Roland Seminar, may 2005 Preliminaries: How should we think about the future? How should future magnitudes be related to present magnitudes? How do these issues bear on how we should think about the “discount rate” understood as the relative ‘value’ of future and present ‘goods’?
There is a very considerable literature about these questions both in economics and in philosophy. Much of this literature is, in my opinion, hopelessly confused. And this confusion is not just a matter of mutual misunderstanding across disciplinary boundaries – although there is a great deal of that. It is also a matter of a failure within disciplines to make relevant distinctions, so that the whole discussion is immersed in a conceptual fog.
Or at least, so it seems to me. I recognize well enough the possibility that the confusion may be a product of my own mind. In which case the attempts I shall make here to clarify things will simply add to the confusion. But I hope that my simple-minded attempt here will at least draw attention to a number of conceptual distinctions – ones that seem to me to be both crucial and all too rarely observed.
The discussion is nested in a general view about how normative analysis should properly be done, and it is best to declare this view right away. I believe that it is crucial in normative analysis to distinguish between issues of normative desirability and issues of feasibility. Proper normative analysis involves a confrontation of feasibility with desirability considerations. I think both economists and philosophers broadly accept this claim – though they tend to deal with it in different ways.
Economists foreground feasibility considerations, and treat desirability as if it were a second-order matter. They focus all their analytic firepower on the consequences of alternative actions (or policies) and treat it as more or less self-evident what should follow in the wake of those consequences. [So for example Milton Friedman asserts that most disagreements on ideological issues are a matter of disagreements about the consequences of alternative policies, not about basic values. Partha Dasgupta makes essentially the same claim about development issues in a recent (long) paper in Economics and Philosophy. Atkinson’s demonstration that simple utilitarianism and Rawlsian maximin generate almost identical redistributional policies is a further example in the same spirit.]
Moral and political philosophers by contrast operate as if the real issues are ones of differing values. It is true that they doff a perfunctory cap at feasibility considerations – “after all”, they say, “ought implies can”. But often enough the O→C constraint is interpreted to suggest that if we can show that something is in principle desirable then feasibility will just follow in its wake! Effectively, feasibility issues are back-grounded and left to lesser minds of a more practical bent to sort out later.1 Sorting out the traffic between the disciplines on this desirability/feasibility issue is no small matter and goes well beyond the scope of this paper. However, the current exercise might be usefully regarded as an application of a kind of method that I think is useful in this regard. And I need to make a couple of remarks about it.
First, I shall treat the issue of normative decision-making in somewhat similar terms to that of rational agent choice between options in a market-place – with the domain of choice understood as involving objects of normative desirability (say, justice and freedom – or friendship and honesty, or whatever). I shall use “normative indifference curves” as a device for reflecting desirability considerations. And I shall talk in terms of a “feasible set” delineating the “opportunity set” that the world presents to the normatively informed agent. This conceptualization requires a distinction between desirability pro tanto and desirability tout court. Desirability pro tanto is reflected in the normative indifference map: desirability tout court reflects the best outcome, subject to feasibility “constraints”. The O→C constraint is therefore seen as a kind of overall optimization requirement – but not as a restriction on ethical reflection generally, which seeks only to specify desirability pro tanto.
Second, although “feasibility” is a term whose role in ‘choice’ situations (whether moral or ‘economic’) economists well understand, it is an unfortunate term in some ways. It carries with it a certain trumping status. After all, if “cannot implies ought not”, then feasibility considerations will do a lot of work in normative analysis and perhaps all the work that can be done. But we ought to accept that even the best-informed social analyst does not know enough to divide the world into feasible and infeasible sets. Perhaps a better term would be “plausibility”. In any event, whatever one’s choice of vocabulary, we ought to accept that ‘feasibility’ is not a category but a metric – it comes in degrees. Some things are clearly infeasible; some are almost certainly feasible. And most possibilities – or at least many of interest -- lie somewhere in between. This fact too is a severe complication because it means that the neat dichotomy between feasibility and desirability, characteristic of standard representations of choice in economics, is no longer sustainable. Here though I am going to finesse this complication, important though it may be. For reasons of conceptual clarity I shall suppose that desirability and feasibility considerations are neatly separable, and see how far we can go in this simplified and highly abstracted setting.
