Вы находитесь на странице: 1из 13
The Coherent Market Hypothesis Chaos theory has received considerable attention in the aftermath of the Crash of 1987. However, an alternative non-linear statistical model is far more useful for understanding both the coherent bull market prior to the crash and the subsequent chaotic market fluctuations in October 1987. Unlike chaos theory, which seeks to forecast the path of stack prices ina deterministic sense, the non-linear statistical model, based on “A Theory of Social Imitation,” can forecast transitions from trendtess (random-walk) markets to periods of coherent, or erderly, price trends and periods of chaotic fluctuations (the panics and crashes associated with abrupt trend reversals). A coherent bull market is characterized by an “‘inversion’' in the historical risk-reward ratio. Over the past 60 years, the stock market has provided an average 10 per cent total yearly return, with a standard deviation of 20 per cent. During coherent bull markets, the return from the stock market averages over 25 per cent, while the standard deviation drops to the neighborhood of 10 per cent. Missing these low-risk opportunities can lead to underperformance, because the rest of the time the market presents more risk than reward, in the form of chaotic markets and periods of irue random walk. This article describes the theoretical basis and practical indicators of coherent markets. The results suggest that both technical and fundamental analysis can add real value to the investment decision-making process. electromagnetic fields in a laser, the phase of cells contracting in live heart tissue and the polarization of political opinion in social groupe—all undergo transitions between siales of macroscopic disorder and more or- derly, or coherent states! This aticle eviews T: MAGNETIZATION of a bar of iron, walk as a special case, Table I summarizes this model and its controlling parameters, ‘Transitions from random-walk markets to pe- riods of crowd behavior are characterized by instability As Figure A illustrates, the normal distribution flattens out into a wide, uniform distribution, and the impact of random news on ‘non-linear models of such state transitions and examines the conditions under which we might expect, in the capital markets, tonsitions from. disorderly, random-walk (efficient-market) states to more ordered states (chaotic and coher ent markets) More specifically, I describe stock market flac. tuations in terms of “A Theory of Socal Imitation.”? Unlike chaos theory, which is based on deterministic models, the Theory of Social Imitation isa non-linear statistical model Tk may be regarded as a statistical version of chaos theory. Furthermore, it inciudes random TFeomote apparel end of ack: ‘stock prices is no longer discounted (damped) quickly. Large, long-lasting (undamped) price fluctuations are likely, posing abnormally high risk to investors. In Figure A, the major market states predicted. by the Theory of Social Imitation are illustrated in terms of “potential wells” and probability distributions. Annualized return may be ‘thought of asa bal in the well, which is buffeted by random forces. Provailing investor sentiment and economic fundamentals determine the shape of the well, while random news buifets the ball (return from a market index), When “normal” sentiment gives way to collective “group-think,” we have to look at non-linear forces to understand market fluctuations. [RNANCIAL ANALYSTS JOURN / NOVEMBER BECEMDCR 156036 Table 1 Non Linear Theory of Social Initation fa=co-'@ ee f' trovowit where fg) ~ probability of anelined rum, g ig) = Sinha + hl = aacsh fea Qi) = Ucn hh) 2g ah gH) and 2 mabe of dees tree 2 igor ceed evr 1h fondanenal bas ete [Maraepaf* movavi ata rowel behavior can lead to a “chootie” mar- ket, as occurred in October 1978 and October 1987, oF to the safest, most rewarding invest- ment opportunities—January 1975 and August 1982, for example. The latter are examples of “coherent” markets. They should be of special interest to investors and are the focus of this auticle Market States Investor sentiment and the prevailing bias in ‘economic fundamentals control the state of the market. When investor sentiment is net conda- ‘ive to “group think” or crowd behavior, the market i likely to be in a random-walk state (efficient market), In such relatively quiet per Figure A Transition from Random Walk to Crow Behovior Teme 0.98 ‘conic Market 005 Cord ichoro wth feck stay beans ae fursmertask~ 22:4 = -0.05) ea 00% _ 0. 