Essays on Decentralized Trade

Nicolaas J. Vriend

Department of Economics of the European University Institute, Florence
April 1993

Thesis submitted for assessment with a view to obtaining the Degree of Doctor of the European University Institute

pdf Complete thesis (PDF format)


  • Acknowledgements v

  • Introduction vii

  • I On the Walrasian Perspective and Decentralized Economies 1

  • II Market-making and Decentralized Trade 21
    Appendices 58

  • III Economics and the Concept of Rationality 77

  • IV Decentralized Trade with Perfectly Rational Agents. 107
    Creating and Trading away Advantageous Opportunities
    Appendices 156

  • References 163


One of the main tasks of economic theory is to explain the outcomes of a decentralized economy. That is, one of its principal aims is to construct logically coherent, comprehensive stories about what is going on in such an economy that are compatible with observed facts. In this sense, explanation of the course of events in a decentralized economy equates to understanding its disparate phenomena (see Aumann [1985]). It might be tempting to interpret the difficulty of economic theory to explain the fluctuations in the aggregate economic activity known as business cycles, as merely a matter of a degree of precision not yet reached in the understanding of decentralized economies. However, the most recent empirical evidence pertaining to Eastern Europe suggests that decentralized trade entails deeper questions that economic theory does not yet offer an answer to.

One of the central problems with which economic theory is concerned remained the same since Smith [1776]: How, why, and when does the `invisible hand' work? Many of the difficulties of decentralized economies seem to be related to problems of information and coordination between individual agents. It may therefore happen that individual agents make decisions that are rational for themselves, while, however, suitable coordination of these individual choices could lead to results that would be preferable for all involved. Clearly, in the former centrally-planned economies these questions were meant to be resolved by making the hand very visible and conspicuous; that was not very successful however. Now, while `perestrojka' has dismantled most of these institutions, and `glasnost' has made everything transparent, it seems that no hand is available, whether visible or not, and that all faith concerning the coordination of economic activities has simply been pinned to the working of completely free prices. Therefore, what is still necessary is a theory that analyzes coordination problems in a decentralized economy. This thesis should be considered as part of a research program whose aim is that of establishing such a theory. The thesis contains four essays on decentralized trade. Although the four essays deal with the same topic, and form a logical sequence, they can be read independently.

The first essay is a discussion of the Walrasian view on the problem of decentralized trade. Fundamental of that approach is the idea that a theory of decentralized trade must necessarily be based upon autonomous agents, i.e., by making assumptions only with respect to the individual agents' preferences and their physical environment. Many economists claim that Walrasian models are capable to explain both the behavior of individual agents and the overall outcome of their actions in a decentralized economy, in which all individual choices may be realized as planned. Actual deviations from such a remarkable and fortunate state, are said to be due to real world complications in the form of external effects, public goods, increasing returns, etc. We delineate the Walrasian perspective through an analysis of the structure of not only the basic Walrasian flexible price model, but also fixed price models, imperfectly competitive models and temporary equilibrium models. Those models are often called non-Walrasian, but we show that they share the same, Walrasian, perspective. We argue that the Walrasian approach in order to explain anything, has to take resort to concepts and structures that transcend the level of the individual agents, and does not succeed in proving what it claims. In other words, the structure of Walrasian models is such that they cannot be considered as ideal representations of decentralized economies. Hence, statements about the acceptability or optimality of decentralized trade cannot be based on theoretical results concerning them.

Our argument is not that the Walrasian models are unrealistic. Frequently these models are considered as a scaffolding for the construction of a theory of decentralized trade. Although this scaffolding is an impressive construction, firmly based upon axioms, we argue not so much that it is however only a scaffolding, but that it might not be the kind of scaffolding necessary for the construction of an economic theory of decentralized trade. We argue that a different methodological point of view, which pays explicit attention to the real interaction between individual agents in a decentralized economy is necessary.

In essay II a formal model of a decentralized economy is presented, in which we consider explicitly some forms of interaction between individual agents. In particular, we do no longer assume that all trade and communication take place centrally through `The Market', a public good kindly provided by the auctioneer, but instead, that the individual agents have to create their own markets. Some of the essential properties of decentralized trade taken into account are that individual agents have a limited knowledge of their economic environment, that such knowledge requires some kind of communication or interaction between these agents, and that individual agents need some information about other agents in order to meet potential trading partners. In this approach, a market is not a central place where a certain good is exchanged, nor is it simply the aggregate supply and demand of a good. A market is constituted by communication between individual agents.

