Nothing excites intellectual curiosity more than the overturning of a time-honoured belief. This is specially so when that overturning is accomplished by scholarly analysis, as distinct from Messianic rhetoric. So when as highly acclaimed a scholar as Amartya Sen challenged the popular belief that famine means shortage of food, it inevitably caused a stir. The famine that had killed two to three million people and brought starvation to millions more in Bengal in 1943 was not, he maintained, a result of shortage of food (Sen 1976, 1977). What's more, he went on to argue, the Bengal famine was by no means unique in this regard. He showed that many contemporary famines in Asia and Africa shared this property of not being caused by reduced availability of food (Sen 1981b). Famine, he concluded, is a case of people not having enough food to eat, but not necessarily of there not being enough food to go around. From this emerged what has come to be known as the 'entitlement approach' to hunger and famine -- an approach that focuses attention on people having or not having enough command over food as distinct from there being or not being enough food to be eaten. Over the years, Sen and others following his lead have extended the reach of the entitlement approach from its initial concern with the genesis of famine. In a recent treatise, for example, Dreze and Sen (1989) have skilfully utilized the insights of this approach to shed radically new light on the policy issues relating to famine relief and the more widespread problem of combating endemic hunger. To many, these insights have forever changed the way they perceive the problems of hunger and famine. I have benefited from helpful comments of Dersch Endale and Amartya Sen on an earlier draft, but I am alone responsible for the views and interpretations contained in the paper. As Robert Solow remarks in his review of the Dreze-Sen book, "It has changed the wav 1 will think about famine relief from now on." (Solow 1991, p.23.) 2 But, for all the adulation it has received, the entitlement approach has not gone unchallenged. Although it is perhaps fair to say that those wh o have delved into Sen's copious writings on this matter have generally come out impressed with his arguments, a significant strand of critical reaction has persisted to this day. A major objective of this paper is to assess the merit of this critical literature with a view to forming a judgement as to where exactly the entitlement approach now stands. However, I am not going to attempt a comprehensive assessment of the whole of the critical literature. It is convenient to divide up this literature into two parts: one that questions the analytical merit of the entitlement approach as a tool for understanding famines in general, and one that takes issue with Sen's empirical analysis of particular famines. The two parts are not necessarily independent, but they are nevertheless distinct. I shall concentrate on the first part, delving into the empirical literature only insofar as issues relating to particular famines are relevant for understanding disputes at the analytical level. I begin by sketching out (in Section 2) the conceptual apparatus of the entitlement approach. In doing so, I draw attention to a certain transition that seems to have occurred between Sen's earliest formulation and the later ones. I attach some importance to this transition, because I believe that the failure to notice it may be responsible at least in part for some of the confusions surrounding the entitlement approach. An attempt is then made in Section 3 to reach a clear understanding of what the entitlement approach really claims, and what it does not. Next, in Section 4, I take up the major criticisms of the entitlement approach, and evaluate their merit in the light of the preceding account of what I believe the entitlement approach to be really about. Section 5 provides a brief summary and some concluding remarks. 2. The Conceptual Apparatus of the Entitlement Approach: The basic unit of analysis is an individual person. For practical purposes, however, the analysis can also be conducted at collective levels such as household, group, or class by using the standard device of assuming a 'representative individual'. 3 The Basic Concepts The analysis is built upon three basic conceptual categories, viz. the endowment set, the entitlement-mapping (or E-mapping, for short), and the entitlement set. The endowment set is defined as the combination of all resources legally owned by a person. In this definition, 'resources' include both tangible assets such as land, equipment, animals, etc., and intangibles such as knowledge and skill, labour power, membership of a particular community, etc. Furthermore, the word legally' has to be interpreted broadly to mean conforming to established social norms and practices, and not merely to what is sanctioned formally by the state. The entitlement set is defined as the set of all possible combinations of goods and services that a person can legally obtain by using the resources of his endowment set. This cryptic definition calls for a little elaboration, however. First, the definition recognizes that from any given set of resources one may be able to obtain many different combinations of final goods and services, although at any point in time a person will be seen to be enjoying only one of those possible combinations, depending on her