The police code of silence in a comparative perspective

Marko Prpić, MA University of Zagreb, Faculty of Education and Rehabilitation Sciences prpic.marcus@gmail.com Robert Peacock, PhD Florida international University, Department of Criminology and Criminal Justice rpeacock@fiu.edu THE POLICE CODE OF SILENCE IN A COMPARATIVE PERSPECTIVE By Marko Prpic University of Zagreb and Robert Peacock Michigan State University THE POLICE CODE OF SILENCE IN A COMPARATIVE PERSPECTIVE Abstract The purpose of this paper is to determine the degree to which the code of silence is a uniform phenomenon, spanning across different countries and police cultures. Samples of police officers from Croatia (1,142 respondents), South Africa (871 respondents), and the United States (664 respondents) evaluated 11 scenarios describing various forms of police misconduct. The results indicate that, in all 11 examined scenarios, there is a statistically significant correlation between unwillingness to report misconduct at work and the country of origin of the examined police officers. Own perception of misconduct and estimates of others’ willingness to report misconduct are significant predictors of the respondents’ own willingness to report misconduct at work. Introduction The code of silence is the unwritten rule according to which police officers never provide incriminating information about their colleagues and, as such, presents a huge obstacle in the effective control of misconduct at work (Mollen Commission, 1994). Since William Westley (1970) first profiled the police profession in his 1950s study of the Gary, Indiana Police Department, the code of silence has been viewed as a persistent outcome in the socialization of police. Westley (1970) and Bittner (1971) made the case that police work was dangerous and officers were socialized to trust only their peers and protect other officers through the “blue wall of silence.” This implicit code has likely existed nearly as long as law enforcement itself, but substantial scholarship on the blue wall of silence did not materialize until the 1990s over a period when the code was increasingly viewed as a leading hurdle to law enforcement reform in both developed and developing countries. Two public inquiries located at opposite ends of the globe were symbolic starting points for this newly heightened concern for the obstacles to addressing police misconduct in the 1990s. The Mollen Commission (1994, 53) found that the code of silence was pervasive throughout the New York Police Department and was the “most significant barrier to effective corruption control;” while the Wood Inquiry in Australia argued that the code of silence was an “incontrovertible and universal product of police culture” that was preventing reform in the New South Wales Police Service (Wood, 1997, 33). Over the same period, scholars and 1 practitioners were finding similar concerns in their studies of twenty-nine former communist countries in Central and Eastern Europe that sought to transition from the secretive world of security-focused militias (milicija) to become modern, democratic police agencies. The parameters of the stated code vary among the police organizations in different countries and even within a singular country or police administration. (Klockars et al., 2000). This is especially true of police in transitional countries, such as Croatia and South Africa. Although the code of silence problem is notably present in such countries, as a subject of serious empirical research, it appears most often in stable democratic countries. This paper provides a comparative approach to the code of silence in the police from three vastly different countries. The paper examines two transitional and specific countries—Croatia and South Africa— compared to the USA. The USA, a country with a historically stable democracy, is the reference point for cross-country comparisons. On the other hand, Croatia is a transitional, as well as post-conflict country, characterized by a high level of corruption. Like Croatia, South Africa is a transitional country facing the legacy of apartheid, as well as a high corruption rate and challenges in the protection of civil rights. Literature Review In terms of scholarship on the role of the code of silence in these agencies, a critical turning point was the July 1996 National Symposium on Police Integrity in the United States that gathered leading scholars and police departments to find new solutions to the “old problem” of police misconduct (Greene, Piquero, Hickman & Lawson, 2004). The symposium planners purposely chose to depart from the past singular concern for individual cases of corruption or misconduct and instead emphasize a new research focus based around the goal of a healthy law enforcement organization and work culture (Klockars, Kutnjak Ivković, and Haberfeld, 2006). This symposium was the starting point of several large research studies on police integrity including the first large-scale tests of what would become known as the Integrity Survey. Klockars and Kutnjak Ivković (2004) designed the theory of police integrity and the Integrity Survey to ameliorate the problems experienced in past studies asking direct questions about police misconduct and the respondents’ involvement in such behavior. This survey would instead measure integrity through police officers’ attitudes, perceptions, and self-reports on how they would respond in hypothetical situations describing integrity challenges. Besides US tests of the integrity instrument, the survey was soon adapted by scholars across the globe including 2 the first international use of the instrument in Croatia (Kutnjak Ivković, & Klockars, 1998). Since the turn of this century, researchers have used the Integrity Survey to assess the health of police agencies across the globe in more than 24 countries including Armenia, Australia, Austria, Bosnia and Herzegovina, Canada, Croatia, Czech Republic, Finland, Hungary, Japan, Malaysia, the Netherlands, Norway, Pakistan, Poland, Romania, Slovenia, South Africa, South Korea, Sweden, Thailand, Turkey, the UK, and the USA. Klockars and colleagues (see, e.g., Klockars and Kutnjak Ivković, 2004; Klockars et al., 1997; Klockars, Kutnjak