Keywords

Introduction

Both Westley (1970) and Bittner (1971) argued that, because police work is dangerous and the danger is unpredictable, police recruits are taught at the police academy and subsequently by their field training officers that they should rely on and trust only their fellow officers. Socialization into the police culture then implies that, if police recruits want to be trusted by their more experienced peers and receive support when they need it, they should participate in the code of silence and keep close to their chest any information about police misconduct committed by fellow police officers.

The first empirical studies of the code in the United States could be traced back to Westley’s study of the police profession (1970). A substantial increase in scholarship exploring the code of silence started in the 1990s, upon revelations that the code of silence is present and is an obstacle in any serious police reform. At the time, the Mollen Commission (1994, p. 53) in New York reported that the code of silence was prevalent in the NYPD and described it as “the most significant barrier to effective corruption control.” Similarly, the Wood Commission in Australia (Wood, 1997) argued that the code of silence is a serious obstacle to a reform of the New South Wales Police Service. This was also the time when Klockars and colleagues (Klockars et al., 1997, 2000; Klockars & Kutnjak Ivković, 2004) proposed the theory of police integrity and developed the accompanying methodology, facilitating empirical studies of the code of silence.

This chapter utilizes the police integrity theory and the related methodology to explore the contours of police integrity in one U.S. police agency. We analyze the degree to which the code in this agency protects various forms of police misconduct, including police corruption, use of excessive force, organizational deviance, and interpersonal deviance. In the process, we also compare not only across different forms of police misconduct, but also across different levels of misconduct severity within each form of misconduct. In the subsequent multivariate analyses, we test the effects of traditional police integrity correlates on the police officers’ adherence to the code of silence. In particular, we explore the influence of organizational factors, such as severity of expected discipline, familiarity with official rules, and expectations of fellow police officers’ estimated willingness to report misconduct, on the police officers’ reluctance to report misconduct.

Studying the Code of Silence

Ever since Klockars and colleagues (Klockars & Kutnjak Ivković, 2004; Klockars et al., 2000, 2004a, b, 2006) proposed this novel way of studying police misconduct, scholars across the world have used it to measure police integrity in general and the code of silence in particular. Specifically, the police integrity theory and the related methodology (e.g., Klockars et al., 1997, 2006; Klockars & Kutnjak Ivković, 2004) have been used to measure empirically the extent of the code of silence in about 30 countries across the world (for an overview, see Kutnjak Ivković, 2015a; see also Klockars et al., 2004a, b; Kutnjak Ivković & Haberfeld, 2015a, b), spanning across continents (e.g., North America, Europe, Asia, Africa, Australia), legal traditions (e.g., civil-law tradition, common-law tradition, Islamic), and levels of economic development (e.g., developed countries, countries in transition).

A number of the countries included in the studies are established democracies, such as Australia (Porter et al., 2015), Austria (Edelbacher & Kutnjak Ivković, 2004), Belgium (Van Droogenbroeck et al., 2019), Britain (Westmarland, 2004, 2005), Canada (Alain, 2004), Finland (Pounti et al., 2004), the Netherlands (Punch et al., 2004), Sweden (Torstensson Levander & Ekenvall, 2004), and the United States (e.g., Klockars et al., 1997, 2004b; Klockars & Kutnjak Ivković, 2004; Kutnjak Ivković et al., 2015). Extant research also included studies from countries in transition, such as Armenia (Khechumyan & Kutnjak Ivković, 2015; Kutnjak Ivković & Khechumyan, 2014), Bosnia and Herzegovina (Datzer et al., 2019; Kutnjak Ivković & Shelley, 2005, 2008), China (Wu & Makin, 2019), Croatia (Kutnjak Ivković & Klockars, 2004; Kutnjak Ivković, 2015b; Kutnjak Ivković et al., 2016), Estonia (Vallmüür, 2015, 2019), Poland (Haberfeld, 2004), Russia (Cheloukhine et al., 2015), Serbia (Peacock et al., 2020), Slovenia (Lobnikar & Meško, 2015), and South Africa (Sauerman & Kutnjak Ivković, 2015).

