One of the most significant shifts in political philosophy and normative economics over the last 40 years has been an increasing emphasis on the value of opportunity. Many of the most influential contributions to normative social theory in this period – including those of Rawls (Reference Rawls1971), Dworkin (Reference Dworkin1981), Arneson (Reference Arneson1989), Cohen (Reference Cohen1989), Sen (Reference Sen1992) and Roemer (Reference Roemer1998) – have focused on the opportunities that are open to individuals, rather than on the outcomes that individuals actually experience. Two common themes in this literature are that it is good for individuals to have more opportunity rather than less, and that opportunities should be distributed fairly among individuals. It seems inescapable that, if these ideas are to be developed, we need a method of comparing the quantities or values of opportunity that a given person has in different economic regimes, and that different people have in a given regime.

A huge amount of intellectual energy has been expended on the problem of constructing satisfactory measures of opportunity. I have been a participant in this enterprise for many years. Gradually, however, I have come to the conclusion that measurements of opportunity, as these have conventionally been understood, are unable to capture some of the features of opportunity that are most valuable in a liberal perspective. Attempts to measure opportunity typically characterize a person's opportunity as a set of options – her opportunity set – from which a choice must be made. The measurement problem is then to compare different sets in terms of the quantity or value of opportunity that they offer.Footnote ¹ But, I now think, opportunity is an open-ended concept: often, we cannot specify in concrete terms what a person does or does not have the opportunity to do, or what the value is of the things that she might do.Footnote ²

An apparently natural approach to the measurement problem is to assess the value of an opportunity set in terms of how far it caters to some range of ‘reasonable’ or ‘normal’ preferences. But one of the main reasons for focusing on opportunities rather than outcomes is to maintain neutrality between different conceptions of the good: the aim is to provide a fair distribution of opportunity and then leave each individual free to pursue her own good as she perceives it. Any normative definition of what is reasonable or normal privileges some conceptions of the good relative to others.

A related problem was first identified by John Stuart Mill (Reference Mill1859/1972). Part of the value of opportunity is in providing a space for the exercise of individuality and originality. But, since the essence of a truly original action is that its value is perceived only by the actor himself, no public standard of opportunity can distinguish between options that are truly valueless and those whose value has yet to be discovered. As Mill puts it:

The idea of using reasonable or normal preferences as a standard runs into a further problem. We now know that, in many situations, people do not have the stable and coherent preferences that rational choice theory has traditionally assumed. It seems that we need a way of accounting for the value of opportunities to act on unstable and incoherent preferences.

A more general problem for the measurement of opportunity is linked to the flaw that Hayek (Reference Hayek1948) identified in the theories of market socialism that were fashionable in the 1930s and 1940s. Real markets rely on a division of knowledge: no individual or agency has access to the whole stock of knowledge that, in its totality, determines the outcome of a market process. This makes it impossible to specify in advance, even probabilistically, what outcomes will result from particular actions by individuals. Thus, opportunity sets cannot be fully defined.

In this paper I offer the outline of a possible solution to these problems of open-endedness. In essence, the idea is to propose a conception of opportunity – opportunity as mutual advantage – which does not require any explicit specification of an individual's opportunity set, and which rests on an understanding of persons as responsible rather than rational agents. The open-endedness of this conception comes at what might seem to be a prohibitive cost – that opportunity, so understood, is not directly measurable. However, the benefits that any individual can derive from his opportunities depend on the resources or real income at his disposal. The problem of defining a fair distribution of opportunity can be rephrased as that of defining a fair distribution of real income. It might seem that problems of open-endedness have just been shunted on, from the domain of opportunity to that of real income. How, it is natural to ask, can we measure the value of a person's real income without considering what he can do with it – that is, by investigating the opportunity set that it induces? But, I shall argue, open-endedness problems are much less troublesome at the level of real income, provided that opportunity as mutual advantage is guaranteed to everyone.

1. REASON, RESPONSIBILITY AND IDENTITY

I use the term opportunity in the sense that is standard in economics and social choice theory: an opportunity for an individual is something that he has the power to bring about, if he so chooses. Some writers, including Steiner (Reference Steiner and Phillips-Griffiths1983), Carter (Reference Carter1999) and Kramer (Reference Kramer2003), use the unqualified term ‘freedom’ for this concept; Sen (Reference Sen1992) calls it ‘effective freedom’. However, the proper definition of ‘freedom’ is a contested issue in philosophy; for my purposes, it is simpler and more transparent to speak about ‘opportunity’. Roughly speaking, the problem I am concerned with is that of assessing the quantity or value of opportunity that a given individual enjoys in given circumstances.

In economic analysis, it is normal to represent each individual as making a choice between options, one and only one of which will come about as a result of that choice. The set of available options is the individual's opportunity set. At first sight, it might seem that what is required is a measure of how well an opportunity set caters to an individual's actual preferences. In economic theory, this is measured by the indirect utility of an opportunity set (that is, the utility that an individual would enjoy if she maximized her utility subject to the constraint of choosing from that set). This is an entirely sensible measure of the value of an opportunity set to an individual whose preferences are given. However, it is not a measure of opportunity in any genuine sense. In effect, it is an ex ante measure of preference-satisfaction, premised on the assumption that the individual will choose according to her given preferences. In assessing a person's opportunities, I take it, we should not take her preferences as given. Rather, we should take a perspective in which preferences are unknown or undecided. We should ask questions of the form: Were the individual to have such-and-such a preference, would she be able to satisfy it?

