Police Legitimacy and Homicide: A Macro-Comparative Analysis

Police Legitimacy and Homicide: A Macro-Comparative Analysis Abstract This study tests the claim that police legitimacy affects the prevalence of homicide. Using a cross-national time-series dataset of 100 countries, I conduct a statistical analysis of the association between the extent to which the public perceives the police as legitimate and the homicide rate. The analysis suggests that police legitimacy has a substantial, negative association with homicide rates, consistent across different sources of homicide data and controlling for a variety of economic, political, and demographic variables. The paper provides evidence that police legitimacy is related to violent behavior, and that this relationship is generalizable across a wide range of contexts, but more pronounced in non-high-income and comparatively unequal countries. The unrest and subsequent events surrounding the 2014 fatal shooting by police of Michael Brown in Ferguson, Missouri, brought to the fore the latent tensions resulting from low levels of trust in the police on the part of a significant section of the American population. Following these events, the US Department of Justice released a scathing report of the Ferguson Police Department outlining a pattern of “unlawful police misconduct and court practices [that] have led to distrust and resentment,” particularly among African Americans (US Department of Justice 2015, 79). The report suggests that this systematic unlawful policing “both reflects and reinforces racial bias” of Ferguson’s police force, thereby undermining “community trust” and “law enforcement legitimacy” (US Department of Justice 2015, 2, 4). These sentiments, which are not unique to Ferguson, have contributed to the growth of the Black Lives Matter movement and to what Tyler, Goff, and MacCoun (2015, 76) label a “sense of crisis in the legitimacy of American policing.” As the enforcers of the rule of law, the police occupy a unique role in society in that they are a highly visible state institution that is also authorized to use coercion. As the police have both extraordinary powers and responsibilities, the public’s perceptions of police legitimacy can be quite influential in shaping citizens’ orientations toward state authority and the law, which in turn is argued to have important consequences regarding the rule of law (Levy, Sacks, and Tyler 2009; Sunshine and Tyler 2003; Tankebe 2009; Tyler [1990] 2006; Tyler and Jackson 2014). The police and the judicial system are state institutions that are charged with the function of upholding the rule of law and social order. They have at their disposal two fundamental strategies to perform this function—coercion and consent. Although not mutually exclusive, historically most legal institutions have primarily focused on coercion, based on the assumption that individuals make calculated decisions to obey the law in order to avoid a range of state sanctions. However, a social order based purely on the state’s capacity to punish or coerce its citizens into complying with its directives is not a sustainable strategy of governance; consequently, many suggest that consent is more important than coercion in maintaining the rule of law (Hall 1994; Lukes 2005; Weber [1922] 1978). Consent refers to the inclination on the part of citizens to voluntarily follow the law regardless of the prospect of being punished by the state. The recognition of the importance of consent concerning issues of law and order has led to a recent shift toward the study of legitimacy—acknowledged as an important factor influencing voluntary compliance with the law (Levy, Sacks, and Tyler 2009; Tankebe 2013). Even though legitimacy is largely considered a complement of—rather than a substitute for—deterrence (Tyler, Goff, and MacCoun 2015), legitimacy has been shown to be more influential than deterrence in shaping law-abiding behavior (Tyler [1990] 2006; Tyler and Jackson 2014). Scholars have established the importance of the legitimacy of the state in promoting legal compliance (Levi and Sacks 2009; Levy, Sacks, and Tyler 2009; Marien and Hooghe 2011; Tyler and Jackson 2014). Most of this research has focused on non-violent legal compliance; however, more recent research has begun to examine the effect of legitimacy on violence, including the prevalence of lethal violence (Dawson 2013; Eisner 2001, 2003; LaFree 1998; Roth 2009). Recent cross-national studies suggest that state legitimacy is negatively associated with homicide rates across a large number of countries (Dawson 2017; Nivette and Eisner 2013), and that it accounts for a greater proportion of the variation in cross-national homicides than levels of income inequality and economic development (Dawson 2017). While analyzing the legitimacy of the state as a whole is important, recent research—particularly the procedural justice literature—has specifically focused on the legitimacy of state legal institutions, namely the police (Hough, Jackson, and Bradford 2013; Jackson et al. 2012, 2013; Sunshine and Tyler 2003; Tankebe 2009; Tyler and Jackson 2014; Tyler [1990] 2006). This literature has recognized that “winning hearts and minds is central to the effective use of [police] authority” (Hough, Jackson, and Bradford 2013, 328). Notably, the importance of police legitimacy has recently been acknowledged in the literature as a significant factor influencing the prevalence of violence, including homicide (Jackson et al. 2013; Kane 2005; Kubrin and Weitzer 2003b; Tankebe 2009; Tyler and Jackson 2014). This view is quickly becoming widespread, as evidenced by the subtitle in a recent Economist article discussing murder rates in American cities, which reads: “Lack of trust in police forces is contributing to a spike in murder rates” (The Economist 2015). However, there is a dearth of studies empirically linking the public’s perceived legitimacy of the police to actual instances of homicide. Most research examines the effect of perceived legitimacy on attitudes toward the use of violence (Jackson et al. 2013; Tankebe 2009; Tyler and Jackson 2014). However, as Tankebe (2009, 261) highlights, attitudes toward violence do not necessarily translate into violent behavior. It is therefore imperative to empirically investigate the relationship between police legitimacy and violent behavior, particularly for studies of homicide. Meso- or macro-level analyses (i.e., using neighborhoods, cities, states, or countries as the units of analysis) can help in this regard by using aggregated levels of legitimacy and homicide statistics to directly examine the impact of police legitimacy on the homicide rate. However, there are currently only two studies that do this at the neighborhood level (Kane 2005; Kubrin and Weitzer 2003b), and none do so cross-nationally. The current study is therefore the first to conduct a cross-national analysis of the effect of police legitimacy on the homicide rate. The analysis suggests that police legitimacy is negatively and robustly associated with homicide rates across a variety of cultural and political contexts, with a stronger relationship in middle- and low-income countries and countries with higher levels of inequality. Police Legitimacy, Crime, and Violence Tyler’s ([1990] 2006) groundbreaking work on why people obey the law highlights the importance of the legitimacy of legal authorities, particularly the legitimacy of the police and the courts, in promoting legal compliance. His theory of procedural justice has given rise to a research tradition examining the effects of police legitimacy on crime; however, most research in this area has focused on either non-violent crime (such as obeying traffic laws, buying stolen goods, petty theft, or tax fraud—see Hough, Jackson, and Bradford [2013]; Jackson et al. [2012]; Marien and Hooghe [2011]; Sunshine and Tyler [2003]) or psychological dispositions toward committing violent crimes (Jackson et al. 2013; Tankebe 2009; Tyler and Jackson 2014), and not actual violence per se. There are few studies that empirically examine the association between police legitimacy and violent behavior (one exception is Papachristos, Meares, and Fagan’s [2012] study of physical confrontations). Concerning empirical analyses investigating the effect of police legitimacy on homicide, a thorough review of the literature revealed only two such studies. Kubrin and Weitzer’s (2003b) study of St. Louis homicides suggests that police misconduct (which decreases the legitimacy of the police) in disadvantaged neighborhoods is related to retaliatory homicides (i.e., using murder as an informal dispute resolution mechanism). Similarly, Kane’s (2005) study of variations in violent crime in New York communities (i.e., police precincts) finds that compromised police legitimacy tends to lead to heightened levels of violent crime, including homicide, in structurally disadvantaged communities. Both studies were conducted at the neighborhood (or meso) level of analysis, and both focus on marginalized communities in particular cities in the United States. It remains to be seen whether this negative relationship between police legitimacy and homicide is generalizable beyond these specific neighborhoods in St. Louis and New York. Consequently, the current study addresses this gap by empirically examining the association between police legitimacy and homicide rates cross-nationally. In doing so, this project undertakes the first macro-level test of the applicability of procedural justice theory to homicide. A cross-national, or macro-comparative, analysis of the effects of police legitimacy on homicide has some advantages. This analytic strategy allows for the use of actual incidences of homicide as the dependent variable, rather than simply measuring attitudes or psychological dispositions toward the use of (deadly) violence. It would be difficult to measure the former at the micro level, which focuses on individuals as the unit of analysis. While meso-level research can also feasibly investigate homicide rates (such as the neighborhood studies cited above), another advantage of macro-level research is that it facilitates an assessment of the generalizability of the results across a wider range of contexts and cultures, an undertaking that is currently lacking in the literature (see Eisner and Nivette 2013; Johnson, Maguire, and Kuhns 2014). Aside from a few notable exceptions (Levy, Sacks, and Tyler 2009; Reisig 2009; Tankebe 2009), most studies of procedural justice theory have focused on Western countries. Cross-national analyses are also well positioned to provide insight given the considerable variation in homicide rates across countries. For instance, some countries, such as Japan and Norway, are consistently less violent, while others, such as Jamaica and Honduras, have perennially high murder rates (United Nations Office on Drugs and Crime 2014). The lowest contemporary homicide rates are approximately 0.5 cases per 100,000 population per year, accounting for a very small proportion (approximately 0.04 percent) of all deaths, while the highest peacetime rates of homicide are around 80 cases per 100,000 population per year, amounting to a major cause of death in some countries (Eisner 2013, 141–43; Smith and Green 2007). A macro-comparative cross-national analysis of police legitimacy also contributes to the integration of micro and macro perspectives. As Eisner and Nivette (2013) lament, there has been limited contact between the cross-national homicide literature and the psychological literature, particularly procedural justice theory. They describe this disconnect as surprising, and all the more so since associations between macro-level variables are generally viewed as reflecting causal relations that typically occur at the micro level (Babones 2014). This is clearly the case in studies analyzing homicide rates, which are aggregate measures of individual acts of violence. There are, of course, some limitations to macro-level studies. For example, unlike micro-level analyses, they cannot measure the extent to which the perpetrators of homicide view the police as legitimate. A study of this nature can only measure whether the average, society-wide level of police legitimacy is associated with the homicide rate; however, in doing so, it provides valuable insight. Indeed, the causal mechanisms outlined below tend to operate at the group or societal level, not at the individual level. For instance, even if an individual’s assessment of the perceived legitimacy of the police is at odds with the prevailing view in their communities, there are nonetheless social forces that pressure conformity. That is, in low police legitimacy environments, peers or family members may pressure individuals to commit retaliatory homicide in response to certain actions. A study of societal-level legitimacy could thus inform micro-level studies through examining the effect of social norms on the decision to commit homicide in certain situations. Causal Mechanisms Linking Police Legitimacy and Homicide Although this study is concerned with examining police legitimacy as a factor influencing homicide, as Eisner and Nivette (2013) remind us, it is important to consider the possibility of reverse (or potentially simultaneous) causation. This model of the reverse causal order—that it is the incidence of homicide that shapes the perceived legitimacy of the police—is known as the instrumental perspective. At its most basic, this perspective suggests that the legitimacy of legal institutions is heavily influenced by performance evaluations—that is, evaluations as to their perceived effectiveness. Concerning crime, violent crime and homicide are considered to have a higher level of public preoccupation and are therefore very influential in affecting citizens’ perceptions of their quality of life (e.g., their level of fear of victimization) (Jang, Joo, and Zhao 2010; Reisig and Parks 2000). As the control of violence and crime is widely considered the primary responsibility of the police, the instrumental model suggests that the legitimacy of the police largely depends upon judgments of their effectiveness in controlling violence. The homicide rate is therefore assumed to be used by the public as an indicator of police performance, and these assessments of police performance are assumed to be strongly influential in determining attitudes surrounding police legitimacy (i.e., the level of trust or confidence in, or public support for, the police). This perspective is corroborated by Jang, Joo, and Zhao (2010) in their cross-national study of 15 countries analyzing the determinants of police legitimacy. They found that “people in higher homicide rate countries reported significantly lower levels of confidence in the police” (Jang, Joo, and Zhao 2010, 58). The instrumental model was also tested by Gau et al. (2012) using neighborhood-level variables in a study conducted of a mid-sized Midwestern American city. Of the 31 neighborhoods analyzed, their results suggest that the homicide rate had a negative, albeit statistically insignificant, effect on police legitimacy. Notably, both empirical studies (Gau et al. 2012; Jang, Joo, and Zhao 2010) are cross-sectional and therefore do not provide conclusive insight into the causal direction of the association. Sociologists suggest, however, that meaning and perception are socially constructed, and therefore do not necessarily correspond with “objective” conditions. That is, the social constructionist argument contends that there is no automatic linkage between the homicide rate and assessments of police legitimacy (Baumer, Messner, and Rosenfeld 2003; Blumer 1971). The procedural justice model, which is the causal mechanism most often identified in the literature linking the legitimacy of state legal institutions (primarily the police, but also the courts) to legal compliance, follows this line of argument (Johnson, Maguire, and Kuhns 2014). That is, it rejects the causal order of the instrumental model, suggesting that it is not the primary mechanism connecting police legitimacy to homicide. Procedural justice theory claims that police legitimacy influences the murder rate, and not the reverse. As Sunshine and Tyler (2003, 534) write: “People are not primarily instrumental in their reactions to the police—in other words judging the police in instrumental [i.e., performance] terms.” Rather, it is largely citizens’ personal experiences of the quality of their interactions with the legal authorities that influence their perception of legitimacy of the police (Sunshine and Tyler 2003; Tyler [1990] 2006). Specifically, this model suggests that procedural justice influences police legitimacy, which in turn affects legal compliance. That is, positive judgments surrounding the perceived fairness of police decisions and the exercise of their authority will result in an increase in the perceived legitimacy of the police, thereby leading to a heightened propensity on the part of the public to obey the law (Sunshine and Tyler 2003; Tyler [1990] 2006). Although the present study does not investigate the first half of this causal chain (i.e., the link between procedural justice and police legitimacy), the procedural justice model is nonetheless a good starting point to discuss possible causal mechanisms through which police legitimacy affects homicide. Originally, the procedural justice model focused on explaining non-violent crimes (such as violating traffic laws) and not on violent crimes such as homicide; however, Jackson et al. (2013) have recently extended the procedural justice model to violent crime. In their study of young, male ethnic minorities in London, they find that increased police legitimacy results in less favorable attitudes toward the use of violence to resolve disputes, to take revenge, or to achieve political objectives. In recognizing the police as the coercive agent of the state, the authors suggest that police legitimacy is intricately connected to the acknowledgment of the state’s rightful monopoly on the use of force in society. Consequently, this results in what the authors label a “crowding out” effect or a “zero-sum relationship” between accepting the state’s monopoly over legitimate violence and the approval of the use of non-state (i.e., vigilante) violence (Jackson et al. 2013, 490). The results suggest that a lack of police legitimacy may lead to increased violence, as citizens would be more willing to take the law into their own hands to resolve conflicts. Relatedly, low police legitimacy is thought to weaken informal social control mechanisms considered crucial in supporting the rule of law (see Kane 2005; Kubrin and Weitzer 2003a; LaFree 1998). Namely, the rule of law is strengthened in societies where the law is enforced not only by police, but also informally by citizens—that is, by family members, peers, schoolmates, colleagues, and neighbors that actively defend and uphold the law by reacting to transgressions by visible disapproval, shaming, or other social sanctions. LaFree (1998, 95), for example, contends that when legal authorities are widely perceived as illegitimate, the general public are less likely to vigorously support and defend the law, and will respond less harshly to those who break the law and those prosecuted by the legal system. This view is corroborated by Desmond, Papachristos, and Kirk (2016, 870), who find that events in a community that lead to decreased police legitimacy result in a decrease in the level of crime-reporting. In the same vein, Jackson et al. (2013, 480) maintain that in low police legitimacy environments, individuals are more likely to deem the use of violence to resolve disputes as acceptable and less likely to sanction others for the same behavior. Moreover, as Kubrin and Weitzer (2003a, 379) highlight, in settings where an oppositional subculture includes low levels of trust or confidence in legal authorities and the law, “residents have weaker cultural support for exerting social control over others” and crime is “less vigorously condemned by residents.” As informal social control mechanisms not only provide an additional deterrent (to formal state sanctions) to committing crimes, but also reinforce the legal order by clarifying the limits of acceptable behavior (“Shame” 2003), their absence can lead to a weakening of the rule of law and increased rates of violence and homicide (LaFree 1998; Nivette and Eisner 2013; Schaible and Hughes 2011). Police illegitimacy may also strengthen informal social control, but as a substitute for, rather than a complement of, formal state control and the rule of law. A low level of police legitimacy may not only lead to the rejection of the state for dispute resolution resulting in a “policing vacuum,” but also stimulate the development of strong alternative informal social control mechanisms to resolve conflict (Kubrin and Weitzer 2003b, 159; Kane 2005, 475). As Kubrin and Weitzer (2003b, 159, 160) argue, police practices—such as inadequate crime control or the abusive treatment of citizens—that decrease police legitimacy in the eyes of the public may result in the development of a “street code” or “cultural codes” that support and legitimate the informal, and often violent, resolution of interpersonal disputes. These codes are often based on an oppositional (sub)culture, where “communities generate distinctive values and beliefs that endorse aggressive behaviour and law violation” (Kubrin and Weitzer 2003a, 379, 380). Anderson’s (1999) ethnographic study of the “code of the street” in Philadelphia is one such example. According to Anderson (1999, 34), the code of the street is the result of a “cultural adaptation to a profound lack of faith in the police and the judicial system.” The code of the street is described as: …a set of informal rules governing interpersonal public behavior, particularly violence. The rules prescribe both proper comportment and the proper way to respond if challenged. They regulate the use of violence and so supply a rationale allowing those who are inclined to aggression to precipitate violent encounters in an approved way. (Anderson 1999, 33) This set of informal behavioral rules, which supplant state law and legal enforcement, encourage displays of toughness, including frequent recourse to violence (or at least the threat of violence), to handle a wide variety of disputes or perceived affronts in order to garner honor and respect. Importantly, adhering to these informal behavioral rules also allows one to avoid appearing weak, thereby decreasing the chances of potential victimization. In these communities or societies, even individuals who strive to be law-abiding citizens often must