3. Democratic Values
Egalitarianism as Prudence
Uncertainty about the share they will get is supposed to induce rational people to opt for an income distribution which only the certainty of getting the worst could make them choose.
A bird in the hand is best if we must have one and if two would be too many.
If the core of Rawls's Theory of Justice was vulgarized à outrance, it could perhaps be summed up thus: Devoid of the vested interests bred by self-knowledge, people opt for an egalitarian society allowing only such inequalities as improve the lot of the least advantaged. This is their prudent option, because they cannot know whether they would do better, or worse, in an inegalitarian society. Refusing to gamble, they take the bird in the hand.
Any sophisticated intellectual construction is inevitably reduced to some easily communicated vulgarization by the time it takes root in the broad public consciousness. Only the most robust arguments, whose core is of one piece, do not in such a process get reduced to pathetic fallacies. An author who needlessly invokes complex solutions to problems which have been assumed away to begin with, soon finds that for example he is publicly reputed to have "proved by game theory" that maximin (maximizing the minimum among alternative outcomes) is the optimal life-strategy for "prudent men," that "the conservative decision rule is to agree to moderately egalitarian social policies" and other words to this effect. Given the value of such terms as "prudent" and "conservative," myths of this type are liable to sway many minds for some time to come, albeit for reasons which Rawls would be the first to disavow.
In his system, the characteristics of the "original position" (ignorance about one's vital particulars coupled with some selective general knowledge of economics and politics), and three psychological assumptions, together determine what people would decide if put in such a position. They will choose Rawls's second principle, notably the part of it enjoining the maximization of the minimum lot in an unknown distribution of lots, or "difference principle." (The case for saying that they will also choose the first principle concerning equal liberty, and bar any more-of-one for less-of-the-other type of compromise between liberty and other "primary goods," is much less open-and-shut, but we will not concern ourselves with that.) The first point at issue is whether the psychological assumptions leading to the maximin choice can properly be made about rational men in general, or whether they represent the special case histories of somewhat eccentric persons.
The end postulated for the rational man is the fulfilment of his life-plan. He ignores its particulars except that it takes a certain sufficiency of primary goods to fulfil it; these goods, then, serve needs and not desires.*37 However, it is hard to see what else makes a fulfilled life-plan into a worth-while end if it is not the expected enjoyment of the very primary goods which go into its fulfilment; they are the means but they must also be the ends.*38 The latter is really implied in their being goods whose index we seek to maximize (rather than merely bring to a level of adequacy) for the least-advantaged. Yet we are told that people are not anxious to have more of them once they have enough for fulfilling the plan. They show no interest in its over-fulfilment! This position is ambiguous, if not downright obscure.
To dispel the ambiguity, one could suppose that people want to fulfil the life-plan, not because of the lifelong access to enjoyable primary goods for which it is a shorthand symbol, but as an end in itself. The life-plan is like climbing Piz Palu which we just want to do, and primary goods are like climbing boots, of no value except as tools. The life-plan either succeeds, or it fails, with no half-way house. It is not a continuous variable, of which it is good to have a little and better to have a lot. It is an either/or matter; we do not want to climb Piz Palu a little, nor can we climb higher than its peak. The lack of interest in more than a sufficiency of primary goods would then make sense, too, for who wants two pairs of boots for climbing one mountain?
This logical consistency between the end and the means (a necessary condition of rationality) would, however, be bought at the price of imputing to rational men much the same absolute view of the life-plan that saints have of salvation. Damnation is unacceptable; salvation is exactly sufficient and nothing else matters besides; it is nonsense to want more salvation. The life-plan is an un-analysable whole. We do not and need not know what the good is of fulfilling it. However, it seems meaningless to wish to more-than-fulfil it, and utter hell to fall short.
