MARSHALL AND KEYNES ON SUPPLY BEHAVIOUR Gerson LIMA1 1. INTRODUCTION In the British Palgrave Dictionary (1987) Prof. Harcourt states that post Keynesian economics has no universal interpretation and analyses three of its strands. One of these, associated with the works by DAVIDSON & SMOLENSKY (1964) and WEINTRAUB (1958), leads to Marshall and to supply and demand analysis. As Davidson quoted in his Preface, for Keynes "the true law relating the aggregate demand and supply functions ... is a vitally important chapter of economic theory". This work is intended to be a contribution in this field: based upon Marshallian principles on producer's behaviour it proposes a notion of supply which is connected with Keynesian features. The proposition is that the notion of supply behaviour and supply curve developed here is the one Keynes searched for on the work of Marshall. Neoclassical theory, following its process of "marginalization" of economic notions pointed out by PASINETTI (1961, page 20), has tried to "appropriate" the notion of supply and demand. However, considering that there is no neoclassical theory of supply outside the world of perfect competition, this appropriation cannot be said to be a success. Accordingly, the broad purpose of this study is to suggest a possible way for emancipating supply and demand analysis, in such a way that it would no longer necessarily be seen as stemming exclusively from a procedure of optimization through differential calculus. This process of marginalization tries to involve Marshall himself, by absorbing his principles on supply into the neoclassical foundations, but this involvement is not stressed here; instead, it is suggested that there are at least two Marshallian approaches to supply behaviour. The first Marshallian approach to supply is well known: it is basically connected with his peculiar definition of short run/long run and with his supply curve, which is constructed on the assumption that price is equal to marginal cost, or more specifically, to his definition of marginal cost. These notions have traditionally been associated with marginalist principles, thus giving Marshall the role of pioneer of neoclassical theory. This approach and the lack of compatibility between different Marshallian approaches are outside the scope of this work. The second Marshallian approach stems from his analysis of classical notions prevailing at his time. This approach, to which economic litterature has deserved less attention than the first one, is concentrated on the description of price formation in the very short run, or daily market period, when real transactions take place and real prices are observed, but there is no clearing: production never matches consumption, which means that in the daily market there is no supply curve and no equilibrium. The specific purpose of this paper is to demonstrate that Marshall's description of the daily market may be developed into a producers' decision-making process which in turn may define a dynamic market model for price and production setting. This model may bring about a long run theoretical status in 1 The Federal University of Paraná, Brazil. An earlier version of this paper was presented at the 1993 Meeting of the National Association of Post-Graduation Courses in Economics, Belo Horizonte, Brazil. which the notion of supply curve emerges, providing thus a general model for supply and demand analysis. A comprehensive empirical example is presented. This paper is organized as follows: section 2 presents Marshallian principles on the very short run, and proposes a simple mathematical formalization of the decision-making process described by Marshall; section 3 develops the decision-making process into a dynamic analytical model for price formation and production setting, from which the supply curve may be brought about, and section 4 develops an empirical example. These sections condense the theoretical discussion and empirical work at the microeconomic level developed in LIMA (1992a) and, finally, section 5 outlines the link between Marshallian microeconomics and Keynesian macroeconomics. Unless specified, page numbers in brackets refer to the Marshall's "Principles". 2. MARSHALLIAN PRINCIPLES ON SUPPLY BEHAVIOUR The Marshallian method of economic analysis may be described by a simple proposition: "the general theory of the equilibrium of demand and supply is a Fundamental Idea" (Preface to the first edition). His model has thus two groups of agents, consumers and producers, who interact with each other in such a way that, under certain conditions, a level of price will be expected to allow the consumption amount to match the production amount. Accordingly, the system must have three basic endogenous variables: consumption, price and production. The final objective is thus to build a market model, based upon the relations between supply and demand, to explain how the levels of those endogenous variables are determined. This is Marshall's subject in Book V of his "Principles of Economics", from which two particularly important assumptions must be stressed. Firstly, competition between agents is said to be "free"; and this proposition has been interpreted by specialists such as SHOVE (1942), HAGUE (1958), and NEWMAN (1960) as meaning that Marshall assumes that imperfect competition prevails; more specifically, GILLEBAUD (1952), MAXWELL (1958), and SHACKLE (1967) propose that Marshall rejects the notion of perfect competition. The intensity of competition is a crucial feature only for neoclassical theory and its discussion is outside the scope of this paper; it is here sufficient to admit, without any pretension to precision, that "free" competition is an indefinite blend of some rivalry and some cooperation. Secondly, although Marshall supposes buyers and sellers to be optimizing agents, he carefully avoids the neoclassical procedure of profit maximization through differential calculus. His unity of analysis on the producer side is, instead of the single producer, the industry of a homogeneous product; when it is necessary to explain the behaviour of the industry, he uses the notion of a "representative firm". Comprehensive analysis of the representative firm may be found in FRISCH (1950), HAGUE (1958), and MAXWELL (1958), with divergent conclusions. Most of this paper is dedicated to the analysis of one principle which seems to