76
M. PARSONAGE AND H . NEUBURGER-J. CAIRNS
Methods of economic evaluation of health care programmes. Oxford, Oxford University Press, 1987. 1 1 . Carr-Hill, R. QuALit Y control: a sensitivity analysis of QALYs. Paper presented at Health Economics Study Group, Brunel University, 1988. 12. Williams, A. Economics of coronary bypass grafting. British Medical Journal. 1985; 291: 326. 13. Gudex, C . QALYs and their use by the health ser-
vice. Discussion Paper 20. Centre for Health Economics, York, 1986. 14. DHSS. Breast cancer screening: the Forrest report. London, HMSO, 1986. 15. Oster, G., Huse, D., Delea, T. and Colditz, G. Cost-effectiveness of nicotine gum as an adjunct to physicians advice. Journal of the American Medical Association. 1986; 256: 1315.
DISCOUNTING AND HEALTH BENEFITS: ANOTHER PERSPECTIVE JOHN CAIRNS Health Economics Research Unit, University of Aberdeen
SUMMARY This paper reviews the argument advanced by Parsonage and Neuburger that the non-monetary benefits of health programmes should be discounted at a lower rate than that used for financial flows. The conceptual issues raised in that paper are discussed and others, such as the tradability of non-monetary benefits and the link between individual and social discount rates, are introduced. The collection and assessment of more evidence is needed before Parsonage and Neuburger’s proposition can be supported. KEY
WORDS-Discounting, time preference, health programmes, non-monetary benefits.
Parsonage and Neuburger (henceforth PN) argue that non-monetary effects should be discounted at a lower rate than that used for financial flows. They recommend, in particular, that health benefits, such as Quality Adjusted Life Years (QALYs), be aggregated with no allowance being made for their differential timing. PN present two types of argument in support of a zero discount rate for health benefits: They argue the case for zero discounting by considering the sources of time preference and arguing that in the context of future health benefits they are unimportant. 2. They reappraise the traditional arguments against departures from a common rate of discount for all effects (most commonly associated with Keeler and Cretin*). 1.
Their discussion is lent an added piquancy by the disclosure, at a recent health economics conference, that the Department of Health and the Treasury have agreed as to the desirability of such
a change in evaluative procedures (Parsonage and Neuburger3). TIME PREFERENCE AND FUTURE HEALTH BENEFITS PN’s discussion emphasises the sources of positive time preference and claims that they are not applicable in the QALY context. They suggest in the section on QALYs and Income ‘that the marginal utility of health improvements has for practical purposes a negligible income elasticity’ and that ‘there seems to be no good prima facie case for saying that a life year saved in a wealthier generation is worth either less or more than a life year saved in the present generation.’ It is not clear whether this latter statement is an empirical or an ethical one. The choice between present and future QALYs will generally also be a choice between beneficiaries A and B. If it is an empirical statement it overlooks the standard argument that
DISCOUNTING AND HEALTH BENEFITS
the opportunity cost of saving a current life year may be higher than that of saving a future life year. As a result, in equilibrium, the value of a current life year saved should be worth, at the margin, more than the value of a future life year saved. PN note that the empirical estimates of time preference over future health are not positive. The evidence on pure time preference and on the income elasticity of the marginal utility of health improvements that they are able to refer to is relatively weak. Their review of the empirical literature is rather selective. In particular, they ignore a series of econometric studies of individual behaviour (see, for example, Viscusi and Moore4 and Lawrance’). If we can assign monetary values to the nonmonetary effects which reflect the appropriate shadow prices, the arguments for not using a common discount rate would seem to disappear. This might be taken to imply that the rationale for a zero discount rate for QALYs is that it furnishes a closer approximation to the true stream of relative value. In other words the summation of the adjusted QALY stream is closer to the summation of the QALY times shadow price times discount factor than is the discounted stream of QALYs (the summation of QALY times discount factor). Such reasoning is hinted at by PN when they claim later in the section on QALYs and Income ‘If QALYs were measured in monetary terms, their value would rise over time in line with real income.’ It is not the non-monetary nature of the effects as such which is important but rather the degree of confidence we have in our ability to identify appropriate shadow prices. Thus in the case of transportation investments rather than simply adding up the (undiscounted) time savings, we introduce explicit valuations of time and discount the resulting monetary valuations of the time savings, using the same discount rate as for all the monetary effects of the investment. Whichever line of enquiry is followed a common conclusion is reached, namely, that the choice between uniform or differential discounting is largely an empirical one. This conclusion is firmly endorsed by Olsen. The relatively small number of researchers who have questioned the prevailing practice of discounting health benefits have emphasised the question of the tradability of health both with wealth and with future health (see for example