At the opposite extreme to the actual-preference approach is the idea of measuring the size of an opportunity set as a pure quantity, without any reference to human preferences or values. One simple form of ‘pure quantity’ measurement underlies Steiner's (Reference Steiner and Phillips-Griffiths1983) measurement of freedom. This is to treat options as discrete entities and then to count the number of options in an opportunity set. An alternative strategy is to define options in relation to some general space of characteristics, and then to find a measure of the degree of variety, assessed in terms of the dimensions of the relevant characteristics space, among the elements of an opportunity set. This is the approach taken by Carter (Reference Carter1999), who takes the perspective of a kind of folk physics, defining options in terms of the spatial distribution of matter that they induce, and measuring a person's opportunity as the degree to which she can change the distribution of matter in the world. The problem with these approaches, as I argue in Sugden (Reference Sugden2003), is that they are too dependent on particular ways of representing the world – either by cutting up the continuity of our experience into discrete entities that can be counted, or by specifying a particular space of characteristics within which options are located. Typically, the world does not just appear to us as consisting of discrete choice options; that is a representation that we impose on the world. Nor does it just appear to us with a ready-made grid of characteristics. Thus, there is a multiplicity of possible measures of opportunity, each of which can be justified as objective in relation to some corresponding representation of the world; as long as we think in terms of pure quantity, we have no satisfactory reason for selecting one of these measures rather than any other.Footnote ³

If we are to measure the extent of a person's opportunities without taking her preferences as given, and if pure-quantity measures are irredeemably arbitrary, what possibilities are left? Many theorists have started from a general intuition that can be expressed in the following way. In measuring the amount of opportunity that an individual enjoys, we should not take her particular preferences as given, but we should take account of general facts or judgements about the preferability, desirability, choiceworthiness or significance of different options. Thus, one might say, a set of options offers a wide range of opportunity to the extent that it caters to the range of preferences that a person might normally or reasonably be expected to have. The difficulty here is to fill in what is meant by ‘normal’ or ‘reasonable’.

One way of describing this difficulty is to say that, before we can think about what preferences would be reasonable (or normal) for a person, we need some conception of personhood or identity. By reference to what characteristics of a person do we judge what is reasonable for him or her? For anyone with liberal inclinations, this is dangerous ground. Are the same preferences reasonable for the young as for the old? For men as for women? For Saudi Arabians as for Swedes? For Catholics as for atheists? It is all too easy to claim that some preferences, desires or interests are reasonable for some people but not for others, and to conclude that there is no significant lack of opportunity when people are unable to satisfy preferences that have been deemed to be unreasonable or abnormal for them, given the identity that has been attributed to them.

Sen (Reference Sen1999b) expresses a traditional liberal response to this problem under the slogan ‘reason before identity’. A theory of justice, he urges, should be grounded on principles of rationality; it should not assume that the individuals whose interactions are to be governed by the rules of justice are already bound by communal ties. This is not to deny that community membership may be a significant aspect of an individual's sense of identity, but only to claim that a theory of justice should recognize individuals simply as rational agents. Sen contrasts this approach with that of communitarian writers who treat principles of rationality and morality as internal to particular communities. As his slogan signals, Sen believes that rationality and justice are not local matters. If the measurement of opportunity is to conform with these principles, it seems that the underlying conception of personhood should be that of the person as a rational agent.

Sen (Reference Sen1999b, pp. 8–13) sees Rawls as an ally in the cause of ‘reason before identity’, and rightly so. Rawls's (Reference Rawls1971) theory of justice uses a concept of opportunity that is grounded on an understanding of the person as a rational agent. It is a starting point for Rawls that each individual's ‘conception of the good’ is private to her, and that a theory of justice should be neutral between different conceptions. Thus, if a theory of justice requires a measure of opportunity, that measure should not presuppose any particular understanding of what is desirable or valuable in human life. Rawls implements this principle by proposing a ‘thin theory of the good’. Linking personhood with rationality, he defines each individual's good in terms of her own ‘rational plan of life’; but he respects neutrality by leaving the content of that plan unspecified. The only assumption is that certain ‘primary goods’ are required for the execution of any such plan. These goods comprise ‘rights and liberties’, ‘powers and opportunities’, ‘income and wealth’ and self-respect (pp. 62, 90–95, 395–399). Later in this paper, I will try to resolve some of the difficulties that are swept under the carpet by Rawls's practice of treating income (or its capitalization as wealth) as if it were a physical commodity like rice rather than as what it is, a summary statistic generated in a market economy. For the moment, however, it is sufficient to record that Rawls's intention is to measure opportunity indirectly, in terms of resources that any individual would need in order to execute a rational plan. My proposal will use a similarly indirect approach, but without the emphasis on rationality.

Sen's (Reference Sen1992, Reference Sen1999a) theory of capabilities is an attempt to assess opportunities directly. As in Rawls's theory, rationality plays a central role. Sen's theory starts with the assessment of individual well-being. Such an assessment, Sen claims, must be concerned with the quality of whatever is ‘constitutive of a person's being’ (Reference Sen1992, p. 39). These constitutive elements are functionings. A person's actual state of being is represented as a vector of functionings. The set of alternative functioning vectors from which the person can choose is her capability set. The extent of her opportunities is to be assessed in terms of the properties of her capability set.

