All good in the neighbourhood? Exploring the role of local conditions for political trust and corruption perceptions within a minority context

ABSTRACT

Members of a minority population are expected to evaluate political authorities through a different spectrum than majority members. Still, only a few studies have explored the role of local conditions as determinants for political attitudes among a linguistic minority population. This study focuses on the role of linguistic homogeneity, local economic conditions and community size as determinants for corruption perceptions and political trust at the individual level among members of a linguistic minority population: the Swedish-speaking Finns. Moreover, this study differs empirically between evaluations of the local and national authorities. Combining data from a panel survey collected by Barometern (2019–21) with varying contextual-level factors in a multilevel model, we show that evaluations of local authorities are related to local contextual factors, in particular municipal unemployment rates and median incomes, while national evaluations are largely unrelated to local conditions. In contrast to theoretical expectations for an in-group bias among minority members, however, we find no effect for linguistic homogeneity on corruption perceptions and political trust after controlling for economic conditions. The results make an important contribution to our understanding about the role of local conditions for political attitudes among a well-assimilated minority population.

KEYWORDS:

• corruption perceptions
• political trust
• local
• multilevel
• linguistic minority

INTRODUCTION

Minorities are integral parts of most European societies. Still, members of a minority population, be it an ethnic or linguistic minority, are often perceived as being dependent on a benevolent majority population for prosperity, and even security (Binder et al., 2009; Koev, 2019; Simon et al., 2001). Consequently, a minority population is expected to put more support into the governing institutions out of a necessity to safeguard against the ‘tyranny of majority’ (Wilkes & Wu, 2018, p. 22). Nevertheless, members of minority populations can still be in strong positions locally, both numerically and proportionally, creating a local environment where they can prosper without facing prejudices even if the national situation might be challenging. The shifting local environments in which members of a minority population live are thus expected to have consequences for the formation of political attitudes amongst members of the minority population (Karv et al., 2022; Wilkes & Wu, 2017). Even so, studies focusing on the role of local environments for the formation of political attitudes within a minority population are still somewhat lacking in the literature.

Two essential elements of a well-functioning political system are that the political processes are perceived as fair and the political authorities as trustworthy, setting the normative foundations for the political community (Marien & Werner, 2019; Petrzelka et al., 2013; Tyler, 2003). However, some amount of public scepticism could arguably also contribute with something positive for the political community as it helps keeping the authorities accountable (Norris, 2011; Salminen & Ikola-Norrbacka, 2010). Still, when there is a deeper and widespread belief among the population that politicians and bureaucrats are corrupt and untrustworthy it is expected to undermine the foundations of the political system (Easton, 1965; Linde & Erlingsson, 2013). Hence, for a well-functioning political system, perceptions about the fairness and trustworthiness of political processes and the political authorities are key (Rahn & Rudolph, 2005). In this study, we will use two distinct evaluations of the fairness and trustworthiness of local and national political authorities among a linguistic minority population. The research purpose of the subsequent study is to explore the role of local level determinants for two related types of political evaluations, in the hope of also shedding some light on the complicated relationship between the two closely entangled but theoretically distinct concepts: corruption perceptions and political trust.

The research aim is twofolded. First, we aim to explore corruption perceptions and political trust in relation to two institutional levels, local and national, among a linguistic minority population in Finland. Research has shown that trust in local authorities is often higher than trust in national authorities (Denters, 2002; Fitzgerald & Wolak, 2016; Hansen & Kjaer, 2020; but see Petrzelka et al., 2013) and that the higher levels of government tend to be perceived as being more corrupt (François & Meon, 2021). This has been explained by the fact that local governments have greater possibilities to act in accordance with public preferences while national politics could be perceived as having no impact on everyday lives (Stein et al., 2021). However, here Finland seems to have constituted an exception in a Western European comparison (Fitzgerald & Wolak, 2016, p. 135; Salminen & Ikola-Norrbacka, 2010, p. 655; but see Karv, 2020).

Second, we aim to explore whether, and the extent to which, corruption perceptions and political trust are related to the local level surroundings of the minority population, and thus contextually bound, irrespective of institutional level. The local situations of the members of the linguistic minority population in Finland, that is, the Swedish-speaking Finns, also differs extensively within the country (Karv et al., 2022). Given the research purpose, we will combine survey data gathered by Barometern (Lindell, 2019; 2021) with local level characteristics across the Swedish-speaking communities in Finland. Through this procedure, we can answer our overarching research question: to what extent is political (dis)satisfaction within a minority setting related to local level characteristics? Our results show that local factors are indeed associated with corruption perceptions and political trust across a minority population, making an important contribution to the broader research field about the determinants of political attitudes within a minority.

THE SYSTEM IMPORTANCE OF CORRUPTION PERCEPTIONS AND POLITICAL TRUST

As corruption is perceived as a widespread negative phenomenon throughout the world, with expected negative effects on various types of politics attitudes and political behaviour, the topic has attracted substantial interest among scholars. One of the most established definitions of corruption is that it relates to the misuse of public power for private gain (Beeri & Navot, 2013, p. 713; Karklins, 2005, p. 4). Corruption at the national level has been repeatedly linked with negative sentiments towards politicians in cross-country comparisons (Anderson & Tverdova, 2003; Bowler & Karp, 2004; Mishler & Rose, 2001), repeatedly argued to threaten the legitimacy of democratic processes (Pharr, 2000; Seligson, 2002). Corruption perceptions are on the other hand something that exists in the minds of the citizens, sometimes based on one’s own personal experiences, experiences of friends or family, media reports of scandals, from hearsay or from a general distrust in the political community. It is thus crucial to acknowledge that corruption perceptions are not the same as actual corruption. However, according to Melgar et al. (2010, p. 120), high corruption perceptions in a society fosters a ‘culture of distrust’ with possibly more devastating effects than actual corruption. Hence, widespread disbelief in the political community contributes to growing negative sentiments about the fairness of the political processes, political institutions and the political system (Easton, 1965; Norris, 2011).

