Does prior misconduct predict drug use by police?

ABSTRACT

Recent research has yielded important insight into the characteristics and behaviours of officers who commit misconduct across the UK, USA and Australia. However, there remains little research into illicit substance use by sworn police. Considering substance use by police through the lens of police misconduct, this paper employs correlation and regression analyses using a sample of 489 sworn officers, to analyse whether individual types of misconduct, or the overall volume of misconduct committed, were predictive of an officer returning a positive drug test during departmental random drug testing. Findings suggested that neither individual types of misconduct, or the overall number of complaints received by officers were useful in predicting a positive drug test. However, a current or historical association with a known offender held a strong relationship with returning a positive drug test. Pivotally, holding a second job (outside of the agency) was also a significant predictor of a positive drug test. The findings of this research hold important implications for drug testing policy and procedure of policing agencies, and for understanding which factors may suggest that an officer is at risk of drug use.

KEYWORDS:

• Misconduct
• drug testing
• corruption
• mental health

Introduction

Recent research suggests that prior misconduct, or criminality outside of the workplace, may be instructive in assessing the risk of career ending misconduct by police (Greene et al. 2004, Kane and White 2009, White and Kane 2013, Donner 2019, Cubitt et al. 2020, Cubitt and Birch 2021, Cubitt et al. 2022, Simpson and Kirk 2022, Cubitt 2023). While it is increasingly accepted that analysing the misconduct history of officers may help police agencies measure the likelihood of generalised career-ending misconduct, this approach has not yet been applied to specific types of serious misconduct, including the use of illicit substances by serving police. Given the powers afforded to police, and the historically prohibitive legislation enforced for drug use in the community, drug use by police is important to understand. The present research therefore considers whether prior instances of misconduct are predictive of an officer testing positive for an illicit substance during departmental random drug testing protocol.

Drug use by police is prohibited by law enforcement agencies around the world. In countries such as England, the United States of America, and Australia it is typically considered to be serious misconduct, the type that could result in dismissal. Pivotally, when analysing serious police misconduct, prior literature tends to aggregate all types of misconduct that may result in dismissal (Kane and White 2009, White and Kane 2012, Cubitt et al. 2020). These studies frequently find that officers who fit certain characteristics, or have a history of certain specific types of misconduct, are prone to general serious misconduct. However, there is little information on whether these correlates hold true when analysing specific types of serious misconduct. Given that drug use by police is viewed as misconduct by Western policing agencies, and is subject to prevention and management approaches, the present research seeks to understand the correlates and predictors of officers returning a positive random drug test to help inform and refine these approaches. In undertaking this task, we first review the literature on police misconduct, and particularly career ending misconduct, before going on to consider what we currently know about drug use by serving police.

Literature review

Research into police misconduct

Given the historical lack of variation in individual characteristics and background among police, demographic correlates of police misconduct have been difficult to tease out (Grant and Grant 1996). However, male officers are generally considered more prone to misconduct than women (Greene et al. 2004, McElvain and Kposowa 2008, Kane and White 2009, White and Kane 2013, Gaub 2020), while younger officers are also thought to pose a higher risk of misconduct (Greene et al. 2004, McElvain and Kposowa 2008, Rojek and Decker 2009, White and Kane 2013). However, employment features may also be instructive, both prior to joining and while employed as a police officer. For example, officers with a history of organisational discipline appear at an elevated risk of misconduct (Greene et al. 2004, Kane and White 2009, Donner 2019, Simpson and Kirk 2022), while higher rank is associated with lower risk of committing misconduct (Hickman et al. 2004, Kane and White 2009, Gaub 2020). Officer tenure is more complex, with some finding longer tenure at a department to be a risk factor (McElvain and Kposowa 2008, Wolfe and Piquero 2011, Cubitt et al. 2020), and others finding it to be protective (Kane and White 2009, Simpson and Kirk 2022). Recent research has also supported the notion of peer and network effects in promoting misconduct (Quispe-Torreblanca and Stewart 2019, Wood et al. 2019, Cubitt 2021).

When considering the relationship between broad categories of prior workplace deviance and career ending misconduct, White and Kane (2012) found that citizen complaints of misconduct held a significant association with, not only career ending misconduct, but an officer being fired within a short timespan after that complaint. Further, Kane and White (2009) found that officers who had a history of workplace deviance, or disciplinary action in prior careers, were associated with serious misconduct. Pivotally, Kane and White (2009) are among the only examples of research to analyse positive drug tests by police. While they were only able to analyse prior disciplinary histories as broad categories, rather than individual types, they found that a history of employment discipline held a small but statistically significant relationship with failing a subsequent drug test. This research suggested that histories of misconduct were important to understand, but also that there may be potential value in measuring more granular detail on individual misconduct types.

More recently, research into police misconduct has focused on individual domains, or types of non-career ending misconduct, and whether they might suggest that an officer is likely to be ongoingly misconduct prone. For example, Donner (2019) found that prior minor misconduct held a relationship with further workplace deviance, specifically misconduct that featured the misuse of police powers, records, and vehicles. Similarly, Cubitt et al. (2022) found that the misuse of authority or powers, harassment and vilification, and improper use of force held a strong relationship with career ending misconduct. Other research has considered an array of individual types of misconduct, finding that their relationship with career ending misconduct varied depending on the severity of the prior misconduct (Cubitt et al. 2020).