In Development as Freedom, Sen sums up his approach by saying: ‘The basic concern . . . is with our capability to lead the kinds of lives we have reason to value’ (Reference Sen1999a, p. 285; there is a similar summary in Sen, Reference Sen1992, p. 81). Sen uses the ‘reason to value’ formula throughout that book. It expresses the idea that reason takes priority over identity: capabilities are to be assessed by the standards of reason, not by reference to preferences or evaluations that are currently professed within a given community. ‘Reason to value’ is essential for one of the recurring themes of Sen's work – the idea that a theory of well-being should discount the effects on a person's actual desires of adaptation to persistent deprivation. Sen wants to be able to say that a person who suffers from objective deprivation can have a low level of well-being even if, as a result of psychological adaptation, she does not feel unhappy about her circumstances and has no desire to change them (Reference Sen1992, p. 55; Reference Sen1999a, pp. 62–63). His way of saying this is to say that the deprived person lacks the capability to lead the kind of life she has reason to desire, whether or not she in fact desires it.

Clearly, a theory of this kind requires some external standpoint from which the reasonableness of a person's actual desires can be assessed. This is not just a matter of dealing with the particular problem of adaptation. If the concept of capability is to be put to use, we need to be able to compare capabilities across individuals and across regimes, and that requires some standard of comparison. Sen's approach is to treat such comparisons as calling for ‘reasoned evaluation’. In the context of collective choice (as contrasted with individual moral judgement), he sees reasoned evaluation in terms of open public debate and democratic decision-making: ‘[T]here has to be some kind of reasoned “consensus” . . . This is a “social choice” exercise, and it requires public discussion and a democratic understanding and acceptance’ (Reference Sen1999a, pp. 78–79). I take him to mean that, for the purposes of collective choice, the extent of a person's opportunities is to be understood as the degree to which her capability set allows her to choose from a range of alternative, rationally desirable life-plans, and that questions about what is and is not rationally desirable are to be resolved by democratic decision-making after public debate.

This proposal throws further light on Sen's understanding of the priority of reason over identity. The implication seems to be that, in order for a person to claim that some opportunity is valuable for her, it is not enough that she might desire to use it. She should be able to show, by appeal to principles that can be acknowledged by rational fellow-citizens, that she has reason to desire this. Sen wants to free us from the unwanted demands of our communities (or of those self-proclaimed communities which claim us as their members); but his proposals do not release us from what may be the equally unwanted demands of public reason.

One might think, as I do, that even Rawls's thin theory of the good makes excessive demands on behalf of reason. Why, in assessing the extent of a person's opportunity, do we need to refer to rational plans of life at all? Rawls's theory is intended to be neutral between the alternative rational plans that an individual might choose to follow; but what about people whose desires are too incoherent to be integrated into any rational system of preferences?

There is now a great deal of evidence that individuals’ decision-making behaviour often fails to reveal coherent preferences. In many decision contexts, individual behaviour displays systematic patterns that are contrary to the theory of rational choice, but explicable by psychologically based theories of mental processing.Footnote ⁴ Advocates of rational choice theory often claim that such behaviour is the result of ‘errors’ that, in principle, can be identified and corrected, to reveal individuals’ ‘true’ underlying preferences. In many cases, however, that claim lacks empirical support. For example, it is well-established that the preferences that are revealed in people's choices over pairs of options differ systematically from those that are revealed in their separate monetary valuations of the same options, but it is far from clear which (if either) of these preferences is ‘correct’ – or even what a judgement of correctness is supposed to mean. In the light of such findings, it is difficult to justify the assumption that each individual's preferences determine her rational plan of life. But if that assumption cannot be made, why is opportunity valuable?

Some readers may be tempted to answer that it isn't. They may think that any proposal to measure opportunity against a background of incoherent preferences must itself be incoherent. That thought derives plausibility from the common practice in normative economics and moral philosophy of treating the possession of coherent preferences as a necessary condition for being an agent at all. Thus, for example, Philip Pettit (Reference Pettit2006, pp. 138–139) claims that ‘[t]he fundamental tenet of our common sense psychology of human agents is that agency involves acting to realize various goals in a way that is sensible in light of the apparent facts’. On Pettit's account, agency requires a system of beliefs and desires that are mutually coherent in roughly the sense that they satisfy the axioms of conventional decision theory. An individual whose preferences fail to satisfy these coherence properties, and who does not acknowledge such incoherencies as errors that can and should be corrected, is simply not an agent. From here, it is a short step to the thought that such an individual has disqualified himself from the status of a person whose preferences should be – or perhaps even can be – respected: if he lacks an integrated sense of value, what is there to be respected?

Another manifestation of the equation of agency with coherent preferences is the practice of modelling the choices of a dynamically inconsistent individual as if they were made by two or more distinct ‘selves’, each with its own (internally consistent) preferences, interacting strategically with one another.Footnote ⁵ Normative analysis then proceeds by privileging the preferences of one of those selves, and by characterizing any action by the other selves as weaknesses of will. It is as if economics and decision theory, and related branches of moral philosophy, have no way of recognizing the continuing identity of a person whose preferences are incoherent.

