The Crisis in Economic Theory

The publication of these two books is a landmark in the development of economic theory. Singly and jointly, they represent a fundamental challenge to the reigning neoclassical orthodoxy. The more sophisticated practitioners of that theory have long recognised that it is in deep trouble, but have stuck to it because of the lack of a viable alternative, on the principle that you can’t beat something with nothing. Whatever objections one may have to neo-classical economics, it certainly is something: a highly-developed and formalised body of thought which has been applied to a wide range of practical and theoretical issues. One would hesitate before saying that the work of Marx, Schumpeter or Herbert Simon was ‘nothing’, and the ‘sophisticated practitioners’ referred to above would certainly not make any such judgment. Yet they would tend to say that their writing, though not lacking in insights, is amorphous, their ideas unformalised and unformalisable. They might well agree with Lord Robbins, who is reported to have said of Schumpeter’s Capitalism, Socialism and Democracy that it was a piece of ‘supremely intelligent after-dinner talk’. In any case, they would insist on the global and general character of these theories, and the absence of testable hypotheses to be derived from them.

The two books under review make the most important contributions for decades towards the construction of plausible alternatives to the neo-classical mode of theorising. Nelson and Winter offer an ‘evolutionary’ alternative which derives in equal amounts from Schumpeter and Simon. From the first they take the idea that competition is a process that involves winners and losers, not just a feature of markets which have so many firms that each must take the prices as given. From the second they take the idea of bounded rationality or ‘satisficing’, a neologism coined to provide an alternative to the postulate of maximisation at the core of neo-classical theory. Roemer offers a different heterodoxy, by focusing on the Marxist notions of class and exploitation. To understand these non-neo-classical phenomena he uses standard neo-classical tools, no doubt to the surprise of many and the dismay of some of his Marxist readers. The two departures from orthodoxy therefore have nothing in common with each other. Nelson and Winter affirm that they have not found the Marxist notion of class useful, while Roemer draws extensively upon the theoretical tools to the demolition of which their book is devoted.

Here is a first approximation of what the textbook orthodoxy looks like. A market economy is made up of firms and households. Firms produce outputs using inputs produced by other firms and labour offered by households. Households buy goods with the payment for labour services offered and with income derived from shares in firms. Firms and households make choices, the former guided by profit maximisation, the latter by their preferences. Firms must decide how much they want to produce and how to produce it. With respect to the first question, it is assumed that the firm is so small that it is unable to affect the prices of the product it makes or of the input it buys: it has no monopoly power. With respect to the second, it is assumed that the firm faces a well-defined set of production possibilities, i.e. distinct input combinations that will give the same output. Among these the firm chooses the input combination with the lowest total costs. Households must decide how much labour to offer and how much to buy of which goods, within the constraints set by their income and the ruling prices. The central theorem of neo-classical economics states that given these assumptions, and other more technical postulates, there exists a set of prices that will clear all markets, i.e. a set of prices that will induce the consumers to buy exactly the amount of each good that producers are induced to make. A second theorem, almost equally important, states that if firms and consumers only interact via the market mechanism, the equilibrium will also be an optimum, in the sense that it would not be possible to improve the situation of one household without making another worse off. A third theorem states that any optimum can be achieved in this way, by suitable redistribution of shares in firms.

Actual applications of this abstract theory deviate from it in a number of ways, either by imposing more structure on the economic relations or by modifying some of the assumptions. For instance, by specifying the nature of the production possibilities (‘the production function’) one can derive that workers will be paid according to their marginal product. Some authors refer to payment below marginal product as ‘exploitation’, although this is not the usual neo-classical approach to normative issues. The standard way of handling normative questions of distribution is, rather, to assume the existence of a ‘social welfare function’ which is determined by the political system. Another modification is to give up the ambition of understanding how everything is connected with everything else, as in general equilibrium theory. Thus macro-economics considers aggregate entities such as industries or even the economy as a whole, rather than innumerable small firms. Partial equilibrium situations take into account the behaviour of a few firms considered as price-setters rather than passive price-takers. One may also introduce uncertainty, in the form of probability distributions entertained by the firm, and time, in the form of intertemporal markets. Technical change may be incorporated by multiplying the production function by some constant and letting that constant increase over time. The state is introduced to resolve the market failures that may arise when firms and consumers interact outside the market – for example, when firms create pollution for each other or for the consumers.

