1. INTRODUCTION
Economics plays a dual role in the formulation of public policies. Not only does it seek to predict and/or explain the consequences of such policies, but it also analyses the ethical issues involved in evaluating alternative policies in terms of their social desirability. It is true that, from time to time, economists have expressed scepticism regarding the role of economists in ethical assessments of social policies.Footnote 1 Despite such reservations, however, welfare economics, the area of economics that studies the normative bases of public policies, has come to stay as an integral and important part of economics. One can discern two distinct recent trends here. First, there seems to be growing recognition of the importance of the study of ethical issues relating to social and economic policies.Footnote 2 Second, the ethical basis of welfare economics seems to have become much richer and broader. Traditional welfare economics was basically utilitarian in nature though it did not always rely on the sum-of-utilities criterion of classical utilitarianism. This has changed significantly over the last few decades and welfare economists have given much attention to various non-utilitarian values, such as individual rights, freedom and the opportunity for choice, procedural fairness and individual responsibility.Footnote 3 The purpose of this paper is to develop a general framework in which the wide range of ethical concerns that are now being explored in welfare economics can be conveniently organized and discussed. It is not our intention to prove new results in this framework. Rather, our purpose is to contribute to conceptual clarity by highlighting the structure of different categories of ethical considerations that figure in welfare economics and other related studies.
The plan of the paper is as follows. Part 2 is devoted to a detailed examination of the different aspects or features of extended social states, which may be ethically relevant. Part 3 introduces and illustrates several types of ethical evaluation of extended social states that are exclusively based on different aspects of outcomes. Part 4 illustrates various types of non-consequentialist evaluations of extended social states. Part 5 contains a summary of our discussion and presents some concluding remarks. We have kept the discussion in Parts 2–5 deliberately less formal. The formal framework underlying this discussion is presented in the Appendix. In addition to providing precise formulation of many of the ideas and concepts discussed informally in the main text, the formal framework in the Appendix also serves a substantive purpose: it allows us to distinguish precisely the many different forms of non-consequentialism, which are rather difficult to capture in an informal discussion.
2. ETHICAL EVALUATION OF EXTENDED SOCIAL STATES
Consider a given society N = {1, 2,. . .,n}, where 1, 2,. . ., n are the individuals in the society. The set of all conceivable physically specified outcomes (or, ‘physical outcomes’ for short) for the society will be denoted by X. We use the term ‘physical outcome’ in the ordinary sense of the term while recognizing that the ordinary usage involves some ambiguities (we take up this issue later), and we use the terms ‘physical outcome’ and ‘physical consequence’ interchangeably.
Our basic concern is with the values that may enter an individual's ethical evaluation of extended social states (the term extended social states will be defined more precisely later). We now outline the different factors that such an ethical evaluator (EE) may consider to be relevant for her evaluation of extended social states.
2.1 The history
In our framework, the society is associated with a given history; the history includes a description of the situation in the society at the point of time where the analysis starts as well as a description of all that has happened in the society prior to that point of time. The history of the society is given and unalterable, though, of course, one can ask counterfactual questions about the society with a different (hypothetical) history. In the standard general equilibrium model of a private ownership economy, the history is given simply by a specification of the initial endowments that the different consumers have inherited from the past, but, in other contexts, one may like to use a much richer description of what has happened prior to the point of time where the social decision process starts. Though the society cannot choose its history, the history can play an important role in the evaluation of extended social states.
We would like to clarify that our notion of history is different from the notion of history as it is used in game theory. In game theory, the ‘history’ of a non-initial node in a game tree denotes the unique path that connects the initial node of the tree to the specific node under consideration. Our notion of history is unrelated to this game theoretic notion of history; we use the term ‘history’ as a primitive term.Footnote 4
Let H denote the set of different conceivable histories of the society. Elements of H will be denoted by h, h′, etc. In light of the clarificatory remark in the previous paragraph, H is to be regarded as exogenously given and independent of game trees.
2.2 The institutional framework
The individuals in our society function in a given institutional framework. An institutional framework may have been inherited from the past and the society may decide (explicitly or implicitly) to continue with the inherited institutional framework. Alternatively, the society may decide, explicitly or implicitly, to change the inherited institutional framework. In the appendix, we introduce a formal definition of an institutional framework as a normal game form. From an intuitive point of view, however, institutional frameworks, to be denoted by G, G′, etc. can be understood broadly as alternative specifications of the rules and regulations that govern the society and that determine what the individuals may or may not do and the physical outcomes that will result corresponding to any given configuration of strategies or actions that the individuals may choose. Thus, in a society, where the political, legal and social institutions incorporate the individual's right to practice any religion of his or her choice, the institutional framework will specify what penalties an employer will incur if he/she discriminates against an employee for adopting a particular religion. In a society, where individuals have the right to get free medical care, the rules specify the procedure that an individual must follow to get free medical care if she is ill and the redress that she is entitled to if she is not given such medical care in a timely fashion.
A change in social policy can be thought of as a switch from one institutional framework to another. The change may be a ‘radical’ change, such as a change from central planning to free markets or a change from share cropping to complete absence of share cropping. Alternatively, it may involve ‘relatively minor’ changes, such as a change in the rates of income taxes or a change in the legally prescribed share of the tenant from half to two-thirds.
Given a history h, let X(h) be the set of all physical outcomes, which are feasible given the history h, and let 
${\cal G}$(h) be the set of all institutions which are physically feasible given the history h.Footnote 5 Note that, given a particular history, certain institutions may be infeasible. Thus, a society that starts with a very small historically given initial endowment of resources and a low level of technology may find it impossible to have an institutional framework where every worker is assured of a generous minimum wage and other employment benefits. Note that an institutional framework which may be feasible given a history h may not be physically feasible for a different history h′.
2.3. Strategies
Given a history h ∈ H, and given an institutional framework ${\cal G}$(h), the individuals choose their respective strategies or plans for actions. The specific strategies that may be chosen by the individuals in the society may, by themselves, be relevant for the ethical evaluation of social affairs, so we shall incorporate such strategies directly in the description of extended social states that may be evaluated ethically.
2.4 Outcomes and personal well-being
The notion of the outcome or consequence of a social policy is rather ambiguous. It is possible to take the position that any aspect of extended social states that may be relevant for an individual's ethical assessment should be a part of the description of the outcome or consequence. Taking this line of thinking to its logical end, one can say that, if the history of the society or the institutional framework or the strategies adopted by the individuals matter for an ethical observer's ethical evaluation of the affairs in the society in any way, then it should also be a part of the description of the consequence or the outcome. While this claim has some plausibility, this broad use of the term does not conform to its use either in ordinary language or in much of the literature in philosophy. Further, even if we incorporate everything that is relevant for the ethical observer's evaluation in the notion of an outcome, there will still remain a conceptual need to distinguish between the different ingredients of this all-encompassing notion of an outcome. We have, therefore, chosen to retain the distinction between the final outcomes for the society on the one hand and the society's history, the institutional framework, and the strategies of the individuals on the other hand.
So far as outcomes are concerned, we would like to make a distinction between two distinct aspects of an outcome, namely, the physical outcome and the personal well-being of the individuals. In the economic literature on the competitive market mechanism, we have the distinction between an ‘allocation’, which is a specific version of our notion of physically specified outcomes, and the utilities that the individuals derive from the allocation, the utility of an individual being a specific interpretation of the individual's well-being. Three points should be noted here. First, in general, the physical outcome need not necessarily be just an allocation as it is defined in the standard theory of general equilibrium. Thus, if we like, the heights and weights of children in the society, the statistics regarding the illness of people, the crime rates, etc. can all be parts of the description of a physical outcome for the society. Second, while economists have typically identified the utility of an individual with her well-being, the concept of utility itself has been interpreted in different ways at different times and by different writers.Footnote 6 Thus, while retaining the same term, ‘utility’, the intuitive content of the term has been changed from time to time to mean, variously, pleasure and the absence of pain, happiness, and fulfilment of desires or preference satisfaction, the last interpretation being the one usually used in modern microeconomic theory. Finally, there are conceptions of personal well-being, which are very different from the utility-based conception of personal well-being, no matter which interpretations of utility one decides to adopt. An important example of such conceptions is the notion of well-being due to Nussbaum (Reference Nussbaum1988, Reference Nussbaum2000) and Sen (Reference Sen1985, Reference Sen1987), where the well-being of an individual is based on the bundle of functionings achieved by the individual and her capability set (i.e. the set of all functioning bundles available to her), functionings being the ‘doings’ and ‘beings’ that people value (see Sen Reference Sen1985, Reference Sen1987).
For our purpose, it is not necessary to commit ourselves to a specific conception of an individual's personal well-being. It is, however, important not to blur the distinction between an individual's well-beingFootnote 7 and her ethical preferences.Footnote 8 An individual may subscribe to the ethical position that there should not be much inequality in income distribution in the society or that tigers have a right to survive and the society must not allow them to become extinct, but these ethical preferences have very little to do with the individual's personal well-being as such (this is not to deny that, if an individual has sympathy for tigers, then their survival may be a relevant consideration for the individual's well-being). If j happens to be the ethical evaluator (EE) of social affairs and i is an individual in the society under consideration, then it is plausible to say that, in arriving at his ethical evaluation of social states, j should take into account i‘s personal interest or well-being and it is also plausible to say that j may use her own value judgements regarding the ethical undesirability of inequality of income distribution or the extinction of tigers, but it is not at all clear why, in giving her ethical evaluations of the social state,j should consider i‘s value judgements about the inequality in income distribution or i‘s value judgements about the extinction of tigers. What the ethical evaluator needs is information about the individuals’ interest or well-being as distinct from their ethical preferences/judgements regarding social states.
We assume that, in general, the well-being of each individual i depends on the institutional structure and the strategies used by all the individuals in addition to depending on the physical outcome that emerges as the result of these strategies, given the institutional structure. More formally, we assume that, for every individual i, there exists an exogenously given real-valued function wi defined over the class of all ordered triples (G, s, x), such that, for some history h ∈ H, G ∈ ${\cal G}$(h), s = (s 1, . . ., sn) is a configuration of the individuals’ strategies that can arise in the institutional framework G, and x is the physical outcome that emerges, given the institutional set up G, and the strategies, s 1, . . ., sn of the individuals. Intuitively, wi(G, s, x) indicates the level of well-being that individual i achieves from (G,s,x). It is usual to assume that the well-being of an individual is determined exclusively by the physical outcome. Therefore, our assumption that an individual's well-being may also depend directly on the institutional structure and the strategies adopted by all individuals needs some explanation.
A little reflection will show that, at least under some conceptions of well-being, an individual's well-being may directly depend on the institutional structure and the strategies of all individuals. Consider the conception of an individual's well-being in terms of utility, where utility is interpreted as achieved happiness. Consider a change in the institutional structure, which gives the individuals a much wider range of strategies implying a much greater degree of freedom. It is possible that some individuals may find this increased freedom burdensome by itself and this may reduce their utility independently of any change in the physical outcome that emerges finally. An extreme case of this possibility is illustrated in the film Sophie's Choice, where the Nazi officer gives the mother the choice of which of her two children must die. Independently of the final outcome that one of her two children will die, the fact that she has to choose which of her two children must die is, in itself, a direct cause of the mother's anguish. In this case, to assume that the mother's utility depends exclusively on the final physical outcome will miss an important causal factor (namely, the decision-making mechanism itself) that affects the mother's utility.
The strategies of people can also affect an individual's utility, apart from any influence these strategies may have on the physical outcome. Thus, if individual i is a dissident living under a totalitarian regime, i‘s neighbour, with whom i is not on friendly terms, may denounce him or i‘s friend may denounce him. In both cases, the punishment that i receives will be the same, but i‘s unhappiness may be much greater in the latter case. In general, there does not seem to be any compelling reason for ruling out the possibility that an individual's utility may be affected directly by the strategies that people adopt or the institutional framework, as well as by the physical consequence or outcome. Finally, we would like to note that, in general, we do not assume that, given a history, an institutional framework G which is feasible for the given history, and the individuals’ well-being functions, the individuals will choose their strategies so as to ‘maximize’ their respective well-being levels. Under many of the interpretations of well-being (e.g. the interpretation in terms of functionings and capability), the rigid link between an individual's well-being and her choices that is often taken for granted in standard economic theory, does not exist.
Consider a triple (G, s, x), andw(G, s, x), such that G is an institutional framework, s = (s 1, . . ., sn) is a vector of strategies available to the individuals in the institutional framework G, x is the physical outcome corresponding to G and s, and w(G, s, x) = (w 1(G, s, x), . . ., wn(G, s, x)) is the vector of the individuals’ well-being levels corresponding to G, s and x. Then we shall call w(G, s, x) the well-being outcome corresponding to G and s, and we shall call 
$\epsfbox{art/x}$ ≡ (x, w(G, s, x)) an outcome corresponding to G and s. We shall call 
$({G,s,\epsfbox{art/x}})$ an extended social state. For brevity, we shall denote extended social states by 
$\epsfbox{art/a}$, 
$\epsfbox{art/b}$, $\epsfbox{art/a}$′, $\epsfbox{art/b}$′, etc. For all h ∈ H and all extended social states = $({G,s,\epsfbox{art/x}})$, we say that a is h-feasible if and only if G ∈ G(h). For all h ∈ H, let Yh denote the set of all h-feasible extended social states.
2.5 Ethical evaluation of extended social states
We assume that, for every given history h, the EE evaluates the h-feasible extended social states, and that such evaluation is represented by a reflexive and transitive binary relation 
$\succurly{177}_h$ defined over Yh. For all $\epsfbox{art/a}$, $\epsfbox{art/b}$ ∈ Yh, $\epsfbox{art/a}$$\succurly{177}_h$$\epsfbox{art/b}$ denotes that, given the history h, the EE considers the h-feasible extended social state a to be ethically at least as desirable for the society as the h-feasible extended social state $\epsfbox{art/b}$ . Our concept of the EE's ethical evaluation of alternative feasible extended social states is akin to the Bergson (Reference Bergson1938, Reference Bergson1954)–Samuelson (Reference Samuelson1947) concept of a social welfare function (see also Little's (Reference Little1952) discussion of the concept of social welfare) and Harsanyi's (Reference Harsanyi and Williams1982) concept of an individual's ‘moral preferences’), which represent an individual's value judgements about social welfare, as distinct from Arrow's (Reference Arrow1951) notion of a social welfare function as a method of aggregating people's opinions/judgements/preferences.Footnote 9 Note that, for each history, the EE's ranking is defined over the set of extended social states which are feasible for that history. Our formal framework does not allow ‘across-history’ comparisons of the following type: the ethical desirability of $\epsfbox{art/a}$ given the history h is greater than the ethical desirability of $\epsfbox{art/a}$ ′ given the history h′, but, if one likes, our formal framework can be easily adapted to allow across-history comparisons by the EE.
An extended social state is a specification of an institutional framework, a vector of strategies, a physical outcome, and a vector of individual well-being levels, and, for each history h ∈ H, the EE ranks the h-feasible extended social states. The different ethical theories regarding the evaluation of extended social states differ with respect to which of these five features – the institutional framework, the strategies, the physical outcome, the well-being vector and history – are regarded as directly relevant for the EE's rankings or evaluations. There is wide consensus, at least among economists, that the individual well-being levels are directly relevant for the EE's evaluations; the difference seems to arise mainly about the role of the other features. But what exactly is meant by saying that some of these other features are directly relevant for the EE's rankings? Note that, in our model, each individual's well-being function is a function of the institutional framework, the strategies and the physical outcome. Therefore, the institutional framework, the strategies and the physical outcome, may have indirect relevance for the EE's rankings via their possible effects on the individuals’ well-being levels if, as is usually the case, the EE's evaluations of the h-feasible extended social states take the vector of individual well-being levels into consideration. The institutional framework, the strategies, the physical outcome and the history may, however, impact the EE's evaluations of extended social states directly, and therefore, those features can have direct relevance for the EE's rankings, or, equivalently, the EE's rankings may be directly dependent on those features. Let us illustrate the notion of ‘direct dependence’ or ‘direct relevance’ by considering a specific feature, namely, the institutional framework.
The EE's evaluations are directly institution-dependent if and only if there exist h ∈ H and h-feasible extended social states $\epsfbox{art/a}$, $\epsfbox{art/b}$, $\epsfbox{art/a}$′ and $\epsfbox{art/b}$′ such that $\epsfbox{art/a}$ and $\epsfbox{art/a}$ ′ are identical with respect to the strategies and also outcomes figuring in them, and, similarly, $\epsfbox{art/b}$ and $\epsfbox{art/b}$ ′ are identical with respect to the strategies and outcomes, but the ranking of $\epsfbox{art/a}$ and $\epsfbox{art/b}$ in terms of $\succurly{177}_h$ is not exactly analogous to the ranking of $\epsfbox{art/a}$ ′ and $\epsfbox{art/b}$ ′ in terms of $\succurly{177}_h$.
Intuitively, if the EE's evaluations are directly institution-dependent, then, for some history h, the EE takes into account the institutional framework in ranking some h-feasible social states. We have intuitively illustrated the notion of ‘direct institution-dependence’. In general, one can analogously formulate the notions of direct dependence of the EE's evaluations on features other than individual well-being levels (namely, the history, the institutions, the strategies and the physical outcome), or different combinations of these other features; precise definitions of these different types of direct dependence of the EE's evaluations on features other than individual well-being levels are given in the Appendix. These notions help clarify the exact nature of the differences between various ethical positions that people may take when they go beyond individual well-being levels in their assessment of social affairs or public policies.
3. EVALUATION OF EXTENDED SOCIAL STATES, WHICH ARE EXCLUSIVELY UTILITY-FOCUSED, EXCLUSIVELY WELL-BEING-FOCUSED OR CONSEQUENTIALIST
We now consider how different ethical theories relating to the evaluation of extended social states fit in our framework, depending on what they assume about the dependence of ethical evaluations on different subsets of the set of features that we have discussed earlier, namely, the history, the institutional framework, the strategy vector, the physical outcome, and the well-being vector. In principle, one can think of an elaborate taxonomic scheme, consisting of a large number of different categories of ethical theories; we discuss them in a formal fashion in the Appendix. We note here only some of the most conspicuous of these categories.
We begin with utilitarianism, which identifies the well-being of a person with her utility and ranks extended social states exclusively in terms of the sum total of all individuals’ utilities, corresponding to the different extended social states. Note that the notion of utility itself can have different interpretations, such as happiness, preference satisfaction, etc. though, in the recent economic literature, utility has been usually interpreted as preference satisfaction. Depending on which interpretation of utility is adopted, one can have different versions of utilitarianism.
Utilitarianism is just one example of a broader class of evaluation rules that can be called exclusively focused on utility: under these rules, extended social states are ranked exclusively on the basis of the well-being vectors figuring in them, with the well-being of an individual being identified with her utility. Evaluation rules which are exclusively focused on utility include rules of evaluation other than utilitarianism. For example, when the EE evaluates extended social states on the basis of the minimums of the individual utilities corresponding to different extended social states, the evaluation is exclusively focused on utility, but is not utilitarian.
Traditional evaluation methods used in economics have been exclusively focused on the well-being of individuals, with well-being being interpreted as utility. But, in recent decades, welfare economics has gone beyond the notion of individual well-being as utility and has developed other notions of individual well-being. An important example of this is to be found in Sen's (Reference Sen1985, Reference Sen1987) functioning and capability approach to individual well-being.Footnote 10 If we take a broader notion of individual well-being, we can introduce the concept of evaluation with exclusive focus on personal well-being: the EE's evaluation of extended social states is exclusively focused on personal well-being if and only if extended social states are ranked exclusively on the basis of the well-being vectors figuring in them.
If the EE's evaluation is exclusively focused on personal well-being, then the only information that matters for him in comparing extended social states is the information about the individuals’ personal well-being. It is important to note that our notion of ‘exclusive focus on personal well-being’ is different from the notion of welfarism as it is often used in the recent literature on welfare economics and social choice. In this literature, the term welfarism is typically used to indicate the case where the EE's evaluation of extended social states depends exclusively on information about the individuals’ utilities. While one can use any terminology so long as the terms are clearly defined, this specific use of the term welfarism is a little misleading. Utility, in some sense or other, is not the only interpretation that one can put on the concept of an individual's welfare or personal well-being. There are other plausible interpretations of the notion of personal welfare or well-being. If the EE's evaluations exclusively focus on personal well-being in any of these other senses, there is no reason why one should not call such evaluations welfaristic as well. At the same time, we recognize that the current usage of the term welfarism to mean exclusively utility-centred ethical evaluation of extended social states is here to stay. So, to avoid confusion, we have used the term ‘exclusive focus on personal well-being’ to indicate the general case where the EE's evaluation is exclusively based on consideration of the individuals’ personal well-being, no matter what the notion of personal well-being may be.
At this point, it might be of interest to consider an example of an evaluative criterion which is focused on personal well-being but not focused on utility.
Example 3.1. The functioning and capability approach of Sen (Reference Sen1985, Reference Sen1987) and Nussbaum (Reference Nussbaum1988, Reference Nussbaum2000) conceives an individual's well-being as being determined by the individual's achieved bundle of functionings and his capability set, i.e. the set of all functioning bundles available to the individual. Suppose the EE's evaluations are exclusively focused on personal well-being and he subscribes to the Sen–Nussbaum notion of personal well-being. Then his evaluations are not exclusively focused on utility though they are exclusively focused on personal well-being.
Note that, given an extended social state $({G,s,\epsfbox{art/x}})$, the information about the achieved functioning bundle of an individual constitutes a part of the description of the physical outcome x, which figures in $\epsfbox{art/x}$, but the information required to identify an individual's capability set (i.e. the set of functioning bundles available to the individual), cannot be normally found in the description of the outcome $\epsfbox{art/x}$ since the realized outcome, by itself, does not give us much idea about the opportunities that the individual has for choosing a functioning bundle. Which other components of $({G,s,\epsfbox{art/x}})$ contain the information required to identify the capability set of an individual will clearly depend on how we conceive the capability set. When the functioning bundle that an individual can attain depends on other individuals’ strategies together with his own strategy, one can think of various ways of defining the capability set, though none of them appears to be entirely satisfactory (see Pattanaik and Xu Reference Pattanaik, Xu, Gotoh and Dumouchel2009 for a discussion of this issue). Depending on how we identify the capability set of an individual, the capability set of the individual may be determined by the institutional framework (see Pattanaik and Xu Reference Pattanaik, Xu, Gotoh and Dumouchel2009). But an ethical evaluator, who regards an individual's well-being as being determined by her achieved functioning bundle and her capability set, can still be exclusively focused on personal well-being if the only way in which the institutional framework, the strategies, and the physical outcome can influence the EE's evaluations is through their effect on the individuals’ well-being.
The above evaluations are examples of what we call consequentialist evaluations of extended social states. The EE's evaluations of extended social states are consequentialist if and only if extended social states are ranked exclusively based on their respective outcomes: if the EE's evaluation of extended social states depends exclusively on the outcomes and neither any of the other features (i.e. the history, the institutional framework and the strategy vector) in isolation nor any combination of those features has direct relevance for the EE's evaluation.Footnote 11
For illustrative purposes, we consider a prominent example of an evaluative criterion which is consequentialist but not focused on personal well-being.
Example 3.2. Consider the following reasoning, where the well-being of an individual is identified with her utility.
Usually, in evaluating extended social states, I would go by the sum of the utilities of all individuals in the society. But, in the case of the extended social states $\epsfbox{art/a}$ and $\epsfbox{art/a}$ ′, I find that, though the sum of utilities is higher in $\epsfbox{art/a}$ than in $\epsfbox{art/a}$ ′, the physical outcomes x and x′ in $\epsfbox{art/a}$ and $\epsfbox{art/a}$ ′, respectively, differ only with respect to individual i‘s religion, which I consider to be a matter of i‘s private life and i‘s utility from x′ is higher than her utility from x. Therefore, in comparing $\epsfbox{art/a}$ and $\epsfbox{art/a}$ ′, I shall completely ignore the utilities of individuals other than i and, going by i‘s utility, I shall rank $\epsfbox{art/a}$ ′ higher than $\epsfbox{art/a}$.
The reasoning above basically reflects the intuition underlying Sen's (Reference Sen1970) famous condition of minimal liberalism, which Sen considered to be a necessary condition for the existence of rights, especially rights to liberty in personal matters . Note that, besides the utilities of the individuals, the features of the physical outcomes also matter for the EE. So, while the EE is a consequentialist, his evaluations are not exclusively focused on utility.
4. NON-CONSEQUENTIALIST EVALUATIONS OF EXTENDED SOCIAL STATES
Consequentialist evaluations of extended social states focus exclusively on the outcomes figuring in extended social states. Sometimes, however, the EE's evaluations of extended social states may not be consequentialist. When the EE's evaluations of extended social states are not consequentialist, we say the EE is a non-consequentialist (or the evaluation is non-consequential). Non-consequentialism arises whenever the history, institutions or strategies, or some combination of them have direct relevance for the EE's evaluations apart from any significance they may have because of their effect on the outcomes. Thus, the source of non-consequentialism can be diverse (for a detailed classificatory scheme for different types of non-consequentialism, see the Appendix). We consider below only a few examples of non-consequentialist evaluation of extended social states.
Example 4.1. Historic injustices against specific social groups, such as the African Americans in the USA and the lower castes in India, are often evoked as a justification for positive discrimination in favour of those groups. This has been an intensely debated issue (see, among others, Thomson Reference Thomson1973; Simon Reference Simon1974; Sher Reference Sher1975, Reference Sher1979; Goldman Reference Goldman1976; Cohen Reference Cohen2009). For our purpose, however, it is enough to note that, if the EE accepts this justification for affirmative action, her ranking of two extended social states, one with affirmative action favouring a specific social group and the other without such affirmative action, can be very different depending on whether or not there is a history of some injustice committed against the group. The EE's evaluation will then manifest non-consequentialism arising from direct history-dependence.
It may be worth noting that what Nozick (Reference Nozick1974) calls the ‘historical principles of justice’ are not necessarily examples of what we have called direct history-dependence of ethical evaluations. This is because Nozick uses the term ‘history’ to include all the circumstances (i.e. history as we use the term, the institutions, and the strategies that the players use) that lead to the final outcome. Nozick's position can involve non-consequentialist evaluations of various types – history may play a direct role, strategies may play a direct role, and/or institutional frameworks may play a direct role in the evaluations.
Example 4.2. Consider two extended social states, $\epsfbox{art/a}$ and $\epsfbox{art/b}$, which are identical in all respects except for that, in $\epsfbox{art/a}$, women have the legal right to join the army but choose not to join the army, while, in $\epsfbox{art/b}$, women are prevented by law from joining the army. Many people, who believe that gender-based legal discrimination is intrinsically ethically reprehensible, would rank $\epsfbox{art/a}$ higher than $\epsfbox{art/b}$ . That would be an instance of non-consequentialism arising from direct institution-dependence.
Example 4.3. Various theories of distributive justice based on responsibility advocate a ‘reward principle’ in which redistributions are based on individual responsibility (see, among others, Cohen Reference Cohen1989; Arneson Reference Arneson1990; Roemer Reference Roemer1998; Fleurbaey Reference Fleurbaey2008; Fleurbaey and Maniquet Reference Fleurbaey, Maniquet, Arrow, Sen and Suzumura2009): individuals should be compensated ‘only for those welfare deficits which are not in some way traceable to the individual's choices’ (Cohen Reference Cohen1989: 914), and ‘distributive justice does not recommend any intervention by society to correct inequalities that arise through the voluntary choice or fault of those who end up with less, so long as it is proper to hold the individuals responsible for the voluntary choice or fault behavior that gives rise to the inequalities’ (Arneson Reference Arneson1990: 176). Suppose the EE takes this position. Then the EE may not ethically approve of financial assistance from the government to people facing foreclosure of their homes if they face such foreclosure because they secured high mortgages from banks by overstating their incomes, and the EE may approve of such assistance if the people facing foreclosure of their homes do so because of circumstances beyond their control. In that case, the EE's evaluation of extended social states will involve non-consequentialism arising from direct strategy-dependence.
Clearly, depending on the roles that the history and/or various components in an extended social state play in the evaluation of extended social states, we can have different types of non-consequentialism. As we indicate in the Appendix, in principle one can think of as many as 127 different types of non-consequentialism. It is not practicable to give an example of each of these 127 categories of non-consequentialism. The above examples illustrate just a few of the important forms of non-consequentialism.
From our discussion, it is clear that the distinction between consequentialism and non-consequentialism depends crucially on the convention regarding the use of the term ‘consequence’ or ‘outcome’. As we noted earlier, in principle it is possible to take the position that anything that matters for the EE's evaluation of extended social states should be built into the notion of consequences. If one takes this position, then, of course, it would no longer make any sense to distinguish between consequentialism and non-consequentialism. We have two observations on this position. First, even if one chooses to broaden the concept of a consequence in this fashion, it will still be necessary to discuss why any of the features, which figure in our framework, namely, the history, the institutional structure, the strategies, and what we called an outcome, should (or should not) be considered directly relevant for the EE's evaluations, and, hence, should (or, should not) be a part of the wider notion of a consequence. Second, given that the term consequence has been widely used in a somewhat narrow sense, there may be a case for not departing from this long-established tradition unless the departure contributes to conceptual clarity or analytical convenience. In view of both these considerations, we have chosen to retain the narrow conception of consequence and, hence, the distinction between consequentialism and non-consequentialism.
5. SUMMARY AND CONCLUSION
In this paper, we have introduced a conceptual framework, which can accommodate the wide range of value judgements used in ethical evaluations of extended social states and which can be used to differentiate the different categories of value judgements by referring to the type of information on which they may be based. In the process, we have tried to conceptualize the notions of consequentialism, non-consequentialism, exclusive focus on personal well-being, exclusive focus on utility, etc. in operational ways. We believe that our framework and our discussions of different types of ethical criteria that may be used in evaluating extended social states can contribute to conceptual clarity about the ethical basis of welfare economics.
To put our analysis and conceptual framework in perspective, we note that, while the conventional normative economic analysis of alternative social policies has yielded numerous valuable insights, it has relied on too narrow an ethical foundation insofar as, not only has it been consequentialist, but it has also exclusively focused on utility usually interpreted as preference satisfaction and, in addition, has often assumed that utility depends only on allocations (allocations being defined as in general equilibrium theory). This has proved restrictive in several ways. First, it ignores the fact that the institutional framework, the strategies adopted by individuals, and the physical outcomes often have intrinsic ethical importance for many people, independently of their effect on the well-being of individuals, and it also ignores the role of history in people's ethical judgements about social affairs. Second, it does not distinguish between an individual's personal well-being and the fulfilment of her ethical preferences (e.g. her ethical preferences for a more equitable allocation), nor does it distinguish between an individual's impulsive or fleeting desires and her considered preferences. Third, the assumption that an individual's utility depends only on the allocation ignores the fact that institutions and the actions of people may directly impact an individual's utility, apart from their effect on the physical outcome. In recent decades, many of these restrictive features have come under increasing criticism in welfare economics and behavioural economics. In particular, a number of writers have sought to broaden the ethical basis of welfare economics in different ways by introducing explicitly ethical considerations such as individual rights, freedom and procedural fairness.
Our main purpose in this paper has been to develop a comprehensive structure which can accommodate in a systematic fashion this wider range of ethical concerns regarding public policies and which will be useful in identifying the exact nature of the differences that exist between different types of ethical theories regarding the assessment of social affairs. In doing so, we have extended the conceptual framework of some earlier contributions, such as Pattanaik and Suzumura (Reference Pattanaik and Suzumura1996), who focused on the roles of rights and outcomes in social welfare judgements, and Suzumura and Xu (Reference Suzumura and Xu2001, Reference Suzumura and Xu2003), who incorporate freedom and opportunities of choice into evaluations of social alternatives. In the latter contribution, the notion of consequentialism is visualized as the evaluation that focuses exclusively on social alternatives and ignores the information about freedom and opportunities of choice completely. It may be noted that the notion of social alternatives is taken as a primitive in the contribution by Suzumura and Xu (Reference Suzumura and Xu2001, Reference Suzumura and Xu2003), and consequently, depending on one's take on the notion of social alternatives, their notion of consequentialism may not coincide with the notion of consequentialism elaborated in this paper.
APPENDIX
Let N = {1, 2, . . ., n} denote the society, where 1, 2,. . ., n are the individuals in the society; let X denote the set of all conceivable physically specified outcomes (physical outcomes for short) for the society; and let H denote the set of different conceivable histories of the society, the elements of H being denoted by h, h′, etc.
Starting with a given history, the society decides what institutional framework it will have. We visualize an institutional framework as a normal game form, where the society, N, is the set of players.
For all h ∈ H, let X(h) be the set of all physical outcomes, which are feasible, given the history h. For all h ∈ H, let ${\cal G}$(h) be the set of all institutions, that is, all normal game forms G = (N, S 1, . . ., Sn, f), such that: (i) N, the society, is the set of players; (ii) for every i ∈ N, Si is the (non-empty) set of strategies available to player i; and (iii) f is the outcome function, which, for every s ∈ S ≡ S 1 × S 2 × . . . × Sn, specifies exactly one physical outcome x in X(h).Footnote 12${\cal G}$(h) is interpreted as the set of all game forms or institutional frameworks which are physically feasible, given the history h. Note that a game form which may be feasible for a given history h may not be feasible for another history h′.
Given a history h ∈ H, and given an institutional framework G ∈ ${\cal G}$(h), the individuals choose their respective strategies. We assume that, for every i ∈ N, there exists an exogenously given real valued well-being function wi defined over the set of all ordered triples (G, s, x) such that, for some h ∈ H, G ∈ ${\cal G}$(h), s is an n-tuple of strategies figuring in G, and x is the physical outcome specified by the function f for s. Let h ∈ H, G ∈ ${\cal G}$(h), s be an n-tuple of strategies figuring in G ∈ ${\cal G}$(h), and x be the physical outcome corresponding to s. We denote (w 1(G, s, x), . . ., wn(G, s, x)) by w(G, s, x) and we say that $\epsfbox{art/x}$ ≡ (x, w(G, s, x)) is the outcome corresponding to (G, s). For every h ∈ H, let Yh be the set of all triples $({G,s,\epsfbox{art/x}})$, such that G ∈ ${\cal G}$(h), s is an n-tuple of strategies figuring in the game form G, and $\epsfbox{art/x}$ = (x, w(G, s, x)) is the outcome corresponding to (G, s). A triple $({G,s,\epsfbox{art/x}})$ such that, for some h ∈ H, $({G,s,\epsfbox{art/x}})$ ∈ Yh is called an extended social state. Extended social states will be denoted by $\epsfbox{art/a}$, $\epsfbox{art/a}$′,$\epsfbox{art/b}$,$\epsfbox{art/b}$′ etc. For each h ∈ H, let $\succurly{177}_h$ be a reflexive and transitive, but not necessarily connected, binary relation over Yh; $\succurly{177}_h$ is interpreted as the EE's ranking of the various extended social states in Yh, given the history h.Footnote 13
Let $\epsfbox{art/a}$ = $({G,s,\epsfbox{art/x}})$and $\epsfbox{art/b}$ = (G ′, s ′, $\epsfbox{art/x}$′) be two extended social states. We say that $\epsfbox{art/a}$ and $\epsfbox{art/b}$are identical with respect to institutions (I-identical) iff G = G′; and we say that $\epsfbox{art/a}$ and $\epsfbox{art/b}$ are different with respect to institutions (I-different) iff G ≠ G′. Similarly, we have the notions of $\epsfbox{art/a}$ and $\epsfbox{art/b}$ being: identical with respect to strategies (ST-identical); different with respect to strategies (ST-different); identical with respect to outcomes (O-identical); different with respect to outcomes (O-different); and identical with respect to well-being levels (W-identical)
Definition 7.1 (Consequentialism and Non-consequentialism). We say that the EE's ethical evaluations of extended social states are consequentialist (or, equivalently, the EE is a consequentialist) if and only if, for all h, h′ ∈ H, all $\epsfbox{art/a}$, $\epsfbox{art/b}$ ∈ Yh, and all $\epsfbox{art/a}$′, $\epsfbox{art/b}$′ ∈ Yh ′, if $\epsfbox{art/a}$ and $\epsfbox{art/a}$ ′ are O-identical and $\epsfbox{art/b}$ and $\epsfbox{art/b}$ ′ and $\epsfbox{art/b}$ ′ are O-identical, then [a $\succurly{177}_h$$\epsfbox{art/b}$iff $\epsfbox{art/a}$′$\succurly{177}_{h}$′$\epsfbox{art/b}$′] and [$\epsfbox{art/b}$$\succurly{177}_h$$\epsfbox{art/a}$ iff $\epsfbox{art/b}$′$\succurly{177}_{h}$′$\epsfbox{art/a}$′]. The EE's evaluations of extended social states are said to be non-consequentialist (or, equivalently, the EE is said to be a non-consequentialist) iff those evaluations are not consequentialist.
Consequentialism can take various special forms, and we define some of them next.
Definition 7.2 (Exclusive Focus on Personal Well-Being of individuals). We say that the EE's evaluation of extended social states shows exclusive focus on the personal well-being of individuals (or, equivalently, the EE is exclusively focused on the personal well-being of individuals) iff, for all h, h′ ∈ H, all $\epsfbox{art/a}$, $\epsfbox{art/b}$ ∈ Yh, and all $\epsfbox{art/a}$′, $\epsfbox{art/b}$′ ∈ Yh ′, if $\epsfbox{art/a}$ and $\epsfbox{art/a}$ ′ are W-identical and $\epsfbox{art/b}$ and $\epsfbox{art/b}$′ are W-identical, then [$\epsfbox{art/a}$$\succurly{177}_h$$\epsfbox{art/b}$ iff $\epsfbox{art/a}$′$\succurly{177}_{h}$′$\epsfbox{art/b}$′] and [$\epsfbox{art/b}$$\succurly{177}_h$$\epsfbox{art/a}$ iff $\epsfbox{art/b}$′, $\succurly{177}_{h}$′, $\epsfbox{art/a}$′].
Definition 7.3 (Exclusive Focus on Utility). The EE's evaluation of extended social states shows exclusive focus on utility iff it shows exclusive focus on the personal well-being of individuals and the EE considers the personal well-being function of every individual to be that individual's utility function.
If the EE interprets utility as happiness (resp. preference satisfaction), then exclusive focus on utility constitutes exclusive focus on happiness (resp. exclusive focus on preference satisfaction).
Definition 7.4 (Utilitarianism). The EE is a utilitarian if and only if the EE's evaluations are exclusively focused on utility and, for all h ∈ H and for all $\epsfbox{art/a}$ = (G, s, (x, w (G,s,x))) and $\epsfbox{art/a}$′ = (G ′,s ′,(x ′,w(G ′,s ′,x ′))) in Yh, $\epsfbox{art/a}$$\succurly{177}_h$$\epsfbox{art/a}$′ iff 
$\mathop \sum \nolimits_{i \in N} w^i (G,s,x) \ge \mathop \sum \nolimits_{i \in N} w^i (G^{\prime},s^{\prime},x^{\prime})$.
