Journal Editor & Reviewers

Behavioral Finance and Technical Analysis 5
Interpreting Data From an Experiment on Irrational Exuberance, Part A:
Applying a Cusp Catastrophe Model and Technical Analysis Rules
Henry O. Pruden, Ph.D.; Dr. Bernard Paranque; Dr. Walter B

The Deviation Oscillator (DO) 13
Saleh Nasser, CMT

Momentum Leads Price: A Universal Concept With Global Applications
Timothy W. Hayes, CMT

Editor

Charles D. Kirkpatrick II, CMT
Kirkpatrick & Company, Inc.
Bayfield, Colorado

Associate Editor

Michael Carr, CMT
Cheyenne, Wyoming

A Cusp Catastrophe Model from the behavioral sciences provides a positive scientific theory as to the “why” of behavior in a stock market. Technical market analysis furnishes a nominal theory of rules and principles about “how” a trader or investor may profit from the behavior observed in a stock market.

Introduction

While watching the 1997 McNeil-Lehr Newshour Video depicting the California Institute of Technology Experiment on Irrational Exuberance in securities trading, we became intrigued by the apparent strong parallels between the trading behavior exhibited in the video film and the idealized graphical model of the Cusp Catastrophe Model presented in Christopher Zeeman’s 1977 book Catastrophe Theory Selected Papers, 1972-1997 (See Figure 1). Investigative attention was focused upon the fold or cusp in the model which captures the transition from bullish behavior to panic conditions. As outside observers, we agreed with all of the comments made by Professor Charles Plott (McNeil Lehrer, 1997) of Cal Tech during the experiment, except for one critical exception. Professor Plott claimed that there was no way that a participant or observer could have predicted the break in prices and, hence, no way for any of the players who were in the experiment to have capitalized upon the bubble’s burst. We disagreed with Professor Plott’s conclusion. Instead, we hypothesized that according to Cusp Catastrophe Theory a “dissipative gradient” occurs in the behavior pattern of the trades in the experiment just before the catastrophic plunge in price. Hence, the break in price action could have been anticipated. Furthermore, we anticipated that close examination of the price behavior within the zone of the Cusp, or threshold, would reveal behavior patterns that could be profitably analyzed and interpreted according to the rules and indicators of technical market analysis.

Figure 1. A Cusp Catastrophe Model of a Stock Exchange

Technical market analysis and “behavioral finance” are similar in their roots. Both are rooted in the assumption that man acts for behavioral reasons in ways that, by the standards of classical economics, may seem irrational. Both approach the study of markets to identify patterns of human behavior that uncover opportunities for profits. “Technical market analysis” has existed as a practice in real world financial markets for over a century. It, too, has theoretical roots in psychology and sociology which are often overlooked by the practical men and women of action who practice investing, trading and analysis. If we envision a theory-application spectrum, we can see “behavioral finance” occupying the theoretical pole while technical market analysis occupies the practical applications end of the spectrum.

Arguably practitioners and students of technical market analysis championed the center stage of behavioral finance long before the arrival of what recently has become known as “behavioral finance.” In his 1969 book, Stock Market Behavior: The Technical Approach to Understanding Wall Street, Dr. Harvey Krow defined technical analysis as synonymous with “behavioral finance.” In his preface to that book, Dr. Krow identified three competing schools of thought in finance: fundamental analysis, the random walk, and the behaviorist. Technical analysis fell within the behaviorist or behavioral school, concluded Dr. Krow.

The prominence of behavioral finance grafted to technical analysis was boosted by October 9, 1993 issue of the Economist magazine. In that article, the author Matt Ridley observed the linkage between technical analysis and “behavioral finance.” Mr. Ridley stated that a combination of computer horsepower and mathematical brainpower had made it possible to find new sources of profit in the forecasting of financial markets. As the author stated: What the new mathematicians are mining for is not inefficiencies in the flow of information but something entirely different. They have found new meat in the familiar fact that traders are a diverse bunch; by unearthing some of its previously unrecognized effects...the most popular idea for explaining it has to do with the heterogeneity of traders in particular, the fact that people reason differently about the information they receive, that they have different time horizons...and that they have different attitudes to risk... The efficient-market theory is...right that efficiency will delete time-arbitrage opportunities based on who does not have information, but wrong to conclude that therefore the market cannot be beaten.

Ridley emphasized, “Prices do contain hints of what they will do next. Computers have resuscitated chartism.”

As the Ridley article emphasized an appreciation that technical analysis was evolving through the attempt to predict prices using computers to study market behavior. For example, moving average timing and break out signals produced profits by more than by chance.

As Ridley concluded, “Chartists – who prefer to be called technical analysts – justify their techniques with quite reasonable arguments about the behavior of investors. They do not claim to predict the behavior of the index so much as the behavior of the people who trade in the market...a rising price is a band wagon.” And there are models from behavioral science that capture band wagons. Technicians were studying the behavior of people who make markets run. Behavioral Finance and Technical Analysis Interpreting Data from an Experiment on Irrational Exuberance, Part A: Applying a Cusp Catastrophe Model and Technical Analysis Rules Henry O. Pruden, Ph.D. Dr. Bernard Paranque Dr. Walter Baets Visiting Scholar Head, Finance and Professor of Complexity and Information Department Knowledge Management Behavioral Finance: Cusp Catastrophe Model

“Catastrophe theory is a new mathematical method for describing the evolution of forms in nature. It was created by Rene Thom who wrote a revolutionary book Structural Stability and Morphogenesis in 1972, expanding the philosophy behind the ideas. It is particularly applicable where gradually changing forces produce sudden effects. We often call such effects catastrophes, because the lack of intuition about the underlying continuity of the forces makes the very discontinuity of the effects so unexpected. The remarkable thing about the results is that, although the proofs are sophisticated, the elementary catastrophes themselves are both surprising and relatively easy to understand, and can be profitably used by scientists who are not expert mathematicians” (Zeeman, 1977). A Catastrophe Theory Model modified for the explanation of the evolution/ revolution of behavior in the securities market can be classified in the realm of behavioral finance. (See Thaler, 1993; Statman, 1998 and Pruden, 1989). An early model of the Cusp Catastrophe Model modified to explain speculative crashes appeared in Zeeman (1976, 1977). Later, Pruden (1979) expanded upon Zeeman’s use of the Cusp Model version of Catastrophe Theory to allow for “buying stampedes” as well as “selling panics.” Pruden (1980) also established connections between the Cusp Catastrophe Model and technical market analysis. Whereas the Catastrophe Theory Model, like other models from the behavioral sciences, provides a positive scientific theory as to the “why” of behavior in the stock market, technical market analysis furnishes a nominal theory of rules and principles about “how” a trader or investor may profit from the behavior observed in the stock market. Hence, the presupposition is that behavioral science models that explain the stock market behavior provide solid scientific foundations upon which to base the principles and practices of technical market analysis.

The Cusp Catastrophe shown in Figure 1 offers a unique three-dimensional graphic model for structuring two independent and one dependant variable. It furnishes a basis for classifying and interrelating price trends and sentiment variables, thereby enhancing logical clarity and empirical predictability. Implicit in the model is a fourth temporal dimension.

EQUILIBRIUM SURFACE

The Cusp Catastrophe model posits two parallel surfaces. The upper behavior or equilibrium surface is represented by a price index such as the Dow Jones Industrial Average. This behavior surface is further subdivided into a top sheet representing bullish behavior and a bottom sheet reflecting the dominance of bearish behavior. Each point on the behavior surface is an equilibrium juncture between supply and demand, even though incremental and transitory. Near the center of the behavior surface of the model lies the Cusp Catastrophe’s most interesting feature – a fold curve or cusp. What this suggests is that there is no equilibrium (horizontal range) available until the top sheet is reached after a buying stampede or the bottom sheet is reached after a selling panic. Notice that the abstract model shows the behavior surface curving over to a threshold point, after which comes the panic sell-off. In the Cusp Catastrophe model this all-important juncture along the top sheet is known as the “dissipative gradient.”

CONTROL SURFACE

The market price or the equilibrium behavior surface is the dependent variable. The independent, predictor or control variable, which accounts for the index or to which the index may be ascribed, lies on the control surface below. In Figure 2, the independent, predictors are shown as the emotional forces of fear and greed.

The model featured in Figure 1 presents fear and greed as opposing factors. The relative power of these two opposing forces is what animates market behavior. The gradual changing relationship between fear and greed gives rise to sudden discontinuations in price behavior when thresholds are reached and panics or stampedes ensue.