be especially important for Marshall: the time component, or the dynamic movement of the endogenous variables towards their equilibrium levels. The purpose here is to stress that there is a Marshallian decision-making model which contains in a latent form a natural dynamic component which can be isolated and formalized mathematically in an explicit form. This 2 analysis is intrinsically associated with the Marshallian notions of gravitation and equilibrium. In its simplest form, the Marshallian model, or more generically the supply-and-demand model, is composed of at least three simultaneous equations, because the endogenous variables are at least three: consumption, price, and production. In Marshall terms (Preface to the first edition), one must be "sure that he has enough, and only enough, premisses for his conclusions (i. e. that his equations are neither more nor less in number than his unknowns)". One of the equations is the demand curve, which provides a relation between price and consumption. The second relation could be the equilibrium condition: the equality between the amount demanded and the amount produced. Considering that equilibrium is unfeasible in the real world, Marshall proposes that the market clears for any commodity only under "normal" conditions, defining as normal a theoretical supply curve in which "the supply price of any amount of that commodity may be taken to be its normal expenses of production" (page 285). The normal supply curve is thus the third relation searched for. Marshall breakes normal conditions into two categories: the short and the long run. In the short run the industrial production capacity is given, and in the long run all factors are variable and may be adapted to the level of demand. However, there is no clear line between long and short run: "Nature knows no absolute partitions of time into long periods and short" (Preface to the first edition); in any period of time "price is governed by the relations between demand and supply" (page 314). Beyond the short and the long run there is also the "very short run", or the "daily market" period, in which the Marshallian normal conditions cannot be observed. In the daily market period production only ever matches consumption incidentally: market does not necessarily clear in actual day-to-day transactions. There is no supply curve in the daily market; production is given and hence "Market values are governed by the relation of demand to stocks actually in the market" (page 309), with less influence from cost. This is the subject of this section: the Marshallian principles on producer behaviour in the daily market period. 2.1. The Marshallian Decision-Making Model It is in the daily market that actual trading takes place, including the purchase of those production factors which will be used in the next period. It is during the daily market that decisions made on price and production are transformed into real transactions: observable values, those collected for statistical purposes, are those realized in this period. Considering that in the daily market period price may be adapted to existing conditions, but production cannot because it always takes time, each of these variables may follow different decision patterns, which would be partially independent of each other. Accordingly, Marshall proposes a model of daily market supply behaviour in which producers follow a decision-making process divided into two components, the supply price and the production decision: "Production and marketing are parts of the single process of adjustment of supply to demand" (Industry and Trade, page 181). 2.1.1. Decision-Making on Supply Price Marshall states that each firm interacts with its customers proposing daily market prices which are not equilibrium values; the market would only be in equilibrium when and if the exogenous variables stopped varying. Prices are never determined by cost or demand alone: 3 particularly in the daily market prices are mainly governed by the available stock in relation to the demand (page 290), in such a way that "As a general rule, the shorter the period which we are considering, the greater must be the share of our attention which is given to the influence of demand on value" (page 291). Despite playing a less important role, cost of production is also considered in the producers' decision-making process on price, associated with the principle that "The amount which each farmer or other seller offers for sale at any price is governed by his own need for money in hand" (page 277), money which is necessary to finance at least part of the next production cycle. Given that the market is a permanent entity, a daily market never occurs only once, and the explanation for the sellers' behaviour in the daily market period must be searched for in the long run: "a controlling influence over the relatively quick movement of supply price during short periods is exercised by causes in the background which range over a long period; and the fear of spoiling the market often makes those causes act more promptly" (page 313). For example, after a fall in demand the seller will not necessarily cut prices to the point at which all stock could be sold, either because "Each man fears to spoil his chance of getting a better price later on from his own customers" (page 311), or because "he is more or less in fear of incurring the resentment of other producers, should he sell needlessly at a price that spoils the common market for all" (page 311). Marshall describes the relationship between buyers and sellers in daily transactions as a "higgling (sic) and bargaining" process in which both parties try to show some degree of indifference so as to induce a reduction in the rival's targets. Sellers "will not show at once that they are ready to accept that price" (page 277) which will eventually be accepted; buyers insinuate that they are not interested in the purchase, they "pretend to be less eager than they really are" (page 277). On the other hand, the statement about the relationship between many competitors is that "though everyone acts for himself, his knowledge of what others are doing is supposed to be generally sufficient to prevent him from taking a lower or paying a higher price than others are doing" (page 284). This approach to price formation stems from the classical side of Marshall; it may also be found in SMITH (1776) who had already remarked that in the daily market selling prices depend upon the need of the seller "to get immediately rid of the commodity" (page 57). This means that if the seller is not in immediate need of cash he does not sell, or he would sell at whatever price only an amount sufficient to solve his immediate cash problems. If the seller does not sell all the quantity available, then there will be a rise in stock, in such a way that in the next decision it will probably be more important to "get rid" of the commodity, but again without excessively cutting prices in order not to "spoil the market". Therefore, "not spoiling the market" is associated with inventories, and the stock imposes a downward pressure on prices: to greater inventory, smaller price. These notions have probably been developed by classical authors for the general purpose of explaining why the daily market price is not determined exclusively by demand, why there is no daily market supply curve which would simply be given by the existing stock, and why there is no vertical daily market supply curve. The same purpose may be identified in some other post-Marshallian propositions; for example the "user cost" by KEYNES (1936), the "convenience yield" by KALDOR (1939), the "normal stock" by SAMUELSON (1948, page 268), and the "shadow inventory" by BLINDER (1982). All these approaches seem to be intended to explain the reason why the supply curve cannot be a vertical line, why 4 producers do not offer all available stock, accepting the price imposed by consumers, clearing thus the market in such a way that the inventory level would always be null. Alternatively, given the fact that some inventory almost always exists, some authors have proposed that it results from a decision to retain stocks, and not necessarily from a decision not to sell all that has been produced. The purpose of stockpiling in this case would be twofold: to keep the price to a certain level for a certain period of time, and to "smooth" the production changes, preventing excessively large variations. More precisely, in the neoclassical approach some stock necessarily results from the hypothesis of the maximization of the present value of future profits, apparently first suggested by SMITHIES (1939): if demand is expected to grow, then it is more profitable to build up stock for selling in the future, and vice versa. One competing idea, perhaps introduced by KIRMAN & SOBEL (1974), is to propose inventory as a strategic variable in game models. These neoclassical propositions lead to the notion that stock exists because there is a policy for it, or a demand for it, and the demand for stock, or the investment in stock, has generally been associated with a target or a planned inventory level. In the neoclassical tradition two approaches have more frequently been proposed to deal with inventories: the "buffer stock/production smoothing" and the (S,s) models; WILKINSON (1989) and BLINDER & MACCINI (1991) present comprehensive surveys on this subject. However, it is not difficult to realize that if producers can make decisions on price and production, then the market model has as many equations as variables and the neoclassical equation of the demand for inventory would therefore be redundant or, worse, inconsistent. At the macroeconomic level the existence of two particular inventories may be noticed. Firstly, a reserve fund may be seen as a stock of money which would be a corollary for the Keynes' speculative reason for holding money, as noted by ARROW, KARLIN & SCARF (1958). It is worthwhile to observe the inverse relation: the greater the reserve of money, the lower the interest rate. Secondly, the Phillips curve is a relation between wage and another macroeconomic inventory: unemployment. PHILLIPS' (1958) proposition is that "the rate of rise (of prices) being greater the greater the excess demand” (page 283). When demand is growing small inventory levels allow for larger price variations. A corollary is that, as a general rule, large stocks are associated with low price levels. Therefore, also at the macroeconomic level there is an inverse relation between price and inventory which has been the subject of much empirical analysis, such as that presented by WILKINSON (1989). To sum up, the supply price is actually observed only in the daily market, and it is a function of the cost of production and the stock then available; at the beginning of each daily market period the stock is given, and if cost were not considered then the price would be determined exclusively by demand. Nonetheless, whoever says "that the price is governed by demand (...) does not claim strict accuracy" (page 290). In Marshall's approach neither the cost nor the demand which is embodied in the inventory are alone sufficient to explain the price: they are both necessary. The supply price may thus be expressed as a function of the cost W and the inventory E which comes from the previous period: supply price: Pt = f (Wt , Et-1) (1) where the derivative in relation to costs (dP/dW) is positive and (dP/dE) is negative. 5 2.1.2. Decision-Making on Production In Marshall's approach the producer's decision-making on production is the complementary component of supply behaviour. It is important to stress first of all that industrial production is naturally complex and normally takes time; nonetheless, all decisions on production are transformed into reality in the daily market, through the purchasing and hiring of factors, which is the only event observed for statistical purposes. In the Marshallian approach production decision is primarily based upon the profit margin, both expected and previously observed in the daily market period: the higher the profit, the higher the total capital allocated to the branch, increasing production in the short run, and capacity in the long run. This process may be identified with the notion of capital mobility. In some daily market period the supply price could be such that the margin would be "insufficient"; however, in a long run average the profit margin must be considered at least as "acceptable": if the price is not "sufficient to cover in the long run a fair share of the general costs of the business, production will gradually fall off" (page 313). The profit margin is defined (page 313) as the surplus over direct costs; it is the return on the total capital, and