By these and numerous other adaptations neo-classical theory has evolved into an incredibly elaborate, increasingly rigorous, ever-expanding construction. Yet the whole edifice rests on shaky foundations. As Nelson and Winter put it, ‘orthodoxy builds a rococo logical palace on loose empirical sand.’ Their basic objection is simple: orthodoxy neglects or misconceives the role of information in economic life. This refers not only to the fact that information-gathering is costly, and that there is a trade-off between collecting information and using it. It refers more fundamentally to the fact that we do not know what that trade-off is. Leif Johansen, whose recent death was a great loss to the economic profession, once offered the following illustration: ‘It is like going into a big forest to pick mushrooms. One may explore the possibilities in one limited region, but at some point one must stop the exploration and start picking because further exploration as to the possibilities of finding more and better mushrooms would defeat the purpose of the outing. One must decide to stop on an intuitive basis, i.e. without actually investigating whether further exploration would have yielded better results.’ In other words, there is a process of search that stops when one has found something that is good enough, or ‘satisfactory’, without it being assumed that what has been found is in any way ‘optimal’.

From this and other considerations Nelson and Winter argue that the behaviour of firms cannot be explained as a process of optimal adjustment to market conditions. Instead, they see firms as adopting a set of fixed routines that remain in operation as long as they do not give results deemed ‘unsatisfactory’. This applies immediately to the ‘choice’ of technique. Nelson and Winter deny that firms have costless access to techniques other than the one they are currently using. To acquire knowledge about other techniques is costly, and the firm will not do so without some inducement. That inducement cannot be the prospect of profit, since one cannot know whether the other techniques will in the event prove profitable. It is adversity that stimulates the search for new techniques. Necessity is the mother not only of invention, but also of what is usually called ‘substitution’ – i.e. the switch from one set of inputs to another within the same ‘production function’.

The argument also applies to higher-level routines. If, for instance, the routine technique leads to unsatisfactory performance, then a routine for searching for new techniques is switched on. This will be a rule of thumb telling the firm to spend some time looking at what other firms are doing and some time looking for fresh techniques that are not too distant from what it is currently doing. There is no guarantee that this search rule is optimal, i.e. that the imitation-innovation mix is the one which is associated with the greatest expected profits. Yet if the rule has performed reasonably well in the past, the firm will use it to solve its current problem. Nelson and Winter are of course well aware of the fact that in large modern corporations innovation is carried on as a matter of routine, i.e. that it does not need adversity to trigger it off. Here the argument would be that the level of innovative activity, i.e. the budget for research and development within the firm, is determined by satisficing rather than maximising, as is the allocation of this budget between imitation and innovation.

The orthodox, ‘Chicago’ answer to this line of argument goes as follows. The assumption that firms are capable of maximising may be unrealistic. It may well be that firms actually make their decisions by following routines or rules of thumb which reflect internal compromises or historical accidents rather than a systematic effort at profit-maximising. Yet for predictive purposes this does not matter, since the competitive market will eliminate all firms other than those which happen to be using a profit-maximising routine. In this analogy with natural selection, firms correspond to organisms and routines to genes. The optimal routines spread among the population of firms thanks to the expansion of the successful firms, which also take over or are imitated by the unsuccessful ones. Whether or not the firms have maximising intentions, they will tend to exhibit maximising behaviour, which is what matters for most purposes.

