PROCEDURAL JUSTICE, HISTORICAL INSTITUTIONALISM, AND CONSEQUENCES OF POLICING FOR PUBLIC OPINION

The ever-growing body of scholarship on procedural justice, especially in the United States, has profoundly informed our understanding of how actions by police shape public opinion about police and the criminal justice system more broadly (Reference TylerTyler, 1990; Reference Tyler and HuoTyler & Huo, 2002). The argument at the center of this research is that public perception of police and the willingness to collaborate and comply with the law are affected by how officers perform their duties. If members of the public are satisfied with encounters and perceive them as just, feelings of trust and legitimacy will increase; otherwise the two may dwindle (Reference Hough, Jackson, Bradford, Myhill and QuintonHough et al., 2010; Reference Mazerolle, Bennett, Davis, Sargeant and ManningMazerolle et al., 2013). In response to unjust treatment, the public may also feel alienated from law enforcement and become cynical about the legal system (Reference Aas and BosworthAas & Bosworth, 2013; Reference Lerman and WeaverLerman & Weaver, 2014; Reference Nivette, Eisner, Malti and RibeaudNivette et al., 2015). European studies on procedural justice similarly find that fair, transparent, and respectful policing is generally related to higher levels of trust and legitimacy (Reference Hough, Jackson and BradfordHough et al., 2013a, Reference Hough, Jackson, Bradford, Tankebe and Liebling2013b; Reference Mazerolle, Bennett, Davis, Sargeant and ManningMazerolle et al., 2013; Reference TylerTyler, 2011).

Empirical research on procedural justice in Europe has significantly enriched our understanding of policing and its public perception, but most studies have either considered the continent as a homogenous whole or they exclusively focused on established democracies (Reference Hinton, Newburn, Hinton and NewburnHinton & Newburn, 2009). Moreover, whereas much of the research in the United States used police contact as a starting point to understand how policing affects trust and legitimacy, actual police contact was rarely the focus of European studies. In the present study, we pay attention to these two issues in particular. Critically, however, while we build on existing studies, we focus on a point that has not been prominent in prior research—that the perception of police and the effects police have on the public are shaped by history and political context.

Historical criminologists have long pointed to the inextricable links between state formation and police, but the analyses have mainly considered the western and northern parts of Europe (Reference Bayley and TillyBayley, 1975; Reference EmsleyEmsley, 2007). Our emphasis is placed on the importance of considering the political histories of policing in other parts of the continent to understand the heterogeneous experiences with law enforcement and the effects those experiences may have on the public. In modern Europe, these issues are especially relevant at the political boundaries drawn in the aftermath of World War II between communist states and liberal democracies.

Furthermore, our perspective is informed by historical institutionalism (a strand of so-called neo-institutionalism), merged with procedural justice theory into a multilevel theoretical formulation of police legitimacy. Much historical institutionalist work is concerned with explaining the sources of societal change (e.g., Reference SkocpolSkocpol,  1979), but our interest is in understanding how continuity of institutional forms in the aftermath of societal transformation might shape public attitudes in the present. Key to this is path dependence, whereby social institutions are embedded in longer-term processes that tend to reproduce them, albeit in modified form (Reference Pierson, Skocpol, Katznelson and MilnerPierson & Skocpol, 2002; Reference ThelenThelen, 1999). Our claim is that policing is, to some degree, isomorphic with some of the older institutional arrangements that linger in the wake of political transitions. Traces of an older politico-legal regime render civilian attitudes about fairness and legitimacy more strongly determined by their individual experiences with police, relative to societies that have not experienced similar (or at the very least, recent) transformations.

A central proposition of historical institutionalism is that societies develop along different institutional lines, which gives the state a privileged role in explanatory frameworks. We take this to mean that the relationship of the individual to the state—and consequently between the individual and representatives of the state, as in the police—varies meaningfully depending on forms of institutional dominance. Reference GiddensGiddens (1979) defines institutions as “deeply-layered practices constitutive of social systems” (p. 65) and describes them as “sedimented in time-space” (p. 80). Post-communist countries are formerly dominated by political institutions, suggesting police fulfill a coercive function that ensures perpetuation of the regime. From an institutionalist perspective, this coercive and potentially corrupt legacy can be manifested in the present by generally low levels of perceived fairness and legitimacy with respect to police. Established democracies, by comparison, are dominated by legal institutions, which shape civilian attitudes in a manner suggesting generally high levels of perceived fairness and legitimacy.

Aside from expectations about the overall level of fairness and legitimacy, we also consider whether civilian attitudes are more or less strongly correlated with individual experiences with law enforcement in post-communist countries than in established democracies. Given the expected low levels of perceived fairness and legitimacy in the former, we would also posit that positive experiences with law enforcement—because they defy expectations of coercive or unfair treatment—might actually correspond with higher levels of perceived fairness and legitimacy relative to the same experiences in established democracies. We next describe some of the key aspects of policing in post-communist countries, both before and after the fall of Communism, and discuss ways in which the public experience of contact with police may have a different effect relative to their effects on the public in established democracies.

POLICING AND THE POST-COMMUNIST DIVIDE

Policing in communist countries has been intimately linked with politics. Not responsible solely for public safety, police had the mandate to maintain party rule by clamping down on dissent (Reference Marenin, Caparini, Fields and MooreMarenin & Caparini, 2005; Reference Mawby and KadarMawby, 2001). The most infamous effort to do so was the formation of secret police services, such as Securitate in Romania and Służba Bezpieczeństwa in Poland. These services safeguarded the political order and acted as the repressive tool of the ruling party (Reference MawbyMawby, 2000; Reference StanStan, 2009). The network of secret surveillance was vast and involved tens of thousands of informers. In practice, this meant that police were the “sword of the revolution,” and that the law was subordinated to regime survival (Reference Aitchison, Bradford, Jauregui, Loader, Steinberg and AitchisonAitchison, 2016). A similar but less expansive lack of restraint may have persisted into the post-communist era as some officers, despite lustration efforts, have assumed high-ranking positions following the democratic revolutions or they have in other ways influenced the practice of policing on the ground (Reference GaleottiGaleotti, 1993; Reference StanStan, 2009; Reference WelshWelsh, 1996).