Ivković, and Haberfeld, 2004a) developed the Integrity Survey as part of a theory of police integrity that incorporated the code of silence as a key dimension of the theory. The survey measured the code of silence through questions on the respondents' willingness to report a wide variety of police misconduct (e.g., Klockars et al., 2000; Pagon and Lobnikar, 2000; Micuccui and Gomme, 2005; Klockars, Kutnjak Ivković, and Haberfeld, 2006; Marche, 2009; Rothwell and Baldwin, 2007a, 2007b; Schafer and Martinelli, 2008; Kutnjak Ivković and Shelley, 2008; Wolfe and Piquero, 2011; Long, Cross, Shelley, and Kutnjak Ivković, 2013; Kutnjak Ivković, Haberfeld, and Peacock, 2013; Kutnjak Ivković and Sauerman, 2013). In their first survey of 3,250 police officers from 30 US local police agencies, Klockars and colleagues (2000) showed that the contours of the code vary greatly with different forms of police corruption described in the questionnaire. The results showed that there are also major differences in the contours of the code not only across scenarios, but also across police agencies (Klockars et al., 2000). Even when Klockars and colleagues used the police integrity questionnaire to survey police officers in three agencies of integrity (Klockars, Kutnjak Ivković, and Haberfeld, 2006), they found that officers in one agency adhered to the code much more than the officers in the others. The results of other US studies provided additional evidence of the diverse adherence to the code of silence across different police agencies (e.g., Marche, 2009; Micuccui and Gomme, 2005; Rothwell and Baldwin, 2007a, 2007b; Schafer and Martinelli, 2008; Wolfe and Piquero, 2011) and even within police agencies (e.g., Greene at al., 2004). The most consistent observation across studies has been finding that the code of silence was negatively related to the perceived seriousness of misconduct (Klockars et al., 1997, 2000; Kutnjak Ivković, Haberfeld, and Peacock, 2013; Kutnjak Ivković, Haberfeld, and Peacock, 2015; Kutnjak Ivković and Sauerman, 2013; Long, Cross, Shelley, and Kutnjak Ivković, 2013). Other studies suggest that the code of silence may vary by officer rank (e.g., Huon et al, 1995; Weisburd and Greenspan, 2000). When there are differences, the supervisor code of silence 3 usually was found to be narrower than the line officer code (e.g., Huon et al., 1995; Klockars, Kutnjak Ivković, and Haberfeld, 2006; Kremer, 2000; Kutnjak Ivković and Klockars, 1998; Kutnjak Ivković and Shelley, 2008; Pagon and Lobnikar, 2000; Weisburd and Greenspan, 2000). Comparative studies of the code of silence are rare. In the first such study, Klockars, Kutnjak Ivković and Haberfeld (2004) analyzed the code in 12 countries, and found significant differences among countries. They wrote that, in Croatia, Hungary, Pakistan, Poland, and South Africa, the code of silence is so strong that officers are actually estimating just how unwilling most officers are to report the misconduct described in the scenarios (Klockars, Kutnjak Ivković, Haberfeld, 2004). More recent comparative research conducted in ten countries (Armenia, Australia, Croatia, Estonia, Russia, Slovenia, South Africa, South Korea, Thailand, USA) indicates that, according to the results, the countries could be divided into four categories (Kutnjak Ivković, Haberfeld, 2015). The first category consists of Australia, Estonia, Slovenia, South Africa and the USA, whose mean values are “on the reporting side of scale” (Kutnjak Ivković, Haberfeld, 2015). The second category involves Armenia, Croatia and Thailand, whose mean values are also on the reporting side of scale, but are somewhat lower and closer to the midpoint of the scale. This would imply that codes in Armenia, Croatia and Thailand seem to be somewhat stronger than the codes in Australia, Estonia, Slovenia, South Africa and the USA. The results also imply that the codes are somewhat stronger in the third category of countries (South Korea). The fourth category shows that the code of silence is by far the strongest in Russia, whose mean values in all scenarios are predominantly grouped on the nonreporting side and the midpoint of the scale (Kutnjak Ivković, Haberfeld, 2015). The Code of Silence and Disciplinary Fairness In the 1990s, Kutnjak Ivković and Klockars (1996, 1998) sought to shed greater light on the code of silence by testing three theoretical models for a potential relationship between the code of silence and disciplinary fairness. Specifically, they proposed the discipline indifference model, the simple deterrence model, and the simple justice model. The discipline indifference model does not expect a relationship between the code and disciplinary fairness (Kutnjak Ivković and Klockars, 1998). This model assumes that officers consider other factors that interfere with their weighing of discipline fairness and the code of 4 silence. For example, if a police agency rarely disciplines for misconduct, police officers will accurately perceive that the certainty of discipline is low and the severity of the discipline would be considered less of a risk. In work cultures with a strong code of silence, the threat of discipline, particularly if exercised sporadically, may do very little to change the police officers' intentions. Therefore, the discipline indifference model assumes that the percentage of police officers willing to report misconduct would be unchanged regardless of differences in their perceptions on whether the discipline is too harsh, fair, or too lenient. The simple deterrence model expects that the code of silence and discipline severity are negatively related. The harsher the discipline, the less likely police officers will participate in the code of silence (Kutnjak Ivković and Klockars, 1998). This is a simplified model because it takes into account only the severity of discipline and disregards the certainty and celerity of discipline. Thus, the prediction of the simple deterrence model is that the percentage of police officers who adhere to the code of silence is the lowest for the officers who perceive discipline to be too harsh and the highest for the officers who perceive discipline to be too lenient. The last of the three proposed models is a simple