Figure 2. Dissipative Gradient

CUSP MODEL IN OPERATION

Now let us imagine Figure 1 in operation. The flow of the market index takes place over a smooth surface composed of equilibrium points. Changes in the control variables, fear and greed, have unique responses on the behavior surface. The dynamic process of the model causes the index to seek out local points of stable, albeit temporary, equilibrium.

Starting at a bear market low, where the market index is on the lower attractor sheet, the level of greed (demand) is suppressed by the level of fear (supply). Mounting greed (e.g., expectation of higher prices) gradually overcomes fear until the edge of the sheet is reached, at which point the market breaks out of an upside reversal pattern via a catastrophe jump to the top sheet as the mood of the market becomes decidedly bullish. The index then flows along a rising channel on the top sheet until the bullish potential is exhausted. At that point, both greed and fear are high. Finally, as fear overcomes greed the market index is pushed to a threshold on the top sheet, then the price index plunges to the bottom sheet via a bearish catastrophe jump.

Catastrophe Theory analyzes equilibrium and its breakdown. As such, it is ideally suited for understanding the stock market where price movements result from the balances and imbalances between buying power and selling pressure, which in turn are animated by the forces of greed and fear. Applications of Catastrophe Theory can be qualitative in nature. Catastrophe Theory does not pretend to render pinpoint or unalterable predictions far in advance. The theory does not negate the art of interpretation.

In Catastrophe Theory the prior history of behavior states of the market is required to predict the future. This undercuts the assumptions of the “random walk” or efficient market hypothesis. Catastrophe Theory underscores the relevance of the historical, chart approach to analyzing the market. The Cusp model encompasses duality and opposition. There is room for a greed axis and a fear axis. It brings the opposition between bullish versus bearish sentiments into clear relief.

The Cal Tech Experiment on Irrational Exuberance

When the internet bubble burst there was a massive opportunity to make serious money through short selling or at least avoiding losing money already earned. A predictive theory that would have alerted a trader to the potential collapse would have been extremely valuable.

The Cusp Catastrophe Model could have been that predictive theory. The Cusp-Catastrophe Model is based upon behavioral science/behavioral finance to explain types of non-linear, discontinuous behavior. It is especially models behavior of rapid change, such as a stock market bubble bursting. Catastrophe theory has revealed that sudden change and behavior extremes are not only natural and interrelated but, if one were to see the early warning signs, a collapse would be predictable.

The Cusp Catastrophe Model posits that behavior is driven by fear and greed. In the case of a stock bubble, price climbs along the top layer of the Cusp Model. Eventually the speculative excess reflects increasing nervousness and starts to de-escalate, moving toward the drop-off at the cusp. Within the cusp itself there exists the small, incremental change downward, the “dissipative gradient,” that marks the beginning of the collapse. Afterward behavior then suddenly drops off the cusp and falls vertically in rapid collapse (See Figure 4). Market behavior explainable by the Cusp Catastrophe Model was evident after the U.S. stock market run up in late 1999 and early 2000. The market had become extremely overpriced and signs of nervousness started to appear. There was one market session in which the NASDAQ dropped over 500 points only to make a surprising and outstanding recovery back before the end of the session. The nervousness depicted by this sudden and dramatic price drop was a sign of the impending collapse. The dot.com bubble having reached the cusp, it was period for a catastrophic decline.

THE EXPERIMENT

Using the Cusp Catastrophe as a framework, we interpreted the research data from a Cal-Tech Experiment on Irrational Exuberance that was produced by WGBH television and shown on PBS (McNeil Lehrer, 1997). The Cal Tech experiment furnished empirical data to “test” propositions derivable from the Cusp Catastrophe Model. The experiment likewise offered an opportunity to extract and highlight several nominal rules/indicators of technical analysis that fit with the logic of the Cusp Model. The indicators of technical analysis that fit with the Cusp Model were then also applied to the data of the experiment in an effort to anticipate and profit from the catastrophic decline in price that followed the bursting of the speculative bubble created during the experiment. To further our research efforts, we stopped the video of the Cal Tech Experiment at key junctures in order to photograph the charts that had recorded the behavior of the traders during the experiment (See Figure 3 and Figure 4). We blew up the pictures from the video that showed the transition from a bull market to a bear market. Our interpretation of the expanded photos of behavior led us to conclude that the Cusp Catastrophe Model coupled with technical analysis principles and indicators could have explained the experimental data and exploited the trading action generated by this laboratory experiment on irrational exuberance. The data indicated that, as anticipated, a dissipative gradient precedes a catastrophic collapse in price.

Professor Charles Plott, California Institute of Technology, conducted his experiment with well trained, knowledgeable Cal Tech students. The students had experience with similar experiments but they had not been exposed to the exact parameters of irrational exuberance experiment. All the students were by definition very bright, very rational individuals who were oriented toward making the most profit available within the context of the risk they perceived. Trading was done on a closed network of computers and students were allowed to buy or sell one stock. The better they traded, the more money they could make up to several hundred dollars. The students who held the stock at the periodic divided payment junctures were the ones who would win the game. The stock being traded was a fictitious oil company and to make things simple it had only one oil well. As oil was pumped from the well, the stockholder student was paid dividends at pre-set intervals. When the oil ran out, the company was basically worthless.

Figure 3. The Overall Results of the Experiment

Figure 4 depicts the overall results of the experiment. It shows the average value of the company as it depletes the oil (line A) and the absolute maximum value of the company with oil at its maximum potential market value (line B). Line A should represent the average stock price and line B should be the maximum price over achieved. Stock prices beyond line B were not rational because everyone in the game knew that the value above line B was beyond any underlying asset value. At set intervals, the dividends were paid. These dividend payment intervals are shown in Figure 4 by the dashed vertical lines. All participants in the experiment knew this information before the game was played. The prices established by the buyers and sellers in the experiment did not drop as would have been expected from the logic of rational economic analysis of the situation even though all players were rational and had the same information. The traders in the Cal Tech experiment persistently traded at a prices that were greater than the fundamental value indicated the company was worth. As the experiment progressed, the traders in the experiment ignored the average value line and then, surprisingly, crossed the maximum value line. The students in the experiment paid for the stock in the experiment well beyond what even the most optimistic investor should have paid. Apparently, chasing dividend distribution dates, they continued to trade was based upon the greater fool theory. It was rational to buy overvalued stock so long as someone else would buy their overvalued stock later on, after the dividend had been collected, thus allowing the trader to continue to profit from dividends with little risk.

Eventually, as the oil well neared depletion, the market began to show signs of nervousness. This nervousness by players in the experiment was very evident on the videotape since Professor Plott had tied price bids to purchase the stock to lower sounds on the musical scale. The high notes on the chart reflected sell offers while the low notes were bid orders (See Figure 5). As the market neared extreme upside valuations, there arose heightened nervousness evidenced by a striking increase in the intensity of the lower notes. Both sellers and buyers were shifting their expectations downwards apace with lowering tones and the sound volume level increased rapidly. Such a change in the sound of the market, the sentiment, has been often noted by traders on the floor of the exchange as a harbinger of a reversal of price trend. As the experiment progressed the buy offers that were well below the existing price began to increase, although the price level itself stabilized into a horizontal trend channel. Ultimately there occurred a sudden, sharp drop – the catastrophic jump – in the transaction price in the experiment. The market changed suddenly and swiftly; sentiment flipped from bullish to bearish as the price plunged to its underlying economic asset value.

It should be mentioned that this experiment was conducted without exogenous factors. There were no news or media reports, no external noise, and no one was allowed to voluntarily enter or leave the game. These restrictions may have contributed to the stability of the price data along a horizontal trend channel rather than prompting price to oscillate upward and downward as time progressed.

Figure 4. Applying Technical Analysis

Even though Professor Plott asserted that the sudden and dramatic shift that was not predictable, looking at the data with the aid of the Cusp Catastrophe Theory reveals that there was a tip-off before the tumble. This tip-off was to be expected by the curve of the “dissipative gradient” at the cusp of the model.

