The History of Macroeconomics Viewed Against the Background of the Marshall-Walras Divide

Transcription

1 DRAFT The History of Macroeconomics Viewed Against the Background of the Marshall-Walras Divide Michel De Vroey Université catholique de Louvain IRES 3, Place Montesquieu 1348 Louvain-la-neuve Belgium May 2003 A paper presented at the April HOPE Conference on The IS-LM Model: Its Rise, Fall and Strange Persistence at Duke University

2 1 1. Introduction The demise of the IS-LM model and its replacement by stochastic intertemporal equilibrium models is probably the most salient change that occurred in macroeconomics during the last quarter of the twentieth century, a shift that many commentators have viewed as a Kuhnian scientific revolution. My aim in this paper is to analyse it against the background of what I call the Marshall-Walras divide. According to the latter, the contributions of Marshall and Walras to economic theory are less complementary than usually believed. Instead, the Marshallian and the Walrasian approaches are considered alternative research programmes. This, I will claim, has a direct bearing on the unfolding of macroeconomics: the change that has taken place is the invention of Walrasian macroeconomics and its substitution to Marshallian macroeconomics. This claim is not entirely original. It was voiced by some authors who studied what at the time was called Monetarism Mark I and Monetarism Mark II, associated respectively with Friedman s and Lucas s Phillips curve models (Friedman 1968, Lucas 1972). 1 The specific contribution of my paper is twofold. First, it gives a more systematic study of the Marshall- Walras difference and, second, it covers a scope that is wider than the mere confrontation of Friedman s and Lucas s standpoints. 2 In section 2, I survey a few papers aiming at assessing the present state of macroeconomics. Preliminary clarification tasks are undertaken in the next three sections. They bear on the meaning and scope of macroeconomics, the Keynesian label, and the contents of the neoclassical synthesis. In section 6, I justify my claim as to the Marshall-Walras divide. Finally, in section 7, I use it as a clue for explaining the recent developments in macroeconomics. 2. A survey of recent critical assessments My starting point is Blanchard s and Woodford s surveys of the history of macroeconomics written on the occasion of the turn of the century. Blanchard (2000) divides the history of macroeconomics in three stages: pre 1940, a period of exploration, from 1940 to 1980, a period of consolidation, and since 1980, a new period of consolidation. He views the passage from the IS-LM model to real business cycle models as a smooth evolutionary process. One stage within it is the dynamization of the static IS-LM model by authors such as Modigliani, Friedman, and Tobin who were able to recast some of 1 For example, Hoover noted that Friedman, as one important monetarist, differs from the new classicals on a fundamental point of methodology: he is a Marshallian; they are Walrasians [(1984) 1990, 528].

3 2 its components consumption, investment and financial decisions within an expectations perspective. A next step is the introduction of rational expectations. The last step consists of the consideration of imperfections which many macroeconomist saw as central to an explanation of macroeconomic fluctuations (2000: 11). Thus, Blanchard hardly pictures the evolution of macroeconomics in terms of scientific revolution. On the surface, the history of macroeconomics appears as a series of battles, revolutions and counterrevolutions. But this would be the wrong image. The right one is of a surprisingly steady accumulation of knowledge (Blanchard 2000: 2). He just admits to the existence of a temporary crisis by the end of the 1970s. At the time, the introduction of the rational expectations seemed to mark a breach, yet this assumption soon turned out to be acceptable by all economists, so that the early impression proved false. 3 Any divide between New Keynesians and New Classicists was short-lived. In the early 1980s, macroeconomic research seemed divided in two camps, with sharp ideological and methodological differences. Real business cycle theorists argued that fluctuations could be explained in a fully competitive model, with technological shocks. New Keynesians argued that imperfections were of the essence. Real business cycle theorists used fully specified general equilibrium models based on equilibrium and optimisation under uncertainty. New Keynesians used small models capturing what they saw as the essence of their arguments, without the paraphernalia of fully specified models. Today, the ideological divide is gone. Not in the sense that underlying ideological differences are gone, but in the sense that trying to organise recent contributions along ideological lines would not work well. As I argued earlier, most macroeconomic research today focuses on the macroeconomic implications of some imperfection or another. At the frontier of macro-economic research, the field is surprisingly a-ideological (Blanchard 2000: 39). Contrary to Blanchard, Woodford (1999) finds it useful to structure his account on the idea of a succession of revolutions and counter-revolutions. He starts with a few reflections on the birth of macroeconomics (i.e. the study of business fluctuations in the early decades of the 20 th century) to move on with the Keynesian Revolution and the neoclassical synthesis. After a section on the great inflation and the crisis of Keynesian economics, he goes on studying the 2 On the latter theme, see De Vroey (2001). 3 At the time the introduction of rational expectations was perceived as an attack on the received body of macroeconomics. But, with the benefit of hindsight, it feels much less like a revolution that like a natural evolution (Blanchard 2000: 10).