In other papers, I have tried to explicate such a concept of identity. The essential idea is that personhood consists, not in a coherent system of preferences or in a rational plan of life, but in an attitude towards one's past, present and future actions which acknowledges those actions as one's own and accepts responsibility for their consequences. A person with such an attitude can value opportunities for herself as a continuing entity, whether or not her preferences remain constant (Sugden, Reference Sugden2004a, Reference Sugden2006b, Reference Sugden2007). This is a conception of the person as a responsible agent.

As a responsible agent, an individual is entitled to claim that whether or not she has a plan of life, and if she does, whether or not it is rational, is no one's business but her own. She values opportunity for herself because (looking ahead to future decision problems) she wants it to be the case that she will then be able to do as she then desires. She is ready to accept the consequences of acting on those desires, whatever they may turn out to be. If she recognizes that other people value their opportunities just as she values hers, and that one person's opportunities can be another person's constraints, she will see that there are limits to the opportunities that she can fairly claim. But as long as she respects other people's fair entitlements, why should she be expected to justify to them how she uses her own? As a summary of this conception of personhood, I offer a slogan of my own: responsibility before reason.

2. OPPORTUNITY AS MUTUAL ADVANTAGE

I shall now propose and defend a conception of opportunity that is appropriate for responsible agents: opportunity as mutual advantage. Opportunity, so understood, is not directly measurable: we cannot say whether one person's opportunities are greater or less than another's. However, there is a sense in which we can say whether, within a given economy, all feasible opportunities have been made available.

In economics, the conventional normative criterion is that of preference-satisfaction. A person is deemed to be well-off, we might say, to the extent that she experiences the outcomes that she in fact desires. (I use the term ‘well-off’ as a generic term for any normative judgement of a person's economic position, to be used in evaluating alternative economic regimes.) According to the capability approach, a person is well-off to the extent that her opportunity set provides her with a wide range of options that she has reason to desire. As a first approximation to the concept of opportunity that I am proposing, consider the following formulation: a person is well-off by being able to choose whatever she might desire and be willing to pay for.

Notice that this formulation contains no ‘to the extent that’ clause: the absence of such a clause signifies that no claim is being made about the measurement of opportunity. The clause ‘whatever she might desire’ introduces a significant element of open-endedness into the definition of opportunity by decoupling it from any particular pattern of desires. In order to test whether this criterion is satisfied for any individual, we do not have to ask what she in fact desires. Nor do we have to ask what she has reason to desire. We can postulate any desire, whether hers or not, whether reasonable or unreasonable, whether normal or abnormal, and ask whether, were she to have that desire, she would have the power to satisfy it.

The clause ‘and be willing to pay for’ signifies that each individual's opportunities are defined relative to some prior assignment of entitlements. The idea is that each individual has the opportunity to do whatever he might desire to do with those things to which he has entitlements. Further, each group of individuals has the opportunity to do whatever its members unanimously desire to do with those things to which, taken together, they have entitlements. For example, suppose Annie owns a bag of flour and Bill owns a keg of beer. Then Annie and Bill have the opportunity to exchange the flour and the beer if they both so choose. Another way of putting this is to say that if Annie were to desire the keg of beer, and were willing to pay the price necessary to acquire it, then it would be possible for her to get the beer by paying the price. The ‘necessary’ price is defined reciprocally, in terms of Bill's opportunities to do what he wants with his entitlements. Annie can get the beer if she is willing to give up something that Bill is willing to accept in return. This reciprocity introduces another element of open-endedness into the definition of opportunity: one person's opportunities cannot be specified independently of other people's desires. The term ‘opportunity as mutual advantage’ expresses this idea.

As a first step in understanding this concept of opportunity, consider a simple exchange economy in which all goods are private.Footnote ⁶ Each individual has an initial endowment of each good; the sum of these endowments is equal to the total stock of goods in the economy. A standard way of assessing opportunities would be to identify each individual's opportunity set (which, in a market economy, would be the set of bundles of goods whose value at market prices is no greater than that of the individual's endowments). But opportunity as mutual advantage is defined at an earlier conceptual stage, at which the terms on which goods can be exchanged have not yet been determined. What is required is that, whatever individuals’ desires turn out to be, each individual will be able to get what he desires and is willing to pay for.

This criterion is satisfied if, whatever individuals’ desires turn out to be, those individuals will be free to trade at the prices of a competitive equilibrium. I give a full proof of this proposition in Sugden (Reference Sugden2004a). That paper also shows how competitive equilibrium can be brought about by the actions of a small number of professional traders; the only rationality that is required of ordinary consumers is that, at any given moment, they buy goods only at the lowest prices currently posted by traders and sell only at the highest.

