Legal Cynicism and System Avoidance: Roma Marginality in Central and Eastern Europe

he Roma are Europe’s largest minority group and face extensive discrimination across the continent. Drawing on a survey of Roma and non-Roma households in twelve Central and Eastern European countries, we analyze the extent to which legal cynicism, as a cognitive frame, is connected to the avoidance of helpful social institutions. We thus expand existing research on legal cynicism to focus on individuals’ contacts with potentially helpful institutions that can buffer inequality. We conclude that the interplay of legal cynicism and system avoidance, which have provided deep insights into the reproduction of structural disadvantage in American cities, also provide us with international insights into the causes of inequality and minority disadvantage across hundreds of towns in Central and Eastern Europe. In this way, legal cynicism and system avoidance work to reproduce durable inequality.


Introduction
In 2011, a municipality in northern Romania built a concrete wall around a predominantly Roma neighborhood-which one neighborhood resident poignantly called "a wall to hide our poverty" (The Economist 2015). Romania's antidiscrimination body went on to fine the town and ordered that the wall be removed. Yet while the mayor paid the fine, he went on to win a landslide reelection in the subsequent year (Estrin 2012). Rather than removing the wall, the mayor drew on urban renewal funds to invite students to paint graffiti over it. He then insisted that the wall was a work of art, and as a result could not be removed. This is not an isolated account. The Roma minority, also known as the Romani and disparagingly referred to as "gypsies," face structural exclusion and discrimination throughout Europe, including discrimination by state and legal actors (World Bank 2015). Yet sociological research increasingly points to how marginalization is a product not only of discrimination by outside actors, but also the cultural and cognitive processes that result and lead to detachment and estrangement from formal institutions-including those that may offer relief, recognition, and buffer inequality (Alexander 2008;Desmond et al. 2016;Bell 2017;Asad and Clair 2018;Lamont 2018;Small et al. 2010). We anticipate that these findings would prove particularly salient in the case of the Roma. Across the EU, policies have been focused on institutional and legal reforms designed to promote social inclusion and reduce barriers to formal equality: yet comparatively little attention has been paid to the cultural and cognitive dimensions through which Roma minority members are likely to draw on institutional supports. Roma marginalization, in other words, may be influenced by perceptions of institutions that could otherwise buffer inequality.
As a result, we require a model that accounts for both discriminatory contexts and cultural orientations (Small et al. 2010). To do so we draw on two concepts that, in the context of the United States, have emerged as salient for understanding minority members' orientations toward formal institutions. These are measures of "legal cynicism" (including the belief that legal institutions are effective for addressing everyday problems; see Hagan 1994;Sampson and Wilson 1995;Sampson 2012;Sampson et al. 2018), and of individuals' "system avoidance" (namely their avoidance of potentially helpful state and non-state institutions; see Goffman 2009;Brayne 2014;Haskins and Jacobsen 2017;Desai et al. 2019;Asad 2020).
In so doing, we link legal cynicism, as a cultural orientation towards legality, to behavioral correlates, namely system avoidance. We identify both legal cynicism and system avoidance as embedded in situational contexts, rather than exclusively a result of negative individual experiences with criminal justice officials. We find that the shared cognitive orientation of legal cynicism is connected with a greater likelihood of system avoidance in marginalized contexts. This understanding of inequality suggests that both cultural frames and behavior regarding state institutions are embedded in community expectations (Asad 2020), which need not be directly connected with reported individual-level experiences with justice officials. We further hypothesize that legal cynicism and system avoidance work together to exacerbate and institutionalize inequality when people refrain from drawing on available institutional resources (Tilly 1999).
In work to date, both legal cynicism and system avoidance have mainly been identified among impoverished and largely minority neighborhoods and individuals in the United States. Yet Roma and African Americans have, as Matache and West (2018) argue, "crossed similar paths, as white policymakers continued to employ similar tactics to maintain white normativity, social power, and privilege." By focusing on legal cynicism and system avoidance among European Roma, we extend existing analyses of the relationship between marginality and engagement with legal institutions both geographically and conceptually. This offers an opportunity for further analytical comparison and theory expansion (Wagner and Berger 1985).
Across twelve countries, we investigate whether local township conditions shape the degree to which people are skeptical about law and legality, as well as the degree to which they avoid formal institutions. In so doing, this paper advances our understanding of the structural deprivation and exclusionary treatment of Europe's largest ethnic minority group, offering a view of marginality that is cultural, cognitive, and contextual. Drawing on an extensive cross-national survey of Roma and non-Roma households in East-Central Europe, we find that when living in towns where they experience higher levels of discrimination and segregation, members of the Roma minority are more cynical about legality and legal institutions. We further find that this legal cynicism has behavioral correlates, with individuals avoiding potentially helpful institutions such as hospitals or banks, with likely effects for inequality. Taken together, we find that legal cynicism and system avoidance combine to generate a marginalization process that links the neighborhood contexts in which the Roma live with cognitive frames of skepticism over law and legality, which culminate in an avoidance of helpful institutions. Through this dual process, we argue, durable and entrenched inequality is reproduced (Tilly 1999).