Corruption perceptions are closely linked with theories about procedural fairness (Tyler, 2003; van den Bos & Wilke, 1998), and many studies have explored the impact of corruption on political trust, most perceiving corruption levels as a determinant for various levels of political trust (Anderson & Tverdova, 2003). A trust judgement reflects an individual's assessment about the trustworthiness of someone or something and is thus relational but seldom unconditional (Levi & Stoker, 2000, p. 476), and political trust is therefore here perceived as an evaluative judgement directed towards a political object (e.g., van der Meer, 2010). Political trust has been established as one of the most system important evaluative political attitudes and has thus been described as a glue that holds the political system together simultaneously as it makes the political system work efficiently (van der Meer, 2010, p. 518). Declining levels of political trust is thus considered as a virus able to threaten the well-being of any type of political system, as low-trusting citizens are less likely to follow laws, vote during elections and participate in the society (Grönlund & Setälä, 2007; Marien & Hooghe, 2011). Hence, political trust could even be perceived as a success criterion for societies (Listhaug & Ringdal, 2008, p. 131).

We argue that corruption perceptions reflect a more system important evaluation of the well-being of a political system than political trust. If a citizen perceives a politician as not being trustworthy, it is not farfetched to expect that same citizen to perceive that same politician as being corrupt. As Uslaner noted, ‘corruption lies at the heart of distrust of government’ (Uslaner, 2017, p. 303). In earlier studies, corruption perceptions have been viewed as both a cause and effect of political trust (Della Porta, 2000; Marien & Werner, 2019), showing low-trusting citizens to be more likely to perceive politicians as corrupt (Van de Walle, 2008; Wroe et al., 2013) and that fairness perceptions affect political trust (Marien & Werner, 2019; Uslaner, 2017). It is thus likely that an individual might not express any feelings of trust towards a political object while still perceiving the same political object not to be corrupt.

THE ROLE OF LOCAL CONDITIONS

Political attitudes are affected by the social surroundings of individuals and the quality of local life (Campbell et al., 1954; Fitzgerald & Wolak, 2016; Karv et al., 2022; Rahn & Rudolph, 2005; Wolak, 2018). According to Reeves and Gimpel (2012, p. 509), citizens use the observations they make in their everyday lives to shape their opinions. Studies have also shown that citizens are prone to use cognitive shortcuts, for example, heuristics, when asked to make judgements about politics (Anderson, 1998; Rudolph, 2017). Consequently, as the high quality of local government has been described as one of the success factors across the Nordic countries (Haveri, 2015) it has contributed to fostering high levels of trust in the local authorities and low levels of corruption perceptions (Salminen & Ikola-Norrbacka, 2010). Local governance has traditionally played an influential role in Nordic societies, with the substantial levels of local autonomy constituting one of the main characteristics of the Nordic political system (Haveri, 2015, p. 139). Given the strong role of local governments in the Nordic countries, we expect the local conditions to have a strong influence on the formation of attitudes towards both the local and national political authorities.

Language constitutes an important ethno-cultural marker, which in turn determines whether an individual is part of the minority or the majority group within a political community (Hwang, 2017). The relation between belonging to a linguistic minority group and political trust is not straightforward but context dependent, as some minorities are in a disadvantageous position while some enjoy significant legal protections which in turn is expected to affect their political evaluations. Thus, given a local or national context where there are significant legal protections, members of a linguistic minority group might be expected to express more positive political attitudes, as for instance has been shown with the French-speaking Canadians (Hwang, 2017) and the Swedish-speaking Finns (Bäck, 2019; Grönlund & Isaksson, 2008).

The population structure of local communities affects the characteristics of a community in several ways. Following the social psychology literature, individuals are generally expected to have an in-group bias, that is, a general preference to interact with individuals like themselves (Tajfel, 1982). Moreover, there is a large array of studies showing that communities that are more homogeneous are associated with more social stability and better general performance (Alesina et al., 1999; Anderson & Paskeviciute, 2006; McLaren, 2017; Patsiurko et al., 2013). Thus, in more ethnically or linguistically homogeneous communities’ political trust is expected to be higher as both the public administration and the elected incumbents are more likely to constitute members of the citizens’ wider ethnical or linguistic in-group (Rahn & Rudolph, 2005). Some scholars therefore suggest that the reason for this is that population heterogeneity breeds conflict and as a result creates political systems that are more unstable (Anderson & Paskeviciute, 2006, p. 785). In line with these findings, research about intergroup relations between minority and majority members have argued that members of minority groups are expected to feel less secure than members of the majority group (Simon et al., 2001, p. 312) and are thus sensitive to perceived power imbalances within the community (Binder et al., 2009). It follows therefore that the higher the relative size of a minority population the more secure a minority member should feel.

