Quiet Revolution in Welfare Economics- by Michael Albert and Robin Hahnel
Some ... readers may have had the same childhood fantasy as myself: a walk in the landscape alone where from any point you could come to any other point without ever having to go uphill, because the landscape contours were changing all the time as a function of your path. A modified fantasy ... might be the following. From any point in the landscape you may move effortlessly to any point situated at a lower level, regardless of any hills that might lie between. Also suppose that as you go from one point to a lower-level point, the landscape changes so that some of the points formerly situated at a higher level now become lower level points. An intriguing question would then be to ask which points are attainable ... from a given initial point ai. Obviously a point b would be accessible if and only if there is a sequence a1, a2, ...an, b such that a2 is downhill from a1, a3 downhill from a2, ..., and b downhill from an. This, however, is not very helpful unless we also specify how, as a result from the step from ai to ai+1, the set of points downhill from ai+1 and the set of points downhill from any other point as well, are changing .... We have to specify how the possibly possible changes when we realize one of the possibles.... In our fantasy ... possibilities are constantly appearing and disappearing: some may be lost forever, others may turn up again.
-John Elster, Logic and Society
THE ISSUE has never been whether or not tastes change. Rather the question is whether or not economists should concern themselves with such changes. Milton Friedman justifies traditional nonconcern as follows:
Despite qualifications, economic theory proceeds largely to take wants as fixed, This is primarily a case of division of labor. The economist has little to say about the formation of wants; this is the province of the psychologist. The economist's task is to trace the consequences of any given set of wants. The legitimacy of any justification for this abstraction must rest ultimately, in this case as with any other abstraction, on the light that is shed and the power to predict that is yielded by the abstraction. 1
While we agree with Friedman that "the proof is in the pudding," we believe economists should have much to say about preference formation and that our ability to compare the welfare properties of different economic systems hinges largely on our ability to do so.
Aristotle explained long ago: Our actions flow from our character and our character is formed by our actions. In the words of a modern-day economist: "Choice depends on tastes and tastes depend on past choices." 2 Or, as a former Chinese leader might have put it: Where do preferences come from? Do they come from the sky? Do they come from the sky filtered only through the clouds of social and cultural institutions? Put this way, our answer is no. They come from human nature filtered not only through the "clouds" of social and cultural institutions but also directly through the "thunderheads" of economic institutions, which profoundly influence other cloud formations as well.
The fact that "we are what we make ourselves through our daily activities" entails the view that human characteristics and consciousness, and thereby preferences as well, are "produced," in part, by our previous activities. And these previous activities certainly include our previous economic activities. Someone who "consumes" piano lessons as a youngster becomes more likely to "prefer" classical music concerts to rock concerts and to "manifest" preferences for an FM radio than an Am radio later in life. One who "consumes" erector sets in childhood is more likely to "choose" an engineering career; whereas someone who "consumes" dolls with all the appropriate props is more likely to "manifest" preferences for the work activity of a parent and housekeeper. One who played Little League baseball as a youth in the 1950s was more likely to spend an entertainment dollar attending professional baseball games than professional soccer matches as an adult in the 1970s. As Elster explains,
"this holds up a promise of improvement and a danger of inconsistency; a tool for self-education and manipulation alike. The crucial feature of this approach to human action is that it brings together two aspects often thought to be incompatible: intentionality and causality. Choice according to consistent preferences is an important part of what we mean by intentionality, whereas the laws governing the endogenous change of preferences are purely causal operating 'behind the back' of the actor concerned. The actor may, of course, try to anticipate the impact of present behavior upon future preferences and make his [or her] present choice in light of his [or her] evaluation of these consequences; indeed this is what Aristotle tells us to do." 3
In chapter 2 we began to see how recognizing the "human characteristic transforming" effects of the production process can change our view of the welfare properties of production under conditions of private enterprise. 4 In this chapter we review others' treatments of the related issue of endogenous preferences and begin to explain the significance of the concept.
John Stuart Mill found the Benthamite version of classical utilitarian theory wanting because it precluded discussing the phenomenon of preference development by insisting that "quantum of pleasure being equal, pushpin is as good as poetry." For Mill, human progress was practically synonymous with the development of ever-superior preference structures, hence his reluctance to forgo comparison between preferences themselves. So while Mill claimed to remain within the utilitarian tradition:
He nonetheless insisted on broadening the Benthamite interpretation of "the largest sense" to conform with his own firmly held belief paraphrased by Bertrand Russell:
I regard utility as the ultimate appeal on all ethical questions; but it must be utility in the largest sense, grounded on the permanent interests of man as a progressive being. 5
Mill attempted to do this by treating pleasure as having a qualitative as well as quantitative aspect.
As a matter of fact, we consider some tastes better than others: We do not merely hold that some tastes are ours and other tastes are other people's. We do not even always consider our own tastes the best: We may prefer bridge to poetry, but think it is better to prefer poetry to bridge. 6
It would be absurd that while, in estimating all other things, quality is considered as well as quantity, the estimation of pleasure should be supposed to depend on quantity alone .... It is quite compatible with the principles of utility to recognize the fact that some kinds of pleasure are more desirable and more valuable than others ... there is no known Epicurean [i.e., utilitarian) theory of life which does not assign to the pleasures of the intellect, of the feelings and imagination, and of the moral sentiments, a much higher value as pleasures than to those of mere sensation. 7
Mill has been frequently ridiculed for inconsistency by the likes of Sidgwick for failing to understand that the classical utilitarian principle precludes noncomparable, qualitative distinctions between pleasures 8 . But we find it more productive to interpret Mill as arguing that certain preferences are qualitatively superior in the sense that they allow for the enjoyment of greater quantities of pleasure than do "less-cultured" preferences.
A cultivated mind finds sources of inexhaustible interest in all that surrounds it; in the objects of nature, in the achievements of art, the imaginations of poetry, the incidents of history, the ways of mankind, past and present, and their prospects in the future. 9
In this interpretation, Mill felt no need to discard the utilitarian principle as the final arbiter, but felt increases in pleasure resulted not only from more complete fulfillment of existing desires but also through the development of new systems of desires that could be deemed superior in that they were capable of yielding greater quantities of pleasure through their fulfillment. In other words, instead of being "fuzzy headed," perhaps Mill recognized the importance of preference development as well as preference fulfillment to a valuative methodology and merely wished to insist on an adequate analysis of the factors affecting preference development as well as fulfillment for utilitarianism to be a useful valuative methodology. On the other hand, Mill concluded that however important preference development was in general, economic institutions had little effect in this regard. Mill focused his arguments on the importance of considering preference development when evaluating political and cultural institutions. Mill's analysis of preference development and the reasoning process necessary to justify his conclusion that preferences could be treated as exogenous to the economic system is carefully explained by Gintis in his chapter on Mill. We merely offer Gintis' conclusions:
Mill resolves this problem ... that welfare increases not only on the basis of fixed wants, but in the development of wants themselves ... by postulating that the system of personality structures [upon which preferences depend] develops according to an internal social mechanism, logically as well as causally unaffected by the social [economic] mechanisms devoted to the satisfaction of manifest preferences at a particular point in time. 10
Mill does not assert preferences are exogenous to political and cultural institutions. Indeed, Gintis argued convincingly that Mill's most noteworthy interpreters have committed a gross oversight in not viewing Mill's writings in social philosophy and recommendations concerning political institutions designed to preserve "individual liberty" as largely motivated by a concern that political and cultural institutions be appropriately fashioned to ensure optimal preference development. The relevance of political and cultural institutions for Mill stems from his belief that:
the attributes of superior preference structures are sufficiently objective and identifiable [that] ... if the individual is "unimpeded" in his activity toward preference development, the systems of preference structure will themselves develop in a proper manner 11 .
Mill sees repressive political and cultural institutions as the only possible impediments to the exercise of individuality and emulation that are the driving forces of his "internal mechanism" for the development of preferences. Gintis' critique of Mill goes straight to the point:
In broadest perspective ... the model must be operationally interpreted as saying there are no functionally supported socioeconomic institutions directly concerned with preference development. 12
That is, once political and cultural institutions have been recognized as structures relevant to individual preference development, why are economic institutions excluded? Gintis' more general point is that there is a bias in evaluating economic institutions solely on the basis of their efficiency in providing material goods as judged by today's preferences. Once we have asserted preference development as a social goal, economic institutions must be judged from this point of view as well. Gintis drives the point home in the citation below by observing that were a specific kind of spiritual development the only social goal, highly "inefficient" precapitalist institutions would receive high marks, while presumably "efficient" capitalist institutions would receive low scores (not that Gintis means to champion this particular goal).
