Soros’ Theory of Reflexivity: a critical comment

Tony Lawson

Cambridge University, Faculty of Economics, Sidgwick Avenue, Cambridge, CB3 9DE, UK. Tony.


In a recent book George Soros (2009)1On March 20th 2010, three weeks or so before attending a conference hosted by George Soros’ Institute for New Economic Thinking (in … Continue reading proposes a new paradigm for economic thinking. In doing so he reveals himself to be somewhat wary of the emphasis of contemporary mainstream economists on methods of mathematical modelling, and stridently critical of certain prominent substantive economic theories, most especially those developed under the heads of rational expectations, the efficient markets hypothesis, and equilibrium theory.

Soros’ contribution, I believe, contains insight. However in arguing his case Soros draws several questionable contrasts and these likely detract from the force of his insights. In this short note I briefly question some of Soros’ background claims and assessments, and examine the contribution that remains once, or if, various unsustainable, and in fact unnecessary, claims are jettisoned. A critical commentary on the sustainable part of Soros’ contribution then follows.

A central argument here is that Soros’ criticisms do not quite go far enough. Though the shortcomings may seem minor, they bear implications regarding the sorts of contribution likely to be encouraged within Soros’ recent venture, systematised as the International Network for Economic Thinking (INET). Despite Soros’ intentions in sponsoring INET to improve the state of modern academic economics, I fear that unless the nature and implications of Soros’ insights are better recognised, the net result, at the level at which it matters, i.e., that at which the discipline goes awry, will just be more of the same, a perpetuating of the most fundamental errors.


George Soros’ central contention, the core of his proposed ‘new paradigm’, is advanced under the head of reflexivity. Human beings, Soros observes, seek both 1) to understand reality and 2) to influence reality. He refers to the former as the cognitive function and to the latter initially as the participating function and, more recently, as the manipulative function. The essential point for Soros is that there is a two way exchange: 1) from reality (including humans and their interactions) to our understanding of it, and 2) from each human being armed with an understanding, back to (the rest of) reality. For Soros the term reflexivity is basically used to express or capture this two way relationship.

Soros’ concern is that both functions of this two-way relationship can be in operation at the same time, leading them to interfere with each other:

When both functions operate at the same time they may interfere with each other. For the cognitive function to produce knowledge it must take social phenomena as independently given; only then will the phenomena qualify as facts to which the observer’s statement may correspond. Similarly, decisions need to be based on knowledge to produce the desired results. But when both functions operate simultaneously, the phenomena do not consist only of facts but also of intentions and expectations about the future.

(Soros 2009, p. 4)

Soros’ central contention does indeed contain insight, and it takes us beyond the standpoint of much modern economics. But there is a need for some clarification.

The nature of the problem

It is important to appreciate the real nature of the problem to which Soros is pointing. In effect Soros is identifying factors that contribute to rendering social reality inescapably open; he is indicating that the future, because dependent on what we are all doing, not least in anticipating and reacting to each other, is yet to be determined and inherently unpredictable.

I shall use the term closure or closed system to express a situation in which an event regularity, or correlation occurs; it is a condition for successful prediction. In the absence of conditions supporting event regularities the system can be said to be open.

The future, and indeed the present, of social reality are indeed open and unpredictable, and Soros emphasises reflexivity as one of the reasons why. However, in correctly identifying the feature of reflexivity Soros wrongly supposes thereby both that assessments about social phenomena are necessarily more fallible or imperfect than claims about natural phenomena, and also that a science of social phenomena and/or capable everyday human activity, are impossible.

Very briefly, all knowledge is fallible and imperfect whatever it is about. Even the seemingly most explanatorily powerful of natural scientific claims have usually been found to be inadequate in some ways in due course. Soros is surely right to emphasise that “our understanding of the world in which we live is inherently imperfect”. But it confuses matters to suggest that such imperfection arises “because we are part of the world we seek to understand” (p. 3). Not only is all knowledge fallible, and our understanding of other worlds, say outer space, or that of the dinosaurs, imperfect, but it is not clear that our having being part of anything necessarily complicates our understanding of it relative to our understanding of anything else. Thus medical research into the human body is often seemingly comparably successful to some of the non-human or non-social sciences. Moreover it can reasonably be argued that we often have a greater insight into the social realm, i.e., that realm of phenomena whose existence necessarily depends at least in part on us, just because it depends on us, and is intrinsic to our activities.