Separately, organisational or environmental factors appear to have some influence on misconduct. Huff et al. (2018) found that organisational protocols around vetting, and accountability policy were associated with lower levels of misconduct, while Boateng et al. (2021) found that officers employed by small agencies were less likely to commit violent crime. Recently, having a presence on police databases, either for prior criminal charges or intelligence relating to potential criminal behaviour, has been shown to be predictive of career ending misconduct (Cubitt 2023). Factors such as structural disadvantage, and population mobility may also hold some interaction with police misconduct (Kane 2002).

Drug use by sworn police

Internationally, substance use has been acknowledged among law enforcement officers for several decades (Dietrich and Smith 1986, Micezkowski and Lersch 2002, Gorta 2009). However, most research focusing on drug use by police is not empirical, and emerges from countries with more liberal drug laws, such as the USA and Canada (Hathaway and Erickson 2003, Pardo 2014). In Australia, focus turned to officer use of illicit substances during the Royal Commissions into police corruption in the mid-1990s (Fitzgerald 1989, Wood 1997). The findings of these inquiries, and similar inquiries throughout Australia, resulted in the majority of law enforcement agencies implementing drug and alcohol testing policies by the early to mid-2000s, in which every sworn officer is subject to random drug testing procedures.

While there can be little doubt that a proportion of police use illicit drugs, research into the characteristics of these officers, and whether they reflect officers who commit other types of career ending misconduct, is limited. What research there is suggests that the high-stress nature of policing may place officers at an increased risk (Smith 2007). It is possible that the police subculture of socialising with alcohol as a form of stress reduction, and the noted link between this culture and officers’ sense of loyalty (Dietrich and Smith 1986), may also be transferrable to illicit substance. In 2005 an Australian police oversight body, the Police Integrity Commission (PIC), interviewed academics and practitioners, including 21 Police Superintendents, regarding substance use by police. The findings of their research suggested that officers may treat substance use as a way to relieve stress during burnout (Police Integrity Commission 2005). Their report concluded that substance use featured, at least in part, as a coping mechanism for stress. A relationship well-established in literature on other populations (Willis and Hirky 1996, Sinha 2001, 2008).

It has long been acknowledged that police may be at higher risk of mental health issues, particularly trauma-related issues, than other groups within the community by virtue of their duties (Foley and Massey 2018). While, separately, experiences of trauma emerging from law enforcement duties have been associated with maladaptive coping strategies such as substance use (Cross and Ashley 2004). Although debriefing and social support are known mitigating factors, they only serve to marginally diminish an already elevated risk of traumatic stress among officers (Carlier et al. 2011). Given the noted co-occurrence of traumatic stress injuries and substance use (Jacobsen et al. 2001), it is important to note the potential for the influence of mental health issues in substance use by police.

Criminological theory and officer substance use

While we review the literature on police misconduct, as it is considered within policing agencies to be a form of misconduct or deviance, the theoretical approach to this phenomenon may differ. Early research into misconduct often interpreted findings through the lens of rotten apples and rotten barrels, to refer to individual misconduct prone officers, and misconduct prone peer groups respectively. The notion of rotten apples emerged from the United States, Sherman (1974) offered several typologies of police corruption, suggesting that a small number of officers engage in corruption, which tends not to substantially impact on the community or the policing agency. Expanding on this idea, Sherman (1978) discussed the ‘rotten barrel’, which was then extended by Punch (2003) to characterise groups of officers who commit misconduct. However, while these approaches are useful in interpreting the structure of misconduct when it occurs individually or in groupings, it is limited in its explanation of the factors contributing to officer deviance.

More recent research on misconduct extends this inquiry to peer influences on misconduct, including organisational explanations, finding some support for the idea that misconduct is not only concentrated among a small number of officers, but those officers have a contagious effect on others (Ouellet et al. 2019, Quispe-Torreblanca and Stewart 2019, Cubitt 2021). Each of these theoretical lens’, in part, lend themselves to understanding why officers might use illicit substances. Whether it is the association with a maladaptive workplace culture, increased occupational stress and the strain, or an environment that implicitly rewards acts of misconduct.

Approaches such as opportunity (Kappeler et al. 1998), personality typifications (Niederhoffer 1967), ecological effects of neighbourhoods (Klinger 1997, Kane 2002), and officer impulsivity and reduced deterrence (Pogarsky and Piquero 2004) have all been employed to explain misconduct. However, social learning theory (Akers 1998) and general strain theory (Agnew 1992, 2006) remain the most commonly used theoretical frameworks in this field. Social learning theory builds on the foundational ethnographic research of policing (e.g. Bittner 1970, Westley 1970). These studies tend to focus on the occupational norms, informal code, and general police culture, all of which are suggested to rewarded misconduct and corruption (Skolnick 2002, Chappell and Piquero 2004).