A positive evaluation of the performance of local authorities is expected to nurture positive sentiments towards the local government. Subsequently, citizens living in a local community with well-functioning political institutions and trustworthy political representatives should be more positive towards the local authorities in comparison to a community characterized by the opposite (Fitzgerald & Wolak, 2016; Reeves & Gimpel, 2012). One of the most used measurements for the performance of local authorities concerns the local economy, and political trust has been repeatedly shown to be associated with both the local and national economic performance (Fagerland Kroknes et al., 2015; Hetherington & Rudolph, 2008; Rahn & Rudolph, 2005; Weinschenk & Helpap, 2015). Thus, the level of political trust tends to rise when the outlook is brighter. Likewise, studies have shown that people who perceive the national economy as bad have a greater tendency to think that corruption is more prevalent across government (de Lancer Julnes & Villoria, 2014). Hence, it could be expected that the economic performance or status of a political community might be a factor equally affecting corruption perceptions within a community. Scholars have also shown that assessments of the performance of the local economy shape the general impression of the national economy (Reeves & Gimpel, 2012), as individuals do notice first hand when family and friends are losing their jobs. Hence, even if the local economic conditions are expected to have a more direct impact on how the local authorities are being evaluated, the local economic conditions are also expected to be used as an information shortcut when citizens are asked to evaluate the performance of national authorities.

Another strand of community characteristics that have been explored in relation to various types of political attitudes are the effects of living in an urban or rural environment. Thus, at least since Lipset and Rokkan (1967) it has been believed that locations matter for the formation of political attitudes and that there is a territorial dimension to politics. Research has found that living in the periphery might foster distrust towards national politics, as living in more rural communities is associated with lower levels of trust in national political institutions (Mitsch et al., 2021; Stein et al., 2021). Nevertheless, trust in the local political institutions is expected to be higher in more rural communities, as it is easier to both understand and influence local politics within smaller communities (Dahl & Tufte, 1973; Denters, 2002; John, 2001; Stein et al., 2021). Similarly, larger communities have been shown to be associated with more negative corruption perceptions (e.g., Bergh et al., 2017). Hence, location and size of the community are also important factors to account for. In line with the findings from the literature, we have created four hypotheses to guide the remaining part of this study:

Hypothesis 1: Political trust and corruption perceptions among a minority population are more negative in linguistically more heterogeneous communities.

Hypothesis 2: Political trust and corruption perceptions among a minority population are more negative in economically worse performing communities.

Hypothesis 3: Political trust and corruption perceptions among a minority population are more negative in larger urban communities.

Hypothesis 4: Local contextual factors are better at predicting varying levels of political trust and corruption perceptions towards the local than the national government.

We hereafter proceed with presenting the research design guiding the empirical part of this study.

DESIGN, METHOD AND DATA

In order to test our hypotheses, we employ panel-based survey data gathered by Barometern (2019–21), with respondents derived from the Swedish-speaking communities in Finland, for example, Swedish-speaking Finns. Finland can be described as a high-trusting society, characterized by well-functioning political institutions and a trusting population (Bäck & Kestilä, 2009; Listhaug & Ringdal, 2008; Salminen & Ikola-Norrbacka, 2010; Söderlund, 2019). Thus, public authorities in Finland are perceived as being both honest and trustworthy (Salminen & Ikola-Norrbacka, 2010, p. 654). Moreover, Finland has since 1995 been constantly ranked as one of the least corrupt countries in the world according to the Corruption Perceptions Index (Transparency International, 2021). Moreover, Finland is together with the rest of the Nordic countries often celebrated as a role model for its clean and honest government (Erlingsson & Kristinsson, 2020; Zook, 2009). However, while classic forms of so-called ‘every-day corruption’ that ordinary citizens might encounter in their daily lives (e.g., bribes to doctors or police officers) are exceedingly rare in Finland, more sophisticated and hard-to-detect forms of elite corruption based on favouritism and networking are judged to be considerably more common (Salminen & Ikola-Norrbacka, 2010). Since many Finns lack a personal experience of corruption and corruption scandals are relatively scarce, one could argue in line with Van de Walle (2008, p. 2) that this ‘invites respondents to broaden their frame of reference to whatever factor they wish when giving an opinion on corruption’.

Studies have shown that there are apparent variations across the different parts of Finland when it comes to various types of political attitudes (Grönlund & Isaksson, 2008; Karv, 2020; Karv et al., 2022). For instance, studies have shown the Swedish-speaking population to be slightly more politically trusting than the majority speaking Finns (Bäck, 2019; Grönlund & Isaksson, 2008). The Swedish-speaking population of Finland constitutes roughly 5.5 percentage of the total population, with the Swedish language enjoying equal rights with Finnish, according to the Finnish Constitution (§ 2.17). This means that a Swedish-speaking Finn can demand to get service in Swedish when dealing with state authorities, such as the police, anywhere in Finland. The Swedish-speaking population is geographically located in the coastal mid-Western and Southern parts of Finland, even constituting the majority population in several of the 293 municipalities in Finland. In a Finnish context, language has historically always constituted a cleavage as the Swedish-speaking communities has distinct interests related to the Swedish language (Westinen, 2015, p. 282). Even though the Swedish-speaking Finns constitute a linguistic minority in Finland that is regarded as a highly assimilated minority group, members of the Swedish-speaking community strongly identify with the minority community (Sandberg, 2019). A minority population is often represented at the political arena by a political party with the sole purpose of guarding the interests of the minority, and these can even obtain a proportionally large influence. The explanation for this is that minority members are more prone to political participation in order to safeguard their positions (Branscombe et al., 1999; Koev, 2019; Stokes, 2003; but see Sandovici & Listhaug, 2010). Hence, also in the case of Swedish-speaking Finns a significant majority still vote for the only language party in the Finnish political system, the Swedish People’s Party (SPP). Hence, ‘being a Swedish-speaking Finn explains which party you vote for and being a Finnish-speaking Finn explains primarily which party you do not vote for’ (Westinen, 2015, p. 293).