We in fact find ourselves to be extraordinarily capable beings qua social beings, very adept at understanding and coping with the everyday complexities of social life. We are very skilful at acquiring language, negotiating the market place, motorways, forms of international travel, and communications, new technologies, the host of human institutions, and so forth. In fact, on examination, our ability to navigate a rather complex social reality is revealed to be really quite extraordinary (requiring many years of training).

How come we are so skilful at negotiating the world around us given that the latter is essentially open? We are so just because reality, both (the non-social) natural and social, is also structured. Although both realms are open in the sense that, away from experimental laboratories, event regularities (regularities of the form whenever event x then event y) rarely occur, behind surface phenomena in each domain lie deeper causal forces. Indeed it is often just because a multiplicity of causal forces governs any outcome that regularities at the level of outcomes or events rarely occur. Now the underlying causes in (the non-social) natural and social realms alike will often be relatively enduring, and frequently identifiable. For example, underlying and governing the paths of autumn leaves are causal factors such as gravity and aerodynamic forces; underpinning the symptoms of mad cow disease is the prion; and underpinning social practices are social structures such as collective practices, rule systems, social relations, and the like. And it is through seeking knowledge of these underling structures that science and capable human practice alike are feasible.

Science, I have elsewhere argued (1997, 2003, 2010) is characterised precisely by the move from surface phenomenon of interest to identifying the underlying cause(s). In other words, its central goal is causal explanation. It is because the social realm, like the (non-social) natural realm, is structured that this move is as possible in the former realm, rendering a social science entirely feasible.

It is also precisely because reality, both (non-social) natural and social, is so structured, that capable everyday behaviour is feasible, despite outcomes or events being unpredictable in advance. For underlying and grounding social practices are causal structures such as social rules, and relations. And capable human agency is possible through people drawing on their knowledge of underlying social (and natural) structures, and adapting as they go along.

Thus when, for example, an individual embarks on a motoring journey, it is usually enough to know the rules of the Highway Code, how to drive, the causal forces represented by other vehicles, the way gravity operates, and so forth. Specifically, it is not necessary to predict in advance each configuration of vehicles etc., to be encountered throughout the journey. The latter feat would usually, of course, be impossible, because, despite its highly structured nature, even the motorway is ultimately an open system.

So what is the problem?

If openness of social (and non-social natural) systems proves not to be undermining either of the possibility of social science or of capable day-to-day human agency, what is the problem to which Soros is pointing with his notion of reflexivity?

I have indicated that social science and daily activities are viable because they are concerned with, and draw upon, understandings of the deeper structures. In particular human daily activities are conditioned upon these, and result from our abilities to adapt to immediate circumstances.

The sorts of problems that most interest Soros, however, arise because of a concern not (or not just or primarily) with (identifying and understanding) underlying causal structures, but rather and centrally with predictions of specific concrete outcomes. These outcomes are exactly the sort of phenomena to which financial speculation and investment are oriented; the focus is typically on future relative prices or relative price movements. However, because the financial system is open—and as we have seen Soros’ insights systemised under the heading of reflexivity give reason to suppose that the financial system is in fact as open as any—such concrete outcomes are just the sort of thing that cannot typically be known or successfully predicted in advance, even if or when a knowledge of some significant underlying structures has been achieved.

Soros’ criticisms

We are now in a position to more clearly understand the real nature Soros’ concerns relating to modern economics. Despite Soros various suggestions to the contrary, the fundamental problem to which he draws attention does not depend on any natural/social distinction or on a science/non-science opposition. If there is a contrast of relevance to be drawn it is between open and closed systems. Soros’ concern boils down to the impossibility of successfully anticipating concrete outcomes in an open system, except maybe by chance, coupled with an apparent failure of many academic economists to recognise this.

In open social systems, such as the financial world, uncertainty is always rife; and specific outcomes such as relative prices cannot possibly be known in advance. Yet, as Soros notes, modern mainstream economics is awash with assumptions of perfect foresight or rational expectations, of efficient markets, and of market equilibrium, all, at least in the manner they are typically employed, essentially premised on the successful predictability of future outcomes; all supposing that in an open system actual outcomes can be effectively anticipated. Indeed all such endeavour in effect treats open systems as if they are actually closed.

