«Edited by ANNE MASON Research Fellow, Centre for Health Economics University of York and ADRIAN TOWSE Director, Ofﬁce of Health Economics Radcliffe ...»
On the right-hand side we ﬁnd a more tentative and open-ended line of enquiry. Welfare, in any of the senses on the left-hand side, counts as one – but only one – of a set of relevant characteristics of individuals or groups. Groups need to be there because of the possibility that one may wish to include some characteristics of relationships between individuals in the idea of the social good.
In this speciﬁc sense, extra-welfarism might allow ‘society’ to be more than a collection of atomistic individuals. Although the right-hand side approach may entail maximising some function of these diverse entities, it may not go so far (stopping, for example, at listing and measuring the variables in question as in ‘cost–consequences analysis’), or it may take a fuzzier approach (‘satisﬁcing’, for example). Its eclecticism is attractive to any who want to associate the idea
64 THE IDEAS AND INFLUENCE OF ALAN WILLIAMS
The opening up of so disparate a range of considerations has important consequences. For example, following the right-hand column in a context in which ‘being free’ is a major element of the objective function might prove more antagonistic to government regulation and ‘interference’ than any such case mounted on a Paretian platform. It is likely to provoke two kinds of response from left-leaning economists (in the columnar sense). The ﬁrst of these two responses is to say that the entire scheme is preposterously ambitious; far better to exercise some modesty by recognising that the economist’s role is restricted to a narrower set of preference-based considerations that we may call ‘economic welfare’ and the rest are someone else’s business. The counter to this is twofold. First, it may be – and it often actually is – business that no one else is around to pick up. Second, economists rarely hedge their advice, whether about the operation of markets in general, or of those in or connected with the healthcare sector, by disclaimers about all the possibly relevant considerations they are ignoring. The modesty, in short, is a sham and the rest of the world is invited to accept policy prescriptions that appear, at least on some occasions, as grotesque, like the appropriateness, to use Uwe Reinhardt’s telling example (Reinhardt, 1998), of an idea of efﬁciency that has poor and sickly baby Smith consuming a small fraction of the health care consumed by rich and healthy baby Chen.
The other kind of response from the left-leaning (in the columnar sense) economist is to say that all the ‘extras’ that right-leaning economists want to allow, at least in principle, are as admissible on the left as they are on the right, in that there are preferences over them all – we can even have preferences over preferences – and utility numbers are readily available to be assigned to them by well-known experimental methods. There is but a short solipsistic step to the reductio ad absurdam of supposing that every attitude is a ‘preference’ and every difﬁcult choice is a utilitarian trade-off – as though there really were no difference between the questions ‘what do I, a person with preferences, want to do?’ and ‘what ought I to do, I, having preferences but wanting to be behave morally?’ or as if there were not a possible – even frequent – conﬂict between the answers to these two questions. It also blithely ignores the genuine
RESOURCE ALLOCATION IN HEALTH CARE 65puzzle: what is so special about ‘preferences’ that we want to erect our idea of social welfare entirely upon them? Moreover, even if we do not go beyond preferences, we ought surely to care about their ethical character. The same kind of solipsistic thinking leads to the absurd Panglossian conclusion that all is at all times for the best in the world, otherwise it would have changed.
If it didn’t, then the expected utility of the world changed must have been less than the expected utility of the world kept constant; the gainers evidently could not compensate the losers. And the costs of reducing the transaction costs that may have impeded some transactions must have been higher than the combined surplus to be had from making the transactions that were not made – otherwise those costs would have been incurred and the transactions thus enabled would have been executed (Culyer, 1984).
A clear danger in the right-hand column arises from the risk of doublecounting or over- and under-accounting.Whereas there are standard procedures for avoiding these biasing distortions in conventional welfare economics – transfers and sunk costs are not confused with opportunity costs; increased property values are not added to the value of shortened commuting times – the opportunities for missing such effects on the right is much enhanced by the plethora of possible elements. Avoidance requires clarity about the elements and an understanding of how they may be empirically related.
Step (d) is the point at which there seems to be a clear separation of ‘welfarism’ and ‘extra-welfarism’. I am using the term ‘extra-welfarism’ throughout in order to differentiate it from a more extreme ‘non-welfarism’ (Kaplow and Shavell, 2001; Dolan and Olsen, 2002) which appears to deny the relevance of any part of the welfarist programme (and preferences in particular) in helping to identify the social good.
e. Welfare is a function (only?) of individual ‘preferences’ (these may be ‘true’, ‘underlying’, ‘stated’... preferences) or / versus The constituents of the social good are context-dependent and determined by an ‘ethical authority’ for that context (e.g. the ‘minister of health’) In step (e) we explicitly lay bare the roots of welfarism in preference relations, with the tentative suggestion that welfare is dependent only on the satisfaction of preferences, with some acknowledgment of the fact that ‘preferences’ are not self-evident and generally have to be inferred via fairly precarious chains of reasoning from behaviours of one kind or another. The basis of welfare in preferences is, however, essentially stipulative and a priori. It is therefore a serious embarrassment to be told, as we increasingly are, that the measurement of preference is an empirical will-o’-the-wisp (Kelman, 2005). On the righthand side we may still retain welfare (despite doubts) in the preferential sense, but its retention is conditional on the nature and circumstances of the decisions to be made. The ethical authority comes not from a stipulative
66 THE IDEAS AND INFLUENCE OF ALAN WILLIAMScondition about what welfare is, but from an ethical arbiter – or a process through which such an arbiter is to be identiﬁed. In the one case, the ethics come from the characteristics of the axioms, including their factual truth; in the other from the legitimacy of the arbiter – such as whether she has been elected. Essentially, the right-hand side substitutes a process (a stipulator or a method of ﬁnding one) for an algorithm.