As an illustration of the essential idea, consider an exchange economy with just two individuals, i and j, and two goods, 1 and 2. Person i is endowed with the bundle (z_{i 1}, z_{i 2}); person j is endowed with (z_{j 1}, z_{j 2}). This pair of bundles is the allocation z. There is an equilibrium price p for good 2 (expressed in units of good 1) at which the two goods can be exchanged. Each person chooses what trades to make at this price. Thus, i's opportunity set contains every bundle (x_{i 1}, x_{i 2}) such that x_{i 1} + px_{i 2} ≤ z_{i 1} + pz_{i 2}; and similarly for j. Since p is an equilibrium price, the market clears. The holdings of i and j after trade are denoted by (x_{i 1}*, x_{i 2}*) and (x_{j 1}*, x_{j 2}*); this allocation is denoted x*. This equilibrium has three significant opportunity-related properties. First, entitlements are respected: each individual is able to retain his initial endowment if he so chooses. Second, exchange is non-discriminatory: each individual is able to trade on the same terms as the other. Third, all opportunities for mutually advantageous transactions are exhausted. To see why, take any feasible allocation x of the two goods between the two individuals other than the allocation x* actually reached by trade, and consider whether a transaction which substituted x for x* would be mutually advantageous. In the alternative allocation x, i's bundle is (x_{i 1}, x_{i 2}) and j's is (x_{j 1}, x_{j 2}). Since, by assumption, this allocation is feasible, the market value of the goods it contains can be no greater than the combined market value of the two individuals’ endowments. But each individual's opportunity set contained every bundle whose market value was no greater than that of his own endowment. Thus, either (x_{i 1}, x_{i 2}) was in i's opportunity set, or (x_{j 1}, x_{j 2}) was in j’s. The implication is that at least one of the two individuals had the opportunity to undertake his component of the transaction that would have led to x, but chose not to do so. In this sense, a transaction which took the two individuals from x* to x would not be mutually advantageous.

The property that all opportunities for mutually advantageous transactions are exhausted is the analogue in my framework of Pareto efficiency in traditional welfare economics; that competitive equilibrium has the former property is analogous with the first fundamental theorem of welfare economics. However, my concept of ‘mutual advantage’ does not require any assumptions about individual rationality. In my analysis, a putative transaction among a set of individuals (the parties to that transaction) is mutually advantageous if and only if each party would choose to undertake his component of it. Thus, in saying that an actual transaction is ‘advantageous’ to a person, no claim is being made about how that transaction promotes her well-being, or about its consistency with her supposed rational plan of life; all that is being said is that, at the moment at which the relevant choice is made, it is made willingly. And that, I submit, is as it should be in a theory that is grounded on responsibility rather than rationality.

I shall say that an economic regime satisfies the principle of opportunity as mutual advantage if, whatever individuals’ desires turn out to be, entitlements will be respected, exchange will be non-discriminatory, and all opportunities for mutually advantageous transactions will be exhausted. The clause ‘whatever individuals’ desires turn out to be’ is crucial. This is not a criterion for appraising the efficiency with which given preferences are satisfied; what is being appraised is the extent to which individuals are able to satisfy whatever preferences they may come to have. Thus, it is a principle of opportunity and not simply of preference-satisfaction. Notice, however, that this criterion is not framed in terms of given opportunity sets. Considering each individual in isolation, the only condition that is imposed on his opportunity set is that it must contain his initial endowment. Apart from this, each person's opportunities to act on his desires are conditional on the desires of other people.

Opportunity as mutual advantage is concerned with what people are able to do (that is, with their opportunities) and not with the absence of restrictions on their actions (that is, with their negative liberty). To see this, consider an exchange economy, and consider the principle of negative liberty that prohibits restrictions on mutually agreed exchanges. The absence of such restrictions is neither necessary nor sufficient to generate a competitive equilibrium. It is not necessary, because equilibrium can be reached without everyone's trading with everyone else. (For example, as I show in my 2004a paper, equilibrium can be reached when all trade is intermediated by professional traders.) It is not sufficient, because the achievement of equilibrium requires the integration of information that is distributed among many individuals. Markets are complex institutions through which, under favourable conditions, this integration can be achieved. Freedom of exchange is surely an important factor in the emergence and survival of market institutions, but a competitive market involves much more than this form of negative liberty.

The idea of opportunity as mutual advantage belongs to the tradition of classical liberalism. Related ideas can be found in many of John Stuart Mill's economic and political writings. In On Liberty, Mill proposes a principle of individual liberty. According to this principle, everyone should enjoy:

This is a principle of negative liberty, intended to protect individuals in the pursuit of what they take to be their own interests, however irrational their judgements of those interests may be. It implicitly presupposes some prior assignment of entitlements, relative to which the concept of ‘harm’ to others is defined. (For example, it gives each individual the freedom to choose how to distribute his income between alternative forms of consumption, but consuming other people's goods without their consent counts as harm.) Similarly, the ‘liberty of combination’ protects groups of individuals in the joint pursuit of what they take to be their mutual interests.

One might have expected the liberty of combination to include freedom of trade. Perhaps surprisingly, Mill treats the ‘doctrine of Free Trade’ as a distinct idea, albeit with ‘equally solid’ foundations. This doctrine is that ‘both the cheapness and the good quality of commodities are most effectually provided for by leaving the producers and sellers perfectly free, under the sole check of equal freedom to the buyers for supplying themselves elsewhere’. Mill seems to be distinguishing between sellers of goods and final consumers. Selling a good is treated as a ‘social act’, legitimately subject to regulation in the public interest, while consumption belongs to the sphere of individual liberty. Thus, a prohibition on the sale of alcohol would infringe individual liberty, while the imposition of health and safety regulations on employers does not (pp. 150–151). Here, Mill seems to be equivocating between negative liberty and opportunity. He wants to say that the principle of individual liberty is infringed, not only by laws against the consumption of alcohol, but also by laws against its sale by business firms; but he seems also to want to say that that principle does not apply to the actions of firms. The underlying thought might perhaps be better expressed in terms of opportunity as mutual advantage: each individual should have the opportunity to consume alcohol, subject to his paying whatever costs are necessary to induce others to produce and sell it.

