**CHAPTER 4 Modelling road safety in twenty-one countries**

**4.9 France**

For France, a model of the fatality rate was constructed using the all-occupant seat belt wearing rate, as well as dummy variables, two of which represent changes in speed and other laws affecting traffic. The results are shown in Table 4.51.

Table 4.51 Regression results for predicting the French fatality rate

*Regression Statistics*

Multiple R 0.9956359

R Square 0.9912909

Adjusted R Square 0.9904992 Standard Error 1.7799184

Observations 49

ANOVA

*df* *SS* *MS* *F* *Significance F*

Regression 4 15866.50849 3966.6271 1252.048 1.09017E-44

Residual 44 139.3968178 3.1681094

Total 48 16005.9053

*Coefficients Standard Error* *t Stat* *P-value* *Lower 95%* *Upper 95%*

Intercept 54.021026 0.663007 81.478870 0.000000 52.684824 55.357228

Seat belts -0.473259 0.013317 -35.53831 0.000000 -0.500097 -0.446421

Dumpre65 10.314777 1.221076 8.447287 0.000000 7.853861 12.775693

50urban -1.447701 0.865666 -1.672355 0.101552 -3.192336 0.296935

03laws -2.170571 0.804584 -2.697756 0.009860 -3.792103 -0.549039

Figure 4.157 shows the pattern of the fatality rate is fairly accurately predicted by the model.

Figure 4.157 Actual/predicted French fatality rate

Fatalities per billion safety-wgt vkt

Actual Predicted

0 10 20 30 40 50 60 70

2029 2027 2025 2023 2021 2019 2017 2015 2013 2011 2009 2007 2005 2003 2001 1999 1997 1995 1993 1991 1989 1987 1985 1983 1981 1979 1977 1975 1973 1971 1969 1967 1965 1963

Figure 4.158 shows the components of the French fatality prediction/forecast. The major influence is the increase in seat belt wearing. From the early 90s, the reduction in the urban speed limit to 50 km/hr starts to have an effect, as do new traffic laws in 2003.

Figure 4.158 Components of the French fatality rate prediction

0

Impact of seatbelts + dummies + urban 50km + 2003 laws Impact of seatbelts + dummies + urban 50km

Fatality rate Impact of seatbelts Impact of seatbelts + dummies

Fatalities per billion safety-wgt vkt

Figure 4.159 shows that the modeling also produces a fairly accurate prediction of the level of fatalities over time.

Figure 4.159 Actual/predicted French road deaths

Actual Predicted

0

Chapter 4 • Modelling road safety in twenty-one countries The safety weights applied in France were cars, light commercial vehicles and other vehicles 1.0, motorcycles 20.0, buses 1.5, trucks 2.0 and mopeds 10.0. The injury weighted vkt calculation has motorcycles at 30.0 and mopeds at 15.0.

In France, the injury rate (road injuries per billion injury-weighted vkt) moved in sync with the fatality rate, except early in the period. This is shown in Figure 4.160.

Figure 4.160 French injury and fatality rates

0 200 400 600 800 1000 1200 1400

2011 2009 2007 2005 2003 2001 1999 1997 1995 1993 1991 1989 1987 1985 1983 1981 1979 1977 1975 1973 1971 1969 1967 1965 1963

Fatalities per billion safety-wgt vkt

Injury rate Fatality rate

Injuries per billioninjury-wgt vkt

0 10 20 30 40 50 60

A model of the injury rate was constructed using the fatality rate, two dummy variables and a time trend prior to 1988. The regression results are shown in Table 4.52.

Table 4.52 Regression results for predicting the French injury rate

*Regression Statistics*

Multiple R 0.9988514

R Square 0.9977042

Adjusted R Square 0.9974955 Standard Error 14.2974998

Observations 49

ANOVA

*df* *SS* *MS* *F* *Significance F*

Regression 4 3908795.43 977198.85 4780.383 2.00111E-57

Residual 44 8994.414026 204.41850

Total 48 3917789.844

*Coefficients Standard Error* *t Stat* *P-value* *Lower 95%* *Upper 95%*

Intercept 69.276109 8.461770 8.186953 0.000000 52.222533 86.329685

FatalRate 14.688031 0.252530 58.163545 0.000000 14.179090 15.196971

Dumpre66 -71.865078 11.343584 -6.335306 0.000000 -94.726570 -49.003587

Dumpre73 -4.832204 0.831540 -5.811152 0.000001 -6.508062 -3.156346

Timepre88 10.245268 0.612795 16.718926 0.000000 9.010262 11.480274

Figure 4.161 shows the pattern of the injury rate is fairly accurately predicted by the model.

Figure 4.162 shows that the modeling also produces a fairly accurate prediction of the level of injuries over time.

Figure 4.161 Actual and predicted French injury rate

Actual Predicted

Injuries per billion injury-wgt vkt

2029

Figure 4.162 Actual and predicted French road injuries

Actual Predicted

2029 100 000 150 000 200 000 250 000 300 000 350 000 400 000 450 000

Chapter 4 • Modelling road safety in twenty-one countries

### 4.10 Germany

For Germany, a model of the natural log of the fatality rate was constructed using the all-occupant seat belt wearing rate and measures of alcohol deaths per total deaths and real petrol price, as well as a dummy variable representing speed control measures and a time trend prior to 1979. The results are shown in Table 4.53.