Roma Inequality
Sociological research on the Roma identifies this minority as a "sub-proletariat" that faces social and economic exclusion (Ladanyi and Szelenyi 2006). The Roma were historically accused of committing atrocious crimes and were slaves in parts of Eastern Europe for five centuries until final emancipation in the mid-19th century, within a decade of the emancipation of African-Americans (Crowe 2007;Thornton 2014). Their systemic marginalization continued thereafter: Roma were targeted during the Holocaust, with 250,000 to 500,000 European Roma killed during World War II. In the post-communist period that followed, resurgent nationalism saw Roma labeled as reminders of a "backwards" past that countries orienting to the West were seeking to downplay or deny (Sendroiu and Mogosanu 2019; see also Brubaker 1996, Eley andSuny 1996). 1 The Roma are now the largest non-immigrant ethnic minority in Europe. Of an estimated 10-12 million Roma currently living in Europe, six million live within the European Union (EU), and most are EU citizens (EC 2015). Yet the Roma remain excluded from most economic, political, and social institutions (UNDP 2015). About 90 percent of Roma live in households below their respective European nations' operational definitions of poverty (UNDP 2015). Only 15 percent of European Roma graduate from high school or vocational school (UNDP 2015), 20 percent of Roma are not covered by national health insurance programs (FRA 2012), while about two-thirds of Roma are without paid employment (UNDP 2015). As with African-Americans in the United States, Roma are disproportionately represented in European prisons, and disproportionately more likely to suffer ill-treatment while incarcerated (Press Association 2014; European Roma Rights Center 2018). The Roma face similar patterns of exclusion and discrimination both within and outside the European Union (FRA 2012).
Given this context, it may not be surprising that Roma community members often report skepticism of state institutions that have reproduced racial stigma . Research shows this often leads them to develop close ties with more proximal individuals (Sendroiu and Upenieks 2020). But this puts them in a "contradictory situation," since these proximal others are similarly marginalized, and they collectively feel alienated from distant yet potentially helpful institutions (Málovics et al. 2019a). As a result, some scholars advocate that better representation in public spaces and institutions would allow some redress for Roma alienation (Fraser 2016;Málovics et al. 2019b). 2

The Broader European Context
The relationship of the Roma to European institutions and norms, and with it integration, is a hot-button issue across Europe. Surveyed residents are divided about integration, as are politicians and legal authorities, with the main French police union asserting that "[t]hese are people who sell themselves, who racketeer, who construct criminal networks, and their way of life is totally incompatible with that of our modern societies" (Erlanger 2013). And the matter of integration into Europe is made more complex by the fact that Roma are European by birthright, whether from Romania or from Belgium. Indeed, Roma integration has been further complicated with the advent of the European Union, which ostensibly allows mobility for Eastern European Roma-yet this mobility is deeply circumscribed in practice, and further complicated by power differentials across European countries (Thornton 2014). 3 Éric Fassin (2010Fassin ( , 2015 persuasively argues that this is precisely why European municipalities often withhold services from Roma neighborhoods: making their lives difficult serves as a substitute for deportation. Since the 1990s and the end of the Cold War, discrimination against the Roma has been a pressing policy concern across European countries and agencies (Gheorghe 1991;Vermeersch 2012;Sendroiu and Mogosanu 2018). Over the past decade, EU countries have doubled down on their efforts to remedy the discriminatory treatment of Roma. In 2011, the European Commission (EC) called on all member countries to develop national strategies for Roma integration, which were then adopted in 2013 (EC 2015). Since then, yearly reports have been issued by the European Commission assessing member states' national integration strategies (see, for instance, EC 2013). Some of these policy interventions have successfully alleviated Roma marginality (Méreiné Berki et al. 2017). Yet on the whole, such initiatives are rare and mostly ineffective (Sigona and Trehan 2009;Creţan and Powell 2018; see Creţan and Turnock 2008 for exceptions; see Creţan and Light 2020 for Roma marginality during the COVID-19 pandemic).
Indeed, policy interventions that are designed to address Roma belonging in Europe can also have unanticipated consequences, often by demarcating the Roma minority as not belonging within the polity (Yuval-Davis et al. 2017;Yuval-Davis et al. 2018). This is particularly evident in Barker's (2017) research on the status of the Roma minority in Sweden. Barker (2017) finds that while heavy-handed interventions by police and other state officials may protect migrant Roma from physical violence, it can lead to further deprivation by marking them as excluded from the polity.
This provides the context for the current study. Past research indicates that Roma are distrustful of law and legality, and are similarly distrustful of formal institutions, suggesting that they are a "subaltern counterpublic" (Fraser 2016). Yet we know little about whether any such distrust might be part of a broader process through which Roma inequality persists. In what follows, we explore whether, and how, Roma marginalization, across locations, is connected with perceptions of law, legality, and formal institutions.

Legal Cynicism and System Embeddedness
Two decades ago, Sampson and Bartusch (1998) developed the concept of "legal cynicism" to focus on what happens in communities where individuals are distrustful of law and legal institutions. In their landmark study drawn from an extensive sample of 343 Chicago neighborhood clusters, Sampson and Bartusch (1998) built on sociological control theories of crime that emphasized the demoralizing and anomic results of state disinvestment in African-American neighborhoods. They found that skepticism about law and legal agents are products of living in situations of concentrated disadvantage (Sampson and Bartusch 1998). They concluded that rather than pre-existing subcultural explanations, the reason that people in largely African-American neighborhoods were skeptical about law was due to the "structural context of disadvantage and resource exploitation across neighborhoods" (Sampson and Bartusch 1998: 784).
Since that time, research has linked legal cynicism to a range of behavioral outcomes. For instance, Kirk and Papachristos (2011) find that despite improvement in economic conditions otherwise expected to reduce crime, homicide rates remained persistently high in Chicago neighborhoods with high levels of legal cynicism. They explain this finding by showing that legal cynicism leads to a reduced willingness to call the police to settle disputes, which in turn results in greater violence in such neighborhoods (Kirk and Papachristos 2011). In other words, residents become "more likely to presume that the law is unavailable or unresponsive to their needs" Papachristos 2011: 1203).
Central to legal cynicism research is the relationship between culture and formal institutions. Sampson and Bartusch's (1998) pioneering work refers to legal cynicism as a "cognitive landscape" in which institutions are mistrusted. This has since been elaborated as "a cultural orientation in which the law and the agents of its enforcement are viewed as illegitimate, unresponsive, and illequipped to ensure public safety" (Kirk and Matsuda 2011: 443). Desmond et al. (2016: 872) further build on this cultural approach to institutions to argue that in legally cynical environments, high-profile episodes of police violence "likely contribute to that very cynicism by being incorporated into the community's collective memory" (see also Savelsberg and King 2007).
These cultural dynamics help explain when residents continue to call on the police, and when they avoid doing so: in their analyses of longitudinal data from Chicago, Hagan et al. (2018) find that in neighborhoods of high legal cynicism, residents persist in calling the police as a result of automatic and unconscious forms of cognition, and McCarthy et al. (2020) tie a history of legal cynicism with collective memory of police violence, and thereby explain the greater rates of police complaints in these contexts. Recent qualitative research across US cities finds that residents do a fair amount of cultural work to explain their continued reliance on the police, which they reconcile through narratives of recognition and strategic needs to explain their continued reliance on the police to resolve problems (Bell 2016(Bell , 2017Campeau et al. 2020;Levi et al. 2020). And notably, the effects of legal cynicism appear to not be limited to the US: studies of war-torn states in the Balkans, Sudan, and Iraq, have found that legal cynicism can lead to further cycles of conflict and ethnic resentment (Hagan and Rymond-Richmond 2008;Ivkovich and Hagan 2011;Hagan et al. 2016;Kaiser and Hagan 2018).
Meanwhile, research on system embeddedness and avoidance demonstrates that risk perceptions and concerns over state officials also influence the degree of reliance on institutional supports. Through longitudinal analyses, Brayne (2014) finds that earlier criminal justice contact then leads individuals to avoid formal institutions such as hospitals, banks, employers, and schools, since they all keep records and so are perceived as engaging in surveillance. Similarly, Haskins and Jacobsen (2017) find that incarcerated parents are less likely to engage with their children's schooling. We note that in contrast to work on legal cynicism, these are not all state (or even public) institutions. Yet these are institutions that can help buffer marginality and can be helpful in times of need.
Of course, avoiding formal institutions is not only the result of prior criminal justice contact. Underlying this avoidance is people's precarity, perceived and objective vulnerability, and risk perceptions. This explains the avoidance of surveilling institutions by those with criminal records. Yet it also explains why noncitizens, for whom the threat of deportation and punishment loom large, seek illegibility and invisibility by avoiding state institutions (Desai et al. 2019;Asad 2020). Asad (2020) thereby points to the diversity of strategies and degrees of legibility that vulnerable individuals deploy when engaging with the state, and thus provides a framework for analyzing the degree of system embeddedness in which they engage. These can, in turn, reproduce inequality and "detach noncitizens from mainstream institutions useful for their own or their family's material and social well-being" (Asad 2020:136).