This view [that economic institutions do not influence preference development] is especially curious in the case of the economist, who is more than willing to accept the position that social organization around issues of material development can further individual goals. But in what sense does an institution like "Alcoholics Anonymous," as an example of an institution directed toward a form of self-development that cannot be achieved by individuals in isolation, differ from a steel plant, as a form of material development that cannot be achieved by individuals in isolation? In short, this asymmetrical treatment of material and developmental goals, and of the degree to which social organization can promote their advancement, leads to an inherent bias in favor of the Robbins judgment in any analysis.
We could in fact parody Mill's argument by reversing it. Suppose a society in which individual development goals, say "spiritual growth," held primacy. A social welfare analyst would then devote the major part of his time exhibiting the precise form social organization would take toward the end of individual spiritual growth, noting that while economic security is also of considerable importance, this could be achieved by allowing individuals the "liberty" of forming informal, functionally unsupported groupings to further their material ends. It should be no surprise that he arrives at a "traditional" nonindustrial society as an ideal type! 13
How a thinker as humanistic in his orientation and as broad in his views as Mill could recognize the crucial importance of preference development to valuative methodology as well as the influence of social and cultural institutions on preference development without recognizing that economic institutions influence preference development as well, might seem hard to explain. On the other hand, belief in the separability of functions, specifically regarding the economic institutions of competitive capitalism, appears elsewhere in Mill's thought. Just as Mill asserted that distribution could be separated from efficiency in private enterprise market economies, he assumed that preference development could be separated from efficiency in fulfilling preferences.
In chapter 5 we will argue for a more holistic paradigm. For many purposes labeling institutions "political," or "cultural," or "economic" is helpful, but for other purposes such labels obfuscate important relationships. While it might be the case that particular political, cultural, or economic institutions exert an influence only within "their spheres," this should not be presumed, much less taken for granted, as a consequence of their labeling. In some cases "economic" institutions might perform functions other than those traditionally thought of as economic functions. Mill believed this was not the case for the economic institutions of private enterprise and competitive markets regarding preference development. The traditional welfare paradigm presumes this is not the case for "economic" institutions in general with respect to the "cultural" function of preference development. While we will argue against the "separability" assumption in both the particular case of private enterprise market institutions and the case of economic institutions in general, for now we simply wish to thank John Stuart Mill for pointing out the importance of preference development for welfare theory.
While C. C. von Weizsacker and Robert Pollak are not the only modernday neoclassical economists to treat endogenous preferences, they are among a very small group who have broken the taboo against questioning the origins of preferences. Trygve Haavelmo broached the subject in his doctoral dissertation in 1944. 14 Georgescu-Roegen, Strotz, and Hasanyi introduced the subject in the economic journals in the 1950s. 15 C. C. von Weizsacker called attention to the potential significance of the issue for modern neoclassical theory in an article in 1971. 16 Peter Hammond, C. Luch, and Marcel Boyer among others have written occasionally on the subject more recently. 17 But it was Robert Pollak who, almost singlehanded, kept the subject alive in neoclassical journals during the 1970s. 18 C. C. von Weizsacker posed the problem that had troubled few since J. S. Mill as follows:
There are influences on tastes which may have little to do with those variables economists are usually concerned about. From the point of view of the economist we can call them exogenous influences. We may acknowledge their existence, but we need not investigate them very much further. But other influences on tastes may depend more or less directly on certain economic variables like, for instance, the tendency of firms to increase their profits through advertising, or the consumption patterns of other consumers, or the consumer's past experience with consumer goods. If these influences on tastes are not taken into account, we clearly would be in danger of making false predictions due to mis-specification of certain parameters in our model. 19
In this book, we will not be concerned with advertising at all, and we treat interconsumer externalities as simply a subcategory of external effects in general. But the influence and implications of "consumer's past experience with consumer goods" on their present preferences directly concerns us. Von Weizsacker proceeds to pinpoint the relevant question for welfare theory even more precisely:
The efficiency or Pareto optimality properties of general equilibrium models rest on the assumption of fixed tastes. Are these efficiency theorems relevant to a world with changing tastes? Surely, purely exogenous changes of tastes (from the point of view of the economist) will essentially not alter these theorems .... But what about Pareto optimality if endogenous variables of the economic system influence tastes? 20
Finally, von Weizsacker evaluates the status of traditional welfare theory regarding this subject in a way we wholeheartedly support:
I think it is fair to say that this question has not yet been investigated in sufficient depth by economists. Yet it appears to be one of the important problems in the discussion about the social relevance of economic theory. Does the model with fixed tastes give insights into the problems of an economy with endogenously changing tastes? One of the reasons why economists did not very deeply discuss this question may be that their present concepts of Pareto optimality and efficiency possibly are not flexible enough to cope with endogenously changing tastes. It may become necessary to change the conceptual framework of our theory; we may be forced to ask almost philosophical questions about the concepts we use. In every discipline, there is always a reluctance to engage in this kind of activity. 21
Or, since addressing the issue requires changing the traditional paradigm, it has been more convenient to pretend the subject is of little interest. Unfortunately, the manner in which von Weizsacker pursued the issue in his article permitted subsequent neoclassical treatments to effectively "bury" the welfare implications he sought to emphasize.
Von Weizsacker presents a model with only two goods, and assumes tastes are only influenced by consumption in the previous period. He conjectures "although the mathematics would become more complicated, I presume the relaxation of these two assumptions would not change the substance of the argument" 22 Ahmad EI-Safty proves this conjecture correct. 23 But while von Weizsacker's conclusions are not sensitive to the number of goods or the length of the adaptive process, they are sensitive to his assumption that the adaptive process is myopic habit formation. EI-Safty and Pollak both demonstrate that von Weizsacker's conclusions based on a myopic habit formation modeling of the adaptive process are not generalizable. In the words of EI-Safty, "the long-ran demand functions can be rationalized by a utility function if and only if the short-run utility function is such that any good that experiences learning or taste change is separable from all other goods." 24 And Pollak concludes, "Von Weizsacker's long-run utility function approach to the evaluation of welfare is technically possible only in a narrow class of cases. 25 Since EI-Safty and Pollak's criticisms are technically correct, it is clear that a myopic habit formation conceptualization of endogenous preferences is of very limited relevance to welfare theory.
With hindsight, for two reasons choosing to model endogenous preferences as myopic habit formation appears to have been unfortunate, at least regarding uncovering the implications of endogenous preferences for welfare theory. First, "habit formation" is, more than anything else, relevant to convexity; it complicates matters concerning theorems that establish the existence of long-run demand functions and the existence of equilibria in market economies.
Especially Pollak, in subsequent treatments, has focused on these problems. But existence, and optimality of anything that might exist, are two separable concerns. We are inclined to agree with Pollak that "habit formation" is of little additional relevance to welfare theory, but this does not mean that other kinds of endogenous preferences are not of great relevance to welfare theory. 26 By concentrating on habit formation von Weizsacker also made more likely subsequent focus on the theory of empirical estimation of consumer demand rather than the welfare implications of endogenous preferences. In other words, von Weizsacker's choice of habit formation as the form of endogenous preferences to analyze leads to largely irrelevant technical complications and a focus on the positive rather than the normative implications of endogenous preferences, all quite contrary to his own desires.
Second, and more importantly, by focusing on "myopic adaptive behavior" rather than on what we might term "knowledgable" or "purposeful preference molding" von Weizsacker created a context in which the most significant welfare theoretic implications of adaptive behavior are absent by supposition and a situation in which the existence of long-run "meta" utility functions that rationalize long-run demand functions exist only under highly restrictive conditions. Creating myopic habits, an individual fails to recognize that consumption of a good now increases the desire for that same good later. Purposefully molding preferences, an individual recognizes that present choice of consumption will both fulfill present preferences (more or less efficiently) and change future preferences in a particular direction (that may be more or less convenient in light of expected conditions of future availability). In both cases previous consumption affects future preferencesmaking the latter "endogenous." But in the first case the individual only recognizes the "preference fulfillment effect," while in the second case he or she recognizes the "preference fulfillment effect" and the "preference development effect" as well. In the first case individuals are presumed oblivious to the causal laws of their adaptive behavior. In the second case, they are granted perfect foresight of all the implications of their choices. 27 It should cause little surprise that individual behavior that fails to recognize the preference-development effects of choices can be rationalized by a long-run utility function only under singular conditions. But there is no reason to suppose this problem will affect purposeful preference molding. As will become apparent from our own treatment in chapter 6, the dynamic of purposeful preference development poses far more significant welfare theoretic questions than myopic habit formation.