Table 4.53 Regression results for predicting the log of the German fatality rate

*Regression Statistics*

Multiple R 0.9984244

R Square 0.9968514

Adjusted R Square 0.9964765 Standard Error 0.0547951

Observations 48

ANOVA

*df* *SS* *MS* *F* *Significance F*

Regression 5 39.92456776 7.9849135 2659.4218 2.26935E-51

Residual 42 0.126104993 0.0030025

Total 47 40.05067275

*Coefficients Standard Error* *t Stat* *P-value* *Lower 95%* *Upper 95%*

Intercept -0.804536 0.317519 -2.533819 0.015102 -1.445315 -0.163756

Seat belts -0.009564 0.000690 -13.86142 0.000000 -0.010956 -0.008172

Alcohol 1.522264 0.097363 15.635008 0.000000 1.325778 1.718749

Pre79time 0.037706 0.004399 8.572283 0.000000 0.028829 0.046582

Speeddum -0.115843 0.022924 -5.053436 0.000009 -0.162105 -0.069581

Petrol -0.429742 0.077112 -5.572983 0.000002 -0.585360 -0.274125

Figure 4.163 shows the pattern of the fatality rate is fairly accurately predicted by the model.

Figure 4.163 Actual/predicted German fatality rate

Fatalities per billion safety-wgt vkt

2029 2027 2025 2023 2021 2019 2017 2015 2013 2011 2009 2007 2005 2003 2001 1999 1997 1995 1993 1991 1989 1987 1985 1983 1981 1979 1977 1975 1973 1971 1969 1967 1965 0 10 20 30 40 50 60 70 80 90

Figure 4.164 shows the components of the German fatality prediction/forecast. Major influences are the increase in seat belt wearing and the ‘learning’ effect prior to 1979. From the mid-1990s, alcohol and speed controls start to have an effect.

Figure 4.164 Components of the German fatality rate prediction

0

Fatalities per billion safety-wgt vkt

Impact of seatbelts + alcohol + petrol price + speed Impact of seatbelts + alcohol + petrol price

Fatality rate Impact of seatbelts Impact of seatbelts + alcohol Impact of seatbelts + alcohol + petrol price + speed + dummies

2029

Figure 4.165 shows that the modeling also produces a fairly accurate prediction of the level of fatalities over time.

Figure 4.165 Actual/predicted German road deaths

2029

Actual Predicted

0

Chapter 4 • Modelling road safety in twenty-one countries The safety weights applied in x were cars, light commercial vehicles and other vehicles 1.0, motorcycles 10.0, buses 1.5, trucks 2.0 and mopeds 5.0. The injury weighted vkt calculation has motorcycles at 7.0 and mopeds at 3.5.

In Germany, the injury rate (road injuries per billion injury-weighted vkt) moved roughly in sync with the fatality rate over the period. This is shown in Figure 4.166.

Figure 4.166 German injury and fatality rates

2011 2009 2007 2005 2003 2001 1999 1997 1995 1993 1991 1989 1987 1985 1983 1981 1979 1977 1975 1973 1971 1969 1967 1965 310 810 1310 1810 2310

Fatalities per billion safety-wgt vkt

Injury rate Fatality rate

Injuries per billion injury-wgt vkt

0 10 20 30 40 50 60 70 80 90

A model of the injury rate was constructed using the fatality rate, time trends and a dummy variable. The regression results are shown in Table 4.54.

Table 4.54 Regression results for predicting the German injury rate

*Regression Statistics*
Multiple R 0.99675371

R Square 0.993517959

Adjusted R Square 0.992900622 Standard Error 36.26263056

Observations 47

ANOVA

*df* *SS* *MS* *F* *Significance F*

Regression 4 8465095.400 2116273.850 1609.360 0.000

Residual 42 55229.092 1314.978

Total 46 8520324.492

*Coefficients Standard Error* *t Stat* *P-value* *Lower 95%* *Upper 95%*

Intercept 417.65914 29.59346 14.11322 1.54E-17 357.93711 477.38116

FatalRate 20.00221 1.45506 13.74665 3.83E-17 17.06578 22.93864

Pre82time -6.00430 5.05837 -1.18700 0.24189677 -16.21251 4.20390

Dum99on 46.07058 27.16758 1.69579 0.09732697 -8.75581 100.89697

Time9910 -9.68380 2.86109 -3.38466 0.00155487 -15.45770 -3.90989

Figure 4.167 shows the pattern of the injury rate is fairly accurately predicted by the model.

Figure 4.168 shows that the modeling also produces a fairly accurate prediction of the level of injuries over time.

Figure 4.167 Actual and predicted German injury rate

0

Actual Predicted

Injuries per billion injury-wgt vkt

Figure 4.168 Actual and predicted German road injuries

2029

Actual Predicted

0 100 000 200 000 300 000 400 000 500 000 600 000 700 000

Chapter 4 • Modelling road safety in twenty-one countries