Journal of Technical Analysis (JOTA). Issue 57 (2002, Winter)

JOURNAL

Analysis

ofTechnical

Market Technicians Association, Inc. A Not-For-Profit Professional Organization ■ Incorporated 1973

SM

Winter-Spring 2002

Issue 57

2JOURNAL of Technical Analysis • Winter-Spring 2002

THE JOURNAL OF TECHNICAL ANALYSIS EDITOR & REVIEWERS 4

THE ORGANIZATION OF THE MARKET TECHNICIANS ASSOCIATION, INC. 5

CHARLES H. DOW AWARD WINNER • MAY 1993

CHARLES DOW LOOKS AT THE LONG WAVE 6Charles D. Kirkpatrick II, CMT


INFORMATION, TIME AND RISK 9William X. Scheinman


THE QUANTIFICATION PREDICAMENT 17Timothy W. Hayes, CMT


AUTUMN PANICS: A CALENDAR PHENOMENON 22Christopher Carolan

CHARLES H. DOW AWARD WINNERS • MAY 1999

CORPORATE INSIDERS’ BIG BLOCK TRANSACTIONS 26Eric Bjorgen and Steve Leuthold

CHARLES H. DOW AWARD CO-WINNER • MAY 2001

STOCK SELECTION: A TEST OF RELATIVE STOCK VALUES REPORTED OVER 17-1/2 YEARS 30Charles D. Kirkpatrick II, CMT


SIGN OF THE BEAR 35Peter G. Eliades


IDENTIFYING BEAR MARKET BOTTOMS AND NEW BULL MARKETS

Paul F. Desmond 38

JOURNAL of Technical Analysis • Winter-Spring 2002 • Issue 57

Table of Contents

You may have noticed the new cover and the new name “Journal of Technical Analysis” to our revered MTA publication. There have been other changes as well.

First, you have a new editor. Hank Pruden, your previous editor, first managed the MTA Journal in 1993, almost nine years ago. What a wonderful nine years forthe Journal. He and Dave Upshaw and many others as reviewers produced over those years a professional and useful Journal for MTA research that has flourishedamongst a group of practitioners not normally associated with research itself. This was an admirable, arduous, and unrewarded feat, one dedicated to the MTAand its professional image, and one that humbles me. Thank you Hank and Dave and all you others for your hard work over those past nine years.

Second, you may have noticed that we now have two finance professors as manuscript reviewers. We hope to entice even more. As we upgrade the articles in theJournal to satisfy more stringent criteria for content and analysis, it is imperative that we include those from the academic world to help us. We have begun to doso and welcome Professors Avner Wolf and Julie Dahlquist to our circle.

Third, we begin this editorial reign with a collection of all the Charles H. Dow award papers including this year’s. This may seem presumptuous at first, becauseI have twice won the award myself, but in truth, the award has never been given much notice within the MTA or elsewhere. It is an award that recognizes goodwriting and research, and as such, should be a cornerstone for this publication that attempts to do likewise. Thus, we have reproduced each winning article in itsoriginal form with updates when necessary. Only one of the winners, Bill Scheinman, is no longer with us; the remaining winners are still very much in the businessof technical analysis. We hope these papers will provide inspiration for you not only to compete for the Dow Award but also to provide us with additional researchstudies that we can share with our members and the rest of the investment world.

Charles D. Kirkpatrick II, CMT, Editor


1. Standards of Judgment

A submitted or nominated work will be judged according to the following:

a. The work is based upon the concepts of technical analysis.

b. The work is either original or is a significant extension of an estab-lished work of technical analysis.

c. The subject matter is substantive. Solid research and analysis are im-perative.

d. The work is practical and enhances the understanding of market action.A market forecast will not, by itself, be considered for the Award. Thepresentation of an analytical method or trading system is expected toinclude the results of applying the technique to specific past data ac-cording to generally accepted standards of testing.

e. The strength and clarity of writing are superior.

2. Nominations or Submissions of Published Works

Papers written especially for the Award or works published betweenJanuary 1 and December 31 of the prior year may be submitted or nomi-nated.

Nominations must be made in writing and a copy of a nominated paperor book must accompany them. There is no fee for nominations or sub-missions. Nominations and submissions are to be sent to Charles H.Dow Award, Journal of Technical Analysis, 74 Main Street, 3rd Floor,woodbridge, New Jersey 07095.

3. Style

The text must be a succinct and conclusive presentation of the subject.The charts, tables, and figures should be used to exemplify or to supple-ment the text and should not be the primary means of conveying thewriters’ points.

4. Papers

A submitted paper must not contain less than 1,500 or more than 4,000words. A paper shall not contain more than 10 charts, tables, or figures

CHARLES H. DOW AWARD GUIDELINES

Creative market technicians are invited to submit their best papers for the annual Charles H. Dow Award for excellence in technical analysis. Editorsand publishers of technical analysis are invited to nominate for the Award an outstanding work of technical analysis published in 2000. The Charles H.Dow Award, sponsored by Market Technicians Association, Inc. (MTA), Barron’s, and Dow Jones Newswires will be given to the work that breaks newground or makes innovative use of established techniques in the spirit of pioneering market technician, Charles H. Dow.

The Charles H. Dow Award is presented annually at the MTA’s Annual Seminar. The winning author will receive a personal Award, will be recognizedin Barron’s, and will be invited to discuss the paper at the Annual Seminar or at a monthly meeting of the MTA. The publication or a summary may bepublished in the MTA Journal, the MTA newsletter and/or the MTA website. Dow Jones Newswires will make copies of the paper available for distribu-tion to the public through various media. A perpetual plaque including the author’s name with those of previous recipients of the Charles H. Dow Awardwill reside at the MTA office in New Jersey. At the discretion of the judges, the authors of runner-up papers will receive personal awards. No cash awardwill be given to any award winner or runner up.

GUIDELINES

total. Submissions must be typed on white bond paper, 8.5" by 11" size,double-spaced, in black ink.

Charts, tables, and figures should be placed in appropriate sections ofthe text. When it is not possible to do so, they must be presented onwhite paper, 8.5" by 11" size. Charts, tables, and figures must be indi-vidually labeled in numerical sequence. They shall be submitted in cam-era-ready format and may be presented in color. Seven complete copiesof the paper must accompany a nomination or submission. Each hard-copy submission or nomination must be accompanied by a least onedisc copy, preferably in Word-Excel format, that includes all graphics.Nominations or submissions of books or lengthy articles must includeat least seven copies, non-returnable. Each must include at least sevensynopses of the work, no more than three pages in length, that summa-rize intent, methodology, and conclusions.

Hardcopy submissions or nominations of lengthy articles must includeat least one disc copy as described above for papers.

5. Deadline

The last day for nominating or submitting publications is February 28.Entries received after that date may be accepted at the discretion of thejudging panel.

6. Judging Panel

The judging panel will include at least three past winners of the CharlesH. Dow Award, selection preference given to the three most recent win-ners. The past winner of longest standing will rotate out of the judgingpanel each year to be replaced by the latest Award winner.

In addition, the judging panel will include no more than one votingrepresentative from each of Barron’s, Dow Jones Newswires, and theMTA. Members of the Board of Directors of the MTA, excepting theeditorial board of the Journal of Technical Analysis, shall not be eli-gible for the judging panel.


EDITOR

Charles D. Kirkpatrick II, CMTKirkpatrick & Company, Inc.

Bayfield, Colorado

ASSOCIATE EDITOR

Michael CarrCheyenne, Wyoming

Connie Brown, CMTAerodynamic Investments Inc.

Pawley's Island, South Carolina

Matthew ClaassenPrudential Financial

Vienna, Virginia

Julie Dahlquist, Ph.D.St. Mary's UniversitySan Antonio, Texas

J. Ronald Davis, CMTGolum Investors, Inc.

Portland, Oregon

Cynthia KaseKase and Company

Albuquerque, New Mexico

Cornelius LucaBridge Information Systems

New York, New York

John McGinley, CMTTechnical Trends

Wilton, Connecticut

Michael J. Moody, CMTDorsey, Wright & Associates

Pasadena, California

Jeffrey Morton, MD, CMTPRISM Trading Advisors

Missouri City, Texas

Kenneth G. Tower, CMTUST Securities

Princeton, New Jersey

Avner Wolf, Ph.D.Bernard M. Baruch College of the

City University of New YorkNew York, New York

PRODUCTION COORDINATOR

Barbara I. GompertsFinancial & Investment Graphic Design

Marblehead, Massachusetts

MANUSCRIPT REVIEWERS

JOURNAL of Technical Analysis • Winter-Spring 2002 • Issue 57

Journal Editor & Reviewers

The JOURNAL of Technical Analysis is published by the Market Technicians Association, Inc., (MTA) 74 Main Street, 3rd Floor, Woodbridge, NJ 07095. Itspurpose is to promote the investigation and analysis of the price and volume activities of the world's financial markets. The JOURNAL of Technical Analysis

is distributed to individuals (both academic and practitioner) and libraries in the United States, Canada, Europe and several other countries. The JOURNAL

of Technical Analysis is copyrighted by the Market Technicians Association and registered with the Library of Congress. All rights are reserved.

PUBLISHER

Market Technicians Association, Inc.74 Main Street, 3rd Floor

Woodbridge, New Jersey 07095


THE ORGANIZATION OF THE MARKET TECHNICIANS ASSOCIATION, INC.MEMBER AND AFFILIATE INFORMATION

MTA Member

Member category is available to those “whose professional efforts arespent practicing financial technical analysis that is either made available tothe investing public or becomes a primary input into an active portfoliomanagement process or for whom technical analysis is a primary basis oftheir investment decision-making process.” Applicants for Member mustbe engaged in the above capacity for five years and must be sponsored bythree MTA Members familiar with the applicant's work.

MTA Affiliate

MTA Affiliate status is available to individuals who are interested intechnical analysis and the benefits of the MTA listed below. Most impor-tantly, Affiliates are included in the vast network of MTA Members andAffiliates across the nation and the world providing you with commonground among fellow technicians.

Dues

Dues for Members and Affiliates are $200 per year and are payablewhen joining the MTA and annually on July 1st. College students mayjoin at a reduced rate of $50 with the endorsement of a professor.

Applicants for Member status will be charged a one-time applicationfee of $25.

MEMBERS AND AFFILIATES

■ have access to the Placement Committee (career placement)

■ can register for the CMT Program

■ may attend regional and national meetings with featured speakers

■ receive a reduced rate for the annual seminar

■ receive the monthly newsletter, Technically Speaking

■ receive the Journal of Technical Analysis, bi-annually

■ have access to the MTA website and e-mail network

■ have access to the MTA lending library

■ become a Colleague of the International Federation of TechnicalAnalysts (IFTA)

JOURNAL SUBMISSION GUIDELINES

We want your article to be published and to be read. In the latter re-gard, we ask for active simple rather than passive sentences, minimal syl-lables per word, and brevity. Any paper longer than 20 pages, double-spaced, will be returned. Charts and graphs must be cited in the text, clearlymarked, and limited in number. All equations should be explained in simpleEnglish, and introductions and summaries should be concise and informa-tive.

1. Authors should submit, with a cover letter, their manuscript and sup-porting material on a 1.44mb diskette or through email. The coverletter should include the authors’ names, addresses, telephone numbers,email addresses, the article title, format of the manuscript and charts,and a brief description of the files submitted. We prefer Word for docu-ments and *.jpg for charts, graphs or illustrations.

2. As well as the manuscript, references, endnotes, tables, charts, figures,or illustrations, each in separate files on the diskette, we request thatthe authors’ submit a non-technical abstract of the paper as well as ashort biography of each author, including educational background andspecial designations such as Ph.D., CFA or CMT.

3. References should be limited to works cited in the text and should fol-low the format standard to the Journal of Finance.

4. Upon acceptance of the article, to conform to the above style conven-tions, we maintain the right to make revisions or to return the manu-script to the author for revisions.

Please submit your non-CMT paper to:Charles D. Kirkpatrick II, CMT7669 CR 502Bayfield, CO [email protected]

JOURNAL OF TECHNICAL ANALYSIS

■ Annual subscription to the JOURNAL of Technical Analysis for nonmem-

bers: $50 (minimum two issues).

■ Single issue of the JOURNAL of Technical Analysis (including back is-

sues): $20 each for members and affiliates and $30 for nonmembers.

■ 56 issues of all articles from to January 1978 to Fall 2001 of the Jour-

nal are available on a single CD. The cost for MTA Members/Affili-

ates, students and academics is $95 and $295 for non-Members/Affili-

ates. To order, log on to www.mta.org and download the Journal CD

order form.


Using the stock market principles outlined by Charles H. Dow, howcould we look at the long wave in stock prices?

Dow published The Wall Street Journal beginning in 1889 and, unfor-tunately, died in 1902. He wrote during a period of generally rising stockprices from the depression lows in the 1870s to the then all time high in1901. During that period Dow formulated his theory of the stock market.It consisted of two important components: the cyclical nature of the mar-kets and in the longer cycle, the “third wave,” the need for confirmationbetween economically different sectors, specifically the industrials and therailroads.

Following an earlier analogy between the stock market and ocean wavesduring the tidal cycle, Dow hypothesized in his famous Wall Street Journal

editorial of January 4, 1902:

“Nothing is more certain than that the market has three well-definedmovements which fit into each other. The first is the variation due to localcauses and the balance of buying and selling at that particular time. Thesecondary movement covers a period ranging from 10 days to 60 days,averaging probably between 30 and 40 days. The third movement is thegreat swing covering from four to six years.”

Some technicians, especially cycle analysts, would quibble with thesimplicity of Dow’s breakdown since there is evidence of other waves withperiodicity between 40 days and four years. However, cycle analysts wouldalso have to acknowledge that Dow’s breakdown is certainly accurate,though perhaps not inclusive, and that the periods he mentions are, re-markably, still the dominant cyclical movements today.

But Dow stopped short at the four- to six-year cycle, essentially thebusiness cycle. He assumed that stock prices had an underlying uptrendabout which these cycles oscillated. This was consistent with his experi-ence at the time. Stock prices (see chart A, Dow Jones Industrial, 1885-1902) had wild gyrations during the late 19th Century, but the underlyingtrend was generally upward. He undoubtedly would have added a fourthwave, or “long wave,” had he lived to see the 1929-32 crash.

Aside from recognizing that the stock market had a pattern, which isthe basis for technical analysis, Dow also recognized, in his theory of con-firmation between the Industrial Average and the Railroad Average, thatthere must be an economic rationale for any signals given by the stock

market price action. Most pure technicians conveniently overlook thisbecause it diverges from a strict price analysis. Unfortunately, investmentanalysts have evolved into three camps since Dow – technicians, funda-mentalists and academics – and as seems to be the way of human nature,they generally disregard the other’s work to reinforce their own identity.However, Dow was above all that, (or at least before it), and consideredthe economic rationale for a cyclical turn in the stock market just as impor-tant as the technical.

In the post-1929 era, we now know that the underlying long-term up-trend in stock prices can be severely interrupted. From looking at stockprices going back several hundred years we also note that the 1929-1932decline was not an anomaly. It occurs with frightening regularity, roughlyevery 40 to 60 years (see Chart B, Dow Jones Industrial, Reconstructed,1700-1940). We call this cycle the “long wave” and ponder on how Dowwould have analyzed it.

As an aside, there are still many analysts, especially academics, whobelieve that the long wave is imaginary. Their thesis is based on the as-sumption that markets don’t have a “memory.” They argue that today’sprices are totally independent of yesterday’s, of last week’s, of last year’sand certainly of 50 years ago prices. Furthermore, since Fourier trans-forms and other sophisticated mathematical techniques have been unableto identify with certainty such cyclicality, it probably doesn’t exist. On theother hand, new experiments, especially those with non-linear mathemat-ics, are beginning to knock down the “no memory” thesis. Edgar Peters, inhis book Chaos and Order in the Capital Markets, suggests that the stockmarket has at least a four year memory. Professors McKinley and Lo fromWharton and MIT have demonstrated that stock price action is inconsis-tent with a “no memory” thesis and are now using non-linear mathematicsto study prices. Professor Zhuanxin Ding from the University of Califor-nia has shown that stock prices act as if they had long memories. Evensimple moving averages, as studied by two professors at the University ofWisconsin, Dr. William Brock and Blake LeBaron, can generate profitabletrading signals from prices alone, an inconsistency with the “no memory”thesis. The Economist wrote in a special section on the Frontiers of Fi-

nance on October 9, 1993:

CHARLES DOW LOOKS AT THE LONG WAVE

Charles D. Kirkpatrick II, CMT


Chart ADow Jones Industrial Average

January 1885 - December 1902 (Linear Regression Trend)

Chart BDow Jones Industrial Average

1700-1940 (Reconstucted)


“This was a shock for economists. Might chartists, that disreputable bandof mystics, hoodwinking innocent fund managers with their entrail-gazingtechniques and their obfuscatory waffle about double-tops and channelbreak-outs, be right more often than by chance? How could it be?”

Whether we believe in price memory or not, charts of stock prices sincethe South Sea Bubble in 1720 show that there are obviously times whenthe stock market experiences enormous, speculative rises and subsequent,disastrous declines. These major events occur at periods considerably longerthan Dow’s four- to six-year movements. Furthermore, when we look atother economic data, such as commodity prices, GNP (even U. S. PostOffice revenue), etc., we see the same long-term periodicity.

How would Charles Dow have looked at this long wave price action forsignals? Probably he would have begun by looking only at the highs andlows of each four- to six-year cycle. Intermediate-term motion would belargely irrelevant to the long wave. Simplistically, he would likely havestated that the long wave was up when the tops and bottoms of the four-year cycles were making new highs, and conversely, when the tops andbottoms were making new lows, the long wave was declining.

In the last 60 years, this approach would have missed the 1929 crash,but the ultra long-term investor would have sold his stocks in 1930 whenthe 1929 low, a four-year cycle low, was broken. It would also have toldthe investor in 1950 that the long wave was turning upward, that it wastime to invest in the stock market. Unfortunately, there would have beenseveral false signals. For example, in the 1970s, two four-year cycle lowsbroke below previous four-year lows, wrongly suggesting that the longwave was headed down again. Also, in the 1930s, after the initial bottomin 1932, several four-year cycle lows were broken between 1937 and 1949,suggesting that the long-term cycle was turning down again after havingbottomed in 1932.

False signals also occurred in Dow’s original work on the four-yearcycles and are the reason for his turn to confirmation between the Indus-trial and the Railroad averages. He based his confirming signals on theeconomic assumption that expansion in industrial profits could be a tem-porary anomaly but not if the produced goods were being shipped, by rail-roads, to customers. A confirmation between the two averages in eitherdirection suggested that the new trend was real.

Unfortunately, over the long wave, the theory of industrials versus rail-roads breaks down. First, over time, railroads are not always the principalform of transportation for goods (How do you ship the service industry?and how about canals in the 1830s?), and second, the apparent cause forthe long-wave has more to due with capital formation, debt and moneythan with industrial production.

Money has a price too – the interest rate. Interestingly, interest rates

over the past several hundred years have also had a long wave that hascorresponded in period, if not in turning points, with the stock market (seeChart C, U.S. Long-Term Interest Rates, Reconstructed, 1700-1940). Forthis reason, we assume Dow would have looked to the interest rate marketfor confirmation of a trend change in the long-term stock market.

Looking at interest rate trends, however, is not as simple as looking fora confirmation in trend between industrials and rails. Long wave interestrate cycles do not overlap precisely with long wave stock price cycles (seeChart D, U.S. Long-Term Interest Rates & Dow Jones Industrial Average,Reconstructed, 1700-1940). They will not “confirm” a move to new highsor lows as the rails will the industrials. It is important that one understandmore about the history of the long wave direction in interest rates before asignal can be confirmed for the stock market.

