Optimising the length of random breath tests: Results from the Queensland Community Engagement Trial

Abstract

Research suggests that the length and quality of police–citizen encounters affect policing outcomes. The Koper Curve, for example, shows that the optimal length for police presence in hot spots is between 14 and 15 minutes, with diminishing returns observed thereafter. Our study, using data from the Queensland Community Engagement Trial (QCET), examines the impact of encounter length on citizen perceptions of police performance. QCET involved a randomised field trial, where 60 random breath test (RBT) traffic stop operations were randomly allocated to an experimental condition involving a procedurally just encounter or a business-as-usual control condition. Our results show that the optimal length of time for procedurally just encounters during RBT traffic stops is just less than 2 minutes. We show, therefore, that it is important to encourage and facilitate positive police–citizen encounters during RBT at traffic stops, while ensuring that the length of these interactions does not pass a point of diminishing returns.

Keywords

Police legitimacy police performance procedural justice random breath tests (RBTs)traffic stops

Introduction

Traffic law enforcement is a routine function of police agencies and is the most visible aspect of policing (Fildes, 1995). Police spend a significant amount of their time on traffic activities and the most common contact that citizens have with police is during traffic-related operations (Hoover, Dowling, & Fenske, 1998). Indeed, research suggests that approximately 25% of time that police officers spend in contact with citizens is related to traffic (Liederbach, 2005; Liederbach & Frank, 2003). In Australia, one of the most common ways for citizens to come into contact with police is during routine random breath tests (RBTs).

Police conduct about 12 million RBTs across Australia every year (Ferris et al., 2013). RBTs are undertaken in a number of different contexts: large ‘Booze Bus’ operations on main arterial routes, ‘pop up’ stationary operations involving seven to 10 officers on suburban streets and main throughways, and ad hoc traffic stops. All of these RBT encounters involve police engaging with citizens, asking them for a breath sample and then either sending the driver on their way or detaining them for further testing. Yet, despite the high frequency and visibility of RBTs and traffic law enforcement, very little research has explored the dynamics and outcomes of police–citizen interactions during traffic stops.

Our paper starts from the premise that the quality of interactions between police and citizens (see Murphy, 2009; Reisig & Lloyd, 2009; Sunshine & Tyler, 2003; Tyler, 2004) is important within the context of traffic law enforcement. The extant literature on police legitimacy shows that when police treat people with dignity and respect, convey trustworthy motives, give ‘voice’ to citizens and are neutral in their decision making, citizens will have more positive perceptions of police and are more likely to comply with their requests (see Mazerolle, Bennett, Davis, Sargeant, & Manning, 2013a). We know, however, that these types of procedurally just encounters take longer than those interactions that do not adopt the principles of procedural justice (Mazerolle, Bennett, Antrobus, & Eggins, 2012). In the context of RBTs, we wonder how long is too long, how short is too short, and what is ‘just right,’ or what we call the ‘Goldilocks’ length of encounter time that optimises citizen perceptions of police.

The business (customer satisfaction), public health (e.g. patient waiting times) and tourism (e.g. waiting times for theme park rides) literatures are replete with research showing that the length of encounter or waiting times influences citizen satisfaction with services provided. For example, Roslow, Nicholls, and Tsalikis (1992) assess the twin elements of customer satisfaction in the context of banking – time and quality – finding that waiting time and time spent ‘consuming’ the service were both more important factors in shaping citizen satisfaction than the quality of the service delivered. In the policing literature, the one body of research that provides a hint that there may be, in fact, an optimal length in police operations is the work stemming from the hot spots literature. The Koper Curve, for example, demonstrates that 15 minutes is the optimal time that police officers should spend in a hot spot location in order to maximise the crime control benefits (Koper, 1995; Telep, Mitchell, & Weisburd, 2012).

Our paper, using data collected as part of the Queensland Community Engagement Trial (QCET), examines whether there is an optimal length – a Goldilocks effect – for police encounters with citizens during procedurally just RBT traffic stops. In Australia, police systematically undertake RBTs during traffic encounters in order to identify people driving under the influence of alcohol. We use this point of contact between police and citizens to assess whether a similar Koper Curve exists in relation to optimising positive perceptions of police during traffic stop encounters. We begin this paper by reviewing the literature regarding traffic policing and what we know about encounter length. We pay attention to what we know about police optimising the length of time of encounters, and we also highlight the extant literature on the quality of the exchange between police and citizens. We introduce QCET and describe how the randomised field trial sought to assess the impact of procedurally just encounters between police and citizens. We then outline the methodology of the current paper and present our results. Finally, we discuss the theoretical and policy implications of our research.

Background literature

Police officers spend a significant proportion of their working time on traffic enforcement. Liederbach (2005), for example, reported that managing traffic problems in urban areas of Cincinnati, Ohio involved the greatest proportion of police–citizen encounter time in nine of 14 suburban policing areas. Using Systematic Social Observation methods, Liederbach showed that the proportion of time spent on traffic tasks ranged from 20% to 49% of a normal shift. Traffic stops in rural police areas were similarly the most time consuming task, occupying about 25% of police time spent interacting with citizens (Liederbach & Frank, 2003).

Some traffic encounters, such as ‘stop and talks,’ can be very short interactions ranging from 20 to 30 seconds (see Famega, 2009). Other types of police interventions, such as those involving police response to domestic violence, robbery or break and enter calls, are typically much longer and more complex than traffic encounters. The amount of time that police spend on different policing operations is clearly shaped by the complexity of the event. Yet, the policing literature is virtually silent on the question of how length of time spent on a police task might shape outcomes. We do not know, for example, whether longer times spent by police at burglarised homes might lead to higher levels of citizen satisfaction with police or greater likelihood of a clearance than shorter times.

Chris Koper's (1995) seminal work on hot spots policing is foundational for understanding the importance of time spent on police operations. Using observational data from the Minneapolis Preventive Patrol Experiment (see Sherman & Weisburd, 1995) to assess the impact of patrol presence on residual deterrence, Koper's (1995) study examined the varying lengths of time police spent in crime hot spots. He found that police officers needed to spend at least 10 minutes stopped (or parked) in a hot spot in order to significantly increase the residual deterrence effect of hot spots policing. Koper's (1995) research considered police stops in crime hot spots that varied from a simple drive through to police patrol presence of up to 20 minutes. He found that the optimal length for police presence in hot spots was between 14 and 15 minutes, with diminishing returns observed from this point (Koper, 1995, p. 663). A more recent study conducted in Sacramento, California similarly found that police need to spend 15 minutes in hot spots in order to deter crime (Telep et al., 2012).