justice model that predicts that police officers' motivation to report misconduct is linked with their desire to see misconduct disciplined appropriately or justly (Kutnjak Ivković and Klockars, 1998). Therefore, this model expects that police officers who perceive discipline as too harsh will stick to the code of silence. However, the model is silent about the relation between those who view discipline as fair and those who view discipline as too lenient. Kutnjak Ivković and Klockars (1998, p. 335) argued that: "The motive of simple justice would offer no reason for increased reporting of misconduct under conditions of excessive leniency.” Thus, the prediction of the simple justice model is that the percentage of police officers who adhere to the code of silence is higher for the officers who perceive discipline as too harsh than for officers who perceive discipline as fair. At the same time, there is no prediction for the relation among police officers who perceive discipline as fair or too lenient. The three proposed models for evaluating the code of silence were first tested on Croatian police (Kutnjak Ivković and Klockars, 1998) and followed up on with samples from Bosnia and Herzegovina, the Czech Republic (Kutnjak Ivković & Shelley, 2005; 2010), and the United States (Long, Cross, Shelley, and Kutnjak Ivković, 2013). These studies on the code of silence were performed using the first version of the integrity survey that was generally focused on corrupt acts, but scholars have increasingly turned to a second version of the integrity survey 5 that sought judgments on a broad spectrum of misconduct spanning code of conduct violations to excessive force incidents (see Klockars and Kutnjak Ivković, 2004). Kutnjak Ivković and Sauerman (2013) were the first to test the three models on a South African police survey. The cross-cultural comparisons of the results of tests of the three models do not suggest a uniform picture for the relation of the code and disciplinary fairness across respondents who evaluated discipline as too harsh and those who evaluated it as fair. In Croatia (Kutnjak Ivković and Klockars, 1998) and the Czech Republic (Kutnjak Ivković and Shelley, 2010) the simple justice model was largely supported (10 out of 11 scenarios in the Croatian study fit the model; 10 out of 10 scenarios for supervisors and 7 out of 10 scenarios for non-supervisors in the Czech study). The simple model found much weaker support in Bosnia and Herzegovina (4 out of 10 scenarios fit the model; Kutnjak Ivković and Shelley, 2005), the United States (2 out of 4 scenarios fit the model; Long, Cross, Shelley, and Kutnjak Ivković, 2013), and South Africa (6 out of 11 scenarios fit the model; Kutnjak Ivković and Sauerman, 2013). The results were more uniform when the relation between the code and disciplinary fairness were compared for the respondents who evaluated discipline as fair and those who evaluated discipline as too lenient. In all four studies (Kutnjak Ivković and Klockars, 1998; Kutnjak Ivković and Sauerman, 2013; Kutnjak Ivković and Shelley, 2005, 2010) the results suggest that most of the scenarios fit the discipline indifference model (8 out of 11 in Croatia, Kutnjak Ivković and Klockars, 1998; 7 out of 10 for the Czech respondents, Kutnjak Ivković and Shelley, 2010; 8 out of 10 for the Bosnian respondents, Kutnjak Ivković and Shelley, 2005; and 10 out of 11 for the South African respondents, Kutnjak Ivković and Sauerman, 2013). In the few cases in which differences were found among those judging discipline as fair and too lenient, the findings mostly suggested the simple deterrence model (e.g., Kutnjak Ivković and Sauerman, 2013; Kutnjak Ivković and Shelley, 2005), the simple justice model (e.g., Kutnjak Ivković and Klockars, 1998), or both (Kutnjak Ivković and Shelley, 2010). Methodology Klockars and Kutnjak Ivković (2004) developed the theory of police integrity and the accompanying methodology (e.g., Klockars and Kutnjak Ivković, 2004; Klockars, Kutnjak Ivković, and Haberfeld, 2006). One dimension of the theory focuses on various techniques used by police administrations to detect and research misconduct in the agency (Klockars, Kutnjak Ivković, 2004). 6 The theoretical approach is combined with a new methodology that facilitates the examination of police integrity as an organizational concept. It is precisely this organizational/occupational approach to the study of police integrity that involves questions of fact and opinion that can be explored directly, without arousing the resistance that direct inquiries about corrupt behavior are likely to provoke. With this approach, it is possible to ask nonthreatening questions about officers' knowledge of agency rules and their opinions about the seriousness of particular violations. Although it is possible to use an administrative/individual approach to measure corrupt behavior, the methodological obstacles are significant. Using the organizational/occupational approach, how seriously officers regard misconduct, how amiable they are to supporting discipline, and how wiling they are to tolerate misconduct in silence can easily be measured. The police integrity questionnaire contains 11 hypothetical scenarios describing a range of police misconduct. Four scenarios describe the use of excessive force, ranging from verbal abuse to the abuse of deadly force (Scenario 4, Scenario 6, Scenario 7, and Scenario 11). Another two scenarios describe falsification of an official report (Scenario 10) and failure to execute a felony warrant (Scenario 2). Five scenarios include examples of police corruption (Scenario 1, Scenario 3, Scenario 5, Scenario 8, and Scenario 9). Every scenario is accompanied by a series of seven questions, each measuring the normative reflections of police integrity. These questions focus on factual issues and opinions, exploring the respondents' views of misconduct seriousness, appropriate and expected discipline, and willingness to report misconduct. The dependent variables in our study are based on questions asking the respondents to assess whether they would report misconduct (“Do you think YOU would report a fellow police officer who engaged in this behavior”). Possible answers were provided in the form of a Likert scale ranging from 1 to 5, where 1 signifies “definitely not”, and 5 “definitely yes”. Later, for the requirements of bivariate analysis, the own reporting variable was recoded into two categories: “would not report” and “would report”. The independent variable in the study most relevant for our topic is the country itself. The created independent variable included three countries, the USA, Croatia and South Africa. In addition, we have included several organizational variables. 