International efforts to reform policing have been ongoing since the early 1990s and were spearheaded by organizations such as the United Nations. Yet police forces have remained politicized during post-communist times (Reference BayleyBayley, 2006). Compared to established democracies, police have engaged in higher levels of corruption and have had lower levels of professionalization (Reference Aitchison, Bradford, Jauregui, Loader, Steinberg and AitchisonAitchison, 2016; Reference Alvazzi del Frate and van KesterenAlvazzi del Frate & van Kesteren, 2004). At the same time, due to the balkanization of countries across the continent after the fall of the Iron Curtain, police forces were implicated in ethnic divisions that left some groups more dominantly represented in law enforcement relative to others—and this may have further contributed to distrust in police. In post-conflict societies that experienced a violent transition, such as countries in former Yugoslavia, militarization and ethnic repositioning of police have been distinguishing characteristics that made law enforcement in times of peace more challenging; it also made it harder for police and the public to trust one another (Reference HillHill, 2010; Reference Kutnjak IvkovićKutnjak Ivković, 2009; Reference Kutnjak Ivković and ShelleyKutnjak Ivković & Shelley, 2005).

From the vantage point of procedural justice theory, secrecy, lack of transparency, and unchecked power are all characteristics that make the public distrustful of law enforcement (Reference TylerTyler, 1990). As a result, the public may be reluctant to be in contact with police because they are less likely to expect fair and transparent treatment, and the interactions may be less satisfactory when they happen. For these reasons, the effects of police contact on public opinion may be greater as police officers can operate with greater discretion and less accountability (Reference Beck and ChistyakovaBeck & Chistyakova, 2002). In addition, negative experiences are known to have a greater impact on individuals than positive ones (Reference Baumeister, Bratslavsky, Finkenauer and VohsBaumeister et al., 2001; Reference SkoganSkogan, 2012), which suggests that, if unsatisfactory encounters are more common, there may be higher correlation between the extent and nature of police contact and public attitudes.

A critical point is that the aftermath of the dissolution of totalitarian regimes lingers. It takes time to establish a functioning democratic society and a police force that reflects democratic values. Reference Hinton, Newburn, Hinton and NewburnHinton and Newburn (2009, p. 7) make this point eloquently when they write, “legal codes that were designed to maintain political order and control need substantial post-transition updating so as to provide effective, political and civil freedoms and human rights, a process that can take years.” They go on to add that “police involvement in repression associated with authoritarian regimes has cast a long shadow over police-community relations, generating enduring patterns of mutual suspicion, fear, and mistrust” (p. 8). In fact, the notion that police forces emerging from the collapse of Communism needed reform has underlied the effort to promote democratic policing globally—policing that is accountable to law, monitored by independent bodies, and responsive to the needs of ordinary citizens (Reference BayleyBayley, 2006; Reference Pino and WiatrowskiPino & Wiatrowski, 2006).

We must also consider that the sudden transition experienced by post-communist countries in Europe may have shaped how police forces do their job and how they relate to the public. Building on Durkheim's idea that anomie results from rapid social change, studies have found evidence that higher crime rates may be an outcome of major political shifts (Reference Pridemore and KimPridemore & Kim, 2006; Reference StamatelStamatel, 2009). Even though these studies have not considered policing in particular, the general societal levels of anomie may have also been reflected in the practice of police work. The morale of police officers, the lack of resources and low salaries, increasing crime rates in some instances, as well as the need to adjust to the new political system may have affected the quality of policing and how police interacted with the public. A study based in post-communist Russia, for instance, identified predatory policing—officers using their authority for own material gain—as pervasive (Reference Gerber and MendelsonGerber & Mendelson, 2008). The weakening of states during democratic transitions also made police vulnerable to corruption and infiltration of organized crime (Reference HajdinjakHajdinjak, 2002; Reference HolmesHolmes, 2009).

It is important to note that policing in established European democracies is not ideal as police also engage in unfair and discriminatory practices, especially in interactions with minority and immigrant populations (Reference FassinFassin, 2013). But police had more time and infrastructure to adopt the principles of democratic policing. This notion is supported by research that shows higher overall satisfaction with police and the way police deal with preventing and controlling crime in western relative to the eastern and central parts of the continent (Reference Alvazzi del Frate and van KesterenAlvazzi del Frate & van Kesteren, 2004). At the same time, the public reports higher levels of trust in legal institutions and trust in police in established European democracies (Reference Kutnjak IvkovićKutnjak Ivković, 2008; Reference KääriäinenKääriäinen, 2007). These findings reinforce our point that change happens along with historical continuity. Similar points have been made elsewhere. A recent study on trust in institutions noted that “the ‘specter’ of the communist past is still haunting the eastern part of the continent” (Reference Schaap and ScheepersSchaap & Scheepers, 2014, p. 91). More broadly, when discussing changes in policing in modern democracies, Reference Jones and NewburnJones and Newburn (2002, p.142) highlighted that “much current criminology tends to exaggerate the degree of change, and underplay the extent of continuity, in seeking to explain the transformations taking place in contemporary policing systems.”

Other studies have performed country-level analysis using data from the 2010 ESS Trust in Justice module, the dataset we also choose for the current study. While these studies have not examined police contact as consequential for police fairness and legitimacy across the post-communist divide, their findings show a regional gap reflecting the importance of political history. In one of the first analyses using the Trust in Justice module data, the least positive public evaluations of fairness and legitimacy were detected in post-communist countries (Reference Hough, Jackson and BradfordHough et al., 2013a, Reference Hough, Jackson, Bradford, Tankebe and Liebling2013b). In a more recent analysis, Reference Mehozay and FactorMehozay and Factor (2017) have examined normative beliefs as predictors of police legitimacy, and found, similar to other research, that clusters of countries with the lowest levels of perceived legitimacy generally include post-communist countries. In contrast to those studies, we focus on police contact and the large body of prior research demonstrating the importance of physical encounter for public perceptions of trust and legitimacy (Reference Mazerolle, Bennett, Davis, Sargeant and ManningMazerolle et al., 2013).

