The Value of Improved Road Safety

The Value of Improved Road Safety


L Hultkrantz, VTI and Örebro University, SE; G Lindberg, VTI, SE; C Andersson, Umeå University, SE


We estimate the values of private or public urban road safety measures. Our lower-bound estimates are robust to scope (scale) and hypothetical bias, but exceed the currently used CBA values.


The value of statistical life (VSL) is of major importance to cost-benefit assessment of road infrastructure investments, road maintenance planning, and to traffic control decisions, such as limitation of speed. It seems however to be of a ghostly nature that escapes precise empirical measurement. A major problem blurring the image catched by various preference-revelation instruments is the size bias, i.e. that measures of willingness-to-pay (WTP) for road safety are unreasonably insensitive to the size of the risk reduction. When the WTP for a larger risk reduction is almost the same as for a smaller risk reduction, the VSL will be inversely proportional to the size of the risk reduction. Therefore the WTP per unit of risk reduction, which is the VSL, can be more or arbitrarily set to any number within a wide range, i.e., it will be high for a small risk reduction and low for a large reduction. While similar problems have been encountered for other benefits as well, they seem to be more difficult to overcome for safety improvements that result in small size probability reductions (Carson et al. 2001). A related problem is the collinearity of risks of accidents with severe and less severe consequences, respectively. This makes it difficult to estimate separate values per risk unit of fatalities and non-fatal injuries both from revealed and stated preference data (Viscusi 2003).

Several other problems add to the obscurity of VSL estimates. The bulk of studies are based on stated-preference methods that generally are plagued by hypothetical bias, leading to exaggeration of WTP, as responses reflect attitude or intention (?yes, maybe?), rather than real commitment (?yes, sure?). Another source of upward bias is that enhanced safety often is framed as a private good (like a safety cushion), although the VSL estimates will be used to evaluate actions performed by the public sector (like road improvements).

There still is hope that better preference-elicitation methods will develop, although some researchers have suggested that the core problem is in cognitive constraints that cannot be circumvented by improved instruments. In this paper, however, we report the results of a contingent valuation study in Sweden that takes scale bias as a fact. Instead of targeting a central-value estimate of VSL, we attempt to find a lower-bound estimate of what we call the value of a serious statistical accident, VSSA. This is based on a conservative assessment of the WTP for a risk reduction eliminating fatal and serious-injury accidents . We search for values from respondents reporting high confidence in their answers, and estimate values within both private and public good contexts. Dividing this value with the whole baseline risk of fatal and serious-injury accidents gives the lower-bound VSSA estimate. Based on the results corresponding to a public safety program, we conclude that an upward revision of the combined unit value of fatalities and serious injuries used in Swedish infrastructure planning is warranted. Our estimates do indeed imply that the willingness to pay of fully confident respondents is proportional to the size of the risk reduction, but this result is not statistically significant.

The result for the private good is close to the VSL suggested by previous studies. However, unlike these assessments this is a lower-bound estimate and the scope of VSSA is wider, as it applies also to serious injuries. As infrastructure planning is made in the public-good context, we recommend the public-good value estimate for use in CBA models. Within this context we also estimate the effect of a provision condition intended to reduce free-riding strategic response from 'conditional cooperators'.

For road accidents in Sweden in general, and for road accidents in the urban areas of the city in which this study was made, there are approximately seven times as many reported seriously injured victims as the number of persons killed. The ?Vision Zero? public-safety program for the city that was valued in this study was supposed to prevent two fatalities and fourteen serious injuries per year. The annual social benefit of such a program is according to the presently used CBA-models 76 million SEK. With the lower-bound public-program VSSA estimate of our study, the value of the corresponding ?bundle? is 119 million SEK. Thus, although this study does not give support for an increase of the currently used VSL value, it does imply that the benefits accruing from safety measures that target accidents with mortal or serious-injury outcomes on average are larger than the currently assessed values.


Association for European Transport