I should perhaps add, before proceeding that the failure to distinguish desirability from feasibility considerations is one of the more common and confusing aspects of the discussion among economists. More of this anon.
Two Useful Distinctions: In his interesting essay on “Discounting the Future” in Ethics out of Economics, John Broome makes a distinction at the outset between “plans made by an individual on her own behalf” and “plans made on behalf of the public as a whole”. In fact, there seem to be two distinctions at stake here – and both of them are worth emphasizing. They are:
the distinction between individual and collective action
the distinction between the inter-temporal allocation of benefits within a life (call this the inter-temporal question, T) and the allocation of benefits between lives across generations (call this the inter-generational question, G)
The first of these distinctions is familiar, but it is differently understood between scholars of different stripes. Public finance economists tend to understand it in one way; public choice theorists in another; moral philosophers in yet another; and so on. In relation to the discount rate, for example, it is worthwhile keeping a sharp distinction between two questions:
the ‘preference’ question – what preferences should I have as between present and future persons (or consumption)?
the policy question -- what policies towards future persons (or consumption) should I vote for?
I think much of the misunderstanding between philosophers and economists on discounting matters arises from a confusion over which of these questions one is addressing. I shall try to sustain that claim below. Much more could be said about the individual/collective distinction: much more needs to be said. Indeed, there is hardly an aspect of normative political theory that does not engage it in one form or another. But here is not the place to pursue the issue – or even hint at its complexity.
One thing ought though be added – which is that inter-temporal and inter-generational outcomes emerge from a combination of collectively determined policy and private individual action. The interaction between policy and private action is an issue that should not be overlooked. In particular, some policies that seem attractive may have their effects undone by compensating private action, even by individuals who think that the policies in question “do good things”.2 The second distinction deals specifically with the discount rate, however exactly understood, and deserves special consideration here. There are some philosophers (Derek Parfit the most notable perhaps) who are inclined to the view that we should think of ourselves in future periods as different selves, related to our present self in much the same way as close friends and relatives are related at a point in time. I do not find that view plausible. It seems to me that attitudes to one’s own future are best understood as issues of ‘prudence’, whereas attitudes to persons who will live (and die) after you invoke considerations of ‘justice’. If I fail to save for my old age, I hurt myself (admittedly in the future); if I fail to deal appropriately with future persons, I hurt others. Certainly, my own moral intuitions about these two cases are rather different. Millian liberalism would seem to draw a sharp distinction, and economists are probably disposed to follow Mill in this, at least.
These two distinctions allow us to draw the obvious two by two matrix of possibilities, and is helpful to do this in keeping track of the issues one is addressing. At both individual and collective levels, the distinction between G and T questions arises. I can save for my old age; or I can save for bequest. Policies can be devised for increasing savings for retirement, or increasing the position of future generations. In all cases, increases in savings in the present seem likely to be required to satisfy the relevant end – and policies (or actions) undertaken for one purpose may have implications for the other purpose.
object of choice
Pro Tanto Desirability: The first and perhaps most important issue to be clear about is what the object of ultimate normative concern is exactly. If we are not clear about this, then we are likely quickly to go astray.
I assert that in the discounting context, the object of desirability is the “best” allocation of whatever is to be allocated across time/generations. This at least is common across the T and G issues. This may seem to be stating the obvious, but it has an important implication: namely that the rate of discount as such has no ultimate normative significance whatsoever! The rate of discount is a relative price -- or perhaps a relative value. It tells us the worth of future magnitudes in terms of present magnitudes – but it does not tell us what the allocation of magnitudes across time either is, or ought to be.
[It is interesting to reflect that “just price theory” has virtually disappeared from the study of economics – even in economics normative variant. Just price theory as an account purporting to explain the level of prices has long since been discredited: we now analyse the determination of prices in terms of ‘levels of demand’ and ‘conditions of supply’ across the many actors that participate in the market in question. Conceivably, the attitudes of demanders (actual or potential) to what constitutes a ‘just price’ may exercise some effect on actual prices, but if so this possibility never receives much attention in the classic textbook treatments.
