AccidentAnalysis and Prevention 51 (2013) 93--97 Contents lists available at Sciverse ScienceDirect Accident Analysis and Prevention journal homepage: Comparison of crashes during public holidays and regular weekends Sabreena Anowara, Shamsunnahar Yasmina, Richard 3 Department of Civil Engineering, McGill University, Montreal, Quebec, Canada 0C4 '3 Chair in Road Safety Management, Faculty ofBusines5, Economics and Law, La Trobe University, Melbourne, Victoria 3086, Australia ARTICLE INFO ABSTRACT Article history: Received 30 June 2012 Received in revised form 11 September 2012 Accepted 31 October 2012 Keywords: Public holidays Logistic regression Speeding Drink-driving Seatbelt Enforcement Publicity campaigns Traffic collisions and fatalities during the holiday festive periods are apparently on the rise in Alberta, Canada, despite the enhanced enforcement and publicity campaigns conducted during these periods. Using data from 2004 to 2008, this research identifies the factors that delineate between crashes that occur during public holidays and those occurring during normal weekends. We find that fatal and injury crashes are over--represented during holidays. Amongst the three risky behaviors targeted in the holiday blitzes (driver intoxication, unsafe speeding and restraint use), non--use of restraint is more prevalent whereas driver intoxication and unsafe speeding are less prevalent during holidays. The mixed results obtained suggest that it may be time to consider a more balanced approach to the enhanced enforcement and publicity campaigns. (C) 2012 Elsevier Ltd. All rights reserved. 1. Introduction 1.1. Background Motor vehicle collisions are a major concern in many devel- oping and developed countries. For instance, recent Canadian data showed that a total of 2767 fatalities and 194,177 injuries occurred on the roads as a result of motor vehicle collisions in 2007 (Transport Canada, 2007). In the Canadian Province ofAlberta alone, nearly 400 people are killed and more than 27,000 people are injured in over 112,000 motor vehicle collisions each year (Alberta Transportation, 2006). The annual social cost ofmotor vehicle col- lisions to Albertans is estimated at $4.68 billion or 2.4% ofAlberta's gross domestic product. Therefore, much more work needs to be done to make our roads safer for all users at all times. With regard to time, traffic collisions and the ensuing fatal- ities during the statutory holiday festive periods are apparently on the rise in both developing and developed countries (Anowar et al., 2009, 2012). For example, a total of 6937 collisions occurred in 1999 during the holidays and long weekends which killed 39 people in Alberta but the total number of crashes escalated to 11,337 in 2008, with 43 people killed (Alberta Transportation, 1999, 2008). Although collisions during statutory holidays represent Corresponding author. Tel.: +61 3 94791267; fax: +61 3 9479 3283. E--mail addresses: sabreena.anowar@mail.mcgill.ca (S. Anowar), (S. Yasmin), r.tay@latrobe.edu.au, rtay@ucalgary.ca (R. Tay). 0001-4575/$ -- see front matter (C) 2012 Elsevier Ltd. All rights reserved. only a small percentage (less than 10%) of the total motor vehi- cle collisions occurring in Alberta, the number of fatal collisions occurring during statutory holidays is found to be higher than those during non--holidays. Overall, the average number of fatal collisions for these holidays (1.11 per day) is approximately 18% higher than the non-holiday rate (0.94 per day). The average num- ber of people killed per day on Albertan roadways during these holidays is also higher than the rest of the year (Anowar et al., 2012). Consequently, there are more aggressive police enforcement activities and publicity campaigns targeted at drinl<-driving, speed- ing and other risky driving behaviors during these festive holidays in Alberta and worldwide (Alberta Transportation, 2006; Transport Canada, 2001: Pilkington, 2000: Watson et al., 2002; Alsop and Langley, 2000). Moreover, traffic fatalities and enforcement activ- ities during these long weekends often attract disproportionately more media and public attention. A sample of news headlines in Alberta shows that this issue is a concern for rural and urban com- munities, large municipalities and small towns, and printed and electronic media: "Christmas Eve crash near Mundare kills three, orphans baby" (Edmonton journal, 27/12/2010). "Long weekend means police patrol roads" (Channel 880 News, 21/4/2011). "Royal Canadian Mounted Police (RCMP) will hunt Speeders on Easter long weekend" (Calgary Herald, 21/4/2011). "The Labour Day long weekend proved once again to be deadly on Alberta's roadways" (Crowsnest Pass Promoter, 4/9/2008). 