Who gets the blame and who gets the credit? Policing, assistance, and political trust among the Roma in Europe

ABSTRACT

In many European countries the Roma are the largest minority group, and research often highlights their heightened exposure to discrimination, harassment, and even abuse during interactions with the state, and the majority group. In contrast, many governments have assistance programs targeted to the Roma in an effort to boost integration. However, there is strikingly little systematic quantitative research on how these experiences are related to the trust that the Roma place in political institutions. This study addresses this gap by using EU-MIDIS II data from 9 European countries to assess the relationship between political trust and the experiences the Roma have with the police and assistance programs. Our analyses show that different experiences relate to trust in institutions differently: interactions with the police, either by being stopped or assaulted are most strongly associated with lower trust in the police but also reduce trust across institutions. In contrast, local governments may have the most to gain from those who report having received assistance based on their minority membership, with other institutions receiving modest trust benefits. Finally, both overall personal experiences of discrimination and the perceived extent of discrimination in their country are associated with low levels of trust in the institutions.

KEYWORDS:

• political trust
• police
• social services
• the Roma
• EU-MIDIS

Introduction

Political trust, the belief that the government is benevolent, responsible, reliable, and capable, has many known societal benefits. High political trust is linked to economic growth (Fukuyama 1995), increased support for the welfare state (Rothstein and Stolle 2008), environmental policies (Fairbrother 2016), and stable functioning democratic regimes (Mishler and Rose 2001). Yet despite these benefits, little is known about the political trust among minority groups in Europe, or what factors might be influencing them. When this research topic is addressed it typically is analyzed through the lens of immigrant minorities in relation to trust in the police (see, for example: Czymara and Mitchell 2022; Kääriäinen 2007; Röder and Mühlau 2012). Still, despite being the largest minority group in many European countries, relatively little is known about how much the Roma trust different political institutions or what factors relate to their political trust.

Scholarship on political trust often highlights institutional quality as a main factor for different levels of trust across societies. A key aspect of institutional quality is how well norms of procedural justice are followed by street-level representatives of the state, which include the police and the people that work at agencies that distribute welfare state benefits. Of the previous scholarship on the Roma, what is often highlighted are the various forms of discrimination, harassment, and even violence that this group faces as a function of their minority status in Europe. This is true of their interactions with various state agencies, but also in daily life with other citizens (European Union Agency for Fundamental Rights 2022). The violation of norms of procedural justice suggest that these types of experiences should reduce trust across political institutions since people generalize their experiences (Fairbrother et al. 2022). Conversely, many EU countries have established assistance programs which target Roma populations in an effort to boost integration and reduce inequality. Being a recipient of this type of assistance, in contrast, might bolster perceptions of institutional quality and result in higher trust across political institutions. In addition to this, being part of a minority that receives higher levels of harassment because of their ethnicity may have an additional negative effect on trust in institutions.

As an analytical guide to investigate these propositions, we ask: Who gets the blame and who gets the credit when the Roma have interactions with various arms of the state? However, we begin with an overview of the Roma in Europe, outlining their special position in society and different explanations for their documented disparate outcomes compared to other groups and how this should impact their trust in different political institutions.

The Roma in Europe

While the number of Roma in Europe is unknown and contested, both the Council of Europe and the European Union provide a similar estimated population of 10–12 million Roma in the continent, approximately 6 million of them living in the European Union (Council of Europe 2012; European Comission 2020). The term ‘Roma’ was adopted at the first World Roma Congress held by activists in London in 1971 to include diverse groups like the Roma, Sinti, Kale, Romanichels, Boyash/Rudari, Askhali, Egyptians, Yenish, Dom, Rom and Abdal, as well as Travelers populations. This is also the official term used by both the European institutions and most of the member states. The common origin of this diverse population is supported by historic, linguistic and genetic studies, pointing to the fact that, against common misconceptions, many Roma have been living in the same geographic areas and have mixed with the other local native populations in different degrees for generations (Gamella and Carrasco-Muñoz 2017; Matras 2015; Orton et al. 2019) which has resulted in considerable variation in their social forms (use of the language, religion affiliation, economic conditions, housing, and others). Furthermore, their countries of residence present important variation in their political history (e.g. dictatorships and democracy development), the orientation of the welfare state and public policies (those addressed to the general population, but also the policies explicitly designed for the Roma), in addition to other cultural and economic factors that have shaped the Roma experience in Europe.

Despite the diversity of historical contexts, the Roma have suffered a common history of repression and persecution by the state and related institutions (i.e. churches) with political power. The Roma were, for example, enslaved for centuries in Romania and other countries (Hancock 2005); they were one of the main targeted groups of the Nazi Holocaust (Hancock 2005); they were the unique target of the Gran Redada (‘the Great-Round-up’), to imprison all the Roma in Spain in the eighteenth century (Gómez Alfaro 1993); and recently they were recognized as victims of forced sterilization in countries like Sweden in the 1950s and Czechoslovakia after 1966 (Minister of Culture 2015; Zampas and Lamačková 2011). The Roma are also the victims of direct acts of violence by the non-Roma (gadjo) majority groups in the form of pogroms, lynchings and arson attacks, among other forms of aggression (OHCHR 2015). Indirect forms of violence include the systematic effort in official policies and legislation to assimilate the Roma to the general population, intending to remove their cultural practices and suppress their identification as Roma (Barany 2002).

Contemporarily, reliable statistical information about the Roma is still scarce and limited in scope and quality. Nevertheless, the available data about the Roma in Europe indicates that compared with the other social groups the Roma are among the most disadvantaged in economic terms, and likely to be those suffering from racism in its different forms. Data about the living conditions indicate that the Roma suffer from higher levels of poverty and inequality. In the countries included in this study, 80% are estimated to live under the at-risk-of-poverty threshold of their country (estimated as being below the 60% of the median income), one in three live in households without running water, every tenth without electric power, and more than one quarter experienced hunger at least one night in the last month (FRA 2018). From these sources it would be very difficult to identify countries were the Roma enjoy the best social conditions. For instance, Spain is one of the countries with better indicators in some dimensions (like housing services or access to health care) but falls very short in many other indicators (poverty rates, labor market participation and access to paid work, educational attainment, and others).

Most recently the European Union is taking steps to address some of these historical inequalities and abuses, by creating a strategic framework for equality, inclusion and participation targeted at the Roma (European Comission 2020). These measures come in response to varying levels of social support provided by the EU member states to the Roma living within their borders. How much money each member state dedicates to Roma support initiatives is difficult to track, but the section of the European Social Fund dedicated to the integration of marginalized communities such as the Roma reports having spent 1.5 billion Euros and including 3.9 million participants. The variation in how well these resources are reaching the Roma across countries are reflected in the EU-MIDIS II survey (FRA 2017). For example, the proportion of Roma that report having received support in the in various forms such as housing, education, job training, or financial assistance ranges from 52% in Hungary to 23% in Bulgaria.

Theory: institutional racism, support, and trust

The literature on political trust suggests that the success of these social programs to assist the Roma will depend, at least in part, on the trust relationship between the Roma and state entities. Some base line level of trust is often viewed as a pre-requisite for economic activity (Fukuyama 1995), so even assuming there are ample funds to assist those in need, if the Roma believe that the government is discriminatory, corrupt, dishonest, incompetent, or otherwise dysfunctional; it will be very difficult to make any material impact. To explain different levels of political trust, researchers often point to the quality of institutions themselves as being responsible for building or eroding trust since governments with high institutional capacity and adherence to the rule of law are associated with high levels of trust among their citizenry (Mishler and Rose 2001; Rothstein and Stolle 2008). Indeed, just as distrusting trustworthy institutions can contribute to societal problems, blindly trusting institutions that may cause one harm is similarly problematic (Norris 2022). In this literature, the specific situation of minorities is often omitted, so little is known about how institutions are perceived by marginalized groups or how their encounters experiencing discrimination, or conversely being the recipient of social assistance may impact their trust.

As a point of departure, we recognize that historically rooted structural and institutional racism amount to profound violations of norms of procedural justice. In reference to the historic experiences of the Roma in Europe, the concept of historical trauma is useful, and refers to ‘a complex and collective trauma experienced over time and across generations by a group of people who share an identity, affiliation, or circumstance’ (Mohatt et al. 2014). The specific form racism facing the Roma population, anti-gypsyism (Howard and Vajda 2016), are comprised of long-term intergenerational experiences of unfair treatment which engenders distrust. This is often referred to as the ‘cultural explanation’ of trust (Dinesen 2012), and is typically used to explain trust differences between immigrants and natives, but in this case it would predict low levels of trust in political institutions overall due to being socialized in separated minority communities. However, recent research suggests that political trust is often particularized to different institutions and emphasizes the importance of two key mechanisms: the perceived fairness on the parts of the state that are charged with enacting regulations, and interactions with street-level bureaucrats (i.e. the police) (Czymara and Mitchell 2022; Fairbrother et al. 2022; Schnaudt, Hahn, and Heppner 2021). In both, the police play an outsized role in the perceptions of legitimacy people have of the state, both in terms of the type and frequency of their interactions.

To illustrate with a classic example, the 1922 report The Negro in Chicago, issued by the Commission on Race Relations was designated to produce recommendations to avoid race riots like those that had occurred a few years prior. It pointed out that distrust in the police caused by the discriminatory practices of officers were a major contributor to the riots, and recommended better policing practices to prevent future violence (Chicago Commission on Race Relations 1922). A century later, similar racist practices by the police have inspired the Movement for Black Lives (M4BL), and new reports with more detailed recommendations have been produced (Phelps, Robertson, and Powell 2021) based on the idea that better policing practices would improve race relations. If true, then the police have the potential to build trust, as Bell (2017) suggests, ‘police officers treating people with dignity and respect, behaving in a neutral, nonbiased way, exhibiting an intention to help, and giving people voice to express themselves and their needs in interactions’, act as the main mechanism to promote political trust and facilitate positive interethnic relations.

It’s also possible that this concept places too much of the responsibility for political trust on the police. Sampson and Bartusch (Sampson and Bartusch 1998), elaborating on the classical concepts of anomie, normlessness and social deviance, used the term ‘legal cynicism’ to describe situations where the communities express cynicism about the legitimacy of laws and the ability of police to do their job in an effective nondiscriminatory manner. In their analysis, legal cynicism was explained by structural neighborhood-level differences like concentrated economic disadvantage, residential stability, and concentrated immigration which minimize the contribution of interpersonal factors, such as procedural justice or policing practices. This idea was also expressed in The Negro in Chicago report referenced previously, wherein most of the recommendations are about other structural aspects like education, employment, housing, mass media, public opinion, and other social sectors (like churches, unions or clubs), implying that a wider and more comprehensive anti-racist agenda is needed to build trust among marginalized groups.

If too much importance is placed on the police, then research should turn to other community intervention practices to assess their trust building or eroding potential. Recently, it is the case that governments are offering services targeted to the Roma in an effort to offset some of the consequences of these historically structured inequalities. While the trust literature would argue that the effectiveness of these programs are at least to some extent dependent on the level of political trust of the people receiving them (Mensah and Adams 2020), both the EU and respondents to this survey report that at least some services are reaching the Roma in the countries included in this study. Still, much of the literature that focuses on how trust can be increased have typically focused on strengthening institutions (Mishler and Rose 2001) or more recently, increasing levels of direct democratic participation (Christensen 2019). It is also the case that having positive experiences such as receiving help on the basis of a persons’ historically marginalized status should restore some legitimacy to the state. In other words, the spending targeted towards the Roma should restore some of the trust that has been eroded by the mechanisms of structural exclusion and legal estrangement. There is a burgeoning literature which investigates this claim especially in relation to welfare state benefits (Kumlin and Haugsgjerd 2016), and that welfare state interventions can be especially important for trust during times of crisis (Ellinas and Lamprianou 2014). Indeed, recent research has shown that increased social spending on infrastructure is followed by an increase in trust in local government (Li and Mayraz 2017), which suggests that this type of intervention can have an effect. However, they also show that the positive impact of economic stimulus on trust was mostly restricted to local government suggesting that ‘who gets the credit’ in the form of higher trust should also be particularized. We incorporate this question, with an assessment of whether the respondents have been the recipient of some type of assistance on the basis of being Roma, is associated with different types of political trust.

This interplay between long-term structural influencers and individual level experiences would predict lower levels of political trust in the police and other social institutions at the aggregated level among racialized and minoritized populations, because of marginalization and structural exclusion (Sampson and Bartusch 1998; Schroedel et al. 2020), but also even lower levels of trust in the police among victims of procedural injustice at the individual level. Furthermore, parsing out whether it is the discriminatory experiences with state actors versus non-state actors (i.e. landlords, employers, etc.) should be import since recent research in Canada shows that the relationship between discrimination experiences and political trust is complex and dependent on the racialized groups having the experiences (Wilkes and Wu 2019). In line with the part of the prediction at the aggregate level, the EU-MIDIS report (FRA, 2017: 100) compares the levels of political trust among the Roma with the general population based on data from the European Social Survey (2014) for the four countries with comparative data on the Roma (the Czech Republic, Hungary, Portugal and Spain). In their analysis, the lower levels of trust in the police are reported by the Roma in three of the four countries (Hungary levels are among the same in both populations). Nevertheless, levels of trust in the legal system were similar to those for the general population. We test the second part of this assertion by exploring the association between experiences of procedural (in)justice (measured as individual experiences with the police) and trust in the police, we also consider if those experiences are related to the level of trust in other political institutions that are not connected with the police and the legal system, and we do so while taking other types of discrimination into account.

To summarize and return to the question of ‘who gets the blame and who gets the credit?’ when the Roma interact with different arms of the state; we believe that the strength of the relationships between experiences and political institutions will be informed by both the historical context of the Roma in Europe as well as the experiences with the institution in question. This would lead us to expect that experiences of procedural (in)justice such as corruption will be associated with lower levels of trust in the local municipal officials, but also in the local police, national politicians, and the legal system in general. We see this as reasonable because of how institutional actors have behaved in the past toward this group, leading them to attribute their experiences to all institutional actors. In contrast, we expect the more positive interactions of receiving assistance to be particularized to local level municipal officials and the police. This is partly because of the previous literature on welfare state interventions we outlined above, and partly because the social services programs are relatively new in comparison to the long history of the Roma in Europe, meaning that the recipients of those benefits will only attribute that experience to those agencies they interact with often. We also investigate the possibility that some of the trust eroding effects of negative experiences with state institutions can be offset by having received social assistance (i.e. does the credit outweigh the blame?).

Data and methods

The population estimated for the 9 countries covered in this study varies considerably in size from a max of about 2.5 million Roma in Romania to a minimum of about 30 thousand in Croatia (average estimates as a percentage of the total population: 0.5% Portugal, 0.8% Croatia, 1.6% Greece, 1.6% Spain, 1.9% Czech Republic, 7.5% Hungary, 8.6% Romania, 9% Slovakia, and 10% Bulgaria) (Council of Europe 2012). One common issue with researching minority populations such as the Roma are the relatively small sample sizes that are produced by traditional random sampling procedures that are used frequently in values and attitudinal research. Also, in many general social surveys, the Roma cannot be identified as they do not include identifying information about the ethnicity of the interviewees. To address this, we analyze data from the second European Minorities and Discrimination Survey (EU-MIDIS II). This survey oversampled minority groups in the EU, and the Roma were studied in 9 European countries (Greece, Portugal, Croatia, Spain, Czech Republic, Slovakia, Hungary, Romania, and Bulgaria) with a total of 7947 respondents. The oversampling procedure varied depending on the targeted group, with a combination of different methods to ensure greater coverage.¹ All interviews occurred face to face and were computer assisted taking place between September 2015 and October 2016.

Dependent variables

We include four measures of trust in political institutions, two that capture respondents’ evaluations of local level institutions and two that capture evaluations national level institutions. The two items at the local level measure trust in the police and trust in municipal officials, and the two at the national level measure trust in the legal system and trust in politicians. These questions read as follows:

‘Using this card, please tell me on a scale of 0–10 how much you personally trust each of the [COUNTRY] institutions I read out. 0 means you do not trust an institution at all, and 10 means you have complete trust.

• the local (municipal) authorities in the place where you live

• [COUNTRY]’s police?

• [COUNTRY]’s legal system?

• [COUNTRY]’s politicians?’

Each of these items was measured on an 11 point scale (0 =  ‘No trust at all’, 10 =  ‘Complete trust’).

While our analysis is primarily interested in how different experiences influence trust in political institutions differently, we also created an institutional trust index by averaging the four items to assess how some experiences are related to trust in political institutions overall. The survey response items are correlated moderately high within individuals between .55 and .73 and have a high Cronbach’s alpha (.86) giving confidence to the reliability of the political trust index item.

Key independent variables

To measure who gets the blame, our key independent variables are measures of whether the respondents have been stopped by the police (within the last year or last five years), whether they have had experiences of corruption in the form of a request or the expectation of a bribe from a police officer, judge or inspector, and whether they report having been assaulted by the police. These three measures capture, albeit crudely (1) whether the respondent has had an interaction with the police, (2) how recently this interaction has occurred and (3) two measures of mistreatment. To measure who gets the credit we include whether the respondents report having received support on the basis of their minority status.

These questions read as follows:

• In the past 5 years in [COUNTRY] (or since you have been in [COUNTRY]), have you ever been stopped, searched, or questioned by the police? IF LESS THAN 5 YEARS IN COUNTRY, ‘SINCE YOU’VE BEEN IN THE COUNTRY’

• And has this happened to you in the PAST 12 MONTHS?

• In the past 5 years in [COUNTRY] (or since you have been in [COUNTRY]), has a police officer ever physically assaulted you because of your: Roma background / ethnic minority background? By this I mean something like being pushed, hit or kicked. IF LESS THAN 5 YEARS IN COUNTRY, ‘SINCE YOU’VE BEEN IN THE COUNTRY

• In the past 5 years did any government official in [COUNTRY], for example a customs or police officer, a judge or an inspector, ask you or expect you to pay a bribe for his or her services?’

• Have you ever received any of the following types of help or support from a public institution or NGO in [COUNTRY] because you are from an/a: Roma background / ethnic minority background]?

Controls

In an attempt to isolate the relationships between trust and their experiences with our key independent variables, we include other predictors which should erode political trust such as whether they have reported discrimination in work, healthcare, housing, or schooling on the basis of their skin color ethnic origin or religion. Furthermore, to separate the respondents’ experiences with these institutions from their overall evaluations of the state-of-affairs in their country, we include a scale item measuring their perceived prevalence of discrimination in their country on the basis of skin color, ethnic origin or immigration background, and religion. While this measure is different from other studies where the overall state-of-affairs evaluations involve metrics of levels of perceived democracy (Hooghe, Marien, and Oser 2017), we believe it is a useful in this case, and is arguable a better control because it is particular to discrimination. As additional control variables we include age, gender, whether they are engaged in paid work, and education. Last, since recent longitudinal evidence points to social isolation (both real and perceived) as a predictor of political trust we include number of people in their household and marital status² (Langenkamp 2022).

We analyze these data with OLS linear regression that include individual level post stratification weights calculated for the Roma target group. Due to the large amount of variation on the political trust items between countries, we include country dummies, but models without these dummies can be found in the appendix. We begin by reporting the relationships that experiences with the state have with respondents’ overall trust in political institutions captured by the political trust index. To assess who gets the blame and who gets the credit, we run separate models for each of the four trust dimensions. Since each trust item is measured on the same scale, and the models are otherwise identical, we put these relationships into perspective by plotting the point estimates and confidence intervals next to each other in a dot-whisker plot. For simplicity in comparison, we report the four items and their main predictors and control variables of interest. We also assess the possibility that the relationship between political trust and trust eroding experiences with discrimination, or the police may be offset by having received services by interacting those predictor variables.³ Plots were made using the sjPlot package in R (Lüdecke 2022).⁴

Descriptive results

Average trust in political institutions varies considerably between countries (Table 1) and across trust dimensions (Figure 1). For example, trust is highest in local government in most countries with the highest average trust in Portugal (mean 5.73 sd 1.75) and Romania (mean 5.75 sd 2.98). On average the Roma in Portugal also express the lowest average trust in an institution, the country’s politicians (mean 2.18 sd 2.11). The variation within these descriptive statistics show that the Roma are not a monolith in their trust evaluations, either in terms of their country context or the political institution type they are evaluating. This is also represented in the Roma’s average institutional trust for the scale item, which is highest in Romania (mean 5.02 sd 2.71) and lowest in Spain (mean 3.49 sd 2.63).

Figure 1. Density plot of different political trust measurements among the Roma including their means and standard deviations. Data: EUMIDIS-II.

Table 1. Descriptive statistics of the political trust variables. Data: EUMIDIS-II.

	Overall	Bulgaria	Croatia	Czech Republic	Greece	Hungary	Portugal	Romania	Slovakia	Spain
	(N = 7947)	(N = 1078)	(N = 538)	(N = 817)	(N = 508)	(N = 1171)	(N = 553)	(N = 1408)	(N = 1098)	(N = 776)
Political trust index
Mean (SD)	4.27 (2.51)	3.80 (2.39)	4.13 (2.42)	4.28 (2.43)	5.10 (1.98)	4.54 (2.59)	3.97 (1.77)	5.02 (2.71)	3.87 (2.39)	3.49 (2.63)
Missing	97 (1.2%)	4 (0.4%)	0 (0%)	1 (0.1%)	9 (1.8%)	32 (2.7%)	25 (4.5%)	8 (0.6%)	12 (1.1%)	6 (0.8%)
Trust in police
Mean (SD)	4.63 (3.06)	4.49 (3.19)	5.01 (3.17)	4.25 (2.82)	5.63 (2.70)	4.83 (2.91)	3.98 (2.63)	5.38 (3.14)	3.93 (2.92)	4.07 (3.24)
Missing	162 (2.0%)	11 (1.0%)	7 (1.3%)	3 (0.4%)	13 (2.6%)	42 (3.6%)	30 (5.4%)	11 (0.8%)	36 (3.3%)	9 (1.2%)
Trust in municipal officials
Mean (SD)	5.19 (2.98)	5.14 (3.31)	4.15 (2.95)	5.41 (2.87)	5.17 (2.23)	5.38 (3.04)	5.73 (1.75)	5.75 (2.98)	5.44 (2.97)	3.71 (3.05)
Missing	200 (2.5%)	12 (1.1%)	6 (1.1%)	4 (0.5%)	30 (5.9%)	41 (3.5%)	27 (4.9%)	15 (1.1%)	43 (3.9%)	22 (2.8%)
Trust in the legal system
Mean (SD)	4.11 (2.98)	3.17 (2.95)	4.28 (2.97)	4.14 (2.73)	5.66 (2.46)	4.55 (2.93)	3.70 (2.37)	4.98 (3.14)	3.32 (2.85)	3.58 (2.98)
Missing	328 (4.1%)	31 (2.9%)	11 (2.0%)	16 (2.0%)	72 (14.2%)	67 (5.7%)	43 (7.8%)	29 (2.1%)	34 (3.1%)	25 (3.2%)
Trust in politicians
Mean (SD)	3.03 (2.73)	2.30 (2.58)	3.00 (2.79)	3.12 (2.47)	3.51 (2.13)	3.30 (2.74)	2.18 (2.11)	3.90 (2.91)	2.90 (2.86)	2.48 (2.70)
Missing	288 (3.6%)	22 (2.0%)	6 (1.1%)	38 (4.7%)	54 (10.6%)	60 (5.1%)	38 (6.9%)	19 (1.3%)	30 (2.7%)	21 (2.7%)

Analytic results

We first assess whether experiences with policing and receiving assistance are related to overall levels of trust. Figure 2 shows the point estimates and confidence intervals of the OLS model with the political trust index dependent variable. Both having been stopped by the police (−.47 within the last year, −.43 1 year or longer) and those who report being assaulted by the police (−.62) report lower overall trust in political institutions. In contrast, those who report having received some type of governmental support report higher average trust (.37). These results suggest that interactions with street level bureaucrats are both positively and negatively associated with political trust overall. However, the largest association with lower overall political trust is having had personally experienced discrimination (−.64), this is true despite these experiences not being directly associated with the institutions in question (e.g. work and housing as opposed to the police). This association between discrimination experiences and trust remains even controlling for the perceived levels of discrimination (−.34). Surprisingly, the estimate for corruption experience is relatively small (−.40), which runs counter to the political trust literature that highlights the importance anti-corruption measures as central for high levels of political trust.

Figure 2. Dot whisker plot of OLS coefficients with country level dummies (not shown). Data: EUMIDIS-II.

Despite the high Cronbach’s alpha, we can imagine that experiences with the state can relate to trust in political institutions differently. Interestingly, whereas experiences with corruption did not have a strong relationship with the political trust index, it is the police that seem to get the blame when people have these experiences (more so than the legal system or local government, for example −.45 versus −.37 and −.15 respectively). Trust is reduced in all institutions when stopped by the police, and like the previous analysis this relationship is stronger if respondents have been stopped more recently. Intuitively, the largest negative association is between those who report having been assaulted by the police and their trust in that institution, but interestingly there are what could be considered ‘spill-over’ effects, with local government and the legal system which explain the negative association with assaults and overall political trust in Figure 3. There are trust benefits across all political institutions when people are the beneficiaries of some type of service (police = .27, legal system = .22 municipality = .56, politicians = .31). This finding is in line with our expectation that local officials have the most to gain from providing services to the Roma, but the positive associations between having received benefits and the other institutions cannot be overlooked and are positive ‘spill-over’ effects. Finally, discrimination experiences have a similarly consistent, albeit negative, relationship with trust across political institutions included in the analysis. And, with a weaker relationship, also perceived levels of discrimination (Table 2).

Figure 3. Dot and whisker plot of coefficients from models with different dependent variables. Data: EUMIDIS-II.

Table 2. OLS regression models across different trust items. Data: EUMIDIS-II.

	Trust index	Police	Legal system	Municipal officials	Politicians
(Intercept)	4.55 ***	5.20 ***	3.77 ***	6.28 ***	2.93 ***
	(4.11–4.99)	(4.66–5.74)	(3.24–4.31)	(5.74–6.82)	(2.43–3.42)
Discrimination experiences	−0.64 ***	−0.60 ***	−0.65 ***	−0.71 ***	−0.52 ***
	(−0.77 – −0.51)	(−0.75 – −0.44)	(−0.80 – −0.49)	(−0.87 – −0.56)	(−0.67 – −0.38)
Perceived levels of discrimination	−0.34 ***	−0.35 ***	−0.32 ***	−0.38 ***	−0.33 ***
	(−0.40 – −0.28)	(−0.43 – −0.28)	(−0.39 – −0.24)	(−0.45 – −0.30)	(−0.39 – −0.26)
Corruption experiences	−0.26	−0.45 *	−0.37	−0.15	−0.07
	(−0.57–0.05)	(−0.83 – −0.08)	(−0.75–0.01)	(−0.54–0.23)	(−0.41–0.28)
Stopped: Over 1 year ago	−0.43 ***	−0.43 **	−0.24	−0.38 **	−0.53 ***
	(−0.66 – −0.20)	(−0.71 – −0.15)	(−0.51–0.04)	(−0.66 – −0.10)	(−0.79 – −0.28)
Stopped: within the last year	.47***	−0.56 ***	−0.34 **	−0.43 ***	−0.52 ***
	(−0.66 – −0.29)	(−0.79 – −0.33)	(−0.57 – −0.12)	(−0.66 – −0.20)	(−0.73 – −0.31)
Police assaulted	−0.62 **	−1.41 ***	−0.24	−1.04 ***	0.27
	(−1.03 – −0.21)	(−1.91 – −0.91)	(−0.74–0.26)	(−1.55 – −0.53)	(−0.19–0.74)
Received support	0.37 ***	0.27 ***	0.22 **	0.56 ***	0.31 ***
	(0.25–0.49)	(0.13–0.42)	(0.08–0.37)	(0.41–0.70)	(0.17–0.44)
Observations	7212	7163	7017	7125	7054
R^2/R^2 adjusted	0.111/0.108	0.090/0.087	0.098/0.095	0.106/0.103	0.084/0.081

* p < 0.05 ** p < 0.01 *** p < 0.001.

All models include control variables not shown: household size, age, gender, engaged in paid work, education, marital status and country dummies.

It could also be that the associations between political trust and the trust eroding predictors (for example: having discrimination experiences or having had interactions with the police) are offset by having received services on the basis of the respondents’ ethnic background. To test this possibility, we interacted these terms.⁵ Figure 4 presents the predicted values of the interaction terms of discrimination experiences on the left, and encounters with the police on the right, with having received support. The figure on the left shows no substantial interaction effects on trust, with those having had experiences of discrimination being substantially less trusting than those who have not, regardless of whether they have also received social services. The figure on the right shows a slightly more substantial drop between those who have been stopped by the police for those that have received social services than those that have not, but also that the relationship between those that have been stopped by the police within the last year is similarly pronounced for those that have received social services. It should be noted that in both figures, the predicted values of trust are highest for those that have received social services and those that have not experienced discrimination or been stopped by the police.

Figure 4. Plot of interaction effects on political trust: discrimination experiences and having received support (left), stopped by police and having received support (right). Data: EUMIDIS-II.

Discussion

The results from this analysis show a complex picture about which types of experiences are related to which types of political trust amongst the Roma in Europe. For example, interactions with the police, and having received support are related to trust across political institutions (both negatively and positively, respectively). While these relationships are not causal in nature, they suggest that these types of encounters with the state have a generalizing effect across political institution type, and are strongest just after the experience occurs which support findings in other contexts with panel data (Fairbrother et al. 2022). In other words, everyone gets the blame/credit.

We also find that those that have experienced discrimination in interactions with other social organizations related to, but not necessarily representative of, political institutions (i.e. work and housing) generalize these experiences to political institutions as well. In fact, this is the largest individual level predictor of overall political trust, which highlights the importance of discrimination in various aspects of society in eroding trust. These findings support the institutional quality thesis (Mishler and Rose 2001; Rothstein and Stolle 2008), even among the countries in our sample which could be characterized by comparatively low quality of governance, and a sub-population that is historically marginalized. Unfortunately, the trust building impacts of having received social services were not able to mitigate the trust eroding impacts of having been stopped by the police, or discrimination. We take this finding as evidence that to restore trust among marginalized populations like the Roma, governments need to work toward a broad anti-racist agenda that incorporates both strengthening norms of procedural justice, anti-discrimination efforts, as well as social services aimed at reducing inequalities.

However, incorporating measures of the type of interaction shows particularized relationships to the political institution in question. Our analyses show that when respondents report being assaulted, or having had an experience with the police or judicial system they evaluate as corrupt, these experiences are most strongly related to the police and not to local governments, the legal system, or politicians. This is positive since having negative experiences will not necessarily impact their trust in political institutions that operate on more aggregate levels. These results support recent research that suggest that people particularize their experiences, especially when they represent breaks in norms of procedural justice (Czymara and Mitchell 2022; Schnaudt et al. 2021).

While not the central focus of this study, we find it worth mentioning that there is considerable variation across the 9 countries included in this study as is shown in Table 1. This descriptive finding is, in our opinion, itself interesting considering the long history of oppression coupled with their contemporarily marginalized position in society. Given this history (and conventional wisdom about this group) we would expect lower levels of political trust across the board. While not being able to draw conclusions from these differences we would just highlight that in many countries trust in political institutions, particularly local government, is comparatively high, meaning that trust should not be a prohibiting factor in launching more robust social assistance programs in the future.

Taken together with the importance of political trust in impacting behavior amongst marginalized groups (Schroedel et al. 2020), this is an encouraging finding. In the future, data that incorporate the type of oversampling procedures used in the EU-MIDIS II which include more countries would allow use of multilevel models to analyze these differences to determine what factors at the societal level are related to political trust among the Roma. In addition to the limited number of countries included, there are other data related limitations to this study which should be taken into consideration. Many variables (such as information about respondent’s parents’ birthplace) are not available, meaning we are unable to assess differences between those who are newly arrived in a country from those who have spent long periods of time, or their entire lives there (Czymara and Mitchell 2022). This also means that we are unable to include any regional or local level predictors, which theoretically should be important given the high levels of segregation the group is subjected to. Additionally, we are unable to make comparisons in trust between the Roma and the majority, or otherwise ‘native’, groups in these countries.

Despite these limitations, our contributions to the growing field of political trust remains. However, perhaps most importantly, we offer the first systematic assessment of political trust amongst the Roma, a group for whom trust is disproportionately important in terms of their relationship with the state and rectifying past injustices. As we have mentioned, the Roma are the largest ethnic minority group in Europe yet there is a glaring gap in (particularly quantitative) research assessing their attitudes. While this study helps to close this gap, considerably more research is needed.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Additional information

Notes on contributors

Jeffrey Mitchell

Jeffrey Mitchell is a sociologist and postdoctoral researcher at the Department of Sociology at Umeå University in Sweden. His research focuses on a broad range of topics including prejudice, trust in political institutions, and the effects of social policies and policing.

Daniel La Parra-Casado

Daniel La Parra-Casado is a Professor of Sociology at the Department of Sociology, University of Alicante. His research focuses on social stratification and health, with an intersectional focus (ethnicity, migrant status, social class, and gender).

Notes

1 For a more detailed description of the sampling procedure see the Second European Union Minorities and Discrimination Survey Technical Report. https://search.gesis.org/gesis_bib/gesis-bib-153654.

2 For specific wordings of all variables in the questionnaire see the appendix.

3 It is also possible that these relationships are dependent on where they take place. To test this, we include interaction models with the main predictor variables and country dummies, and present these plots in the appendix, but we find little variation in the relationships across countries in the sample.

4 The data are publically available and the R code for data cleaning and analysis are available at https://osf.io/f4chb/.

5 We also include a model with the effects of having been assaulted by the police and having received support in the appendix. Similar to the results presented here, having received support does not offset the trust eroding effects of having been assaulted by the police.