Research and analysis

Annexe 2

Published 28 July 2020

RQII

What are the main methodological issues in deriving a Value of a Life Year (VOLY) and what approaches exist in literature for addressing these?

1. Introduction

1.1. Background and coverage of this chapter

In this research question (RQ), we will detail key methodological issues in deriving a VOLY and the potential approaches for addressing these factors. We defer the equivalent analysis of the issues relating to deriving a VOLY from Willingness-To-Pay for a Quality Adjusted Life-Year (WTP-QALY) values until RQIII. To establish the scope of this RQ, we distinguish between 2 types of methodological issues – the first relates to VOLY-specific concerns, specifically related to the of communication of life expectancy gains including risk communication, while the second relates to stated preference surveys in general. We will focus on the former but do acknowledge that certain of the more general concerns relating to the latter will impact directly on a VOLY valuation and so will also assess the evidence with respect to these particular factors^{[footnote 1]}. Whilst these factors have already been considered to some degree in RQI due to their fundamental influence on VOLY valuation we examine some studies in more detail to explore their implications. As this RQ draws primarily on the VOLY literature reviewed elsewhere in the report, no further literature search strategy was employed.

Thus, this RQ will be structured around 3 key themes. Firstly, we will examine the methodological challenges related to how the gain in life expectancy has been specified in surveys and the approaches taken to explain this to survey respondents. Specifically, we cover how life expectancy gains are described and communicated, including how the link between life expectancy gains and risk reductions are communicated. This is germane to describing the hypothetical market in which the monetary valuation task is carried out.

Secondly, whist all features of a stated preference survey have the potential to impact on the valuation of any public good^{[footnote 2]}, we consider 2 key aspects to be particularly influential in the specific case of valuing life expectancy gains: the payment vehicle^{[footnote 3]}, used to elicit the WTP (for example, tax, increase in cost of living) including the duration and timing of this payment (for example, ongoing, over a defined period of, say, 10 years or a one-off payment); and the different ways in which a survey may be administered (survey modes) that might be used to elicit the valuation.

Finally, we outline alternative valuation approaches. Whilst all VOLY valuation studies to date (see RQI) have elicited WTP from respondents, there are well-documented potential difficulties associated with directly eliciting WTP. A solution to this issue might be to consider an alternative methodology, a relative valuation approach. We will present the current rationale and evidence with respect to these approaches, including their main advantages and disadvantages.

The final section will make recommendation for how the issues identified above could be addressed in the specification of any new primary research. Based on the literature review, recommendations will also focus on establishing the feasibility of applying a WTP-based, direct valuation approach to the VOLY.

2. Specifying gains in life expectancy within the hypothetical market

2.1. Life Expectancy Gains

Generally, as noted in Chapter RQI in the direct valuation of a VOLY one of 3 approaches is adopted: either to ask respondents to directly value a specific life expectancy gain^{[footnote 4]}; to value a reduction in their risk of dying over an N year period; or a combination of the 2. A priori, if respondents had a good understanding of the relationship between risk reductions and the resulting change in life expectancy, the distribution of risk reductions is communicated to them and understood and preferences are known by the individual, it should not matter which approach is taken since the policy output generated is equivalent. Whilst it is not possible to quantitatively establish the impact that this difference has had, it is likely that the extremely wide range of estimates observed in RQI are at least in part explained by the different approaches taken in different studies.

It is also the case that practice has not remained stable across the period of VOLY studies. Early studies tended to specify only the duration of the life expectancy gain and, in fact, this approach continues to be used in some studies (Nielsen, 2010; Vlachokostas et al., 2010; Johannesson and Johansson, 1997; Chilton et al., 2004; Ara and Tekesin, 2017; Grisolia et al., 2018).

The – possibly advantageous – simplicity of the former approach, is illustrated by these quotes from Chilton et al. (2004)^{[footnote 5]}.

Chronic Mortality Scenario (p.5):

Faster ageing; Some chemical in the air may cause wear and tear on our bodies, so that people living in areas with more pollution may age faster and die younger than people in low pollution areas. Some experts think that the average person in Britain might lose about a month of life in this way. Others think that the average loss of life might be as much as a year

Chronic Mortality Valuation Question (p.6):

X MONTHS MORE LIFE IN NORMAL HEALTH: By reducing the general level of air pollution that causes wear and tear and faster ageing everyone could live longer. That would mean that you {and everyone else in your household} could expect to live about X months longer in your {their} normal state of health.

However, the problem with this approach is that such a change in life expectancy might be construed as a certain “add-on” at the end of life. This concern has been backed up by empirical evidence from some studies that find differences in values when presenting the same health gain in different formats (Morris and Hammitt 2001; Desaigues et al., 2007; Grisolia et al., 2018). Perhaps more strikingly, though, is that Chilton et al. (2004) found that the estimated VOLY dropped considerably when ‘normal health’ was replaced by ‘poor health’ in the valuation scenario (e.g. from £27,630 to £7,280 for a VOLY estimated from a one-month extension in life expectancy). It is unknown how respondents perceived their health state to be in other studies but if it was perceived as in poor health (in the absence of explicit information) then there is reasonable chance that at least some VOLY estimates elicited under this particular framework are (possibly significant) underestimates. This is over and above any additional discounting that may have been applied by respondents to account for the (perceived) timing of this gain.

