The Role of Economic Valuation Of

The Role of Economic Valuation of

Health Effects in UN/ECE LRTAP Work:

History, Debate, Outstanding Issues

David Pearce*

Prepared for the UK DETR/UN ECE Symposium on

THE MEASUREMENT AND ECONOMIC VALUATION

OF HEALTH EFFECTS OF AIR POLLUTION

London, Institute of Materials, February 19-20, 2001

*Professor of Environmental Economics, University College London and Rapporteur, UN ECE Network of Experts on Benefits and Economic Instruments (NEBEI).

• Contact:

1 Introduction

Scientific concern with transboundary acidifying pollutants dates back to the 1960s. International action dates from the early 1970s, mainly in the field of research, and it is the research that provided the basis for the 1979 Geneva Convention on the Long Range Transport of Air Pollution (LRTAP). As is well known, LRTAP did not set targets, offered no timetable for emission reductions and contained no means for compliance. It is only with the first Protocol, the 1985 Helsinki Protocol on sulphur emissions, that firm action began to be taken and various actions of ever-widening geographical and pollutant coverage subsequently ensued - see Table 1.

Table 1The Convention on LRTAP and its Protocols

• Signature does not necessarily mean ratification.

The initial stimulus to LRTAP and its Protocols came from concern about the acidification of lakes in Scandinavia, and later by concern about forest damage. The focus of LRTAP was therefore not surprisingly very much on ecosystem damage arising from air pollutants. Links between air pollution and human health had already been suggested for some considerable time, but health was not a driving force for LRTAP. It was quite distinctly ecosystem based. It is only fairly recently that health-based concerns have had greater influence on the process while, at the same time, economic approaches based on cost-benefit analysis have also been produced. The LRTAP approach remains primarily ecosystem-effect driven. Ironically, the cost-benefit studies produced for UNECE have been dominated by health effects, reflecting the way in which the economics had been developed outside the UNECE framework. Thus it is useful to reflect a little on how cost-benefit approaches had evolved, almost exclusively in the USA.

The links between air pollution and health had, of course, been known for some considerable time and had driven national legislation, for example in the USA. Works such as those by Lester Lave and Eugene Seskin for the USA (Lave and Seskin, 1977) had been influential in not only generalising the epidemiology of air pollution but also in showing that economic cost figures could be attached to the health impacts. Lave and Seskin's work suggested that health impacts alone justified the abatements costs of air pollution, at least for stationary source air pollution. Costs exceeded benefits for mobile sources. Freeman (1982) reached very similar conclusions: overall, benefits exceeded costs but, disaggregated, stationary source benefits exceeded costs and costs exceeded benefits for mobile sources. Portney (1990) reanalysed the situation for the US Clean Air Act and was far more cautious about the benefit-cost comparison, largely because the costs of control had escalated although, in Portney's view, there was substantial scope for reducing those costs through the use of better policy instruments. Continued reliance on 'best available technology' (BAT) has the unfortunate effect of making pollution control far more expensive than its need be (Pearce, 2000). In a generously funded study of US air pollution policy, US EPA (1997, 1999) analysed retrospective and prospective costs and benefits of air pollution policy and concluded that there were extremely high benefit-cost ratios, e.g. 44 for the central estimate of benefits and costs in the retrospective study (US EPA, 1997) and more modest but very attractive ratios in the prospective study (about 4:1)(US EPA 1999, Portney, 2000). Moreover, EPA regards these as probable underestimates. In turn, the benefits are dominated by health benefits (99% if damage to children's IQ is included). The EPA's analysis has, however, been subjected to very critical analysis (Lutter, 1998; Sieg et al., 2000). The significant feature of all these studies is that, whatever one's view of the individual estimates of benefits and costs, health benefits dominate. Ecosystem impacts tended to be confined to agriculture (crops).

Returning to the LRTAP context, then, the cost-benefit input has been dominated by health effects, whereas the LRTAP process remains largely driven by ecosystem effects. The main explanation for this is that cost-benefit approaches developed originally in the health context and, whereas the UNECE process has sponsored highly original work on the determination and estimation of 'critical loads', only minor resources have been allocated to research on economic benefits. Accordingly, the economics input has had to rely on the available literature, rather than being able to generate primary estimates of benefits. However, it is also the case that measuring ecosystem benefits is difficult. There is widespread agreement that ecosystems can be seen as holistic assets which generate a range of valuable services. Work on estimating the benefits of some of those services is extensive - e.g. recreational value - but on the more generalised life-support functions there is, so far, very little work. Questionnaire-based approaches ('stated preference') probably hold out most promise, but there is also a long way to go in valuing services in terms of what would have to be provided if those services were lost. The difficulty of estimating ecosystem benefits is revealed in the cost-benefit studies that do exist in the Europe-wide context.