7 First, we measured how serious police officers consider the respective behaviors (“How serious do YOU consider this behavior to be?”). Possible answers ranged on a Likert scale from 1 to 5, with 1 signifying “not at all serious”, and 5 “very serious.” For the purpose of the bivariate and regression analyses, the variable was dichotomized into “would not report” and “would report”. Second, we measured the respondents’ estimates of actual discipline (“If an officer in your agency engaged in this behavior and was discovered doing so, what, if any, discipline do YOU think WOULD follow?”). The answers offered to the respondents depended upon the legal norms in each country. In the United States, the options were: “none”, “verbal reprimand”, “written reprimand”, “period of suspension without pay”, “demotion in rank” and “dismissal” (Kutnjak Ivković et al., 2015). In Croatia, the options were: “written reprimand”, “fine up to 10% of salary”, “fine up to 20% of salary”, “no promotion”, “no career advancement”, “reassignment to a lower position” and “dismissal” (Kutnjak Ivković, 2015). In South Africa, these options were: “none”, “verbal reprimand”, “written reprimand”, “reassignment to a different position”, “demotion in rank” and “dismissal” (Kutnjak Ivković and Sauerman, 2015). In each country, the ends of the scale (“none” and “dismissal”) are the same. We coded all other possible answers as one category (“some discipline other than dismissal”). Because few respondents selected either “none” or “dismissal” in some scenarios, we had to drop a category in these scenarios (e.g., in scenarios 1 and 7, the category “dismissal” was omitted; in scenarios 2, 3, 4, 9 and 10, the category “none” was omitted). Third, to assess the perceptions of fairness, we have created the variable discipline fairness for each scenario by subtracting the value for the expected discipline from the value for the appropriate discipline. When the result was zero, we assumed that the respondents evaluated expected discipline as fair. When the result was negative, we assumed that the respondents evaluated expected discipline as too harsh. When the result was positive, we assumed that the respondents evaluated expected discipline as too lenient. Fourth, we measured the respondents’ views about colleagues’ willingness to report unacceptable behavior (“Do you think MOST POLICE OFFICERS IN YOUR AGENCY would report a fellow police officer who engaged in this behavior?”). Possible answers were ranged on a 5-point Likert scale from 1 to 5, where 1 represented “definitely not”, and 5 “definitely yes”. With the purpose of later analyses, the variable was dichotomized into “would not report” and “would report”. Finally, our analyses include demographic variables as well. 8 First, the work experience was measured as the length of service in the police (“How many years have you been a police officer?”). Possible answers were: “less than 1 year”, “1-2 years”, “3-5 years”, “6-10 years”, “11-15 years”, “16-20 years”, and “over 20 years”. For the purposes of further analyses, the answers were recoded into: “up to 5 years”, “6-15 years” and “over 15 years”. Second, we also included a measure of the respondent's supervisory status. The answers were “yes” and “no.” The purpose of the paper is to determine the degree to which the code of silence is a uniform phenomenon spanning across different countries and police cultures. In particular, we would like to explore a connection between readiness to report and country of origin. Hence, our null hypothesis is that there is no statistically significant relation between the code of silence and the country of origin. Sample As there are 17,985 independent police organizations operating in the United States (BJS, 2011), it would be difficult to measure police integrity across the entire population (Kutnjak Ivković et al., 2015). Instead of collecting data from the entire population, Kutnjak Ivković and colleagues (2015) collected the data in 11 police agencies which generally represent the range of police employees in large, medium and small cities in the USA. The potential respondents were notified by email, and 37.4% of them participated in the study (Kutnjak Ivković et al., 2015). 9 Table 1: USA respondents' demographic characteristics Number of respondents Percent of respondents 445 168 72.6% 27.4% 91 113 298 366 59 14.6% 12.3% 32.0% 39.5% 6.4% 314 94 21 74 66 55.2% 16.5% 3.7% 13.0% 11.6% Supervisory role Non-supervisor Supervisors Length of service Up to 5 years 6-10 years 11-15 years 16-20 years Over 20 years Type of assignment Patrol Detective/investigative Community policing officer Special (vice, juvenile, etc.) Administrative The South African sample (Kutnjak Ivković and Sauerman, 2015) consisted of 871 police officers, surveyed in 2010-2012. The response rate was 87.5% (Kutnjak Ivković, Sauerman, 2015). The sample includes police officers from three different police agencies, across the South African provinces. 10 Table 2: South African respondents' demographic characteristics Supervisory role Non-supervisor Supervisors Length of service Up to 5 years 6-10 years 11-15 years 16-20 years Over 20 years Type of assignment Patrol Detective/investigative Communications Traffic Administrative Community policing Other Gender Male Female Number of respondents Percent of respondents 544 324 62.5% 37.5% 157 230 72 117 295 18.0% 26.4% 8.3% 13.4% 33.9% 243 298 50 36 112 65 76 27.9% 33.2% 5.7% 4.1% 12.9% 7.5% 8.7% 611 254 70.8% 29.2% The Croatian sample was a stratified sample of 996 respondents (Kutnjak Ivković, 2015). The sample was selected to represent 20 police administrations (four categories) and about 200 police stations (three categories). The authors created a 4 × 3 table by cross-tabulating four categories of police administrations with three categories of police stations (Kutnjak Ivković, 2015). The response rate was 88% (Kutnjak Ivković, 2015). 