As the title of their book indicates, Nelson and Winter go along with this evolutionary approach, but they use it mainly to provide an alternative to, rather than a justification for, the neo-classical theory. Although they believe that processes of search and selection are of the essence of economic development, they deny that the selection of successful firms by the market tends to produce optimal behaviour. Again, the basic argument is very simple. A routine is never optimal in itself. It can only be optimal relative to a given state of the economic environment. If that environment undergoes rapid change, the optimality criteria may change accordingly. A firm that uses a routine that is optimal according to the current criterion, and in consequence is favoured by the market, may see its fortunes decline when the environment changes in such a way as to produce new criteria. We are, in short, dealing with adaptation to a moving target. True, this holds equally for natural selection, which also operates in a constantly changing environment. Yet here there are reasons for thinking that the speed of the adjustment process (the rate of mutations and of their selection) is so great as to permit the organisms to track the slowly changing environment. In economic and social life, this does not appear to be the case.

The Nelson-Winter attack on optimality is therefore a two-pronged one. The argument from satisficing is that firms cannot optimise ex ante, since they do not have and cannot get the information that would be required. Specifically, they would need an optimal amount of information, but this leads to a new optimisation problem and hence into an infinite regress. On the other hand, we cannot expect firms to maximise ex post, since the elimination of the unfit does not operate with the same speed and accuracy as it does in natural selection. Taken together, these two arguments strike at the root of neo-classical orthodoxy.

Of the two arguments, the second is clearly an empirical one, turning on the relative speeds of the moving target and the process of tracking it. To advance from the hypothesis that the adaptation may be unable to catch up with the target, to the positive statement that it will be unable to do so, Nelson and Winter develop a series of simulation experiments which, from the methodological point of view, form the most striking part of their work. A typical experiment might be summarised as follows. A given simulation run is characterised by the presence of two sorts of firm, imitators and innovators. Both engage in research and development, but the former spend their funds looking at what other firms are doing while the latter track the ‘latent productivity growth’ yielded by scientific progress. In both cases the return on investment depends to some degree on chance, hence each run has to be carried out many times to ensure the robustness of the results. Firms that happen to strike it lucky obtain resources with which to expand. The extent to which they actually expand depends on such factors as the ease of external financing, the degree of aggressiveness in the industry etc. These factors are allowed to vary between runs, so that one can see how the relative success of imitators and innovators is shaped by their environment – and by chance. What is observed is that the fortunes of individual firms wax and wane. Firms do not stay happy and prosperous for ever. Moreover, in different versions of a given run, the same firm may follow different trajectories, as a result of the stochastic element. Yet in this incessant flux some moderately stable patterns can be seen to emerge. For example, the proportion of imitating and innovating firms settles down to a fairly constant value. Economic forces constrain the population of firms, not the individual firm. Conceptually, this means that the neo-classical concept of equilibrium is replaced by the more realistic concept of a steady state.

These are, I believe, the central elements of the Nelson-Winter argument, but the book ranges from subtle theoretical analyses of the nature of choice to highly explicit mathematical modelling, from the theory of the firm to the theory of bureaucratic agencies. It is very engagingly written, and conveys extremely well the dilemma that must haunt any social scientist worth his salt: the necessity of choosing between realism and simplicity as guides to theory construction. If it has a weakness, it is mainly an expository one. Approximately half the book is a slightly revised version of a number of articles written over the last decade or so. The corresponding chapters are not well integrated with one another, nor with the less technical half of the work. This is not to say that the book lacks unity, since the general framework sketched above is present throughout. It does, however, lack a sense of progression, in that the chapters do not follow logically upon each other. Yet, accepting the authors’ own argument that the best may be the enemy of the good, we should be grateful for what they have offered us, which is certainly ‘good enough’.

John Roemer’s book addresses an entirely different set of questions. It takes a long, hard look at some classical concepts in Marxism and, in doing so, transforms them completely. Before I go on to explain and discuss his views, I should mention that I am hardly a neutral reader, since I am both thanked in the preface and cited on the cover. This ought to make me lean over backwards to find flaws in the book, but the worst I can say is that it is in important ways incomplete.