Although the strength of the code seemed to vary across countries or clusters of countries, at least some presence of the code of silence has been detected in every country covered by empirical studies to date. As Klockars and colleagues (Klockars et al., 2004a, p. 13) summarized the findings of their 14-country comparison, “[p]erhaps the most dramatic finding that emerges from examining the contours of integrity concerns the worldwide prevalence of the code of silence.” To illustrate how diverse the contours of the code could be across the world, Klockars and colleagues (Klockars et al., 2004a, p. 17) further elaborated and emphasized that, in about one-third of the countries included in the study, “not a single incident out of the 11 incidents described in the survey would be very likely to be reported.” On the other hand, in about one-third of the countries included in the study most, but not all of the incidents would not be protected by the code.

After they designed the police corruption questionnaire and collected the U.S. data, Klockars and colleagues (Klockars et al., 2000, p. 10) argued that the police corruption questionnaire addresses just one aspect of police integrity and that “the second generation of this survey” (i.e., the police integrity questionnaire) will provide coverage of other forms of police misconduct as well. This comparative 14-country study—utilizing the police corruption questionnaire—revealed substantial differences in the extent of the code of silence in terms of protecting different forms of police corruption (Klockars et al., 2004a). The subsequent 10-country study, based on the police integrity questionnaire (Kutnjak Ivković & Haberfeld, 2015b), provided further support to the claim that the extent of the code of silence varies greatly across the world and empirically confirmed that this conclusion holds not only for police corruption, but for other forms of police misconduct as well. However, no study to date of which we are aware has used the third version of the survey, one that incorporates an even wider range of scenarios—scenarios including examples of police corruption, use of excessive force, organizational deviance, and interpersonal deviance—to study the contours of the code of silence.

Correlates of the Code of Silence

Organizational Correlates

The theory of police integrity (Klockars et al., 1997, 2000, 2004a, b, 2006) is organizational in nature, stating that the police agency has a critical role in shaping the level of police integrity among its employees. To that end, as Klockars and Kutnjak Ivković (2004) were developing the theory and the related methodological approach, they introduced measures of these organizational components.

The first dimension of the theory focuses on the official rules and the way in which they are made by the administration, communicated to the police officers, and supported by them (Klockars et al., 2000, 2004a, b, 2006). The specific questions measuring this first dimension of the theory focus on the police officers’ perceptions of misconduct seriousness and their familiarity with the official rules. Starting with the first survey conducted using this theoretical and methodological approach (Klockars et al., 2000), studies have consistently shown that the extent of the code of silence and the police officers’ own evaluations of misconduct seriousness are strongly and negatively related (Cheloukhine et al., 2015; Haberfeld, 2004; Hickman et al., 2016; Khechumyan & Kutnjak Ivković, 2015; Klockars et al., 1997, 2004a, b, 2006; Kutnjak Ivković et al., 2013, 2018; Kutnjak Ivković & Khechymian, 2013; Kutnjak Ivković, Peacock, & Haberfeld, 2016; Kutnjak Ivković & Sauerman, 2013; Kutnjak Ivković, Haberfeld, Cajner Mraović, et al., 2019; Lobnikar & Meško, 2015; Long et al., 2013; Lim & Sloan, 2016; Maskály et al., 2019; Pagon & Lobnikar, 2000; Peacock et al., 2020; Porter & Prenzler, 2016; Vallmüür, 2015; Westmarland, 2006; Wu & Makin, 2019). This was the case not only within each country studied, but also across countries (e.g., Andreescu et al., 2012; Huberts et al., 2003; Klockars et al., 2004a; Kutnjak Ivković & Haberfeld, 2015a, b; Pagon et al., 2000). The most recent comparative study (Kutnjak Ivković & Haberfeld, 2015b) showed that the examples of police misconduct evaluated as the least serious, such as the acceptance of gratuities and a verbal abuse of citizens, are more likely to be protected by the code of silence than the examples of police misconduct evaluated as the most serious, such as stealing from a crime scene and abusing deadly force.