11 Table 3: Croatian respondents' demographic characteristics Number of respondents Supervisory role Non-supervisor Supervisors Length of service Up to 5 years 6-10 years 11-15 years 16-20 years Over 20 years Type of assignment Patrol Detective/investigative Communications Traffic Administrative Community policing Other Gender Male Female Percent of respondents 742 175 81.0% 19.0% 113 91 298 366 59 12.3% 9.8% 32.0% 39.5% 6.4% 335 162 53 47 17 96 197 36.9% 17.9% 5.8% 5.2% 1.8% 10.6% 21.8% 809 107 88.2% 11.8% The Croatian sample has the highest percentage of non-supervisors (81%; Kutnjak Ivković, 2015), followed by the U.S. sample (72.6%; Kutnjak Ivković et al., 2015) and the South African sample (62.5%; Kutnjak Ivković and Sauerman, 2015). The South African sample is somewhat less experienced (44% have 10 or fewer years of experience; Kutnjak Ivković, Sauerman, 2015), while the Croatian sample (22%; Kutnjak Ivković, 2015) and the U.S. sample (27%; Kutnjak Ivković et al., 2015) have about one-quarter of the respondents with 10 years or experience or fewer. While the Croatian and U.S. police officers are most frequently assigned to patrol, their South African counterparts are mostly engaged in detective/investigative assignments (Kutnjak Ivković, 2015, Kutnjak Ivković, Sauerman, 2015, Kutnjak Ivković et al., 2015). Results Results of Bivariate Analysis We began the analysis by researching the bivariate relation between the dependent variable (country of origin) and their own willingness to report in all 11 cases. 12 Every scenario shows a statistically significant relation between willingness to report and the country of origin (Table 4). In other words, the bivariate analyses suggest that the country is strongly linked with the respondents' expressed willingness to report. Hence, the results of bivariate analyses indicate that we should reject the null hypothesis. Post-hoc tests, according to the Marascuillo procedure (Glass, Hopkins, 1984), yielded statistically significant differences between the respective pairs in all 11 of the examined scenarios. It is interesting that, in 10 of the 11 scenarios, the differences in the Croatia-South Africa pairings are statistically significant including in the scenario describing the use of deadly force—the most serious use of excessive force scenario. In 9 out of 10 scenarios with statistically significant differences, Croatian police officers seemed more likely to say that they would adhere to the code of silence than their South African counterparts. Similarly, there were statistically significant differences in 9 out of 11 scenarios between the expressed willingness to report between Croatian and US respondents. In all but one scenario, Croatian respondents seem to be more likely to say that they would not report. The only scenario in which the U.S. respondents seemed more likely to stick to the code of silence is Scenario 1 which describes the acceptance of free gifts. Finally, in 9 out of 11 scenarios there were also statistically significant differences between the expressed willingness to report by the South African and the U.S. respondents. However, the differences were not uniformly spread out; in some scenarios, the U.S. respondents were more likely to say that they would not report, while in others, South African respondents were more likely to say the same. 13 Table 4: Bivariate analysis of willingness to report by country Croatia Would not report Rank South Africa Would not Rank report USA Would not report Rank Chi-Square Scenario 1 (Gifts) 24.1% 1 48.8% 10 59.8% 10 216.320*** Scenario 2 (Warrant) 52.2% 6 44.0% 8 61.7% 11 46.825*** Scenario 3 (Burglary) 46.7% 5 34.7% 7 20.0% 3 126.887*** Scenario 4 (Deadly Force) 44.0% 4 32.6% 6 13,0% 1 172.164*** Scenario 5 (Supervisor 2) 37.6% 2 17.2% 3 22.8% 5 109.379*** Scenario 6 (Hurts) 56.2% 8 16.8% 2 22.0% 4 392.569*** Scenario 7 (Verbal) 53.8% 7 32.2% 5 45.8% 9 91.392*** Scenario 8 (Alcohol) 58,5% 9 30.4% 4 41.2% 8 157.996*** Scenario 9 (Auto) 61.9% 11 55.9% 11 40.9% 7 70.496*** Scenario 10 (Report) 60.3% 10 47.1% 9 24.6% 6 202.527*** Scenario 11 (Watch) 37.6% 2 15.7% 1 14.3% 2 170.529*** * p<0.05; ** p<0.01; *** p<0.001 14 Significant pairs CRO v. USA SA v. USA CRO v. SA SA v. USA CRO v. SA CRO v. USA SA v. USA CRO v. SA CRO v. USA SA v. USA CRO v. USA CRO v. SA CRO v. USA SA v. USA CRO v. SA SA v. USA CRO v. SA CRO v. USA SA v. USA CRO v. SA CRO v. USA SA v. USA CRO v. USA SA v. USA CRO v. SA CRO v. USA CRO v. SA Phi .287 .134 .220 .259 .205 .389 .188 .248 .165 .280 .258 Table 5: Logistic regression of expressed willingness to report by organizational variables and country Scenario 1 (Gifts) Supervisory status1 Length of service2 6-15 years Over 15 years Own seriousness3 Expected discipline4 Some discipline Dismissal Discipline fairness5 Too harsh Too lenient Others' reporting6 Country7 Croatia South Africa Constant Model Nagelkerke R Squared Scenario 2 (Warrant) Scenario 3 (Burglary) Scenario 4 (Deadly Force) B .597 s.e. .150 Odds 1.816*** B .556 s.e. .124 Odds 1.743*** B .645 s.e. .148 Odds 1.905*** B .331 s.e. .177 Odds 1.392 .386 .387 1.316 .178 .185 .125 1.472* 1.473* 3.729*** .668 .531 2.330 .156 .157 .155 1.950*** 1.700*** 10.283*** .496 .579 1.914 .173 .176 .209 1.641** 1.785** 6.782*** .490 .498 1.580 .208 .211 .168 1.632** 1.645** 4.856*** 1.463 .185 4.321*** 1.370 .141 3.937*** 1.752 .130 5.765*** 1.773 .159 5.889*** -1.999 1.158 2.542 .392 .230 .131 .136*** -1.030 3.184*** .937 12.707*** .757 .228 .209 .110 .357*** 2.553*** 2.132*** -1.470 1.261 3.010 .236 .197 .182 .230*** 3.530*** 20.289*** -1.238 1.373 3.930 .244 .332 .209 .290*** 3.946*** 50.883*** .282 1.134 -3.609 .153 .158 .262 1.326 3.108*** .027*** .145 .136 .237 .950 1.615*** .029 -.399 -.025 -3.059 .169 .174 .296 .671* .976 .047 -.408 .227 -2.930 .239 .248 .333 .665 1.255 .053*** .571 -.052 .480 -3.544 .343 .539 * p<0.05; ** p<0.01; *** p<0.001 1 Reference category = "No"; 2 Reference category = "Over 5 years"; 3 Reference category = "Not serious"; 4 Reference category = "None"/"Some Discipline“; 