Abstract
The reasons for the lack of communication between economists and econophysicists are clarified by giving an outline of the methodological differences between economics and econophysics, which has emerged as a new field of research over the last decade. Theorists working in econophysics see economic and financial phenomena through the ‘lens’ of statistical physics. So far, there has been no real dialogue between economists and econophysicists because both communities adhere to their own methodological precepts.
Keywords
Introduction
Over the last decade, a considerable number of physicists has started applying concepts that originate in physics to economic phenomena. The term ‘econophysics’ is now used to describe these works. This new field of research is not a new ‘fashion discipline’ within the set of existing economic theories; rather, it is a new way of thinking about economic and financial systems directly inspired by statistical physics. A growing number of papers on econophysics have been published in journals devoted to physics and statistical mechanics. Several series of academic meetings (such as conferences1 and workshops) dedicated to this topic are regularly organized and, moreover, new PhD programs in econophysics have recently appeared in some universities. Today, econophysics appears to be a new step in the history and evolution of the physical sciences.
This study places the main differences between economics and econophysics in a methodological perspective. It shows that economics is an empirical and micro perspective discipline whereas econophysics is an empiricist and macro founded field. This kind of analysis allows us to understand better the lack of dialogue between econophysicists and economists.
Econophysics, a new field of research
The term ‘econophysics’, as a specific label and conceptual practice, was first coined by the physicist H. Eugene Stanley in 1996 in a paper published in Physica A (Stanley et al., 1996). As the name suggests, econophysics presents itself as a hybrid discipline, which can be defined as ‘a quantitative approach using ideas, models, conceptual and computational methods of statistical physics applied to economic and financial phenomena’ (Burda et al., 2003, 1). Although econophysics is mainly focused on financial economics, a growing body of work in macroeconomics is also emerging (see Shinohara and Gunji, 2001 or Donangelo and Sneppen, 2000). Econophysics is directly related to the development of the so-called ‘complexity science’ that emerged during the 1990s (Rickles, 2007). Economic systems are obvious candidates for treatment in terms of this ‘complexity science’ because they are composed of multiple components (agents) interacting in such a way as to generate the macro- properties of economic systems and subsystems (Rickles, 2008). These macro- properties can be characterized in terms of statistical regularities.
The influence of physics on economics is not new. A number of authors have studied the ‘physical attraction’ exerted by physics on economists (Mirowski, 1989), (Ingrao and Israel, 1990). But as McCauley (2004) points out, in spite of these theoretical and historical links between physics and economics, econophysics represents a fundamentally new approach that differs from previous influences. Its practitioners are not economists taking their inspiration from the works of physicists to develop their discipline, as has occurred repeatedly in the history of economics, when physicists have tried to give an economic meaning to the concepts coming from physics. Econophysics is very different, since this field is developed by physicists who are going beyond the boundaries of their discipline, studying various problems usually studied by social sciences, in light of their own methods. Econophysicists are not attempting to integrate concepts of physics into economics as it exists today.2 Instead they are seeking to ignore, even to reject this discipline in an endeavour to replace the theoretical framework that currently dominates it with a new framework derived directly from statistical physics. ‘Econophysics is not like academic economics. We are not trying to make incremental improvements in theory; we are trying instead to replace the standard models with something completely new’ (Mirowski, 1989, 5).
Therefore, one might ask to what degree economics should be affected by the emergence of this new field. The answer is quite simple: economists cannot be indifferent to the emergence of econophysics, because this new approach claims to study economic reality and it offers new predictions about it. Since models coming from econophysics often provide better predictions, they could take the place of economic models in the future.
Indeed, since Levy processes can be seen as a theoretical generalization (Schoutens 2003) of the improved Gaussian frameworks used in finance, econophysicists can find, in specific conditions, the same results as those offered by financial economists. In this perspective, stable Levy framework provides an adequate solution3 to capture observed market effects (anomalies such as fat tails or volatility clustering) that are not explained or predicted by the mainstream. Of course, economists know and use Levy processes, but they only work with nonstable Levy processes because they have a finite variance.
Stable Levy processes have been used in economics and finance in the 1960s but their statistical properties (infinite variance) led economists to abandon these types of models. As explained Fama (1976), ‘the cost of rejecting normality for securities returns in favour of stable nonnormal distributions are substantial and it behoves us to investigate the stable nonnormal hypothesis further ’. In order to describe the leptokurticity4 of financial distribution, economists decided to incorporate within the existing models the possibility that extraordinary events can occur. However, econophysics is not an incorporation of a ‘leptokurtic dimension’ but rather an attempt to use a specific model which has been abandoned in the 1960s by economists. Stable Levy processes have been reintegrated into economists by econophysicists who developed what we called truncated Levy processes offering a stable Levy processes with a finite variance.