depends also on the turnover of floating capital (page 374). Among many alternatives, the profit margin may be defined as the difference between the supply price P and the cost of production W: profit margin: Mt = (Pt - Wt) (3) Despite the fact that it is adequate for temporal analysis on the same industrial branch, for empirical comparisons between different branches this definition may be misleading; in this case it should be modified by information on the different fixed capital and on the turnover observed in each branch. In theoretical terms it may be expected that factor mobility, especially financial capital mobility, would prevent sectoral production constraints: given a reserve fund, if capital is free to move, then all productive sectors have the financial capital they consider appropriate for the running level of production, obtaining thus an unconstrained, non-uniform, profit margin. Additionally, Marshall states that production also depends upon inventories: a rise in stock is perceived as a fall in demand, and thus each company, acting alone or in accordance with its competitors, diminishes production in order to avoid excessive stockpiling and further pressure of stock on price. This stems from Marshall's remark that in order to prevent the inventory from becoming excessive, producers decide, "either from a regard to their own interest or by formal or informal agreement with other producers, to suspend production for fear of further spoiling the market" (page 311). On the contrary, if stocks fall producers understand that demand has risen and then each company plans to invest and increase production, trying to acquire for itself the largest possible share of the new demand. To sum up, the production decision component of supply behaviour may be stated as a function of the previously realized profit margin M and the stock E then available: production decision: Qt = h (Mt-1, Et-1) (4) where the derivative in relation to profits (dQ/dM) is positive and (dQ/dE) is negative. Besides Marshall's contribution, the existence and the importance of a relation between inventory and production decision has been studied in many theoretical and empirical works, under different approaches among which the "stock adjustment" by HICKS (1965) and the production smoothing procedures must be remembered. 6 Therefore, the complete Marshallian daily market model has as many equations as endogenous variables: producers set price P through the supply price equation which states that price is a function of the exogenous variable cost, and of inventory E which has a natural accounting equation associated with it; consumers decide on the amount demanded D through the equation of the total demand; and finally producers decide the amount Q to produce through the equation of the production decision, which is based on the profit margin M defined, for example, by equation (3), and on inventory E. There is no supply curve in the daily market; this would be a redundant equation. 2.2. Concluding Remarks on Marshall The necessary and sufficient condition for equilibrium is market clearing: quantities supplied and demanded must be equal. In contrast, Marshall assumes that in the real daily market production and consumption never actually match. In the daily market there is no clearing, which is the same as to say that in the daily market, when real transactions are carried out, there is no equilibrium between supply and demand; accordingly, there is no daily market supply curve. This distinction between non-clearing daily market and clearing equilibrium market is important also because data on real transactions are observed and collected for statistical and analytical purposes only in the daily market when there is no equilibrium between supply and demand: real data are disequilibrium data. Marshallian producers make decisions when the market is outside equilibrium; however, it would be inappropriate to say that he assumes a disequilibrium approach: the lack of equilibrium in his model is associated with actual transactions realized during the daily market period, and this feature stems basically from the behaviour of the exogenous variables, costs and demand, which are continuously varying. Marshall states that the market is always moving towards equilibrium; it would reach this situation only if the exogenous variables were constant during a certain minimal period of time. Marshallian equilibrium is a theoretical construction, an abstraction; quoting Adam Smith "and other economists" as he says, he points out that "the normal or natural value of a commodity is that which economic forces tend to bring about in the long run (...) if the general conditions of life were stationary for a run of time long enough" (page 289). ROBINSON (1965) states that "The short period is here and now, with concrete stocks of means of production in existence. Incompatibilities in the situation (...) will determine what happens next. Long-period equilibrium is not at some date in the future; it is an imaginary state of affairs in which there are no incompatibilities in the existing situation, here and now" (page 101). In this line BOGGIO (1987) observes that, in classical terms, natural rates and prices are centres of gravitation towards which actual endogenous variables tend to move; in Marshall's words equilibrium positions are "centres about which the amount and the price tend to oscillate" (page 289). Marshall uses the image of a stone hanging by a string (page 288) to illustrate this phenomenon: if the stone is shifted from its equilibrium position the force of gravity will bring it back, after some oscillation. Considering that exogenous shocks induce "movements partially rhythmical and partially arbitrary" (page 288), the conclusion is that gravitation around equilibrium positions is a random phenomenon. Gravitation is thus a component of economic variables which stems from the process of adjustment of producers to ever-changing demand and cost conditions. 