Koper (1995) and Telep et al. (2012) did not take into account what police actually did during their stop. Yet, we know from the Sherman and Weisburd (1995) study, from where Koper derived his data, that officers varied in what they actually did in the hot spots. In a personal communication from onsite observations of the hot spots during the Minneapolis Hot Spots Experiment, George Kelling described the differences in officers' behaviors: ‘… some were reading newspapers or sunning themselves while sitting on the patrol car while others were engaging citizens in friendly interactions in community-policing style’ (Sherman & Weisburd, 1995, p. 634). In the language of business research (see Roslow et al., 1992), this foundational hot spots study failed to explore the twin concerns of time and quality in shaping outcomes.

The silence in the policing literature on exploring the influence of time and quality of police interventions is evident in the published findings of the first hot spot experiment (see Sherman & Weisburd, 1995), the related study findings (see Koper, 1995) and the Sacramento replication (see Telep et al., 2012). Indeed, Braga's (2007) systematic review of hot spot experiments highlights the weakness of the literature, concluding that we need to know more about community reaction to police enforcement action (see also Braga, 2005). Policing scholarship assumes, for example, relative consistency in the quality of police interactions for community policing (Weisburd, Gill, Bennett, & Telep, 2012), foot patrol (see Ratcliffe, Taniguchi, Groff, & Wood, 2011) and problem-oriented policing (Weisburd, Eck, Hinkle, & Telep, 2008). By contrast, Taylor et al.'s (2011) experiment conducted in Jacksonville, Florida sought to differentiate the influence of different types of hot spots policing tactics (problem-oriented policing, saturation patrol and the control condition). They found that it was not just the time spent on a policing activity that was important, but also what police did while in the hot spot that shaped the crime control outcomes (Taylor et al., 2011).

Beyond crime control outcomes, the burgeoning literature on police legitimacy and procedural justice in policing (see Mastrofski, Snipes, & Supina, 1996; Mazerolle, Bennett, Davis, Sargeant, & Manning, 2013b; Sunshine & Tyler, 2003; Tyler, 2004; Tyler & Fagan, 2008) clearly shows that the manner in which police interact with citizens – whether during foot patrol, hot spots policing or problem-oriented policing – can influence citizen perceptions of police. We know, for example, that when individuals perceive police as procedurally just, they are more likely to think the police are doing a better job (see Hinds & Murphy, 2007; Kochel, Parks, & Mastrofski, 2011; Sunshine & Tyler, 2003), perceive police as legitimate, and thus comply and cooperate with police (Mazerolle, Bennett, et al., 2013b; Sunshine & Tyler, 2003; Tyler, 2004). When police apply the principles of procedural justice, citizens are more pleased with both the interaction and the outcome of the interaction (Mastrofski et al., 1996; Tyler & Fagan, 2008). Individuals are also more positive in their assessment of police when police officers are fair, open and respectful in their encounters (Myhill & Bradford, 2012).

The police legitimacy literature pertaining to traffic stops is consistent with the general legitimacy literature. Wells (2007), for example, shows that procedural justice has an impact on drivers' perceptions of police during police–driver encounters at traffic stops. In his evaluation of the Lincoln, Nebraska Police Department's Quality Service Audit, Wells (2007) examined survey data from three groups of citizens who had contact with the department's patrol officers: crime victims, drivers involved in crashes and citizens who received citations. The results suggest that police actions that reflect procedural justice and outcome-oriented behaviour (e.g. professionalism, competence and contact-specific behaviour) are important to citizens' overall ratings of police officers during their encounters. Wells identified that in all three contexts, procedural justice was more important than police officers' outcome-oriented behaviours.

Similarly, Tyler and Folger (1980) conducted phone interviews with a random sample of households in Evanston, Illinois, where they asked participants whether they had been stopped by police during the past five years for questioning or for a traffic violation. Within their sample, 36% had been stopped by police and 9% had been stopped more than once. Over 90% of the contacts with police involved traffic or motor vehicle violations (Tyler & Folger, 1980). They found that those who reported that they were treated fairly by the police gave more positive evaluations of their specific encounter with the police and police generally than did those who reported that they were unfairly treated. The impact of fair treatment was observed when the police cited the participant for both traffic and motor vehicle violations.

Robin Engel (2005) also identified that procedural justice and perceived fairness of the outcome were significant predictors of citizens' perceptions of injustice during police traffic stops (Engel, 2005). Using the Police Public Contact Survey (PPCS), designed to examine citizens' interactions with police, Engel analysed a sample of US residents as part of a supplement to the 1999 National Crime Victimization Survey. Of the respondents who reported contact with police in the last year (n = 17,720; 22% of the 80,543 citizens who completed the PPCS), 40% (n = 7054) indicated that the contact with police occurred during traffic stops. Engel's (2005) results suggest that participants' perceptions of both procedural and distributive justice were influenced by several factors: the initial reason for the traffic stop, whether the police used force during the encounter, and whether either the vehicle or the citizen was searched. Those citizens stopped for traffic offences other than speeding were 3.1 times more likely to believe they were treated unjustly by police compared to those stopped for speeding (Engel, 2005). Participants who had force used against them by police, or who were searched, or had their vehicle searched, were less likely to perceive that the police were procedurally just.

The extant literature also shows that the behaviour of police officers while undertaking traffic enforcement affects the number of complaints made about police organisations. Johnson (2004) found, for example, that if individuals, when stopped for traffic enforcement, perceived the actions of police officers as fair and respectful, the police service may be able to reduce the number of complaints made and increase the support that they receive from the community. He suggested that the number of complaints from drivers to police agencies could be reduced by training police officers to communicate more effectively and appropriately when interacting with individuals stopped for traffic enforcement purposes (Johnson, 2004).

Perceptions of police during police–citizen encounters often vary by who initiates the encounter. In most cases, traffic stops are initiated by police. Research shows, however, that citizens generally report more positive attitudes towards the police when they initiate the contact compared to situations where the police initiate the contact (Audija & Giever, 2010; Cheurprakobkit, 2000). Regardless of who initiates the contact, citizens report that if their encounter with police is positive, they are more likely to hold positive views generally of the police (Chow, 2010, 2012; Worrall, 1999). Conversely, Skogan (2006) found that having a bad experience is four to 14 times more influential on citizen attitudes towards police performance and legitimacy than having a good experience (see also Schafer, Huebner, & Bynum, 2003).