Individual liberty and free trade are both implicated in Mill's feminist work, The Subjection of Women (Reference Mill1869/1988). Mill's use of these ideas illustrates some of the differences between opportunity as mutual advantage and Sen's theory of capabilities. Recall that the capability approach deals with the issue of adaptation by appealing to judgements about what it is reasonable to desire. Mill is at least as conscious of adaptive desires as Sen is. He argues that the current (that is, mid-Victorian) status of women is a survival from the ‘primitive state of slavery’ that was once common for both sexes; indeed, the legal status of a married woman is still, quite literally, that of a ‘bondservant’ (pp. 6, 32). In making this claim, so shocking to his readers, Mill has to confront the fact that most mid-Victorian women do not perceive themselves as unfree. How can this be? Mill's answer is that, because all but the most brutal men want their wives to be willing ‘favourites’ rather than forced slaves, men have systematically (if unconsciously) sought to ‘enslave their minds’: women have been trained to develop those self-abnegating dispositions that are most attractive to, and most useful for, men (p. 15). Using one of his favourite gardening metaphors, Mill says: ‘in the case of women, a hot-house and stove cultivation has always been carried on of some of the capabilities of their nature, for the benefit and pleasure of their masters’ (p. 22). But, despite his belief that women's desires are an adaptation to a form of slavery, Mill does not argue for the emancipation of women on the grounds that they have reason to desire a different form of life. Instead, he appeals to the principles of liberty and free trade.

Mill proposes the abolition of legal restrictions on women's access to education and employment, and speaks out against the ‘tyranny of opinion’ that enforces informal restrictions. He proposes that the legal status of marriage should be that of a contract between equals, similar to a business partnership (pp. 41–42). In such a regime, men and women will be able to make free choices about whether or not to marry. In responding to opponents who argue that it is essential for society that women marry and have children, and that this is proof that the natural vocation of a woman is that of wife and mother, Mill uses the economic logic of free markets. Goods and services that really are valued can always be supplied in the market, so long as those who demand them are willing to pay the costs of supply. If men have a sufficiently strong desire for women to be housewives and mothers, the terms of freely negotiated marriage partnerships will be such that women find it worthwhile to agree to them, given the opportunities available to them in a free labour market. In a free society, men will have to pay the price of the services they want, or go without. Mill makes a telling comparison with the former practice of using press gangs to force merchant seamen to work in the Navy:

Notice that Mill does not need to address the question of what the natural vocation of women is, or of whether women have reason to desire a mode of life other than that of wives and mothers. Nor does he need to predict what proportion of women will choose to marry in a free society, or on what terms marriage partnerships will form. It is sufficient for his proposal that marriage and the labour market should both respect the principle of opportunity as mutual advantage. We should allow individuals to make their own choices in that regime – and accept such consequences as may follow.

I have tried to show that the concept of opportunity as mutual advantage is coherent, and that it presupposes neither individual rationality nor any particular conception of the good. But the reader may still ask: What is good about this kind of opportunity?

I need an answer that acknowledges the open-endedness of opportunity. Thus, I shall not appeal to particular conceptions of the good. I shall not ask what is good for the world, viewed from a neutral point of view (whatever that might be). Nor shall I appeal to ideas about what it is rational for an individual to desire. I shall offer reasons why any individual, thinking about his own interests from his own point of view, might conclude that opportunity as mutual advantage is good for him.

There are two steps to this. First, I need to show each person that, taking his own and other people's entitlements as given, and taking his own perspective on what is good for him, he benefits from opportunities to do as he likes. That does not seem an extravagant claim. It is easy for any of us to point to other people and claim that they would benefit from being prevented from engaging in what we think is foolish behaviour. My hunch is that, deep down, most people who support paternalistic restrictions on choice are thinking about paternalism in this third-party way. They are imagining that, because the restrictions are intended only to prevent foolish actions, people like themselves (who, of course, are not fools) will not be affected. They are also imagining that whoever designs the regulations will share their own sense of what is foolish, rather than belonging to the party of fools. For these reasons, it makes a big difference to the terms of political debate if each person separately considers what is good for her rather than all trying to arrive at a collective judgement about what is good for all of them together.

To be clear: I do not claim that, as a matter of conceptual necessity, each person must see herself as benefiting from opportunities to do as she likes. It is possible to be paternalistic in an intrapersonal way: at one moment, or in one state of mind, an individual may want there to be external restrictions on what she can do at a different moment or in a different state of mind. I suspect that this form of paternalism is less common than many philosophers and decision theorists suppose. But whatever the truth about this, the bottom line is that the advocate of opportunity as mutual advantage has to convince each person separately that she benefits from being free to do as she likes. In other papers, I have tried to develop a framework for such an argument (Sugden Reference Sugden2006b, Reference Sugden2007).

The second step is to show each person that, taking her own and other people's entitlements as given, and taking her own perspective on what is good for her, it is good for her overall that other people have opportunities to do as they like. The simplest and most powerful reason is contained in the idea of mutual advantage. If opportunity is understood in terms of mutual advantage, the opportunities of different individuals are related to one another in a positive way. Your freedom to seek out and take advantage of opportunities for mutual benefit encompasses your freedom to seek out and take advantage of opportunities to benefit me. If everyone has opportunity in this sense, I am part of an economic system that is full of people who can expect to be rewarded for finding ways of benefiting me. In a world in which knowledge is dispersed – in which I do not know in advance who might be able to benefit me, or what form those benefits might take – that is enormously to my advantage.