2 Europe-wide cost-benefit studies

Health benefits, in the form of reduced premature mortality and reduced morbidity, figure prominently in cost-benefit studies of actual and proposed European Directives on air quality control. Table 2 shows a selection of studies relating to air pollutants and reveals that health benefits account for a minimum of one-third and a maximum of nearly 100 per cent of overall benefits from pollution control. Moreover, in most cases these benefits exceed the costs of control by considerable margins. Health benefits therefore 'drive' positive benefit-cost results, as they do in the USA studies.

The European studies suggest that benefits exceed costs even for scenarios defined in terms of 'maximum technologically feasible reduction' (MFR) of pollutants, i.e. scenarios in which the most pollutant-reducing technologies are used. Such scenarios should be characterised by very high marginal abatement costs at very high levels of pollution reduction, precisely the context where one would expect incremental benefits to be less than incremental costs. While the benefit cost ratio does appear to fall for such scenarios relative to other more modest abatement targets, the reduction is not dramatic and benefits continue to exceed costs. Thus, AEA Technology (1999) finds a benefit cost ratio of 2.17 for a MFR scenario, compared to 2.87 for practical targets based on the relevant Protocol. The incremental benefit cost ratio of going from Protocol targets to 'MFR' targets is 1.6. It is possible, of course, that such ratios are correct. But, given that ecosystem benefits are generally excluded from the analyses, it is possible to argue that 'true' benefit-cost ratios are higher still[1]. If so, there is an uneasy tension between these very high benefit-cost ratios and the fact that they remain very high even when the limits of technology are being introduced.

Table 2Health benefits as a percentage of overall benefits in recent cost-benefit studies

Note to Table 2: we have selected results using VOSL (value of statistical life) rather than 'VOLY' (value of a life year) since the latter are not correctly estimated in the studies that also provide VOLY results. See text for discussion.

3 Why are health benefits high?

It would not be correct to say that there is any consensus on the scale of economic health benefits from air pollution control. There are those who think the current studies are 'about right', those who think they seriously overstate benefits, and those who would argue that they seriously understate benefits. The underlying equation in all of the studies is simply stated as:

Hij = bij.Vj.P…[1]

where

Hij= the health effect, j, from pollutant i, aggregated across the relevant population

bij= the dose-response coefficient relating pollutant i to effect j.

Vj= the willingness to pay (accept) to avoid (tolerate) the health effect.

P = population at risk.

It follows that disputes about the scale of health benefits must reflect differences of view about the epidemiology (the dose-response coefficients), the 'unit values' applied to health (the willingness to pay), the population at risk, or some combination of these factors. This symposium is concerned with the first two, although it should not be concluded that estimating the population at risk is a controversy-free area. Since we have expert papers on the issues of epidemiology and valuation, the rest of this paper is concerned only with 'signalling' the main sources of the debate. It turns out that the economic issue of what monetary value can be applied to air pollution health effects is intimately connected with the epidemiology.

3.1 Acute effects

One of the surprising features of the epidemiological literature is that there is limited evidence of what the expected gain in lifetime actually is from risk reductions, or, conversely, what period of life is 'lost' due to exposure to air pollution. We know that acute air pollution has a 'harvesting' effect, i.e. it tends to have its major effects on older people. Two issues arise: what is the period of life lost by this group, and how would that period of life lost be valued? We shall hear later some evidence that suggests the period life lost, i.e. the time at which death would otherwise occur, could be a matter of days only (see David Maddison's presentation and Maddison (2000)). If this is true, there is an issue for economic valuation, namely, what is the correct notion of economic value to apply and what are the implications for policy? If lifetime foregone is a matter of days only, it is arguable that policy should not be aimed at removing acute episodes of pollution. On the other hand, those at risk from acute episodes may have a high willingness to pay to reduce those episodes. That high willingness to pay may reflect their perception of the effect of the increased risk (e.g. they may believe the effect has a greater impact on life expectancy), or they may be aware of the limited impact and may still value it highly. It clearly matters what the perception of the age group at risk actually is. The available evidence suggests that individuals' willingness to pay to avoid health risks declines with age, and we shall hear some of that evidence in the symposium. There are also some questions about which air pollutants should be targeted if we do agree that policy measures need to be taken. There is a widespread consensus that the smaller particles carry the highest cost burden but it is not always clear that the epidemiology has sorted out the interactive effects of multi-pollutant contexts. So long as pollutants are correlated, then a cautious approach would involve selecting just one 'representative' pollutant and avoid the attempt, made in most cost-benefit studies, to aggregate individual effects from different pollutants. It is possible that the process of adding up the effects of different pollutants may have contributed to the large scale of the benefits found in cost-benefit studies.