abide by, or at least work within, these rules as a survival strategy (Anderson 1999). Kane (2005, 474) suggests that when the police are widely perceived as illegitimate, then “attempts to mobilize the police in response to violence or potential violence [by residents who would normally be inclined to do so] may seem both futile [e.g., residents may fear harassment by the police] and dangerous [residents may fear reprisals from community members who learn of their cooperation with the police].” Nisbett and Cohen’s (1996, xv) “culture of honor” of the American South or Gray’s (2003, 18) “badness-honour” phenomenon in Jamaica are analogous to Anderson’s code of the street in Philadelphia in that in these environments it becomes imperative to develop a reputation for strength and toughness through demonstrations of readiness to resort to violence to defend against predation. Relatedly, Fiske and Rai (2015) suggest that most homicides are morally motivated in the sense that the killers believe they are not only morally justified in their actions, but in many cases they also understand to have a moral obligation or responsibility to commit homicide in certain situations. As these situations are dictated by local prevailing cultural norms that regulate social relationships, in environments with low police legitimacy that have developed an oppositional culture (such as the “code of the street” or “culture of honor” described above), murder is widely considered a normatively appropriate response, if not a normative imperative, for a larger range of situations. Similarly, Kubrin and Weitzer (2003b) contend that retaliatory homicide is more prevalent where it is culturally supported, as evidenced by their examples of murderers broadly boasting about the homicides they committed, assuming that others would agree with the morality of their actions, and by mothers imploring their sons to kill, rather than call the police, in responding to certain transgressions. Therefore, within these social and cultural contexts (i.e., environments with low police legitimacy), we expect elevated levels of homicide, as it is viewed as an appropriate method, if not a moral obligation, to use as a form of retaliation and to resolve a wider range of interpersonal disputes. In sum, low police legitimacy is expected to result in more favorable attitudes toward the use of violence, to weaken pro-state informal social control mechanisms considered essential in promoting the rule of law, and to strengthen informal social control mechanisms that supplant, rather than support, the law. These mechanisms are in line with the long-standing theoretical tradition advanced by Hobbes ([1651] 1958) suggesting that greater levels of violence and homicide will occur in societies where individuals do not surrender their right of retribution to the state and instead pursue their own dispute resolution mechanisms (see also Black 1983; Nivette 2014). Methodological Design In order to conduct an initial test of these causal mechanisms, this study draws upon cross-national time-series data to examine the relationship between police legitimacy and homicide. I perform statistical analyses using a compiled dataset of all countries with available data, which spans the period from 1995 to 2014. The dataset is unbalanced, with the primary models averaging between 2.1 and 2.8 observations per country and time intervals of different durations between the observations, while 17 percent to 29 percent of all countries in the models have only a single observation. Furthermore, both police legitimacy and especially homicide are sluggish; that is, they change only gradually over time.1 Consequently, typical macro-comparative time-series modeling strategies that place an emphasis on within-country variation (such as fixed or random effects models) are less suitable given the data structure.2 Moreover, as there is no time period common to all countries, it is not possible to run panel-corrected standard error models.3 I therefore use OLS pooled regression with cluster-robust standard errors to control for potential bias resulting from autocorrelation (that is, from pooling multiple observations of the same country at different time points).4 Measuring Police Legitimacy As with the general concept of legitimacy, there is no consensus surrounding the conceptualization and operationalization of police legitimacy (Johnson, Maguire, and Kuhns 2014; Tankebe 2013). It is important to make the distinction between two different conceptions of legitimacy: subjective or perceived legitimacy (also known as empirical legitimacy) and objective legitimacy (also known as normative legitimacy) (Hinsch 2010; Hough, Jackson, and Bradford 2013; Tyler, Goff, and MacCoun 2015). Subjective or perceived legitimacy follows a Weberian, or beliefs-based, conceptualization of legitimacy. This conceptualization can be defined as the extent to which citizens believe the police (or other state institutions) to be legitimate—that is, the public’s assessment surrounding the rightfulness of the police’s authority in the execution of their extraordinary coercive powers to enforce the law and promote order (see Tyler, Goff, and MacCoun 2015, 87). It is therefore based on perceptions and subjective evaluations, which may—but not necessarily—be related to “objective” criteria such as the crime rate or even the extent to which the police act in accordance with the law. Conversely, objective or normative legitimacy is based on an evaluation of legal authorities by a set of universal criteria that in principle could be applied to all societies. These analyses often involve evaluations as to whether laws are in accordance with general social norms and moral standards, and the degree to which legal institutions (including the police) work within the parameters set out by the law (Beetham [1991] 2013; Hough, Jackson, and Bradford 2013). According to this notion of legitimacy, even if a legal system or institution has overwhelming support from the citizenry, if the system or institution does not meet these criteria, it will be deemed as lacking legitimacy. This study examines perceived (or subjective) legitimacy of law enforcement as a causal factor influencing homicide. That is not to say the objective conceptualization of legitimacy is unimportant; however, in the context of the present study, the notion of perceived legitimacy is better matched to the causal mechanisms described above. Namely, it is the perception and subjective evaluations of police legitimacy on the part of the population that are identified as the catalysts driving the police legitimacy–homicide relationship. Tyler’s earlier work measures subjective police legitimacy in two ways: 1) the perceived obligation to obey the police; and 2) the public’s overall trust and confidence in the police (Tyler [1990] 2006, 28–29, 45). However, Bottoms and Tankebe (2012) and Tankebe (2013, 105) suggest that the obligation to obey cannot necessarily be equated to legitimacy, in that the obligation to obey is a much larger concept that may derive from a variety of factors, of which legitimacy is only one, and may not result from perceptions of legitimacy at all. They follow Weber ([1922] 1978), who maintains that legal compliance as a result of legitimacy is distinct from compliance based on other motivations such as coercion, expediency, or custom. This aligns with Gau et al.’s (2012, 341) argument that trust in police, and not the perceived obligation to obey the law, is the primary operative element of police legitimacy. Consequently, I use trust in police, specifically confidence in the police, as a measure of police legitimacy. To measure the level of confidence in the police, I draw upon the World Values Survey (WVS 2015) and its companion survey, the European Values Study (EVS 2015). I use the survey question inquiring about the degree of confidence the respondent has in the police.5 The usable response categories range from 1 to 4 (1 = “None at all”; 2 = “Not very much”; 3 = “Quite a lot”; and 4 = “A great deal”).6 I take the mean of all answers of respondents within a country as a measure of the overall belief in police legitimacy in that country for each survey year. Several Global Barometer project partners (Afrobarometer 2017; Arab Barometer 2017; Asian Barometer 2017; Latinobarómetro 2017) collect comparable data on the same question, albeit with minor differences in response category wording.7 These wording differences could be interpreted as resulting in different distances between response categories, and therefore call into question the statistical comparability of the data. For this reason, the models limited to the WVS/EVS data remain the primary models of the analysis; however, the Global Barometer data are included to evaluate the robustness of the results (roughly doubling the number of observations). Refer to table 1 for summary statistics of the police legitimacy variable using the WVS/EVS data. It reports a difference of 1.75 units of police legitimacy between the highest and lowest scores. As the analysis below investigates the effect of police legitimacy across diverse economic and political contexts, table 1 also presents the distribution of police legitimacy scores for different economic and political country groupings.8 Although high-income countries have on average higher police legitimacy scores and less variation as compared to the group of middle- and low-income countries, there is nonetheless a difference of 1.31 police legitimacy units between the highest and lowest high-income scores. The top scoring high-income countries are Finland, Denmark, Norway, Australia, Canada, New Zealand, Ireland, Singapore, and Turkey9 (all with average police legitimacy scores across all available years greater than 3.0). The high-income countries with the lowest average police legitimacy scores are Argentina, Mexico, Trinidad and Tobago, and Russia, with the US score falling roughly in the middle of the high-income pack. Interestingly, the highest average police legitimacy score belongs to Vietnam, followed by Jordan and Uzbekistan, all non-high-income, higher-inequality, and non-democratic countries. Notably, countries with a higher level of income inequality have, on average, higher mean police legitimacy scores and less within-group variation, while the groups of democratic and non-democratic countries have nearly identical mean police legitimacy scores and roughly comparable score ranges. The lowest police legitimacy score of 1.79 belongs to Pakistan (2012), classified as a democracy that year. Table 1. Summary Statistics of the Police Legitimacy Variable Police legitimacy variable N Mean Standard deviation Min Max All observations 249 2.57 0.36 1.79 3.54 High-income countries 121 2.70 0.29 1.97 3.28 Non-high-income countries 128 2.45 0.39 1.79 3.54 Low-inequality countries 120 2.51 0.41 1.79 3.54 High-inequality countries 129 2.62 0.31 1.99 3.28 Democracies 192 2.56 0.34 1.79 3.28 Non-democracies 57 2.59 0.45 1.87 3.54 Police legitimacy variable N Mean Standard deviation Min Max All observations 249 2.57 0.36 1.79 3.54 High-income countries 121 2.70 0.29 1.97 3.28 Non-high-income countries 128 2.45 0.39 1.79 3.54 Low-inequality countries 120 2.51 0.41 1.79 3.54 High-inequality countries 129 2.62 0.31 1.99 3.28 Democracies 192 2.56 0.34 1.79 3.28 Non-democracies 57 2.59 0.45 1.87 3.54 Table 1. Summary Statistics of the Police Legitimacy Variable Police legitimacy variable N Mean Standard deviation Min Max All observations 249 2.57 0.36 1.79 3.54 High-income countries 121 2.70 0.29 1.97 3.28 Non-high-income countries 128 2.45 0.39 1.79 3.54 Low-inequality countries 120 2.51 0.41 1.79 3.54 High-inequality countries 129 2.62 0.31 1.99 3.28 Democracies 192 2.56 0.34 1.79 3.28 Non-democracies 57 2.59 0.45 1.87 3.54 Police legitimacy variable N Mean Standard deviation Min Max All observations 249 2.57 0.36 1.79 3.54 High-income countries 121 2.70 0.29 1.97 3.28 Non-high-income countries 128 2.45 0.39 1.79 3.54 Low-inequality countries 120 2.51 0.41 1.79 3.54 High-inequality countries 129 2.62 0.31 1.99 3.28 Democracies 192 2.56 0.34 1.79 3.28 Non-democracies 57 2.59 0.45 1.87 3.54 Homicide Rate Data The dependent variable, the natural log of the homicide rate (homicides per 100,000 population), is taken from the United Nations Office on Drugs and Crime (UNODC) Homicide Statistics database (UNODC 2015). This dataset draws from both criminal justice and public health data sources, including the United Nations Survey of Crime Trends and the Operations of Criminal Justice Systems, national government sources, international and regional agencies, and the World Health Organization (see UNODC 2013). It is designed to be comparable across countries and over time, and it includes data that are considered the most reliable that most closely align with UNODC’s definition of homicide as an “unlawful death purposefully inflicted on a person by another person” (UNODC 2013, 109). I use all available homicide rate data for the years 2000–2013. To assess the robustness of the results, I use the World Health Organization’s (WHO) Global Health Estimates as an alternate source of homicide data (WHO 2011, 2014a). These data utilize a standardized definition of homicide across countries and have comprehensive country coverage. The figures are based primarily on vital registration data, but provide estimates and adjustments through statistical modeling for countries with either incomplete or missing data (see WHO 2014c).10 Although the WHO data is used in the UNODC Homicide Statistics database described above, the majority (roughly two-thirds) of the UNODC data are derived from other sources. Accordingly, the WHO dataset is sufficiently different from the UNODC dataset for use as an alternate measure of the homicide rate to assess the robustness of the findings. I use the most recently revised waves of homicide rate data for the years 2000, 2004, 2008 and 2012.11 I perform relatively contemporaneous analyses with the UNODC homicide rate data, by either matching the homicide statistics to the WVS/EVS year or, if that is not possible, plus or minus one year (resulting in data covering the years 1999 to 2014). This procedure generates a total of 171 observations (or country-years) for 82 countries, representing 83.3 percent of the world’s total population. The analysis that includes the Global Barometer data is entirely contemporaneous and provides 375 observations for 91 countries, representing 83.5 percent of the global population. As the WHO homicide data are only available for the years 2000, 2004, 2008, and 2012, I match these homicide data to the four most recent WVS (and two most recent EVS) waves. Accordingly, the 2012 homicide data are applied to the survey years 2010–2014, the 2008 homicide data to the survey years 2005–2009, the 2004 homicide data to the survey years 2000–2004, and the 2000 homicide data to the survey years 1995–1999. This method provides a total of 249 observations for 88 countries, representing 87.0 percent of the global population, while the model that incorporates the Global Barometer data contains 486 observations for 100 countries, representing 88.7 percent of the world’s population. Control Variables To provide a more thorough analysis of the effect of police legitimacy on homicide, the models include five control variables. The first two control variables are economic factors—economic development as measured by the natural log of the gross domestic product (GDP) per capita (Feenstra, Inklaar, and Timmer 2015) and income inequality as measured by the Gini coefficient (Solt 2016).12 Several studies have found a negative relationship between economic development and homicide; it is posited that citizens in rich countries are generally afforded a larger range of opportunities that makes homicide less attractive (LaFree 1999; Neumayer 2003; Trent and Pridemore 2012, 623). Concerning income inequality, many studies suggest that it is positively associated with homicide (LaFree 1999; Nivette 2011; Trent and Pridemore 2012). The effect of inequality is thought to follow the logic of relative deprivation, where frustration caused by the inability to achieve socially prescribed goals may lead to violent aggression, including homicide (see Gurr 1970; Merton 1938). The three other controls are political and demographic variables. I control for the level of democracy, as measured by the Polity IV dataset (Centre for Systematic Peace 2004), as it is thought to be associated with lower levels of homicide. Democracy may affect homicide through institutionalizing conflict along social divides (e.g., class, religious, ethnic, linguistic, or ideological divides) by enabling peaceful political competition and changes in government (Lipset 1963). Moreover, non-democratic regimes are more likely to use violence and violate human rights for political ends, potentially leading to the desensitization to, and modeling of, violence among the citizenry (Neumayer 2003; Nivette 2011; Stamatel 2009, 1428). Ethnic heterogeneity is included as a demographic control variable using Fearon’s (2003) measure of ethnic fractionalization. Ethnic heterogeneity has been found to be positively associated with homicide rates (Avison and Loring 1986; LaFree 1999; Trent and Pridemore 2012), with some suggesting that it potentially decreases levels of social integration and/or fosters feelings of antagonism toward members of other ethnicities, leading to increased levels of homicide (Avison and Loring 1986; Neapolitan 1997). The final control variable measures the proportion of the total population who are males between 15 and 24 years of age (United Nations 2013).13 This demographic group is considered the most likely to commit homicide, and previous studies have shown a positive relationship between young males and homicide (Hirschi and Gottfredson 1983; LaFree 1999; Nivette and Eisner 2013; Trent and Pridemore 2012). Data for all control variables correspond to the WVS, EVS, or Global Barometer survey year. Variable Correlations The correlation matrix (table 2) lists the zero-order correlations of all independent variables using the UNODC homicide rate data. Prior to controls, police legitimacy, GDP per capita (log), income inequality, ethnic heterogeneity, and the youth male population are all fairly strongly correlated with the homicide rate (i.e., with correlation coefficients between 0.48 and 0.61), while democracy is the only variable in the model with a relatively weaker zero-order correlation. Although these correlations suggest that the effect of police legitimacy is roughly as important as the more established cross-national correlates of homicide found in the literature, it is essential to determine whether this finding stands when controlling for the effects of the other independent variables. Table 2. Zero-Order Correlation Matrix (UNODC homicide data) Homicide rate (log) Police legitimacy Democracy GDP/capita (log) Income inequality Ethnic diversity Young male pop Homicide rate (log) 1.0000 Police legitimacy −0.5029 1.0000 Democracy −0.1247 −0.0810 1.0000 GDP/capita (log) −0.4774 0.2974 0.4972 1.0000 Income inequality 0.6138 −0.2047 −0.2026 −0.5457 1.0000 Ethnic diversity 0.5187 −0.3019 −0.1159 −0.3927 0.4921 1.0000 Young male pop 0.5493 −0.2262 −0.5176 −0.8036 0.6454 0.4575 1.0000 Homicide rate (log) Police legitimacy Democracy GDP/capita (log) Income inequality Ethnic diversity Young male pop Homicide rate (log) 1.0000 Police legitimacy −0.5029 1.0000 Democracy −0.1247 −0.0810 1.0000 GDP/capita (log) −0.4774 0.2974 0.4972 1.0000 Income inequality 0.6138 −0.2047 −0.2026 −0.5457 1.0000 Ethnic diversity 0.5187 −0.3019 −0.1159 −0.3927 0.4921 1.0000 Young male pop 0.5493 −0.2262 −0.5176 −0.8036 0.6454 0.4575 1.0000 (obs = 171) Table 2. Zero-Order Correlation Matrix (UNODC homicide data) Homicide rate (log) Police legitimacy Democracy GDP/capita (log) Income inequality Ethnic diversity Young male pop Homicide rate (log) 1.0000 Police legitimacy −0.5029 1.0000 Democracy −0.1247 −0.0810 1.0000 GDP/capita (log) −0.4774 0.2974 0.4972 1.0000 Income inequality 0.6138 −0.2047 −0.2026 −0.5457 1.0000 Ethnic diversity 0.5187 −0.3019 −0.1159 −0.3927 0.4921 1.0000 Young male pop 0.5493 −0.2262 −0.5176 −0.8036 0.6454 0.4575 1.0000 Homicide rate (log) Police legitimacy Democracy GDP/capita (log) Income inequality Ethnic diversity Young male pop Homicide rate (log) 1.0000 Police legitimacy −0.5029 1.0000 Democracy −0.1247 −0.0810 1.0000 GDP/capita (log) −0.4774 0.2974 0.4972 1.0000 Income inequality 0.6138 −0.2047 −0.2026 −0.5457 1.0000 Ethnic diversity 0.5187 −0.3019 −0.1159 −0.3927 0.4921 1.0000 Young male pop 0.5493 −0.2262 −0.5176 −0.8036 0.6454 0.4575 1.0000 (obs = 171) Multicollinearity between the independent variables is tolerable—all variance inflation factor (VIF) scores are 4.3 or less. However, GDP per capita (log) and the proportion of the population that are young males are the only two variables with VIF scores above three, and table 2 indicates that these two variables have the strongest bivariate correlation in the matrix (with a correlation coefficient of −0.80). To test for potential bias as a result of this strong correlation, I ran the models below excluding the young male population variable, then again excluding the GDP per capita (log) variable (results not shown). Their exclusion do not substantially change the results of the effect of police legitimacy, and as such both variables are included in the models. Results Following convention (see Babones 2014), I report the statistical significance of the results of the OLS pooled regression models of the homicide rate. Nonetheless, as the country samples are non-random (i.e., based on data availability) and represent between 83 percent and 89 percent of the world’s total population, the direction and magnitude of each variable’s effect provides valuable information about the sample independent of its statistical significance. Table 3 lists the results of the models that analyze the effect of police legitimacy on the homicide rate. All models that include the police legitimacy variable show a strong and negative regression coefficient, implying that countries with higher levels of police legitimacy tend to have lower homicide rates. Models 1 through 3 use the UNODC homicide data. Model 1 shows that, of the control variables, only income inequality and ethnic diversity achieve statistical significance (at a threshold of p < 0.05). However, with the exception of democracy, the effects of the remaining control variables are in the expected direction. Model 2 indicates that police legitimacy has a statistically significant, negative relationship with the homicide rate. As the