Evaluations of whether the example of police misconduct violates official rules constitute another measure of the first dimension of police integrity. The relationship between the police officers’ familiarity with official rules and their willingness to stick to the code of silence has not been explored as frequently. However, the results show that familiarity with official rules is not such a strong predictor (e.g., Kutnjak Ivković et al., 2018; Peacock et al., 2020; Van Droogenbroeck et al., 2019). In a U.S. study exploring the effects of familiarity with official rules on the code of silence (Kutnjak Ivković et al., 2018), familiarity with official rules was significant in about one-half of the scenarios, but the effects disappeared in multivariate models in which other organizational variables have been included. Similarly, in a study of Belgian police officers, Van Droogenbroeck et al. (2019) discovered that familiarity with official rules was completely mediated through the perceived willingness to report by fellow police officers.

The second dimension of the theory focuses on the control mechanisms (Klockars et al., 2000, 2004a, b, 2006). Questions about the appropriate discipline and expected discipline (i.e., the discipline respondents anticipate that the police agency would mete out) measure this dimension (Klockars et al., 2000, 2004a, b, 2006). The severity of the expected discipline seems to be negatively related to the strength of the code of silence (e.g., Kutnjak Ivković et al., 2018; Kutnjak Ivković, Haberfeld, Cajner Mraović, et al., 2019; Kutnjak Ivković & Shelley, 2008; Lim & Sloan, 2016; Peacock et al., 2020; Wolfe & Piquero, 2011).

The third dimension of the theory focuses on the code of silence. The related methodology not only incorporates a question measuring the code of silence, but also contains a question measuring police officers’ views about their peers’ adherence to the code of silence. Specifically, the question measures perceptions about the most police officers’ willingness to report misconduct . In extant literature, this measure was viewed as the measure of organizational culture within the police agency (e.g., Porter & Prezler, 2019), a deviant climate within the police organization (e.g., Lim & Sloan, 2016), and endorsement of peers (Long et al., 2013). Extant research uniformly shows that in multivariate models the estimates of peers’ willingness to report are the most consistent and the strongest predictor of the respondents’ own expressed willingness to report (Hickman et al., 2016; Kutnjak Ivković et al., 2018; Kutnjak Ivković et al., 2019; Lim & Sloan, 2016; Long et al., 2013; Peacock et al., 2020; Van Droogenbroeck et al., 2019).

Individual Correlates

The effect of several individual correlates has been previously tested in multivariate models of the code of silence. When included in the models, supervisory status is typically correlated with the respondents’ own expressed willingness to report in most scenarios (Kutnjak Ivković et al., 2018; Kutnjak Ivković, Haberfeld, Cajner Mraović, et al., 2019; Long et al., 2013; Peacock et al., 2020), although some studies found it to be less significant (e.g., Lim & Sloan, 2016; Van Droogenbroeck et al., 2019). Assignment was also included in a few studies and yielded mixed results (e.g., Lim & Sloan, 2016; Long et al., 2013). In multivariate models, the length of service was typically found to be a non-significant predictor (Hickman et al., 2016; Kutnjak Ivković et al., 2018; Kutnjak Ivković, Haberfeld, Cajner Mraović, et al., 2019; Peacock et al., 2020; Van Droogenbroeck et al., 2019), occasionally displaying significance in some specific scenarios (e.g., Long et al., 2013).

These individual correlates also included demographic characteristics. Respondents’ gender was rarely included in the multivariate models; when it was included, it was consistently shown to be unrelated to the respondents’ willingness to report (Hickman et al., 2016; Van Droogenbroeck et al., 2019). Police officers’ education is rarely explored in these models. Lim and Sloan (2016) found that it had a significant effect in only one out of the four types of scenarios. Similarly, Datzer et al. (2019) found that the educational level is not related to the police officers’ expressed willingness to report, while the type of education (i.e., police-related education vs. general education) is.