3.2 Chronic effects

In the context of chronic air pollution impacts, we know that if they are included in the cost-benefit studies they have the effect of greatly increasing the health benefits from air pollution control. Acute pollution episodes obviously risk omitting chronic effects, i.e. reduced life expectancy due to the health effects of fairly continuous exposure over long periods of time. To date we still have only a handful of epidemiological studies that permit chronic effects to be estimated (Pope et al, 1995. For reference to some more recent work see Künzli et al. 2000). As we shall hear later, Künzli et al. (2000) suggest that, using the relationships established in these studies, there is a significant impact of air pollution on chronic mortality. Maddison (1998) applies the same US methodology to the UK for an hypothetical wholesale elimination of particulate matter, and estimates that the change in the conditional life expectancy of the 80+ age group is 1.1 months; that for the 70-79 age group is 2.1 months, and the 60-69 age group is 3.0 months. Rabl (2000), again using the same studies, estimates that a 50-70% reduction in particulate matter (PM2.5)(about 10 ug/m3) would decrease loss of life expectancy in the European Union and the USA by around 6 months.

The results of these studies thus appear to be consistent in suggesting that life expectancy is reduced by a matter of months, but this should not be too surprising since the same methodology, based on the Pope et al . (1995) study, is being used. What does seem clear, however, is that if these are the likely health effects then what needs to be valued in economic terms of the periods of life expectancy lost. This raises the issue of whether we should be using the 'value of a statistical life' (VOSL) or the value of a life year lost (VOLY).

4Valuing 'lives'

A value of statistical life is an aggregated version of individuals' willingness to pay (WTP) to reduce a risk. In its simplest form

where ΣWTPi = sum of individual WTPs for the change in risk over N individuals exposed to the risk;

Δr = the change in risk;

ΣiΔri = number of statistical lives gained (lost).

A VOLY on the other hand is the sum that an individual is WTP to extend life expectancy by one year. The two values - VOSL and VOLY - should bear some relationship since the person at risk must have some idea of remaining life expectancy. In expressing a WTP to reduce risk, then, they should be accounting for the remaining life period available to them.

One approach to estimating the VOLY is to regard it as the annuity which when discounted over the remaining life span of the individual at risk would equal the estimate of VOSL. Thus, if the VOSL of, say, £1.5 million relates to traffic accidents where the mean age of those involved in fatal accidents is such that the average remaining life expectancy would have been 40 years, then

VOLY = VOSL/A

where A = [1-(1+r)-n]/r, n is years of expected life remaining and r is the discount rate[2].

These VOLY numbers can then be used to produce a revised VOSL allowing for age. At age 60, for example, suppose life expectancy is 15 years. The VOSL(60) is then given by

VOSL(60) = VOLY/(1+r)T-60

where T is life expectancy. In the case indicated, this would be, at a 1% discount rate and a 'standard' VOSL of Є1 million:

VOSL(60) = (30,460).(13.87) = Є422,480.

The result is that the age-related VOSL declines with age and this appears to accord with the findings of the economic valuation literature that WTP declines with age, at least beyond a point of around 70 years of age.

There are several reasons for doubting the usefulness of the VOLY approach when it is based on a VOSL. First, it is unclear that this approach to VOLYs accounts for the value that individuals place on time itself. Arguably, the less time left, the higher the WTP for that remaining unit of time. This may account for the finding that WTP does not vary much with age until around 70. Second, the VOLY 'model' produces a WTP estimate that declines with age over all ages, and this is again not consistent with what we know about WTP and age.