--- category: literaturenote citekey: wongeffectquantifiedroadsafety2010 title: Is the effect of quantified road safety targets sustainable? authors: "Wong, S.C.; Sze, N.N." year: 2010 date: 2010-00-00 2010 doi: 10.1016/j.ssci.2009.12.020 publication: Safety Science url: "https://doi.org/10.1016/j.ssci.2009.12.020" zotero_key: WNXEHCUZ zotero_storage: DS6F8H6M collections: transpordi_planeerimine folder: 001_artiklid/Road Safety Indicators firstAuthor: "Wong, S.C." status: converted --- ![](_page_0_Picture_1.jpeg) Contents lists available at [ScienceDirect](http://www.sciencedirect.com/science/journal/09257535) # Safety Science journal homepage: [www.elsevier.com/locate/ssci](http://www.elsevier.com/locate/ssci) ![](_page_0_Picture_5.jpeg) ## Is the effect of quantified road safety targets sustainable? S.C. Wong a,\*, N.N. Sze b - aDepartment of Civil Engineering, The University of Hong Kong, Pokfulam Road, Hong Kong - bDepartment of Transport & Planning, Faculty of Civil Engineering & Geosciences, Delft University of Technology, Delft, The Netherlands #### article info #### Article history: Received 4 August 2009 Received in revised form 22 October 2009 Accepted 26 December 2009 Keywords: Quantified road safety target Organisation for Economic Co-operation and Development Before-and-after evaluation Long-term effect Road fatality Treatment-comparison group approach #### abstract Recent studies have shown that the setting of road safety targets is associated with a substantial reduction in road fatalities in the short-term. Although such targets may not themselves be responsible for the reduction in fatalities, they serve as a useful measure of the intention and commitment of road authorities to formulate timely road safety measures that lead to the achievement of the target. A quantified target is regarded as one of the key components of a road safety strategy. However, few studies have examined the degree of commitment and attention of road authorities to such targets in the long-term. In this study, we revisit the effectiveness of the quantified road safety targets set by the Organisation for Economic Co-operation and Development (OECD) countries, but with a different method, time scale, and group of comparison countries. We not only evaluate the associations between quantified targets and road fatality levels, but also measure the changes in the time-series trends in road fatalities over the long-term. The results indicate that all seven treatment countries analyzed experienced desirable changes in the time-series trend in road fatalities during the period under study, which implies an increase in road safety improvement over time. > -2009 Elsevier Ltd. All rights reserved. #### 1. Introduction Quantified road safety targets have been established by many Organisation for Economic Co-operation and Development (OECD) countries to motivate and monitor actions to reduce road accidents, injuries, and fatalities [\(OECD Scientific Expert Group, 1994,](#page-6-0) [2008\)](#page-6-0). Comprehensive road safety strategy frameworks have been formulated to stimulate the development of objective-related, cost-effective, and practical measures that contribute to target achievement ([Elvik, 1993; Allsop, 2000, 2009; ETSC, 2001, 2003a\)](#page-5-0). However, a road safety target needs to be both achievable and challenging. Targets that are too ambitious can de-motivate, whereas those that can be achieved without a high level of implementation of all of the relevant measures can induce complacency. Establishing an appropriate target that strikes a balance among ambition, capability, and social acceptability relies on the efficient assessment of traffic exposure and risk levels and the contribution of each road safety measure [\(ETSC, 2003b](#page-6-0)). To measure the degree to which a quantified road safety target has been realized is challenging, and only a few studies have estimated the effectiveness of national road safety targets in a rigorous manner [\(Elvik, 2001; Elvik and Vaa, 2004; Wong et al., 2006](#page-6-0)). [Elvik](#page-6-0) [\(2001\)](#page-6-0) deployed a treatment-comparison group approach to evaluate the effectiveness of setting regional and national road safety targets in which the effects of confounding factors were controlled. The choice of appropriate comparison countries determines the reliability of an analysis. [Elvik's \(2001\)](#page-6-0) study relied on the researcher's expertise and knowledge to select a single comparison country. [Wong et al. \(2006\)](#page-6-0) proposed a systematic approach to the selection of a group of comparison countries based primarily on historical road fatality trends. Their study suggested that the creation of quantified road safety targets in 14 OECD countries during the 1981–1999 period was associated with significant fatality reductions over a 3-year period, with an overall fatality reduction of 17.4%. However, whether the association between quantified road safety targets and fatality reduction persists in the long-term is