The purpose of this research, which was conducted between August 2014 and January 2015, was to provide guidance on approaches for valuing injury and mortality risks distilled from international literature on theory and practice. It was also to consider whether the current approach to estimating the value of statistical life (VOSL) in transport evaluations in New Zealand, established in 1991, is still appropriate, and also whether and how to fill gaps in current knowledge, regarding: • the relative value of safety and other non-market impacts of transport • whether VOSL should vary with age and life expectancy of those at risk • whether VOSL in transport should be aligned to value of safety in other risk contexts. The research focused on the human or welfare cost of injuries and fatalities, not the financial costs such as medical expenses and lost productivity. These items are included in transport appraisals but are out of scope for this report. Their value is dwarfed by the human costs of fatal and non-fatal injuries, reflecting the pain, suffering and aversion to the risk of such injuries. The research benefited from a number of meta-analyses of valuing injury risks conducted by the Organisation for Economic Co-operation and Development (OECD) and elsewhere. New Zealand’s transport VOSL values fatalities prevented according to the population’s mean willingness to pay (WTP) for small reductions in the risk of fatality (ie the marginal rates of substitution of wealth for risk of death). The literature review indicates this approach is appropriate, and is even called ‘the standard method’ in some publications. New Zealand was one of the first countries to adopt this approach. Since then, the techniques of nonmarket valuation used to infer values of risk changes have advanced considerably and now offer scope to value multiple facets of risk and other factors in evaluation of investment choices. The literature indicates that key determinants of the VOSL are the level of income of the affected population and the starting level of risk. There is international evidence that WTP to reduce involuntary risks is greater than for those which people assume voluntarily and exercise some control over. Since its establishment in 1991, New Zealand’s VOSL has been updated with reference to the ordinary wage rate as a proxy for changes in income. However, over the same period the number of road deaths in New Zealand has reduced from about 19 per 100,000 population to less than 7 in 2014, changing the underlying risk. Since the income and starting level of risk have both changed over time, the value determined in 1991 no longer reflects the current public value of reducing risk. On the basis of this review, a new study to update the VOSL estimate for current conditions in New Zealand would be timely. This would present opportunities for extending the survey to obtain new information with uses in both transport and other policy applications. However, the recommendations made here are primarily from the perspective of values of use for transport assessments. The findings on the three research questions can be summarised as follows. 1 Should valuations of non-market impacts within the transport sector be aligned? The relative value of VOSL with other non-market values in transport analysis is most apparent in the literature with respect to the value of non-fatal injuries, the value of non-acute risks of long-term health impairment or premature death (environmental health risk) and the value of travel time. Non-fatal injuries in New Zealand are valued at a fraction of the VOSL, at 10% for serious injury involving hospitalisation and 0.4% for minor injuries. This fraction was determined in a previous study in New Zealand and corroborated with a similar fraction found in two studies overseas. The evidence base is thin but consistent, and an updated VOSL could confirm those fractions and explore the feasibility of broadening the range of non-fatal injury categories beyond two. Non-acute risks of morbidity and mortality have arisen overseas particularly with respect to transport’s contribution to emissions and air quality deterioration which aggravates ill-health of those with respiratory conditions and can also shorten lifespans. This gives rise to the need for values to account for risk of premature mortality at some time in the future, and also for impaired living in the interim years. Overseas stated preference studies have shown how this can be done to establish relative values of VOSL for acute risk of immediate fatality and latent risk of long-term life years lost, although few countries have yet put this into practice in their transport appraisals. It could be worthwhile considering estimating such relativities for New Zealand, if data are sufficient to establish the links between emissions, air quality, ill-health and fatalities. The relative values of fatality risk and time have received relatively little attention in the international literature. As base risk and income are the principal drivers of VOSL, and the value of time varies with user type and trip characteristics, it is uncertain whether linking VOSL and value of time through the same survey vehicle will guarantee the production of consistent relative values of the two impacts. Further research in this area is therefore required to understand whether such an approach is possible. In addition, there are current issues in the valuation of travel time that need to be resolved independently of revising the VOSL. 2 Should VOSL vary with age and life expectancy of those at risk? The literature is ambiguous over whether and how VOSL varies with age or should be adjusted for life expectancy. Some suggests VOSL rises to middle age and then declines, but others suggest individuals’ risk aversion and WTP for safety increases in older age cohorts, reflecting their greater accumulated wealth and fewer expected years in which to use it. For many ex ante transport assessments that affect risks and save lives of anonymous individuals across a network or system, age-related VOSL is irrelevant and a single VOSL suffices. However, for choices between age-cohortspecific policies (such as child restraints versus older driver policies) it would be useful for VOSL to reflect the public’s preferences for reducing risks to different age groups. The literature suggests child-specific policies may require a separate VOSL that is unlikely to be less than that for adults, which may reflect consideration of life expectancy or simply affection for children. A directly