Background The objective of the current study is to determine to what extent the reduction of Chile’s traffic fatalities and injuries during 2000–2012 was related to the police traffic enforcement increment registered after the introduction of its 2005 traffic law reform.
Methods A unique dataset with assembled information from public institutions and analyses based on ordinary least square and robust random effects models was carried out. Dependent variables were traffic fatality and severe injury rates per population and vehicle fleet. Independent variables were: (1) presence of new national traffic law; (2) police officers per population; (3) number of traffic tickets per police officer; and (4) interaction effect of number of traffic tickets per police officer with traffic law reform. Oil prices, alcohol consumption, proportion of male population 15–24 years old, unemployment, road infrastructure investment, years’ effects and regions’ effects represented control variables.
Results Empirical estimates from instrumental variables suggest that the enactment of the traffic law reform in interaction with number of traffic tickets per police officer is significantly associated with a decrease of 8% in traffic fatalities and 7% in severe injuries. Piecewise regression model results for the 2007–2012 period suggest that police traffic enforcement reduced traffic fatalities by 59% and severe injuries by 37%.
Conclusions Findings suggest that traffic law reforms in order to have an effect on both traffic fatality and injury rates reduction require changes in police enforcement practices. Last, this case also illustrates how the diffusion of successful road safety practices globally promoted by WHO and World Bank can be an important influence for enhancing national road safety practices.
Statistics from Altmetric.com
In 2004, WHO along with the World Bank (WB) launched the ‘World report on road traffic injury prevention’1 and globally promoted several road safety measures to tackle the roots of road crashes and avoid their tragic consequences. After 12 years of parliamentary and governmental debate, and following closely the recommendations set by the ‘World Report’ in 2005, Chile approved a comprehensive traffic law reform which via decree regulation was enacted at the end of 2006 (Urzúa J, personal communication). This reform included measures such as: (i) mandatory seat belt use for all vehicle occupants, (ii) introduction of child restraints, (iii) day-running lights for motorcycles, (iv) banning the use of cell phone while driving, (v) stricter requirements to obtain driver licenses to drive light vehicles and (vi) higher fines for driving under the influence of alcohol.2 Statistics indicate that traffic fatality and severe injury rates per 100 000 population declined by 13% and 5%, respectively, between 2006 and 2012. However, the role of Chile's traffic law reform—and by extension, those of the recommendations set by WHO and the WB—in this downward trend remains unknown.
While a vast literature suggests that road traffic legislation reforms per se may help to prevent the occurrence of traffic crashes and its consequences,3–10 others have argued that this type of measure is ultimately sterile or at best short-lived.11 ,12 The latter has been particularly noticeable in Latin American countries since in many cases they lack the necessary resources to implement legislative changes.13–15 However, other literature focus on institutional mechanisms16 ,17 which offer more precise explanations of the potential impact of legal reform. These studies have argued that road safety legislation reforms can only be effective in reducing crashes and traffic fatalities when they are followed up with particular types of institutional mechanisms, such as police enforcement (ie, stationary or patrolling18 ,19) or the creation of road safety agencies,1 as well as with the introduction of technological devices, such as red-light cameras,20–22 and automatic speed enforcement,23 ,24 youth-focused intervention programmes25 and/or diffusion of public campaigns.26–28
In this paper, we contribute to this body of research by using Chilean data between 2000 and 2012 to empirically assess the effects of (a) traffic law reform, (b) police officers per capita, (c) police traffic enforcement and (d) the interaction of traffic law reform with police traffic enforcement on traffic and morbidity rates. Such an assessment is evidently highly relevant for Chile as a policy impact evaluation, but it also has both theoretical and methodological implications for the field of study at large. First, unlike other federal countries such as Brazil, Canada or the USA, where traffic reforms may have not been universal, Chile, as a centralised country, provides an opportunity to examine an entire national jurisdiction unit over time. The latter therefore can be regarded as a ‘natural policy experiment’.29 Furthermore, it is worth noting that under Chile's administrative regime, in which a single national police body is responsible for patrolling