77
Goodin’). This potentially fruitful way to address the problem is not investigated by PN. ETERNAL DELAY Popular wisdom has it that to use a lower discount rate for health benefits as compared to costs has certain rather undesirable consequences, namely that projects will always improve in evaluative terms if they are delayed. PN challenge the applicability of this alarming finding. Their primary criticism is that Keeler and Cretin’ are applying criteria inappropriately. Let us concentrate on the case where there is a budget constraint or at any rate where not all projects can be undertaken even though they all have positive net present values or, in a costeffectiveness context, all meet acceptable standards of cost-effectiveness. This is the situation which occurs in practice. PN’s advocacy of the cost per QALY as a criterion, for determining whether the project goes ahead now, is mistaken in that the cost per QALY gives no information as to the capital requirements of the investment. As a result it is an inadequate guide to decision-making in situations of capital rationing. But PN are correct when they argue that it is not relevant to ask what effect differential timing has on the benefit-cost ratio of the project. Keeler and Cretin’ assume that an improvement in the benefit-cost ratio for a project resulting from delaying its start would be adequate grounds for postponing the project. Under conditions of capital rationing (with no prospect of carrying funds over from one year to another) and given that the objective is to maximise health benefits from a given budget, decisions with respect to the timing of projects should depend on the opportunity cost in terms of other projects foregone. Assume that if project m is postponed until the future time period this will displace some other future project g and will permit an additional current project s. Whether or not postponement of project m is a good idea will depend on the overall change in health benefits (not just coming directly from m but also from projects g and s). Thus the cost per QALY or benefit-cost ratio information on project m starting now or in the future is not adequate to enable a decision over timing to be taken. A higher benefit-cost ratio for project m in the future might suggest postponement, but if project
NbUBUKtibK-J.
s which replaces it now is inferior to the displaced project g, overall health benefits may be reduced. A further quite distinct reason for not endorsing cost per QALY measures as the sole or primary criterion arises from a concern as to the adequacy of single cost per QALY figures for all scales of investment, for different locations and so on. It is time to admit confidence intervals to our league tables.
EFFECT OF ZERO DISCOUNT RATE FOR BENEFITS The authors argue that changing the discount rate for QALYs will have no effect on the level of investment in health but rather will alter the relative cost-effectiveness of different procedures. They provide an illustration which shows procedures, where the benefits are fairly long-lasting, moving up the QALY league table. But the overall attractiveness of the marginal investment will probably be increased as a result of the use of a zero discount rate for health benefits. It seems unlikely that the projects which were previously just rejected are not now acceptable. Of course this does not necessarily imply that the appropriate budget will be increased, but it does not seem unreasonable to suggest that some upward pressure will be exerted on the budget.
IMPLICATIONS FOR OTHER AREAS OF PUBLIC EXPENDITURE Are the arguments which have been presented general with respect to all non-monetary benefits (and costs?) or more specifically to health benefits? If the difficulties involved in setting shadow prices are central to the argument for a zero discount rate the implications for other areas will depend on the view taken as to whether there are many other areas of public expenditure raising comparable difficulties. Some environmental benefits would appear to be comparable and thus on this reasoning might also be candidates for zero discounting. The difficulty of shadow pricing may be closely linked to the question of the tradability of the good in question. It is easier to devise shadow prices for other non-monetary effects such as time savings which are more readily tradable (faster
CAIKNS
modes of transport, labour saving devices, marginal labour supply decisions etc.).