unknown. This motivates us to revisit the subject using a different method, time scale, and group of comparison countries. All OECD member countries are anticipated to be candidates for comparison groups for the evaluation of the effectiveness of targets set in seven OECD countries, and a regression approach is used for the qualification testing and association measurement. We measure the changes in both the average fatality levels and the time-series trends in road fatalities in the long-term. ## 2. Data This study investigates the sustainability of the effects of national quantified road safety targets – identified through a \* Corresponding author. Tel.: +852 2859 1964; fax: +852 2559 5337. E-mail addresses: [hhecwsc@hkucc.hku.hk](mailto:hhecwsc@hkucc.hku.hk) (S.C. Wong), [nnsze@graduate.hku.hk](mailto:nnsze@graduate.hku.hk) (N.N. Sze). number of deliverables and research reports – set in all OECD countries during the period 1970–2000 [\(Elvik, 2001; Elvik and](#page-6-0) [Vaa, 2004; OECD Scientific Expert Group, 2008\)](#page-6-0). Table 1 presents the years in which quantified road safety targets were set (and the corresponding target year) by the national governments of OECD countries. Fourteen OECD countries first established quantified road safety targets during the 1973–1997 period, whereas 16 OECD countries (Austria, Belgium, Canada, Czech Republic, Germany, Greece, Ireland, Italy, Japan, Korea, Luxemburg, Mexico, Portugal, Slovak Republic, Switzerland, and Turkey) did not set a target or set only one target after 2000. The fatality data used in this study are taken from World Road Statistics (WRS) and the International Road Traffic and Accident Database (IRTAD) for the period 1963–2006 ([IRF, 2008; IRTAD,](#page-6-0) [2008](#page-6-0)). It should be noted that not all OECD countries have adopted the standard definition of road fatalities given by the United Nations (as shown in Appendix I), and thus a correction factor has been applied to adapt these data to a 30-day definition for reliable comparisons across countries and periods ([IRTAD, 1998](#page-6-0)). Iceland Table 1 Setting of national road safety targets in all OECD member countries (in chronological order). | Country | Notation | Target
set | Target
year | Subsequent target
set | |-------------------|----------|---------------|--------------------------|--------------------------| | Finland | FI | 1973 | 1979 | 1977 | | Netherlands | NL | 1985 | 2000 | 1986 | | United
Kingdom | GB | 1986 | 2000 | 1999 | | Norway | NO | 1987 | 1993 | | | Denmark | DK | 1988 | 2000 | 1998 | | Sweden | SE | 1989 | 2000 | 1996 | | New Zealand | NZ | 1990 | 2001 | 1999 | | Spain | ES | 1992 | 1999 | | | Australia | AU | 1993 | 2001 | 1997 | | Hungary | HU | 1993 | 2000 | | | Poland | PL | 1996 | 2001 | | | United States | US | 1996 | 2008 | | | France | FR | 1997 | 2002 | | | Iceland | IS | 1997 | 2000 | | | Austria | AT | | No target set until 2001 | | | Canada | CA | | | | | Czech Republic | CZ | | | | | Greece | GR | | | | | Ireland | IE | | | | | Italy | IT | | | | | Luxemburg | LU | | | | | Portugal | PT | | | | | Slovak Republic | SK | | | | | Switzerland | CH | | | | | Belgium | BE | | | | | Germany | DE | | | | | Japan | JP | | | | | Korea | KR | | | | | Mexico | MX | | | | | Turkey | TR | | | | and Luxemburg have rather low annual fatality numbers, and are thus removed from subsequent analyses because of the possible bias associated with the instability of low fatality rates [\(Elvik,](#page-6-0) [2001](#page-6-0)). As certain initiatives are usually put in place when a quantified safety target is first announced by a national government, the road safety performance in the year in which a target is set is subject to that target. Thus, the year preceding the implementation of the quantified safety target is set as the reference year. We employ the treatment-comparison group approach for the effectiveness evaluation ([Hauer, 1997; Wong et al., 2005, 2006](#page-6-0)), using all OECD member countries (listed in Table 1) as the comparison group. Although certain countries (e.g., Belgium and Germany) did not establish quantified safety targets in the study period, the OECD countries share similar characteristics in terms of level of economic development, commitment to road safety achievement, development of a road management system, and establishment of a robust strategy framework. This provides a reliable base for the comparison group. To ensure that the measures are independent, for each treatment country that set a target in a particular year, countries that did not set a target until the end of the corresponding after-period are selected as candidates for the comparison group. The target year is taken as the final year of the after-period. However, when a new target is set before the expiration of the existing target, the year preceding the setting of the new target is selected as the final year instead. For example, Norway set a road safety target for 1993 in 1987, and thus the reference year is set at 1986 and the