In any case, von Weizsacker draws the following generalization from his treatment of myopic habit formation:
We therefore have to ask whether it is reasonable for a society to build its decisions on its present preferences or on some other criterion. If present preferences are strongly influenced by myopic thinking, by lack of imagination how a different world would look, we should not accept these preferences as the last word. 28
While true, we would classify this conclusion as obvious. In essence it says: If individuals incorrectly estimate the effects of different choices on their well-being, leaving individuals to make their own choices in all likelihood will not maximize their well-being. 'Me possibility of "myopic habit formation" simply points out another way this might happen. But von Weizsacker recognizes the importance of what we regard as the more interesting question of "purposeful preference molding" as well as the philosophical dilemma it poses in his conclusion:
It will require another paper to discuss this problem of intertemporal decision making in view of anticipated and planned changes of tastes. Society ... may have to opt for a metapreference in favor of challenging the prevailing preferences of its members and to ask if and how these preferences can perhaps be improved .... This does not, of course, mean that there should be some person or group of persons who from their "superior" point of view dictate the values of society. 29
In our view, he also foresees the only acceptable answer to the dilemma:
Society's decisions must rest on the preferences of all its members. But we have to acknowledge and make use of the fact that preferences are partly the product of people's environment [emphasis added]. 30
Robert Pollak does not pursue von Weizsacker's plea for "another paper ... to discuss ... intertemporal decision making in view of anticipated and planned changes in tastes." As he explains:
Those who favor incorporating taste formation and change into economic analysis fall into two groups whose intersection is almost empty. One is primarily interested in the welfare implications of changing tastes, the other in the analysis of household behavior .... The recent impetus to incorporate taste formation and change into economic analysis has come primarily from those interested in household behavior rather than welfare, and the principle focus of this work has been empirical demand analysis. 31
Pollak concentrates primarily on implications of myopic habit formation for the positive theory of demand estimation. Because this subject does not concern us, we ignore Pollak's important contributions in this area. But, as a secondary concern, Pollak also addresses welfare implications of endogenous preferences. Since his conclusions regarding welfare theory are so different from ours, it is important to review his results.
Pollak argues specifically that von Weizsacker's conclusions regarding endogenous preferences are both technically and substantially unsustainable. He specifically rejects the alternative approach that would avoid the objections he poses to von Weizsacker's treatment. He concludes that if preferences are endogenous, welfare conclusions are greatly complicated and we must be more cautious about drawing welfare conclusions. But for Pollak, nothing definitive can be concluded. If anything, endogenous preferences render the whole project of welfare theory obsolete. We will deal with these points one by one.
First, Pollak claims von Weizsacker's approach to welfare analysis is technically inadequate:
Von Weizsacker argues that the long-run utility function is the appropriate criterion by which to judge the welfare effects of changes in consumption. Since we have shown, contrary to his conjecture, that when there are more than two goods the long-run utility function exists only in special cases, it cannot serve as a general welfare criterion. Revealed preference arguments provide no help; the counterpart of the nonexistence of the long-run utility function is violation of the long-run version of the strong axiom of revealed preference. At most, then, the long-run utility function may be the appropriate welfare criterion only for a narrow class of cases in which it exists. 32
Pollak is technically correct. But the "existence" problem he demonstrates is a direct consequence of modeling endogenous preferences as myopic habit formation. Since that is the form of endogenous preferences von Weizsacker treated, Pollak's objection is well taken as far as von Weizsaker's "conjectures" are concerned. But Pollak's "impossibility theorem" regarding nonexistence of long-run utility functions except for a narrow class of cases is not universally applicable to other formulations of endogenous preferences, and to our own formulation in particular. 33
Pollak goes on to argue, however, that von Weizsacker's long-run utility function is not an appropriate criterion for judging the welfare effects of changes in consumption, even when it does exist. For von Weizsacker the long-run utility function defines an indifference map in which a consumption vector, q(0), will be on a higher indifference surface than another consumption vector, q(n), if and only if the consumer can move from q(0) to q(n) in a finite sequence of steps where she or he feels better off at each step along the way. Pollak explains why he rejects von Weizsacker's indifference map as a welfare criterion via the following example:
A nonsmoker might prefer to remain a nonsmoker rather than smoke three packs of cigarettes a day, but he might choose to smoke half a pack a day rather than abstain completely. After becoming accustomed to smoking half a pack a day, the individual might prefer to remain a light smoker rather than smoke three packs a day, but he might choose to smoke a pack a day rather than continue at half a pack a day. By this process, the myopic nonsmoker is led to become a heavy smoker. This scenario is entirely consistent with von Weizsacker's assumptions, yet I am loath to conclude that the individual is better off at q(n) than q(0). 34
Pollak generalizes as follows:
I interpret the individual's willingness to move from q(0) to q(n) in a sequence of small steps when he is unwilling to do so in a single large step as indicative of his failure to understand the habit formation mechanism and not of the underlying superiority of q(n). 35
Once again, we agree with Pollak. Von Weizsacker's conception of a long-run utility function that is appropriate for welfare purposes is unacceptable-ironically, particularly in a myopic habit formation context. But this does not mean no other conception of a long-run utility function is appropriate for welfare purposes in context of a different conception of endogenous preferences.
Pollak finally considers such an alternative formulation of long-run utility functions:
Since von Weizsacker's long-run utility function approach to the evaluation of welfare is technically possible only in a narrow class of cases and is conceptually unsatisfactory even for those, it is desirable to consider alternative approaches. One such alternative, one which von Weizsacker suggests in his concluding paragraph, is to view the problem in an intertemporal. framework. That is, instead of focusing on the one-period utility function, U(q(t), q(t-1)), assumed to be the same in every period, we could evaluate welfare in terms of the intertemporal utility function
V(q) = W[U(q(1), q(0)),...U(q(t), q(t-1))].
The proposed welfare test, in other words, is whether the individual-taking full account of the impact of his present consumption on his future tastes-would be willing to undertake a particular change. 36
But rather than accept this alternative formulation, Pollak rejects it.
The principal difficulty with this approach is that it is schizophrenic. The habit formation model is tractable because the individual is assumed to be myopic-he fails to recognize the impact of his current consumption on his future tastes. If he could be persuaded to recognize the effects of habit formation, then it would certainly be appropriate to base welfare comparisons on the intertemporal utility function, but then his demand behavior would be far more complex than that predicted by the model of "myopic" habit formation. If an individual insists on being myopic, it is less clear that the intertemporal utility function is the appropriate welfare criterion, but it is tempting to take a paternalistic view and argue that it is. However, it is difficult to reconcile this approach to welfare with an approach to demand analysis based on myopic habit formation. 37
Frankly, Pollak's double fixation on the phenomenon of myopic habit formation and demand analysis seem to have rendered him myopic regarding conscious or "purposeful preference molding" and welfare theory. A model with identical utility functions in all periods that evaluate consumption of a commodity more highly if more of it was consumed in the previous period is quite appropriate to analyzing the effects of changing initial consumption levels on demand under conditions of habit formation. Specialization of addiction is hardly a surprising result. But the model Pollak rejects appears to us obviously suited to analyzing the welfare effects of all manner of recognized preference development effects. What do his objections amount to?
Obviously, "myopic habit formation" and "purposeful preference molding" are different behavioral phenomena. While it does not appear unreasonable to consider the possibility that both occur in the economy as a whole and even in the behavior of a single individual who recognizes the "preference development effects" of some of his or her choices but not others, we need not debate the point. Pollak concedes all that is necessary by admitting "it would certainly be appropriate to base welfare comparisons on the intertemporal utility function" if people recognize what we call the "preference development effects" of their choices. We might point out that the tradition of welfare theory has always been to focus on the implications of rational behavior of economic actors. In this light to focus on purposeful preference molding appears quite logical if we are concerned with the implications of endogenous preferences for welfare theory. 38 After all, to find that myopic and rational behavior are "schizophrenic" is hardly surprising and in no case a reason to reject analyzing rational behavior, even if one wishes, as Pollak did, to analyze the effects of myopic behavior as well.
Pollak's second objection is that demand behavior would be "far more complex" in the case of conscious preference molding than myopic habit formation. We can only respond: So what? The additional complexity is only objectionable in the sense that we should not try to incorporate it in our formal analyses if it proves so intractable that it prevents drawing any conclusions. As the results of part 3 demonstrate, this is not the case. If we assume away nonconvexities and stipulate perfect foresight-timehonored traditions among welfare theorists-both the technical and substantive objections posed by Pollak to assuming long-run "meta" utility functions disappear. But even if the situation did prove so much more complex that it was intractable, if conscious preference molding exists and has important effects, welfare theorists should take great pains to qualify conclusions derived from formal models that are unable to incorporate the phenomenon. And developing ways to incorporate conscious preference molding into our formal models should rank high on the list of projects challenging welfare theorists if they wish to be taken seriously by those trying to evaluate economies and formulate appropriate policies.
Finally, Pollak believes it is difficult to reconcile a conscious preference molding approach to welfare theory with a myopic habit formation approach to demand analysis. We believe this reveals a misunderstanding of the welfare project. There are good reasons to believe that analyses of rational and actual behavior will differ-perhaps markedly. 39 In any case, even if there were a "conflict of interest," as welfare theorists we could hardly accept Pollak's priorities.