However, the relationship between substance use and stress suggests that strain theory may be instructive in understanding drug use by police. According to Agnew (1992), there are three main components of general strain theory; strain is most commonly attributed to failing to achieve positively valued goals, it is caused when valued stimuli are withdrawn, and may result from negative stimuli being presented or often anticipated. Pivotally, since its inception, general strain theory has been used to explain many types of deviant behaviour (Agnew and DeLisi 2012), including substance use (Stogner and Gibson 2011). Stogner and Gibson (2011) found that strain that was associated with health problems held a relationship with the onset, and at times frequency of substance use. Bishopp and Boots (2014) extended this idea to suicidal ideation, employing a strain theory approach to find that the stress of exposure to traumatic events held an association with suicidal ideation among police. While peripherally associated, together these findings point to a potential relationship between strain associated with health issues, such as trauma exposure, and the possible onset of substance use.

Law enforcement is a stressful occupation (Adams and Buck 2010), however it is not only the exposure to traumatic events or violence during operational policing duties that have been found to hold a relationship with strain and deviance. Rather, certain types of strain inherent in organisational policing environments may influence deviance. Bishopp and colleagues (2016) demonstrated a relationship between organisational stressors emerging, not only from exposure to operational policing, but also from fatigue and at times the prospect of internal investigations. Misconduct appears to be one of several possible responses to the occupational stress and burnout associated with policing (Paoline 2004, Bishopp et al. 2020). A conclusion supported by the findings from Australian police oversight bodies (Police Integrity Commission 2005), and by the limited research available featuring self-reported data from officers, including those found to have used illicit substances (Cubitt and Judges 2018).

General strain theory appears to be central to the limited research available considering theoretical applications to police misconduct. The present research considers drug use by sworn police, a more specific type of misconduct than has been considered among prior studies implementing theoretical lens’ to this phenomenon. However, the literature suggests that stress, and strain, may be central factors in maladaptive or deviant behaviour among police, while separately also suggesting that there may be a relationship between strain and the onset of substance use. It is therefore important to consider whether strain may not only apply to more general police misconduct, but also to more specific maladaptive behaviours, including drug use by police.

The purpose of this research

While frameworks have been developed by researchers such as Kraska and Kappeler (1988) to identify and evaluate substance use among law enforcement, these approaches do not consider whether drug use may be preceded by a history of misconduct. At present, each policing agency in Australia implements a random drug testing protocol. The purpose of the present research is to understand whether, among the administrative data available to policing agencies, there are consistent antecedents to an officer returning a positive random drug test. In particular, if a series of minor misconduct consistenly precedes such a positive test. If a positive random drug test tends to co-occur with a consistent series of antecedents, there may be important implications for policing agencies, in particular for intervention and tailored misconduct prevention activities which focus on substance use.

Importantly, the data used here do not allow us to measure whether misconduct is causally linked to the drug use, or vice versa. Rather this study considers the correlates of a positive random drug test. The positive drug test is used as the outcome in this research principally because it is the detection point for police agencies, and signals an important intervention point. If police are able to better anticipate drug use by officers prior to a positive drug test, they may be able to better implement an intervention featuring support mechanisms. However, at present, there has been no consideration of whether administrative data held by police could assist in understanding risk of a positive drug test by an officer. Among the data held by police, misconduct records may be pivotal in this task. The present research therefore considers two key questions; (1) whether the complaint history of an officer could be predictive of a subsequent positive departmental drug test during random testing, and (2) whether there are implications among these findings that could help improve drug testing policy among police agencies.

Methodology

A note on Australian jurisdictions and drug testing protocol

Australian police operate in a unique context. Compared with the UK, where there are 45 police forces (Police UK 2024), and the United States, where there are more than 17,000 local and State agencies with law enforcement functions (Gardner and Scott 2018), there are eight state or territory agencies, and one federal police within Australia (Barker 2019). This environment means that each individual police agency in Australia is responsible for metropolitan, suburban and regional areas, and tailoring their approach to account for each. These duties are typically undertaken by between 200, and 350 operational police per 100,000 community members with the exception of the Northern Territory where the numbers are notably higher (Productivity Commission 2023). Every law enforcement agency in Australia has a policy position against the use of any illicit substances, including androgenic steroids, and the abuse or misuse of pharmaceutical substances (AFP 2023, NSW Police Force 2022, Police Force (Member Testing) Regulations 2011, Police Service Administration (Alcohol and Drug Testing) Amendment Bill 2003). All random drug tests that violated these policies during the reference period were included for analysis.

Among Australian policing agencies, testing positive for an illicit substance is considered serious misconduct, and grounds for dismissal. These procedures emerged from the recommendations of formal government inquiries into policing. Sworn officers are required to submit to random alcohol or drug testing when requested, any time, 24 hrs a day, 7 days per week. Any officer selected to be tested is required to submit a urine sample, if an officer refuses to submit a sample upon random selection, they may be subject to the same disciplinary action as those who have tested positive. While this protocol exists for alcohol alongside drug use, officers who returned a positive alcohol test were excluded from the present analysis.