Our paper adds to the extant literature by illustrating how the length of time police officers spend in a police operation, as well as the quality of the encounter, can influence the outcome of policing activities involving the public. Holding the quality of the police–citizen encounter constant, we focus on understanding whether length of encounter time shapes citizen perceptions of police. Using data from the QCET, we explore whether there is an optimal length of time during procedurally just RBT encounters that shape positive views of police. The sections below provide details on trial methods as well as the analytic method used to explore the relationship between encounter length and citizen views of police. In this paper, the procedurally just encounter – the experimental arm of the trial – is a constant. Yet, within this experimental arm, we assess the influence of varying dosage time (i.e. the time spent during the police–citizen encounter) on citizen perceptions of police procedural justice and police performance.

Method

QCET is a randomised field trial that directly tests the impact of police–citizen encounters on citizen views of the police. Mazerolle and her colleagues have described details of QCET in several previous papers and technical reports (Mazerolle, Antrobus, Bennett, & Tyler, 2013; Mazerolle et al., 2011, 2012). In short, the field trial involved the random allocation of 60 RBT stationary operations (akin to roadblocks) to either the standard RBT operation (control condition) or an experimental condition.

RBTs in Australia involve police randomly pulling over blocks of seven to 10 drivers at any one time and then using a calibrated instrument to test whether each driver is driving under the influence of alcohol. The police officer instructs each driver to open the window; the officer then presents the driver with a short plastic tube attached to a small handheld machine and instructs the driver to blow into the tube until told to stop.

Routine RBT police–citizen encounters in Australia are very systematic and generally devoid of anything but compulsory communication (i.e. the mandated message that officers are required to give to the driver for the purpose of obtaining a specimen of breath for the breath test). The length of police–citizen contact during a standard RBT averages less than 20 seconds for drivers whose initial reading is negative¹ and involves minimal verbal exchange between the police officer and the driver. If the driver blows over the alcohol limit (.05 grams per 100 millilitres of blood for full licence holders in Queensland), the police officer asks the driver to step out of his or her car and then processes the driver in an onsite caravan using more sophisticated breath testing equipment. In Queensland, RBT interactions comprise a significant amount of in-person traffic enforcement, with Queensland Police Service officers conducting over 3.3 million RBTs, and 22,662 random blood or saliva tests, in 2011–2012 in order to detect alcohol and other drug driving offences (Queensland Police Service, 2012).

The 60 RBT operations in QCET involved police pulling over between 300 and 400 cars per operation. Each operation was conducted on major suburban streets in Brisbane, the capital city of Queensland. The police provided each driver in the field trial with a sealed envelope, and let drivers know that the envelope contained a survey developed by researchers at The University of Queensland, that the survey was voluntary, and could be completed and returned at a later time (see Mazerolle et al., 2012) using a self-addressed, pre-paid envelope identifying the date and location of the RBT operation.

The business-as-usual encounter (control condition) involved the police conducting routine RBT interactions with citizens during traffic roadblocks. By contrast, in the experimental condition, traffic police used a script that operationalised the key elements of procedural justice: neutrality, treating people with dignity and respect, conveying trustworthy motives and giving citizens voice during decision making (see Mazerolle et al., 2012). Officers were encouraged to cover the elements of procedural justice conveyed in the script in a conversational way instead of reading the script to the driver verbatim. All participants in the trial were provided with a survey regardless of the outcome of the RBT. For both the control and the experimental conditions, drivers who blew over the blood alcohol concentration limit were processed on site as per standard operating procedures. These motorists were also provided with a survey pack.

Monitoring the integrity of the conditions delivered is critical to all experiments and crucial for minimising threats to the internal validity of the trial (Cook & Campbell, 1979). The nature of RBT operations in QCET provided exceptionally easy access to monitor the delivery of the control (business-as-usual) and the experimental (procedurally just) conditions. Researchers and a sworn Inspector-level police officer (dedicated to oversee the field execution of QCET) walked up and down the line, listening to the conversations between police and citizens and ensuring that all officers delivered the treatment as planned. The Inspector pulled officers off the line if they got sloppy with the script, had a bad attitude or tried to subvert the trial (see Mazerolle et al., 2014). We propose, therefore, that the treatment integrity of QCET was exceptionally high, because of both the operational design of the encounter and the resources invested to monitor the treatment fidelity.

Key findings from QCET show that the procedurally just encounter (experimental condition) led citizens to report changes in their attitudes to driving under the influence of alcohol (Mazerolle et al., 2012), have more positive perceptions of police performance, be more willing to comply with police directions, be more satisfied with police (Mazerolle et al., 2012) and shape more positive generalised perceptions of police (Mazerolle et al., 2012). Building on these earlier studies of QCET, we examine in this paper whether variations in the experimental encounter length influenced citizen perceptions of procedural justice and police performance. Like Koper's (1995) study of the optimal time that police need to spend in hot spots to reduce crime, we assess whether there is a Goldilocks encounter length in the context of procedurally just traffic stops.

Measuring encounter length

On average, seven to 10 officers per RBT operation delivered either the standard (N = 30 RBT operations) or experimental intervention (N = 30 RBT operations). The Inspector and research staff monitored the delivery integrity for each of the 60 operations, recording the length of time that officers spent engaged in delivering both the treatment and control conditions. Recall that an RBT traffic road block uses seven to 10 officers to pull over seven to 10 drivers in a single line, all at the same time. With this operational pattern, all officers spend equal time with all drivers pulled over in a block. In all 60 RBT operations, researchers used a stop watch to time every block of drivers. An encounter length time was therefore calculated per line of seven to 10 cars and then averaged to calculate the encounter length time per RBT operation. With the road block line of some seven to 10 officers all needing to stay on the same time schedule of moving individual cars through the roadblock, it was easy for the research team to record the duration of each encounter per line. These durations were then averaged to generate an average encounter length per operation. Given that the RBT operation was the unit of random assignment (see also Mazerolle et al., 2012), coupled with the fact that we did not have a way of linking timed encounters to returned surveys, we collated information on the average length per RBT operation (i.e. for each of the 60 RBT operations).²

The average encounter length for operations in the experimental condition was 99.11 seconds (SD = 30.01; median = 95.48; range = 45.75–160.06 seconds; see Figure 1) and 25.51 seconds (SD = 4.84; median = 24.44; range = 17.62–36.32 seconds) for the control condition. Variations in encounter length were generated by variability in the way officers per operation built rapport with drivers, the weather conditions (quicker when it was raining) and the location of the RBT operation itself (quicker in busy traffic conditions). As was expected – and reflective of the precision in the execution of the experimental and control conditions – the relationship between condition and encounter length is r = .88. Indeed, the longest control group encounter was 36.32 seconds, and the shortest experimental group encounter was 45.75 seconds. As such, encounter length can perfectly discriminate between experimental and control conditions, indicative of a well-executed experiment.

Sample – Experimental group only

A total of 8985³ drivers from the experimental condition (as well as 12,000 drivers from the control condition) were given a survey, with an addressed postage paid envelope for its return, and requested to return their completed survey to The University of Queensland. In this paper, we examine the variability in encounter lengths for the experimental group only for two reasons: First, the control condition did not use the principles of procedural justice and generated no variability in encounter length (as expected). Second, the logic of our paper is to hold the nature of the procedurally just encounter condition as a constant (experimental condition) and explore whether the length of the encounter time had any influence on variations in citizen perceptions of police. We know that the procedural justice script was delivered with a high degree of integrity due to the attention paid to monitoring the dialogue of the officers in both the control and treatment conditions. What we do not know is whether the variability in encounter length influenced citizen perceptions above and beyond that of the experimental script.

For the experimental condition, 1097 of 8985 drivers returned surveys distributed at RBT operations by Queensland Police Service officers (12.20% response rate). This response rate is similar to a study that used a telephone survey to investigate Queensland citizens' attitudes to drink driving and random breath testing (14% response rate; see Freeman & Watson, 2009; Watson & Freeman, 2007). The experimental condition sample was 50.23% male (n male = 551, n female = 534, 12 participants did not report gender) and participants' ages ranged from 17 to 89 years with a mean of 47.13 years (SD = 14.62 years). The majority of the sample was married or de facto (n = 671; 61.17%), employed either full or part time (n = 745, 67.92%) and classified themselves as Australian (n = 513; 46.76% – 822 or 74.90% were born in Australia). These statistics show that the QCET sample is comparable to the population of Queensland in general, based on the 2011 population census (Queensland: 50.4% female; 59.6% married or de facto; 59.0% working; 73.7% born in Australia) (Australian Bureau of Statistics, 2012). Further information regarding the sample characteristics is available in Mazerolle et al. (2012).

Outcome measures

In this paper, we explore the relationship between encounter length within the experimental arm of the trial and two key outcome measures gathered from the driver survey: respondents' perceptions of police procedural justice and their perceptions of police performance. The survey included five items that made up a scale assessing participants' perceptions of police procedural justice (scale M = 3.59, SD = 0.80, Cronbach's alpha = .91),⁴ and five items that made up a scale assessing participants' perceptions of police performance (scale M = 3.59, SD = 0.78, Cronbach's alpha = .94).⁵ Participants ranked their responses to items for these two scales on a 5-point scale from strongly disagree to strongly agree, and from very poor job to very good job, respectively. There were minimal missing data in the scales that were created (<6%). Missing data were deleted listwise in scale creation.

Analytic approach

Our paper focuses on the variations in encounter length time for the 30 RBT experimental operations. The 30 experimental operations thus serve as the unit of analysis in this paper. We used Generalized Additive Models (GAMs) (Hastie, Tibshirani, & Friedman, 2001) as a means of modeling the relationship between the response variables and functions of the independent variable. We used GAMs as they allowed us to model nonlinear effects. In our case, we wanted to explore whether there was a specific window of time (or a ‘hump’) where encounter length might show significant effects on respondent perceptions. GAMs allow us to directly model this effect, providing more insight than if using a linear model, which would only tell us if the effect of encounter length increased on average with time. In GAMs, the nonlinear effects are modeled using cubic smoothing splines with predetermined degrees of freedom. These are fit using an iterative back-solving algorithm. For clarification, we give a general description and comparison between GAMs and linear models. In a linear model, data are modeled as

y_{i} = (X β) i + ɛ_{i}

The GAM is

y_{i} = α + f (X_{i}) + ɛ_{i}

where the function(s) f are cubic smoothing splines of some predetermined degree of freedom, estimated using a back-fitting algorithm. Smoothing splines place a knot at every observation and the best fit is a compromise (depending on the degrees of freedom) between a smooth line interpolating every observation and a linear term, as in an Ordinary Least Squares model. This differs from spline basis function regression where basis functions are calculated for a predetermined set of knots over the range of the data, and the fit is determined using least-squares regression. The degree of flexibility possible in a GAM allows for a nonlinear effect to be manifest, as in the case where encounter length may have little to no effect up to a certain time, and then diminish afterwards. The resulting model can be decomposed into the linear and nonlinear components and each tested for statistical significance (for a more thorough discussion, see Hastie & Tibshirani, 1990).⁶

The GAMs were fit in R (R Core Team, 2013) using the GAM package (Hastie, 2013) for the responses, using a set of spline basis functions of varying degrees of freedom from 1 (simple linear regression) to 20. The Akaike Information Criteria (AIC; Akaike, 1973) for each of these models are recorded and the model with the lowest AIC is selected for use in inference. For the two outcome measures in our paper – perceptions of police performance and perceptions of procedural justice – both scales selected the model with 9 degrees of freedom as having the lowest AIC.

Results

Perceptions of police performance

Using the modeling approach described above, we present the model fit statistics for police performance. These results are shown in Table 1.

Table 1.

Model fit statistics for police performance.

	Degrees of freedom	Non-parametric degrees of freedom	Non-parametric F statistic	p
Intercept	1
S(Encounter Length, 9)	1	8	2.106	0.032
Null deviance	Null deviance degrees of freedom	Residual deviance	Residual deviance degrees of freedom	AIC
618.15	1012	603.77	1003	2372.57

Figure 2 presents the main effects (measured from − .4 to + .4) from the GAM for perceptions of police performance with their 95% confidence intervals (dotted lines) plotted against encounter length (solid line) as measured in seconds. Along the x axis of the graph is a ‘rug plot’ where every tick mark represents an RBT operation.