after-period is 1987–1993. Hence, countries with no target set or that set a target after 1993 are selected as potential candidates for the comparison group for Norway. Denmark set a road safety target for 2000 in 1988, and so the reference year is set as 1987. However, a new target was set in 1998. The after-period for the target set in 1988 in Denmark would thus be 1988–1997. Hence, countries with no target or that set a target after 1997 are selected for analysis (as shown in Table 2). Because the timing of the target setting by the 16 OECD countries that set targets after 2000 is unclear, the treatment countries with an after-period up to 2000 and beyond (e.g., Hungary, 2000; Poland, 2001; United States, 2008; and France, 2002) may have no suitable comparison group for analysis. Consequently, only eight OECD countries with quantified road safety targets set during the 1973–1993 period are assessed. Table 2 presents the details of these countries. ## 3. Qualification test for the comparison groups We formulate the comparison groups using a rigorous qualification test to compare the historical fatality figures between a treatment country and a corresponding group of comparison countries. Let the number of fatalities in year y in the treatment country t be nty (with y = k, k + 1, k + 2, ... , -1, 0, 1, 2, ... , m, where k Table 2 Observation periods and groups of comparison countries for each of the eight treatment countries that set a target during the period 1973–1993. | Treatment country | Reference year | Before-period | After-period | Potential comparison countries | |-------------------|----------------|---------------|--------------|------------------------------------------------| | FI | 1972 | 1963–1971 | 1973–1976 | NL, GB, NO, DK, SE, NZ, ES, AU, HU, US, FR, Cˆ | | GB | 1985 | 1975–1984 | 1986–1998 | Cˆ | | NO | 1986 | 1976–1985 | 1987–1993 | PL, US, FR, Cˆ | | DK | 1987 | 1977–1986 | 1988–1997 | Cˆ | | SE | 1988 | 1978–1987 | 1989–1995 | FR, Cˆ | | NZ | 1989 | 1979–1988 | 1990–1998 | Cˆ | | ES | 1991 | 1981–1990 | 1992–1999 | Cˆ | | AU | 1992 | 1982–1991 | 1993–1996 | Cˆ | and m are equal to the length of the before- and after-periods, respectively) and the number of fatalities in the comparison country c be ncy . We first normalize the number of fatalities relative to the number in year 0 (y ¼ 0 denotes the reference year, or the year preceding that in which a quantified target is set) to obtain the scaling factors $$f_{ty} = n_{ty}/n_{t0}$$ $$f_{cy} = n_{cy}/n_{c0}, \quad c = 1, 2, ..., C_p,$$ where Cp is the number of potential comparison countries. For example, the before- and after-periods for Norway are 1976–1985 and 1987–1993 respectively (as shown in [Table 2](#page-1-0)). k is equal to 10 and m is equal to 7, and the reference year is set at 1986. Thus, the number of fatalities in 1986 is used as the base for the estimation of the scaling factors for Norway and the corresponding comparison countries for the period 1976–1993. We apply a regression approach to the qualification test to examine the appropriateness of a comparison country. The following equation represents the trend of the scaling factors $$ln f_i(y) = \alpha_i y, \text{ with } i = t, \forall c,$$ where ln [fi(0)] = 0 by default, as fi(0) = ni0/ni0 = 1, and ai represents the rate of change in the scaling factor for country i. The logarithmic transformation of the scaling factor avoids the occurrence of negative values of f. We now define a dummy variable to signify a comparison country c as dc. Let ac = at + bcdc, where bc represents the difference between the rates of change in the scaling factors of the comparison country c and the treatment country t. Substituting into the foregoing equation, we have $$ln f_i(y) = (\alpha_i + \beta_i \delta_i) y, \text{ with } i = t, \forall c,$$ where dt = 0. We then conduct a linear regression analysis with the dependent variable ln fi(y) and two independent variables y and diy for y = k, k + 1, k + 2, ... , -1, under the assumption of normally distributed residuals. We do not need the data point at y = 0, because it is automatically satisfied by default. In the regression equation, at approximates the rate of change in the scaling factor of the treatment country. If at is not statistically significant, then there is no statistical evidence to indicate a fatality trend in the years before a quantified target was set. bc ; c ¼ 1; 2; ... ; C denotes the difference in the rates of change between the treatment country and the comparison country c. If bc is statistically significant, then the fatality trend in country c deviated substantially from that of the treatment country in the years before the target was set. Hence, c is not selected at a 95% level of confidence. Table 3 shows the results of the qualification test. Among the countries that set targets at an earlier stage, 11 (magnitudes of coefficient range from 0.001 to 0.015) out of the 27 potential comparison countries for Finland pass the qualification test at the 95% confidence level, with the trend in the number of fatalities (coefficient = 0.024) being statistically significant at the 1% level. Four (Canada, coefficient = 0.014; Italy, 0.001; Portugal, -0.004; Belgium, -0.012) out of 15 potential comparison countries for the United Kingdom pass the qualification test at the 95% confidence Table 3 Results of the comparison group qualification tests (coefficient estimates). | Comparison country | Treatment country | | | | | | | | |--------------------|-------------------|-------------|------------|------------|-------------|------------|------------|-------------| | | FI | GB | NO | DK | SE | NZ | ES | AU | | NL | 0.016 | X | X | X | X | X | X | X | | GB |