This Chapter

Extant research on the code of silence has demonstrated that organizational variables are critical predictors, whereas individual characteristics are not such strong predictors of the police officers’ expressed willingness to report. However, all prior studies of the code of silence have utilized the first police corruption questionnaire or the second police integrity questionnaire, but none have used the new version of the police integrity questionnaire that contains scenarios representing four types of police misconduct. This chapter follows in the footsteps of the traditional police integrity approach and incorporates traditional measures of police integrity in the analyses. At the same time, it expands the existing literature by exploring the effects of these critical variables in the scenarios describing not only police corruption and the use of excessive force, which have traditionally been used in the police integrity literature, but also in the scenarios of organizational deviance and interpersonal deviance, which have not been explored previously in the literature.

Methodology

Sample

In 2018/2019, we have surveyed police officers from a U.S. medium-size municipal police agency serving an urban community. The sample includes 148 police officers, both line officers and supervisors. For a detailed description of the characteristics of our sample and police agency, please see Chap. 1.

Measures

The analyses in this chapter are based on the results of the new version of the police integrity questionnaire (Kutnjak Ivković, Haberfeld, Cajner Mraović, et al., 2019) that includes scenarios dealing with police corruption, use of excessive force, organizational deviance, and interpersonal deviance. Upon reading the description of each scenario, the respondents were asked seven identical questions, targeting the respondents’ perceptions of misconduct seriousness, their familiarity with the official rules, views about appropriate and expected discipline, and expressed willingness to report misconduct. For details, please see Chap. 1.

Dependent Variable

Our measure of the respondents’ adherence to the code of silence—their own willingness to report misconduct—is based on the question asking the respondents to assess whether they would be willing to report the misconduct described in the scenario. The respondents could have selected an answer from a 5-point Likert-type scale, ranging from 1 = “definitely would not report” to 5 = “definitely would report.” The responses were ultimately collapsed into two categories; values of 1 and 2 were coded as a 1 (i.e., adhere to the code of silence) and values of 3 through 5 were recoded as a 0 (i.e., not adhering to the code of silence). However, for the most severe scenarios depicting corruption (theft from a burglary scene) and excessive force (shooting a suspect in the back), we coded the variable differently: values 1 through 3 were coded as 1 (i.e., adhering to the code of silence) and values 4 and 5 were coded as 0 (i.e., not adhering to the code of silence).

The decision to code the answers in the most severe scenarios differently was made for three reasons. First, the bulk of the prior literature studying the code of silence has opted to collapse the variables in this manner. Thus, to make our results comparable with prior research we scaled our analyses the same way. Second, the distribution of willingness to report was definitively not uniform nor consistent across scenarios, which causes methodological problems for estimating multivariate effects. In other words, leaving the ordered version of the dependent variable almost ubiquitously violated the proportional odds assumption that is at the key of ordered logistic regression models (Long & Freese, 2014). Finally, when we looked at the distribution of responses, there was evidence of a Jenks break in the data, whereby there were two homogenous subgroups within the data that represented the respondents better than the initial five categories. This is not uncommon in Likert data, where participants frequently make a distinction without a difference (Davis, 1987).

Organizational Independent Variables

We have included several organizational variables in our models. The respondents were first asked to provide their own evaluations of misconduct seriousness. Possible answers to this question ranged on a 5-point Likert-type scale from 1 = “not at all serious” to 5 = “very serious.”

The respondents were also asked whether they think that the behavior violates official rules. Possible answers to the question about familiarity with the official rules ranged on a 5-point scale from 1 = “definitely not” to 5 = “definitely yes.”

There were also two questions about the discipline. The first question asked the respondents to share their views of the appropriate discipline for each example of police misconduct, while the next question asked them about their estimates of the expected discipline. In both questions, the respondents could have selected one of the six possible answers: 1 = “none” [no discipline], 2 = “verbal reprimand,” 3 = “written reprimand,” 4 = “period of suspension,” 5 = “demotion in rank,” 6 = “dismissal.” For our multivariate models, we recoded the expected discipline into three categories: no discipline (1 = “no discipline” or 2 = “verbal reprimand”), some discipline (3 = “written reprimand,” 4 = “suspension,” and 5 = “demotion in rank”), and dismissal (6 = “dismissal”).