There is nothing irrational per se in imputing an uncompromising, saintly mentality to people engaged in devising distributive institutions; saints can be as rational or as irrational as sinners. The problem is rather that, unlike salvation which has profound meaning and content for the believer, the life-plan is emptied of content if it must be abstracted from command over primary goods (i.e. if the latter are to be stopped from serving as ends); can it still be sustained that it is the goal of the rational man to fulfil it, though it looks an unexplained eccentricity to want to do so? Besides this, it is hardly worth mentioning that interpreting the life-plan as an ultimate end, and an all-or-nothing affair at that, is forbidden by Rawls's own view that it is a mosaic of sub-plans which are fulfilled separately and perhaps also successively (see chapter VII), i.e. not an indivisible goal in which you either succeed or fail.
The significance of this question resides in the role three specific psychological assumptions are called upon to play in making rational people "choose maximin." Take the last two first. We are told (1) that "the person choosing... cares very little, if anything, for what he might gain above the minimum stipend" (p. 154), and (2) that he rejects alternative choices which involve some probability, however minute, that he might get less than that, because "the rejected alternatives have outcomes that one can hardly accept" (p. 154). If these two assumptions were to be interpreted literally, the choosers would behave as if they had the single-point objective of climbing to a chosen mountain-top. They would go for a critical quantity (index number) X of primary goods like for a pair of nailed boots; less would be useless and more pointless.
If, in addition, they knew that opting for a society with a maximin-governed distribution of primary goods (income) would in fact produce for its least-advantaged members the critical stipend X, they would choose it regardless of the relative probabilities of getting a bigger, equal or smaller stipend in other kinds of societies. If worse alternatives are simply unacceptable and better ones leave you cold, it could not possibly matter how probable they are. Your maximand is discontinuous. It is the single number X. If you can get it at all, you take it. Talking of a "maximin" strategy and of "choice in the face of uncertainty" is the very paradigm of the red herring.
(What happens if a maximin-principled society turns out not to be rich enough to assure for everybody a high enough minimum stipend, such as X, sufficient to let them fulfil their life-plans? Rawls is satisfied that since such a society is both reasonably just and reasonably efficient, it can safely guarantee X for everybody [pp. 156 and 169]; the certitude of X, then, is a preferred alternative to facing incertitude.
This, of course, is as it may be. A society may be efficient, yet quite poor—the successive Prussias of Frederick William I and of Erich Honecker would probably both fit this bill—and people in the original position have no clue whether the efficient and just society they are about to devise might not be quite poor, too. James Fishkin takes the view that if a society can guarantee everybody's satisfactory minimum, it is a society of abundance "beyond justice."*39 On the other hand, if the stipend guaranteed by enacting maximin fell short of the critical X, people could not both regard the meagre guaranteed stipend as one they "can hardly accept" yet rationally choose it in preference to non-guaranteed, uncertain but more acceptable alternatives.)
If uncertainty is to be something more in Rawls's theory than a redundant catch-word, a passport to the fashionable land of decision theory, his life-plan and his two psychological assumptions about the minimum stipend (i.e. that less is unacceptable and more unnecessary) must not be taken literally. Though primary goods fulfil "needs and not desires," we must firmly recall that they are consumable goods and not tools; that no matter how little or how much of them people have, they are never indifferent to having more; and that there is no significant discontinuity, no void above and below the satisfactory minimum stipend, but rather an intense "need" for primary goods below and a less intense "need" above it, so that the index of primary goods becomes a proper maximand, a fairly closely spaced schedule of alternative numbers, fit to be ordered consistently, instead of one lonely number. Rawls wishes the theory of justice to be a particular application of the theory of rational choice; if his assumptions are taken at face value, all occasion for choice is shut out in advance; we must interpret them more loosely so that they leave room for genuine alternatives.*40
Having done so, we find that we have in fact glimpsed the outline of the utility function of the people concerned (despite Rawls's protestations that they behave as if they had none). It conforms to the conventional supposition of diminishing marginal utility at least in the neighbourhood of a level X of primary goods. (There is a presumption, arising from Rawls's remarks, that it conforms to it in more distant ranges, too.) If people were oblivious of this, they could not be conscious of the greater or lesser acceptability of various stipends of primary goods, and would not feel an imperative "need" to get at least so much, nor a much less compelling "need" to get more. Unless they had some such awareness of the relative intensity of their "needs" (or desires?), they could not rationally evaluate mutually exclusive uncertain prospects of getting different lots of primary goods, except for judging that one prospect was infinitely valuable and the others were worthless.