5 Reference category = "Fair"; 6 Reference category = "Would not report"; 7 Reference category = "USA"; 15 .667 Scenario 5 (Supervisor 2) Supervisory status1 Length of service2 6-15 years Over 15 years Own seriousness3 Expected discipline4 Some discipline Dismissal Discipline fairness5 Too harsh Too lenient Others' reporting6 Country7 Croatia South Africa Constant Model Nagelkerke R Squared Scenario 6 (Hurts) Scenario 7 (Verbal) Scenario 8 (Alcohol) B s.e. Odds B s.e. Odds B s.e. Odds B s.e. Odds .521 .186 1.683** .912 .184 2.488*** .908 .179 2.480*** .788 .185 2.198*** -.383 -.166 1.002 .220 .227 .150 .682 .847 2.725*** -.277 -.395 1.327 .222 .228 .158 .758 .674 3.770*** -.384 -.293 1.992 .236 .237 .167 .681 .746 7.331*** .329 .153 1.626 .217 .223 .158 1.390 1.166 5.085*** 1.984 .256 7.275*** 2.094 3.842 .285 .172 8.115*** 46.604*** 1.540 .281 4.666*** 1.734 1.596 .251 .364 5.663*** 4.935*** -2.021 1.823 3.474 .489 .296 .154 .133*** -2.229 6.192*** 1.268 32.264*** 3.842 .378 .406 .172 .108*** 3.553** 46.604*** -2.214 .866 4.187 .488 .446 .173 .109*** 2.377 65.846*** -1.904 1.649 3.784 .296 .401 .161 .149*** 5.201*** 43.976*** -.375 .298 -2.817 .190 .205 .340 .202 .216 .337 .757 2.369*** .033*** -.487 .353 -3.880 .195 .198 .372 .614** 1.424 .021*** .109 .676 -4.341 .198 .206 .357 1.115 1.996** .013*** .637 .687* 1.347 .060 -.278 .862 -3.397 .735 .753 * p<0.05; ** p<0.01; *** p<0.001 1 Reference category = "No"; 2 Reference category = "Over 5 years“; 3 Reference category = "Not serious"; 4 Reference category = "None"/„Some Discipline“; 5 Reference category = "Fair"; 6 Reference category = "Would not report"; 7 Reference category = "USA"; 16 .743 Scenario 9 (Auto) Supervisory status1 Length of service2 6-15 years Over 15 years Own seriousness3 Expected discipline4 Some discipline Dismissal Discipline fairness5 Too harsh Too lenient Others' reporting6 Country7 Croatia South Africa Constant Model Nagelkerke R Squared Scenario 10 (Report) Scenario 11 (Watch) B s.e. Odds B s.e. Odds B s.e. Odds .471 .160 1.602** .446 .153 1.562** .934 .221 2.545*** .090 .243 2.620 .218 .217 .206 1.094 1.275 13.732*** .009 -.036 2.694 .203 .206 .226 1.009 .965 14.788*** -.644 -.603 1.230 .257 .271 .172 .525 .547 3.422*** 1.727 .161 5.624*** 1.280 .140 3.596*** 1.974 .319 7.202*** -1.298 1.335 3.749 .311 .259 .232 .273*** 3.800*** 42.476*** -1.347 1.426 4.448 .293 .259 .290 .260*** 4.161*** 84.474*** -2.430 1.953 3.773 .387 .678 .182 .088*** 7.049** 43.506*** -.702 -.504 -3.197 .173 .185 .293 .667 .495*** .604** .041*** -.464 .017 -3.277 .174 .182 .310 .664 .629** 1.017 .038*** -.134 .431 -1.091 .243 .260 .317 .641 .875 1.539 .336** * p<0.05; ** p<0.01; *** p<0.001 1 Reference category = "No"; 2 Reference category = "Over 5 years“; 3 Reference category = "Not serious"; 4 Reference category = "None"/„Some Discipline“; 5 Reference category = "Fair"; 6 Reference category = "Would not report"; 7 Reference category = "USA"; 17 Results of Regression Analyses To tease out the potential country effect on the code of silence when other variables are controlled for, we performed a logistic regression of the respondents' own willingness to report. Instead of summarizing the results for each scenario, we provide a summary of findings across all 11 scenarios. We first start the analysis by exploring organizational factors (perceptions of seriousness, expected discipline, discipline fairness, and others' reporting), followed up by the respondents' demographic characteristics (supervisory position, length of service), and last but not least important - country. By far the most important predictor of the respondents' expressed willingness to report is their assessment of whether other police officers would report (Table 5). In fact, estimated others’ willingness to report is identified as a statistically significant predictor of willingness to report in all eleven scenarios. Accordingly, those respondents who think that others would report misconduct are also more likely to state their willingness to report such behavior. Another important organizational variable is the respondents' own assessment of misconduct seriousness. Again, own perception of the seriousness of misconduct is a statistically significant predictor of their own willingness to report. Those respondents who evaluated behavior to be serious also appear more likely to say that they would indeed report it. The anticipated disciplinary measure is also identified as an important predictor for own reporting with this variable showing statistical significance in all 11 scenarios. However, for methodological reasons (too few respondents), not all categories were used in all scenarios (Table 5). When “no discipline” was used as a reference category (Scenario 1: free meals, gifts from merchants, Scenario 5: supervisor offers holiday for errands, Scenario 6: officer strikes prisoner, Scenario 7: verbal abuse of motorist, Scenario 8: cover-up of police DUI accident), the results suggested that, compared to the respondents who thought that no discipline would follow, the respondents who thought that some discipline would follow were more likely to say that they would report. When “some discipline“ was used as a reference category (Scenario 2: failure to arrest friend with warrant, Scenario 3: theft of watch from crime scene, Scenario 4: unjustifiable use of deadly force, Scenario 6: officer strikes prisoner, Scenario 8: cover-up of police DUI accident, Scenario 9: auto body shop 5% kickback, Scenario 10: false report of drug possession, Scenario 11: Sgt. fails to halt beating of child abuser), compared to the respondents who thought that “some discipline” would follow, the respondents who thought that dismissal would follow were more likely to say that they would report. 