THE CAPABILITY APPROACH AS A HUMAN DEVELOPMENT PARADIGM AND ITS CRITIQUES

Transcription

1 THE CAPABILITY APPROACH AS A HUMAN DEVELOPMENT PARADIGM AND ITS CRITIQUES Nuno Martins Faculty of Economics and Management, Portuguese Catholic University, Porto, Portugal Keywords: capability approach, human development, theory of justice, welfare and measurement Cite as: Martins, N. (2012), The Capability Approach as a Human Development Paradigm and its Critiques, Encyclopedia of Life Supporting Systems, Paris, UNESCO, pp Contents 1. Introduction The capability approach Theories of justice Equality Justification of moral principles and weighting different functionings Rules for aggregating individual welfare into social welfare Prescription and description Social measurement, justice and pragmatism Welfare economics and the criticism of the capability approach Conclusions Summary In the capability approach, advanced by Amartya Sen and Martha Nussbaum, human well-being is evaluated in terms of capabilities and functionings. This approach moves beyond perspectives that analyse well-being in terms of primary goods, resources or utility. The capability approach has been used to address the topic of human development, while providing an alternative framework to others in political philosophy and welfare economics. This article will describe the capability approach, as a human development paradigm, and contrast it with alternative frameworks within welfare economics and political philosophy, which have been adopted by authors who are critical of the capability approach. 1

2 1. Introduction Amartya Sen (1992) argues that theories of justice can be seen in terms of: a criterion (or social notion, or a rule of social judgement ) of justice; and a space (or variable ) to which this criterion is applied. The use of a criterion to assess spaces has been one of the main features of much political philosophy. Some examples of spaces to which criteria are applied are the space of primary goods in John Rawls s difference principle (to which the maximin criterion is applied), or the space of utilities in utilitarianism (with which maximising criteria are afterwards combined). The capability approach can be characterised as an approach aimed at providing not a criterion, but a space for evaluating equality, the space of capabilities (instead of utility, or primary goods). This means that the capability approach is not a complete theory of justice, for it does not address one of the two elements that a theory of justice must address, namely, a criterion, or a a rule of social judgement, that enables us to go from the information on the individual, to an assessment of the social welfare of all individuals. Often, there are two different motivations for obtaining a rule of social judgement that enables aggregating individual welfare: (a) a concern with social justice regardless of pragmatic concerns with obtaining a complete ordering of social outcomes and a simple measure of the latter, and (b) a concern with social justice combined with pragmatic concerns about obtaining a complete ordering of social outcomes and a simple measure of the latter. While capability theorists have been concerned with both cases, the Human Development Programme in particular has followed the latter perspective, attempting to reach a complete ordering of countries in terms of their level of human development. The criticisms of the capability approach as a human development paradigm have been made from essentially those two perspectives. On the one hand, we have criticisms of those concerned with social justice regardless of pragmatic concerns with obtaining a complete ordering of social outcomes. These criticisms arise especially within the domain of political philosophy. On the other hand, we have the criticisms of those who are concerned with concrete measurement, and with obtaining a complete ordering of social outcomes. These criticisms arise especially within the domain of welfare economics. Both these criticisms, arising within political philosophy, and within welfare economics, will be addressed here. But before doing so, it is now necessary to explain in more detail the capability approach as a human development paradigm. 2. The capability approach The capability approach was first proposed in 1979 by Amartya Sen in a lecture titled Equality of What?, reprinted in Sen (1982). However, a similar perspective had already been proposed by Sen in 1978 (in an article reprinted in Sen, 1984, titled Ethical Issues in Income Distribution: National and International, chapter 12 of the 1984 collection), where Sen uses the term primary powers instead of capabilities. The term primary powers shows the motivation behind the capability approach, in 2

3 which Sen tries to go beyond John Rawls s (1971) theory of justice, and its use of primary goods as the relevant space in which to assess inequality. Sen disagrees with the use of the space of primary goods, proposed by Rawls (1971), as the appropriate space for assessing inequality. The reasons for this is that primary goods are only means to well-being, not the end we are seeking, and that different human beings will obtain different levels of well-being when endowed with the same level of primary goods. Sen (1982) uses the example that a disabled person is disadvantaged relatively to others, and hence equality of primary goods would deliver inequality of well-being in this case. Hence, because of the diversity between human beings, equality of primary goods leads to inequality of well-being. This critique of Rawlsian primary goods as the space to judge well-being can be extended to other approaches that focus on goods, resources, commodities or material conditions as the space for assessing inequality see also Ronald Dworkin (1981a, 1981b) on the equality of resources. Sen (1982) notes how the utilitarian approach takes into account differences between human beings, for different human beings might have different utility functions. However, utilities reflect the person s mental metric, not underlying needs that are essential given the nature of human beings. Sen uses the example that a person who is hard to please should not be entitled to a higher level of goods due to having such a demanding taste. Thus, Sen argues that human well-being should be assessed in terms of human functionings, where a human functioning is what a given human being is or does. The notion of human functioning has Aristotelian roots, which have been developed especially by Martha Nussbaum (1988, 1992), but have also been acknowledged by Sen (1999: 289). Furthermore, the capability approach provides a multidimensional perspective on human well-being, since it focuses on various human functionings. However, Sen (1982, 1999) argues that equality should be assessed taking into account not only achieved functionings, but rather the potential to achieve. This means that equality must be evaluated in the space of potential functionings, which Sen and Nussbaum (1993) designate as capabilities. Thus, capabilities refer to what a human being can be or do. The capability approach, developed by Sen and Nussbaum, has generated a vast literature, developing it in theoretical and practical directions see, for example, Sabina Alkire (2002), Flávio Comim, Mozaffar Qizilbash and Sabina Alkire (2008), Séverine Deneulin (2009), Wiebke Kuklys (2005), Ingrid Robeyns (2005), or Vivian Walsh (2000, 2003, 2008). In supporting this view, Sen s arguments against utilitarianism and Rawls s emphasis on primary goods reflect an ontological concern with what is essential to human beings (and perhaps ultimately rest upon this ontological dimension). Sen s claim that a disabled person is disadvantaged relatively to others, or the claim that a person who is hard to please is not entitled to a higher level of welfare on grounds of such a demanding taste, do reflect some notion of what is essential to the nature of human beings, and implies an ontological account of human being and human functioning see also see Nuno Martins (2006, 2007a, 2007b), Matthew Smith and Caroline Seward (2009) and Ilse Oosterlaken (2011) on the ontological aspects concerning the capability approach. 3

4 In this sense, the persuasiveness of Sen s argument springs to a great extent from this concern with what is essential to human beings. This conception is in line with an Aristotelian essentialist (and realist) account of human being. In fact, Martha Nussbaum, who together with Sen advanced the capability approach Nussbaum (2000, 2003) develops precisely an Aristotelian approach Nussbaum (1992). 3. Theories of justice The capability approach is aimed at providing a framework for analysing social arrangements from an ethical point of view. This ethical evaluation can be seen as part of some underlying theory of justice, even though it is not a complete theory of justice. In order to understand the relation between the capability approach and theories of justice, the first issue to be addressed here concerns the various theories of justice that have been proposed. In fact, the criticisms of the capability approach as a human development paradigm within the domains of political philosophy are usually connected to different views of justice. So this topic is essential to help us understand the criticism of the capability approach as a human development paradigm. Sen (1992) suggests analysing theories of justice in terms of two elements: a criterion or social notion to be used in the ethical evaluation of social states; and a space or variable to which this criterion (or social notion) is applied. A space is some variable (such as resources, goods, commodities, liberties, rights, wealth, income, or capabilities) in terms of which individual welfare can be assessed. A criterion is a rule of social judgement, which enables us to go from a judgement about individual welfare (made in some space) to a judgement about social welfare, by combining the information on different individuals. In this context, it is important to stress that the capability approach is not (or at least is not yet) a complete theory of justice. For a theory of justice must include: a criterion (or social notion) of justice; and a space to which this criterion is applied. While the space allows an engagement on distributive considerations, by seeing how a given space or variable is distributed between different persons, the criterion must allow us to aggregate the information in order to obtain a measure of overall welfare. Sen s original formulation of the capability approach proposes the use of capabilities as the space, but nothing is said concerning the criterion which enables the aggregation of the information concerning capabilities, in order to compare competing situations from the point of view of social welfare and social justice. Rawls s difference principle, on the other hand, contains both a space viz., Rawlsian primary goods and an aggregative criterion to apply to this space the maximin rule. The capability approach, in Sen s original formulation, suggests capabilities as a space where freedoms and the differences between human beings are accounted for. But the question of which criterion to use is left open. Sen (2009) eventually came to criticise Rawls s overall theory of justice, while providing an alternative view of justice. But that was not the central concern of the capability approach, which is concerned with a description of the space in which equality is to be assessed. 4