While our conceptual distinction between post-communist countries and established democracies in Europe has practical and conceptual usefulness, there are countries for which such a classification is not straightforward. While the Federal Republic of Germany became a bastion of liberal democracy soon after World War II, the German Democratic Republic was under the tight grip of the Soviet Union and the communist party. The state was ruthless in protecting the regime and omnipresent through its network of informants (Reference GiesekeGieseke, 2014). But after the unification in 1990, Germany was quick to invest resources to ensure a democratic transformation of its police forces in the east. As a result, policing structure and function soon started to resemble its counterparts in the west (Reference Shelley and MawbyShelley, 1999).

Southern European countries—Greece, Spain, and Portugal—have also experienced authoritarian dictatorships after World War II. Some of the characteristics that marked countries of the communist bloc also applied to these countries until the 1970s, including political police, surveillance, human rights violations, and repression. In that regard, the way the public perceived and experienced police had a great deal in common with their communist neighbors. Yet the democratization these countries experienced since the 1970s afforded them decades of democratic development and integration with international organizations that promoted police reform. In other words, they had time to foster a more accountable and democratic law enforcement system. For these reasons, we include Greece, Spain, and Portugal among the established democracies in our analysis, details of which are described next.

DATA AND METHODS

The data used in this study are from the European Social Survey (ESS), specifically, the Trust in Justice module available in the fifth round of data collection (2010). In each round, the ESS uses multistage probability sampling to select a sample representative of all individuals age 15 or older, with a target sample size of 1500 and a minimum target response rate of 70% in each country (European Social Survey, 2009). Translation and administration of the questionnaire is performed by a team overseen by a national coordinator, with trained interviewers conducting face-to-face interviews of approximately 1 hour in length. The Trust in Justice module used in this study is one of many rotating modules that, along with the core module, is administered to every participant in the round.

The ESS was fielded in 28 countries in the 5th round, although the present analysis is limited to the 26 countries comprising continental Europe, the United Kingdom, and the Russian Federation, while excluding Israel and Cyprus. (These countries are listed by region in Appendix A, Supporting Information) The ESS meets rigorous standards for cross-national comparability and is widely considered one of the highest quality datasets for international research (Reference Fitzgerald, Jowell, Harkness, Braun, Edwards, Johnson, Lyberg, Mohler, Pennell and SmithFitzgerald & Jowell, 2010). With appropriate weighting, the ESS provides nationally representative as well as comparative estimates of quantities of interest. It is therefore well suited to compare cross-national estimates of police contact, fairness, and legitimacy. The construction of the Trust in Justice module was largely informed by procedural justice theory. More details about conceptual underpinnings and the development of survey measures in the module can be found elsewhere (European Social Survey, 2011).

Measures

The dependent variables of interest are measures of fairness and legitimacy at the respondent level, each of which is conceptualized as a second-order latent variable composed of three first-order latent variables or subconstructs. Our conceptualization of procedural justice is aligned with Reference Hough, Jackson and BradfordHough et al. (2013a, Reference Hough, Jackson, Bradford, Tankebe and Liebling2013b; see also Reference Jackson, Bradford, Hough, Kuha, Stares, Widdop, Fitzgerald, Yordanova and GalevJackson et al., 2011), part of the team which devised the measures included in the Trust in Justice module.Footnote ¹ (Item wording and grouping by subconstructs are provided in Appendix B.) Fairness of the police refers to the belief of the public that police are effective and competent, and there is a presumption of fairness and respectfulness in treatment. Operationally, fairness of police comprises police effectiveness (three items), distributive fairness (two items), and procedural fairness (three items). Legitimacy of police refers to the belief of the public that they are obliged to respect police authority, on the grounds that police share their values and act lawfully. This is operationalized with subconstructs duty to obey (three items), moral alignment (three items), and perceived legality (two items). The latent variable methods employed to create the measures of fairness and legitimacy are described in the next section, and at the relevant point, we comment on alternative latent variable methods used to judge sensitivity. We also return to conceptualization later for extended discussion.

At the country level, the independent variable of interest is an indicator of whether the country is post-communist or part of established European democracies. Eleven countries are classified as post-communist, with the balance of 15 countries classified as established democracies. At the respondent level, the independent variables of interest are measured from reports of police contact and satisfaction. Survey participants were asked the following yes/no question: “In the past 2 years, did the police in [country] approach you, stop you or make contact with you for any reason?” As shown in Table 1, one-third (34%) of respondents reported experiencing a police-initiated contact. Those who answered affirmatively were then asked the follow-up question: “How dissatisfied or satisfied were you with the way the police treated you the last time this happened?” Respondents chose from five ordered response categories, ranging from very dissatisfied (0) to very satisfied (4). As indicated in Table 1, the average police contact is intermediate between “neither dissatisfied nor satisfied” and “satisfied” (mean = 2.5).Footnote ²

Table 1. Descriptive statistics

Note: N = 51,340. Means of binary variables are shown as percentages. Respondent-level variables are weighted (design weight) whereas the country-level variable is unweighted. Note that for all outcome measures, a higher value indicates a higher level of fairness and legitimacy.

^a Second-order latent variable from a principal component analysis.

^b First-order latent variable from a generalized partial credit model.

^c Sample is limited to respondents who report a police contact.