7 3. AN ANALYTICAL APPROACH TO MICROECONOMICS In a few words, the dynamics of the Marshallian decision-making model is as follows: variations in the exogenous variables on the consumers' side, such as their income, lead to shifts in the demand curve, then in consumption, and then in inventories. Variations in inventories induce producers to change price and production simultaneously, each one searching for the best profit allowed for him by the intensity of rivalry and the degree of cooperation between them. Both price and production are thus connected to inventories, and much research work has been conducted on one or other relation: price vs inventories, or production vs inventories. In the literature of economics there is little analysis of the simultaneous relations of inventories with price and production. However, this is not a new strand; theoretical contributions are presented in DUMENIL & LEVY (1987), KIRMAN & SOBEL (1974), and in HAY (1970). Additionally, econometric works have been developed by MILLS (1962) and by his critics STEUER & BUDD (1968). Particularly interesting are the empirical findings of KAWASAKI, McMILLAN & ZIMMERMANN (1982), as their work is based upon a statistical method which is similar to a reduced model, but is independent of any particular structural market model. Their conclusion is that firms do react to any shift in the level of inventories, setting new price and production values, in such a way that convergence to a new equilibrium situation is assured. Based upon these principles, this section proposes and develops the following DYNAMIC ANALYTICAL MARKET MODEL demand: Dt = ao - a1 Pt + a2 Ft supply price: Pt = bo + b1 Wt - b2 Et- production decision: Qt = ho + h1 Mt- - h2 Et- where Mt = Pt - Wt is the profit margin; and Et = Et-1 + Qt - Dt is the inventory. There are five endogenous variables: D, P, Q, M, and E, and five corresponding equations. The exogenous variables are income-related F and cost-related W. The main feature of the model is the supply behaviour, which is a generalization of the Marshallian decisionmaking model on supply price (equation 1) and on production (equation 4). Considering, as suggested by Marshall (page 281), that current values are consequences of previous decisions, instead of basing their decisions only on their last realized profit and on their available inventory, in this model producers consider a series of the past values of these endogenous variables, represented by the lag structure given by , and . The lag structure is a priori unknown, but it may be determined by econometrics of the model. This is not a disequilibrium schedule, it is a dynamic model. Given a situation of equilibrium, after an external shock in time t the model brings about the values of all endogenous variables in time t+1, t+2, t+3, and so on. If the external shock was given only once, then successive adjustments made by producers would make endogenous variables follow a damped oscillatory trajectory in the direction of a new equilibrium situation. 8 However, considering that in the real world exogenous shocks are many and randomly distributed, the trajectory is always disturbed, in such a way that the equilibrium is never actually attained; actual values oscillate around the equilibrium positions. Special equilibrium values for endogenous variables are neither expected nor assumed ad hoc by this model; for example, inventory is in no circumstances presumed to be null. 3.1. Equilibrium and Gravitation Considering that the equilibrium situation cannot be attained, it may only be imagined as a theoretical construction; it could only be observable if and when all exogenous variables have stopped varying. Then, after some time has elapsed, the effects of external shocks over all endogenous variables would have vanished. The structural market model would therefore be in a motionless position, which is the equilibrium status. The model has thus one theoretical equilibrium position in time t which is associated with the values for the exogenous variables in time t. Therefore, the model actually gravitates around theoretical equilibrium positions. Actually, observed economic values are always daily market values; long run equilibrium values are abstractions from reality which must necessarily be based on those daily market values. Long run equilibrium values are neither decided nor imposed by producers: they must be theoretically deduced from daily observed values. The process by which actual values of economic variables are kept away from their equilibrium status is the gravitation phenomenon, which depends on how fast producers can and want to adjust production to consumption. Measured by the spread between actual and long run equilibrium values, gravitation is the link between reality and economic theory. This process of adjustment of production is required as a consequence of exogenous shocks given to the system, from both demand and cost sides. Considering that these shocks are random, gravitation is also random and must have zero mean; it is a disturbance term. This also means that gravitation adds to the behaviour of an economic endogenous variable a trend component which may be isolated and eliminated because it is just a noise without economic meaning. Moreover, gravitation is not tractable through time-series analysis because it has a non-systematic, impredictable behaviour. As LIMA (1992b) proposes, gravitation is an error component of economic variables, and consequently parameters obtained by direct econometric estimation using actual series will probably be biased. Alternatively, gravitation may be seen as an omitted varible, implying that there will probably be autocorrelation in the residuals. Therefore, the econometric method should eliminate the gravitational component of endogenous variables. 3.2. The Supply Curve The main feature of the analytical decision-making model is supply behaviour which has been divided into two separate equations which is associated with the short run: the supply price and the production decision. However, price and production are mutually dependent, via inventories, and when in the theoretical situation of equilibrium one is connected with the other by a special, stable, theoretical relation which may be identified as the supply curve. The short run supply behaviour is such that changes in the level of inventories induce changes in both price and production, and in the same direction, because coefficients have the same negative signal in the supply price equation and in the production decision equation. 