The Conceptual Model: Legal Cynicism and Inequality
In a broad sense, legal cynicism offers a theory rooted in the relationship between culture and institutions (Swidler 2009), in which material conditions of state disinvestment and malign treatment lead to cultural skepticism about legal institutions and legality more broadly. We therefore conceptualize and model legal cynicism as (1) a feature of locations that are marked by state disinvestment and disempowerment. In turn, we argue that (2) legal cynicism correlates with system avoidance, as a cultural frame that conditions the degree to which residents are willing to engage with state institutions, including those which may be helpful to them.
Our first goal is therefore to test the links between legal cynicism and system embeddedness. The legal cynicism literature establishes that skepticism about law and legal officials has effects for the tenor and volume of individuals' engagement with the state (e.g., Kirk and Matsuda 2011;Desmond et al. 2016). Further building on the burgeoning literature on system embeddedness (Brayne 2014;Desai et al. 2019;Asad 2020), we posit that legal cynicism will correlate to engagement with an array of institutions that could potentially offer assistance to individuals.
Building on current research analyzing the effects of each concept, we theorize the combination of legal cynicism and system avoidance as a mechanism that reproduces inequality through an interplay of cultural landscapes, perceptions, and institutional avoidance. We suggest that findings regarding system embeddedness that we see in recent research, and evidence that individuals hesitate to call on state agents in situations of heightened legal cynicism, is consistent with the broader view that in marginalized contexts, legal cynicism and system avoidance are intensified, with likely effects for durable inequality (Tilly 1999). This further resonates with research on the collateral consequences of punishment, which shows that state and non-state institutions can have effects for the reproduction of inequality and marginalization (Harris et al. 2010;Uggen and Stewart 2011;Kirk and Wakefield 2018).
In the absence of longitudinal data, we cannot test a temporal model whereby some locales lead to greater legal cynicism and system avoidance. In this paper, we instead test for the existence of associations between contextual factors, legal cynicism, and system avoidance-and we conduct this analysis cross-nationally. We argue that legal cynicism and system avoidance are mechanisms through which disadvantage is maintained. Combining research on legal cynicism with the concept of system avoidance, we first analyze the predictors of individual legal cynicism, mapping these cognitive frames onto contextual measures of marginality. We then use this framework to predict the contexts in which individuals are likely to avoid formal institutions.

The Survey
Our analyses draw on the 2011 Regional Roma Survey conducted by the United Nations Development Programme (UNDP), the World Bank, and the European Commission (Ivanov et al. 2012). This survey was meant to map out the socio-economic conditions of Roma living in Europe. It did so by comparing Roma households with nearby non-Roma households using questions spanning health, money management, and education. In all these areas, Roma were considerably disadvantaged compared to their non-Roma neighbors. The study design allows comparisons at individual and ecological levels of Roma and non-Roma households in a number of different countries, and while the data are not representative of all Roma in Europe, it is "as representative as possible of those Roma who face social exclusion and risk marginalization" (Ivanov et al. 2012: 11). The sample consists of 20,018 Roma and 9,782 non-Roma households drawn from 12 Eastern and Central European countries, namely: Albania, Bosnia and Herzegovina, Bulgaria, Czech Republic, Slovakia, Montenegro, Croatia, Hungary, Macedonia, Moldova, Romania, and Serbia.
We use multi-level models to assess the relationship between legal cynicism and other variables of interest, because our models include variables at the individual, municipal, and country levels. In the first set of analyses, we predict legal cynicism using mixed-effects Poisson regressions, which is the recommended modeling strategy when predicting count variables, as is the case for legal cynicism. 4 In turn, for ease of interpretability, the coefficients from these models were produced as incidence rate ratios (IRR). These analyses were conducted with the mepoisson command in Stata, version 14.
To model system embeddedness and avoidance, we used multilevel (individual, municipal, and country) mixed effects logistic regressions. This is the recommended modeling strategy for predicting binary outcomes such as system avoidance. These analyses were performed using the melogit command in Stata, version 14. The coefficients from this model, meanwhile, are presented as odds ratios (OR). These are preferable to log odds coefficients, because they take on a similar interpretation to IRRs, where the number of units above or below 1 give an estimate of the relative odds of being in one category relative to the reference category.
Our analyses focus on a survey module on "Individual Status and Attitudes," which was administered to a randomly chosen adult member of survey households. Removing those with missing data on key variables (less than 1% of the sample), we retain an analytic sample of 11,447 individuals (with the N decreasing slightly to 11,422 for analyses predicting system avoidance). We have 393 municipalities in our analytic sample. On average, these towns have 30.36 respondents and size ranges from 2 to 599 respondents. As an additional robustness check (to ensure that smaller town sizes do not skew average town variables), we tested our models using a restricted sample made up only of towns with more than 10 respondents; findings remained unchanged and so here, we present models from the complete sample. Table 1 contains descriptive statistics for the variables in our analysis.