Having rejected von Weizsacker's conclusions and the alternative approach that appears so promising to us, what are Pollak's conclusions regarding the implication of endogenous preferences for welfare theory? Pollak draws an important distinction between the kinds of preferences appropriate for demand analysis and welfare analysis:
In demand analysis the objects of choice are vectors of private decision variables, Q, and preferences over them depend on a vector of predetermined "state variables," Z; we call such a preference ordering "conditional."...Welfare analysis must compare the individual's well-being in alternative situations which differ with respect to the state variables as well as the private decision variables....Welfare analysis requires us to redefine the objects of choice to include not only the private decision variables but also the state variables which, from the standpoint of conditional preferences, are predetermined. We call an ordering over such an augmented set of alternatives an "unconditional preference ordering."... The additional information contained in the unconditional preference ordering is irrelevant to demand analysis, but for welfare evaluation it is indispensable. 40
For Pollak the problem lies precisely in the fact that while "unconditional preferences are necessary for welfare analysis they cannot be inferred from market behavior or any other conditional choices." 41 In his view, this means very little can be said about "taste change and welfare analysis." 42
To his credit Pollak recognizes a disturbing implication:
Taste formation ... poses ... difficult problems for welfare analysis. Variable tastes undermine the normative significance of the fundamental theorem of welfare economics which asserts (in a precise sense and under fairly stringent assumptions) that in competitive equilibrium everyone gets what he wants, subject to the constraints imposed by technology, resources, and the satisfaction of the wants of others. However if tastes are sufficiently malleable, then this may be no more than a corollary of the more general proposition that people come to want what they get. 43
But Pollak's narrow interpretation of endogenous preferences as myopic habit formation leads him to mistakenly conclude that the sets of "private decision variables" and "state variables" are necessarily disjoint. Yet this is not the case for the kinds of endogenous preferences most relevant to welfare analysis. Pollak interprets state variables as follows:
State variables may be the individual's own past consumption (habit formation) or the consumption of others (interdependent preferences). They may be environmental variables ... or goods or services provided by the government .... Or they may be socioeconomic variables or demographic variables. 44
Obviously, state variables influence the satisfaction an individual will attain from different "Private decisions." And obviously, the values of state variables in the present cannot be affected by present choices. But, for Pollak, individuals do not believe themselves capable of affecting values of state variables at all. Pollak is correct that habits formed myopically are similar to the consumption of others, supplies of public goods, and values of demographic variables in that all exist beyond the conscious control of individual decision makers. But if preference development effects of consumption are recognized, one's own past consumption is still a "state variable" in that it is part of the "state" of affairs that cannot be willed away yet influences the satisfaction obtainable from different present choices. However, it is not a variable beyond the reach of "private decision making." This is precisely the interesting feature of endogenous preferences as far as the theory of rational choice and welfare analysis is concerned. Yet it seems to escape Pollak entirely. We must turn to Herb Gintis to find someone who
1. Accepts von Weizsacker's challenge to "write another paper" treating the "problem of intertemporal decision making in view of anticipated and planned changes of tastes"
2. Accepts the obviously relevant conception of long-run utility functions that Pollak rejected
3. Understands that the importance of endogenous preferences for welfare theory lies in their status as state variables that are not beyond the ability of individual decision makers to influence
Whereas Pollak pursued the positive implications of "myopic habit formation" and denied any serious normative implications, Gintis pursued the normative implications of "purposeful preference molding" and insisted that they dramatically affect traditional conclusion 45 Gintis begins by exploring precisely what the modern conception of an individual's "manifest" preference structure is, formulating three useful conceptions of preference ordering along the way:
The Incomplete Knowledge Preference Ordering (IKPO) is determined as follows: the individual is isolated at the beginning of time period t, presented with a price structure over commodities and factors, and directed to choose a bundle of commodities and factor supplies based on this precise structure, satisfying an income constraint. By varying the price structure presented to the individual, we arrive at the Incomplete Knowledge Preference Ordering, in the usual manner of revealed preference theory. The Complete Knowledge Preference Ordering (CKPO) is obtained by allowing the individual to "experience" each bundle compatible with a given price structure, asking him ex post which he actually preferred. In both cases, we shall ask the individual to assume, for the purposes of choice, that his preference structure cannot be altered over the time period in question. Thus we shall refer to the IKPO and the CKPO as instantaneous preference structures.
Economic theory normally posits the stability of individual preference structures; this is a factually incorrect interpretation of the individual's own intentions ... the individual not only possesses an IKPO and a CKPO at time t, but ... has certain definite ideas about changing this structure during the time period in question, arriving at another preference structure at time t+1. In short, we assume the individual has at least a vague set of "preferences on preference structures" themselves. The source of these preferences may be briefly described. Some will be based on the individual's desire to change his skills, some on the desire to acquire new tastes, some to improve the moral or aesthetic character of his preference structure, and some perhaps to correct past mistakes in preference structure formation. Thus the individual may "prefer" an IKPO which values a certain commodity more highly than his present structure .... For example, he might consider that an IKPO that values poetry more highly than his present IKPO desirable in that the "cost" of poetry is small, and in that persons who do prefer poetry seem to feel the benefits are great .... Clearly the activity involved in moving from one preference structure to another will involve the outlay of certain skills, and the foregoing of certain types of consumption, as well as certain other subjective costs and perhaps benefits. All in all, it is assumed the individual weighs all these factors in deciding a course of action for the period t; he will consume, produce, and alter his preference structure. Thus in a larger context, we can amalgamate the IKPO for period t with the above valuation in preference-on-preference structures into one grand preference ordering for period t. We shall call this the individual's Complete Preference Structure (CPS). 46
The obvious question is where does the revealed preference structure of modern neoclassical welfare theory fit into this framework? Gintis replies:
It cannot be the CKPO, as revealed preference theory does not assume the individual has perfect knowledge. Nor is it the IKPO, as revealed preference theory, in recording individual choices, does not ask the individual to assume his preference structure cannot be altered over the time period. It would seem, rather, that the CPS coincides with the "revealed" preference structure of traditional theory, as both are determined merely by bidding the subject to exhibit his choices at time t. 47
Next, Gintis concludes that "preference structures described in revealed preference theory must change over time, and in particular will tend to be intransitive.' 48 We might add there is no longer reason to assume revealed preferences would display even the more fundamental property of simple consistency! The source of these undesirable properties is that preferences are now seen to depend on previous economic activities or the actual choices of previous consumption and work bundles. In acting upon the dictates of my present IKPO and income constraint, I consume different goods and engage in different work activities, thereby changing my IKPO in the future. It is as if I became a "new person," and just as neoclassical theory never expected to find different individuals' revealed preferences characterized by consistency or transitivity, there is no more reason to expect this of a single individual's revealed preferences. 49
First we summarize the parts of Gintis' critique of neoclassical welfare theory we agree with and build upon. Then we present the parts we object to and improve upon.
Preferences depend on previous economic choices because preferences are now seen to be explicitly mediated (parameterized) by an individual's human characteristics where, in turn, these human characteristics are seen to be partially produced by (functions of) previous choices of all kinds, including economic choices traditionally conceived as choices of consumption bundles and work activities. A mathematical formalization of these relationships is presented as model 1 in chapter 6. It does no more than precisely express:
1. The relationship between individuals' previous work and consumption activity choices and their human characteristics. We envision these characteristics as various personality traits, talents or skills, knowledge, and values or attitudes toward what they understand is going on in the world.
2. The parametric relationship between these human characteristics and the enjoyment, pleasure, satisfaction, utility, or well-being individuals obtain from any consumption bundle or work activity in the present-that is, their instantaneous welfare function.