This makes this dataset highly suitable for testing hypotheses regarding the impact of linguistic homogeneity on political trust and corruption perceptions since we have some communities where Swedish-speaking Finns enjoy a majority position and others where they are a minority. Moreover, there are also extensive differences in terms of the local economic conditions and types of communities in which the Swedish-speaking population lives. The Barometer data concerning political trust were gathered in late 2019 while the data on corruption perceptions and attitudes were gathered later in early 2021. The fact that the data for our dependent variables were gathered in two separate surveys, over one year apart, reduces the chance that the political trust- and corruption-related questions contaminated each other through the psychological phenomenon known as ‘priming’ (Tourangeau et al., 2000). Hence, if we find similar results for both types of dependent variables, we can be quite sure that the results are not a random fluke but a relatively stable attitude.

Political trust is often measured with items asking the respondent to rate the trustworthiness of various actors and institutions (Levi & Stoker, 2000). Even if trust is a relational concept, the trust assessments of various political institutions in Finland are often similar (Bäck, 2019). Thus, a trusting or distrusting individual is often prone to express similar sentiments irrespective of political object. In order to measure differences between trust evaluations, we have therefore created a local and a national trust index. The respondents were asked the following standard question: Personally, how much trust do you have in the following institutions? With response options being Trust completely, Trust somewhat, Neither trust nor distrust, Distrust somewhat, Distrust completely and Don’t know. To create the local trust index, we combined trust assessments of Municipal politicians and Municipal officials (in the respondent’s home municipality, Cronbach’s alpha = 0.859). To create the national trust index, we combined trust assessments of Political parties, National parliament and National government (Cronbach’s alpha = 0.872). We used Pearson’s R as a correlation estimate, with two-tailed tests of significance, to explore the binary relation between the two indices in order to show that the respondents were generally able to differ between the two institutional levels (r = 0.552, p < 0.001). Hence, even though there is a strong correlation between the trust assessments, this shows that the respondents can differ between the levels.

Corruption perceptions are usually measured by asking the respondents to assess the likeliness of political actors and institutions being involved in corrupt behaviour (François & Meon, 2021). In order to measure corruption perceptions, we have created two distinct corruption perception indices, one related to the local and the other related to the national authorities. Here we asked the respondents: In general, how common do you think it is with corruption such as bribery, embezzlement, nepotism and cronyism, conflicts of interest and the like in the following types of institutions and organizations? With response options being Very uncommon, Quite uncommon, Neither common nor uncommon, Quite common, Very common and Don’t know. To create the local corruption perceptions index, we combined assessments of Political decision-makers in your home municipality and Civil servants and public employees in your home municipality (Cronbach’s alpha = 0.918). To create the national corruption perceptions index, we combined assessments of Political decision-makers at national level and Civil servants and public employees at national level (Cronbach’s alpha = 0.880). We also here checked for the binary relation between the indices (r = 0.797, p < 0.001), and here the correlation was stronger. This was no surprise given that this is considered a more system important evaluation, and hence less variation between the levels was expected. All four dependent variables are coded to range from 0 to 10, with a higher value reflecting either higher trust or that corruption is perceived as more common (see Appendix S1 in the supplemental data online for original language survey items).

As our research aim is to assess the impact of various local factors on trust and corruption assessments toward the local and national authorities, we have included several possible local contextual-level predictors in our analyses. To assess the effects of a linguistically homogenous environment we have created a variable reflecting the proportion of Swedish-speaking Finns at the municipality level. We have also created a variable reflecting the strength of the SPP in the municipal council to control for the role of a politically homogeneous local environment (Karv et al., 2022). To explore the role of local economic conditions, municipal unemployment rates and median incomes are used as proxies (see also Bergh et al., 2017; Stein et al., 2021; Weinschenk & Helpap, 2015). To differentiate between different types of local communities, a variable reflecting the population size and a variable developed by Statistics Finland for describing the degree of urbanization across municipalities are also included (Karv et al., 2022). The latter variable is constructed by dividing the proportion of the municipal population living in urban settlements and the population of the largest urban settlement into urban, intermediate and rural municipalities. Finally, we also control for from which part of Finland the respondent is. All our municipality-level variables originate from Statistics Finland and the years 2019–20 (Statistics Finland, 2021a, 2021b), each coded as ranging between 0 and 1. We further decided to include only respondents from municipalities with enough responses in our analyses in order to make more robust cross-municipal comparisons, with 19 responses constituting the lower threshold for inclusion, giving us a total of 31 municipalities (Hoyle & Gottfredson, 2015) (see Appendices S2 and S3 in the supplemental data online for municipal level values). As we assess individuals and municipalities at distinct analytical levels, we will use a statistical method enabling us to explore the effects of both levels simultaneously at the next stages.

ANALYSIS

We begin with presenting an overview of the relation between our four indices and our main local level predictors in Table 1. Here all the local predictors are categorized in order to enable a comparative overview, creating categories consisting of roughly the same number of municipalities. If we start with the local corruption perceptions, it seems clear that the role of local characteristics differs depending on the analytical level. A stronger SPP, higher proportion of Swedish-speaking Finns, lower unemployment rates, higher median income and living in a more rural community all seems to be related to more positive corruption perceptions in the relation to the local authorities. However, in relation to the national authorities the results are more mixed. Shifting focus to political trust, the results are more similar irrespective of analytical level, with a stronger SPP, higher proportion of Swedish-speaking Finns, lower unemployment rates, higher median income, living in a smaller and more rural community all related to higher levels of political trust.

Table 1. Corruption perceptions and political trust based on municipality factors.

		Corruption perception		Political trust
Contextual variable		Local (N = 2221)	National (N = 2232)	Local (N = 2645)	National (N = 2642)
Mean		4.04 (2.70)	4.38 (2.67)	5.81 (2.17)	5.52 (1.91)
Ratio of Swedish-speaking Finns (%)		0.007**	0.023*	0.00***	0.072
	Low (0–10)	4.15 (2.63)	3.99 (2.66)	5.52 (1.92)	5.57 (1.91)
	Medium (10–50)	3.96 (2.52)	4.46 (2.57)	5.75 (2.13)	5.32 (1.91)
	Strong (50+)	4.02 (2.81)	4.55 (2.71)	5.99 (2.30)	5.60 (1.90)
SPP strength (excluding Åland Islands)^a		0.001***	0.040*	0.00***	0.012*
	Some representation	4.14 (2.60)	4.18 (2.70)	5.64 (1.97)	5.57 (1.91)
	Largest party, no majority	4.04 (2.65)	4.43 (2.60)	5.68 (2.16)	5.32 (1.90)
	Majority	3.96 (2.78)	4.61 (2.71)	5.96 (2.33)	5.51 (1.95)
Unemployment rates (%)		0.002**	0.094	0.00***	0.00***
	Low (0–5)	3.72 (2.73)	4.33 (2.67)	6.39 (2.33)	5.75 (1.89)
	Medium (5–10)	4.10 (2.69)	4.40 (2.71)	5.64 (2.06)	5.43 (1.91)
	High (10+)	4.45 (2.52)	4.31 (2.40)	5.45 (2.20)	5.55 (1.97)
Median income (€)		0.002**	0.131	0.00***	0.002**
	Low (under 35,000)	4.19 (2.70)	4.38 (2.70)	5.55 (2.10)	5.47 (1.96)
	Medium (35,000–42,500)	4.02 (2.70)	4.55 (2.61)	5.76 (2.26)	5.36 (1.93)
	High (over 42,500)	3.79 (2.68)	4.19 (2.68)	6.23 (2.12)	5.75 (1.78)
Type of community		0.012*	0.38	0.00***	0.073
	Urban	4.08 (2.63)	4.19 (2.66)	5.59 (2.02)	5.45 (1.91)
	Intermediate	4.24 (2.78)	4.75 (2.66)	5.65 (2.22)	5.54 (1.87)
	Rural	3.79 (2.74)	4.45 (2.69)	6.40 (2.33)	5.66 (1.94)
Community population		0.082	0.617	0.00***	0.139
	Low (under 7000)	4.11 (2.66)	0.53 (0.26)	6.25 (2.29)	5.52 (1.95)
	Medium (7000–20,000)	3.72 (2.72)	0.58 (0.26)	6.05 (2.30)	5.66 (1.88)
	High (over 20,000)	4.21 (2.68)	0.56 (0.27)	5.54 (2.04)	5.44 (1.91)
Region		0.006**	0.069	0.00***	0.005**
	Ostrobothnia	3.84 (2.64)	0.55 (0.27)	6.10 (2.28)	5.53 (1.96)
	Turunmaa	4.48 (2.68)	0.56 (0.26)	5.05 (2.12)	5.48 (1.84)
	Uusimaa	4.11 (2.71)	0.56 (0.27)	5.67 (2.05)	5.45 (1.91)
	Åland islands	3.97 (2.81)	0.58 (0.26)	6.21 (2.15)	5.96 (1.68)

Note: Standard deviations are shown in parentheses. One-way analysis of variance (ANOVA) test of variance: *p < 0.05; **p < 0.01; ***p < 0.001. All results are weighted according to gender, age, education and municipality.

^aSPP is not active in the Åland Islands.

To explore further these relationships, we apply linear multilevel regression models. Multilevel models enable us to account for variations at two analytical levels, making it possible to identify the impact of local-level factors while simultaneously accounting for individual-level characteristics (Gelman & Hill, 2006; Steenbergen & Jones, 2002). Consequently, a two-level multilevel regression model is applied in order to explore to what extent individual-level (level 1) variations in the two types of corruptions perceptions and political trust are affected by varying local factors identified at the municipality level (level 2) (Heck et al., 2014). Furthermore, we decided to use unweighted data in our regression models, which means that the results are not necessarily representative of the target population (Swedish-speaking Finns). However, we control for the factors to which the sample was skewed and also check the robustness of our results by re-analysing the data with added weights and robust standard errors, as recommended by Solon et al. (2015).

We begin by calculating the intraclass correlation coefficient (ICC) to explore how much of the variation in corruption perceptions and trust that is between instead of within municipalities (Gelman & Hill, 2006, p. 448). The ICC values are calculated by dividing the covariance intercept with the total covariance from our null model, for example, within and between municipality variations. Starting with the ICC values for corruption perceptions, the ICC for the local level was 0.015 and for the national level 0.019. This indicates that 1.5% of local corruption perceptions and 1.9% of national corruption perception variations is expected to occur between municipalities. Continuing with political trust, the ICC for local trust was 0.108 and for national trust 0.016. Suggesting that 10.8% of local trust and 1.6% of national trust variation is expected to occur between municipalities. The ICC values are quite similar in terms of the national level and local corruption perceptions, while local trust seems to be somewhat more contextually bound.

