Abstract
The debate on the appropriate discounting of future outcomes has a long history, with normative deliberation identifying several different considerations. It is suggested here that only two considerations are appropriate in the discounting of health outcomes. First, discounting ought to account for the possibility of future catastrophic events. Second, it is arguably legitimate for democratically elected governments to weight disproportionately their current constituency over future constituencies. It is adjudged that accounting for both of these factors should lead to a maximum annual discount rate of 0.5%, substantially lower than the rates currently recommended by public authorities.
Introduction
In the assessment of policy interventions, discounting future benefits is a subject of longstanding debate. Several great figures in intellectual history thought that valuing distant outcomes differently to more proximate ones is a psychological error that ought to play no part in normative reasoning. Adam Smith, for instance, wrote in 1759 that the ‘pleasure which we are to enjoy ten years hence, interests us so little in comparison with that which we may enjoy today’ … but the impartial ‘spectator does not feel the solicitations of our present appetites. To him the pleasure which we are to enjoy a week hence, or a year hence, is just as interesting as that which we are to enjoy this moment.’ 1
Indeed, in terms of the generation of welfare, it is not obvious why a benefit experienced in the future should be valued differently from that same benefit experienced today, but in policy appraisal discounting future outcomes is broadly perceived as good practice. The National Institute for Health and Clinical Excellence (NICE) in the UK, for example, currently recommends a discount rate of 3.5% be applied to health outcomes when assessing health care interventions. It is likely that the discount rate used by NICE was informed by the UK Treasury which sets out a framework for public policy appraisal. 2 Explicitly, it states that the ‘discount rate is used to convert all costs and benefits to “present values”, so that they can be compared. The recommended discount rate is 3.5%.’ Several experts believe that UK government departments do not, in general, discount health benefits at 3.5% (Michael Parsonage, Michael Spackman, personal communications) and that, therefore, NICE is an anomaly with respect to this practice. Even if this is the case, the breadth of NICE's remit renders its recommendations of considerable import and to discount at such a rate weighs heavily against our future health and that of future generations. For instance, a discount rate of 3.5% implies that a gain of one year of life now equates to a gain of approximately four years 40 years from now, and about 30 years 100 years from now. For policies that offer future benefits, discounting at this level is problematic.
My aim is to offer a challenge to discounting health outcomes at such a high rate and to suggest an alternative based on normative reasoning. Articles on discounting tend to be written with the use of formal terminology and algebraic proofs. As a consequence, they are read by a quite narrow grouping of specialists. Discounting, however, is largely a matter of ethics and judgement, so the relevant components of discounting and the parameters attached to those components ought to be of broad interest.
Causes of excessive discounting
According to Drummond and colleagues, a number of analysts reporting early cost-effectiveness analyses in the New England Journal of Medicine in the 1970s used a 5% discount rate on outcomes. 3 No real justification was given for this rate but from this apparent example of first mover advantage, 5% became standard practice in health economic evaluation for many years. In their influential work, Lipscomb and colleagues recommended that results using two rates be reported as standard in health economic analyses: 3% and 5%. 4 They recommended 3% by arguing that this best reflects the shadow price of capital (which might be more relevant for discounting costs than health outcomes) and 5% because that was the rate used conventionally. Since such leading figures in health economic evaluation advocated discounting of at least 3%, it is not hard to understand why such high rates of discounting on health outcomes prevailed.
As to the 3.5% discount rate recommended by NICE, as aforementioned it would be too much of a coincidence to say that this was not informed by the UK Treasury. Such a Treasury recommendation does have a normative foundation, which is worth examining in relation to health and welfare.
The normative components of a health outcomes discount rate
The UK Treasury's figure of 3.5% is composed of: 2% to represent the combined effect of the elasticity of the marginal utility of consumption and the growth in per capita consumption; a pure time preference rate of 0.5%; and a catastrophic risk premium of 1%. The first two factors were posited by Ramsey in 1931, even though he acknowledged that discounting enjoyments ‘is ethically indefensible and arises merely from the weakness of the imagination.’ 5 Nonetheless, let us examine each of the three components in turn.
Marginal utility of consumption and growth in per capita consumption
If people are better off in the future than they are now, a unit increase in the consumption of most goods may be worth less, due to declining marginal utility. However, when health is the good the situation becomes more complex. It is possible to surmise that if health levels improve in the future, then each additional increment of health will indeed be worth less due to declining marginal utility. 6 Moreover, if health levels are expected to improve over time, then it may be deemed appropriate to discount future health for the sake of intergenerational equity. However, given such factors as increasing obesity levels and resistance to antibiotics, indefinite improving average population health cannot be easily assumed.
Even if health levels improve, people may attach an increasing marginal utility to each health increment if their material circumstances improve concurrently because a higher income may help them to squeeze more enjoyment out of each unit of health. Similarly, improved health may make it feasible for people to enjoy better each additional unit of other consumption goods, although those goods may legitimately be prone to declining marginal utility, rendering it ambiguous in these circumstances as to whether the future marginal utility of those goods is positive or negative.