A further justification of these choices of human characteristics will be offered in chapter 5 as an important part of our own treatment. However, the above verbal formulation of these relations is sufficient to suggest a problem more damaging to traditional welfare theory than the previous observation that commonly presumed characteristics of revealed preferences are not justified. Gintis was the first to pose the question whether Pareto optimality in terms of "manifest preferences" has any "moral" significance if preferences are endogenous. The essence of traditional welfare theory was summarized previously as "high marks are to be given to economic systems that display a close 'fit' between the relative terms on which economic goods are made available and people's relative preferences for those goods." But theorists presumed the "close fit" had something to recommend it because they assumed the fit was achieved by an adjustment of the relative supplies of different goods to people's relative desires for those goods. The possibility that a "close fit" between desires and supplies is achieved by an adjustment of people's desires to conditions of relative supply undermines the worth of neoclassical welfare theory's principle criterion for judging the performance of economies. Gintis put it this way:
To justify a set of economic institutions on the basis of the manifest preferences of individuals who were formed through their preparation for participation in these institutions involves making an arbitrary cut in a system of causal linkages. Not only are preference orderings formed on the basis of economic organization, but this organization clearly erects certain functional prerequisites, certain structural uniformities in patterns of preference orderings in the absence of which the system as a whole would fail to cohere. The apparent mutual determination of a social structure (and economic structure in particular) and individual personality systems thus presents severe problems for a purely atomistic welfare theory. 50
The point is that Pareto optimality , judged according to manifest preferences, might signify stability rather than merit. If an economic system adjusts provisions to people's desires, a "close mesh" is meritorious. But if people get what they want because they adjust their desires to what is made available by the economic system, a "close mesh" may only indicate that the system is stable since people's desires and expectations are being fulfilled. Ironically, endogeneity of preferences seems to transform Pareto optimality from an evaluative concept into an analytic concept. It is possible to demonstrate that rational individuals, who recognize their preferences are endogenous, will adjust to changes in the relative terms of supply of different goods not only by changing the relative amounts of those goods they consume, but by changing their preferences for those commodities as well. They will diminish their desires for commodities whose terms of availability they believe will become more difficult, and augment their desires for commodities whose terms of availability they believe will become easier by changing consumption and work activity choices in earlier time periods so as to change their future human characteristics and, thereby, their future preferences. This result is formulated and proved in chapter 6 as Theorem 6.2, and destroys an important "separability" result that is a cornerstone of at least the partial equilibrium version of traditional welfare theory. Gintis summarized the conclusion as follows:
... the movement of preference structures over time will in part depend on both commodity price structure and wage structure. As the individual moves from one IKPO to another on the basis of expectations as to the prices of commodities and remuneration for work skills that they accentuate to various degrees, the assumption of fixed preferences and the resulting implementation of Pareto optimality conditions in the determination of wages and prices in themselves alter the preference structures originally posited. This observation radically alters the interpretation and the validity of the axioms of welfare economics. 51
Of course, if rational individuals take into account that present consumption and work activities not only fulfill present preferences but also change future preferences, whereas traditional welfare theory does not recognize this "preference development" effect, traditional theory will obviously misestimate the welfare effects of different actions. It follows that policy recommendations of traditional theory based on erroneous predictions of the overall welfare effects of different choices should not be given much weight. A proof of the proposition that in its present form traditional welfare theory consistently misestimates welfare effects of different economic choices if preferences are endogenous, originally derived by Gintis, is formulated and proved in chapter 6 as Theorem 6. 1.
The phenomenon of rational adjustment of preferences, toward wanting what is presumed will be most readily available and away from wanting what is presumed will be difficult to obtain, should not be confused with another way that "utility" might be a function of price-namely the "snob effect," or "keeping up with the Joneses," or Veblen's concept of "conspicuous consumption." 52 The clear difference between the phenomena can be appreciated by noticing that the direction of the individual's adjustment is exactly opposite in the two cases. Whereas the phenomena we are discussing lead rational individuals to diminish their consumption of goods whose future prices are expected to be high, conspicuous consumption leads individuals to increase consumption of goods with high prices. One could conclude that both phenomena occur in the real world and that the effects of both are significant. On the other hand, whereas it has been relatively easy for neoclassical welfare theory to dismiss the "snob effect" as "irrational" behavior in which the individual "pays" for snobbery regardless of how prevalent the motivation might be, the phenomenon we refer to is characterized precisely by its rationality, and therefore, at least according to neoclassical tradition, by its prevalence if not universality.
Up to this point we are willing to follow Gintis in exploring the consequences of endogenous preferences. However, Gintis traveled down a path we find less interesting than a path branching in a different direction. Gintis' observations are not incorrect or devoid of interest, but in our opinion they do not reveal the fundamental deficiency of traditional welfare theory's treatment of preferences. We present Gintis' argument first and motivate the interpretation we will pursue throughout the book second. Gintis proceeded by observing that a considerable amount of uncertainty is involved in the concept of the Complete Preference Structure (CPS).
Firstly, it is based on the IKPO rather than the CKPO. Secondly, the process by which the new preference structure is achieved may lead to a new structure, differing more or less from that at which the individual originally aimed. Thirdly, the "costs" and "benefits" of undertaking the activity leading to the change in preference structure are only incompletely known to the individual, particularly in that he cannot experience beforehand the process of preference structure change .... Lastly, and perhaps most importantly, he can know only indirectly and by inference how satisfied he might be with the new preference structure. 53
But Gintis rejected moving toward "abstraction" from the problems of uncertainty on the following basis:
It is tempting to postulate a "complete knowledge" counterpart to the individual's CPO. This would involve positing an individual with known CKPO and complete knowledge of his "capacities," moving him to every possible CKPO for the next period, and registering his "preferred CKPO." In fact, this type of "gedanken experiment" abstracts from the human situation to such an extent as to be useless in understanding the dynamics of preference systems development. 54
We agree with Gintis that this type of "gedanken experiment" abstracts mightily from the human situation. To assume that all individuals can know how their preferences depend on their various human characteristics, and how the development of those characteristics is influenced by consumption and work choices, and also the conditions of future availabilities of all economic goods, violates two important aspects of our human condition. First, as much as we might try to estimate from others' verbal and nonverbal behavior what it would be like to be different from what we are, it must always remain somewhat impossible for us to know such a thing. Our conception of the realm of human possibilities outside our own experience is inevitably influenced and limited by our own experience. Unfortunately, the postulation of a preordering of preferences-a complete, reflexive, and transitive ordering of preferences themselves-ignores this difficulty.
Second, there are good reasons to believe, particularly in light of endogenous preferences, that the future is not determinate. Granting that preferences themselves are moldable removes half of the "fixed data" of traditional general equilibrium models and makes even more likely the existence of more than one path that our economic world might take from any initial situation. It is the misfortune of the assumption of perfect knowledge of the future terms of availabilities of economic goods that it ignores this problem of many different possible futures.
However, in our opinion Gintis made a tactical error in formulating the essence of his critique of traditional welfare theory around the problems of uncertainty that endogenous preferences certainly do create. 55 In our view, emphasizing the additional problems of uncertainty posed by endogenous preferences ultimately limits one's critique of particular economic institutions to highly technical grounds. We may agree with the sentiments Gintis expressed in the passage just quoted, but we feel it is possible to grant traditional welfare theory all the perfect knowledge assumptions it requires, no matter how unreflective of the human situation they may be, and still formulate a critique of the traditional approach. Moreover, we feel that doing so gets us to the heart of the traditional paradigm, whereas focusing on additional problems of uncertainty does not.
Abstracting from the imperfect knowledge that is very much part of the human condition is entirely within the tradition of welfare theory, and, in our opinion, justified. To think otherwise is to misunderstand the nature of welfare theory and confuse the purpose of welfare theory with that of positive economic theory. The fact that the knowledge required by the perfect knowledge version of the CPO stretches credulity well beyond the breaking point is good reason to reject the concept as useful in explaining actual human behavior. But that is not the essential purpose of welfare theory. The major purpose of welfare theory is to analyze the implications of rational behavior, and the extent to which this informs aspects of actual human behavior is secondary. How interesting the question of what the welfare effects of perfectly informed rational behavior in different institutional environments would be, compared to many other interesting questions, is, of course, a matter of debate. But that is the question welfare theory has always chosen to address.
Gintis maintained, on the other hand, that "the attractiveness of modern welfare economics lies in the simplicity of its underlying normative assumptions." 56 By this he means that the Robbins Principle appears unobjectionable in not requiring elaborate ethical justification. Gintis asserted:
The common argument against the inclusion of preference change in welfare theory relies on the inadvisability of the economist's "dabbling in value judgments." ...This, in fact, lies at the heart of the "revolution" in welfare economics, of which the Robbins judgment, the Hicks-Kaldor Scitovsky analysis, and the Bergsonian social welfare function mark the high points. 57
Based on this belief, Gintis stated his purpose:
I shall show that both the intuitive nature and irreducibility of welfare judgments are unwarranted in that they are derived from a simplistic and ultimately reductionist formulation of the individual choice situation and the nature of individual preference ordering. 58
He proceeded by observing:
We have seen that the statement, "individuals themselves know best what they want" cannot be taken to mean that they have complete knowledge-for the statement that the IKPO and the CKPO coincide is factual and not normative. Moreover, it is clearly false. 59
Gintis then stated the obvious conclusion:
Implicitly, the Robbins Principle exhorts us to assume the CPO as the final judge and arbiter for the individual in spite of the fact that it may not be the best instrument in some, or even most, cases. 60
Unfortunately, Gintis then drew precisely the secondary conclusion that, in our opinion, welfare theorists should be most careful to studiously avoid:
It is conceivable, for instance, that the science of psychological testing be developed to the point where the character-development goals of the individual can in certain circumstances be better determined by this method than by naive preference on preference structures. Yet this occurrence would in no way affect the validity of this aspect of the Robbins Principle. For we are here faced with another "normative identification": we must treat the CPO as if it were the best guide and arbiter, independent of whether this is true in each particular case, or even in most cases. While the factual inaccuracy of this normative identification in no way invalidates the value judgment involved, it can by no stretch of the word be referred to as a "fundamental intuition" as is commonly asserted. 61
But justifiable fear of the implications of this kind of reasoning is precisely what makes the neoclassical "value judgment" behind the Robbins Principle so appealing! It is the notion of the sanctity of the individual that is the chief appeal of traditional welfare theory, not its seeming simplicity, or ethical neutrality. And the main reason that this notion is so attractive is that it appeals to people's healthy antitotalitarian instinct that people have the right to choose for themselves what kind of people they want to become and what kind of preferences they wish to develop. And this is, indeed, a "fundamental intuition!" Gintis' reasoning might be misread by some to imply support for preference structure choice by expert testing rather than by people's own choice should the former appear more "accurate" - whatever that would mean in such a situation. We believe that to paint welfare theory into the kind of comer described in the previous quotation would place a new welfare paradigm on the worst possible ground.
But to conclude, Gintis summarized what he accomplished by focusing on the problem of uncertainty in preference development:
I have emphasized that the seeming immediacy of neoclassical welfare economics results from a peculiarly simplistic representation of the individual choice situation and the relation of the individual to his social environment .... The assumption of "tastes as data" in the formation of social policy is neither a tautology, nor is it a factual statement in light of the violability of preference structures; it is yet another value judgment couched in terms of an "as if' statement. The new definition states that an individual is "better off' in one period, provided he would have chosen to be in that position in the preceding period. Such an individual need not be happier, better off, more secure, more satisfied, or even alive-but as economists we are instructed to place him there, insofar as it is possible.
The importance of this presupposition cannot be overstated; without this premise, the theory of allocation efficiency becomes inoperative from the point of view of decision making. Indeed, Pareto optimality has no meaning in the absence of this judgment; unable to say that one individual should be in a preferred position, we could hardly generalize to all. One could yet analyze conditions of "efficiency," but there would remain no reason to prefer an efficient state to any other .... We conclude that the methodological purity which in part accounts for the striking popularity of modern welfare economics is ephemeral. The premises are indeed simple in statement, but involve a number of separate judgments each by no means obviously true, and each wanting for justification. Doubtless none but the dogmatic could emerge from a genuine attempt at such justification without realizing the need to reintroduce a host of considerations not today considered a part of welfare economics proper .... Having shown welfare axioms to involve complex and questionable value premises, rather than standing as simple intuitions, it becomes plausible to search for replacements. 62
In Gintis' view, "few would be willing to leave the abstract and analytic treatment of Little and Samuelson" 63 unless already convinced that those waters were as ethically muddied as the ones they were headed for.
But according to Gintis, what are the implications of endogenous preferences for the conclusions of traditional welfare theory? It is one thing to argue that endogenous preferences signal the need for reconstruction of traditional welfare theory-for a different way of posing our valuative questions. It is another thing to argue that any of our conclusions regarding the welfare properties of different sets of economic institutions will be altered as a consequence of treating preferences as endogenous.
When Gintis finally ignored the difficulties he went to such pains to emphasize, and proceeded to "assess the ability of neoclassical analysis to come to terms with welfare problems in a situation where preferences change, but where there is complete knowledge on the part of the individuals as to the welfare effects of these changes," 64 what did he conclude? Gintis stated an equilibrium price structure will yield a Pareto optimal outcome in a competitive market economy under the usual assumption 65 That is, if all the traditional assumptions are retained, except preferences are treated as endogenous rather than exogenous, any perfectly competitive capitalist economy will still have a general equilibrium that will be Pareto optimal. But Gintis argued this is of little benefit because "there is no reason to believe the solution to the multi-period problem to be unique." 66 And, therefore, "the only realistic recourse is to posit a collective economic instrument (e.g., a State Planning Board) to replace the Invisible Hand in the determination of expected prices" 67 in light of multiple 'self-fulfilling' sets of price expectations compatible with maximization and market clearance.'' 68
Presumably, Gintis meant multiple equilibria "compatible with individual maximization," since if he meant social maximization in the sense of Pareto optimal , a Central Planning Board to choose the "correct" price path from the many equilibrium price paths would not be needed to guide individual maximizing behavior toward achieving social optimality-any path would be fine. In other words, Gintis surmised that endogenous preferences implied a competitive private enterprise market economy could take many equilibrium paths, at least some of which would not be Pareto optimal . We see no other way to interpret his words consistent with the conclusion that some sort of State Planning Board would be required to replace the invisible hand to choose the socially desirable as distinct from socially undesirable equilibrium price vector(s).
If Gintis were correct that some of the multiple equilibria might not be Pareto optimal, we would agree with him that the logical necessity of inserting a State Planning Board to guarantee Pareto optimality would carry us far indeed from the original conception of a competitive market economy. Especially since the knowledge required by the Planning Board to carry out its appointed function in such a system would have to be the complete and perfect kind envisioned in some models of Centrally Planned Economies. 69 But we will prove in chapter 6 that all equilibria envisioned by Gintis in a purposeful preference molding vision of endogenous preferences under the assumption of perfect foresight must necessarily be Pareto optimal if they exist. And, therefore, if endogenous preferences are to play a role in a substantive criticism of the welfare properties of competitive market economies, the criticism must be of a different nature.
A major part of our work treats both human characteristics and preferences as endogenous. Therefore, it is useful to summarize and evaluate the contributions of previous treatments of preference development.
1. Preference development as well as preference fulfillment should be analyzed as part of a meaningful welfare theory.
2. Individual preferences are not only endogenous to the major cultural, political, and kinship institutions of society, but to previous economic choices made by the individual, and therefore, presumably to any aspect of the economic structure that influences those choices.
3. At least in a partial equilibrium setting, the "separability axiom" of neoclassical welfare theory that permits independent treatment of supply and demand is untenable. As soon as individuals recognize their preferences are endogenous, it becomes irrational for them to ignore consideration of future expected prices when making decisions today that will have predictable effects on future preferences.
4. For welfare theorists to ignore the preference development effects of consumption and work choices and focus exclusively on the preference fulfillment effects leads to systematic misestimations of welfare effects of economic choice.
5. The Robbins judgment underlying traditional welfare theory, that economists should take tastes as givens, is by no means intuitively obvious, but requires a "number of separate judgments each by no means obviously true, and each wanting for justification."
How damaging is all this to traditional theory? Previous treatments make clear that taking preference development seriously raises difficulties that have long been ignored by most welfare theorists. To that extent, previous treatments of preference development are certainly "troublesome" for welfare theory as it has been practiced. But for any of us with a highly activated instinct for efficiency-which should be expected among economists as a product of our training rituals - the question arises whether taking preference development seriously will change any fundamental conclusions of welfare theory. If not, the task can be undertaken by those so inclined, at their leisure, while the rest continue with other tasks untroubled by doubts that the premises underlying their work are eroded.
The majority of those who have taken the time to treat preference development have concluded that doing so does not change any fundamental conclusions of welfare theory. The intuition of J. S. Mill was that private enterprise market institutions posed no difficulties for optimal preference development and efficient preference fulfillment. modern neoclassical theorists such as von Weizsacker, Pollak, El Safty, and Hammond who have taken the trouble to extend the traditional welfare paradigm to incorporate endogenous preference development have not been primarily concerned with welfare implications but with implications for the analytical theory of demand. Reviewing their work one would conclude that while endogenous preferences pose a number of technical complications, they can be adequately treated through natural extensions of neoclassical theory. No neoclassicist writing in the 1970s and 1980s has exhibited any concern regarding the optimality properties of competitive, private enterprise market equilibria, only concern with the existence of various utility functions and the stability and interpretation of various demand relations when preferences are treated as endogenous. Certainly nothing in their writing suggests panic over the traditional welfare paradigm!
Herb Gintis, on the other hand, came to very different conclusions. His treatment of preference development led him to conclude that the logical/philosophical basis of the Robbins Principle is far more complicated than neoclassical welfare theory would have us believe. He also argued it is fundamentally unacceptable, requiring a reconstruction of welfare theory. Finally, he argued that treatment of preferences as endogenous casts considerable doubt on the welfare properties of private enterprise market institutions.