homicide rates are logged, the results of model 2 suggest that for every one-unit increase observed in police legitimacy, the homicide rate decreases by approximately 65 percent on average.14 Even with a difference of slightly less than two units between the highest and lowest average police legitimacy scores, the findings of this model suggest that, cross-nationally and net of controls, police legitimacy has a substantial effect on the homicide rate. In model 3, which includes the Global Barometer data, the effect of police legitimacy is similar, notwithstanding the substantially higher number of observations. Table 3. OLS Pooled Regressions of the Homicide Rate (UNODC and WHO homicide data logged) UNODC homicide data WHO homicide data Regressors Model 1 Model 2 Model 3 Model 4 Model 5 Model 6 Model 7 Police legitimacy −1.057*** −1.122*** −0.720*** −0.716*** −0.642*** (0.246) (0.234) (0.258) (0.265) (0.240) Democracy 0.028 0.007 0.016 0.022 0.010 0.010 0.019 (0.032) (0.029) (0.032) (0.027) (0.026) (0.030) (0.026) GDP/capita (log) −0.105 0.048 0.086 −0.471*** −0.352*** −0.239* −0.276* (0.174) (0.159) (0.156) (0.117) (0.126) (0.141) (0.143) Income inequality 0.045*** 0.047*** 0.045** 0.044*** 0.043*** 0.043*** 0.027* (0.017) (0.016) (0.019) (0.014) (0.014) (0.016) (0.015) Ethnic diversity 1.149** 0.799* −0.099 1.198*** 1.050** 1.029** 0.493 (0.497) (0.447) (0.592) (0.447) (0.432) (0.441) (0.512) Young male pop 14.753 14.600* 35.101*** 0.569 2.525 11.176 26.669*** (9.268) (8.410) (10.298) (7.400) (7.032) (8.356) (9.860) Constant −1.339 0.143 −1.292 3.618** 4.361*** 2.538 2.244 (2.297) (1.988) (1.907) (1.509) (1.465) (1.672) (1.878) Observations 171 171 375 249 249 171 486 No. of countries 82 82 91 88 88 82 100 R-squared 0.474 0.569 0.633 0.525 0.566 0.597 0.548 UNODC homicide data WHO homicide data Regressors Model 1 Model 2 Model 3 Model 4 Model 5 Model 6 Model 7 Police legitimacy −1.057*** −1.122*** −0.720*** −0.716*** −0.642*** (0.246) (0.234) (0.258) (0.265) (0.240) Democracy 0.028 0.007 0.016 0.022 0.010 0.010 0.019 (0.032) (0.029) (0.032) (0.027) (0.026) (0.030) (0.026) GDP/capita (log) −0.105 0.048 0.086 −0.471*** −0.352*** −0.239* −0.276* (0.174) (0.159) (0.156) (0.117) (0.126) (0.141) (0.143) Income inequality 0.045*** 0.047*** 0.045** 0.044*** 0.043*** 0.043*** 0.027* (0.017) (0.016) (0.019) (0.014) (0.014) (0.016) (0.015) Ethnic diversity 1.149** 0.799* −0.099 1.198*** 1.050** 1.029** 0.493 (0.497) (0.447) (0.592) (0.447) (0.432) (0.441) (0.512) Young male pop 14.753 14.600* 35.101*** 0.569 2.525 11.176 26.669*** (9.268) (8.410) (10.298) (7.400) (7.032) (8.356) (9.860) Constant −1.339 0.143 −1.292 3.618** 4.361*** 2.538 2.244 (2.297) (1.988) (1.907) (1.509) (1.465) (1.672) (1.878) Observations 171 171 375 249 249 171 486 No. of countries 82 82 91 88 88 82 100 R-squared 0.474 0.569 0.633 0.525 0.566 0.597 0.548 Robust standard errors in parentheses. *** p < 0.01 ** p < 0.05 * p < 0.1 Table 3. OLS Pooled Regressions of the Homicide Rate (UNODC and WHO homicide data logged) UNODC homicide data WHO homicide data Regressors Model 1 Model 2 Model 3 Model 4 Model 5 Model 6 Model 7 Police legitimacy −1.057*** −1.122*** −0.720*** −0.716*** −0.642*** (0.246) (0.234) (0.258) (0.265) (0.240) Democracy 0.028 0.007 0.016 0.022 0.010 0.010 0.019 (0.032) (0.029) (0.032) (0.027) (0.026) (0.030) (0.026) GDP/capita (log) −0.105 0.048 0.086 −0.471*** −0.352*** −0.239* −0.276* (0.174) (0.159) (0.156) (0.117) (0.126) (0.141) (0.143) Income inequality 0.045*** 0.047*** 0.045** 0.044*** 0.043*** 0.043*** 0.027* (0.017) (0.016) (0.019) (0.014) (0.014) (0.016) (0.015) Ethnic diversity 1.149** 0.799* −0.099 1.198*** 1.050** 1.029** 0.493 (0.497) (0.447) (0.592) (0.447) (0.432) (0.441) (0.512) Young male pop 14.753 14.600* 35.101*** 0.569 2.525 11.176 26.669*** (9.268) (8.410) (10.298) (7.400) (7.032) (8.356) (9.860) Constant −1.339 0.143 −1.292 3.618** 4.361*** 2.538 2.244 (2.297) (1.988) (1.907) (1.509) (1.465) (1.672) (1.878) Observations 171 171 375 249 249 171 486 No. of countries 82 82 91 88 88 82 100 R-squared 0.474 0.569 0.633 0.525 0.566 0.597 0.548 UNODC homicide data WHO homicide data Regressors Model 1 Model 2 Model 3 Model 4 Model 5 Model 6 Model 7 Police legitimacy −1.057*** −1.122*** −0.720*** −0.716*** −0.642*** (0.246) (0.234) (0.258) (0.265) (0.240) Democracy 0.028 0.007 0.016 0.022 0.010 0.010 0.019 (0.032) (0.029) (0.032) (0.027) (0.026) (0.030) (0.026) GDP/capita (log) −0.105 0.048 0.086 −0.471*** −0.352*** −0.239* −0.276* (0.174) (0.159) (0.156) (0.117) (0.126) (0.141) (0.143) Income inequality 0.045*** 0.047*** 0.045** 0.044*** 0.043*** 0.043*** 0.027* (0.017) (0.016) (0.019) (0.014) (0.014) (0.016) (0.015) Ethnic diversity 1.149** 0.799* −0.099 1.198*** 1.050** 1.029** 0.493 (0.497) (0.447) (0.592) (0.447) (0.432) (0.441) (0.512) Young male pop 14.753 14.600* 35.101*** 0.569 2.525 11.176 26.669*** (9.268) (8.410) (10.298) (7.400) (7.032) (8.356) (9.860) Constant −1.339 0.143 −1.292 3.618** 4.361*** 2.538 2.244 (2.297) (1.988) (1.907) (1.509) (1.465) (1.672) (1.878) Observations 171 171 375 249 249 171 486 No. of countries 82 82 91 88 88 82 100 R-squared 0.474 0.569 0.633 0.525 0.566 0.597 0.548 Robust standard errors in parentheses. *** p < 0.01 ** p < 0.05 * p < 0.1 Models 4 through 7 of table 3 present the results using the alternate source of homicide statistics—the WHO homicide data. Comparing model 4 to model 1, the effect of GDP per capita (log) is larger and achieves statistical significance. This discrepancy may be explained by the potential bias inherent in the WHO homicide data outlined above—namely that, among other sources, it draws upon gross national income data in formulating homicide estimates in countries with incomplete vital registration data, thereby potentially artificially inflating its statistical impact on the homicide rate. The effect of police legitimacy in model 5 is comparable to its effect in model 2, with the results indicating a statistically significant negative relationship with the WHO homicide rate. According to the model, with each unit increase observed in police legitimacy, the homicide rate decreases, on average, by roughly half.15 To more accurately compare the results of both homicide data sources, model 6 draws upon the WHO homicide data, while limiting the number of observations to those in the UNODC data models. This does not substantively change the effect of police legitimacy as compared to model 5. Likewise, model 7 indicates that the effect of police legitimacy also does not substantively change when the additional observations based on the Global Barometer data are included. To examine the relative importance of the effect of police legitimacy, table 4 lists the partial and the squared semi-partial correlation coefficients for each variable (using the UNODC homicide data). The partial correlation coefficient measures the correlation of a variable with the homicide rate after controlling for the effect of all the other independent variables. The results suggest that, unlike the zero-order correlations presented in table 2, police legitimacy has the strongest correlation with the homicide rate after controls. Furthermore, an examination of the squared semi-partial correlation coefficients reveals that the incremental effect of police legitimacy is the largest of all the independent variables in explaining the variation in the cross-national homicide rate.16 This suggests that the zero-order correlation between police legitimacy and the homicide rate not only holds, but becomes more pronounced with the presence of controls, and that, of all the variables included in the models, police legitimacy is the most consequential in accounting for the variation in homicide rates across countries. Table 4. Correlations with UNODC Homicide Rate (logged) Variable Partial Semi-partial2 Police legitimacy −0.4243 0.0947*** Democracy 0.0378 0.0006 GDP/capita (log) 0.0354 0.0005 Income inequality 0.3578 0.0633*** Ethnic diversity 0.2001 0.0180** Young male pop 0.1695 0.0128** Variable Partial Semi-partial2 Police legitimacy −0.4243 0.0947*** Democracy 0.0378 0.0006 GDP/capita (log) 0.0354 0.0005 Income inequality 0.3578 0.0633*** Ethnic diversity 0.2001 0.0180** Young male pop 0.1695 0.0128** (obs = 171) *** p < 0.01 ** p < 0.05 * p < 0.1 Table 4. Correlations with UNODC Homicide Rate (logged) Variable Partial Semi-partial2 Police legitimacy −0.4243 0.0947*** Democracy 0.0378 0.0006 GDP/capita (log) 0.0354 0.0005 Income inequality 0.3578 0.0633*** Ethnic diversity 0.2001 0.0180** Young male pop 0.1695 0.0128** Variable Partial Semi-partial2 Police legitimacy −0.4243 0.0947*** Democracy 0.0378 0.0006 GDP/capita (log) 0.0354 0.0005 Income inequality 0.3578 0.0633*** Ethnic diversity 0.2001 0.0180** Young male pop 0.1695 0.0128** (obs = 171) *** p < 0.01 ** p < 0.05 * p < 0.1 As past studies have mostly focused on Western countries, models are run separately for high-income and low- and middle-income countries, as well as democratic and non-democratic countries, to examine the relative importance of police legitimacy in non-Western countries (see table 5).17 Although a comparison of models 1 and 2 suggests that police legitimacy has a negative effect on homicide across different categories of economic development, the effect size of police legitimacy is greater in non-high-income than in high-income countries (whose effect does not achieve statistical significance).18 Interestingly, these findings suggest that the police legitimacy-homicide relationship is more pronounced in middle- and low-income countries. Concerning the political context, police legitimacy has a significantly negative effect on homicide across both political categories, though the effect of police legitimacy is slightly stronger in non-democracies (model 4) as compared to democracies (model 3).19 Finally, given previous research on disadvantaged communities, models 5 and 6 use the Gini coefficient as a proxy of the proportion of the total population considered marginalized. The populations of countries with greater levels of inequality are assumed to contain a higher proportion of marginalized people. The results suggest that the effect of police legitimacy on homicide is more pronounced (and only achieves statistical significance) in countries with higher levels of income inequality. Taken together, the findings indicate that the negative association found between police legitimacy and the homicide rate is robust across political contexts, with a stronger relationship in low- and middle-income (i.e., primarily non-Western) countries and in countries where a higher proportion of the total population are marginalized. Table 5. OLS Pooled Regressions of the Homicide Rate (UNODC homicide data logged) High-income countries Non-high-income countries Democracies Non-democracies Low-inequality countries High-inequality countries Regressors Model 1 Model 2 Model 3 Model 4 Model 5 Model 6 Police legitimacy −0.467 −0.975*** −0.657** −1.099*** −0.131 −0.885*** (0.506) (0.272) (0.307) (0.291) (0.515) (0.291) Democracy −0.003 0.027 −0.185** 0.012 −0.055*** 0.065** (0.041) (0.031) (0.083) (0.039) (0.017) (0.030) GDP/capita (log) −0.576 0.185 0.164 −0.221 −0.472* 0.128 (0.463) (0.190) (0.194) (0.304) (0.254) (0.191) Income inequality 0.026 0.055** 0.072*** −0.021 −0.013 0.063** (0.021) (0.022) (0.013) (0.025) (0.026) (0.024) Ethnic diversity 1.570*** 0.177 0.360 0.608 0.891* 0.042 (0.571) (0.603) (0.450) (0.829) (0.455) (0.570) Young male pop 9.672 9.174 13.468 4.723 −3.913 23.138* (12.818) (10.896) (9.171) (14.175) (6.894) (12.080) Constant 5.803 −0.819 −0.883 6.146 6.413*** −2.384 (4.705) (2.482) (2.107) (3.929) (2.147) (2.555) Observations 97 74 136 35 95 76 No. of countries 45 45 61 24 41 46 R-squared 0.510 0.455 0.698 0.378 0.547 0.508 High-income countries Non-high-income countries Democracies Non-democracies Low-inequality countries High-inequality countries Regressors Model 1 Model 2 Model 3 Model 4 Model 5 Model 6 Police legitimacy −0.467 −0.975*** −0.657** −1.099*** −0.131 −0.885*** (0.506) (0.272) (0.307) (0.291) (0.515) (0.291) Democracy −0.003 0.027 −0.185** 0.012 −0.055*** 0.065** (0.041) (0.031) (0.083) (0.039) (0.017) (0.030) GDP/capita (log) −0.576 0.185 0.164 −0.221 −0.472* 0.128 (0.463) (0.190) (0.194) (0.304) (0.254) (0.191) Income inequality 0.026 0.055** 0.072*** −0.021 −0.013 0.063** (0.021) (0.022) (0.013) (0.025) (0.026) (0.024) Ethnic diversity 1.570*** 0.177 0.360 0.608 0.891* 0.042 (0.571) (0.603) (0.450) (0.829) (0.455) (0.570) Young male pop 9.672 9.174 13.468 4.723 −3.913 23.138* (12.818) (10.896) (9.171) (14.175) (6.894) (12.080) Constant 5.803 −0.819 −0.883 6.146 6.413*** −2.384 (4.705) (2.482) (2.107) (3.929) (2.147) (2.555) Observations 97 74 136 35 95 76 No. of countries 45 45 61 24 41 46 R-squared 0.510 0.455 0.698 0.378 0.547 0.508 Robust standard errors in parentheses. *** p < 0.01 ** p < 0.05 * p < 0.1 Table 5. OLS Pooled Regressions of the Homicide Rate (UNODC homicide data logged) High-income countries Non-high-income countries Democracies Non-democracies Low-inequality countries High-inequality countries Regressors Model 1 Model 2 Model 3 Model 4 Model 5 Model 6 Police legitimacy −0.467 −0.975*** −0.657** −1.099*** −0.131 −0.885*** (0.506) (0.272) (0.307) (0.291) (0.515) (0.291) Democracy −0.003 0.027 −0.185** 0.012 −0.055*** 0.065** (0.041) (0.031) (0.083) (0.039) (0.017) (0.030) GDP/capita (log) −0.576 0.185 0.164 −0.221 −0.472* 0.128 (0.463) (0.190) (0.194) (0.304) (0.254) (0.191) Income inequality 0.026 0.055** 0.072*** −0.021 −0.013 0.063** (0.021) (0.022) (0.013) (0.025) (0.026) (0.024) Ethnic diversity 1.570*** 0.177 0.360 0.608 0.891* 0.042 (0.571) (0.603) (0.450) (0.829) (0.455) (0.570) Young male pop 9.672 9.174 13.468 4.723 −3.913 23.138* (12.818) (10.896) (9.171) (14.175) (6.894) (12.080) Constant 5.803 −0.819 −0.883 6.146 6.413*** −2.384 (4.705) (2.482) (2.107) (3.929) (2.147) (2.555) Observations 97 74 136 35 95 76 No. of countries 45 45 61 24 41 46 R-squared 0.510 0.455 0.698 0.378 0.547 0.508 High-income countries Non-high-income countries Democracies Non-democracies Low-inequality countries High-inequality countries Regressors Model 1 Model 2 Model 3 Model 4 Model 5 Model 6 Police legitimacy −0.467 −0.975*** −0.657** −1.099*** −0.131 −0.885*** (0.506) (0.272) (0.307) (0.291) (0.515) (0.291) Democracy −0.003 0.027 −0.185** 0.012 −0.055*** 0.065** (0.041) (0.031) (0.083) (0.039) (0.017) (0.030) GDP/capita (log) −0.576 0.185 0.164 −0.221 −0.472* 0.128 (0.463) (0.190) (0.194) (0.304) (0.254) (0.191) Income inequality 0.026 0.055** 0.072*** −0.021 −0.013 0.063** (0.021) (0.022) (0.013) (0.025) (0.026) (0.024) Ethnic diversity 1.570*** 0.177 0.360 0.608 0.891* 0.042 (0.571) (0.603) (0.450) (0.829) (0.455) (0.570) Young male pop 9.672 9.174 13.468 4.723 −3.913 23.138* (12.818) (10.896) (9.171) (14.175) (6.894) (12.080) Constant 5.803 −0.819 −0.883 6.146 6.413*** −2.384 (4.705) (2.482) (2.107) (3.929) (2.147) (2.555) Observations 97 74 136 35 95 76 No. of countries 45 45 61 24 41 46 R-squared 0.510 0.455 0.698 0.378 0.547 0.508 Robust standard errors in parentheses. *** p < 0.01 ** p < 0.05 * p < 0.1 Although the findings suggest a strong association between police legitimacy and homicide, they do not provide insight into the causal order of the association. However, given the data structure (particularly that the variables are slow to change and there is minimal within-country variation in the dataset) and in the absence of a suitable instrumental variable, the task of assessing reverse or simultaneous causality is difficult. Unsurprisingly, models using time lags of varying durations and those that invert the independent and dependent variables (not shown) are inconclusive in this regard. Therefore, the possibility of reverse or simultaneous causation cannot definitively be ruled out. Discussion The quantitative analyses reveal that the public’s perceived level of police legitimacy is related to homicide rates at the country level. They provide strong evidence of a significant and negative relationship between police legitimacy and homicide, suggesting that countries with a higher level of police legitimacy tend to have much lower homicide rates controlling for various economic, political, and demographic variables. This relationship is robust across homicide data sources (i.e., both criminal justice and public health sources) and the number of countries and observations included in the models. The statistical impact of police legitimacy on the homicide rate is considerable, surpassing that of the other variables modeled, including income inequality, which is considered by some as the most important cross-national determinant of homicide (see Trent and Pridemore 2012). This study also provides evidence that the negative relationship between police legitimacy and lethal violence is generalizable across a wide range of contexts. The results suggest that previous findings of a negative relationship between police legitimacy and homicide in certain disadvantaged American urban neighborhoods are applicable on a much broader scale—that is, the association holds using country-level data and nationally representative samples. Of course, as the data are aggregated, they do not permit a test of whether this association is consistent across the entire population or whether it is more acute in specific population subsections, which is a task for future research using disaggregated data. Conversely, the data do permit an assessment of the police legitimacy–homicide relationship across various country groupings. The finding that the relationship is stronger in countries with relatively higher levels of inequality points to the possibility that marginalized communities are important drivers of this association. Moreover, the strong association between police legitimacy and homicide in low- and middle-income countries implies that there is a potentially fruitful opportunity to further investigate this relationship in non-Western countries, particularly as this group of countries has greater variation in average levels of police legitimacy. Another contribution of this study is the empirical investigation of the relationship between the aggregate level of police legitimacy and homicide rates at the country level. As such, the analysis integrates micro- and macro-level research by testing the applicability of (the latter half of) the procedural justice theory causal mechanism at the macro level. As mentioned, micro-level studies have identified the importance of police legitimacy in influencing attitudes toward the use of violence, rather than actual violent behavior. As it is possible, as Tankebe (2009) suggests, that attitudes toward violence may not be perfectly correlated with violent behavior, the present study addresses this limitation by employing homicide levels, rather than attitudes toward the use of violence, as the dependent variable. Accordingly, it fills an important gap in the procedural justice literature by providing strong empirical support for the claim that, beyond more favorable attitudes toward the use of violence, lower levels of police legitimacy are related to higher rates of homicidal violence. This study also differs from past procedural justice research in that it assesses the public’s overall perception of police legitimacy (by examining the average level of police legitimacy within a given society), rather than the level of the perceived legitimacy of the police held by the perpetrators of acts of violence. This aligns with the causal mechanisms outlined above that underscore the importance of social norms, not only in shaping the individual’s perception of the police through socialization, but also in exerting social pressure to potentially respond to certain situations with violence. The finding that societal levels of police legitimacy have a robust negative relationship with homicide rates therefore makes a significant contribution to the literature. Furthermore, it opens the door to future social-psychological research examining whether average levels of police legitimacy within a given society have an effect on the homicide rate independent of the views of police legitimacy held by those who commit murder. Although the statistical findings provide evidence that police legitimacy is related to homicide, they provide limited insight into the causal mechanisms driving this relationship. The results indirectly support the causal mechanisms described above—namely, that low police legitimacy results in an increased willingness to resort to violence to resolve disputes through weakening pro-state informal social control mechanisms and/or strengthening informal social control mechanisms that supplant the rule of law. While these mechanisms are not mutually exclusive, further research is needed to determine their broader relevance and whether other causal mechanisms are involved. Moreover, the results do not rule out the possibility of reverse or simultaneous causation. Although there is a strong theoretical case for the causal order presented in this paper, it is nonetheless essential to empirically assess the causal direction. However, given the current data limitations, until such time as more longitudinal data become available or a suitable instrumental variable is devised, attempts to empirically determine the causal order will remain challenging. In the meantime, future research examining the effect of the homicide rate on police legitimacy in a model including all known causes of the latter as controls could prove insightful. Assuming that the causal direction flows from police legitimacy to homicide, an important implication of these findings is that increasing police legitimacy, particularly in countries where it is comparatively low, could potentially go a long way in reducing homicide rates. This implies that police forces would be well advised to engage in policing practices that increase their perceived legitimacy among the citizenry and, by the same token, to stop engaging in practices that decrease their perceived legitimacy. For example, the recent Department of Justice reports of both the Ferguson and Baltimore Police Departments (US Department of Justice 2016; 2015) cite egregious instances of systematic “stop-and-search” practices, which have been shown to target visible minorities, lead to few charges, and undermine the legitimacy of the police in both the United States and Canada (Meng, Giwa, and Anucha 2015; Tyler, Goff, and MacCoun 2015). The findings of this study provide renewed reason to end these policing practices. Additionally, countries—particularly low- and middle-income countries—with much lower police legitimacy scores than either Canada or the United States may also have problems such as widespread corruption of the police and/or underfunded police forces, leading to weakened police capacity and possibly police force disengagement, which potentially contribute to higher homicide rates via the causal mechanisms described above. Although police corruption and weak police forces are not new issues, the present analysis underscores their importance in suggesting that, if left unresolved, these problems may contribute to increased levels of lethal violence. Finally, the results support recent empirical research that maintains that, in addition to socio-economic and political variables, cultural factors such as social values and attitudes are an important determinant of cross-national homicide rates (Lappi-Seppälä and Leht 2014). This study not only contributes to the growing body of empirical evidence corroborating this perspective, but identifies a specific social attitude—the public’s overall disposition toward the legitimacy of law enforcement—as being closely linked to the homicide rate across countries. These findings imply that further research investigating the effects of particular social norms and attitudes on the prevalence of violence is a potentially promising avenue of social inquiry. Footnotes 1 The correlation coefficient between the homicide rate and its scores lagged five years is 0.96. For police legitimacy, this correlation coefficient is 0.93. 