Article Text

Download PDFPDF

The effect on collisions with injuries of a reduction in traffic citations issued by police officers
  1. Étienne Blais1,2,3,
  2. Marie-Pier Gagné1
  1. 1School of Criminology, University of Montreal, Quebec, Canada
  2. 2Interuniversity Research Centre on Enterprise Networks, Logistic and Transportation, University of Montreal, Quebec, Canada
  3. 3Quebec National Institute for Public Health (INSPQ), Quebec, Canada
  1. Correspondence to Professor Étienne Blais, School of Criminology, University of Montreal, C.P. 6128, succursale Centre-ville, Montréal, QC H3C 3J7, Canada; etienne.blais{at}


Objective To assess the effect on collisions with injuries of a 61% reduction in the number of traffic citations issued by police officers over a 21-month period.

Methods Using descriptive analyses as well as ARIMA intervention time-series analyses, this study estimated the impact of this reduction in citations issued for traffic violations on the monthly number of collisions with injuries.

Results Simple descriptive analysis reveals that the 61% reduction in the number of citations issued for traffic violations during the experimental period coincided with an increase in collisions with injuries. Results from the interrupted time-series analyses reveal that, on average, eight additional collisions with injuries occurred every month during which the number of tickets issued for traffic violations was lower than normal. As this pressure tactic was applied for 21 months, it is estimated that this situation was associated with approximately 184 additional collisions with injuries: equivalent to 239 traffic injuries (either deaths, minor or serious injuries).

Conclusion In the province of Quebec, police officers are an important component of road safety policy. Issuing citations prevents drivers from adopting reckless driving habits such as speeding, running red lights and failing to fasten their seat belt.

  • Deterrence
  • driver
  • evaluation
  • interrupted time series
  • law enforcement
  • policy
  • public health
  • traffic injuries
  • traffic violations

Statistics from

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.

According to the WHO approximately 1.3 million people die each year from traffic injuries and some 20–50 million sustain injuries.1 Averages based on the 2001–7 data show that, in the province of Quebec, Canada, there are, each year, more than 650 motor vehicle fatalities and more than 49 000 traffic injuries that require either immediate hospitalisation or some kind of medical care.2 Despite noticeable improvements over the past 30 years, drinking and driving and speeding are still the two major causes of traffic injuries.2 3

The legal system has been used worldwide to deal with problems associated with reckless driving habits. The legal approach to reducing the losses from collisions caused by these prohibited driving practices is based on a set of official rules and regulations formalised in traffic safety and criminal codes. These codes define as well as prohibit behaviours known to increase the risk of traffic collisions. Failure to comply with such codes would result in such punishments as fines, demerit points, licence suspensions or even jail, the threat of which is believed to act as a deterrent to the prohibited behaviours.4

The term deterrence denotes instances in which individuals refrain from an illegal act in response to the perceived threat of punishment. The deterrence doctrine assumes that the effectiveness of the legal threat is a function of the perceived certainty, severity and celerity of punishment in cases of offence.5 The greater the likelihood of arrest, prosecution, conviction and punishment, the stronger the effect of the legal threat.4 6 Evidence from systematic reviews shows that collisions with injuries decrease as the risk of being arrested increases as a result of police enforcement campaigns such as random breath testing programmes, sobriety checkpoints, random road-watch programmes and the installation of automatic control devices such as speed and red light cameras.4

Although scientific evidence highlights the need to increase the risk of arrest in order to strengthen the deterrent produced by the threat of legal punishment,4 6 few studies have investigated the opposite effect, ie, situations characterised by a significant reduction in the likelihood of being arrested for traffic violations by police officers. In February 1976, the Finnish police went on strike for approximately 2 weeks. As a result, the percentage of gross speed violations (exceeding the speed limit by at least 10 km/h) went up by 50–100% and the SD of speeds increased by 20%. Although these results suggest that the risk of collisions should have increased, collisions data were not analysed.7

In an attempt to predict the monthly frequency of collisions in the province of Quebec, Canada, Gaudry and colleagues8 have estimated the effect on various categories of collisions of major disruptions in the enforcement activities of two police forces representing approximately one-third of all the province's police forces at the time of the study: the Service de Police de la Ville de Montréal and the Sureté du Québec. These disruptions were characterised, among other things, by substantial reductions in the number of arrests for traffic violations. Based on their findings, the authors concluded that such disruptions are not likely to increase collisions (all p values were greater than 5%).