Moreover, for some years, new degree programs devoted to econophysics have been developed by physics departments in several universities (Kutner and Grech, 2008) and we are now seeing the first reflections on the potential contributions of econophysics to political decision-making (Artemi, 2008). So far, there has been no real dialogue between economists and econophysicists because both communities still adhere to their own methodological precepts (Farmer and Lux, 2008). By giving an outline of the methodological differences between the two disciplines, this study aims to clarify the reasons for the lack of communication between economists and econophysicists (Rosser, 2003).
A data-driven approach
Econophysicists claim to describe the world as it is, and they consider that economists have developed an understanding based on false a priori beliefs about the social world. Within this context, econophysicists reject such concepts as ‘utility’, ‘perfect rationality’, and ‘equilibrium’ because they are seen as inappropriate a priori beliefs. According to Keen (2003), these key concepts have no empirical basis. The concept of equilibrium, for example, is a central organising principle in economics.5 It appears as an a priori belief that provides a ‘standardized approach and a standardized language in which to explain each conclusion’ (Farmer and Geanakoplos, 2009, 17). More precisely, the concept of equilibrium is used by economists, primarily, as an analytic tool.6
Although, this concept was brought to economics from physics,7 econophysics is mainly derived from contemporary statistical physics in which the concept of equilibrium does not necessarily play a key role.8 Of course the notion of equilibrium is often used by econophysicists, but not necessarily in the a priori manner that economists use it. In econophysics, this equilibrium is regarded as a potential state of the system and not necessary as the goal of the system.9 For econophysicists, ‘there is no empirical evidence for equilibrium’ to be seen as a final state of the system (Mirowski, 1989, 6) and equilibrium is then considered as a state of the world rather than used as an analytical tool.
In line with their empiricist approach, econophysicists also reject all forms of normalization or data mining often used in economics in order to standardize data (Backhouse and Morgan, 2000). By data mining, I mean the activity of fitting a wide variety of models to the data in the opportunistic hope of finding one that fits well.
Data mining is a rational attitude: Why, for example, should we take into account abnormal data observed very rarely in the past? Data mining can then be seen as a ‘filter ’ as Hoover and Perez (2000, 207) wrote, ‘the lenses and filters of the astronomers or the statistics of the econometricians are in fact the ones that would reveal the aspect of the truth that interests us’. Despite the rational justification for data mining, econophysicists do not use data mining to filter data and they strongly reject this kind of ‘apriorism’ (McCauley, 2006). For them, there is no ‘abnormal data’ but only data about reality. However, let us mention that, regardless of what they might assert, the idea that econophysicists do not have any a priori beliefs about the world is, paradoxically an a priori proposition itself (Schinckus, 2010). From this, we can infer that, whereas economics is an empirical field, econophysics is, by contrast, an empiricist discipline. The distinction between an empirical and an empiricist perspective refers to the use of data: although both perspectives are based on data, the perspective of econophysics entails the more physical belief, that data must not be standardized but rather used as they have been measured, whereas the perspective of economics recognizes that data mining can be useful in order to have significant data.10 Moreover, empiricism leads to a physical reductionism that can be found in the works of econophysicists when they claim that only the language and the methodology of physics should be used in describing complex economic systems.
Uncertainty and risk are two key concepts in economics and the distinction between them has been proposed by Knight (1921). Risk is quantifiable and it describes a situation which is governed by a well specified probability distribution. On the contrary, uncertainty refers to a situation in which random outcomes cannot be attributed to a probability distribution. In economics and especially in financial economics, uncertainty is often reduced to risk which is measured by a statistical moment. In a sense, models coming from econophysics may also broaden the notion of uncertainty (risk) in economics (Schinckus, 2009). Indeed, econophysicists have developed this concept in several ways. They often use the concept of entropy to characterize the idea of uncertainty [I should say risk] in economics and finance. As Dionisio et al. (2005) explain, ‘entropy is a measure of dispersion, uncertainty [risk], disorder and diversification used in dynamic processes, in statistics and information theory; and it has been increasingly adopted in financial theory’. The application of entropy to economics is relatively recent and it provides a diversity of models for the study of uncertainty: Gibbs entropy (Chakrabati and Kajal, 2000), Tsallis entropy (Takahashi, 2007), Shannon entropy (Dionisio et al., 2005) or Rényi entropy (Brody et al., 2007). Dionisio et al. (2005, 161) write that ‘the use of entropy as a measure of uncertainty [risk] in economics appears to have many potentialities and a vast field of development, both in theoretical and empirical work’. Unlike economists who often reduce uncertainty to risk given only by one perception (statistical dispersion), econophysicists use this concept in a more explorative way by studying several of its dimensions.