4 3 three successive waves of attacks that were launched against Keynesian theory: monetarism, rational expectations and the new classical economics, and, finally, real business cycle theory. In his last section, he ponders the emergence of the new neoclassical synthesis. Three of Woodford s many sharp remarks are worth evoking. The first is his assessment of monetarism. In his opinion, the latter has won the day in bringing monetary policy and expectations to the forefront. Yet he considers that, from a methodological viewpoint, it has been absorbed within the Keynesian paradigm. «The Keynesian model of the 1970s came to incorporate the most important of the monetarist insights, thus achieving a new synthesis» (Woodford 1999: 18). The second point concerns the relationship between new classical models à la Lucas and real business cycle models. The latter were built up using the intertemporal Walrasian generalequilibrium methodology of the former. However, an important methodological difference exists between them, Woodford notes, in that the former are parables with no claim to direct empirical confrontation while the latter claim to explain the real world The real business cycle literature offered a new methodology, both for theoretical analysis and for empirical testing. It showed how such models [of the Lucas type] could be made quantitative, emphasising the assignment of realistic numerical parameters values and the computation of numerical solutions to the equations of the model, rather than being content with merely qualitative conclusions derived from more general assumptions. The equilibrium business cycle models of Lucas had really only been parables; they could not be regarded as literal descriptions of an economy, even allowing for the sort of idealisation that all models of reality have. Real business cycle models are instead quantitative models, that are intended to be taken seriously as literal depictions of the economy, even if many details are abstracted from. The literature emphasises the numerical predictions of the models, when parameter values are assigned on the basis of measurement of the relevant aspects of an actual economy (Woodford 1999 : 25-26). Finally, as far as the future prospects of macroeconomics are concerned, while laudatory about the real business cycle approach, Woodford is nonetheless of the opinion that its impact has not been decisive and that the Keynesian viewpoint is again taking the upper hand. Rejoining Blanchard, he believes that the future of macroeconomics lies in incorporating Keynesian features into real business cycle models (Woodford 1999 : 29). The same point is to be found, in an even more overt way, in Goodfriend and King (1997), a paper aiming at making the case that in the last ten years «macroeconomics is moving toward

5 4 a New Neoclassical Synthesis» (1997 : 231). To this end, they compare four episodes of the development of macroeconomics, the neoclassical synthesis, monetarism and rational expectations, real business cycles and, finally, the new neoclassical synthesis. Like Blanchard and Woodford, Goodfriend and King praise the real business cycle programme for its micro-foundations and its allowing a comparison of alternative policies on the basis of measures of utility benefits or costs. Like them, they argue that it must be enriched with the considerations of imperfections, the very purpose of new neoclassical synthesis models. Under this heading, they put models that «range from the flexible, small models of academic research to the new rational-expectations policy model of the Federal Reserve Board» (1997 : 232). These models, which Goodfriend and King view as close in spirit to the neoclassical synthesis that prevailed in the 1960s, combine Keynesian elements, especially sticky prices and imperfect competition, and real business cycle elements intertemporal optimisation, rational expectations, and their integration in stochastic dynamic model (1997 : 232). These two components, they claim, are compatible because of their shared reliance on microeconomics (1997 : 256). We call the new style of macroeconomic research the New Neoclassical Synthesis because it inherits the spirit of the old synthesis discussed in section 2. New Neoclassical Synthesis offer policy advice based on the idea that price stickiness implies that aggregate demand is a key determinant of real economic activity in the short run. New Neoclassical Synthesis models imply that monetary policy exerts a powerful influence on real activity. This has both positive and normative implications. From a positive point of view, the central conclusion is that economic fluctuations cannot be interpreted or understood independently of monetary policy. This is true notwithstanding the fact that the real business cycle model at the core of New Neoclassical Synthesis assigns a potentially large role to productivity, fiscal policy or relative price shocks. From a normative perspective the New Neoclassical Synthesis says that aggregate demand must be managed by monetary policy in order to deliver efficient macroeconomics outcomes (Goodfriend and King 1997: 255-6). 4 Let me now turn to two studies, which are more critical of the new classical approach, by Blinder and Lipsey respectively. Blinder 1988 offers a fierce criticism of new classical economics, arguing that: There was no anomaly, that the ascendancy of new classicism in academia was instead a triumph of a priori theorising over empiricism, of intellectual aesthetics over 4 Cf. Hairault (1999: 616).

