SWOV homepage
Nederlands  

The value of statistical life in road safety

 

Accident costs are an important part of the costs of traffic and transport. An evaluation of accident costs can help in allocating road safety budgets more efficiently and rationally. The costs of fatal accidents are, however, difficult to calculate, as much of the damage done by fatalities is 'immaterial'. Still, it is important to assign monetary value to fatalities to be able to include this most harmful accident type in the evaluation of accident costs. Some valuation methods, estimated values and factors that affect these estimates are presented in this article.

 

Assigning monetary value to a fatality requires the economic valuation of a so-called statistical life. It is essential to be aware of the fact that not the life of a specific individual is to be valuated, but one life in a large population. The value of a statistical life (VOSL) in road safety is therefore the value of the saving of one life in a large population of road users, i.e. the value of a reduction in fatality risk.

 

Valuation

 

The VOSL can be estimated by determining how much people are willing to pay for a reduction in the fatality risk. For example, a large group of individuals are on average willing to pay 25 Euro for a risk reduction from 7 to 4 fatalities in a population of 100,000. Then the estimated VOSL would be 0.8 million Euro ([25/3]*100,000).

Thus, the willingness to pay (WTP) method is based on individual preferences. These preferences can be determined by stated or revealed preference methods. A stated preference method uses a questionnaire. In this, the respondents are asked how much they are willing to pay for certain goods (such as a reduction in the fatality risk). A revealed preference method estimates the value of the good by relating it to goods on the market. For example, the price of an airbag can be related to the risks on a fatal accident with and without an airbag. The value of the reduction in fatality risk can thus be calculated.

From the 1970's onwards, many studies have tried to estimate the VOSL in road safety using stated and revealed preference methods. Such studies were carried out in different countries and in different years, and have resulted in a wide range of estimates. The estimates of a statistical life value show vast differences, going all the way from 150,000 up to 30 million 1996 US dollars.

 

Factors affecting VOSL estimates

 

Researchers at the Vrije Universiteit Amsterdam studied literature on the economic valuation of statistical life in road safety. They collected the results from 25 different studies and analysed them using statistical methods. They used this 'meta-analysis' to determine whether there are factors that systematically affect the estimates of the value of life. Some of their results are presented in this article. The full results are reported in a discussion paper of the Tinbergen Institute, institute of economic research.

From the studies, 71 VOSL estimates in road safety were collected, together with the characteristics of these studies. On the basis of this database, a descriptive analysis and meta-analyses were carried out. The descriptive analysis confirmed that the VOSL estimates show large differences. A more intriguing result was that some of the highest stated preference estimates of the VOSL are over 1,000 times the Gross Domestic Product (GDP) per head of population of that country. This means that people would under certain conditions be willing to pay more for increased safety than their available budgets would ever allow !

Meta-analysis was carried out to determine whether the large differences in the 71 VOSL estimates could be explained by a number of 'standard explanations' from literature. Among these are, for example, the GDP per head of population and the fatal accident rates of the countries concerned, and in the year of data collection. Standard explanations also involve expectations that stated preference studies yield higher estimates than revealed preference studies. And studies that value safety as a private good are expected to present higher estimates than studies valuing safety as a public good. The result of this meta-analysis was that these, and some other 'standard' variables hardly have any, or no explanatory power. In other words: other explanations than the standard ones should explain the large variety found in VOSL.estimates.

Initial fatality risk

 

In stated preference studies, questionnaires present the fatality risk to the respondent, as well as the reduced fatality risk after an investment in road safety. Literature indicates that the VOSL is affected by this presented initial risk level and the reduction of this risk as the result of a safety improvement. It can be expected that the WTP increases with increasing size of the risk reduction. In other words: the more 'difference it makes', the more people are willing to pay. It can also be expected that the WTP for a given reduction in fatality risk increases with increasing level of the initial risk. In other words: the higher the initial fatality risk, the more money people are willing to pay for a given reduction in this fatality risk. Vice versa: the lower the initial risk level, the lower the WTP for a given reduction. The relation between the willingness to pay for a given reduction in fatality risk and the initial risk levels can be visualised in a graph.

 

The abovementioned theory has not been used in the existing literature to explain the large variance in VOSL estimates, presumably because the initial risk levels are already very small. It is typically assumed that in this low range of initial risks, the exact risk level does not affect anymore the amount of money people are willing to pay for a given risk reduction; this amount is constant. This can also be seen in the figure.

Nonetheless, the researchers at the Vrije Universiteit Amsterdam tested whether the initial risk levels and the risk decline in several studies could explain the variance in VOSL estimates. It can be concluded from this meta-analysis that the VOSL does indeed depend on the initial risk level and the risk decline. In other words: people really do take the risk levels mentioned in the questionnaires into account.

 

To conclude

 

It can be concluded that a statistical life does not have one, common value, not even from this theoretical, economical perspective. The VOSL will be different in different situations. The value depends, among other things, on the initial level of the fatality risk, and on the risk reduction considered. These variables have a far better explanatory power than more general background variables such as GDP per head of population.

In the Netherlands, the VOSL has until now been estimated by means of the 'human capital method'. The economic criticism on this method is that it is not based on individual preferences. An empirical follow-up of the research project will be conducted by researchers of the Vrije Universiteit Amsterdam and SWOV. The results are expected to be values of a statistical life for the Netherlands, which will be estimated by making use of individual preferences.

The full discussion paper (number 00-089/3) can be downloaded from www.tinbergen.nl

The willingness to pay for a given reduction in fatality risk as a function of the initial risk levels (in 1996 U.S. dollars, logarithmic scale).

 

SWOV Research Activities 16 - April 2001

Knowledge base