Applying the Cusp-Catastrophe Model

What occurred in the Cal Tech exercise was ipso facto an experiment demonstrating the elements and efficacy of the Cusp-Catastrophe Model. The market moved along an elevated course until it met with a bifurcation point. That point was the maximum expected value or line B on Figure 3. The group of students participating, motivated by greed, collectively decided to continue buying despite shrinking oil well resources. This build up of a speculative bubble can be seen as taking place along the top sheet of prices on the Cusp Catastrophe model. The price moved smoothly along the upper level of the sheet until the cusp. Then as fear started to play a stronger role while buying intentions were becoming exhausted (fear was overcoming greed as a collective motive) the threshold of the cusp was reached. Thereafter, the price plunge in the experiment can be explained as the dropping off from the upper level of the cusp in a dramatic swoon to lower level as the selling panic, the downward catastrophe jump erupted.

APPLYING TECHNICAL ANALYSIS

The Cusp Catastrophe Model itself and the application of the Cusp Catastrophe Model to the Cal Tech Experiment on Irrational Exuberance spotlighted the efficacy of six principles of technical analysis and trading that are well known but often overlooked or under-appreciated by technicians and traders. These principles of technical market analysis and trading are:

• Fear vs. Greed Juxtaposed
  

• Trading Range Channels Along Tops and Bottoms
  

• Descending Price Peaks: Dissipative Gradient
  

• Catastrophic Panics Causing Price Gaps
  

• Mental Discipline Needed to Win the “Greater Fool” Game

These six principles could play an analytical role alerting a trader, a participant in the Cal Tech Experiment, when to abandon playing the “greater fool theory” game. These principles of technical analysis and trading were instrumental in the diagnosis of the “dissipative gradient” and thus the prognosis of the decline (See Figure 4).

FEAR VS. GREED JUXTAPOSED

The Cal Tech experiment vividly revealed the classical role of sentiment as the musical notes depicting bids and offers reflecting sentiment shifted downward before the downward slide in price (Figure 5). The anticipatory role of sentiment played on key. Additionally, the picturing of the fear and greed variables sentiment as opposing forces in the Cusp Model was brought into dramatic display by two high notes versus the low notes in the Cal Tech experiment. The Cusp Model revealed a new and powerful way for practicing technicians to display and interpret indications of sentiment.

A tight trading range of channel of price behavior, predicted by the Cusp Model, was created by the student investors in the Cal Tech experiment (Figure 6). That the price behavior adhered so closely to a linear trend channel was surprising to the author, who expected to see broader and more jagged up and down swings in price. The linear trend channel of price behavior occurring during the experiment upheld the technical analysis practice of drawing linear trend lines of support and resistance. The breaking of a “support line” drawn along the horizontal price bottoms of the price channel created during the experiment constituted the crossing of the Cusp and the onset of the catastrophic downward plunge in price.

FIGURE 5. FEAR VS. GREED JUXTAPOSED

FIGURE 6. TRADING RANGE
Descending Price Peaks: Dissipative Gradient

Our expectation of a pattern of “descending price peaks” within the trend channel but before the price break was a key reason why we had disagreed with Professor Plott’s assertion that the break in price was unpredictable and unbeatable. As he opined, there was simply no way to get out on the way down. However, technical analysis with aid of the Cusp Model’s, “dissipative gradient” of descending price peaks led us to expect a window of opportunity that would alert a few astute traders to exit before the crash. The evidence from the Cal Tech Experiment on Irrational Exuberance confirmed that expectation: the downward plunge in price at the end of the Cal Tech experiment was predictable (Figure 7).

Descending price peaks were long ago recognized by such technical analysts as Richard D. Wyckoff as a reliable pattern for prognosticating of behavior for lower prices to come. The repeated attempts to rally which failed to reach previous price levels (i.e., lower price peaks) showed that demand was reaching exhausting. Greed/bullish sentiment was no longer supporting the elevated price, hence a price drop was about to occur.

Figure 7. Descending Prices Peaks

Figure 8. Catastrophe Panic Causing Price Gaps

The pattern of descending price peaks occurred in the experiment it was reminiscent of the right-hand side of the classic price-reversal patterns analysis employed by technical analysts. For example, within the classic head-and-shoulders top formation, the technical-analyst-trader is counseled to enter a short position on the third rally or pullback to the neckline of price support. Prior rallies to higher prices would have been to the right shoulder and to the head of that formation. In sum, the Cusp Catastrophe Model reveals the “triple descending peaks” pattern as a powerful technical tool. In our judgment, technicians tend to overlook and underappreciate the pattern of descending peaks as a tip off of weakness and harbinger of panic.

Once the panic decline gets underway, the scramble of offers to sell coupled with the withdrawal of bids to purchase, leads to repeated “air pockets” or gaps in price on the way down in price Data from the experiment revealed a gapping phenomenon more pronounced and prevalent than the time honored “break away,” “measuring” and “exhaustion” gap trio of technical analysis (Figure 8). The “breakaway gap” corresponds to the catastrophe jump across the threshold.

Figure 9. Mental Discipline Needed to Win the “Greater Fool” Game

The sixth technical analysis principle enumerated above, “mental discipline needed to win the ‘Greater Fool’ game” has more to do with trading discipline than with chart reading (Figure 9). When commenting on the rational versus irrational behavior of the student-investors in the experiment, Professor Plott observed that each person who was playing the game to win was acting rationally. One is tempted to amend Professor Plott’s statements with the words “individually rational but collectively irrational.” To win the game, the student trader had to engage in the risky behavior of buying. Those who did not participate in the game could not earn the all-important dividend reward available to those who did play. The optimal winning mental discipline would have been to play the game in order to have a chance to win and “continue to play with confidence until the sentiment-mood started to shift. Then the trader-student game participant had to depart from the game as price behavior as evidenced by the descending price peaks that follow after the early warning signals of sentiment.” With the empowerment given to an active-aggressive trader by underlying behavioral finance theory like the Cusp Catastrophe Model, and the five technical analysis principles explained above, the participant in the game could have played the game with confidence until the end of the opportunity for profit.

Summary and Discussion

Catastrophe Theory began with the ideas of Rene Thom in the early 1960s. Both the mathematics and the applications were present from the beginning, each stimulating the other, as can be seen in Thom’s classic book on Structural Stability and Morphogenesis. The concept was then popularized by the various works of Christopher Zeeman, most notably in his 1977 book, Catastrophe Theory: Selected Papers 1972-77. Then in 1979 Pruden described the logical linkages between Catastrophe Theory, a dimension within behavioral finance and technical market analysis. The Cusp Model of Catastrophe Theory presents a behavior path flowing along the top sheet until a threshold, a cusp, is reached as the underlying emotions shift toward fear overcoming greed, then suddenly the market price will jump downward. The clues given by the shift toward dominance by supply over demand during the latter stages of a trading range will presage the development of a bearish trend in stock prices. But up until that transition phase threshold point, the market would remain high and delay its descent until remaining pockets of demand were exhausted.

Let us fast forward to the stock market shown in the Cal Tech Experiment and then zoom in on the behavior brought about by bullish emotions vs. bearish emotions during the latter phases in a trading range market. The moral of the jump story applies directly. During the trading range the alternating price swings up and down reflect the struggle between greed and fear, so the analyst - trader must respect the fact that the market could jump either way...its behavior is bimodal. So to be effective the trader must remain neutral until the testing phase on the right hand side of a stock market trading range. Before we reach a conclusion regarding future trend direction, the market should be allowed to define the line of least resistance and then and only then should a position, long or short, be entered. The breakdown was anticipated by the descending price peaks, “the dissipative gradient,” as shown on the Cusp Catastrophe Model. Behavior changes gradually before the breakdown. On the top sheet of the cusp catastrophe model one can see a slight curling over of the behavior path. Stock market behavior would likely show, for example, a series of descending peaks in price. Similarly on the bottom sheet of the cusp model one can see a gradual curling upward of the bottom behavior sheet before the upward jump (breakout) by a series of ascending bottoms during the ending stages or the right hand side of a trading range. These descending price peaks and ascending price bottoms are powerful, but under-appreciated, technical tools.

References

• Krow, H. (1969), Stock Market Behavior: The Technical Approach to Understanding Wall Street, Random House

• Pruden, Henry O., (1999), “Life Cycle Model of Crowd Behavior,” Technical Analysis of Stocks and Commodities.

• Pruden, Henry O., (1979), “Catastrophe Theory: A Model for Stock Market Behavior,” Market Technicians Association Journal.

• Pruden, Henry O., (1980). “Catastrophe Theory: A Practical Application,” Market Technicians Association Journal.

• Pruden, Henry O. (1995), “Behavioral finance: What is it?” Market Technicians Association Newsletter and MTA Journal, September.

• Pruden, Henry O., (2003), “Catastrophe Theory and Technical Analysis Applied to a Cal Tech Experiment on Irrational Exuberance,” Managerial Finance Journal.

• Ridley, M., “Survey: Frontiers of Finance,” The Economist, (October 9, 1993).

• Rogers, Everett M., and F. Floyd Shoemaker (1971). Communications of Innovations, Free Press.

• Schwager, Jack D., (1996). Schwager On Futures: Technical Analysis, John Wiley & Sons.

• Shiller, Robert J., “Stock Prices and Social Dynamics” in Thaler, Richard H. ed, (1993) Advances in Behavioral Finance, Russell Sage Foundation

• Statman, Meir., (1998), “Behavioral Finance,” Contemporary Finance Digest, Financial Management Association.

• Thaler, Richard H., ed. (1993). Advances In Behavioral Finance, Russell Sage Foundation.

• Thom, Rene., (1972), Stabilite Structurelle et Morphogenese, New York: Benjamin Press

• Wyckoff, Richard D. in Charting The Market: The Wyckoff Method, Jack Hutson, Ed. Technical Analysis of Stocks and Commodities.

• Zeeman, E.C., (1977), Catastrophe Theory: Selected Papers, 1972-1977, Reading, Mass: Addison-Wesley Publishing Company (1977), p. 1.

• Zeeman, E.C., (1976), “Catastrophe Theory,” Scientific American, pp. 65- 83.

• Zeeman, E.C., (1974), “On the Unstable Behavior of Stock Exchanges,”
  

About the Authors

DR. HENRY O. PRUDEN

Hank Pruden is a professor in the School of Business at Golden Gate University in San Francisco, California where he has been teaching for 20 years. Hank is more than a theoretician, he has actively traded his own account for the past 20 years. His personal involvement in the market ensures that what he teaches is practical for the trader, and not just abstract academic theory.

He is the Executive Director of the Institute of Technical Market Analysis (ITMA). At Golden Gate he developed the accredited courses in technical market analysis in 1976. Since then the curriculum has expanded to include advanced topics in technical analysis and trading. In his courses Hank emphasizes the psychology of trading and as well as the use of technical analysis methods. He has published extensively in both areas. Hank has mentored individual and institutional traders in the field of technical analysis for many years. He is presently on the Board of Directors of the Technical Securities Analysts Association of San Francisco and is past president of that association. Hank was also on the Board of Directors of the Market Technicians Association (MTA). Hank has served as vice chair, Americas IFTA (International Federation of Technical Analysts): IFTA educates and certifies analysts worldwide. For eleven years Hank was the editor of The Market Technicians Association Journal, the premier publication of technical analysts. From 1982 to 1993 he was a member of the Board of Trustees of Golden Gate University.

Professor Pruden is a visiting scholar at Euromed Marseille Ecole de Management, Marseille, France during 2004-2005.

DR. BERNARD PARANQUE

Bernard Paranque is a doctor of economics ( University of Lyon Lumi re - 1984) and holds the “Habilitation ˆ diriger les recherches” (1995). He began his career as an associate economist in an accountancy firm in 1984. In 1990, he joined the “Banque de France” (French Central Bank) business department. From 1990 to 2000 he produced papers on the financial structure of non-financial companies (www.ssrn.com). He was a representative of the Banque de France in the European Committee of Central Balance Sheet Offices between 1993 and 2002.

In 1999, he was on secondment from the Banque de France to the Secretary of State to SMEs’ where he was in charge of the “business financing” department. He was also a member of the French delegation to the SMEs’ working party of the Business and Environment Committee of the OECD. His research refer to the “ conomie des conventions” and are focused on the financial behavior of the non-financial organization and the promotion of specific tools and assessment procedures designed to enhance SMEs’ access to financing.

He is co-author with Bernard Belletante and Nadine Levratto of “Diversit conomique et mode de financement des PME” published in 2001. He is also the co-author of “Structures of Corporate Finance in Germany and France” with Hans Friderichs in” JahrbŸcher fŸr National konomie und Statistik,” 2001.

He is associate researcher of the CNRS team IDHE-ENS Cachan in Paris and member of the New York Academy of Science. He joins Euromed Marseille Ecole de Management as Professor of Finance and Head of the “Information and finance” department.

DR. WALTER BAETS

Walter R. J. Baets is Director Graduate Programs at Euromed Marseille - Ecole de Management and Distinguished Professor in Information, Innovation and Knowledge at Universiteit Nyenrode, The Netherlands Business School. He is also director of Notion, the Nyenrode Institute for Knowledge Management and Virtual Education. Previously he was Dean of Research at the Euro-Arab Management School in Granada, Spain. He graduated in Econometrics and Operations Research at the University of Antwerp (Belgium) and did postgraduate studies in Business Administration at Warwick Business School (UK). He was awarded a Ph.D. from the University of Warwick in Industrial and Business Studies.

He pursued a career in strategic planning, decision support and IS consultancy for more than ten years, before joining the academic world, first as managing director of the management development centre of the Louvain Universities (Belgium) and later as Associate Professor at Nijenrode University, The Netherlands Business School. He has been a Visiting Professor at the University of Aix-Marseille (IAE), GRASCE (Complexity Research Centre) Aix-en-Provence, ESC Rouen, KU Leuven, RU Gent, Moscow, St Petersburg, Tyumen and Purdue University. Most of his professional experience was acquired in the telecommunications and banking sector. He has substantial experience in management development activities in Russia and the Arab world.

His research interests include: Innovation and knowledge; Complexity, chaos and change; The impact of (new information) technologies on organisations; Knowledge, learning, artificial intelligence and neural networks; On-line learning and work-place learning.

He is a member of the International Editorial Board of the Journal of Strategic Information Systems, Information & Management and Systems d’Information et Management. He has acted as a reviewer/evaluator for a number of International Conferences (e.g. ECIS an ICIS) and for the EURACE programme. He has published in several journals including the Journal of Strategic Information Systems, The European Journal of Operations Research, Knowledge and Process Management, Marketing Intelligence and Planning, The Journal of Systems Management, Information & Management, The Learning Organization and Accounting, Management and Information Technologies. He has organised international conferences in the area of IT and organizational change.

Walter Baets is the author of “Organizational Learning and Knowledge Technologies in a Dynamic Environment” published in 1998 by Kluwer Academic Publishers, and co-author with Gert Van der Linden of “The Hybrid Business School: Developing knowledge management through man- agement learning,” published by Prentice-Hall in 2000. Along with Bob Galliers he co-edited “Information Technology and Organizational Transformation: Innovation for the 21st Century Organization” also published in 1998 by Wiley. In 1999, he edited “Complexity and Management: A col- lection of essays,” published by World Scientific Publishing. Recently he co-authored “Virtual Corporate Universities,” published 2003 by Kluwer Academic

Introduction

The major aim of the Deviation Oscillator, or DO, is to track minor changes in the strength of a trend. It usually does not track major reversals; however, it can be very suitable with countertrend corrections. The DO moves in an unbounded range above and below a zero level and it can be used alone and/or with other indicators. Its objective is to detect weakening bulls or bears as soon as possible. Sometimes sellers begin to weaken, while the price is still declining; the DO will recognize this weakness and will begin showing some bullish tendencies, even before prices begin to rise.

This indicator is derived from a chart with three moving averages; a moving average of the close, a moving average of the high, and another one of the low. It can be observed that the MA of the close deviates between the MA of the high and the one of the low. When prices rise the MA (close) approaches the one of the high, when prices decline, it approaches the moving average of the low. This confirms the notion that during a rise the price usually closes near the high of the day, and vice versa. Based on this observation, the DO was created. Thus, the DO calculates the deviation of the moving average of the close of a certain issue from the moving average of the high and from that of the low. Whenever the moving average of the close of a certain period deviates from the low towards the high it indicates strength. When it deviates from the high towards the low it indicates weakness. The DO is very useful when used with other indicators like MACD, momentum, and the stochastic oscillator. This paper explains its calculation, its basic interpretation, how it can be used in combination with other indicators.

The most important aspect of this indicator is divergences. When a divergence occurs it means that a countertrend move should occur. The DO can be used along with momentum as a confirming indicator, and to filter some of its bad signals since at times DO diverges with momentum. It can also be used with MACD as a setup. A MACD buy signal will be triggered when accompanied by a positive divergence between DO and the price. This gives superior results as opposed to using MACD crossovers alone.

Another oscillator that was extracted from DO is the “RCDO”. It is the Rate of Change of the Cumulative function of DO. This oscillator is mainly used for overbought and oversold conditions. This oscillator and its uses are also explained in this paper.

The Calculation

1. Calculate a moving average of the close, a moving average of the high and a moving average of the low. These calculations use simple moving averages and a time span of 20 days.

2. Calculate the distance between the moving average of the high and the moving average of the close (MA(high)- MA (close)). The greater the difference, the closer MA (close) goes towards MA (low).

3. Calculate the distance between the moving average of the close and the moving average of the low (MA (close) - MA (low)).

Figure 1 shows two lines that intersect with each other. When line 1 (MA high-MA close) crosses line 2 (MA close-MA low) to the upside, then the moving average of the close is closer to the moving average of the low than that of the high. When line 1 crosses line 2 to the downside, the moving average of the close is nearer to the moving average of the high.

Figure 1

Now, to extract the Deviation Oscillator, line 2 is subtracted from line 1: Deviation Oscillator (DO) = [MA (high) - MA (close)] - [MA (close) - MA (low)] This oscillator deviates above and below a zero line. Breaking the zero line to the upside means that prices are getting closer to the lows and vice versa. To make visual inspection easier, the scale is inverted. Thus rises and declines of the DO will accompany rises and declines in prices.

Figure 2

Microsoft chart with DO, after inverting the y-axis. Note that trading on zero crossovers is not recommended as it suffers from whipsaws. Now a break of the zero line to the upside (after the scale is inverted) means that the moving average of the close is closer to MA high than MA low and vice versa. A short cut for the calculation: to avoid inverting the scale, a simpler calculation can be used. (MA close - MA low) - (MA high - MA close). We will not have to invert the scale by using this calculation.

Use of the Deviation Oscillator

BASIC INTERPRETATION

Zero Crossover

Buying when DO crosses above the zero line and selling when it crosses below it proved to be a losing technique, resulting in a total loss of 27.05% for the 30 Dow stocks from 1999-2003. Results improved when a buffer zone was placed at 0.2 and -0.2. Thus the buy signal was not triggered until the upper buffer zone was broken to the upside and the position was closed when a violation of the lower boundary occurred. The loss was reduced to 20%. The results were worst when shorts were added; covering shorts and buying longs above zero and closing longs and building shorts below zero with a loss of 35.5% which was reduced to a loss of 24% by using a buffer zone.

Divergences

This is the most important aspect of the Deviation Oscillator. Even when using DO with other oscillators, divergence analysis is employed. Divergence is very important as it shows that there is hidden weakness or strength in the market that is not apparent in the price action.

• First type of divergence occurs when DO is rising and the price is still declining (positive divergence) or the DO curve is declining while the price is rising (negative divergence). This means (in the case of a positive divergence) that despite that prices are still declining; the closing price is getting closer to the high. The 20-day MA of the close is moving away from the MA of the low and approaching that of the high. The decline is losing its strength, as buyers are able to bring the closing price away from its lows. If the price is declining and DO rising, buy at a breakout of a minor top, with a stop loss below this top or below the nearest minor bottom, depending on risk tolerance. Use this divergence as a setup and buy at a breakout. Usually such a breakout will not be false because it was preceded by some strength. If another indicator confirms this positive divergence, the signal will be stronger. The same holds true when DO declines while the price is still rising. It means that the bears are getting stronger as they are able to bring the closing price away from the highs. One of the most bullish signals appears when DO rises vertically, while the price is still in a trading range or slightly declining.
  

• Second type of divergence appears more often: it occurs when the price makes a lower low while the DO follows a higher low (in the case of a positive divergence), or when the price forms a higher high, while the DO triggers a lower high (in the case of negative divergence). A positive divergence in this case means that during the second bottom the MA of the close was nearer to the MA of the high than during the first bottom. The price violated support but with weaker sellers.
  

Figure 3

Philip Morris (MO) shows a very interesting story. During March 2003, the DO witnessed a positive divergence with the price. During April and May, the stock’s price began to form a higher low, confirming the previous divergence. In May, while the price was trading sideways, the DO moved sharply upwards, hinting of a continuation of the rise. At the end of the stock’s rise, from 18 June, to early July, the DO began to move downwards, while the stock was still rising, signaling potential weakness that came later.

Using the Deviation Oscillator with Other Indicators

USING THE DEVIATION OSCILLATOR WITH MOMENTUM

The Deviation Oscillator can be used as a confirming indicator for momentum divergences. Usually, when momentum witnesses a positive divergence, this divergence will be more meaningful if it is confirmed by a similar divergence in DO. A divergence triggered by both momentum and DO is a strong signal. A positive divergence in this case means that the decline is decelerating; and the MA of the close is getting closer to the MA of the high.

Another way to use DO along with momentum is to track divergences between both indicators. Sometimes the price makes a lower low, confirmed by momentum, which also triggers a lower low formation. This action might not be confirmed by DO, which follows a higher low formation, thus diverging with momentum. Such divergences are very useful as they are usually followed by a minor reversal in the trend. The same can occur at a market peak, when momentum continues making a higher high, while DO follows a lower high formation. It was found, however, that positive divergences give better results than negative divergences.

Testing was done on the 30 Dow stocks to see how positive divergences between DO and momentum affected the price:

• 75.8% of the time a positive divergence between DO and momentum led to a rise in price of more than 5%.
  

• 12.9% led to movements less than 5%, either positive or negative. The divergence had no effect.
  

• 11.29% of the time, momentum was a better indicator and a decline exceeded 5% occurred after a positive divergence.

A positive divergence here means that momentum was triggering a lower low, while DO was showing a higher low formation. The buy signal should be triggered the day after the divergence occurs, or after an up day with high volumes for more confirmation. A stop loss could be placed below the latest minor bottom. Using candlesticks patterns for buying can also be useful. The logic behind such a divergence is as follows: the price is declining, and still accelerating to the downside, however, MA (close), which is already very near to MA (low), begins to move towards MA (high). In other words, buyers are getting stronger as they are able to close the market away from its lows.

Figure 4

The chart of Newmont Gold (NEM) shows a positive divergence between DO and Momentum during July 2002. As the price was making a lower low, momentum confirmed this weakness, while DO triggered a higher low formation. Prices rose afterwards from around 24.5 to 30. During October and early December 2002, both DO and Momentum diverged with the price action, thus confirming each other’s strength. As the price was trying to find support near its first bottom, both indicators witnessed a higher low formation. A healthy rise followed afterwards. In the beginning of April 2003, as price was slightly rising from 25, DO witnessed a sharp rise to the upside, indicating that prices are moving rapidly to their highs. Momentum was rising but at a slower pace. Usually a sudden rise in DO is considered as a very bullish action as it means that the closing price is moving quickly towards the highs.

USING THE DEVIATION OSCILLATOR WITH MACD

One very useful technique is to use DO as a setup for buying and use MACD crossovers for actual buying and selling signals. The MACD crossover method generates buy and sell signals when the MACD line crosses above or below its signal line. The MACD is one of the indicators that give early signals when a new trend is underway. This is a very good merit of the MACD, but obviously a merit that often suffers from whipsaws. Eliminating bad MACD signals by using another indicator enhances our trading results. Using D.O. along with MACD reduces whipsaws.

The tactic used is to buy when the MACD line crosses above its signal line only after a positive divergence occurred between the DO and the price. Exit on the first bearish MACD crossover. The main drawback of this system is that it exits early. Traders and technicians can find better exits by additional research.

Main Rule for combining D.O. and MACD: Buy when the DO creates a positive divergence with prices and the MACD triggers a bullish crossover. Exit after the first MACD bearish crossover. This tactic can also be used with other indicators. The reason why the DO can be used as a setup when it triggers divergences with prices is that its divergences are even more meaningful than momentum divergences. To increase objectivity, three distinct buy signals can be defined, dependent on the behavior of MACD.

“One cross” system is a positive divergence occurs between the DO and the price, and then MACD triggers a buy signal after or during the second DO bottom. The MACD thus gives only one bullish crossover after the divergence between the DO and the price. The buy is executed when the MACD signals a bullish crossover after the DO positive divergence and exit at the first bearish crossover.

Figure 5

The chart of JP Morgan Chase shows the “One cross” buy signal. During February 2002, a positive divergence occurred between DO and the price action. Following this divergence, a bullish MACD crossover occurred (see the vertical line). Only one MACD buy signal occurred after the DO divergence. The two vertical lines show the buy and sell signal. As stated, the exit signal is triggered with the first MACD bearish crossover after a buy is signaled. “Two cross” system occurs when MACD triggers two buy signals. The first crossover coincides with the first DO bottom, while the second MACD crossover coincides with the second DO bottom. During this time, the DO triggers a higher bottom, while the price follows a lower bottom formation (positive divergence). What really happens is that during the first DO bottom the MACD gives a buy signal. During the second DO bottom, an MACD bearish crossover, followed by a new bullish crossover occurs. The trick of this tactic is that it gives us a false bearish crossover; however, using DO in conjunction with MACD will eliminate such a whipsaw. Even if the trader is whipsawed by selling at the MACD bearish crossover, he will quickly re-enter with the new buy signal as it coincides with a positive divergence in the DO.

Thus, the buy signal will be triggered at the second MACD bullish crossover after a positive divergence between the DO and the price. Exit will take place at the first bearish MACD crossover.

The rationale of the “two cross” system is that more strength appeared before the actual buy signal. A bullish MACD crossover during the first DO bottom tells us that there was more strength in the market than the first “one cross” system. Obviously, as the market makes a lower low, the MACD will trigger a bearish crossover, which will be followed by a new bullish crossover while the DO diverges with prices.

Figure 6

The S&P 500 chart shows an example of the “two cross” system. During February and March 2003, DO witnessed a positive divergence with price action. In February, during the first bottom, MACD triggered a bullish crossover. As the price declined during March, MACD witnessed a bearish crossover, followed by a new bullish crossover, which coincided with a higher DO bottom. This is a positive divergence between DO and the price action, followed by a MACD buy signal. The only difference here is that the MACD witnessed two bullish crossovers instead of one. The letters “A” and “B” on the chart show the two MACD crossovers. The two vertical lines show the buy and exit signal. This trade was profitable, however, prices continued moving to the upside after the exit was triggered. As was mentioned previously, the main drawback of this system is that it gives a premature exit signal.

“Cross and a test” system should be expected to give the best results as the MACD witnesses a bullish crossover during the first bottom, but does not witness a bearish crossover afterwards. During the second bottom, and while the DO is positively diverging with prices, the MACD line declines slightly to test its signal line, before rising again. The buy signal is triggered as soon as a positive divergence between the DO and the price is identified and the MACD line moves upwards after testing its signal line. The logic of this system is tha
there was strength from the beginning (bullish MACD crossover during the first bottom) but the temporary weakness was much less than that of the “two cross” system, as the MACD line did not witness a bearish crossover. It only tested its signal line and moved upwards again. The buy will be triggered the second day as the MACD line begins moving upwards once again.

Figure 7

During February and March 2003, McDonald’s chart shows the DO witnessed a positive divergence with price. MACD triggered a bullish crossover during the first bottom, and a test between the MACD line and its signal line took place during the second bottom. No bearish crossovers occurred. The buy was triggered at point B and the exit signal took place with the first bearish crossover afterwards. This system is very profitable and it has the merit of being objective. Its drawback is the premature exit signals that often occur.

The DO and MACD system was tested across the 30 Dow stocks from 1999 through 2003 with relatively good results. The “two cross” and “cross and a test” gave a superior results compared to the “one cross” in terms of % profit per trade. On the other hand, the “one cross” was better in terms of % of profitable trades than the “two cross”. Most of the buy signals fell in the category of “one cross” and “two cross.” A small percentage triggered the “cross and a test” system. Usually when the MACD gives a buy signal and begins to decline again, it will witness a temporary bearish crossover before witnessing a new buy signal.

Overall, 41.4% of the buy signals were triggered by the “one cross” system; 48.7% were triggered by the “two cross”; while only 9.7% were “cross and a test.”

Testing results of the DO versus MACD

RCDO

RCDO is an oscillator extracted from the Deviation Oscillator that is mainly used for overbought/oversold purposes. The RCDO is the Rate of Change (ROC) of the cumulative function of the Deviation Oscillator, in this paper a 5-day ROC is used. RCDO smoothes the Deviation Oscillator. While the cumulative function alone can be used to show longer-term trends, the cumulative curve, does not provide actionable information. Taking its ROC makes it more sensitive. The result is an oscillator that moves slower than the normal Deviation Oscillator but can show overbought and oversold conditions when used in combination with other oscillators.

The ROC of the Cumulative Deviation Oscillator moves faster and when used with the stochastic oscillator it serves as a confirming indicator for overbought and oversold conditions. Obviously, the indicator can be used in many ways and with other indicators; here it used as a confirming indicator for the stochastic and Bollinger Bands. When all three indicators confirmed each other, a buy was triggered. The upper boundary of the Bollinger Bands was used as an exit signal. Obviously, exit signals were not perfect, but the important thing is that nice profitable moves followed buy signals. The technician should find other exit tactics than the ones used here, especially when profitable moves occur.

A buy signal is triggered when the stochastic reaches oversold, the RCDO reaches oversold, and the price tests the lower boundary of the Bollinger Bands. There is a small problem here. The RCDO is unbounded, unlike the stochastic, so oversold can be identified after a certain area is touched at least two times. There are no pre-defined levels to be used as oversold and overbought.

Figure 8

Novellus (NVLS) shows the price, the DO, the RCDO, and the stochastic. During December 2000- September 2001, there was three times where the stochastic and the RCDO reached oversold, while the price was testing the BB’s lower boundary. A rise followed this situation in all three instances. To summarize the trading rules:

• BUY when the stochastic oscillator reaches oversold, the RCDO reaches oversold and the price touches the lower boundary of the ollinger Bands.
  

• EXIT either when the price touches the upper boundary OR when it violates the 20 days moving average to the downside (after rising from the lower boundary and breaking the moving average upwards). Stop loss should be placed below the lower boundary of the Bollinger Bands. This system was tested on the Dow 30 stocks and the results were promising. While 22% of the trades were breakeven trades (plus or minus 5%), 62.5% of the trades were profitable with an average profit of around 14% per trade, while 15% were losing trades with an average loss of 9.1% per trade.: All of the testing was done with visual inspection. More precise testing might be slightly different from the results presented here.

Conclusion

The DO is a tool that can be added to the technician’s arsenal. It is by no means a full system of its own. As all other technical indicators, the DO is a subjective indicator that can be interpreted differently from one technician to another. This paper did not examine weekly and monthly charts, but at times, major divergences are revealed in long-term charts. Figure 9 shows the weekly S&P 500. Bullish action in the weekly DO occurred during October 1990 and Jan 1991, as it was rising sharply, hinting that a big rise in the stock market might be under way.

Figure 9

The DO can be used as a confirming indicator or as a setup for other technical tools (as shown with the DO-MACD system). The default time period for daily charts is a 20-day simple moving average for the high, low, and close. Changing this time period can be done to increase or decrease the sensitivity of the indicator.

It is important to note that one of the strongest signals that the DO presents occurs when price is moving sideways (or slightly declining) while the DO witnesses a vertical move to the upside. This action indicates that the moving average of the close is quickly running towards the moving average of the high, despite the apparent stability in price action. This is a very strong signal of a potential strength.

A final note that is worth repeating here is that the DO proved to be more useful in signaling price strength. A positive divergence is much more important than a negative divergence. DO is best used as a tool to enter the market; however, exits should be based on other tools. Metastock Formulas

Biography

Saleh Nasser, CMT is the Chief Technical Analyst for Commercial International Brokerage Company (CIBC) in Cairo, Egypt. His main job is to recommend to brokers and investors of the company what to buy/sell and how to manage their positions. CIBC is the largest brokerage firm in Egypt. Saleh began working in the stock market using technical analysis in 1997 at United Brokerage Corporation, a small brokerage company based in Cairo. He then worked as Chief Technical Analyst for Fleming CIIC, also in Cairo, Egypt. Saleh is the head of the education committee in the Egyptian Society of Technical Analysts (ESTA), as well as the Treasurer and Member of the Board of Directors of ESTA. He conducted many courses and seminars, teaching brokers, investors, and undergraduate students about technical analysis. Saleh was graduated from Cairo University in 1995 with a BA in Economics.