0.018 | X | X | X | X | X | X | X | | NO | 0.004a
| X | X | X | X | X | X | X | | DK | 0.004a
| X | X | X | X | X | X | X | | SE |
0.032 | X | X | X | X | X | X | X | | NZ | 0.032 | X | X | X | X | X | X | X | | ES | 0.015a | X | X | X | X | X | X | X | | AU | 0.008a
| X | X | X | X | X | X | X | | HU | 0.035 | X | X | X | X | X | X | X | | PL | | X | 0.058
| X | X | X | X | X | | US | 0.007a
| X |
0.018 | X | X | X | X | X | | FR | 0.037 | X |
0.032 | X | 0.012a | X | X | X | | AT | 0.031 | 0.022
| 0.057
| 0.037
| 0.004a | 0.049
| 0.062
| 0.025 | | CA | 0.014a | 0.002a | 0.035
| 0.015a | 0.028a | 0.019a
| 0.064
| 0.041 | | CZ | 0.002a |
0.018 |
0.044 |
0.020 | 0.015a |
0.032 |
0.073 | 0.075 | | GR | 0.048 | 0.047 | 0.011a | 0.008a
| |
0.033 | 0.012a
| 0.098 | | IE | 0.063 | 0.020
| 0.064
| 0.021
| 0.006a | 0.042
| 0.055
| 0.051 | | IT | 0.001a
| 0.001a |
0.032 |
0.016 | 0.013a |
0.051 |
0.039 | 0.073 | | PT | 0.06 | 0.004a
|
0.031 | 0.016 | 0.053 | 0.009a
|
0.020 | 0.086 | | SK | – | – | – | – | – | – | – | – | | CH | 0.009a | 0.016
| – | 0.027
| 0.001a
| 0.054
| 0.073
| 0.042 | | BE | 0.051 | 0.012a
|
0.038 |
0.018 | 0.014a |
0.032 |
0.052 | 0.046 | | DE | 0.002a
|
0.043 |
0.068 |
0.060 |
0.024 |
0.072 |
0.026 | 0.081 | | JP | 0.008a
| 0.027 | 0.006a
| 0.015 | 0.053 | 0.008a | 0.017
| 0.095 | | KR | | 0.095 | 0.049 | 0.044 | 0.123 | 0.074 | 0.048 | 0.127 | | MX | – | – | – | – | – | 0.086 | 0.034 | 0.117 | | TR | 0.022 | 0.043 | 0.040 | 0.066 | 0.079 | 0.013a | 0.020
| 0.078 | | Selected | 11 | 4 | 2 | 2 | 8 | 4 | 1 | 0 | | a | 0.024** | 0.028**
| 0.011* |
0.008 | 0.031**
| 0.023** | 0.042** | 0.069**
| | Goodness of fit | 35.28** | 57.35** | 48.12** | 43.95** | 50.98** | 51.25** | 28.54** | 28.07** | X: not considered as a comparison group candidate. –: not applicable due to missing data. \* Statistically significant at the 5% level. \*\* Statistically significant at the 1% level. a Comparison group passed the test at the 95% confidence level. level, with a decreasing trend in the number of fatalities (-0.028) that is significant at the 1% level. Two (Greece, 0.011; Japan, 0.006) out of 18 potential comparison countries for Norway pass the qualification test at the 95% confidence level, with an increasing trend in the number of fatalities (0.011) that is significant at the 5% level. Among the countries that set targets at an intermediate stage, two (Canada, 0.015; Greece, -0.008) out of 14 potential comparison countries for Denmark pass the qualification test at the 95% confidence level, with no significant trend found in the number of fatalities. Eight (France, 0.012; Austria, 0.004; Canada, 0.028; Czech Republic, 0.015; Ireland, 0.006; Italy, 0.013; Switzerland, -0.001; Belgium, 0.014) potential comparison countries for Sweden pass the qualification test at the 95% confidence level, with a decreasing trend in the number of fatalities (-0.031) that is significant at the 1% level. Among the countries that set targets at a later stage, four comparison countries for New Zealand (Canada, -0.019; Portugal, 0.009; Japan, 0.008; Turkey, 0.013) pass the qualification test at the 95% confidence level, with an increasing trend in the number of fatalities (0.023) that is significant at the 1% level. One comparison country for Spain (Greece, -0.012) passes the qualification test at the 95% confidence level, with an increasing trend in the number of fatalities (0.042) that is significant at the 1% level. Finally, there is no valid comparison country for Australia, although the decreasing trend in the number of fatalities (-0.069) is found to be significant at the 1% level. The application of this rigorous qualification test assures the validity of the comparison groups for the evaluation of the effectiveness of the quantified road safety targets reported in the next section. ![](_page_3_Figure_6.jpeg) #### Table 4 Results of the association measures between target setting and road fatality levels. | Treatment country | Coefficient estimates | | Comparison group | Goodness-of-fit test | Previous studies | | |-------------------|-----------------------|--------------|--------------------------------------------|----------------------|-------------------|--------------------| | | a | a | | | Elvik (2001) | Wong et al. (2006) | | | | | | | Percentage change | | | FI | 0.016*** | 0.278***
| NO, DK, ES, AU, US, CA, CZ, IT, CH, DE, JP | 23.61*** | 0.3%