9 Therefore, these two equations may be combined: focusing on the production decision, first the stock E is replaced by its expression taken from the supply price equation. The inventory will thus be eliminated in the production decision equation, which will become a function of M, P and W. After that, the profit margin M is replaced by its definition, which states that it is the difference between P and W. These substitutions lead eventually to one expression where Q is a function of P and W, such as for example: Q*t = Bo + B1 P*t - B2 Wt (5) where asterisks indicate equilibrium values. This equation is the line of simultaneous theoretical equilibrium of price and production which may be defined as the supply curve: it results from the behaviour of producers when deciding on price and production, which means that it contains only price and production levels that producers consider interesting for them. Differently from what neoclassical theory states, the supply curve is not a relation of causality; this would happen only if firms were allowed to decide production based upon a given price level, or vice versa. Instead, firms have some discricionary power over market prices and can modify them by varying production, and vice versa. In any case, the final theoretical equilibrium levels of price and production depend also upon consumers, via their demand curve. The supply curve is a theoretical construction connecting imaginary equilibrium levels of price and production; it cannot actually be observed because real values are only those computed from the daily market when there is no supply curve. This dynamic model can therefore describe the daily (actual) and the long run equilibrium (theoretical) economic behaviour. Considering that the supply curve is derived from its two components, the supply price and the production equation, its slope depends upon the propensity to invest and on the simultaneous reactions of all companies in relation to individual inventories shifts. Formally: (dQ/dP) = [(Q/M) x (M/P)] + [(Q/E) x (E/P)] where (dQ/dP) is the long run supply slope. While the position (height) of the supply curve depends on the technology of production and on input prices, represented by W in the model, the slope is subject to some physical features, such as perishability, technology of production and distribution, management, financial strength, and ultimately upon businessmen's characteristics, such as proficiency, reactions to expectations formation, selling eagerness, knowledge of the market, cash availability, and rivalry. Competition is in turn conditioned by objective considerations such as the availability of funds, and the notion that it makes no sense to an individual producer to carry the stock of the whole industry. With the exception of the direct effect of profit on production, represented by the first square brackets in the expression above, and the given physical restrictions, Marshall condensed all those behavioural features which influence the supply slope into the single notion of "not spoiling the market". One important characteristic of those features is that they are non-separable ex post; it is unfeasible to identify empirically the effect of any one of them, thus isolating it from the others, because one cannot have statistical information on them: most of them are not measurable, as BLINDER (1990) has already observed. Incidentally, neoclassical theory adds to that list of behavioural features one more hypothesis: that the process of profit 10 maximization can be described by differential calculus. Following this approach, modern neoclassical analysis such as that developed in BRESNAHAN (1989) applies differential calculus for maximization simultaneously condensing those behavioural features into notions such as "index of competitiveness" or "conjectural variation". SCHMALENSEE (1988), more specifically, states that besides maximization the supply slope "emerges from what may be a complex pattern of behaviour" (page 650). However, the fact that the individual effect of each of those features on the supply slope cannot be identified remains untouched if one multiplies or not the composed effect by an additional hypothesis - the profit maximization - about the behaviour of producers. Ironically, this means that the neoclassical effort in developing economic theory may possibly be reduced to a simple change in names, replacing "not spoiling" by "complex pattern" without adding new information on supply behaviour. Except if game theory can find a miraculous device to measure empirically how competitors interact, and not simply state it as a hypothesis unsuited for any empirical test, it looks improbable that neoclassical theory will someday elaborate a theory of supply behaviour. To sum up, the supply curve is a synthetic tool, describing a set of possible equilibrium points of the economy; a situation to which the economy would tend to run, if exogenous variables have stopped varying. The slope is the core of the supply curve and arises from the psychological behaviour of producers, which is constrained by some permanent and temporary conditions imposed by institutions of real life. No special value, ad hoc defined, is expected; the slope is a composition of parameters of the structural model and, like them, it cannot be determined ex ante. The slope of the supply curve is a resultant of "a complex pattern of behaviour"; instead of theoretically anticipated, it must and can be estimated by econometrics. 4. AN EMPIRICAL APPLICATION 4.1. Main Results In order to give empirical support to the analytical approach proposed here, this section presents briefly the results obtained in a study on the Brazilian cement industry. The market model is that presented in section 3, and the exogenous variables are: F, the Gross Fixed Capital Formation, and W, the estimated direct cost of production. The econometric method is that presented in LIMA (1992b). The reduced model is presented below, where D1 refers to a dummy variable associated with the fact that during the first half of the period of analysis production was constrained by available industrial capacity (“t” statistics are in brackets): ESTIMATED REDUCED MODEL CONSUMPTION: Dt = 9.896 - 0.312 Wt + 30.234 Ft-1 - 2.958 D1 (5.17) (-2.91) (13.07) (-3.76) R2 = 0.98 DW = 1.90 F (3,12) = 162.48 PRICE: 11 Pt = 16.791 + 1.102 Wt-2 + 11.967 Ft-1 (6.33) (6.04) (2.32) R2 = 0.87 DW = 2.24 F (2,12) = 40.64 PRODUCTION: Qt = 9.362 - 0.304 Wt + 31.048 Ft-1 - 2.960 D1 (4.69) (-2.72) (12.87) (-3.60) R2 = 0.98 DW = 1.94 F (3,12) = 158.10 PROFIT MARGIN: Mt = 8.910 + 0.908 Wt-2 + 16.337 Ft - 13.970 Ft-2 (2.62) (4.21) (2.17) (-2.08) R2 = 0.71 DW = 2.24 F (3,11) = 9.16 INVENTORY: Et = 0.310 + 0.076 Wt-1 - 1.618 Ft + 1.372 Ft-1 - 0.404 D1 (0.88) (4.56) (-2.28) (1.93) (-2.81) R2 = 0.93 DW = 2.09 F (4,11) = 36.71 INVENTORY (NO INTERCEPT): Et = 0.086 Wt-1 - 1.415 Ft + 1.378 Ft-1 - 0.296 D1 (7.13) (-2.13) (1.96) (-3.92) In the next step the reduced equilibrium model was calculated, supposing theoretically that all exogenous variables were fixed at their levels in time t and that D1 is always zero. Formally: REDUCED EQUILIBRIUM MODEL consumption: D*t = 9.896 - 0.312 Wt + 30.234 Ft price: P*t = 16.791 + 1.102 Wt + 11.967 Ft production: Q*t = 9.362 - 0.304 Wt + 31.048 Ft profit margin: M*t = 8.910 + 0.908 Wt + 2.367 Ft inventory: E*t = 0.086 Wt - 0.037 Ft Considering that the demand equation and the supply price equation are just-identified, their parameters could be derived directly from the coefficients of the reduced equilibrium model following an indirect least squares procedure. Nonetheless, in this example the equilibrium series, which are free of gravitation, were used in the second stage of the econometric work to compute the structural relations between the endogenous variables of the market model, then bringing about the estimates: SUPPLY PRICE: 12 P*t = 18.00 + 15.82 Wt - 169.7 E*t (15.9) (5.3) (-4.9) R2 = 0.97 DW = 2.03 F (2,12) = 224.3 PRODUCTION DECISION: Q*t = - 33.89 + 2.57 M*t - 6.58 E*t (-985) (1639) (-754) R2 = 1.00 DW = 2.51 F (2,12) > 10000 DEMAND CURVE: D*t = 14.65 - 0.28 P*t + 33.62 Ft (7241) (-4258) (13284) R2 = 1.00 DW = 1.97 F (2,12) > 10000 Finally, the long run equilibrium supply curve (expression 5), which is theoretically derived as a combination of the supply price and the production decision equations, may also be seen as an outcome of an ILS procedure that brings about the expression below. SUPPLY CURVE: Q*t = - 34.6 + 2.61 P*t - 3.20 Wt 4.2. A Commentary on Price Stickiness The cement industry has frequently been taken as a major example of oligopoly, and the Brazilian industry cannot be seen as an exception to this rule. One could then model price formation in this market as if firms follow the rigid mark up principle of the "orthodox" post-Keynesian theory. In this situation price would be fully sticky in accordance with the notion that price stickiness may be considered as a safe indication of the existence of a cartel. Suppose thus that a particular model assumes price P to be a simple linear function of production costs W. Using the same econometric method one could therefore obtain the reduced equation for the price: Pt = 17.328 + 0.803 Wt-1 + 0.691 Wt-2 (7.29) (2.66) (2.45) R2 = 0.88 DW = 2.02 F (2,12) = 45.10 In the next step one would derive the price rule: Pt = 17.328 + 1.494 Wt In this simplified model there would be no room for production decision: price is given for consumers who decide how much to buy and consequently how much will be produced. Therefore, the supply curve would be the same as the price rule. Apparently, actual data does not allow for rejection of the Keynesian orthodox hypothesis that the mark up principle prevails in the Brazilian cement market. This simple model takes as an assumption that, at least in the short run, producers set prices without regard to demand. 13 However, by assuming price to be demand-insensitive this model just cannot test the hypothesis whether demand influences prices or not. In contrast, the general approach proposed here can do the thrick. In fact, econometric results obtained from the general model reflect the market features one could expect from this industrial branch: supply is quite elastic and Brazilian cement prices are, if not fully, certainly "very" sticky. As estimated with the values of exogenous variables for 1986, price elasticity of supply is 4,0 while the elasticity of price in relation to demand variations is only 0.21. For an economic expansion as great as 10%, the cement price would rise by only roughly 2%. Actual data gives support to the hypothesis that demand do influences prices, but also reveals that this influence is low when compared with production costs. The supply-and-demand model can deal with the notion of price stickiness without assuming a rigid mark up pricing policy. Price rigidity can actually be tested instead of assumed. Furthermore, the "historical conditions" which could explain why short run mark up rates vary must necessarily be embodied in the levels of exogenous variables of the general model, supposing of course that no relevant exogenous variable has been ommitted. The supply-and-demand model, which is not a neoclasssical device, is more general than the rigid mark up pricing policy. 5. A LINK BETWEEN MARSHALL AND KEYNES This paper develops the Marshallian principles on the daily market, when real transactions are made and real prices are observed, to build a dynamic model which may explain both the theoretical imaginary equilibrium situation and the actual movements around it. Furthermore, the model allows for the deduction of the theoretical supply curve, whose slope depends on a complex behaviour of producers, thus reproducing the supply-anddemand model, independently of the hypotheses of profit and utility maximization. It is demonstrated that supply-and-demand is not necessarily a neoclassical concern. This approach is in line with the "anti-Say" Marshallian notion that the economy is pushed by individual producers who adjust supply to demand, following a convergent decisionmaking process. The connection with Keynes and macroeconomics may be briefly outlined, as follows, in addition to the statement by Harcourt and the references to the Keynesian user cost and to the Keynesian speculative reserve fund as a kind of inventory. Firstly, the analytical approach to supply behaviour has given no ad hoc special equilibrium levels; it leads freely to some theoretical equilibrium solution: a free "intrinsically economic" solution. This means that the interaction of supply and demand will determine not only the theoretical equilibrium levels of price, production and inventory, but also the actual industrial capacity utilization rate, the level of the reserve fund, the employment level and so on. The analytical approach assures that, given the economic policies at the micro and macroeconomic levels, there will always be some theoretical "market