Sample for this research

Data for this research were available from one policing agency in Australia, the agency has been de-identified. Data reflected officers who had tested positive for the use of illicit substances between 2005 and 2017. All officers who had tested positive, and had completed the administrative dismissal process, were included for analysis, resulting in a sample size of n = 89 officers. In other words, if an officer had tested positive for an illicit substance, however the departmental misconduct procedure managing them had not yet been completed, they were not included in this research. It is possible that an officer may retain their position after returning a positive random drug test, however all those included in the present sample had either resigned during the investigation or misconduct management process, or had been dismissed. To form a comparison group, and to fit the analytical methodology, n = 400 officers who had not tested positive for illicit substances were sampled. Comparison officers were selected on the criteria that they fit the same distribution of calendar years in which they finished their academy training, sex, and age as the group who tested positive to illicit substances.

Data description

This research intended to undertake an empirical analysis of the correlates of drug use among police, with the purpose of considering whether these events can be better anticipated, and whether there are opportunities for focused policy responses introducing intervention and support mechanisms. For this reason, the data included for analysis principally related to data available within policing agencies, with a focus on complaints against police. This study used a de-identified secondary dataset featuring information on sworn police who had tested positive for illicit substances, and a comparison group who had not. The outcome variable in this research referred to whether an officer had tested positive for an illicit substance (binary variable). Demographic variables included sex (binary variable), rank (categorical variable), and their duty type (categorical variable). While the mean age of officers, their mean lengths of service and years spent at final rank are reported in the summary statistics, because positive drug tests were not equally distributed across temporal intervals, and in order to allow for greater insight into at-risk time periods of officers, for the regression analysis these variables were broken into categories. For age, these categories reflected each 10 years, i.e. 20–29, 30–39 and so on, and for length of service, and years at final rank, it was each five years, i.e. 0–4, 5–9 and so on.

Information on the complaint histories of officers were also included. For each individual type of complaint that had been substantiated against an officer, a continuous count variable was included (See Table 2 in the results section for list of included substantiated complaint types). Separately, variables reflecting the total number of complaints (continuous variable) and the number of remedial management actions taken in response to complaints (continuous variable) were included. Given recent research finding a relationship between the volume of unsubstantiated complaints received by an officer and risk of serious misconduct (Cubitt et al. 2020), a continuous variable reflecting the number of unsubstantiated complaints received by officers was included. Finally, a variable was included reflecting whether each officer had previously been considered for serious misconduct, defined by a minimum standard of being considered for dismissal. Among those who had tested positive for an illicit substance, while some had prior instances of serious misconduct, none of their prior misconduct was for either substance use or possession.

Analytical approach

Given the amount of data available, and limited prior research in this area, the present research is best considered exploratory. It was first important to compare the available data. Summary statistics were provided for each variable and a simple correlation analysis was then undertaken between the complaint variables available, and positive drug tests. This correlation analysis had two purposes. Correlations were computed to determine the relationship that individual types of complaints held with a positive drug test, and to help measure the presence of collinearity for subsequent modelling. Given the differing structure of variables, a heterogeneous correlation matrix was produced (Brown and Benedetti 1977, McGrath and Meyer 2006, Babchishin and Helmus 2016). Variance Inflation Factor (VIF) was also used to test for collinear variables (Thompson et al. 2017). Prior research suggesting that a threshold VIF value of 5 signalled collinear variables, however recent research has warned against setting a threshold VIF value for establishing multicollinearity (O’brien 2007, Kalnins and Hill 2023). As such, while a rule of thumb was not used to benchmark collinearity, the threshold value was broadly considered to indicate the presence of collinear variables within the analysis.

While the sample of officers who had tested positive for using illicit substances was not large, it constituted every officer who had across a 12-year period from this agency. The logistic regression is a commonly employed statistical method using police data when analysing misconduct (Kane and White 2009, Gaub 2020). Given the sample size available, and the research questions, this research estimated a logistic regression using the binary outcome variable of a positive test for illicit drug use. The previously described demographic, and complaint related variables were included as independent variables in the analysis, with the exception of the variables used to match the groups (the age and sex variables), and those identified through VIF analysis as introducing multicollinearity. Prior to proceeding with the model diagnostics, we used the Box-Tidwell test (Osborne 2015) to evaluate the linearity of the continuous independent variables and log odds, each of these variables met the assumption, and the analytical process progressed. When computing the logistic regression, the reference categories were selected in view of the prior literature. For example, prior literature on police misconduct suggests that being early career and inexperienced were important, while working patrol duties also holds a relationship with misconduct. As such, reference categories for required variables focused on junior ranks, lengths of service reflecting inexperience, and patrol duties. This procedure allows for an understanding of whether certain complaint types were predictive of drug use, or indeed whether officers at certain career stages were more associated with drug use.

A Receiver Operating Characteristic (ROC) curve was used to identify the accuracy of the logistic regression, through the Area Under the Receiver Operating Characteristic (AUROC) curve. The ROC curve identifies the true positive rate of classification (y-axis), compared with the false positive rate (x-axis) at any threshold value. The AUROC, which we refer to in simple terms as the accuracy of the model, represents the probability that a randomly selected case will be accurately classified. These analyses were undertaken using the statistical analysis software, R version 4.2.0, and the ‘dplyr’, ‘pROC’, ‘PerformanceAnalytics’, ‘car’ and ‘ggplot2’ packages.