| N/Aa | | GB | 0.018***
|
0.066 | CA, IT, PT, BE | 54.98*** |
5.0% | N/Aa | | NO | 0.011*** | 0.319***
| GR, JP | 14.75*** | +6.2% |
4.7% | | DK | 0.006
| 0.140**
| CA, GR | 7.16*** | +16.5% | 37.7%***
| | SE | 0.025***
| 0.096*** | FR, AT, CA, CZ, IE, IT, CH, BE | 199.85*** | 24.8%***
| +6.1% | | NZ |
0.002 | 0.244***
| CA, PT, JP, TR | 8.68*** | 18.5%**
| 8.8%**
| | ES | 0.025*** | 0.515***
| GR | 9.18*** | 13.8%*
| 15.3%***
| a Extensive analysis of the first target set in FI and the target set in GB was not carried out in [Wong et al. \(2006\).](#page-6-0) ## 4. Association between target setting and fatalities ## 4.1. Effect on the number of annual fatalities For the selected countries c, we assume that bc = 0, that is, that the trends in fatalities between the treatment country (had a road safety target not been imposed) and the corresponding comparison group are homogeneous, as there is no statistically significant evidence that this is not the case. We then specify the following regression equation. $$ln f_i(y) = \alpha y + a g_i(y)$$ , with $i = t, \forall c$ . We denote the impact of setting a quantified target by gi(y), where gi(y) = 1 when i = t and y = 1, 2, 3, ... , m, and gi(y) = 0 otherwise. This is to measure the absolute change in the number of fatalities that is attributed to the existence of a quantified safety target. As shown in Fig. 1, a indicates the immediate effect of target setting on the number of fatalities, where - (i) a > 0 implies an absolute increase in the fatality level and - (ii) a < 0 implies an absolute decrease in the fatality level. The results of the effectiveness evaluations are presented in Table 4. Of the seven treatment countries, six demonstrate an average reduction in fatalities after target setting, of which the reduction in four (Finland, coefficient = -0.278; Norway, -0.319; Denmark, 0.140; Spain, -0.515) is statistically significant at the 1% level, and that in New Zealand (-0.244) is statistically significant at the 5% level. In contrast, the number of fatalities in Sweden displays a significant increase (coefficient = 0.096) at the 1% level. This is not the first time that the changes in number of fatalities associated with quantified road safety targets have been assessed. [Elvik \(2001\)](#page-6-0) attempted to address this issue using a single comparison country, while [Wong et al. \(2006\)](#page-6-0) examined the matter using a group of comparison countries that were selected from among the OECD member countries that had set targets, qualified through rigorous statistical testing. In this study, we revisit this interesting issue using a group of comparison countries selected from all OECD member countries, but employ the target year as the end of afterperiod and a regression approach to hypothesis testing. Table 4 presents the association estimates of this study and those of the aforementioned two earlier studies. The results of this study are generally comparable with those of [Elvik \(2001\)](#page-6-0). Of the seven treatment countries evaluated in this study, four demonstrate fatality changes that are consistent with those of Elvik's Fig. 1. Illustration of the immediate effect of target setting. model. Although the directions of the fatality changes in Norway \* Statistically significant at the 10% level. \*\* Statistically significant at the 5% level. \*\*\* Statistically significant at the 1% level. and Denmark differ between the current study and that of Elvik, the changes are not statistically significant in the latter model. Both the current study and that of [Wong et al. \(2006\)](#page-6-0) employ systematically selected and validated groups of comparison countries for the effectiveness evaluation. However, in this study we employ a longer after-period for observation. As shown in [Table](#page-3-0) [4](#page-3-0), all five treatment countries investigated in both models demonstrate consistent fatality changes. The results of the effectiveness evaluations in the three studies are consistent, despite the fact that the question is addressed here using a different method and time scale and possibly a different set of comparison countries. This result confirms the reliability and robustness of the association measures in the model used in this study. #### 4.2. The effect of target setting We also attempt to determine whether the effect of target setting on fatality reduction diminishes over time by using the following regression equation $$\ln f_i(y) = \alpha_i y + (a + by)g_i(y), \text{ with } i = t, \forall c,$$ where gi is a dummy variable that signifies target setting, as described in Section [4.1](#page-3-0). We assume that the impact of setting a quantified road safety target takes the linear form a + by, where a denotes a constant increase or decrease in the number of fatalities with respect to that in the reference year and b denotes the change in the trend in the number of fatalities over the years. Fig. 2 shows the possible impacts of target setting, where - (i) a > 0 and b > 0 