0.06 = 1% eames ; Eee ee. 3 oak — ffs ee | aE an zi ve aor 1" ‘ mile mee Zam. z Ueticiet marley =o : real undamentas 39! Crane FO 006 = e moe ‘Anngaized Retarn 8 0.05 001 an ‘on 001 i Penta Wel Bi True Ren Wath (Eeicore act ‘neta urate Sieh =o [ANANCIAL ANALYSTS JOURNAL /NOVEMNER DECEMER 180 1.97 Glossary ‘Coherent Behavior A state of (macroscopic) order inva complex physical or socal system made up of large number of component parts where each pari ree w actindependentl. For am ple, crder may arse in the orientation of mle ules in a magnet, the synchronization of light waves in 2 laser or the polarization of opinicns ‘na tocal group. Perea of coherent behevior inthe stock market occur when the annualized return from 2 market index is greater than the | annualized volalty of the index, ceflocting 2 strong bias or orderly trend in price fluctua ‘ions (Chaos Theory: Also known as catsetrophy theory (made popular by the mathematician Rene Them), abranch of matiematics concerned vith the deterministic solution of abrupt cr discon ‘invous (Le, chaotic) changes im the state of a system, such as the buckhing of a beam or ‘trbulent (as oppose to lamina) ow ofa lad Jn the stock market, chs theory seeks to fore- cst ina deterministic sence the future pat of stock prices. incuding large. abrupt price shanges that ocar during pans and crashes Nom-Linear Statistical Model: A mathematical ‘model that forerasts the conditions under which foe may expect trations fom a sate of mace rostopic ciscrder (random walk} to more oF erly states (coherent betavio: with stable pce trends) as well as abrupt (chectc) tend rever sals (e., panics and crashes). Unlike deteraun- ‘ste chaos theory, a non-linear statistical model seeks to define the prolly distri govern ‘ing stock market luctuations, rather than the spect future path o stock paces. Uniform Distibution: A specific typeof probabil: iy distribution which best repregente stock mar. {et behavior during the unstable tanstion from 4 tue randomwalk market to 9 coherent or chaotic market, Unlike the bell-shaped normal distibuton governing pre Ructatins in o tive rondemevak mat, the unform prska tity distbuton is much wider and ater, impiing that large, longlsting price Matus. sens are roe ely daring the pre of sta bits Ina aoe makets “sda” prot bit dimuon may ec. where exes tremor key than the centr ofthe asta ton tinction ‘Theory of Social Imitation: One particular non- Tru state mode, which can be ppd the tock art to oven teint fon periods afte tam alk tn ceheren! pce frends ane the chaotic fetutionsasrchted wh panics and cashes. This theory icles Tandon walk at 9 special cst ol ay be thought oft» eal for of Shoe tao cds, the stock market is least likely to outper- form fixed dellar and fixed income alternatives. ‘The combination of strong positive fundamen- tals and investor sentiment conducive to crowd behavior leads to the safest, most rewarding slate—a coherent bull market. In these market periods, a fully invested position is necessary 10 avoid the risk of underperforming the market averages. When the fundamental bias is neither strongly bullish nor bearssh during a period of crowd behavior, the result is a chaotic mar- kket-—2 dangerous, “quasiefiicient” market in which random news is discounted quickly, but ‘with a bias, and sentiment may ewiteh abruptiy from bullish te bearish (@s during the Crash of 9), Tm addition to these most prevalent market states, a coherent bear market (highly negative fundamentals with crowed behavior) is theareti- cally possible. This may be the best model of the 1973-74 bear market and the Crash of 1923, which anfolded ever a period of several years, Fortunately, coherent bear markets are histor. cally rare, Bearish fundamentals. usually dampen investor enthusiasm, and prices tend 0 driftlower, in a encom walk, over an extended bear market Figure A shows that when crowd behavior prevails, market fluctuations will tend to follow ‘a bimodal distribution. Under these conditions, ‘market action can best be described as @ biased andom walk or as “quasieficient." Ona short term basis, new developments will be dis: ‘counted quickly, but investors will tend to r2- spond to good news while ignoring bad news (or vice versa). Under these conditions, senti- ment may also, by chance, switch abruptly from ‘one state fo its mirror image, producing anom- algus short-term volatility ‘During pericds of crowd behavior, the market is highly sensitive to any underlying bias in economic fundamentals, Somewhat’ bearish fundamentals may increase the negative lobe of the bimodal distribution and decrease the posi- tive. Under these conditions, if investor senti- ment is bullish, even small random events may be enough to push the ball (market index re- turn) across the potential barrier in the center of the well separating the bullish from the bearish state. Strong bullish fundamentals, however, canelfectively suppress the negative lobe of the probability distribution, yielding a high ex- pected return with a low standard deviation— i.e., a coherent market [HINVANGIAL ANALYSTS [OUINAL NOVEMBER DECEMBER (50 038 Edgar Peters has recently presented evidence ‘of biased random walks.