That argument alone does not establish that other people's opportunities to seek mutual advantage among themselves always works to my benefit. But, I suggest, the general tendency of markets is for mutually advantageous transactions to work to the benefit of third parties. Why? Because if I am to be able to benefit from transactions with other people, those people must be in a position to supply me with things that I value. And so I have an interest in their being able to create value among themselves. One of the most fundamental insights of economics – and one that most people, most of the time, find hard to accept – is that the market is a positive-sum game. Each of us achieves what he wants, not by gaining what other people lose, but as part of a network of transactions in which other people achieve what they want.

3. OPPORTUNITY AND REAL INCOME

So far, I have argued that opportunity as mutual advantage is in the long-term interests of each person, viewed as a responsible agent and with her entitlements taken as given. It is now time to consider issues of distribution. Attempts to measure opportunity have usually been motivated by concerns about its distribution. Theorists have wanted to be able to explicate the idea of equality of opportunity, or to specify some minimum level of opportunity that should be guaranteed to all members of a society. Intuitively, the idea of equalising opportunity across individuals, or of defining a social minimum, does not seem incompatible with an understanding of opportunity as mutual advantage. But, as I have argued, opportunity as mutual advantage is not a quantity that can be measured: it is too open-ended a concept for that. So how can the distribution of such opportunity be analysed?

It seems clear that, if opportunity cannot be measured directly, distributional analysis must be based on the measurement of entitlements. I shall now argue that, in a regime which respects opportunity as mutual advantage, a person's entitlements can be measured in a way that reflects – at least in a rough and ready way – the extent of opportunity that that person enjoys.

To introduce the basic ideas that underlie my proposal, I use an example from everyday life in which it is easy to see both the importance of measurement and the open-endedness of opportunity. National pension schemes are intended to guarantee a minimum standard of living to individuals past the age of retirement. Pensions are paid in money, but money in itself – pieces of paper, numerical entries in a bank account – is of no use to anyone. What matters is the purchasing power of money. The purchasing power of a pension is an opportunity set – the set of alternative bundles of goods that it can buy. In this sense, any assessment of the real value of a pension requires the measurement of opportunity sets. The universal old-age pension provided by the National Insurance system in the UK is currently pegged to the retail price index, the objective being to maintain the ‘real’ value of a pensioner's income in the face of unpredictable changes in the economic environment. The question I propose to ask is this: What real opportunities are guaranteed by an index-linked pension?

To avoid unnecessary complications, I switch to a stylised formal model in which an index-linked pension can be represented. This model has a sequence of discrete periods t = 0, 1, . . . . There are n goods, all of which (it is assumed for simplicity) are on sale in all periods. The vector of prices for these goods in any given period t is denoted by the vector p_t = (p ₁[t], . . ., p_n[t]). Prices are set at the start of each period, before consumers make their purchases, and remain constant throughout the period. In each period t, there is a set of consumers, each of whom allocates a fixed income (which may be different for different consumers) between spending on the various goods. Summing over all consumers, the vector of total quantities bought of the n goods in period t is q_t = (q ₁[t], . . ., q_n[t]). It is now possible to define a ‘chained’ price index I_t for each period. For any price vector p and any quantity vector q, let p·q denote the total cost of buying the quantities q at the prices p. The price index is normalized by setting I ₀ = 1. Then I ₁ = (p₁·q₀ / p₀·q₀)I ₀, I ₂ = (p₂·q₁ / p₁·q₁)I ₁, and so on.

This index can be interpreted as follows. For any period t, a bundle of goods can be defined to be representative of consumption in that period if it contains the same proportionate mix of goods as is found in total consumption in the same period; a consumer who buys a representative bundle is a representative consumer. Now consider a bundle that is representative of consumption in period 0, and which has a money value of I ₀ (that is, 1) at the prices of period 0. Then I ₁ is the cost of buying the same bundle at the prices of period 1. If we consider a bundle that is representative of consumption in period 1 and which has a value of I ₁ at the prices of period 1, I ₂ is the cost of buying that same bundle at the prices of period 2. And so on.

To represent an index-linked pension in this model, consider an individual who retires at the end of period 0. The base value of his pension is set in period 0 at some money value y ₀. Then in each period t > 0, as long as he survives, he receives a money income of I_ty ₀. (For simplicity, I assume that the retiree has no other source of income, and that income cannot be saved from period to period.) We can ask what, in terms of concrete consumption opportunities, the retiree is being guaranteed.

Let us define the base bundle q₀* as that bundle of goods that is representative of consumption in period 0, and that has a value of y ₀ at the prices of period 0. In other words, the base bundle is the bundle that a representative consumer would have bought in period 0 from an income equal to the base value of the retiree's pension. The rule for updating the pension at the start of period 1 guarantees the retiree the opportunity to buy this base bundle in that period. This guarantee will be implemented by giving the retiree a money income of I ₁y ₀ in period 1. We can now define q₁* as that bundle of goods that is representative of consumption in period 1, and that has a value of I ₁y ₀ at the prices of period 1. The rule for updating the pension at the start of period 2 guarantees the retiree the opportunity to buy q₁* in that period. This guarantee will be implemented by giving the retiree a money income of I ₂y ₀ in period 1. And so on.