The study is not designed to reveal causal relationships—it is instead descriptive and exploratory. We nevertheless include several control variables to adjust for obvious sources of confounding of the relationship between police contact and perceptions of fairness and legitimacy. The first set includes gender and age as males and youth are much more likely to engage in law-breaking behaviors (Reference Britt, Farrington, Kazemian and PiqueroBritt, 2019; Reference Gartner and TonryGartner, 2011). Further controls include ethnic minority and citizenship statuses because of the disparate law enforcement treatment of minority populations (Reference TonryTonry, 1997). The last set of control variables—education and income—accounts for the fact that persons with lower socioeconomic status are more likely to live in socially disadvantaged areas and thus have a greater risk of experiencing involuntary police encounters (Reference Sampson, Morenoff and Gannon-RowleySampson et al., 2002). Although this literature is largely based on U.S. samples, these factors also tend to be associated with attitudes toward police and the criminal justice system (Reference Brown and ReedBrown & Reed, 2002). In terms of measurement, gender, whether a person belongs to a minority ethnic group in the country, and citizenship status were dichotomous indicators. Control variables measured continuously include age, educational attainment, and household income. The latter two variables possess ordinal rather than interval-ratio metric, but because the mean outcomes are approximately linear in the responses, they are treated as having interval-ratio metric in the regression models.

Analysis

The fairness and legitimacy subconstructs are measured from 16 total items, each with dichotomous or ordinal response coding such that a higher value represents more fairness or legitimacy. The starting point for the analysis is a set of six generalized partial credit models for each first-order construct (police effectiveness, distributive fairness, procedural fairness, duty to obey, moral alignment, and perceived legality), in order to extract the latent ability scores. This is followed by principal components analysis to quantify each second-order construct (fairness of police, legitimacy of police). The factor scores resulting from the latter are then used as dependent variables in a pair of multilevel regression models in which police contact and satisfaction (as well as all respondent-level regressors) are interacted with a dummy indicator for established democracies. The approach is described in more detail below, but Figure 1 is provided as a conceptual guide.

Figure 1. Conceptual diagram. Note: This diagram provides a guide for the analysis. The first-order latent constructs (police effectiveness, distributive fairness, etc.) are ability scores extracted from generalized partial credit models of the observed indicators. Definitions of the observed indicators are provided in Appendix B. The ability scores are then subjected to principal components analysis to obtain the second-order fairness and legitimacy constructs, which are treated as the dependent variables in the analyses reported below

Latent variable models

The generalized partial credit model (GPCM) is a generalization of item response theory (IRT) to polytomous variables (Reference MastersMasters, 1982; Reference MurakiMuraki, 1992; for related models, see Reference Hambleton and SwaminathanHambleton & Swaminathan, 1985; Reference SamejimaSamejima, 1969). The GPCM is a probability model yielding estimates of item characteristics (“discrimination,” “difficulty”) and latent respondent characteristics (“ability”) which underlie response patterns to a set of ordered measures. Formally, the probability that respondent i (of N individuals) chooses ordered response k (of K responses) to item j (of J items) is written using the logistic distribution function:

$\mathrm{Pr}\left(y_{\mathrm{ij}}=k\right)=\frac{\exp \left(\sum _{s=0}^k{a}_j\left({\theta }_i-b_{\mathrm{js}}\right)\right)}{\sum _{t=0}^K\left(\exp \left(\sum _{s=0}^t{a}_j\left({\theta }_i-b_{\mathrm{js}}\right)\right)\right)}$

Notice there are three distinct outputs from the model. First, the parameter θ_i represents a respondent's “ability,” or their position on a latent continuum. A respondent with large θ_i has a systematic tendency to choose higher response categories on all of the items. The estimated value of this person-specific parameter is chosen in such a way as to maximize the likelihood of each subject's observed response pattern, and is the key output that is used for further analysis (described below).Footnote ³ Second, the parameter a_j represents an item's “discrimination,” or the strength of the association between item j and latent ability θ_i. If a_j is large, then subjects with high ability score higher on the item than subjects with low ability, whereas if a_j is small, subjects with high and low ability score similarly. These item-specific parameters are analogous to factor loadings in factor analysis. Third, the parameter b_js represents the item's “step difficulty,” or the level of latent ability beyond which the probability that item j is equal to k (vs. k – 1) exceeds 0.5. If b_js is large, then relatively few subjects choose response k on item j, whereas if b_js is small, many subjects choose the response. These item- and response-specific parameters resemble response thresholds from an ordered logit model. (For interested readers, coefficients from the GPCMs are provided in Appendices C and D.)

The key output from the GPCM is the estimated latent ability score, ${\hat{\theta }}_i$, which in the present study indicates the degree to which a respondent “endorses” the fairness and legitimacy items with respect to the police, on average. Latent ability is assumed to be continuous and normally distributed with zero mean and estimated variance. Because it possesses equal-interval metric, it is suitable for further analyses rooted in normal-error processes. We therefore subject the latent ability scores to two principal component analyses—one for the three fairness subconstructs (police effectiveness, distributive fairness, procedural fairness) and one for the three legitimacy subconstructs (duty to obey, moral alignment, perceived legality). The resulting factor scores are then used as dependent variables in multilevel regression models, described next. We refer to the resulting factor scores as ${\hat{\theta }}_i^{*}$ to distinguish them from the subconstruct latent ability score estimates, ${\hat{\theta }}_i$, in the description of the multilevel models below.

Multilevel regression models

Because the 51,340 respondents in the ESS are nested in 26 countries, the main portion of the analysis relies on two-level hierarchical models. A simplified version of the model includes a dummy indicator for police contact, a dummy indicator for established democracies, and their interaction (control variables are implied, as are interactions between the established democracy indicator and the control variables):

${\hat{\theta }}_i^{*}=\alpha +\beta {\text{Contact}}_{\mathrm{ij}}+\gamma {\text{Established}}_j+\delta \left({\text{Contact}}_{\mathrm{ij}}\times {\text{Established}}_j\right)+u_j+e_{\mathrm{ij}}$

Given the interaction, the main effect β represents the mean difference in ${\hat{\theta }}_i^{*}$ between respondents in post-communist countries who experienced police-initiated contact in the previous 2 years versus respondents in post-communist countries who did not. The interaction coefficient δ is a contrast pertaining to residents of established democracies, such that $\beta +\delta$ represents the mean difference in ${\hat{\theta }}_i^{*}$ between respondents in established democracies who experienced police contact versus respondents in established democracies who did not. If post-communist countries and established democracies exhibit the same tendencies with respect to the correlation of police contact with fairness and legitimacy, δ will not differ significantly from zero.Footnote ⁴

Given the relatively small number of countries in the sample, we did not control for country-level covariates in the multilevel regression analysis (Reference Bryan and JenkinsBryan & Jenkins, 2016). In addition, for generalizable inference using the ESS when respondent-level data are combined from all countries, normed probability weights are required. Two such weights were incorporated into the analysis. The design weight is a respondent weight that adjusts for inclusion probabilities to ensure that the sample of respondents in any single country is generalizable to the population of that country. The population weight is a country weight that adjusts for differences in total population size to ensure that the combined sample is generalizable to all 26 countries.