‘ Peters concludes that “pure random walk theory doesnot apply to the capital markets. The capital markots instead follow a biased random walk.” The Theory of Secial Imitation suggests that, at times, the market is in a true random-walk state, IF these periods of true random walk are eliminated, the remaining periods of crowd behavior may be expected t exhibit even greater persistence (le, return trends) than Peters’ data indicate. ‘This conclusion would have important implica: tions for market timing and asset allocation. Techniques such as tactical asset allocation may not be futile in periods of coherent markets However, when true random-walk markets pre vail, trend-tollowing strategies and relative. strength approaches will nat be effective, ‘A Theory of Social Imitation More than 10 years ago, A. Woodcock cbserved that “certain events, ‘such as steck market crashes and suciden structural failures. are to tally unpredictable. But that could change if a revolutionary mathematical theory lives up to its promise.” The revolution referred to was tho emerging field of catastrophe, or chaos, theory. ‘Chaos theory in general includes 2 wide vari- ety of non-linear models, all of which are deter iinistic. This poses « problem in terms of the practical application of chaos theory to the stock ‘market, inasmuch a8 the market is heavi fluenced by random noise. On Wall Street. the element of chance must be an integral pait of any new theory of stock price Factuations. The Theory of Social Imitation is a non-linear statistical model based on the Fokker Planck equation. (See the appendix for mathemavical details) As it intudes both a stastical version of catastrophe theory and the random valk a3 special cases, it avoids the limitations of deter- rinisticchacs theory models. Its therofore well ‘quipped to deal with tock market fluctuations ‘The Theory of Socal Imitation, developed by orl Callen ane Don Shapero, extends the well mown Ising model of ferzomagneticm to the phenomenon of polarization of opinion in social groups." Wolfgang Weidlich ofthe University of Stuttgart originally proposed this appreach to describe intense polarization of opinions in so- al groups, such 25 aroso in the French student revolution of the 1940s” The Ising Model To appreciate the Ising, model, consider a bar of iron in which individual molecular magnetic spins point either up or down. IF the bar is hot, random bombardment by neighboring mole: cles will cause molecular orientation to become disordered. At times, just by chance, mare molecules may point up than down, or versa. The macroscopic magnetic field will fluc: tuate randomly around 2270, ina state of disc der. When the temperature of the iron bar lowered below a critical point, the magnetic interaction between adjacent molecules begins to become stronger than the random thermal forces. If, by chance, asinall cluster of molecules begins (0 orieat in ‘one direction, neighboring molecules wil tend to fellow suite. Soon numer fous large clusters may form. Each cluster will have molecules aligned, but some clusters will align in one direction and some in another. On ‘a macroscopic level, there may be large, long: lasting magnotic field fluctuations. On average, hhowover, the net fold strongth will still be zero, unless there is an external bas tending to align the clusters in one way rather than another. When a bar of iron is susceptible to magne zation, an external magnetic force will tend to align most clusters in one direction, Random thermal forces will still cause some variations in the macroscopic magnetic field, but the fuctua- tions will be stable around some large net value. In this situation, the iron has a high degree of ‘order. The orientation of the magnetic fel, however, could “lip” as a result of extemal forces or even Just as a matter of chance. Polarization in Social Groups Callen and Shapero suggested that the Ising ‘model could be used to describe the behavior of ‘a wide variety of socal groups—tfish aligned in schools, birds flying in flocks, fireflies ashing. in unison and people conforming te the dictates cf fads and fashions, Their Theory of Social Imitation is based on the assumption that, on a ‘macroscopic levei, individuals in a group be- have in a manner similar to the molecules in & tear of ion. Under some conditions, the individ uals will actindependently of each other. Under the right conditions, however. the same individ- vals’ thinking may become polarized—i.e., the individuals will act as a crowd and individual ‘ational thinking wil be replaced by a collective group think.” [RNANCIAL ANALYSTS ]OURNAL NOVEMDEX DSCEMBER 198/038 [As Charles Mackay observed in the 19th een: tury: “Men it has well been said, think in herds; will be seen that they go mad in herds, while they recover their senses slowiy, and one by cone.” Likewise, a bar of iton susceptible to magnetization by an exiernal impulse (magnetic field) will become strongly polarized for an exienided period of time and siowly return tothe unpolarized state, long after the extemal influ: fence has passed In general, transitions from disorder to order tend to share the same macroscopic characteris tics, whether the subsystems are from physics, chemistry, biclogy or sociology. Even though the magnet and social groups have vaslly dit ferent subsystems, their macroscopic behavior shares the same properties of all transitions from disorder to order. These have been de- scribed by Haken and are summarized in the ‘quantilative models in the appendix."