MONASH UNIVERSITY AUSTRALIA FURTHER MODELLING OF SOME MAJOR FACTORS INFLUENCING ROAD TRAUMA TRENDS IN VICTORIA: 1990-96 FURTHER MODElliNG OF SOME MAJOR FACTORS INFLUENCING ROAD TRAUMA TRENDS IN VICTORIA: 1990-96 by Stuart Newstead Max Cameron Sanjeev Narayan Report No. 129 April 1998 Il MONASH UNIVERSITY ACCIDENT RESEARCH CENTRE MONASH UNIVERSITY RESEARCH CENTRE REPORT DOCUMENTATION PAGE Report No. 129 Report Date Apri11998 ISBN 0732607094 Pages 23 + Appendices Title and sub-title: Further modelling of some major factors influencing road trauma trends in Victoria: Author(s) Newstead, S.V.,Cameron, M.H. & Narayan, S. 1990-96 Type of Report & Period Covered GE~RJ\L, 1990-1996 Sponsoring Organisation - This project was funded through the Centre's baseline research program for which grants have been received from: Department of Justice Royal Automobile Club of Victoria Ltd. VicRoads Transport Accident Commission Abstract: Based on previous work that has estimated the contribution of some major factors in reducing road trauma in Victoria over the period 1990-1993, this project has made use of the statistical analysis methods developed to extend these estimates to 1996. The major factors considered in the study have stemmed from the results of a number of studies in Victoria which have evaluated the effects of countermeasures and other factors which appear to be responsible for the substantial reduction in road trauma since 1989. The factors for which contributions have been estimated were: • • • • Increased random breath testing, supported by mass media publicity Speed cameras, supported by mass media publicity Reduced economic activity Reduced alcohol sales • Improvements to the road system through treatment of accident black spots The percentage change in road trauma levels, as measured by serious casualty crash numbers, due to each factor has been estimated for each year over the period 1990-1996. Models linking variations in serious casualty crashes to various factors were computed using monthly crash data from the years 1983 to 1996. Subsequently, the contributions of random breath testing, speed camera tickets issued, levels of road safety television publicity, unemployment rates and alcohol sales to the reduction in the number of serious casualty crashes were estimated for the period 1990-96. A method of separately estimating the effect of accident blackspot treatments and desegregating this from the trend was described and applied. Key Words: (IRRD except where marked*) statistical analysis, accident frequency, breath test, speed camera, advertising, economics, alcohol usage, accident black spot Reproduction of this page is authorised. FURTHER MODELLING OF SOME FACTORS INFLUENCING ROAD TRAUMA TRENDS IN VICTORIA III IV MONASH UNIVERSITY ACCIDENT RESEARCH CENTRE Contents 1. INTRODUCTION 1.1 2. 1 REVIEW OF WORK COMPLETED TO DATE 1 INCLUSION OF CRASH DATA TO THE END OF 1996 IN THE MODELS ..•................................3 2.1 FACTORS INCLUDED IN THE MODELS 2.1.1 Economic Measure 3 3 2.1.2 Factors Relevant to High Alcohol Hour Crashes 2.1.3 Factors Relevant to Low Alcohol Hour Crashes 2.2 DISAGREGGATION AND COMBINATION OF THE RESULTS 2.3 EFFECT OF ACCIDENT BLACK SPOT TREATMENTS AND OTHER FACTORS 2.4 SUMMARY OF ESTIMATED CONTRlBUTIONS 4 12 14 14 16 3. DISCUSSION .....................................................•................•......•..•...........•............•.........•..........•.....•..... 18 4. CONCLUSION .........................................................................................................................•...•.........19 5. FURTHER WORK RECOMMENDED ..................................................•.........•.......•................•.....•... 20 6. 21 ACKN 0WLEDG MENTS ..........................•.........•...•...............••............................................................ 7. REFERENCES .........••.........•.....••.......................................••.........................................•......................... 21 FURTHER MODELLING OF SOME FACTORS INFLUENCING ROAD TRAUMA TRENDS IN VICTORIA V List Of Tables TABLE 1 : EFFECTS OF UNEMPLOYMENT RATE, NUMBER OF RANDOM BREATH TESTS, ALCOHOL SALES (ALL VICTORIA), AND ADSTOCK OF TAC DRINK-DRIVING PUBLICITY ON SERIOUS CASUALTY CRASHES DURING HIGH ALCOHOL HOURS OF THE WEEK. MELBOURNE AND COUNTRY VICTORIA 1983 - 1992 12 TABLE 2 : EFFECTS OF UNEMPLOYMENT RATE, NUMBER OF SPEED CAMERA TINS ISSUED, AND ADSTOCK OF TAC "SPEEDING" AND "CONCENTRATION" PUBLICITY ON SERIOUS CASUALTY CRASHES DURING LOW ALCOHOL HOURS OF THE WEEK. MELBOURNE AND COUNTRY VICTORIA 1983 - 1996 TABLE 3: ESTIMATED REDUCTIONS IN SERIOUS CASUALTY CRASHES ATTRIBUTABLE TO VARIOUS FACTORS - VICTORIA, ALL HOURS, 1990-96 TABLE 4 : ESTIMATED REDUCTIONS MAJOR FACTORS: VICTORIA 1990-96 VI IN SERIOUS CASUALTY MONASH UNIVERSITY ACCIDENT RESEARCH CENTRE 13 14 CRASHES ATTRIBUTABLE TO 17 List Of Figures FIGURE 1: PERCENTAGE CHANGE FROM 1987, VICTORIA. FATALITIES AND SERIOUS INJURIES 1 FIGURE 2: UNEMPLOYMENT 3 RATE IN MELBOURNE AND THE REST OF VICTORIA 1983-96 FIGURE 3A : NUMBER OF BUS-BASED RANDOM BREATH TESTS PER MONTH. MELBOURNE (MSD) 1989-1996 4 FIGURE 3B : NUMBER OF BUS-BASED RANDOM BREATH TESTS PER MONTH. REST OF VICTORIA (ROV) 1986-1996 5 FIGURE 4A(I) : TAC ROAD SAFETY TELEVISION ADVERTISING - MONTHLY ADSTOCK BY THEME: MELBOURNE REGION 1989 - 1996 6 FIGURE 4A(II) : TAC ROAD SAFETY TELEVISION ADVERTISING - MONTHLY ADSTOCK BY THEME: COUNTRY VICTORIA 1989 - 1996 6 FIGURE 4B(I) : DRINK-DRIVING 7 FIGURE 4B(II) : DRINK-DRIVING ADSTOCK: MELBOURNE REGION 1989 - 1996 ADSTOCK: REST OF VICTORIA 1989 - 1996 7 FIGURE 4C(I): SPEED AND CONCENTRATION AD STOCK : MELBOURNE REGION 1989 - 1996 .... 8 FIGURE 4C(II): SPEED AND CONCENTRATION ADSTOCK: REST OF VICTORIA 1989 - 1996 FIGURE 5: INDEX OF ALCOHOL SALES IN VICTORIA, 1983 - 1996 8 9 FIGURE 6A : MONTHLY NUMBER OF TRAFFIC INFRINGEMENT NOTICES ISSUED FOR SPEEDING OFFENCES DETECTED BY SPEED CAMERAS. MELBOURNE 1989 - 1996 10 FIGURE 6B : MONTHLY NUMBER OF TRAFFIC INFRINGEMENT NOTICES ISSUED FOR SPEEDING OFFENCES DETECTED BY SPEED CAMERAS. REST OF VICTORIA 1989 - 1996 10 FIGURE 7 : ESTIMATED REDUCTIONS MAJOR FACTORS: VICTORIA 1990-96 FURTHER MODELLING IN SERIOUS CASUALTY CRASHES ATTRIBUTABLE OF SOME FACTORS INFLUENCING ROAD TRAUMA TRENDS IN VICTORIA TO 17 VII List of Appendices APPENDIX A: PERCENTAGE REDUCTIONS IN SERIOUS CASUALTY CRASHES ATTRIBUTABLE TO VARIOUS SOURCES. VICTORIA 1990-1993 RELATIVE TO 1988 APPENDIX B : ESTIMATION BLACKSPOT TREATMENTS OF SERIOUS CASUALTY CRASH REDUCTIONS VIII MONASH UNIVERSITY ACCIDENT RESEARCH CENTRE DUE TO ACCIDENT EXECUTIVE SUMMARY Based on previous work that has estimated the contribution of some major factors in reducing road trauma in Victoria over the period 1990-1993, this project has made use of the statistical analysis methods developed to extend these estimates to 1996. The major factors considered in the study have stemmed from the results of a number of studies in Victoria which have evaluated the effects of countermeasures and other factors which appear to be responsible for the substantial reduction in road trauma since 1989. The factors for which contributions have been estimated were: • • • • Increased random breath testing, supported by mass media publicity Speed cameras, supported by mass media publicity Reduced economic activity Reduced