Outils et évaluation des politiques de l'environnement

Valuing mortality impacts


It is increasingly common to include estimates of value of statistical life (VSL) in analyses of proposed policies that affect people’s mortality risks. While such VSL estimates have often been derived using methods that, for example, compare wage differentials between risky and non-risky jobs, such methods may be inappropriate to assess the value of very different environmental, health and transport risks affecting the general population.

Environmental pollution, for example, typically affects the youngest or the oldest part of the population the most (rather than male workers in their prime years, whom wage risk studies are based on) and mortality results from long-term pollution exposure and exacerbation of pre-existing medical conditions (rather than accidental deaths in the workplace). The wage-risk studies also face the problems of separating between actual and perceived risks and other factors that cause variation in wages.

Therefore, a growing body of research use stated preference methods instead (contingent valuation or choice modelling), asking people directly or indirectly for their willingness to pay to reduce such risks. OECD has taken stock of this literature and done a meta-analysis of the VSL estimates; seeking to explain the variation in the estimates from differences in study designs (including the way risk changes are displayed), characteristics of risk (type and size of risk, baseline risks, latency, etc.), socio-economic characteristics (age, income, gender, health status, etc.) and other variables derived from the studies and from other available statistics.

The analysis is presented in the report Mortality Risk Valuation in Environment, Health and Transport Policies and provides an in-depth discussion of how VSL estimates best can be used in policy assessments. The report is an outcome of the PIMAVE project – Policy Implications of Meta-Analysis of Value-of-Statistical-Life Estimates.

Great emphasis was placed on sorting out the estimates that were most appropriate for inclusion in the meta-analysis. All the authors of the original studies that we managed to contact were invited to express their opinion on whether or not a particular estimate from their survey was suitable for inclusion in a meta-analysis – and a large number of authors kindly responded. In numerous cases, the authors indicated that (some of) their estimates were not suited for inclusion – for example because the survey had been carried out to test a particular methodological issue, not being designed to be representative of the general population.

This Excel file contains all the information extracted from the original studies for this project, and indications of which estimates the author(s) have indicated that should not be included in a meta-analysis. As can be seen from the Excel file, a number of other estimates were also excluded from the final meta-analysis, according to the following criteria:

  • number of observations behind a particular estimate VSL was lower than 100; and/or
  • number of observation in the survey overall was lower than 200; and/or
  • sample population was clearly not representative of the general population in the geographical area of the survey, e.g. because the sample only included persons with specific professions; and/or the sample population had an average income that clearly differed from the general population (e.g. if a sample in a low-income country only included car owners); and/or the sample population had an age composition that differed significantly from the average adult population in the geographical area of the survey.

In addition, surveys that didn’t provide information on the magnitude of the risk change that respondents were asked to place a value on were excluded from the meta-analysis. In the Excel file, the sheet called “Meta-data” shows the information extracted from the original surveys, and the sheet called “Stata-data” shows the input as it was included in the final meta-analysis, using version 11 of the “Stata” programme.

A preliminary version of the meta-analyses was presented in the document "Valuing Lives Saved From Environmental, Transport and Health Policies: A Meta-Analysis of Stated Preference Studies".

A large number of additional meta-regression models were tested, and many sensitivity analyses were undertaken. These anaylses were reported in documents written by Vincent Biausque, The Value of Statistical Life: A Meta-Analysis, and by Henrik Lindhjem and Ståle Navrud, Meta-analysis of stated preference VSL studies: Further model sensitivity and benefit transfer issues. Drawing on these analyses, a number of  implications for policy-makers were drawn out and a “user’s guide” for how best to include VSL estimates in cost-benefit analyses was prepared.

Additional data screening, meta-regressions and sensitivity analyses were done and are presented in the book Mortality Risk Valuation in Environment, Health and Transport Policies. Based on the additional analyses, further refinements to the advice on how to use VSL in policy assessments are also included in the book.




Working papers and similar documents


Additional information from the European Commission

 This project was financed with the financial assistance of the European Union. The views expressed herein and in the various reports can in no way be taken to reflect the official opinion of the European Union.


For more information, please contact Nils Axel Braathen.


Documents connexes


Also AvailableEgalement disponible(s)