This guarantee goes some way towards assuring the retiree that he will be able to maintain the standard of living that he would have been able to enjoy in period 0 with an income of y ₀. To see the nature of the guarantee, consider the sequence of guaranteed bundles q₀*, q₁*, q₂*, . . . . Recall that q₀* is what a representative consumer would have bought in period 0 from an income of y ₀. It follows from the definition of I ₁ that q₀* can be bought in period 1 from I ₁y ₀, the retiree's income in that period. However, a representative consumer in period 1 with an income of I ₁y ₀ would in fact buy q₁*. In this statistical sense, we can say that a representative consumer ‘prefers’ q₁* to q₀*. So the change in the guarantee from q₀* (the guarantee for period 1) to q₁* (the guarantee for period 2) has been validated by the ‘preference’ of a representative consumer. And similarly for all subsequent changes in the guarantee. This chain-linking of representative preferences ensures that the real value of the guarantee, assessed according to the preferences of representative consumers, will not fall over time. Chain-linking also ensures that the guarantee tracks changes in technology and changes in preferences in the relevant population, to the extent that these changes are reflected in the composition of average consumption. Thus, the rule for defining guaranteed bundles is well-defined ex ante, but the composition of those bundles is not fixed in advance.

However, if the guarantee provided by the pension is understood only in terms of the rule for updating its money value, it has serious limitations. The most obvious limitation is that there is no formal assurance that the retiree will have the opportunity to consume anything other than representative bundles. Thus, he has no assurance that, if his preferences differ from those of his fellow-consumers, he will be able to satisfy those preferences. In reality, of course, a retiree can expect his opportunity set in each period to contain much more than the guaranteed bundle: it will contain every bundle that can be bought on the market in that period from a budget equal in market value to the guaranteed bundle. But it is important to recognize that the extent of these additional opportunities is not taken into account in the method of measurement that determines the money value of the pension in each period. The explicit guarantee of index-linking is complemented by an implicit expectation that, in each period, the market will offer a wide range of alternative consumption options.

The inability of the explicit guarantee to respond to preference diversity is not its only limitation. Not only is there no assurance that the guaranteed bundles will be tailored to the particular preferences of the retiree; there is no assurance that they will be tailored to anyone's preferences. To see why this so, imagine that the economy of my model is centrally planned. The reader has probably assumed that I have been modelling an unplanned market economy, but nothing that I have said so far has actually required this. So suppose the economy is organized in the following way. At the start of each period, government planners decide the prices at which goods are to be sold; given these data, consumers are free to spend their incomes as they choose. One might hope that the planners would try to simulate the workings of a market economy by setting the price of each good equal to its marginal cost and then planning production to meet the resulting demand. But suppose they are ignorant of the principles of market socialism. Instead, they produce goods in exactly the same proportions in every period, independently of consumers’ preferences. In each period, the planners set the prices of the various goods so that all markets can be expected to clear: they respond to changes in demand only by changing prices, never by changing production. If markets clear, the mix of goods bought by a representative consumer must be the same as the mix of goods produced. Thus, the planners’ production strategy ensures that the composition of a representative bundle is the same in all periods. This implies that there can be no change in the guaranteed bundles: we have q₀* = q₁* = q₂* = . . . . There is no assurance that these guaranteed bundles have any correspondence with average preferences.

By thinking about what an index-linked pension would guarantee in the absence of a well-functioning market, we see that conventional understandings of pension entitlements are premised on an expectation that (roughly) competitive markets operate, and that this will continue to be the case. This expectation allows pensions to be specified ex ante by well-defined rules without too much strain from problems of open-endedness.

What lessons should be drawn from this extended example? Recall that the starting point for the example was the problem of assessing the extent of opportunity provided by a pension entitlement. What has emerged, I suggest, is that this assessment can usefully be broken down into two components. The first component is a form of measurement, represented in the example by the value of y ₀, the base value of the pension in the prices of period 0. In more general terms, the measure is of real income. The second component is not so much a measurement as a precondition – the precondition that competitive markets operate. I introduced the example of the pension as illustrating two features of opportunity that are potentially in conflict – the need for measurements of it, and its open-endedness. In the case of the pension, I suggest, much of what we might want of a measurement of opportunity is achieved by a real-income measure – given the background condition of competitive markets. In itself, however, the real-income measure takes little account of the open-endedness of opportunity. That is taken into account through the background condition, by virtue of the way in which markets respond to unanticipated changes in preferences and technology. And that pattern of response can be characterized as respect for opportunity as mutual advantage.

Generalizing from this example, my proposal is that we should not try to measure opportunity directly. Instead, we should understand opportunity in terms of mutual advantage, and accept its open-endedness and non-measurability. But how much a person can do with such opportunity is partly determined by her initial entitlements, just as how much a retiree can do with his pension is partly determined by its size. Thus, a normative theory of opportunity needs to be concerned with the distribution of entitlements as well as with mutual advantage; and that requires some measurement of entitlements. However, that measurement should not be understood as a self-contained assessment of the opportunities to which a person's entitlements give her access. For many purposes, it may be sufficient to measure entitlements in terms of real income.Footnote ⁷