As stated, the alternative approach is to ask respondents to value only the risk reduction(s), to either calculate a (statistical) life year indirectly from a one-period risk reduction/VPF (e.g. Alberini et al. (2006)) or directly (e.g. Desaigues et al., 2007 based on Krupnick et al.’s [2002] study). The main strength of this method is that it conveys to respondents the notion that a gain in life expectancy is not an add-on at the end of life. However, variants of the method have been shown to suffer from problematic response patterns that cannot be reconciled with theoretical predictions. The most common of these is scope insensitivity, whereby WTP for a risk reduction is not proportional to the size of the risk reduction. For example, in Krupnick et al (2002), respondents were asked to provide an annual WTP for risk reductions each year of 1 in 1000 and 5 in 1,000 reductions in the risk of death. The ratio of WTPs for the 2 risk reductions was small (in the range 1.4-1.6) and certainly not in approximate proportion to the ratio of the risk reductions, as would be predicted by theory. This insensitivity to scope (of the risk change) is pervasive and well known, particularly in the VSL literature. For example, Lindhjem et al.’s (2011) meta-analysis regressions indicate an elasticity of VSL with respect to risk reduction between -0.5 and -0.25, even in restricted samples. This is statistically significant and implies that WTP is significantly less than proportional to the size of the risk reduction.

It has been argued that insensitivity to scope may be driven by poor communication of risk. However, Desaigues et al. (2007) and other studies in RQI employed the most commonly accepted risk communication method (risk grids in which respondents are shown for example a 10,000 square grid, with a small proportion of shaded squares to convey visually a respondent’s risk of dying) and yet they still encountered scope insensitivity. This suggests that the problem of insensitivity to scope of the risk change may not be resolvable by standard risk communication methods alone. Some studies chose instead to communicate this risk in a more indirect manner^{[footnote 6]}.

We have outlined some of the key issues with the 2 main approaches: describing a change in life expectancy, and describing a change in risk. These issues may have motivated those studies that provided respondents with both explicit, quantitative information on the risk change as well as the gain in life expectancy. A major purpose of these particular studies is methodological in nature, specifically set up to explore whether differences in the way that life expectancy is communicated affects WTP. Extrapolating these findings for policy purposes is not straightforward and will be considered later in this report, and so in what follows we simply report the main findings.

Morris and Hammitt (2001) compared WTP for identical health outcomes expressed as life expectancy gains and risk reductions respectively. It was found that respondents expressed higher WTP for the benefit expressed as a life expectancy gain and the life expectancy format performed better than the risk reduction format with respect to sensitivity to scope of the benefit.

Desaigues et al., (2007) compared the following survey variants (among others); 1) mortality risk reduction (no information about life expectancy specified as a duration) 2) increase in life expectancy (no information about risk) and 3) a combination of 1 and 2. Based on debriefing and discussions with respondents they concluded that the concept of life expectancy is better understood than the risk of dying, but that many respondents reported a zero willingness-to-pay as the gain in life expectancy was too short^{[footnote 7]}.

More recently, Grisolia et al., (2018) compared 3 methods for presenting identical health outcomes; A) mortality risk reduction, B) increase in life expectancy and C) an increase in the probability of reaching an individual’s full lifespan. They conclude that the most effective way of communicating is using increases in life expectancy (B) rather than risk reduction per se (A)^{[footnote 8]}. This is based on the goodness of fit of the econometric model^{[footnote 9]}, significance of relevant parameters and the convergence to existing literature in Ryen and Svensson’s (2015)^{[footnote 10]}. In contrast, they find frame (C) to be particularly problematic: the majority of respondents appear to need to be compensated to choose a scenario with an increased probability of achieving a full life span and they suggest that respondents may have had difficulties in grasping the meaning of this framing method. The fact that the study finds such a large variation across the 3 methods of presenting the health outcomes does warrant some caution about the validity of the survey results. Another potential problem is that the study was carried out in the context of cardiovascular mortality risk. This allowed the researchers to offer gains in life expectancy from 5 to 13 months which in most other contexts would be unrealistic^{[footnote 11]}, reducing its generalisability for policymaking. In addition, the proposed gain in life expectancy would be delivered by a combination of change in diet (customized for the individual), exercise and expenditures on food. This is very different from a standard VPF/VOLY study, not least because there are costs involved in exercise (not mentioned or taken into account in the study). A final validity concern is that it is not clear (in the paper or the survey instrument reported) whether changes in costs and diets would need to take place for the rest of the respondent’s life to achieve the full effect which is crucial information for elicitation of preferences.

It is difficult to arrive at a definitive conclusion, but it would seem that the case for deploying either one method or the other is weak. Based on this, it would seem that, where possible, the most defensible approach is likely to be one that incorporates both methods. Intuitively, at least, this should make it clearer to respondents that the life expectancy gain is not an ‘add on’ at the end of life and enable a clearer link to be made in the respondent’s mind as to the nature of the benefit a small risk reduction (or series of risk reductions) can bring to them personally (an increase in chance of survival over their lifetime).

2.2. Communicating the link between life expectancy gains and risk reductions

2.2.1. Survival curves

In earlier studies, as noted, the link between life expectancy gains and risk reductions was not made explicit. However, since then, a number of studies (Nielsen, 2010; Desaigues et al., 2011^{[footnote 12]}; Ara and Tekesin, 2017) presented survival curve diagrams and emphasised that it was the change in life expectancy being valued. The primary purpose of the visual aid was to convey intuitively (i) how underlying perturbations in the survival function change life expectancy, (ii) the on-going nature of the gain and thereby to reduce or eliminate the use of the (incorrect) heuristic that the gain in life expectancy is an ‘add- on’ at the end of life, in poor health.