The concept of exploitation in Marxism serves two different functions. First, it points to one of the two main reasons for criticising capitalism, the other being capitalism’s tendency to inhibit the free development of the individual’s creative powers. Secondly, it enters into an explanation of the class struggle, the intuition being that the exploited tend to organise themselves against the exploiters. A central concern of Roemer’s book is to bridge the gap between these two aspects of exploitation, i.e. to show that the income distribution which on normative grounds can be said to be exploitative is also one that will motivate people to collective action.

According to the traditional Marxist concept of exploitation, a person is exploited if he works more hours than the labour time embodied in the goods he can buy for his income. This concept of exploitation is the topic of the first two-thirds of Roemer’s book: in the last third he moves on to a more general theory which is intended to be valid even in conditions under which the traditional notion is ill-defined. It should be said that even when labour values are well defined, it would be very difficult to calculate them and hence very difficult to draw the exact dividing line between the exploited and the exploiters. If it could be shown, however, that exploitation is systematically correlated with some observable feature which will in fact tend to mobilise people into collective action, then the gap will have been bridged. This is exactly what Roemer does. He shows that exploitation status is systematically correlated with class membership, and notably that an individual who is forced to sell his labour-power is also exploited. There are those who will think this is obvious, but Roemer also shows that the converse implication, which might appear equally obvious, does not hold.

Roemer’s theory is set within the framework of general equilibrium theory. Individuals start out with different endowments of the non-labour means of production and equal amounts of labour-power. They are assumed to have goals, which in one set of models are to minimise labour time subject to a subsistence constraint, and in another set to maximise income subject to a constraint on the length of the working day. To achieve these goals, they engage in market transactions. In equilibrium – i.e. under the set of prices that allows all markets to be cleared – some agents will find that they work more hours than are embodied in what they consume (according to the subsistence models) or in the consumption goods they could buy with their income (according to the other models). They form the set of exploited agents: the set of exploiting agents is defined in an analogous way.

The models are further distinguished by the kind of market transactions they allow. If this distinction is combined with a distinction between the two goals which agents can be assumed to have, we get four different ways of looking at exploitation. In the first model, agents are supposed to minimise labour time by engaging in exchange of non-labour commodities. In other words, there is a commodity market, but neither a labour market nor a credit market. By definition such an economy cannot contain classes, which are defined by the buying and selling of labour-power, yet there can be exploitation. This is an interesting result, since it disproves the frequently stated view that ‘exploitation takes place at the point of production,’ substituting for it the idea that exploitation occurs at the point of exchange. True, the possibility of exploitation without class is, with one exception, mainly a logical one. In the real world it would soon enough bring about class formation. The exception is international trade, which may bring about exploitation by ‘unequal exchange’ even if there is no employment relation.

In the second model we retain the assumption of working for subsistence, but add a labour market. This allows Roemer to prove his central Class-Exploitation Correspondence Theorem, to the effect, broadly speaking, that those who hire others to work for them are exploiters while those they hire are exploited. To speak less broadly, we must look at how Roemer defines classes. The members of a class are characterised by the mix of activities they will perform in equilibrium – i.e. when they do as well as they can at the equilibrium prices. There are three types of activity: working for oneself, selling one’s labour-power and hiring others to work for oneself. A big capitalist is someone who when optimising only hires other people; a small capitalist one who must work for himself and also hire others; a petty bourgeois one who only works for himself; a semiproletarian someone who when optimising works part of the time for himself and part of the time for others; and a proletarian one who must sell all his labour-power. The Class-Exploitation Correspondence Theorem says that all members of the first two classes are exploiters, all members of the last two exploited, while the petty bourgeoisie may contain exploiters as well as exploited agents.

The third model replaces the labour market with a credit market, and shows that the resulting economies are essentially identical, thus confirming the neo-classical dictum that it doesn’t matter whether capital hires labour or labour hires capital. This also provides an additional demonstration of the fact that there can be exploitation without an employment relation.