alcohol sales • Improvements to the road system through treatment of accident black spots The percentage change in road trauma levels, as measured by serious casualty crash numbers, due to each factor has been estimated'for each year over the period 1990-1996. Models linking variations in serious casualty crashes to various factors were computed using monthly crash data from the years 1983 to 1996. Subsequently, the contributions of random breath testing (RBT), speed camera tickets issued, levels of road safety television publicity, unemployment rates and alcohol sales to the reduction in the number of serious casualty crashes were estimated for the period 1990-96. A method of separately estimating the effect of accident blackspot treatments and desegregating this from the trend was described and applied. The major contributors and the apparent percentage reduction in serious casualty crashes due to each measure/factor were estimated as: • Speed camera operations (principally speeding tickets issued): 10-11% each year • "Speeding" and "concentration" television advertising: 5-7% each year • Drink-driving program (bus-based RBT together with "drink-driving" publicity campaigns) • Reduced alcohol sales: 9-10% each year 3% in 1990 6% in 1991 7% in 1992 9% in 1993 8% in 1994 9% in 1995 10% in 1996 FURTHER MODELLING OF SOME FACTORS INFLUENCING ROAD TRAUMA TRENDS IN VICTORIA IX • Reduced economic activity (measured by unemployment rates): 2% in 12% in 15% in 16% in 14% in 10% in 10% in 1990 1991 1992 1993 1994 1995 1996 • Accident Black Spot treatments 1.6% in 2.5% in 3.2% in 5.3% in 6.2% in 6.2% in 5.6% in 1990 1991 1992 1993 1994 1995 1996 The anti-speeding and drink-driving programs together are estimated to have contributed reductions in serious casualty crashes of at least 22-25% during these seven years. Including the accident blackspot treatments, the overall contribution of road safety initiatives is estimated to have risen from 23% reduction in 1990 to nearly 30% reduction in 1993-1996. X MONASH UNIVERSITY ACCIDENT RESEARCH CENTRE 1. INTRODUCTION Since 1989 there has been a remarkable reduction in deaths and injuries on Victorian roads. Figure 1 shows a decline in fatalities in Victoria from a peak of 776 in 1989 to 417 in 1996. As well as the substantial decrease in road fatalities in Victoria since 1989, there have also been large reductions in serious injuries (Figure 1). A relative plateau in the levels of both fatalities and serious injuries has, however, been observed from around 1992 through to 1996. FIGURE 1: Percentage change from 1987. Victoria, fatalities and serious injuries 120 100 '0 ~ 'ii ••• •• lL .!! > l!••t•• C 60 80 40 20 o 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 Ve.r I_Fatal~ies SeriousinjuriesI These initial large reductions in deaths and injuries after 1988 have been attributed to the implementation of various safety programs which were introduced commencing in September 1989, and to the downturn in the economy which occurred during the period under consideration. It is important to estimate the contribution of each of these programs and if possible the mechanism by which they achieved their reductions, so that they can be fine tuned for further gains and allocation of future resources can be made on the basis of the best available information. This project aims to build on previous work which has estimated the contribution of some of major road safety programs, along with economic factors, which have lead to the dramatic reduction in road trauma in Victoria since 1989. 1.1 REVIEW OF WORK COMPLETED TO DATE A study by Newstead et al (1996) has estimated the contribution of some major factors in reducing road trauma, as measured by serious casualty crashes (SCCs), over the period 1989-94. This updated previous work covering the period 1989-93 (Newstead et al 1995) FURTHER MODELLING OF SOME FACTORS INFLUENCING ROAD TRAUMA TRENDS IN VICTORIA 1 which developed a method of combining the results of a number of studies in Victoria which have evaluated the effects of countermeasures and the factors which appear to be responsible for the substantial reduction in road trauma since 1988. The factors for which contributions were estimated were: • • • • Increased random breath testing, supported by mass media publicity New speed cameras, supported by mass media publicity Reduced economic activity (as measured by unemployment rate) Reduced alcohol sales • Improvements to the road system through treatment of accident black spots The original studies upon which the work to date has been based have modelled Victorian road trauma trends for various road user groups at various times of the week (Cameron et al 1992a, Cameron et al 1992b, Cameron et al 1993). These studies have defined four strata in terms of area and time into which crashes in Victoria can be divided for modelling road trauma trends. These four strata are; the Melbourne metropolitan area (the capital city of the State of Victoria) and the rest of Victoria, with each area being considered in two time divisions, "high alcohol hours" (HAH) and "low alcohol hours" (LAH) (HAH and LAH are defined by Harrison 1990). The method of analysis used has been that of multivariate log-linear regression, which involves relating measures of road safety programs and economic effects, along with general trend and monthly variation, to the observed road trauma series via a regression equation. This method has proved useful in being able to establish the particular influence each factor in the regression equation has on the outcome measure (vis. road trauma). The general form of the fitted regression equation is where SCC; is the number of serious casualty crashes observed in month i, TREND is a linear trend factor, FEB, MAR, ... ,DEC are monthly dummy variables to account for regular seasonal variation, FACTOR1,FACTOR2, ... are measures of the road safety program, economic or social factors of interest and (a,b, ...,g) are parameters of the model which were estimated by multivariate log-linear regression. A separate regression equation of the form of equation (1) was fitted to the monthly serious casualty crash series each of the four strata defined above. Factors included in each regression equation were chosen based on the considerable experience built up from detailed analysis in past evaluation studies. This project aims to re-assess the contribution these factors have had on Victorian road trauma trends up to 1996. This has been achieved by extending the results of New stead et al (1996) to cover the period 1989 to 1996 in order to estimate the contribution of each major factor to reducing serious casualty crashes over this time. 2 MONASH UNIVERSITY ACCIDENT RESEARCH CENTRE 2. INCLUSION OF CRASH DATA TO THE END OF 1996 IN THE MODELS This section begins by detailing the trends in the factors considered in the models through to 1996. These factors were then used to model the monthly number of SCCs over the period 1983 to 1996 using the model structures of New stead et al (1995). 