tastes and preferences. The entitlement set refers to all the possible combinations, not just the one actually being enjoyed. Second, resources may be used in many different ways to obtain the final goods and services. For example, a farmer may use his land, labour, and other resources to produce the food he wants; a labourer may exchange his labour power to secure his food; a fisherman may first use his labour, equipment and fishing boat to produce a catch of fish and then exchange it to get the rice he wants; an unemployed person may use his resource of 'citizenship of a welfare state' to claim a transfer of state funds in the form of unemployment benefit. These acts of production, exchange, and transfer are all different ways of using one's resources. Third, the manner in which a person uses his resources must have the sanction of the law of the The need for adopting such a broad interpretation arises from the fact that while legal ownership in the modern sense is often non-existent in traditional societies, especially in relation to landed property, some notion of ownership nonetheless exists based on conventions. 4 land, again interpreted in the broadest sense of the term; thus, the commodities he can obtain through looting (by using his muscle power in an illegal fashion) are not counted as part of the entitlement set. The entitlement mapping, or E-mapping, is simply the relationship between the endowment set on the one hand and the entitlement set on the other. Roughly speaking, it shows the rates at which the resources of :he endowment set can be converted into goods and services included in the entitlement set. For example, an E-mapping includes, for the farmer, the input-output ratios in farm production; for the labourer, the ratio betwc en money wage and the price of food i.e., the real wage rate; for the fisherman, both the input-output ratio in fishing and the relative price of fish and rice; and for the unemployed person, the rate of unemployment benefit. Thus, an E-mapping would in general have three broad components: a production component containing various input-output ratios (or, more generally, production functions), an exchange component made u p of rates of exchange involved in trading ' , and a transfer component. The following diagram shows the relationship among the three basic concepts: E-mapping Endowment _. Entitlement Note that wage employment i.e., the trading of labour-power, is a part of the exchange mapping. Ghose (1982) seems to have overlooked this point when he suggested that Seis framework of 'exchange entitlement' needs to be broadened to include analysis based on 'employment entitlement'. 4 Strictly speaking, the exchange mapping includes, in addition to rates of exchange, also any restriction that may exist on trading, such as quantity rationing, involuntary unemployment, etc. It is worth noting that the transfer component includes only those transfers to which a person is legally entitled -- for example, social security provisions of the state. This leaver out not only illegal transfers (such as, stealing and looting), but also non-entitlement transfers, such as charity. Although there is nothing illegal about receiving charity, it is rot counted as part of entitlement mapping for the simple reason that one is not legally entitled to charity, whatever may be one's view about the poor's moral entitlement to it. The general point is that entitlement analysis is concerned with legal as distinct from moral entitlement. 5 Next follows the concept of entitlement failure, which is derived from the three basic concepts and plays a crucial role in the analysis of famines. A person is said to suffer from the failure of food entitlement when her entitlement set does not contain enough food to enable her to avoid starvation in the absence of non-entitlement transfers, such as charity. It means that no matter how a person may reallocate resources to obtain the food she wants, she cannot get the minimum amount needed to escape starvation. A famine occurs when a large number of people within a community suffer from such entitlement failures at the same time. It is useful to note at this stage one important aspect of the causal structure that binds these concepts together. A moment's reflection will show that while all three basic categories can in principle affect each other, there nevertheless exists an important asymmetry between entitlement on the one hand and endowment and E-mapping on the other. In the case of both endowment and E-mapping, the definitions allow for the effect of exogenous factors i.e., it is granted that either of them may change without any prior change in any of the other two categories , but the same is not true of entitlement. Since the entitlement set is derived by applying E-mapping on the endowment set, it is only through changes in either endowment or Emapping that any change in entitlement can occur. Note that this is not a theory or a hypothesis, but simply a logical implication of the definitions. It then follows that 'entitlement failure', and thus famine, can only occur through some adverse change in either endowment or E-mapping or both. This leads to the useful organising principle that all possible causes of famines can be classified into two broad groups: one that affects the endowment set and the other that affects the entitlement mappings. In a sense, this organising principle can be regarded as the core of the entitlement approach; we shall come to appreciate its significance as we proceed further. For some purposes, a slightly different way of classifying the causal factors may be useful. Noting that E-mapping consists of three different kinds of relations, viz. production, exchange, and transfer, one can identify four distinct sources of entitlement failure. These are: endowment loss, For example, endowment may change exogenously when a farmer happens to inherit the land of his father, or E-mapping can change exogenously when, for instance, adverse whether reduces the crop output, or the government raises the price of rationed food, etc. 6 production failure, exchange failure, and transfer failure. For people who do not rely primarily on exchange to obtain their staple food, entitlement failure would occur through the first two of the four channels. This case is described by Sen as direct entitlement failure. When exchange is involved, then any one of the first three channels may act as the conduit of entitlement failure. For example, a fisherman may lose his boat (an endowment loss) which will prevent him from catching the fish that he must exchange in order to get his staple food, rice; or, his boat may be intact but his catch of fish may still be too inadequate (production failure) to be exchanged for the minimum amount of rice he needs; or, both endowment and production may remain intact, and yet he may not get enough rice because the relative price of fish has slumped (exchange failure). In all these cases, a trade entitlement failure will be said to have occurred. The point of this dichotomy between direct and trade entitlement failures is to draw attention to the fact that the genesis of famines may be very different as between subsistence and exchange economies. Direct entitlement failures have traditionally been the major cause of famine in the subsistence-oriented peasant economies of the past. But in the modern exchange economies, famines caused by trade entitlement failures is a very distinct possibility. Indeed, one of Sen's major contributions to our understanding of hunger has been to demonstrate how this distinctive mechanism has been at work in modern-day famines. Transition in the Conceptual Framework The preceding discussion has been based on the formulation presented by Sen in his book Poverty and Famines and in his subsequent writings. In both terminology and content, this formulation differs somewhat from Sen's earliest presentations of the entitlement approach (for example, in Sen 1977;. The newer framework is, in my view, more complete and consistent. But it is my impression that the transition that has taken place in the conceptual It should be noted that unlike the dichotomy between endowment failure and mapping failure, this dichotomy between direct and trade entitlement failures is neither disjoint nor exhaustive. It is not disjoint because both direct and trade entitlement failures can occur due to endowment loss or production failure. And it is not exhaustive because it leaves out the possibility of transfer failure. 7 framework has gone largely unnoticed, so that while commenting on the entitlement approach people still often cling to the older framework. This has sometimes resulted in a misunderstanding about the nature and objective of the entitlement approach, leading also to unwarranted criticisms. It is therefore necessary to clarify how the formulation has changed over time. It is useful to begin by considering the term 'exchange entitlement' — a term we have not used so far (for reasons to be explained below). It was in the language of 'exchange entitlement' — rather than 'entitlement', without the qualifier 'exchange' - that Sen originally launched his analysis of famines; specifically, famine was described as the failure of exchange entitlement. The term was defined as follows: "With an initial endowment x of commodities (including labour), the exchange entitlements offered by a particular set of market configurations (in addition to direct production possibilities) can be seen as the set S(x) of all commodity bundles that can be acquired starting from x. (Formally, therefore, the set of exchange entitlements can be seen as a mapping S(.) from a given person's endowment vectors to availability sets of commodity vectors.)" (Sen 1977, p.34) Two features of this definition are worth noting. First, Sen seems to suggest that 'exchange entitlement' stands for both S(x) and S(.) — the entitlement set and the entitlement mapping respectively, as we have called them. The first sentence points to entitlement set, but the second sentence, within parenthesis, seems to point to entitlement mapping. Secondly, the definition of exchange entitlement seems to exclude the 'production' channel of converting endowments into entitlements. Although formally both the set and the mapping were implied by the term, it is clear from Sen's subsequent remarks that his stress was on the mapping interpretation. One of the clearest examples is the following statement: "Even in an exchange economy, starvation can result from the loss of assets (including health) rather than exchange entitlement variations." (Sen 1977, p.35; emphasis added.) Sen is drawing a distinction here between starvation caused by loss of assets and starvation caused by exchange entitlement variation. But the need for this distinction would not arise if the term exchange