The confusing aspect between long wave interest rates and the stockmarket is that sometimes both can be moving in the same direction andsometimes each can be moving in opposite directions. This is becausestock prices have a corporate profit or growth component, as well as aninterest rate or alternative investment component. In the former, stockprices rise as a result of economic growth, industrial expansion and profit-ability along with interest rates; in the latter, stock prices rise as an alterna-tive investment to falling yields on fixed income securities. The latter, aswe shall see, is more dangerous.

When we look at the evidence over the past several hundred years wesee alternating periods of rising and falling interest rates. These are called“secular” moves and have to do with the expansion and contraction of capitaland debt.

Notice in Chart D that the peak in interest rates always precedes thelong wave peak in stock prices by many years. When interest rates and thestock market are both rising together, the industrial growth component isdominant. The period after interest rates peak is when stock prices rise asan alternative investment. During that period declining interest rates forceyield-conscious investors into alternative investments of lesser quality inorder to maintain yield. Since stocks are the most risky and least qualityinvestments, they become the final alternative, especially when their pricecontinues to appreciate as a result of increasing cash flow into the stockmarket. The recent conversion of government-guaranteed CD deposits intostock mutual funds is typical during this period. Unfortunately, it eventu-ally leads to the declining long wave in stock prices.

Each declining stock market wave has occurred only during a seculardecline in interest rates. Over the past several hundred years, you won’tsee a long wave decline in stock prices while interest rates are rising. De-clining interest rates at first can cause a financial speculation and an enor-mous rise as yield is chased through lesser quality, but eventually declin-

Chart CU.S. Long-Term Interest Rates


Chart DU.S. Long-Term Interest Rates & Dow Jones Industrial Average



ing interest rates are unhealthy for the long wave in stock prices. With thisin mind, Dow would likely have developed the following confirmationrules for the long wave in stock prices:

1. When four-year stock price cycles reach new highs and business-cycleinterest rates are rising, the long wave is rising.

2. When four-year stock price cycles break below previous lows and busi-ness-cycle interest rates are rising, the long wave is rising.

3. When four-year stock price cycles break above previous highs and busi-ness-cycle interest rates are declining, the long wave has been given awarning but is still rising.

4. When four-year stock price cycles break down below previous lowsand business-cycle interest rates are declining, the long wave is declin-ing.

5. After a decline, the long wave will not turn up until business cycleinterest rates also turn up.

Using this set of rules, let’s walk through the past 75 years using theaccompanying Chart E of long-term U.S. interest rates and the Dow JonesIndustrial Average since 1900.

From Dow’s death in 1902 both interest rates and the stock market rose.

Chart EU.S. Long-Term Interest Rates & Dow Jones Industrial Average

1900-Present

According to rule #1, the long wave was rising. Interest rates peaked in1920 and declined through 1946. Declining interest rates are a warning tobe confirmed later by a breakdown in the stock market. Thus, under rule#4, when the stock market broke to new lows in August 1930 (DJIA monthlymean = 231), it confirmed the long wave downturn.

During the 1930s and 1940s, while the initial bottom in 1932 turnedout to be the actual bottom, the gyrations were large and the stock markettrend generally flat. Interest rates declined until the end of World War II.Any upward breakout had to be taken skeptically (rule #5).

Finally, in March 1950, interest rates broke above their earlier busi-ness-cycle high (rule #5 and #1). Since rising interest rates are alwaysaccompanied by a rising stock market long wave, this was the buy signal.The DJIA was 249 at the time.

In the 1970s, the stock market broke below its prior four-year cyclelows in 1970 and in 1974. However, interest rates were still rising and thusthe long wave was still rising (rule #2).

Interest rates finally peaked in September 1981. This was a warning(rule #3), similar to the interest rate peak in 1920, that the long wave wasending. Currently, the stock market has yet to break below a previousfour-year cycle low and thereby confirm a new decline in the long wave.The last four-year low was 2340 in the DJIA in 1990*. Should it be brokenbefore a higher low is established, we will have confirmation of the down-turn in the long wave.

Would Charles Dow have looked at the long wave in this manner? Wedon’t know. But his principle of first observing price action simplisticallyand then confirming it with other markets, using some economic justifica-tion, gives us an excellent background for analysis of the long wave andteaches us to remain broad-minded and rational. His legacy is more thanjust a stock market theory. It is a way of thinking that transcends the nar-row confines and pettiness of much investment analysis.

March 30, 1994

*Note: 8064 in the DJIA in 2001, the NASDAQ has already begun itslong wave decline.

CDK, 2002


“The nature of risk is highly sensitive to whether we act before or afterwe have all the information in hand. This is just another way of sayingthat risk and time are only opposite sides of the same coin, because theavailability of information increases with the passage of time. Thus, risk,time and information interact upon one another in complex and subtleways.”

From keynote address by Peter L. Bernstein upon receiving the InauguralDistinguished Scholar Award from the Southwestern Economic Association, Dallas,

March 4, 1994.

The reader should keep in mind that any discussion of the financialmarkets is of necessity a discussion of constantly changing statistics andother data. This article was originally written in May 1994 and was sub-mitted to the MTA Journal at that time. Therefore, while the data usedherein were current as of May 20, 1994, such data applied to any specificsituation described may no longer be applicable. The same caveat appliesto the Sequel, which was written and submitted on November 11, 1994market close, and briefly discusses how each of the theories or methodsdescribed herein worked or failed to work during the period subsequent toMay 20,1994.

SYNOPSIS

This article outlines the core theories of Charles H. Dow and EdsonGould. Three of Gould’s methods used to forecast stock prices, which arebased on quantifying investor psychology, are described and then illus-trated using current data. Several forecasts are then made based on howGould’s three methods and those of the author combine, in the author’sopinion, to operate in current financial markets. Future levels of interestrates, stock prices, an industry group, the technology sector, as well as twoindividual stocks, are estimated. A sequel, written six months after the origi-nal article was submitted, discusses how the forecasts turned out.

DOW’S THEORIES

The granddaddy of all stock market technical studies is the Dow Theory,which was originated by Charles H. Dow around the turn of the century.According to Dow, major bull or bear trends are indicated when the Dow

Jones Industrial and Transportation averages, one after the other, set newhighs or lows. A divergence between the indices often indicates a potentialturning point in the underlying trend of the stock market. Dow set the stagefor the later theories, still used and elaborated on by market analysts today,of what may broadly be defined as divergence analysis. That is the study ofdivergences among and between a broader universe of indices and indica-tors than were available to Dow. Dow’s theory was used in the context ofhis basic commandment: “To know values is to know the meaning of themarket.”l But Dow also said that wise investors, knowing values above allelse, buy them when there is no competition from the crowd. Indeed, theybuy them from the crowd during periods of mass pessimism, and sell themto the crowd in return for cash during late stages of advancing markets.The stock market as a whole, said Dow, “represents a serious, well-consid-ered effort on the part of farsighted and well-informed men to adjust pricesto such values as exist or which are expected to exist in the not too remotefuture.”2

GOULD’S THEORIES

Edson Gould, who first studied the Dow Theory, was a practicing mar-ket analyst for more than fifty years between the early 1920s and late 1970s.His main focus was on forecasting the stock market. Though a student ofphysics and the harmonics of music, as well as business cycles and Greekcivilization, each of which he believed helped explain certain aspects ofhow the stock market behaved, he came to believe, after reading GustaveLeBon’s classic, The Crowd,3 that “the action of the stock market is noth-ing more nor less than a manifestation of mass crowd psychology in ac-tion.”4

The methods and techniques Gould utilized in his service, Findings &

Forecasts, attempted to “...integrate the many economic, monetary and psy-chological factors that set the level and cause the changes in stock prices.”5

He regarded the economic factors as important but typically late so far asthe stock market is concerned. He regarded the monetary factors as crucialfor the stock market and typically early. Whereas, he believed that, “Of allthree sets of factors, the psychological factors are by far the most impor-tant – in fact, the dominant factors affecting the cyclical swings of stockprices.”6

THESIS

It follows from the above that one of the most important aspects of allin successfully analyzing the stock market is measuring investor sentiment.The consensus view, the most difficult factor of all to gauge accurately,can be glimpsed at times – and only in part-through not only such transac-tion-based data as put/call ratios, premiums and open interests, but alsopoll-based data such as the weekly Investors Intelligence reports of whatpercentage of investment advisors are bullish or bearish. Whereas the au-thor regularly screens such data for extremes, the theories and methodswhich are derived from Gould and are discussed below are, in an of them-selves, measures of the behavior of the investment crowd and, in his opin-ion, more practically useful in making and implementing investment deci-sions. And inasmuch as they are also applied to the monetary factors, abond market opinion is derived therefrom, as well.

The index and stock charts used to illustrate this article are of:

1. Treasury Bonds Nearest Futures, Monthly

2. Treasury Bonds Nearest Futures, Weekly

3. New York Stock Exchange Financials Index, Weekly Standard & Poor's40 Utility Stock Composite, Weekly

4. Standard & Poors 400 Industrial Stocks Composite, Monthly

5. Drug Shares Index, Weekly Close (Sum of BMY, LLY, MKC, MRK,PFE, UPJ x 4.50541)

6. TXB-Hambrecht & Quist Technology Stock Index Less CBOE Bio-technology Stock Index, Weekly Average

7. Merck (MRK), Weekly

8. U.S. Robotics (USRX), Daily

GOULD’S METHODS AND TECHNIQUES

Edson Gould is, perhaps, best known for his monetary rule and valua-tion barometer: His Three-Step-and-Stumble Rule states that, “Wheneverany one of the three rates set by monetary authorities – the rediscount rate,

INFORMATION, TIME AND RISK

William X. Scheinman



the rate for bank reserve requirements, and margin requirements on stocks– increases three times in succession ... invariably ... the stock market hassubsequently ... suffered a sizable setback.”7 Whereas his Senti-Meter is,“the ratio of the Dow Jones Industrial Average to the average rate of an-nual cash dividends paid on that average. “8 When the Senti-Meter reads$30 per $1 of dividends or more it indicates a high and risky market. Areading of $15 or less indicates a relatively low and cheap market.

Lesser known and, perhaps, too arcane for many, the author has foundthat three of Gould’s methods and techniques are more practically usefulin helping decide when and at what levels a given stock or price index is“too” high, or “too” low and what constitutes a sentiment extreme. Withthis background in mind, let’s examine Gould’s theories and applicationsof Resistance Line Measurement, Unit Measurement and the Rule of Three,as well as the author’s theory of the Cut-in-Half principle and its oppo-sites.

RESISTANCE LINE MEASUREMENT

According to Gould, “ ... the market continually reveals a quantum ofmass psychology comprising time and price. It follows that a sharp declinein a short period of time generates as much bearishness as a slow and mi-nor decline over a long period of time.”9 This theory, then, is based onthree principal determinants of crowd psychology in the market place: pricechange itself, elapsed time to achieve it and the perceived amount of risk.

The resistance line theory attempts to measure these three elements of masspsychology mathematically, weighing both the vertical price change and thehorizontal elapse of time. This measure of potential risk or reward must be keyedoff whatever the investor regards as any pair of prices which consist of an im-portant high and low of the particular security’s price history. Four of the chartswhich are discussed below illustrate how the resistance lines are applied.

The theory is that a trendline rising at one-third (or two-thirds) the rateof an advance movement is likely to produce resistance to subsequent de-cline, but, if violated, the decline will accelerate from the point of penetra-tion. Similarly, a trendline declining at one-third (or two-thirds) the rate ofa decline movement may provide resistance to a subsequent advance, but,if penetrated, the advance will accelerate from that point. Sometimes theseresistance lines work, sometimes they don’t; they are not foolproof. Butthe author uses resistance lines because they seem to be more accuratethan ordinary trendlines and – most importantly – because they can be drawnbefore the subsequent price action takes place.

LONG TREASURY BONDS

Let’s see how resistance line theory may be helpful this year in gaugingwhen Treasury Bonds Nearest Futures, which have been falling mostlysince their September 1993 peak, etch a major low. Inasmuch as these Trea-suries, Monthly (Chart 1) made a major low in 1981 at 55.156 and morethan doubled it at the 1993 high of 122.313, the most important set ofresistance lines derive from that low and that high. Referring to the chart,we observe that the May 11, 1994 low of 101.125 slightly broke the rising2/3 speed line before reversing upward to close May 20 at 105.000. Impor-tant here are the facts that this same resistance line approximately definedeach of the 1987, 1990 and 1991 Treasury lows. Translated into an opinionon May 20, this means that Treasuries won’t decisively break par this year.Should they do so, it might imply the onset of a renewed inflationary cycle.

Moreover, the second of Gould’s methods, the unit measurement prin-ciple, also helps to determine the importance of the early-May intradaylow of 101.125.

UNIT MEASUREMENT

This technique is sometimes helpful in estimating terminal phases ofadvances and declines, of both individual stocks and market indices. In

other words, what constitutes a price which is “too” high or “too” low. Itsmeasurements are expressed in terms of bull and bear “units.” A bull unitconsists of the number of points of an initial advance by a stock or priceindex following the bottom of a prior important decline, succeeded by asubsequent reaction which, however, remains above that bottom and thenis followed by a second advance that goes beyond the first one. A bear unitis formed in the same manner but in the opposite direction. These mea-surements sometimes portend the length of an overall advance (or decline)and indicate levels at which a trend may meet resistance, or, at times, anextreme reversal.

Chart 1Treasury Bonds Nearest Futures, Monthly

Chart 2Treasury Bonds Nearest Futures, Weekly & Bear Unit Count


Price action with the primary trend frequently “works off” units threetimes (sometimes four times), in accordance with the Rule of Three, thebasis of which is discussed below. In other words, for a move with thetrend, expect three units, but be prepared for the fourth. One other impor-tant point about unit measurement is that recognition of the 2-unit level, bya sharp reaction from it, often indicates that following such a reaction thesecurity will go all the way and work off three, or four units. Whereasrecognition of that level which is defined by 2-l/3 units, without recogni-tion of the 2-unit level (by resistance from it), is grounds for caution, espe-cially for trend followers, since that is often the hallmark of a contratrendmove.

Now we are equipped to develop a second opinion about Treasury BondsNearest Futures, Weekly, which is illustrated in Chart 2. Referring to thechart, we observe that these Treasuries etched a bear unit of 4-7/8 pointsby their initial September 7-22, 1993 decline from 122.313 to 117.438.According to unit measurement theory, then, the contratrend, upward reac-tion from the 2-bear unit level of 112.563, which was reached at the No-vember 23 low of 112.031, implied that Treasuries would go down all theway – to work off either 3 or 4 bear units to 107.688, or 102.813, respec-tively. With the actual intraday May 11 low of 101.125, this close – lessthan 2 percent away-recognition of the theoretically maximum 4-bear unitcount, also leads to the conclusion that that was a low in Treasuries ofmajor importance.

Chart 3NYSE Financials with Resistance Lines

& S&P 40 Utilities with Unit Count

INTEREST-SENSITIVE STOCK MARKET INDICES

Because of the importance of the monetary factors, ideally the resis-tance line and unit measurement theories should also be reflected in inter-est-sensitive stock market indices. Sure enough, the New York Stock Ex-change Financials and Standard & Poors 40 Utilities indices (both on Chart3) did, so far, in 1994 faithfully reflect both resistance line measurementand unit measurement theory, respectively. Referring to the chart, we ob-

serve that the Financials’ week of April 8, 1994 low of 200.01 and all sub-sequent lows, which were higher (itself a positive divergence), reversedupward above the rising 2/3 speed resistance line from the 1990 low. Gouldalways said that the ability of a price index to stay above its rising 2/3speed resistance lines during reactions was the hallmark of a powerful ad-vance.

Whereas the S&P 40 Utilities, which etched a bear unit of 10.29 pointsby the September 17-October 15, 1993 decline from 189.49 to 179.20,worked off a fairly precise 4-bear unit count to 148.33, compared to theactual May 13 low of 146.85. Close enough. Moreover, these Utilities alsorespected their rising 2/3 speed resistance line from the 1981 low, whichapproximated this 4-bear unit count.

It logically follows from each of these two theories that should the afore-said risk parameters of these three interest-sensitive indices – Treasuries,NYSE Financials, S&P 40 Utilities – be decisively downside penetratedon a closing basis that the bear market in bonds not only had more to go onthe downside but also that stocks might have begun a bear market as well.However, the author does not believe that is the case at May 20, 1994, aswe examine next.

Chart 4S&P 400 Industrials, Monthly with Grand Bull Unit Count from 1982

A STOCK MARKET ROAD-MAP

Gould also said that over longer periods of time unit measurement wasuseful, too: “A ‘grand unit’ is, as the name implies, a big unit sometimestaking months to complete and years to confirm.”10 We think this theoryhas been remarkably accurate since the stock market’s 1982 low and that itis relevant now. Referring to Standard & Poors 400 Industrials (Chart 4)we observe that during the 14-month long 1982-1983 advance from 113.08to 195.25, a grand bull unit of 82.19 points was etched – the l0-month-long1983-1984 decline to 167.64 not exceeding the 1982 low.

Thereafter, the S&P 400 steadily rose until hitting the 1986 peak of282.87, which was less than 2 percent above the 2-bull unit count at 277.46.The subsequent 12 percent reaction to that year’s September low of 252.07constituted recognition that the unit measurement principle was operative,and that the S&P 400 would go on to work off at least 3 or 4 bull units.


From the 1986 low, the S&P 400 gathered steam and began to acceler-ate in 1987, reaching the 3-bull unit count of 359.65 in June. That levelwas potentially an important peak level in accordance with the theory – expect three units. During the next two months the S&P 400 overshot the3-bull unit count but peaked 9.7 percent higher (intraday) in August, fol-lowed by the crash.

From the 1987 crash low, stocks steadily rose until hitting the July 1990peak of 438.56 which was less than one percent below the 4-bull unit countto 441.84. That was a perfect fourth and final move, according to Gould’sunit measurement theory. During the next three months stocks fell by 21percent.

Of current relevance, in the author’s opinion and experience, is thatsometimes unit counts will work off a double set of units, i.e., 6 or 8 units.This appears to be the current case for the S&P 400 Industrials, which,rising from the October 1990 low of 345.79, recognized the five-bull unitcount to 524.03 repeatedly last year by resisting further advance. How-ever, by late-1993 that level was decisively exceeded. This means to usthat the theory is saying the stock market should continue to rise until reach-ing at least the 6-bull unit count to 606.22, before the bull market whichbegan from the 1982 low is over.

In 1994 the S&P 400 Industrials advanced further to reach the 560.88level in February, before reacting to the April 20th low of 507.36, a drop of9.5 percent. Referring again to Chart 4, we further observe that during theFebruary-April reaction the rising 2/3 speed resistance line from the 1990low, which during April was at the 500 level, was effective in defining thatmonth's low. This means we believe current risk from the May 20th closeof 530.88 approximates four percent, say 510, whereas potential reward –to 606.22 – would be a gain of 14 percent. Those seem like good odds.

THE RULE OF THREE

Now, we examine the third of Edson Gould's theories, the Rule of Three.For reasons about which people have speculated for thousands of years,the number “three” and “four” have a meaning of finality about them. Forexample, Aristotle said, the “Triad is the number of the whole, inasmuchas it contains a beginning, a middle and an end.” This concept may bedeeply rooted in the natural family unit of father, mother and child, whichis given religious expression in the concept of the Holy Trinity. The finan-cial markets, which, after all, reflect human emotions, also frequently actin the same way. Sometimes there is a fourth movement, which usually ischaracterized as a “now and never” action, climactic in nature. (Threestrikes you're out; four balls take a walk). That financial markets and indi-vidual stocks typically – but not always – move in a series of three or foursteps is apparent in both very short-term moves as well as those encom-passing months and even years.