estimated value of a life year has potential application to policies that affect the risks of long-term chronic health conditions and foreshortened future lifespans, such as decisions on emissions or the promotion of active modes like cycling and walking that can raise risks of accidental injury while reducing long-term health risks. Such a value would replace the current practice of calculating a value per life year as the annualised equivalent of the VOSL, as if VOSL is the net present value of a stream of future life years. The VOSL is not currently calculated as a net present value and the annualised value implies a constant value over time which is probably incorrect. Directly estimated values per life year are therefore preferable. This recommendation attaches lesser priority to putting a dollar value on quality weighted constructs such as quality adjusted life year (QALY) or disability adjusted life year (DALY), as these are not created equal, and reflect a wide variety of health states that would be difficult to apply in transport assessments. Where health outcomes are easier to predict, as in choices for medical interventions, there may be more merit in valuing QALYs or DALYs, although the implicit assumption that all QALYs and DALYs are the same when they are in fact products of varying longevity and health status is challenging for valuation and even where monetary value is not used, as in cost-effectiveness analysis. Empirical literature shows that gains in longevity are valued higher than gains in life quality, so valuing longevity without complications of quality adjustment would be useful. It could also be useful to estimate value per life year at varying levels of life quality (perhaps with broad categorisation) to account for those conditions that affect life quality strongly with relatively low change in longevity. 3 Should VOSL be aligned between injury sectors in New Zealand? Because VOSL is affected by base risk and income of the affected population, a VOSL is not likely to be the same in transport and other risk domains, such as workplace safety or health interventions. Ideally therefore separate VOSLs should be determined for each domain reflecting its particular characteristics of risk. However, a transport VOSL could provide an anchor or benchmark for these other risk areas, with appropriate adjustment if the value relativity between different risks is appropriately estimated. Because VOSL is estimated for road transport, with a perception of individual control associated with a lower VOSL than in cases of involuntary risk, the transport VOSL is unlikely to overstate the societal value of averting other public safety risks and can be viewed as a lower bound for value in those other settings. The method chosen to estimate the new transport VOSL would be a stated preference valuation study. It could be either a contingent valuation survey (as in 1989/90 and in 1998/99) to estimate public WTP for specific scenarios of risk reduction, or choice modelling in which relative preferences for different travel attributes are estimated. Benefit transfer, or the use of existing studies from elsewhere applied to New Zealand situations, is not ideal, as it depends on the population preferences in one country and context being closely similar to those in the other. Whichever method is chosen, it should be tailored to the information needed to inform practical policy and project appraisal. This would be done in developing the survey questionnaire, once the scope of the survey and the policy issues it needs to serve have been determined. Estimation of attribute values will be useful only where the data on these attributes are available or will be available in the future. In brief the recommendations of this report are: 1 New Zealand’s approach to valuing fatal and non-fatal injuries in transport in terms of public WTP for small changes in risk is appropriate and widely used across OECD countries. 2 A principal determinant of VOSL is baseline risk which has changed since the New Zealand VOSL was established in 1991, so an update is now overdue. 3 As VOSL varies with risk characteristics, in principle each area of public policy on injury risk (eg transport, workplace safety, health services, leisure) should have its own customised VOSL, but as a default the transport VOSL can be used in these areas, perhaps adjusted by a mark-up (or down) informed by the relativities of different VOSLs in analogous contexts from overseas or surveys. 4 The practice of valuing non-fatal injury as a fraction of the VOSL (as in New Zealand transport appraisals) is also appropriate, albeit based on thin empirical literature which it would be useful to reestimate when VOSL is updated. 5 The literature suggests little reason to expect fixed relativities between values of safety and other nonmarket effects of transport; the VOSL and the values of time have both been updated in line with wages so their relativity will not have changed over time. However, further research in this area is required to understand whether linking VOSL and value of time through the same survey vehicle will guarantee the production of consistent relative values of the two impacts. 6 Literature is ambiguous on whether and how WTP-based VOSL varies with the age of the affected population but it does show differences in value of reducing acute risks of immediate injury and latent risks of long term impacts on life expectancy, so in updating VOSL consideration should be given to direct valuation of these different types of risk to calculate new VOSL and value of statistical life year (VSLY), and to identification of whether VSLY varies with different broad levels of quality of life arising from an injury. 7 A stated preference valuation method is preferable for New Zealand because of the flexibility it offers in testing risk scenarios and less formidable data requirements than the principal alternative of revealed preference methods, but further detailed consideration should be given to stated preference design and the choice between contingent valuation and choice modelling approaches. Time, environmental health impacts and non-fatal injury could all be valued by stated preference study, but not necessarily the same study which could be overloaded by multiple questions. 8 In updating VOSL, all extensions to valuing life years or different risk contexts are dependent on data available being sufficient to build credible risk scenarios and on there being practical policy uses for the values obtained. (Author/publisher)
Abstract