2 Although random effects models can be estimated without dropping cases and the results are substantively similar to those presented below, they are almost entirely driven by between-country variation. 3 Alternatively, it is possible to run feasible generalized least squares models adjusting for panel heteroskedasticity and autocorrelation as a robustness check. The results (not shown) are substantively similar to those presented below. However, as these models are forced as a result of different time interval lengths between observations, and because panel heteroskedasticity is likely to be a minor concern given the low number of observations per country (see Babones 2014), I present the pooled OLS models. 4 This is performed using the “cluster” option in STATA. As a robustness check for autocorrelation, I rerun the models using Prais-Winsten estimation (not shown), which produces nearly identical results to those shown below. 5 Survey question wording: “I am going to name a number of organizations. For each one, could you tell me how much confidence you have in them: is it a great deal of confidence, quite a lot of confidence, not very much confidence or none at all?: The police” (WVS 2015). 6 Note that I have reversed the order of the original WVS/EVS response categories in order to make the interpretation of the results more intuitive. All other answer categories (“Not applicable,” “No answer,” “Don’t know,” “Missing,” and “Not asked in survey”) are excluded from the analysis. 7 With response options such as “no,” “little,” “some,” and “a lot of confidence,” or “no trust,” trust to “a limited extent,” “a medium extent,” and “a great extent.” 8 High-income countries are defined as those with a GDP per capita of $12,000 or greater. Countries with high levels of inequality are defined as those with a Gini coefficient greater than 35. Democracies are defined by countries with a Polity 2 score of 6 or greater (see Centre for Systematic Peace 2004). 9 Turkey’s police legitimacy scores are taken prior to the Erdogan presidency. 10 The WHO estimation procedure draws upon variants of three variables included in the present statistical analysis (the Gini coefficient, the proportion of the total population who are males aged 15 to 24, and gross domestic product per capita) (WHO 2014b). Consequently, the coefficients of these three control variables will likely be biased (i.e., increasing the likelihood of Type 1 error—i.e., statistical false positives) in the models using the WHO homicide data presented below. However, the coefficient of the focal independent variable, police legitimacy, should not be biased by this estimation technique. 11 While the WHO homicide data are technically not longitudinally comparable due to estimation differences across years, as the focus of the present analysis is primarily on between-country variation, I contend that their use as an alternate data source to assess the robustness of the results is defensible. 12 As Gini coefficients are missing for some years, where possible in the WVS/EVS models, I replace a country’s missing year with its most recent available score (which are a maximum of nine years apart), resulting in an additional 14 to 18 observations. As the models with and without these additional Gini scores produce nearly identical results, the former are presented to maximize the number of observations. 13 As the data are only available at five-year intervals, I perform linear interpolation and extrapolation (using the “ipolate” command in STATA) for the missing values. 14 The proportional change in the homicide rate is 0.35 (i.e., a 65 percent decrease) for every one-unit increase in police legitimacy. This is calculated by taking the exponentiated value of the coefficient of police legitimacy in model 2 (−1.057). 15 The proportional change in the homicide rate is 0.49 (i.e., a 51 percent decrease) for every one-unit increase in police legitimacy in model 5. 16 The squared semi-partial correlation coefficient indicates the marginal increase in proportion of the explained variance (R2) in the models by adding each independent variable to a model composed of all other variables. 17 These results are based on the UNODC homicide data. 18 A one-unit increase in police legitimacy results, on average, in a 37 percent decrease in the homicide rate for high-income countries (model 1) and a 62 percent decrease for non-high-income countries (model 2). 19 In democracies the homicide rate decreases on average 48 percent with each unit increase in police legitimacy, while in non-democratic countries the homicide rate decreases on average 67 percent for every unit increase in police legitimacy. About the Author Andrew Dawson is an Associate Professor of Sociology at York University, Glendon Campus. His recent research focuses on the relationship between state legitimacy and violence, and his latest publications have appeared in The Sociological Quarterly, Social Science History, and World Development. References Afrobarometer . 2017 . “Afrobarometer Data.” Afrobarometer, accessed January 19 2017. http://www.afrobarometer.org/. Anderson , Elijah . 1999 . Code of the Street: Decency, Violence, and the Moral Life of the Inner City , 1st ed . New York : W. W Norton . Arab Barometer . 2017 . “Arab Barometer Data.” Arab Barometer, accessed March 10 2017. http://www.arabbarometer.org. Asian Barometer . 2017 . “Asian Barometer Data.” Asian Barometer (sent via email request), accessed March 15 2017. http://www.asianbarometer.org/. Avison , William , and Pamela Loring . 1986 . “ Population Diversity and Cross-National Homicide: The Effects of Inequality and Heterogeneity .” Criminology 24 ( 4 ): 733 – 49 . Google Scholar CrossRef Search ADS Babones , Salvatore J. 2014 . Methods for Quantitative Macro-Comparative Research . Thousand Oaks, CA : SAGE . Google Scholar CrossRef Search ADS Baumer , Eric , Steven Messner , and Richard Rosenfeld . 2003 . “ Explaining Spatial Variation in Support for Capital Punishment: A Multi-Level Analysis .” American Journal of Sociology 108 ( 4 ): 844 – 75 . Google Scholar CrossRef Search ADS Beetham , David . [ 1991 ] 2013. The Legitimation of Power , 2nd ed. , Political Analysis. New York : Palgrave Macmillan . Google Scholar CrossRef Search ADS Black , Donald . 1983 . “ Crime as Social Control .” American Sociological Review 48 ( 1 ): 34 – 45 . Google Scholar CrossRef Search ADS Blumer , Herbert . 1971 . “ Social Problems as Collective Behavior .” Social Problems 18 : 298 – 306 . Google Scholar CrossRef Search ADS Bottoms , Anthony , and Justice Tankebe . 2012 . “ Beyond Procedural Justice: A Dialogic Approach to Legitimacy in Criminal Justice .” Journal of Criminal Law and Criminology 102 ( 1 ): 119 – 70 . Centre for Systematic Peace . 2004 . “Polity IV Annual Time-Series, 1800–2013.” Accessed June 6 2014. http://www.systemicpeace.org/inscrdata.html. Dawson , Andrew . 2013 . “ The Social Determinants of the Rule of Law: A Comparison of Jamaica and Barbados .” World Development 45 : 314 – 24 . Google Scholar CrossRef Search ADS ——— . 2017 . “ The Belief in State Legitimacy and Homicide: A Cross-National Analysis .” The Sociological Quarterly 58 ( 4 ): 552 – 75 . Google Scholar CrossRef Search ADS Desmond , Matthew , Andrew V. Papachristos , and David S. Kirk . 2016 . “ Police Violence and Citizen Crime Reporting in the Black Community .” American Sociological Review 81 ( 5 ): 857 – 76 . Google Scholar CrossRef Search ADS The Economist . 2015 . “Midsummer Murder.” August 15, pp. 19–20. Eisner , Manuel . 2001 . “ Modernization, Self-Control and Lethal Violence: The Long-Term Dynamics of European Homicide Rates in Theoretical Perspective .” British Journal of Criminology 41 ( 4 ): 618 – 48 . Google Scholar CrossRef Search ADS ——— . 2003 . “ Long-Term Historical Trends in Violent Crime .” Crime and Justice; A Review of Research 30 : 83 – 142 . Google Scholar CrossRef Search ADS ——— . 2013 . “What Causes Large-Scale Variation in Homicide Rates.” In Aggression in Humans and Other Primates: Biology, Psychology, Sociology , edited by Hans-Henning Kortüm and Jürgen Heinze , 137 – 61 . Berlin and Boston : de Gruyter . Eisner , Manuel , and Amy Nivette . 2013 . “Does Low Legitimacy Cause Crime? A Review of the Evidence.” In Legitimacy and Criminal Justice: An International Exploration , edited by Justice Tankebe and Alison Liebling , 308 – 25 . Oxford : Oxford University Press . Google Scholar CrossRef Search ADS EVS . 2015 . “European Values Study, 1981–2008, Longitudinal Data File.” GESIS Data Archive, accessed April 2 2017. http://www.europeanvaluesstudy.eu/. Fearon , James D. 2003 . “ Ethnic and Cultural Diversity by Country .” Journal of Economic Growth 8 ( 2 ): 195 – 222 . Google Scholar CrossRef Search ADS Feenstra , Robert C. , Robert Inklaar , and Marcel P. Timmer . 2015 . “ The Next Generation of the Penn World Table .” American Economic Review 105 ( 10 ): 3150 – 82 . www.ggdc.net/pwt. Google Scholar CrossRef Search ADS Fiske , Alan Page , and Tage Shakti Rai . 2015 . Virtuous Violence: Hurting and Killing to Create, Sustain, End, and Honor Social Relationships . Cambridge : Cambridge University Press . Gau , Jacinta M. , Nicholas Corsaro , Eric A. Stewart , and Rod K. Brunson . 2012 . “ Examining Macro-Level Impacts on Procedural Justice and Police Legitimacy .” Journal of Criminal Justice 40 : 333 – 43 . Google Scholar CrossRef Search ADS Gray , Obika . 2003 . “Badness-Honour.” In Understanding Crime in Jamaica: New Challenges for Public Policy , edited by Anthony Harriott , 13 – 47 . Kingston, Jamaica : University of the West Indies Press . Gurr , Ted Robert . 1970 . Why Men Rebel . Princeton, NJ : Princeton University Press . Hall , John A. 1994 . Coercion and Consent: Studies on the Modern State . Cambridge : Polity Press . Hinsch , Wilfried . 2010 . “ Justice, Legitimacy, and Constitutional Rights .” Critical Review of International Social and Political Philosophy 13 ( 1 ): 39 – 54 . Google Scholar CrossRef Search ADS Hirschi , Travis , and Michael Gottfredson . 1983 . “ Age and the Explanation of Crime .” American Journal of Sociology 89 ( 3 ): 552 – 84 . Google Scholar CrossRef Search ADS Hobbes , Thomas . [ 1651 ] 1958. Leviathan: Parts I and II . New York : Macmillan Publishing Company . Hough , Mike , Jonathan Jackson , and Ben Bradford . 2013 . “Legitimacy, Trust, and Compliance: An Empirical Test of Procedural Justice Theory Using the European Social Survey.” In Legitimacy and Criminal Justice: An International Exploration , edited by Justice Tankebe and Alison Liebling . Oxford: Oxford University Press . Jackson , Jonathan , Ben Bradford , Mike Hough , Andy Myhill , Paul Quinton , and Tom R. Tyler . 2012 . “ Why Do People Comply with the Law? Legitimacy and the Influence of Legal Institutions .” British Journal of Criminology 52 ( 6 ): 1051 – 71 . Google Scholar CrossRef Search ADS Jackson , Jonathan , Aziz Huq , Ben Bradford , and Tom R. Tyler . 2013 . “ Monopolizing Force? Police Legitimacy and Public Attitudes towards the Acceptability of Violence .” Psychology, Public Policy, and Law 19 ( 4 ): 479 – 97 . Google Scholar CrossRef Search ADS Jang , Hyunseok , Hee-Jong Joo , and Jihong Zhao . 2010 . “ Determinants of Public Confidence in Police: An International Perspective .” Journal of Criminal Justice 38 : 57 – 68 . Google Scholar CrossRef Search ADS Johnson , Devon , Edward R. Maguire , and Joseph B. Kuhns . 2014 . “ Public Perceptions of the Legitimacy of the Law and Legal Authorities: Evidence from the Caribbean .” Law & Society Review 48 ( 4 ): 947 – 78 . Google Scholar CrossRef Search ADS Kane , Robert J. 2005 . “ Compromised Police Legitimacy as a Predictor of Violent Crime in Structurally Disadvantaged Communities .” Criminology 43 ( 2 ): 469 – 98 . Google Scholar CrossRef Search ADS Kubrin , Charis E. , and Ronald Weitzer . 2003 a. “ New Directions in Social Disorganization Theory .” Journal of Research in Crime and Delinquency 40 ( 4 ): 374 – 402 . Google Scholar CrossRef Search ADS ——— . 2003 b. “ Retaliatory Homicide: Concentrated Disadvantage and Neighbourhood Culture .” Social Problems 50 ( 2 ): 157 – 80 . Google Scholar CrossRef Search ADS LaFree , Gary . 1998 . Losing Legitimacy: Street Crime and the Decline of Social Institutions in America, Crime & Society . Boulder : Westview Press . ——— . 1999 . “A Summary Review of Cross-National Comparative Studies of Homicide.” In Homicide: A Sourcebook of Social Research , edited by M. Dwayne Smith and Margaret A. Zahn , 125 – 45 . London : Sage . Lappi-Seppälä , Tapio , and Martti Leht . 2014 . “ Cross-Comparative Perspectives on Global Homicide Trends .” Crime and Justice 43 ( 1 ): 135 – 230 . Google Scholar CrossRef Search ADS Latinobarómetro . 2017 . “Latinobarómetro Data.” Latinobarómetro Corporation, accessed February 20 2017. http://www.latinobarometro.org. Levi , Margaret , and Audrey Sacks . 2009 . “ Legitimating Beliefs: Sources and Indicators .” Regulation & Governance 3 ( 4 ): 311 – 13 . Google Scholar CrossRef Search ADS Levy , Margaret , Audrey Sacks , and Tom R. Tyler . 2009 . “ Conceptualizing Legitimacy, Measuring Legitimating Beliefs .” American Behavioral Scientist 53 ( 3 ): 354 – 75 . Google Scholar CrossRef Search ADS Lipset , Seymour Martin . 1963 . Political Man: The Social Bases of Politics . Anchor Books ed. Garden City, NY : Doubleday & Company . Lukes , Steven . 2005 . Power: A Radical View , 2nd ed . New York : Palgrave Macmillan . Marien , Sofie , and Marc Hooghe . 2011 . “ Does Political Trust Matter? An Empirical Investigation into the Relation between Political Trust and Support for Law Compliance .” European Journal of Political Research 50 : 267 – 91 . Google Scholar CrossRef Search ADS Meng , Yunliang , Sulaimon Giwa , and Uzo Anucha . 2015 . “ Is There Racial Discrimination in Police Stop-and-Searches of Black Youth? A Toronto Case Study .” Canadian Journal of Family and Youth/Le Journal Canadien de Famille et de la Jeunesse 7 ( 1 ): 115 – 48 . Merton , Robert K. 1938 . “ Social Structure and Anomie .” American Sociological Review 3 ( 5 ): 672 – 82 . Google Scholar CrossRef Search ADS Neapolitan , Jerome L. 1997 . Cross-National Crime: A Research Review and Sourcebook . Westport, CT : Greenwood Press . Neumayer , Eric . 2003 . “ Good Policy Can Lower Violent Crime: Evidence from a Cross-National Panel of Homicide Rates, 1980–97 .” Journal of Peace Research 40 ( 6 ): 619 – 40 . Google Scholar CrossRef Search ADS Nisbett , Richard E. , and Dov Cohen . 1996 . Culture of Honor: The Psychology of Violence in the South . New Directions in Social Psychology Boulder : Westview Press . Nivette , Amy . 2011 . “ Cross-National Predictors of Crime: A Meta-Analysis .” Homicide Studies 15 ( 2 ): 103 – 31 . Google Scholar CrossRef Search ADS ——— . 2014 . “ Legitimacy and Crime: Theorizing the Role of the State in Cross-National Criminological Theory .” Theoretical Criminology 18 ( 1 ): 93 – 111 . Google Scholar CrossRef Search ADS Nivette , Amy , and Manuel Eisner . 2013 . “ Do Legitimate Polities Have Fewer Homicides? A Cross-National Analysis .” Homicide Studies 17 ( 1 ): 3 – 26 . Google Scholar CrossRef Search ADS Papachristos , Andrew V. , Tracy L. Meares , and Jeffrey Fagan . 2012 . “ Why Do Criminals Obey the Law? The Influence of Legitimacy and Social Networks on Active Gun Offenders .” Journal of Criminal Law and Criminology 102 ( 2 ): 397 – 440 . Reisig , Michael D. 2009 . “ Procedural Justice, Police Legitimacy, and Helping the Police Fight Crime: Results from a Survey of Jamaican Adolescents .” Police Quarterly 12 ( 1 ): 42 – 62 . Google Scholar CrossRef Search ADS Reisig , Michael D. , and Roger B. Parks . 2000 . “ Experience, Quality of Life, and Neighborhood Context: A Hierarchical Analysis of Satisfaction with Police .” Justice Quarterly 17 : 607 – 30 . Google Scholar CrossRef Search ADS Roth , Randolph . 2009 . American Homicide . Cambridge, MA : Belknap Press of Harvard University Press . Schaible , Lonnie , and Lorine Hughes . 2011 . “ Crime, Shame, Reintegration, and Cross-National Homicide: A Partial Test of Reintegrative Shaming Theory .” The Sociological Quarterly 52 : 104 – 31 . Google Scholar CrossRef Search ADS “Shame.” 2003 . “ Shame, Stigma, and Crime: Evaluating the Efficacy of Shaming Sanctions in Criminal Law .” Harvard Law Review 116 ( 7 ): 2186 – 2207 . CrossRef Search ADS Smith , Delores , and Katherine Green . 2007 . “ Violence Among Youth in Jamaica: A Growing Public Health Risk and Challenge .” Pan American Journal of Public Health 22 ( 6 ): 417 – 24 . Google Scholar CrossRef Search ADS Solt , Frederick . 2016 . “ The Standardized World Income Inequality Database .” Social Science Quarterly 97 ( 5 ): 1267 – 81 . SWIID Version 5.1. Google Scholar CrossRef Search ADS Stamatel , Janet P. 2009 . “ Correlates of National-Level Homicide Variation in Post-Communist East-Central Europe .” Social Forces 87 : 1423 – 48 . Google Scholar CrossRef Search ADS Sunshine , Jason , and Tom R. Tyler . 2003 . “ The Role of Procedural Justice and Legitimacy in Shaping Public Support for Policing .” Law & Society Review 37 ( 3 ): 513 – 48 . Google Scholar CrossRef Search ADS Tankebe , Justice . 2009 . “ Self-Help, Policing and Procedural Justice: Ghanaian Vigilantism and the Rule of Law .” Law & Society Review 43 ( 2 ): 245 – 70 . Google Scholar CrossRef Search ADS ——— . 2013 . “ Viewing Things Differently: The Dimensions of Public Perceptions of Police Legitimacy .” Criminology 51 ( 1 ): 103 – 35 . Google Scholar CrossRef Search ADS Trent , Carol , and William Alex Pridemore . 2012 . “A Review of the Cross-National Empirical Literature on Social Structure and Homicide.” In Handbook of European Homicide Research: Patterns, Explanations, and Country Studies , edited by Marieke C A Liem and William Alex Pridemore , 111 – 35 . New York : Springer . Google Scholar CrossRef Search ADS Tyler , Tom R. [ 1990 ] 2006. Why People Obey the Law . Princeton, NJ : Princeton University Press . Tyler , Tom R. , Phillip Atiba Goff , and Robert J. MacCoun . 2015 . “ The Impact of Psychological Science on Policing in the United States: Procedural Justice, Legitimacy, and Effective Law Enforcement .” Psychological Science in the Public Interest 16 ( 3 ): 75 – 109 . Google Scholar CrossRef Search ADS Tyler , Tom R. , and Jonathan Jackson . 2014 . “ Popular Legitimacy and the Exercise of Legal Authority: Motivating Compliance, Cooperation, and Engagement .” Psychology, Public Policy, and Law 20 ( 1 ): 78 – 95 . Google Scholar CrossRef Search ADS United Nations . 2013 . “Population by Age Groups—Male.” United Nations Department of Economic and Social Affairs, Population Division, accessed June 13 2014. http://esa.un.org/unpd/wpp/Excel-Data/population.htm. United Nations Office on Drugs and Crime . 2014 . “Statistics—Historic Data,” accessed December 15 2014. https://www.unodc.org/unodc/en/data-and-analysis/statistics/historic-data.html. US Department of Justice . 2015 . “Investigation of the Ferguson Police Department.” US Department of Justice Civil Rights Division—March 4, 2015, accessed July 2 2016. https://www.justice.gov/sites/default/files/opa/press-releases/attachments/2015/03/04/ferguson_police_department_report.pdf. ——— . 2016 . “Investigation of the Baltimore City Police Department.” US Department of Justice Civil Rights Division—August 10, 2016, accessed August 16 2016. https://www.justice.gov/opa/file/883366/download. UNODC . 2013 . “Global Study on Homicide.” United Nations Office on Drugs and Crime, accessed June 22 2015. http://www.unodc.org/documents/gsh/pdfs/2014_GLOBAL_HOMICIDE_BOOK_web.pdf. ——— . 2015 . “UNODC Homicide Statistics.” United Nations Office on Drugs and Crime, accessed June 22 2015. https://data.unodc.org/. Weber , Max . [ 1922 ] 1978. Economy and Society: An Outline of Interpretive Sociology . Berkeley : University of California Press . WHO . 2011 . “Disease and Injury Country Estimates, 2004–2008.” World Health Organization, accessed June 18 2015. http://www.who.int/healthinfo/global_burden_disease/estimates_country_2004_2008/en/. ——— . 2014 a. “Cause-Specific Mortality Estimates 2000–2012.” World Health Organization, accessed June 18 2015. http://www.who.int/healthinfo/global_burden_disease/estimates/en/index1.html. ——— . 2014 b. “Estimating Global Homicide Deaths.” World Health Organization, accessed March 24 2016. http://www.who.int/violence_injury_prevention/violence/status_report/2014/report/Estimating_global_homicide_deaths.pdf. ——— . 2014 c. “WHO Methods and Data Sources for Country‐Level Causes of Death 2000–2012.” World Health Organization, Department of Health Statistics and Information Systems, accessed July 22 2015. http://www.who.int/healthinfo/statistics/GlobalCOD_method.pdf. WVS . 2015 . “World Values Survey Longitudinal 1981–2014 Data.” World Values Survey Association, accessed April 2 2017. http://www.worldvaluessurvey.org/. © The Author(s) 2018. Published by Oxford University Press on behalf of the University of North Carolina at Chapel Hill. All rights reserved. For permissions, please e-mail: journals.permissions@oup.com. This article is published and distributed under the terms of the Oxford University Press, Standard Journals Publication Model (https://academic.oup.com/journals/pages/about_us/legal/notices) http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png Social Forces Oxford University Press