The fourth model reinstates the labour market but assumes that the goal of the agents is accumulation rather than subsistence. This, clearly, is the canonical model for capitalist exploitation and class formation. Exploitation now has to be defined somewhat differently, in terms of the labour content embodied in the goods that agents could buy with their income and not in terms of what is embodied in the given subsistence bundle of goods. Since the agents can use their income to buy many different bundles of goods, which need not add up to the same total labour content even if they add up to the same price, there is an indeterminacy in the concept of exploitation, leading to a ‘grey area’ of agents who are neither exploited nor exploiters. These are the agents whose income would allow them to buy some bundles embodying more labour than they perform, and other bundles embodying less. The class structure also is somewhat different, in that the class of non-working capitalists disappears. The Class-Exploitation Correspondence Theorem now says that anyone who must hire labour to optimise is an exploiter and that anyone who must sell his labour-power is exploited. The grey area is included in the petty bourgeoisie, which also contains some exploiting and some exploited agents.

An important feature of these models is that classes are defined modally, by what they must do rather than by what they actually do. A capitalist is someone who must hire labour to optimise, even if for some reason he chooses not to optimise. A second-generation Rockefeller does not become a worker by taking a manual job in one of the family factories: he remains a capitalist as long as he does not also give away his fortune and inheritance rights. Nor does the self-proletarianised student turn into a worker by virtue of the fact that he or she could do better by becoming, say, a self-employed doctor or lawyer. This feature of the definition makes good sociological sense. We want the notion of class to be useful in the explanation of social conflict and collective action, and there are good reasons for thinking that solidarity exists between those who must sell their labour-power and does not extend to those who for some reason do so without having to.

The conceptual – and political – interest of these models is beyond doubt. True, they are highly abstract, and might seem somewhat irrelevant to actual capitalist economies, characterised by oligopolies, cartels, collective bargaining and massive state intervention. Yet I believe that in order to understand exploitation in these more complex settings, it is a useful if not indispensable starting point to see how it emerges in the simple competitive models. A more serious objection is that Roemer does not fully address the issue of what is wrong about exploitation, and the possibility that even a socialist society might find exploitation not only necessary but unobjectionable.

Consider first ‘the Nozick problem’. In a hypothetical economy in which initially all endowments are equally distributed, some of the agents might decide not to use all their income on consumption goods, but to use part of it to buy means of production. On what grounds could one then stop them from buying the labour-power of others who preferred to spend their whole income on consumption? Should the socialist police step in and block such ‘capitalistic acts between consenting adults’? Take the following, from a novel by Vasili Grossman, quoted by Alec Nove in The Economics of Feasible Socialism:

I wanted since childhood to open a shop, so that any folk could come in and buy. Along with it would be a snack-bar, so that the customers could have a bit of roast meat, if they like, or a drink. I would serve them cheap, too. I’d let them have real village food. Baked potato! Bacon-fat with garlic! Sauerkraut! I’d give them bone-marrow as a starter ... A measure of vodka, a marrow-bone, and black bread of course, and salt. Leather chairs, so that lice don’t breed. The customer could sit and rest and be served. If I were to say all this loud, I’d have been sent straight to Siberia. And yet, say I, what harm would I do to people?*

The customers would, indeed, benefit, and so would those who would rather be waiters or cooks than factory workers. The cooks and waiters might well be exploited, in the technical sense of the term, but who cares? If all the parties concerned prefer exploitation to its absence, how could there be anything wrong in it? True, there are many circumstances in which one might invoke false consciousness, asymmetries of information or lack of organisation among the exploited to argue that they do not really prefer exploitation, but it would be pure wishful thinking to argue that one or more of these conditions will always obtain when people freely enter into an exploitative arrangement. At best one could argue that people would not allow the inequalities thus created to become cumulative.

The Nozick problem arises in socialism. It would be wildly off the point to say of capitalist economies that most capitalists have bought their means of production with income derived from their own labour. Yet even in capitalism one might want to deny that exploitation is in any way morally wrong. In order to do so one would have to abandon the assumption that all people have the same labour endowments, and assume that some individuals are innately more gifted for, say, managerial functions. Although this makes it impossible to define exploitation as I have done, since there is no homogeneous labour that can serve as a measuring rod for the labour content of goods, we may still talk of exploitation in a looser sense, if some people earn more than others without putting in more effort or having special needs that justify their getting more. Yet one might argue that this ‘exploitation’ is mutually beneficial, and morally quite unobjectionable.

Taking our cue from Marx, we may ask whether the capitalist is robbing the worker and, if so, whether there is anything wrong in his doing so. Marx argued that although the capitalist does rob the worker, he also ‘forces the production of surplus-value and thus helps create what is to be deducted’. In other words, if the capitalist manager was not there to organise production, there would be nobody who could steal the surplus, but nor would there be any surplus to steal. Hence the ‘incentive problem’ which is at the core of the Rawlsian theory of justice: how can it be wrong to reward someone for a task that he would not have performed in the absence of a promise of a reward? If the workers gain from being exploited, by getting a part of the surplus which is made possible by the managerial talent of the capitalist, how can one complain about the capitalist appropriating the rest of the surplus?

Unlike the Nozick problem, the incentive problem allows for a fairly easy answer. The objection I have just set out rests on the idea that ‘ought implies can’: if something is impossible, one cannot promote it as just and use it as a reference point to condemn something else as being unjust. And, so the objection goes, since it is not possible to have the managerial function without the reward to managers, the desired alternative of high productivity without exploitation simply is not available. According to this argument, the actual state of high productivity with exploitation should instead be compared to the alternative of a non-exploitative economy with low productivity, which nobody would rationally prefer. The objection fails because it is not appropriate to read ‘can’ in the sense of ‘historically feasible’. Some of the more utopian aspects of Marxism should be rejected on the grounds that they are biologically impossible: it is pointless to condemn capitalism on the grounds that not everybody is a Leonardo da Vinci. The problem of managerial effort, however, is a quite different one. I do not know any plausible argument for the view that the need for material incentives to call forth the exercise of skills is an ineradicable part of human nature. On the contrary, I believe that Marx was right in his view that the use of skills is so rewarding in itself that incentives are not required. The fact is not that a reward is necessary to elicit these skills but that those who possess them can hold society to ransom by threatening to withhold their skills unless they are highly rewarded. Of course, there are exceptions, but by and large I believe this to be true. And if it is, those who invoke the incentive problem to justify inequalities do so in bad faith, at least if they argue that the problem will persist for ever, not just in a transition period.

These last remarks relate to questions which are central to the last third of Roemer’s book, where he presents his ‘general theory’ of exploitation. It is general in the sense that it is well-defined even when the notion of labour content is not, thus freeing the theory of exploitation from its dependence on the labour theory of value. It is also general in the sense of not offering any specific mechanisms to explain the emergence of exploitation. To my mind, the inconveniences of generality in the latter sense offset any advantages to be had from generality in the first sense. The theory is general, but at a high cost. Also, it is not a general theory of exploitation and class, since the framework used in this part of the book does not allow us to define classes and relate them to exploitation.

The general notion of exploitation is that a group is exploited if it would be better off were it to withdraw from society with some part of the (tangible and intangible) means of production. Exactly which part is specified in a set of withdrawal rules, the distinction between the various possible rules corresponding to a distinction between different forms of exploitation. Feudal exploitation is defined by stipulating that the agents are allowed to withdraw with their own means of production. In other words, agents are feudally exploited if they could do better on their own with what they have, hence force must be used to keep them from doing so. Roemer points out that the neo-classical notion of exploitation, briefly mentioned above, is in fact equivalent to feudal exploitation. Thus when neo-classical theorists state that under competitive conditions there is no exploitation in capitalism, what they are really asserting is the absence of feudal exploitation.

Capitalist exploitation occurs when the agents would be better off were they to withdraw with their per capita share of society’s tangible means of production. It is not obvious that there is capitalist exploitation in capitalism. If the workers were to withdraw with their per capita share of the tangible means of production, they might find that they were worse off because of the lack of (intangible) managerial skills. This, in Roemer’s terminology, would be a case of socialist exploitation. A group of agents is socialistically exploited if it would be better off were it to withdraw with its per capita share of the tangible and intangible means of production, i.e. both machinery and skills. He refers to this as socialist exploitation because he believes that it is a widespread phenomenon in the ‘socialist’ countries, although he is also aware of the fact that many of the inequalities in these countries might be due to ‘status exploitation’, a phenomenon which is closely related to feudal exploitation.