solution". This solution cannot ensure full-employment as a principle. Instead, such as the involuntary inventory of finished goods, the involuntary industrial idleness and the involuntary stock of the speculative fund, the involuntary unemployment - the Keynesian unemployment - may be expected to be the rule and may be determined empirically. This approach looks like a legitimate out-of-equilibrium Keynesian feature. It is not an equilibrating device, in the sense that a situation with some kind of "justice" would eventually emerge. It allows for a 14 super-aggregate analysis in which historical circumstances - captured by the model through the levels of the exogenous variables - do matter. Secondly, the formal link with macroeconomics will be given by the production (investment) decision equation, which is in fact a factors demand equation, financial capital certainly being the most important of them. Investment is a function of the profit rate and depends on the mobility of financial capital: given a speculative reserve fund, if financial capital is free to move, then all productive sectors may generally have the financial capital they are worth of. Further discussion on the theme and an empirical application of this supply-and-demand approach may be found in LIMA (1993), where special emphasis is given to the analysis of the Brazilian inflationary process. To sum up, mobility of capital is the interface between the real industrial productive sector and the monetary Keynesian reserve fund, between the real and the monetary sides of economics. Capital mobility provides the link between Marshallian microeconomics and Keynesian macroeconomics. REFERENCES ARROW, K. J., KARLIN, S. & SCARF, H. (1958), "Studies in the Mathematical Theory of Inventory and Production". Stanford University Press. BLINDER, A. S. (1982), "Inventory and Sticky Prices: More on the Microfoundations of Macroeconomics". American Economic Review, vol. 72, pp. 334-48. BLINDER, A. S. (1990), "Price Stickiness in Theory and Practice". American Economic Review, Papers and Proceedings, vol. 81, pp. 89-99, 1991. BLINDER, A. S. & MACCINI, L. J. (1991), "The Resurgence of Inventory Research: What Have We Learned?". Journal of Economic Surveys, vol. 5, pp. 291-328. BOGGIO, L. (1987), "Centre of Gravitation", in The Palgrave Dictionary of Economics. MacMillan, reprint of 1988. BRESNAHAN, T. F. (1989), "Empirical Studies of Industries with Market Power", in SCHMALENSEE, R. & WILLIG, R. D. (editors), "Handbook of Industrial Organization", Volume II. Elsevier. DAVIDSON, P. & SMOLENSKY, E. (1964), "Aggregate Supply and Demand Analysis". Harper & Row. DUMENIL, G. & LEVY, D. (1987), "The Dynamics of Competition: A Restoration of the Classical Analysis". Cambridge Journal of Economics, vol. 11, pp. 133-64. FRISH, R. (1950), "Alfred Marshall's Theory of Value". Quarterly Journal of Economics, vol. 64, pp. 495-524. GILLEBAUD, C. W. (1952), "Marshall's Principles of Economics in the Light of Contemporary Thought". Economica, vol. 19, pp. 111-30. HAGUE, D. C. (1958), "Alfred Marshall and the Competitive Firm". The Economic Journal, vol. 68, pp. 673-90. 15 HARCOURT, G. C. (1987), "Post-Keynesian Economics", in The Palgrave Dictionary of Economics. MacMillan, reprint of 1988. HAY, G. A. (1970), "Production, Price and Inventory Theory". American Economic Journal, vol. 60, pp. 531-45. HICKS, J. R. (1965), "Capital and Growth". Oxford University Press. KALDOR, N. (1939), "Speculation and Economic Stability". Review of Economic Studies, vol. 7, pp. 6-48. KAWASAKI, S., McMILLAN, J. & ZIMMERMANN, K. F. (1982), "Dis-equilibrium Dynamics: An Empirical Study". American Economic Review, vol. 72, pp. 992-1004. KEYNES, J. M. (1936), "The General Theory of Employment, Interest and Money". MacMillan. KIRMAN, A. & SOBEL, M. J. (1974), "Dynamic Oligopoly with Inventories". Econometrica, vol. 42, pp. 279-87. LIMA, G. P. (1992a), "Une Analyse Critique des Fondements Théoriques et Empiriques de la Courbe d'Offre". PhD dissertation, University of Paris X. LIMA, G. P. (1992b), "The Equilibrium Component of Dynamic Models", Regional Meeting of the Brazilian Econometric Society, University of S. Paulo, October 1992. LIMA, G. P. (1993), "Origem e Robusteza do Processo Inflacionário Brasileiro" (The Origin and the Ability-to-Resist of the Brazilian Inflationary Process), Discussion Paper #14, Department of Economics, University of Brasília, June 1993, and Cadernos de Direito Tributário e Finanças Públicas da Revista dos Tribunais, July/sept 1993, pp. 161-76. MARSHALL, A. (1890), "Principles of Economics". MacMillan, 8th edition, printing of 1986. MARSHALL, A. (1919), "Industry and Trade". Augustus M. Kelley, printing of 1970. MAXWELL, J. A. (1958), "Some Marshallian Concepts, Specially the Representative Firm". The Economic Journal, vol. 68, pp. 691-8. MILLS, E. S. (1962), "Price, Output, and Inventory Policies". John Wiley. NEWMAN, P. (1960), "The Erosion of Marshall's Theory of Value". Quarterly Journal of Economics, vol. 74, pp. 587-601. PASINETTI, L. L. (1961), "Structural Change and Economic Growth". Cambridge University Press, edited in 1981. PHILLIPS, A. W. (1958), "The Relation between Unemployment and Rate of Change of Money Wage Rates in the United Kingdom, 1861- 1957". Economica, vol. 25, pp. 283-99. ROBINSON, J. (1965), "The General Theory after Twenty-Five Years". Collected Economic Papers, vol. III, pp. 100-2. SCHMALENSEE, R. (1988), "Industrial Economics: An Overview". The Economic Journal, vol. 98, pp. 643-81. SHACKLE, G. L. S. (1967), "The Years of High Theory". Cambridge University Press. 16 SHOVE, G. F. (1942), "The Place of Marshall's Principles in the Development of Economic Theory". The Economic Journal, vol. 52, pp. 294-329. SMITH, A. (1776), "The Wealth of Nations". The Modern Library, New York, 1937. SMITHIES, A. (1939), "The Maximization of Profits over Time with Changing Costs and Demand Functions". Econometrica, vol. 7, pp. 312-8. STEUER, M. D. & BUDD, A. P. (1968), "Price and Output Decisions of Firms - A Critique of E. S. Mill's Theory". Manchester School of Economic and Social Studies, vol. 36, pp. 1-25. WEINTRAUB, S. (1958), "An Approach to the Theory of Income Distribution". Chilton, New York. WILKINSON, M. (1989), "Aggregate Inventory Behavior in Large European Economies". European Economic Review, vol. 33, pp. 181-94. Curitiba, Brazil, May 1996. 17