Baker et al. (2014) reports the results from a structured debriefing exercise to qualitatively investigate the survival curve approach. The study concluded that most respondents appeared to understand the continuous nature of the life expectancy gain over time and that the gain was in some way uncertain. As such, using the survival curve diagram was an important step forward and meant that such respondents were not constructing their WTP responses for the certainty of extra months/years at the end of life i.e. an “add-on”. A more concerning finding was that that many respondents conflated quantity of life expectancy gains with quality of life improvements, and seemed to be adding in quality of life when stating the WTP. According to the authors, the reference to quality of life could partly be explained by the fact that the survey was in an environmental context and it would not be an incorrect assumption to think that there would be environmental benefits as well, that could in turn enhance quality of life. In addition, the concept of quality of life is more familiar and concrete than gains in life expectancy. Overall, this study highlighted that careful descriptions of the context in which life expectancy gains are achieved helped people to understand the good, but there was a potential for additional, irrelevant^{[footnote 13]} factors to be brought to bear in respondents’ valuations.

2.2.2. Differing distributions of risk reductions

In addition, to explain the ongoing nature of a gain in life expectancy, Hammitt (2007) emphasises that changes in life expectancy can be generated by any number of perturbations in an individual’s survival function, for instance: a) a reduction in risk of death during the coming year, b) an ongoing reduction over the next 20 years, or c) a reduction that only takes effect in later years of life. With this in mind, the specification of such gains, and respondents’ understanding of how they are derived, is an empirical issue.

2 papers (Nielsen et al., 2010 and Hammitt and Tuncel, 2015) focussed on developing a method to communicate how different distributions of risk reduction can generate the same gain in life expectancy. These studies hypothesised that whilst a gain in life expectancy (in normal health) may be considered a ’good thing’, people may nevertheless have preferences over these distributions. These preferences may, in turn, lead to different valuations of the same gain in life expectancy. The consequences for valuation were not directly addressed: instead, a relative valuation method (see sub section XX below) was used to estimate the strength of preference between different ways of generating a given gain in life expectancy (a,b and c above).

The significant finding from Nielsen et al. (2010) was that most respondents displayed a marked preference for one way of generating a particular gain in life expectancy rather than another. The preferences for 3 main orderings of the life expectancy programmes were distributed more or less evenly across this subject pool. In Hammitt and Tuncel (2015), the pattern of preference orderings is similar to that obtained by Nielsen et al. (2010) except that in Nielsen et al. (2010), many fewer respondents expressed indifference over the 3 perturbations^{[footnote 14]}.

It was acknowledged earlier that while the primary methodological concern with respect to eliciting a monetary VOLY must be its explanation in order to maximise respondents’ understanding of what is a very complex good, other areas of the hypothetical market and/or survey administration have the potential to impact significantly on the validity of any survey-based VOLY. In particular, we single out the payment vehicle (and its duration) – for which there is very little evidence in the literature regarding its effect on a VOLY per se. We also consider the mode of administration – for which a few studies are available from the wider mortality and morbidity valuation literature.

3. Additional potential biases with respect to valuing life expectancy gains

3.1. Payment vehicle

We do not review the issue of payment vehicle bias in stated preference surveys in general, since this issue has been extensively discussed and researched in the wider stated preference literature. We focus instead on aspects of the payment vehicle central to VOLY valuation. We do, however, note that obvious candidates such as voluntary contributions or taxes – common in environmental valuation studies - are not suited to the unique manner in which life expectancy gains are delivered i.e. ongoing and over a persons’ lifetime (regardless of the perturbation in hazard rates that apply in any one particular case). It would not be credible to expect voluntary donations to be maintained over such a protracted time period and tax systems also change, sometimes significantly, over a person’s lifetime^{[footnote 15]}.

This feature has been recognised in VOLY elicitation, in that a consensus appears to have evolved in the literature to use payment vehicle involving ongoing payments. This approach also has the advantage that a person’s budget constraint is less likely to bite relative to a one-off payment. However, unfortunately, no studies to date have compared results from a one-off payment and an ongoing payment, so any observations on this issue must be tentative. This is at least in part due to the fact that discounting behaviours will affect both values but almost certainly in different ways.

In the VPF literature, a number of studies have presented respondents with a 10 year risk reduction and asked them to value that over a 10 year period, see e.g. Krupnick et al., (2002); Alberini et al. (2004); Krupnick (2007); Alberini et al. (2006). Alberini et al. (2002) also used graphs to show to relationship between risk reduction and payment frequency. Other variants include valuing annual risk reductions with monthly payments (Leiter and Pruckner, 2009) or with an unspecified payment frequency (Hammitt and Graham, 1999; Andersson, 2007). Vassanadumrongdee and Matsuoka (2005) specified an annual payment for a latent benefit (in 10 years’ time). Turning to the VOLY literature, one approach has been utilised in which respondents are asked to assume that cost will be incurred through a (permanent) increase in the cost of living (Nielsen et al., 2010; Chilton et al., 2004). This approach is flexible and appears to be accepted by respondents although its efficacy in generating reliable values has not been proven explicitly^{[footnote 16]}. Other variants of this include WTP per month out of normal salary (Desaigues et al. 2011), a monthly tax for the rest of ones’ life (Vlachokostas et al, 2010), a payment each year out of a normal budget constraint or monthly household budget (Ara and Tekesin, 2017 and Chanel and Luchini, 2008,2014 respectively) or one payment per year staring immediately (Alberini et al., 2006). Grisolia et al. (2018) do not specify the payment duration. Other isolated examples include a one-off payment now (Johannesson and Johansson, 1996) and risk-free cigarettes (Hammar and Johansson-Stenman, 2004). It is not possible to disentangle the effect on the basis of the current literature, given the potential confounds and interactions between other aspects of the surveys for example, elicitation format, sample characteristics etc., making comparisons between different payment formats difficult. Nevertheless, a general consensus is apparent in that most VOLY studies utilise a recurring payment (either monthly or annually) for the rest of a respondent’s life.