5 Thus, in his original formulation of the capability approach, Sen does not intend to replace the whole Rawlsian theory, but only a small part of this theory. In fact, Sen does not challenge the Rawlsian liberty principle, which requires that each person is to have an equal right to the most extensive basic liberty compatible with a similar liberty for others Rawls (1971). Sen s concern is related to the difference principle, or more precisely, to the particular way of applying this principle: the use of an index of primary goods when applying the maximin criterion. Sen is mainly preoccupied with the fact that, when using the maximin criterion, one only takes account of the primary goods themselves, and not of how they affect people. So to go beyond an exclusive focus on primary goods regardless of how they affect people, need not lead to the rejection of the totality of the Rawlsian theory. As Sen explicitly acknowledges, [t]he focus on basic capabilities can be seen as a natural extension of Rawls s concern with primary goods, shifting attention from goods to what goods do to human beings. (Sen 1982: 368). In fact, Sen s capability approach is a perspective which shares some resemblances with Rawls s, like the stress of the importance of going from the space of achieved welfare (either measured in utilities, resources, income or wealth) to the space of freedoms, opportunities and possibilities that individuals have in a given social arrangement. However, unlike Rawls s theory, the capability approach looks not only at goods or resources, but also at the way in which they impact on human beings, given the diversity between individuals. The key difference between Rawls and Sen is that while Rawls adopts a prescriptive approach, Sen adopts a descriptive approach, to justice see also Sen (2009), and the contrast between transcendental theories of justice, and comparative theories of justice. Criteria, which Sen also terms as social notions (Sen 1992:25), combining formula (Sen 1992:74), or rules of social judgement (Sen 2002:273), include not only Rawls s maximin, which consists in furthering the interests of those who are in a worse position, but also other concepts such as the leximin, or the efficiency of the social arrangement. While the maximin criterion consists in maximising the interests of the individual in the worse position, leximin (or lexicographic maximin) consists in first maximising the interests of the individual in the worse position, and then maximise the interests of the second individual in the worse position, and so on. In utilitarianism, on the other hand, the criterion of efficiency involves a maximisation exercise in the space of overall utility, the latter usually defined as the sum of utilities (notice that when the maximisation criterion is applied to sums, such as the sum of utilities, it is usually named as sum-maximisation ). Another approach within utilitarianism consists in using von Neumann-Morgenstern (1944) expected utility functions, following the perspective advocated by Harsanyi (1975). John Nash (1950) also suggests a solution in the context of a bargaining problem, which has the interesting aspect that such solution is independent of the units in which the variable (for example, utility) is defined. Moreover, Nash (1950) proves that this solution is the only solution with such property. The Rawslian maximin criterion, on the other hand, is applied to the space of primary goods (which include wealth, income, rights, liberties, opportunities and 5

6 the social bases of self respect). Even though Rawls originally applied the maximin criterion to a space of primary goods, this criterion can be applied to any space (for example, in welfare economics this criterion has been applied to the space of utilities) note also that both the efficiency criterion and the maximin criterion involve maximisation, but the former applies such maximisation to a variable constructed taking into account the welfare of all individuals, while the latter defines such variable with respect to the worst-off individual only. So while the classical utilitarian theory of justice consists in applying the summaximisation criterion to the space of utilities, the Rawlsian theory of justice implies applying the maximin criterion to the space of (Rawlsian) primary goods. Note that this is only a part of the second principle of the Rawlsian theory of justice (the difference principle ). Besides, the Rawlsian theory of justice also has a first principle where the relevant space is liberty. Different combinations of criterions (or social notions) applied to different spaces (or variables) would deliver different informational bases for a theory of justice. Hence the maximin criterion and the efficiency criterion (here defined as summaximisation ), the von Neuman-Morgenstern expected utility framework or the Nash solution to bargaining problems, could be applied to spaces other than utility. These spaces could be rights, liberties, opportunities, duties, wealth, income, or any other variable, where the different combinations would deliver different theories of justice. Within political philosophy, the perspectives of Rawls, and utilitarianism, have been the main opponents of the capability approach. Other alternative perspectives have been proposed as well, such as Ronald Dworkin s, who advocates equality of resources. Also, G. A Cohen (1993) has criticised Sen s capability approach for promoting an athletic conception of the human agent, where life is valuable only when agents actively choose something, exercising a functioning. However, it is important to note that the differences between the capability approach and these perspectives concern the space, and not the criterion, to be used. 4. Equality Theories of justice can contain more than one criterion, where different criteria are applied to different spaces. For example, the criterion of equality may be applied in conjunction with some other criteria, even though both are applied to different spaces. In fact, Sen (1992) notes that the criterion of equality has the particular characteristic of being present in nearly all theories of justice (applied to some space), and possibly combined with other criteria that are applied to different spaces. Sen (1992: xix) argues that [...] a common characteristic of virtually all the approaches to the ethics of social arrangements that have stood the test of time is to want equality of something something that has an important place in the particular theory. So the criterion of equality is applied to some space or variable in a wide range of approaches. 6