traffic, coordination of enforcement should be less problematic and therefore a traffic law reform might uniformly intensify it. Second, as noted by Elvik et al,6 while there is a significant amount of research on the efficacy of road traffic legislation reforms and the institutional mechanisms associated with them in the USA, the Commonwealth and European countries, there remains a remarkable paucity of similar research exploiting cross-jurisdictional time-series variation in other regions of the world, and particularly in South America. Third, this case offers an explicit test of national pathways through which the influence of global forces such as the policy diffusion of the ‘World Report’ can ultimately contribute to changes in national territories can play out.30 Last, from a methodological point of view, the rich cross-region time-series data available for Chile allow us to control for the impact of unobserved region and/or time-specific determinants of traffic incidents such as climate variation with the help of random effects models.31 Indeed, Chile comprises a wide range of weather conditions extending across 38° in latitude. Simple ordinary least square (OLS) estimates of the impact of traffic reform, police officers per capita or police traffic enforcement might be otherwise confounded with unobserved determinants of traffic crashes such as climate variation within regions during 2000–2012. Hence, using Chilean data permits us to evaluate the robustness of the literature of road safety institutional mechanisms, expand our road safety knowledge of the South American region more generally and of this country more particularly, and contribute to our understanding of how, in the realm of road safety policies, global factors impact national territories.
Figure 1 shows a steady decline in Chilean traffic fatality and severe injury rates (per 10 000 vehicles) between 2000 and 2012. Specifically, traffic fatalities have dropped by roughly 52% ((7.97–3.83)/7.97) and severe injuries by 56% ((38.01–16.53)/38.01). This figure also shows a less pronounced decline in traffic fatality (per 100 000 population) and severe injury rates (per 10 000 population). Particularly, traffic fatalities under this indicator dropped by approximately 21% ((11.02–8.75)/11.02) and severe injuries by 28% ((5.25–3.77)/5.25).
Table 1 documents cross-regional variation in traffic fatality and serious injury rates in 2000 and 2012, the two end-year sample points for both rates: vehicle fleet and population. These numbers suggest significant cross-regional variation in fatality and severe injury rates in the year 2000. In 2000, five regions, O'Higgins, Maule, Biobío, La Araucanía and Los Lagos, had an average of 12.1 traffic fatality rates per 10 000 vehicles, while, with the clear exception of the Magallanes region, all the other regions had an average of 6.81. A very similar pattern is observed in fatalities per 100 000 population. We observe the Magallanes region to be the exception with the lowest rate of 4.56. On the other hand, O'Higgins, Maule, Biobío, La Araucanía, Los Lagos and Atacama had an average of 14.4, and all the other regions an average of 9.49. Regarding severe injury rates for the same year, we observe that Atacama, Coquimbo, Biobío, La Araucanía, Los Lagos and Aysén had an average of 46.18 per 10 000 vehicles, whereas the group composed of Tarapacá, Antofagasta, Valparaíso, O'Higgins, Maule and the Metropolitan region had an average of 34.80. For this particular rate as well Magallanes has a very low rate of 24.40. When we analyse the severity rate per population, Atacama is the region with the highest rate with an average of 79.17, Los Lagos region had the lowest rate with 44.27 and all the other regions had an average of 53.4. The first four columns for the year 2012 indicate that variation between and within regions has decreased overall.
Thus, the regional data point to overall decreases in traffic fatality and severe injury in both vehicles and population rates, and a convergence across regions. One potential explanation for these parallel trends could lie in the enactment of a comprehensive traffic law reform in 2006, which may have impacted police traffic enforcement. However, as we can observe in figure 1, a decreasing tendency was evident before the enactment of this law, particularly for fatalities per 10 000 vehicles and severe injuries per 1000 vehicles. There are other elements that, as noted in the literature, should be considered since traffic laws changes might be confounded with other significant factors or not strong enough to influence variation in these rates. For instance, several studies have indicated that increasing oil prices,32 ,33 alcohol prices,34 ,35 infrastructure development6 ,36 or unemployment37 ,38 can also help produce a downwards tendency. Furthermore, a more direct pathway might also be found in changes in police enforcement practices, particularly in the Chilean context where the centralised nature of this state ensures that these practices are uniformly applied across regions.