Limitations

The de-identified agency from which these data emerge randomly tests for alcohol intoxication via breath test, through the same randomised methods as drug testing. While it may have been beneficial include these data for analysis, and to compare officers who tested positive for alcohol intoxication to those who tested positive for illicit substances, ultimately these data were not available. Previous research in this area identified noteworthy predictors of alcohol and other drug use, including education, income, employment, and marital status (Parker et al. 1995), it is likely that this information could have been useful in the present research however none of these variables were available for analysis.

Data entry error is a consistent limitation when using information emerging from policing agencies around the world. Data entry is typically undertaken manually by time-poor staff, resulting in considerable risk for inaccuracy. Data cleaning, and cross referencing was undertaken as a component of this research, prior to analysis, however inaccuracies remain possible. It is also a limitation of these data that they are observational and therefore they cannot be used to evaluate causality. While we cannot measure whether any of the data in this research directly contributed to the positive drug test, we can consider the antecedents or a positive drug test, and the strength of correlation held by the available data. However, it is important to acknowledge that while we measure and discuss correlates, we cannot comment on causal relationships with a positive drug test by an officer.

Results

Summary statistics and correlation

Officers who tested positive for illicit substances were typically men in their early 30’s, due to the matching criteria the same was true of the comparison group. Table 1 suggests that the most common rank in each group was Senior Constable, followed by Constable, however a greater proportion of the comparison group held more senior ranks than in the positive drug test group, and were more likely to be undertaking general duties. Importantly, a considerably greater proportion of the positive drug test group held secondary employment (outside of the police) than the comparison group (22.2% vs 7.6%). They were also more likely to accrue a larger total number of complaints, and to experience more remedial management action in response to minor instances of misconduct.

Table 1. Summary statistics of the positive drug test and comparison groups.

Demographics		Positive drug test group (n = 89)	Comparison group (n = 400)
Age (years)		33.1 (22–61)	34.2
Sex (%)	Male	92.2	91.8
	Female	7.8	8.2
Rank (%)	Probationary	8.9	1.5
	Constable	33.7	21.0
	Senior constable	53.9	59.5
	Sergeant	1.8	14.5
	Commissioned officer	1.8	3.5
Duty type (%)	General (patrol) duties	68.5	54.3
	Specialist duties	31.5	45.7
Years at rank (years)		5.1 (1–19)	6.28 (1–21)
Length of service (years)		10.13 (1–27)	12.9 (1–34)
Secondary employment (%)		22.2	7.6
Total number of complaints		7.9	5.6
Number of unsubstantiated complaints		3.8	3.6
Instances of management action		2.6	1.8

Among the group who tested positive for illicit substances, the most common class of drug reported to be present in the test was stimulants, in 64 percent of tests. This was followed by cannabis, present in 43 percent of tests. While there were 3 separate cases in which anxiolytic medication, opioids or Gamma-hydroxybutyric acid were present in the tests. Given these proportions, it is important to note that there were some officers who tested positive for more than one substance.

With the exception of complaints relating to mishandling property or exhibits, and disobeying directions from a superior, the positive drug test group appeared to engage in more misconduct across each complaint type (Table 2).

Table 2. Number of substantiated complaints accrued per officer for the positive drug test and comparison groups.

Complaint type	Positive drug test group (n = 89)	Comparison group (n = 400)
Prior serious misconduct	0.20	0.07
Drink driving	0.06	0.03
Firearm misuse	0.08	0.02
Improper association	0.09	0.01
Performance at court	0.29	0.22
Allegation of corruption	0.03	0.01
Issue with investigation	0.36	0.30
Customer service	0.63	0.36
Harassment or bullying	0.12	0.11
Untruthfulness	0.24	0.11
Misuse of powers	0.09	0.08
Disobey direction of superior	0.07	0.08
Unauthorised absence	0.07	0.06
Mishandling of property or exhibits	0.12	0.13
Misuse of computer systems	0.19	0.12
Breach code of conduct	0.13	0.05

Variance inflation factor and correlation analysis

An initial model was estimated using all available variables, the VIF was then computed. Three variables achieved the threshold value above 5, these were the total number of complaints received by an officer (6.62), the number of unsubstantiated complaints (5.43), and the number of management actions taken against an officer (5.82). These values met the threshold proposed in the literature indicating multicollinearity. Given the intention of this research, and the considerable risk that including these three variables posed to the integrity of the statistical modelling, the latter two variables were removed from further analysis, with the variable reflecting the overall number of complaints received by officers retained. Upon removing these variables, the VIF was recomputed for the produced model, with all variables achieving a value below 2.4, indicating that multicollinearity concerns had likely been resolved.