94 S, Anowar et al. and Prevention 51 (2013) 93-97 Holidays are meant to be times of enjoyment and festivity. Unfortunately, these times also have the image as a time for par- tying, drunkenness, speeding and other reckless driving behaviors. Holidays are also associated with a large increase in recreational private travel resulting in longer trip distances, and more travel in rural and unfamiliar environment. Supposedly, owing to these fac- tors, in many countries ofthe world, holiday periods are commonly viewed as times of heightened danger on the roads resulting in fatal and injurious traffic collisions. Hence, additional resources are frequently employed during public holidays to boost enforcement and publicity campaigns. However, these factors are also over- represented during regular weekends and relatively little research has been done on identifying the road safety issues related to specif- ically public holidays. 1.2. Objectives and scope ofstudy In this paper, a logistic regression model will be estimated to identify the factors contributing to crashes during public holidays and long weekends. In particular, we aim to determine whether crashes during public holidays are more severe than any regular weekends and whether the factors contributing to crashes during public holidays are different from those contributing to weekend crashes. More importantly, our results will also provide valuable insight on whether the increased enforcement activities and pub- licity campaigns during the holidays are used efficiently to address the correct road safety problems. 1.3. Literature review Road crashes during the major holiday periods attract intense media interest. Nonetheless, research studies focusing on analysing the contributory factors of the road crashes are relatively few, and mostly examine specific holidays, crash types or behaviors. For example, the Australian Transport Safety Bureau (ATSB, 2003, 2006) conducted two studies focusing on holiday accidents. The goal of both studies was to examine the characteristics of fatal crashes occurring during the national holiday periods. The annual trends in road fatality numbers for two of the major statutory holiday periods, Christmas and Easter, were examined and compared with the remainder of the year. Interestingly, both studies found that the observed differences of fatality rates between holiday and non- holiday periods were generally small in size and not statistically significant. A similar research initiative was undertaken by the American state of Missouri to identify the magnitude, severity and char- acteristics of holiday traffic crashes 2003). The study analyzed crashes occurring during the following statutory holi- days: Memorial Day, Fourth ofjuly, Labor Day, Thanksgiving Day, Christmas and New Year Day. However, no comparison was made between holiday and non-holiday crashes or between holiday and regular weekend crashes. Bloch et al. (2004) used crash data of 14 major holidays and spe- cial occasions in California to compare the rise in alcohol related fatal and injury crashes during holidays with that of the non- holiday periods. They employed the Poisson regression modeling technique (log-linear and logistic), controlling for the seasonal dif- ferences in terms of days of week and months of the year. The results of the study suggested that drinking and driving was more of a concern during the winter holiday seasons than the summer ones. Farmer and Williams (2005) used data for the years 1986-2002 to determine which days of the year tend to experience a relatively higher number ofdeaths. They observed that six ofthe ten days with the greatest number ofdeaths occurred near these major American holidays: Independence Day, Christmas, New Year, and Labor Day. The authors attributed such high numbers of crash deaths to the probable combination ofincreased recreational travel, alcohol con- sumption, and excessive speeding during holidays. Amongst other possible reasons for the increased fatalities during holidays sug- gested were: travel on rural unfamiliar roads, driver distractions and fatigue, which all resulted in the increased likelihood ofdrivers committing errors. In another study, Alsop and Langley (2000) specifically focused on the Christmas road tolls. They used the negative binomial and binomial regression techniques to examine the temporal trends in the number of fatalities during the Christmas holiday festivities in New Zealand. Their results indicated that the road toll neither decreased nor improved significantly over the years. The authors argued that the lack of statistically significant increase in Christ- mas fatalities could be viewed as a positive outcome, given the large increases in population and number ofcars driven. Presumably, the average individual risk might have reduced over time. On the other hand, a lack ofstatistically significant decrease in Christmas fatali- ties could not be viewed as a positive outcome, given the increased emphasis placed on this period by traffic safety agencies. Besides statutory holidays, the effect of weekdays and weel -- ends were also explored in several studies since traffic patterns during weekdays and weekends were quite different and crashes during weekends tended to be more severe (Yau, 2004; Gray et al., 2008; Barua and Tay, 2010; Quddus et al., 2010; Christoforou et al., 2010; Rifaat et al., 2011). According to these authors, much of the traffic during weekends consisted of discretionary travel, involved more drivers who had been drinking, speeding and driving while fatigued. However, very little research was found that examined the relative crash risks between holidays and weekends or the differ- ences in the factors contributing to crashes during these two types of non--work days. 2. Methodology 2.1. Logistic regression model Recall that the aim of the research is to determine the factors that are different between crashes that occur during statutory holi- days (including long weekends) and those crashes occurring during normal weekends. Since the dependent variable is discrete and dichotomous in nature, the binary logistic regression is an appropri- ate technique to identify the different factors contributing to these two types ofcrashes. In this study. the binary response is defined as: 1, ifcrash occured during statutory holidays ym 0, ifcrash occured during regular weekends (1) Let. and denote the probability of crash occurring during statutory holiday periods and regular weekends, respectively. McFadden (1981) shows that under the standard logistic distribution, the closed form solution ofthe prob- abilities will be: eXl3(l3iXin) (2) where is a vector of measurable characteristics that determine outcome i; is a vector of estimable parameters. The best estimate of)? could be obtained by maximizing the log likelihood function: Pn(i) fl - Zwin (1 -- Pm): - (3) i=1 In this study, Stata version 11 is used for model development and estimation. S. Anowar et al. //lccidentAnalysis and Prevention 5] (2013) 93-97 95 Note that there are two common binary or dichotomous models: the binary logistic model used in this study and the binary probit model which assumes that the error terms are normally distributed. Many studies have found that the results obtained from both these models are verysimilar (Maddala, 1988; l(ennedy, 2001; Greene, 2003). The binary logistic model is chosen in this study for con- venience. lt is also more commonly used than the probit model (Kennedy, 2001 1 Moreover, some researchers have chosen to use random effects or the random coefficient logit model or mixed logit model instead of the fixed effects model (Milton et al., 2008; Anastasopoulos and Mannering, 2009: Kim etal., 2010). Random effects model are often used when the data contain repeated measures and/or to account for driver heterogeneity. These issues, however, are not a concern in this study because neither panel data nor repeated measures are used and the unit of analysis is a crash event. Moreover, pre- liminary analyses using random coefficient models found that the estimates of the variances ofthe random coefficients were statisti- cally insignificant. 2.2. Data The data used in this study is obtained from Alberta Transporta- tion and Infrastructure. it should be noted that in Alberta, traffic crash data is compiled by the Office ofTraffic Safety, Alberta Trans- portation from police reports collected and maintained by the Royal Canadian Mounted Police in the rural areas and by local municipal police forces in larger cities of the province. In Alberta, any crash resulting in injury or property damage costing more than $1000 would be required by law to be reported to the police. The crash records contain the common types of information on the collision, including the time, location and severity ofcollisions as well as data on the driver, crash type, vehicle, environment and any special road features at the crash locations. Data on crashes during the weekends and statutory holidays for the years 2004-2008 were extracted from this provincial database. For this study, the holidays considered were: New Year, Family Day long weekend, Easter long weekend, Victoria Day long week- end, Canada Day, August long weekend, Labor Day long weekend, Thanksgiving long weekend, Remembrance Day and Christmas. These ten holidays were chosen because the crashes occurring during these holidays were routinely reported and highlighted in Alberta Transportation's Annual Collision Reports. The week- end crashes comprised those crashes that occurred during regular weekends excluding statutory holidays. The final data sample consisted of 125,416 crashes for the five--year period and of these, 27.8% occurred during statutory holidays and the rest occurred during regular weekends. Based on the information available in the dataset, 15 factors were selected for analysis. These factors included crash charac- teristics, environmental conditions, operational characteristics and driver characteristics. Following some preliminary analyses, three statistically insignificant factors were excluded and 12 factors were retained in the final analysis. The descriptive statistics of the vari- ables included in the final model are reported in Table 1. Note that several factors that were widely used in the literature on crash frequency analyses were not included in this study since our focus was on delineating between crashes occurring on regu- lar weekends and public holidays. For example, although exposure would be significant in determining crash frequency, no theoretical reason existed to hypothesize that exposure should be a signifi- cant factor in our model. The effects of traffic flow on crash risks during weekends and public holidays would likely be very simi- lar. Moreover, exposure data were not available for most of the crash locations. On the other hand, although data for other vari- ables, such as weather, were available, they were not included in Table 1 Difference in crash profiles Variables Statutory holidays Weekends Crash severity Fatal crash 0.5 0.4 injury crash 17.0 16.5 PDO 82.5 83.1 Occurrence time Morning (6:00 am--12;00 pm) 20.5 19.5 Mid-day (12:00 pm--6:00 pm) 403 40.0 Evening (6:00 pm--12:00 am) 28.8 28.0 Night (12:00 am--6:00 am) 10.3 12.5 Municipality Urban 67.4 69.6 Rural 32.6 30.4 Location Intersection 56.0 56.8 Non-intersection 44.0 43.2 Road class Highway 28.5 26.2 Non--highway 71.5 73.8 Number of vehicles Single--vehicle 38.4 39.1 Two-vehicle 57.6 57.0 More than two vehicles 4.0 3.9 Crash type Struck--object 31.2 31.3 Off-road 11.0 12.0 Angular 18.9 18.2 Sideswipe 8.2 8.4 Rear--end 22.6 22.4 Head--on 0.9 0.9 Other collisions 7.3 6.9 Driver familiarity Albertan 92.1 92.6 Non-Albertan 7.9 7.4 Light condition Daylight 60.0 57.8 Dark without artificial light 54.4 53.0 Dark with artificial light 14.0 15.7 Unknown light condition 5.3 5.8 Driver condition Normal 93.5 92.5 Drunk 4.8 5.5 Fatigued 0.7 0.7 Other driver condition 1.2 1.3 Speed of vehicle Safe 92.4 91.9 Unsafe 7.6 8.1 Seat--belt use Restrained - 90.5 90.5 Non--restrained 3.4 3.2 Seat--belt use unknown 6.2 6.3 the model because their effects on crash risks during weekends and public holidays were expected to be very similar. Preliminary anal- yses also found that the estimated coefficients were statistically insignificant. Since all the contributing factors were categorical in nature, sev- eral dummy variables were used to represent each ofthese factors. Note that one of the dummy variables had to be used as the ref- erence. The estimates obtained for the other variables were then interpreted with reference to the default or reference case. For example, for the number ofvehicles factor, the reference case used was single vehicle and the estimates for the two and more than two vehicle crashes were analyzed and interpreted relative to single vehicle crashes. 