Claxton and colleagues suggest that the value of health might be growing in consumption terms because the health of the population is expected to grow more slowly than its consumption of other goods and services. 7 Therefore, over time, the welfare gain from better health will increase relative to the welfare gain from the increase in other consumption and thus the social value of health in terms of overall consumption will increase. Thus, at the margin, people will invest more in health than in other goods over time because the marginal utility of health relative to the marginal utility of other goods will increase. However, this does not suggest that the absolute (as opposed to relative) marginal utility of health will increase. As noted above, it could plausibly decrease and thus, if the interest is the effect on wellbeing, the declining value of health (if this were to occur), rather than its increasing relative value, is of relevance and would have a positive impact on the discount rate.
Overall, we do not know whether per capita health will grow, nor whether the marginal utility of health will increase, decrease or stay the same. It would seem prudent to remain neutral and assume that the discount factor for these considerations is zero. This position is reinforced by Parfit's claim that constant marginal utility should be assumed for health-related outcomes because they are of equal value whenever they occur. 8 Broome concurs with respect to what he calls ‘constant wellbeing goods’, such as life and health, stating that ‘on average, saving one person's life in one hundred years will presumably add just as much well-being to the world as saving one person's life now.’ 9
Pure time preference
A second reason for valuing a future good less than in the present is that of pure time preference, i.e. people prefer to have things sooner rather than later. There have been numerous studies conducted to uncover pure time preference rates over a range of outcomes. Many of these were reviewed by Frederick and colleagues, who found that with outcomes such as money, pain, health and life years, implicit annual time preference rates vary enormously, from −6% to infinity, although high discount rates predominate. 10 They imply that it is inadvisable to use these rates in policy analysis because preferences may be subject to visceral influences, such as intemperance, impatience, hunger, desire and cravings that can be categorized as transient fluctuations in tastes. Moreover, although respondents in preference elicitation studies are typically instructed to assume that delayed rewards will be delivered with certainty, it is unclear whether they can accept this assumption, because delay is perhaps unavoidably associated with uncertainty. Further, although preference surveys may imply that respondents demonstrate a declining marginal utility for life years, 6 these studies say nothing about whether distant life years do actually produce less welfare than more proximate ones at the time that the life years are experienced. Frederick and colleagues remark that given all the confounding factors, it is unclear how much of the imputed discount rates from respondent surveys reflect pure time preference, and that it is possible that once the confounders are controlled for, no pure time preference would remain, which is the position taken in this essay. 10
However, although Stern, 11 for example, argues that future generations should, if it were possible, have full representation in current decisions, and thus ought to have the same claim on our ethical attention as current generations, it may be unavoidable for democratic governments to disproportionately reflect to some extent the preferences of individuals who are presently members of the body politic. 12 If it is deemed acceptable to discount for this factor, the rate chosen is necessarily based upon a judgment. It is suggested here that a discount factor of no more than 0.3% be used to account for this consideration, which, on its own, implies that the value that a policy maker attaches to a year of life 100 years hence is roughly 75% of the value of a year of life today.
Catastrophic risk premium
There is a sound basis for placing less value on future goods due to the possibility of a catastrophic event occurring. That is, there is a chance that human beings will not exist, say, 100 or 1,000 years from now. Stern adopted the perspective of a global social planner, and estimated that if a discount factor to account for the uncertainty of existence is set at 0.1%, then this allows for a 10% chance of the extinction of the human race through whatever cause within the next 100 years. 11 Given that humans have survived the past two million years, this chance of extinction appears high. However, some international and country-specific events might occur over the next century that may not kill everyone (pandemics, wars, asteroids, economic crises, etc.), but could affect a large proportion of the population, and may lead to health care services being drawn away from chronic care interventions that would have otherwise generated more health benefit. It seems reasonable and reasoned to set a discount factor of no more than 0.2% to account for the possible occurrence of catastrophic events.
All in all, the position taken here is that the total discount rate on health outcomes should not exceed 0.5% (0.3% due to concern for the preferences of current constituents plus 0.2% to account for the possibility of catastrophic events).
Conclusion
Discounting is value laden, and therefore the rates chosen to account for the above factors will always be based on judgement. Indeed, Paul Samuelson, the father of the discounted utility model, himself commented that several features of the model are completely arbitrary. 13 Through what is intended as a reasoned approach it is proposed here that the total discount rate on health outcomes should not exceed 0.5% per annum. Compared to the current convention of in excess of 3%, a rate of 0.5% would allow for a more legitimately favourable evaluation of policies that impact on our future health and welfare, and that of future generations. It is worth remembering the words of David Hume, who wrote in 1739 that ‘there is no quality in human nature which causes more fatal errors in our conduct than that which leads us to prefer whatever is present to the distant and remote.’ 14