This latter situation is the central focus of Soros’ consternation. And his criticisms remain legitimate, despite the questionable nature of some of the contrasts he makes in motivating them. Thus Soros is correct when he writes:

I contend that rational expectations theory totally misinterprets how financial markets operate. Although rational expectations theory is no longer taken seriously outside academic circles, the idea that markets are self correcting and tend towards equilibrium remains the prevailing paradigm on which the various synthetic instruments and valuation models which have come to play such a dominant role in financial markets are based. I contend that the prevailing paradigm is false and urgently needs to be replaced.

The fact is that participants cannot base their decisions on knowledge. The two-way, reflexive connection between the cognitive and manipulative functions introduces an element of uncertainty or indeterminacy into both functions. That applies both to market participants and to the financial authorities who are in charge of macroeconomic policy and are supposed to supervise and regulate markets. The members of both groups act on the basis of an imperfect understanding of the situation in which they participate. The element of uncertainty inherent in the two-way reflexive connection […] cannot be eliminated, but our understanding, and our ability to cope with the situation, would be greatly improved if we recognized this fact.

(Soros 2009, pp. 6–7)

Elsewhere Soros in similar vein, and with good reason, takes issue with the ‘efficient market hypothesis’:

The prevailing interpretation of financial markets—the Efficient Market Hypothesis (EMH)—has been well and truly discredited by the Crash of 2008. The current financial crisis was not caused by some exogenous factor—like the formation or dissolution of an oil cartel—but by the financial system itself. This puts the lie to the assertion that financial markets tend towards equilibrium and deviations are caused by external shocks. But the alternative theory of how markets work that I am proposing—the theory of reflexivity—has not taken its place. It has not even received serious consideration by the economics profession.

(Soros 2009, p. 216)

Critical commentary

It could reasonably be asked why I have bothered to run through all this if Soros’ main claims are found to stand up, despite some questionable background assessments offered in their support.

A first answer is precisely to pre-empt any (or combat any unvoiced) rejection of Soros’ insights on the grounds that the arguments made in their support are suspect.

But actually I want to go further. By clarifying the grounds of Soros’ conclusions I think it is possible to show that Soros mostly does not go far enough in his criticisms; or that when he does so, there are grounds for his doing so in a rather bolder fashion than he appears hitherto to have felt is justified. Amongst other things, the comments which follow are significant for questioning whether Soros’ proposed new International Network for Economic Thinking is, in seeking to improve the state of modern economics, truly getting at the crux of the problem.

It seems evident from the passages just reproduced, that Soros takes theories and/or assumptions like “rational expectations” and the “efficient markets hypothesis” to be serious claims about the economy, albeit false ones. This is understandable given that he is a practitioner in financial markets, and is on the receiving end, as it were, of the activities of those who follow economic theorising.

However, it should by now be clear that these sorts of theories are not so much causes of the sorts of failings of modern economics with which Soros is concerned, as manifestations. Because social reality is, for the sorts of reasons given by Soros in his theory of reflexivity, everywhere open, and because the emphasis on modelling and event prediction presupposes that social reality is everywhere closed, the sorts of theories that can be entertained by mathematical modellers are to significant degree necessarily unrealistic. Whatever conceptions that modellers come up with, core features will necessarily be fundamentally implausible.

To formulate matters somewhat more generally, the emphasis on mathematical modelling and event prediction in modern economics, presupposes that closed systems of event regularities or correlations everywhere occur (notice, incidentally, that these closures are as much presupposed or required by the ‘newer’ approaches to mathematical economics, those often referred to as non-linear modelling, complexity modelling, agent-based modelling, model simulations, [most of] behavioural or neuro-economics, and so on, as they are by the more traditional forms of micro, macro and econometric modelling). Such event regularities would be guaranteed is social reality consisted of systems of isolated atoms. By atom here I mean an entity that exercises its own separate, independent, and invariable effect, whatever the context. By positing such atoms it follows that each time some triggering event or condition X occurs the same induced response, Y say, follows, and this response is the actual outcome unless some other countervailing force interferes; to theoretically discount the possibility of such an interference is the role of the assumption that the atoms act in isolation. Although, the twin assumptions of atomism and isolationism are not strictly necessary, they do provide the guarantee that most modellers seek, and it is difficult to think of any other assumptions that can do the trick. In any case it is easy enough to see that it is theories implicitly couched in terms of worlds of isolated atoms, that dominate the actual economics literature (see e.g., Lawson 2003 for an extensive analysis of all this.)

There are many reasons for supposing that the presumptions of atomism and closure are invalid as characterisations of the basic nature of social reality. Soros’ account of reflexivity, as noted, is a specific, if important, example. It is because all economic theories attached to formalistic modelling endeavours must be consistent with worlds of isolated atoms that all are necessarily unrealistic. The problem is not intellectual failing or ideology at the level of economic theory, but philosophical failing or ideology at the level of scientific method and (implicitly at least) social ontology

In fact, it is important to appreciate that the theories of which Soros is so critical were originally produced not because their formulators sought to capture something about the economic world but merely because they capture a form of internal consistency at the level of modelling practice.

For example, John Muth’s (1961) intention in forming the rational expectations hypothesis, was merely to constrain the predictions attributed to human individuals or firms that are modelled to be consistent with the predictions generated by the economic model within which the same agents are theorised. As such the proposal is more a technique for consistency in modelling than anything else. As Muth himself formulated the ‘hypothesis’:

expectations, since they are informed predictions of future events, are essentially the same as the predictions of the relevant economic theory […] The hypothesis can be rephrased a little more precisely as follows: that expectations of firms (or, more generally, the subjective probability distribution of outcomes) tend to be distributed, for the same information set, about the prediction of the theory (or the “objective” probability distributions of outcomes).

(Muth 1961, p. 316)

Any assertion that these expectations (and model in which they are imposed) are essentially correct, is a step that is additional to assuming rational expectations. Actually, just because social reality is open and economic models are constructed on the presumption that it is closed, it is no surprise that typically the projections of formalistic models are found everywhere to be wide of the mark. It follows that the processes of rendering the expectations of agents modelled consistent (in the noted sense) with these model projections ensure that these agents’ expectations too are typically wide of the mark. Thus the presumption by so many commentators that the hypothesis of rational expectations is effectively equivalent to assuming that expectations are accurate is ironic in the extreme.

The hypothesis, I repeat, is not about realisticness but consistency. If the modeller changes the model specifications, and the modelled expectation generating mechanism changes too, to retain consistency. That is about all that can be said of the hypothesis.

Of course, certain ideologically inclined economists have sought to exploit this property. Specifically models have been formulated in terms of variables with certain values termed ‘natural rates’, and where, by assumption/design, the variables in question are allowed to depart from the natural rates only through the modelled agents making mistakes. Thus, as part of the model’s design specifications governments are able to influence the variables in question, to get them to depart from natural rates, only through fooling people. Into this framework rational expectations are introduced rendering ineffective even this limited role granted to government. But all this once more constitutes theoretical specifications additional to the rational expectations hypothesis itself.

In similar fashion any interest in equilibrium theorising in modern economics is motivated by concerns that have little to do with relevance or realisticness and everything to do with modelling consistency (see e.g., Lawson 2005, 2006). Economic models often comprise sets of equations, each of which is notoriously found to have little relation to what happens in the real world. Despite their lack of realisticness, however, one question that keeps economists occupied with such models, is whether the different equations of a system of such formulations are mutually consistent—in the sense that there ‘exists’ a vector of values of some variable, say prices, that is consistent with each and all the equations. Such a model ‘solution’ has tended to be called an equilibrium. This, though, if to repeat, is not a hypothesis about the world but merely a (possible) property that a set of equations may or may not be found to possess.

Of course, if not all economists always recognise this, those who have contributed to the formulation of equilibrium economics are often clear enough. Consider, for example, Frank Hahn on the matter:

[…] it cannot be denied that there is something scandalous in the spectacle of so many people refining the analyses of economic [equilibrium] states which they give no reason to suppose will ever, or have ever, come about. It probably is also dangerous. Equilibrium economics […] is easily convertible into an apologia for existing economic arrangements and it is frequently so converted.

(Hahn 1970, pp. 88–89)

Or consider Huw Dixon who summarises matters as follows:

At its most general, we can say that ‘equilibrium’ is a method of solving economic models. At a superficial level, an equilibrium is simply a solution to a set of equations.

(Dixon 1990, p. 356)

Nor can much more be claimed for, or of, the efficient markets hypothesis. Consider for example the manner of its defence by William Sharpe, a Noble Memorial Prize Winner in economics, writing in his famous article on capital asset pricing. Here he acknowledges that the claims required to generate the conclusion that capital markets are ‘efficient’ are:

[…] highly restrictive and undoubtedly unrealistic assumptions. However, since the proper test of a theory is not the realism of its assumptions, but the acceptability of its implications, and since these assumptions imply equilibrium conditions which form a major part of classical financial doctrine, it is far from clear that the formulation should be rejected[…].

(Sharpe 1964, p. 434)

The more careful of the advocates of modelling have recognised all along that substantive formulations are not intended to be realistic, of course. As Hahn acknowledges in his “Intellectual Retrospect”:

The great virtue of mathematical reasoning in economics is that by its precise account of assumptions it becomes crystal clear that applications to the “real” world could at best be provisional. When a mathematical economist assumes that there is a three good economy lasting two periods, or that agents are infinitely lived (perhaps because they value the utility of their descendants which they know!), everyone can see that we are not dealing with any actual economy. The assumptions are there to enable certain results to emerge and not because they are to be taken descriptively.

(Hahn 1994, p. 246)

The problem is that practical modellers just seem unable, in the main, to grasp or accept this, or they somehow think that it does not matter. Soros clearly does recognise that the lack of realisticness and relevance of modern economics matters. The reason I run through all this is simply to bring home the fact that the theories or assumptions that so concern Soros are effectively of secondary importance compared to, and indeed constructed just to be supportive of, the practice of mathematical modelling per se. It is the overwhelming emphasis on mathematical modelling in modern economics that is the real cause of the problems that concern him.

Implications for Soros

So the fundamental point I am wanting to draw out here is that because the modern insistence on methods of mathematical deductive modelling presuppose, for their usefulness, a ubiquity of closed systems, whereas the social world in which we live is quintessentially open, the emphasis on mathematical modelling itself is subject to the sorts of criticisms that Soros makes of its recently most prominent theories.

Moreover, not only are the substantive theories and questions derivative of the formalist emphasis, but, like all fashions, they come and go with some rapidity. However, the mathematical deductive modelling emphasis itself, the ultimate source of all the problems, remains largely unchallenged (see Lawson 2009b).

Indeed Soros notes in a passage reproduced above that the rational expectations hypothesis is ‘no longer taken seriously’. So too equilibrium theorising is out of vogue. But, as I say, the mathematical emphasis of the discipline remains. My concern, however, is that Soros does not fully, or consistently, appreciate that it is the insistence on formalistic modelling per se that is the source of all the problems, and that the endeavour to produce new forms of mathematical modelling, incorporating revised economic theories, will necessarily mean merely more of the same.

In this light I cannot help but note that most of the members of the advisory board of Soros’ new Institute for New Economic Thinking have been, and continue to be, mathematical modellers. There is a real danger, I fear, that many and perhaps even most of the recipients of the funds of the Institute will seek to produce more of the same: more mathematical models and techniques of economic modelling, if no doubt justified as being different sorts of mathematical models and different types of modelling methods to those currently on the scene. If this is indeed the outcome then the enterprise will very likely be found to be as irrelevant as projects that have prevailed all along.

At times Soros (2009) does seem critical of the emphasis on mathematical modelling per se in economics. But relevant criticism appears to be advanced only with notable hesitancy and often a degree of ambiguity. For example following passages reproduced above, Soros questions why his theory of reflexivity “has not even received serious consideration by the economics profession” (p. 216). Soros rightly recognises that this is because his theory is seemingly inconsistent with the project of mathematical modelling, but he seems to want to combine his theory with a formalistic modelling approach nonetheless. Specifically Soros writes:

Those who are most sympathetic to my views explain to me that my theory is not getting more attention because it cannot be formalized and modelled. But that is exactly the point I am trying to make: Reflexivity gives rise to uncertainties that cannot be quantified and probabilities that cannot be calculated. Frank Knight made that point a century ago in Risk, Uncertainty and Profit, and John Maynard Keynes recognized it, too. Yet market participants, rating agencies, and regulators alike came to depend on quantitative models in calculating risks.

(Soros 2009, p. 217)

So far so good. But then Soros adds:

One question I am seeking an answer to is whether it is possible to model reflexivity, or whether one should continue using quantitative models but take reflexivity into account by adding a margin for error due to incalculable uncertainties. My hunch is that we need to do both. Reflexivity cannot be modelled in the abstract, but it should be possible to model specific instances of it, such as the effect of the willingness to lend on real estate prices. At the same time, quantitative models may be useful for calculating the risks that prevail in near-equilibrium conditions, while remembering, particularly for regulatory purposes, that conditions may occasionally veer quite far away from equilibrium. These are questioned that need to be explored.

(ibid., p. 217)

Here I think Soros reveals himself to be ultimately insufficiently critical of the emphasis of the formalistic modelling endeavour. What does it even mean to be in “near-equilibrium conditions”? As Soros points out himself over and again, social reality is what we make it; it is path dependent. There is no set of outcomes that would prevail if we were not involved. Nor does it make sense to make reference to counterfactual states of affairs that could not come about (for example those in which we all have perfect foresight or rational expectations) and which have no bearing on anything that happens. Soros is rightly critical of those who respond to his contribution by merely suggesting that his “theory of reflexivity merely states the obvious, namely that market prices reflect the participants’ biases”, adding:

That is an obvious misunderstanding of my theory, which holds that mispricing in financial markets can, in specific circumstances and ways, affect the fundamentals that market prices are supposed to reflect.

(p. 216)

But a matter about which the qualifier “supposed” seems to reveal recognition is that there is no obvious meaning to the notion of fundamentals here, not if fallible and mistaken ideas and practices constitutively influence whatever happens. There is just a social system moving (or being moved) along, with all of us collectively drawing upon, and inadvertently or otherwise, reproducing, as well as transforming aspects of, that which has gone before in a non-predetermined way, in a manner in part captured by Soros’ theory of reflexivity2For a systematic elaboration of this process of transformation see Lawson 1997, 2003..

Soros continues by anticipating that behavioural economics and evolutionary systems theory constitute new ways of economic thinking, but is rightly cautious about their formalistic orientation. However, he finishes up a discussion of these projects writing:

[…] I am eager to understand better the connection between evolutionary systems theory and reflexivity. I posed the question at the Santa Fe Institute, which is devoted to the study of complexity, but I have not yet found the answer. That is another question I wish other people would think about.

(Soros 2009, p. 222)

Complexity theory as studied at the Santa Fe Institute is of course but another form of mathematical deductive modelling (if of a largely non-linear variety). It will thus be unable to meet Soros’ concerns. Yet Soros seems not to appreciate this. Evolutionary theory per se, in contrast, seems entirely an appropriate framework for Soros’ notion of reflexivity, albeit the evolutionary theory of Darwin rather than the very distant cousin found in modern economics. This, though, is too long a story to enter into here (though see Lawson 2003, especially chapters 5 and 103Let me, however, make a very brief comment. It is worth recalling the central and great Darwinian insight is that a subset of members of a … Continue reading).

Final comments

In advancing his theory of reflexivity George Soros often appears to rest his case on various contrasts which, I have suggested, are not only overstated or incorrect but also inessential to his argument. Because of the importance of his central claims, and because his book seems to be widely influential, I have taken the time to indicate that Soros’ main contentions stand up anyway. Hopefully, this will prevent these insights being rejected either by those sympathetic to Soros’ position but unconvinced by the surrounding arguments, or by those opposed to Soros’ central contentions and who might seek to use questionable (supplementary) arguments to detract from them.

In particular, I have suggested that Soros’ contrast between the nature of (non-social) natural and social phenomena as formulated is too sharp and not especially relevant to the analysis, and nor is his science/non-science differentiation.

To be fair to Soros, there are times when I believe he appreciates all this himself. Thus towards the end of his 2009 book he comments:

Here, I must introduce a cautionary note about my own argument. I am troubled by the sharp distinction that I have drawn between human affairs and natural phenomena. Such sharp dividing lines are not characteristic of nature, but of human efforts to make sense of an infinitely complicated reality. This is also in accordance with my postulate of radical fallibility.

(Soros 2009, p. 222)

And a few lines later he adds:

I am willing to admit that reflexivity does not meet the currently accepted standards of scientific theory […]. I contend, however, that we must either modify the standards or study financial markets in a non-scientific way. The former may be difficult, because it would involve a loss of status for economists.

Here Soros is right to suggest that we must modify the accepted standards, but wrong about his concept of reflexivity not meeting the important ones. Soros’ conception expresses an efficacious causal process that Soros himself demonstrates, throughout his writings, to be empirically grounded. As I have argued over again elsewhere the only standards that require modifying are those of commentators who misconstrue the nature of science (see e.g., Lawson 1997, 2003, 2010).

Outside the discipline of economics many mock the pretentiousness and irrelevance of the emphasis of modern economists on formalistic practices4Nor is this a recent phenomenon, see for example Richard Parker (1993) or David Howe (2000). We might recall, too, commentaries in the March 1994 … Continue reading

If there is a central moment to the practices of the successful sciences it involves first identifying ‘interesting’ and typically puzzling phenomena, and then unearthing their causes. Just because such causal explanation involves moving from one type of phenomenon (the one to be explained) to a typically very different type of phenomenon (the causal factor[s] responsible), mathematical reasoning cannot do the job. Rather at such a point in the explanatory process more creative forms of human endeavour, like reasoning by analogy and metaphor and guess work, come to the fore.

These ways of proceeding may conform little to the typical economist’s (misplaced) ideas of fundamental scientific practice, but if, in response to Soros’ initiatives, a (greater) take up of such causal explanatory research were the case, the result would likely nonetheless actually represent something of a return of economics to scientific ways.


  • Dixon, Huw. 1990. “Equilibrium and Explanation”. In Foundations Of Economic Thought, ed. John Creedy. Oxford: Basil Blackwell.
  • Hahn, Frank H. 1970. “Some adjustment problems”. Econometrica, n°38 January, pp. 1–17, reprinted in Hahn 1984.
  • Hahn, Franl H. 1984. Equilibrium and Macroeconomics. Oxford: Basil Blackwell.
  • Hahn, Frank H. 1994. “An Intellectual Retrospect”. Banca Nazionale del Lavoro Quarterly Review, vol. XLVIII n°190, pp. 245–258.
  • Howell, David. 2000. The Edge of Now: New questions for democracy and the network age. London: Macmillan.
  • Lawson, Tony. 1997. Economics and Reality. London and New York: Routledge
  • Lawson, Tony. 2003. Reorienting Economics. London and New York: Routledge.
  • Lawson, Tony. 2005. “The. Confused. State of Equilibrium Analysis in Modern Economics: an. Ontological. Explanation”. Journal for Post Keynesian Economics, vol. 27, n°3 Spring, pp. 423–444.
  • Lawson, Tony. 2006. “Tensions in Modern Economics. The case of Equilibrium Analysis”. In Equilibrium in Economics: Scope and Limits, ed. Valeria Mosini. London and New York: Routledge.
  • Lawson, Tony. 2008. “Social Explanation and Popper”. In Popper and Economic Methodology: Contemporary Challenges, Tom Boylan and Paschal O’Gorman (eds). London and New York: Routledge.
  • Lawson, Tony. 2009a. “Applied Economics, Contrast Explanation and Asymmetric Information”. Cambridge Journal of Economics, vol. 33, n°3 May, pp. 405–420.
  • Lawson, Tony. 2009b. “The Current Economic Crisis: its Nature and the Course of Academic economics”. Cambridge Journal of Economics, vol. 33, n°4 July, pp. 759–788.
  • Lawson, Tony. 2010. “Economics and Science”. The Transatlantic: Journal of Economics and Philosophy, vol. 1, pp. 8–13. Online version at:, accessed on 20 th March 2013.
  • Muth, John F. 1961. “Rational Expectations and the Theory of Price movements”. Econometrica, vol. 29 n°3, July, pp. 315–335.
  • New Scientist. October 31 1992, pp. 26–31
  • Observer Magazine. 20th September 1992, p. 7
  • Parker, Richard. 1993. “Can Economists Save Economics?”. The American Prospect, vol. 4 Issue 13, March 21th.
  • Sharpe, William. 1964. “Capital Asset Prices: A Theory of Market Equilibrium Under Conditions of Risk”. The Journal of Finance, vol. XIX n°3 September, pp. 425–442.
  • Soros, George. 2009. The Crash of 2009 and What it Means: The New Paradigm for Financial Markets. New York: Public Affairs.


Modern academic economics is not explanatorily successful. Although this situation has long been evident, it has become widely recognised only with the onset of the recent economic crisis. With the situation now widely acknowledged, however, various initiatives have been launched in the hope of achieving something better. One highly significant such initiative has stemmed from George Soros’ insight that whilst reflexivity is a widespread feature of social reality, the sorts of economic theories that have dominated academic economic output are fundamentally inconsistent with it. Here I argue that whilst Soros’ contribution contains insight, it does not go far enough, and risks having the insights contained neglected due to the inclusion by Soros of various questionable assessments that are in any case unnecessary to the central argument.

JEL codes: A1, B4, G1


  • reflexivity, economic modeling, open and closed system, consistency, rational expectations hypothesis, equilibrium, efficient markets hypothesis, science


1 On March 20th 2010, three weeks or so before attending a conference hosted by George Soros’ Institute for New Economic Thinking (in Kings College Cambridge starting April 8 th), I set off for a lecture tour of Japan. Having inadvertently left my intended travel reading behind, and finding George Soros’ (2009) book at Heathrow airport, I purchased it. In so doing, I was mindful both of the forthcoming conference, and that a close friend had repeatedly recommended Soros’ book over the previous year as an important read. Indeed the latter is the case. This paper is basically an engagement with Soros produced in the confines of various forms of transit as I travelled to and from, and around various parts of, Japan. No doubt the style of the paper, and the fact that there are few references other than to Soros (2009) and to my own writing, reflects that confinement/context.
2 For a systematic elaboration of this process of transformation see Lawson 1997, 2003.
3 Let me, however, make a very brief comment. It is worth recalling the central and great Darwinian insight is that a subset of members of a population may come to flourish relative to other members simply because they possess a feature, which others do not, that renders them relatively suited to some local environment. Fundamentally, the question of the intrinsic worth of those who flourish most is not relevant to the story.We can briefly recall an example from biology. Consider the case of the varying fortunes of spotted grey and dark moths against an environment of UK industrialisation. Prior to the nineteenth century the spotted grey was more common than the dark moth. When resting on the lichen covered trees in their habitat the spotted grey moth was effectively invisible to birds, unlike the dark moth which was easily spotted against the light coloured trees and eaten. With XIXth century industrialisation, however, pollutants killed the lichen on the trees in certain areas and rendered the bark of trees in the relevant vicinities a dark colour. Both types of moth continued to rest on trees. But with the spotted grey now more easily recognisable to birds, there was a shift in the relative proportions of the two populations from the spotted grey towards the darker variety. In a sense the pollution-darkened barks protected the darker moths from the danger of the moth-seeking birds.A significant feature of this process is that certain individuals are found to fare better than others just because they are of a type, or possess a trait, relatively suited to their local environment, not because they are successful in any wider or absolute or more laudatory sense.In the social realm, as I have elsewhere argued (Lawson 2003) the variety to be selected amongst typically exists amongst certain human practices of interest, and the selecting environment is the sum total of all the practices. The latter determine which of the variety are selected as successful. Which are selected can usually only be discovered a posteriori. This applies to new processes of production, etc., but also to speculations where the ‘winning bets’ in financial systems depends precisely on the sum total of what everyone chooses to do. It also applies to the longevity of mainstream mathematical modelling project, which has clearly been unsuccessful in terms of results, but has been selected on grounds more to do with the acceptability of formalism per se in a given institutional context. It also applies to the specific (formalistic) theories within this explanatorily unsuccessful project, which get selected according to some needs of the academic environment that are found to be determining a posterior (all this, of course, is a very long story, one elaborated in detail in Lawson 2003, chapter 10). Hopefully these few comments are enough to indicate that an evolutionary framework is indeed a likely relevant one for Soros concept of reflexivity that is so tuned, as it is, to processes of human interaction in an inescapably open social system.
4 Nor is this a recent phenomenon, see for example Richard Parker (1993) or David Howe (2000). We might recall, too, commentaries in the March 1994 edition of The Times Higher Education Supplement which carried the heading “No reality please. We’re economists”. The (U.K.) Observer Magazine (20th September 1992, p. 7) concluded that “there’s no such thing as economics. It’s all voodoo, bluff and pseudo-science”. New Scientist (October 31 1992, pp. 26-31) even carried sketches of economists forecasting economic variables by reading lines on the palms of their hands and of econometricians pushing numbers around while blindfolded.