3. Results and discussion The estimation results of the binary logit model are reported in Table 2. Overall, the model fitted the data relatively well, with a very large chi-square statistic and very small p-value. Note that 96 S. Anowar et al. /Accident/inolysis and Prevenrio.-151 (2013) 93-97 Table 2 Estimation results of the binary logistic regression. Variables Coefficient Std. err. t-Stat Odds ratio Main variables Crash severity (reference: PDO) Fatal crash 0.1848 0.0928 1.99 1.2030 lnjury crash 0.0440 0.0180 2.44 1.0449 Driver condition (reference: normal) Drunk -0.0558 0.0316 -1.76 0.9457 Speed of vehicle (reference: safe) Unsafe -0.0420 0.0243 -1.73 0.9588 Seat-belt use (reference: restrained) Non-restrained 0.0909 0.0371 2.45 1.0951 Unknown 0.0519 0.0268 1.93 1.0532 Control variables Occurrence time (reference: morning (6.00 am-1 2.00 pm)) Night(12:0O am--6:00 am) --0.1744 0.0224 -7.80 0.8399 Mid-day (12:00 pm-6:00 pm) -0.0328 0.0144 -2.28 0.9677 Municipality (reference: urban) Rural 0.1103 0.0240 4.60 1.1157 Location (reference: intersection) Non-intersection 0.0346 0.0181 1.91 1.0352 Road class (reference: non-highway) Highway 0.0847 0.0212 4.01 1.0884 Number of vehicles (reference: single vehicle) Two--vehic|e 0.0659 0.0216 3.05 1.0681 More than two vehicles 0.0736 0.0381 .1.93 1.0764 Crash type (reference: struck-object) Off-road --0.l297 0.0228 -5.68 0.8783 Angular 0.0823 0.0231 3.56 1.0857 Rear-end 0.0455 0.0215 2.12 1.0465 Other collisions 0.0911 0.0271 3.36 1.0954 Driver familiarity (reference: Albertan) Non--Albertan 0.0555 0.0237 2.34 1.0570 Light condition (reference: daylight) Artificial light --0.0700 0.0200 -3.51 0.9323 Unknown --0.0860 0.0283 -3.04 0.9175 Constant --1.0498 0.0215 -48.74 -- Number of observations 125,416 Log likelihood at zero Log likelihood at convergence --73.943.01 Chi square (20) 356.50 P-value <0.001 a positive estimated coefficient will indicate that the correspond- ing variable increases the likelihood of a crash occurring during public holidays rather than regular weekends, whereas a negative estimated coefficient will indicate the reverse. 3.1. Main independent variables In our analysis. the main independent variables considered are crash severity. driver intoxication. unsafe speeding and restraint use because these are the most highlighted issues related to holiday crashes in the media and much of the enforcement activities and publicity campaigns are focused on deterring drivers from these driving infringements. It should be noted that the results obtained on the outcomes are only correlational and do not imply any causal- ity. Hence, care should be exercised in interpreting the results and their implications. It is evident from the results shown in Table 2 that both fatal and injury outcomes are more prevalent during statutory holi- days than weekends, and this finding is consistent with the general beliefthat the roadways are more dangerous during statutory holi- days (Farmer and Williams, 2005). Moreover, our results also show that non-use of restraints (seat--belts) by vehicle occupants (driver and/or passenger) is higher in crashes during holiday periods which can be partly attributed to the lower proper restraint use dur- ing leisure trips as observed by Okamura et al. (2010). On the other hand, both driving while impaired and driving at an unsafe speed are found to be less prevalent in holiday crashes, albeit, only marginally significant (90% confidence level). With respect to policy implications, our results showed that rel- ative to regular weekends, non-use of seatbelt was more prevalent whereas drink-driving and speeding were less prevalent during public holidays. Hence. policy makers' might want to consider focussing more on seatbelt use during their holiday blitzes and tar- geting drink--driving and speeding more during regular weekends. Note that our model was only able to identify the factors that were more prevalent in crashes occurring during holidays than crashes occurring during regular weekends but not the effectiveness of the enforcement or publicity per se. The results, however, would enable us to identify potential target areas and set the right priorities for future enforcement and publicity campaigns. 