RESULTS

To explore mean police contact and satisfaction by country, in Figure 2, we plot predictions from intercept-only multilevel models. These are ranked in a so-called caterpillar plot from the lowest mean to the highest mean, with different markers denoting established democracies and post-communist countries. In each graph, the Pearson correlations between an indicator for established democracies and mean levels of police contact and satisfaction exceed +0.4, and even with only 26 countries, both correlations are statistically significant (p < 0.05). The clear pattern is for respondents in established democracies to experience more police contact, but also to report higher mean satisfaction with those contacts. Specifically, established democracies exhibit more prevalent police contact by an average of 10 percentage points, and higher mean satisfaction by an average of 0.3 on a five-point scale.

Figure 2. Caterpillar plots of country-level police contact. Note: Police contact is binary, so the country-level mean is a proportion. Satisfaction with contact is ordinal, with response categories ranging from 0 (very dissatisfied) to 4 (very satisfied). Pearson's correlation coefficients with an indicator for established democracies are 0.48 (police contact; p < 0.05) and 0.56 (satisfaction with contact; p < 0.01). Spearman's rank correlation coefficient between the country means of police contact and satisfaction is 0.76 (p < 0.001)

Close inspection indicates that in each graph, 10 of the 15 established democracies (67%) are higher than the global median (i.e., rank >13); whereas 8 of the 11 post-communist countries (73%) are lower than the global median (i.e., rank ≤13). What is also worth noting is the range of mean police contact and satisfaction. For example, respondents in Bulgaria report the least prevalent police contact (16%) whereas respondents in Finland report the most (50%). On the other hand, residents of Ukraine experience the lowest mean satisfaction with police contact (mean = 1.7, intermediate between “dissatisfied” and “neither dissatisfied nor satisfied”), whereas in Sweden, mean satisfaction is highest (mean = 3.0, equal to “satisfied”).

In Figure 3, we explore differences in fairness and legitimacy in a similar fashion. In each graph, 11 of the 15 established democracies (73%) possess a higher mean than the global median; whereas 9 of the 11 post-communist countries (82%) are lower than the global median. Because fairness and legitimacy are standardized principal components, the units of the y-axis are standard deviations. Residents of Ukraine thus have mean fairness and legitimacy scores that are one full standard deviation lower than the global mean.

Figure 3. Caterpillar plots of country-level fairness and legitimacy, by region. Note: Spearman's rank correlation coefficient between the country means of fairness and legitimacy is 0.83 (p < 0.001)

Results of the multilevel regression models are available in Table 2. The indicator for established democracies is interacted with all variables, with the summing of interaction effects and the appropriate main effects shown in the table. To simplify interpretation of the results, the coefficients pertaining to established democracies and post-communist countries are shown in separate columns. Note that positive coefficients signify higher ratings of fairness and legitimacy, whereas negative coefficients represent lower ratings.

Table 2. Multilevel regression model results for the correlation of police contact with fairness and legitimacy, by region

Note: Estimates are weighted at the respondent level (design weight) as well as the country level (population weight). Included but not shown are an intercept and dummy indicators for the number of items in the dependent variable with missing values. Shaded coefficients are those that significantly differ across regions (p < 0.05). A positive coefficient indicates a higher level of fairness and legitimacy. The coefficient for police contact is a contrast between respondents who report police contact with which they were “very dissatisfied,” versus respondents who report no police contact at all. The coefficient for satisfaction then denotes the incremental relationship of contact satisfaction (from “very dissatisfied” to “very satisfied”) with the fairness and legitimacy outcomes, conditional on experiencing police contact.

^* p < 0.05.

^** p < 0.01.

^*** p < 0.001 (two-tailed tests).

Before turning our attention to police contact and satisfaction, our key variables, we describe the control variables. Generally, individuals who are older, who are noncitizens, who have less education, and who have more income exhibit higher levels of fairness and legitimacy than their younger, noncitizen, better-educated, and lower-income counterparts. With the exception of education and legitimacy, this is true irrespective of region. Interestingly, minorities report a higher level of fairness and legitimacy, but only in post-communist countries. In the case of fairness, the regional difference is statistically significant. While there is no gender difference in perceptions of fairness, women report significantly lower levels of legitimacy in established democracies, but significantly higher levels in post-communist countries. This regional difference is itself statistically significant. We return to ethnicity and gender at a later point, when we find in some sensitivity analyses that regional differences in both fairness and legitimacy become stronger.

The coefficients for police contact and satisfaction have uniform direction and significance in both outcome models and in both established democracies and post-communist countries, suggesting respondents who have different experiences with the police differ significantly in their perceptions of fairness and legitimacy, irrespective of region. First, recall the police contact coefficients capture the difference in ratings of fairness and legitimacy between respondents who report being “very dissatisfied” with their most recent police contact and respondents who report no police contact at all. These coefficients are consistently negative and significant, indicating that the most-dissatisfied police contacts correspond with erosion in fairness and legitimacy compared to no police contact. Second, the satisfaction coefficients quantify differences in fairness and legitimacy conditional on police contact, by contrasting respondents who differ in their reported level of satisfaction with their last police encounter. These coefficients are consistently positive and significant, which indicates that most-satisfied police contacts correspond with higher ratings of fairness and legitimacy than most-dissatisfied police contacts.