® Stock Market Fluctuations Unlike @ bar of iron, the stock market is an “open” system. It zoquires a continuous flow of money to maintain a transition from a disor- dered to 8 more ordered state, just as the laser requires an external pump to create the popula- tion inversion needed to maintain a fresh sup- ply of electrons ready to emit light. lf the flow of ‘energy in a laser is insufficient, it will only emit normal, “random.” light. One ofthe surprising features of the laser transition, however, is that it is remarkably similar to the magnet’s siate change, even though the underlying process is vastly difforent and the laser is far from thermal equilibrium.” assume that industry groups in the stock market axe analogous to the molecules in a bat cf iron, and that the retura from the stock ‘market is proportional to the difference between the number of industry groups trending higher and the number trending lower, Market returns may either fluctuate randomly around zero (as in the overheated bar ofiroa) o:, under the right conditions, they may exhibit a high degree of polarization, leading to a large net difference between gainers and losers and corresponding. Taxge market moves. The Ising model, as applied to the stock market, has thiee key inputs. The first is senti- ment—k—a measure of whether the level of “group think” is above or below e citical tran sition threshold. The second, fundamental bi- as—h—is a measure of externel preference Figure B Major Market States {amental [Gehan Fonte atictee sat tow i, gh eva tne =a a, edt it ae ows [Ean eae toward bullish or bearish sentiment. The third— n-represents the number of degrees of free- dom that exist. For the stock market, [interpret faa the number of industry groups. While k and h may vary widely, n remains relatively Constant. (It is assumed to be 186 in sampie calculations.) Figure B summarizes the market slates ex- pected as a function of prevailing investor sen- fiment and fundamentals. Below the critical transition threshold (k = 2), the random walk slate prevails. Above the transition threshold, coherent bull markets occur if fundamentals are strongly positive, while chaotic markets are likely if fundamentals don’t provide a clear direction for investors, Tabie Il presents theoretical expected returns and standéed deviations for a range of combine- tions of sentiment and fundamentals. For equal changes in fundamentals, h, expected refum increases in a norvdinear fashion. In a random- Table RskReward Frccas Tas Soninere Tusainenls Expetel Strand sire) Qo fetus _Devazion Rendon ke 0a TO Wak 18 ° tw re be Ymiton 2000 Su 20 3 tie 2 be OB 220 Thm te 2200 to eB 22 Sham he 22 a es 2200 hms a re Coherent 32000 Tak Bar 3200 as 8 novnaer oecenmen 1100 40 Figure © Random Wak (Eifcient Market? Be Annuaiued Rots Seatner(h) = 18 fndhmeras fy = eapCEN eum» @ sande GeEREN = walk market, the impact of a change in funda mentals 3 considerably less than it io during period of crowd behavior. Furthermore, in co” herent markets, the magnitude of the expected return is more than twice its standard deviation, suggesting a potential quantitative definition of coherent behavior in capital mackets Random Walk is Just the First Step! The non-linear model given in Table I can be considerably simplified for the special case when sentiment is not conducive to crowd be- havior—ic., k <2." If we further essume that fundamentals aze neutral (h ~ 0), the probatil- ity distrbution of returns, f(q) can be expressed a3 follows: fq) = Wig siexpt — 4g. 0 where the variance is #=1@- kin. 2 ‘This isthe “normal distribution corresponding toa snapshot in time of a random walk process. In general, the random walk involves time. dependent diffusion (widening) ofthe probabil ity distribution. The nowlinear model in Table 1 represents a stationary solution to the Fokker Planck equation (which is a generalized form of the Langevin equation of Brownian motion) Time-dependent, non-linear solutions are treated ae random walks involving both the usual time-dependent diffasion and a coherent cérft toward che most stable stationary state." When nonlinear effects are significant, tan sition probabilities of steps up oF down are no longer equal. In effect, steps in one direction ‘may be larger and more likely chan steps in the TBIAMCIAL ANALYSTS JOURNAL, ‘other direction, This amounts to a biased ran- dom walk, where the bias may behave either in coherent or chaotic fashion depending on pre- veiling sentiment and fundamentals, While the non-linear model reduces to the usval random. walk as a special (linear) case, the general non- linear, time-dependent situation i quite com- plex Figure illustrates both a potential well and a probability distribution for the special case of a true random-walk market. The annualized re- turn from a market index, q, can be thought of asa ball tapped in the potential well, bulfeted by random forces. The probability ditsibution is determined in part by the shape of the well. In 2 true random-walk state, the well has a single bottom near zero, representing a disordered state, Theoretically, random-walk markets ‘could provide cither a small, stable bullish re- turnora smal, stable bearish return, depending fon fundamentals. Historically, however, ran dom-walk markets have provided a stable neg~ ative return and are most frequently associated with bear markets. Transition to Crowd Behavior ‘The variaace in the normal probebilty disti- bution in Equation () becomes very large as k approaches two, the critical transition thresh- Old. In this situation, the normal distribution no longer applies. The rarrdom-walk model i no longer valid during the transition to crowd behavior An instability occurs at transition, In a ran dom-walk market, the motion of @ ball in the potential well would be heavily damped; thats, the effects of each random pash (piece of news) fon the ball (return) would die out (be dis- ovesaen oecennea v0 41 Figure D_ Urata Trnsitiont Amz Rom (3) Samaneoug “Sentmeat = 20; ena) counted) quickly. At transition, however, the level of damping drops dramatically. The ball is free to swing from one extreme to another within the potential well. This implies a highly inefiicient market in which lange, long.lasting sentiment swings must be expected. Figure D illustrates the potential well and probability distribution associated with the un: stable transition from random walk to crovd ‘behavior. The potential well forthe transition to ‘crowd behavior is nearly fat over a wide range cf expected returns. Nearly anything can hap- pen in a period of instability. The case shown is {for neutral fundamenials; even a slight funda- rental bias would tend to skev the distribution strongly in the direction of the bias. Chaotic Markets As k increases above the critical threshold of two, the Ising model predicts a double-bottem ‘potential well and bimodal probability distritu- tion. Thats, a high degree of polarization exists among investors, but without a strong funda- 1 eapeted ean = sana vito = 6% ‘mental biss, there is no clear indication as to ‘whether the crowd will stabilize in a bullish or bearish state. Furthermore, there is @ possibiiry of abrupt sentiment shifts (chaotic Auctuations) from bullish to bearish, of vice versa. The probability of a large sentiment shift is sgreatesi when prevailing investor sentiment runs counter to a small external bias in funda- ‘mentals. Figure F ilustates, for example, the potential well and probability distribution appii Gble to the period immediately prior to the Crash of 87. ‘The climate prior to the crash was conducive to crowd behavior; this was indicated by the ‘extremes in market volume and breadth, Fun damentals were somewhat bearish, a3 areault of rising interest rates throughout 1987 as well as tunasually high valuations by historical norms. ‘The hike in the discount rate on September 4, 1987 clearly signalled the Federal Reserve's in: tent to further tighten monetary conditions. At ‘the same time, the market had risen more than 25 per cont during the prior 12 months, and Figure E —Chaote Marke* 7 3 8} pom ooh bom 2 ous t on # -o0e om F 8 oa oat LZ o a 7 Tae Anoatzed Return) Sane) = Fe naam)» 89 cpeaelecum = ON sun Ean = FINANCIAL ANALYETS JOURNAL NOVEMOER DECEMBER 190042 Figure Coherent Bit Nake c Te vot Tiss gock f og 4-003 ow £ 3 oon L . Tra E 005. 0.01 - x oe ~Be © +5 somal enim) bullish sentiment prevailed, The market, viewed as a Brownian particle, was in the less probable state. While a bullish state is quite possible when crowd behavior is combined with bearish fun: O and <0), the normal distibution bifurcates, or splits into a symmetrical bimodal distnitution. Therefore Equation (A2) may be regarded as a statistical version of chaos theory. In general, chaos theory involves a wide va riety of potential functions, The Theory of Social Imitation, summarized in Table I, is another ‘example of statistical chaos. Hiere the potential function is moze complex than the simple exam- ple in Equation (Al) and requires numerical solution by computer. Footnotes 1. HL Haken, Synergctis, Non-qui@rium Phase Trar- sitions and Sef Organiatior in Physics, Chemistry, and Biolegy (Now York: Springer-Verlag, 1978) 2 E, Callen and D, Shapero, “A Theory of Soc Imitation,” Piysis Tey 27 (1974), No. 7. 3. T. Vaga, ‘Siock Marit Fluctuations,” letter, Physics Tuay 32 (1978), No.2, p. 80. 4 E. Peters, “Fractal Stricice inthe Capital Mar- ets," Flan Analysts Journal Guly’August 1589), 5. A. E.R, Woodcock, “Catastrophe Theory: Pre dicing the Unpredictable,” Machite Design, Feb- rey 10,1977 6. CNeoGiennger, “The Butterly Effet: Can tiny tremors cause market chaos? (And cin computer modeling predict i)," Ovals, Third Quarter 1680, 7. W. Wellich, “The Statistical Description of Pe- larization Phenomena in Society.” Frits Journal NOvRNMER-OECENAR 180 48

Вам также может понравиться