The correlation analysis suggested that the relationship between complaint variables was limited (Figure 1). Of the variables available for analysis, there were only three that featured a relationship stronger than r = 0.4. The strongest correlation was found between complaints relating to an officer’s performance at court, or customer service (meaning interactions with members of the public) and the total number of complaints received (r = 0.49, p < 0.01 and r = 0.48, p < 0.01 respectively). Further, officer length of service held a relationship with the years spent at final rank, and the total number of complaints received (r = 0.43, p<0.01 and r = 0.42, p < 0.01 respectively). However, these were the only relationships that yielded a correlation greater than r = 0.4, and while they bear note, they are still considered to be moderate-weak relationships.

Figure 1. Correlation matrix for data on officer complaint histories and positive drug tests.

Pivotally, for the central component of this research, there were no complaint variables featuring a correlation with a positive drug test stronger than r = 0.18 (p < 0.01), suggesting that, on first impression, there was little relationship between the prior misconduct variables and a subsequent positive drug test. Further, the raw number of complaints received by officers was not significantly associated with returning a positive drug test. However, it is notable that the variable holding the strongest correlation with an officer testing positive for illicit substances was a complaint of an improper association. In other words, while the effect size was weak, a substantiated finding that an officer had an association with a known criminal held a weak correlation with that same officer testing positive to an illicit substance.

Logistic regression

A logistic regression was estimated to predict positive drug tests among officers. The logistic regression featured a strong rate of prediction, an AUROC of 0.832. An AUROC of greater than 0.7 is considered to represent a noteworthy prediction rate (Grogger et al. 2021), meaning that the logistic regression appeared to be a robust prediction model, and suggesting that there were variables included in these data that were quite good at discriminating which officers would and would not test positive for illicit substances. The ROC curve for the logistic regression is provided as Figure 2.

Figure 2. Receiver operating characteristic curve for logistic regression predicting a positive drug test by officers.

Logistic regression results (presented in Table 3) supported the findings of the correlation analysis in suggesting that, when measuring the likelihood that an officer may test positive for illicit substances, historical misconduct was not useful. Of the 15 complaint types included for analysis, only one yielded a statistically significant interactions with a positive drug test, having an improper association with a known offender. Further, while the specific types of complaints did not appear to be useful, the same was true of the overall volume of complaints. The variable reflecting the total number of complaints received by an officer did not hold a notable relationship with a positive random drug test. This means that using information emerging from an officer’s misconduct history was likely a poor method for assessing risk of a positive drug test.

Table 3. Logistic regression estimates and odds ratios.

Variable		Log-odds	SE	OR
Rank (ref. Probationary Constable)	Constable	−1.66	0.94	0.19
	Senior constable	−2.46	1.11	0.09*
	Sergeant	−4.41	1.51	0.01**
	Commissioned officer	−3.59	1.89	0.03
Duty type (ref. General duties aka Patrol)	Specialist duties	0.02	0.40	1.02
Years at final rank (0-4 ref.)	5–9 years	−0.59	0.68	0.55
	10+ years	−0.75	1.11	0.47
Length of service (0-4 years ref.)	5–9 years	0.36	0.58	1.44
	10–14 years	1.05	0.98	2.86
	15–19	0.79	1.19	2.21
	20–24 years	1.31	1.34	3.74
	25+ years	−0.31	1.62	0.72
Secondary employment		1.33	0.47	3.78**
Total number of complaints		0.11	0.06	1.11
Complaint variables
Prior serious misconduct		0.19	0.63	1.21
Drink driving		−0.07	1.02	0.93
Firearm misuse		0.77	0.86	2.16
Improper association		2.09	0.99	8.09*
Performance at court		0.19	0.34	0.81
Allegation of corruption		0.96	1.36	2.62
Issue with investigation		−0.15	0.29	0.86
Customer service		−0.04	0.27	0.96
Harassment or bullying		−0.04	0.47	0.96
Untruthfulness		0.36	0.39	1.44
Misuse of powers		−1.26	0.95	0.78
Disobey direction of superior		0.23	0.62	1.26
Unauthorised absence		−0.73	0.79	0.48
Mishandling of property or exhibits		−0.77	0.53	0.46
Misuse of computer systems		−0.22	0.49	0.80
Breach code of conduct		0.97	0.61	2.63

**p < 0.01, *p < 0.05.

Similarly, demographic variables did not yield notable relationships. Although junior officers appeared more likely to return a positive drug test, variables reflecting years of service did not hold a notable relationship. However, pivotally, officers who held secondary employment, unrelated to being a police officer were at significantly greater odds of testing positive for an illicit substance.

Discussion

While recent research has focused on the correlates of career ending misconduct among police, empirical research considering the correlates of specific types of serious misconduct, such as drug use among officers, is limited. The present research offers unique insight into the characteristics and correlates associated with officers who are detected using illicit substances.