At the following stage, we include municipality-level factors (level 2) in our multilevel model as covariates to assess empirically to what extent they can account for the variations across municipalities in corruption perceptions and political trust. Before proceeding with the regressions, we begin by controlling for the risk of multicollinearity with our variables (Field, 2009, p. 224). The risk for multicollinearity is obvious, as we included two similar types of measurements for each contextual type in our initial overview in order to get a more comprehensive picture about the relation between context and our dependent variables. Hence, we decided to use ratio of Swedish-speaking population as an indicator of linguistic homogeneity, unemployment rates as an indicator of local economic conditions and type of community to differentiate between municipal types. The tests conducted to see if these variables met the assumption of collinearity indicated that multicollinearity should not be an issue with these variables (Swedish-speaking population, Tolerance = 0.347, variance inflation factor (VIF) = 2.88; Unemployment rates, Tolerance = 0.339, VIF = 2.95; Type of Community, Tolerance = 0.536, VIF = 1.86).

At the final stage, we have further included four individual-level predictors (level 1) in the model to assess whether our contextual-level findings are affected when the model also accounts for individual-level predictors. On the individual level, we control for gender, age, and education, and we further include a quadratic term for age to control for possible non-linear effects. Previous studies also suggest that SPP voters tend to be more trusting of local and national-level decision-makers (Grönlund & Isaksson, 2008), which is why we also control for this aspect in our analyses. Gender and SPP support are coded as dummies where 1 indicates that the respondent is female/tends to vote for SPP, while age and education range from 0 to 1, where a higher value indicates that the respondent is older/has a higher education (see Appendix S4 in the supplemental data online for overview of the relation between our indices and individual-level variables).

We begin our regression analyses with corruption perceptions, presented in Table 2. The first two columns in Table 2 show the results for perceptions of corruption among local authorities. None of our contextual variables is statistically significant in model 1, although our unemployment rates indicator is borderline significant (B = 0.878, p = 0.052). When we add our individual-level predictors in model 2, however, our proxy for local economic performance turns significant and positive (B = 0.716, p = 0.049), suggesting that local corruption is perceived as more common in municipalities with higher unemployment rates, a finding that supports our Hypothesis 2. Being male, highly educated and a SPP supporter are significant predictors of a lesser tendency to perceive the local authorities as corrupt. Moreover, we find that age has a non-linear effect on corruption perceptions suggesting that they first increase as we move up the different age groups and then decline for the oldest respondents.

Table 2. Multilevel results: corruption perceptions.

	Local		National
Independent variable	Model 1	Model 2	Model 1	Model 2
Municipality level
Swedish-speaking Population (%), (0–1)	0.344 (0.351)	0.007 (0.271)	0.442 (0.407)	0.120 (0.294)
Unemployment Rates (%), (0–1)	0.878 (0.452)	0.716 (0.364)*	0.737 (0.514)	0.517 (0.388)
Type of Community (0 = urban, 1 = rural)	0.009 (0.223)	−0.088 (0.183)	0.204 (0.253)	0.041 (0.194)
Individual level
Gender (0 = male, 1 = female)		0.337 (0.111)**		0.232 (0.108)*
Age (0–1)		4.207 (0.920)***		4.451 (0.899)***
Age ^2 (0–1)		−5.093 (0.944)***		−4.381 (0.922)***
Education (0–1)		−0.877 (0.190)***		−1.825 (0.186)***
SPP support (dummy: 0 = no, 1 = yes)		−0.707 (0.122)***		−0.450 (0.120)***
Random effects
Residual (Null model)	6.876 (0.199)	6.876 (0.199)	6.659 (0.192)	6.659 (0.192)
Residual	6.889 (0.200)	6.510 (0.197)	6.662 (0.192)	6.243 (0.189)
Intercept (Null model)	3.859 (0.085)	3.859 (0.085)	4.140 (0.090)	4.140 (0.090)
Intercept	3.251 (0.375)	3.795 (0.387)	3.477 (0.427)	4.125 (0.400)
ICC	0.007	0.00	0.015	0.002
Log-likelihood	−5785.0623	−5147.0412	−5795.8798	−5129.1414
Municipalities	31	31	31	31
Respondents	2424	2185	2444	2196

Note: OLS, unstandardized coefficients; Multilevel, maximum likelihood estimates; Covariance type, unstructured. Standard errors are shown in parentheses: *p < 0.05; **p < 0.01; ***p < 0.001.

The two latter columns in Table 2 show the results for perceptions of corruption among national authorities. Here, on one hand, we see that none of our contextual predictors in either model is significant, suggesting that the unemployment rate is a better predictor of attitudes towards local authorities, a finding that supports our Hypothesis 4. Our individual predictors, on the other hand, remain almost unchanged from our previous model. The biggest differences in comparison to the local corruption perceptions model is that the coefficient for education is considerably larger than in the previous model while the SPP support coefficient is smaller. In other words, education matters more when it comes to explaining national corruption perceptions while party support matters less.