3.2. Survey mode comparisons

It is beyond the scope of this study, to identify the most appropriate survey administration mode, not least since the most appropriate will, to an extent, be driven by the particular survey instrument. Instead, we focus on the most likely candidates going forward. As noted in RQI, most of the VOLY studies have been carried out either on the Internet or as a face-to-face interview, either by members of research team or a professional market research company. These seem reasonable alternatives but the obvious trade-off here is the potential for a VOLY to be generated from a large, representative sample (online) versus a potential increase in control (in respect of who responds to the survey and the quality of their answers/degree of engagement – in face-to-face elicitations).

A review by Lindhjem and Navrud (2011) of SP studies comparing internet with other modes observed that they generally do not find substantial difference. The majority of welfare estimates were more or less equal, although somewhat lower for the internet surveys. They note that there was no clear evidence of substantially lower quality or validity of Internet responses, but that they are often confounded by measurement and sample composition effects. 4 stated preference studies found no significant difference between WTP estimates from internet and face-to-face modes: Covey et al. (2010); Nielsen (2011), Lindhjem and Navrud (2008); and Lee et al. (2016) across different contexts. Covey et al. (2010) reported the results of a stated preference study concerning rail safety that was carried out on both a face-to-face and internet basis and found that there was a close correspondence between the findings of the 2 studies. In particular, both studies indicated that respondents placed roughly the same value on prevention of a rail fatality - regardless of the precise cause of the fatality (e.g. train derailment or collision) - if the potential victims were behaving responsibly (including roughly the same value for prevention of a fatality in a multiple-fatality accident as in a single-fatality case). However, if potential victims were behaving irresponsibly (e.g. trespassers or drunks) then in both the face-to-face and internet surveys the implied values were reduced by more than 50%^{[footnote 17]}. Nielsen (2011) estimated a VOLY in the context of air pollution whilst the Lindhjem and Navrud (2008) and Lee et al. (2016) studies are both in the context of environmental goods.

In contrast, WTP from internet surveys was found to be either higher than face-to-face interviews in Canavari et al. (2005) and lower than face to- face interviews in Marta- Pedroso et al. (2007) and Berrens et al. (2003)). The studies were carried out in the context of pesticide ban in food, preserving an area and climate change, respectively. In the context of eliciting expert opinion on carbon capture storage, Baker, Bosetti et al. (2014) found different results from the 2 different modes examined (internet and face- to-face, but were unable to provide any clear indication of which method might be preferred. Similarly, Goldenbeld, C. and De Craen, S., (2013) asked respondents to judge different road safety measures. Results indicated that online respondents were less inclined to give socially desirable answers and were less inclined to use more extreme ratings in their opinions about measures. Interestingly, contrary to their expectations, face-to-face respondents did not tend to give more positive answers in judgment of road safety measures.

Several studies have compared the socio-demographic differences between respondents in stated preference surveys using internet and other collection modes, and report some sample differences (Berrens et al. 2003, Canavari et al. 2005, Marta-Pedroso et al. 2007, Nielsen 2011, Windle et al. (2011)). However, Windle et al. (2011) and Goldenbeld and De Craen (2013) found no consistent sociodemographic differences across studies.

An updated review of the effects of survey administration mode is merited before any recommendation could be made although, as a general point, it would seem that, whatever mode is used the questionnaire must be thoroughly tested using standard cognitive questionnaire design tools such as focus groups and in-depth interviews or verbal protocols to reduce obvious sources of error introduced by the complexity of the good being valued.

We now turn to the final theme of this RQ, possible alternative valuation approaches.

4. Relative valuation approach

The main alternative to direct, or absolute, valuation in the mortality and morbidity valuation literature to date is a relative valuation approach. This approach encompasses a variety of formats but the general principle is to avoid asking respondents for monetary values and instead ask them to trade-off characteristics of the cause of death for example, the risk itself (known as a ‘risk-risk’ trade-off, for example Magat et al., 1996 and McDonald et al., 2016) or the number of people affected (known as ‘matching’, for example Chilton et al., 2002). The underlying rationale is that it is arguably cognitively simpler for people to trade ‘like-for-like’ as opposed to money and risk directly.

4.1. Risk-risk trade-off

The Risk-Risk trade-off method was developed by Viscusi et al. (1991), and extended in Magat et al. (1996).^{[footnote 18]} this method, participants are presented with 2 risk-change scenarios, a reduction (or increase) in the risk of dying by cause A, and a reduction (or increase) in the risk of dying by cause B (see below for some examples of causes A and B used in the literature). Participants select their preferred option, and the size of the risk reductions are then altered until the participant is indifferent between the 2 scenarios. The researchers can then infer the preference for avoiding fatality by cause A relative to the preference for avoiding fatality by cause B. These ‘relative’ values can be applied to an absolute monetary ‘peg’ (e.g. a ‘roads’ VPF) to generate WTP-based monetary values for the prevention of fatalities in other contexts. Van Houtven et al. (2008) set out the formal underpinnings of the relationship between risk-risk trade-offs and WTP values. Recently, Rheinberger et al. (2018) used a DCE design that embedded the concept of a Risk-Risk trade-off to disentangle the fatality and injury components of the WTP to avoid a statistical accident, demonstrating a useful new methodology that bridges direct and indirect elicitation of the value of safety improvements.