CONCLUSION Economists must decide whether the theoretical arguments for differential discounting, or for at least challenging uniform discounting, are sufficiently convincing for us to accept that we face an empirical question. One element to consider is whether the eternal delay argument remains persuasive. If it is accepted that we face an empirical question we should consider carefully what relevant evidence is available, how good it is, what further evidence is required and how we might best obtain it. Some but not very much work is underway. But the problems faced are formidable (see for example Cairns8). Not the least of which are fundamental questions as to the adequacy of the economist’s conventional model of time preference (see Ainslie9 and Loewenstein and Prelec lo). The collection and evaluation of rather more evidence is warranted before most health economists will share PN’s radical conclusion that ‘In practice, for most purposes, it is appropriate to use a zero discount rate for future health benefits’. Although this bold assertion appears in their introduction, such caution ultimately tempers even PN’s radicalism when, in the section on uncertainty, they advocate the use of different discount rates ‘particularly where the benefits of a proposed policy stretch well into the future’. Finally, there is an important missing step in the argument, not addressed in this paper and only barely by PN’: the one that takes us from individual preferences over future health states to an appropriate social discount rate for use in resource allocation in the NHS.
REFERENCES 1. Parsonage, M. and Neuburger, H. Discounting and health benefits. Health Economics. 1992; 1: 2. Keeler, E. B. and Cretin, S . Discounting of life-
saving and other non-monetary effects. Management Science. 1983; 29: 300-6. 3. Parsonage, M. and Neuburger H. Discounting and QALYs. Paper presented at the Health Economists’ Study Group, Aberdeen, 1991.
DISCOUNTING AND HEALTH BENEFITS
4. Viscusi, W. K . and Moore, M. J. Rates of time preference and valuations of the duration of life. Journal of Public Economics. 1989; 38: 297-317. 5. Lawrance, E. C. Poverty and the rate of time preference: evidence from panel data. Journal of Political Economy. 1991; 99: 54-77. 6. Olsen, J. P. Is the time preference rate for health different? Paper presented at the Second World Congress on Health Economics, Zurich, 1990. 7. Goodin, R. E. Discounting discounting. Journal of Public Policy. 1982; 2: 53-72.
79
8. Cairns, J. A. Health, wealth and time preference. Discussion Paper 07/91. Health Economics Research .Unit, Aberdeen, 1991. 9. Ainslie, G. Derivation of ‘rational’ economic behavior from hyperbolic discount curves. American Economic Review. 1991; 81: 334-40. 10. Loewenstein, G . and Prelec, D. Negative time preference. American Economic Review. 1991; 81: 347-52.
M. PARSONAGE AND H . NEUBURGER-J. CAIRNS
Methods of economic evaluation of health care programmes. Oxford, Oxford University Press, 1987. 1 1 . Carr-Hill, R. QuALit Y control: a sensitivity analysis of QALYs. Paper presented at Health Economics Study Group, Brunel University, 1988. 12. Williams, A. Economics of coronary bypass grafting. British Medical Journal. 1985; 291: 326. 13. Gudex, C . QALYs and their use by the health ser-
vice. Discussion Paper 20. Centre for Health Economics, York, 1986. 14. DHSS. Breast cancer screening: the Forrest report. London, HMSO, 1986. 15. Oster, G., Huse, D., Delea, T. and Colditz, G. Cost-effectiveness of nicotine gum as an adjunct to physicians advice. Journal of the American Medical Association. 1986; 256: 1315.
DISCOUNTING AND HEALTH BENEFITS: ANOTHER PERSPECTIVE JOHN CAIRNS Health Economics Research Unit, University of Aberdeen
SUMMARY This paper reviews the argument advanced by Parsonage and Neuburger that the non-monetary benefits of health programmes should be discounted at a lower rate than that used for financial flows. The conceptual issues raised in that paper are discussed and others, such as the tradability of non-monetary benefits and the link between individual and social discount rates, are introduced. The collection and assessment of more evidence is needed before Parsonage and Neuburger’s proposition can be supported. KEY
WORDS-Discounting, time preference, health programmes, non-monetary benefits.
Parsonage and Neuburger (henceforth PN) argue that non-monetary effects should be discounted at a lower rate than that used for financial flows. They recommend, in particular, that health benefits, such as Quality Adjusted Life Years (QALYs), be aggregated with no allowance being made for their differential timing. PN present two types of argument in support of a zero discount rate for health benefits: They argue the case for zero discounting by considering the sources of time preference and arguing that in the context of future health benefits they are unimportant. 2. They reappraise the traditional arguments against departures from a common rate of discount for all effects (most commonly associated with Keeler and Cretin*). 1.