We consider trade in a homogeneous commodity. Firms decide upon their effective supplies, and may create their own markets by sending a number of information signals communicating their willingness to sell. Meeting of potential trading partners is arranged in the form of shopping by consumers. Consumers visit only those agents of whom they are informed that these are firms possibly offering them a trading opportunity. We consider a Symmetric Nash Equilibrium in the firms' strategies. A necessary condition for existence of a SNE is that the production technology has decreasing returns to scale. Moreover, the costs of making a market and/or producing for the market must not be too high relative to the price of the commodity for each possible extent of the market at all levels of production. The latter might be due not only to the level of prices and costs as such, but also to a too small number of potential buyers in the economy or a too large number of competing firms. In a SNE the economy splits up in a number of possibly overlapping, imperfectly competitive markets. Trading opportunities are stochastic, and both firms and consumers may be rationed at the same time, i.e., markets are not orderly.

The developed framework appears to lend itself well to study many typical phenomena of decentralized economies, such as the emergence of central markets, the role of middlemen, and price-making. There are, however, also some limitations inherent to this approach. For example, we analyze a static, one-period model, and we assume that firms do have objective knowledge about the aggregate demand and the aggregate signaling activity. Typical of a decentralized economy is the locally interacting character of such a system, and the fact that most interaction takes place on a repeated basis. This puts forward important difficulties as to the modeling of an individual agent's actions.

Therefore, essay III discusses not the scaffolding, but the foundations of an economic theory of decentralized trade. The starting-point of economics is the homo oeconomicus who, given his preferences, pursues his self-interest, seeking to do the best he can. Often it is said that the fundamental characteristic of the homo oeconomicus is his rationality. Rational behavior in economics means that an individual agent chooses the most preferred action in his perceived opportunity set, where opportunities are defined such that all perceived costs and benefits are taken into account. Hence, rationality is necessarily constrained to be an essentially contentless notion in economics. We discuss several alternative approaches to rationality sometimes found in the literature, such as expected utility theory and theories of bounded rationality, and show that these are inconsistent with the economic approach.

By postulating that all actions of economic agents are rational, rationality is no longer a powerful, independent, explanatory factor of individual behavior. We argue that the homo oeconomicus is not the explanandum of economics. On the contrary, the rational individual agent, taking into account his preferences and perceived opportunities, is the explanans of economic theory. The substantive interest of economists is in the aggregate outcomes of the actions of many of these agents. Hence, a crucial point concerning economic models is the modeling of the agents' perceived opportunities. We argue that it may be fruitful to consider the perception of opportunities as an endogenous process, i.e., to assume that this perception depends strictly upon the preceding sequence of actions. Then, in a formal model, actions will be a function of perceptions, and perceptions a function of earlier actions. As a result, one gets a sequence analysis of actions in which perceptions or expectations do not appear explicitly, and one could model each rational individual agent's actions as a function of previous actions and outcomes.

In essay IV we analyze a dynamic model of decentralized trade that bears some resemblance to the model of essay II, with its explicit communication structure, incorporating the insights from essay III concerning the modeling of the actions of individual agents. A model of decentralized trade is considered with two types of agents: firms that produce a given commodity, and consumers who repeatedly wish to purchase one unit of that commodity. Consumers `shop around', while firms may attract the attention of potential customers by sending information signals. Trading opportunities perceived by the locally interacting individual agents change endogenously from period to period; either due to a change in underlying circumstances or to a change in the perception of these circumstances. Important is the interaction between these two, i.e., between the dynamics of learning and the dynamics of economic forces as such. We use a combination of Genetic Algorithms and Classifier Systems to model each individual agent as a `machine'. One of the important advantages of these Artificial Intelligence techniques is that they make it possible to analyze in how far the market provides sufficient structure to tie down the set of possibly perceived opportunities, thus constraining the behavior of the individual agents.

We run a simulation of the model with 50 firms and 5000 consumers for 2000 periods. A macroeconomic 'equilibrium' characterized by comparatively steady aggregates is approached, in which competition appears to lead to coordination of economic activities, communication by firms and patronage by satisfied consumers play an important role, and high communication expenditures are a significant source of macroeconomic inefficiency. The microeconomic distributions underlying the aggregates show strong differences between the market shares of firms, and between the shopping strategies of the consumers. Nevertheless, on average all firms offer an identical service rate, and the costs incurred per unit sold are also identical for all firms, while the rate of shopping success is equal for all consumers. The emerging regularities show that the market does tie down the set of possible beliefs of the agents, and does constrain their actions. Hence, further developments of this approach seem promising.

home top

Nick Vriend,
Last modified 2012-12-07