We have also included another organizational variable that measures police officers’ estimates of most police officers’ willingness to report. The question asked the respondents to predict how likely most police officers in their police agency would be to report misconduct. The respondents could have selected one answer from a 5-point Likert-type scale, ranging from 1 = “definitely would not report” to 5 = “definitely would report.”

Individual Independent Variables

We have included several variables measuring respondents’ demographic characteristics into the questionnaire: their length of service, gender, assignment, supervisory status, and education.

Analytic Strategy

The analytic strategy for this chapter proceeds in two phases. In the first phase, we look at the contours of the code of silence across each of the scenarios. Specifically, we look at the distribution of the code of silence for each of the scenarios. In the second stage of the analyses, we build multivariate regression models that can predict the effects of the organizational independent variables on the code of silence, net of the other independent variables in the model.

Unlike larger datasets that can include many independent variables, we are limited by our sample size (N = 148). Logistic regression models require quite a bit of data to yield consistent and stable results. This is because the maximum likelihood estimator that is at the heart of the logistic regression makes assumptions about asymptotic normality, which is frequently violated without sufficient data (Ngunyi et al., 2014). While several recommendations have been developed for determining how many cases per variable are needed to include in a model, generally 20 observations per independent variable are required (Tabachnick & Fidell, 2019). In this case, our analyses would be limited to seven independent variables.

However, recent advancements in computational processing power allow scholars to make use of alternative estimation techniques in the family of machine learning, notably penalized regression models (Dezeure et al., 2015). These models add an extra constraint to the estimation equation such that the addition of more variables is more difficult (Friedman et al., 2010). We utilized LASSO regression models—one of the three primary forms of penalized regression (i.e., ridge regression, LASSO regression, and elastic net regression). The LASSO stands for the least absolute shrinkage and selection operator that estimates a penalty term (λ), which is a sum of all of the absolute coefficients in the model (Friedman et al., 2010). The benefit of the LASSO model is for the situations in which there is a high ratio of k (number of independent variables) to n (number of observations). Thus, the LASSO regression approach will allow us to use all of the information to develop the best subset of predictors to explain the dependent variable (Kammer et al., 2020). We use an extension of the LASSO model that allows for causal inferences to be made from the data, using the xpologit package in Stata 16.

At all stages of the multivariate analyses, we include all the independent variables and control variables that could be associated with a person’s willingness to report misconduct. Given the design of the questionnaire, which seeks to promote honest reporting among officers by minimizing the risk that they could be identified through their survey responses, we have a limited number of control variables to select from. Specifically, we include gender (“female” vs. “male”), assignment (“patrol officers” vs. “other assignments”), length of employment, education (as a 4-level ordinal variable from high school, associate’s degree, bachelor’s degree, or master’s degree or higher), and supervisory status (“yes” vs. “no”). The LASSO estimator will consider whether all variables in the model, independent and control, bring additional explanatory power to the model. In the tables below, there are no parameter estimates for these control variables as they do not yield additional explanatory power beyond the independent variables because they increase the value of the penalty term, λ.

Results

Contours of the Code of Silence

The respondents’ expressed willingness to report misconduct varies greatly across the scenarios (Table 2.1), suggesting that the code of silence does not protect all misconduct equally. On the one hand, the respondents were least likely to say that they would protect a theft from a burglary scene, shooting the suspect in the back, and a supervisory failure to stop the beating (Ranks 1 to 3 in Table 2.1). The percent of the respondents who would protect such misconduct in silence is less than 20% and, for the two scenarios, actually well below 10% (Adhering to the Code of Silence in Table 2.1). On the other hand, the respondents were most likely to say that they would not report a colleague who yelled at co-workers, falsely reported being sick, and accepted gifts (Ranks 10–12 in Table 2.1). Over two-thirds of the respondents said that they would cover up such behaviors (Adhering to the Code of Silence in Table 2.1). The respondents’ willingness to report misconduct is closely related to their perceptions of how serious the misconduct is (Table 2.1): there was a very strong correlation between the ranking of scenarios based on the respondents’ willingness to report and the ranking of scenarios based on the respondents’ assessments of scenario seriousness (Spearman’s rho = .993, p < .001).