Police Legitimacy and Homicide: A Macro-Comparative Analysis

Social Forces , Volume Advance Article – May 24, 2018

Loading next page...
 
/lp/ou_press/police-legitimacy-and-homicide-a-macro-comparative-analysis-a1X6hCWhZw
Publisher
Oxford University Press
Copyright
© The Author(s) 2018. Published by Oxford University Press on behalf of the University of North Carolina at Chapel Hill. All rights reserved. For permissions, please e-mail: journals.permissions@oup.com.
ISSN
0037-7732
eISSN
1534-7605
D.O.I.
10.1093/sf/soy043
Publisher site
See Article on Publisher Site

Abstract

Abstract This study tests the claim that police legitimacy affects the prevalence of homicide. Using a cross-national time-series dataset of 100 countries, I conduct a statistical analysis of the association between the extent to which the public perceives the police as legitimate and the homicide rate. The analysis suggests that police legitimacy has a substantial, negative association with homicide rates, consistent across different sources of homicide data and controlling for a variety of economic, political, and demographic variables. The paper provides evidence that police legitimacy is related to violent behavior, and that this relationship is generalizable across a wide range of contexts, but more pronounced in non-high-income and comparatively unequal countries. The unrest and subsequent events surrounding the 2014 fatal shooting by police of Michael Brown in Ferguson, Missouri, brought to the fore the latent tensions resulting from low levels of trust in the police on the part of a significant section of the American population. Following these events, the US Department of Justice released a scathing report of the Ferguson Police Department outlining a pattern of “unlawful police misconduct and court practices [that] have led to distrust and resentment,” particularly among African Americans (US Department of Justice 2015, 79). The report suggests that this systematic unlawful policing “both reflects and reinforces racial bias” of Ferguson’s police force, thereby undermining “community trust” and “law enforcement legitimacy” (US Department of Justice 2015, 2, 4). These sentiments, which are not unique to Ferguson, have contributed to the growth of the Black Lives Matter movement and to what Tyler, Goff, and MacCoun (2015, 76) label a “sense of crisis in the legitimacy of American policing.” As the enforcers of the rule of law, the police occupy a unique role in society in that they are a highly visible state institution that is also authorized to use coercion. As the police have both extraordinary powers and responsibilities, the public’s perceptions of police legitimacy can be quite influential in shaping citizens’ orientations toward state authority and the law, which in turn is argued to have important consequences regarding the rule of law (Levy, Sacks, and Tyler 2009; Sunshine and Tyler 2003; Tankebe 2009; Tyler [1990] 2006; Tyler and Jackson 2014). The police and the judicial system are state institutions that are charged with the function of upholding the rule of law and social order. They have at their disposal two fundamental strategies to perform this function—coercion and consent. Although not mutually exclusive, historically most legal institutions have primarily focused on coercion, based on the assumption that individuals make calculated decisions to obey the law in order to avoid a range of state sanctions. However, a social order based purely on the state’s capacity to punish or coerce its citizens into complying with its directives is not a sustainable strategy of governance; consequently, many suggest that consent is more important than coercion in maintaining the rule of law (Hall 1994; Lukes 2005; Weber [1922] 1978). Consent refers to the inclination on the part of citizens to voluntarily follow the law regardless of the prospect of being punished by the state. The recognition of the importance of consent concerning issues of law and order has led to a recent shift toward the study of legitimacy—acknowledged as an important factor influencing voluntary compliance with the law (Levy, Sacks, and Tyler 2009; Tankebe 2013). Even though legitimacy is largely considered a complement of—rather than a substitute for—deterrence (Tyler, Goff, and MacCoun 2015), legitimacy has been shown to be more influential than deterrence in shaping law-abiding behavior (Tyler [1990] 2006; Tyler and Jackson 2014). Scholars have established the importance of the legitimacy of the state in promoting legal compliance (Levi and Sacks 2009; Levy, Sacks, and Tyler 2009; Marien and Hooghe 2011; Tyler and Jackson 2014). Most of this research has focused on non-violent legal compliance; however, more recent research has begun to examine the effect of legitimacy on violence, including the prevalence of lethal violence (Dawson 2013; Eisner 2001, 2003; LaFree 1998; Roth 2009). Recent cross-national studies suggest that state legitimacy is negatively associated with homicide rates across a large number of countries (Dawson 2017; Nivette and Eisner 2013), and that it accounts for a greater proportion of the variation in cross-national homicides than levels of income inequality and economic development (Dawson 2017). While analyzing the legitimacy of the state as a whole is important, recent research—particularly the procedural justice literature—has specifically focused on the legitimacy of state legal institutions, namely the police (Hough, Jackson, and Bradford 2013; Jackson et al. 2012, 2013; Sunshine and Tyler 2003; Tankebe 2009; Tyler and Jackson 2014; Tyler [1990] 2006). This literature has recognized that “winning hearts and minds is central to the effective use of [police] authority” (Hough, Jackson, and Bradford 2013, 328). Notably, the importance of police legitimacy has recently been acknowledged in the literature as a significant factor influencing the prevalence of violence, including homicide (Jackson et al. 2013; Kane 2005; Kubrin and Weitzer 2003b; Tankebe 2009; Tyler and Jackson 2014). This view is quickly becoming widespread, as evidenced by the subtitle in a recent Economist article discussing murder rates in American cities, which reads: “Lack of trust in police forces is contributing to a spike in murder rates” (The Economist 2015). However, there is a dearth of studies empirically linking the public’s perceived legitimacy of the police to actual instances of homicide. Most research examines the effect of perceived legitimacy on attitudes toward the use of violence (Jackson et al. 2013; Tankebe 2009; Tyler and Jackson 2014). However, as Tankebe (2009, 261) highlights, attitudes toward violence do not necessarily translate into violent behavior. It is therefore imperative to empirically investigate the relationship between police legitimacy and violent behavior, particularly for studies of homicide. Meso- or macro-level analyses (i.e., using neighborhoods, cities, states, or countries as the units of analysis) can help in this regard by using aggregated levels of legitimacy and homicide statistics to directly examine the impact of police legitimacy on the homicide rate. However, there are currently only two studies that do this at the neighborhood level (Kane 2005; Kubrin and Weitzer 2003b), and none do so cross-nationally. The current study is therefore the first to conduct a cross-national analysis of the effect of police legitimacy on the homicide rate. The analysis suggests that police legitimacy is negatively and robustly associated with homicide rates across a variety of cultural and political contexts, with a stronger relationship in middle- and low-income countries and countries with higher levels of inequality. Police Legitimacy, Crime, and Violence Tyler’s ([1990] 2006) groundbreaking work on why people obey the law highlights the importance of the legitimacy of legal authorities, particularly the legitimacy of the police and the courts, in promoting legal compliance. His theory of procedural justice has given rise to a research tradition examining the effects of police legitimacy on crime; however, most research in this area has focused on either non-violent crime (such as obeying traffic laws, buying stolen goods, petty theft, or tax fraud—see Hough, Jackson, and Bradford [2013]; Jackson et al. [2012]; Marien and Hooghe [2011]; Sunshine and Tyler [2003]) or psychological dispositions toward committing violent crimes (Jackson et al. 2013; Tankebe 2009; Tyler and Jackson 2014), and not actual violence per se. There are few studies that empirically examine the association between police legitimacy and violent behavior (one exception is Papachristos, Meares, and Fagan’s [2012] study of physical confrontations). Concerning empirical analyses investigating the effect of police legitimacy on homicide, a thorough review of the literature revealed only two such studies. Kubrin and Weitzer’s (2003b) study of St. Louis homicides suggests that police misconduct (which decreases the legitimacy of the police) in disadvantaged neighborhoods is related to retaliatory homicides (i.e., using murder as an informal dispute resolution mechanism). Similarly, Kane’s (2005) study of variations in violent crime in New York communities (i.e., police precincts) finds that compromised police legitimacy tends to lead to heightened levels of violent crime, including homicide, in structurally disadvantaged communities. Both studies were conducted at the neighborhood (or meso) level of analysis, and both focus on marginalized communities in particular cities in the United States. It remains to be seen whether this negative relationship between police legitimacy and homicide is generalizable beyond these specific neighborhoods in St. Louis and New York. Consequently, the current study addresses this gap by empirically examining the association between police legitimacy and homicide rates cross-nationally. In doing so, this project undertakes the first macro-level test of the applicability of procedural justice theory to homicide. A cross-national, or macro-comparative, analysis of the effects of police legitimacy on homicide has some advantages. This analytic strategy allows for the use of actual incidences of homicide as the dependent variable, rather than simply measuring attitudes or psychological dispositions toward the use of (deadly) violence. It would be difficult to measure the former at the micro level, which focuses on individuals as the unit of analysis. While meso-level research can also feasibly investigate homicide rates (such as the neighborhood studies cited above), another advantage of macro-level research is that it facilitates an assessment of the generalizability of the results across a wider range of contexts and cultures, an undertaking that is currently lacking in the literature (see Eisner and Nivette 2013; Johnson, Maguire, and Kuhns 2014). Aside from a few notable exceptions (Levy, Sacks, and Tyler 2009; Reisig 2009; Tankebe 2009), most studies of procedural justice theory have focused on Western countries. Cross-national analyses are also well positioned to provide insight given the considerable variation in homicide rates across countries. For instance, some countries, such as Japan and Norway, are consistently less violent, while others, such as Jamaica and Honduras, have perennially high murder rates (United Nations Office on Drugs and Crime 2014). The lowest contemporary homicide rates are approximately 0.5 cases per 100,000 population per year, accounting for a very small proportion (approximately 0.04 percent) of all deaths, while the highest peacetime rates of homicide are around 80 cases per 100,000 population per year, amounting to a major cause of death in some countries (Eisner 2013, 141–43; Smith and Green 2007). A macro-comparative cross-national analysis of police legitimacy also contributes to the integration of micro and macro perspectives. As Eisner and Nivette (2013) lament, there has been limited contact between the cross-national homicide literature and the psychological literature, particularly procedural justice theory. They describe this disconnect as surprising, and all the more so since associations between macro-level variables are generally viewed as reflecting causal relations that typically occur at the micro level (Babones 2014). This is clearly the case in studies analyzing homicide rates, which are aggregate measures of individual acts of violence. There are, of course, some limitations to macro-level studies. For example, unlike micro-level analyses, they cannot measure the extent to which the perpetrators of homicide view the police as legitimate. A study of this nature can only measure whether the average, society-wide level of police legitimacy is associated with the homicide rate; however, in doing so, it provides valuable insight. Indeed, the causal mechanisms outlined below tend to operate at the group or societal level, not at the individual level. For instance, even if an individual’s assessment of the perceived legitimacy of the police is at odds with the prevailing view in their communities, there are nonetheless social forces that pressure conformity. That is, in low police legitimacy environments, peers or family members may pressure individuals to commit retaliatory homicide in response to certain actions. A study of societal-level legitimacy could thus inform micro-level studies through examining the effect of social norms on the decision to commit homicide in certain situations. Causal Mechanisms Linking Police Legitimacy and Homicide Although this study is concerned with examining police legitimacy as a factor influencing homicide, as Eisner and Nivette (2013) remind us, it is important to consider the possibility of reverse (or potentially simultaneous) causation. This model of the reverse causal order—that it is the incidence of homicide that shapes the perceived legitimacy of the police—is known as the instrumental perspective. At its most basic, this perspective suggests that the legitimacy of legal institutions is heavily influenced by performance evaluations—that is, evaluations as to their perceived effectiveness. Concerning crime, violent crime and homicide are considered to have a higher level of public preoccupation and are therefore very influential in affecting citizens’ perceptions of their quality of life (e.g., their level of fear of victimization) (Jang, Joo, and Zhao 2010; Reisig and Parks 2000). As the control of violence and crime is widely considered the primary responsibility of the police, the instrumental model suggests that the legitimacy of the police largely depends upon judgments of their effectiveness in controlling violence. The homicide rate is therefore assumed to be used by the public as an indicator of police performance, and these assessments of police performance are assumed to be strongly influential in determining attitudes surrounding police legitimacy (i.e., the level of trust or confidence in, or public support for, the police). This perspective is corroborated by Jang, Joo, and Zhao (2010) in their cross-national study of 15 countries analyzing the determinants of police legitimacy. They found that “people in higher homicide rate countries reported significantly lower levels of confidence in the police” (Jang, Joo, and Zhao 2010, 58). The instrumental model was also tested by Gau et al. (2012) using neighborhood-level variables in a study conducted of a mid-sized Midwestern American city. Of the 31 neighborhoods analyzed, their results suggest that the homicide rate had a negative, albeit statistically insignificant, effect on police legitimacy. Notably, both empirical studies (Gau et al. 2012; Jang, Joo, and Zhao 2010) are cross-sectional and therefore do not provide conclusive insight into the causal direction of the association. Sociologists suggest, however, that meaning and perception are socially constructed, and therefore do not necessarily correspond with “objective” conditions. That is, the social constructionist argument contends that there is no automatic linkage between the homicide rate and assessments of police legitimacy (Baumer, Messner, and Rosenfeld 2003; Blumer 1971). The procedural justice model, which is the causal mechanism most often identified in the literature linking the legitimacy of state legal institutions (primarily the police, but also the courts) to legal compliance, follows this line of argument (Johnson, Maguire, and Kuhns 2014). That is, it rejects the causal order of the instrumental model, suggesting that it is not the primary mechanism connecting police legitimacy to homicide. Procedural justice theory claims that police legitimacy influences the murder rate, and not the reverse. As Sunshine and Tyler (2003, 534) write: “People are not primarily instrumental in their reactions to the police—in other words judging the police in instrumental [i.e., performance] terms.” Rather, it is largely citizens’ personal experiences of the quality of their interactions with the legal authorities that influence their perception of legitimacy of the police (Sunshine and Tyler 2003; Tyler [1990] 2006). Specifically, this model suggests that procedural justice influences police legitimacy, which in turn affects legal compliance. That is, positive judgments surrounding the perceived fairness of police decisions and the exercise of their authority will result in an increase in the perceived legitimacy of the police, thereby leading to a heightened propensity on the part of the public to obey the law (Sunshine and Tyler 2003; Tyler [1990] 2006). Although the present study does not investigate the first half of this causal chain (i.e., the link between procedural justice and police legitimacy), the procedural justice model is nonetheless a good starting point to discuss possible causal mechanisms through which police legitimacy affects homicide. Originally, the procedural justice model focused on explaining non-violent crimes (such as violating traffic laws) and not on violent crimes such as homicide; however, Jackson et al. (2013) have recently extended the procedural justice model to violent crime. In their study of young, male ethnic minorities in London, they find that increased police legitimacy results in less favorable attitudes toward the use of violence to resolve disputes, to take revenge, or to achieve political objectives. In recognizing the police as the coercive agent of the state, the authors suggest that police legitimacy is intricately connected to the acknowledgment of the state’s rightful monopoly on the use of force in society. Consequently, this results in what the authors label a “crowding out” effect or a “zero-sum relationship” between accepting the state’s monopoly over legitimate violence and the approval of the use of non-state (i.e., vigilante) violence (Jackson et al. 2013, 490). The results suggest that a lack of police legitimacy may lead to increased violence, as citizens would be more willing to take the law into their own hands to resolve conflicts. Relatedly, low police legitimacy is thought to weaken informal social control mechanisms considered crucial in supporting the rule of law (see Kane 2005; Kubrin and Weitzer 2003a; LaFree 1998). Namely, the rule of law is strengthened in societies where the law is enforced not only by police, but also informally by citizens—that is, by family members, peers, schoolmates, colleagues, and neighbors that actively defend and uphold the law by reacting to transgressions by visible disapproval, shaming, or other social sanctions. LaFree (1998, 95), for example, contends that when legal authorities are widely perceived as illegitimate, the general public are less likely to vigorously support and defend the law, and will respond less harshly to those who break the law and those prosecuted by the legal system. This view is corroborated by Desmond, Papachristos, and Kirk (2016, 870), who find that events in a community that lead to decreased police legitimacy result in a decrease in the level of crime-reporting. In the same vein, Jackson et al. (2013, 480) maintain that in low police legitimacy environments, individuals are more likely to deem the use of violence to resolve disputes as acceptable and less likely to sanction others for the same behavior. Moreover, as Kubrin and Weitzer (2003a, 379) highlight, in settings where an oppositional subculture includes low levels of trust or confidence in legal authorities and the law, “residents have weaker cultural support for exerting social control over others” and crime is “less vigorously condemned by residents.” As informal social control mechanisms not only provide an additional deterrent (to formal state sanctions) to committing crimes, but also reinforce the legal order by clarifying the limits of acceptable behavior (“Shame” 2003), their absence can lead to a weakening of the rule of law and increased rates of violence and homicide (LaFree 1998; Nivette and Eisner 2013; Schaible and Hughes 2011). Police illegitimacy may also strengthen informal social control, but as a substitute for, rather than a complement of, formal state control and the rule of law. A low level of police legitimacy may not only lead to the rejection of the state for dispute resolution resulting in a “policing vacuum,” but also stimulate the development of strong alternative informal social control mechanisms to resolve conflict (Kubrin and Weitzer 2003b, 159; Kane 2005, 475). As Kubrin and Weitzer (2003b, 159, 160) argue, police practices—such as inadequate crime control or the abusive treatment of citizens—that decrease police legitimacy in the eyes of the public may result in the development of a “street code” or “cultural codes” that support and legitimate the informal, and often violent, resolution of interpersonal disputes. These codes are often based on an oppositional (sub)culture, where “communities generate distinctive values and beliefs that endorse aggressive behaviour and law violation” (Kubrin and Weitzer 2003a, 379, 380). Anderson’s (1999) ethnographic study of the “code of the street” in Philadelphia is one such example. According to Anderson (1999, 34), the code of the street is the result of a “cultural adaptation to a profound lack of faith in the police and the judicial system.” The code of the street is described as: …a set of informal rules governing interpersonal public behavior, particularly violence. The rules prescribe both proper comportment and the proper way to respond if challenged. They regulate the use of violence and so supply a rationale allowing those who are inclined to aggression to precipitate violent encounters in an approved way. (Anderson 1999, 33) This set of informal behavioral rules, which supplant state law and legal enforcement, encourage displays of toughness, including frequent recourse to violence (or at least the threat of violence), to handle a wide variety of disputes or perceived affronts in order to garner honor and respect. Importantly, adhering to these informal behavioral rules also allows one to avoid appearing weak, thereby decreasing the chances of potential victimization. In these communities or societies, even individuals who strive to be