Moving on to political trust, we find very similar results as in our previous models, albeit with the opposite sign in front of the coefficients. ‘Unemployment rates’ is the sole significant negative contextual predictor in the first two columns (B = −1.629, p = 0.006 in model 2), suggesting that trust in local authorities is significantly lower in municipalities with higher rates of unemployment. Meanwhile, we find no significant contextual effects for trust in national authorities. Concerning our individual-level predictors, we find that women tend to be less trusting of local authorities, and that the highly educated and SPP supporters are more trusting of both local and national authorities. Education once again seems to matter more when it comes to explaining trust in national authorities (see Table 3 for results).

Table 3. Multilevel results: political trust.

	Local		National
Independent variable	Model 1	Model 2	Model 1	Model 2
Municipality level
Swedish-speaking Population (%), (0–1)	−0.091 (0.526)	0.012 (0.483)	0.142 (0.255)	0.333 (0.186)
Unemployment Rates (%), (0–1)	−1.648 (0.645)*	−1.629 (0.595)**	−0.551 (0.328)	−0.344 (0.251)
Type of Community (0 = urban, 1 = rural)	0.025 (0.319)	0.051 (0.293)	−0.148 (0.161)	−0.056 (0.126)
Individual level
Gender (0 = male, 1 = female)		−0.197 (0.081)*		0.069 (0.073)
Age (0–1)		−1.624 (0.680)*		−2.587 (0.609)***
Age ^2 (0–1)		2.513 (0.693)***		2.611 (0.623)***
Education (0–1)		0.391 (0.138)**		0.983 (0.123)***
SPP support(dummy: 0 = no, 1 = yes)		0.256 (0.092)**		0.241 (0.082)**
Random effects
Residual (Null model)	4.008 (0.108)	4.008 (0.108)	3.310 (0.089)	3.310 (0.089)
Residual	4.010 (0.108)	3.741 (0.109)	3.308 (0.089)	3.072 (0.089)
Intercept (Null model)	6.028 (0.134)	6.028 (0.134)	5.691 (0.059)	5.691 (0.059)
Intercept	6.824 (0.538)	6.547 (0.528)	5.939 (0.271)	5.524 (0.262)
ICC	0.067	0.057	0.011	0.001
Log-likelihood	−5925.6907	−5026.7887	−5639.472	−4761.5069
Municipalities	31	31	31	31
Respondents	2791	2407	2791	2404

Note: OLS, unstandardized coefficients; Multilevel, maximum likelihood estimates; Covariance type, unstructured. Standard errors are shown in parentheses: p < 0.05; **p < 0.01; ***p < 0.001.

To further test the robustness of our findings, we redo our analyses using alternative configurations and indicators for our variables of interest, including weights and robust standard errors (for the results of some of these analyses, see Appendices S5–S8 in the supplemental data online). Adding weights and robust standard errors to our standard models results in that ‘Unemployment rates’ is no longer statistically significant following the conventional threshold, however, it still has the by far largest coefficients in comparison to the other contextual predictors and it is significant on the p < 0.1 level in model 2 for trust in local authorities. We also try out an alternative indicator for local economic conditions, ‘Median income’, and it produces very similar results, however, this time it is highly significant (p < 0.01) when it comes to explaining local political trust in the models where weights are included, suggesting that trust in local authorities is weaker in municipalities with a lower median income. Likewise, we try an alternative indicator for the nature of the community (logged population size), but the results remain substantially unchanged. Additionally, we controlled for the respondents from the Åland Islands, since their social and political context differs much from the rest of Swedish-speaking Finland (Karv et al., 2022, p. 9), however, this did not change the results.

In sum, while the economic indicators are not always significant in our models, especially regarding corruption perceptions, they generally point in the same direction, namely that local corruption is perceived as more common and trust in local authorities is weaker in municipalities where local economic conditions are more challenging.

DISCUSSION

Given our findings, we did not find support for our first hypothesis, that political trust and corruption perceptions among a minority population should be more negative in linguistically more heterogeneous communities. Even though we did find some evidence that both the local and national levels of trust and corruptions perceptions differed based on the linguistic compositions of the municipalities, we were not able to confirm it statistically. Hence, even though earlier studies have found that the political and linguistic contexts in which minorities live matters for individual level political efficacy (Karv et al., 2022), the same does not hold for political trust and corruption perceptions. This was a bit surprising, given the usually strong correlation between political efficacy and political trust (Geurkink et al., 2020). However, we were able to confirm our second hypothesis, that political trust and corruption perceptions among a minority population should be more negative in economically worse performing communities. Our analyses were able to show that trust was higher and corruptions perceptions more positive the better the local economic conditions in the municipality. This also after changing the measurement of economic performance, adding further robustness to our finding. However, this only related to the evaluations of the local authorities. When it comes to the national level evaluations, the role of the municipality level economic performance was not as clear. Hence, citizens trust their local authorities for other reasons than they trust their national authorities (Fitzgerald & Wolak, 2016). Still, it might also be that it is easier for more homogeneous communities to create favourable economic conditions, as familiarity breeds trust and thus fosters cooperation, possibly leading to better economic performance (e.g., Alesina et al., 1999; Patsiurko et al., 2013). In that case, favourable economic conditions should be an indirect effect derived from linguistic homogeneity.