Each of these entropies defines a particular statistical framework that can be used as a function of the reality being studied. By providing this diversity, econophysicists offer a collection of operational instruments in conditions of uncertainty. All these explorative approaches can enhance the existing economic models of uncertainty.
A macro perspective approach
Economists and econophysicists focus on mathematical modelling at the expense of a realistic depiction of human motivation and action. But there is a deep difference in the way they model economic reality. Both disciplines use a similar atomistic approach. However, economics is biased towards methodological individualism and economists focus on static characteristics of individuals by analyzing their behaviour in terms of personal characteristics (expressed in terms of utility functions, risk aversion, etc.). Within this framework, mainstream economists focus their attention on individuals and, often, they ignore their interactions. Individuals are thereby ‘disembedded’ (Granovetter, 1985) from the systems in which they act. In neoclassical economics, rationality appears to be fundamentally causal and explains the (microscopic) behaviour of each individual (Mongin, 2002) without taking into account the interaction with the economic system (Colander et al., 2008, 8).
Adam Smith, who was a founding father of neoclassical economics, claimed that, by pursuing his own self-interest, an individual would be led, as if by an invisible hand, to promote the wider interests of society. Their faith in this invisible hand may have discouraged neoclassical economists from studying the interactions of individuals. Instead, they have focused their attention on models of individual rational choice. As a result, neoclassical economic theory has little to say about such matters as systemic financial risk or the origins of financial crises.
Because economic activity is interactive in essence, the perspectives of econophysics are appropriate for understanding the connections between the component parts of economic systems (firms, banks, households). Moreover, the various characteristics of these components can be attributed to their position within the economic system, which governs their interactions with other elements.
Econophysicists are prepared to regard market participants (including traders, speculators, and hedgers) as unthinking atoms that obey statistical laws. Therefore, they can avoid the difficult task of theorising about the individual psychology of investors (Brandouy, 2005). Econophysicists do not care about rational agent theory and they take no account of the personal characteristics of the individuals. Therefore, they can offer only a macro description of the system (i.e. financial markets). In this perspective, an approach that appears to be a fruitful one in modelling the behaviour of financial markets examines the interactions of cellular automata. Each of these automata has an embedded set of decision rules and all interactions of the automata do often produce results that are characteristic of financial markets.
Donangelo and Sneppen (2000), as well as Shinohara and Gunji (2001), have explained the emergence of money by studying the dynamics of exchange in a system composed of many interacting heterogeneous agents. Whereas Donangelo and Sneppen (2000) have used non-Gaussian statistics to quantify fluctuations in exchanges with particular reference to anomalous Hurst exponent,11 Shinohara and Gunji (2001) developed a reciprocity model in which the interactions between agents are, on one hand, asynchronous12 and on the other hand, specific to the type of agent.
This diversity of interactions and possibility of applying them to economics could lead economists to complete the existing economic models in order to understand better complex interactions between all elements (firms, banks, households) of economic systems (Farmer and Geneakoplos, 2009). In this perspective, econophysics can then be seen as a complementary field to economics.
Conclusion
The methodological gap emphasized in this paper allows a better understanding of why econophysicists and economists have no real dialogue: beyond their theoretical differences, they have a very different epistemic approach to the world. Economics is an empirical and micro perspective discipline whereas econophysics is an empiricist and macro founded field. Even if this lack of dialogue between economists and physicists can be connected with the closed nature of these two disciplines,13 some authors (Jovanovic and Schinckus, 2010) claim that a dialogue and perhaps even a theoretical connection could be achieved in the future. If that happens, the theoretical and practical consequences will be real, because the framework they have developed describes uncertainty in a very different way from the theoretical explanations proposed by neoclassical economics. In this perspective, all economic decisions and performance measures, therefore, are directly affected by the emergence of econophysics and the aim of this study was to present the main differences between this new field and the economic mainstream.