Chart 5Drug Shares Index, Weekly (Sum of BMY, LLY, MKC, PFE, UPJ x 4.5841)

DRUG SHARES INDEX

However, to simply illustrate the Rule of Three we next examine theDrug Shares Index (Chart 5), a composite of Bristol Meyers, Lilly, MarionMerrill Dow, Merck, Pfizer and Upjohn. Between January 8, 1992 andAugust 13, 1993 the Drugs dropped 42.8 percent in a classic bear market,which consisted of four steps down. Also, helping define the fourth step asthe final one was the fact that the August 13, 1993 low of 1011.46 closelyapproximated the 3-bear unit count from the 1992 peak, at 1012.93.

After rallying 20 percent from the 1993 low, to 1217.04 on January 14,1994, the Drugs came down again to etch a successful test of last year’slow, at the April 15, 1994 low of 1014.84. In other words, we are confidentthat a classic double bottom has been put in place for this group. Addition-ally, as illustrated later, another yardstick of extreme investor behavior tar-geted both Lilly and Merck as having etched final lows last year and thisyear.

TECHNOLOGY AND GROWTH STOCKS

No discussion of the stock market would be complete without address-ing the role of the technology and growth sectors. They are important notonly because they often represent the fastest growing companies, butalso, as I stated in my book which was first published in 1970, “...glamour/growth stocks which, because they are highly volatile – ordinarily two-and-a-half times more so than those in the DJIA – are favorite vehicles ofsophisticated investors.”11 This volatility provides greater time opportu-nity than is available in the behavior of most other stocks.

Edson Gould, “put together the first ‘glamour average’ back in 1960,”12

though, surprisingly, the pamphlet, A Vital Anatomy, from which we’vealso earlier quoted various Gould statements about his theories and meth-ods, says nothing whatsoever about “glamour” stocks. Having originallygotten this idea from Gould in the late 1960s I created my own “GlamourPrice Index,” which consisted of the stocks of eleven highly-regarded, well-known, technology-oriented companies.l3 However, in the most recent


edition of my book, I noted that in recent years I’ve scrapped my original“Glamour” and several other technology- or growth-based indices in favorof the more representative Hambrecht & Quist Technology Stock Index14

and its sub-index of even more rapidly growing, smaller companies, theH&Q Growth Stocks Index. But inasmuch as the H&Q indices includestocks in the biotechnology sector, which I believe march to a differenttune than other growth and technology types, I also have created two otherindices which consist of the numerical values of each of the respectiveH&Q indices less the CBOE Biotechnology Stock Index. Hence, in thetechnology and growth sectors, we examine these five different indices:

1. H&Q Technology Stock Index, which is comprised of the publiclytraded stocks of 200 technology companies, broadly defined in fivebasic groups: Computer Hardware, Computer Software, Communica-tions, Semiconductors, Health Care (within which is a Biotechnologysub-index). The index was originally conceived in the 1970s as a price-weighted index. In 1985 it was reconstructed and market capitalizationweighted. Changes in the index occur as mergers, acquisitions and fail-ures dictate – not infrequently.

2. H&Q Growth Stock Index is a subset of the Technology Index and iscomprised of all companies in the Technology Index which have an-nual revenues of less than $300 million. Companies are removed everyJanuary if they have passed $300 million in revenues.

3. CBOE Biotechnology Stock Index

4. TXB Index, which is the H&Q Technologies Excluding Biotechs

5. GXB Index, which is H&Q Growth Stocks Excluding Biotechs

THE TXB INDEX

Of these five indices, the author thinks the TXB Index is both the mostrepresentative of the overall technology sector as well as being the mostorthodox in reflecting investor psychology We examine it next. Referringto Chart 6, we observe that between their respective 1990 lows and 1994highs (through May 20, 1994), whereas the DJIA gained almost 71 percentand the Dow Transportations rose more than 131 percent, the TXB Indexmore than tripled. So much for volatility!

We also observe that at its March 18, 1994 peak, the TXB Index com-pleted a third step up from its 1991 low. In accordance with the Rule ofThree, this allows for either the possibility that that was a final step, orallows for the emergence of a fourth and final higher high, after the currentreaction is over. We favor the latter possibility and believe Gould’s twoother theories provide well-defined potential risk and reward parametersfor the outcome we envisage.

Chart 6TXB Index, Weekly Average &

Dow Jones 30 Industrials and 20 Transportations

As to risk in the TXB Index, which at May 20, 1994 was down 14-l/2percent from its March 18 peak, it must not break below the rising 2/3speed resistance line from the 1991 low, in order to maintain its bullishuptrend, in accordance with Resistance Line Theory. Inasmuch as TXBclosed at 303.15 on May 20, and the aforesaid 2/3 speed line was nearingthe 287 level, that means we think that risk of this date approximates 5percent.

Whereas potential reward of a possible fourth and final rise of the TXBIndex we think may be estimated through the Unit Measurement method.Referring again to Chart 6, we further observe that the TXB Index etched abull unit of 99.96 points by its initial advance from the September 20, 1991low of 112.29 to the April 24, 1992 high of 212.25. The 2-bull unit level of312.21 was briefly recognized by its one-week reaction from near that levelin early 1994, before advancing to the higher March 18 all-time high. As-suming then, that the aforesaid resistance line risk parameter holds on thecurrent reaction, we believe that potential reward from the May 20 level isabout 35 percent to the 3-bull unit count at 412.17.

These sound like favorable odds of 7-to-1 between possible risk andreward, in the author’s opinion. I note, too, that an overhead trend, whichis projected through the 1992, 1993 and March-1994 peaks and which alsoparallels the rising 2/3 speed resistance line, approximates the 400 level byyear-end 1994, as well. In other words, the author believes that the TXBIndex will rise by about one-third before this sector is vulnerable to a bearmarket.

BEAR MARKET

After the reward area is approximated, that’s from where I think a bearmarket in technology and growth, as well as one for the stock market, over-all, may begin. That there will likely be a bear market between now and1995 is suggested by the facts that every single “5” year in this century hasbeen an up year, which means that there “should” be an intervening bearmarket before the 1995 bull market begins. However, an alternative sce-


nario is simply that it will take between now and year-end 1994 for thereward area to be reached.

If that proves to be the case and the stock market rises to record levelsand nears the potential reward areas we have outlined herein, by year-end1994 (possibly narrowly extending into early 1995), that would be the fourthconsecutive up year – a possible “final” up year, according to the Rule ofThree. In that event, 1995, especially if perceived by “too” many as alwaysan up year since it is a “5” year, would then set the stage for it to becomethe first down “5” year during the past century, in the author’s opinion, i.e.,1995 happens in second-half 1994.

Chart 7Merck, Weekly: 4-Step 1992-1993 Decline & 1994 Test at Cut-In-Half Level

THE CUT-IN-HALF RULE AND ITS OPPOSITE

The fourth gauge for measuring investor extremes is conceptually thesimplest of all – the Cut-In-Half Rule and its opposite. Briefly stated, whenan important stock or price index loses 50 percent of its value, a rally oreven major reversal often originates from near that level. Keep in mindthat the Cut-In-Half Rule and a 50 Percent Retracement are quite different.For example, two stocks each base at the level of 50 and both rise to 100. Ifone declines to 75, before advancing once again, it has retraced 50 percentof its advance from 50 to 100. However, if the other one drops back to 50from 100, it has been “cut-in-half.” A textbook example of the Cut-in-HalfRule is shown in Chart 7 of Merck, which we discuss below.

Why the Cut-In-Half Rule and its related spinoffs often work is prob-ably because the investor crowd quantifies 50 percent off the top as “too”cheap. Whereas the opposite is that after an important stock or index doublesit often runs into trouble. At that point, investors tend to take at least someprofits. But since some indices, individual stocks, commodities and inter-est-bearing securities are more volatile than others, this same yardstick issometimes extended on the way up to a triple, quadruple, quintuple, oreven a sextuple, (with low price stocks sometimes squaring their lows).Whereas, on the way down there is sometimes a double cut-in-half (off 75percent), or – more rarely – a triple cut-in-half (87-l/2 percent off the top).Keep in mind, however, that in applying these Cut-In-Half yardsticks aspotential long entry points, one should be satisfied that the company’s bal-

ance sheet is not in serious question.

STOCK SELECTIONS

Though it can be repeatedly demonstrated that these four theories ofinvestor behavior are constantly operating in all financial markets, in theauthor’s opinion, it does not necessarily follow that one can readily usethem in every instance. Sometimes the units are not readily discernible andthe resistance lines don’t work. Moreover, sometimes there is a fifth stepin an overall advance or decline movement, which appears to contradictthe Rule of Three – though a case might be made that such a fifth steprepresents an undercut (or overcut) test of the fourth step.

However, after using these theories over time to make day-to-day in-vestment decisions, I have found that they are valuable when discernibleand add confidence to a decision. That is particularly the case when morethan one theory appears to be operative in a given situation.

Chart 8US Robotics, Daily with Doubling Levels & Unit Measurement Counts

For example, referring to Chart 7 of Merck (MRK @ 30-l/4), we can ob-serve that when it closely approximated its theoretical cut-in-half level of 28-9/32 during August 1993, at the actual low of 28-5/8, on a fourth (and presumablyfinal) step down from the January 3, 1992 peak of 56-9/16, it appears to haveconstituted a classic buying juncture. Thereafter rallying to 38 by January 5,1994, Merck subsequently tested last year’s low at this year’s April 15, 18 lowsboth at 28-1/8. This makes me confident that the cut-in-half level was, or ap-proximated, a final low, especially since there is no great mystery about Merck’sfundamentals and fifty percent off the top seems a reasonable – if not “too”great – a discount for those investors critical of the Clintons’ health care plans.

More volatile technology and growth stocks sometime reflect these theoreti-cal principles of how crowd behavior plays out in the financial markets in anextraordinary way. For example, referring to Chart 8 of U.S. Robotics (USRX @30-3/4), a world-wide leader in data communications, we can observe that afterdoubling the late-1991 low of 12-l/4 by the early-1992 high of 24-1/4, Roboticsdropped sharply (off 45 percent). Whereas the early-1993 high of 25-1/2 almostdoubled the summer-1992 low of 13-3/8. Then the March 1993 low of 17 was


slightly more than doubled at the October high of 35-1/4, whereas the subsequentdecline to 23 worked off an almost-perfect 4-bear unit count to 23-1/4.

Moreover, this year’s high of 46 (on March 8) was a perfect double,from which a reaction has begun, with a bear unit of 5-1/4 points alreadyetched and confirmed (by a lower low), and more recently appearing torecognize the 3-bear unit level of 30-1/4, by the May 10-11 lows of 29-1/4,as the new low from which to key off. That the stock of a single companycould have gone through so many extreme bull and bear moves, in such ashort period of time, shows not only that Alvin Toffler’s “Future Shock”15

has arrived on Wall Street but also that traditional Wall Street research isincapable of dealing with it effectively. The arrival of “future shock,” whatsome now call the information age, also presents a challenge to stock mar-ket technicians – to do their homework in order to stay ahead of the curve.

May 22, 1994

SEQUEL

At Market Close November 11,1994:

What Happened During the Subsequent Six Months

Treasury Bonds Nearest Futures (Charts 1 and 2) perfectly testedtheir May 11 low at their virtually identical July 11 low of 100.0625 –compared to the May 11, 100.2500 (the numerical value of the Futures areabout one point lower than shown on the chart because the Nearest Futureshad rolled over from June’s to September’s, and currently December’s).Thereafter Treasuries rallied back to the August 5 high of 105.21875, thenslowly eroded until par was broken at the September 22 close of 99.40625.At that point we conceded the 13-year long uptrend in Treasuries was clearlybroken and that the major trend inference of bonds should be assumed asbeing down. By November 11, Treasuries slumped even more, closing at96.0625. Moreover, since this break of the grand resistance line of Trea-suries also took out the 4-bear unit count level, it implies to us that ulti-mately at least six bear units will be worked off. That level is 93.0625.

However, we never changed our positive stock market opinion becausethe two interest-sensitive stock market bellwethers we mostly rely on re-mained intact, notwithstanding the break in bonds.

The NYSE Financials (Chart 3), which had hit an intraday low of 199.95on April 4, closing May 20 at 214.27, slightly exceeded the 220 level dur-ing four days in June, then also slowly eroded until closing November 11exactly at 199.56. While this does constitute a break of its resistance line,and hence is clearly negative as of November 11, it seems such an obvious“test” of the April 4 low, that it is conceivable to us the Financials may beable to mount at least a weak rally from here.

We draw this tentative conclusion because the third interest-sensitiveindex, Standard & Poors 40 Utility Stock Index (also on Chart 3), whichclosed November 11 at 148.51 – still above its May 13 low – we don’tthink will take out that level. Not only has the 4-bear unit count level of148.33 been repeatedly and successfully tested during 15 trading days inOctober and November but is also defined by these Utilities’ rising 2/3speed resistance line from the March 1980 low. That is about as preciserecognition of a Gouldian-defined risk parameter as it ever gets! Naturally,this also means that a decisive break of it would undoubtedly require somechange in our current stock market opinion.

Our Stock Market Road-Map for Standard & Poors 400 Industri-

als (Chart 4 and Chart 9), successfully tested the April 20 low (507.30) atthe June 27 low of 511.90, thereafter rising to an all time high of 564.50 onOctober 31. Closing November 11 at 550.87, potential reward has nowmoved down to only 10 percent whereas near-term risk remains about 4percent (the rising 2/3 speed resistance line moving up to about 527). Notquite as good odds as on May 20.

The Drug Shares Index (Chart 5) was up almost 18 percent at Novem-

ber 11 from May 20 and we think is headed substantially higher. Thoughgaining almost 30 percent from its April low at the November 11 close of1316.70, we think the Drugs will work off at least three bull units, a furthergain of 25 percent from here. Three bull units were worked off on the waydown, so why not three bull units on the way up?

Chart 9S&P 400 Industrials, TXB Index (Technology Less Biotechs)

The TXB Index (Technologies Less Biotechs) (Chart 6 and Chart 9), atits June 23 daily close of 293.97, never broke below its 2/3 speed resis-tance line risk parameter and subsequently rose 28.9 percent from that lowto 378.83 on November 9. Obviously the odds of further gain from herehave sharply deteriorated, potential remaining reward only a possible ad-ditional 8.8 percent, in our opinion. We have chosen to deal with this changeof the odds by building cash as specific technology and growth compo-nents reach their individual, respective potential reward zones.

Merck (MRK @ 36-3/4) (Chart 7) hit a recovery high of 37-5/8 onNovember 10 and we believe is headed into the 44-45 zone. That is de-fined by both a bull unit count and an overhead declining 1/3 speed resis-tance line.

Whereas U.S. Robotics (USRX @ 38-3/4) (Chart 8) worked off a fourthbear unit at its June 2nd low of 24, then etched a new bull unit of 5-1/2points by its subsequent initial rise to 29-1/2. USRX went on to slightlyexceed the 3-bull unit count of 40-1/2. at the November 9 high of 42-1/4.The maximum upside potential we see from here, is a 4-bull unit count to46, which would also be a prospective double top with the early-1994 peak.

CONCLUSION

I believe that this real-time experience in using the Gouldian theoriesamply demonstrates both their usefulness as well as their drawbacks, thoughonly scratching the surface of their potential applications. Their key ad-vantages are that Gould’s quantifications of investor sentiment help one toboth reach and act upon specific investment conclusions on a case by casebasis, without being held hostage to an endless, self-imposed debate aboutwhat to do.


REFERENCES

1. Why Most Investors Are Mostly Wrong Most of the Time, W. X.Scheinman, 1991, Fraser Publishing Company (p. 139)

2. Scheinman, ibid

3. The Crowd, G. LeBon, 1896, Fraser Publishing Company (1982)

4. A Vital Anatomy, E. Gould, (Undated), Anametrics, Inc.

5. Gould, ibid

6. Gould, ibid

7. Gould, ibid

8. Gould, ibid

9. Gould, ibid

10. Gould, ibid

11. Scheinman, ibid

12. Gould, ibid

13. Scheinman, ibid

14. Hambrecht & Quist Technology and Growth Indices, Michael De Wittand Shiela Ennis, Hambrecht & Quist Incorporated, January 1993

15. Future Shock, A. Toftler, 1971, Bantam Doubleday

BIBLIOGRAPHY

Numbers■ Jung, C.G., Collected Works of C.G. Jung, General Index, (Volume 20,

pp. 485-489, “Numbers”), Princeton University Press, 1979

■ Menninger, K., Number Words and Number Symbols; A Cultural His-

tory of Numbers, The M.I.T Press, 1970

■ Von Franz, M-L, Number and Time, Northwestern University Press,1974

Technology■ Veblen, T., Imperial Germany and the Industrial Revolution, Transac-

tion Publishers (1990 reprint)

William X. Scheinman, was a registered investment advisor since 1968. He movedfrom Wall Street to Reno, Nevada in December 1974 where he advised financial insti-tutions worldwide. One of the founding members of the Market Technicians Associa-tion, Bill was the founder of the African-American Students Foundation, Inc. whichbetween 1956 and 1961 brought more than 1,000 students from all over Africa to theUnited States to attend institutions of higher learning. He was the founder of theAfrican-American Leadership Foundation, Inc. Bill died on May 24, 1999.


“This indicator has always produced huge profits! In fact,

you would have doubled your money in just six months!”

Such a claim could be a sales pitch. It could also be an analyst’s enthu-siasm about some work just completed. But in either case, such claimsappear to be meeting increasing skepticism, perhaps because enough haveproven to be based more on fiction than quantifiable fact, perhaps becauseenough investors have been burned by indicators that have failed to panout when put to real-time use, perhaps because the combination of ever-strengthening computing power and ever-increasing program complexityhave made excessive optimization as easy, and dangerous, as ever.

In any case, the need to quantify accurately and thoroughly is greaterthan ever. Honest and reliable quantification methods, used in the correctway, are needed for increased research credibility. They are needed to im-part objectivity. They are needed for effective analysis and for the soundbacking of research findings. The alternative is the purely subjective ap-proach that uses trendlines and chart patterns alone, making no attempt toquantify historical activity. But when the quantification process fails todeliver, instead producing misleading messages, the subjective approachis no worse an alternative – a misguided quantification effort can be worsethan none at all. The predicament, then, is how to truly add value throughquantification.

THE CONCERNS

The major reason for quantifying results is to assess the reliability andvalue of a current or potential indicator, and the major reason we haveindicators is to help us interpret the historical data. The more effective theinterpretation of historical market activity, the more accurate the projec-tion about a market’s future course. An indicator can be a useful source ofinput for developing a market outlook if quantitative methods back its re-liability.

But for several reasons, quantification must be handled with care. Theinitial concern is the data used to develop an indicator. If it’s inaccurate,incomplete, or subject to revision, it can do more harm than good, issuingmisleading messages about the market that’s under analysis. The data shouldbe clean and contain as much history as possible. When it comes to data,more is better – the greater the data history, the more numerous the likeoccurrences, and the greater the number of market cycles under study.

This leads to the second quantification concern, and that’s sample size.The data may be extensive and clean, and the analysis may yield an indica-tor that foretold the market’s direction with 100% accuracy. But if, forexample, the record was based on just three cases, the results would lackstatistical significance and predictive value. In contrast, there would befewer questions regarding the statistical validity of results based on morethan 30 observations.

The third consideration is the benchmark, or the standard for compari-son. The test of an indicator is not whether it would have produced a profit,but whether the profit would have been any better than a random approach,or no approach at all. Without a benchmark, “random walk” suspicionsmay haunt the results.1

The fourth general concern is the indicator’s robustness, or fitness –the consistency of the results of indicators with similar formulas. If, forexample, the analysis would lead to an indicator that used a 30-week mov-ing average to produce signals with an excellent hypothetical track record,

how different would the results be using moving averages of 28, 29, 31, or32 weeks? If the answer was “dramatically worse,” then the indicator’srobustness would be thrown into question, raising the possibility that thehistorical result was an exception to the rule rather than a good example ofthe rule. An indicator can be considered “fit” if various alterations of theformula would produce similar results.