We found support, although inconclusive, for our third hypothesis, that political trust and corruption perceptions among a minority population should be more negative in larger urban communities. The levels of political trust were higher, and this especially toward the local authorities, while corruption perceptions towards the local authorities were also more favourable in the smaller, and more rural, communities. However, these initial findings were not confirmed after further analyses. Our findings from the multilevel regression models suggest that the local economic performance was able to predict individual-level corruption perceptions of and trust in local authorities, and this also after controlling for individual-level socio-demographic factors. Nevertheless, in relation to the evaluation of national authorities the local contexts were insignificant for explaining the variations. This confirms our fourth and final hypothesis, that local contextual factors are better at predicting varying levels of political trust and corruption perceptions towards the local than the national government. We further found that the local level was evaluated more favourably than the national level in terms of both corruption perceptions and political trust, which contradicts some of the earlier findings from Finland (Fitzgerald & Wolak, 2016; Salminen & Ikola-Norrbacka, 2010). Hence, members of linguistic minority seem to evaluate their local authorities in the same way as members of the majority population. Thus, there does not appear to be any ‘strength in numbers’ factor at play when it comes to corruption perceptions and political trust at the local level across the Swedish-speaking communities.

Moreover, SPP supporters were more likely to be politically trusting and possess a favourable view of the corruption level, irrespective of institutional level. In terms of political trust, this broadly confirms earlier findings (Grönlund & Isaksson, 2008), while it adds new insights into the relation between supporting SPP and corruption perceptions. Whether this relation between supporting a minority party and having more favourable corruption perceptions is unique for this given context or whether it also says something about the larger phenomenon is outside the scope of the current study. Based on our findings, and in relation to broad research question stated in the introduction, we believe that we have shown that some extent of political (dis)satisfaction within a minority setting is, indeed, related to local level characteristics.

LIMITATIONS

There are several limitations with this study, with the perhaps largest being the lack of comparable survey data measuring corruptions perceptions and political trust among the Finnish-speaking population. Hence, we are not here able to make any larger suggestions regarding the minority–majority relation in Finland and to what extent the local characteristics affect corruption perceptions and political trust more generally across Finland. Therefore, we do not know whether our findings relate solely to the minority population our whether this is a larger national phenomenon. Moreover, even though we were able to control for many of the probable local level determinants, we did not have sufficient data for exploring the role of actual corruption across the municipalities. Assessments from local experts on the extent of corruption in different municipalities would be a good complement to this data and would allow us to examine if there are any commonalities between elite and citizen perceptions (e.g., Erlingsson & Kristinsson, 2016).

Finally, although we use panel-based survey data we do not have data on the same survey items for the same respondents across multiple time points, which would allow us to track potential changes across time. This would enable us to investigate if changes in our dependent variables are related to changes in our independent variables, and hence draw more robust conclusions concerning causality. Furthermore, given that we are using survey data collected from two separate surveys, although collected from the same survey panel, it makes making first-hand comparisons between the contextual effects on political trust and corruption perceptions a bit more challenging. Even as this might be considered to constitute a minor limitation for making comparative inferences from our statistical analyses we do not, however, believe this to affect the reliability of our statistical results to any problematic extent.

CONCLUSIONS

In this study, we have explored how members of a minority population evaluates local and national authorities while further establishing the role of municipality-level determinants for explaining corruption perceptions and political trust. We expected that in local contexts where the minority population had a stronger position, the local authorities would be perceived more favourably. Furthermore, we expected the local authorities to be seen more favourably the more economically prosperous the community. Finally, we expected the local conditions to be more related to evaluations of local than national authorities. Our findings show that in Finland, the members of the minority population clearly differ between the local and national authorities, and that the local economic conditions can explain some of the cross-municipal variations in local political trust and local corruption perceptions. However, the role of linguistic homogeneity and community size did not seem to be relevant when controlling for the role of economic conditions. What then to make of these findings?

One potential policy implication derived from our findings is that the key to fostering trust in local political authorities is not necessarily to encourage the creation of linguistically homogenous minority enclaves. This is also in line with earlier findings from the social capital literature, which has even found that the most socially trusting members of the French-speaking community in Canada can be found in the most linguistically diverse communities (Bridgman et al., 2021). Consequently, the key factor for fostering trust among a linguistic minority population is the same as for members of the majority population, improve the local economic outlook (Weinschenk & Helpap, 2015). The local political authorities seem to be judged based on their ability to fulfil this task, and when doing so they are subsequently able to earn the trust of the local population.

In Finland, the role of municipalities is changing as a new administrative level in Finland is set to be implemented. This will transfer responsibility of the social- and healthcare sector from municipalities to a new regional administrative level, simultaneously transferring a substantial amount of the local budgets to the aforementioned. This will create a situation where the locals will have a hard time knowing where the responsibility lies, and thus extending the width between the demos and the kratos when it comes to the most important public service provided in Finland, healthcare. This will subsequently also directly affect the Swedish-speaking Finns, as the influence of SPP is set to decline when decision-making is transferred from municipalities with a Swedish-speaking majority to the larger regional institutional structures were the Swedish-speaking Finns will only, in most regions, constitute a small minority. When the political influence of SPP declines, it means that the Swedish-speaking population in these regions will become even more dependent on a benevolent majority in order to safeguard their status. This over time might have far-reaching consequences for the Swedish-speaking population in Finland. On a more general note, the role of local governments in the everyday lives of Finnish citizens is visibly declining, which might contribute to an alignment between how the local and national levels of government are being evaluated in the future.

DISCLOSURE STATEMENT

No potential conflict of interest was reported by the authors.

Additional information

Funding

The research was made possible by funding and support from Svenska Litteratursällskapet i Finland (The Society of Swedish Literature in Finland) (projects Det politiska förtroendet i Svenskfinland (T.K.); and Smolk i ankdammen? Uppfattningar och attityder angående korruption i finlandssvenska kommuner (F.M.)).