Economic modeling and complex thinking
The case of financialized commodity markets
By Bertrand Munier, Professor, Institut d’Administration des Entreprises de Paris and GRID-GREGOR
Affiliate Professor, New York University Polytechnic Institute
Chief Economist, momagri
We are reproducing below excerpts from an article by momagri Chief Economist Bertrand Munier that was published in the study “Agir et Penser en Complexité with Jean-Louis Le Moigne – Témoignages de mises en actes” conducted by D. Genelot and M.J. Avenier1. As it showcases the limits of traditional economic tools to understand financialized commodity markets, the work provides much insight for the current discussions on agriculture, whose markets are particularly unstable.
Momagri Editorial Board
“The revolution in Finance [happened] as it was realized that asset prices
show large volatility that does not reflect anything about fundamentals.
[…]. Macroeconomics [did not] keep up with [this] revolution,
[…] to the great detriment of contemporary macroeconomics”.
Former United States Secretary of the Treasury
Professor and President Emeritus, Harvard University
Interview given to The Financial Times, 2008
Traditional economic analysis represents all markets as tools to adjust consumption plans to production plans. The most basic version of such model is the static Walrassian version. In this almost universal version of straight competition in perfect markets, every good or service has a unique equilibrium price, which matches global supply and demand. Such amazing result is evidently due to the hypotheses put forward, in particular that regarding the “trial and error” work leading to a price determination before consumers and producers are allowed to trade. It is thus not possible––given the assumptions of “price takers” in this market system––to have multiple prices for a same good or service. The static version can be limited to this summary description.
In the inter-temporal version of the model, one must also explain how players can sense, through anticipation, prices for the following cycles, because otherwise the model would be underdetermined. The logical extension of the approach then demands that each player has the ability to anticipate the equilibrium price for the following cycle. The added extension regarding risk is immediate: One simply assumes that each player anticipates the probability distribution of the price for the following cycles or, in a slightly lighter version, the hope for such distribution. This is the assumption of rational anticipations. In such a system, any price fluctuation necessarily has an exogenous cause––modification of available resources, of production start-up technology (through natural hazard in agriculture for instance), or of consumers’ preferences.
Alternatively, in a more “dynamic” version where the end of each cycle plays out the equilibrium of the next one, we can also conceive a temporary “error” of anticipation by at least one player, the price then fluctuating––at least for the given cycle––and the model’s intrinsic negative feed-back driving the system back to equilibrium, at least in general cases. Experience shows that such models can only generate, in general macroeconomic equilibrium (the so-called “computable general equilibrium dynamic” models), limited price fluctuations of approximately three to five percent from one cycle to the next. This is much higher than what we are observing in automotive markets, but far lower than what is recorded in mineral or agricultural commodity markets, whose price annual averages have been wildly fluctuating by 25 to 30 percent since the 2003 and 2004 years. Between March and early December 2008, wheat prices declined by close to 60 percent in daily quotes, and even more in spot quotes. There is a serious gap between the observation of commodity markets and their models provided by the micro- or macro-economic Walrassian general equilibrium of supply and demand. With regard to partial equilibrium––a deterioration of the previous one––of global supply and demand for a given product, it only allows for pseudo-scholarly and after-the-fact forecast for the markets considered here. This was also the case for the American mortgage market, the so-called “subprime” crisis, or the economic downturn itself.
As we broaden the study of commodity markets, it seems obvious to acknowledge––as did Summers, Greenspan, Eichengreen, Lux, Rogoff, Stiglitz and many others––that traditional economic tools do not permit to understand the development of not only financial asset markets––a long-known fact––but also of financialized markets, such as real estate markets in some countries, and, especially since the early 2000s, the major commodity markets. The objective of the following section is to stress why the concept of complex system then allows providing an analysis framework to improve the comprehension of such markets.
We can draw two types of conclusions from the above-mentioned reasoning, one of an epistemological and historic nature of economic thinking, and another of a practical nature in more immediate terms.
In practical terms, we will first note that economic forecast as per the 1960s meaning is, for the type of asset markets considered in this article, highly dependent on the course of the system. Any bifurcation will render illusive a forecast based on usual econometric methods––regardless of their sophistication. Some official organizations would probably be wise to invest greater resources elsewhere than in such type of misconception. We are also aware of the instability of variances and co-variances in financial markets, and of the controversial nature of a “mean reversion” for economic quantities. The psychology of markets would indeed deserve a wider audience.
Today, the major issues for researchers include the type of expected volatility, of diagnostic and, if possible, of forecasting the type of phenomenon of anticipation polarization, and, lastly, of possible market turning points. First let us observe that, on these very issues, the mainstream economic analysis is silent. In addition, we have a paucity of work on these topics. We need global general interaction models regarding the volatility expected from given economic and/or customs policies. Such models do not exist, even if acceptable substitutes were designed here and there. Thus the momagri model (2009, then 2010-11, until an improved version 3)2.
There is very little specific material on the forecast of expectation polarization, as the data is more related to cognitive psychology and social psychology. Regarding forecasting eventual turning points, the first advances were derived from physics (Sornette, 2003). These are nevertheless crucial issues for both economic policy makers and market players.
From an epistemological standpoint, complex thinking is primarily a method that generates a critical appreciation of the many recent developments in economic analysis––information asymmetry, risk and uncertainty analysis and formation of anticipations, among others––occurrences that sometimes are heading in the same direction as this complex thinking, even if some preceded it here and there. Of course, we must resist comparing THE “economic thinking” (broadly set at the end of the 19th century/early 20th century with the reasoning by Walras and Pareto, thus ignoring the developments occurred during a century), and THE “complex thinking”, whose most developments came many decades later and profited from a scientific experience that cannot compare with that of the previous thinking. In compliance with this condition, the contribution of complex thinking will not be considered as means to castoff all current economic models, but as a source for the rebuilding and potential enrichment of traditional models (Smith, 2002; Markose, 2003) for economic sub-groupings (markets, economies and branches) to which the latter are not adapted. That is what we tried to show in this article regarding the cases of financialized commodity markets. We can probably learn even more to handle, as much as possible, future crises that would be similar to the ones that peppered news reports, from the 1980s to the 2008 major downturn.
This is the bet we took with Jean-Louis Le Moigne in 1976 when we worked together on the basis of our respective research on the decision du create the GRASCE3 (Le Moigne, 1974; Munier, 1970, 1972). Admittedly, many locks have yet to be broken, but the bet has not only lost his timeliness, but been given an increase powerful value due to the events over the past few years.
1 Chapter 14, pages157 to 168, L’Harmattan publishers.
2 The acronym momagri stands for “MOuvement pour l’organisation Mondiale de l’AGRIculture”. It is a Paris-based but internationally defined think tank. The model, which was designed to inform international policy makers on global agricultural strategies, is actually a system of models organized around a classic general equilibrium model bordered by partial models generating equilibrium changes that approximate successive stages of imbalance. Additional details are available from a first article (Munier, 2010) or a published work (Munier, ed., 2011). Some 15,000 equations are providing a reasonable meaning acknowledged by the practitioners currently involved in international agricultural economic monitoring.
3 Initially named Groupe de Recherche en Analyse de Système et Calcul Economique (1976-1992). The acronym has been kept but added with the later terms we currently know. However, the intended combination of economic and complex issues remains prevalent within the team.