Third, cross-region comparison of the police contact and satisfaction coefficients indicate stronger correlations in post-communist countries than in established democracies. For example, mean ratings of fairness among respondents with very dissatisfied police contacts in established democracies is lower by 0.68 compared to their peers in established democracies with no police contact. In post-communist countries, on the other hand, the mean is lower by 0.89—about 1/3 larger in magnitude in post-communist countries relative to established democracies, and judged against a lower baseline. The same pattern of results holds with respect to ratings of legitimacy—the coefficient for police contact is larger by almost 1/2 in post-communist countries. Furthermore, the incremental improvement in fairness and legitimacy corresponding to the level of satisfaction is significantly larger (and more than 50% larger in magnitude) in post-communist countries than in established democracies.Footnote ⁵

To better understand the relationships indicated in Table 2, Figure 4 provides cross-region comparisons of fairness and legitimacy by police contact experiences. These are predictive margins which hold all other regressors constant at their observed values. The figure shows, first, that being “very dissatisfied” with prior police contact corresponds with significantly lower ratings of fairness and legitimacy relative to no police contact or to “very satisfied” police contact. Respondents with very-satisfied police contacts also report significantly higher ratings than respondents with no police contact. We thus find that police contact experiences are correlated with fairness and legitimacy in such a way that positive experiences are characterized by the highest ratings and negative experiences are characterized by the lowest ratings, irrespective of whether respondents reside in established democracies or post-communist countries.

Figure 4. Predictive margins of country-level fairness and legitimacy, by region and police contact. Note: Within-region differences in predictive margins are all statistically significant (p < 0.001), such that mean fairness and legitimacy are highest among respondents with very satisfied police contact, and lowest among respondents with very dissatisfied police contact. Except for mean fairness conditional on very satisfied police contact, cross-region differences in predictive margins are also statistically significant (p < 0.05), indicating mean fairness and legitimacy are higher in established democracies than in post-communist countries

It can also be seen in the figure that the correlations between police contact experiences and ratings of fairness and legitimacy are stronger in post-communist countries than in established democracies. This can be seen from the fact that the predictive margins differ by a larger margin in post-communist countries. Nor can this be accounted for by the fact that the variables under consideration have larger variances in post-communist countries than in established democracies, as the tests of coefficient significance are consistently larger in post-communist countries relative to established democracies. In summary, while police contact experiences are correlated with fairness and legitimacy in the same way in both established democracies and post-community countries, the correlation is significantly stronger in post-community countries.

In a series of analyses which are not shown, we probe the sensitivity of the foregoing results in a variety of ways. First, we included additional country-level covariates to adjust for obvious sources of confounding of the regional differences in the relationship of police contact and satisfaction with fairness and legitimacy. These were the logged homicide rate, per capita gross domestic product, and Gini coefficient of income inequality, which were included in the model one at a time as well as jointly.Footnote ⁶ Importantly, the existence of the interaction reported in Table 2 was unchanged. We also adjusted for measures of political unrest, by controlling for the total number of casualties due to terrorism from 2001 to 2010 as measured in the Global Terrorism Database (Reference LaFree and DuganLaFree & Dugan, 2007), as well as the volume of political protest from 1990 to 2004 as measured in the World Handbook of Political Indicators IV (Reference Jenkins, Taylor, Abbott, Maher and PetersonJenkins et al., 2018).Footnote ⁷ Doing so also did not alter our conclusions from Table 2.

Second, we explored alternative ways to classify countries than the established democracy versus post-communist dichotomy. For example, we substituted an index of democratic governance known as a polity score from the Polity IV Project (Reference Marshall, Gurr and JaggarsMarshall et al., 2018).Footnote ⁸ We alternatively incorporated individual and aggregate ratings of political rights and civil liberties reported annually in Freedom in the World (Freedom House, 2020).Footnote ⁹ We also measured the length of time elapsed since a country was admitted into the European Union.Footnote ¹⁰ An advantage of these measures is they can be treated as continuous rather than dichotomous measures of democratic strength or longevity. In both instances, the same pattern of results from Table 2 is replicated, and the interactions remained strongly significant. Namely, the correlation between police contact and fairness/legitimacy was more strongly negative in less democratic countries (indicating “very dissatisfied” police contact erodes fairness and legitimacy relative to no contact at all), while the correlation between satisfaction and fairness/legitimacy, conditional on police contact, was more strongly positive in less democratic countries (such that “very satisfied” police contact improved fairness and legitimacy over no police contact). Thus the substitution of continuous measures of democratic governance confirmed our findings from a cruder, dichotomous classification.

Third, we followed the recent advice of Reference Heisig and SchaefferHeisig and Schaeffer (2019) to estimate random coefficients on all of the level-1 regressors involved in a cross-level interaction (which is all 8 respondent-level regressors, in our case). Even with only 26 countries, the random effects were precisely estimated, and the interaction of region with police contact and satisfaction persisted. Fourth, we employed a structural equation model to obtain the first- and second-order latent variables, rather than the combination of generalized partial credit models and principal components analysis. Doing so yielded the same pattern and significance of results, but with an important exception. Besides the interactions of region with police contact and satisfaction, the only other interaction was between region and ethnicity, which was statistically significant in both the fairness and legitimacy models. Minorities reported significantly higher levels of fairness and legitimacy, but only in post-communist countries. In established democracies, by comparison, minorities did not differ from their counterparts.

Fifth and finally, because policing scholars might differ in their conceptualizations of fairness and legitimacy (a point to which we return in the closing section), we treated each of the six subconstructs as a dependent variable in a separate multilevel model (regression results for the six subconstructs are shown in Appendix E). We did this to ensure the findings do not hinge on combining the subconstructs in the way we do to form measures of fairness and legitimacy. Police contact experiences were significantly correlated with all six subconstructs in the same direction as previously reported, irrespective of region. Namely, “very dissatisfied” police contact was correlated with erosion in the fairness and legitimacy subconstructs relative to no police contact, and higher levels of satisfaction (conditional on police contact) were correlated with higher levels of the fairness and legitimacy subconstructs. Furthermore, the interaction of police contact experiences with region was fully replicated in four of the six models, and partially replicated in a fifth. The single subconstruct indicating no such regional interaction was distributive fairness, which we believe is a notable finding in its own right.