law-abiding citizens often must abide by, or at least work within, these rules as a survival strategy (Anderson 1999). Kane (2005, 474) suggests that when the police are widely perceived as illegitimate, then “attempts to mobilize the police in response to violence or potential violence [by residents who would normally be inclined to do so] may seem both futile [e.g., residents may fear harassment by the police] and dangerous [residents may fear reprisals from community members who learn of their cooperation with the police].” Nisbett and Cohen’s (1996, xv) “culture of honor” of the American South or Gray’s (2003, 18) “badness-honour” phenomenon in Jamaica are analogous to Anderson’s code of the street in Philadelphia in that in these environments it becomes imperative to develop a reputation for strength and toughness through demonstrations of readiness to resort to violence to defend against predation. Relatedly, Fiske and Rai (2015) suggest that most homicides are morally motivated in the sense that the killers believe they are not only morally justified in their actions, but in many cases they also understand to have a moral obligation or responsibility to commit homicide in certain situations. As these situations are dictated by local prevailing cultural norms that regulate social relationships, in environments with low police legitimacy that have developed an oppositional culture (such as the “code of the street” or “culture of honor” described above), murder is widely considered a normatively appropriate response, if not a normative imperative, for a larger range of situations. Similarly, Kubrin and Weitzer (2003b) contend that retaliatory homicide is more prevalent where it is culturally supported, as evidenced by their examples of murderers broadly boasting about the homicides they committed, assuming that others would agree with the morality of their actions, and by mothers imploring their sons to kill, rather than call the police, in responding to certain transgressions. Therefore, within these social and cultural contexts (i.e., environments with low police legitimacy), we expect elevated levels of homicide, as it is viewed as an appropriate method, if not a moral obligation, to use as a form of retaliation and to resolve a wider range of interpersonal disputes. In sum, low police legitimacy is expected to result in more favorable attitudes toward the use of violence, to weaken pro-state informal social control mechanisms considered essential in promoting the rule of law, and to strengthen informal social control mechanisms that supplant, rather than support, the law. These mechanisms are in line with the long-standing theoretical tradition advanced by Hobbes ([1651] 1958) suggesting that greater levels of violence and homicide will occur in societies where individuals do not surrender their right of retribution to the state and instead pursue their own dispute resolution mechanisms (see also Black 1983; Nivette 2014). Methodological Design In order to conduct an initial test of these causal mechanisms, this study draws upon cross-national time-series data to examine the relationship between police legitimacy and homicide. I perform statistical analyses using a compiled dataset of all countries with available data, which spans the period from 1995 to 2014. The dataset is unbalanced, with the primary models averaging between 2.1 and 2.8 observations per country and time intervals of different durations between the observations, while 17 percent to 29 percent of all countries in the models have only a single observation. Furthermore, both police legitimacy and especially homicide are sluggish; that is, they change only gradually over time.1 Consequently, typical macro-comparative time-series modeling strategies that place an emphasis on within-country variation (such as fixed or random effects models) are less suitable given the data structure.2 Moreover, as there is no time period common to all countries, it is not possible to run panel-corrected standard error models.3 I therefore use OLS pooled regression with cluster-robust standard errors to control for potential bias resulting from autocorrelation (that is, from pooling multiple observations of the same country at different time points).4 Measuring Police Legitimacy As with the general concept of legitimacy, there is no consensus surrounding the conceptualization and operationalization of police legitimacy (Johnson, Maguire, and Kuhns 2014; Tankebe 2013). It is important to make the distinction between two different conceptions of legitimacy: subjective or perceived legitimacy (also known as empirical legitimacy) and objective legitimacy (also known as normative legitimacy) (Hinsch 2010; Hough, Jackson, and Bradford 2013; Tyler, Goff, and MacCoun 2015). Subjective or perceived legitimacy follows a Weberian, or beliefs-based, conceptualization of legitimacy. This conceptualization can be defined as the extent to which citizens believe the police (or other state institutions) to be legitimate—that is, the public’s assessment surrounding the rightfulness of the police’s authority in the execution of their extraordinary coercive powers to enforce the law and promote order (see Tyler, Goff, and MacCoun 2015, 87). It is therefore based on perceptions and subjective evaluations, which may—but not necessarily—be related to “objective” criteria such as the crime rate or even the extent to which the police act in accordance with the law. Conversely, objective or normative legitimacy is based on an evaluation of legal authorities by a set of universal criteria that in principle could be applied to all societies. These analyses often involve evaluations as to whether laws are in accordance with general social norms and moral standards, and the degree to which legal institutions (including the police) work within the parameters set out by the law (Beetham [1991] 2013; Hough, Jackson, and Bradford 2013). According to this notion of legitimacy, even if a legal system or institution has overwhelming support from the citizenry, if the system or institution does not meet these criteria, it will be deemed as lacking legitimacy. This study examines perceived (or subjective) legitimacy of law enforcement as a causal factor influencing homicide. That is not to say the objective conceptualization of legitimacy is unimportant; however, in the context of the present study, the notion of perceived legitimacy is better matched to the causal mechanisms described above. Namely, it is the perception and subjective evaluations of police legitimacy on the part of the population that are identified as the catalysts driving the police legitimacy–homicide relationship. Tyler’s earlier work measures subjective police legitimacy in two ways: 1) the perceived obligation to obey the police; and 2) the public’s overall trust and confidence in the police (Tyler [1990] 2006, 28–29, 45). However, Bottoms and Tankebe (2012) and Tankebe (2013, 105) suggest that the obligation to obey cannot necessarily be equated to legitimacy, in that the obligation to obey is a much larger concept that may derive from a variety of factors, of which legitimacy is only one, and may not result from perceptions of legitimacy at all. They follow Weber ([1922] 1978), who maintains that legal compliance as a result of legitimacy is distinct from compliance based on other motivations such as coercion, expediency, or custom. This aligns with Gau et al.’s (2012, 341) argument that trust in police, and not the perceived obligation to obey the law, is the primary operative element of police legitimacy. Consequently, I use trust in police, specifically confidence in the police, as a measure of police legitimacy. To measure the level of confidence in the police, I draw upon the World Values Survey (WVS 2015) and its companion survey, the European Values Study (EVS 2015). I use the survey question inquiring about the degree of confidence the respondent has in the police.5 The usable response categories range from 1 to 4 (1 = “None at all”; 2 = “Not very much”; 3 = “Quite a lot”; and 4 = “A great deal”).6 I take the mean of all answers of respondents within a country as a measure of the overall belief in police legitimacy in that country for each survey year. Several Global Barometer project partners (Afrobarometer 2017; Arab Barometer 2017; Asian Barometer 2017; Latinobarómetro 2017) collect comparable data on the same question, albeit with minor differences in response category wording.7 These wording differences could be interpreted as resulting in different distances between response categories, and therefore call into question the statistical comparability of the data. For this reason, the models limited to the WVS/EVS data remain the primary models of the analysis; however, the Global Barometer data are included to evaluate the robustness of the results (roughly doubling the number of observations). Refer to table 1 for summary statistics of the police legitimacy variable using the WVS/EVS data. It reports a difference of 1.75 units of police legitimacy between the highest and lowest scores. As the analysis below investigates the effect of police legitimacy across diverse economic and political contexts, table 1 also presents the distribution of police legitimacy scores for different economic and political country groupings.8 Although high-income countries have on average higher police legitimacy scores and less variation as compared to the group of middle- and low-income countries, there is nonetheless a difference of 1.31 police legitimacy units between the highest and lowest high-income scores. The top scoring high-income countries are Finland, Denmark, Norway, Australia, Canada, New Zealand, Ireland, Singapore, and Turkey9 (all with average police legitimacy scores across all available years greater than 3.0). The high-income countries with the lowest average police legitimacy scores are Argentina, Mexico, Trinidad and Tobago, and Russia, with the US score falling roughly in the middle of the high-income pack. Interestingly, the highest average police legitimacy score belongs to Vietnam, followed by Jordan and Uzbekistan, all non-high-income, higher-inequality, and non-democratic countries. Notably, countries with a higher level of income inequality have, on average, higher mean police legitimacy scores and less within-group variation, while the groups of democratic and non-democratic countries have nearly identical mean police legitimacy scores and roughly comparable score ranges. The lowest police legitimacy score of 1.79 belongs to Pakistan (2012), classified as a democracy that year. Table 1. Summary Statistics of the Police Legitimacy Variable Police legitimacy variable N Mean Standard deviation Min Max All observations 249 2.57 0.36 1.79 3.54 High-income countries 121 2.70 0.29 1.97 3.28 Non-high-income countries 128 2.45 0.39 1.79 3.54 Low-inequality countries 120 2.51 0.41 1.79 3.54 High-inequality countries 129 2.62 0.31 1.99 3.28 Democracies 192 2.56 0.34 1.79 3.28 Non-democracies 57 2.59 0.45 1.87 3.54 Police legitimacy variable N Mean Standard deviation Min Max All observations 249 2.57 0.36 1.79 3.54 High-income countries 121 2.70 0.29 1.97 3.28 Non-high-income countries 128 2.45 0.39 1.79 3.54 Low-inequality countries 120 2.51 0.41 1.79 3.54 High-inequality countries 129 2.62 0.31 1.99 3.28 Democracies 192 2.56 0.34 1.79 3.28 Non-democracies 57 2.59 0.45 1.87 3.54 Table 1. Summary Statistics of the Police Legitimacy Variable Police legitimacy variable N Mean Standard deviation Min Max All observations 249 2.57 0.36 1.79 3.54 High-income countries 121 2.70 0.29 1.97 3.28 Non-high-income countries 128 2.45 0.39 1.79 3.54 Low-inequality countries 120 2.51 0.41 1.79 3.54 High-inequality countries 129 2.62 0.31 1.99 3.28 Democracies 192 2.56 0.34 1.79 3.28 Non-democracies 57 2.59 0.45 1.87 3.54 Police legitimacy variable N Mean Standard deviation Min Max All observations 249 2.57 0.36 1.79 3.54 High-income countries 121 2.70 0.29 1.97 3.28 Non-high-income countries 128 2.45 0.39 1.79 3.54 Low-inequality countries 120 2.51 0.41 1.79 3.54 High-inequality countries 129 2.62 0.31 1.99 3.28 Democracies 192 2.56 0.34 1.79 3.28 Non-democracies 57 2.59 0.45 1.87 3.54 Homicide Rate Data The dependent variable, the natural log of the homicide rate (homicides per 100,000 population), is taken from the United Nations Office on Drugs and Crime (UNODC) Homicide Statistics database (UNODC 2015). This dataset draws from both criminal justice and public health data sources, including the United Nations Survey of Crime Trends and the Operations of Criminal Justice Systems, national government sources, international and regional agencies, and the World Health Organization (see UNODC 2013). It is designed to be comparable across countries and over time, and it includes data that are considered the most reliable that most closely align with UNODC’s definition of homicide as an “unlawful death purposefully inflicted on a person by another person” (UNODC 2013, 109). I use all available homicide rate data for the years 2000–2013. To assess the robustness of the results, I use the World Health Organization’s (WHO) Global Health Estimates as an alternate source of homicide data (WHO 2011, 2014a). These data utilize a standardized definition of homicide across countries and have comprehensive country coverage. The figures are based primarily on vital registration data, but provide estimates and adjustments through statistical modeling for countries with either incomplete or missing data (see WHO 2014c).10 Although the WHO data is used in the UNODC Homicide Statistics database described above, the majority (roughly two-thirds) of the UNODC data are derived from other sources. Accordingly, the WHO dataset is sufficiently different from the UNODC dataset for use as an alternate measure of the homicide rate to assess the robustness of the findings. I use the most recently revised waves of homicide rate data for the years 2000, 2004, 2008 and 2012.11 I perform relatively contemporaneous analyses with the UNODC homicide rate data, by either matching the homicide statistics to the WVS/EVS year or, if that is not possible, plus or minus one year (resulting in data covering the years 1999 to 2014). This procedure generates a total of 171 observations (or country-years) for 82 countries, representing 83.3 percent of the world’s total population. The analysis that includes the Global Barometer data is entirely contemporaneous and provides 375 observations for 91 countries, representing 83.5 percent of the global population. As the WHO homicide data are only available for the years 2000, 2004, 2008, and 2012, I match these homicide data to the four most recent WVS (and two most recent EVS) waves. Accordingly, the 2012 homicide data are applied to the survey years 2010–2014, the 2008 homicide data to the survey years 2005–2009, the 2004 homicide data to the survey years 2000–2004, and the 2000 homicide data to the survey years 1995–1999. This method provides a total of 249 observations for 88 countries, representing 87.0 percent of the global population, while the model that incorporates the Global Barometer data contains 486 observations for 100 countries, representing 88.7 percent of the world’s population. Control Variables To provide a more thorough analysis of the effect of police legitimacy on homicide, the models include five control variables. The first two control variables are economic factors—economic development as measured by the natural log of the gross domestic product (GDP) per capita (Feenstra, Inklaar, and Timmer 2015) and income inequality as measured by the Gini coefficient (Solt 2016).12 Several studies have found a negative relationship between economic development and homicide; it is posited that citizens in rich countries are generally afforded a larger range of opportunities that makes homicide less attractive (LaFree 1999; Neumayer 2003; Trent and Pridemore 2012, 623). Concerning income inequality, many studies suggest that it is positively associated with homicide (LaFree 1999; Nivette 2011; Trent and Pridemore 2012). The effect of inequality is thought to follow the logic of relative deprivation, where frustration caused by the inability to achieve socially prescribed goals may lead to violent aggression, including homicide (see Gurr 1970; Merton 1938). The three other controls are political and demographic variables. I control for the level of democracy, as measured by the Polity IV dataset (Centre for Systematic Peace 2004), as it is thought to be associated with lower levels of homicide. Democracy may affect homicide through institutionalizing conflict along social divides (e.g., class, religious, ethnic, linguistic, or ideological divides) by enabling peaceful political competition and changes in government (Lipset 1963). Moreover, non-democratic regimes are more likely to use violence and violate human rights for political ends, potentially leading to the desensitization to, and modeling of, violence among the citizenry (Neumayer 2003; Nivette 2011; Stamatel 2009, 1428). Ethnic heterogeneity is included as a demographic control variable using Fearon’s (2003) measure of ethnic fractionalization. Ethnic heterogeneity has been found to be positively associated with homicide rates (Avison and Loring 1986; LaFree 1999; Trent and Pridemore 2012), with some suggesting that it potentially decreases levels of social integration and/or fosters feelings of antagonism toward members of other ethnicities, leading to increased levels of homicide (Avison and Loring 1986; Neapolitan 1997). The final control variable measures the proportion of the total population who are males between 15 and 24 years of age (United Nations 2013).13 This demographic group is considered the most likely to commit homicide, and previous studies have shown a positive relationship between young males and homicide (Hirschi and Gottfredson 1983; LaFree 1999; Nivette and Eisner 2013; Trent and Pridemore 2012). Data for all control variables correspond to the WVS, EVS, or Global Barometer survey year. Variable Correlations The correlation matrix (table 2) lists the zero-order correlations of all independent variables using the UNODC homicide rate data. Prior to controls, police legitimacy, GDP per capita (log), income inequality, ethnic heterogeneity, and the youth male population are all fairly strongly correlated with the homicide rate (i.e., with correlation coefficients between 0.48 and 0.61), while democracy is the only variable in the model with a relatively weaker zero-order correlation. Although these correlations suggest that the effect of police legitimacy is roughly as important as the more established cross-national correlates of homicide found in the literature, it is essential to determine whether this finding stands when controlling for the effects of the other independent variables. Table 2. Zero-Order Correlation Matrix (UNODC homicide data) Homicide rate (log) Police legitimacy Democracy GDP/capita (log) Income inequality Ethnic diversity Young male pop Homicide rate (log) 1.0000 Police legitimacy −0.5029 1.0000 Democracy −0.1247 −0.0810 1.0000 GDP/capita (log) −0.4774 0.2974 0.4972 1.0000 Income inequality 0.6138 −0.2047 −0.2026 −0.5457 1.0000 Ethnic diversity 0.5187 −0.3019 −0.1159 −0.3927 0.4921 1.0000 Young male pop 0.5493 −0.2262 −0.5176 −0.8036 0.6454 0.4575 1.0000 Homicide rate (log) Police legitimacy Democracy GDP/capita (log) Income inequality Ethnic diversity Young male pop Homicide rate (log) 1.0000 Police legitimacy −0.5029 1.0000 Democracy −0.1247 −0.0810 1.0000 GDP/capita (log) −0.4774 0.2974 0.4972 1.0000 Income inequality 0.6138 −0.2047 −0.2026 −0.5457 1.0000 Ethnic diversity 0.5187 −0.3019 −0.1159 −0.3927 0.4921 1.0000 Young male pop 0.5493 −0.2262 −0.5176 −0.8036 0.6454 0.4575 1.0000 (obs = 171) Table 2. Zero-Order Correlation Matrix (UNODC homicide data) Homicide rate (log) Police legitimacy Democracy GDP/capita (log) Income inequality Ethnic diversity Young male pop Homicide rate (log) 1.0000 Police legitimacy −0.5029 1.0000 Democracy −0.1247 −0.0810 1.0000 GDP/capita (log) −0.4774 0.2974 0.4972 1.0000 Income inequality 0.6138 −0.2047 −0.2026 −0.5457 1.0000 Ethnic diversity 0.5187 −0.3019 −0.1159 −0.3927 0.4921 1.0000 Young male pop 0.5493 −0.2262 −0.5176 −0.8036 0.6454 0.4575 1.0000 Homicide rate (log) Police legitimacy Democracy GDP/capita (log) Income inequality Ethnic diversity Young male pop Homicide rate (log) 1.0000 Police legitimacy −0.5029 1.0000 Democracy −0.1247 −0.0810 1.0000 GDP/capita (log) −0.4774 0.2974 0.4972 1.0000 Income inequality 0.6138 −0.2047 −0.2026 −0.5457 1.0000 Ethnic diversity 0.5187 −0.3019 −0.1159 −0.3927 0.4921 1.0000 Young male pop 0.5493 −0.2262 −0.5176 −0.8036 0.6454 0.4575 1.0000 (obs = 171) Multicollinearity between the independent variables is tolerable—all variance inflation factor (VIF) scores are 4.3 or less. However, GDP per capita (log) and the proportion of the population that are young males are the only two variables with VIF scores above three, and table 2 indicates that these two variables have the strongest bivariate correlation in the matrix (with a correlation coefficient of −0.80). To test for potential bias as a result of this strong correlation, I ran the models below excluding the young male population variable, then again excluding the GDP per capita (log) variable (results not shown). Their exclusion do not substantially change the results of the effect of police legitimacy, and as such both variables are included in the models. Results Following convention (see Babones 2014), I report the statistical significance of the results of the OLS pooled regression models of the homicide rate. Nonetheless, as the country samples are non-random (i.e., based on data availability) and represent between 83 percent and 89 percent of the world’s total population, the direction and magnitude of each variable’s effect provides valuable information about the sample independent of its statistical significance. Table 3 lists the results of the models that analyze the effect of police legitimacy on the homicide rate. All models that include the police legitimacy variable show a strong and negative regression coefficient, implying that countries with higher levels of police legitimacy tend to have lower homicide rates. Models 1 through 3 use the UNODC homicide data. Model 1 shows that, of the control variables, only income inequality and ethnic diversity achieve statistical significance (at a threshold of p < 0.05). However, with the exception of democracy, the effects of the remaining control variables are in the expected direction. Model 2 indicates that police legitimacy has a statistically significant, negative relationship with the homicide rate. As the homicide rates are logged, the results of model 2 suggest that for every one-unit increase observed in police legitimacy, the homicide rate decreases by approximately 65 percent on average.14 Even with a difference of slightly less than two units between the highest and lowest average police legitimacy scores, the findings of this model suggest that, cross-nationally and net of controls, police legitimacy has