implies an immediate increase in the fatality level, and that the increase is rising; - (ii) a > 0 and b < 0 implies an immediate increase in the fatality level, and that the increase is diminishing; - (iii) a < 0 and b > 0 implies an immediate decrease in the fatality level, and that the decrease is diminishing; - (iv) a < 0 and b < 0 implies an immediate decrease in the fatality level, and that the decrease is rising. For example, if the regression analysis gives a significant negative a and a positive b, then target setting had a significant immediate impact in reducing fatalities, but the effect diminishes substantially over time. Table 5 shows the results of the regression analysis for the absolute changes in number of fatalities, controlling for impacts on the time-series trend. Of the seven treatment countries, four demonstrate an initial reduction in fatalities after target setting, one of which (Spain, coefficient = -0.307) is significant at the 1% level. ![](_page_4_Figure_17.jpeg) Table 5 Results of the association measures between target setting and time-series fatality trends. | Treatment | Coefficient estimates | | | Comparison | Goodness | | |-----------|-----------------------|----------------------------|---------------------|---------------------|-------------|--| | country | a | Treatment effects,
g(y) | | group | of-fit test | | | | | a | b | | | | | FI | 0.016*** | 0.026 | 0.101
| NO, DK, ES, AU, | 16.63*** | | | | | | | US, CA, CZ, IT, CH, | | | | | | | | DE, JP | | | | GB | 0.017***
| 0.096 | 0.024**
| CA, IT, PT, BE | 40.11*** | | | NO | 0.012*** | 0.059 | 0.066*** GR, JP
| | 15.55*** | | | DK |
0.005 | 0.025 | 0.032*
| CA, GR | 5.98*** | | | SE | 0.025***
| 0.205*** | 0.028* | FR, AT, CA, CZ, IE, | 136.63*** | | | | | | | IT, CH, BE | | | | NZ | 0.001
| 0.039
| 0.042*
| CA, PT, JP, TR | 6.99*** | | | ES | 0.027*** | 0.307*** | 0.048** | GR | 8.58*** | | \* Statistically significant at the 10% level. In contrast, Sweden (coefficient = 0.205) demonstrates an initial increase, significant at the 1% level, in road fatalities after target setting. The findings of the changes in fatality rates are interesting. Six out of the seven treatment countries illustrate a stronger trend in fatality reduction (or a weaker trend in fatality increase) that is either statistically or marginally significant, regardless of the absolute decrease or increase in the number of fatalities after target setting. It is noteworthy that the aforementioned initial increasing trend in fatalities in Sweden diminishes (coefficient = -0.028) over the target period at the 10% level of significance. The United Kingdom (coefficient = -0.024), Norway (coefficient = -0.066), Denmark (coefficient = -0.032), and New Zealand (coefficient = -0.042) all demonstrate a remarkable tendency to anticipate a reduction in fatalities at the 5%, 1%, 10%, and 10% levels, respectively, although no significant absolute change in fatalities occurred after target setting in these countries. Finally, Spain (coefficient = -0.048) demonstrates a substantial absolute reduction in fatalities that is significant at the 5% level. In sum, setting a quantified road safety target is found to be associated with the desired change in the road fatality trend in the intermediate term by diminishing the rate of increase, changing an increase to a decrease, or yielding a greater reduction rate. However, a possible drawback of the treatment-comparison group before-after evaluation approach is that the potential effect of confounding variables on the number of fatalities in the afterperiod may not be perfectly controllable. For instance, a country may have been hit by economic recession immediately after the implementation of a target, which would affect the road safety performance and number of fatalities. Alternatively, a country may set an ambitious target because it has already for some time had a strong ambition to improve road safety, and the target setting does not change this fact. It would not be surprising if such a country continued to improve road safety more effectively than other countries. Apparently, these scenarios are above any effect of a quantified target, and may not be fully revealed by using a comparison group that is qualified using the before-period number of fatalities. ## 5. Discussion Our previous study [\(Wong et al., 2006](#page-6-0)) revealed a strong association between road safety targets and fatality reduction, with Fig. 2. Illustration of the impact of target setting on fatality trends. an overall fatality reduction of 17% three years after target setting \*\* Statistically significant at the 5% level. \*\*\* Statistically significant at the 1% level. in nine OECD countries. This study reinforces this finding, but goes onto examine the induction force of quantified safety targets on the trend in the number of fatalities in the long run. The results suggest that a stronger trend in fatality reduction (or a weaker trend in fatality increase) is associated with quantified road safety targets during