Combining previous results could suggest that significant reductions in the certainty of being arrested for traffic violations will not translate into increased collisions. However, previous studies have failed to consider two major issues in macro-level deterrence research. The first issue is the proper identification of the spatial unit of aggregation that best estimates the potential for the transmission of sanction-related information by social networking across territorial areas.9 10 Most studies specify a geographical area as the unit of analysis and they assume, by design, that estimates of the level of threat may be effectively transmitted across space over a specified temporal period. This assumption is debatable because it is neither geographical nor political areas that guide police deployment and restrict enforcement behaviour but rather the administrative jurisdiction in which they operate. Accordingly, major disruptions in Sureté du Québec and Service de Police de la Ville de Montréal services are likely to affect the road toll of their respective jurisdictions but not that of the whole province.

A second issue concerns conditions under which the threat of legal sanctions will act as a deterrent. With the exception of one case, all perturbations included in the study of Gaudry et al8 lasted one day or less. On the other hand, several deterrence studies point out that a minimal arrest risk must be reached and maintained long enough to allow potential offenders to consider information about the legal threat when they evaluate the cost of engaging in illegal activities.6 9 11 For instance, the Montreal Police Department unit in charge of traffic safety had to double citations for a period of 24 months in order to reverse the upward trend in collisions with injuries in their jurisdiction.12

This study is therefore designed to further our understanding of the effects of a reduction in the number of traffic citations on collisions with injuries. Between November 2003 and July 2005, unionised police officers of Quebec City, Canada, conducted a pressure tactic consisting of reducing the number of citations issued for traffic violations in order to accelerate the negotiation of their new work contract.i Using interrupted time-series analyses techniques, we test whether this situation worsened the road toll. In keeping with the recent developments in macrolevel deterrence research, we operationalise the unit of spatial aggregation based on the independent variable (ie, the group whose behaviour serves as the hypothesised mechanism); the unit used is thus the police jurisdiction instead of a geographical or political jurisdiction.



Data on collisions and arrests were provided by the Société de l'Assurance Automobile du Québec (SAAQ), the Quebec crown corporation responsible for road safety, driver licensing and permits and motor vehicle registration. The SAAQ receives all police reports on collisions and administers a system of tickets, demerit points and licence suspensions.

The SAAQ defines a traffic collision as any event in which damages are caused by a moving vehicle. The police officer on duty at the scene of a collision must submit an electronic collision report form to the SAAQ within 8 days. Insurers are obliged to inform the SAAQ when a collision has not been reported to the police. Coroners must also provide the SAAQ with copies of their investigations.2

In the province of Quebec, traffic violations are either specified in the Road Safety Code or in the Canadian Criminal Code and, in both cases, they are enforced by police officers. Violations such as driving while impaired by alcohol and reckless driving causing injury are offences under the criminal code, whereas other violations—such as speeding, running a red light and failing to fasten seat belts—are offences under the Road Safety Code. The latter offences result in fines and demerit points. Apprentice and regular licence holders are respectively allowed 4 and 15 demerit points. Demerit points remain in the driver's record for 2 years after the citation was issued. The accumulation of demerit points above the preset threshold will automatically lead to suspension of the licence.13

Variables under analysis

Our main dependent variable—monthly collisions with injuries for Quebec City—is an aggregate of: fatal collisions resulting in the death of at least one vehicle occupant within 8 days of the incident; collisions resulting in serious injuries leading to the hospitalisation of at least one individual; and collisions with minor injuries that do not require hospitalisation but may require other types of medical care.

To avoid any problems regarding validity, two control series are used: ‘collisions with material damages of $1000 or more for Quebec City’ is used as a non-equivalent dependent variable, and ‘collisions with injuries for the Province of Quebec’ (excluding those occurring in the jurisdiction of the Quebec City Police Department) is used as a non-equivalent no-treatment control group time series.14 15

Two sets of independent variables were used for the purpose of our analyses. First, the pressure tactic variable was coded as a step function: time units included in the intervention period (also referred to as the experimental period) were coded 1 (November 2003 to July 2005), with units in the control period coded 0 (January 2001 to October 2003, then from August 2005 to October 2007). We also created an interaction term ‘time×intervention’, which is a continuous variable counting the number of months during the intervention. These variables can respectively estimate abrupt and gradual changes in the series due to the pressure tactic.14 Second, we considered the number of arrests. This variable served to estimate the dose–response relationship between the number of arrests and collisions with injuries. According to the SAAQ, more than 75% of all citations were handed out for speed-limit violations in Quebec City during the period under investigation.13

Statistical analysis

The analyses were based on a total of 81 monthly temporal units, respectively, 60 and 21 in the control and experimental periods. In the first stage of the analysis, we computed the average number of collisions and arrests per month from January 2001 to September 2007 for both periods.

We next conducted integrated moving average (ARIMA) intervention time-series models to evaluate the effect on collisions with injuries of the reduction in the number of arrests.16 17 Intervention ARIMA models estimate the effect of an intervention at a specific point in time while controlling for trends before the intervention. Unlike linear regression models, ARIMA models can take into account autocorrelations between units of analysis.