Figure 1Summary Results From Hypothetical Indicator Tests

These results contain an impressive-looking EXCEPTION to the rule ...Number Moving

of Average Accuracy Gain/Annum

Trades (Periods) Buy Level Sell Level Rate (%) (%)

40 70 100 110 50 11.2

39 71 99 111 50 11.3

37 72 98 112 65 15.1

37 73 97 113 52 10.1

36 74 96 114 50 9.8

These results would all be good EXAMPLES of the rule ...

50 20 15.6 8.6 55 11.8

49 21 15.8 8.4 56 12.0

48 22 16.0 8.2 56 12.1

47 23 16.2 8.0 57 12.1

46 24 16.4 7.8 56 12.0

Buy-Hold Gain/Annum 6.3

Moreover, the non-robust indicator may be a symptom of the fifth con-cern, and that’s the optimization process. In recent years, much has beenwritten about the dangers of excessive curvefitting and over-optimization,often the result of unharnessed computing power. As analytical programshave become increasingly complex and able to crunch through an ever-expanding multitude of iterations, it has become easy to over-optimize.The risk is that armed with numerous variables to test with minuscule in-crements, a program may be able to pick out an impressive result that mayin fact be attributable to little more than chance. The accuracy rate andgain per annum columns of Figure 1 compare results that include an im-pressive-looking indicator that stands in isolation (top) with indicators thatlook less impressive but have similar formulas (bottom). One could havefar more confidence using an indicator from the latter group even thoughnone of them could match the results using the impressive-looking indica-tor from the top group.

What follows from these five concerns is the final general concern ofwhether the indicator will hold up on a real-time basis. One approach is tobuild the indicator and then let it operate for a period of time as a real-timetest. At the end of the test period, its effectiveness would be assessed. Toincrease the chances that it will hold up on a real-time basis, the alterna-tives include out-ofsample testing and blind simulation. An out-of-sampleapproach might, for example, require optimization over the first half of thedate range and then a real-time simulation over the second half. The resultsfrom the two halves would then be compared. A blind-simulation approachmight include optimization over one period followed by several tests ofthe indicator over different periods.

Whatever the approach, real-time results are likely to be less impres-

THE QUANTIFICATION PREDICAMENT

Timothy W. Hayes, CMT



sive than results during an optimization period. The reality of any indica-tor developed through optimization is that, as history never repeats itselfexactly, it is unlikely that any optimized indicator will do as well in thereal-time future. The indicator’s creator and user must decide how muchdeterioration can be lived with, which will help determine whether to keepthe indicator or go back to the drawing board.

TRADE-SIGNAL ANALYSIS

With the general concerns in mind, the various quantification methodscan be put to use. The first, and perhaps most widely used, is the approachthat relies on buy and sell signals, as shown in Figure 2.2 When the indi-cator meets the condition that it deems to be bullish for the market in ques-tion, it flashes a buy signal, and that signal remains in effect until the indi-cator meets the condition that it deems to be bearish. A sell signal is thengenerated and remains in effect until the next buy signal. Since a buy sig-nal is always followed by a sell signal, and since a sell signal is alwaysfollowed by a buy signal, the approach lends itself to quantification asthough the indicator was a trading system, with a long position assumed ona buy signal and closed out on a sell signal, at which point a short positionwould be held until the next buy signal.

Figure 2

The method’s greatest benefit is that it clearly reveals the indicator’saccuracy rate, a statistic that’s appealing for its simplicity – all else beingequal, an indicator that had generated hypothetical profits on 30 of 40 tradeswould be more appealing than an indicator that had produced hypotheticalprofits on 15 of 40 trades. Also, the simulated trading system can be usedfor comparing a number of other statistics, such as the hypothetical perannum return that would have been produced by using the indicator. Theper annum return can then be compared to the gain per annum of the bench-mark index.

But the method’s greatest benefit may also be its biggest drawback. Nosingle indicator should ever be used as a mechanical trading system – asstated earlier, indicators should instead be used as tools for interpretingmarket activity. Yet, the hypothetical and actual can be easily confused.Although the signal-based method specifies how a market has done be-tween the periods from one signal to the next, they are not actual records ofreal-time trading performance. If they were, the results would have to ac-count for the transaction costs per trade, with a negative effect on tradingresults. Figure 3 summarizes the indicator’s hypothetical trade results be-fore and after the inclusion of a quarter-percent transaction cost, illustrat-ing the impact that transaction costs can have on results. The more numer-ous the signals, the greater the impact.

Also, as noted in the results, another concern is the maximum draw-down, or the maximum loss between any consecutive signals. But again,as long as it is clear that the indicator is for perspective and not for dictat-ing precise trading actions, indicators with trading signals can provide usefulinput when determining good periods for entering and exiting the marketin question.

ZONE ANALYSIS

In contrast to indicators based on trading signals, indicators based onzone analysis leave little room for doubt about their purpose – they don’teven have buy and sell signals. Rather, zone analysis recognizes black,white and one or more shades of gray. It quantifies the market’s perfor-mance with the indicator in various zones, which can be given such labelsas “bullish,” “bearish” or “neutral” depending upon the market’s per an-num performance during all of the periods in each zone. Each period in azone spans from the first time the indicator enters the zone to the nextobservation outside of the zone. Unlike the signal-based approach, the in-dicator can move from a bullish zone to a neutral zone and back to a bull-ish zone. An intervening move into a bearish zone is not required.

Figure 3Summary Results For Indicator In Figure 2 — No Transaction Costs

Value Line Geometric $ 574,104 1/24/72 – 5/30/96Last Profit Number Days Gain Model Buy/HoldSignal Current of Per Per Batting Gain Per Gain Per $10,000"Sell" Trade Trades Trade Trade Average Annum Annum Investment

5/07/96 -2.9% 240 37 1.9% 50% 18.1% 4.8% $574,104

Maximum Drawdown: -4.68%

Summary Results For Indicator In Figure 2 — Including Transaction CostsOf A Quarter Percent Per Trade

Value Line Geometric $173,271 1/24/72 – 5/30/96Last Profit Number Days Gain Model Buy/HoldSignal Current of Per Per Batting Gain Per Gain Per $10,000"Sell" Trade Trades Trade Trade Average Annum Annum Investment

5/07/96 -3.4% 240 37 1.4% 45% 12.4% 4.8% $173,271

Maximum Drawdown: -4.68%

Zone analysis is therefore appealing for its ability to provide usefulperspective without a simulated trading system. The results simply indi-cate how the market has done with the indicator in each zone. But this typeof analysis has land mines of its own. In determining the appropriate lev-els, the most statistically-preferable approach would be to identify the lev-els that would keep the indicator in each zone for roughly an equal amountof time. In many cases, however, the greatest gains and losses will occur inextreme zones visited for a small percentage of time, which can be prob-lematic for several reasons:

1. if the time spent in the zone is less than a year, the per annum gain canpresent an inflated picture of performance;

2. if the small amount of time meant that the indicator made only onesortie into the zone, or even a few, the lack of observations would lendsuspicion to the indicator’s future reliability;

3. the indicator’s usefulness must be questioned if it’s neutral for the vastmajority of time.

A good compromise between optimal hypothetical returns and statisti-cal relevance would be an indicator that spends about 30% of its time inthe high and low zones, like the indicator in Figure 4. For an indicator withmore than four years of data, that would ensure at least a year’s worth oftime in the high and low zones and would make a deficiency of observa-tions less likely. In effect, the time-in-zone limit prevents excessive opti-mization by excluding zone-level possibilities would look the most im-


pressive based on per annum gain alone.Another consideration is that in some cases, a closer examination of

the zone performance reveals that the bullish-zone gains and bearish-zonelosses occurred with the indicator moving in particular directions. In thosecases, the bullish or bearish messages suggested by the per annum resultswould be misleading for a good portion of the time, as the market mightactually have had a consistent tendency, for example, to fall after theindicator’s first move into the bullish zone and to rise after its first moveinto the bearish zone.

Figure 4

It can therefore be useful to subdivide the zones into rising-in-zone andfalling-in-zone, which can have the added benefit of making the informa-tion in the neutral zone more useful. This requires definitions for “rising”and “falling.” One way to define those terms is through the indicator’s rateof change. In Figure 5, which applies the approach to the primary stockmarket model used by Ned Davis Research, the indicator is “rising” in thezone if it’s higher than it was five weeks ago and “falling” if it’s lower.Again, the time spent in the zones and the number of cases are foremostconcerns when using this approach.

Figure 5

Alternatively, “rising” and “falling” can be defined using percentagereversals from extremes, in effect using zones and trading signals to con-firm one another. In Figure 6, for example, the CRB Index indicator is“rising” and on a sell signal once the indicator has risen from a troughwhereas it’s “falling” and on a buy signal after the indicator has declinedfrom a peak. Even though the reversal requirements resulted from optimi-

zation, the indicator includes a few poorly-timed signals and would berisky to use on its own. But the signals could be used to provide confirma-tion with the indicator in its bullish or bearish zone, in this case the samezones as those used in Figure 4. For example, in late 1972 and early 1973the indicator would have been rising and in the upper zone, a confirmedbearish message. The indicator would then have peaked and started to loseupside momentum, generating a “falling” signal and losing the confirma-tion. That signal would not be confirmed until the indicator’s subsequentdrop into its lower zone.

Figure 6

The chart’s box shows the negative hypothetical returns with the indi-cator on a sell signal while in the upper zone, and on a buy signal while inthe lower zone. In contrast to the rate-of-change approach to subdividingzones, this method fails to address the market action with the indicator inthe middle zone. But it does illustrate how zone analysis can be used to inconjunction with trade-signal analysis to gauge the strength of an indicator’smessage.

SUBSEQUENT-PERFORMANCE ANALYSIS

In addition to using signals and zones, results can be quantified by gaug-ing market performance over various periods following a specified condi-tion. In contrast to the trade-signal and zone-based quantification meth-ods, a system based on subsequent performance calculates market perfor-mance after different specified time periods have elapsed. Once the long-est of the time periods passes, the quantification process becomes inactive,remaining dormant until the indicator generates a new signal. In contrast,the other two approaches are always active, calculating market performancewith every data update.

The subsequent-performance approach is thus applicable to indicatorsthat are more useful for providing indications about one side of a market,indicating market advances or market declines. And it’s especially usefulfor indicators with signals that are most effective for a limited amount oftime, after which they lose their relevance. The results for a good buy-signal indicator are shown in Figure 7, which lists market performanceover several periods following signals produced by a 1.91 ratio of the 10-day advance total to the 10-day decline total.

In its most basic form, the results might list performance over the nextfive trading days, 10 trading days, etc., summarizing those results with theaverage gain for each period. However, the results can be misleading ifseveral other questions are not addressed. First of all, how is the averagedetermined? If the mean and the median are close, as they are in Figure 7,then the mean is an acceptable measure. But if the mean is skewed in one


direction by one or a few extreme observations, then the median is usuallypreferable. In both cases, the more observations the better.

Secondly, what’s the benchmark? While the zone approach uses rela-tive performance to quantify results, trade-signal analysis includes a com-parison of per annum gains with the buy-hold statistic. Likewise, the sub-sequent-performance approach can use an all-period gain statistic as abenchmark. In Figure 7, for instance, the average 10-day gain in the DowIndustrials has been 2% following a signal, nearly seven times the 0.3%mean gain for all 10-day periods. This indicates that the market has tendedto perform better than normal following signals. That could not be said ifthe 10-day gain was 0.4% following signals.

Figure 7

Percent Change Of Dow Industrials Following 1.91 Ratio Of 10-dayAdvances To 10-Day Declines

Trading Days LaterSignal 10-DayDate A/D 5 10 22 63 126 252

06/23/47 1.96 -0.1 2.9 5.3 0.3 0.1 3.7

03/29/48 2.05 2.2 3.2 5.8 11.2 4.0 0.6

07/13/49 2.06 1.4 1.9 3.5 7.0 15.2 28.4

11/20/50 2.01 1.5 -1.7 -1.4 10.0 9.8 18.8

01/25/54 2.00 0.5 1.1 0.3 8.3 18.2 36.4

01/24/58 2.00 -0.1 -0.4 -3.1 0.6 10.3 31.4

07/10/62 1.98 -1.4 -2.0 0.9 0.0 14.0 21.5

11/07/62 1.91 2.4 3.5 4.8 10.3 17.3 21.1

01/13/67 1.94 1.4 1.1 2.6 2.9 5.6 6.9

08/31/70 1.91 1.1 -1.8 -0.5 3.9 15.5 17.9

12/03/70 1.95 1.5 1.7 3.6 11.1 14.1 5.0

12/08/71 1.98 1.0 3.5 6.2 10.6 10.4 20.2

01/08/75 1.98 2.8 2.7 12.0 20.9 37.2 41.4

01/06/76 2.05 2.5 6.6 8.3 12.7 11.3 10.9

08/23/82 2.02 0.2 2.6 3.9 14.6 22.6 34.0

10/13/82 2.03 1.9 -0.9 2.4 6.7 13.9 24.6

01/21/85 1.93 1.3 2.3 1.4 0.4 7.6 20.1

01/14/87 2.19 2.9 6.3 7.3 10.7 22.1 -5.4

02/04/91 1.96 4.7 5.8 6.9 6.1 7.8 16.7

01/06/92 1.99 -0.5 1.7 1.8 1.5 4.3 3.4

Median 1.4 2.1 3.6 7.7 12.6 19.4

Mean 3.1 2.0 3.6 7.5 13.1 17.9

Mean – All Periods 0.2 0.3 0.7 2.0 4.0 8.1

% Cases Higher 80 75 85 100 100 100

% Cases Higher All Periods 56 58 60 63 67 70

Signals based on 10-day total of NYSE advances over 10-day total of NYSE declines. Concept courtesy of

Dan Sullivan, modified by Ned Davis Research.

A third question is how much risk has there been following a buy-sig-nal system, or reward following a sell-signal system? Using a buy-signalsystem as an example, one way to address the question would be to list thepercentage of cases in which the market was higher over the subsequentperiod, and to then compare that with the percentage of cases in which themarket was higher over any period of the same length. Again using the 10-day span in Figure 7 as an example, the market has been higher after 75%of the signals, yet the market has been up in only 58% of all 10-day peri-ods, supporting the significance of signals. Additional risk information couldbe provided by determining the average drawdown per signal – i.e., themean maximum loss from high to low following signals. The mean for the10-day period, for example, was a maximum loss of 0.7% per signal, sug-

gesting that at some point during the 10-day span, a decline of 0.7% couldbe considered normal. The opposite approaches could be used with sell-signal indicators, with the results reflecting the chances for the market tofollow sell signals by rising, and to what extent.

Along with those questions, the potential for double-counting must berecognized. If, for example, a signal is generated in January and a secondsignal is generated in February, the four-month performance following theJanuary signal would be the same as the three-month performance follow-ing the February signal. This raises the question of whether the three-monthreturn reflects the impact of the first signal or the second one. Moreover,such signal clusters give heavier weight to particular periods of marketperformance, making the summary statistics more difficult to interpret. Prob-lems related to double-counting can be reduced or eliminated by adding atime requirement. For the signals in Figure 7, for instance, the conditionmust be met for the first time in 50 days – if the ratio reaches 1.92, drops to1.90, and then returns to 1.92 two days later, only the first day will have asignal. The time requirement eliminates the potential for double-countingin any of the periods of less than 50 days, though the longer periods stillcontain some overlap in this example.

Figure 8

Performance Of Dow Industrials Following Initial Index Confirmation(Joint 52-week Highs For The First Time In A Year)

26 Weeks Later 39 Weeks Later 52 Weeks Later

Confirming % Mean % All % Mean % All % Mean % All LatestIndex Cases Higher % Gain Periods Higher % Gain Periods Higher % Gain Periods Close

New York Utilities 7 100 8.79 5.53 100 13.59 7.98 100 16.62 10.35 5/12/95

World Composite 6 100 8.47 5.85 80 9.74 8.88 100 12.91 11.78 9/15/95

Weekly New Highs 9 89 7.79 3.53 78 10.09 5.27 100 14.41 6.98 3/31/95

NYSE Weekly Volume 10 70 6.64 3.53 67 5.16 5.27 89 6.91 6.98 7/14/95

S&P 500 Composite 22 73 5.13 3.53 73 9.92 5.45 82 14.78 7.47 2/10/95

NYSE Composite 20 63 4.02 3.72 68 8.10 5.68 79 13.29 7.60 2/10/95

AMEX Composite 9 67 3.53 5.53 67 8.20 7.98 78 13.62 10.35 3/31/95

OTC Composite 9 56 3.77 5.53 67 7.73 7.98 78 12.38 10.35 3/17/95

Dow Transports 22 77 5.26 3.53 73 8.62 5.45 76 9.99 7.47 4/13/95

S&P High-Grade Index 12 67 5.46 3.53 75 9.73 5.27 75 10.77 6.98 2/17/95

S&P Industrials 12 58 1.66 3.53 58 4.34 5.27 75 10.15 6.98 2/10/95

NYSE Financials 11 45 0.59 3.53 55 4.86 5.24 73 10.06 6.92 4/07/95

Dow Utilities 23 70 6.00 3.27 65 7.63 5.05 73 9.18 6.95 5/05/95

Weekly A/D Line 12 58 2.44 3.53 67 5.30 5.27 73 7.31 6.98 4/13/95

S&P Low-Priced Index 11 55 1.26 3.53 40 2.88 5.27 70 7.31 6.98 7/14/95

Value Line Composite 14 50 1.35 3.53 50 3.74 5.27 69 6.26 6.98 4/13/95

Confirmation occurs, and a case identified, when the DJIA and the index in question both reach 52-week highs, the first such joint occurrence in at least a year. Table is sorted based on percentage of casesin which the index was higher over the subsequent 52-week periods (column shaded). “% All Periods” isthe DJIA’s mean gain for all 26, 39, and 52 week periods starting with the beginning of the data series inquestion. Table updated through 4/04/96.

Another application of subsequent-performance analysis is shown inFigure 8, which is not prone to any double-counting. The signals requirethat three conditions are met, all for the first time in year – the Dow Indus-trials much reach its highest level in a year, another index must reach itshighest level in a year, and the joint high must be the first in a year. Thesignificance for the various indices can then be compared in conjunctionwith their benchmarks – i.e., the various all-period gains. Figure 9 uses 12of those indices to show how subsequent performance analysis for bothbuy signals and sell signals can be used together in an indicator. For eachtime span, the chart’s box lists the market’s performance after buy signals,after sell signals, and for all periods.


REVERSAL-PROBABILITY ANALYSIS

Finally, the subsequent performance approach is useful for assessingthe chances of a market reversal. In Figure 10, the “signal” is the market’syear-to-year change at the end of the year, with the signals (years) catego-rized by the amount of change – years with any amount of change, thosewith gains of more than 5%, etc. In this case, the subsequent-performanceanalysis is limited to the year after the various one-year gains. But theanalysis takes an additional step in assessing the chances for a bull marketpeak within the one- and two-year periods after the years with market gains,or a bear market bottom within the one- and two-year periods after theyears with market declines.

Figure 9

This analysis requires the use of tops and bottoms identified with ob-jective criteria for bull and bear markets in the Dow Industrials. The rever-sal dates show that starting with 1900, there have been 30 bull marketpeaks and 30 bear market bottoms, with no more than a single peak and asingle trough in any year. This means that for any given year until 1995,there was a 31% chance for the year to contain a bull market peak and a31% chance for the year to contain a bear market bottom (30 years withreversals / 95 years).

Figure 10

Using this percentage as a benchmark, it can then be determined whetherthere’s been a significant increase in the chances for a peak or trough in theyear after a one-year gain or loss of at least a certain amount. The chart’sboxes show the peak chances following up years and the trough chancesfollowing down years, dividing the number of cases by the number of peaks

or troughs. For example, prior to 1995, there had been 31 years with gainsin excess of 15% starting with 1899. After those years, there was a 52%chance for a bull market peak in the subsequent year (16 following-yearswith peaks / 31 years with gains of more than 15%). The chances for apeak within two years increased to 74%, which can be compared to thebenchmark chance for at least one peak in 61% of the two-year periods(since several two-year periods contained more than one top, this is not theexact double of the chances for a peak in any given year).

A major difference in this analysis is that in contrast to signals andzones, which depend upon the action of an indicator, this approach de-pends entirely on time. Each signal occurs after a fixed amount of time(one year), with the signals classified by what they show (a gain of morethan 5%, etc.). Depending upon the classification, the risk of a peak ortrough can then be assessed.

CONCLUSION

Each one of these methods can help in the effort to assess a market’supside and downside potential, with the method selected having a lot to dowith the nature of the indicator, the time frame, and the frequency of oc-currences. The different analytical methods could be used to confirm oneanother, the confirmation building as the green lights appeared. An alter-native would be a common denominator approach in which several of theapproaches would be applied to an indicator using a common parameter(i.e., a buy signal at 100). Although the parameter would most likely beless than optimal for any of the individual methods, excessive optimiza-tion would be held in check. But whatever approaches are used, it needs tobe stressed that each one of them has its own means of deceiving. By betterunderstanding the potential pitfalls of each approach, indicator develop-ment can be enhanced, indicator attributes and drawbacks can be betterassessed, and the indicator messages can be better interpreted.

The process of developing a market outlook must be based entirely onresearch, not sales. The goal of research is to determine if something works.The goal of sales is to show that it does work. Yet in market analysis, thelines can blur if the analyst decides how the market is supposed to per-form, then selling himself on this view by focusing only on the evidencethat supports it. What’s worse is the potential to sell oneself on the value ofan indicator by focusing only on those statistics that support one’s view,regardless of their statistical validity. As shown by the various hazardsassociated with the methods described in this paper, such self-deception isnot difficult to do.

Our goals should be objectivity, accuracy, and thoroughness. Using asound research approach, we can determine the relative value of using anyparticular indicator in various ways. And we can assess the indicator’s valueand role relative to all the other indicators analyzed and quantified in asimilar way. The indicator spectrum can then provide more useful inputtoward a research-based market view.

FOOTNOTES

1. Reference to Burton Malkeil's A Random Walk Down Wall Street, whichargues that stock prices move randomly and thus cannot be forecastedthrough technical means.

2. The charts that accompany this paper were produced with the Ned DavisResearch computer program.

Since winning the third Dow Award in 1996, Tim Hayes has expanded upon"The Quantification Predicament" in writing his first book, "The Research DrivenInvestor," published in November 2000 by McGraw-Hill. A Global Equity Strate-gist for Ned Davis Research, Tim and his team have developed numerous U.S. andglobal asset allocation indicators and models in recent years, while also develop-ing global market and sector ranking systems and indicators based on 18 marketsectors in 16 countries.


The crash of the Hong Kong stock market in October 1997, with itsobvious parallels to similar events in the U.S. in 1987 and 1929, once againraises the specter of October as a dark and ominous month for stocks. Is itmerely a coincidence that these three crashes all occurred in October? Isthere a timing pattern among autumn panics useful to market participants?This article expands upon the observation, originally contained in Chapter1 of the author’s book, The Spiral Calendar1, outlining the correlation be-tween the lunar calendar and the stock market panics of 1929 and 1987.This paper examines how the 1997 Hong Kong panic conforms to thatearlier model, as well as examines the great autumn panics of the 19thcentury. Finally, a look at the peculiar international character of panics,and its implications for the possible causes of these panics.

DEFINITION OF TERMS

Panic. The focus of this article is on short-term equity market panics.The crashes of 1929 and 1987 are the obvious examples. I define thesepanics as one-to-three day, free-fall drops of approximately 20% in themajor averages. The term “panic” is preferred over “crash” as the defini-tion of panic stresses the suddenness and irrationality of the event. Panicswere originally ascribed to the god Pan simply because there were no ob-vious fundamental causes for their occurrence.

Collapse. Collapse is used to signify the larger macro market declinelasting weeks or months within which the panic occurs. An example wouldbe the Hong Kong panic of October 1997, occurring within the larger Asian

equity and currency collapse that ran from July 1997 to January 1998.Annual Lunar Calendar. The annual lunar calendar used here is based

on the Babylonian calendar, which was the model for the later Jewish cal-endar. This annual lunar calendar labels the date of the first new moonfollowing the spring equinox as month one, day one; or 1-1. The followingdate is 1-2. The date of the second new moon after the spring equinox is 2-1, etc. The difficulty with annual lunar calendars, and one of the reasonsfor their abandonment, is that the solar year does not have an even numberof months. Thus, some years in an annual lunar calendar have 12 months,others 13. For our purposes, which focus on the Autumn months, this issueis inconsequential. All calculations use Eastern Standard Time to deter-mine the dates of the lunar phases.

In 1992, this author demonstrated how the panic dates of “Black Tues-day,” October 29, 1929, and “Black Monday,” October 19, 1987 occurredon the same annual lunar calendar date, 7-28. Additionally, the other simi-lar points in the comparisons of those two years, the spring lows, summerhighs and autumn failure highs all occurred within one day on the lunarcalendar. Figure 1 shows those years in a chart aligned with the lunar cal-endar, where similar lunar dates are juxtaposed above each other. The pan-ics are marked with arrows. The other similar features are denoted withdashed lines. The chart also includes Hong Kong’s Hang Seng Index forthe panic year 1997.

These price moves are extraordinarily large over a very short period oftime. Are these panics the largest such declines, or do we selectively re-member the October panics and forget those of other months? A scan ofdaily data of the Dow Jones Industrial Average from 1915, the Hang SengIndex from 1980, The Japanese Nikkei Index from 1950, and the GermanDAX Index from 1960 for the 10 largest, single-day percentage drops isshown in table 1. Seven of those ten declines were days associated withone of the three panics. Two of the others, the spring 1989 declines in theHong Kong market, were tied to a fundamental news event, the Tiananmencrisis in China. The final entry is from the German market during the “mini-crash” of October 1989, an October event similar to the others, but smallerin magnitude. The point to stress here is that in their breadth and ferocity,these panics lie outside the boundaries of normal price action.

There are no other comparable one-to-three day declines of this magni-tude in the data. They represent the very largest percentage drops in thedatabase. This is not normal market behavior. What else ties these eventstogether? The panics occupy virtually identical positions on the annuallunar calendar.

Table 1Largest 1 day % decline

1 26-Oct-87 -33.33% Hang P2 18-Oct-87 -22.61% DJIA P3 5-Jun-89 -21.75% Hang4 20-Oct-87 -14.90% Nikkei P5 28-Oct-97 -13.70% Hang P6 28-Oct-29 -12.82% DJIA P7 16-Oct-89 -12.81% DAX8 29-Oct-29 -11.73% DJIA P9 19-Oct-87 -11.12% Hang P

10 22-May-89 -10.78% Hang

AUTUMN PANICS:A Calendar Phenomenon

Christopher Carolan


Figure 2Autumn Panics, Lunar Aligned


Table 2Autumn Panic 1 day % change

Lunar Month 7 7 7 7 (8)

Lunar Day 27 28 29 30 (1)

DJIA 1929 -12.8% -11.7% 12.3% 5.8%

DJIA 1987 Closed -22.6% 5.8% 10.1%

Hang Seng 1997 -5.8% -13.7% 18.8% -3.7%

Table 2 shows the percentage declines for each panic in the key four-day time span around the lows. The lunar dates 7-27 and 7-28 are the “darkdays,” encompassing the various Black Tuesdays of N.Y. in 1929 and HongKong in 1997, and the Black and Blue Mondays in N.Y. in 1987 and 1997respectively. In each case, lunar date 7-28 marked the end of the panic andthe next two days, 7-29 and 7-30 (or 8-1, some lunar months have 29 days,others 30) saw significant retracement rallies. Table 3 groups the data intotwo-day segments and includes the percentage of these retracement rallies.This table shows the striking similarity of these panics and how that simi-larity conforms to the annual lunar calendar.

Table 3Autumn Panic 2 day % change

Lunar Month 7 7 % retrace

Lunar Days 27-28 29-30

DJIA 1929 -23.6% 18.9% 62%

DJIA 1987 -22.6% 16.6% 57%

Hang Seng 1997 -18.7% 14.4% 61%

Table 4Spike Low-New Moon Differential

Panic Low EST 8-1 New Moon Diff in Hrs.

1929 29-Oct 14:45 1-Nov 8:00 -65

1987 20-Oct 11:30 22-Oct 13:26 -50

1997 29-Oct 9:15 31-Oct 6:01 -45

Table 4 pinpoints the precise timing of the panic lows on the lunar cal-endar. The timing from 1929 is gathered from the news accounts that de-scribed stock prices as rallying sharply off their lows in the last fifteenminutes of trading on Black Tuesday, October 29. The 1987 and 1997 timesare from available databases for the Dow Industrials and are corrected toEastern Standard Time. The table also shows the date and time of the near-est lunar phase, the eighth new moon on the annual lunar calendar, as wellas the difference in hours between the stock market’s low and the moon’sphase. The timing of these three great panic lows is within twenty-fourhours of each other. In other words, all three lows fall within the same one-half of one percent of the calendar year.

A REVIEW OF THE PRE-1915 AUTUMN PANICS

The Panic of 1907

The so-called panic of 1907 does not fit our short-term panic criteria.There was no market decline of approximately 20% in the span of one-to-three days. The largest, singleday declines were 3% in the Dow Jones In-dustrial Average during the collapse. There was a collapse and coincidentbanking panics, most of which occurred in October of that year. Sobel, inPanic on Wall Street2, describes the ending of the collapse. J.P Morgan puttogether his plan to save the banking system on November 3-5, 1907, 7-28through 7-30 on the annual lunar calendar. After being closed for ElectionDay on November 5 (7-30), stocks rallied strongly on lunar 8-1. The crisiswas over. The timing of the end of the crisis is consistent with the lunarpanic model. The day Morgan realized the banking system was not goingto fail, he put into motion a plan to save the banks, which ultimately ar-

rested the decline. That day was lunar 7-28, the same date as the lows ofthe later 20th century panics.

The Crash of 1873

September 18 and 19, 1873 were labeled “Black Thursday” and “BlackFriday” in the collapse of 1873. The Friday selling took prices of majorstocks 5 to 25% percent below Thursday’s already collapsed levels. Thispanic was considered the greatest on Wall Street until 1929. The news ac-counts describe the same type of free fall and despair as the 20th centurycounterparts. The annual lunar calendar dates of “Black Thursday” and“Black Friday” were 6-27 and 6-28, one month earlier, but exactly thesame lunar days as the 20th century examples. News accounts describe atemporary bottom late on Friday. Saturday, September 20 brought renewedselling and the closure of the exchange after a shortened two-hour tradingday. The stock exchange remained closed for a week thereafter. Though onMonday September 22 prices rose sharply in trading in the streets. Thetiming of the 1873 Autumn panic is consistent with the 20th century re-sults, though exactly one month earlier.

The Crash of 1857

The collapse of 1857 was not a stock market free-fall in the sense of the20th century panics outlined above. It was a very sharp drop in stocks overa period of nine weeks, accompanied by a number of runs on banks, persis-tent pressure on the banking system, and sharply rising interest rates. Also,it was international in scope, a facet we’ll address later. Though the sellingin the equity markets did not climax in a free-fall panic, the pressure on thebanking system did, as the N.Y. banking panic broke out on October 13and mayhem continued for two days thereafter. Sobel, in the Panic on Wall

Street3, quotes George Strong writing on October 15. “Wall Street blue

with collapse. Everything flaccid like a defunct Actina.” On the annuallunar calendar, October 13 and 14, 1857 are 7-27 and 7-28, the same “darkdays” as the 20th century examples.

CAUSATION

The correlation between the annual lunar calendar and the timing of the three20th century panics as well as the supportive data from the 19th century doesnot prove that an annual lunar calendar position is the cause of those panics. Afew examples of anything cannot statistically prove a hypothesis. However, itshould be realized that each occurrence is not a 50-50, or true/false proposition.If the Hong Kong panic had occurred on any of the 360 days of 1997 other thanlunar 7-27, 7-28, 6-27, or 6-28; then this model would be effectively discred-ited. Yet the 1997 Hong Kong panic climaxed 5 hours after the timing of the1987 panic and 20 hours after the 1929 panic on the lunar calendar.

Previous theories explaining panics have not fared well when the nextpanic came along. In the 19th century, it was widely believed that panicsoccurred in October specifically because banks’ cash positions were weak-ened as farmers were paid for the new crop. Yet today, agriculture makesup a much smaller fraction of the world economy than before, yet Octoberpanics are still with us. The Federal Reserve System was set up in thebelief that if banking panics were prevented, stock market panics wouldcease to exist as well. That causal theory was disproved by the 1929 crash.The 1929 panic was blamed on low margin levels, yet 1987 happened any-way. In 1987, the finger was pointed at program trading. However, the1997 panic occurred without any appreciable role by program traders.

The lunar calendar model of panics, alone among theories, not onlysurvived the next panic intact, but its basic tenet was remarkably affirmedby the precise timing of the 1997 low.

The timings of financial collapses do not show a pattern. The 1997Asian collapse began in July, while the crisis of 1987 and the collapse of1857 began in August. The 1929 and 1873 examples began in September.Yet in each case, the start of the collapse did not result in immediate wide-


spread panic. Those panics seem to wait for a particular time period on thecalendar, the 27th and 28th days of the autumn lunar months, usually Oc-tober, but in one instance September.

THE INTERNATIONAL QUESTION

The international character of financial crises has been a difficult prob-lem for those who have sought to ascribe causes to collapses and panics.Kindleberger, in Manias, Panics and Crashes writes, “Time and again,

observers like Juggler, Mitchell and Morgenstern have observed that fi-

nancial crises tend to be international, either running parallel from coun-

try to country or spreading by one means or another from the country where

they originate to other countries.4” And “What is remarkable is that secu-

rities prices do the same even when only a few securities can be said to be

truly international, that is, are traded on several markets, their prices joined

by arbitrage. In 1929 all stock markets crashed simultaneously; the same

was largely true in October 1987...It is striking that share prices behaved

in parallel almost sixty years apart, even though share prices were thought

not to have been integrated in the 1920s as they were in the 1980s.5”

The panics of 1987 and 1997 highlighted the international quality ofpanics. Traders the world over saw these markets dive and then rally inunison. In this wired world, that interconnection is not so extraordinary,though Kindleberger is surprised by the international nature of the 1929collapse. An examination of the 1857 collapse is more revealing.Kindleberger notes, “What is striking is the concentrated nature of the

crises...Clapham observes that it broke out almost at the same moment in

the United States, England, and Central Europe, and was felt in South

America, South Africa and the Far East.6” Aside from the internationalnature of the macro collapse, the 1857 collapse affords a unique, controlleddatabase of market behavior in the “dark days” of lunar 7-27 and 7-28 ontwo continents. In 1857, the Atlantic cable linked America with Englandby telegraph. In the early days of the collapse, the telegraph cable failedand all communication was done by ship for the remainder of the crisis.The London Times and The New York Times from the period leading up toand through the N.Y. banking panic provide striking evidence of two mar-kets in distress. Wall Street began its rally from the depths of the collapseon October 13, 1857 (lunar 7-27) at the same time the banking panic brokeout in N.Y. Table 5 is reprinted from The New York Times of October 14,illustrating the sharp rise in prices underway as contrasted with the lows ofOctober 13. I’ve added the column on the right showing the month’s-endprices. Some issues had made their lows earlier in September, but otherswere at or near their lows on October 13. What’s clear is that prices beganto rally from their depressed levels on lunar 7-27, coincident with the out-break of the banking panic. This sequence parallels the 1987 experience,when U.S. bond prices began a sharp rally from their lows on lunar 7-27,coincident with the outbreak of the stock panic.

Table 5N.Y. Stocks, October 1857

Aug. 22 Low Since Oct. 13 Oct. 14 Oct. 31

N.Y Central 77 50 58 61 65

Galena 86 52 55 62 64

Rock Island 90 53 58 64 66

Delaware 115 77 78 81 94

Panama 90 60 65 67 72

Reading 67 24 33 36 30

N.Y States Sixes 112 90 90 90 97

Missouris 78 59 60 67 68

Virginias 90 66 67 81 78

Ill. Cen. Bonds 98 50 51 59 68

Erie’s 1883 80 50 50 50

Erie Share 28 7 8 11 13

Source N.Y. T imes, Oct 14. fractions omitted

At this same time, Europe was aware of, and sharing in, the collapse inAmerica. In the week leading up to October 13, the bank of England raisedtheir discount rate twice, while Paris, Hamburg and Amsterdam each raisedtheir rates once. Though debt and equity prices traded down sharply, therewas no free-fall panic. London stocks and debt bottomed decisively onOctober 13 at the beginning of the trading day. The London Times of No-vember 2, 1857 summarized the events of October and printed the table ofprices labeled here as Table 6. To that table is added the date of the month’slow for each security. Here is the commentary accompanying the table.“The range of Consols (government debt) has been unusually extensive,showing a difference of 4 percent between the highest and lowest prices,although at the conclusion (of the month) the market has returned to theprecise position in which it stood at the commencement...In railway sharesthe fluctuations have also been violent, and the rebound, except in a fewcases has not been equal to that of the funds.”

The London Times offered this account of the trading in debt on Octo-ber 13 in its October 14 edition. “The fluctuations in the funds today (Oct.

13) have again been most rapid and extensive. The market opened with a

great weakness at a fall of nearly one and a quarter percent from the heavy

prices of last evening. But there was subsequently a considerable reaction

and a more healthful tone became apparent in all departments of busi-

ness.” Now, here’s the account of stock trading on October 13 from The

New York Times of October 14. “The stock market this afternoon advanced

from 1 to 3 percent, the conviction being general that the basis of business

would be changed tomorrow and that a large amount of money held in

abeyance since the panic first paralyzed confidence will be set free now

that the worst is known...”


Table 6London Stocks, October 1857Sep. 30 Oct. High Oct. Low Oct. 31 Low Date

British Debt

Consols 90.25 90.5 86.5 90.5 13-Oct

Railway Shares

Brighton 102.5 103.5 100 103 13-Oct

Caledonian 85 86 76.5 83.5 29-Oct

Eastern Counties 57.5 58.5 51.5 54.5 13-Oct

Great Northern 98 98.5 92 93.5 13-Oct

Great Western 54.5 55.25 50.5 51 29-Oct

London & NW 97.5 98.5 93.5 96.75 13-Oct

Midland 83 83.25 79.5 83.25 14-Oct

Lancashire & Yorkshire 95.5 97.25 91.5 93.25 13-Oct

North Staffordshire 13.25 13.75 12.75 13.75 15-Oct

South Eastern 65.5 66.25 61.5 64 14-Oct

South Western 90.5 91 87.5 89.5 13-Oct

North Eastern Berwick 92.5 93.5 88.5 91.5 13-Oct

D.tto York 73.5 80 75.5 78.5 13-Oct

Source: London Times

The cause of the market low in New York on October 13 is ascribed tothe banking panic, yet London bottomed on the same day. The selling,motivated by fear, was pervasive on both side of the Atlantic leading up toOctober 13. That selling ceased and a vigorous rally commenced on thesame day, continents apart, with neither market having access to any timelyinformation from the other. Word of the N. Y. banking panic did not reachLondon until October 26, and was then reported in The Times the follow-ing day.

The sudden, international cessation of distress selling that is a hallmarkof 20th century panics also occurred in the crisis of 1857, at a time when

no timely communication existed. The international character of panics hasbeen a stumbling block to those who subscribe to local, “fundamental”causes for these panics. Contrarily, a lunar-based model for panics wouldseem to require an international manifestation of the phenomenon. If themoon is affecting market participants, it should affect them the world over.All the panic examples cited here, from 1857 through 1997, have beeninternational, yet the dearth of communication technology in 1857 pro-vides a datum that cannot be explained as a serial reaction. The interna-tional character of panics is distinctively supportive of the lunar model.