the target periods in the OECD countries studied, which indicates an increase in road safety achievement over time. However, as the estimates of the fatality count may be sensitive to the length of the observation periods selected ([Partyka, 1984](#page-6-0)), this study tracks the effect of quantified safety targets on fatality counts across the entire period for which a target is in effect. One could argue that only an ambitious road safety target can stimulate the commitment of policymakers to achieve the target and thus result in a considerable reduction in fatalities. A recent study compared the performance of the ''Vision Zero" goal set by the Swedish government with that of other apparently less rigorous targets. The results indicated that such a goal is rational, motivating, and action guiding, and can achieve the desired level of road safety performance ([Rosencrantz et al., 2007](#page-6-0)). Our findings indicate that a more substantial fatality reduction is realized in the long run among most countries that set targets. However, to achieve the goal of sustainable, long-term road safety, more experience and effort needs to be accumulated during the process to ensure that the initial positive effect of target setting does not diminish. The target of a low number of fatalities or level of risk can become a spiritual substitute or philosophy, rather than a tool to motivate and monitor road safety actions or strategies. More modest and short-term targets may be more relevant for achieving long-term reductions in road fatalities. Setting up interim milestones that are associated with a long-term target, conducting periodic progress reviews, and developing sub-targets in particular areas or for different road entities are all necessary ([ETSC, 2003a,b;](#page-6-0) [Allsop, 2009](#page-6-0)). In addition, the development of different practical theories is essential to the establishment of an appropriate target or goal to enable the realization of the desired satisfaction and achievement levels [\(Locke and Latham, 2002\)](#page-6-0). Our findings suggest that the effect of road safety targets did not deteriorate in the periods under study, and was associated with favorable changes in fatality trends in the long run. Hence, there is no significant evidence that the targets that were set were irrational or unrealistic. Rather, the results indicate timely research and development before the setting of road safety targets by the national governments of the OECD countries. It must also be emphasized that a comprehensive strategy framework is critical to the success of a rational quantified road safety target ([OECD Scientific Expert Group, 1997; Elvik, 1999;](#page-6-0) [Loo et al., 2005, 2007](#page-6-0)). The Norwegian government has developed a comprehensive system of road safety management using a broad set of objectives to realize the national road safety target. Within this safety management framework, a set of sub-targets is put in place for various road safety indicators, followed by the setting of an overall target. This has formed the basis for the development of relevant cost-effective road safety measures. However, research reveals that although the target is supported by a well-conceived and seemingly attractive road safety management system, the overall target of a one-third reduction in fatalities set by the Norwegian government has yet to be achieved. A robust institutional framework and strong political foundations are critical to the success of the proposed road safety management system [\(Elvik, 2008\)](#page-6-0). Hence, more effort is required to pool resources to achieve a sustainable reduction in road fatalities, despite the existence of a cost-effective road safety planning system. The intermediate factors that might facilitate the realization of a sustainable reduction through target-oriented road safety measures are worth exploring. It is difficult to determine all of the possible factors, circumstances, and developments that were anticipated in a country from the setting of a quantified target. Fortunately, this benchmarking issue has been addressed by a comprehensive study at the European level [\(Wegman et al., 2008](#page-6-0)). ### 6. Conclusion In this study, a regression approach is applied to evaluate the effect of quantified road safety targets in the long-term. The results suggest that target setting in seven OECD countries is associated with a substantial positive effect on the time-series trend in road fatalities during the period in which the targets were in effect. Although the research gap regarding the sustainability of the effect of quantified road safety targets is addressed in this study, the interactions among possible influential factors, including the ambitiousness of the target, the existence of long-term commitments, the use of sub-targets, the presence of detailed road safety programs, and national income and expenditure on road safety programs, on the association between quantified road safety targets and fatality reduction