What is more relevant here, though, is that ‘just price’ has almost entirely ceased to be a category in the normative analysis of market operation. In part, this is because the prices of particular goods seems too narrow a category for normative evaluation – what is required is an assessment of the market system as a whole, and the relative prices of goods that emerge from that system are just one aspect of that assessment. But it is, I think, also recognized that the price of a good is not as appropriate a feature of evaluation as the quantity of that good consumed. If the price of X is “inappropriate” the inappropriateness is to be traced to the feature that consumption of X is too high or too low, and this relative to other goods that might have been substituted for X. Or at least, that is the contention I want to investigate – and I shall use the discount rate as the vehicle for that investigation.
In any event, if just price analysis has almost entirely disappeared, there seems to be one notable exception – and this is the case of the discount rate. The question then is whether the exception is appropriate – whether there is something distinctive about the discount rate that justifies its special status as an application of “just price” methods – or whether that status is just a vestige of an analytic method that once enjoyed favour but has long since been discarded (and properly so) in every other context!]
As already suggested, one useful way of formulating pt-desirability is in the form of relevant “ethical indifference (EI) curves” and this is the method I shall use. I shall begin with the IG case, and then later turn to the question of whether (and if so how) the IT case might be different. I shall grossly simplify the IG case by abstracting from differential numbers. Comparison of alternative possible societies with differing populations – like questions of whether numbers of persons saved under alternative actions – raise interesting issues that I think can usefully be finessed. I do not think such issues are intrinsic to IG questions, though I do not doubt that they arise in many cases. In the interests of further simplification, I shall treat the generational choice as one between the interests of the present generation and the next generation – and consider the choice between their respective ‘interests’. That is, the two axes of the EI map I shall draw will be Up and Uf: I take it that the object of normative concern is the ‘appropriate’ weighing up of the interests of these two ‘generations’ and I shall treat each generation as being represented by the average utility level in each.
Now, since the IG question is, at least on its face, an issue of inter-personal justice, where the two “persons” are merely representatives of their birth cohort, I believe it natural to draw on the EI map that is standardly used in the inter-personal justice context. [See Roemer (2004) for an elegant example.] In the standard inter-personal setting, there is an axiom of “anonymity” routinely applied, which states that these indifference curves ought to be symmetric around the 450 line. In other words, in justice terms, (U1,U2) ought to be considered equivalent to (U2,U1). In the IG case, this anonymity principle becomes the “principle of IG-neutrality”, but its geometric representation remains as in the standard justice cases. A specific implication is the requirement that along the 450 line, the slope of any EI curve should be -1 (if the EI curve is continuously differentiable).
Two extremes have become familiar in the relevant literature: simple utilitarianism (where aggregate utility, Up + Uf, is the only object of normative concern); and maximin (where the indifference curves form as right-angled lines with the apex along the 450 line. I assert that neither of these extremes is acceptable as a principle of inter-generational justice. The right principle of IG justice seems to me to be some form of ‘prioritarianism’ of which utilitarianism and maximin are the limiting cases. I think utilitarianism is unacceptable because it implies that in any case where there is a positive rate of return, the present should impoverish itself in order to make for maximal savings and hence maximal consumption in the future. This seems to be the problem that Koopmans identified – though Koopmans seems to have believed that its solution required a systematic preference for the present over the future. In fact, what it requires is no sacrifice of the principle of IG neutrality – just a prioritarian rather than a utilitarian approach.
At the other extreme, maximin seems to me to be inadequate because it gives no relevance at all to inter-temporal rates of return. A very tiny sacrifice by the present generation that will give huge benefits in the future can never be justified, if future generations are going to be better off than the present generation. Obversely, a tiny sacrifice by the future that gives enormous returns in the present cannot be justified if the future is going to be less well-off than the present. On the maximin criterion we can make the benefit-cost ratio as large as we like and it will still be irrelevant – equality of generational outcomes is what justice requires, independent of circumstances, (unless making one generation better off also makes the other better off).