Data on (i) traffic fatalities and (ii) severe injuries were obtained from Chile's National Commission of Road Safety of the Ministry of Transport. This organisation has compiled, from police reports, an extensive dataset of traffic fatalities and injuries in Chile from 1972 to the present. Complete data for all 13 regions are available from 2000 to 2012. This provides us with a balanced sample with N=169 region-year observations. Fatalities are classified according to status (driver, passenger, motorcyclist, bicyclist and pedestrian), age, sex and time and day of accident. Following the Chilean Penal Code, police officers classify injuries as severe (ie, injury which prevents the person from performing acts, which constitute such person's usual and customary daily activities, at least 30 days or more immediately following the occurrence of the injury), less severe (injury which prevents the individual from returning to previous activities up until 30 days immediately following the occurrence of the injury) and minor (injury which does not require immediate medical attention). Data on vehicle fleet and population were both obtained from the National Institute of Statistics. Traffic fatalities and serious injuries per 10 000 vehicles and per 100 000 population were the outcome variables assessed in this study, for a total of four dependent variables. By analysing these variables separately, we may be able to identify the specific impact of each policy in similar, but not identical, outcomes. Last, by introducing the two denominators (vehicle fleet and population), we facilitate international comparisons of the results as well as test robustness across models.
‘Traffic law reform’ was coded as a yearly dummy variable, ‘0’ being the 2000–2005 period when the law had not been approved, and ‘1’ being the 2006–2012 period when the reform was in force. ‘Police per 100 000 population’ is the national number of police officers, using data from United Nations Office on Drugs and Crime. ‘Police traffic enforcement’ is measured as the number of traffic tickets processed by local courts, using data provided by the Ministry of Justice, divided by the national number of police officers, using data from United Nations Office on Drugs and Crime. It must be noted that due to security reasons, the Chilean Police Force (Carabineros de Chile) only provides information regarding the number of police officers at the national level. Thus, the operationalisation of ‘Police traffic enforcement’ is number of tickets per region and year divided by the national number of police officers per year. While the literature suggests number of police officers9 ,39 or number of traffic tickets40 as separate proxies of police traffic enforcement, we actually propose a combination of these two factors to capture more accurately this road safety measure. First, since police officers have duties other than patrolling traffic, we recognise that an increase in the number of police officers cannot be assumed to automatically lead to a commensurate increase in traffic ticketing. Second, because significant changes in the total number of police officers can actually trigger any absolute increase in the number of traffic tickets, we acknowledge that an increase of traffic ticketing per se may be cofounded with the number of police officers, and thus cannot be assumed to have a clear-cut effect on traffic fatalities or severe injuries. In order to assess the effect of the road safety institutional mechanism for the Chilean case, an interaction variable of ‘Traffic law reform’ and ‘Police traffic enforcement’ was introduced; this variable is labelled ‘Road safety institutional mechanism’. This effect captures how traffic fatality and morbidity rates may be sensitive to police traffic enforcement, a practice which in turn is sensitive to law changes, given the centralised character of Chile's political regime.
In consideration of previous literature, several control variables were introduced: (1) ‘Oil price average’ measures the average price of four types of combustible, 93, 95 and 97 octanes and diesel, using data from the National Commission of Energy. (2) ‘Unemployment’ is measured using data from the National Institute of Statistics. (3) ‘Percentage of young male population’ measures the male population between 15 and 24 years old as a proportion of the total population, using data from the National Institute of Statistics. (4) ‘Alcohol consumption’ measures national alcohol consumption per year per population, using data from the Drugs and Alcohol Prevention and Rehabilitation National Service. (5) ‘Road infrastructure’ measures the total amount of both public and private infrastructure spending allocated to the design, construction and maintenance of public roads, using data from the Ministry of Public Works.
Table 2 provides a summary of the dependent, independent and control variables used in the models.
Using OLS and robust random effect (RRE) models, values of the four dependent variables were examined from 2000—when there was enough information to have a balanced sample—to 2012. These models were chosen in order to test the robustness of each independent variable of interest. Following Bell and Jones41 and Shor et al,36 we suggest that by carrying out RRE models, rather than fixed effects models or OLS models, more information could be accounted for and thus a more precise testing could ultimately be provided. In this way, we can capture information regarding regions even though they are time-invariant variables. Our choice of RRE models has the advantage of allowing us to capture the effects of a theoretical important time invariant variable ‘number of police officers per 100 000’ population, which we cannot assess using a fixed effect model. Last, it is important to notice that since simple random effects models can be sensitive to heteroskedasticity, the robust option, which is facilitated by the software Stata, was introduced to control for this type of anomaly. We also use piecewise RRE in order to test more accurately the effect of police traffic enforcement after the enactment of the law in two different periods, 2006–2012 and 2007–2012. The rationale for distinguishing these two periods is based on the implementation process of this law. While the law was approved in December 2005, it is important to note that in December 2006, various decrees regulating implementation specifics of the norms approved by this law reform (ie, technical characteristics of helmets of cyclists, child restraints or hands free mobile phone devices) were enacted and thus the law could have affected police traffic enforcement in either 2006 or 2007. In order to decrease the type I error associated with the variables introduced in each model, we set two significance levels, p<0.05 and p<0.01.
Table 3 shows the results of four statistical OLS models and four RRE models, estimating the effectiveness of the current overall ratings in predicting traffic fatality rates during 2000–2012. With the exception of ‘Traffic law reform’, which is a dichotomous variable, we log-transform our dependent, independent and control variables to account for their skewed distribution. We observe five important results across models: (i) neither traffic law nor police traffic enforcement appears to be independently associated with traffic fatality rates or severe injury variations in the Chilean case. (ii) The results for the interaction effect between ‘Traffic law reform’ and ‘Police traffic enforcement’ suggest that it captures a significant decreasing effect on both traffic fatality and severe injury rates for 2000–2012. The size of the impact of this institutional mechanism is 7.1% (for OLS and RRE models) when analysing fatalities per vehicle fleet, and 8.3% (for OLS and RRE models) when analysing fatalities per population. This institutional mechanism is also associated with a decrease of 7.2% in severe injuries per vehicle (for OLS and RRE models), and 7.5% in severe injuries per population (for OLS and RRE models). (iii) We observe that ‘Police per 100 000 population’ is negatively and significantly associated with fatalities per vehicle in both OLS and RRE models, and severe injuries per population in the OLS model. (iv) With regard to ‘Percentage of young male population’, and in accordance with the literature, a significant effect is captured by the two RRE models that predict severe injuries. That is, the smaller this population is the more likely a reduction of severe injuries can be observed. (v) ‘Alcohol consumption’ is only significant when predicting severe injuries per population in the RRE model, suggesting that a decline in this variable reduces the overall severity of injuries.
The ‘Road safety institutional mechanism’ is statistically significant at p<0.01 in five models. The three models in which the effect of this variable is significant at p<0.05 are: the OLS fatality per vehicle model, the RRE severe injury per vehicle model and the RRE of severe injuries per population. Theoretically, we are inclined to choose all the RRE models over OLS because residuals are not independently and identically distributed. Second, we also prefer the RRE model over OLS because the variable region can be absorbed by this model specification. In this case, for instance, weather variation, which is associated with traffic events, may also affect police enforcement, and therefore indirectly influence traffic events. Since we do not have access to weather information, RRE is the preferable option. But in any case, in the eight models, the introduction of the ‘Road safety institutional mechanism’ is consistent in both significance and size when predicting traffic fatalities and severe injuries. Therefore, we can suggest that the introduction of a traffic law reform in Chile had a positive impact on police traffic enforcement since a steadier increase is registered after this legislative change.
Table 4 shows the results of eight piecewise RRE models for the periods 2006–2012 and 2007–2012 (OLS results are not displayed but are extremely similar to the ones presented here. These results are available on request). This analytical strategy allows us to account more accurately how the road safety institutional mechanism operated. Here we observe three important results: (i) Police per 100 000 population does not seem to be independently associated with traffic fatality rate nor severe injury variations in these two periods. (ii) Regardless of which denominator is chosen to assess severe injury rates variation, police traffic enforcement is associated with significant reductions of 34% for the 2006–2012 period, and 37% for the 2007–2012 period. This also suggests an effective intensification of this measure after the year 2007. (iii) In terms of traffic fatalities, we also notice a decline of these rates, in both periods for both denominators. However, the range is much larger because during the 2006–2012 period, fatalities per vehicle are associated with a reduction of 42% (p<0.01), whereas fatalities per population decrease by 21%, but this result is not statistically significant (p>0.05). With regard to the 2007–2012 period, we observe that the reduction has increased to 79% for fatalities per vehicles and 59% for fatalities per population, both significant at p<0.01, reinforcing the suggestion that police traffic enforcement was significantly intensified after the enactment of the law and that this had the intended effect.
The current study provides evidence of how, in a centralised country, a combination of traffic law reform with police enforcement can steadily decrease both traffic fatalities and severe injuries. The statistical analysis suggests the constitution of an effective institutional mechanism, police enforcement intensification, following closely after the enactment of Chile's traffic law reform. However, three limitations of the current study need to be mentioned in order to temper the analysis. First, Chile unfortunately does not have information on exposure to road risk, measured as kilometres travelled, and therefore we acknowledge this to impose some caution on how results should be analysed. Nevertheless, by having introduced two different rates (per vehicle fleet and population), we are confident that the results are likely to be robust. Second, we lacked information to analyse variation in both type of traffic infractions and vehicle safety, and thus some elements of the downward trend observed from 2000 onwards and for both postenactment law periods were unfortunately not fully captured. However, our RRE models indirectly accounted for this lack of information. Third, as has been pointed out in the literature,41–46 the classification of death and injuries made by police officers, relative to those made by health professionals, can be biased and thus we understand how these particular results may not be as precise as one would like them to be. Particularly, as it has been pointed out by WHO,42 a more precise analysis of road traffic fatalities should consider a 30-day definition, since research has shown that a considerable number of people who die as a result of a crash succumb to their injuries within 30 days of sustaining them. In this regard, it would be important to carry out further analysis using health service datasets in order to test our fatality and severe injury models.
There are four elements that need to be considered further in order to understand how this case unfolded: first, the traffic law reform included a whole range of norms regulating the behaviours of drivers, passengers, cyclists and pedestrians. As a result, the National Police was able to expand its repertoire of traffic control strategies since a whole new array of road risky behaviours could be addressed. Second, we observed that, as some literature had suggested,11 ,12 traffic law reforms can be sterile if these instruments are not supported by other actions. In this particular case, we actually notice that the traffic law reform did not have a significant independent impact on traffic fatalities or severe injuries for the 2000–2012 period. Nevertheless, since Chile is a centralised country, we did observe that the police force, sensitive to this law change, increased its controls, and thus a reduction of both traffic fatality and a severe injury rates was observed after the enactment of the reform. In other words, while we registered the presence of police traffic enforcement during the period before the enactment of the law in 2006,—but it seems to have been insufficient in explaining the downward tendency observed for these years—it is only after the implementation of the legislation that police enforcement seemed effective. At any rate, this institutional mechanism among parliament, central government and police institution is ultimately fundamental to reduce traffic fatality and morbidity rates, since in centralised countries police enforcement practices are sensitive to macro political outcomes. The institutional mechanism sheds light on why a decreased tendency in traffic fatalities and injuries within and among regions was observed. Third, our observation of an important intervening road safety institutional mechanism implies that legislative changes may not necessarily or automatically translate into immediate traffic fatalities reductions. This is particularly relevant when we observe that police traffic enforcement, when analysing traffic fatality by population rates, is only significant for one of the two analysed postenactment periods. The following hypothesis could be advanced to understand why the law did not impact police enforcement practices immediately: various decrees, which were regulating the implementation specifics of some of the new norms—such as stipulating required safety characteristics for the production of helmets for cyclists, child restraints and hands free mobile phones devices—were in fact enacted at the end of 2006. Thus, police officers could only enforce road users’ behaviours associated with the correct use of these devices once the final regulation of these elements was actually in force. Last, in order to understand why road safety legislation—which was both debated and held up more than 12 years in parliament—was ultimately approved in 2005, it is important to acknowledge the influence of the ‘World Report’ launched in 2004. The timely introduction of this report in the parliamentary debate helped both to document the need of tackling a serious global challenge, as well as to inform law makers of the existence of scientifically-proven successful road safety measures. While some of the report's recommended measures were considered (ie, mandatory use of child restraints, or banning cell phone use while driving) and others left behind (ie, mandatory use of helmets for cyclists on all roads, not only on urban roads; reducing the accepted blood alcohol level, not only increasing penalties for driving under the influence of alcohol), nevertheless, it was useful in increasing the sense of urgency in the need to target the road safety challenge that Chile had left unattended. This element highlights how a particular type of global forces can ultimately have a positive impact on local practices. To conclude, the current study contributes to understanding road safety processes in a region to which more research effort should be directed, given that its traffic fatality and injury rates are among the highest in the world.
What is already known on this subject
While a vast literature suggests that road traffic legislation reforms per se may help to prevent the occurrence of traffic crashes and their consequence, others have argued that this type of measure is at best short-lived.
Others studies have argued that road safety legislation reforms can only be effective in reducing crashes and traffic fatalities when they are followed up with particular types of institutional mechanisms such as police enforcement or the creation of road safety agencies.
What this study adds
By using Chilean data, we evaluate the robustness of the literature of road safety institutional mechanisms, by particularly focusing on police enforcement, and expand our road safety knowledge of the South American region.
Empirical estimates from instrumental variables in ordinary least square and robust random effects models suggest that the enactment of the traffic law reform in interaction with police control is significantly associated with a decrease of 8% in traffic fatalities and 7% in severe injuries.
Piecewise regressions of robust random effects models after the enactment of the traffic law reform suggest that police traffic enforcement reduced severe injuries by 35% and 37% for the 2006–2012 and 2007–2012 periods, respectively, and traffic fatalities by 59% for the 2007–2012 period.
We contribute to our understanding of how, in the realm of road safety policies, global factors, such as the World report on road traffic injury prevention, promoted by WHO and the World Bank, positively impact national territories.
We would like to thank two anonymous IP reviewers and Felipe Salazar Tobar for their valuable insights and suggestions. Their contribution at different stages of this research was fundamental for improving this work. We would also like to acknowledge the National Commission of Traffic Safety of the Government of Chile (CONASET) for providing traffic fatality and severe injury data to carry out the analysis.
Contributors Each author made substantial contributions to all of the following: (1) the conception and design of the study and interpretation of data, (2) drafting the article and revising it critically for important intellectual content and (3) final approval of the version to be submitted. Finally, JINM collected the data and carried out the statistical analysis.
Funding This work was supported by the Social Sciences and Humanities Research Council of Canada Award Number 752-2013-1936.
Competing interests None.
Provenance and peer review Not commissioned; externally peer reviewed.
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.