The Risk-Risk trade-off has also been applied in the context of life expectancy gains (Nielsen et al., 2010; Hammitt and Tuncel, 2015). As the focus of these studies was to examine the feasibility of asking respondents to trade off differing risk distributions (as opposed to life expectancy gains per se) the researchers carried out their studies in a context-free setting. From the perspective of viability, in both cases reported the method generated plausible results with respect to preference rankings^{[footnote 19]}, although notably, WTP values were not elicited in either case, given the overall purpose of the 2 studies.

Whilst reasonably numerous, studies using this method have focused on reporting the empirical results, making it difficult to assess the robustness of the methodology. An exception to this is the methodological investigation by Clarke et al. (1997) in which the results from the standard gamble method, time trade-off method, and risk–risk trade-off method were compared in the context of (Painless) death from a natural disaster and Gaucher Diseases (metabolism disorder). Utilising a test–retest methodology, they found that the results from the risk–risk trade-off method performed worst with respect to reliability. However, in the context of non-fatal and fatal injuries (traffic), Nielsen et al. (2018) found that estimates from a risk–risk trade-off study can be improved by employing a pre-survey learning experiment in which respondents make incentivised risky choices and also using a frame that focuses on the total risk or risks that respondents face.

4.2. Chained method (VPF)^{[footnote 20]}

The chained approach was developed in the context of the VPF in response to the failure of the CV method to show strong scale sensitivity (Beattie et al. 1998; Carthy et al. 1999). The method is based on 2 steps - which the authors argue are conceptually manageable - the first of which involves estimating the marginal rate of substitution between wealth and risk of a non-severe non-fatal injury and the second step involves estimating the relative utility loss for death and the non-severe non-fatal injury.

To assess the internal consistency of the method, Carthy et al. (1999) deployed 2 alternative formats, direct and indirect. Under the former, VOSL estimates are obtained by chaining the marginal rate of substitution between wealth and risk of one non-fatal injury of moderate severity to the relative utility loss for death and that non-fatal injury. In the latter, a similar process is followed except that the VOSL is based on estimates of the marginal rate of substitution between wealth and risk of the moderately severe non-fatal injury obtained by chaining the marginal rate of substitution between wealth and risk of a non-fatal injury of lesser severity to the relative utility losses for death and the moderately severe injury and the relative utility losses of the 2 injuries. Whilst the results of this internal consistency test were on the face of it disappointing (whereby the mean estimates from the latter were some 60-70% larger), the authors argue that at least some of the discrepancy is likely to be due to a compounding of errors in the indirect approach. Perhaps more importantly, the overall results did not suffer from pervasive scope insensitivity, which had been present in the earlier study (Beattie et al., 1998) and provided the main motivation for the development of this ‘new’ approach.

4.3. The J-Value approach

This approach aims to derive the value of a gain in life-expectancy from what is essentially a simple lifetime “work/leisure” choice model. However, in a report commissioned by the HSE, Spackman (2009) argues that the J-Value approach is “…too simplistic to be a competitor to the methods now established in the UK and elsewhere for the valuation of fatality risks.” This conclusion is substantially reinforced by the argument presented in Jones-Lee & Chilton (2017).

5. Summary

This review examined some of the key methodological issues with respect to VOLY elicitation, drawing on the wider literature where relevant. With respect to communicating life expectancy gains, whilst it is not fully resolved, there is a general consensus to elicit WTP for life expectancy gains between 1 and 12 months. In some cases, this is based on epidemiological evidence, in others it is less clear. Whatever the case, these durations seem to be accepted by respondents (even though the range of values elicited varies widely, as discussed in RQI.

More importantly, it seems that neither asking about life expectancy gains or risk reductions alone is entirely satisfactory, both bring with them a different set of problems. To avoid the perception that it is an ‘add on’ at the end of life steps have been taken to elicit preferences over underlying risk reductions, although this has not yet been done in the context of WTP. Whilst a relative valuation approach is possible, there is no strong evidence in the literature to suggest it would fare better than a direct WTP approach to elicit a VOLY, either through following procedures already established in the literature or developing new procedures for eliciting WTP using a chained approach or assessment and valuation of different risk reduction distributions.

Another consensus appears to have arisen with respect to the duration of the payment in that, to mirror the on-going and permanent nature of the risk reductions and to avoid biting budget constraints, payment should be elicited (either monthly or annually) either over a long period (perhaps 10 years) or over person’s lifetime. There is less consensus on the payment vehicle itself with some studies utilising sustained increases in the cost of living and others expenditures from salary or normal budget constraints. There is little evidence, if any, to support for one-off payments.

The evidence on survey administration mode is mixed and there are very few studies that compare different modes within a study. Cross study comparisons of survey mode effects are difficult, if not impossible.

Bibliography

Alberini, A. (2017). Measuring the economic value of the effects of chemicals on ecological systems and human health, OECD Environment Working Papers, No.116, OECD Publishing, Paris.

Alberini, A., Cropper, M., Krupnick, A. and Simon, N.B., 2004. Does the value of a statistical life vary with age and health status? Evidence from the US and Canada. Journal of Environmental Economics and Management, 48(1), pp.769-792.

Alberini, A., Hunt, A. and Markandya, A., 2006. Willingness to pay to reduce mortality risks: evidence from a 3-country contingent valuation study. Environmental and Resource Economics, 33(2), pp.251-264.

Andersson, H., 2007. Willingness to pay for road safety and estimates of the risk of death: Evidence from a Swedish contingent valuation study. Accident Analysis & Prevention, 39(4), pp.853-865.

Ara, S. and Tekeşin, C., 2017. The Monetary Valuation of Lifetime Health Improvement and Life Expectancy Gains in Turkey. International journal of environmental research and public health, 14(10), p.1151.

Baker, E., Bosetti, V., Jenni, K. and Ricci, E., 2014. Facing the experts: survey mode and expert elicitation. CMCC-Euro Mediterranean Centre for Climate Change.

Baker, R., Bartczak, A., Chilton, S. and Metcalf, H., 2014. Did people “buy” what was “sold”? A qualitative evaluation of a contingent valuation survey information set for gains in life expectancy. Journal of Environmental Management, 133, pp.94-103.

Beattie, J., Covey, J., Dolan, P., Hopkins, L., Jones-Lee, M., Loomes, G., Pidgeon, N., Robinson, A. and Spencer, A. (1998). On the contingent valuation of safety and the safety of contingent valuation: Part 1 – caveat investigator. Journal of Risk and Uncertainty, 17, 5-25.

Berrens, R.P., Bohara, A.K., Jenkins-Smith, H., Silva, C. and Weimer, D.L., 2003. The advent of Internet surveys for political research: A comparison of telephone and Internet samples. Political analysis, 11(1), pp.1-22.

Cameron, M., Gibson, J., Helmers, K., Lim, S., Tressler, J. and Vaddanak, K., 2010. The value of statistical life and cost–benefit evaluations of landmine clearance in Cambodia. Environment and Development Economics, 15(4), pp.395-416.

Canavari, M., Nocella, G. and Scarpa, R., 2005. Stated willingness-to-pay for organic fruit and pesticide ban: an evaluation using both web-based and face-to-face interviewing. Journal of Food Products Marketing, 11(3), pp.107-134.

Carthy, T., Chilton, S., Covey, J., Hopkins, L., Jones-Lee, M., Loomes, G., Pidgeon, N. and Spencer, A.. (1999). On the contingent valuation of safety and the safety of contingent valuation: Part 2 - The CV/SG “Chained” Approach, Journal of Risk and Uncertainty, 17(3), 187-213.

Chanel, O. and Luchini, S., 2008. Monetary values for air pollution risk of death: A contingent valuation survey.

Chanel, O. and Luchini, S., 2014. Monetary values for risk of death from air pollution exposure: a context-dependent scenario with a control for intra-familial altruism. Journal of Environmental Economics and Policy, 3(1), pp.67-91.

Chilton, S., Covey, J., Hopkins, L., Jones-Lee, M., Loomes, G., Pidgeon, N. and Spencer, A., 2002. Public perceptions of risk and preference-based values of safety. Journal of Risk and Uncertainty, 25(3), pp.211-232.

Chilton, S., Covey, J., Jones-Lee, M., Loomes, G. and Metcalf, H., 2004. Valuation of health benefits associated with reductions in air pollution. Defra publication PB, 9413.

Chilton, S.M. and Hutchinson, W.G. (1999). Some further implications of incorporating the warm glow of giving into welfare measures: A comment on the use of donation mechanisms by Champ et al., Journal of Environmental Economics and Management, 37(2), 202-209.

Chilton, S., Jones-Lee, M., Kiraly, F., Metcalf, H. and Pang, W., 2006. Dread risks. Journal of Risk and Uncertainty, 33(3), pp.165-182.

Clarke, A.E., Goldstein, M.K., Michelson, D., Garber, A.M. and Lener, L.A., 1997. The effect of assessment method and respondent population on utilities elicited for Gaucher disease. Quality of Life Research, 6(2), pp.0-0.

Covey, J., Robinson, A., Jones-Lee, M. and Loomes, G., 2010. Responsibility, scale and the valuation of rail safety. Journal of Risk and Uncertainty, 40(1), pp.85-108.

Desaigues, B., Ami, D., Bartczak, A., Braun-Kohlová, M., Chilton, S., Czajkowski, M., Farreras, V., Hunt, A., Hutchison, M., Jeanrenaud, C. and Kaderjak, P., 2011. Economic valuation of air pollution mortality: A 9-country contingent valuation survey of value of a life year (VOLY). Ecological indicators, 11(3), pp.902-910.

Desaigues, B., Rabl, A., Ami, D., My, K.B., Masson, S., Salomon, M.A. and Santoni, L., 2007. Monetary value of a life expectancy gain due to reduced air pollution: Lessons from a contingent valuation in France. Revue d’économie politique, 117(5), pp.675-698.

Dolan, P., Jones-Lee, M. and Loomes, G., 1995. Risk–risk versus standard gamble procedures for measuring health state utilities. Applied Economics, 27(11), pp.1103- 1111.

Goldenbeld, C. and De Craen, S., 2013. The comparison of road safety survey answers between web-panel and face-to-face; Dutch results of SARTRE-4 survey. Journal of safety research, 46, pp.13-20.

Grisolía, J.M., Longo, A., Hutchinson, G. and Kee, F., 2018. Comparing mortality risk reduction, life expectancy gains, and probability of achieving full life span, as alternatives for presenting CVD mortality risk reduction: A discrete choice study of framing risk and health behaviour change. Social Science & Medicine.