Momentum leads price. This Dow Theory tenet is the force behind an ever expanding universe of applications, from overbought/oversold and trend-following indicators to trading systems that use moving rates of change to compare individual stocks and market sectors. The benefits of using relative price momentum for U.S. stock and sector selection have received academic attention and have been applied to active portfolio management. But can relative price momentum be applied to global strategy, allocation, and selection decisions? This paper demonstrates that the answer is “yes,” pointing to real-time performance data in detailing the methodologies behind systems that use momentum to rank market indices and sectors from different countries.

The global application of relative momentum has far-reaching implications – the potential uses of momentum as a leading indicator of price changes extend well beyond national borders. While major market moves tend to be global in scope, some markets are stronger than other markets in these moves and some sectors are stronger than other sectors, enabling investors to maximize bull market profits by identifying the leaders and avoiding the laggards. Market profits can be realized by allocating to the right country, sector profits can be realized by allocating to the right sector, and the optimal mix is allocation to the right sector within the right country. An awareness of the weakest markets and sectors can benefit portfolio managers with equity exposure mandates or traders who can profit by shorting in a bear market. The challenge is to identify emerging strength and weakness with accuracy and consistency.

How relative momentum facilitates this process, and why “momentum leads price,” is that it detects acceleration and deceleration. For example, if two cars leave a starting point with the same acceleration, and if the drivers stop accelerating at the same time, the cars will subsequently coast in tandem, all else being equal. But if one driver accelerates, or keeps the pedal on the metal for a longer period of time, his car will coast further. If the driver of the second car starts to accelerate while the first car is coasting, that car could take the lead. When the field includes dozens of cars – i.e., markets or sectors – identifying the accelerators can be next to impossible in the absence of a proven systematic approach, one that has not only performed well in a back-test period, but also stood the test of real time.

The Global Market Ranking

In the late 1990s, we set out to determine if a system based on relative price momentum could be developed using primary indices for markets around the world. Such a system could help identify relative strength shifts among regions, between developed and emerging markets, and among specific countries. We hypothesized that a methodology used successfully to rank U.S. sectors and groups would be effective in ranking the market indices. Specifically, we used a Ned Davis Research approach that applies a momentum composite to each sector or group in a list, sorts the list from the highest momentum composite to the lowest, and applies a strategy that overweights those with strong momentum and underweights those with weak momentum. The use of a momentum composite (for example, the total of four-week, 10-week, and 40-week rates of change) is intended to reduce whipsaws and allow the system to more accurately identify significant changes in momentum.

After identifying 42 markets with sufficient data history for back-testing, including indices for 22 developed markets and 20 emerging markets, we developed a proprietary momentum formula and tested various strategies for overweighting and underweighting. Cognizant that overfitting can lead to disappointment when a system becomes operational, we tested the momentum formulas and strategies in-sample over a period of about two years from early 1994 to early 1996, developed the system, and then confirmed its effectiveness in an out-of-sample period from early 1996 to August 1999. But the biggest test would be the real-time test, the test of whether the system would provide value-added information when put to use in real-time.

Before reviewing the results of that test, let’s take a look at how the system works. Ranking the markets from strongest to weakest by applying the same weekly momentum composite to each market, we developed a system that demonstrates the tendency for momentum to lead price. The optimal results showed that when a market has gained relatively strong momentum, that market has tended to enjoy subsequent and persistent relative strength. Specifically, a market gains overweight status (i.e., it’s a buy) when it rises into the top three of the 42-market ranking, remaining an overweight until it drops out of the top 13. A market drops to underweight status when it falls into the bottom three, remaining an underweight until it rises out of the bottom 13.

Chart 1

(Click Table to Enlarge)

The system is thus symmetrical, as the relative momentum strength needed for a market to gain overweight status is of the same magnitude as the relative momentum weakness needed for a market to enter the underweight category. Accordingly, the number of overweight markets has tended to be in line with the number of underweight markets. We estimate that over the course of the simulated and real-time periods, markets have remained in the overweight category for an average of 26 weeks, while the average holding period for the underweights has been 17 weeks.

Chart 1 features the equal-weighted composites of the overweights and underweights, constructed with weekly rebalancing. The performance of the overweights and underweights can be compared to the performance of the equal weighted 42-market composite, which is represented by the dashed line in the chart. As indicated in the top row of Table 1, the overweights have gained 23% per annum over the entire time frame, outpacing the universe gain per annum of 4%. They have beaten the universe in 60% of the weeks, as indicated in the second row. Most significantly, the overweights not only outperformed during the in-sample and out-of-sample periods, but also in real-time, passing the test of its ability to identify markets with relatively strong momentum and thus persistent relative strength. Also throughout, they have performed well on a risk-adjusted basis, as indicated by an information ratio of more than 1.0. This ratio is calculated by dividing the annualized average of the one-week excess returns (the performance of the overweights less the performance of the universe) by the annualized standard deviations of those excess returns, thereby measuring the excess return per unit of risk.

When comparing the per annum performance of the underweights to the universe per annum gain, we would want to see persistent underperformance. But in fact, the underweights have performed about as well as the universe over the entire period and have actually outperformed in real time, with no greater tendency to outperform in a given week than to underperform. Clearly the system’s value is its ability to identify winning markets, not losing markets. Among the system’s most useful indications was its identification of strong relative momentum among emerging markets in late 2001 and 2002, supporting the case for emerging markets, in the aggregate, to outperform developed markets. And for individual markets, one of the ranking’s best real-time calls was its overweighting of Argentina in June 2002, a time of political instability and economic worry. The market, however, sensed that the fears were overdone and began to rally. The market has remained overweight ever since, and at this writing had gained 109% over the period, as indicated in Table 2 featuring the Global Market Ranking. Over the same period, the 42-market universe gained 3%.

Chart 2

Of course, not all of the overweights continue to perform well. Some exit the list with a loss. But a reassuring quality of the system is that, by staying with strong markets for big gains and dumping decelerating markets with nothing worse than a small loss, it holds to the adage, “let profits run and cut losses short.” And as exemplified by the Argentina example, it cares not about why a market is doing what it’s doing, or whether it’s doing what it should be doing based on “the fundamentals.” Rather, its only objective is to identify relatively strong momentum and get on board, without the fundamental distractions. The Global Sector Ranking

The next challenge was applying the concept and approaches to global sectors, an undertaking that started with the development of a sector line-up. For this we used the capitalization-weighted, U.S. dollar-priced Dow Jones sector data, replicating the global sectors on an equal-weighted basis. By giving all of a sector’s stocks the same weight, a single stock with an exceptionally high market-cap will not dictate a sector’s relative strength changes. We required that a sector include at least four stocks large enough to surpass our market-cap cut-off, resulting in 130 sectors. After a 2002 rebalancing, the line-up included 131 sectors, with 18 industry classifications and 16 countries represented. The breakdown is shown in Table 3.

With the sectors established, we set out to develop a system using the same methodologies used to create the market ranking, formulating a momentum composite assigned to each sector, ranking the sectors in descending order from 100 to 02 , testing in-sample to determine the optimal strategy for overweighting and underweighting, and then confirming the reliability of the system in an out-of-sample period. Like the market ranking strategy, the resulting sector ranking strategy overweights a sector when its relative momentum is one of the strongest, then holds the sector even as the relative momentum recedes. It does the opposite for underweighting. This system provides another demonstration of how momentum leads price, illustrating that after a sector experiences unduly strong momentum, its relative strength is likely to persist even after the relative momentum has started to worsen. Specifically, a sector earns overweight status when it reaches a ranking of 95 and remains overweight until it drops to 60 or lower. A sector becomes an underweight when it drops below 5 and remains underweight until it rises to 40 or higher.

Thus like the market ranking, the sector ranking system is symmetrical. But the sector system is faster than the market system, as indicated by estimated average holding periods of 13 weeks for the overweights and 12 weeks for the underweights. As indicated in Chart 2 and Table 4, other differences are the longer in-sample period (January 1985 to January 1993), longer out-of sample period (January 1993 to May 2001), and shorter real-time period (May 2001 through June 2003).

In the case of this ranking system, it’s the underweights that have shined in real-time. While the composite of the overweights has dropped -10% per annum in the real-time period, the universe has lost -13% per annum, a modest excess return. And the overweights have outperformed in 55% of the real-time weeks. Meanwhile, the underweights have substantially underperformed, losing -26% per annum in real time. And they have outperformed in only 41% of the real-time weeks, with a decisively negative information ratio.