f. The relevant preferences are (only?) those over goods and services or (possibly?) the characteristics of goods and services or / versus While preferences may be important, other considerations (feelings?
capacities? capabilities? values? interests? perceptions? relationships?
processes?) may also have signiﬁcance The left-hand side of step (f) has two variants: preferences are considered to be either over goods and services or over characteristics of goods and services (examples might be their effectiveness, convenience and responsiveness). On the right-hand side such preferences may also be important, but other factors may also be deemed relevant such as the employment conditions of those delivering the services or whether the service delivery team treats patients as partners in a treatment and rehabilitation process. Some utilitarians on the left include ‘process’ in principle and that is a somewhat less restrictive form of welfarism than that delineated in the table.
g. Preferences can be revealed (approximately) either through market transactions or through experiments or / versus Some elements may be revealed (approximately) through transactions or experiments, others may be assessed better through other methods such as deliberation (or with a combination of methods) Step (g) concerns the measurement of the sources of welfare. On the left, preferences may be inferred from market transactions or through experiments in either real (relative) terms or by a monetary yardstick – willingness to pay or willingness to accept. On the right, the same techniques may be applied and inferences drawn for preferences deemed relevant; some of them may also prove suitable for measuring and valuing non-preference elements (for example, limits placed on the application of the ‘rule of rescue’ or the discovery of distributional weights from economic experiments); but there will often be other considerations that may require altogether different mechanisms of empirical inference and appraisal through, for example, consultation and negotiation (for example, the short-term service impact of a new addition to the drug formulary on general practitioners (GPs) and pharmacies and its managerial implications; or working out in committee an agreed formula for
RESOURCE ALLOCATION IN HEALTH CARE 67ranking needs; or whether a proposed allocation mechanism may be seen as an embarrassing political symbol or precedent).
More fundamentally, experimental economics and psychology have revealed major ambiguities in what we may understand by ‘preference’. These partly impact on the meanings one attaches to ‘rationality’, and partly suggest that ‘real’ preferences (if they can be said to exist at all) are virtually impossible to reveal. They also suggest that whatever it is that experiments and like mechanisms reveal is highly contingent on circumstances – not merely the circumstances of the birth and environment of the subject of the experiment but the immediate circumstances under which whatever is being revealed is revealed – such as framing effects and the like (Kahneman and Tversky, 2000).
This opens up the tantalising possibility of systems for ordering public policy options that make no pretence at accuracy in their portrayal of preferences or people’s states of mind and that might even go further and be the ranking decisions of committees based upon whatever virtues are deemed acceptable by the ‘authority’. It seems an attractive quality of a public health policy mechanism that it ought to aim in some way to reﬂect what seem to be the values of stakeholders (especially the general public and those most directly affected by health care – patients and their families), for which decision makers would be accountable, and which would resemble utilities in their form and structure but would not necessarily be seeking to represent ‘preferences’. The challenge is to decide whether welfarism or extra-welfarism is better able to embody such values.
h. Preference equates to ‘utility’ or / versus Preference-based utility is but one consideration Step (h) contains several causes of potential confusion. Because of the greater inclusivity of the right-hand column, the utility measures appropriate to the left-hand column may also be relevant on the right. Moreover, one can also have utility measures of non-preference entities and, when this is the case, the differences between the two columns become blurred. The problem here is that ‘utility’ is doing more work than it can normally bear. For example, in
addressing the matter of health outcomes one might:
● ask the following types of people to assign utility numbers according to a
conventional experimental method to states of health-related quality life:
— doctors — patients with the speciﬁed condition — all patients — taxpayers — informal carers ● where the subjects may be members of the class described and asked to assign their own values or be non-members but asked to assign numbers
68 THE IDEAS AND INFLUENCE OF ALAN WILLIAMSputting themselves in the shoes of those indicated (revealing something resembling the ‘ethical preferences’ of Harsanyi (1955)). In either case the experiment is designed to measure preferences assign utility numbers in the foregoing experiment and its variants ●
according to the subjects’ view of:
— the desert of the probable beneﬁciaries — the productivity of the probable beneﬁciaries — the likely income generated for the service providers.
In each of these cases, the term ‘utility’ may be entirely appropriate in, say, the conventional sense of being normative and requiring ordering systems to be used that are characterised by completeness, transitivity, continuity, convexity and non-satiation. But the normative signiﬁcance is quite different according to who is doing the assigning, whose utility is being measured – if anyone’s – and according to whether utility is being used as an index of preference, or health, or need, or of some other desideratum that enables decision makers to prioritise uses for resources.
i. Decisions can be informed by the agenda-setting/question-provoking, taxonomising methods of CEA etc Step (i) identiﬁes an underrated virtue of economics, which is its taxonomising capability – the categorisation of elements for a decision maker to consider.