Hammar, H. and Johansson-Stenman, O., 2004. The value of risk-free cigarettes–do smokers underestimate the risk?. Health economics, 13(1), pp.59-71.

Hammitt, J. (2007). Valuing changes in mortality risk: lives saved vs. life years saved, Review of Environmental Economics and Policy: 1: 228-240.

Hammitt, J.K. and Graham, J.D., 1999. Willingness to pay for health protection: inadequate sensitivity to probability?. Journal of risk and uncertainty, 18(1), pp.33-62.

Hammitt, J.K. and Tunçel, T., 2015. Preferences for life-expectancy gains: Sooner or later?. Journal of Risk and Uncertainty, 51(1), pp.79-101.

Hynes, S. and Hanley, N., 2006. Preservation versus development on Irish rivers: whitewater kayaking and hydro-power in Ireland. Land Use Policy, 23(2), pp.170-180.

Johannesson, M. and Johansson, P.O., 1996. To be, or not to be, that is the question: an empirical study of the WTP for an increased life expectancy at an advanced age. Journal of risk and uncertainty, 13(2), pp.163-174.

Jones-Lee, M. and Chilton, S. (2017). Principal limitations of the j-value model. International Journal of Business Continuity and Risk Management, 7(3), 222-232.

Krupnick, A., 2007. Mortality-risk valuation and age: stated preference evidence. Review of Environmental Economics and Policy, 1(2), pp.261-282.

Krupnick, A., Alberini, A., Cropper, M., Simon, N., O’brien, B., Goeree, R. and Heintzelman, M., 2002. Age, health and the willingness to pay for mortality risk reductions: a contingent valuation survey of Ontario residents. Journal of risk and Uncertainty, 24(2), pp.161-186.

Lee, C.K., Kim, T.K. and Mjelde, J.W., 2016. Comparison of preservation values between Internet and interview survey modes: the case of Dokdo, South Korea. Journal of Environmental Planning and Management, 59(1), pp.22-43.

Leiter, A.M. and Pruckner, G.J., 2009. Proportionality of willingness to pay to small changes in risk: the impact of attitudinal factors in scope tests. Environmental and Resource Economics, 42(2), pp.169-186.

Lindhjem, H. and Navrud, S. (2008). Internet CV surveys-a cheap, fast way to get large samples of biased values. MPRA Paper 11471, University Library of Munich, Germany.

Lindhjem, H. and Navrud, S., 2011. Are Internet surveys an alternative to face-to-face interviews in contingent valuation?. Ecological economics, 70(9), pp.1628-1637.

Magat, W.A., Viscusi, W.K. and Huber, J., 1996. A reference lottery metric for valuing health. Management Science, 42(8), pp.1118-1130.

Marta-Pedroso, C., Freitas, H. and Domingos, T., 2007. Testing for the survey mode effect on contingent valuation data quality: A case study of web based versus in-person interviews. Ecological economics, 62(3-4), pp.388-398.

McDonald, R.L., Chilton, S.M., Jones-Lee, M.W. and Metcalf, H.R., 2016. Dread and latency impacts on a VSL for cancer risk reductions. Journal of Risk and Uncertainty, 52(2), pp.137-161.

Morris, J. and Hammitt, J.K., 2001. Using life expectancy to communicate benefits of health care programs in contingent valuation studies. Medical Decision Making, 21(6), pp.468-478.

Morrison M, Blamey R, Bennett J, 2000. Minimizing payment vehicle bios in contingent valuation studies. Environmental and Resource Economics 16: 407-422.

Nielsen, J.S., 2010. Approaching the value of a life year: empirical evidence from a Danish contingent valuation survey. Nationaløkonomisk Tidsskrift, 148(1), pp.67-85.

Nielsen, J.S., 2011. Use of the Internet for willingness-to-pay surveys: A comparison of face-to-face and web-based interviews. Resource and Energy Economics, 33(1), pp.119-129.

Nielsen, J.S., Chilton, S. and Metcalf, H., 2018. Improving the risk–risk trade-off method for use in safety project appraisal responses. Environmental Economics and Policy Studies, pp.1-26.

Nielsen, J.S., Chilton, S., Jones-Lee, M. and Metcalf, H., 2010. How would you like your gain in life expectancy to be provided? An experimental approach. Journal of Risk and Uncertainty, 41(3), pp.195-218.

Rheinberger, C.M, Schlapfer, F. and Lobsiger, M. 2018 A novel approach to estimating the demand value of public safety, Journal of Environmental Economics and Management 89 285–305.

Ryen, L. and Svensson, M. (2015). The willingness to pay for a quality adjusted life year: a review of the empirical literature. Health Economics, 24: 1289–1301.

Spackman, M. (2009). Review of the J-value Literature – Final Report. The Health and Safety Executive. London, NERA Economic Consulting.

Van Houtven, G., Sullivan, M.B. and Dockins, C., 2008. Cancer premiums and latency effects: A risk tradeoff approach for valuing reductions in fatal cancer risks. Journal of Risk and Uncertainty, 36(2), pp.179-199.

Vassanadumrongdee, S. and Matsuoka, S., 2005. Risk perceptions and value of a statistical life for air pollution and traffic accidents: evidence from Bangkok, Thailand. Journal of Risk and Uncertainty, 30(3), pp.261-287.