For various reasons I am less impressed by this model of exploitation than by the various models based on market exchange. For one thing, I believe that the general approach is inherently ill-suited to serve as a definition of exploitation. It is part and parcel of our intuitive notions about exploitation that it involves some form of interaction between exploiters and exploited, and thus some form of causal relation. We know, however, that the notion of causality cannot be exhaustively captured by statements about counterfactuals, by assertions about the fate of the workers under the various, hypothetical withdrawal conditions. The truth of ‘If A had not occurred, B would not have occurred’ is neither a sufficient nor a necessary condition for the truth of ‘A caused B,’ as is shown by the phenomena of preemptive causation and correlation. This does not exclude the possibility that counterfactual statements may be a useful characterisation of many, perhaps most important cases of exploitation, but they cannot serve as a definition in the sense of giving necessary and sufficient conditions. For another thing, the model suffers from an inherent vagueness, in that it has too little structure to help us understand how exploitation emerges and who is exploited by whom. In particular, I was not convinced by Roemer’s attempt to link up the theory of exploitation with historical materialism. The link would presumably consist in a demonstration that forms of exploitation rise and fall according to their tendency to promote or fetter the growth of the productive forces, but no argument is offered for any such connection.

This being said, the model, by highlighting the many different sources of social inequality, does offer a very helpful perspective in which we can think about the problem of distributive justice. Feudal exploitation occurs because men do not own their labour power, socialist exploitation because they do. Capitalist exploitation occurs because some people only own their labour-power. I believe these insights will have to be incorporated into any general Marxist theory of justice, although I have only the vaguest idea of what such a theory would look like.

Let me end by asking the obvious question: could the two heterodoxies under review somehow be integrated with each other? Nelson and Winter devote two chapters to normative analysis and policy discussion. The result is often interesting and illuminating, but also confirms their own view that ‘a normative theory consistent with an evolutionary approach to positive theory almost certainly will be complex and messy.’ I for one would argue that parts of Marxism are already closer to the satisficing than to the maximising framework, notably with respect to the theory of technical change, and that one might also consider an extension of the satisficing point of view to the analysis of exploitation. Again, I have no idea of what this would look like, except that it would be likely to be quite messy. That in itself ought not to be an objection. If the evolutionary framework of satisficing, induced search and dynamic selection is a viable and coherent alternative to neo-classical thinking, we should be happy to sacrifice the artificial elegance of orthodoxy for a more earthy realism. The question is whether the alternative is fully viable. Almost against my will, I tend to think it is not. Satisficing is more than ‘nothing’: it is almost ‘something’ – but not enough to beat neo-classical theory into retreat. At the core of satisficing there is an element of ad hoc-ness from which at present I see no means of escaping. It is created by the fact that no explanation is forthcoming of why people have the aspiration levels they have – why some people spend ten minutes looking for a good place for mushrooms and others spend two hours. The level at which different people ‘satisfice’ remains unexplained. Hence the theory will tend to lack predictive power, in spite of its obvious descriptive adequacy.

I am not saying that this is an inherent flaw that will never be overcome. We may hope that the work of the satisficing school of theorists will be integrated with the work of cognitive psychologists, to produce a simple and robust explanation of aspiration levels. Meanwhile, however, orthodoxy reigns, not because of factual support, but because of the weakness of the contenders. In this respect, Roemer’s work is more subversive, since it uses the analytical tools of neo-classical economics to undermine its normative and sociological assumptions. Instead of firms and households, Roemer offers us classes; instead of the ‘social welfare function’ he proposes exploitation as the criterion of justice. His book, in my opinion, is a step towards realism without loss of rigour. At present, the satisficing school offers realism at the expense of rigour. If and when that weakness is overcome, the joint impact of the two challenges to orthodoxy may well prove fatal.