DISCUSSION AND CONCLUSION

Procedural justice theory has profoundly informed our understanding of how interactions with police influence public opinion. The main contribution of our study, informed by procedural justice and historical institutionalism, is the consideration of how political histories may continue to shape policing practices and public opinion about police even after political regimes have changed. This topic has not fared prominently in research on procedural justice, especially not cross-nationally. Our argument is based on what observers of policing in post-communist countries are well aware of—that the pace and depth of reforms was lacking, and that the communist legacy remains influential in how police perform their work. The argument reflects what Reference Beck, Robertson, Hinton and NewburnBeck and Robertson (2009, p. 50) concluded when commenting on police reform in post-Soviet Russia: “much of the police reform has in fact been superficial or involved the continuation—and even the reinforcement—of Soviet trends and traditions, rather than a break from past practices and principles.”

There is little doubt, however, that post-communist countries in Europe have witnessed a significant move toward practices and principles of democratic policing, yet our findings suggest lingering differences compared to established democracies. We found that countries formerly under communist governments had consistently and significantly lower prevalence of and satisfaction with police contact. Perhaps not surprisingly given their experiences with authoritarian regimes, Southern European countries found themselves in the middle between the post-communist and established democracies. Despite police contact being associated with attitudes toward fairness and legitimacy across all countries, we also documented that the association was stronger in the post-communist region. These findings are generally in line with research that finds mistrust and fear of police to be more common in Eastern and Central Europe (e.g., Reference Gerber and MendelsonGerber & Mendelson, 2008; Reference Schaap and ScheepersSchaap & Scheepers, 2014).

In light of our results, we should return to the question to what extent have the police in transition left their undemocratic past behind them? Efforts at reform according to principles of democratic policing likely have helped police transition into becoming a more democratic organization, although evidence and opinions about the impact of police reforms in European countries during transition to democracy is mixed (Reference KádárKádár, 2001). It is also perhaps indicative of a positive change that differences between post-communist countries and established democracies we found in our data are not vast, suggesting they might be converging. However, with increasing powers granted to police as they are tasked with policing the refugee crisis, it is not certain that differences will continue to narrow (Reference Aas and BosworthAas & Bosworth, 2013). When police are given expanded rights to use force, with little oversight and independent monitoring, the trust that the public has in law enforcement—and its fairness and legitimacy—may deteriorate.

We should also note that even though trust in governments has generally been increasing, with some variation, in post-communist countries, public perception of law enforcement is tainted by issues that post-communist countries continue to face to a greater extent than their Western and North European counterparts. While arguably less of an issue relative to the early years of democratic transition, corruption in the public sector is still relatively high and the public continues to see corruption as a significant problem (Transparency International, 2019). Law enforcement represents the state most visibly and the public support it enjoys is closely linked to how the public perceives government institutions more broadly (Reference Kutnjak IvkovićKutnjak Ivković, 2008). In that respect, the extent to which the public in post-communist countries will perceive fair and legitimate authority is not only related to how police perform their jobs, but also to how the government is perceived. In that sense, if the public has low opinion about the government, it may also have low expectations of the services the government provides, including law enforcement. As long as government-related issues like corruption continue to plague post-communist countries, increase in perceptions of police fairness and legitimacy may be slow to materialize.

Even though trajectories of policing and police–public interactions since the fall of Communism, as we argued, are similar and path dependent, there is also idiosyncrasy in the way police forces have changed and the extent to which they remained the same. When describing the emergence of East European capitalism, Reference Stark and BrusztStark and Bruszt (2001, p. 1130) commented that “the ruins of communism were not a tabula rasa” but, they also added, path dependency is not entirely past dependency. Social and political actors creatively use the resources at their disposal to assemble systems with unique characteristics. In that respect, despite arguing that the results we have observed in our study are a product of similar political pasts, they also reflect idiosyncratic local circumstances. Comparative knowledge about police contact and public opinion about law enforcement would benefit from research that digs deeper, using case studies, into the nature of policing, and its change, during the period of democratic transition.

A longer historical view indicates other historical differences between and within regions. Parts of Europe that used to be under the Austro-Hungarian Empire, such as Austria, Hungary, and Croatia had perhaps as many similarities before the second half of the twentieth century as they had differences afterward. Under Franz Joseph, police forces were established and have taken on some of the contours of policing that were preserved later on, such as centralized governance and a quasi-militaristic command structure (Reference MawbyMawby, 2000). After World War II, however, police forces have kept some of their original features, but the countries under the Soviet influence were molded to maintain the domination of their respective communist parties. This influence was propagated by a standardized system of policing that was “developed in Moscow and exported throughout the socialist societies” (Reference Shelley and MawbyShelley, 1999, p. 75). The result was significant uniformity in the function, structure, and legitimacy of police across communist Europe—the political function was emphasized, alongside a militaristic structure, whereas legitimacy arose from the party, not from the people (Reference MawbyMawby, 2000).

Furthermore, the post-communist divide is not the only conceptual distinction that can be applied to the European continent. Another classification distinguishes between countries in the core and periphery, and is based on a nation's relationship to the global market economy (Reference WallersteinWallerstein, 2004). According to the world systems theory, core countries are those which are commonly associated with the wealthier, more highly industrialized, capitalist West. The periphery, on the other hand, comprises the Central and Eastern European countries (and the Russian Federation), most of which have experienced the transition to a market economy after emerging from Soviet influence in the last three decades. Our classification between post-communist countries and established democracies maps almost exactly onto the core-periphery distinction (Reference BabonesBabones, 2005; Reference Chase-Dunn, Kawano and BrewerChase-Dunn et al., 2000).