a substantial effect on the homicide rate. In model 3, which includes the Global Barometer data, the effect of police legitimacy is similar, notwithstanding the substantially higher number of observations. Table 3. OLS Pooled Regressions of the Homicide Rate (UNODC and WHO homicide data logged) UNODC homicide data WHO homicide data Regressors Model 1 Model 2 Model 3 Model 4 Model 5 Model 6 Model 7 Police legitimacy −1.057*** −1.122*** −0.720*** −0.716*** −0.642*** (0.246) (0.234) (0.258) (0.265) (0.240) Democracy 0.028 0.007 0.016 0.022 0.010 0.010 0.019 (0.032) (0.029) (0.032) (0.027) (0.026) (0.030) (0.026) GDP/capita (log) −0.105 0.048 0.086 −0.471*** −0.352*** −0.239* −0.276* (0.174) (0.159) (0.156) (0.117) (0.126) (0.141) (0.143) Income inequality 0.045*** 0.047*** 0.045** 0.044*** 0.043*** 0.043*** 0.027* (0.017) (0.016) (0.019) (0.014) (0.014) (0.016) (0.015) Ethnic diversity 1.149** 0.799* −0.099 1.198*** 1.050** 1.029** 0.493 (0.497) (0.447) (0.592) (0.447) (0.432) (0.441) (0.512) Young male pop 14.753 14.600* 35.101*** 0.569 2.525 11.176 26.669*** (9.268) (8.410) (10.298) (7.400) (7.032) (8.356) (9.860) Constant −1.339 0.143 −1.292 3.618** 4.361*** 2.538 2.244 (2.297) (1.988) (1.907) (1.509) (1.465) (1.672) (1.878) Observations 171 171 375 249 249 171 486 No. of countries 82 82 91 88 88 82 100 R-squared 0.474 0.569 0.633 0.525 0.566 0.597 0.548 UNODC homicide data WHO homicide data Regressors Model 1 Model 2 Model 3 Model 4 Model 5 Model 6 Model 7 Police legitimacy −1.057*** −1.122*** −0.720*** −0.716*** −0.642*** (0.246) (0.234) (0.258) (0.265) (0.240) Democracy 0.028 0.007 0.016 0.022 0.010 0.010 0.019 (0.032) (0.029) (0.032) (0.027) (0.026) (0.030) (0.026) GDP/capita (log) −0.105 0.048 0.086 −0.471*** −0.352*** −0.239* −0.276* (0.174) (0.159) (0.156) (0.117) (0.126) (0.141) (0.143) Income inequality 0.045*** 0.047*** 0.045** 0.044*** 0.043*** 0.043*** 0.027* (0.017) (0.016) (0.019) (0.014) (0.014) (0.016) (0.015) Ethnic diversity 1.149** 0.799* −0.099 1.198*** 1.050** 1.029** 0.493 (0.497) (0.447) (0.592) (0.447) (0.432) (0.441) (0.512) Young male pop 14.753 14.600* 35.101*** 0.569 2.525 11.176 26.669*** (9.268) (8.410) (10.298) (7.400) (7.032) (8.356) (9.860) Constant −1.339 0.143 −1.292 3.618** 4.361*** 2.538 2.244 (2.297) (1.988) (1.907) (1.509) (1.465) (1.672) (1.878) Observations 171 171 375 249 249 171 486 No. of countries 82 82 91 88 88 82 100 R-squared 0.474 0.569 0.633 0.525 0.566 0.597 0.548 Robust standard errors in parentheses. *** p < 0.01 ** p < 0.05 * p < 0.1 Table 3. OLS Pooled Regressions of the Homicide Rate (UNODC and WHO homicide data logged) UNODC homicide data WHO homicide data Regressors Model 1 Model 2 Model 3 Model 4 Model 5 Model 6 Model 7 Police legitimacy −1.057*** −1.122*** −0.720*** −0.716*** −0.642*** (0.246) (0.234) (0.258) (0.265) (0.240) Democracy 0.028 0.007 0.016 0.022 0.010 0.010 0.019 (0.032) (0.029) (0.032) (0.027) (0.026) (0.030) (0.026) GDP/capita (log) −0.105 0.048 0.086 −0.471*** −0.352*** −0.239* −0.276* (0.174) (0.159) (0.156) (0.117) (0.126) (0.141) (0.143) Income inequality 0.045*** 0.047*** 0.045** 0.044*** 0.043*** 0.043*** 0.027* (0.017) (0.016) (0.019) (0.014) (0.014) (0.016) (0.015) Ethnic diversity 1.149** 0.799* −0.099 1.198*** 1.050** 1.029** 0.493 (0.497) (0.447) (0.592) (0.447) (0.432) (0.441) (0.512) Young male pop 14.753 14.600* 35.101*** 0.569 2.525 11.176 26.669*** (9.268) (8.410) (10.298) (7.400) (7.032) (8.356) (9.860) Constant −1.339 0.143 −1.292 3.618** 4.361*** 2.538 2.244 (2.297) (1.988) (1.907) (1.509) (1.465) (1.672) (1.878) Observations 171 171 375 249 249 171 486 No. of countries 82 82 91 88 88 82 100 R-squared 0.474 0.569 0.633 0.525 0.566 0.597 0.548 UNODC homicide data WHO homicide data Regressors Model 1 Model 2 Model 3 Model 4 Model 5 Model 6 Model 7 Police legitimacy −1.057*** −1.122*** −0.720*** −0.716*** −0.642*** (0.246) (0.234) (0.258) (0.265) (0.240) Democracy 0.028 0.007 0.016 0.022 0.010 0.010 0.019 (0.032) (0.029) (0.032) (0.027) (0.026) (0.030) (0.026) GDP/capita (log) −0.105 0.048 0.086 −0.471*** −0.352*** −0.239* −0.276* (0.174) (0.159) (0.156) (0.117) (0.126) (0.141) (0.143) Income inequality 0.045*** 0.047*** 0.045** 0.044*** 0.043*** 0.043*** 0.027* (0.017) (0.016) (0.019) (0.014) (0.014) (0.016) (0.015) Ethnic diversity 1.149** 0.799* −0.099 1.198*** 1.050** 1.029** 0.493 (0.497) (0.447) (0.592) (0.447) (0.432) (0.441) (0.512) Young male pop 14.753 14.600* 35.101*** 0.569 2.525 11.176 26.669*** (9.268) (8.410) (10.298) (7.400) (7.032) (8.356) (9.860) Constant −1.339 0.143 −1.292 3.618** 4.361*** 2.538 2.244 (2.297) (1.988) (1.907) (1.509) (1.465) (1.672) (1.878) Observations 171 171 375 249 249 171 486 No. of countries 82 82 91 88 88 82 100 R-squared 0.474 0.569 0.633 0.525 0.566 0.597 0.548 Robust standard errors in parentheses. *** p < 0.01 ** p < 0.05 * p < 0.1 Models 4 through 7 of table 3 present the results using the alternate source of homicide statistics—the WHO homicide data. Comparing model 4 to model 1, the effect of GDP per capita (log) is larger and achieves statistical significance. This discrepancy may be explained by the potential bias inherent in the WHO homicide data outlined above—namely that, among other sources, it draws upon gross national income data in formulating homicide estimates in countries with incomplete vital registration data, thereby potentially artificially inflating its statistical impact on the homicide rate. The effect of police legitimacy in model 5 is comparable to its effect in model 2, with the results indicating a statistically significant negative relationship with the WHO homicide rate. According to the model, with each unit increase observed in police legitimacy, the homicide rate decreases, on average, by roughly half.15 To more accurately compare the results of both homicide data sources, model 6 draws upon the WHO homicide data, while limiting the number of observations to those in the UNODC data models. This does not substantively change the effect of police legitimacy as compared to model 5. Likewise, model 7 indicates that the effect of police legitimacy also does not substantively change when the additional observations based on the Global Barometer data are included. To examine the relative importance of the effect of police legitimacy, table 4 lists the partial and the squared semi-partial correlation coefficients for each variable (using the UNODC homicide data). The partial correlation coefficient measures the correlation of a variable with the homicide rate after controlling for the effect of all the other independent variables. The results suggest that, unlike the zero-order correlations presented in table 2, police legitimacy has the strongest correlation with the homicide rate after controls. Furthermore, an examination of the squared semi-partial correlation coefficients reveals that the incremental effect of police legitimacy is the largest of all the independent variables in explaining the variation in the cross-national homicide rate.16 This suggests that the zero-order correlation between police legitimacy and the homicide rate not only holds, but becomes more pronounced with the presence of controls, and that, of all the variables included in the models, police legitimacy is the most consequential in accounting for the variation in homicide rates across countries. Table 4. Correlations with UNODC Homicide Rate (logged) Variable Partial Semi-partial2 Police legitimacy −0.4243 0.0947*** Democracy 0.0378 0.0006 GDP/capita (log) 0.0354 0.0005 Income inequality 0.3578 0.0633*** Ethnic diversity 0.2001 0.0180** Young male pop 0.1695 0.0128** Variable Partial Semi-partial2 Police legitimacy −0.4243 0.0947*** Democracy 0.0378 0.0006 GDP/capita (log) 0.0354 0.0005 Income inequality 0.3578 0.0633*** Ethnic diversity 0.2001 0.0180** Young male pop 0.1695 0.0128** (obs = 171) *** p < 0.01 ** p < 0.05 * p < 0.1 Table 4. Correlations with UNODC Homicide Rate (logged) Variable Partial Semi-partial2 Police legitimacy −0.4243 0.0947*** Democracy 0.0378 0.0006 GDP/capita (log) 0.0354 0.0005 Income inequality 0.3578 0.0633*** Ethnic diversity 0.2001 0.0180** Young male pop 0.1695 0.0128** Variable Partial Semi-partial2 Police legitimacy −0.4243 0.0947*** Democracy 0.0378 0.0006 GDP/capita (log) 0.0354 0.0005 Income inequality 0.3578 0.0633*** Ethnic diversity 0.2001 0.0180** Young male pop 0.1695 0.0128** (obs = 171) *** p < 0.01 ** p < 0.05 * p < 0.1 As past studies have mostly focused on Western countries, models are run separately for high-income and low- and middle-income countries, as well as democratic and non-democratic countries, to examine the relative importance of police legitimacy in non-Western countries (see table 5).17 Although a comparison of models 1 and 2 suggests that police legitimacy has a negative effect on homicide across different categories of economic development, the effect size of police legitimacy is greater in non-high-income than in high-income countries (whose effect does not achieve statistical significance).18 Interestingly, these findings suggest that the police legitimacy-homicide relationship is more pronounced in middle- and low-income countries. Concerning the political context, police legitimacy has a significantly negative effect on homicide across both political categories, though the effect of police legitimacy is slightly stronger in non-democracies (model 4) as compared to democracies (model 3).19 Finally, given previous research on disadvantaged communities, models 5 and 6 use the Gini coefficient as a proxy of the proportion of the total population considered marginalized. The populations of countries with greater levels of inequality are assumed to contain a higher proportion of marginalized people. The results suggest that the effect of police legitimacy on homicide is more pronounced (and only achieves statistical significance) in countries with higher levels of income inequality. Taken together, the findings indicate that the negative association found between police legitimacy and the homicide rate is robust across political contexts, with a stronger relationship in low- and middle-income (i.e., primarily non-Western) countries and in countries where a higher proportion of the total population are marginalized. Table 5. OLS Pooled Regressions of the Homicide Rate (UNODC homicide data logged) High-income countries Non-high-income countries Democracies Non-democracies Low-inequality countries High-inequality countries Regressors Model 1 Model 2 Model 3 Model 4 Model 5 Model 6 Police legitimacy −0.467 −0.975*** −0.657** −1.099*** −0.131 −0.885*** (0.506) (0.272) (0.307) (0.291) (0.515) (0.291) Democracy −0.003 0.027 −0.185** 0.012 −0.055*** 0.065** (0.041) (0.031) (0.083) (0.039) (0.017) (0.030) GDP/capita (log) −0.576 0.185 0.164 −0.221 −0.472* 0.128 (0.463) (0.190) (0.194) (0.304) (0.254) (0.191) Income inequality 0.026 0.055** 0.072*** −0.021 −0.013 0.063** (0.021) (0.022) (0.013) (0.025) (0.026) (0.024) Ethnic diversity 1.570*** 0.177 0.360 0.608 0.891* 0.042 (0.571) (0.603) (0.450) (0.829) (0.455) (0.570) Young male pop 9.672 9.174 13.468 4.723 −3.913 23.138* (12.818) (10.896) (9.171) (14.175) (6.894) (12.080) Constant 5.803 −0.819 −0.883 6.146 6.413*** −2.384 (4.705) (2.482) (2.107) (3.929) (2.147) (2.555) Observations 97 74 136 35 95 76 No. of countries 45 45 61 24 41 46 R-squared 0.510 0.455 0.698 0.378 0.547 0.508 High-income countries Non-high-income countries Democracies Non-democracies Low-inequality countries High-inequality countries Regressors Model 1 Model 2 Model 3 Model 4 Model 5 Model 6 Police legitimacy −0.467 −0.975*** −0.657** −1.099*** −0.131 −0.885*** (0.506) (0.272) (0.307) (0.291) (0.515) (0.291) Democracy −0.003 0.027 −0.185** 0.012 −0.055*** 0.065** (0.041) (0.031) (0.083) (0.039) (0.017) (0.030) GDP/capita (log) −0.576 0.185 0.164 −0.221 −0.472* 0.128 (0.463) (0.190) (0.194) (0.304) (0.254) (0.191) Income inequality 0.026 0.055** 0.072*** −0.021 −0.013 0.063** (0.021) (0.022) (0.013) (0.025) (0.026) (0.024) Ethnic diversity 1.570*** 0.177 0.360 0.608 0.891* 0.042 (0.571) (0.603) (0.450) (0.829) (0.455) (0.570) Young male pop 9.672 9.174 13.468 4.723 −3.913 23.138* (12.818) (10.896) (9.171) (14.175) (6.894) (12.080) Constant 5.803 −0.819 −0.883 6.146 6.413*** −2.384 (4.705) (2.482) (2.107) (3.929) (2.147) (2.555) Observations 97 74 136 35 95 76 No. of countries 45 45 61 24 41 46 R-squared 0.510 0.455 0.698 0.378 0.547 0.508 Robust standard errors in parentheses. *** p < 0.01 ** p < 0.05 * p < 0.1 Table 5. OLS Pooled Regressions of the Homicide Rate (UNODC homicide data logged) High-income countries Non-high-income countries Democracies Non-democracies Low-inequality countries High-inequality countries Regressors Model 1 Model 2 Model 3 Model 4 Model 5 Model 6 Police legitimacy −0.467 −0.975*** −0.657** −1.099*** −0.131 −0.885*** (0.506) (0.272) (0.307) (0.291) (0.515) (0.291) Democracy −0.003 0.027 −0.185** 0.012 −0.055*** 0.065** (0.041) (0.031) (0.083) (0.039) (0.017) (0.030) GDP/capita (log) −0.576 0.185 0.164 −0.221 −0.472* 0.128 (0.463) (0.190) (0.194) (0.304) (0.254) (0.191) Income inequality 0.026 0.055** 0.072*** −0.021 −0.013 0.063** (0.021) (0.022) (0.013) (0.025) (0.026) (0.024) Ethnic diversity 1.570*** 0.177 0.360 0.608 0.891* 0.042 (0.571) (0.603) (0.450) (0.829) (0.455) (0.570) Young male pop 9.672 9.174 13.468 4.723 −3.913 23.138* (12.818) (10.896) (9.171) (14.175) (6.894) (12.080) Constant 5.803 −0.819 −0.883 6.146 6.413*** −2.384 (4.705) (2.482) (2.107) (3.929) (2.147) (2.555) Observations 97 74 136 35 95 76 No. of countries 45 45 61 24 41 46 R-squared 0.510 0.455 0.698 0.378 0.547 0.508 High-income countries Non-high-income countries Democracies Non-democracies Low-inequality countries High-inequality countries Regressors Model 1 Model 2 Model 3 Model 4 Model 5 Model 6 Police legitimacy −0.467 −0.975*** −0.657** −1.099*** −0.131 −0.885*** (0.506) (0.272) (0.307) (0.291) (0.515) (0.291) Democracy −0.003 0.027 −0.185** 0.012 −0.055*** 0.065** (0.041) (0.031) (0.083) (0.039) (0.017) (0.030) GDP/capita (log) −0.576 0.185 0.164 −0.221 −0.472* 0.128 (0.463) (0.190) (0.194) (0.304) (0.254) (0.191) Income inequality 0.026 0.055** 0.072*** −0.021 −0.013 0.063** (0.021) (0.022) (0.013) (0.025) (0.026) (0.024) Ethnic diversity 1.570*** 0.177 0.360 0.608 0.891* 0.042 (0.571) (0.603) (0.450) (0.829) (0.455) (0.570) Young male pop 9.672 9.174 13.468 4.723 −3.913 23.138* (12.818) (10.896) (9.171) (14.175) (6.894) (12.080) Constant 5.803 −0.819 −0.883 6.146 6.413*** −2.384 (4.705) (2.482) (2.107) (3.929) (2.147) (2.555) Observations 97 74 136 35 95 76 No. of countries 45 45 61 24 41 46 R-squared 0.510 0.455 0.698 0.378 0.547 0.508 Robust standard errors in parentheses. *** p < 0.01 ** p < 0.05 * p < 0.1 Although the findings suggest a strong association between police legitimacy and homicide, they do not provide insight into the causal order of the association. However, given the data structure (particularly that the variables are slow to change and there is minimal within-country variation in the dataset) and in the absence of a suitable instrumental variable, the task of assessing reverse or simultaneous causality is difficult. Unsurprisingly, models using time lags of varying durations and those that invert the independent and dependent variables (not shown) are inconclusive in this regard. Therefore, the possibility of reverse or simultaneous causation cannot definitively be ruled out. Discussion The quantitative analyses reveal that the public’s perceived level of police legitimacy is related to homicide rates at the country level. They provide strong evidence of a significant and negative relationship between police legitimacy and homicide, suggesting that countries with a higher level of police legitimacy tend to have much lower homicide rates controlling for various economic, political, and demographic variables. This relationship is robust across homicide data sources (i.e., both criminal justice and public health sources) and the number of countries and observations included in the models. The statistical impact of police legitimacy on the homicide rate is considerable, surpassing that of the other variables modeled, including income inequality, which is considered by some as the most important cross-national determinant of homicide (see Trent and Pridemore 2012). This study also provides evidence that the negative relationship between police legitimacy and lethal violence is generalizable across a wide range of contexts. The results suggest that previous findings of a negative relationship between police legitimacy and homicide in certain disadvantaged American urban neighborhoods are applicable on a much broader scale—that is, the association holds using country-level data and nationally representative samples. Of course, as the data are aggregated, they do not permit a test of whether this association is consistent across the entire population or whether it is more acute in specific population subsections, which is a task for future research using disaggregated data. Conversely, the data do permit an assessment of the police legitimacy–homicide relationship across various country groupings. The finding that the relationship is stronger in countries with relatively higher levels of inequality points to the possibility that marginalized communities are important drivers of this association. Moreover, the strong association between police legitimacy and homicide in low- and middle-income countries implies that there is a potentially fruitful opportunity to further investigate this relationship in non-Western countries, particularly as this group of countries has greater variation in average levels of police legitimacy. Another contribution of this study is the empirical investigation of the relationship between the aggregate level of police legitimacy and homicide rates at the country level. As such, the analysis integrates micro- and macro-level research by testing the applicability of (the latter half of) the procedural justice theory causal mechanism at the macro level. As mentioned, micro-level studies have identified the importance of police legitimacy in influencing attitudes toward the use of violence, rather than actual violent behavior. As it is possible, as Tankebe (2009) suggests, that attitudes toward violence may not be perfectly correlated with violent behavior, the present study addresses this limitation by employing homicide levels, rather than attitudes toward the use of violence, as the dependent variable. Accordingly, it fills an important gap in the procedural justice literature by providing strong empirical support for the claim that, beyond more favorable attitudes toward the use of violence, lower levels of police legitimacy are related to higher rates of homicidal violence. This study also differs from past procedural justice research in that it assesses the public’s overall perception of police legitimacy (by examining the average level of police legitimacy within a given society), rather than the level of the perceived legitimacy of the police held by the perpetrators of acts of violence. This aligns with the causal mechanisms outlined above that underscore the importance of social norms, not only in shaping the individual’s perception of the police through socialization, but also in exerting social pressure to potentially respond to certain situations with violence. The finding that societal levels of police legitimacy have a robust negative relationship with homicide rates therefore makes a significant contribution to the literature. Furthermore, it opens the door to future social-psychological research examining whether average levels of police legitimacy within a given society have an effect on the homicide rate independent of the views of police legitimacy held by those who commit murder. Although the statistical findings provide evidence that police legitimacy is related to homicide, they provide limited insight into the causal mechanisms driving this relationship. The results indirectly support the causal mechanisms described above—namely, that low police legitimacy results in an increased willingness to resort to violence to resolve disputes through weakening pro-state informal social control mechanisms and/or strengthening informal social control mechanisms that supplant the rule of law. While these mechanisms are not mutually exclusive, further research is needed to determine their broader relevance and whether other causal mechanisms are involved. Moreover, the results do not rule out the possibility of reverse or simultaneous causation. Although there is a strong theoretical case for the causal order presented in this paper, it is nonetheless essential to empirically assess the causal direction. However, given the current data limitations, until such time as more longitudinal data become available or a suitable instrumental variable is devised, attempts to empirically determine the causal order will remain challenging. In the meantime, future research examining the effect of the homicide rate on police legitimacy in a model including all known causes of the latter as controls could prove insightful. Assuming that the causal direction flows from police legitimacy to homicide, an important implication of these findings is that increasing police legitimacy, particularly in countries where it is comparatively low, could potentially go a long way in reducing homicide rates. This implies that police forces would be well advised to engage in policing practices that increase their perceived legitimacy among the citizenry and, by the same token, to stop engaging in practices that decrease their perceived legitimacy. For example, the recent Department of Justice reports of both the Ferguson and Baltimore Police Departments (US Department of Justice 2016; 2015) cite egregious instances of systematic “stop-and-search” practices, which have been shown to target visible minorities, lead to few charges, and undermine the legitimacy of the police in both the United States and Canada (Meng, Giwa, and Anucha 2015; Tyler, Goff, and MacCoun 2015). The findings of this study provide renewed reason to end these policing practices. Additionally, countries—particularly low- and middle-income countries—with much lower police legitimacy scores than either Canada or the United States may also have problems such as widespread corruption of the police and/or underfunded police forces, leading to weakened police capacity and possibly police force disengagement, which potentially contribute to higher homicide rates via the causal mechanisms described above. Although police corruption and weak police forces are not new issues, the present analysis underscores their importance in suggesting that, if left unresolved, these problems may contribute to increased levels of lethal violence. Finally, the results support recent empirical research that maintains that, in addition to socio-economic and political variables, cultural factors such as social values and attitudes are an important determinant of cross-national homicide rates (Lappi-Seppälä and Leht 2014). This study not only contributes to the growing body of empirical evidence corroborating this perspective, but identifies a specific social attitude—the public’s overall disposition toward the legitimacy of law enforcement—as being closely linked to the homicide rate across countries. These findings imply that further research investigating the effects of particular social norms and attitudes on the prevalence of violence is a potentially promising avenue of social inquiry. Footnotes 1 The correlation coefficient between the homicide rate and its scores lagged five years is 0.96. For police legitimacy, this correlation coefficient is 0.93. 