I do not intend here to attempt a lengthy defense of ‘prioritarianism’ – or, as an older generation of economists would have put it, a “paretian social welfare function”. The arguments used in the context of single period inter-personal justice seem to carry over to the case where the persons in question happen to be from different generations. And in any event, those with utilitarian or maximin inclinations can be accommodated within the broad family as limiting extremes.
One clear and simple implication of the prioritarian formulation is that there is no single discount rate that is ideal! It is just a mistake to talk of the discount rate as if the appropriate relation between generations could be captured in a single number. The appropriate social discount rate is a function of the share of total flourishing between generations. Or better put, the “ideal” discount rate and the “ideal” relative position of the generations will be mutually determined. And they will in such a way that the discount rate will “favour the present over the future” only when the present does less well than the future. It will favour the future over the present only when the present is doing better than the future. In the world of pt-desirability, then, there is an inverse relation between discount rate and IG utility shares: positive discounting of the future is equivalent to the future doing better than the present. This is just a statement about the shape of the inter-generational EI-map.
[We should add that in the limiting cases the discount rate is irrelevant. In the maximin case, assuming convex opportunity sets, we should aim at equal positions for all generations, whatever the rate of return. In the utilitarian case, if the rate of return is positive we should allocate everything above subsistence to the future: if the rate of return is negative we should allocate everything to the present.]
So far, I have focused attention on the IG case. The IT case invokes, as I have said, rather different moral intuitions – at least in me. Considerations of prudence certainly require that the individual not consume so profligately in the present that she is destitute in her dotage. They equally require that the agent not imperil her survival in the present by miserly habits of accumulation now in the interests of future opulence. So the EI map for this case is likely to avoid falling below a certain threshold in either period. Beyond that minimal requirement, however, it is not self-evident that there is anything inappropriate in having a ‘bias’ towards either the present or the future (ie that the IT EI curves must have a slope of 450 when present and future consumption are equal). An IT EI map that exhibited a preference for having a splurge sometime (perhaps indifferent between whether the splurge occurs now or in the future) for example, doesn’t seem at all objectionable – whereas the analogous IG case may well be.
Feasibility considerations enter into determination of overall desirability in two ways. First, and most simply, the world presents for ethical choice the actual possibilities for the present and the future – not just counterfactual ‘opportunity sets’ but actual ones as best we can discern them. For our purposes here, I shall make the simplest possible assumption – namely, that there is such an opportunity set, and that it presents a rate of trade between present and future that is independent of the allocation between present and future. The overall ideal is therefore just the point at which the relevant pt-indifference curve is tangent to the real-world opportunity set; and this will be where the pt-discount rate and the rate of return are equal. Denote this point N*. In this simplified setting, the pt-discount rate is determined entirely by feasibility considerations, because the rate of return is constant. In a more plausible setting where the rate of return is a function of the distribution of resources between present and future, then the rate of discount and the rate of return will be equalised at a point at which pt-desirability considerations (the shape of the IG-indifference map) do bear. In specifying the tc-desirability point [Uc*, Uf*], there will be an appropriate ‘discount rate’ that is equal to the marginal opportunity cost of present-generation utility in terms of future generation utility.
The second aspect of the feasibility story involves the possibility (likelihood) that agents in making their inter-generational choices will not adhere to the requirements of justice. The general presumption in this respect is that the current generation will pay inadequate heed to the ethical claims of the future generation. So – the thought goes – individuals will locate at a point on the possibilities frontier that is not the ideal one: Call this point E*; the characteristic anxiety is that E* will involve future utility less than Uf*.
There are then two possible responses to this ‘motivational failure’ aspect. One is to attempt to alter agent’s preferences. Call this the “preaching” strategy. The other is to modify the relative prices between present and future generation’s utility so that with no change in actual preferences, individuals will be induced to locate at N* rather than E*. Call this the “policy strategy”. Moral philosophers are, predictably perhaps, more disposed to the former; economists have long been pre-occupied with the latter. In both cases, of course, the preacher and the policy-maker are assumed to have the normatively correct perception of the tc-ideal. They just have different conceptions of the best technology to get there (or as close as is feasible).