6 5 observation and, in some measure, of conservative ideology over liberalism. It was not, in a word, a Kuhnian scientific revolution. If this is so, it helps explain a phenomenon that a Kuhnian would find puzzling: macroeconomics is already in the midst of another revolution which amounts to a return to Keynesianism but with a much more rigorous theoretical flavour. (Blinder 1988: 110) Like Woodford, Blinder thinks that no important conceptual issues separate IS-LM Keynesians and monetarists. 5 Drawing a distinction between the positive and normative components of Keynesianism, he considers that the difference between Keynesian and monetarists resides in the last of these two dimensions. The division of Keynesian economics into positive and normative component is central to understanding both the academic debate and its relevance to policy. Positive Keynesianism is a matter of scientific judgement. A positive Keynesian believes that both monetary and fiscal policy can change aggregate demand, that fluctuations in aggregate demand have real effects, and that prices and wages do not move rapidly to clear markets. No policy prescriptions follow from these beliefs alone. Normative Keynesians add both value judgement and political judgement to the preceding list. A normative Keynesian believes that government should use its leverage over aggregate demand to reduce the amplitude of business cycles. He or she is probably far more interested in filling in cyclical throughs than in shaving off peaks. These normative propositions are based on judgements that (a) macroeconomic fluctuations significantly reduce social welfare, (b) the government is knowledgeable and capable enough to improve upon free-market outcomes, and (c) unemployment is a more important problem than inflation. The long and to some extent continuing battle between Keynesians and monetarists, you will note, has been primarily fought over the normative issues particularly (b) and (c). Thus, by my definition, most monetarists are positive Keynesians, but not normative Keynesians. The briefer, but more intense, debate between Keynesians and new classicals had, by contrast, been fought primarily over the tenets of positive Keynesianism (Blinder : 112-3). 6 In the same vein, Lipsey (2000) admits to the existence of a new classical revolution yet regret its arising. To many Keynesians, the new classical programme replaced messy truth by precise error (2000: 76). It expresses just a change in taste. In no way, can the success of the new programme be ascribed to the failure of the old one. To him, most of the charges against the monetarist-keynesian paradigm are unjustified. 7 5 The same view is held by Snowdon and Vane, who write that monetarist influences were absorbed within the existing framework leading to a Keynesian-monetarist synthesis (Snowdon and Vane 1996: 386). 6 As to the reasons of the success new classicism, Blinder claims that they are «rooted in the sociology of science, in attachment to theory, and in ideology not in empiricism» (1988 : 117). 7 He alike views the difference between monetarists and Keynesian as being mainly rhetorical (Lipsey 2000: 68).

7 6 The papers surveyed above have in common to have been written by authors who in one way or another can be considered Keynesians. 8 Either they are frankly opposed to real business cycle models, as is the case of Blinder and Lipsey, or they believe in the possibility of introducing Keynesian features into them. In order to have a balanced survey, I should examine papers recounting the history of macroeconomics from the new classical viewpoint. However, such papers seem conspicuous by their absence, except for Lucas well-known criticisms of Keynesian theory. 9 Of course, several papers presenting the real business cycle method exist, e.g. Plosser (1989), Prescott ([1986] 1994), Prescott (1998), yet they hardly bother to refer to previous paradigms in any detail. The underlying rationale, I surmise, is that, to them, scientific economics, involving the equilibrium method, on the one hand, and adequate mathematical and measurement tools, on the other, emerged only recently that is, with Lucas. Before, at least one of these two components was lacking. 10 In their opinion, little interest is then to be gained from returning to the pre-scientific stage of knowledge. No need exists for history of economics. 3 Back to square one: What is macroeconomics? None of the authors examined devote much time on the above question. Blanchard, and Woodford concur in stating that it consists of the study of fluctuations (Blanchard 2000: 2; Woodford: 1999: 1). At present, macroeconomics deals with fluctuations; as a result, it is claimed that any study of fluctuations is macroeconomics. I disagree. On the one hand, I do not consider the early reflections on the business cycle as belonging to macroeconomics. On the other hand, if this definitional stance is taken, it turns out that the IS-LM model does not belong to macroeconomics! The micro/macro distinction is usually ascribed to Ragnar Frisch. According to Machlup ([1960] 1973: 99), he coined it in his essay Propagation Problems and Impulse Problems in Dynamic Economics (1933: 172). Machlup notes that, contrary to most later writers who have treated general equilibrium analysis as microeconomics, Ragnar Frisch is one of the minority in regarding the theory of general equilibrium as a part of macro-economics, because it is concerned with the whole economic system in its entirety ([1960] 1973: 98). However, Frisch s stance has not been taken up. The result is a messy state of affairs. What is now called macroeconomics is concerned with a whole economy, yet not every analysis 8 Other interesting pieces are Drèze (2001), Danthine (1997), Snowdon and Vane (1996), Hairault (1999). 9 Cf. Lucas ([1980] 1981, [1978] 1981, [1977] 1981), Lucas and Sargent ([1979] 1994). See also De Vroey (2003b). 10 As stated by Lucas, While Keynes and the other founders of what we now call macroeconomics were obliged to rely on Marshallian ingenuity to tease some useful dynamics out of purely static theory, the modern theorist is much better equipped to state exactly the problem he wants to study and then to study it (Lucas 1987: 2). See also his Nobel lecture (1996: 669).