Table 2.1 Contours of the code of silence
Full size table

There is also a substantial variation in the respondents’ expressed willingness to report within each type of misconduct. Within police corruption , stealing from a crime scene was expected to be much less protected by the code than accepting gratuities (Table 2.1). While about one-half or more police officers would protect the acceptance of gratuities (65%) and running errands for a supervisor (50%), only a small proportion of police officers would tolerate a theft from a crime scene in silence (5%). There is also a substantial variation within the three scenarios describing the abuse of force . In particular, the majority of the respondents would cover a police officer who verbally abused a citizen (57%), but only less than 20% would protect a supervisor who did not stop the beating and less than 10% would not report a fellow officer who abused deadly force (Table 2.1). Within scenarios describing organizational deviance , only a minority of the officers would protect false overtime reporting in silence (37%), while about one-half of the respondents (52%) would cover a fellow officer’s DUI accident and the overwhelming majority (73%) would protect a fellow police officer who falsely reported sick. The variation existed in the scenarios of interpersonal deviance as well; a minority of the respondents (33%) would not report an officer who was telling sexist jokes, but the majority would not report an officer who was spreading false rumors in the agency (58%) or yelling at co-workers (86%).

The Effects of Police Integrity Measures on the Code of Silence

We next looked at the multivariate results from the LASSO models to explain what factors affect officers’ adherence to the code of silence. To aid in the interpretation of the results, we group the results based on the type of misconduct (i.e., corruption, excessive force, organizational deviance, and interpersonal deviance).

Police Corruption

The results from all scenarios dealing with police corruption are generally consistent with prior police integrity literature (Table 2.2). To begin with, respondents who believe that other police officers are more likely to adhere to the code of silence are also more likely to say that they would adhere to the code themselves. However, this effect is not consistent in magnitude across all types of scenarios, whereby the effect of perceived others’ adherence to the code is the weakest—yet still significant and rather large in magnitude—for receiving gifts and the strongest for supervisory corruption.

Table 2.2 The code of silence and police integrity theory estimates
Full size table

Likewise, we see that either a recognition that a particular act is a violation of policy (odds ratio (OR) = 0.42, p < .001) or an evaluation of an act as serious (OR = 0.36, p < .001) is negatively associated with the code of silence. The direction of the effects for these two variables is consistent for supervisory errands, but only significant for recognition that this act is a violation of policy (OR = 0.66, p < .05). We cannot estimate these parameters for the situation depicting the theft from a burglary scene because the act is almost universally noted as a violation of policy and something that is seen as very serious by our respondents. Yet, some officers report they will adhere to the code of silence even in this case (OR = 43.88, p < .001).

We also explored the effect of perceptions of discipline severity on the police officers’ adherence to the code of silence. The results suggest that in all three corruption scenarios, compared to the officers who indicated that some intermediate discipline is expected, officers who expected no discipline were more likely to say that they would adhere to the code of silence, with the effect being most pronounced for the scenario depicting theft from a burglary scene (OR = 16.59, p < .001). This suggests that officers who expect no discipline are unlikely to come forward and report these sorts of behaviors. On the other hand, the respondents who expected dismissal were only significantly more likely to say that they would adhere to the code of silence in the case depicting supervisory corruption (OR = 2.85, p < .001).

Use of Excessive Force

Next, we look at the results for the scenarios depicting situations of excessive force , also presented in Table 2.2. Again, in all three scenarios, we see that the belief that others will adhere to the code of silence is a robust predictor of officers saying that they would personally adhere to the code of silence. Furthermore, the magnitude of this effect across all three excessive force scenarios is among the strongest effects for this variable across all scenarios in this chapter.