Their discussion is lent an added piquancy by the disclosure, at a recent health economics conference, that the Department of Health and the Treasury have agreed as to the desirability of such
a change in evaluative procedures (Parsonage and Neuburger3). TIME PREFERENCE AND FUTURE HEALTH BENEFITS PN’s discussion emphasises the sources of positive time preference and claims that they are not applicable in the QALY context. They suggest in the section on QALYs and Income ‘that the marginal utility of health improvements has for practical purposes a negligible income elasticity’ and that ‘there seems to be no good prima facie case for saying that a life year saved in a wealthier generation is worth either less or more than a life year saved in the present generation.’ It is not clear whether this latter statement is an empirical or an ethical one. The choice between present and future QALYs will generally also be a choice between beneficiaries A and B. If it is an empirical statement it overlooks the standard argument that
DISCOUNTING AND HEALTH BENEFITS
the opportunity cost of saving a current life year may be higher than that of saving a future life year. As a result, in equilibrium, the value of a current life year saved should be worth, at the margin, more than the value of a future life year saved. PN note that the empirical estimates of time preference over future health are not positive. The evidence on pure time preference and on the income elasticity of the marginal utility of health improvements that they are able to refer to is relatively weak. Their review of the empirical literature is rather selective. In particular, they ignore a series of econometric studies of individual behaviour (see, for example, Viscusi and Moore4 and Lawrance’). If we can assign monetary values to the nonmonetary effects which reflect the appropriate shadow prices, the arguments for not using a common discount rate would seem to disappear. This might be taken to imply that the rationale for a zero discount rate for QALYs is that it furnishes a closer approximation to the true stream of relative value. In other words the summation of the adjusted QALY stream is closer to the summation of the QALY times shadow price times discount factor than is the discounted stream of QALYs (the summation of QALY times discount factor). Such reasoning is hinted at by PN when they claim later in the section on QALYs and Income ‘If QALYs were measured in monetary terms, their value would rise over time in line with real income.’ It is not the non-monetary nature of the effects as such which is important but rather the degree of confidence we have in our ability to identify appropriate shadow prices. Thus in the case of transportation investments rather than simply adding up the (undiscounted) time savings, we introduce explicit valuations of time and discount the resulting monetary valuations of the time savings, using the same discount rate as for all the monetary effects of the investment. Whichever line of enquiry is followed a common conclusion is reached, namely, that the choice between uniform or differential discounting is largely an empirical one. This conclusion is firmly endorsed by Olsen. The relatively small number of researchers who have questioned the prevailing practice of discounting health benefits have emphasised the question of the tradability of health both with wealth and with future health (see for example
77
Goodin’). This potentially fruitful way to address the problem is not investigated by PN. ETERNAL DELAY Popular wisdom has it that to use a lower discount rate for health benefits as compared to costs has certain rather undesirable consequences, namely that projects will always improve in evaluative terms if they are delayed. PN challenge the applicability of this alarming finding. Their primary criticism is that Keeler and Cretin’ are applying criteria inappropriately. Let us concentrate on the case where there is a budget constraint or at any rate where not all projects can be undertaken even though they all have positive net present values or, in a costeffectiveness context, all meet acceptable standards of cost-effectiveness. This is the situation which occurs in practice. PN’s advocacy of the cost per QALY as a criterion, for determining whether the project goes ahead now, is mistaken in that the cost per QALY gives no information as to the capital requirements of the investment. As a result it is an inadequate guide to decision-making in situations of capital rationing. But PN are correct when they argue that it is not relevant to ask what effect differential timing has on the benefit-cost ratio of the project. Keeler and Cretin’ assume that an improvement in the benefit-cost ratio for a project resulting from delaying its start would be adequate grounds for postponing the project. Under conditions of capital rationing (with no prospect of carrying funds over from one year to another) and given that the objective is to maximise health benefits from a given budget, decisions with respect to the timing of projects should depend on the opportunity cost in terms of other projects foregone. Assume that if project m is postponed until the future time period this will displace some other future project g and will permit an additional current project s. Whether or not postponement of project m is a good idea will depend on the overall change in health benefits (not just coming directly from m but also from projects g and s). Thus the cost per QALY or benefit-cost ratio information on project m starting now or in the future is not adequate to enable a decision over timing to be taken. A higher benefit-cost ratio for project m in the future might suggest postponement, but if project
NbUBUKtibK-J.
s which replaces it now is inferior to the displaced project g, overall health benefits may be reduced. A further quite distinct reason for not endorsing cost per QALY measures as the sole or primary criterion arises from a concern as to the adequacy of single cost per QALY figures for all scales of investment, for different locations and so on. It is time to admit confidence intervals to our league tables.