Dependent Variables
Measurement of legal cynicism has varied across research studies to date. Sampson and Bartusch (1998) developed the following scales for their legal cynicism index: "Laws were made to be broken"; "It's okay to do anything you want as long as you don't hurt anyone"; "To make money there are no right and wrong ways anymore, only easy ways and hard ways"; "Fighting between were not provided with a list of answers, and instead their first three spontaneous responses were marked; 66 percent of the sample responded no one. The measure we use here is the percentage of individuals (ranging 0-1) in a town who named an institution (e.g., local government, church, humanitarian organization) rather than responding "no one." * * Both Roma and non-Roma respondents were asked "Some people have certain opinions about different nations and cultures. Below is a set of statements reflecting certain opinions, stereotypes and prejudice about Roma. We would like to know your opinion about them. Please tell us which of them you find justified and whichnot." Respondents were the provided with a list of statements, from "Roma are cheerful and enjoy life" to "Roma are lazy." As a conservative measure of anti-Roma stereotypes, we coded all those who judged one of these stereotypical statements to be "entirely justified" as holding anti-Roma stereotypes. * * * Respondents chose among multiple options of dominant ethnic groups in the town; Roma coded as 1; all else = 0 friends or within families is nobody else's business"; and "Nowadays a person has to live pretty much for today and let tomorrow take care of itself." Relying on the same Chicago neighborhood data as Sampson and Bartusch (1998), Kirk and Papachristos (2011) combine measures of disregard for state legal norms along with evaluations of state institutions. Legal cynicism is thus measured in terms of agreement that "laws are made to be broken," "the police are not doing a good job in preventing crime in this neighborhood," and "the police are not able to maintain order on the streets and sidewalks in the neighborhood." There is a .75 correlation between this scale and the earlier Sampson and Bartusch (1998) index. Later work by Kirk and Matsuda indicates that this measurement of legal cynicism remains fundamentally about residents' perceptions of state agents rather than moral cynicism broadly, so that even the measure of "laws are made to be broken" is included "under the assumption that perceptions of the broader legal system influence resident cooperation with the police" (2011: 454). We build on the core of this thinking to develop our legal cynicism scale for the context of Roma marginality in Central and Eastern Europe. The Regional Roma Survey includes a module of questions on the attitudes of respondents. Translated into English, the questions we rely on ask respondents to indicate in which of the following situations they think it can be acceptable to: (a) claim government benefits to which a person is not entitled, (b) give a bribe to achieve what he/she wants, (c) not pay taxes that are required, (d) for an official to accept a bribe in the course of their duties, and (e) to steal food if one's family is going hungry. As with the history of attempts to capture the phenomenon of legal cynicism in the United States, the questions we rely on emphasize skepticism over state actors and the legitimacy of legal norms for everyday life ). These questions also reflect the local contexts of Central and Eastern Europe, in which benefits, bribery, and corruption are more salient cultural and political perceptions of official state and legal malfeasance, even if levels of corruption vary widely (Zaloznaya 2017).
We conducted further tests to confirm whether these items can be combined into an index. In particular, we found that the Cronbach's Alpha for these items (0.787) supports their use to form an index of legal cynicism. We drop the last item because it is a poor conceptual fit with the other items (this increases the Alpha score to 0.801). The first set of models uses this cynicism index as a dependent variable. For each variable within the index, highly agreeing with any one of the statements = 2, somewhat agreeing = 1, and not agreeing = 0. The cynicism index therefore ranges between 0 and 8, with 0 being low cynicism and 8 high cynicism.
Meanwhile, the second set of models we estimate predict avoidance of major social institutions. We take on board Asad's (2020) view that hesitation over system embeddedness is not limited to those with past criminal justice contact, and instead is a product of marginality and perceived risk over legibility. Yet in an effort to study both legal cynicism and system embeddedness, we here seek to capture a wide range of formal institutions, both within and beyond the state. As a result, we adopt Brayne's (2014) earlier measures of system avoidance that encompass both state and non-state institutions, and examine whether in the past year the respondent saw a doctor when needed, made use of financial institutions, attended school or was employed in paid work. We combine these as a measure of avoidant behavior. If a respondent avoids any of these social systems, they are coded as 1; no system avoidance = 0. We note that 48% of the sample do not engage in system avoidance, 43% avoid just one system, almost 8% avoid 2 systems, and only 1% avoid all of the above. Methodologically, coding system avoidance as a binary variable therefore allows us to prevent the small percentage of extreme avoiders from skewing our findings.

Independent Variables
Our first independent variable is ethnicity (Roma = 1, non-Roma = 0). We also include past experiences of discrimination, which are coded as having experienced discrimination = 1, or not having experienced discrimination = 0, no matter the source of the discrimination. As a result, this can capture negative experiences of discrimination that will be more likely to be experienced by the Roma-i.e., discrimination based on ethnic grounds-and forms of discrimination experienced by both Roma and non-Roma respondents, based on age, gender, disability, or other reasons.
The next three variables are contextual at the town level, which we created by aggregating individual responses to survey questions. We build on prior work from Suttles (1972), Kapsis (1978), Sampson and Bartusch (1998) At the country level, we include perceived levels of crime from a UNDP survey on Roma vulnerability, in order to test whether levels of cynicism and system avoidance increase in contexts with more crime. 5 We also include GDP in millions of dollars and EU status in order to test for whether richer countries, or those functioning within the EU framework of rights and minority protections, will have less cynicism and system avoidance.
All our models control for standard demographic variables: gender (female = 1, male = 0), age, and years of education. They also control for self-reported individual income, which we standardized at the township level, with values more than three standard deviations away from the mean recoded to equal three.
Finally, in the models with system avoidance as the dependent variable, legal cynicism is treated as an individual-level independent variable.