USES

Put simply, every market participant should have his calendar markedwith the “dark days” of lunar 7-27 and 7-28. Even better, everyone shouldcalculate the time of the eighth new moon and subtract 55 hours from thatpoint. A time window of plus or minus twelve hours from that point is thelunar calendar model for an Autumn panic’s low point. There may not beanother October stock panic for sixty years or longer. And the lunar modeloffers no clues as to which years will see a panic. Yet there can be nodoubt, as the trillions of dollars lost during these panics makes plain, mar-ket intelligence that can pinpoint when an unfolding panic will climax isinvaluable. In 1997, as worldwide markets became unglued in October, thelunar calendar model provided by 1987 and 1929 pointed to late Monday,October 27 as the ideal low point. The dramatic early Tuesday morninglow of October 28 demonstrated the model’s effectiveness in real time.

Calendars are complex mechanisms. Calendar research must recognizethe importance of both lunar and solar calendars. The annual lunar modelfor panics points to the 27th and 28th days of the lunar month as the darkdays, yet that is only true in the autumn season, the 6th or 7th lunar month.Past studies that purport to find no lunar relationship in markets have treatedall lunar phases alike, lumping spring and fall together as well as summerand winter. Likewise, specific seasonal analysis tends to ignore the con-current lunar calendar. Those who dismiss that October may be a roughmonth for stocks cite that overall, it is not the worst month for stocks sta-tistically, falling on average .5% since 1915. A proper approach to calen-dar research suggest that distinctions should be made among Octobers basedon the lunar calendar. Here’s the lunar distinction. When there is no fullmoon between October 3 and 19 inclusive, the Dow has been up 1.5% inOctober since 1915. In those years with a full moon between those dates,the Dow’s average change is a loss of 1.9%. Seasonal analysis should rec-ognize the lunar distinctions and vice versa. The annual lunar calendarmakes those distinctions. When autumn panics are viewed though its prism,the results are remarkable.

FOOTNOTES

1 Carolan. The Spiral Calendar. New Classics Library. 1992

2 Sobel. Panic On Wall Street. Dutton. 1988. pp 318-320

3 ibid. p 106

4 Kindleberger. Manias Panics and Crashes. Basic Books. 1989. p 131

5 ibid. p 131

6 ibid. p. 143

© 1998 Christopher CarolanCalendar Research Inc.PO Box 41, Gainesville GA 30503 USA770/718-0032, [email protected]


SUMMARY

■ The Leuthold Group has been compiling corporate insiders big blocktransactions since 1982 in order to gauge the sentiment of the “smartmoney.”

...”Big Block” transactions is defined as those involving more than

100,000 shares or have a total transaction value greater than

$1,000,000.

■ Corporate insiders have been selling at levels approaching historicalextremes recently. The short-term outlook for the stock market is be-ginning to look negative by this measure.

...Increased levels partially due to SEC’s 1997 code revision, which

shortens the holding period of restricted shares. Additionally, a lower

long-tenn maximum capital gains tax rate (20%) was effected in ’97

and has likely resulted in increased selling by insiders.

■ Since 1983, when net selling measured in dollars has reached histori-cally high levels, the stock market performed poorly over the next 12months.

...Normalizing the data allows a better historical perspective by ad-

justing for the growth of the stock market over time. When normalized,

the current high level of dollar volume of net selling, while still high, is

significantly below levels of 1983, 1989 and the selling extremes of

mid-1998.

■ When net selling (measured in dollars) reaches historically low levels,the stock market has demonstrated significant above average perfor-mance over the next 12 months.

. . .The 10 week moving average is particularly useful to signal bear

market bottoms. This measure has signaled “net buying” from corpo-

rate insiders a total of three times in the last 15 years; all were within

weeks of bear market bottoms.

■ The number of net buy/sell transactions is also currently near historichighs. Even when the data is normalized, the number of selling trans-actions have been increasing since 1991.

■ Conclusion: Quantitatively testing the normalized historical data forinsiders net transaction levels measured in dollars confirms that whenhistorically high or low extreme levels are hit, they offer excellent trig-ger points for asset allocators and market timers.

INTRODUCTION

The efficient market hypothesis holds that the market discounts infor-mation as soon as it is made public. While the degree of efficiency in theU.S. stock market is debatable, most would agree that corporate insiderspossess superior knowledge about their own company’s prospects for thefuture. Clearly insider trading laws prohibit using “material, non-publicinformation” for financial gain, but hunches about the success or failure ofa new product line, for instance, can come into play when a corporate in-sider decides whether to accumulate or sell company stock.

Consequently, monitoring the significant buying and selling of com-pany insiders should lend some insight concerning a firm’s financial healthand growth prospects...insight that might not be gleaned from the latest

CORPORATE INSIDERS’Big Block Transactions

Eric Bjorgen and Steve Leuthold

CHARLES H. DOW AWARD WINNERS • MAY 1999

quarterly statements. It then follows, that since the stock market is thesum of all public firms, the aggregate buying and selling patterns of allinsiders from these firms should lend insight about the future prospects ofthe stock market. This study examines the merits of this assumption.

Since 1982, the Leuthold Group has been tracking corporate insiders

big block transactions on a weekly basis. It is one of the components thatwe use in the weekly Major Trend Index’s “Sentiment” category. We arethe only research firm that we are aware of that compiles this type of data.The SEC makes information on insider’s transactions available weekly.By law, all corporate insiders (and beneficial owners who hold 10% ormore of outstanding shares) are required to file Form 4 by the l0th day ofthe month following a transaction. Each week the latest filings are com-piled and published in Vicker’s Weekly Insider. This is where we begin.

COMPILING THE INSIDERS BIG BLOCK DATA

For our purposes of gauging insiders’ sentiment, we ignore small trans-actions – focusing only on big block transactions involving buying/sellingmore than 100,000 shares or those transaction with a total value of $1 mil-lion or more. We ignore the transactions of corporations, foundations, trustsand other institutional shareholders, since these transactions are often mo-tivated by factors that have nothing to do with the financial prospects of acompany.

The resulting list of buy and sell transactions are logged and thensummed up to yield a weekly aggregate net dollar amount of buys vs.sells. Most of the time the net selling is much greater than the net buying,but this is not always the case. We then tabulate the aggregate net num-

ber (frequency) of buys vs. sells meeting the above criteria. An exampleof a recent week’s list of qualifying transactions appears in the appendix toillustrate the individual transactions that we look at in deriving a weeklyreading.

Chart 1Net Weekly Amount & 10-Week Moving Average


PART A: THE RAW DATA

The Chart 1 depicts the weekly net dollar amount of insiders’ buyingand selling activity (vertical bars) and the 10-week average (lower line) vs.the S&P 500 on semilog scale. Note: on this chart and all that follow,

points above the zero axis represent net selling and the very infrequent

points below the axis represent net buying.

Observations

■ Over the course of the last sixteen years, the weekly data (representedby the bars) shows that weeks of net selling outnumber weeks of netbuying by about 12:1. This is primarily because insiders’ sell transac-tions include the sale of stock resulting from the exercise of options,although no corresponding buy transaction occurs when options areissued. In the last seven years, the sell/buy ratio has climbed evenhigher (to about 50:1), partly due to increasing use of options to com-pensate corporate insiders.

■ The 10-week average (lower line) passes below the zero axis into netbuying territory on only three occasions within the last 16 years (markedwith arrows). Each time this occurred, a bear market bottom had oc-curred within a short period, sometimes within weeks. Intermittentweeks of strong net buying and low net selling accompanied these bearmarket lows, which has made this measure an excellent “buy” signal atthe lows of the three significant down markets since 1983.

■ Since the late 1990 signal, the 10-week average has not been in netbuying territory, but there haven’t been any bear markets either. Theclosest the average came to net insider buying was in early 1995 (indi-cated by the dashed arrow). This four-year low in net selling could nothave been better timed, coming right at the end of an 18 month consoli-dating market.

■ On the sell side, this non-normalized series is more difficult to inter-pret. Because the collective wisdom of all corporate insiders is thoughtto be a forward-looking stock market indicator, high levels of sellingshould theoretically precede stock market corrections. However, asoption issuance and market capitalization has increased, the 10-weekaverage and weekly data has shown a strong tendency to drift upward.Fixed “sell” trigger levels that worked ten years ago are commonplacetoday. Normalizing the data avoids this problem (see part B of thisstudy).

■ For now, let it suffice to observe the market’s behavior subsequent tosharp upward “spikes” in the 10-week average. As indicated on thechart, a rapid increase in insider net selling often precedes or coincides

with market weakness and price volatility, but generally the signalshave been 6 to 12 months early for the “major” corrections of 1987 and1990. For example, insider net selling peaked in March 1987, but themarket did not crash until October. Also, the spike in the fall of 1989was immediately followed by a choppy market (which many investorswould have been happy to avoid), but the S&P 500 did not collapseuntil August 1990. A more comprehensive look at this indicator as asell signal appears in part B where this series is normalized.

■ The latest reading through April 7, 1999 shows the 10-week average is

now in a rising trend, but is still 33% below the all time high recorded inlate May 1998. Also note that the week or 3/24/99 net aggregate insider

selling posted the second highest single week reading ever ($2.8 billion).

While the dollar amount measures the magnitude of insider’s transac-tions, the net number of transactions measures the breadth of net sells/buys. Normally the two data series move together (e.g. when the dollaramount of net selling rises, the number of net “sells” also rise). But this isnot always the case. Occasionally the weekly net dollar amount of netselling surges, but the net number of sell transactions remains flat. Thisindicates that there were one or more very large transactions during thatparticular week. For instance:

...In July 1989, weekly dollar volume soared to $2.1 billion while netnumber of sells actually fell from the preceding week. This was theresult of insider Carl Icahn liquidating $1.3 billion worth of Texacoshares, a company for which he had served as an officer.

...During a weekly reporting period in May 1995, two significant insid-ers at Duracell sold big blocks of shares, accounting for 90% of thesoaring $1.5 billion volume that week. But the number of net insidersales that week was down from the levels of previous weeks.

...In March of 1998, Bill Gates and Paul Allen sold shares of Microsofttotaling over $1.6 billion during a two week period, accounting for abouthalf of the all-time record dollar volume of insiders sales reported thatweek ($3.2 billion). But unlike the previous two cases, net number ofsells also hit a record high. What are insiders transactions revealingabout the stock market when you have high levels of conviction (indi-cated by record dollar amount) and broad consensus (record number ofnet sells)?

...Historically it has meant that a market peak may soon be at hand.

The current 10-week moving average is also in a rising trend. In terms of

weekly number of net sells, the all time high was nearly beat on the weekending March 24th, when a net 406 individual corporate insiders were

Chart 2Weekly Number Of Net Buy /Sell Transactions & 10 Week Averages

Chart 3Net Weekly Dollar Amount & 10-Week Average

(Normalized By Total Stock Market Capitalization)


selling big blocks of their company’s stock (previous 1998 record still stands

at 410 net sells).

...Because non-normalized data has a tendency to drift upward as mar-ket capitalization and total number of traded issues increases, it is nec-essary to normalize the data if any comparisons are to be made betweentoday and sixteen years ago. In part B we normalize the dollar volumedata in order to provide a better historical perspective.

PART B: THE NORMALIZED DATA

Chart 3 is similar to the first chart we presented except that the data isnormalized as a percentage total equity market capitalization. This willallow more meaningful comparison over time and keep the tendency forthe data to drift upward held in check.

■ On the buy side, normalization does not significantly change the origi-nal conclusions. No amount of normalization will change net selling tonet buying. Since late 1990, the 10-week average has not returned intonet buying (positive) territory, but it still came fairly close in early 1995.

■ However the net selling peaks of the 1980s now prominently stand out.In fact, when normalized, the current nominal reading for the 10-weekaverage now falls short of peaks recorded in 1983, 1989, 1993 and 1997.But the rapid increase of net selling so far in 1999 is approaching his-torical extremes through the most recent reading. The normalized data

still indicates that the recent high levels of insiders’ big block selling may

have negative implications for the stock market.

■ Since 1983, upward surges in insiders net selling appear to precedeperiods of market weakness, however the chart above shows that theresults are not entirely consistent. In 1983, 1987, 1993 and late 1989(1990 bear market), surges in net selling did foreshadow intermediatecorrections that were quickly followed by a longer term rallying mar-ket. But the three other peaks that occurred during the 1990s weresoon followed by periods of market consolidation. The most recent setof sell signals occurred during Q2 of 1998, and provided a timely exitsignal for the market declines that occurred the following quarter.

■ The dashed horizontal lines on the chart at -.07% (seven hundredths ofa percent) and -.01% (one hundredth of a percent) represent points atwhich selling reached historically high and low extremes. Since 1983,the 10-week average has moved outside this range only about 14% ofthe time.

■ The normalized data on the chart seems to provide some strong evi-dence (at least visually) between high levels of net selling (above the.07% line) and subsequent inferior market performance. On the otherside of the coin, the few instances of insider net buying have demon-strated a record of signaling bear market lows. Even when net sellingfalls to historically low levels (below the dashed .01% line) it was agood time to start buying stocks.

This study wouldn’t be complete without some quantitative evidence tosupport the visual evidence. In Part C we test the relationship between highand low levels of net selling and subsequent market performance.

PART C: TESTING THE DATA

The table below shows subsequent market performance for differentlevels of dollar volume of insider net selling. We show the price perfor-mance of the S&P 500 in 3, 6, 9 and 12 month time periods when theinsider selling 10-week average reaches historical extremes on a normal-ized basis (refer to previous chart). As the dashed lines on the normalized10-week average dollar chart indicates, extremes are signaled when net

selling falls below one-hundredth of a percent of total market capitaliza-tion (bullish) and when net selling reaches seven-hundredths of a percentof total market capitalization (bearish). The 10-week average is now within10% of the bearish line and in a rising trend. Only time will tell where themarket goes from here, but the current high level of net selling gives risefor concern about what insiders are collectively revealing about the out-look for the stock market in the coming year.

When Insider’s Net Selling... The Stock Market’s Average Return Has Been....

3 Month 6 Month 9 Month 12 Month

Reaches Historically High Levels -0.6% 3.5% 7.0%* 0.3%*

Is Within Historically Normal Range 3.7% 7.2% 10.7% 14.8%

Reaches Historically Low Levels 6.0% 10.4% 13.5% 17.7%

* Returns For 1998 Signal Pending

■ Performance subsequent to high levels of net selling was below “nor-mal range” in all time frames. Average 3-month subsequent perfor-mance when in “bearish” range was a loss of 0.6% compared to 3.7%average gain when in “normal” range.

■ At low net selling levels, subsequent market performance better than“normal range” in all time frames. Outperformance optimized at 3-month horizon, but consistently outperforms by 280-320 b.p.’s overlonger time horizons.

■ This study covers 1983 to date covering roughly 848 weeks. Duringthis time, the normalized 10-week average dollar amount has spent atotal of 74 weeks in the historically low selling (bullish) range or 8.7%of the time. The average has spent 48 weeks in the historically highselling (bearish) range, representing 5.7% of the time. Following in-siders buying and selling cues over this sixteen-year time span wouldhave been profitable for the market timer or asset allocator.

■ Recent periods of high net selling include the mid 1998 signals thatoccurred as the market was in decline. Not enough time has elapsed toevaluate the longer-term merits of these latest readings, but the pre-liminary data shows the mid 1998 signals were followed by below av-erage performance in the 3-month range, but not over the longer timeframes. But what occurs in the next 3-6 months will determine if theselatest signals were indeed as productive as they have been in the past.

PART D: CONCLUSIONS

■ The 10-week average of dollar volume of net insider sells seems towork extremely well in identifying bear market bottoms. When net sell-ing has hit historical lows; it has identified the market bottoms of 1984,1987 and 1990 within several weeks.

■ The 10-week dollar average also has a proven track record of signalingperiods of impending market weakness when insider net selling reacheshistorically high levels (like now).

. . . Normalizing the data to reflect the growth of the stock market overtime is necessary to identify the key selling triggers on a historical ba-sis.

■ Testing the data confirms the value of tracking trends of insiders’ ag-gregate buying/selling behavior. Using the historical extremes as ac-tion points to increase or decrease equity holdings has worked well.

■ Net dollar volume and net number of transactions often move together,but occasionally an unusually large single transaction can swell dollarvolume without causing the number of net sells to rise. When thisoccurs, the number of net sells becomes significant as a confirmingindicator.

■ Our measure of insider buys/sells is indicating that since mid-1997, insid-

These two extremes can be viewed as action signals. Do someselling at .07% above, and be a buyer at .01% and below.


ers have been selling their stock at historically high levels in both nominaland normalized terms. Based on the historical relationship between levelsof net selling and subsequent market performance, insiders may be sig-naling that the road ahead for the stock market may be rocky in the com-

ing year.

. . . However, part of the increase in insider net selling is due to recentchanges in SEC restrictions about the length of holding periods for re-stricted shares. This has encouraged insiders to sell more freely thanbefore. Additionally, reductions in the maximum capital gains rate mayhave resulted in increased selling by corporate insiders. At this point, itis difficult to tell how much this is contributing to current high levels ofnet selling.

Written May 1998



INTRODUCTION

In the 1960s and 1970s, as the ability to use computers became morewidespread, a number of experiments were performed on stock market andcorporate data to determine the best variables required to select stocks.These experiments were crude by today’s standards, but in their innocence,these analysts discovered many truths and dispelled many myths. For ex-ample, one experiment with Price/Earnings ratios (P/E) suggested, ironi-cally, that contrary to beliefs held even today, the best P/E was a high one,not a low one – that stocks with high P/E’s tended to outperform those withlow P/Es1. Experiments like these allowed analysts to focus on the value ofa number of variables that heretofore had been too complicated or timeconsuming to pursue. At that time, when the random walk theory, betatheory, efficient market hypothesis (EMH) and capital asset pricing model(CAPM) were gaining in popularity, Robert A. Levy published a book2 andan article in the Journal of Finance3 based on his Ph.D. thesis at AmericanUniversity4 that showed how well–performing stocks, i.e. those with rela-tive price strength, continued to perform well and that poorly performingstocks likewise continued to perform poorly. Levy’s theory was not origi-nal. The theory of relative price strength had been around for a long time.5

However, Levy, with the new aid of computer power, added some nuancesand calculations that had not previously been used and found them to bevery successful. Since they tended to refute the then popular theory thatthe stock price action was entirely random (the Efficient Market Hypoth-esis), his conclusions were subject to considerable criticism; his calcula-tions and statistical evidence were severely condemned; and finally, hisresults were left in the dust of academic vitriol.6 Though long forgottennow by most analysts, his theories nevertheless have been kept alive by afew. For almost twenty years a model run in real time (‘live’) and pub-lished every week for 17-1/2 years, based largely on these theories, hasshow his calculations to have been and continue to be useful in selectingstocks with higher–than–market post performance.

THE TEST

Most computer-ized experiments andstock market modelsare calculated usingwhat is called “opti-mization.”7 In the at-tempt to find vari-ables that are impor-tant in determiningthe future of stockprices, most experi-ments use past dataand adjust the vari-

ables and their parameters to find a ‘fit’ between those variables and stockpost-performance. This is called “forced optimization.” Most discussionthen centers around how closely the results fit the data, how sophisticatedthe statistical methods were, and why the results occurred as they did, for-getting that the results may have no usefulness in the future. Some com-

puter ized- t radingmodel builders avoidforced optimizationby spliting their datainto several parts.They perform theirexperiments on oneor more parts, andthen test the resultsagainst the otherparts. However, thebest and most con-

vincing test of any theory is to see if it works by itself using completelyunknown data. This is what this study accomplished weekly over 17-1/2years.