entitlement were to refer to the set S(x), for in that case starvation caused by loss of assets would also be called starvation due io exchange entitlement variation. By making this distinction then, Sen must be implying that exchange entitlement variation refers only to the shift in S('). In other words, the term exchange entitlement is to be equated with the mapping S('), as distinct from the set S(x). The second feature of the definition — namely, the exclusion of the production channel - - also gets further support from various remarks of Sen. For example, "... famines can certainly take place without shifts in exchange entitlement. An example is a famine affecting people who typically eat what they produce, e.g. hunters, or peasants in an economy with little exchange." (Sen 1977, p. 35; emphasis original.) These two features of the original formulation imply a certain restriction on the concept of 'failure of exchange entitlement' (FEE). Insofar as exchange entitlement refers only to the mapping, starvation caused by the loss of assets would not qualify as a case of FEE; similarly, insofar as the production channel is excluded from the mapping, starvation suffered by direct producers of food due to a crop failure would not count as FEE. Accordingly, when famine is said to be caused by the failure of exchange entitlement, some categories of famine will be left out of reach of the entitlement approach. As we shall see later in the paper, this implicin restriction has been responsible for a good deal of confusion about the real message of the entitlement approach. It should be noted that the restrictive nature of this formulatior had a certain redeeming logic in the particular context in which it arose The context was the great Bengal famine of 1943, and Sen was arguing that the proximate cause of this famine was neither loss of production nor loss of assets, but adverse shift in people's command over food in the market place. Given this hypothesis, there was no great harm in excluding production loss from the analytical framework. Also, it made sense to stress the mapping interpretation of exchange entitlement, even though formally one could refer to both the set and the mapping in the same breath. There was no inconsistency in doing the latter, because if loss of assets (x) is disregarded then any variation in S(x) must come solely from variation in S('), and any variation in S(') must be reflected fully in a corresponding variation in S(x). 9 So, for the purposes of that specific empirical analysis, his formulation of exchange entitlement was not particularly problematic. But as Sen later extended his analysis to other instances of famine, where the loss of both production and assets had played a more prominent role, the need for a more general framework became obvious. It was then no longer possible to exclude production; and, moreover, if entitlement variation caused by loss of assets was to be allowed, then it would have been singularly confusing to describe both S(x) and S(') by the same term. Both these concerns are taken care of in the formulations presented in Sen's later writings, beginning with Sen (1981a, 1981b). In the first place, production is explicitly included, along with exchange and transfer, in the definition of mapping. Secondly, two distinct terms are now employed to refer to the set and the mapping. The precise manner of making this distinction has itself seems to have undergone some change over the years. The latest position, as spelt out in Dreze and Sen (1989, pp. 9-10, 23), seems to be as follows. The set is now described as the 'entitlement set' or just 'entitlement' (leaving out the qualifier 'exchange'), while the mapping is described by 'exchange entitlement', or 'exchange entitlement mapping', or simply 'E-mapping'. Famine is now defined as 'entitlement failure', rather than 'exchange entitlement failure' as in the past, thus leaving no room for doubt that famines caused by both endowment loss and mapping failure belong to the domain of entitlement analysis. We have followed this formulation closely in this paper, with the exception that the qualifier 'exchange' has been eliminated not only from the description of the set but also from that of the mapping. This has been done mainly to avoid any confusion that might arise from the term's original association with a more restricted framework of analysis. To recapitulate, the original framework was restricted in two ways: 1) entitlement failure was seen to arise solely from variation in entitlement There is also a second, essentially semantic, reason for avoiding it. Since exchange is only one of the three components of mapping (the other two being production and transfer), it does not seem very illuminating to use 'exchange entitlement' as a synonym for 'entitlement mapping'. Sen of course argues that production can also be seen as an exchange — to wit, an exchange with nature; but that still leaves out transfer, which is typically unilateral in character and thus rather hard to be seen as an act of exchange. 10 mapping, and (2) entitlement mapping was defined so as to exclude production. Both these restrictions were removed in the subsequent generalized framework, in which (1) entitlement failure was seen to arise from changes in both endowment set and entitlement mapping, and (2) entitlement mapping was defined comprehensively to include production, exchange, and transfer.