevä (2009) found that local conditions do matter for welfare attitudes in Finland and Sweden, although it should be noted that their focus was on attitudes towards national welfare arrangements. Their findings imply that future studies on how local characteristics shape attitudes towards local welfare are more likely to find context effects if the focus is on countries larger than the Netherlands.

Secondly, when determining which context conditions to include in our study, we self-evidently relied on empirical research on welfare attitudes. As previous studies have demonstrated that unemployment rates, the share of immigrants or ethnic minorities, the political climate and the level of urbanisation are relevant for national welfare attitudes, we expected them to also be applicable to attitudes towards local social policies. This, however, was not the case, which suggests that other municipal conditions may be more relevant in this regard. For example, regional cultures (cf. Van der Waal and Houtman, 2011), or regional differences in religious denominations (cf. Jaime-Castillo et al., 2016), could be relevant. Also, our theoretical scenarios were rather general, while actual scenarios might be more conditional, especially in case of counterarguments such as the consensualism and reactive mechanisms. The absent effect of the political climate in our analysis might very well be the net result of those two mechanisms being at play for different subsets of the population. In addition, more concrete policy-related elements, such as satisfaction with, or the effectiveness of, specific policies, may be more relevant for attitudes towards local social policies than the ideological differences across municipalities in our focus (cf. Van Oorschot and Meuleman, 2012). This is for future research to uncover.

Lastly, in the specific case of our test of the hunkering down hypothesis, we had to rely on a very crude measure due to data limitations. While a measure of the precise municipal share of non-Western ethnic minorities would be ideal, the best measure available is a crude, three-category version, which may have underestimated this effect. Furthermore, the literature offers a counterargument to Putnam’s (2007) hunkering down thesis, claiming that a higher level of ethnic diversity may increase the (perceived) risk of income loss. This, in turn, may result in higher support for protection-based welfare policies, known as the compensation effect (Eger and Breznau, 2017; Finseraas, 2008). In cases of the positive compensation effect and negative hunkering-down effect both being at play (for different subsets of citizens), they would have cancelled each other out in our analyses.

Needless to say, where the multilevel modelling approach used in this study provides a clear picture of the relevance of individual and contextual variables, alternative methodological approaches, especially more qualitative ones, could provide more insights into the hows and whys of our findings. Research in this sense is clearly called for.

Overall, this is the first study to explore whether preferences towards local social policies are shaped by local context conditions. While local worlds of welfare do exist in the Dutch case – preferences towards local social policies clearly vary across municipalities – the various context conditions conventionally used in welfare-attitude research cannot account for that pattern. Future studies could explore and uncover alternative explanations for the existence of local worlds of welfare attitudes.

Figures

Figure 1

The relationship between income and support for protection-based local social policies by unemployment rate

Figure 2

The relationship between employment status and support for protection-based local social policies by unemployment rate

Figure 3

The relationship between vote choice and support for protection-based local social policies, by social-democratic political climate

Figure 4

The relationship between vote choice and support for cohesion-building local social policies, by Christian-democratic political climate

Figure 5

The relationship between vote choice and support for activation-based local social policies, by liberal political climate

Table 1

Descriptive variables

	Min.	Max.	Mean	SD	n
Dependent variable
Support for protection-based local social policies	0	1	0.45		2,654
Support for cohesion-building local social policies	0	1	0.21		2,654
Support for activation-based local social policies	0	1	0.33		2,654
Individual-level variables
Gender (female)	0	1	0.53		2,703
Age	18	100	52.9	17.50	2,703
Employed	0	1	0.49		2,696
Income	0	12,475	3.10	1.66	2,462
Education
Low	0	1	0.26		2,697
Middle	0	1	0.34		2,697
High	0	1	0.40		2,697
Vote choice
Vote for social-democratic party	0	1	0.31		2,145
Vote for Christian-democratic party	0	1	0.19		2,145
Vote for liberal party	0	1	0.33		2,145
Vote for other party	0	1	0.18		2,145
Municipal-level variables
Unemployment rate	2.9	9.2	4.87	1.17	2,730
Political climate
Social-democratic	0	1	0.40		2,730
Christian-democratic	0	1	0.08		2,730
Liberal	0	1	0.53		2,730
Share of non-Western ethnic minorities
<3%	0	1	0.11		2,682
3–10%	0	1	0.40		2,682
>10%	0	1	0.49		2,682
Urbanisation	1	4.78	3.18	1.07	2,730

Source(s): Dutch Local Election Studies, 2018 (own calculations)

Table II

Multilevel logistic regression analyses of 1,913 citizens in 336 Dutch municipalities in 2018, with the dependent variable ‘local social policy preferences’ (entries are log-odds regression coefficients; standard errors in parentheses)