2.1 FACTORS INCLUDED IN THE MODELS The following section describes the factors, both road safety and other, which have been considered in modelling road trauma in Victoria over the time span under consideration. Whilst these factors are identical to those used in the work of Newstead et al (1996), it is considered of value to again present an overview of these in this report to show their continuing trends to the end of 1996. Knowledge of these trends will be useful in interpreting the results of the analysis presented further on. 2.1.1 Economic Measure A selection between the economic measure used by Newstead et al (1996), number employed, and the one used by Newstead et al (1995), unemployment rate, had to be made to be included in the main model. It was discovered using residual analysis for the estimated models that unemployment rate and the number employed showed similar random patterns and dispersion. Consequently, it was decided to use unemployment rate as the measure of economic activity since its interpretation is easier in the context of road trauma. Figure 2 shows unemployment rates for both Melbourne and the rest of Victoria over the period 1983 to 1996. Of note in Figure 5 is that, whilst the unemployment has risen sharply and peaked over the 1990-93, 1994-1996 has seen a recovery in the economy with unemployment rates falling during these years. FIGURE 2: Unemployment Rates in Melbourne (MSD) and the rest of Victoria (ROV), 1983-96 0.14 0.12 0.10 ~ 'E t 0.08 ." o Q. e 0.06 :> 0.04 0.02 i i ~ .•. .. " ~ ~ ~ ~ ~ 1 I-MSD --ROV I Month '•.. " N '" on 0.00 •. ;;; S is ill \\l '" .!J .., .!J .!J .!J .!J FURTHER MODELLING " ~ .., OF SOME FACTORS INFLUENCING ROAD TRAUMA TRENDS IN VICTORIA 3 2.1.2 Factors Relevant to High Alcohol Hour Crashes The major road safety initiative targeted at crashes in high alcohol hours (HAH) in Victoria since 1989 has been the random breath testing program (RBT) supported by a high profile advertising with drink-driving themes sponsored by the Transport Accident Commission (TAC). The program of increased RBT using highly visible booze-buses commenced in September 1989 with car based testing being predominant form of RBT before this time. With the introduction of the bus-based RBT program, a corresponding decline in car-based operation was observed. The Victoria Police discontinued recording activity in the car-based RBT program from January 1sI 1996 but rather, recorded this information under Preliminary Breath Test (PBT) statistics. Accordingly, it is unknown to what extent car based RBT activities have been conducted since this date. Hence, in this work, only the bus-based testing is considered in modelling HAH crashes. Past studies (Newstead et al 1995, Cameron et al 1992a, 1993a,b) have found the most appropriate measure of the RBT enforcement activity in relation SCC numbers to be the number of tests conducted. Figures 3a and 3b shows the number of bus-based RBTs conducted per month from 1983 to 1996 in metropolitan Melbourne and in the rest of Victoria respectively. In metropolitan Melbourne during 1996, the level ofRBT activity has remained similar to that during 1994 and 1995, with around 100,000 tests conducted per month, except for the final months in 1995 and initial months in 1996 and where the level of RBT aCtivity had gone beyond 120,000 tests per month. In the country Victoria the total number of RBTs have fallen to around 30,000 per month during 1996, which is around 2025,000 tests per month less than in previous years. FIGURE 3a : Number of Bus-based Random Breath Tests per month Melbourne (MSD) 1989 - 1996 I 140000 120000 " a: ~ .z"cE ~ '0 I 100000 80000 60000 ID 40000 20000 ~lllhl"" " ~lil ...• ...• '"N m '" C: 0 ::;: ~cn~m U) ~m~m ~CX)~Q) ~ ::;: >'('11 >. (") U') U) o..N a.. ~ LO C: >-m Q.m ~m~m >. 0.. ('l') ~ ~~ ~ ~~ gl\la; 11l 11l 11l m ia;!m Month 4 ~ ~ MONASH UNIVERSITY ACCIDENT RESEARCH CENTRE FIGURE 3b : Number of Bus-based Random Breath Tests per month Rest of Victoria (ROV) 1989 - 1996 50000 45000 40000 35000 {!!. m 30000 a: ~ 25000 •• .., E :i 20000 15000 10000 5000 0 i 0~ ~ IV o!on 0> N '" ....• ....• '3 1'; tlc: U 'I' '1£ C I'IV '3 c: Sl 5l '3 '3 S,l lli er;> er;> 0> ....• ....• " tl 'I' 0'3'I' .., ~ N ~ ~ ~ Supporting the RBT program in Victoria has been an intense publicity campaign sponsored by the TAC, of which one of the main themes has been drink-driving. Until the end of 1993, the levels of Adstock generated in metropolitan Melbourne and the rest of Victoria were reported as the same. During 1994, different levels of advertising were reported as placed in each region reflecting the Police campaign on drink-driving in Country areas and a series of advertisements specifically aimed at country people. Figures 4a, 4b and 4c show the levels of all TAC road safety advertising Adstock, TAC drink-driving Adstock and TAC speed and concentration Adstock respectively current in each month from 1989 to December 1994 in Melbourne and the rest of Victoria (for a definition of Adstock, see Broadbent 1979). Both in metropolitan Melbourne and country Victoria the level of awareness of TAC road safety advertising to have been maintained at significant levels throughout 1995 and 1996, although at slightly lower levels than the previous years. The levels of advertising specifically on the drink-driving theme in both regions have risen marginally in 1996 periods as compared to the other years. FURTHER MODELLING OF SOME FACTORS INFLUENCING ROAD TRAUMA TRENDS IN VICTORIA 5 FIGURE 4a(i) :TAC Road Safety Television Advertising - Monthly Adstock by theme Melbourne region il 0i; _ ~ ~ IL 20000 !l: lil Z Z 2000 4000 8000 14000 6000 12000 10000 16000 18000 1 ; m ~"~il ~ 2 ; g ~ Z IL 2 I_Drink Driving ~ ~ : ~ ¥ ~"~ill"lil ~ Z IL 2 ~ Z IL oSpeed ~ g mConcentration & & & ~ ~"~il ~ Z IL 2 CS eat Belt. _Fatigue m ~ ~ ~ ~•• z~ 2 _Motorcycle_j FIGURE 4a(ii) :TAC Road Safety Television Advertising - Monthly Adstock by theme Country Victoria region 18000 If ::E 12000 6000 10000 4000 8000 14000 16000 % <I( ~ ~ lil o g 2000_ .., N .IL .. '" ., IZ Lz"s 0( :e ~ ..>~ g !l:~ 0; ;J; N 0 IL ~ " ~ !l:~ 1> 1 > Belts _Fatigue _Motorcycles 6 .Country People 0( 0( ~ I MONASH UNIVERSITY ACCIDENT RESEARCH CENTRE FIGURE 4b(i) :Drink-driving Adstock : Melbourne region 1989-1996 8000 7000 6000 5000 4000 ::i 3000 0... u .!! > •• 10 •• 1000 2000 0 0> i •• ;, "~M•lil~>>>-• . '~~.cZ0z>-" ~It)u. ~ ~ % ~~ h~ t1 t 0> Cl> &l 15 a,; "0>-•• Cl> > h 0> ~ ~ FIGURE 4b(ii) :Drink-driving Adstock: Country Victoria region 1989-1996 i 3000 6000 u .!! 0:( > •• ... ." •• 2000 1000 5000 4000 FURTHER MODELLING OF SOME FACTORS INFLUENCING ROAD TRAUMA TRENDS IN VICTORIA 7 FIGURE 4c(i) :Speed and Concentration Adstock: Melbourne region 1989-1996 III > ~...: 'Cl cc Gi S 9000 5000 4000 III 7000 6000 8000 3000 2000 1000 :;..,.,.N~'".0z0;~0.,., '" .<0 ..., .., l\l 0 '" ~l\l ::\. Cl' '" ::\. FIGURE 4c(ii) :Speed and Concentration Adstock : Country Victoria region 1989-1996 ~ 'Cl (.l cc .s I> II ...: III Gi 4000 3000 8000 T 6000 2000 5000 7000 t I :; 0 ~ ~ ~z ..., , .., N .l\l ::> .'" .0; J. .,. <0 <0 1000 ::\. 0.., 0::\.., '" '" ::\. ~l\l '" 0 8 MONASH UNIVERSITY ACCIDENT RESEARCH CENTRE An index alcohol sales in Victoria has been included as an explanatory variable and represents an exposure measure of the potential to drink-drive. Figure 5 shows the monthly index of alcohol sales for the period 1983-1996. Whilst the index shows a downward trend over the years 1990-1996 there is perhaps some evidence of the series leveling from about 1994 onwards. FIGURE 5: Index of Alcohol Sales on Victoria, 1983-96 400 350 •• 300 'C ~ :; .5 250 •• 11 iii III 200 '0 ~ 8 cc '0 150 )( 11 ~ 100 50 0 .