Consider next Rawls's first psychological assumption about "sharply discounting estimates... of probabilities" (p. 154). People (still in the original position) are required to choose between principles which determine types of society, which in turn entail particular income distributions, under each of which they could find themselves drawing any one of the different lots of primary goods which reward differently situated people in that type of society. They can, as we know, choose an equal distribution, or maximin (likely involving some inequality), or one of a possibly large number of feasible distributions, many of which will be more inegalitarian than maximin.*41 We also know that maximin dominates equality,*42 i.e. that no rational and non-envious person will choose the latter if he can choose the former. Other than that, however, the mere requirement of rationality leaves the remaining choices wide open as between maximin and more unequal distributions. People are uncertain what their own lot would be in each, and have no objective data at all for guessing. They are, nonetheless, said to choose one and take their chances under it.
Since they are rational, the distribution they do choose must have the property that the utilities of the alternative lots that can be drawn under it, each multiplied by the probability (0 ≤ 1) of drawing that particular lot, yield a larger total sum than would any other feasible distribution. (For "yield" one may wish to substitute "are thought to yield.") This is merely a corollary of the definition of rationality. In technical language, we would say "it is analytic that the rational man maximizes the mathematical expectation of utility."*43 The limiting case of uncertainty is certainty, where the probability of drawing a given lot is 1 and that of drawing any other lot is 0. The rational man can then be said to be simply maximizing utility and never mind its probability.
Rawls is free to assert that his parties are "sceptical" and "wary of probability calculations" (pp. 154-5). If they do choose in the face of uncertainty, which is what they have been put in the original position to do, their choices amount to imputing probabilities to outcomes, no matter whether they do it sceptically, confidently, anxiously or in any other state of emotion. We are even free to insist that they do no such thing. All that matters is that their behaviour would make sense if they did. If their conduct cannot be described in such terms, the assumption of their rationality must be given up. We can say, for instance, that people attach a probability of 1 to drawing the worst lot and probabilities smaller than 1 but greater than 0 to drawing each of the better lots; but we cannot in the same breath say that they are rational. If they were, they would not implicitly contradict the axiom that the odds of drawing all the lots add up to one.
It is easy enough to accept that if rational people were certain of drawing the worst lot under any income distribution, they would choose the one which had the "best worse" (maximin). This would always be the best play in a game where they could choose the distribution and the opposing player (their "enemy") could assign them their place within it, for he would be sure to assign them the worst one.*44 Rawls says both that people in the original position reason as if their enemy was going to assign them their lot (p. 152), and that they should not reason from false premises (p. 153). Presumably, the fiction of an enemy is intended to convey, without quite saying so, that people act as if they imputed a probability of 1 to the worst lot. In fact, maximin is designed to deal with the assumed certainty that our opponent will make moves that help him most and hurt us worst, but conveying this without saying so does not make the idea sensible in a situation where there is no enemy, no competing player, no opposing will, in short, where there is no game, only gratuitously introduced game-theory language.
Each person in the original position knows without a doubt that any unequal distribution of lots must by its nature contain some lots that are better than the worst one, and that some people will draw them. What can make him sure that he won't? He has "no objective ground," nor any other cause for reasonable belief, that he has no chance of being one of these people. But if the better lots do have non-zero probabilities, the worst one cannot have a probability of 1, or else the odds would not add up. Hence whatever rational people may choose in the original position, they do not choose maximin except by a fluke (in the course of "randomizing" in a mixed strategy?), so that the likelihood of unanimous choice is as good as nil and the theory is aground.*45
A straightforward way to refloat it would be to jettison rationality. This would be all the more tempting as real people are not obliged to be rational. They are quite capable of tying themselves up in amazing logical inconsistencies. They can both accept and contradict a given axiom (such as the one that if one outcome is certain, the others must be impossible). Freed of the harsh and perhaps unrealistic discipline of rationality, they can be supposed to behave any way the theorist may fancy. (For instance, in his numerous writings on the theory of risky choices, G. L. S. Shackle substituted poetic and pretty suggestions about human nature in place of the arid calculus of probability and utility. The "liquidity preference" of Keynesian economics is at bottom also a resort to suggestive poetry. Many theories of producers' behaviour rely on assumptions of non-rationality—full cost pricing, "growth" and market share objectives, rather than profit maximization, are well-known examples.) Once conduct need no longer conform to a central maximization assumption, "anything goes," which is precisely the weakness of such approaches, though this need not prejudice their suggestiveness and teachability.
It takes only a modicum of poetic licence to impart the idea that it is a sensible thing to vote for a type of society in which you would not come to great harm even if your particular place in it were designated by your enemy. Thus is a non-rational, impressionistic case established for maximin, the egalitarian bird in the hand as the counsel of conservatism, prudence and moderation.
Perhaps without realizing that he has moved on to non-rational territory, Rawls bolsters this case, in the spirit of his reflective equilibrium, by two related arguments. Both appeal to our intuition and he seems to regard both as decisive. One is the strain of commitment: people will refuse to "enter into agreements that may have consequences they cannot accept," especially as they will not get a second chance (p. 176). This is a puzzling argument. If we play "for real," we may of course lose what we stake. We do not get it back to play with again. In this sense, we never get a second chance, though we keep getting other chances in subsequent plays. They may be worse ones, in that we enter them weakened by the loss of our stake in the first play. Poker and business do have this cumulative character, where nothing fails like failure and chance favours the longest purse; pure games of chance and games of skill do not. Admittedly, if we draw a poor lot of primary goods, under the assumptions of the Theory of Justice, we will not get a chance to draw again in our and our descendants' lifetime. Social mobility is ruled out. Yet there is still a multitude of other gambles ahead, where we can be lucky or unlucky. Some of them, such as the choice of wife or husband, having children, changing jobs, may be as decisive for the success or failure of our "life-plan" as the "stipend of primary goods" we have drawn. Naturally, a low stipend may affect our chances in these gambles.*46 Gambling for the lifetime stipend is, therefore, sure to be one of the most important gambles we ever face, which should by rights be an argument for, and not against, applying to it the rules of rational decision making.
If we know at all what we are doing, the term (for a lifetime, for all posterity) over which a given lot of primary goods, once drawn, is to last us, must of course be built into our valuation of each such lot from the worst to the best. It is precisely its lifetime term which explains why it is our entire life-plan which determines the relative intensity of our "need" for various-sized lots of primary goods. If drawing the lot of a dim-witted, idle beggar means living his life till we die, we are bound to weigh the risk of it very carefully. Our mathematical expectations of the utility of the lots among which there is such a repulsive one, must already reflect all our dread of this prospect. It seems double counting that, re-baptized "strain of commitment," it must reflect the same dread a second time.*47
No doubt we weigh the risk of death seriously. Death, whatever other prospects it may hold, in our culture is taken to exclude a second chance at earthly life. But it is obviously wrong to assert that the "strain of commitment" to an unacceptable outcome makes us refuse the risk of death. Our everyday peacetime life is abundant proof that we do not refuse it. Why would the risk of living a dim, idle and beggarly life be different in kind? It must all depend on our assessment of the probabilities characterizing the risk and of the attractiveness of the possible rewards we can earn by taking the risk. The "strain of commitment," if there is one, is a legitimate consideration entering into these assessments. As a separate and overriding consideration, it is at best poetry.
Finally, it is incomprehensible to be told that good faith would stop us from accepting the strain of commitment, since if we took a given risk and lost (e.g. voted for a very inegalitarian income distribution and found ourselves in bottom place), we might not be able or willing to pay up (i.e. to accept the bottom place). If someone lets me bet him a million dollars which (unlike "Bet-a-million Gates") I do not have, I am acting in bad faith and he is acting rashly. But the "original position" of Rawls is not credit betting. If I turn out a dim bottom-person in the society I chose and which treats such persons badly, there is no obvious way in which I can "default." How do I refuse to honour my bet and play my allotted role of a dim bottom-person given that I am one? How do I extort from the more privileged members of my inegalitarian society a satisfactory minimum stipend and an agile brain? Considering that I could not if I would (and that as a dim person I may not even want to), the fear of my own default will not stop me. Good or bad faith, weakness of will and shame at not honouring my bet do not enter into it.
A separate informal argument contends that people will choose maximin, i.e. a tempered egalitarian distribution favouring the worst-placed, in order to make their decision "appear responsible to their descendants" (p. 169, my italics). Now it is one thing to be responsible and another to appear, to be seen to be so (though the two may overlap). If I want to do what I think is best for my descendants and never mind how my decision will look to them, I am acting as if I were a principal. In seeking to do as well for them as I would for myself, I might allow for their utility (say, the time-pattern of their "need" for primary goods) to be different from mine. My rational decision, however, must still correspond to the maximization of expected utility, except that it is my best guess of their utility I will try to maximize. If maximin is not rational for me, it does not become rational for my descendants either.
If, on the contrary, my concern is how my decision will look, I am acting as an employee or a professional adviser would rationally act for his principal. In addition to the latter's interest, he would consider his own. It is difficult to devise conditions in which the two are certain to coincide. For example, if he made a gain for his principal, his own reward, fee, salary or job security might not increase proportionally. If he made a loss, his own loss of job or reputation as a responsible treasurer, trustee or manager might be more than proportional. As his assessment of the ex ante risk entailed in an ex post gain need not be the same as that of his principal, it cannot even be said that if instead of acting selfishly, he tried to maximize his principal's gains he would be acting (i.e. taking the same gambles) as would the principal.*48 In general, it is unlikely that if he maximized his expected utility, he would also be maximizing that of his principal, or vice versa. The two maxima will tend to diverge, the decisions of the employee being usually biased to ward off possible blame and to conform to conventional wisdom; the principal for whom he is acting cannot know that this conduct does not maximize his utility but only that of the employee.
If maximin, a bird in the hand and selling your uncertain birthright for a guaranteed mess of pottage were asserted often enough to be the responsible thing to do, the employee would rationally have to opt for them if his maximand was best served by appearing responsible to his principals, like Rawls's contracting parties who want to appear responsible to their descendants. Here, then, is a fairly successful deduction of moderate egalitarianism from rationality. Rawls has accomplished this at the cost of having parents arrange the future of their children with a view, not to the latter's best interests, but to what would probably make them look prudent in their children's eyes. Some parents no doubt do behave like this, and some might even help install the welfare state in order that their children should praise their forethought;*49 but on the whole the argument hardly looks strong enough to explain the terms of a unanimous social contract and to support a whole theory of justice.
Notes for this chapter
John Rawls, "Reply to Alexander and Musgrave," Quarterly Journal of Economics, 88, 1974.
Cf. the diagnosis of Benjamin Barber, "the instrumental status of primary goods is compromised" (Benjamin Barber, "Justifying Justice: Problems of Psychology, Measurement and Politics in Rawls," American Political Science Review, 69, June 1975, p. 664). His reason for finding this, though, is different from mine.
James Fishkin, "Justice and Rationality: Some Objections to the Central Argument in Rawls's Theory," American Political Science Review, 69, 1975, pp. 619-20.
Formally a believer faced with the alternatives of going to heaven or to hell (and who knows neither purgatory, nor degrees of heaven from first to seventh), would be exercising rational choice by opting to go to heaven. However, the surrounding assumptions render the choice problem trivial, or rather phoney.
This must obviously remain the case no matter how much Rawls's first principle (equal liberty, whatever that may mean) and the second part of his second principle (positions open to talents) restrict the set of feasible distributions by hindering the occurrence of very small and very large incomes (pp. 157-8)—a hindrance we may well admit for purposes of argument, without conceding that Rawls has established its likelihood.
For completeness, we may add that if maximin dominates equality, it must also dominate income-distributions intermediate between maximin and equality, i.e. all distributions more egalitarian than itself.
A frequently committed howler is to confuse the mathematical expectation of utility with the utility of the mathematical expectation. (The coincidence of the two would permit the statement that the marginal utility of income was constant.) A related howler is to double-count the utility function and the attitude to risk, as in "he does not maximize utility because he has an aversion to risk," as if risk-aversion were not just a more colloquial term for characterizing the form of his utility function. Cf. Rawls's version of the argument in favour of maximizing average utility: "if the parties are viewed as rational individuals who have no aversion to risk" (p. 165, my italics), "prepared to gamble on the most abstract probabilistic reasoning in all cases" (p. 166, my italics), but not otherwise, they will maximize the mathematical expectation of utility calculated with the help of Bayesian probability. But in behaving at all sensibly, they must be doing this anyway! If they are averse to risk, they will take one gamble and if they are not, they will take another. If "refusing to gamble" is purported to be rational, it must be capable of being described as the gamble where the sum of the utilities of the possible outcomes, multiplied by their probabilities (which are all zero except for one outcome whose probability is unity), is the highest. It is virtually impossible so to describe the refusal to accept the very small probability of losing a very small sum for the sake of the remaining very high probability of gaining a very large sum, i.e. the requirement is not an empty one.
Probability, as the context should have made clear, is the "subjective" kind of which it is meaningless to say that it is unknown. Only "objective," frequency-type probability tolerates being described as "known" or "unknown," and it tolerates it badly at that!
There is one other way in which people can be represented as "refusing to gamble": we can suppose that they just sit down and cry.
This is analogous to the "fixed-sum game" of dividing a cake among n players where the nth player does the dividing and the n-1 players do the choosing. The nth player is sure to be left with the smallest slice. He will try to make it as big as possible, i.e. divide the cake into equal slices. This is his dominant strategy. If the n-1 players are blindfolded, n has no dominant strategy.
With people knowing no more than that every lot has some non-zero probability of being drawn and all the lots together have a probability of 1 (i.e. one, and only one, of the lots is sure to be drawn), any further logical inference being "discounted" (which is how Rawls expects his parties to reason) it is hard to see what will make their choice determinate, let alone unanimous. The plausible hypothesis seems to be that they will behave like particles in quantum mechanics, and never (short of eternity) reach agreement on a social contract.
If they were allowed to grasp a less inchoate conception of probabilities, e.g. if they could apply the principle of insufficient reason and suppose that failing any indication to the contrary, they were as likely to draw one lot as another, they would have a better chance of reaching agreement on a distribution—which would presumably be more inegalitarian than the one ruled by the maximin "strategy."
Unlike poker or business where a previous loss tends to worsen present chances, certain other risky choices may not be adversely affected. For instance, a low lifetime stipend may not worsen the odds against marrying the right person or having good children.
The very question whether Swiss families are happier than Russian ones is fatuous, although the person who has agreed to draw lots for a place in Russian society does not get a second chance to draw lots for a place in Swiss society.
The prudent man's finding that risk-taking is difficult, especially if it is a risk of losing your stake, is not unlike Sam Goldwyn's celebrated profundity that forecasting is difficult, especially if it is about the future.
"Refusing to gamble" is itself a gamble, and "not making forecasts" is a particular forecast as long as it is unavoidable for today's future to become tomorrow's present. You do not avoid exposure to it by not adjusting to what it might or might not be like. Your adjustment may not be successful. Not adjusting is even less likely to be successful.
Anyone who has had his investments handled by a bank trust department is probably familiar with the phenomenon of "managing wisely but not well." Anyone who has observed the functioning of financial markets dominated by institutions rather than by principals, knows what it means that paid portfolio managers "do not want to be heroes" and "do not stick their necks out," buying when everybody else is buying and selling when everybody else is selling.
If parents thought that children were going to grow up less able, less provident and less resilient than themselves, they might consider that a welfare state would be genuinely better for them than an inegalitarian state. The parents might then want to install it straightaway, either because they could not trust their children to recognize their best interests, or because the choice of state had to be made right now for all posterity. However, Rawls does not use this line of paternalistic argument.
End of Notes
Return to top