	Model 1	Model 2	Model 3	Model 4	Model 5	Model 6	Model 7	Model 8
Local social policy preferences	Protection-based	Protection-based	Protection-based	Protection-based	Cohesion building	Activation-based	Protection-based	Activation-based
Constant	–0.319 (0.286)	–0.128 (0.400)	0.150 (0.527)	–0.847 (0.500)	–0.951 (0.630)	–0.156 (0.526)	–0.268 (0.407)	–0.710 (0.571)
Individual-level variables
Female	0.257* (.105)	0.271* (0.115)	0.280* (0.118)	0.236* (0.126)	0.076 (0.133)	–0.317* (0.128)	0.286* (0.115)	–0.316* (0.127)
Age	0.012** (0.004)	0.012** (0.004)	0.012** (0.004)	0.016** (0.004)	–0.008 (0.005)	–0.011* (0.005)	0.012** (0.004)	–0.011* (0.005)
Employed	0.023 (0.122)	0.036 (0.134)	–0.391 (0.485)	–0.023 (0.153)	0.059 (0.168)	–0.001 (0.137)	0.046 (0.135)	–0.003 (0.137)
Income	–0.116*** (0.035)	–0.121*** (0.033)	–0.136 (0.109)	–0.133** (0.043)	–0.017 (0.041)	0.143** (0.041)	–0.123*** (0.033)	0.139*** (0.040)
Education
Middle	–0.228 (0.137)	–0.256 (0.166)	–0.271 (0.172)	–0.171 (0.176)	0.277 (0.195)	–0.005 (0.191)	–0.257 (0.168)	–0.001 (0.189)
High	–0.570*** (0.137)	–0.598*** (0.147)	–0.638*** (0.153)	–0.495** (0.158)	0.141 (0.183)	0.448* (0.193)	–0.588*** (0.148)	0.445* (0.194)
Vote choice
Social-democratic party				0.235 (0.260)	0.148 (0.239)	–0.764** (0.234)		–0.758** (0.233)
Christian-democratic party				–0.256 (0.252)	0.394 (0.272)	–0.594* (0.274)		0.210 (0.190)
Liberal party				–0.003 (0.202)	–0.156 (0.229)	–0.035 (0.242)		0.545** (0.192)
Municipal-level variables
Unemployment rate		–0.045 (0.052)	0.091 (0.093)	0.108 (0.068)	–0.089 (0.074)	–0.014 (0.076)	0.013 (0.059)	–0.030 (0.080)
Municipal political climate
Social-democratic political climate				–0.404 (0.266)
Christian-democratic political climate					0.132 (0.824)			–0.124 (0.258)
Liberal political climate						–0.424 (0.248)		–0.164 (0.142)
Share of non-Western ethnic minorities
3–10%				–0.003 (0.204)	0.128 (0.326)	–0.105 (0.200)	–0.045 (0.177)	–0.125 (0.231)
>10%				–0.298 (0.227)	0.145 (0.351)	0.209 (0.238)	–0.272 (0.193)	0.184 (0.351)
Urbanity								0.012 (0.115)
Cross-level interactions
Unemployment rate × employment			–0.081 (.096)
Unemployment rate × income			–0.003 (0.019)
Left political climate × left-wing voting				0.239 (0.372)
Left political climate × Christian-democratic voting				0.507 (0.396)
Left political climate × liberal voting				0.048 (0.298)
Christian political climate × left-wing voting					–1.023 (1.105)
Christian political climate × Christian-democratic voting					–0.071 (0.908)
Christian political climate × liberal voting					–0.139 (1.108)
Liberal political climate × left-wing voting						0.530 (0.336)
Liberal political climate × Christian-democratic voting						0.557 (0.368)
Liberal political climate × liberal voting						0.160 (0.322)
Random effects
Municipal-level variance		0.204 (0.079)	0.172 (0.078)	0.151 (0.092)	0.356 (0.248)	0.180 (0.077)	0.197 (0.078)	0.179 (0.077)
Log-likelihood	–1592.71	–1583.55	–1578.21	–1159.59	–886.80	–1090.66	–1571.63	–1092.74
n individual	1,913	1,913	1,913	1,913	1,913	1,913	1,913	1,913
n municipality	336	336	336	336	336	336	336	336

Note(s): * p < 0.05; p < **0.01; p < ***0.001 (two-sided)

Reference categories: Female (ref. = Male); Employed (ref. = Unemployed); Education (ref. = Low); Vote choice model 5 (ref. = social-democratic party); Vote choice model 6 (ref. = Christian-democratic party); Vote choice models 7 and 8 (ref. = liberal party); Share of non-Western ethnic minorities (ref. = <3%); Municipal political climate model 5 (ref. = social-democratic political climate); Municipal political climate model 6 (Christian-democratic political climate); Municipal political climate models 7 and 8 (ref. = Liberal political climate)

Notes

1. Survey items on national governments’ responsibilities are generally formulated as follows: “Should it be the government's responsibility to provide welfare type [x] to societal group [x]?” (For an overview of such items, see Svallfors, 2004, p. 124).

2. For a more elaborate conceptual explanation, please consult “measuring attitudes towards decentralised social policies” on p. 3.

3. Estimating models 1–8 using a continuous variable instead of a categorical variable for level of education does not alter the results.

4. A robustness check that estimated the interaction terms Unemployment rate*Employed and Unemployment rate*Income simultaneously did not yield substantially different results compared to models 2 and 3. These analyses are available upon request.

5. Estimating models 5 (protection-based), 6 (cohesion-building) and 8 (activation-based) with all independent variables does not lead to a different conclusion.