Correlates of substance use by police

There were some demographic similarities among officers who test positive for illicit substances and those who commit other types of career ending misconduct. Prior research has extensively considered the role of age and sex in police misconduct, finding that misconduct prone officers were more likely to be young (Greene et al. 2004, McElvain and Kposowa 2008, Rojek and Decker 2009, White and Kane 2013), and male (Greene et al. 2004, McElvain and Kposowa 2008, Kane and White 2009, White and Kane 2013, Gaub 2020). The present research took the approach of matching officers using these two variables, to allow measurement of whether complaint variables, or overall volume of prior misconduct, were important antecedents to a positive drug test. As a result, the sex and age variables were not included for regression analysis, as they were not true independent variables. Despite this, cautious note can be made that the demographics of officers who used illicit substances followed a similar trend to those who commit serious misconduct, although perhaps more exaggerated. For example, in analogous jurisdictions in Australia and the UK, the proportion of male officers who commit other types of serious misconduct tended to be around 80–85 percent, on average they were in their mid to late 30’s (Cubitt et al. 2020, Cubitt 2023). While the sample of officers who used illicit substances fit a similar age group, the proportion who were male was higher.

Historically, research has found that as the rank of an officer increased, the risk of committing misconduct decreased (Hickman et al. 2004, Kane and White 2009, Gaub 2020). This notion has often been related, at least in part, to higher ranking officers spending less time interacting with members of the public by virtue of their duties shifting further from operational policing and toward administration. However, when applying this logic to officers returning positive drug tests, it did not hold. There were no ranks that held a significantly greater association with positive drug use, while officers who undertook specialist duties were equally likely to test positive to illicit substances as general duties (patrol) officers. In totality, these findings suggested that career characteristics were not useful in assessing risk of drug use by police.

The role of secondary employment, holding a second job in addition to being a police officer, has been subject to little research. Cubitt et al. (2020) previously found that secondary employment predicted career ending misconduct. Here, officers who held secondary employment were 3.8 times more likely to return a positive drug test than an officer who did not. These findings support emerging evidence that secondary employment may not only be a significant risk for career ending misconduct among police, but it may also indicate increased risk of drug use by officers.

In answer to the first research question, based on the results of the analysis, it appears that prior minor misconduct did not hold a strong association with an officer returning a positive drug test, except for those who had a recorded improper association. In other words, a positive drug test was much more likely to occur among those who had an association with a known offender. However, none of the remaining individual types of misconduct were useful as predictors of whether an officer would test positive for an illicit substance. Further, the volume of complaints received by an officer, a common metric used to evaluate whether an officer was a risk for a serious misconduct event, also yielded no meaningful relationship with a positive drug test. This was an important finding, while the volume or type of misconduct committed by an officer may be useful in tailoring generalised misconduct prevention (Chalfin and Kaplan 2021, Sierra-Arrevalo and Papachristos 2021), or assessing risk of other types of misconduct (Kane and White 2009, Cubitt et al. 2020), neither appear to hold a relationship with the likelihood of a positive drug test. However, where an officer held secondary employment, or if they had associated with a known offender, they were significantly more likely to return a positive drug test.

Interpreting findings in view of theory

Prior research has offered strain theory to help explain why police may commit misconduct (Paoline 2004, Bishopp et al. 2020). Indeed, officers who have committed serious misconduct have attributed their behaviour to increased strain both within and outside of the workplace (Cubitt and Judges 2018), and pivotally, personal financial strain was a common factor among these officers. It is possible that the secondary employment variable in this research serves as a proxy measure of financial strain. Given the relationship between this variable and a positive drug test, this finding may support extending the application of strain as a theoretical lens specifically to drug use by police. However, in this instance rather than strain emerging from burnout directly associated with the duties of a police officer, it may emerge from burnout in the presence of additional, secondary employment outside of the agency, and potential financial stress. If this is the case, it is important that agencies allowing secondary employment among their officers identify the reasons given by an officer for taking on a second job, and perhaps the type of employment undertaken. For example, there are likely to be high-risk secondary employment activities that expose officers to a greater degree of burnout (shift work), or association with known offenders (private security).

While these findings offer tentative support for Strain Theory as an important lens through which to consider drug use by police, it should be noted that this research is limited to using proxies to infer strain. In other words, emerging from the empirical focus of this analysis, and the use of information readily available to police agencies, there were few variables included for analysis that could directly test whether strain held a role in drug use by police. Rather, proxies for strain were used to infer the possibility. While a similar finding could be made for social learning theory, another popular theoretical lens for misconduct behaviours among police, the findings of the statistical modelling yielded a more notable relationship with strain theory than other possible interpretations. However, despite the limited certainty emerging from use of proxy variables rather than direct measures, the inherent stress associated with policing and the associated strain experienced by officers, in combination with these findings, support the need for further research into the role of strain in drug use by police.

It is therefore recommended that an important next step is an empirical test of the relationship between strain, as it is defined previously in this paper, and drug use by police. The present research has provided empirical evidence of a relationship between data available to police and instances of detected drug use, by including data based on availability to police and usefulness for developing policy responses. Pivotally, one of the variables that yielded a significant interaction with officer drug use was a proxy measure for strain. Future research could use strain theory as a lens for the design of data collection, analysis and interpretation of findings, to establish whether the tentative findings from the present paper, that strain holds a central role in drug use by police, are indeed true.