As Figure 2 shows, there is a general positive slope for encounter length, indicating that individual citizens' positive perceptions of police performance increase as the length of the encounter increases. However, the only positive significant effect is for a period of time between 116 and 118 seconds (close to the 2-minute mark) (see Table 2). This is where the solid (main effect) and dotted lines (95% confidence intervals) all sit above zero.

Table 2.

Police performance – positive significant effect sizes.

Time (s)	Main effect size	CI	p	N
116.13	0.124	(0.018, 0.230)	0.026	34
116.42	0.122	(0.017, 0.227)	0.026	37
118.11	0.106	(0.002, 0.210)	0.048	45

This result suggests that RBT encounter lengths that lasted for the shortest periods of time (i.e. a minute or a minute and a half) were probably of insufficient time to fully deliver the procedurally just script. Short encounters, for example, might be too quick to elicit voice or to convey trustworthy motives. The other two elements of procedural justice – dignity and respect, and neutrality – are likely easier to convey within a short encounter length. QCET, however, was not designed to tease out which of the four procedural justice elements delivered the best outcomes.

Our results also show that the drop-off in positive perceptions of police performance for long encounters (i.e. those lasting longer than two minutes) suggests that police chatting for too long with drivers is likely to have a backfire effect: People are happy to be engaged in a procedurally just way, but if the dialogue goes on too long, people do not report as positive a perception of police performance. Put another way: To achieve the just right length of a procedurally just traffic encounter – what we call the Goldilocks optimal length of encounter – police need to keep RBT traffic stop encounters to just under two minutes.

Perceptions of police procedural justice

The model fit statistics for police procedural justice are shown in Table 3.

Table 3.

Model fit statistics for police procedural justice.

	Degrees of freedom	Non-parametric degrees of freedom	Non-parametric F statistic	p
Intercept	1
S(Encounter Length, 9)	1	8	3.227	0.001
Null deviance	Null deviance degrees of freedom	Residual deviance	Residual deviance degrees of freedom	AIC
635.85	1020	615.92	1011	2403.439

Figure 3 presents the main effects from the GAM for perceptions of police procedural justice with their 95% confidence intervals plotted against encounter length.

Like Figure 2, Figure 3 shows that there is a general, positive slope for encounter length, indicating an increase in individual citizens' positive perceptions of procedural justice the longer the encounter length. Figure 3 also shows that there are two areas where the confidence interval clearly excludes zero, which indicates encounter length times that generate significant positive effects on perceptions of police procedural justice. These occur at between 110 and 116 seconds and then again at 151 seconds (see Table 4).

Table 4.

Police procedural justice – positive significant effect sizes.

Time (s)	Main effect size	CI	p	N
109.83	0.184	(0.034, 0.335)	0.020	36
111.25	0.183	(0.048, 0.319)	0.010	38
116.13	0.114	(0.007, 0.221)	0.040	34
116.42	0.108	(0.002, 0.213)	0.049	38
151.02	0.270	(0.085, 0.456)	0.006	46

We note that the second, and largest, significant effect is at the 151 second mark. However, this is a single point in time, whereas the continuous window of significant effects between 110 to 116 seconds suggests a more stable result (we elaborate on this issue below). We also note that the 110–116 second window is also significant for Police Performance. Taken together, we propose that the just right, Goldilocks length of time for procedurally just traffic stops is between 110 and 116 seconds (just under 2 minutes, about 1 minute 50 seconds).

The second significant and positive result for longer encounter lengths shown in Figure 3 (and to a lesser extent, Figure 2) warrants further inquiry. The result suggests that both perceptions of police performance (Figure 2) and procedural justice (Figure 3) are positive at the extreme end of longer encounter lengths. While the estimated effects for the longer time period do show significant effects in Figure 3, these are possibly due to either an outlier, or boundary effects from the fit. When we examined the three RBT operations that seemed to suggest that there is an added benefit for some extra-long encounters, we found that all three occurred on the same weekend (12, 13 and 14 of February 2010). When we examined the ‘within operation' recordings of encounter length, we could not identify anything to suggest that the recordings were done incorrectly. Indeed, the researchers monitoring the three operations were experienced, and there was nothing unusual about the officer supervision on those dates. The weather was neither very hot nor very cold. Children return to school after the summer break at the end of January, so the mid-February weekend is unremarkable. The only explanations that we can propose as to why this cluster of operations generated a somewhat anomalous finding are that perhaps the group of officers rostered on for that weekend were especially engaging, or maybe it was because it was the Valentine's Day weekend and people were just generally happier.

Discussion and conclusion

Police officers spend a significant amount of their time interacting with citizens on traffic-related activities. We know that as much as 50% of officer time is spent on traffic tasks (Liederbach, 2005; Liederbach & Frank, 2003) and that over 90% of contacts that citizens have with police involve traffic or motor vehicle violations (Tyler & Folger, 1980). In Australia, the most common measure for detecting and deterring drink driving is through the RBT roadside testing of drivers (Bates, Soole, & Watson, 2012).

Given the fact that traffic stops are often negative or neutral experiences for drivers, the QCET team designed and executed a randomised field trial that used RBT operations to test the impact of procedurally just dialogue on citizen perceptions of police. In this paper, we used the QCET data to assess whether there is an optimal length of time that police should be engaged in procedurally just dialogue with drivers. We hypothesised that if the encounter length was too short, the police would not be able to properly engage drivers. We argued that a short, yet procedurally just, encounter may not sufficiently elicit voice or convey trustworthy motives. Yet, we also thought that an encounter that went too long might annoy a driver and undo the positive first few minutes of the encounter.

Our research extends the policing literature in two ways: Is there a threshold effect of time spent by police during police interventions (Koper, 1995), and does this threshold hold when the procedurally just quality of the encounter is held constant? We know, for example, that citizens have more positive views of police when police treat citizens with dignity and respect, convey trustworthy motives, give voice to citizens and are neutral in their decision making (see Mazerolle, Bennett, et al., 2013a). However, in our paper, we ask whether there is a point at which the police can over-do engaging with people using procedurally just dialogue.

Our results suggest that there is a general, positive relationship between the length of the procedurally just encounter and perceptions of police: The longer the procedurally just encounter, the more positive the perceptions that citizens have of police, as measured by perceptions of police performance and perceptions of procedural justice. Yet, we also show that there is an optimal length of time (the Goldilocks amount of time) for eliciting positive perceptions of police. We show that if police are using procedurally just dialogue, there is an ideal length of time where the encounter generates the best results in terms of citizen perceptions of police performance and procedural justice. Engaging with citizens during RBT encounters for about one minute 50 seconds is the best length of time, whereas a short, 1 minute procedurally just dialogue is too quick.

Prior research suggests that it is important for police to include the elements of procedural justice in their traffic-related encounters with citizens (Engel, 2005; Tyler & Folger, 1980; Wells, 2007). However, the results of our study suggest that it is also important for police officers to consider the length of this interaction. If police officers spend too long on the interaction, citizens are likely to react negatively to what they undoubtedly perceive as time wasting activity on the part of the police. Indeed, our results show that if police detain citizens for too long during RBT encounters, there is likely a backfire effect: All the good work of the first 2 minutes of a procedurally just RBT encounter is washed away if police chat to people for too long, because people want to get on their way. As measured by citizen perceptions of police performance and procedural justice, people seem to get annoyed with police if police talk to them too much. Notwithstanding the three outlier cases over the Valentine's Day weekend – where very long encounters did not show a backfire effect – the key message is that procedurally just encounters need not go overboard.

While this study suggests that there is an optimal time range for procedurally just police–citizen interactions during routine traffic stops, we recognise some important limitations of our study. First, our study did not record the time of each of the 20,000 plus encounters across all 60 operations. Instead, researchers timed the duration of encounters for the lines of seven to 10 officers and then averaged these duration times to generate the average encounter length time for each of the 60 RBT operations. As such, our study has no way of assessing the precise relationship between what an individual driver experienced in terms of their encounter length and the impact on their survey responses.

Second, we note that QCET achieved a low response rate for both the experimental and control survey respondents. While analysis of these low response rates suggests that they do not alter the substantive findings (see Antrobus, Elffers, White, & Mazerolle, 2014), it is clearly the case that a higher response rate is important to reassure scholars and practitioners alike that the differences between the two groups are not biased in any way.

Third, QCET was not designed to disentangle the four components of procedural justice (voice, trustworthy motives, dignity and respect, and neutrality). As such, we have no way of knowing which of the four ingredients might have been short-changed in the briefer experimental encounters. For example, the officers working in the experimental condition might have delivered the full script as required, yet failed to spend sufficient time to give the driver enough voice. We surmise that the ingredients of dignity and respect and neutrality were most likely to be conveyed in all of the experimental encounters, even the shortest ones. We suggest, however, that the shorter encounter lengths most likely failed to elicit voice, or were, perhaps, unable to convey trustworthy motives to the degree that was possible within the context of the longer encounter lengths. These are, however, only speculative ideas that would need to be tested under new field trial conditions.

Fourth, our study only included RBT encounters. By design, we have not included any other type of encounter that police experience on a daily basis with citizens. As such, our study is not about suggesting that there is a 2-minute optimal encounter length for police–citizen interactions. This 2-minute finding is, of course, only relevant to RBTs. We do suggest, however, that there is likely an optimal encounter length time for different types of police operations (hot spots, problem-oriented policing, foot patrol), in different types of encounter situations (such as dealing with domestic violence, break and enter, robberies, assaults). Clearly, further trials are needed to locate the optimal dosage time across different types of problems and police operations.

Despite these limitations, our study provides some evidence that there is a need for police officers to invest in procedurally just dialogue while interacting with citizens during RBTs, traffic stops and other types of encounters more generally. However, if police officers spend too long talking with citizens during the interaction (even using procedurally just dialogue), citizens' perceptions of police performance and procedural justice can be compromised. Too much engagement could be perceived by citizens as time wasting. Thus, police officers should engage with citizens in a procedurally just manner, yet ensure that they manage the interaction so that the length of the interaction is long enough to be effective, but not too long either.

Footnotes

Acknowledgements

The authors thank the team of researchers from The University of Queensland (Institute for Social Science Research) and Griffith University who participated in a variety of ways to bring this trial to fruition. The partnership between the research team and the Queensland Police Service is particularly acknowledged. The views expressed in this material are those of the authors and are not those of the Queensland Police Service. Responsibility for any errors of omission or commission remains with the authors. The Queensland Police Service expressly disclaims any liability for any damage resulting from the use of the material contained in this publication and will not be responsible for any loss, howsoever arising, from use or reliance on this material.

Funding

The research reported in this paper was funded, in its entirety, by the Australian Research Council (ARC) Centre of Excellence in Policing and Security (CEPS) (ARC Grant SR0700002).

Notes

Professor Lorraine Mazerolle is the Foundation Director and a Chief Investigator in the Australian Research Council (ARC) Centre of Excellence in Policing and Security (CEPS). She is an ARC Laureate Fellow and Research Professor in the Institute for Social Science Research (ISSR) at The University of Queensland (UQ) and leads a team of highly talented, multi-disciplinary scholars from the CEPS node at UQ. She is the recipient of numerous US and Australian national competitive research grants on topics such as community regulation, problem-oriented policing, police technologies, civil remedies, street-level drug enforcement and policing public housing sites. Professor Mazerolle is a Fellow of the Academy of Experimental Criminology, immediate past President of the Academy, foundation Vice President of the American Society of Criminology Division of Experimental Criminology and author of scholarly books and articles on policing, drug law enforcement, regulatory crime control, displacement of crime and crime prevention.

Dr Lyndel Bates is a Lecturer with the Centre for Accident Research and Road Safety – Queensland (CARRS-Q), School of Psychology and Counselling, at the Queensland University of Technology. She has over a decade of experience in road safety-related research and extensive teaching experience at a tertiary level in criminology and criminal justice. Her research interests include road law enforcement and road safety, graduated driver licensing and the use of research to improve and inform policy.

Dr Sarah Bennett is a research fellow in the ARC Centre of Excellence in Policing and Security (CEPS) at the Institute for Social Science Research (ISSR). Sarah is an experimental criminologist with experience in running multi-site randomized controlled trials with police in Australia and the UK and was the Project Leader for the Queensland Community Engagement Trial. Sarah's research interests include legitimacy and policing, restorative justice, pathways to preventing offending and the impact of crime on victims. Sarah is a Fellow of the Academy of Experimental Criminology (AEC) and recipient of the distinguished AEC Young Scholar Award and Nigel Walker Prize (Cambridge University).

Dr Gentry White graduated from University of Missouri in 2006 with a PhD in Statistics. He was a VIGRE Postdoctoral Fellow at North Carolina State University and an Associate Postdoctoral Fellow at the Statistical and Mathematical Sciences Institute from 2006 to 2009. His areas of research include spatial data analysis and state-space modelling. Dr White is currently a Research Fellow with the Australian Research Council (ARC) Centre of Excellence in Policing and Security (CEPS), based at The University of Queensland.

Mr Jason Ferris is a senior research fellow in the Australian Research Council (ARC) Centre of Excellence in Policing and Security (CEPS) at the Institute for Social Science Research (ISSR) based at The University of Queensland. Jason has over 10 years of public health research experience. He has a well-established publication record with a strong focus on alcohol and drug research, sexual health and adolescent health. With a master's degree in biostatistics, he has well developed and expansive quantitative methods skills. He is also highly experienced in varying data collection methods. He provides statistical analysis, guidance and mentorship to research staff as well as external consultancies.

Dr Emma Antrobus is a Research Fellow in the ARC Centre of Excellence in Policing and Security (CEPS) at the Institute for Social Science Research (ISSR) based at The University of Queensland. Emma has a background in psychology and jury decision making, particularly in relation to the special measures involved with child witness testimony. Her current research interests are focused on randomised controlled trials examining the impact of legitimate policing, and interventions for young people at risk.

References

Akaike, H. (1973). Information theory and an extension of the maximum likelihood principle. In B. N. Petrov, & F. Csaki (Eds.), Proceedings of the second international symposium on information theory (pp. 267–281). Budapest: Akademiai Kiado.

Antrobus, E., Elffers, H., White, G., & Mazerolle, L. (2014). Non-response bias in randomized control experiments in criminology: Putting the Queensland Community Engagement Trial (QCET) under a microscope. Evaluation Review. Advance online publication. doi: 10.1177/0193841X13518534.

Audija

Giever

(2010) The influence of citizen interaction the police on crime-reporting behavior: Its manifestations among university students. Kriminologija and Socijalna Integracija 18(2): 45–61.

Australian Bureau of Statistics (2012) 2011 census of population and housing basic community profile: Queensland (Cat No. 2001.0), Canberra: Australian Bureau of Statistics.

Bates

Soole

Watson

(2012) The effectiveness of traffic policing in reducing traffic crashes. In: Prenzler

(ed) Policing and security in practice: Challenges and achievements, Houndmills, UK: Palgrave Macmillan, pp. 90–109.

Braga

A. A.

(2005) Hot spots policing and crime prevention: A systematic review of randomized controlled trials. Journal of Experimental Criminology 1: 317–342.

Braga

A. A.

(2007) The effects of hot spots policing on crime. Campbell Collaboration Library of Systematic Reviews 3(1): 1–94. doi: 10.4073/csr.2007.1.

Cheurprakobkit

(2000) Police-citizen contact and police performance: Attitudinal differences between Hispanics and non-Hispanics. Journal of Criminal Justice 28(4): 325–336.

Chow

(2010) Police-public relations: Perceptions of the police among university students in a western Canadian city. International Journal of Criminology and Sociological Theory 3(2): 496–511.

10.

Chow

(2012) Attitudes towards police in Canada: A study of perceptions of university students in a western Canadian city. International Journal of Criminal Justice Sciences 7(1): 508–523.

11.

Cook

T. D.

Campbell

D. T.

(1979) Quasi-experimentation: Design & analysis issues for field settings, Chicago, IL: Rand McNally College Publishing.

12.

Engel

R. S.

(2005) Citizens' perceptions of distributive and procedural injustice during traffic stops with police. Journal of Research in Crime and Delinquency 42(4): 445–481. doi: 10.1177/0022427804272725.

13.

Famega

C. N.

(2009) Proactive policing by post and community officers. Crime & Delinquency 55(1): 78–104. doi: 10.1177/0011128707310160.

14.

Ferris

J. A.

Mazerolle

King

Bates

Bennett

Devaney

(2013) Random breath testing in Queensland and Western Australia: Examination of how the random breath testing rate influences alcohol related traffic crash rates. Accident Analysis and Prevention 60: 181–188.

15.

Fildes

(1995) Driver behaviour and road safety. In: Brewer

Wilson

(eds) Psychology and policing, Hillsdale, NJ: Lawrence Erlbaum Associates, pp. 31–61.

16.

Freeman

J. E.

Watson

B. C.

(2009) Drink driving deterrents and selfreported offending behaviours in a sample of Queensland motorists. Journal of Safety Research 40(2): 113–120.

17.

Hastie, T. (2013). Generalized Additive Models (R Package Ver. 1.08) [Statistical software]. Retrieved from http://cran.r-project.org/web/packages/gam/index.html.

18.

Hastie

Tibshirani

R. J.

(1990) Generalized additive models (Vol. 43), London: Chapman & Hall.

19.

Hastie

Tibshirani

R. J.

Friedman

J. H.

(2001) The elements of statistical learning: Data mining, inference, and prediction, New York, NY: Springer.

20.

Hinds

Murphy

(2007) Public satisfaction with police: Using procedural justice to improve police legitimacy. The Australian and New Zealand Journal of Criminology 40(1): 27–42.

21.

Hoover

L. T.

Dowling

J. L.

Fenske

J. W.

(1998) Extent of citizen contact with police. Police Quarterly 1(3): 1–18. doi: 10.1177/109861119800100302.

22.

Johnson

(2004) Citizen expectations of police traffic stop behavior. Policing: An International Journal of Police Strategies and Management 27(4): 487–497.

23.

Kochel

T. R.

Parks

Mastrofski

S. D.

(2011) Examining police effectiveness as a precursor to legitimacy and cooperation with police. Justice Quarterly 30(5): 895–925. doi: 10.1080/07418825.2011.633544.

24.

Koper

C. S.

(1995) Just enough police presence: Reducing crime and disorderly behavior by optimizing patrol time in crime hot spots. Justice Quarterly 12(4): 649–672.

25.

Liederbach

(2005) Addressing the “elephant in the living room”: An observational study of the work of suburban police. Policing: An International Journal of Police Strategies and Management 28(3): 415–434.

26.