As each series has a unique structure, ARIMA models are developed using a three-stage iterative process: (1) identification; (2) estimation and (3) diagnostic. Before performing these stages, ARIMA modelling requires each series to achieve stationarity. The augmented Dickey–Fuller test was conducted to identify the presence of non-stationarity and to determine whether series required differentiation (stochastic trend) or the integration of a trend variable (deterministic trend) to be stationary.

Identification involves analysis of autocorrelation and partial autocorrelation matrices to identify autoregressive and moving average parameters. The predictive validity of these parameters is subsequently estimated in the estimation stage. As each series exhibited seasonality, seasonal parameters were first identified and error terms were then re-inspected to identify other lags with significant Q-statistics. In the diagnostic stage, errors terms were plotted to make sure that they followed a white noise process. In all models, Q-statistics are not significant (p>0.05). Error terms are normally distributed, and the homoskedasticity assumption is respected. Once the models have been well specified, the independent variables are included in the model. All analyses are conducted with STATA 10.0 and the TREND module in SPSS 15.0.


In Quebec City, between January 2001 and September 2007, there were on average 146.65 collisions with injuries and 612.68 collisions with material damages only per month (table 1). There were on average 3344.39 arrests per month for traffic violations. On average, during the pressure tactic, arrests for traffic violations fell by 61%, and collisions with injuries and material damages increased by 3.73 and 7.02%, respectively.

Table 1

Monthly average for collisions and arrests for the experimental and control periods

Even though table 1 reports steeper increases in the number of collisions in Quebec City than in the rest of the province, these results preclude drawing any valid conclusions about the effect of the reduction in traffic citations on collisions. Pre-intervention trends and seasonal effects could account for the increase in collisions.15

In ARIMA intervention time-series analyses with adjustment for autocorrelation, we found a significant increase in collisions with injuries in Quebec City (table 2). As shown by the intervention variable coefficient (B=8.75; p<0.02), collisions with injuries increased by approximately eight units per month following the beginning of the pressure tactic. Analyses based on the number of citations as the independent variables also show similar evidence, ie, a negative relationship between arrests and collisions with injuries (B=−0.05; p<0.01) (not shown). For collisions with material damages only, we found a reduction of approximately 64 collisions (B=−63.58; p<0.01) per month and we observed a significant and gradual increase in the trend (B=1.12; p=0.01). Collisions with injuries for the rest of the province of Quebec did not display any abrupt or gradual change following the introduction of the pressure tactic.

Table 2

Results of the ARIMA intervention time-series analyses

Although our results suggest that a reduction in the intensity of police arrests has a negative impact on the road toll, we ran two additional series of analyses to strengthen the validity of our results. First, the Dickey–Fuller may incorrectly indicate the presence of a stochastic trend while the latter is deterministic in nature, especially when the series exhibit signs of a structural change.18 We therefore conducted our analyses with a trend term in the collisions-with-injuries model to test whether the differentiation of the series may have influenced our results. Results were similar with a coefficient of 9.06 (p=0.02) for the intervention variable (not shown).

Second, socioeconomic indicators were not integrated in our models because these data were not available on the basis of police jurisdictions. To overcome potential validity threats related to history, simple selection and construct validity, we used a strategy known as ‘removed treatment’. It implies estimating the impact of one intervention first by considering the ‘before’ with the in-treatment periods and second by considering the in-treatment with the ‘afte’ treatment period.14 Results were similar independently of the selected period; the intervention variable was associated with a monthly rise of approximately eight to 10 collisions with injuries (p<0.05 in both models) (table 3).

Table 3

Results of the ARIMA intervention time-series analyses using the removed treatment strategy


Following the pressure tactic initiated by the unionised police officers of the Quebec City Police Department, which was characterised by an average reduction of 61% in the number of citations handed out for traffic violations, Quebec City experienced a significant increase in collisions with injuries. Results were consistent across all models (models with stochastic and deterministic trends, as well as models using the ‘removed treatment’ strategy). As there were, on average, 8.75 additional collisions with injuries (table 2) during this exceptional policing condition lasting 21 months, approximately 184 collisions with injuries could have been prevented if the pressure tactic had not taken place (or approximately 239 traffic injuries, because there were on average 1.3 individuals per collision causing injuries).