Non-consequentialism can take different forms. To distinguish these different forms, we introduce formal definitions of the different ways in which the EE's evaluations may be directly dependent on different non-empty subsets of the set of features consisting of the history, the institutional structure and the strategy vector.
Definition 7.5.
(i) The EE's evaluations are directly history-dependent if and only if there exist distinct h, h′ ∈ H and $\epsfbox{art/a}$, $\epsfbox{art/b}$
$\in Y_h \mathop \cap \nolimits^ Y_{h^{\prime}},$ such that not [($\epsfbox{art/a}$$\succurly{177}_h$$\epsfbox{art/b}$ iff $\epsfbox{art/a}$$\succurly{177}_{h}$′$\epsfbox{art/b}$) and ($\epsfbox{art/b}$$\succurly{177}_h$$\epsfbox{art/a}$ iff $\epsfbox{art/b}$$\succurly{177}_{h}$′$\epsfbox{art/a}$). The EE's evaluations are directly institution-dependent iff there exist h ∈ H and $\epsfbox{art/a}$, $\epsfbox{art/b}$,$\epsfbox{art/a}$′, $\epsfbox{art/b}$′ ∈ Yh, such that $\epsfbox{art/a}$ and $\epsfbox{art/a}$ ′ are ST-identical and also O-identical; $\epsfbox{art/b}$ and $\epsfbox{art/b}$ ′ are ST-identical and also O-identical; and $\epsfbox{art/a}$ and $\epsfbox{art/a}$ ′ are I-different or $\epsfbox{art/b}$ and $\epsfbox{art/b}$ ′ are I-different]; and not [($\epsfbox{art/a}$$\succurly{177}_h$$\epsfbox{art/b}$ iff $\epsfbox{art/a}$′$\succurly{177}_h$$\epsfbox{art/b}$′) and ($\epsfbox{art/b}$$\succurly{177}_h$$\epsfbox{art/a}$ iff $\epsfbox{art/b}$ ′ $\succurly{177}_h$$\epsfbox{art/a}$′)] The definition of direct strategy-dependence is exactly analogous to that of direct institution-dependence.
(ii) The EE's evaluations are directly (history+institution)-dependent iff there exist distinct h, h′ ∈ H, $\epsfbox{art/a}$, $\epsfbox{art/b}$, ∈ Yh, and $\epsfbox{art/a}$ ′, $\epsfbox{art/b}$ ′ ∈ Y h′, such that [$\epsfbox{art/a}$ and $\epsfbox{art/a}$ ′ are ST-identical and also O-identical; $\epsfbox{art/a}$ and $\epsfbox{art/a}$ ′ are ST-identical and also O-identical; and $\epsfbox{art/a}$ and $\epsfbox{art/a}$ ′ are I-different or $\epsfbox{art/b}$ and $\epsfbox{art/b}$ ′ are I-different]; and not [($\epsfbox{art/a}$$\succurly{177}_h$$\epsfbox{art/b}$ iff $\epsfbox{art/a}$ ′ $\succurly{177}_{h}$′$\epsfbox{art/b}$ ′) and ($\epsfbox{art/b}$$\succurly{177}_h$$\epsfbox{art/a}$ iff $\epsfbox{art/b}$ ′ $\succurly{177}_{h}$′$\epsfbox{art/a}$ ′). The definitions of direct (history+strategy)-dependence and direct (institution+strategy)-dependence are similar to that of (history+institution)-dependence.
(iii) The EE's evaluations are directly (history+institution+strategy)-dependent iff there exist distinct h, h′ ∈ H, $\epsfbox{art/a}$, $\epsfbox{art/b}$, ∈ Yh, and $\epsfbox{art/a}$ ′, $\epsfbox{art/b}$ ′ ∈ Y h′, such that [$\epsfbox{art/a}$ and $\epsfbox{art/a}$ ′ are O-identical and $\epsfbox{art/b}$ and $\epsfbox{art/b}$ ′ are O-identical; $\epsfbox{art/a}$ and $\epsfbox{art/a}$ ′ are I-different or $\epsfbox{art/b}$ and $\epsfbox{art/b}$ ′ are I-different; and $\epsfbox{art/a}$ and $\epsfbox{art/a}$ ′ are ST-different or $\epsfbox{art/b}$ ′ and $\epsfbox{art/b}$ ′ are ST-different]; and not [($\epsfbox{art/a}$$\succurly{177}_h$$\epsfbox{art/b}$ iff $\epsfbox{art/a}$ ′$\succurly{177}_{h}$′$\epsfbox{art/b}$ ′) and ($\epsfbox{art/b}$$\succurly{177}_h$$\epsfbox{art/a}$ iff $\epsfbox{art/b}$ ′ $\succurly{177}_{h}$′$\epsfbox{art/a}$ ′)].
If the EE is a non-consequentialist, then her evaluations must involve at least one of the seven different types of direct dependence introduced in Definition 7.5, namely, direct history-dependence, direct institution-dependence, direct strategy-dependence, direct (history+institution)-dependence, direct (history+strategy)-dependence, direct (institution+ strategy)-dependence, and, finally, direct (history+institution+ strategy)-dependence. It is, however, possible for the EE's evaluations to involve simultaneously any combination of these seven different types of direct dependence. Thus, in principle, our framework allows us to distinguish 127 (= 27 − 1) mutually exclusive categories of non-consequentialist evaluations, the categories being specified in terms of the possible sources of such non-consequentialism.