It is worth emphasizing that the two strategies make rather different appeals to language. The philosopher is disposed to say that at E* (from the point of view of pt-desirability) the discount rate that individuals are using is “too high” – ie the slope of the actual indifference curves at E* exhibits a higher preference for the present than does the slope of the EI curve through E*. The economist by contrast is inclined to say that the discount rate is “too low” in the sense that, in order to achieve N*, we would need to change relative prices so that we achieve a discount rate equal to the slope of the actual indifference curves through N*. There will have to be a subsidy on transfers to the future -- financed by an income tax, say – so that N* emerges as the chosen equilibrium. At N*, the actual indifference curve involves a discount rate (equal to the perceived rate of return) higher than that exhibited by the EI curve through N*.
[THIS IS A CASE WHERE A PICTURE IS WORTH A THOUSAND WORDS. I SHALL DRAW THE RELEVANT PICTURE WHEN I PRESENT THE PAPER. IN THE MEANTIME, I LEAVE THOSE WHO ARE FAMILIAR WITH SUCH THINGS TO DEVISE THE RELEVANT DIAGRAM THEMSELVES.]
The simple bottom line here is that the philosopher/preacher will be inclined to say that the central problem is that the discount rate is too high – whereas economists will be inclined to say that it isn’t high enough. But any apparent disagreement hangs on a confusion. Both are agreed that the real problem is that there will be too much consumption by the present generation and not enough by the future – that something has to be done to shift resources in favour of future generations. In short, all this confusion could be avoided if discussion focused on the ‘quantities’ rather than the ‘prices’. After all, on the view propounded here, it is the quantity domain that is the domain of basic normative concern.
There are several other ‘feasibility’ questions that might reasonably be posed at this point. One involves the political as opposed to the technical feasibility of policies designed to favour the future. Future generations may not vote (they might if the generations are close enough together); and if they do, and one is reliant on this fact to ensure that their interests are defended, there seems little reason to suppose that the future might exploit the present no less than the present exploiting the future. I think there is an answer to this challenge, but I shall not explore that answer here. I shall just acknowledge the challenge and move on. If one were persuaded that the challenge were an insuperable one, one might be led to the view that ‘preaching’ is the only possible response!
Another feasibility issue involves specifying in rather greater detail the policies that were supposed to secure the inter-temporal transfer to future generations. Clearly increased savings policies in themselves are not sufficient, because individuals could plausibly simply consume their lifetime accumulations later than otherwise in their lives. That is, policies that solve an IT problem do not necessarily solve an IG problem. What would be required is some policy that specifically encourages inter-personal transfers across generations. Such policies must change relative prices: for government to do more accumulation runs the risk of being infra-marginal and therefore undone by adjustments in individual bequest behaviour. [The arguments at stake here are more or less analogous to those that appear in the public debt literature, and the social security literature. Public provision of future benefits will simply crowd out private provision unless marginal incentives change.]
It is intriguing to me that those who profess to be most concerned about the lot of future generations are also those most horrified at the abolition of estate taxes (to say nothing of proposals for bequest subsidies.) Yet there are only two vehicles available to secure intergenerational transfers: increased bequests at the private level, and increased accumulation of durable assets in the public sector. And the crowding out point already mentioned indicates why the latter might be a rather ineffective strategy, unless pursued at a scale that drives public bequest to zero (an exigency that could hardly be thought to be advantageous to the future!) The argument may well be that bequest subsidies will have undesirable consequences in terms of the distribution of future generations’ well-being. That concern is not an irrelevant one: it would be somewhat odd if one were concerned about intergenerational justice but not intra-generational. This observation suggests in turn that analysis in terms of generations as an aggregation might be too course-grained to capture one’s moral intuitions. Perhaps in analysing intergenerational effects one ought to specify the properties of each ‘generation’, not just the ‘average well-offness’ but also the degree of equality within the generation. Even so, “inter-generational’ preferences are not typically personally indifferent in the way that IG justice presupposes. That is, each person in generation 1 is attached to specific persons in generation 2 in a non-neutral fashion and in a manner that tends to be reflected in individual bequest behaviour. Methods of transferring resources across generations will have to reckon with the fact that individual bequest is the chief method by which such transfers are effected – at least at the margin – and that individual bequest may not be as malleable as an IG justice enthusiast would like.