EFFECT OF ZERO DISCOUNT RATE FOR BENEFITS The authors argue that changing the discount rate for QALYs will have no effect on the level of investment in health but rather will alter the relative cost-effectiveness of different procedures. They provide an illustration which shows procedures, where the benefits are fairly long-lasting, moving up the QALY league table. But the overall attractiveness of the marginal investment will probably be increased as a result of the use of a zero discount rate for health benefits. It seems unlikely that the projects which were previously just rejected are not now acceptable. Of course this does not necessarily imply that the appropriate budget will be increased, but it does not seem unreasonable to suggest that some upward pressure will be exerted on the budget.
IMPLICATIONS FOR OTHER AREAS OF PUBLIC EXPENDITURE Are the arguments which have been presented general with respect to all non-monetary benefits (and costs?) or more specifically to health benefits? If the difficulties involved in setting shadow prices are central to the argument for a zero discount rate the implications for other areas will depend on the view taken as to whether there are many other areas of public expenditure raising comparable difficulties. Some environmental benefits would appear to be comparable and thus on this reasoning might also be candidates for zero discounting. The difficulty of shadow pricing may be closely linked to the question of the tradability of the good in question. It is easier to devise shadow prices for other non-monetary effects such as time savings which are more readily tradable (faster
CAIKNS
modes of transport, labour saving devices, marginal labour supply decisions etc.).
CONCLUSION Economists must decide whether the theoretical arguments for differential discounting, or for at least challenging uniform discounting, are sufficiently convincing for us to accept that we face an empirical question. One element to consider is whether the eternal delay argument remains persuasive. If it is accepted that we face an empirical question we should consider carefully what relevant evidence is available, how good it is, what further evidence is required and how we might best obtain it. Some but not very much work is underway. But the problems faced are formidable (see for example Cairns8). Not the least of which are fundamental questions as to the adequacy of the economist’s conventional model of time preference (see Ainslie9 and Loewenstein and Prelec lo). The collection and evaluation of rather more evidence is warranted before most health economists will share PN’s radical conclusion that ‘In practice, for most purposes, it is appropriate to use a zero discount rate for future health benefits’. Although this bold assertion appears in their introduction, such caution ultimately tempers even PN’s radicalism when, in the section on uncertainty, they advocate the use of different discount rates ‘particularly where the benefits of a proposed policy stretch well into the future’. Finally, there is an important missing step in the argument, not addressed in this paper and only barely by PN’: the one that takes us from individual preferences over future health states to an appropriate social discount rate for use in resource allocation in the NHS.
REFERENCES 1. Parsonage, M. and Neuburger, H. Discounting and health benefits. Health Economics. 1992; 1: 2. Keeler, E. B. and Cretin, S . Discounting of life-
saving and other non-monetary effects. Management Science. 1983; 29: 300-6. 3. Parsonage, M. and Neuburger H. Discounting and QALYs. Paper presented at the Health Economists’ Study Group, Aberdeen, 1991.
DISCOUNTING AND HEALTH BENEFITS
4. Viscusi, W. K . and Moore, M. J. Rates of time preference and valuations of the duration of life. Journal of Public Economics. 1989; 38: 297-317. 5. Lawrance, E. C. Poverty and the rate of time preference: evidence from panel data. Journal of Political Economy. 1991; 99: 54-77. 6. Olsen, J. P. Is the time preference rate for health different? Paper presented at the Second World Congress on Health Economics, Zurich, 1990. 7. Goodin, R. E. Discounting discounting. Journal of Public Policy. 1982; 2: 53-72.
79
8. Cairns, J. A. Health, wealth and time preference. Discussion Paper 07/91. Health Economics Research .Unit, Aberdeen, 1991. 9. Ainslie, G. Derivation of ‘rational’ economic behavior from hyperbolic discount curves. American Economic Review. 1991; 81: 334-40. 10. Loewenstein, G . and Prelec, D. Negative time preference. American Economic Review. 1991; 81: 347-52.