Legal Cynicism
Table 2 contains findings from multi-level mixed-effects Poisson regression models predicting legal cynicism. Model A includes our individual level control variables, all of which are significant. Being female (0.84, p < .001), having more years of education (0.93, p < .001), or being older (0.99, p < .001) all decrease the likelihood of being cynical. On the other hand, having a higher income compared to the rest of one's township increases the likelihood of being cynical about legal institutions (standard income = 1.05, p < .001). We speculate that income is positively associated with our measures of legal cynicism because several of them focus on the economic benefits of disregarding legal norms. As a result, these may be tapping into ideals of hierarchic self-interest and market competition (Coleman 1987;Hagan et al. 1999;Hövermann et al. 2015).
In Model B, where we introduce effects for discrimination and ethnicity, we see that both measures are significant and strongly increase the odds of legal cynicism. Having experienced discrimination thus produces a 29 percent increase in the odds of legal cynicism (p < .001). Meanwhile, being Roma results in a 59 percent increase in the odds of legal cynicism (p < .001).
Our town and country level variables are also significant in predicting cynicism. In Model C, we see that living in a town with anti-Roma stereotypes is related to a strong increase in the odds of legal cynicism (1.67, p < .01). Living in a locale with helpful institutions also increases the odds of legal cynicism (2.03, p < .001), though as we will see in the next model, these town effects work differently for members of different ethnic groups. We note that the measure of helpful institutions is very broad, a town-level average of individuals within a town who were able to access any institutional help at all, whether from church, NGOs, and so on. Being able to access such institutions may, however, reduce the need to rely on legal systems, or indeed may cast legal institutions-which may be corrupt and ineffective-in a particularly negative light. For the Roma, however, helpful institutional contexts can be empowering even if not directed to them specifically (see also Sendroiu and Upenieks 2020), as we will see in the models described below. Finally, living in a place where Roma are dominant does not have a statistically significant effect on cynicism, though again this will be different once we look at how this measure works among the Roma specifically.
At the country level, we see in Model C that neither country level perceptions of criminality nor living in the EU are significant predictors of legal cynicism. GDP, meanwhile, is a significant predictor of legal cynicism (p < .001), but with an incidence rate ratio of 1, we have little sense of the direction of the effect. Though we cannot say with certainty, this does suggest that legal cynicism is an outcome of the more local contexts where individuals live, as well as the particular forms of marginality that individuals experience, rather than country level factors. This also resonates with neighborhood-level differences in legal cynicism within US cities .
In order to test this relationship between local disadvantage and marginality, we introduce interactions between being Roma and our three town-level variables in Models D through F. We do this in order to see whether it is specifically for the Roma that these disadvantaged locales produce legal cynicism and, in subsequent models, system avoidance. In these models focused on legal cynicism, we indeed see that the cross-level interactions between being Roma and townlevel variables are all significant in predicting legal cynicism. The interaction between being Roma and living in towns with helpful institutions is associated Err.   (Sendroiu and Upenieks 2020), but here we see that for the Roma specifically, these institutions are nonetheless protective, associated with a decrease in cynicism. Meanwhile, the interactions between being Roma and either towns with anti-Roma stereotypes or Roma dominated towns are both related to increases in the odds of legal cynicism (1.77 and 1.56, respectively, p < .001). We thus see how, for the Roma, living in more disadvantaged towns characterized by more anti-Roma stereotypes, or simply where Roma are the majority population, is a meaningful predictor of cynical frames. This is especially interesting in the case of towns where Roma are dominant. Recall from Model C that these Roma dominant towns do not have a statistically significant effect for the sample more broadly. Through the interaction, we thus see that it is specifically for the Roma that living in Roma dominated towns is damaging and produces more legal cynicism.

System Embeddedness and Avoidance
In the second part of our analysis, we investigate whether legal cynicism and its predictors are in turn related to system avoidance. Model G in Table 3 shows that of our individual-level controls, being female (0.62, p < .001), having more education (0.86, p < .001) and having a higher income (0.81, p < .001) all decrease the odds of system avoidance. Some of these patterns are notably different than the predictors of legal cynicism. In particular, income is associated with an increase in cynicism but a decrease in avoidance, likely because income itself is connected with integration to education and work.
These two measures-legal cynicism and system avoidance-are nonetheless significantly related. We see in Model H that legal cynicism is statistically significant in predicting an increase in the odds of system avoidance (1.11, p < .001). This holds when we add ethnicity and discrimination to the analysis in Model I, though both are strong predictors of system avoidance (Roma = 2.15, p < .001; discrimination = 1.73, p < .001).
We add in our town and country level variables in Model J. At the town level, we add helpful institutions, anti-Roma stereotypes, and Roma dominated towns. At the country level, we add perceptions of crime, GDP and EU status. Among these six variables, GDP is the only one that has a statistically significant effect on system avoidance, though with an odds ratio of 1 (p < .05) such that the direction of the relationship is difficult to interpret.
That on their own, as main effects, the town-level variables are not significant speaks to a difference between legal cynicism and system avoidance in that they appear not to affect avoidance of helpful institutions for the whole sample. When     we introduce our cross-level interactions in Models K through M, we nonetheless see that living in highly marginalized contexts will affect Roma individuals' system avoidance. This is, in particular, the case for Roma living in Romadominated towns. In other words, just as we saw in the legal cynicism models above, Roma living in Roma-dominated towns experience greater odds of system avoidance (1.33, p < .05). As a result, we conclude that legal cynicism and system avoidance emerge and are intensified in these contexts, with the specification that this only seems to be the case for those who belong to this marginalized minority group.