Ultimately, these findings tentatively support the role played by strain in officer drug use. Policing agencies may benefit from considering whether the type or nature of secondary employment is likely to increase the strain experienced by an officer, or whether the intention to undertake an external job indicates increased strain outside of (and otherwise unknown to) the workplace. However, it is important that, where possible, future research include direct measures of strain when analysing substance use, and more broadly maladaptive behaviour, by police.

Implications

Misconduct management and detection of drug use

Traditionally, substance use by police has been viewed as a significant organisational risk due to the potential effect on officers’ performance and the subsequent liability for law enforcement agencies (Harvey 1991). While drug testing may also be used as a form of social control over the behaviour and performance of police officers (Brunet 2005), it is also frequently viewed as a means of improving potential performance deficiencies brought on by substance use (Daley and Ellis 1994). However, there is little evidence that drug testing effectively addresses other aspects of drug-related corruption (Micezkowski and Lersch 2002), such as the development of associations with known offenders, and the susceptibility to blackmail. From this perspective, while police are drug tested, it is important that positive tests are treated as serious instances of misconduct, not least because the functions of democratic policing, and the legitimacy of police are dependent on officers being held to (at a minimum) the standards of the legislation that they enforce. However, it is also possible that, if we know the factors that correlate with officers testing positive for illicit substances, there are opportunities for better management strategies.

From a complaint management perspective, this research may mean that there are opportunities for being more vigilant in relation to officers with secondary employment, and officers who have at one time associated with known criminals. These types of analyses offer opportunities to adapt complaint management approaches to either better detect substance use, or to better support officers to reduce the risk of substance use (or both). There are opportunities for these types of analyses to guide the development of more suited management, detection and support opportunities specifically for drug use by police. Alternatively, that specific complaint types did not hold relationships with a positive drug test is an important implication for asset distribution, and effort put into detecting officers who may be using illicit substances. Using complaints as a metric for risk in undertaking prevention or detective activity may be misplaced effort, while it may be more effective to increase organisational surveillance over who has secondary employment and why.

Finally, it is possible that these findings could be used to improve random drug testing. For example, on face value, the randomisation of drug testing could be weighted to test officers that fit the characteristics found in this analysis more frequently. A type of weighted random testing, rather than true random testing. For this to be effective, every time an officer returned a positive drug test, their data could be added to the analysis to ensure that the findings remained current, and the weighted random testing remained accurate. However, while this approach may yield increased rates of detection, caution is recommended. If we are to assume that there have been officers who have used drugs but have successfully avoided random drug testing regimes, through either luck or implementing a methodology for avoiding detection, a weighted random testing approach designed using the characteristics of those who have been detected is unlikely to be effective. There are clear trade-offs to amending detection methodologies for drug use by police, and it is important to note that the findings of the analyses in this research reflect only those who have been caught, while we cannot be certain of the characteristics of those who avoid detection.

The role of stress and trauma exposure

Prior findings have noted that stress and mental health issues may be associated with misconduct (Cubitt and Judges 2018), while separately, occupational stress and trauma exposure have been linked with problematic alcohol use among police (Police Integrity Commission 2005, Smith 2007). In other populations, substance use as a coping mechanism for stress has been well established over time (Willis and Hirky 1996, Sinha 2001, 2008). The culture of internalising, or avoiding acknowledgement of problematic behaviour may be a component of a culture of silence (Kutnjak Ivkovic & Shelley, 2008). It is reasonable to consider that this culture extends to problematic behaviours emerging from stress, particularly in the overtly masculine subculture of policing, in which identifying oneself as having trauma-related mental health issues is often associated with a loss of power and opportunity (Young et al. 2012).

While this is not the first research to acknowledge stress as a potential factor in substance use by police (Cross and Ashley 2004, Cubitt and Judges 2018), it is an important factor in the implications of this research. Police likely experience a higher rate of trauma exposure than other groups within the community (Foley and Massey 2018). While there is no research that has yet established a causal interaction between trauma exposure and substance use among police, it is an important direction for future research. However, it is also clear that when substance use is detected through the misconduct process, it is too late at least for the employment of that officer. Policing agencies have successfully provided support frameworks for officer stress through work environment management training and the availability of psychological and medical assistance (Pillay and Claase-Schutte 2004). The findings of this research could be used as a guide to which groups within police agencies that could benefit from increased support, organisational surveillance, or both. For example, given that secondary employment featured a strong interaction with an officer testing positive for using an illicit substance, it could be that agencies adopt a process of approving secondary employment in the presence of increased support for that officer, or increased organisational surveillance.

Conclusion

The implementation of successful drug testing regimes is important to the legitimacy of modern policing agencies. This research provides some insight into a severely under researched area, and pivotally, it suggests that positive drug tests were not associated with a single type of prior misconduct, or the volume of complaints received by officers – two commonly used metrics to measure risk. Rather, there was a strong relationship between having a second job (outside of the police) and returning a positive drug test. While officers who have had an association with a known criminal also held a strong relationship with a positive drug test. These findings hold important implications for how random drug testing procedure is implemented, the known risk factors for substance use among police, and how agencies can evaluate who is at risk of returning a positive drug test.

Disclosure statement

No potential conflict of interest was reported by the author(s).