merits further exploration. A propensity analysis identifying the factors that contribute to the likelihood that an appropriate road safety target and a successful road safety management system will be implemented to realize a target is also an interesting topic for future research. ## Acknowledgements We would like to thank two anonymous reviewers for their constructive and insightful comments on an earlier version of the paper. The work described in this paper was supported by grants from the Research Grants Council of the Hong Kong Special Administrative Region of China (Project No.: HKU7176/07E) and the Outstanding Researcher Award from the University of Hong Kong. #### Appendix I Correction factors for road fatalities applied in the IRTAD database to cope with the 30-day definition [\(IRTAD, 1998](#page-6-0)). | Country | Fatality definition | Correction factor | |----------------|----------------------|-------------------| | Austria | 3 days | 1965–1982: +15% | | | | 1983–1991: +12% | | Czech Republic | Until 1979: 24 h | Until 1979: +30% | | France | 1970–1992: 6 days | 1970–1992: +9% | | | | 1993–2000: +5.7% | | Greece | Until 1995: 3 days | Until 1995: +15% | | Hungary | Until 1975: 48 h | Until 1975: +20% | | Italy | 7 days | +8% | | Japan | Until 1992: 24 h | Until 1992: +30% | | Korea | 3 days | +15% | | Portugal | 24 h | +30% | | Spain | Until 1993: 24 h | Until 1993: +30% | | Switzerland | Until 1991: >30 days | Until 1991:
3% | | Turkey | 24 h | +30% | ## References Allsop, R.E., 2000. Strategies and targets for reducing death and injury in road traffic accidents. In: Reading in Injury Prevention and Control, 3rd National Conference on Injury Prevention and Control, pp. 6–9. Allsop, R.E., 2009. Britain's 11-year road safety strategy beyond midterm and in a European context. International Journal of Sustainable Transportation 3, 141– 159. Elvik, R., 1993. Quantified road safety targets: a useful tool for policy making? Accident Analysis and Prevention 25, 569–583. - Elvik, R., 1999. Can injury prevention efforts go too far? Reflections on some possible implications of Vision Zero for road accident fatalities. Accident Analysis and Prevention 31, 265–286. - Elvik, R., 2001. Quantified Road Safety Targets: An Assessment of Evaluation Methodology. TOI Report 539/2001, Institute of Transport Economics, Oslo. - Elvik, R., 2008. Road safety management by objectives: a critical analysis of the Norwegian approach. Accident Analysis and Prevention 40, 1115–1122. - Elvik, R., Vaa, T., 2004. The Handbook of Road Safety Measures. Elsevier Science Ltd., Oxford, UK. - European Transport Safety Council (ETSC), 2001. Transport Safety Performance Indicators. ETSC, Brussels. - European Transport Safety Council (ETSC), 2003a. Assessing Risk and Setting Targets in Transport Safety Programmes. ETSC, Brussels. - European Transport Safety Council (ETSC), 2003b. Transport Safety Performance in the EU – A Statistical Overview. ETSC, Brussels. - Hauer, E., 1997. Observational Before–After Studies in Road Safety. Elsevier Science Ltd., Oxford, UK. - International Road Federation (IRF), 2008. World Road Statistics. IRF, Geneva. International Road Traffic Accident Database (IRTAD), 1998. IRTAD Special Report: - Definitions and Data Availability. IRTAD, Germany. International Road Traffic Accident Database (IRTAD), 2008. International Road Traffic and Accident Data. IRTAD, Germany. - Locke, E.A., Latham, G.P., 2002. Building a practically useful theory of goal setting and task motivation. A 35-year odyssey. American Psychologist 57, 705–717. - Loo, B.P.Y., Hung, W.T., Lo, H.K., Wong, S.C., 2005. Road safety strategies: a comparative framework and case studies. Transport Reviews 25, 613–639. - Loo, B.P.Y., Wong, S.C., Hung, W.T., Lo, H.K., 2007. A review of the road safety strategy in Hong Kong. Journal of Advanced Transportation 41, 3–37. OECD Scientific Expert Group, 1994. Targeted Road Safety Programmes. OECD, Paris. - OECD Scientific Expert Group, 1997. Integrated Strategies for Safety and Environment. OECD, Paris. - OECD Scientific Expert Group, 2008. Towards Zero: Ambitious Road Safety Targets and the Safe System Approach. OECD, Paris. - Partyka, S.C., 1984. Simple models of fatality trends using employment and population data. Accident Analysis and Prevention 16, 211–222. - Rosencrantz, H., Edvardsson, K., Hansson, S.O., 2007. Vision Zero is it irrational? Transportation Research 41A, 559–567. - Wegman, F., Commandeur, J., Doveh, E., Eksler, V., Gitelman, V., Hakkert, S., Lynam, D., Oppe, S., 2008. SUNflowerNext: Towards a Composite Road Safety Performance Index. SWOV, Leidschendam, The Netherlands. - Wong, S.C., Sze, N.N., Lo, H.K., Hung, W.T., Loo, B.P.Y., 2005. Would relaxing speed limits aggravate safety? A case study of Hong Kong. Accident Analysis and Prevention 37, 377–388. - Wong, S.C., Sze, N.N., Yip, H.F., Loo, B.P.Y., Hung, W.T., Lo, H.K., 2006. Association between setting quantified road safety targets and road fatality reduction. Accident Analysis and Prevention 38, 997–1005.