2 Although random effects models can be estimated without dropping cases and the results are substantively similar to those presented below, they are almost entirely driven by between-country variation. 3 Alternatively, it is possible to run feasible generalized least squares models adjusting for panel heteroskedasticity and autocorrelation as a robustness check. The results (not shown) are substantively similar to those presented below. However, as these models are forced as a result of different time interval lengths between observations, and because panel heteroskedasticity is likely to be a minor concern given the low number of observations per country (see Babones 2014), I present the pooled OLS models. 4 This is performed using the “cluster” option in STATA. As a robustness check for autocorrelation, I rerun the models using Prais-Winsten estimation (not shown), which produces nearly identical results to those shown below. 5 Survey question wording: “I am going to name a number of organizations. For each one, could you tell me how much confidence you have in them: is it a great deal of confidence, quite a lot of confidence, not very much confidence or none at all?: The police” (WVS 2015). 6 Note that I have reversed the order of the original WVS/EVS response categories in order to make the interpretation of the results more intuitive. All other answer categories (“Not applicable,” “No answer,” “Don’t know,” “Missing,” and “Not asked in survey”) are excluded from the analysis. 7 With response options such as “no,” “little,” “some,” and “a lot of confidence,” or “no trust,” trust to “a limited extent,” “a medium extent,” and “a great extent.” 8 High-income countries are defined as those with a GDP per capita of $12,000 or greater. Countries with high levels of inequality are defined as those with a Gini coefficient greater than 35. Democracies are defined by countries with a Polity 2 score of 6 or greater (see Centre for Systematic Peace 2004). 9 Turkey’s police legitimacy scores are taken prior to the Erdogan presidency. 10 The WHO estimation procedure draws upon variants of three variables included in the present statistical analysis (the Gini coefficient, the proportion of the total population who are males aged 15 to 24, and gross domestic product per capita) (WHO 2014b). Consequently, the coefficients of these three control variables will likely be biased (i.e., increasing the likelihood of Type 1 error—i.e., statistical false positives) in the models using the WHO homicide data presented below. However, the coefficient of the focal independent variable, police legitimacy, should not be biased by this estimation technique. 11 While the WHO homicide data are technically not longitudinally comparable due to estimation differences across years, as the focus of the present analysis is primarily on between-country variation, I contend that their use as an alternate data source to assess the robustness of the results is defensible. 12 As Gini coefficients are missing for some years, where possible in the WVS/EVS models, I replace a country’s missing year with its most recent available score (which are a maximum of nine years apart), resulting in an additional 14 to 18 observations. As the models with and without these additional Gini scores produce nearly identical results, the former are presented to maximize the number of observations. 13 As the data are only available at five-year intervals, I perform linear interpolation and extrapolation (using the “ipolate” command in STATA) for the missing values. 14 The proportional change in the homicide rate is 0.35 (i.e., a 65 percent decrease) for every one-unit increase in police legitimacy. This is calculated by taking the exponentiated value of the coefficient of police legitimacy in model 2 (−1.057). 15 The proportional change in the homicide rate is 0.49 (i.e., a 51 percent decrease) for every one-unit increase in police legitimacy in model 5. 16 The squared semi-partial correlation coefficient indicates the marginal increase in proportion of the explained variance (R2) in the models by adding each independent variable to a model composed of all other variables. 17 These results are based on the UNODC homicide data. 18 A one-unit increase in police legitimacy results, on average, in a 37 percent decrease in the homicide rate for high-income countries (model 1) and a 62 percent decrease for non-high-income countries (model 2). 19 In democracies the homicide rate decreases on average 48 percent with each unit increase in police legitimacy, while in non-democratic countries the homicide rate decreases on average 67 percent for every unit increase in police legitimacy. About the Author Andrew Dawson is an Associate Professor of Sociology at York University, Glendon Campus. His recent research focuses on the relationship between state legitimacy and violence, and his latest publications have appeared in The Sociological Quarterly, Social Science History, and World Development. References Afrobarometer . 2017 . “Afrobarometer Data.” Afrobarometer, accessed January 19 2017. http://www.afrobarometer.org/. Anderson , Elijah . 1999 . Code of the Street: Decency, Violence, and the Moral Life of the Inner City , 1st ed . New York : W. W Norton . Arab Barometer . 2017 . “Arab Barometer Data.” Arab Barometer, accessed March 10 2017. http://www.arabbarometer.org. Asian Barometer . 2017 . “Asian Barometer Data.” Asian Barometer (sent via email request), accessed March 15 2017. http://www.asianbarometer.org/. Avison , William , and Pamela Loring . 1986 . “ Population Diversity and Cross-National Homicide: The Effects of Inequality and Heterogeneity .” Criminology 24 ( 4 ): 733 – 49 . Google Scholar CrossRef Search ADS Babones , Salvatore J. 2014 . Methods for Quantitative Macro-Comparative Research . Thousand Oaks, CA : SAGE . Google Scholar CrossRef Search ADS Baumer , Eric , Steven Messner , and Richard Rosenfeld . 2003 . “ Explaining Spatial Variation in Support for Capital Punishment: A Multi-Level Analysis .” American Journal of Sociology 108 ( 4 ): 844 – 75 . Google Scholar CrossRef Search ADS Beetham , David . [ 1991 ] 2013. The Legitimation of Power , 2nd ed. , Political Analysis. New York : Palgrave Macmillan . Google Scholar CrossRef Search ADS Black , Donald . 1983 . “ Crime as Social Control .” American Sociological Review 48 ( 1 ): 34 – 45 . Google Scholar CrossRef Search ADS Blumer , Herbert . 1971 . “ Social Problems as Collective Behavior .” Social Problems 18 : 298 – 306 . Google Scholar CrossRef Search ADS Bottoms , Anthony , and Justice Tankebe . 2012 . “ Beyond Procedural Justice: A Dialogic Approach to Legitimacy in Criminal Justice .” Journal of Criminal Law and Criminology 102 ( 1 ): 119 – 70 . Centre for Systematic Peace . 2004 . “Polity IV Annual Time-Series, 1800–2013.” Accessed June 6 2014. http://www.systemicpeace.org/inscrdata.html. Dawson , Andrew . 2013 . “ The Social Determinants of the Rule of Law: A Comparison of Jamaica and Barbados .” World Development 45 : 314 – 24 . Google Scholar CrossRef Search ADS ——— . 2017 . “ The Belief in State Legitimacy and Homicide: A Cross-National Analysis .” The Sociological Quarterly 58 ( 4 ): 552 – 75 . Google Scholar CrossRef Search ADS Desmond , Matthew , Andrew V. Papachristos , and David S. Kirk . 2016 . “ Police Violence and Citizen Crime Reporting in the Black Community .” American Sociological Review 81 ( 5 ): 857 – 76 . Google Scholar CrossRef Search ADS The Economist . 2015 . “Midsummer Murder.” August 15, pp. 19–20. Eisner , Manuel . 2001 . “ Modernization, Self-Control and Lethal Violence: The Long-Term Dynamics of European Homicide Rates in Theoretical Perspective .” British Journal of Criminology 41 ( 4 ): 618 – 48 . Google Scholar CrossRef Search ADS ——— . 2003 . “ Long-Term Historical Trends in Violent Crime .” Crime and Justice; A Review of Research 30 : 83 – 142 . Google Scholar CrossRef Search ADS ——— . 2013 . “What Causes Large-Scale Variation in Homicide Rates.” In Aggression in Humans and Other Primates: Biology, Psychology, Sociology , edited by Hans-Henning Kortüm and Jürgen Heinze , 137 – 61 . Berlin and Boston : de Gruyter . Eisner , Manuel , and Amy Nivette . 2013 . “Does Low Legitimacy Cause Crime? A Review of the Evidence.” In Legitimacy and Criminal Justice: An International Exploration , edited by Justice Tankebe and Alison Liebling , 308 – 25 . Oxford : Oxford University Press . Google Scholar CrossRef Search ADS EVS . 2015 . “European Values Study, 1981–2008, Longitudinal Data File.” GESIS Data Archive, accessed April 2 2017. http://www.europeanvaluesstudy.eu/. Fearon , James D. 2003 . “ Ethnic and Cultural Diversity by Country .” Journal of Economic Growth 8 ( 2 ): 195 – 222 . Google Scholar CrossRef Search ADS Feenstra , Robert C. , Robert Inklaar , and Marcel P. Timmer . 2015 . “ The Next Generation of the Penn World Table .” American Economic Review 105 ( 10 ): 3150 – 82 . www.ggdc.net/pwt. Google Scholar CrossRef Search ADS Fiske , Alan Page , and Tage Shakti Rai . 2015 . Virtuous Violence: Hurting and Killing to Create, Sustain, End, and Honor Social Relationships . Cambridge : Cambridge University Press . Gau , Jacinta M. , Nicholas Corsaro , Eric A. Stewart , and Rod K. Brunson . 2012 . “ Examining Macro-Level Impacts on Procedural Justice and Police Legitimacy .” Journal of Criminal Justice 40 : 333 – 43 . Google Scholar CrossRef Search ADS Gray , Obika . 2003 . “Badness-Honour.” In Understanding Crime in Jamaica: New Challenges for Public Policy , edited by Anthony Harriott , 13 – 47 . Kingston, Jamaica : University of the West Indies Press . Gurr , Ted Robert . 1970 . Why Men Rebel . Princeton, NJ : Princeton University Press . Hall , John A. 1994 . Coercion and Consent: Studies on the Modern State . Cambridge : Polity Press . Hinsch , Wilfried . 2010 . “ Justice, Legitimacy, and Constitutional Rights .” Critical Review of International Social and Political Philosophy 13 ( 1 ): 39 – 54 . Google Scholar CrossRef Search ADS Hirschi , Travis , and Michael Gottfredson . 1983 . “ Age and the Explanation of Crime .” American Journal of Sociology 89 ( 3 ): 552 – 84 . Google Scholar CrossRef Search ADS Hobbes , Thomas . [ 1651 ] 1958. Leviathan: Parts I and II . New York : Macmillan Publishing Company . Hough , Mike , Jonathan Jackson , and Ben Bradford . 2013 . “Legitimacy, Trust, and Compliance: An Empirical Test of Procedural Justice Theory Using the European Social Survey.” In Legitimacy and Criminal Justice: An International Exploration , edited by Justice Tankebe and Alison Liebling . Oxford: Oxford University Press . Jackson , Jonathan , Ben Bradford , Mike Hough , Andy Myhill , Paul Quinton , and Tom R. Tyler . 2012 . “ Why Do People Comply with the Law? Legitimacy and the Influence of Legal Institutions .” British Journal of Criminology 52 ( 6 ): 1051 – 71 . Google Scholar CrossRef Search ADS Jackson , Jonathan , Aziz Huq , Ben Bradford , and Tom R. Tyler . 2013 . “ Monopolizing Force? Police Legitimacy and Public Attitudes towards the Acceptability of Violence .” Psychology, Public Policy, and Law 19 ( 4 ): 479 – 97 . Google Scholar CrossRef Search ADS Jang , Hyunseok , Hee-Jong Joo , and Jihong Zhao . 2010 . “ Determinants of Public Confidence in Police: An International Perspective .” Journal of Criminal Justice 38 : 57 – 68 . Google Scholar CrossRef Search ADS Johnson , Devon , Edward R. Maguire , and Joseph B. Kuhns . 2014 . “ Public Perceptions of the Legitimacy of the Law and Legal Authorities: Evidence from the Caribbean .” Law & Society Review 48 ( 4 ): 947 – 78 . Google Scholar CrossRef Search ADS Kane , Robert J. 2005 . “ Compromised Police Legitimacy as a Predictor of Violent Crime in Structurally Disadvantaged Communities .” Criminology 43 ( 2 ): 469 – 98 . Google Scholar CrossRef Search ADS Kubrin , Charis E. , and Ronald Weitzer . 2003 a. “ New Directions in Social Disorganization Theory .” Journal of Research in Crime and Delinquency 40 ( 4 ): 374 – 402 . Google Scholar CrossRef Search ADS ——— . 2003 b. “ Retaliatory Homicide: Concentrated Disadvantage and Neighbourhood Culture .” Social Problems 50 ( 2 ): 157 – 80 . Google Scholar CrossRef Search ADS LaFree , Gary . 1998 . Losing Legitimacy: Street Crime and the Decline of Social Institutions in America, Crime & Society . Boulder : Westview Press . ——— . 1999 . “A Summary Review of Cross-National Comparative Studies of Homicide.” In Homicide: A Sourcebook of Social Research , edited by M. Dwayne Smith and Margaret A. Zahn , 125 – 45 . London : Sage . Lappi-Seppälä , Tapio , and Martti Leht . 2014 . “ Cross-Comparative Perspectives on Global Homicide Trends .” Crime and Justice 43 ( 1 ): 135 – 230 . Google Scholar CrossRef Search ADS Latinobarómetro . 2017 . “Latinobarómetro Data.” Latinobarómetro Corporation, accessed February 20 2017. http://www.latinobarometro.org. Levi , Margaret , and Audrey Sacks . 2009 . “ Legitimating Beliefs: Sources and Indicators .” Regulation & Governance 3 ( 4 ): 311 – 13 . Google Scholar CrossRef Search ADS Levy , Margaret , Audrey Sacks , and Tom R. Tyler . 2009 . “ Conceptualizing Legitimacy, Measuring Legitimating Beliefs .” American Behavioral Scientist 53 ( 3 ): 354 – 75 . Google Scholar CrossRef Search ADS Lipset , Seymour Martin . 1963 . Political Man: The Social Bases of Politics . Anchor Books ed. Garden City, NY : Doubleday & Company . Lukes , Steven . 2005 . Power: A Radical View , 2nd ed . New York : Palgrave Macmillan . Marien , Sofie , and Marc Hooghe . 2011 . “ Does Political Trust Matter? An Empirical Investigation into the Relation between Political Trust and Support for Law Compliance .” European Journal of Political Research 50 : 267 – 91 . Google Scholar CrossRef Search ADS Meng , Yunliang , Sulaimon Giwa , and Uzo Anucha . 2015 . “ Is There Racial Discrimination in Police Stop-and-Searches of Black Youth? A Toronto Case Study .” Canadian Journal of Family and Youth/Le Journal Canadien de Famille et de la Jeunesse 7 ( 1 ): 115 – 48 . Merton , Robert K. 1938 . “ Social Structure and Anomie .” American Sociological Review 3 ( 5 ): 672 – 82 . Google Scholar CrossRef Search ADS Neapolitan , Jerome L. 1997 . Cross-National Crime: A Research Review and Sourcebook . Westport, CT : Greenwood Press . Neumayer , Eric . 2003 . “ Good Policy Can Lower Violent Crime: Evidence from a Cross-National Panel of Homicide Rates, 1980–97 .” Journal of Peace Research 40 ( 6 ): 619 – 40 . Google Scholar CrossRef Search ADS Nisbett , Richard E. , and Dov Cohen . 1996 . Culture of Honor: The Psychology of Violence in the South . New Directions in Social Psychology Boulder : Westview Press . Nivette , Amy . 2011 . “ Cross-National Predictors of Crime: A Meta-Analysis .” Homicide Studies 15 ( 2 ): 103 – 31 . Google Scholar CrossRef Search ADS ——— . 2014 . “ Legitimacy and Crime: Theorizing the Role of the State in Cross-National Criminological Theory .” Theoretical Criminology 18 ( 1 ): 93 – 111 . Google Scholar CrossRef Search ADS Nivette , Amy , and Manuel Eisner . 2013 . “ Do Legitimate Polities Have Fewer Homicides? A Cross-National Analysis .” Homicide Studies 17 ( 1 ): 3 – 26 . Google Scholar CrossRef Search ADS Papachristos , Andrew V. , Tracy L. Meares , and Jeffrey Fagan . 2012 . “ Why Do Criminals Obey the Law? The Influence of Legitimacy and Social Networks on Active Gun Offenders .” Journal of Criminal Law and Criminology 102 ( 2 ): 397 – 440 . Reisig , Michael D. 2009 . “ Procedural Justice, Police Legitimacy, and Helping the Police Fight Crime: Results from a Survey of Jamaican Adolescents .” Police Quarterly 12 ( 1 ): 42 – 62 . Google Scholar CrossRef Search ADS Reisig , Michael D. , and Roger B. Parks . 2000 . “ Experience, Quality of Life, and Neighborhood Context: A Hierarchical Analysis of Satisfaction with Police .” Justice Quarterly 17 : 607 – 30 . Google Scholar CrossRef Search ADS Roth , Randolph . 2009 . American Homicide . Cambridge, MA : Belknap Press of Harvard University Press . Schaible , Lonnie , and Lorine Hughes . 2011 . “ Crime, Shame, Reintegration, and Cross-National Homicide: A Partial Test of Reintegrative Shaming Theory .” The Sociological Quarterly 52 : 104 – 31 . Google Scholar CrossRef Search ADS “Shame.” 2003 . “ Shame, Stigma, and Crime: Evaluating the Efficacy of Shaming Sanctions in Criminal Law .” Harvard Law Review 116 ( 7 ): 2186 – 2207 . CrossRef Search ADS Smith , Delores , and Katherine Green . 2007 . “ Violence Among Youth in Jamaica: A Growing Public Health Risk and Challenge .” Pan American Journal of Public Health 22 ( 6 ): 417 – 24 . Google Scholar CrossRef Search ADS Solt , Frederick . 2016 . “ The Standardized World Income Inequality Database .” Social Science Quarterly 97 ( 5 ): 1267 – 81 . SWIID Version 5.1. Google Scholar CrossRef Search ADS Stamatel , Janet P. 2009 . “ Correlates of National-Level Homicide Variation in Post-Communist East-Central Europe .” Social Forces 87 : 1423 – 48 . Google Scholar CrossRef Search ADS Sunshine , Jason , and Tom R. Tyler . 2003 . “ The Role of Procedural Justice and Legitimacy in Shaping Public Support for Policing .” Law & Society Review 37 ( 3 ): 513 – 48 . Google Scholar CrossRef Search ADS Tankebe , Justice . 2009 . “ Self-Help, Policing and Procedural Justice: Ghanaian Vigilantism and the Rule of Law .” Law & Society Review 43 ( 2 ): 245 – 70 . Google Scholar CrossRef Search ADS ——— . 2013 . “ Viewing Things Differently: The Dimensions of Public Perceptions of Police Legitimacy .” Criminology 51 ( 1 ): 103 – 35 . Google Scholar CrossRef Search ADS Trent , Carol , and William Alex Pridemore . 2012 . “A Review of the Cross-National Empirical Literature on Social Structure and Homicide.” In Handbook of European Homicide Research: Patterns, Explanations, and Country Studies , edited by Marieke C A Liem and William Alex Pridemore , 111 – 35 . New York : Springer . Google Scholar CrossRef Search ADS Tyler , Tom R. [ 1990 ] 2006. Why People Obey the Law . Princeton, NJ : Princeton University Press . Tyler , Tom R. , Phillip Atiba Goff , and Robert J. MacCoun . 2015 . “ The Impact of Psychological Science on Policing in the United States: Procedural Justice, Legitimacy, and Effective Law Enforcement .” Psychological Science in the Public Interest 16 ( 3 ): 75 – 109 . Google Scholar CrossRef Search ADS Tyler , Tom R. , and Jonathan Jackson . 2014 . “ Popular Legitimacy and the Exercise of Legal Authority: Motivating Compliance, Cooperation, and Engagement .” Psychology, Public Policy, and Law 20 ( 1 ): 78 – 95 . Google Scholar CrossRef Search ADS United Nations . 2013 . “Population by Age Groups—Male.” United Nations Department of Economic and Social Affairs, Population Division, accessed June 13 2014. http://esa.un.org/unpd/wpp/Excel-Data/population.htm. United Nations Office on Drugs and Crime . 2014 . “Statistics—Historic Data,” accessed December 15 2014. https://www.unodc.org/unodc/en/data-and-analysis/statistics/historic-data.html. US Department of Justice . 2015 . “Investigation of the Ferguson Police Department.” US Department of Justice Civil Rights Division—March 4, 2015, accessed July 2 2016. https://www.justice.gov/sites/default/files/opa/press-releases/attachments/2015/03/04/ferguson_police_department_report.pdf. ——— . 2016 . “Investigation of the Baltimore City Police Department.” US Department of Justice Civil Rights Division—August 10, 2016, accessed August 16 2016. https://www.justice.gov/opa/file/883366/download. UNODC . 2013 . “Global Study on Homicide.” United Nations Office on Drugs and Crime, accessed June 22 2015. http://www.unodc.org/documents/gsh/pdfs/2014_GLOBAL_HOMICIDE_BOOK_web.pdf. ——— . 2015 . “UNODC Homicide Statistics.” United Nations Office on Drugs and Crime, accessed June 22 2015. https://data.unodc.org/. Weber , Max . [ 1922 ] 1978. Economy and Society: An Outline of Interpretive Sociology . Berkeley : University of California Press . WHO . 2011 . “Disease and Injury Country Estimates, 2004–2008.” World Health Organization, accessed June 18 2015. http://www.who.int/healthinfo/global_burden_disease/estimates_country_2004_2008/en/. ——— . 2014 a. “Cause-Specific Mortality Estimates 2000–2012.” World Health Organization, accessed June 18 2015. http://www.who.int/healthinfo/global_burden_disease/estimates/en/index1.html. ——— . 2014 b. “Estimating Global Homicide Deaths.” World Health Organization, accessed March 24 2016. http://www.who.int/violence_injury_prevention/violence/status_report/2014/report/Estimating_global_homicide_deaths.pdf. ——— . 2014 c. “WHO Methods and Data Sources for Country‐Level Causes of Death 2000–2012.” World Health Organization, Department of Health Statistics and Information Systems, accessed July 22 2015. http://www.who.int/healthinfo/statistics/GlobalCOD_method.pdf. WVS . 2015 . “World Values Survey Longitudinal 1981–2014 Data.” World Values Survey Association, accessed April 2 2017. http://www.worldvaluessurvey.org/. © The Author(s) 2018. Published by Oxford University Press on behalf of the University of North Carolina at Chapel Hill. All rights reserved. For permissions, please e-mail: journals.permissions@oup.com. This article is published and distributed under the terms of the Oxford University Press, Standard Journals Publication Model (https://academic.oup.com/journals/pages/about_us/legal/notices)

Journal

Social ForcesOxford University Press

Published: May 24, 2018

There are no references for this article.

You’re reading a free preview. Subscribe to read the entire article.


DeepDyve is your
personal research library

It’s your single place to instantly
discover and read the research
that matters to you.

Enjoy affordable access to
over 18 million articles from more than
15,000 peer-reviewed journals.

All for just $49/month

Explore the DeepDyve Library

Search

Query the DeepDyve database, plus search all of PubMed and Google Scholar seamlessly

Organize

Save any article or search result from DeepDyve, PubMed, and Google Scholar... all in one place.

Access

Get unlimited, online access to over 18 million full-text articles from more than 15,000 scientific journals.

Your journals are on DeepDyve

Read from thousands of the leading scholarly journals from SpringerNature, Elsevier, Wiley-Blackwell, Oxford University Press and more.

All the latest content is available, no embargo periods.

See the journals in your area

DeepDyve

Freelancer

DeepDyve

Pro

Price

FREE

$49/month
$360/year

Save searches from
Google Scholar,
PubMed

Create lists to
organize your research

Export lists, citations

Read DeepDyve articles

Abstract access only

Unlimited access to over
18 million full-text articles

Print

20 pages / month

PDF Discount

20% off