7 In fact, choosing a criterion other than equality does not mean neglecting the latter. Sen emphasises how the main point of many disputes in the fields of ethics and social justice is not concerned with being in favour of or against equality; rather, the crucial issue at stake is which is the focal variable to be equalised. In Ronald Dworkin s contribution, the stress on equality (of resources) is evident. But even when advocating the central role of other criteria (such as freedom from constraints or efficiency), egalitarian concerns are also implied. For example, to argue that Robert Nozick s (1974) libertarian framework is nonegalitarian because it allows for large differences in wealth, income, welfare, utility or holdings of (Rawlsian) primary goods of different individuals, would miss the fact that Nozick s approach demands some sort of equality on other focal variables, namely equality of libertarian rights no one has more right to liberty than any other person. As Sen (1992) argues, Nozick s approach is not simply non-egalitarian, for it implies equality of a given focal variable (viz., libertarian rights), even though it allows for inequality on other variables. The same can be said about the efficiency criterion. Even in the utilitarian approach of maximising the sum-total of the utilities (where the efficiency criterion collapses to the sum-maximisation criterion), egalitarianism is present. Notice that when stressing the criterion of efficiency, utilitarianism applies this criterion to the space of the total sum of utility. In this sense, inequalities may arise between the utilities of different individuals, since the goal is not to equalise them, but to maximise their sum: if the utilities distribution that maximises overall utility is a non-egalitarian one, the goal of maximising the sum will lead to inequality at the level of individual utilities. However, Sen (1992) emphasises that in this utilitarian approach the egalitarian feature is present in the weights attached to each individual in the utilitarian objective function that is to be maximised. This happens because each individual s utility function is assumed to have an equal weight in the sum-total. Therefore, in utilitarianism the criterion of equality runs parallel to the criterion of efficiency, but both are applied to different spaces: while the latter applies to total utility, the former applies to the weights of each individual utility function in the objective overall utility function to be maximised. So egalitarianism does not occur at the level of individual utilities, but with respect to the weights of each individual in the objective overall utility function, since these are often assumed to be equal. Rawls s maximin criterion, on the other hand, stresses the importance of maximising an index of primary goods of the worst-off individual. But when choosing a particular distribution of primary goods, people are assumed to be under a veil of ignorance, and thus they do not know on which position they will end up in the chosen social arrangement. In this approach, even though the criterion applied to the space of primary goods is the maximin (and not equality), egalitarianism is present in the fact that the person can end up in any position of this social arrangement, that is, no privileged place is ensured a priori. Besides, the maximin criterion is to some extent egalitarian, for to maximise the situation of the worst-off individual in every instance reflects a concern with those who suffer the most from inequality. In this sense, the maximin criterion to some extent conciliates maximisation of something with equality. 7

8 Harsanyi (1975) has criticised the maximin criterion, arguing that if an individual would use such criterion in real life situations, such behaviour would lead to absurd situations (for example, in comparing some situation A which contains much better possibilities than situation B except for one scenario that can occur with a very small probability, but is worst than any other case in both A and B, maximin will lead to choose B). Harsanyi (1975) suggests the use of von Neuman-Morgenstern (1944) expected utility functions instead. But note that one thing is to argue that individuals act following von Neuman- Morgenstern expected utility functions in real life situations (as Harsanyi is assuming), another thing is to say that von Neuman-Morgenstern expected utility functions should be the criterion for social judgements. Harsanyi s argument is plausible in what concerns to individual behaviour, but such fact need not imply that it will also be plausible when applied as a criterion of social justice, as opposed to a criterion for individual behaviour. Sen (1992) concludes that different schools of thought have called for equality in a given focal variable as the central exercise, arguing for equality concerning various variables such as income, welfare, on weights of utilities of each person (the case of utilitarianism) or on a class of rights and liberties (the case of pure libertarianism). In this sense, Sen poses as the important question to address not only Why equality, but also the question Equality of What? The very question of Why equality already implies a reference to some feature that is regarded as the variable to be equalised. So when addressing the issue of why does equality matter, one must also select which is the focal variable to be equalised, i.e., one must answer the question Equality of What? Conversely, arguing for equality of some focal variable already implies the recognition of the importance of equality, even though specified in the form of equality of a particular variable. In this sense, the question Why equality? becomes the question Equality of What? At this point, one may wonder why equality appears in these different theories of justice. Sen (1992:17, emphasis in original) suggests as a reason for such that [ ] ethical reasoning on social matters must involve elementary equal consideration for all at some level that is seen as critical, and notes that [p]erhaps this feature relates to the requirement that ethical reasoning, especially about social arrangements, has to be, in some sense, credible from the viewpoint of others potentially all others. Summarising, Sen suggests that theories of justice can fruitfully be seen in terms of: a criterion; and a space to which this criterion is applied. Furthermore, equality is a criterion that is thoroughly present applied to some focal variable, while other criteria applied to different spaces or variables are not incompatible with equality being the criterion for another given variable. The capability approach proposes the space of capabilities as the appropriate space, or focal variable, for assessing equality. Competing approaches to justice and wellbeing, both within political philosophy and welfare economics, differ by proposing other spaces in which to assess equality, and other spaces combined with other criteria. This is the essential difference between the capability approach, as a human 8

9 development paradigm, and the other competing approaches that exist in the literature. 5. Justification of moral principles and weighting different functionings The question that follows is how to select amongst competing criteria and spaces. This raises the question of the process through which moral principles or ethical goals are chosen. In fact, it is because different processes of moral reasoning are used, that different outcomes or results concerning criteria or spaces are reached. Sen s notion of freedom takes into account both opportunities and processes. The process of derivation and moral justification of principles of justice is then an issue to be taken into account, in addition to the discussed problematic of the informational bases of theories of justice. Rawls s discussion of justice, resorting to notions like reflexive equilibrium, original position and veil of ignorance, is well equipped in what concerns the (procedural) derivation and justification of principles of justice. The importance of processes is also present in a different sense. We must distinguish the process through which (moral) principles of justice are derived (or justified), from the ongoing procedures we undertake in our daily affairs. Sen s notion of freedom attaches importance both to consequences and to these processes. So in addition to the process of deriving and justifying moral principles, one must also take into account the process through which a given outcome came about in a given action, regardless of how these procedural (moral) rules of action are themselves derived or justified. The issue of the process through which a given outcome came about is also present when analysing how a theory of justice is implemented in society (again, regardless of how its principles are derived or justified). Concerning this issue, Sen (2009) emphasises the role of democracy, and argues that social choice processes are also crucial for implementing concrete policies. In fact, one could also argue that maybe the choice of a criterion for aggregating capability information should also be a social choice process. Despite of the intention behind the choice of a criterion, and the possible correspondences that can be traced with competing moral values, the choice of a criterion, like the implementation of concrete policies, may ultimately have to rely on social choice processes. In fact, social choice processes are crucial for the legitimacy of the policies that are to be adopted, regardless of the merits of theories of justice that support them. Sen (2009) distinguishes his approach to justice from others also for this reason: because the choice of criteria takes into account democratic processes of social choice, instead of being decided a priori by the political philosopher. The role of social choice and public debate can also be important in addressing an additional complication within the capability approach, and its use as a framework behind human development indicators. Capability information is represented by a vector of potential functionings, not by a scalar. And it may be the case that between two alternatives, one alternative may lead to a higher value in some components of the vector, but a lower value in other components of the vector. This leaves us with 9

10 the problem of which alternative to choose, which depends on which components (that is, which functionings) we value most. Even for a single person comparing two situations, it may be the case that the vector of functionings of one situation contains some components higher, and others lower, than in the alternative situation. We have a problem then in comparing the various options of a single person. And in addition to that, we have the further problem of aggregating the information on various persons. Concerning this second problem, when applying any criterion like the summaximisation criterion to vectors of functionings, one does not have scalars to be summed up. Rather, one will have several vectors of functionings being summed. But if there is no alternative in which all components of the aggregate vector are higher or equal than the components of any other vector, it is not possible to find an option that maximizes all the components of the vector. Even if we decide to apply the maximin criterion to this problem instead, we may also not be able to choose the vector of functionings (or potential functionings) of the worst-off, if no individual has a vector of functionings (or potential functionings) that is dominated by all others (that is, where none of the components of the vector is superior to the corresponding component in the vector of functionings of no other individual). Hence, it may be crucial to have a vector of weights reflecting the importance of each functioning (or potential functioning), in order to obtain a scalar measure of well-being (or advantage) of each person (by multiplying the vector of weights by the vector of functionings). Then, this scalar measure will enable comparing different situations using a unique value of well-being (or advantage). And these scalar values can then be summed in order to obtain the sum-maximisation criterion, or one can pick up the lowest value of each situation, and compare them, in order to use the maximin criterion. But how can the vector of weights, reflecting the importance of each potential functioning, be obtained? Sen argues that only social choice processes and public debate can legitimise this choice of weights. Notice also that some comparisons between vectors of potential functionings might be possible without reducing these into scalars. Sen mentions dominance of a set of capabilities over another as a possible route for comparing sets. Sen defines dominance set-evaluation in the following way: take S and T to be two sets to be compared; then set S dominates set T, given the preference relation R*, in the following situation: S R* T if and only if S S:[#S = #T and k (.): S T, x S : x R*k (x)]. That is, a set dominates S dominates a set T if we can find a subset S of S, and a functional relation between S and T, such that each element of this set S dominates each element of set T which corresponds to it in this functional relation see Sen (1985: 43-44) for elaborations. 10

11 Nevertheless, often these comparisons would have to be partial when none of the two compared sets would dominate the other, for example because we cannot find a subset S or establish the functional relation k(.) and the orderings that would arise would often be incomplete. In this case, to engage in measurement where a complete ordering is attained, the solution is to find a vector of weights through a social choice process and public debate, so that vectors of functionings (or potential functionings) can be converted into scalars using this weighting vector. The scalar measure of individual well-being (or advantage, depending on whether it is the vector of achieved functionings multiplied by this vector of weights or the vector of potential functionings (i.e., capabilities) multiplied by the same vector of weights) will be denoted by Wi. Wi would then be the scalar: Wi = p 1.vi1 + p 2.vi2 + + pm.vim = p. vi The vector vi is the vector of valuation of the m functionings, where these m functionings may be achieved functionings or potential functionings. In the former case, Wi is achieved well-being, and in the latter case Wi is potential well-being, i.e., advantage. In order to understand what vector vi means, it is useful to resort to the following notation Sen provides in Commodities and Capabilities (Sen 1985: 11): xi = the vector of commodities possessed by person i. c(.) = function converting commodity vector into vector of characteristics of those commodities fi(.) = personal utilization function, which is a function converting characteristics of commodities vector into vector of functionings that person i can achieve Fi = the set of utilization functions fi that person i can in fact choose. hi (.) = happiness function of person i, which depends on the functionings vector. 11

12 Sen explains then that when a utilization function fi(.) is chosen, the achieved functions will be given by vector bi, where: bi = fi(c(xi)), while the happiness of the person will be given by ui, where: ui = hi (fi(c(xi))) Notice that this can be interpreted as an utility function, if one interprets utility as happiness, one of the three possible interpretations Sen gives to the concept of utility. The value of a vector of functionings bi is given by the valuation function: vi = vi(fi(c(xi))) The mapping given by vi(.) does not have to be the same as that given by hi(.), for as Sen argues, happiness generated by a life need not coincide with valuing such life. The person s capabilities given his personal features Fi and his command over commodities Xi will be: Qi(Xi)=[bi \ bi = fi(c(xi)), for some fi(.) Fi and for some xi Xi] while the values of well-being that this person can possibly achieve are: Vi=[ vi \ vi = vi(bi), for some bi in Qi] Since we are concerned with valuable functionings here, the relevant vector to be used is then: vi = vi(fi(c(xi))) The vector p = (p 1, p 2,, pm) is the vector of weights attaching to each of the m functionings. Vector p is obtained by some social choice mechanism (and public 12

13 debate) that enables going from various vector pi = (pi1, pi2,, pim) of each i individual of the n individuals to a socially chosen vector p. In the Human Development Index (HDI), this vector is p = (1/3, 1/3, 1/3), while vi is represented using measures of health, education and income. Of course, the motivation for this approach is social measurement, as the purpose of the HDI is providing a measure of human development, not the informational bases for a theory of justice. 6. Rules for aggregating individual welfare into social welfare But going from a vector of valued functionings (or potential functionings) to a scalar measure of well-being (or advantage) is not the only issue to take into account. We also need a criterion of social aggregation to combine with the space the capabilities. One must then distinguish between (a) converting the vector of functionings (or potential functionings) of a person into a scalar measure Wi, and (b) aggregating these levels of individual welfare (or potential welfare) Wi into a measure of social welfare (or potential social welfare), which we may term as SW. Step (a) is about going from the vector of functionings (or potential functionings) to the scalar Wi, while step (b) is about going from the value Wi of each individual to social welfare (or potential social welfare) given by SW. After obtaining scalar Wi for each individual, one could choose as a criterion, for example, to maximise the sum of the Wi of each individual (choosing the efficiency, or sum-maximisation, criterion), or to maximise the value Wi of the worst-off individual (choosing the maximin criterion), for example. Nevertheless, Sen notes how the maximin criterion omits important information, for it only takes into account the welfare of the worst-off individual, and hence it is insensitive to the welfare of all other individuals. Sen explains that the maximin does not take into account how in one scenario the worst-off individual may be slightly better than in another scenario, but all other individuals were considerably better in the second scenario. On the other hand, a summaximisation criterion that simply adds units of welfare of all individuals would take into account information concerning all individuals. However, it would not distinguish between an improvement of the welfare of an individual who is in a state of deprivation, and an improvement of the welfare of an individual who already enjoys a very high level of welfare. A criterion that would perform relatively better given these issues is what can be called the Nash criterion. This criterion consists in obtaining the social welfare function by multiplying the welfare levels of all individuals, and was suggested by John Nash as a solution to a bargaining problem (on which see Nash 1950). In this sense, for a community of n individuals, the Nash criterion could be represented by the functional relation (where SW is social welfare): 13

14 Nash rule: SW = n i=1 Wi Notice that in this notation, the sum-maximisation criterion would be represented by: Sum-maximisation rule: SW = n i=1 W i and the maximin criterion would be: Maximin rule: SW = min {W 1, W 2, W 3,, Wn } One could also define these rules as social welfare functionals, where the domain is the same as before, but the range of which is an ordering, defining R = f (SW). Substituting above one would have: Nash rule: R = f ( i=1 Wi ) n Sum-maximisation rule: R = f ( i=1 W i) n Maximin rule: R = f ( min { W 1, W 2, W 3,, Wn } ) Where R is an ordering of competing social outcomes. Remember that Wi = p. vi.it is important to stress that Wi can be interpreted as the achieved well-being of person i if it is the scalar product of the vector of achieved functionings of individual i by the vector of weights to be attached to each functioning. Then, SW will be social achievement, to which an ordering R will correspond. But Wi can also be interpreted as the well-being freedom of individual i if it is the scalar product of the vector of potential functionings (i.e., capabilities) of individual i by the vector of weights to be attached to each potential functioning. Then, SW will represent the freedom to achieve of the society. Note also that the vector of weights (vector p) is, as discussed before, obtained by a social choice mechanism. The Nash criterion, like sum-maximisation, is sensitive to information about all individuals (for it consists on the multiplication of the welfare levels of all individuals) and takes into account inequality for the product SW of different values Wi increases more when a lower value is increased by a given amount than when a 14

15 higher value is increased by the same amount. In the Nash criterion, the marginal increase in SW for an increase in the welfare level (or potential welfare level) of individual i is the product of the welfare levels (or potential welfare levels) of all other individuals. In this sense, not only every individual is taken into account, as the marginal benefit of increasing the welfare level (or potential welfare level) of any individual depends on how worse-off individual i is comparatively to other individuals. The marginal increase in welfare (or potential welfare) is the higher the lower is the individual whose welfare is increased, and only when all welfare levels are equal, the marginal increase will be equal regardless of whose welfare is increased 7. Prescription and description A useful distinction that can be seen to be implicitly present here is one that questions the purpose to which a criterion is used. For example, while it is clear that the maximin criterion ignores much information (since it excludes all information except that concerning the worst-off individual), one could believe that we should have ethical concerns with the worst-off only, and hence counter-argue that the only informational basis that is important for our purposes is precisely the welfare of the worst-off. The problem here is that one is referring to criterions of social welfare under two purposes: a descriptive purpose; and a prescriptive purpose. The first case can be seen as one where we pursue information about welfare of a community regardless of our ethical goals. The second case is one where we focus on a criterion that privileges our ethical goals. This distinction is very similar to the one Sen makes in his 1980 paper Description as Choice (reprinted in Sen 1982), where Sen notes that selection of characteristics in a description might be value-loaded or not depending on the purpose behind this selection. A criterion always has some descriptive performance, which may or may not suit our prescriptive goals. The claim that the sum-maximisation criterion does not give information about distribution of welfare refers to its descriptive performance, and this claim has nothing to do with what is our ethical goal. But if an equal distribution were an ethical goal, then this same (descriptive) remark would then have prescriptive implications as well. The claim that the maximin criterion only describes the situation of the worst-off is a descriptive claim. But if our ethical concern is with the worst-off, then given this prescriptive purpose using maximin is more adequate than using sum-maximisation, for its descriptive performance (information on the worst-off) encompasses all the information that our prescriptive aim requires. Once one recognises the increase of a social welfare function as an ethical goal, then any criterion has a prescriptive role (insofar as it provides an indication, or description, of this social achievement we intend to foster). In fact, different criteria will imply different variations in well-being under different policy measures, and this 15

16 heterogeneity will mean that different policies will be differently described by different criteria. But this fact, combined with the ethical goal of increasing social achievement, implies that different descriptions (given by different social welfare functions), that result in different evaluations of social welfare, will lead to different ethical assessments of social achievement, and hence will have prescriptive implications as well. In the maximin criterion, the prescriptive recommendation would then be to increase the welfare of the worst-off, and in the sum-maximisation criterion, the prescriptive recommendation would be to increase the welfare of the overall sum. This distinction between the descriptive role and the prescriptive role is crucial to understand the different motivations for the use of criteria or aggregating rules. In social measurement exercises, descriptions such as (i) the sum of individual welfares or (ii) the welfare of the worst-off, provide a picture of social reality. But when these descriptions are combined with the ethical goal of increasing social welfare, then these descriptions become prescriptive criteria. But the information that those aggregating rules provide can be used in both exercises, viz., social measurement and theories of justice: while exercises on social measurement are mainly concerned with description, theories of justice have been essentially concerned with prescription. Also, welfare analysis is always underpinned by ethical presuppositions of underlying theories of justice. Utilitarianism, for example, is the basis for modern microeconomic welfare analysis (where extensive use is made of utility functions). Sen s capability approach, on the other hand, can be seen as the ethical background of the Human Development Indicator (HDI). Nevertheless, much microeconomic welfare analysis and the HDI provide a simple (and somewhat crude) scalar measure of social achievement. This is so because the underlying theory of justice of each exercise is combined with a pragmatic concern with obtaining a simple measure of social achievement, that enables, amongst other things, ranking different social outcomes into a complete ordering. But the pragmatic concern of obtaining a simple (scalar) measure of welfare and a complete ordering of social outcomes (like a well-defined utility function or a ranking of countries according to their HDI) must be separated from the formulation of a criterion of a theory of justice. A criterion may have some pragmatic appeal, but one must separate between (i) criteria for theories of justice, concerned with prescription; and (ii) criteria for social measurement, concerned with descriptive issues. While both these issues are addressed here, one must bear in mind that they are distinct exercises. Furthermore, the emphasis that is being made here on scalar measures of welfare (for example, by using a vector of weights p that enables going from a vector of functionings to a scalar measure of welfare) is in line with these pragmatic concerns, but the reason to use it need not be completely pragmatic. 8. Social measurement, justice and pragmatism The two different motivations for obtaining a criterion are thus: (a) a concern with social justice (either social measurement or theories of justice) regardless of 16

17 pragmatic concerns with obtaining a complete ordering of social outcomes and a simple measure of the latter, and (b) a concern with social justice combined with pragmatic concerns about obtaining a complete ordering of social outcomes and a simple measure of the latter. Sen s capability approach, for example, can be seen as in line with situation (a). Sen always noted the difficulties of aggregating individual welfare, and noted how incompleteness in rankings is a fundamental aspect of reality that cannot be overlooked because of pragmatic concerns with obtaining a simpler indicator of development. The Human Development Indicator (HDI), on the other hand, can be seen as an example of case (b). The Human Development Indicator combines Sen s insights on the multidimensionality of development with the pragmatic concern of obtaining a simple (scalar) measure of welfare and a complete ordering of social outcomes (such as a ranking of countries according to their HDI). These pragmatic exercises must be seen as a useful simplification that provides a starting point for further scrutinity, but at no stage one should regard pursue a aggregating rule with an exclusive concern with simple scalar measures of welfare and complete orderings. Every aggregating rule can incur in problems that arise of oversimplification. This can happen with the sum-maximisation rule, the maximin (and lexicographic maximin) rule. Also, while social measurement exercise might aim at some simplification, theories of justice are not concerned with these. A central problem arises in connection to the situations where the fulfilment of some basic functionings is simply not negotiable. Sen and Nussbaum recognise a strong Aristotelian influence in the capability approach (Sen 1999, Nussbaum 1992, 2000), and a strong case can be made for seeing the capability approach in the context of a realist ontology in line with Aristotelian essentialism. Nussbaum (1992) has argued for an Aristotelian essentialist view on human functionings, where the fulfilment of some basic functionings for human beings is simply not negotiable. One could define a threshold according to an Aristotelian view, where certain minimal requirements must be accomplished, as Nussbaum (2000) argues. Nevertheless, when defining these thresholds, the issue of which functionings are essential to diverse human beings would have to be further discussed as well. This would require an analysis of physical, biological, social and cultural aspects, and a study of the ontological underpinnings the capability approach. In so doing, one must not lose sight of the issue of which needs are essential to diverse human beings, Sen s conception of the opportunity aspect of freedom, and Sen s reference to the process aspect of freedom. These fundamental aspects of Sen s analysis must be taken into account when addressing the problematic of the informational bases of a theory of justice, and any particular criterion. Also, the obsession with obtaining practical (descriptive) scalar measures of welfare or completeness of rankings may bring prejudice to both social measurement and assessments of social justice. As Sen argues: 17

18 In so far as there is genuine incompleteness, disparity, or ambiguity in relative weights, they should be reflected in corresponding ambiguities in the characterization of the weighted value of well-being. This relates to a methodological point, which I have tried to defend elsewhere, that if an underlying idea has an essential ambiguity, a precise formulation of that idea must try to capture that ambiguity rather than lose it. (Sen 1992:48/49, emphasis in original) So rather than ruling out ambiguities, it is important to realise that incompleteness in rankings and aggregation may be unavoidable. 9. Welfare economics and the criticism of the capability approach Economists like Partha Dasgupta (2005, 2007), in contrast, stress the need to engage in trade-offs, while reaching an exact approach to well-being. In fact, Dasgupta (2005, 2007) argues that the ethical underpinnings of welfare economics were already solved by Abram Bergson (1938) and Paul Samuelson (1947) when setting the foundations of economic analysis, through the use of a social utility function which aggregates individual utility functions. However, Sen points out how the use of utility functions, be it individual utility functions, or social utility functions, presuppose complete preferences. But, Sen (1982, 1987, 1997) argues, preferences may be incomplete, as Hilary Putnam (1996) also argues, in which case the Bergson-Samuelson framework, defended by Dasgupta, would be inconsistent, since utility functions presuppose the existence of a complete preference ordering. In fact, in an earlier exchange with Frank Hahn (1991) on which see Meeks (1991) Sen (1991) noted how the central claim of his argument against mainstream economics is that a unique preference ordering is not sufficient to describe human preferences and well-being, and that human preferences may not deliver the axiomatic properties, Sen (1982: 135) notes that are essential to an ordering in mainstream economics, such as being complete (i.e., it is always possible to say which element of a set is preferred), reflexive (i.e., any element of a set is preferred to itself) and transitive (if an element a is preferred to b and b is preferred to c, then a is preferred to c ). Thus, Sen spent much time developing analytical frameworks which allow for the cases where these axiomatic properties may not hold, such as the cases of quasiorderings, where the preference relation is reflexive and transitive, but not necessarily complete, as defined in Sen (1982: 208), or the case of quasi-transitivity, where transitivity exists just for the strict preference relation, and not for the indifference relation, on which see Sen (1982: 142). Sen (1985) advocates a multidimensional approach to well-being, where the latter is represented in terms of vectors of functionings in his capability approach. If the use of scalar measures of well-being leads to a preference relation which is complete, transitive and reflexive, measuring well-being in terms of vectors leads to the 18