Summary of Findings
Taken together (a summary is available in Table 4), we see that Roma residents living in protective locales-in this case, in towns with more helpful institutions-report lower rates of legal cynicism. Meanwhile, living in disadvantaged locales, marked here by the higher presence of anti-Roma stereotypes or by ethnic segregation, increases legal cynicism among Roma respondents. And towns that are Roma-dominant produce increases in both legal cynicism and system avoidance, but this is only the case among Roma respondents. We thereby examine the Roma experience across a spectrum of towns, and cultural and institutional landscapes. Our findings suggest that the most disadvantaged towns-which produce, for the Roma, both legal cynicism and system avoidance-are those which are Roma dominated. Living in these places with individuals who are similarly marginalized seems to have a particularly negative effect for Roma respondents. Conversely, we see that where there are investments in the towns where Roma live-in towns where respondents report the presence of helpful institutions-the Roma are no more likely to avoid institutions than non-Roma respondents.

Discussion: A Situational and Cultural Account of Marginality
The legal cynicism models that underwrite the theoretical arguments of this article highlight a situational approach to explaining the relationship of disadvantaged minorities to law. We find that ethnic segregation of the Roma in the Figure 1. 2011 Regional Roma Survey towns according to institutional efficacy levels marginalized communities in Europe, akin to segregation of African-Americans in poor communities in the United States, produces collectively shared cognitive landscapes of legal cynicism that are also associated with-and behaviorally expressed in-patterns of institutional system avoidance.
This situational approach allows us to map-across twelve countries-the spectrum of towns that have helpful institutions, those that have higher levels of anti-Roma stereotypes, and those that are marked by ethnic segregation. In figures 1-3, we provide visual representations of these town-level attributes, demonstrating the cross-national range of our analyses. These towns provide the ecological basis for our theoretical and empirical extension of legal cynicism and system avoidance to the structural contexts of Central and Eastern Europe.
We therefore elaborate a view of marginality as both contextual and cultural or cognitive, and demonstrate this empirically. Legal cynicism, as a cultural and cognitive frame, is heightened in marginalized contexts that are hostile and divided, marked by a higher degree of anti-Roma stereotypes, ethnic segregation, or a dearth of helpful institutions. In turn, legal cynicism is connected to behavior, namely an avoidance of potentially helpful institutions that is exacerbated for Roma living in marginalized contexts. Altogether, our model focuses on Our findings thus extend a long tradition of research that focuses on situational adaptations in marginalized contexts. Earlier research emphasized situational adaptations, such as Cohen's (1960) study of "delinquent boys" who replace conventional social norms with situational alternatives. More recently, Anderson (1999) demonstrated that while individuals code-switch between local and broader social norms, the code of the street often prevails in marginalized neighborhoods as a way to signal belonging. In much the same way, legal cynicism and system avoidance prevail in marginalized towns across East-Central Europe. Yet we emphasize that our model is explicitly situational. We find that in contrast to much of the political rhetoric, the avoidance of social systems is not a feature of the Roma minority as such-what might be thought of as a subcultural explanation (Sampson and Bartusch 1998)-but is instead related to living as a Roma minority member within ethnically segregated towns.
We recognize that situational adaptations such as cynicism and avoidance may, in some cases, be protective for minority communities (Venkatesh 2006). For instance, in her classic work All Our Kin, Carol Stack (2008[1974: 22) focused on "the adaptive strategies, resourcefulness, and resilience of urban families under conditions of perpetual poverty." But there is an open question of the extent to which these adaptations are helpful for redressing inequality, including through the resources available from mainstream institutions. We find this tension to be at the core of Herbert Gans's work in Urban Villagers (1982). Gans (1982) shows that local adaptations can indeed shore up social support within a community, yet at the same time would decrease members' collective capacities for productively engaging with broader social institutions or individuals outside the community.
Past research on the Roma has also shown this double bind: Roma only feel comfortable reaching out for help from other Roma, but such individuals from their home communities will be equally marginalized (Sendroiu and Upenieks 2020). Roma therefore face a "contradictory situation" whereby they feel strong bonds to similarly marginalized proximal others, but alienated from resourcerich institutions (Málovics et al. 2019). As Gans (1982) showed, this is especially harmful when communities lack the repertoires, habits, or resources needed to engage with social institutions. We see echoes of this in our findings of Roma who live in Roma-majority towns.
Taken together, our findings offer insight into the institutionalization of marginality. It may be that Romani individuals have good reason to be collectively cynical about law and to avoid formal institutions: indeed, our findings reinforce that this is likely reactive to settings that are particularly poor, hostile, or segregated. The substantive outcome, however, is that Roma forsake what are potentially helpful and resource-rich institutions. This patterned and collective decision-making can be a basis for what Tilly (1999) thinks of as durable inequality, and can even reinforce "categorical" views about Roma communities as broadly uninterested in engaging with formal institutions, by ignoring the causes of these behaviors and the inequalities they propagate.

Conclusion
The Roma are said to represent the most significant challenge for social integration in Europe, with as many as 12 million Roma living in destitute conditions across the continent (Spiegel 2014). Yet in contrast to often discriminatory political rhetoric, we know surprisingly little about the operational mechanisms that underlie and perpetuate Roma marginalization (Ladanyi and Szelenyi 2001).
Drawing on research that connects culture, cognitive landscapes, institutions, and Romani studies, we contribute to explaining Roma inequality by linking the ecological contexts in which they live with both their perceptions of law and state officials, and with measures of their engagement with potentially helpful institutions (Kapsis 1978;Small 2006). We find an association between legal cynicism and system avoidance, and we further find that both phenomena are intensified in marginalized contexts.
In particular, we show that collectively shared cognitive frameworks of legal cynicism are associated with Roma being less likely to reach out to formal institutions for health, employment, or banking. Building on previous research we see this combination as an inequality-reproducing mechanism. We argue that the relationship we demonstrate between legal cynicism and system avoidance further illustrates the ways in which legal cynicism can contribute to marginality, leading to greater avoidance of state-and non-state institutions. There are important institutionalizing processes here for understanding Roma marginalization, with historical categories leading Romani individuals to respond to discriminatory and segregated contexts through patterned decision-making that further reinforces inequality by forsaking available institutional resources (Tilly 1999).
We further show that conceptualizations of legality designed to explain crime patterns in Chicago provide us with insight into the causes of inequality and minority disadvantage across hundreds of towns located in Central and Eastern Europe. Romani individuals living in situations of high ethnic segregation are more likely to report legal cynicism, and to avoid institutions. As in the United States, this effect appears to turn on disinvestment rather than subcultural explanations. In towns where respondents report the presence of helpful institutions, Romani individuals are no more likely to avoid institutions than non-Roma residents. This is an intriguing finding, and we anticipate that case study research would allow for more fine-grained, qualitative analyses of how different institutional arrangements and investments have effects for Romani perceptions of legality and their willingness to engage with formal institutions. Further research could also track the relationship between legal cynicism and system avoidance among other groups-including, for instance, poor white communities-as well as among Roma in Western Europe. This would allow for a more robust understanding of the institutional protections that are effective in ameliorating the durable inequality produced in the nexus between marginality, legal cynicism, and system avoidance.
An important limitation of our research is that we conducted this analysis with cross-sectional data. We thus cannot specify the direction of the relationship between legal cynicism and system avoidance. On the one hand, cynicism about law and agents of its enforcement could increase system avoidance. We view this as potentially close to the causal mechanism described in studies to date of system embeddedness, based on distrust, risk perception, or fear (Desai et al. 2019;Asad 2020). Yet it may be that some other factor explains Romani avoidance of institutions, with effects for their views more broadly-including legal cynicism. Further research could work to establish the causal links that may be at work here, whether the link between cynicism and system avoidance is moderated by other factors, or indeed whether the causal direction is flipped, with system avoidance effectively being justified or reinforced through legal cynicism.
By relying on a unique dataset of Roma and non-Roma households, we find that legal cynicism and avoidance of formal institutions by the Roma occur across the diversity of countries in which they live, and can worsen when living in marginalized contexts. Where the Roma experience this contextual marginalization, they express greater legal cynicism and are more likely to avoid formal institutions-including institutions that could potentially mitigate hardship. We thus highlight the relationship between cultural normative landscapes and material institutional supports: neighborhood contexts of disadvantage are associated with increased legal cynicism and system avoidance, and this process generates persistent inequality for the Roma.

Notes
1. More recently, this boundary work against the Roma has been employed in far-right politics (see Creţan and O'brien 2019). 2. In parallel, some argue that this includes ensuring space for insights from Roma researchers within academia (Marushiakova-Popova and Popov 2017). 3. We note that in the US context, the migration of African-Americans north led to ever greater racial disparities, leading to threat and status competition, and challenges posed by economic deindustrialization (Muller 2012;Wilson 1987;Sampson and Wilson 1995). These concentrated inequalities can then have cultural and normative consequences for how group members rely on or seek to avoid legal and official institutions (Sampson and Bartusch 1998;Sampson et al. 2018), further institutionalizing inequalities in disadvantaged and segregated locales (Desmond et al. 2016;Hagan et al. 2018). 4. There is no evidence of overdispersion in legal cynicism (as the mean is greater than the variance), suggesting a Poisson model is preferable to a negative binomial model. 5. We use aggregated country-level answers to the following question: "How would you assess (on a scale of 1 to 5) the level of threat for your household from 'ordinary crime'?" This dataset does not include data for Slovakia or Moldova, and so we assigned these countries the sample average. While we acknowledge the time gap between this crime survey and the Regional Roma Survey, this survey was the fullest available data source of crime perceptions for the countries in our sample. We should also note that findings remain unchanged (models available upon request) when we include 2011 homicide rates in our models as a proxy for crime perceptions. This latter measure is available for the entire sample. Finally, we note that we cannot disaggregate these crime measures further: since many of the townships in our sample are small towns or villages, more fine-grained measures of crime are not available to us across locations.