18 We notice that discipline fairness is another extremely important predictor of willingness of police officers to report unacceptable behavior at work, being statistically significant in all of the 11 examined scenarios. Compared to those who perceived discipline to be fair, those who perceived discipline to be harsh were less likely to say that they would report it. Compared to those who perceived discipline to be fair, those who perceived discipline to be too lenient were more likely to say that they would report in 10 of the 11 scenarios. Our logistic regression models also included two demographic characteristics. Compared to the employees with up to 5 years of work experience, employees with 6-15 years of work experience seemed more likely to say that they would report in only 4 out of 11 scenarios (Scenario 1: free meals, gifts from merchants, Scenario 2: failure to arrest friend with warrant, Scenario 3: theft of watch from crime scene, and Scenario 4: unjustifiable use of deadly force). Compared to the employees with up to 5 years of work experience, employees with over 15 years of work experience were also more likely to say that they would report in 4 out of 11 scenarios (Scenario 1: free meals, gifts from merchants, Scenario 2: failure to arrest friend with warrant, Scenario 3: theft of watch from crime scene, and Scenario 4: unjustifiable use of deadly force). In this paper, supervisory status is recognized as a statistically significant predictor of supervisor readiness to report in 10 out of the 11 scenarios. Compared to line officers, supervisors were more likely to say that they would report. It should be noted that the only scenario in which supervisory status is not significant, and does not affect willingness for own reporting, is the abuse of deadly force scenario. Finally, we explore the independent country effect, with the answers by the U.S. respondents serving as the reference category. The results of our analyses show that the country effect is significant in 9 out of 11 scenarios. The only two scenarios in which country is not a significant predictor of the respondents' willingness to report are those evaluated as the most serious in all three countries (Kutnjak Ivković, 2015, Kutnjak Ivković, Sauerman, 2015, Kutnjak Ivković et al., 2015): abusing deadly force and stealing from a crime scene. In other words, our null hypothesis has been rejected in 9 out of 11 scenarios in which we found the country effect. Compared to the U.S. respondents, Croatian respondents seemed less likely to say that they would report in 5 out of 11 scenarios (Scenario 3: theft of watch from crime scene, Scenario 5: supervisor offers holiday for errands, Scenario 7: verbal abuse of motorist, Scenario 9: auto body shop 5% kickback, Scenario 10: false report of drug possession). While there are 19 statistically significant differences between the South African respondents and the U.S. respondents in 5 scenarios (Scenario 1: free meals, gifts from merchants, Scenario 2: failure to arrest friend with warrant, Scenario 6: officer strikes prisoners, Scenario 8: cover-up of police DUI accident, Scenario 9: auto body shop 5% kickback), police officers from South Africa seemed more willing to say that they would report than police officers from the U.S.A. in 4 out of 5 scenarios. Discussion and Conclusion The results of our analyses point out that the country of respondents’ origin is related to their expressed probability of reporting in all except for the two most serious cases. Specifically, our multivariate models indicate the persistence of different code of silence forms in different countries, even when control variables are included in the model. In other words, our null hypothesis of no country influence on the code of silence has been rejected in all but two scenarios. These two scenarios are the most serious forms of misconduct—stealing from a crime scene and abusing deadly force—and police officers across three countries differed very little in how likely they were to say that they would report. Our findings related to the two most serious scenarios are in line with prior comparative research. In particular, Kutnjak Ivković and Haberfeld (2015) reported how the code of silence is present in 10 countries in their research. However, Kutnjak Ivković and Haberfeld (2015) stated that despite the importance of country as a predictor in a large number of scenarios, it was noticed that the country was not an important predictor in serious situations, such as in scenarios of unjustifiable use of deadly force and scenarios where police supervisors fail to intervene when witnessing police officers under their supervision strike an arrested person. It is obvious that in such situations, there are expectations similar to serious consequences for such unacceptable behavior, and therefore respondents from various countries were all less likely to state that they would not report. Compared to a democratic country such as the United States, Croatia—a country in transition— has a stronger code of silence. On the other hand, a comparison of the U.S. results with the results from another transitional country—South Africa—shows differences as well, but the direction of differences is not uniform. In some scenarios, respondents from South Africa seemed more likely to say that they would report than the respondents from the United States, while in other scenarios, respondents from South Africa seemed to be more likely to say that 20 they would report than their counterparts from the United States, a developed Western democracy. The results demonstrate that the country of origin is an important predictor of code of silence presence. However, one cannot unambiguously conclude that a transitional country always produces a stronger code of silence. The fairness perception of the expected discipline is a significant predictor in all scenarios tested in this study. In relation to respondents who think that the disciplinary measures are too harsh, respondents who consider the disciplinary measure fair were more likely to say that they would report unacceptable behavior at work. This result is in accordance with expectations of one of the theoretical models (Kutnjak Ivković, Klockars, 1998). More specifically, the motivation of police officers for reporting misconduct is also related and influenced by their desire to see misconduct at work disciplined in an appropriate and fair manner, which is in accordance with the work of Kutnjak Ivković and Klockars (1998). This paper represents further progress in police integrity research as it provides a detailed analysis of the code of silence in three countries. However, limitations can be imagined, including