Liederbach

Frank

(2003) Policing Mayberry: The work routines of small-town and rural officers. American Journal of Criminal Justice 28(1): 53–72.

27.

Mastrofski

Snipes

Supina

A. E.

(1996) Compliance on demand: The public's response to specific police requests. Journal of Research in Crime and Delinquency 33(3): 269–305. doi: 10.1177/0022427896033003001.

28.

Mazerolle, L., Antrobus, E., Bennett, S., & Tyler, T. R. (2013). Shaping citizen perceptions of police legitimacy: A randomized field trial of procedural justice. Criminology, 51, 33–64. doi: 10.1111/j.1745-9125.2012.00289.x.

29.

Mazerolle

Bennett

Antrobus

Eggins

(2012) Procedural justice, routine encounters and citizen perceptions of police: Main findings from the Queensland Community Engagement Trial (QCET). Journal of Experimental Criminology 8: 343–367. doi: 10.1007/s11292-012-9160-1.

30.

Mazerolle

Bennett

Davis

Sargeant

Manning

(2013a) Legitimacy in policing: A systematic review. The Campbell Collaboration Library of Systematic Reviews 9(1): 1–146. doi: 10.4073/csr.2013.1.

31.

Mazerolle

Bennett

Davis

Sargeant

Manning

(2013b) Procedural justice and police legitimacy: A systematic review of the research evidence. Journal of Experimental Criminology 9: 245–274. doi: 10.1007/s11292-013-9175-2.

32.

Mazerolle

Bennett

Eggins

Antrobus

White

Davis

(2011) The Queensland Community Engagement Trial: Final Report, Brisbane, Australia: ARC Centre for Excellence in Policing and Security.

33.

Mazerolle

Sargeant

Cherney

Bennett

Murphy

Antrobus

Martin

(2014) Procedural justice and legitimacy in policing. SpringerBriefs in translational criminology, New York, NY: Springer.

34.

Murphy

(2009) Public satisfaction with police: The importance of procedural justice and police performance in police-citizen encounters. The Australian and New Zealand Journal of Criminology 42(2): 159–178. doi: 10.1375/acri.42.2.159.

35.

Myhill

Bradford

(2011) Can police enhance public confidence by improving quality of service? Results from two surveys in England and Wales. Policing & Society: An International Journal of Research and Policy 22: 397–425. .

36.

Queensland Police Service (2012) Annual report 2011–12, Brisbane: Queensland Police Service.

37.

R Core Team (2013) R: A Language and Environment for Statistical Computing [Statistical software], Vienna, Austria: R Foundation for Statistical Computing. Retrieved from http://www.R-project.org/.

38.

Ratcliffe

J. H.

Taniguchi

Groff

E. R.

Wood

J. D.

(2011) The Philadelphia foot patrol experiment: A randomized controlled trial of police patrol effectiveness in violent crime hotspots. Criminology 49(3): 795–831. doi: 10.1111/j.1745-9125.2011.00240.x.

39.

Reisig

M. D.

Lloyd

(2009) Procedural justice, police Legitimacy, and helping the police fight crime: Results from a survey of Jamaican adolescents. Police Quarterly 12(1): 42–62. doi: 10.1177/1098611108327311.

40.

Roslow

Nicholls

J. A. F.

Tsalikis

(1992) Time and quality: Twin keys to customer service satisfaction. Journal of Applied Business Research 8(2): 80–86.

41.

Schafer

J. A.

Huebner

Bynum

(2003) Citizen perceptions of police services: Race, neighborhood context and community policing. Police Quarterly 6(4): 440–468.

42.

Sherman

L. W.

Weisburd

(1995) General detterent effects of police patrol in crime “hot spots”: A randomized, controlled trial. Justice Quarterly 12: 625–648.

43.

Skogan, W. G. (2006). Asymmetry in the impact of encounters with police. Policing and Society: An International Journal of Research and Policy, 16(2), 99–126. doi: 10.1080/10439460600662098.

44.

Sunshine

Tyler

(2003) The role of procedural justice and legitimacy in shaping public support for policing. Law and Society Review 37(3): 513–547.

45.

Taylor

Koper

C. S.

Woods

D. J.

(2011) A randomized controlled trial of different policing strategies at hot spots of violent crime. Journal of Experimental Criminology 7: 149–181. doi: 10.1007/s11292-010-9120-6.

46.

Telep

C. W.

Mitchell

R. J.

Weisburd

(2012) How much time should the police spend at crime hot spots? Answers from a police agency directed randomized field trial in Sacramento, California. Justice Quarterly. (pp. 1–29). doi: 10.1080/07418825.2012.710645.

47.

Tyler

T. R.

(2004) Enhancing police legitimacy. The Annals of the American Academy of Political and Social Science 593: 84–99. doi: 10.1177/0002716203262627.

48.

Tyler

T. R.

Fagan

(2008) Legitimacy and cooperation: Why do people help the police fight crime in their communities? Ohio State Journal of Criminal Law 6: 231–275.

49.

Tyler

T. R.

Folger

(1980) Distributional and procedural aspects of satisfaction with citizen-police encounters. Basic and Applied Social Psychology 1(4): 281–292. doi: 10.1207/s15324834basp0104_1.

50.

Watson

B. C.

Freeman

J. E.

(2007) Perceptions and experiences of random breath testing in Queensland and the self-reported deterrent impact on drink-driving. Traffic Injury Prevention 8(1): 11–19.

51.

Weisburd

Eck

J. E.

Hinkle

J. C.

Telep

(2008) The effects of problem-oriented policing on crime and disorder: A systematic review. The Campbell Collaboration Library of Systematic Reviews 4(14): 1–88. doi: 10.4073/csr.2008.14.

52.

Weisburd

Gill

Bennett

Telep

(2012) Community-oriented policing to reduce crime, disorder and fear and increase legitimacy and citizen satisfaction in neighborhoods (Title registration). Campbell Collaboration Library of Systematic Reviews. Retrieved from http://campbellcollaboration.org/lib/project/202/.

53.

Wells

W. M.

(2007) Type of contact and evaluations of police officers: The effects of procedural justice across three types of police-citizen contacts. Journal of Criminal Justice 35: 612–621. doi: 10.1016/j.jcrimjus.2007.09.006.

54.

Worrall

J. L.

(1999) Public perceptions of police efficacy and image: The ‘fuzziness’ of support for the police. American Journal of Criminal Justice 24(1): 47–66.