Estimating the Willingness‐to‐Pay for Road Safety Improvements

Abstract

The value of road safety is the fundamental input in social cost–benefit analysis of road safety schemes. It is also an increasingly important input in the social evaluation of almost any transport infrastructure project. This value is given by the amount that people are willing to pay for reducing the risk of a becoming a fatal victim or of suffering a serious injury. Traditionally, road safety willingness‐to‐pay has been estimated by means of contingent valuation and other surveys without making explicit reference to a particular travel demand context. The paper advocates the use of stated choice techniques that allow one to recreate the context of a particular trip customized to the respondent’s past experience. For this and other reasons, it is argued that the proposed method is clearly superior to previous methods for estimating people’s willingness‐to‐pay for improved road safety. The paper also provides a summary of the Chilean experience on road safety valuation using stated choice techniques; and it concludes by showing the importance of conducting local studies to elicit people’s willingness to pay for safety.

Acknowledgements

The authors acknowledge the financial support of the Chilean Fund for Scientific and Technological Research (FONDECYT) given to this research through several projects, in particular Project No. 1050672.

Notes

1. The exception could be a safety scheme funded by The World Bank, where some sort of cost–benefit analysis is usually required.

2. Probably a unique experience within the context of developing countries; even more, the social cost–benefit analysis applied in the case of urban transport schemes ranks among the best worldwide.

3. Some of these studies posed a risk–risk trade‐off. However, to arrive at a monetary value, a risk–money trade‐off is necessary sooner or later.

4. Actually, a population is a stock variable, whereas a flow is not.

5.

6. To the authors’ knowledge, the assumption of zero cov(•,•) is more a belief than a proven fact.

7. A statistical death reduction means saving one life, on average, per unit of time (whose life is saved is unknown).

8. A referee suggested incorporating δr as a design variable. However, the problem lies in how to control for the way the individual perceives his/her own risk reduction. Although one can ask different individuals to value different reductions in the number of road accidents, they still process risk probabilities in a way that cannot be controlled for.

9. One referee noted that the two approaches are mutually consistent only when the respondent, when evaluating the number of crashes, has the correct aggregate flow in mind (i.e. s/he would value an extra fatal crash per year differently if s/he were to make the only trip on that road that year, than when millions of trips would be made on that road). In this sense, although a formulation in terms of number of crashes may sound more natural and easy to understand than a formulation in terms of probabilities to most respondents, the cognitive burden may not become any lighter. Unfortunately, whether the yearly flow indeed affects the valuation of an additional crash cannot be tested with the data available for this paper.

10. This section does not discuss how to estimate VRR by means of CV within a road transport context. There is a considerably literature on the issue and readers are referred to the references to Jones‐Lee cited in this paper.

11. All level‐of‐service parameters can be tested for systematic variations in this form.

12. By conventional CV survey, we refer to most CV surveys undertaken to elicit WTP for road safety in the 1980s and 1990s. These include, mainly, the pioneering work of Jones‐Lee et al. and all the surveys described by Schwab and Soguel (1995).

13. The shortcomings of stated preference techniques will not be addressed, since they are common to both CV and SC.

14. These could also be introduced within the CV approach.

15. This may lead someone to think that there could be an inconsistency in (4). As it is a sum of WTP, the more people travel on the route, the higher the WTP for reducing one fatal crash. However, if one is to be consistent with the model in mind, as the number of individuals travelling a route increases, the individual subjective risk reduction brought about by averting one fatal death should decrease. If safety is thought of as any other good in the economy, its value should be related to both the initial level of safety and the magnitude of the safety improvement. Thus, if the magnitude of the improvement diminishes, so does the WTP for it. Depending on this magnitude and the initial level of risk, the value of (4) could be higher or less for a route used by a higher flow and the apparent inconsistency is solved.

16. One of the pilot surveys conducted by Rizzi et al. (1999) gave rise to a paper by Jara‐Díaz et al. (2000). We will not comment on this work since the quality of the data used by Rizzi and Ortúzar (2003) was clearly superior.

17. The referee also noted that the effect of route choices on road safety was ignored.

18. The relevant flow figure for this table was almost 4 million light‐vehicles per year. In a experiment that followed the work described here (Hojman et al., 2005), the vehicle flows were also included in the survey to provide respondents with more information.

19. The UF (Unidad de Fomento) index is elaborated by the Chilean Central Bank and is used for indexation (http://www.bcentral.cl).

20. The exchange rate is also available at the Central Bank site.

21. These findings seem to be at odds with a previous pilot study conducted by Galilea et al. (2000) where a much higher VRR was elicited for Route 5 (mean estimate around US$1.5 million and confidence interval from US$0.9 million to US$3.4 million). However, the sample size of this pilot study was much smaller with only 90 respondents. Notwithstanding, the superior quality of the survey by Hojman et al. (2005) is out of the question.

22. They state, among other reasons: (1) the public good nature of the risk under analysis, and (2) the definition of the payment vehicle. However, they do not give any explanation about whether the survey instrument could affect the outcome of the experiment.

23. The adjustment was made with the inflation calculator at http://www.westegg.com/inflation.

24. Around US$35 000 only at current prices (CITRA, 1996).