the possible non-representative nature of the U.S. and South African police officer samples when compared to the entire police populations in the two countries. As a result, the national differences represented in this study may be limited to the police administrations that participated in the study. Furthermore, a comparative examination of officer perceptions of disciplinary measures may be hindered by the different legal frameworks that officers operate under in these three countries despite our preparation of scenarios to facilitate such country comparisons. Furthermore, the different language adaptations of the questionnaire may be a limiting factor in comparing officer responses, while the legal and social conditions under which the agencies were formed could impact the functioning of the different police administrations and the behavior of the police officers they employ (Kutnjak Ivković, Haberfeld, 2015). Comparisons to date on police integrity research (for instance, Klockars, Kutnjak Ivković, and Haberfeld, 2004; Kutnjak Ivković and Haberfeld, 2015) demonstrate however that the questionnaire is suitable for comparative analyses. 21 References Bittner, E. (1971). The Functions of the Police in Modern Society. Washington D.C.: U.S. Government Printing Office. Glass, G. V., Hopkins, K. D. (1984). Statistical Methods in Education and Psychology. Boston: Allyn and Bacon., p. 391-392. Greene, J. R., Piquero, A. R., Hickman, M. J. & Lawton, B. A. (2004). Police Integrity and Accountability in Philadelphia: Predicting and Assessing Police Misconduct. Retrieved at: https://www.ncjrs.gov/pdffiles1/nij/grants/207823.pdf. Huon, G. F., Hesketh, B. L., Frank, M. G., McConkey, K. M. & McGrath, G. M. (1995). Perceptions of ethical dilemmas: ethics and policing — Study 1, Report Series. Klockars, C.B., Kutnjak Ivković, S. & Haberfeld, M.R. (2000). The measurement of police integrity. Research in brief. Washington, DC: National Institute of Justice. Klockars, C. B., & Kutnjak Ivković, S. (2004). Measuring police integrity. In M. Hickman, A. R. Piquero & J. R. Greene (Eds.), Police Integrity and Ethics. Belmont, CA: Wadsworth. Klockars, C.B., Kutnjak Ivković, S., & Haberfeld, M.R. (2006). Enhancing police integrity. Dordrecht, Netherlands: Springer. Kutnjak Ivković, S., Haberfeld, M.R. & Peacock, R. (2013). Rainless West: the integrity survey’s role in agency accountability. Police Quarterly, 16(2), 148-176. Kutnjak Ivković, S. & Klockars, C. B. (1998). The Code of Silence and the Croatian Police. In Pagon, M. (Ed.). Policing in Central and Eastern Europe: Organizational, Managerial, and Human Resource Aspects. Ljubljana, Slovenia: College of Police and Security Studies. Kutnjak Ivković, S. & O’Connor Shelley, T. (2010). The Code of Silence and Disciplinary Fairness: A Comparison of Czech Police Supervisor and Line Officer Views. Policing: An International Journal of Police Strategies and Management, 33, 548-574. Kutnjak Ivković, S. & O’Connor Shelley, T. (2007). Police Integrity and the Czech Police Officers, International Journal of Comparative and Applied Criminal Justice, 31, 2149. Kutnjak Ivković, S. & O’Connor Shelley, T. (2005). The Bosnian Police and Police Integrity: A Continuing Story, European Journal of Criminology, 2, 428-464. Kutnjak Ivković, S. & Sauerman, A. (2013). Curtailing the code of silence among the South African police. Policing: An International Journal of Police Strategies & Management, 36(1), 175-198. 22 Kutnjak Ivković, S. (2015). Studying Police Integrity in the United States. Measuring Police Integrity Across the World (p. 1-36). New York: Springer-Verlag. Kutnjak Ivković, S. (2015). Police Integrity Croatia. Measuring Police Integrity Across the World (p. 97-124). New York: Springer-Verlag. Kutnjak Ivković, S., Haberfeld, M. (2015). A comparative Perspective on Police Integrity. Measuring Police Integrity Across the World (p. 329-368). New York: Springer-Verlag. Kutnjak Ivković, S., Haberfeld, M. & Peacock, R. (2015). Police Integrity in the United States. Measuring Police Integrity Across the World (p. 295-327). New York: Springer-Verlag. Kutnjak Ivković, S., Sauerman, A. (2015). Police Integrity in South Africa. Measuring Police Integrity Across the World (p. 213-240). New York: Springer-Verlag. Long, M. A., Cross, J. E., O’Connor Shelley, T., & Kutnjak Ivković, S. (2013). The Normative Order of Reporting Police Misconduct: Examining the Roles of Offense Seriousness, Legitimacy, and Fairness, Social Psychology Quarterly, 76, 242-267. Marche, G.E. (2009). Integrity, culture, and scale: An empirical test of the big bad police agency. Crime, Law, and Social Change 51:463-486. Micucci, A. J. & Gomme, I.M. (2009). American police and subcultural support for the use of excessive force. Journal of Criminal Justice, 33, 487-500. Mollen, M. (1994). Anatomy of Failure: A Path for Success, Mollen Report. New York: Commissions to Investigate Allegations of Police Corruption and the Anti-Corruption Procedures of the New York Police Department. Pagon, M. & Lobnikar, B. (2004). Police integrity in Slovenia, in Klockars, C.B., Kutnjak Ivković, S. & Haberfeld, M.R. (Eds.). The contours of police integrity. Thousand Oaks, CA: Sage Publications. Rothwell, G. R. & Baldwin, J. N. (2007). Ethical climate theory, whistle-blowing, and the code of silence in police agencies in the state of Georgia. Journal of Business Ethics, 70 (4), 341-361. Schafer, J.A. & Martinelli, T.J. (2008). First-line supervisor’s perceptions of police integrity. Policing: An international journal of police strategies and management 31(2), 306-323. Weisburd, D. & Greenspan, R. (2000). Police Attitudes Toward Abuse of Authority: Findings from A National Study. Washington, DC: National Institute of Justice. Retrieved at: https://www.ncjrs.gov/pdffiles1/nij/181312.pdf. Westley, W. (1970) Violence and the Police: A Sociological Study of Law, Custom, and Morality. Cambridge, MA: MIT Pres Wolfe, S. E., & Piquero, A. R. (2011). Organizational justice and police misconduct. Criminal Justice & Behavior, 38, 332–353. 23 Wood, J. (1997). Royal Commission into the New South Wales Police Service. Final Report, Vol. 1. Sydney, Australia: Government of the State of New South Wales. 24