We will end in agreement with many of Gintis' conclusions, but often for reasons quite different than those he offers. In chapter 5 we will argue for a reconstruction of welfare theory based on a paradigm that emphasizes personal and, therefore, preference development, but not because we reject the Robbins judgment. In chapter 6 we will argue that an appropriate treatment of preferences as endogenous does play a critical role in altering some of the most fundamental conclusions of traditional welfare theory. But we will also demonstrate that this cannot be for the reason Gintis believedthat some of the multiple equilibria in a competitive economy with endogenous preferences may not be Pareto optimal . Nor is it, as Pollak suspected, because nothing can be concluded about the welfare properties of systems once we countenance purposeful preference development. Rather, we will show that endogenous preferences play a critical role in a substantive critique of the welfare properties of competitive market and other economies, but one that is quite different from what previous analysts of endogenous preferences have imagined. We will also show that endogenous preferences are critical to explaining an important mechanism by which the characteristics of institutional settings affect the characteristics and desires of those who live and make choices within those settings, all of which will allow us to distinguish important welfare properties of different economic systems that are indistinguishable from the perspective of traditional welfare theory.
Traditional welfare theory cannot be faulted for failure to analyze the efficiency properties of different economic institutions-in particular the institutions of competitive markets and private enterprise. Whether one finds the conclusions illuminating, misleading, or both is another matter. But with the exception of inquiring about the distributional flexibility of economic systems, that is the extent of the traditional analysis of the properties and consequences of institutional structures. Specifically, traditional welfare theory fails to ask:
1. How do different institutional structures influence the pattern of preference development?
2. How do different economic institutions affect incentive structures, which in turn help define rational behavior itself?
Or, as Herb Gintis put it "does the set of approved institutionalized decisionmaking mechanisms itself bias the way the individual approaches the problem of personal development, and if so, does it do so in a 'desirable' direction? An acceptable welfare theory must satisfactorily deal with these central issues." 70 This critical "blind spot" in neoclassical theory is well illustrated in Roy Harrod's description of neoclassical methodology:
The method of procedure is to take certain elements of the structure as givennamely the preference lists of individuals for goods and services, the terms on which they are willing to contribute their assistance to production, and the current state of technology .... The object of this procedure would be to provide means of showing how changes in the fundamental data, desires, etc., will govern the course of events. 71
But something is missing from Harrod's list of fundamental data: the institutional context in which technology and preferences are developed in the first place. It is hard to believe anyone would seriously argue that the development of technology or preferences is a random walk with respect to institutional environment. Yet the only alternative interpretations of the traditional view are that the specific institutions of private enterprise and competitive markets are (1) totally neutral, or (2) uniquely optimal with respect to effects on preference development and technological change. Before spelling out the objection, it is worth exploring each of these claims. The most literal interpretation of the traditional approach is that, at most, economic institutions can influence preference development and/or techological. change only minimally, and the impact can be ignored for practical purposes. In this case, changes in preference orderings and technology are taken to occur independently of economic institutions and should simply be "taken as givens" for economic analysis.
If this is not the view of traditional welfare theory, for example, if traditional theorists wish to argue that some undesirable economic institutions thwart or bias preference or technological development, then we must examine the reasons for believing that the particular institutions of private enterprise and competitive markets can be assumed not to influence Harrod's "fundamental data." It is conceivable that though many economic institutions distort preference and technological development, institutions of perfectly competitive capitalism can be ignored as neutral in this regard. Or, it could be that along with their supposed virtues of efficiency and equity flexibility, private enterprise and competitive market institutions optimally affect preference and technological developments as well.
While numerous theorists have argued that the combination of private enterprise and competitive market institutions is felicitous regarding technological progress, 72 very few traditional theorists have considered effects of these institutions on preference development.
The usual argument regarding technology has two parts: (1) private enterprise stimulates innovators by allowing them to reap the rewards of their cleverness, and, (2) competitive markets hasten the spread of clever ideas on pain of being driven out of business. We could make the point that there seems to be something of a contradiction here: If competitive markets maximize the spread of clever ideas, how is it that innovators capture the full rewards of their clevemess? 73 But we choose to highlight the fact that most neoclassical treatments of relations between private enterprise, competitive market institutions, and technological change have focused entirely on the degree of effort they are likely to stimulate without questioning the directions those efforts take. We have already seen that according to the conflict theory of the firm the most profitable technique may not coincide with the most efficient technique. While this reasoning applies to choice of techniques from the set of known technology, it certainly also suggests that under private enterprise we can expect to find a bias in the direction technological change takes. "Profitable" technological change should not necessarily be equated with "efficient" technological change no matter how rapid innovation may be.
As we have seen, traditional welfare theorists "abstract" from the institutional impact on the "fundamental data" of tastes, only to proceed to judge the efficiency of those same institutions in terms of the "fundamental data." John Stuart Mill was exceptional in providing at least a semiexplicit justification for this procedure, and Alfred Marshall expressed his opinion in passing that private enterprise market institutions have an optimal effect on preference development. 74 But most traditional theorists simply take the abstraction for granted.
Of course, the obvious place to look for treatment of this issue is among institutionalist economists. But surprisingly, while institutionalist economic theory has long proclaimed the importance of institutional structures in molding economic behavior, their insights have never been incorporated into welfare theory. A principal reason is that most institutionalist admonitions that institutionalized social pressures strongly influence economic behavior have been formulated as alternatives to explanations in terms of "rational choice." Institutionalist concepts such as "habit," "inertia," "peer pressure," and "invidious comparisons" are usually counterpoised to welfare theoretic concepts such as "individual rational choice" and "opportunity sets." Moreover, descendant from the writings of Thorstein Veblen, institutionalist economists have retained a strong methodological injunction to forswear what others call "normative" economic theory as little more than thinly disguised ideology in order to develop what they envision as truly "nonteleological," "evolutionary" analytical economics.
Regardless of reasons, economists with an institutionalist "bent" have been loath to transgress the preserves of welfare theory-particularly since welfare theory became ever more densely forested with advanced mathematical species. And welfare theorists have seldom considered institutionalist notions to be other than inconsistent with their own approach. We can only lament the consequences of an almost complete lack of cross-fertilization. Recognition that individual rational choice can only be institutionally defined remains far from universal. And construction of a careful analysis of the implications of different economic institutional structures for rational individual choice has long been delayed.
An excellent example of one of the infrequent attempts at cross-fertilization is an article by John Sawyer published in the American Economic Review75 . Sawyer offered some important insights, but regrettably displayed just the kind of attitudes that have precluded incorporation of his insights into welfare theory.
Sawyer began by insisting that "the institutional context cannot be taken as constant or assumed away," and that "the ways in which the formal and informal social structure persistently shapes the behavior for all economic actors" is greatly underappreciated. Moreover, he rejected as totally insufficient the "formal bow to the importance of 'non-economic' factors ... cus tomary in opening chapters of economic textbooks and opening lectures of elementary economics courses." 76 He then observed:
When forced back into sociocultural categories we rapidly discover how little we know ... of how to incorporate such categories of explanation into the systematic analysis of economic processes .... The newer social sciences, however, offer useful concepts .... One of the most useful of these approaches-so-called "functional-structural" analysis, as developed in contemporary sociology-focuses attention on the extent to which social action tends to be "structured," both as to ends and means. Any society necessarily institutionalizes particular goals and values, patterns of social relationship, and ways of doing things requisite to the continued operation of that social system; in sociological language, it prescribes patterns of normative conduct in a system of structured roles .... 77 It is evident that different institutional systems will be more or less encouraging or discouraging to particular patterns of economic activity. 78
These are important insights indeed. Moreover, as Sawyer indicated, they are no longer avant-garde, or the exclusive property of social critics. These propositions have become fundamentals of mainstream sociology. Yet they continue to be almost totally ignored by traditional economic theory. Unfortunately, Sawyer minimized the potential impact of his observations on welfare theory.
First, he focused his insights on technological change rather than preference development. "Our interest here is in the economic drag, the rigidities and resistances to economic development." 79 Next, Sawyer applied his insights concerning the importance of institutional structures only to the noneconomic sphere. He argued for the importance of analyzing effects of social and cultural institutions on "productive" economic behavior, but it did not seem to occur to him that different economic institutions might have differential effects on development of preferences over economic opportunity sets. He certainly failed to extend his insights in this direction. Finally, he recommended that economists must borrow from modern theories of human motivation and behavior of sociologists and psychologists. But he warned that "since we do not yet have the social theory or understanding of the human personality adequate to demonstrate precise interrelationships," it must be recognized that we "clearly cannot offer quantitative solutions." 80 Leaving aside the "precisely" predictable psychological effects of such language on today's proud and highly mathematical economists, Sawyer's instincts distract from the most fundamental implications of his insights.
The problem for welfare theorists is not to explain and predict actual behavior but to explain what would be rational behavior under stipulated conditions. It is enough to recognize these are two very different tasks and point out that welfare theory has, for better or worse, concerned itself with the latter rather than the former, to explain why Sawyer's insights are so easy for welfare theorists to dismiss. But however clear most welfare theorists are about the difference between the two tasks, confusion is rampant within the profession at large. Clarification and motivation for the task pursued by welfare theorists is in order.