One interpretation of the feasibility challenge here is that some preference change might be necessary in order to get much in the way of serious improvement in the IG domain. Whatever the limits on the efficacy of preaching, there are also limits on the efficacy of ‘policy’ – even assuming that the policy determination process is itself impervious to individuals’ preferences. In the general economistic spirit, the best feasible response to the IG justice issue seems likely to involve some combination of preaching and policy technologies!
A solution without a problem?
I have already said that the chief object of concern in this area is the distribution of well-offness across generations -- and not the terms of trade between generations. I have also asserted that much confusion has been courted by focusing attention on those latter terms of trade. But once we do focus attention on the relevant magnitudes, it must surely also occur to us that perhaps all the energy devoted to this issue in the literature is addressed to a non-problem. For I take it that it has been a basic fact of human history over the last three or four centuries at least, that each generation has on average done better than its predecessor. We might be on more secure ground in making this claim by restricting ourselves to the issue within Western countries, and over the last century. But so restricted I take it to be close to uncontestable that on average ‘future generations’ have turned out to do better than the present one, for each generation since let us say 1850. In the aggregate bequest made by each of those generations, no violation of the requirements of intergenerational justice seems to have been made. It is of course difficult to be clear as to what, given the rate of progress on offer, IG justice would actually have required. Perhaps the rate of increase in gdp/head (not itself an ideal measure to be sure) would have justified a yet greater sacrifice by past generations on behalf of future ones, but if so, the case is far from obvious. And it is correspondingly non-obvious why all the ink spilt on this issue was justified.
Of course, it is dangerous to extrapolate the future from the past. Perhaps our own case is different. Perhaps progress has peaked and it will be downhill all the way from here. Such a possibility is certainly imaginable. If so, we had better attend to the moral claims of future generations, because failure to do so may involve major intergenerational injustice. Perhaps.
But the concern that present generations systematically downplay the interests of future generations has turned out to be rather illusory in the past, despite the attention the issue has received from many economists and philosophers. Call this the IG illusion. Perhaps without this illusion, the human race might have fared very badly. Perhaps we owe the onrush of continual progress to it. Perhaps. But I am inclined to the view that the philosophers and economists have misread the situation, and have done so systematically over an extended period. IG justice seems as if it ought to be a problem. But it isn’t – or at least hasn’t been in the past. We might have recognized this fact more readily if we had formulated the ‘problem’ in the right way – by reference to inter-generational well-offness rather than to the discount rate.
Summary: This paper can be summarized by reference to the distinctions it makes – distinctions that I think are useful in moral reasoning more generally, and which seem to me to be all too often overlooked. Not least in the ‘discounting’ context. These distinctions are:
that between desirability and feasibility considerations
that between desirability pro tanto and desirability tout court
that between price and quantity as an object of normative concern
that between ‘preaching’ and ‘policy’
that between inter-temporal and inter-generational allocation issues.
I might here add the distinctions between total and marginal value; and the related distinction between ‘contemplative’, attitude-guiding ethics on the one hand and action-guiding ethics on the other. But I have not foregrounded those distinctions here.
The bottom line that I am led to is that the discounting problem has been systematically mis-specified; and that properly specification invites the thought that it isn’t really a problem at all. And I say this, notwithstanding the existence of doomsayers on all sides – whether of the ‘social security is going bust’ kind, or the ‘running out of oil’ kind, or the ‘environmental catastrophe’ kind. We have not lacked for doomsayers in the past either.
However, my main point here is conceptual, not practical. I think that the discussion around discounting is confused. I have certainly found it so. And I, for my own part, find my attempted de-mystification exercise helpful.
I note that I have not dealt with the role of numbers, or the question of the proper future consumption mix as between say environmental and ‘material’ goods (which clearly exercises many of those who write about the future these days.) This is not because I do not think these issues are interesting and important. It is rather that I do not see them as being intrinsically inter-temporal.
1 For a nice example of an economist’s reaction to this kind of treatment see Dasgupta’s plaint against Sen in the paper already mentioned.
2 I have in mind general prisoner’s dilemma effects. Individuals who voted for the progressive income tax and extensive transfers to the poor might still entirely rationally attempt to minimize their tax obligations!