Additionally, in two of the scenarios we see that officers who acknowledge that these situations are a violation of department policy are less likely to say that they would adhere to the code of silence, except for the situation depicting the supervisor who failed to stop a beating, although this was not significant (OR = 1.18, p > .05). Universally, officers who think these sorts of situations are serious are less likely to say that they would adhere to the code of silence. The magnitude of the effect corresponds with the general rank seriousness of the events. In other words, the strongest effect of seriousness is seen in the situation involving the most serious conduct (i.e., shooting a suspect in the back; OR = 0.01, p < .001), followed by failing to report a beating (OR = 0.05, p < .001), and it is the lowest in magnitude for verbally abusing a citizen (OR = 0.31, p < .001).

Turning to the effect of discipline severity, the pattern of findings is complicated. Notably, compared to the police officers who expected some intermediate discipline, officers who expected no discipline seemed to have a lower likelihood of adhering to the code of silence for the scenario depicting the use of deadly force (OR = 0.01, p < .001), whereas they have a higher likelihood of adhering to the code of silence for the scenario depicting a verbal abuse of a citizen (OR = 1.98, p < .01). There was no significant effect for the scenario depicting the beating of a suspect (OR = 0.60, p > .05). Conversely, compared to the respondents who selected some intermediate discipline, the respondents who selected dismissal were more likely to say that they would adhere to the code of silence in the scenario with the verbal abuse of a citizen (OR = 10.29, p < .001) and in the scenario describing a failure to report a beating (OR = 3.17, p < .01).

Organizational Deviance

Next, we turn to the results for the scenarios depicting organizational deviance presented in Table 2.2. Across scenarios, consistent with prior research, feeling that others will adhere to the code of silence is a consistent, significant, and strong predictor that an officer will adhere to the code of silence in all three scenarios. Additionally, we see the familiar pattern that those officers who find the actions depicted in these scenarios as more serious are less likely to say that they would adhere to the code of silence. Recognizing the actions as a violation of departmental policy shows different effects depending on the scenario. For the scenarios of covering up the DUI crash of a fellow officer (OR = 0.75, p < .05) and false sick report (OR = 0.74, p < .05) the recognition that this is a policy violation decreases adherence to the code of silence. However, recognizing that filing a false overtime report is a violation of department policy is positively related to adhering to the code of silence (OR = 2.45, p < .001).

Unlike prior scenario types, there is only one scenario where the type of expected discipline exerts a significant effect. Notably, we see that, compared to the officers who expected some intermediate discipline, officers who expected no discipline (OR = 27.71, p < .001) were more likely to say that they would adhere to the code of silence in the scenario depicting the false sick report. Additionally, compared to the officers who expected some intermediate discipline, officers who expected dismissal were less likely to say that they would adhere to the code in case of an officer who falsely called in sick to work (OR = 0.25, p < .05). There are no significant, or even consistent, patterns of findings for the expected discipline variables in the other two organizational deviance scenarios.

Interpersonal Deviance

Finally, we look at the results from interpersonal devianc e, shown in Table 2.2. We see the familiar pattern that the anticipated other officers’ adherence to the code of silence is consistently, significantly, and positively related to their own willingness to adhere to the code of silence. We also see that officers who perceive these situations as serious are universally less likely to say that they would adhere to the code of silence. The effect for the variable recognizing the act as a violation of department policy does not yield consistent results: one scenario shows a negative effect (spreading false rumors about coworker, OR = 0.37, p < .001), one has a positive effect (yelling at coworkers, OR = 2.73, p < .001), and one has a null effect (telling sexist jokes, OR = 1.22, p > .05).

The results for discipline again present a complicated pattern of results. We see that, compared to the officers who expected some intermediate discipline, officers who expected no discipline were more likely to say that they would adhere to the code of silence in all three scenarios (Table 2.2). The same pattern of results is seen for dismissal compared to an intermediate discipline, although the effect is only statistically significant for the scenario depicting the spreading of false rumors about a coworker (OR = 8.47, p < .05).