• •• c: M •.... en It) •.... c: .., ...• ::J "5 S SI S,lI; l\l M & :;~ J~ ~ er;> er;> .., ...• ...• .., ::J ::J "5 er;> er;> er;> 1 The most recent modelling of SCC trends in HAH (Newstead et al 1996) also considered the possible effects of the speed camera program on HAH crashes based on the results of work by Rogerson et al (1994) which examined the localised effects of the speed camera program during both high and low alcohol hours. The main findings of this study were observed reductions in casualty crash frequency due to localised speed camera effects during high alcohol hours. Whilst these effects were localised, being examined only within 1 km radius of speed camera sites, it is possible that more generalised speed camera effects were also present in HAH, similar to those measured in LAH by Cameron et al (1992b). Given the established presence of speed camera effects in HAH, it was considered necessary to again allow for the expression of these in the models describing HAH crashes in both Melbourne and the rest of Victoria. Figure 6 shows the number of speed camera TINs issued per month for the period January 1989 to December 1996. Examination of Figure 6 reveals the number of speed camera TINS issued through 1996 to be similar to the two previous years, averaging around 40,000 TINs per month. FURTHER MODELLING OF SOME FACTORS INFLUENCING ROAD TRAUMA TRENDS IN VICTORIA 9 FIGURE 6a: Monthly number of Traffic Infringement Notices Issued for Speeding Offences Detected by Speed Cameras Melbourne 1989 - 1996 " z.!!••'01=••E :> :> ~ 40000 60000 20000 30000 10000 ... 50000 i N ('\' (') (') It) 0:; ts tlc '~ c••\'') :;••c ~ l!,l l!,l ~ 8 ~ .!. Ol 0 .a;> ...., ..., Ol ...., ...., '\' Oi '\' FIGURE 6b : Monthly number of Traffic Infringement Notices Issued for Speeding Offences Detected by Speed Cameras Rest of Victoria 1989 - 1996 " z.;E•• ~ '0 ... :> :> 1= 8000 12000 6000 4000 10000 i 0~ :; :; ~ :; ~~ ...., :> .!. :> N It) (.13 ccc.')i ....,.. (') (') C. Ol .!. tl tl ...., tl 2000"3 ts ..., a;> •• '\' c. III O gtl l!,l ; Oi ;g Ol .. .. ..I. Ol •• Ol '\' 0 10 MONASH UNIVERSITY ACCIDENT RESEARCH CENTRE The model described by equation (1) was fitted to the serious casualty trends in high alcohol hours for metropolitan Melbourne and country Victoria. The major road safety campaigns included in the models were monthly number of RBTs, monthly number of speed camera TINs and monthly road safety media publicity as measured by Adstock. In considering which component of the media publicity to include in the models, the drink-driving component alone proved to be just as strong a predictor of SCC numbers as total publicity, but with a more targeted theme, and hence was included in the model. A linear trend component and monthly seasonal dummies were also included in the models along with unemployment rate and alcohol sales. A summary of the key model parameter estimates for Melbourne and country Victoria is given in Table 1. The models fitted to HAH SCC data for Melbourne and country Victoria, summarised in Table 1, explained 84% and 71% of the monthly variation respectively, based on the fitted model R-squared vales. A statistically significant linear trend was found in the data for metropolitan Melbourne, with the effects of unemployment rate and alcohol sales also being statistically significant. Of the road safety programs, the effects of random breath testing and speed camera TINs were statistically significant. The effect drink-driving publicity was not statistically significant however, with the factor having to be removed from the model because of a high co-linearity between it and the number of RBTs. The high degree of colinearity reflects the fact that the effects of drink driving publicity and number of RBTs are highly confounded meaning individual effects of each factor cannot be estimated independently. In practice the estimated effect of RBT tests in the model is more representative of the combined effect of RBT tests and associated publicity in tandem as a single program. Effects of the modelled factors in country Victoria were quite different from those in Melbourne. Here no statistically significant linear trend was found in the data along with no statistically significant effects due to unemployment rate or speed camera TINs. The effects of alcohol sales and number of RBTs were found to be statistically significant, however the same high co-linearity between RBTs and drink-driving publicity was observed, most likely causing the latter to show no statistically significant association with SCCs, as for metropolitan Melbourne. FURTHER MODELLING OF SOME FACTORS INFLUENCING ROAD TRAUMA TRENDS IN VICTORIA 11 Table 1: of • Effects of unemployment rate, number of random breath tests, alcohol sales (all Victoria), number of speed camera TINs issued, and Adstock of TAC drink-driving publicity on serious casualty crashes during high alcohol hours of the week. Melbourne and country Victoria 1983-96. trend of random bus-based random breath AREA OF VICTORIA alcohol sales during month number unemployment no. drink-driving of of speed Adstock breath camera tests inTINs month (car issued ortests buslogged rate in month 0.7985 0.7084 0.356 -0.375 0.0092 NS Tvalue 0.8892 -8.730 0.140 0.0291 -2.203 NS -2.639 NS++ 0.0228 0.0001 2.301 4.940 0.0004 0.0001 3.736 -4.639 -0.782 -0.0204*** 0.0025*** 0.0003 -0.0155** level 0.4357 -0.3149*** -0.0109* 0.5795*** 0.3067* Estimated (two-tailed) (-0.0293) NS variable Significance (0.0038)NS (0.0001) NS (-0.0068) NS df=151 df=155 * Statistically significant at p<O.05 level; ** Highly statistically significant at p<O.OJ level; *** Very highly statistically significant at p<O.OOJ level; NS = not significant; ++ No estimate available 2.1.3 Factors Relevant to Low Alcohol Hour Crashes Past studies have found the speed camera program to be a major factor influencing SCC numbers in low alcohol hours (LAH), with the most appropriate measure of program activity being in the number of TINs issued for speeding offences detected by speed cameras each month. Like the RBT program, the speed camera program in Victoria has also been supported by a high exposure TAC publicity campaign with the speeding theme. Studies also have found both speeding and concentration publicity campaigns to be a factor influencing SCC numbers in LAH. Figures 4a,b and c, depicting TAC television advertising Adstock, show that during 1996, the level of awareness of advertising with speeding and concentration themes (Figure 4c) remained at significant levels, although at perhaps slightly lower levels in magnitude compared to the two previous years Once again, models of the form of equation (1) were fitted to low alcohol hour SCCs separately in Melbourne and the rest of Victoria, including linear trend, monthly seasonal dummies, unemployment and road safety program measures as model covariates. The two major road safety program measures included in the low alcohol hour models were the 12 MONASH UNIVERSITY ACCIDENT RESEARCH CENTRE number of speed cameras TINs issued per month, and the road safety media publicity as measured by Adstock. Exploratory analysis showed Adstock for both "speeding' and "concentration" themes together showed greater association with monthly SCCs than did just speeding alone. Table 3 summarises the parameter estimates for the LAH models for the two regions of the state. All four main factors fitted were statistically significant predictors in the model of LAH SCCs in Melbourne, with the model explaining 78% of the variation in the data. The number of speed camera TINs issued was a highly statistically significant predictor in the models as were speeding and concentration Adstock and unemployment rate. For LAH SCCs in the rest of Victoria, the fitted model explained 65% of the variation in the data. Unlike the models estimated for HAH SCCs in the rest of Victoria, trend and unemployment rate showed a statistically significant association with SCCs in low alcohol hours. Road safety publicity with the speeding and concentration theme also showed a statistically significant association with LAH SCCs in the rest of Victoria, however the effect of speed camera TINs was not statistically significant, with the factor being removed from the model because of co-linearity problems with the Adstock measure. Table 2: Effects of unemployment rate, number of speed camera TINs issued, and Adstock ofTAC "speeding"and "concentration"publicity on serious casualty crashes during low alcohol hours of the week Melbourne and country Victoria 1983-96. trend AREA OFspeed VICTORIA no. unemployment of rate month issued speeding and concentration Adstock -0.1475* NScamera T valueinTINs 0.0041 -2.916 -0.0013** 0.0090 0.0022 -3.108 -2.648 -0.0196** 0.0002 3.774 0.0015*** 0.0293 -2.546 -4.484 -5.349 -0.2191** NS NS++ Estimated -0.0177** 0.0001 -0.0209*** level logged (two-tailed) variable Significance df=156 df=152 t of * Statistically significant at p<O.05 level; ** Highly statistically significant at p<O.Ollevel; *** Very highly statistically significant at p<O.OOllevel; NS = not significant; ++ No estimate available FURTHER MODELLING OF SOME FACTORS INFLUENCING ROAD TRAUMA TRENDS IN VICTORIA 13 2.2 DISAGREGGATION AND COMBINATION OF THE RESULTS Using the methods of Newstead et al (1995), each of the fitted models have been decomposed to estimate the annual effect of each significant factor on road traumas in each strata for the years 1990-1996. In calculating the influence of each of the factors in the model, a suitable comparison base had to be chosen. Following the methods of Newstead et al (1995), the influence of the road safety programs was compared to a zero base, the zero base representing the situation if the programs had not been in operation (viz. operation levels were zero). For the variables unemployment rate and alcohol sales, comparison with a base level of zero is not meaningful as these measures will in practice never attain this level. Hence for these two factors, their influence was calculated relative to the level of the factors during 1988, the year in which both SCC numbers and unemployment rate were near their maximum and minimum levels respectively. Appendix A details disagreggation of the effects of each of the factors in the models for the four strata, as well as the combination of these results by time of day and region of Victoria. Table 3 shows the overall combination of these results to estimate the total annual contribution of each of the major factors considered to have reduced SCC numbers in Victoria during the years 1990-96. It should be noted that in Table 3, estimates for the effect of RBT and drink driving publicity are not given separately because of the high degree of co-linearity between these factors in the estimated models, as noted above. Instead, the effects of these two factors have been combined into a single measure labelled as the drinkdriving program. Table 3 : 8394 7.9% 13.5% 40.5% 25.3% 9.5% 5.5% 10.4% 1994 1992 1993 1991 1995 1996 8790 5111 5233 8522 5025 4950 8654 5192 5211 9075 8931 5371 5432 10.0% 12.1% 10.9% 11.2% 41.0% 25.0% 7.0% 11.1% 6.1% 14.8% 7.1% 15.6% 6.7% 42.6% 25.3% 9.9% 8.9% 9.4% 7.0% 25.0% 9.6% 6.5% 11.0% 37.9% 25.2% 10.0% 8.8% 10.3% 6.2% 39.7% 39.2% 51'84 5475 5196 11.1% 11.2% 10.2% 5286 Estimated reductions in serious casualty crashes attributable to various factors - Victoria, All Hours, 1990-96 1990 8270 3.0% 9.6% 1.9% 25.8% 8.9% 6219 6136 21.8% 5.0% on of drink-driving program s had remained at 1988 levels 2.3 EFFECT OF ACCIDENT BLACK SPOT TREATMENTS AND OTHER FACTORS In producing the above estimates of the effect that various major factors have had on road trauma in Victoria, monthly serious casualty crash numbers have been modelled as a function of road safety, economic and social factors. As discussed in Newstead et al (1995), 14 MONASH UNIVERSITY ACCIDENT RESEARCH CENTRE the form of the models fitted, shown in equation (1), includes a general linear trend component and monthly dummies. These are included in order to explain residual variation in the data not explained by the factors of interest in the model, so as to achieve a satisfactory fit of the model to the observed data. The roles of the monthly dummies and general linear trend factor in explaining residual variation in the fitted models are quite different. Monthly dummy variables are included in order to explain residual seasonality in the monthly crash data over a year. When included in the models here, the dummies may represent residual variation in the crash data due to, for example; environmental effects such as rainfall and daylight hours, the effects of holiday periods, or the differing number of days in each month. None of these factors were considered explicitly in the fitted models. When considering effects aggregated on an annual basis however, seasonal effects represented by monthly dummies will be the same each year and hence of little interest. In contrast, the role of the general trend factor is to explain residual variation in the data due to factors varying slowly over time in a monotonically increasing or decreasing fashion. Factors not included in the models which may be represented by the included trend component are, for example, population increases, driver licensing increases, gradual and continual improvement of the road network system or, in fact, any other time variant, nonseasonal factor. The trend component of the model represents the average effect of all these other non-seasonal factors not included explicitly in the models. The effect of the general linear trend component can be seen in the change in expected serious casualty crashes from year to year shown in Table 3. One road safety program whose effectiveness is of interest, and is currently represented by a component of the trend in the analysis presented here, is accident blackspot (ABS) treatments. The data available on accident blackspot treatments in Victoria is available only in a highly condensed form from RTA, RCA, VicRoads and Ministry of Transport annual reports, giving the total number of sites treated and total expenditure for each financial year. Ideally, this data would be included as a factor in the SCC data models and the model parameters re-estimated to establish the effectiveness of the ABS program in reducing SCC frequency. Due to the nature of the ABS data available however, this procedure is not practical for a number of reasons. Firstly the data is available only on an annual basis whilst the SCC data is modelled on a monthly basis. This means that, for inclusion in the modelling process, the ABS data would need to be proportioned across the 12 months of the year. To achieve this, an average monthly ABS expenditure or treatment profile would need to be known or assumed. Such a profile is not known and assuming one would be entirely arbitrary. Secondly, the effects of ABS treatments are thought to be cumulative meaning that once each site is treated, it remains effective in reducing crash numbers for many years after. Such a cumulative process, when represented numerically as a covariate for modelling, has a very smooth shape, which is similar to the functional form of the trend component used in the models, fitted in this project. As the trend function and ABS treatment profile are of similar shape, it is not practical to include both factors together as covariates in a single model due to colinearity problems. Inclusion of the ABS data alone in the model would lead to incorrect estimates of effectiveness. This is because the ABS series would take on the role of the trend factor in the model and act as a proxy for all factors not included explicitly in the model, as FURTHER MODELLING OF SOME FACTORS INFLUENCING ROAD TRAUMA TRENDS IN VICTORIA 15 did trend alone. Hence an alternative method of including the effects of ABS treatments in the estimated overall percentage reductions in SCC numbers is required. Appendix B details the process by which the estimated reductions in SCC numbers due to ABS treatments can be obtained. It is assumed firstly that each ABS site treated saves one casualty crash per year from the year of implementation onwards (Cunningham 1993). This assumption may be slightly optimistic for the TAC funded black spot treatment program from 1992 to 1994, however it is felt it still leads to estimates which are generally indicative of the relative countermeasure effectiveness. As the ABS data is given in financial years, it is assumed that all sites were, on average, completed mid year, ie. they were fully effective by the commencement of the calendar year. In this way, a cumulative profile of casualty crashes saved across the years of interest is built up. Casualty crash numbers are then converted to serious casualty crash numbers, for use here, by using the average ratio of serious casualty crashes to all casualty crashes in Victoria for each year. Using the expected number of serious casualty crashes if road safety and other initiatives had remained at 1988 levels from Table 3, along with the estimated annual SCC savings due to ABS treatments, a percentage reduction in SCC numbers due to ABS treatments can then be estimated. Using the multiplicative percentage reduction theory of Newstead et al (1995), the estimated percentage reductions in SCC numbers in Victoria due to ABS treatments over the period 1990-96, relative to 1988, can be factored from the trend component. Sections 8.1 and 8.2 of Newstead et al (1995) give details of the steps involved in factoring out the ABS treatment effects. Appendix B shows that the estimated percentage reductions of ABS treatments on SCC numbers, relative to the 1988 base year, ranged from 1.6% in 1990 to 6.2% in 1995 and then fell to 5.6% during 1996. The reason for the fall during 1996 is two fold, being a combination of a dramatic cut in accident blackspot expenditure after the 1993/94 financial year and a fall in the ratio of the serious casualty crashes to all casualty crashes in Victoria in 1996. 2.4 SUMMARY OF ESTIMATED CONTRIBUTIONS Table 4 gives a summary of the estimated percentage reductions in serious casualty crashes in Victoria attributable to each of the factors considered in the modelling process, along with the contribution of the accident black spot treatments. Shown are the observed, expected and estimated annual percentage reductions in the number of serious casualty crashes over the period 1990-96, along with the individual contribution of each of the major factors considered and the total contribution ofthe road safety programs in each year. 16 MONASH UNIVERSITY ACCIDENT RESEARCH CENTRE 2.5% 7.0% 7.1% 11.1% 9.9% 5.3% 3.2% 10.2% 45.6% 44.0% 4950 1993 1994 1996 1995 5475 10.9% 10.0% 9.4% 26.9% 6.7% 9.5% 29.3% 27.7% 6.2% 6.1% 11.2% 11.0% 6.5% 10.0% 29.4% 5.6% Estimated reductions 1.6% 8.9% 5.0% 1992 5111 42.7% 1991 5025 5211 9099 8770 39.3% 5.5% 15.6% 14.8% 9345 5371 7.9% 8.9% 13.5% 9480 5286 9.6% 5233 9572 5196 42.8% 42.7% 8.8% 10.3% 10.4% 8371 1.9% 29.8% 29.7% 9.6% 23.0% 8585 5192 5184 6219 6136 26.7% 3.0% Table 412.1% :5432 in serious casualty crashes to major factors: Victoria 1990-96. ution of all drink-driving program remained at 1988 tion of road levels safety attributable programs 1990 Figure 7 summarises graphically the information given in Table 4. Each column in Figure 7 shows the percentage reduction in serious casualty crashes attributable to each of four road safety programs considered along with the contributions of the changes in alcohol sales and unemployment rate. The line on Figure 7 gives the total contribution of all the factors considered in reducing road trauma, obtained by multiplying the individual factor effects. Use of multiplicative methods to obtain the total effect was necessary because of the multiplicative model structure used to obtain the effect estimates. FIGURE 7: Estimated reductions in serious casualty crashes attributable to major factors: Victoria 1990-96. Year 1990 1991 1992 1993 1994 1995 1996 0.0 IliiiZJAccident Blackspot -10.0 " tl Ill:0c•• ::J ~ c::::J Unemployment c::::JAlcohol sales -20.0 -30.0 -40.0 _ -50.0 Drink-driving program IiiiiiilSpeed and Concentration Publicity _Speed TINs -+- Total Camera Reduction in SCCs -60.0 FURTHER MODELLING OF SOME FACTORS INFLUENCING ROAD TRAUMA TRENDS IN VICTORIA 17 3. DISCUSSION Results of the analysis presented in Table 3 and Appendix A show a few noteworthy differences in comparison to the results of Newstead et al (1995) covering the years 199093. One of the most notable differences is the estimated magnitude of the effects of the road safety programs on SCCs in the rest of Victoria. Newstead et al (1995) estimated the road safety programs in the rest of Victoria to have lead to reductions in SCCs of the order of 26% in LAH and 20% in HAH. The results presented here, however put these estimates at around only 15% and 14% respectively. There appears to be two principal reasons for these lower estimates. In LAH, whilst the estimated effect of speed and concentration publicity is only slightly lower, speed camera TINs were no longer found to be a significant predictor of SCC numbers in this stratum. Previously, speed camera TINs were estimated to have reduced SCC numbers by around 11%. For HAH crashes in the rest of Victoria, the effects of the drink-driving program has fallen from an estimated 20% reduction to an estimated 12% reduction in the current work. Whilst the influence of alcohol sales on HAH SCCs in the rest of Victoria is much the same, unemployment rate is no longer a significant predictor of SCC numbers in the stratum. The reasons for the differences discussed above are not clear although there are a number of possibilities. It is possible that the effect of the road safety programs in the rest of Victoria has declined significantly over the period after 1993 as people become used to the programs. This may be particularly pertinent to aspects such as media publicity and the speed camera program. In the case of the drink-driving program in the rest of Victoria, the changing result may be a reflection of a change in enforcement strategy not being adequately captured by the program measure currently in use. Rural drink-driving enforcement has changed to the use of satellite cars in combination with booze buses to counter such effects as the 'bush telegraph' and the local population's knowledge of back roads whilst the program measure being used is still only the number of booze-bus tests carried out. Such possibilities would need to be checked by further specific research into the effects of these programs. This might be achieved using methods such as inclusion of factors in the models to test for time trends in the effectiveness of various factors. Another possibility is investigation of the use of other program measures which potentially better reflect the current methods of program operation, as in the case of drink-driving enforcement. Another potential reasons for the differences discussed above is that the modelling procedure is more unstable for SCCs in the rest of Victoria. This is possible given the relatively smaller number of SCCs in comparison to metropolitan Melbourne and the generally lower statistical significance levels of the fitted parameters and R-squared values of the fitted models. A sensitivity analysis on the fitted models would need to be carried out to determine this. Estimates of the relative effects of each modelled factor on SCCs in metropolitan Melbourne obtained here are closely concordant with those obtained by Newstead et al (1995). A few differences however exist in the estimates for HAH crashes in this region. Once again, estimated SCC reductions due to the effects of the drink-driving program are lower here at 20% against 28% previously. This may be partly due to the fact that the effect of speed camera TINs has been included here, where it was not previously, with the effects of speed 18 MONASH UNIVERSITY ACCIDENT RESEARCH CENTRE camera TINs and the drink-driving program not being completely independent. It is interesting to note that the estimated total combined effect of the road safety programs for this stratum is similar here to previous estimates. Like the rest of Victoria, the effect of alcohol sales on HAH crashes in Melbourne is lower here than in previous work. Another point not clear from the modelling work is the relationship between the index of alcohol sales and the drink-driving program. Whilst modelling does not indicate any particular co-linearity between these factors, it is possible a relationship exits in a subtle or indirect way not reflected in the modelling procedure. Subtle relationships may also exist between alcohol sales and unemployment rate, with alcohol sales possibly driven to a certain degree by economic activity which unemployment rate is reflecting. Despite the differences noted above, the overall estimate of the effect each major factor has had on road trauma in Victoria as a whole are generally consistent with previous estimates, particularly in a relative sense. It is again appropriate in discussing the results of the road trauma trends modelling detailed in this report to reiterate the purpose and interpretation principles of the work presented. The general aim of the work presented here is to estimate the relative effectiveness of some major road safety and socio-economic factors in reducing road trauma in Victoria over the period 1990-1996. Obviously this work is not all encompassing in assessing every possible factor which may have contributed to changes in road trauma over this period. Indeed, there is likely to be many successful programs that have not been considered in this work. There are number of reasons for not considering programs other than those appearing here. Firstly, many of the programs not considered are targeted specifically at a narrow range of crashes and/or injuries and hence, whilst they may be highly effective on their target area, their effects are not global enough to measure by the techniques considered here. These last comments highlight the important fact that the methods and results presented here are in no way to replace formal countermeasure evaluation in the traditional manner. Historically, development of the methods presented in this report stemmed from a desire to combine results of a number of formal evaluations to gain an overall picture of the relative effect of each countermeasure that had been in considered in relation to Victoria as a whole during all the periods of the week. To this end, the methods developed were very successful and have remained so in the work presented to date. The methods will remain useful for this purpose in the future also. 4. CONCLUSION Building on the work of a project by Newstead et al. (1995) this project details results of statistical modelling to asses the influence of some major factors influencing road trauma in Victoria over the period 1990-1996. Unlike the previous work, the work presented here considers the effect of the speed camera program on high alcohol hour crashes in Victoria. A number of road safety measures and other factors contributed to the reductions in road trauma in Victoria during the years 1990 to 1996. The major contributors and the apparent percentage reduction in serious casualty crashes due to each measure/factor were: FURTHER MODELLING OF SOME FACTORS INFLUENCING ROAD TRAUMA TRENDS IN VICTORIA 19 • Speed camera operations (principally speeding TINs issued): 10-11 % each year • "Speeding" and "concentration" television advertising: 5-7% each year • Drink-driving program (bus-based RBT together with "drink-driving" publicity campaigns) • Reduced alcohol sales: 9-10% each year 3% in 1990 6% in 1991 7% in 1992 9% in 1993 8% in 1994 9% in 1995 10% in 1996 • Reduced economic activity (measured by unemployment rates): 2% in 1990 12% in 1991 15% in 1992 16% in 1993 14% in 1994 10% in 1995 10% in 1996 • Accident Black Spot treatments 1.6% 2.5% 3.2% 5.3% 6.2% 6.2% 5.6% in in in in in in in 1990 1991 1992 1993 1994 1995 1996 These percentages cannot simply be added up to estimate the total contribution. If more than one contributor is being considered, the percentage reduction of each must be applied in turn. The anti-speeding and drink-driving programs together are estimated to have contributed reductions in serious casualty crashes of at least 22-25% during these seven years. Including the accident blackspot treatments, the overall contribution of road safety initiatives is estimated to have risen from 23% reduction in 1990 to nearly 30% reduction in 1993- 1996. 5. FURTHER WORK RECOMMENDED The results of this research indicate two general areas of further research which should be pursued. (1) Further developments of the methods of road trauma trends modelling. 20 MONASH UNIVERSITY ACCIDENT RESEARCH CENTRE Whilst the general framework of modelling which has been developed in this project and its predecessors has proven extremely valuable in monitoring Victorian road trauma, a number of opportunities for fine tuning and enhancing these methods exist. Potential enhancements may include: • Development of total program measures for countermeasure groups with high interdependency, such as the drink-driving program factors • Development of methods to account for program interactions as part of the modelling process • Development of methods to place confidence limits on the combined percentage reductions estimates. • Detailed study of the relationships between various socio-economic measures (unemployment rates, employment numbers and alcohol sales) and road trauma outcomes. • Sensitivity analysis of the modelling process for road trauma in the rest of Victoria (2) Investigation of the effects of road safety programs in the rest of Victoria. Comparison of the results obtained in this research when compared to previous research indicate a possible reduction has occurred in the effectiveness of the major road safety programs considered in reducing road trauma in the areas of Victoria outside metropolitan Melbourne. Further specific research is needed to determine whether this is indeed true and, if so, to establish the reasons for this diminishing effectiveness and possible ways to counter these. 6. ACKNOWLEDGMENTS David Farrow and Geoff Elston of VicRoads are gratefully acknowledged for supplying the crash data files used in this project. Ron Cook and Paul Williamson of the Traffic Camera Office are acknowledged for supplying data on the number of traffic infringement notices. Grey advertising is thanked for supply of the TARP data for the TAC advertising campaign. 7. REFERENCES BROADBENT, S (1979), "One way TV advertisements work". Journal of the Market Research Society, Vol. 21, No. 3. London. CAMERON, MH, CAVALLO, A, and SULLIVAN, G (1992a), "Evaluation of the random breath testing initiative in Victoria 1989-91: Multivariate time series approach". Report No. 38, Monash University Accident Research Centre. CAMERON, MH, CAVALLO, A, and GILBERT, A (1992b), "Crash-based evaluation of the speed camera program in Victoria 1990-91. Phase 1: General effects. Phase 2: Effects of program mechanisms". Report No. 42, Monash University Accident Research Centre. CAMERON, MH, HAWORTH, N, OXLEY, J, NEWSTEAD, S and LE, T (1993) "Evaluation of Transport Accident Commission road safety television advertising". Report No. 52, Monash University Accident Research Centre. FURTHER MODELLING OF SOME FACTORS INFLUENCING ROAD TRAUMA TRENDS IN VICTORIA 21 CAMERON, MH, and NEWSTEAD, SV (1993) "Evaluation of mass media publicity as support for enforcement". Paper presented at Australasian Drink-Drive Conference, Melbourne, November 1993. CAMERON, MH, NEWSTEAD, SV, and VULCAN, AP (1994), " Analysis of reductions in Victorian road casualties, 1989 to 1992". Proceedings 17th ARRB conference, Part 5, pp165-182. CUNNINGHAM, J. (1993) "Accident savings accruing from ABS programs", Memo, VicRoads, Road Safety Division, 24th May 1993. HARRISON, WA (1990) "Update of alcohol times as a surrogate measure of alcoholinvolvement in accidents". Research Note, Monash University Accident Research Centre. NEWSTEAD, S., CAMERON, M., GANTZER, S. and VULCAN, P. (1995) "Modelling of some major factors influencing road trauma trends in Victoria 1989-93" Report No. 74, Monash University Accident Research Centre. NEWSTEAD, S., GANTZER, S. and CAMERON, M. (1996) "Updated modelling of some major factors influencing road trauma trends in Victoria 1990-94: all crashes and specific crash sub-groups" Monash University Accident Research Centre. ROGERSON, P., NEWSTEAD, S. and CAMERON, M. (1994) "Evaluation of the speed camera program in Victoria 1990-1991. Phase 3: Localised effects on casualty crashes and crash severity. Phase 4: Generalised effects on speed." Report No. 54, Monash University Accident Research Centre. THORESEN, T, FRY, T, HEIMAN, L and CAMERON, MH (1992), "Linking economic activity, road safety countermeasures and other factors with the Victorian road toll". Report No. 29, Monash University Accident Research Centre. 22 MONASH UNIVERSITY ACCIDENT RESEARCH CENTRE APPENDIX A Percentage reductions in serious casualty crashes attributable to various sources Melbourne Statistical Division (MSD) : 1990 -1996 relative to 1988 1553 22.3% 3154 6170 1618 3287 1777 3061 3450 50.8% 1661 1518 1806 3250 3530 12.2% 1992 3017 1734 2964 1968 1 3181 3125 12.5% 11.0% 16.0% 18.6% 10.8% 20.1% 10.2% 48.5% 29.3% 10.9% 47.1% 6.9% 4.4% 19.4% 10.3% 28.8% 29.1% 15.2% 6025 6792 48.0% 46.7% 43.5% 13.6% 10.6% 13.8% 23.2% 10.7% 20.7% 3284 3813 3745 6.4% 15.0% 15.6% 6631 7.2% 5.6% 3348 3555 50.4% 53.3% 49.2% 20.6% 3387 6.1% 6.7% 6473 6320 43.9% 15.3% 45.5% 15.7% 17.4% 20.0% 2007 40.3% 16.3% 14.2% 42.5% 31.4% 13.1% 19.8% 19.7% 14.0% 11.1% 11.3% 31.1% 13.4% 3070 1869 1766 40.2% 38.0% 30.0% 13.7% 38.1% 14.6% 20.3% 43.8% 29.0% 15.4% 10.8% 29.2% 28.9% 6.3% 14.9% 7.8% 8.6% 13.4% 3238 30.8% 30.3% 1770 19.5% 16.9% 2971 1977 19.7% 30.7% 2912 4.7% 2.4% 17.6% 13.5% 5882 9.6% 3.9% 3.2% 4079 26.0% 27.4% 10.0% 33.4% 25.9% 27.8% 2.6% 2102 5.0% 10.1% 1991 APPENDIXA1 1662 24.9% 1233 793 899 19.7% 15.4% 31.2% 14.9% 33.9% 13.4% 22.2% 37.1% 1996 775 796 815 24.6% 1233 18.7% 873 2352 1136 1101 1666 890 15.0% 2317 1703 1.1% 2283 891 1084 1050 15.5% 5.4% 1687 887 16.5% 2.4% 2300 1067 18.1% 1.1% 924 2387 2056 5.1% 15.7% 27.1% 849 1993 1992 35.7% 1991 1995 1994 35.4% 8.5% -1.5% 0.0% 993 12.0% 19.5% 10.3% 2.8% 20.5% 7.5% 7.3% 2369 1738 26.1% 15.2% 13.2% 7.6% 6.3% 14.0% 3.3% 13.5% 2334 28.6% 1118 11.8% 8.0% 2.5% 26.5% 20.7% 6.9% 7.8% 6.2% 14.5% 27.2% 7.7% 6.7% 1823 14.4% 26.6% 888 4.0% 4.4% -3.0% 0.0% 1154 1064 11.0% 12.0% 13.9% 6.2% 7.8% 10.6% 10.6% 23.0% 0.0% 7.0% 14.3% 13.8% 8.2% 6.0% 1.9% .................. Percentage reductions in serious casualty crashes attributable to various sources APPENDIXA2 ROY) 1992 1991 1994 14.6% 5475 2402 8522 2517 8394 2757 8790 24.7% 4788 4135 4387 4294 4581 4248 1995 12.1% 13.2% 18.5% 24.9% 45.3% 16.1% 2581 9075 11.2% 11.6% 14.8% 10.3% 6.2% 2623 2694 2894 4950 5025 41.4% 14.9% 19.3% 46.1% 14.7% 37.9% 8654 40.5% 2476 14.3% 5211 18.4% 25.5% 2574 8931 2859 5432 5233 9.6% 8270 2.7% 1.9% 2970 6136 3166 23.1% 17.5% 29.4% 25.8% 5.0% 10.8% 27.0% 13.6% 18.2% 4288 7.8% 8.0% 14.5% 13.5% 9.6% 10.2% 26.1% 4100 4483 15.1% 32.5% 7.7% 5.5% 4209 12.2% 4683 25.0% 26.6% 14.4% 12.8% 7.9% 16.9% 34.3% 32.7% 4171 1996 1993 10.9% 41.0% 7.1% 11.1% 25.7% 39.7% 11.0% 42.8% 10.4% 15.6% 6.5% 6.1% 19.2% 45.0% 46.0% 2311 19.1% 25.4% 48.4% 16.8% 2639 39.2% 10.0% 25.3% 7.0% 9.5% 36.6% 9.0% 26.3% 34.5% 25.2% 9.4% 13.2% 14.0% 8.8% 10.0% 26.5% 17.2% 36.7% 9.1% 5.8% 12.6% 3.0% 4204 4066 6.8% 1.0% 21.8% 21.5% 22.1% 8.9% 10.2% 6.7% 8.9% 9.9% Percentage reductions in serious casualty crashes attributable to various sources APPENDIXA3 AI..I..VICTORIA APPENDIXB 531 157 0.4515 49 82 375 278 204 257 300 337 6780 246 No. of sites 118 90TVTV 278 204 300 300 7748 7351 8394 8033 40 40 2027 0.3815 0.3850 0.3154 0.3312 1749 0.3234 0.3433 0.3% 40 REDUCTIONS 109.5 234.1 7460.5 7134.1 416.5 9881.9 636.1 566.5 9158.1 90 118 1989 1995 1994 1993 375 82 531 278mid 77.4 257 337 337VEO (assuming 7084 7476 8150 8654 8522 9075 8931 8270 40 157 246 246 130 1288 137.8 287 0.3601 0.3596 651 405 1.9% 4.9% 6.3% 0.7% 0.3138 296.9 6.9% 9.6% 2933 3064 1.5% 0.3239 508.0 9.7% 4.1% 0.2936 1545 2558 3015 0.3087 560.4 611.8 607.3 0.3227 988 50.1 7621.8 145.8 8960.5 8566.5 9501.2 3.2% 6.2% 847.2 899.6 950.9 9974.6 8746.9 476.9 9736.5 1.6% 339.2 8372.2 157 6798.1 18.1 0.9% 90 118 90 1985 1984 1988 1990 531 8790 2.9% 227.3 9.0% 8.9% 5.5% 2.5% 5.3% 5.6% 946.5 1987 1986 1992 1991 1996 375 ESTIMATION OF 7982.1 SERIOUS CASUAL CRASH DUE TO 257 ACCIDENT BLACKSPOT TREATMENTS CASUAL CRASHES SA all sites 204 treated at financial year) Year 1983 TOTAL CAS CRASHES SERIOUS CAS CRASHES EXPECTED* 82 49 APPENOIXB