Footnotes
1
For example, the Annual International Conference on Econophysics or the Annual Colloquium in Econophysics.
2
During past decades, many physics models have been used in economics but these models were mainly used for their mathematical description of physical phenomena. Progressively, these imported models have been integrated in the mainstream (for example the Black and Scholes model). This trend is not observed with econophysics in which economic phenomena are explained in terms of molecular mechanisms or complex interactions. In this perspective, econophysicists do not try to connect their works with the pre-existing economic theory. For an epistemological analysis of this attitude, see Rosser (2003).
3
Clegg (2006) discusses the integration of this empirical fact by econophysics in greater detail.
4
Leptokurticity has generateda lot of debates in finance. For the 1990s, financial economists have progressively begun to study Levy processes which include many classes of stochastic processes such the Wiener process, jump-diffusion processes and pure jumps processes (like Variance Gamma Process, Madan, 1990; Generalized Hyperbolic Process, Eberlein, 1995 or CGMY process, Carr et al., 2000). The class of Levy processes is so general that it refers to several kinds of processes leading economists and econophysicists to use the same word for characterizing different models. Whereas economists use the ‘Levy process’ to define (nonstable) jump diffusion and (nonstable) pure jump models, econophysics use this word to define stable Levy processes. This semantic difference is a source of debates between econophysicists and economists. The first often claim that they offer a new perspective on finance whereas the latter consider that this approach is an old issue in finance. The point is that these two categories of specialists do not talk about the same thing, econophysicists used stable Levy processes whereas economists abandoned these kind of processes in the 1970s (see Fama, 1976 and Jovanovic and Schinckus, 2010 for a historical explanation of this abandonment).
5
The term ‘equilibrium is used in all kinds of contexts: in a Keynesian economics or a Marshallian framework and we would rather deal with the notion of ‘disequilibrium’ if we characterized the Austrian tradition. In this study, I deal with the ‘neoclassical equilibrium’ related to the mainstream economics and which was formulated originally by Walras and developed by Arrow and Debreu. See Farmer and Geanakoplos (2009).
6
Equilibrium has the same status as the particular solution of a differential equation, when this is used in analysing a simple dynamic system. In using such concepts, one need make no assumptions about the rates of convergence nor does one need to make any assumption about the nature of the forcing function, which may have the effect, in the main, of keeping the system a long way from a position of equilibrium (I thank the referees for this clarifying comment).
7
More precisely, the two physical frameworks which directly influenced the notion of economic equilibrium are Newtonian mechanics and thermodynamics. In this perspective, the economic and physical equilibrium appear to have a lot of similarities in their way of describing the phenomena: in both cases, the mathematical tool is optimization with constraint using the method of Lagrange multipliers. But this mathematical isomorphism does not exist with equilibrium in statistical physics. See Mirowski (1989) or Smith and Foley (2008).
8
When econophysicists deal with equilibrium, they use rather a ‘statistical equilibrium’ coming from statistical mechanism (i.e. reconciliation between mechanism and thermodynamics). See Bouchaud (2002).
9
Of course there are some works about disequilibrium in economics but the notion of equilibrium remains the main theoretical reference for these works (there is a disequilibrium because we can evaluate an equilibrium situation). Econo-physicists rather work with nonequilibrium situations.
10
This is a sin to which all scientific disciplines are prone even in physics. However, econophysicists claim, in a more empiricist perspective, to use data as they ‘find’ them.
11
The Hurst exponent refers to the scaling property of the fractal Brownian movement. This statistical parameter is generally used to describe time-dependent phenomena.
12
Each exchange needs a reciprocal situation and must have a particular duration.
13
Pieters and Hans (2002) explored intra and inter-disciplinary communication of economics journals by means of citations analysis. They showed that the first tier of economics journals did not cite articles published in journals of management, marketing, anthropology or psychology between 1995 and 1997. Moreover, according to the Science and Engineering Indicators (Foundation 2000, 103, table 6–54), economics is the most hermetic field of the social sciences, citing more than 87% of intradisciplinary references compared with 50% in sociology for sociology. This is even more self-contained than physics, which cites physics journals in about 80% of references (see Gingras and Schinckus for a bibliometric analysis of the emergence of econophysics).
Christophe Schinckus is Assistant Professor in Finance at the School of Management — University of Leicester. He is also visiting scholar at the London School of Economics.
Correspondence to: Christophe Schinckus, School of Management, University of Leicester, University Road, Leicester, LE1 7RH, +44 (0) 116 223 1788; Email: