KYKLOS, Vol. 5 0 - 1997 - Fasc. 4,561-574 O n George Soros a n d Economic Analysis Rod Cross and Douglas Strachan* ‘Economic theory needs to be fundamentally recon- sidered. There is an element o uncertainty in economic processes that has been largely left unaccounted for. None o f the social sciences can be expected to yield firm results comparable to the natural sciences, and eco- nomics i s no exception. We must take a radically diferent view o f th e role that thinking plays in shaping events’ George Soros (1995, p. 67 ) There is more than a touch of Renaissance Man about George Soros. Whilst completing an undergraduate degree in economics at LSE he came under the influe nce o f Karl Popper. After a spell working as an analyst specialising in arbitrage opportunities betwe en U S and European financial markets, he pro- duced a Popperian treatise on philosophy, ‘The Burden of Consciousness’ ( 1 96 I /2). The Quantum Fund operated by Soros came to be possibly the most prominent investment fund on the world stage. Operating on the basis of leverage, the Quantum Fund earned an average annual yield of 35%, if profits had been reinvested, over the twenty-five years from 1969 (Soros 1995). Soros himself became a major media figu re in Europe after the Fund made $1 billion from taking a short position on ste rling before the exit o f t he C sterling from the ERM on 15 Septemb er 1992 . No t co ntent with just making money, Soros regarded the opera tions o f the Quan tum Fund as a series o f tests f or his theories o f h o w the players in financial mark ets interact with the economic systems i n * University o f Strathclyde and ICMM. Correspondence: Department of Economics, University of Strathclyde, Curran Building, 100 Cathedral Street, Glasgow G4 OLN, Scotland, UK, Tel: 0141- 548-3855/3856/3840, Fax: 0141 -552-5589, E-M ail: [email protected]. 56 1
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
‘Economic theory needs to be fundam entally recon-
sidered. There is an element o uncertainty in economic
pro cess es that has been largely left unaccounted fo r.
None of the social sciences can b e expected to yield firm
results comparable to the natural sciences, and eco-
nomics is no exception. W e must take a radically dif er en t
view of the role that thinking pl ay s in shaping events’
George Soros (1995, p . 67)
There is more than a touch of Renaissance Man about George Soros. Whilst
completing an undergraduate degree in economics at LSE he came under theinfluence of Karl Popper. After a spell working as an analyst specialising in
arbitrage opportunities between US and European financial markets, he pro-
duced a Popperian treatise on philosophy, ‘The Burden of Consciousness’
( 1 96 I /2). The Quantum Fund operated by Soros came to be possibly the most
prominent investment fund on the world stage. Operating on the basis of
leverage, the Quantum Fund earned an average annual yield of 35%, if profitshad been reinvested, over the twenty-five years from 1969 (Soros 1995). Soros
himself became a major media figure in Europe after the Fund made $1 billion
from taking a short position on sterling before the exit of the C sterling from theERM on 15 September 1992. Not content with just making money, Soros
regarded the operations of the Quantum Fund as a series of tests for his theories
of how the players in financial markets interact with the economic systems in
* University of Strathclyde and ICMM. Correspondence: Department of Economics, University of
‘ ... the discrepancy between thinking and reality is not very large and there are forces at play
that tend to bring them close together, partly because people can learn from experience, and
partly because people can actually change and shape socialconditions according to their desires’(Soros 1995,p. 69).
In such situations economic theories that ignore reflexivity might not be toowide of the mark. In far-from-equilibrium onditions, however, which might
be thought of as circumstances in which major changes in the economic
environment take place,
‘ _..he prevailing bias and the prevailing trend reinforce each other until the gap between them
becomes so wide that it brings a catastrophic collapse’ (Soros 1995,p. 70).
An example of a bias that can generate far-from-equilibrium conditions is
provided by errors in valuation.So, in the international lending boom beginning
in the late 1970s,for example, banks judged the borrowing capacity of debtor
countries by looking at debt-GDP or debt service-export ratios. The problem is
that such ratios are reflexive, in that they reflect the lending activities of the
participant banks as well as any ‘fundamentals’ affecting borrowing capacity.
Thus trends of increased lending were established on the basis of reflexive
valuations. Eventually the gap between beliefs about borrowing capacity and
the reality of actual borrowing capacity became so large that the bias underlying
the beliefs came to be realised, and lending collapsed.
Thus reflexivity in economic behaviour is capable of bringing about far-
from-equilibrium situations and regime changes. This, far from being nihilistic,
is an intriguing account of endogenously-generated regime changes. In terms
of economic analysis the suggestion is that the behaviour of economic agents
using rules of thumb such as valuation ratios, and of economic modellers
themselves, need to be introduced into economic models in order to capture the
interaction between thinking and reality. The way forward, then, is to eschew
the reductionist search for finer-grain microfoundations and instead study how
complex economic systems can emerge from the interactions between agents
operating both as thinkers and actors (Soros 199.5,p. 220).
111. RATIONAL EXPECTATIONS AND EFFICIENT MARKETS
From the foregoing discussion it is clear that Soros views the postulates ofrational expectations and market efficiency that have dominated the economics
literature on financial markets in recent years as implausible descriptions of
In rejecting the idea that the uncertainties su rroun ding econo mic behav iour canbe tamed by assum ing that agents have perfect knowledg e of some objective
equilibrium probability distributions, Soros is in the same camp as John
Ma ynard Keynes. Th e untractable nature of the uncertainty surrounding eco-
nom ic decision-taking was stressed in the General Theory (Keyn es 1936, ch. 12)
and in Keynes’ response to his reviewers (Keynes 1937):
‘ .,. the outstanding fact is the extreme precariousness of the hasis of knowledge on which our
estimates of prospec tive yield have to be made . . _ou r nowle dge of the factors which will gov ern
the yield of an investm ent some years hence is usually very slight and often negligible ... if we
speak frankly, we have to admit that our basis of knowledge for estimating the yield ten years
hence of a railway, a copper mine, a textile factory, the goodwill of apatent medicine, an Atlantic
liner, a building in the City of London amounts to little and sometimes to nothing ... in fact,
those who seriously attempt to make any such estimate are often so much in the minority that
their behaviour doe s not govern the ma rket’ (Keynes 1936, pp. 149-150).
Ke yne s then proceeds to point out that the existe nce of securities markets mea ns
that the underlying physical investments
‘ .. _ re governed by the average expectation of those who deal in the Stoc k Exchange as revealed
in the price of shares, rather than by the genuine exp ectations of the professional entrepreneur’(1936, p. 151).
How , then, are the expectations of those dealing in securities d etermined ?
He re there are several similarities between Soro s and K eynes. A first is that,
in both So ros and Keynes, investors follow conv entions or rules of thu mb . In
So ros this is the trend-following behaviour pursued by investors, particularly
prevalent in funds whose performance is measured relative to the rest of the
market, rather than in absolute terms. The expectation implicit in such an
inve stme nt strategy is that the present trend will continue. In K eyn es the trendis called a convention, wh ose essential featu re is the assumption
‘ ... hat the existing state of affairs will continue indefinitely’ (193 6, p. 152).
Such conventions can survive in what Soros would call near-equilibrium
cond itions, and a re comp atible with
‘ ... a considerable measure of continuity and stability in our affairs’ (Keynes 1936, p. 1 52).
Such co nvention s, however, as in Soros, rely o n expectations for which there
‘ ...we are assuming, in effect, that the existing market valuation, however arrived at, is uniquely
correct in relation to our existing knowledge of the facts which will influence the yield of the
investment, and that it will only change in proportion to changes in this knowledge; though,philosophically speaking, it cannot he uniquely correct, since our existing knowledge does not
provide a sufficient basis for a mathematical calculation’ (Keynes 1937, p. 152).
A second area of similarity conc erns what happens in what So ros calls far-from -
equilibrium conditions. In such circum stances trends are broken, and conv en-
tions regarding the correctness of existing market valuations are no longer
tenable. In Keynes such circumstances arise inevitably in systems where
valuations are established as
‘the outcome of the mass psychology of a large number of ignorant individuals’
because there are
‘ .. _no strong roots of conviction to hold it steady’
Keynes’ ‘abnormal times’ correspond to Soros’ ‘far-from-equilibrium condi-
tions’ in which perceptions and reality a re far from being syn chro nised . Fo r
Keynes such chan ges occur
‘ ,.. when the hypothesis of an indefinite continuance of the existing state of affairs is lessplausible than usual even though there are no express grounds to anticipate a definite change’.
Th e implication is that
‘ ... the market will be subject to waves of optimistic and pessimistic sentiment, which are
unreasoning and yet in a sense legitimate where no solid hasis exists for a reasonable calcula-
tion’ (Keynes 1936,p. 154).
A third are a where Soros an d K eynes a re close is in relation to reflexivity. In
Soros the theories which econo mic ag ents use to attempt to understand theirenvironment also change the environm ent when used as a basis fo r decision-
taking. In Keynes this sort of self-referential beh avio ur arises beca use investors
are inevitably invo lved in a gam e requiring each participant to gues s what the
basis of conventional valuation will be in the future. The con ven tiona l valuation
that emerges will depe nd o n a particular participant’s ow n gu ess about wha t
other participants will be guessing, who in turn will be guessing about the
particular particip ant’s guess. Thu s the m arke t valuation will be reflexive. T his
is the basis for Key nes’ striking beauty contest metaphor:
‘professional investment may be likened to those newspaper competitions i n which the
competitors have to pick out the six prettiest faces from a hundred photographs, the prize being
instead on the heuristic devices agents use as decision rules, and how these
decision rules are updated or revised in the light of exp erience. Th e task of
modelling eco nom ic systems is then o ne of populating mo dels with age nts who
are boundedly rational in the sense that they d o not know what any aggregate
probability distributions or equilibria are. Th us instead of assuming that agents
have worked out how the complex system in which they participate behaves,
the way the system itself beh ave s is constructed from the heuristic dev ices used
by the agents.
This approach is pursued in Sargent (1993). Th e starting point is that
‘.._when implemented numerically or econometrically, rational expectations models impute
more knowledge to the agents within the model (who use the ryrtilibriurn probability distrihu-tions in evaluating their Euler equations) than is possessed by an econometrician, who faces
estimation and inference problems that the agents in the model have somehow solved’ (p. 3).
Instead the boundedly rational m odels
‘... expel rational agents from o u r model environments and replace them with ‘artificially
intelligent’ agents who behave like econometricians ... these ‘econo metrician s’ theorise,
estimate and adapt in attempting to learn about probability distributions which, under rational
expectations, they already know’ (p. 3) .
Taking e cono mic agents to behave like econom etricians migh t appear absurd
until it is realised that
‘... n economics, procedures for revising theories in light of data are typically inform al, diverse
and implicit’ (p, 22).
Thu s the distinction b etween e conom ic agents and economists/econometricians
is blurred a nd the So ros self-referential notion of reflexivity is introduced:
‘... uch a system can contain intriguing self-referential loops, especially from the standpointof macroeconomic advisers, who confront the prospect that they are participants in the system
that they ar e mod elling, at least if they believ e that their advice is likely to be conv incing’ (p, 23).
2 . Complexity
In m any sciences the traditional emp hasis has been on searching for finer grain
explan ations of phenom ena. Thu s in 20th century physics there has been an
emp hasis on attem pting to understand the nature an d structure of sub-atomic
particles. In econo mics, albeit a form of alchemy according to So ros, this trait
has been evident in the insistence on micro foundations for macro models.
During the last decade this emphasis has been countered by an increase in
interest in models of complexity, wherein the problem is to work out how
complex systems can be constructed or emerge from basic elements that are
taken as in som e sense irreducible (see Anderson, A rrow and Pines 1988 for
discussion of this approach to economic systems).
Sor os clearly sees this as the way fo r economic analysis to go:
‘... i t is high time to liberate social pheno mena from the straitjacket of natural sc ience, especially
as natural science itself is undergoing a radical change ... analytical science is superseded in
certain fields by the study of complex ity .,. he science of complexity studies open, evolutionary
systems ... all it seeks to do is build models or run simulations ...hut even here I find that the
difference between social and natural phenomena is no t sufficiently recognised ... mostcomputer programm es deal with the evolution of populations ... o study the interaction between
thinking and reality, we need a model of model-builders whose models, in turn, must contain
model-builders ud infinifurn _.. the infinite nesting of models must be brought to closure
some where if the mod els are to serve any practical use ...as a result, the models cannot retlect
reality i n its full complexity’ (Soros 1995,p. 220). ’
Th us although models of complexity are proposed, there is inevitably a deg ree
of arbitrariness surrounding the way such models avoid the infinite regress of
self-reference.
Models of complexity have come to be seen as promising ways to resolve
some of the anomalies surrounding the rational expectations-efficient markets
paradigm regarding financial marke ts. Simp le technical trading rule s of the type
stressed by Keynes and Soros have been shown to out-perform investment
strategies based on the standard paradigm (Brock, Lakonishok and Le Baron
1992, for example). And the standard paradigm implies that no trade should
take place in rational expectations equilibria (Tirole 1982), whereas trading
volumes remain significant even in quiescent financial markets. In models of
complexity these anomalies are resolved by specifying agents who are het-
erogeneous in the sense of using different trading rules. Thus som e traders may
extrapolate trends, others make buck trends thinking that what goes up must
come down, some agents may believe that market prices are determined by
‘fundamental’ economic forces, and yet others may see market prices as
reflecting the weights of the different belief system s of other traders (B rock and
Ho mm es 1996). The heterogeneous agents in such a system are then en dow ed
with a form of artificial intelligence which allows them to ch ange their beliefs
in light of how their initial hunches fare in terms of generating trading profits
(Arthur, Holland, Le B aron, Palmer and Taylor 1994). Thu s trading profits act
as a ‘fitness’ function, and beliefs, and therefore m arket prices, evo lve accord-
ing as to whether particular belief systems allow the traders adopting them to
hysteretic or heterostatic equilibria with the Soros account of the power of
reflexivity in far-from-equilibrium conditions.
VI. SOME CONCLUDING REMARKS
A not uncommon criticism of economic theories is that they are out of touch
with practice, by wh ich is presumably meant reality. Geo rge Soros, unlike many
of the other critics, has made a constructive contribution by articulating a
method of analysis which could put economic theory into closer contact with
reality. In this essay we have argued that, although approaching the issues from
a Popperian background, the Soros approach is resonant with Keynes oneconom ic behaviou r in the face of uncertainty. Thi s line of thought is sometimes
depicted as being nihilistic about the possibility of economic knowledge. We
have argued, on the contrary, that foundations of the Keynes-Soros type are
analytically fecund, as can be seen in recent work on bounded rationality,
com plex econ om ic systems and heterostatic equilibria selected by self-organi-
sation or hysteresis.
RFFERENCbS
Amable, B., J. Henry, F. Lordon and R. Topol (1995). Hysteresis Revisited: a Methodological
Approach, in: R. Cross (ed.), The Nuturul Rate of Uneniployment: Reflections on 25 Years of'
the Hypothesis. Cambridge: Camhridge U.P.
Anderson, P.W. K.J. Arrow and D. Pines (eds.) (1988). The Economy a s an Evolving Complex
System. Massachusetts: Addison-Wesley.
Arrow, K.J. (1988). Workshop on the Economy as an Evolving Complex System: Summary, in:
P.W. A nderson, K.J. Arrow and D. Pines (eds. ),Th e Economy asuri Evolving Comple.rSysteni.
Massachusetts: Addison-Wesley.
Arthur, W.B. J.H. Holland, B. Le Baron, R. Palmer and P. Taylor (1994). An Artificial Stock
Market, m imeo, Santa Fe Institute.
Bak, P., K. Chen, J. Scheinkman and M. Woodford (1993). Aggregate Fluctuations from Inde-pendent Sectoral Shocks: S elf-Organised Criticality i n a Model of Production and Inventory
Dynamics, Richerche Emnotnich e. 47: 3-30.
Brock, W.A., J . Lakonishok and B. Le Baron (1992). Simple Technical Trading Rules and the
Stochastic Properties of Stock Returns, Jourtiul qfFin unce. 47: 1731-1764.
Brock, W.A. and C.H. Hommes (1996).Models of Complexity in Economics and Finance, mimeo,
University of Wisconsin.
Coddington, A. (1983). Keynes iun Econo ~~i i c ~s:he Srurch f o r First Principles. London: Allen &
Unwin.
Cross, R. (1993). On the Foundations of. Hysteresis in Economic Systems, Economics und
Philosophy. 9: 53-74.
Cross, R. (1995). Is the Natural R ate Hypothe sis Consiste nt with Hysteresis'?, in: R. Cross (ed .),The Natural Rate of Uneniployment: Rr:flec/ions on 25 Years of /he Hypothesis. Cambridge:
Keynes, J.M. (1934). Poverty i n Plenty: Is he Economic System S elf-Adjusting’?,Th e Listener, 2 1
Keynes, J.M. (1936). The General Theor j o f E ? n p / o y mmt , nterestand Money. London: Macmillan.Keynes, J.M. ( I 937). The General Theory of Employment, Quarter!\. Journal of Economics. 5 :
Marshall, A. ( I 890). Th e Principles of Economics. London: Macmillan.
Mirowski, P. (1989). More Heat than Light. Cambridge: Cambridge U P .
Sargent, T.J. (1993).Bounded Rationality in Macroeccinomicx Oxford: Oxford U.P.