8 7 bearing on such an object is usually considered macroeconomics. Take Walrasian or neo- Walrasian theory. Beyond doubt, it studies the economy as a whole, yet it is ranked as microeconomics. On the other hand, why counting real business cycles models, which claim a Walrasian or neo-walrasian heritage, as macroeconomics? This definitional puzzle should not be side-stepped. First of all, we should follow Frisch and state that macroeconomics belongs to the domain of general equilibrium since its concern is an economy as a whole instead of isolated sections of an economy. Its specificity is to be concerned with what could be called pragmatic simplified general equilibrium models. 11 Pragmatic general equilibrium models are simplified models of the economy, comprising a small amount of variables. For example, they consider just a few markets, named after those markets that are deemed to be the most important real-world markets. They deal at once with aggregates, by reasoning in terms of representative firms or agents. They embed a few institutions, such as the government and the central bank. They are geared towards addressing policy-issues, such as the level of employment, national output, inflation, government deficit, effects of changes in money supply, etc. They claim to be empirically relevant and to give results which can be tested against reality. In contrast, complex general equilibrium models, as I would characterise Walras s model or the Arrow-Debreu model, are disaggregated models, involving a great number of agents and commodities. They are geared towards more abstract purposes, like demonstrating the logical existence of equilibrium, stability, etc. In this conception, macroeconomics is concerned with mathematical models. Accepting that reasoning in prose is not a model in this sense, the arising of macroeconomics should be ascribed not to Keynes General Theory but to the subsequent models that tried to translate Keynes blurred message into a precise model. Here the IS-LM model stands out. Thus, if this definitional stance is taken, macroeconomics started with the IS-LM model. 12 Imagine next that we are allowed to redefine microeconomics and macroeconomics from scratch. The chances are small that the ensuing configuration would coincide with the existing delineation. My own attempt at such an exercise, which remains close to Frisch s original distinction, is summarised in the upper panel of table Patinkin has been a forerunner of this conception. This is also Woodford s view (1999: 8). See also Farmer (1993 : 1). 12 Several passages of Laidler s book on the Keynesian revolution (1999) hint at my viewpoint, although this is not its main thrust. Let me quote one of them. «Nevertheless, the impression of continuity in the development of macroeconomics which one might get from the very selective survey of reactions to the General Theory presented in the foregoing chapters would be misleadingly incomplete. A break did occur around 1936, and it did center on the General Theory. The break in question, however, arose not directly from that book itself, but from the IS-LM model, which a number of younger commentators found in it and used to expound not just what they took to be its main message, but that of an earlier classical tradition too.» (1999: 303). Other passages are on p. 318 and 324.

9 8 Insert Table 1 First, I separate individual and interactive equilibrium. The former designates a given agent s optimising plan of action when contemplating market participation, the latter a state where individual optimising plans have been made compatible. Next, three different types of interactive equilibrium are distinguished, according to whether their concern is an economy as a whole, an economic sector, i.e. a particular market, or an even smaller entity such as the relationship between groups of agents within a sector, e.g. a firm and its labour pool or a principal and its agents. The first can be associated with general equilibrium analysis, the second with partial equilibrium analysis, the third with game-theoretical models or the type of analysis to be found in search models. Finally, a distinction is drawn between complex and pragmatic models within the category of general equilibrium, as set out above in my characterisation of macroeconomics. Were I allowed to start from scratch, I would define microeconomics against the background of these distinctions as the study of how rational agents make optimising decisions. Thus, its exclusive concern is individual equilibrium. The individual character of the problem is important here it concerns a given decisional unit as well as the fact that of concern is agents plans rather than effective observable decisions. In turn, I would define macroeconomics as being concerned with the economy in its entirety. If this line is taken, the terms of macroeconomics and general equilibrium theory must be considered synonymous. Still, this single category must be divided into two sub-categories according as to whether complex or pragmatic models are under study. Note that macroeconomics so understood does not cover every type of interactive equilibrium. Models pertaining to interactions at a level lower than the economy as a whole, either market or infra-market interactions, are excluded from it. These could be branded as "mesoeconomics". In this perspective, micro and macro should be viewed as intertwined subject matters: macroeconomics needs be founded on microeconomics, microeconomics must lead to macroeconomics. 13 The lower panel of table 1 shows the effective usage of the two terms. Strikingly, the scope of microeconomics is wider than in the ideal taxonomy, since it now refers to all cases, except pragmatic general equilibrium. Only the latter type of models are called macroeconomics. In particular, any complex representation of the economy, as to be found in Walrasian theory, is 13 Mesoeconomics and macroeconomics would be based on the same microeconomic theory.

10 9 considered belonging to microeconomics in spite of its being concerned with the economy as a whole. In other words, general equilibrium and macroeconomics are not synonymous as above, the former designating complex general equilibrium models, the latter pragmatic ones. The reason why real business cycles models are deemed to be macroeconomics now becomes clear : they belong to the pragmatic general equilibrium category. No contradiction is involved with their Walrasian affiliation. 4 Elucidating the Keynesian modifier An interesting feature of Blinder s article is his drawing a distinction between normative and positive Keynesianism, a distinction that other authors do not bother to make. To me, it plays a central role. Let me rephrase it differently by distinguishing between a theoretical or conceptual framework (or apparatus), on the one hand, and the policy cause which it may serve, on the other. That is, a distinction must be drawn between the Keynesian apparatus and the Keynesian policy cause, the former referring to what Blinder calls positive Keynesianism, the latter to his normative Keynesianism. By Keynesian theoretical framework I shall mean the IS-LM model, thus taking the viewpoint that the IS-LM is a fair representation of the General Theory, although of course this can be discussed. But the Keynesian modifier can also be used in reference to what motivated Keynes to write his book. In this line, it can be viewed as a catchword for regrouping authors who think that, for all its virtues, the market economy can exhibit market failures. Schematically, two policy causes can be separated: the laissez-faire and the interventionist or Keynesian cause. The former pleads for the state having a modest role in the economy. In particular, it should refrain from directly intervening in the economy, especially from pursuing an activation policy. In contrast, the Keynesian cause claims that state interventions, in particular demand stimulation, are needed to remedy upon possible flaws. Its defence thus requires that market failures are demonstrated at the theoretical level. In an interview with Foley, Leontief remarked that in his opinion Keynes developed his theory essentially as an instrument to support his policy advice (1998: 122). Put crudely, my viewpoint is that this observation extends beyond Keynes person and applies not only to Keynesians but also to anti-keynesians (in the policy cause sense), be they monetarists or new classicists. Moreover, I think that there is nothing to be blamed about such an attitude. Things would remain simple if there existed a one-to-one relationship between the Keynesian framework and the Keynesian cause on the one hand, and the non-keynesian framework and the anti-keynesian cause, on the other. Yet this is not the case. While the launching of the Keynesian apparatus was motivated by the will to support the Keynesian policy cause, later on authors who were anti-keynesian from a policy perspective Milton Friedman being the

11 10 emblematic example were able to subvert it and use it as a weaponry for the laissez-faire cause. Likewise, while the non-keynesian conceptual apparatus, to be associated with new classical economics, was initiated by authors wanting to defends the anti-keynesian policy cause, this does not exclude the possibility of it becoming subverted in its turn. Put differently, a conceptual apparatus is both flexible and rigid. It is rigid in that it may exclude some phenomena as a matter of premises (market non-clearing in the Walrasian and neo- Walrasian paradigm). Nonetheless, it also features some flexibility. To still refer to this paradigm, for all its being uncongenial to market non-clearing, multiple equilibria and hence underemployment can be introduced within it, thereby offering a vindication of Keynesian policy conclusions The neoclassical synthesis Goodfriend and King claim that the unfolding of macroeconomics has led to the emergence of a New Neoclassical Synthesis close in spirit to the neoclassical synthesis that prevailed in the 1960s. It is therefore important to have a clear grasp of what the old synthesis meant. The term neoclassical synthesis is often traced back to Samuelson s 1955 edition of his Economics textbook, yet to me what he writes there is hardly instructive. 15 For a more substantive explanation, we may turn to Woodford s paper or to Howitt s or Blanchard s entries in the New Palgrave Dictionary. All of them hold the same viewpoint well summarised in the following quotation drawn from Howitt s entry: Since it was widely believed that wages were less than fully flexible in the short run, it seemed natural to see Keynesian theory as applying to short run fluctuations and general equilibrium theory as applying to long-run questions in which adjustment problems could safely be ignored. This view came to be known as the neoclassical synthesis. (Howitt, 1987 : 274). 16 The problem lies in what should be understood by the term synthesis. Strictly speaking, a synthesis refers to the process by which two theoretical analytical frameworks that at a certain stage were considered as antagonistic eventually turn out to be congruent. Hence a merger 14 A further remark to be added to what precedes is that economists are often unaware of the methodological implications of their theoretical practice. So their methodological discourse may be at the antipodes of their effective practice. Again think of Friedman. While his entire work is geared towards the defense of the laissez faire cause and, I repeat, no blame being made he is also the fierce advocate of neutrality and positivism. 15 In recent years 90 percent of American economists have stopped being Keynesian economists or anti- Keynesian economists. Instead they have worked towards a synthesis of whatever is valuable in older economics and in modern theories of income determination. The result might be called neo-classical synthesis and is accepted in its broad outlines by all but about 5 per cent of extreme left wing and right wing writers (Samuelson 1955 : 212). 16 See also Blanchard (1987: 634-5) and Woodford (1999: 10).

12 11 between them proves possible. 17 It must also be the case that the two parties to the merger admit to its effectiveness. A good example of a synthesis, so understood, is the integration of monetarism and the IS-LM model, documented in the papers of Woodford, Blinder and Blinder. 18 It seems to me that the neoclassical synthesis as understood by Howitt et alii does not stand up to this strict understanding of the notion of synthesis. Actually, their conception expresses a Keynesian viewpoint which Classicists have no reason to endorse. The classical model has as much to say about the short period as the Keynesian actually it has a rival theory to propose about it. Even if they have not, classicists should have refused the territorial partition that Keynesians were proposing. Thus, the so-called neoclassical synthesis was not a synthesis in the strict sense of the term but rather a compromise between two approaches that did not want to enter into an open intellectual fight. 6 Marshallian versus Walrasian general equilibrium In his article, Lipsey observes that: In contrast to Keynesian theories, the new [classical] programme was based on a general equilibrium model of the sort developed in the 1950s by Arrow and Debreu ( 2000: 69). Unfortunately, Lipsey does not tell his readers what he exactly means by this remark. For sure, he has in mind that the Keynesian model is not general equilibrium à la Arrow-Debreu, but the point is to see whether this imply that it is not general equilibrium at all. The latter conclusion seems incorrect in so far as it is accepted that Keynes wanted to provide an explanation of unemployment going beyond the exclusive consideration of the labour market. Does this mean that another general equilibrium approach, distinct from the Walrasian, should be considered? This is the opinion that I want to defend. The neglect of the trade technology dimension To make my point, I need to return to basics, and recall the distinction between the issues of the determination and the formation of equilibrium. 19 The first of them is concerned with the 17 The identity of the involved streams should made clear. Thus one should speak of the synthesis between theory A and B. In this respect the neoclassical synthesis appellation is wanting. 18 The well-known Laidler-Tobin debate in the Economic Journal (Laidler 1981, Tobin 1981) confirms the fulfillment of the criterion that enough representatives of the two streams must agree on the existence of a theoretical merger. 19 Sometimes this distinction is made under a different terminology. For example, in his translator's commentary in the English version of Walras' Elements, Jaffé calls emergence or 'establishment' what I suggest to calling formation : "The laws of the emergence or establishment of equilibrium prices refer to the laws of those operations of the market that result in equilibrium, whereas the laws of the determination of equilibrium price take into account 'the ultimate facts and forces which constitute that price' "(1954, p. 501).

13 12 logical existence of equilibrium. Establishing it is the task of the outside theoretician who is supposed to be omniscient about the data of the fictitious economy that is being modelled. The second issue is the problem of how the equilibrium values calculated by the theoretician can effectively be attained as the result of agents interactions. It comprises an institutional dimension related to the organisation of trade, the trade technology enabling the equilibrium outcome to be arrived at. If "a market is an institution for the consummation of transactions", as once stated by Stigler (1965: 245), this institutional dimension cannot be shelved. Yet, for different reasons, economists have gradually turned their exclusive attention to issues pertaining to the logical existence of equilibrium. Let me give an example of this methodological attitude. In 1928 Frank Ramsey solved the problem of how a benevolent planner could find the equilibrium consumption path maximising the intertemporal utility of households. Modern dynamic theory is based on the premise that this result can be extended to a competitive economy. Assuming a given economy as defined by agents and their preferences and firms and their technical constraints, it is claimed that the same equilibrium values will prevail whether this economy it is a planning or a decentralised economy. Such a claim amounts to stating that trade technology does not matter and can be side-stepped without harm. From a history of economics viewpoint, this is an odd standpoint, since one of the claims that Adam Smith wanted to defend in his Wealth of Nations was that a market system was superior to a state-regulated one. The same neglect of the trade technology dimension also explains why, with a few exceptions, Walrasian theoreticians hardly realise that what they call a competitive equilibrium is hardly based on competition (more on this below). As long as it is accepted that the exclusive issue economic theory needs to address is that of the logical existence of equilibrium, the matter is sealed. Walras paved the way to solving it and Arrow and Debreu finalised the job by constructing the benchmark general equilibrium model. No need for an alternative theoretical construction exists. To all intents and purposes most economists have no qualm with such a restricted approach. Yet, I am among those who think that economic theory must also tackle the issue of the institutional conditions needed for reaching equilibrium instead of limiting itself to the issues of the logical existence and stability of equilibrium. Trade technology must be taken in earnest. The Walrasian trade technology 20 A Walrasian economy constitutes a single centralised market, encompassing every agent and every good and service. Its cornerstone is the auctioneer whose main task consists of using the 20 Here and henceforth the term Walrasian will be understood in a broad sense as encompassing neo-walrasian models (the Arrow-Debreu model and its followers) or even models that are often, misleadingly, called non- Walrasian.

14 13 price system in order to make individual trade offers compatible. To this end, he announces price vectors to which agents react by expressing the optimal quantity they offer to trade. General equilibrium (at which every market excess demand is nil or negative, in which case the price is equal to zero) will be arrived at through changes in prices. Unlike most general equilibrium theorists who recognise the auctioneer s role only grudgingly and openly declare their disliking, I think that it is integral to the neo-walrasian research programme. To say it bluntly and at the risk of raising the eyebrows of important commentators of Walras, such as Walker (1996), as well as of many present-day Walrasian theoreticians yet not Lucas (see Lucas 1986) the Walrasian approach cannot dispense with the auctioneer, because it is the only working trade technology assumption on stock. This being granted, the need for making explicit all the whereabouts of the auctioneer-led tâtonnement process cannot be side-stepped. On top of what I already stated, other characteristics are worth mentioning. The formation of equilibrium proceeds in one stroke, involving all goods and agents. At each trade round everything occurs simultaneously. Actual transactions remain suspended until the equilibrium price vector is arrived at. In other words, disequilibrium trading is prohibited. It is furthermore assumed that agents behave well and cooperate with the auctioneer, and that the latter pursues the price-formation process up to its end. Although there is less consensus on this point, I am of the opinion that a Walrasian economy is basically non-monetary, in spite of the existence of a numéraire-good. 21 Moreover, in this type of system, production should be viewed as taking place exclusively on order, since production decision remain virtual until equilibrium is formed. Another point that needs to be clarified is the information structure. In the canonical Walrasian or neo-walrasian model agents are supposed to hold perfect information over the prices announced by the auctioneer, the quality of goods and services, and the states of the world. However, due to the presence of the auctioneer, they are not supposed to be knowledgeable over other agents demand and supply functions, nor over the factors underpinning them. All this still makes for a quite heroic information assumption yet, as will be seen, it is less heroic that the informational assumption underpinning Marshallian theory. 22 While the above traits, except the last one, are well known, I now want to bring out a less considered feature, the communication structure of an auctioneer-led system. The auctioneer economy is a set of bilateral relationships between the auctioneer and isolated individual agents. Before the attainment of equilibrium, agents exclusive social link is with the 21 This is clear for what concerns the Arrow-Debreu model. It is likewise so for Walras Elements in so far as it is accepted that Walras introduction of money in his model of circulation and money does not stand up scrutiny (cf. Bridel 1997). On the other hand, starting with Patinkin (1965), a line of authors have claimed to have successfully introduced money in Walrasian theory, yet this claim has been refuted by Hahn ([1965] 1969) (cf. Bridel 2002). Turning to present-day theory, money is present under the form of cash-in-advance in new classical models, yet again it can be argued that its introduction is contrived. 22 Cf. De Vroey (2003a).

15 14 auctioneer. They do not interact and communicate between them. Nor are they cognisant about excess demand functions and the data that underpin them. An important implication ensues. Whenever a given agent makes a trading offer by responding to the prices announced by the auctioneer, he ignores how many other agents are making an offer similar to his. This means that an agent can be in a monopolistic position whilst being totally unaware and hence incapable of taking advantage of it! In other words, the tâtonnement set-up itself guarantees the "perfectness of competition, whatever the possible monopolistic factors that may be present in the economy. Agents cannot influence prices because the institutional device is such as to preclude it. 23 The auctioneer hypothesis and imperfect competition, it turns out, are incompatible bedfellows. One of Walras strokes of genius was to have started his analysis with considering an entire economy at once, rather than a section of it. What he analysed in his study of a two-goods exchange economy was an entire economy reduced to its simplest possible expression instead of a particular market, i.e. an isolated fraction of the economy. Strictly speaking, it should be asserted that the Walrasian economy comprises no markets at all, in so far as the term of market is understood in its usual Marshallian meaning of a separate institutional set-up. If the market notion were to be used in a Walrasian context at all, it should automatically designate the whole economy. 24 The above are the most salient traits of the Walrasian economy. Is the auctioneer trade technology adequate? The answer is clearly negative. Its main flaw is that, for all its allegedly portraying a decentralised economy, it actually assumes a planning system which rests on a centralised coordination procedure. 25 Its only justification is that no alternatives are available. For example, replacing tâtonnement with non-tâtonnement marks hardly a progress. First, the planning trait is hardly removed. Moreover, if the task at hand is to show how the equilibrium calculated by the economist when doing his logical existence of equilibrium analysis can come into existence, non-tâtonnement still fails since the equilibrium eventually reached differs from that which is obtained in the study of logical existence. 23 De Vroey (1998) argues that they are unable to manipulate the auctioneer. 24 Hence the use of any terminology stating that the market for good X exhibits excess demand is misleading as soon as the market term is understood as referring to a specific separate locus of formation of equilibrium Cf. De Vroey (1999a). 25 This point was already made by Hayek decades ago. What the theory of perfect competition discusses has little claim to be called 'competition' at all and its conclusions are of little use as guides to policy. The reason for this seems to me to be that this theory throughout assumes that state of affairs already to exist which, according to the truer view of the older theory, the process of competition tends to bring about (or to approximate) and that, if the state of affairs assumed by the theory of perfect competition ever existed, it would not only deprive of their scope all the activities which the verb 'to compete' describes but would make them virtually impossible. If all this affected only the use of the word 'competition' it would not matter a great deal. But it seems almost as if economists were deceiving themselves into the belief that, in discussing 'competition', they are saying something about the nature and significance of the process by which the state of affairs is brought about which they merely assume to exist. In fact, this moving force of economic life is left almost altogether undiscussed" (Hayek, 1948: 92-93).

16 15 The relationship between the Marshallian and the Walrasian approaches. The possible positions The standard view about the Marshallian and the Walrasian approaches is that they are complementary. On the hand, they share the same subjective theory of value. That is, they ground market supply and demand functions on agents optimising behaviour. On the other hand, a division of labour seems to be existing between them, with Marshallian theory focusing on the study of isolated parts of the economy while the study of the economy as a whole is the concern of Walrasian theory. However, a difference in style of thought and purpose between the two founders of these traditions and their disciples is acknowledged. As noted for example by Negishi, Marshallian models are practically useful to apply to what Hicks called particular problems of history or experience. On the other hand, Walrasian models are in general not useful for such practical purposes. Walras theoretical interest was not in the solution of practical problems but in what Hicks called the pursuit of general principles which underlie the working of a marker economy (Negishi 1987: 590). To paraphrase, Marshall and Walras have different views as to the purpose of economic theory. To Marshall, it must serve to enlighten particular problems. According to the celebrated expression it is an engine for the discovery of concrete truth. Priority is given to empirical relevance, possibly at the price of some lack in analytical rigor. In contrast, Walras is interested in the more philosophical question of the possibility and efficiency of a decentralised economy, an issue that must definitely be tackled at a higher level of abstraction. Analytical rigor rather empirical relevance is the overriding requirement. The above is the consensus viewpoint. However dissenting voices can also be heard. First, in his celebrated History of Economic Analysis, Schumpeter claimed that Marshall proposed a general equilibrium model, akin to Walras. The argument here is that Marshall s Mathematical Note XXI provides the embryo of a general equilibrium theory. The partial-analysis viewpoint is so much in evidence throughout Marshall s text, and the handy concepts of partial analysis that he forged or refurbished have been so generally received into current teaching that there is some excuse for those who see in Marshall the master of partial analysis and nothing more. All the same, this fails to do justice to the depth and range of Marshall s thought. It is not only that the wider conception of the general interdependence of all economic quantities receives intermittent attention in the Principles: Marshall actually formulated this wider conception in an embryonic way but still explicitly in the notes XIV and XXI of the Appendix. And the Memorials contain a passage (p. 417), rightly emphasised by Mr.

17 16 Shove, that reads: My whole life has been given and will be given to presenting in realistic form as much as I can of my note XXI. It seems fair, therefore to list Marshall also among the builders of the general-equilibrium analysis per se (Schumpeter [1954] 1994: 836). 26 I feel sceptical about Schumpeter s claim. Note XXI shows that Marshall is preoccupied with the phenomenon of joined and composite supply and demand. Clearly an element of interdependency is involved. Yet, in my mind, in order to have general equilibrium analysis the object of analysis must the entire economy rather than a section of it. What Marshall does in his note is to widen the scope of the fragment of the economy that he is considering from a single market to a group of related markets, a research strategy that Friedman was to vindicate in his The Marshallian Demand Curve article ([1949] 1953). Yet this does not amount to analysing the entire economy. Moreover, on cannot content oneself with meta-theoretical comments or declarations of intentions as those found in Marshall s letter quoted above. What counts is what is achieved in the theoretical writings, and here little is to be found. Another dissenting opinion is held by economists who want to champion either Walras or Marshall, while dismissing the other. To wit, Samuelson is of the opinion that Marshall has led economic theory into an impasse. But where Marshall threw off two generations of scholars was in his insistence on having his cake and eating it too. He would try to treat at the same time cases of lessthan-perfect competition and of perfect competition. He would try to achieve a spurious verisimilitude by talking about vague biological dynamics, and by failing to distinguish between reversible and irreversible developments. He would insist on confusing the issue of external economies with the entirely separable (and separate!) issue of varying laws of returns. Marshall was so afraid of being unrealistic that he merely ended up being fuzzy and confusing and confused (1967: 111). In contrast, Friedman is dismissive of the Walrasian theory on the grounds that abstractness, generality, and mathematical elegance have in some measure become ends in themselves, criteria by which to judge economic theory while these are all secondary, themselves to be judged by the test of application. The counting of equations and unknowns is a check on the completeness of reasoning, the beginning of analysis, not an end in itself ([1949] 1953: 91). Authors taking this line refuse to enter into general equilibrium analysis on the grounds that it requires a level of abstraction that is to high for elucidating concrete issues. 27 It is not that 26 Reference is also made to Marshall s observation, which has indeed a Walrasian ring, that however complex the problem may become, we can see that it is theoretically determinate, because the number of unknowns is always exactly equal to the number of equations which we obtain (1920: 855-6). 27 This is for example the standpoint that is present in Bordo and Schwartz s paper presented at this conference (Bordo and Schwartz 2003).

18 17 Marshallian analysis should necessarily be confined to a single markets. It can bear on a cluster of markets of substitutable or complementary goods, yet not on the economy as a whole. To have a relevant enough analysis, large chunks of reality need to be put in the ceteris paribus pond. 28 Friedman and his disciples for all intents and purposes are against general equilibrium because they are Marshallian. However, other economists, while vindicating a Marshallian heritage, have striven to develop Marshallian general equilibrium theory, which they view as radically opposed to Walrasian general equilibrium. The names of Clower and Leijonhufvud ought to be evoked in this respect. Like Friedman, they are against Walrasian general equilibrium theory, yet unlike him, they are not dismissive of general equilibrium per se. In the Marshallian scheme of thought (as in earlier classical doctrine) the central task of economic science is to provide an intellectually satisfying account of the coordination of economic activities in an ongoing economic system comprised of business firms and households whose informational links with one another are provided by markets in which dealers of various sorts - visible counterparts of Adam Smith's "invisible hand" stand ready continuously to trade goods for money or money for goods on terms that each trader varies independently in response to forces that impinge directly upon his own working stocks of money and commodities inventories. In Marshallian analysis, economic agents are conceived to be not so much rational as reasonable. Individual fumble and grope rather than optimise. They are presumed to know little and care less about efficiency except as competition forces them to attend to it. The coordination of economic activities is carried out within particular markets by traders (manufacturers and bankers as well as wholesalers, brokers, and retailers) who either do the task effectively or drift into bankruptcy. As for the coordination of activities among markets, since that is not anyone's specific concern it may or may not be done well. ( Clower [1975] 1984: ). 29 For all it appeal, Clower and Leijonhufvud s programme have never to taken off. To date, their blue-print for a theory has failed to be transformed into a full-fledged theory A stated by Stigler, The general equilibrium theory has contributed little to economic analysis beyond an emphasis on mutual dependence of economic phenomena; the problems are far too complicated to be grasped in toto (1941: 242). 29 The same point is made in Clower and Leijonhufvud ([1975] 1984), where the authors set out the blue-print of a Marshallian general process analysis, viewed as a theory of the decentralised economy, which: (1) lacks a central information-processing and bill-collecting agency; (2) has, instead, middlemen trying to coordinate production and consumption activities in each output market separately; (3) makes the management of stocks of inventories essential to the coordination of these activities; and (4) has the system potentially subject to commercial crises associated with expansions and contractions of the volume of bank and non-bank credit. All this might be J.S. Mill or Alfred Marshall. (Clower and Leijonhufvud [1975] 1984: 217). For a more recent assessment see Leijonhuvfud (1998). 30 For recent works that draw their inspiration from Clower and Leijonhufvud, see Colander (1996).