Viscusi, W.K., 2009. Valuing risks of death from terrorism and natural disasters. Journal of Risk and Uncertainty, 38(3), pp.191-213.

Viscusi, W. K., Magat, W. A., Huber, J. (1991). Pricing environmental health risk: survey assessments of risk-risk and risk-dollar trade-offs for chronic bronchitis. Journal of Environmental Economics and Management, 21, 35-51.

Vlachokostas, C., Achillas, C., Slini, T., Moussiopoulos, N., Banias, G. and Dimitrakis, I., 2011. Willingness to pay for reducing the risk of premature mortality attributed to air pollution: a contingent valuation study for Greece. Atmospheric Pollution Research, 2(3), pp.275-282.

Wakker, P., Deneffe, D., 1996. Eliciting von Neumann-Morgenstern Utilities When Probabilities are Distorted or Unknown. Management Science 42(8), 1131-1150.

Windle, J. and Rolfe, J., 2011. Comparing responses from internet and paper-based collection methods in more complex stated preference environmental valuation surveys. Economic Analysis and Policy, 41(1), p.83.

1. The context of the valuation exercise could have a substantive impact. This issue is, however, considered in detail in RQIV. ↩

2. It is beyond the scope of this literature review to consider SP survey practice in general. Suffice to say, any new primary study tender should carefully justify the preferred approach (such as survey mode e.g. internet, CAPI face- to-face; and elicitation mechanism e.g. open ended, dichotomous choice experiment) since these will vary depending on the complexity of the proposed approach, the resources available and so forth. ↩

3. Payment vehicle bias is not confined to VOLY estimates. However, rather than review this literature in general, we confine ourselves to payment vehicles most suitable for our context, not least since a ‘best’ payment vehicle applicable to all public goods has not as yet been identified. Also note that the payment vehicle is intrinsically linked to the framing. If the risk reduction is presented as a public good, increments in regular taxes seem to be the natural instrument; if it’s taxes would it be an absolute or relative tax raise, etc.. If instead it is a private good then a onetime price increment in a service may be just as credible. ↩

4. Almost without exception, the life expectancy gain offered in all VOLY studies ranges between 1 and 12 months. Whist the reason for this is generally not explained or discussed, Chilton et al. (2004; footnote 4, p.6) note that the gains of 1 month, 3 months and 6 months used in that study were based on available epidemiological evidence which at that time suggested that this was the most appropriate range of values, that is deaths brought forward (from air pollution) typically involved losses of 6 months or less. It also seemed unlikely that any realistic package of measures could increase life expectancy by more than 6 months. ↩

5. The scenarios and associated questions in this study arose as a result of intensive focus group and cognitive testing and the descriptions mirrored the best available scientific consensus with respect to the impacts available at the time. However, these quotes were chosen primarily to illustrate the principle of this approach, and should not be taken to imply anything with respect to its quality viz other studies. ↩

6. For example, as noted in RQI Chanel and Luchini (2008) who described the life expectancy gain as being generated by living in a town with 25 % or 50-100% less pollution than Marseilles, this is likely only to bring with it different problems - even if respondents were able to convert this type of information first into risk reductions and then into life expectancy gains. ↩

7. There is no literature (as far as we are aware) to inform the issues of how well understood the risk of dying is compared to life expectancy. As far as a VOLY is concerned, the main issue that can be accommodated in a survey is to communicate as well as is possible the link between risk reductions and the generation of life expectancy gains and to (try and) clarify that respondents have understood it to an acceptable degree. ↩

8. Overall, the gains in life expectancy did also appear very high compared to the risk reductions offered. ↩

9. In terms of goodness of fit, the best WTP space model is B, with the lowest log-likelihood in absolute terms (-1604.5). The log-likelihood of frame A and C where -2521.6 and -3569.9, respectively. ↩

10. The fact that it converges with the above meta-analysis is perhaps not as significant as it might seem and might be just accidental convergence. The Ryen and Svensson (2015) value is simply the mean of the values from the studies reviewed and hence is not a particularly reliable metric by which to judge the outcome of a particular survey. ↩

11. Such large gains have the potential to impact, at least to a degree, the magnitude of their VOLY estimate (circa £60,000) although the extent of the impact is not known, since VOLYs based on other risk reductions/life expectancy gains were not reported. ↩

12. However, Alberini (2017) criticises the particular survival curves used in this study due to misspecification of life expectancy on the horizontal axis. ↩

13. At least with respect to the policy output being valued. ↩

14. Taken together, these 2 findings may be taken to imply that WTP for these different risk reductions would be more or less equal, suggesting that for policy purposes a VOLY derived from any particular distribution could be extrapolated to the society as whole. We would caution against such a conclusion. ↩

15. Additional problems include, but are not restricted to, the theoretical inequality between the value of a public good under private (voluntary contributions) and public provision (Chilton and Hutchinson, 1999) and the high degree of protest zeros that tax vehicles often generate (Morrison et al., 2000). ↩

16. The usefulness of this approach depends among other things on the variability in living costs. This offers some possibilities for endogenous interpretation of costs. ↩

17. An internet-only non-fatal accident study was also undertaken. It produced internally inconsistent (and hence unpublishable) results. ↩

18. It is worth mentioning that the risk-risk trade-off method is in the end very similar to a standard gamble, of course other utility elicitation approaches could be employed to value health instead of money gambles, see e.g. Wakker and Deneffe (1996) for an example. ↩

19. 2 distinct sample were deployed and 2 different modes of administration. ↩

20. We consider the chained method in relation to WTP-QALY in RQIII. ↩