While we conceptualized fairness and legitimacy in a way consistent with Reference Hough, Jackson and BradfordHough et al. (2013a, Reference Hough, Jackson, Bradford, Tankebe and Liebling2013b; Reference Jackson, Bradford, Hough, Kuha, Stares, Widdop, Fitzgerald, Yordanova and GalevJackson et al., 2011), there is not yet widespread agreement among policing scholars about how best to conceive and measure these constructs, least of all in a comparative context.Footnote ¹¹ We conceptualized perceived fairness as the belief of the public that police are effective and competent, and that there is a presumption of fairness and respectfulness in treatment (police effectiveness, distributive fairness, and procedural fairness). This resembles a certain level of trust as defined by Hough and colleagues. Expectations of fair and effective actions by institutional actors would seem to bear little resemblance to notions of trust which emphasize a degree of emotional vulnerability to uncertainty in social relations. For example, Reference BarbaletBarbalet (2009) views trust as expectations regarding the behavior of others in the absence of pertinent knowledge but carrying with them a risk of betrayal, Reference Giddenswhile Giddens (1991) defines trust as the confidence in the reliability of others which serves as an emotional inoculation against existential anxiety. Interest-based notions of trust emphasize what Reference HardinHardin (1993, Reference Hardin2002) refers to as “encapsulated interest,” or one's judgment that it is in the interest of others to take one's own interests into consideration. Emotional- and interest-based conceptions of trust presume that individuals value continuity in the established relationship, which is of doubtful relevance in one-off encounters like police-citizen interactions.

We conceptualized legitimacy as the belief of the public that they are obliged to respect police authority, on the grounds that police share their values and act lawfully (obligation to obey, moral alignment, and police legality). For other scholars, what we label fairness in this study is closer to how they conceive legitimacy. For Reference TankebeTankebe (2013), legitimacy as a multidimensional construct comprises police effectiveness, distributive fairness, procedural fairness, and lawfulness, but excludes obligation to obey, which is regarded as an outcome rather than a component of legitimacy. This conceptualization is situated in a dialogic theoretical tradition emphasizing properties of both power holders as well as the audience on the receiving end of the decisions of power holders (Reference Bottoms and TankebeBottoms & Tankebe, 2012, Reference Bottoms, Tankebe, du Bois-Pedain, Ulväng and Asp2017). It is also important to acknowledge that the measurement properties of fairness and legitimacy might systematically differ in comparative contexts, especially in developing and post-colonial nations (Reference Johnson, Maguire and KuhnsJohnson et al., 2014; Reference TankebeTankebe, 2008), as well as among minority groups within a population (Reference Kearns, Ashooh and Lowrey-KinbergKearns et al., 2020; Reference Sargeant, Murphy and CherneySargeant et al., 2014).

Increasingly, scholars are examining how procedural justice theory fares across diverse sets of populations and settings. A study based in Australia, for instance, found that ethnicity moderates the association between procedural justice and cooperation with police, while in a London-based study procedural fairness was more important in predicting cooperation among youth who identified with the UK and another country, rather than only the UK (Reference BradfordBradford, 2014; Reference Murphy and CherneyMurphy & Cherney, 2011). Similarly, in China, South Africa, and Ghana, studies have identified a greater role of police effectiveness in predicting police legitimacy and cooperation compared to studies in Anglo-American samples in which members of the public generally give more weight to procedural fairness (Reference Bradford, Huq, Jackson and RobertsBradford et al., 2014; Reference Sun, Wu, Hu and FarmerSun et al., 2017; Reference TankebeTankebe, 2009). In line with our study, this research highlights the notion that aspects of procedural justice matter differently depending on the groups that police engage with—and the characteristics of society in which policing takes place.

Our study is not without limitations. Perhaps most importantly, the data are cross-sectional, which restricts our ability to draw causal inferences from observed associations. A longitudinal dataset would be more fitting for analysis of historical trajectories and path dependence, but such data at the cross-national level are not available. Even though causal inference was not the goal of the study, it is also critical to note that our findings may be susceptible to alternative explanations. Specifically, public opinion about law enforcement may be driving the extent to which members of the public are in contact with police, thus introducing the potential for reverse causality, yet this is unlikely because police contact in our study was initiated by the police, and not by members of the public. In some instances, however, the opinion that members of the public already have about police and law enforcement more generally may influence how satisfied they are with police contacts once they took place (Reference Nagin and TelepNagin & Telep, 2017; Reference Pina-Sánchez and Brunton-SmithPina-Sánchez & Brunton-Smith, 2020).

Because of a relatively small number of countries in the ESS sample, we were not in a position to comprehensively control for country-level differences. Although our results withstand controls for lethal violence and economic inequality, future research would benefit from compiling a database with more countries over more time periods so more extensive controls can be included. Finally, we do not have information about the nature of the police encounter—if, for instance, the encounter involved physical force or other forms of violence. In that sense, we were better positioned to empirically show that history and context matter, rather than conclusively show why they matter. Future research would profit from examining how the interactions unfolded and, especially, if they involved verbal harassment or physical force.

Almost regardless of context, police force is one of the state institutions that is slowest to change. The insular nature of its culture, skepticism toward outside influence, and the solidarity that characterizes relationships between police officers make efforts aimed at reform all but quick and straightforward (Reference BittnerBittner, 1970; Reference Shelley and MawbyShelley, 1999). Similar to the notion of legal cultures, where the focus is on understanding how courts and the legal profession are shaped by local cultural values and norms (Reference Gibson and CaldeiraGibson & Caldeira, 1996), our study highlighted the importance of historical context. We demonstrated the need to consider how attitudes toward, and experiences with, police may be shaped by political history—and that the influence of history may linger long after the political regime has changed. Our aim was to expand procedural justice theory to more explicitly embrace the role of history and politics, and in doing so become a more valuable tool for understanding the nature of stability and change in contemporary research on the interactions between police and the public.

ACKNOWLEDGMENTS

The first version of the study was presented at the conference “Measures and Meanings of Crime and Justice in the Global South” at Rutgers University-Newark. We would like to thank conference attendees for their valuable feedback.