In July 1982, to test variables of relative price strength and relativeearnings growth, a selection and deletion criteria was established, a perfor-mance measurement determined, and a stock list developed (“List 1”). Later,in 1999, a second list (“List 2”) was established using slightly differentcriteria. Each list was reported weekly in Kirkpatrick’s Institutional Mar-

ket Strategist8, and periodically performance results were also reported. Asof December 31, 2000, List 1 had appreciated 5086.6% versus a S&P 500gain of 1087.6% and a Value Line Geometric gain of 221.9% (see Chart I).List 2, during a very difficult and slightly declining stock market, appreci-ated 137.3% versus a S&P 500 gain of 7.41% and a Value Line Geometricloss of 9.99% (see Chart II). The second list also outperformed the originallist which gained 75.19% over the same two year period. Most perfor-mances occurred during a generally rising stock market but none includeddividends, which, though small in most cases, would have made the resultseven more impressive. Transaction costs were not included. Today, at radi-cally discounted levels, commissions are almost negligible costs except inhigh turnover models.

SELECTION CRITERIA

The first, and longest existing test list, List 1, included relative pricestrength, relative earnings growth, and a simple chart pattern as variablesfor stock selection. The selection criteria for List 2 were slightly different.Relative price strength and earnings growth were used but instead of achart pattern, relative price-to-sales ratio (“PSR”) was included to reducethe risk of loss.

The reason for the change in criteria between the second and first testlist was that with the general market having risen since 1982 and the strongstocks having become so volatile, the danger existed that a severe correc-tion would exert even more downward pressure on the list’s performance.For example, in the bear markets of the 1960s and 1970s, relative pricestrength performed well as a selection criteria initially until the very end ofthe general market declines when the strongest stocks tended to decline thesharpest and suffered disproportionately large losses.

To prevent such a loss in an individual stock, in List 1 a simple chartpattern was imposed as a ‘stop’ on negative price action to forcefully de-lete a stock early and prevent it from being caught in a severe decline.

STOCK SELECTION:A Test of Relative Stock Values Reported over 17-1/2 Years

Charles D. Kirkpatrick II, CMT


However, later, through tests of stock price patterns alone, no discernibleadvantage was gained.9 Therefore, to avoid changing the original selectioncriteria for List 1 and thereby interrupting its long record of success, List 2was begun using another approach. Rather than have a price stop to mini-mize loss, the danger of negative performance was minimized in the be-ginning by selecting only those stocks trading at low relative price to sales.Presumably these stocks were trading at bargain prices already. As it turnedout, three additional advantages arose from this model: (1) portfolio vola-tility declined rapidly – portfolio beta was consistently below one, whereasList 1 often had a portfolio beta approaching two, (2) turnover declinedfrom an average holding period of 22 weeks in List 1 to well over a year inList 2, and (3) the size of the portfolio was considerably smaller and moremanageable – near 5 to 15 stocks in List 2 versus up to 80 in List 1.

PERFORMANCE MEASUREMENT

Each week, before making changes to a list, the average percentagegain or loss of each stock in the list was recorded. For example, say the listincluded only stock A and stock B. If stock A was up 5% and stock B up1%, the list was recorded as having risen 3%, the mean of the two stockperformances. These list performances were then accumulated each weekover the test period. The equivalent in the real world would be for a portfo-lio manager to equally invest in selected stocks one week, record its com-bined performance for the next week, and then readjust each stock as wellas add new ones and delete old ones such that for the coming week theportfolio would be equally weighted in each stock. Otherwise, the strongerstocks would accumulate over time into a larger relative position in theportfolio and have an unequal effect upon the portfolio’s total performance.Equally weighting of each stock each week, was the best method to reli-ably measure the criteria used in selecting the stocks.

SPECIFIC SELECTION CRITERIA

Relative Price Strength

Most measures of relative strength weigh a stock’s performance againsta market average or index such as the S&P 500. This is wrong. The addi-tion of a market average only complicates the results. For example, marketaverages are capital–weighted; individual stocks are not. Furthermore, thiskind of measurement makes it difficult to weigh one stock against another,difficult to tell when price strength is changing, difficult to determine com-parative periods, and is difficult to quantify for model building. The bestcalculation for a stock’s relative strength is to measure price performanceequally against all other stocks over some specific time period. Until thearrival of computer power, this kind of calculation was very difficult andtime-consuming. By the 1960s it was not.

Several methods of quantitatively weighing price performance have beenproposed.10 More recently, and since List 1 was begun, for example,Jegadeesh and Titman (1993) used six and twelve month returns held forsix months during the period 1965-1989. Their results demonstrated a post–performance excess return of 12.01%. This evidence tends to confirm Levy’searlier work. However, it was not available when the test model was be-gun. Instead, both List 1 and List 2 used a derivation of Levy’s originalcalculations.

Levy originally calculated the ratio of a stock’s 131-day moving aver-age to its latest price. This ratio was calculated for all stocks. The total listof ratios was then sorted. Each stock was allocated a relative price strengthpercentile between 99 and 0 based on where its ratio fell in the spectrum ofratios. The 0 percentile for the highest ratio (weakest stock) and the 99thpercentile for the lowest ratio (strongest stock).

To make the ratio easier to calculate and to understand, the test listschanged several aspects of Levy’s calculation but not the essence. Rather

than using the ratio of the moving average to the current price, the inversewas used. The ratio of current price to the moving average made the highpercentiles represent the highest relative strength. Thus the 99th percentilerepresented the strongest stock and the 0 percentile represented the weak-est. Second, instead of 131 days of data in the moving average, the testlists used 26 weeks, approximately the same period (131 trading days is26.2 weeks, not including holidays). In this manner, a large amount of datawas not necessary (131 data points per stock versus 26 data points), yet theresulting ratios were equivalent and the effects on the post-performanceminimal. The closing price used each week was the Thursday close.

Relative Earnings Growth

Until this point, it would appear that the study was involved solely withtechnical analysis and price behavior. However, while technical analysishas its weak and strong points, a stock selection method must use all vari-ables that appear to work. Relative earnings growth is one of them.

To a certain extent, “earnings” are a manufactured statistic. They de-pend on many accounting tricks and are not always truthful measures of acompany’s success or failure. Special charges are often later written offagainst earnings, and depreciation is recalculated, or taxes reassessed. Re-ported earnings, therefore, are often subject to controversy and exaggera-tion.9 No one can argue that a stock closed at a certain price (at least withinsome small bound), but analysts often disagree on exactly what a company’sactual earnings may be. This becomes even more complicated when earn-ings are estimated into the future.11 However, earnings reports are watched,especially for surprises, and are acted upon by investors. Tests have shownthat reported relative earnings growth has a positive correlation to the post-performance of a stock.12 Part of this, of course, is because reported earn-ings include any earnings surprise.

To be as sensitive as possible without the effect of seasonality, Levycalculated earnings growth by taking the most recent five quarters of re-ported earnings and measuring the ratio between the latest four quarterstotal to the first four quarters total.13 Thus three quarters overlapped, andthe seasonal tendency of many quarterly reports was eliminated. The ratio,if positive, showed that earnings were growing and by how much and ifnegative that earnings growth was negative and by how much. When com-panies reported losses for any consecutive four quarters, the ratio was notcalculated. Growth then was measured over a relatively short period offive quarters. This same calculation was used in determining the earningsgrowth criteria for both test lists. As with relative price strength, the ratiofor each stock was ranked with the same ratio for all other stocks and apercentile ranking determined whereby those stocks with the highest earn-ings growth were ranked in the highest percentiles, and vice versa for thosewith the lowest earnings growth.

Chart Pattern

As mentioned above, the test required some means to reduce the risk ofan individual stock’s failure. List 1, which was the only one to use a chartpattern, by its nature, was very volatile and its selected stocks very high inprice and valuation. This is only natural when stocks with high relativestrength and high earnings growth are selected. To reduce the danger ofindividual collapse, the use of a simple chart pattern was thought to be thebest method at the time to eliminate those stocks that begin to decline se-verely and before they collapsed.

Computerizing chart patterns, especially twenty years ago, was and stillis a difficult problem.14 The simplest method was to produce a simple point-and-figure chart, one that shows only price reversal points after a predefinedprice magnitude has occurred. To do this, only the magnitude of the pricemove was needed to determine the reversal point. As an example, in manypoint-and-figure charts, a three point reversal magnitude is required for a


price reversal point. If a stock price rises from 50 to 56, then declines to48, since the three points up and down have been met, the reversal pointwas 56, the highest point at which the stock price had risen by at least 3points and reversed by 3 points. This would be called an “upper reversalpoint” since it marked a top in prices. Had the stock only risen to 52 beforedeclining to 48, no reversal point would have been recorded since the stockhad not risen from 50 by the required magnitude of 3 needed to establish areversal point. Conversely, had the stock then declined to 48 and risenback to 55, the price of 48 would have been a “lower reversal point” sincethe stock had declined by at least 3 into 48 and then risen more than therequired 3 immediately afterward. This combined behavior would then haveleft us with a history of an upper reversal point at 56 and a lower reversalpoint at 48. In the chart formula, the last two upper and lower reversalpoints were recorded each week. When prices rose above two upper rever-sal points, the chart was said to be “advancing,” and prices declined belowtwo lower reversal points, the chart was said to be “declining.”

In addition, in the chart formula, a sliding scale of reversal magnitudeswas established to minimize the effect of absolute price differences. Forexample, a 3-point reversal in a 100 dollar stock is less significant than a3-point reversal in a 20 dollar stock. A sliding scale of reversal magnitudesequalized the requirements for a reversal among all stocks.

Rather than be concerned about the actual patterns of the reversal points,List 1 only used the reversal points themselves. Only those advancing stockswere considered for selection, and those stocks in the list that turned downbelow two lower reversal points were eliminated. This provided the ‘stop’needed to protect the portfolio from extraordinary negative events.

Relative Price/Sales Ratio (“PSR”)

Prices are well-known and easily accepted as valid. Annual sales of acompany are also well-known and easily accepted as valid, and when com-bined with prices are an excellent comparative measure of a stock’s value.The higher the price-to-sales ratio, the higher the valuation that investorshave placed on the stock’s future, and also the higher the risk of failure.Lower PSRs suggest lower value placed on a stock’s future. Their advan-tage is that “a small improvement in profit margins can bring a lot to thebottom line, improving the firm’s future P/E. Low PSR stocks are held inlow regard by Wall Street. Those with improving profit margins usuallycatch the Street by surprise.”15 PSRs also include stocks with no earnings(and therefore no P/E). Many studies have shown the value of the PSR.16

O’Shaughnessy (1998) argues that the PSR is the most reliable method ofselecting stock for long term appreciation.17 His method of using the PSR,however, requires that an arbitrary level be established, below which astock is attractive. In List 2 the arbitrary level was disbanded in favor of arelative percentile. First, the ratio was calculated for each stock as the cur-rent weekly close price divided by the last reported four–quarters sales.Next, this ratio for all stocks was then sorted and divided into percentilessuch that the highest was in the 99th percentile and the lowest in the 0percentile. This way, despite the general market level of valuation, a stock’sPSR could be measured against the PSR of all other stocks at the sametime and in the same investment environment

COMBINING CRITERIA INTO MODEL – THE PARAMETERS

Each week the entire list of available U.S. stocks (usually around 5,000)was screened for those stocks at or above the 90th percentile in relativeprice and earnings growth. In List 1 an advancing chart pattern was alsorequired. Any stock not already on the list that met these criteria was addedto the list. When relative price strength declined to or below the 30th per-centile, relative earnings growth declined to or below the top 80th percen-tile, or the stock price pattern broke two previous lower reversal points, the

stock was eliminated from the list. In List 2, the chart pattern was not used,but relative PSR was. The requirement for addition to the list was a rela-tive PSR at or below the 30th percentile. The deletion criteria in List 2were the same as in List 1 except they did not include the relative PSRsince a high level did not necessarily suggest that a stock was facing animpending decline. Additionally, the deletion requirement for relative earn-ings growth was reduced to or below the 50th percentile since earlier expe-rience had shown that a high threshold deleted stocks prematurely.

SPECIFIC RESULTS

Chart III shows the performance of List 1, the S&P 500 and the ValueLine Geometric each year since the inception of the study in 1982. ChartIV shows the more recent total history for List 2 versus List 1, the S&P 500and the Value Line Geometric. List 1, which began its weekly live trial inJuly 1982, gained a total of 5086.6% over the 17-1/2 years versus a 1087.6%gain in the S&P 500 and a 221.9% gain in the Value Line Geometric. Thisgain was 4.37 times the gain in the S&P and 16.11 times the performanceof the Value Line Geometric. During that 17-1/2 year period List 1 hadonly three down years versus three for the S&P 500 and seven for the ValueLine Geometric (see Table A below).

List-2, which began it weekly live trial in January 1999, has had onlytwo years of history to measure. Nevertheless, the results so far have beenimpressive. Over the two years, the list gained 137.3% versus only a 7.41%gain in the S&P 500 and a 9.99% loss in the Value Line Geometric. It hadno down years versus one for the S&P and both for the Value Line, and asmention earlier, its beta and turnover were considerably lower than List-1.

Table APerformance Comparison

List–1, S&P 500, Value Line Geometric & List–2

List–1 S&P VLG List–2

1982 49.1% 26.5% 29.2%

1983 57.6 17.3 22.4

1984 –11.4 0.8 –8.3

1985 33.3 26.1 19.2

1986 20.9 17.8 8.1

1987 11.9 0.1 –11.8

1988 22.6 13.1 14.8

1989 26.5 25.5 10.8

1990 –12.1 –6.4 –24.0

1991 76.8 23.3 23.3

1992 19.4 7.6 11.0

1993 25.7 7.6 10.4

1994 –1.6 –1.6 –6.3

1995 54.5 33.2 19.3

1996 24.5 23.1 13.8

1997 8.2 23.4 17.2

1998 6.5 31.8 –0.5

1999 62.4 19.1 –2.6 59.8%

2000 7.9 –9.8 –7.6 48.5

CONCLUSION

The quantitative analysis of stock selection criteria has diverged in manydirections since the relatively recent widespread use of the computer. Mostanalysis has centered on demonstrating the validity of one or more specificstock market theories and many have shown mediocre results. The methodof testing these results has also fallen into the optimization trap whereby


the “best fit” between data and performance was not tested with new dataand especially with unknown future data.

This study took several variables that had been demonstrated to havevalue in stock selection and in one list, beginning in July 1982, tested theresults “live” each week for 17-1/2 years. The test was done through simu-lating the performance of a hypothetical portfolio, thus adding an elementof practicality not seen in most studies of stock selection, and used a com-bination of technical and fundamental factors without prejudice. These fac-tors measured aspects of a company or its stock on a basis relative to allother stocks and were independent of general market averages except inthe demonstration of performance. The results were exceptionally favor-able for the methods used and demonstrated the usefulness of the variablesemployed. Relative price strength and relative reported earnings growth,when calculated in the manner of this study, showed superior results whencompared to market averages. Since the period over which the study wasdone was one of generally rising stock prices, the final test will be com-pleted only after a major stock market decline. However, considering thelong period over which the study was conducted without adjustment formarket changes, the presumption is that the relative post-performance re-sults of the methods used will continue to exceed average market returns.

FOOTNOTES

1 Avanian and Wubbels (1983)

2 Levy (1968b) – lengthy and doesn’t add much more than Levy (1967)

3 Levy (1967) – This article caused quite a stir in academia because it wasthe first major attempt to refute the efficient market hypothesis.

4 Levy (1966)

5 Bernard (1984), the founder of Value Line, as an example, had success-fully utilized the concepts of relative price strength and relative earningsgrowth since the late 1920s. “dividing the stock’s latest 10–week averagerelative performance by its 52-week average relative price” is the pricemomentum factor used by Value Line. For a recent discussion of the meritsof the Value Line system see Choi (2000).

6 The reaction to Levy’s (1967) article was swift. Michael Jensen (1967) ofHarvard was the first to publish comments. Initially he criticized Levy’smethodology on the basis that the sample was too small and over too shorta period, had a selection bias, and other errors that would overstate theresults. His comment was that Levy’s comment of “the theory of randomwalks has been refuted” was a little too strong. Levy (1968c) then coun-tered with another study including more stocks and a longer time periodthat produced even better results (31% versus the market 10% for 625stocks from July 1, 1962 to November 25, 1966). Finally Jensen andBennington (1970) did their own study, supposedly using Levy’s rules butincluding transaction costs and adjustments for risk, and using 1962 stocksfrom 1926 to 1966, and reported that Levy’s rules resulted in a risk–ad-justed loss. We never hear of Levy’s relative strength work again.

7 See Murphy (1986) and Kaufman (1978)

8 Kirkpatrick (1978-2001)

9 Merrill (1977)

10 The entire concept of past price returns having an effect on future pricereturns has academia in quandary since it tends to cast severe doubt onthe efficient market hypothesis. Many different price return anomalies havebeen reported, some positive and some negative. Long–term and very short–term results tend to be consistently negative. Chopra, Lokonishok, and Ritter(1992), Cutler, Poterba and Summers (1988), De Bondt and Thaler (1985),and Fama and French (1986) show that for holding periods beyond 3 years,the return is negative. Over periods of a month or less, French and Roll(1986) and Lehmann (1990) found negative returns in individual stocksweekly and daily; Lo and MacKinlay (1990) found positive returns weeklyin indices and portfolios but negative returns for individual stocks; andRosenburg, Reid and Lanstein (1985) found negative reversals after a

month. There seems, however, to be a window of about six to twelve monthswhen returns are consistently positive. This was Levy’s hypothesis and ithas now been confirmed by Brush (1983, 1986) and Jegadeesh and Titman(1993). BARRA [see Buckley (1994)] has found the price momentumanomaly in a number of countries, including the US, Japan, the UK, Aus-tralia, and France. Explanations for these anomalies are varied but bestsummed in Chan, Jegadeesh and Lokonishok (1996, 1999).

11 The question of how accurate are reported earnings and especially howaccurate are future earnings forecasts has been widely studied. Niederhoffer(1972) and Cragg and Malkiel (1968) suggest that reported earnings arebetter forecasters of future earnings than analysts forecasts. Indeed, Har-ris (1999) concludes that analyst forecasting accuracy is extremely poor,biased and inefficient. The inaccuracy is mostly the result of random errorand the performance of forecasts vary with both the company characteris-tics and the forecast itself. A whole series of studies has evolved around“earnings surprises” those frequent events when reported earnings differmarkedly from analysts’ expectations. La Porta (1996) has shown that su-perior results can be gained by exploiting these analyst errors becauseexpectations are too extreme. Investors overweight the past and extrapo-late too far into the future. Chan, Jegadeesh and Lokonishok (1996, 1999)speculate that the reason for the relative strength positive anomaly oversix to twelve months is that it takes that long for the analysts to adjust. LaPorta (1996) suggests that it takes several years.

12 Ramakrishnam and Thomas (1998)

13 Levy and Kripotos (1968a)

14 The best and most recent discussion about analyzing chart patterns is inLo, Mamaysky, Wang, and Jegadeesh (2000).

15 Fisher (1996)

16 A number of financial ratios have been used and tested. The most common,of course, is the price-to-earnings ratio (PER). More recently the market–to–book ratio has become popular, and even more recently attention hasreturned to the price-to-sales ratio (PSR). Senchack and Martin (1987)had shown that low PSR stocks tended to outperform high PSR stocks butthat low PER stocks dominated low PSR stocks on both an absolute andrisk-adjusted basis. But recently Barbee (1995) showed in tests from 1979to 1991 that price-to-sales and debt-to-equity had greater explanatorypower for stock returns than did either market-to-book or market-to-eq-uity. Liao (1995) also showed that low PSR stocks avoid the ambiguities ofthe CAPM approach and dominate high PSR stocks and the market.