It should be recognized that for most of the real-time period, the global market trend has been pointed lower, so we have yet to see what the results will look like in a sustained global uptrend. But the out-of-sample results could be an indication, as the market trended higher during most of that period. The overweights gained 29% per annum, with a high information ratio and outperformance in 61% of the weeks. The underweights underperformed but managed to gain 5% per annum. Over the entire time frame, the overweights have gained 23% per annum and outperformed the universe in 59% of the weeks the universe has gained 11% per annum, and the underweights have gained 2% per annum, outperforming the universe in 45% of the weeks.

Even considering the modest real-time outperformance by the overweights, we can be confident that over the long-term, the Global Sector Ranking will prove itself effective in identifying overweights as well as underweights. Such a statement cannot be said about the Global Market Ranking’s ability to identify underweights. Why is the sector ranking more effective in identifying underweights? One possible explanation is the sector ranking’s shorter holding period. Since the sector ranking uses a momentum formula with rates of change that are shorter than those used in the market ranking, it keeps the sector ranking more sensitive to market declines, which typically are shorter than market advances with endings that are typically more climactic, decisive, and faster to complete3

. The slower market ranking has tended to be late in identifying the underweights and late to remove them from underweight status. The market ranking has had far better luck with the overweights, especially during sustained market advances.

Another attribute of the Global Sector Ranking is that it helps answer the question of what’s currently more important, country or sector? At times, the overweights will be dominated by different sectors from a specific country, indicating that allocation to that country is the key to outperformance. At other times, the overweights will be dominated by a sector from different countries, indicating that allocation to that sector is key. But over the long-term, both are important. Peter Hopkins and C. Hayes Miller explained in a 2001 paper that “it is not yet totally clear that geographical importance has permanently diminished on a global basis or that sectoral importance has increased globally from 1992 through 1995, country effects were 1.5 to two times as important as industry effects, but now the importance of country and the importance of sector or industry group are about equal.”4

By identifying opportunities driven by the sector and/or the market, and likewise identifying the risks, the Global Sector Ranking allows for a more comprehensive assessment than could be gained by ranking the markets or sectors separately.

Although the sector ranking’s real-time history is shorter than that of the market ranking, it has several successes to its credit. During the bottoming process of JulyOctober 2002, the ranking detected relative strength among Technology sectors and Canadian sectors. Accordingly, as indicated in Table 5 showing the Global Sector Ranking’s overweights and underweights, the Canadian Technology sector attained overweight status in early November 2002 and had gained 48% as of the end of June 2003. The Canadian Telecom sector has been letting its profits run for an even longer period, gaining 67% from the end of September through June.

In revealing momentum deterioration and assigning underweight status, the ranking proved prescient with its warnings of worse things ahead for European insurance sectors. Between early November 2002 and early February 2003, the insurance sectors for France, Germany, the U.K. and Switzerland all dropped to underweight status, and by mid-March all four were down by more than 20%. By mid-May they had all exited the underweight category, confirming that the risks had diminished for those sectors. In fact, it was European sectors in general that dominated the underweight category in mid-March, then dominating the overweights by mid-May. Sectors from the Pacific region were then pervasive among the underweights.

Uses and Complements

The sector ranking’s information can thus be applied in numerous ways. It can be used to identify a strengthening sector of stocks in a specific country, serving as a screen for investors looking for the strongest stocks in the strongest countries. Conversely, it can be used to identify potential underperformers among a list of buy candidates, warning against buying a stock in a relatively weak sector within an underperforming country. It can also be used strictly along sector lines or country lines.

More broadly, it can be used to gauge the extent to which a sector theme is global in scope, or a country theme regional in scope. For instance, overweight status in the vast majority of Technology sectors would confirm the global scope of that sector’s relative strength, while overweight status in the vast majority of European sectors would carry a strong message of regional relative strength.

The system can also be used for its indications about the global market trend. With a preponderance of defensive sectors, such as Food & Beverage, among the overweights, and higher beta sectors, such as Technology, among the underweights, the system would describe the market as risk averse. With Health Care and Technology sectors dominating the overweights, and Banks and Utilities dominating the underweights, the system would reflect a market preference for growth over value. It could then be determined if those indications were consistent or inconsistent with those provided by a domestic ranking system. The greater the confirmation, the stronger the message of global participation and sustainability.

(Click Table to Enlarge)

Other good complements to the ranking’s indications are breadth measures that can be created using the global sector database, indicators on the breadth of everything from the global universe itself to regions, countries, global sectors, and sectors within individual countries. For example, with the majority of Technology sectors overweight in the sector ranking, a sign that the sector’s relative strength is global in magnitude, we could turn to Charts 3 and 4 (on next page) for confirmation of strength among the stocks within the sector. The top two clips of Chart 3 plot the daily price and relative strength lines for the global Technology sector, the aggregate of all the stocks in the ranking’s eight Technology sectors. And the bottom two clips plot the sector’s advance/ decline line and percentage of issues at 30-day new highs. Chart 4 uses weekly data to feature the global Technology sector along with the percentages of the sector’s stocks that are above their 10-week and 40-week moving averages. The sector’s relative strength would have short-term breadth confirmation with its advance/decline line above it 50-day moving average and moving higher with a rising percentage of 30-day new highs. And the sector would have longer term breadth confirmation with more than 70% of the sector’s stocks above their 10-week moving averages and more than 55% above their 40-week moving averages.

Chart 3

Chart 4

For regions and countries, the Global Market Ranking could serve as an additional overlay, useful for assurance that the sector ranking’s indication of market relative strength has longer-term confirmation. If, for instance, the sector ranking would indicate relative strength among Japanese sectors, we would gain more confidence in the staying power of that relative strength if Japan’s Topix Index would rise to overweight status in the market ranking. More broadly, with sectors from Asia Pacific countries dominating the overweights of the sector ranking, our longer-term confidence would increase if the region’s market indices would dominate the market ranking’s overweights.

Separately or in conjunction with one another, the sector and market rankings are thus useful for identifying changing themes, information that’s essential when developing global market strategy and making allocation recommendations. The rankings can also be used as screens and for input to the decision making process – i.e., whether to buy a certain ETF, a country fund, or a stock in a specific sector. The next generation of these systems would combine a global sector ranking with a ranking of individual stocks that can be bought upon rising to overweight, sold when dropping from overweight, shorted upon falling to underweight, and covered when rising from underweight. Such an executable, long-short system would be especially appealing to hedge funds, demonstrating that in leading price, momentum leads to profits as well.

Conclusion

This paper has demonstrated how relative momentum can be used to develop a ranking of 42 markets around the world, a ranking that has proven effective in identifying winning markets. The paper has also shown how relative momentum can be used to develop a ranking of global sectors, a ranking that has held its own in identifying relative strength in a down market while proving adept at identifying underperformers. Moreover, based on our real time experience using these systems, we can say that they certainly deserve to be the starting point for identifying new themes and emerging leadership. Knowing that a market or sector is emerging as a momentum leader is far more important than knowing why it is doing so, a determination that often can only be made in hindsight. Rather, understanding the market’s current dynamics, that the current differences in relative momentum will lead to differences in performance – that momentum leads price – is the most valuable and useful knowledge that one can have.

Endnotes

1. The Journal of Finance published “Do Industries Explain Momentum?” by Tobias J. Moskowitz and Mark Grinblatt (1999) and “The Profitability of Momentum Strategies by Louis K.C. Chan, Narasimhan Jegadeesh, and Josef Lakonishok (1999). The Rydex Sector Rotation Fund ranks industries based on several measures of price momentum. The fund buys baskets of stocks to replicate the performance of the top-ranked industries.

2. Since 131 sectors are ranked from 100 to 0, more than one sector can have the same rank.

3. According to Ned Davis Research, there have been 33 bull markets since 1900, the median bull lasting 573 days. There have been 32 bear markets, the median lasting 375 days.

4. “Country, Sector, and Company Factors in Global Equity Portfolios” published by The Research Foundation of the Association for Investment Management and Research.

Biography

Timothy Hayes, CMT, is the Global Equity Strategist for Ned Davis Research. Tim oversees the firm’s global and U.S. equity allocation services, authoring the firm’s weekly Stock Market Focus and International Focus publications. He also is editor of the firm’s bi-monthly Investment Strategy. Tim holds the Chartered Market Technician designation and is an MTA member. He has written a book, The Research-Driven Investor, published in November 2000. He is a regular guest on CNBC television and is often cited in The Wall Street Journal and other publications. In 1996, Tim won the Charles H. Dow Award for groundbreaking research in technical analysis.