3.2. Control variables In our study, several factors were included as control variables. We found that holiday crashes were less likely to occur during night-time or mid--day and also less likely to occur under artifi- cial lighting conditions. These results might indicate a difference in consumer travel patterns and high risk times during holidays as compared to normal weekends. Rural areas were over--represented in crashes that occurred during public holidays. Moreover. crashes during long weekends were more likely to involve out-of-province drivers. Long--distance S. Anowar er al. /Accident Analysis and Prevention 51 (2013) 93-97 97 social and recreational travels might occur during national holiday periods and most of these trips would be more likely to take place on high speed and unfamiliar rural roads. Consistent with these findings, we also found that holiday crashes were more likely to occur on highways and at non--intersection locations. Multiple-vehicle (both two and more than two vehicles) crashes were also found to be more prevalent during the holidays. Interestingly, both angular and rear-end crashes were over- represented during the holidays whereas off--road crashes were under-represented. Holiday travelers might not be maintaining enough distance between vehicles and as a result, getting involved in higher number of rear end collisions. Driver distractions by pas- sengers chit-chatting, tending to children etc.) were more likely to be associated with rear-end or an angular crash than single--vehicle crash (Ghazizadeh and Boyle, 2009) and this kind of distraction might happen more frequently during holiday trips than regular weekend trips. 4. Conclusion Holidays are often viewed as times of increased risky driving behaviors on the roads and manyjurisdictions around the world, including Alberta, have invested additional resources to enhance their enforcement and publicity campigns during these periods. However, most of the factors contributing to the alleged increase in crash risks are also present during regular weekends and little research has been conducted to examine the differences between collisions occuring between holidays and regular weekends. This study examined the factors associated with the statu- tory holiday crashes that significantly differed from the factors associated with weekend crashes. A binary logit model was applied to a sample of collision data from Alberta from 2004 to 2008. We found mixed but interesting results from our analy- sis. First, our model showed that both fatal and injury crashes were over--represented during holidays which was consistent with the perception that the roadways were more hazardous during the national holiday periods. Second, amongst the three behav- ior and policy variables (driver intoxication, unsafe speeding and restraint use), non-use of restraint was found to be more preva- lent whereas driver intoxication and unsafe speeding were less prevalent during holidays. These mixed results obtained would suggest that we might need to reconsider how the enhanced enforcement and publicity campaigns should be conducted and to adopt a more balanced approach between holidays and reg- ular weekends as well as among the different risky behaviors targeted. In addition to the main influences on holiday crashes dis- cussed, other factors identified included rural locations, highways, and non-intersection locations, as well as multi--vehicle, angular, rear-end collisions. Moreover, we found that relative to regular weekend crashes, holiday crashes were more likely to involve a driver in an unfamiliar environment. On the other--hand, holiday crashes were less likely to occur during the night and the afternoon, under artificial light conditions, or involved running off the road incidences. References Alberta Transportation, 1999. Alberta Traffic Collision Statistics. Alberta Transporta- tion. Office ofTraffic Safety, Alberta Transportation, Edmonton, Canada. Alberta Transportation, 2006. Alberta Traffic Safety Plan. Alberta Transportation, Edmonton. Alberta Transportation, 2008. Alberta Traffic Collision Statistics. Alberta Transporta- tion. Office ofTraffic Safety, Alberta Transportation, Edmonton, Canada. Alsop, Langley, J., 2000. Dying to go on holiday. Australian and New Zealand. journal of Public Health 24(6), 607-609. Anastasopoulos, P., Mannering, F., 2009. A note on modeling vehicle accident frequencies with random-parameters count models. Accident Analysis and Pre- vention 41 (1 153-159. Anowar, S., Yasmin, S., Tay, R., 2009. Traffic accidents and injuries: turning the fes- tivities of Eids into mourning. in: international Co--operation on Theories and Concepts in Traffic Safety, Leeds, UK. Anowar, S., Yasmin, S., Tay, R., 2012. Severity of single vehicle crashes during holi- days. In: Transportation Research Board Annual Meeting, Washington, DC. Australian Transport Safety Bureau, 2003. The Characteristics ofFatai Crashes during the Christmas/New Year Period. Canberra. Australian Transport Safety Bureau, 2006. Characteristics of Fatal Crashes during National Holiday Periods. Canberra. Barua, U., Tay, R., 2010. Severity of urban transit bus crashes in Bangladeshjournal ofAdvanced Transportation 44 (1), 34-41. Bloch, 5., Shin, H., Labin, S., 2004. Time to party: a comparative analysis of holiday drinking and driving. in: 17th international Conference on Drugs, Alcohol and Traffic Safety, Glasgow. Christoforou, Z., Cohen, S., Karlaftis, M., 2010. Vehicle occupant injury severity on highways: an empirical investigation. Accident Analysis and Prevention 42(6), 1605-1620. Farmer, C., Williams, A., 2005. Temporal factors in motor vehicle crash deaths. injury Prevention 11 (1), 18-23. Ghazizadeh, M., Boyle, L., 2009. influence of driver distractions on the likelihood of rear-end, angular, and single-vehicle crashes in Missouri. Transportation Research Record 2138, 1-5. Gray, R., Quddus, M., Evans, A., 2008. injury severity analysis of accidents involving young male drivers in Great Britain.journai ofSafety Research 39 (5), 483-495. Greene, W., 2003. Econometric Analysis. Prentice Hall, New York. Kennedy, P., 2001. A Guide to Econometrics. MIT Press, Cambridge. Ulfarsson, G., Shankar, V., Mannering, F., 2010. A note on modelling pedestrian injury severity in motor vehicle crashes with the mixed logit model. Accident Analysis and Prevention 42 (6), 1751-1758. Maddala, G., 1988. Limited Dependent and Qualitative Variables in Econometrics. Cambridge University Press, Cambridge. l\/lcFadden, D., 1981. Econometric models of probabilistic choice. in: Manski, C., McFadden, D. A Structural Analysis of Discrete Data with Econometric Applications. The MIT Press, Cambridge, MA. Milton, C., Shankar, V., Mannering, F., 2008. Highway accident severities and the mixed logit model: an exploratory empirical analysis. Accident Analysis and Prevention 40 (1), 260-256. 2003. Missouri Holiday Crashes Report. Missouri State Highway Patrol Statistical Analysis Centenjefferson City. Okamura, K., Mori, K., Mitsui, T., 2010. Factors influencing premature graduation from the use of child restraints in Japan. Accident Analysis and Prevention 42 (2), 403-411. Pilkington, P., 2000. Reducing the speed limit to 20 in urban areas. British Medicaljournal 320, 1150. Quddus, M., Wang, C., lson, S., 2010. Road traffic congestion and crash severity: econometric analysis using ordered response models.}ournal ofTransportation Engineering 136 (5), 424-435. Rifaat, S., Tay, R., de Barros, A., 2011. Effect of street pattern on the severity of crashes involving vulnerable road users. Accident Analysis and Prevention 43 (1). 276-283. Transport Canada, 2001. Vision 2010. Transport Canada, Ottawa. Transport Canada, 2007. Canadian Motor Vehicle Traffic Collision Statistics: 2007. Transport Canada, Ottawa. Watson, B., Siskind, V., Edmonston. C., Tay, R., Sheehan, M., 2002. Evaluation ofthe queensland holidays period road safety trial 2001/2. in: Report to the Evaluation Steering Committee, Queensland Government. Yau, K., 2004. Risk factors affecting the severity ofsingle vehicle traffic accidents in Hong Kong. Accident Analysis and Prevention 36 (3), 333-340.