Many assume the sole scientific purpose of welfare theory-investigation of the implications of individual rationality-is to contribute to an understanding of actual economic behavior. In this view, the aesthetic quality of investigations of abstract rationality may appeal to those who engage in its intricacies and prove sufficiently alluring to guarantee its continuation as an art form. But unless the conclusions of welfare theory can be brought to bear on explanations of actual human behavior, they have no scientific legitimacy. For example, institutionalist economists, following in Veblen's wake, have preferred concepts such as "habit ... .. inertia," and "status" to the "hedonistic calculus" as explanations of actual human behavior. But institutionalists are not alone. Much in modern psychological and sociological literature suggests that a great deal of human behavior cannot be explained by appeals to rationality, hence, the conclusion that welfare theory has but a limited role to play in an analytic theory of human economic behavior.
We do not need to take a stand on this issue. While we believe welfare theory has a role to play in analytical theory because we believe elements of rationality frequently are displayed in individual economic choice, we are more than prepared to admit that role is limited. For us the undeniable importance of welfare economics is as an evaluative theory of economic institutions. If we wish to know what impact an economic institution has on human well-being, we must evaluate the kind of behavior that institution promotes. Does a particular economic institution promote behavior that is socially efficient? Does it promote outcomes in accord with some definition of equity? Does it promote human development that renders people capable of garnering greater satisfaction from the circumstances they will face in the future And when we inquire how institutions "promote" one kind of behavior rather than another, we discover they do so by making some behavior individually rational and other behavior individually irrational.
The important distinction here is that while an economic institution may "promote" some kind of behavior by making it "individually rational," and while such behavior may be socially productive or counterproductive in ways welfare theory spells out with great care, we do not have to believe people always behave in accord with the behavior "promoted" in order to take great interest in the kind of behavior promoted. In a certain sense, the degree to which people follow the dictates of their institutional environment is irrelevant. To the extent that an economic institution has an impact on human welfare, its impact must be measured by the coincidence or discrepancy between the behavior it promotes as individually rational and socially rational behavior. 81 In other words, it is the task of welfare theory to define and compare individually rational and socially rational behavior, irrespective of the extent to which either pattern of behavior coincides with actual behavior. Only in this way can we evaluate the impact of different economic institutions. In this sense, welfare theorists are correct in viewing the details of more or less accurate sociological and psychological theories of actual human behavior as effectively irrelevant to their project. But this does not mean that a theory of rational individual choice from economic opportunity sets does not require a component that analyzes the effects of specifically economic institutions.
It is only because modern sociologists have taken human developmental effects of the institutions that concern them seriously that such analysis is labeled "sociological." In this sense, yes, welfare economics needs a "sociological" component. But by this we mean welfare theorists should "think" in one of the ways sociologists think. Economists should analyze developmental effects of the institutions in whose analysis they specialize. This is quite different from integrating the insights of sociologists concerning social but noneconomic institutions and the determinants of actual human behavior with the insights of economists concerning the efficiency properties of different economic institutions-which is our reading of the research program proposed by Sawyer and others from the "institutionalist" school of economics through the years.
Moreover, in the sense that modern welfare theory has always provided "quantitative" solutions, a welfare theory recognizing the influence of economic structures on choice and preference development can do so as well. In other words, the project we propose need not be any less "quantifiable" than traditional welfare theory. To the extent a theory of rational individual choice as defined by institutional context has anything to contribute to a theory of actual human behavior, sociologists and psychologists would be well advised to borrow from the theories of a new breed of welfare economists! How is that for an incentive compatible mechanism for welfare theorists?
In any case, we will make these issues a major focus of our welfare theory and analysis throughout the rest of this book. Without revealing all our results prematurely, we can outline the logic of why analyzing the effect of economic institutions on preference development is particularly important if welfare theory is to have anything further to say.
The influence of the traditional paradigm on the analysis of preference development can be summarized quite briefly. The most sacred methodological principle of traditional welfare theory is a categorical injunction to economists not to question where preferences come from. Following Gintis we dub this methodological injunction the Robbins Principle and simply note that in cultures considered less civilized it would be called a taboo. Whether we will be able to preserve the notion of the sanctity of the individual while recognizing the importance of preference development remains to be seen. But there is no doubt the traditional paradigm long discouraged analysis of preference development.
The effect of the traditional paradigm on exploring the effects of economic institutions on preference development has been no less damaging. Traditional welfare theory sensitizes us to the efficiency effects of institutional structures-that is, the effect of economic institutions on preference fulfillment. But a "blind spot" in the traditional paradigm has made it all but impossible for traditional theorists to even recognize the potential importance of the effects of economic institutions on preference development.
It is our major contention that exploring the relation between institutional structures and preference development is one of the most fertile fields welfare economists have yet to cultivate. It is also our contention that diminishing returns have set in from extensive cultivation using traditional tools of the relation between familiar institutions and preference fulfillment, and that traditional welfare theory is increasingly unable to distinguish between the major economic systems on welfare theoretic rather than practical grounds. Of course, we must substantiate both these claims as we do in parts 2 and 3 of this book. But assuming, for now, we succeed in arguing our case, the pernicious effect of the traditional paradigm is apparent. The Robbins judgment interpreted as "taboo" has long combined with the institutional "blind spot" to obstruct progress on what we find the most promising frontiers of welfare economics. We close with a brief explanation of what is at stake between the traditional and new welfare paradigms.
Until recently welfare theorists had seriously analyzed only a small number of abstract economies, or combinations of economic institutions. They combined competitive markets and private enterprise production to analyze what is commonly called Perfectly Competitive Capitalism (PCC). They combined public enterprise and central planning to analyze what many call Centrally Planned Socialism (CPS). And they combined competitive markets and employee-managed public enterprise to analyze what some call Workers' Self-managed Market Socialism (WSMS). Prior to the quiet revolution in welfare theory that has been recently unfolding in response to Hurwicz's call for theorists to think of themselves as economic mechanism designers, there were very few other well-defined, abstract, economic models. One exception was the model developed by Oskar Lange, Abba Lerner, and Frederick Taylor. By combining markets with a kind of public enterprise in which a state agency appoints managers and directs them to follow a few simple rules, Lange, Lerner, and Taylor developed a different model of what they called Market Socialism (LLTMS) in a conscious effort to refute the proposition that "socialist" economies could not be efficient. 82 But the work of Lange, Lerner, and Taylor was deemed only a footnote in the long debate about whether or not idealized models of the different major economies could achieve Pareto optimality . 83
We will avoid use of the value-laden words "socialism" and "capitalism," and employ more specific and less "politicized" labels throughout the remainder of this book. We will refer to Private Enterprise Market Economies (PrEMEs), Public Enterprise Centrally Planned Economies (PuECPEs), and Public Enterprise Market Economies (PuEMEs) of two different kinds: Public Enterprise Employee-Managed Market Economies (PuEEMMEs) and Public Enterprise State-Managed Market Economies (PuESMMEs). Regardless of the comparative merits of different definitions of "socialism" and "capitalism," something, no doubt, can be said for "defusing" the discussion.
In sum, welfare theorists have spent considerable time analyzing efficiency properties of a small number of combinations of economic institutions. Though opinion has long been divided, and practical matters are to be considered, it is our belief this "old welfare debate" is essentially over. In later chapters we demonstrate that under similarly generous assumptions, from the perspective of traditional welfare theory, all of the above abstract models are equally capable of generating efficient social outcomes. Moreover, they are all equally "flexible" regarding equity.
While this may come as a surprise to many outside the field, and a few within, it is certainly not unknown to all welfare theorists. And while we save the demonstration for later chapters, we can usefully interpret the conclusion here.
If all the above abstract economies are equally efficient and flexible from the perspective of traditional welfare theory, and if traditional theory defines welfare entirely in terms of efficiency and income flexibility, traditional welfare theory cannot distinguish between these different fundamental economic models. 84 This is not to say traditional welfare theory cannot recognize that the systems are characterized by different economic institutions. But from the perspective of traditional welfare theory the institutional differences do not affect the welfare results.
We believe this is a rather awkward position for welfare theory. It seems to us a reasonable spectator of the welfare theory research project might fail to see the "efficiency" of complicated analyses that are technically taxing, but explain only how "practical" considerations might prevent real world approximations of major economic systems from achieving the equally beneficent results that all achieve in theory. One of the principal advantages of the paradigm and welfare theory we champion is that it can distinguish between welfare properties of these economic models by distinguishing the effects of different institutional "fundamental data" on preference and human development.
Anticipating the findings of parts 2 and 3:
1. Traditional welfare theory, if applied evenhandedly, is unable to distinguish between the performances of the different major economic systems except on the basis of "practical" criticisms.
2. Our new welfare theory, on the other hand, predicts very different results for different major economies and identifies different "welfare theoretic" as well as "practical" problems with each.