Intensive police programmes have helped to make significant reductions in traffic casualties and their severity by increasing arrests for impaired driving, speeding, not fastening the seat belt and running a red light.4 On the other hand, failure to enforce traffic codes can generate adverse effects. Previous studies have failed to find any association between ‘major police perturbations’ and collisions with injuries, but our study, unlike previous studies, integrates two key notions of the general deterrence theory. First, we acknowledge that the level and length of enforcement activities are central issues in macro-level deterrence research.9 11 Just as the provisions of a new law need to be strictly enforced long enough to allow potential offenders to be exposed to punishment threat and then consider the cost of violating them,6 19 a substantial and long-lasting reduction in enforcement activities seems likely to encourage violations and worsen the road toll. Second, we aggregate our data based on police jurisdictions.9 10 The present study is thus especially designed to detect the effect on collisions with injuries of any disruptions in regular police enforcement activities.

Our results combined with information found in Quebec City's main newspapers further highlight the central role threats of sanctions can play as agents of deterrence. Between November 2003 and July 2005, Le Journal de Québec and Le Soleil alone published 18 articles about the conflict between Quebec City's municipal authorities and the police union delegates, three of them focussing precisely on the reduction in traffic citations. The first article on the reduction in the number of arrests was published on 22 October 2004—almost 1 year after the beginning of the pressure tactic. On the other hand, 10 of the 18 articles were published between the months of January and July 2005. If the trend in collisions were dependent on drivers' reactions to media reports, a significant effect should have been observed for the interaction between the intervention and time after intervention variable, which was not the case. Our results are in keeping with studies showing that news about the risk of arrest is diffused very rapidly through interpersonal networks,20 21 and that media and publicity campaigns have limited influence on drivers' behaviours.22

In order to obtain valid results, we estimated the effect of the pressure tactic on collisions with ARIMA intervention time-series analysis. This type of quasi-experimental design is one of the most robust and is widely used to assess and predict outcomes related to public health, psychological and criminal justice interventions.14 15 23 By using such a design and performing appropriate statistical tests, we were able to control, when required, for previous trends in the series, for seasonality and for autocorrelation in the error terms, threats to validity that cannot be controlled for in simple before-and-after comparisons. Furthermore, we have included control series and used the ‘removed treatment’ strategy to increase the robustness of our conclusions.

Despite all the steps taken to ensure the validity of its results, this study has various limitations related to the type of injury collisions occurring in Quebec City. First, it was not possible to determine whether collisions with injuries were indeed related to speed-limit violations, despite the fact that most of the citations were issued for such violations. Overall, 93% of all the collisions with injuries involve minor injuries. Unlike fatal injury collisions, the causes of collisions with minor injuries are not well documented. Second, it was not possible to aggregate data with respect to the severity of injury. We attempted to estimate the effect of the pressure tactic on collisions with serious and fatal injuries, but we did not have enough variance between observations to do so. Third, and for similar reasons, it was not possible to stratify data by age groups to test for the presence of a reference group effect,24 ie, to see whether greater increases in collisions were recorded for age categories of drivers who were less likely to be arrested for a traffic violation.


In this study, we find that police incentives can significantly affect traffic injury trends. A 61% reduction in the monthly volume of traffic citations was associated with an increase of approximately eight collisions with injuries per month in Quebec City. In the province of Quebec, this pressure tactic is regularly used by unionised police officers to accelerate the negotiation of their new work contracts, as the Quebec Labour Code considers police activities an essential service for public safety and, consequently, police officers do not have the right to go on strike. On the other hand, the Council for Essential Services is a body with the power to forbid practices that may endanger public safety. Based on the lessons learnt from our study, public health authorities should be concerned about the pressure tactic investigated in this study.

What is already known on the subject

  • Traffic violations such as speeding, drunk driving and failing to fasten seat belts are important causes of collisions with injuries.

  • Police strikes and major perturbations in police enforcement activities are likely to lead to increases in traffic violations and collisions.

  • The number of traffic tickets issued by Quebec City police officers dropped by approximately 61% during the negotiation of their new collective agreement.

What this study adds

  • Using interrupted time-series analysis, we estimate the effect of such a negotiation strategy (pressure tactic) on collisions with injuries in Quebec City.

  • Our results show that this negotiation strategy was responsible for an additional 184 collisions with injuries, or 239 traffic injuries (either deaths or injuries).

  • Competent authorities should be concerned and should even consider banning this type of negotiation strategy because it jeopardises public health and safety.


The authors would like to thank Steven Sacks and Sybil Murray-Denis for reviewing and editing the manuscript. They are also grateful to Fernand Pichette from the SAAQ for providing data. Finally, they would like to thank the three anonymous reviewers for their precious comments.



  • Funding The results presented in this paper are part of a research programme on police activities and road safety. This programme is funded by the Fonds Québécois de la Recherche sur la Société et la Culture (FQRSC).

  • Competing interests None.

  • Provenance and peer review Not commissioned; externally peer reviewed.

  • i In this study, an arrest for a traffic violation means that the driver was stopped by a police officer and that a citation was issued. In this article, arrests and citations are used interchangeably.