Conclusion

Our results show that the code of silence exists among the police officers in the police agency we study. Such a finding should not be surprising, having in mind that the code of silence is a universal phenomenon that the police integrity research was able to document since the 1990s (e.g., Klockars et al., 2000, 2004a, b; Kutnjak Ivković & Haberfeld, 2015a, b, 2019). A more pressing question is how strong the code of silence is in the police agency included in our case study. Apart from two scenarios describing blatant examples of not only police misconduct, but serious criminal conduct, at least 20% of police officers would protect in silence all other examples of police misconduct in our questionnaire. While this percent may not look high, it very vividly indicates that the code of silence is not only present, but also strong. Whereas this percentage may not seem overwhelming, one-fifth of a police agency could constitute a large number of police officers, and even a single officer’s engagement in misconduct, particularly when it involves the abuse of deadly force and/or racism, may generate strong reactions, including riots, destructions, and a tremendous pushback against police legitimacy and integrity.

Our results show that the code of silence is not a flat prohibition of reporting and that the police officers’ expressed willingness to report is strongly related to how serious they perceive misconduct to be: the more serious they evaluate the misconduct, the less likely they are to say that they would protect it by the code of silence. Although the new police integrity questionnaire has expanded the types of police misconduct included in the questionnaire, our results fit well with the already established line of research focusing primarily on police corruption and the use of excessive force, demonstrating the interconnectedness between the perceptions of seriousness and the expressed willingness to report misconduct (Cheloukhine et al., 2015; Haberfeld, 2004; Hickman et al., 2016; Khechumyan & Kutnjak Ivković, 2015; Klockars et al., 1997, 2004a, b, 2006; Kutnjak Ivković et al., 2013, 2018; Kutnjak Ivković & Khechymian, 2013; Kutnjak Ivković, Peacock, & Haberfeld, 2016; Kutnjak Ivković & Sauerman, 2013; Kutnjak Ivković, Haberfeld, Cajner Mraović, et al., 2019; Lobnikar & Meško, 2015; Long et al., 2013; Lim & Sloan, 2016; Maskály et al., 2019; Pagon & Lobnikar, 2000; Peacock et al., 2020; Porter & Prenzler, 2016; Vallmüür, 2015; Westmarland, 2006; Wu & Makin, 2019).

Perceptions of how other police officers would react—whether they would adhere to the code of silence or not—are one of the strongest and definitely most consistent predictors of the respondents’ own determination to adhere to the code of silence. Our findings of an influential organizational culture are consistent with prior research about the code of silence that has demonstrated the close connection between our own anticipated behavior and the expected behavior of the peer group (Hickman et al., 2016; Kutnjak Ivković et al., 2018; Kutnjak Ivković, Haberfeld, Cajner Mraović, et al., 2019; Lim & Sloan, 2016; Long et al., 2013; Peacock et al., 2020; Van Droogenbroeck et al., 2019). Indeed, police officers feel that they are a part of the police culture and they seem dependent upon what they anticipate that their peers would do.

The effects of the expected discipline severity are uniform neither across all types of misconduct nor within each type of misconduct. On the one hand, in the scenarios describing police corruption and interpersonal deviance, expecting no discipline compared to some intermediate discipline increased the likelihood that officers would say that they adhere to the code of silence. On the other hand, such a relationship was not as prominent or even in the same direction for some of the scenarios describing the use of excessive force and organizational deviance.

Similarly, when we analyzed the effect of discipline severity within each type of misconduct, we found, for example, that, compared to intermediate discipline, dismissal was linked with higher adherence to the code of silence for the scenarios describing an instance of verbal abuse of a citizen and for the scenario describing a failure to report a beating, but not for the scenario describing the abuse of deadly force. These findings are particularly salient given the national conversation around the police use of force, especially deadly force. These results could represent a backlash to the enhanced scrutiny that police officers are currently experiencing. Also, it is troubling that disciplining this type of behavior more seriously could have deleterious effects on officers’ willingness to report this type of misconduct.

While the severity of the expected discipline seems to be an important factor for the respondents’ decision whether to adhere to the code of silence in a number of scenarios, our present analyses do not reveal how fair the respondents evaluated this expected discipline and, even more importantly for the purposes of our study, whether their perceptions of discipline fairness are related to their willingness to report. In the next chapter, we turn our attention to the issue of discipline fairness and its effect on the police officers’ determination to adhere to the code of silence.