17 O’Shaughnessy (1998) also argues for relative price strength as a selec-tion criterion.

REFERENCES

■ Avanian, Alice C. and Rolf E. Wubbels, 1983, Shaking up a corner-stone? Study raises questions on price–earnings ratio importance, Pen-

sions & Investment Age v11n8, 21.

■ Barbee, William C., Sandip Mukherji, and Gary A. Raines, 1996, DoSales–Price and Debt–Equity Explain Stock Returns Better than Book–Market and Firm Size? Financial Analysts Journal v52n2, 56–60.

■ Bernard, Arnold, 1984, How to Use the Value Line Investment Survey:

A Subscriber’s Guide (Value Line, New York).

■ Brush, John S., 1986, Eight relative strength models compared, Jour-

nal of Portfolio Management v13n1, 21–28.

■ Brush, John S., 1983, The predictive power in relative strength & CAPM,Journal of Portfolio Management v9n4, 20–23.

■ Buckley, Ian, 1994, The past is myself, Pensions Management v26v11,93–95.

■ Chan, Louis K. C., Narasimhan Jegadeesh, and Josef Lokonishok, 1999,The profitability of momentum strategies, Financial Analysts Journal

v55n6, 80–90.


■ Chan, Louis K. C., Narasimhan Jegadeesh, and Josef Lokonishok, 1996,

Momentum strategies, Journal of Finanace v51n5, 1681–1713.

■ Choi, James J., 2000, The value line enigma: The sum of known parts?

Journal of Financial & Quantitative Analysis v35n3, 485–498.

■ Chopra, Navin, Josef Lokonishok, and Jay R. Ritter, 1992, Performance

measurement methodology and the question of whether stocks overre-act, Journal of Financial Economics v31, 235–268.

■ Cragg, J. G and Burton G, Malkiel, 1968, The consensus and accuracyof some predictions of the growth of corporate earnings, Journal of

Finance v23n1, 67–84.

■ Cutler, D. M., J. M. Poterba, and L.H. Summers, 1991, Speculativedynamics, Review of Economic Studies v58, 529–546.

■ De Bondt, W. F. M., and R. H. Thaler, 1985, Does the stock marketoverreact?, Journal of Finance v40, 793–805.

■ Fama, E. F., and K. R. French, 1986, Permanent and temporary compo-nents of stock prices, Journal of Political Economy v98, 246–274.

■ Fisher, Kenneth L., 1996, PSRs revisited, Forbes v158n6, 225.

■ French, Kenneth R., and Richard Roll, 1986, Stock return variances:The arrival of information and the reaction of traders, Journal of Fi-

nancial Economics v17, 5–26.

■ Harris, Richard D. F., (1999), The accuracy, bias and efficiency of ana-lysts’ long run earnings growth forecasts, Journal of Business Finance

& Accounting v26n5/6, 725–755.

■ Jensen, Michael C. and George Bennington, 1970, Random walks and

technical theories: Some additional evidence, Journal of Finance v25,469–482.

■ Jensen, MichaelC., 1967, Random walks: Reality or Myth – Comment,Financial Analysts Journal 77–85.

■ Jegadeesh, Narasimhan, and Sheridan Titman, 1993, Returns to buyingwinners and selling losers: Implications for stock market efficiency,Journal of Finance v48n1, 65–91.

■ Jegadeesh, Narasimhan, 1990, Evidence of predictable behavior of se-curity returns, Journal of Finance v45n3, 881–898.

■ Kaufman, Perry J., 1978, Commodity Trading Systems and Methods,John Wiley & Sons, New York.

■ Kirkpatrick, Charles D., II, Kirkpatrick’s Market Strategist (Post Of-fice Box 699, Chatham, MA 02633)

■ LaPorta, Rafael, 1996, Expectations and the cross-section of stock re-turns, Journal of Finance v51n5, 1715–1742.

■ Lehmann, Bruce N., 1990, Fads, martingales, and market efficiency,Quarterly Journal of Economics 105, 1–28.

■ Levy, Robert A. and Speros L. Kripotos, 1968, Earnings growth, P/E’s,and relative price strength, Financial Analysts Journal

■ Levy, Robert A., 1968, The Relative Strength Concept of Common Stock

Forecasting: An Evaluation of Selected Applications of Stock Market

Timing Techniques, Trading Tactics, and Trend Analysis, Investors In-telligence, Larchmont, New York, 318p., illus.

■ Levy, Robert A., 1968, Random Walks: Reality or Myth – Reply, Fi-

nancial Analysts Journal, 129–132.

■ Levy, Robert A., 1967, Relative strength as a criterion for investmentselection, Journal of Finance V22, 595–610.

■ Levy, Robert A., 1966, An Evaluation of Selected Applications of Stock

Market Timing Techniques, Trading Tactics and Trend Analysis, Un-published Ph.D. dissertation, The American University.

■ Liao, Tung Liang, and others, 1995, Testing PSR filters with the sto-

chastic dominance approach, Journal of Portfolio Management v21n3,85–91.

■ Lo, Andrew W., Harry Mamaysky, Jiang Wang, and NarasimhanJegadeesh, 2000, Foundations of technical analysis: Computational al-gorithms, statistical inference, and empirical implementation/discus-sion, Journal of Finance v55n4, 1705–1770.

■ Lo, Andrew W., and Craig A. MacKinlay, 1990, When are contrarian

profits due to stock market overreaction? Review of Financial Studies

v3, 175–206

■ Merrill, Arthur A., 1977, Filtered Waves, Basic Theory: A Tool for Stock

Market Analysts, Analysis Press, Chappaqua, New York.

■ Murphy, John J., 1986, Technical Analysis of the Futures Markets, In-stitute of Finance, New York.

■ Niederhoffer, Victor, and Patrick J. Regan, 1972, Earnings changes,analysts’ forecasts and stock prices, Financial Analysts Journal, 65–71.

■ O’Shaughnessy, James P., 1998, What Works on Wall Street: A Guide to

the Best–Performing Investment Strategies of All Time, McGraw–Hill,New York

■ Poterba J. M., and L. H. Summers, 1988, Mean reversion in stock prices:

Evidence and implications, Journal of Financial Economics v22, 27–59.

■ Ramakrishnam, Ram T. S. and Jacob K. Thomas, 1998, Valuation ofpermanent, transitory, and price irrelevant components of reported earn-ings, Journal of Accounting, Auditing, & Finance v13n3, 301–336.

■ Rosenburg, B., K. Reid, and R. Lanstein, 1985, Persuasive evidence ofmarket inefficiency, Journal of Portfolio Management v11, 9–16.

■ Senchack, A. J., Jr. and John D. Martin, 1987, The relative performanceof the PSR and PER investment strategies, Financial Analysts Journal

v43n2, 46–56.

Charles D. Kirkpatrick II, CMT (AB, Harvard; MBA, Wharton) is the editor ofthe Journal of Technical Analysis and the Academic Liaison for the MTA. For funand to keep current with the markets, he still publishes his 'Market Strategist' letter,which continues to include the selected stock lists mentioned in his Dow Award ar-ticle on relative price strength. In the Fall of 2002 he will be teaching technicalanalysis at Fort Lewis College School of Business in Durango, Colorado where herecently moved to be near his grandchildren and their family.


I view the art of technical analysis and research as an exciting adven-ture. I have often wondered what “Sedge” Coppock was looking for whenhe invented the Coppock Curve. What was Edson Gould researching whenhe discovered the precepts for his “Sign of the Bull”? From my own re-search, I have learned that serendipity, the aptitude for making desirablediscoveries by accident, can play a big part in making meaningful techni-cal discoveries. One of the fantasies that every serious stock market tech-nician has probably entertained is that there must be some kind of indica-tor that will signal us when a major market top is being formed. There aresome effective indicators for identifying market bottoms, but because mar-ket tops tend to be more diffuse, often occurring at different times for dif-ferent indexes, the search for an effective tool to identify major markettops has been, for the most part, a futile one.

In November 1992, I was struck by the apparent lack of volatility in thedaily number of advancing and declining issues on the New York Exchange.Over a period of 21 trading days (the number of trading days in the averagemonth), the highest single day closing advance/decline ratio (simply di-vide up stocks by down stocks on the New York Exchange) was 1.84 andthe lowest was 0.71. At the time, it seemed that was a very small range fora full month of data, so I decided to research further. Rather than use theobserved 1.84 and 0.71 limits as a precedent for further research, the rangewas arbitrarily widened somewhat to 0.65 and 1.95. The first search of thecomputer database attempted to find other time periods of 21 consecutivetrading days when similar “churning” occurred, i.e. when the highest dailyadvance/decline ratio was below 1.95 and the lowest advance/decline ratiowas above 0.65. That might give a clue as to whether the pattern was sig-nificant in any way. The initial research went back to 1966 when the Dowmade its first move towards the 1000 level. The results were stunning.Between 1966 and November 1992 when the pattern first caught my atten-tion, a period of almost 27 years, there were only three other periods whenthe conditions for the pattern were satisfied. Here are the dates when thoseconditions were fulfilled:

January 25, 26, 27, 28 1966

October 17, 18, 21, 22, 24, 25 1968

December 6, 7, 8, 11 1972

Chart 1

The chart above depicts the resolution of the 3 initial patternsdiscovered that ultimately were named The “Sign of the Bear”

It appeared as if technical gold had been struck. Within an average pe-riod of less than a month, the pattern had preceded three of the most impor-tant stock market tops of the past several decades. Equally as important,there were no other instances of the pattern over that 27-year period. Justhow important were the turning points that were preceded by this pattern?

The first day of the 1966 pattern preceded a final market top on theDow Jones Industrial Average on February 9th by 11 trading days. On aninflation-weighted Dow chart, the 1966 top lasted until 1995 as an all-timeDow high.

The first day of the 1968 pattern preceded a final market top on theDow on December 3rd by 27 Trading days. That Dow Jones IndustrialAverage top also corresponded with a major top on the Value Line Com-posite Index, an unweighted index more representative of the averageinvestor’s portfolio. That index would go on to lose approximately 75% ofits value over the next six years.

The first day of the 1972 pattern preceded a final market top on January11th, 1973 by 23 trading days. That top led to one of the sharpest two-yearDow declines in history, almost 50 percent in less than 24 months. Thehigh seen on January 11, 1973 would not be reached again until almost adecade later.

After reviewing those results, the feeling was that something very spe-cial had been discovered. During a period of almost 27 years, there wereonly three occurrences of the pattern and each occurrence led to a majormarket top within, at most, 27 trading days.

Over three decades of market research have made it clear that any pat-tern that appears to have predictive potential should be researched as farback as is practicable. Research of the period between 1940 to 1966 un-covered a total of nine “churning” patterns when the above conditions weresatisfied, namely, the highest daily advance/decline ratio over a 21 dayperiod was below 1.95 while the lowest ratio over that period was above0.65. With the exception of the period from June 1963 to March 1965, theresults were impressive, though not as uniformly dramatic as the post-1965results noted above. Here are all the similar periods noted from 1940 through1965 where the 21 day (or longer) churning pattern occurred, followed bythe number of days in the pattern.

December 30, 31,1952, January 1 1953 (23)

September 22, 23 1955 (22)

May 3, 6, 7, 8, 9, 10, 13 1957 (27)

December 8, 11, 12, 13 1961 (24)

June 7 1963 (21)

February 28 through April 23 1964 (59)

February 4,5 1965 (22) March 16-25 1965 (28)

May 11, 12, 13, 14 1965 (24)

For now, let’s discard the period between June 1963 and February 1965and observe the average results for the remaining six periods. The DowJones Industrial Average, on average, advanced 1.7 % from the close of the21st day in the pattern to the highest subsequent intra-day high after thepattern emerged. On average, it took 10 trading days to reach that high,and the subsequent decline averaged 15.2%. It would be convenient tosomehow eliminate the three instances of the pattern between June 1963and February 1965. We could say that the June 1963 and the February


SIGN OF THE BEAR

Peter G. Eliades


1965 instances barely qualified because they had the fewest number ofconsecutive days (21 and 22, respectively), and that February through April1964 was completely out of character with the other instances because itwas more than double the number of consecutive days of any other in-stance, and eliminate those instances from our examples. But in a strictsense, it would not have been a true reflection of technical history. In anyevent, even those apparent instances of failure were followed by almostimmediate market declines. Those declines, however, were of a minormagnitude.

Overall, these results were deemed to be significant and impressive. Ifthe research had ended there, there would have been sufficient evidence toidentify the pattern as one that closely preceded market tops during theperiod from 1940-1973.

A market historian might note a remarkable commonality in all of theabove periods. Almost without exception, each time the churning param-eters were satisfied over a minimum of 21 trading days, the market waseither at or very close to an all time high or a multi-year high. There isnothing apparent in the definition of the two limits required of the ad-vance/ decline ratio (greater than 0.65 but less than 1.95) over a one monthperiod that would suggest such a result.

As noted initially, the characteristics of this pattern were first noticedin November 1992. The specific pattern which was unfolding then went onfor 48 consecutive trading days from November 9th to December 17th,1992. Between 1992 and 1998, four more instances of the pattern occurred.The updated record of instances of the pattern between 1972 and 1998reads as follows:

November 9 to December 17 1992 (48)

August 20 to September 3 1993 (31)

January 18 to February 3 1994 (33)

April 25 to May 2 1995 (26)

September 12 to September 15 1995 (24)

By 1995, it became obvious that if the pattern was a signpost, a kind offootprint that preceded important market tops, the defining characteristicsof the pattern would have to be refined. The purpose of the refinementswould ideally be to arrive at a tool that was effective in identifying majormarket tops. At the same time, it was important to attempt to avoid thepractice of curve fitting. Going back to the original three patterns that werediscovered, dating between 1966 and 1972, I tried to identify characteris-tics that distinguished those three patterns that worked so very well fromthe patterns that were either apparent failures or patterns that marked onlyminor reversals.

One of the items that appeared significant was the length of the patternbefore the consecutive streak is broken. Intuition would suggest that thelonger the pattern, the greater its potential negative influence. History hasproved otherwise. Once a pattern moves beyond 27-28 market days, it hasfar less a chance of being significant. The 1992 pattern can be eliminatedbecause it was far too long and it did not fit the profile of prior patternswhich saw the Dow going to either multi-year highs or all time highs as thepattern reached 21 days in length. It emerged at a time when the Dow wasmore than five months beyond and almost 6% lower than its previous alltime or multi-year high. It just did not fit into the profile of prior churningpatterns. The April-May 1995 pattern and the September 1995 pattern ap-peared at first glance to qualify as patterns that led to intermediate or long-term tops in the past. As successful predictive patterns of the past werefurther examined, however, a signature that accompanied all the success-ful pattern predictions of major market tops began to become apparent.One important consideration was how the pattern ended. In other words,when the churning streak ended, did it end with a high ratio (above 1.95)

or a low ratio (below 0.65)? The initial three patterns that were discoveredfrom 1966 to 1972, and that worked so remarkably well in identifying majortops, all ended their consecutive streaks with low ratios. In fact, in each ofthose three instances, the end in the streak was conclusive. Either the twoday average advance/decline ratio or the three day average advance/de-cline ratio following the end of the streak was below 0.75. In other words,after at least a full month without one advance/decline ratio below 0.65,there is a distinct and sharp change in the market’s personality. Not only isthere a day with a ratio below 0.65 that breaks the consecutive streak, butfor at least two or three days after the streak ends, there is an average ratiobelow 0.75.

Enough data had now been compiled to formulate a general rule. Thepattern would be dubbed, the “Sign of the Bear.” There were three basicrules that were required to identify a “Sign of the Bear.”

1. There must be a streak of 21-27 consecutive trading days where thedaily advance/decline ratio remains above 0.65 but below 1.95.

2. That consecutive streak must end with a downside break, i.e. with anadvance/decline ratio below 0.65.

3. The downside break in the streak must be confirmed with either a twoday average advance/decline ratio or a three day average advance/de-cline ratio following the end of the streak being below 0.75.

Just around the time the basis for these rules had been formulated, theadvance/ decline data for the period prior to 1940 became available. Thedata were examined with trepidation and with great anticipation. Remem-ber, after the initial discovery of the pattern, the follow-up research wentonly back to 1940, the limit of our database at the time. There was not oneinstance of the pattern from 1940 until December 1952. Looking at thenewly acquired data, I was again stunned by the results. There was not oneinstance of the pattern in the decade of the 1930s, just as there had beennone in the 1940s. Working backwards from December 1952, take a wildguess when the first appearance of the pattern occurred. The dates wereJuly 19 and 20, 1929. That’s right! Just over six weeks before the mostfamous top of the 20th century, the pattern occurred and the three basicrules were met. A “Sign of the Bear” had appeared. It would not be seenagain until December 1961, over 32 years later. Surely, this appearance ofthe pattern in a completely different time period would do away with anynotion that there was “data mining” or curve fitting performed on the ini-tial data that formed the basis of the research.

What is the rationale that explains why the patterns defining a “Sign ofthe Bear” should result in major market tops? I believe that searching for21 day periods without one daily ratio less than 0.65 would obviously di-rect the computer to periods of market strength, periods where the marketwent for at least a full month without a big down day. At the same time, thecomputer is directed to periods of investor complacency – a full monthwithout a meaningful day of selling. Now add the requirement that therealso be no daily ratio higher than 1.95 and the computer would be directedto periods of market strength and bullish sentiment, but not the kind ofupside breadth (advance/decline ratios higher than 2-1) which is usuallyrequired to sustain a healthy market advance. Voila! It’s just the combina-tion that a technician might look for at a market top. The final requirementis one that almost all technicians learn sooner or later. Require confirma-tion of your pattern. Unless there is a sharp turnaround to the downside asrequired by rules number 2 and 3, the pattern might be relatively innocu-ous. Once that confirmation occurs, history tells us the market is in trouble.

The final challenge was to test the theory in real time. There were sev-eral patterns in the 1990s, but until 1998, they all failed to meet the threerequirements necessary for a “Sign of the Bear.” Finally, on April 6, 1998,the 3 requirements were satisfied for the first time in almost 26 years. As


this paragraph is being written in 2000, we know that April 1998 proved tobe not only a major top for the daily advance/decline line of the NY Ex-change, it was also the all-time high on the Value Line Composite Index(Geometric). That high has not been approached to this day, even thoughits sister index, the Value Line Composite Index (Arithmetic) has sincegone to new all-time highs. It has long been our contention that the geo-metric Value Line is superior to the arithmetic one in giving a true pictureof the average share of stock. Chart 2 shows the daily advance/decline lineof the New York Stock Exchange with an arrow pointing to the April 1998“Sign of the Bear” signal. Once again, in real time, the “Sign of the Bear”gave a virtually perfect signal for a change in the market’s personality.

Chart 2

This chart depicts the April 1998 “Sign of the Bear” signal in relation tothe Daily Advance/Decline Line of the New York Stock Exchange

On September 18, 2000, another “Sign of the Bear” signal was con-firmed. It was only the second signal since December 1972. It came just 5trading days from a new all-time high on the N. Y. Composite Index. Therehave now been two signals generated within 30 months of each other. Theclosest previous signals were the ones generated in January 1966 and De-cember 1968, thirty-three months apart. There are not enough results tomake a statistically informed judgment, but there is a suggestion from theprior signals that the “Sign of the Bear” is an indication of not merely apotential major top, but perhaps also an important secular change in theoverall market from long term bull to market underperformance for manyyears to come. It is difficult to understand how a simple pattern of only 1-2 months duration could predict the future course of the market for years tocome, but examine the final chart below. The 1929 signal marked a Dowtop that would not be exceeded for over a quarter of a century. The 1961signal preceded a top and then the final run up of only 35% before the 1966top which was not convincingly exceeded for over 16 years. The 1968 top,as was explained earlier, led to a decline of around 75% in the averageshare of stock or mutual fund, and the 1973 top led to one of the sharpestDow declines of the 20th Century, a top that would not be significantlyexceeded until a decade later. How will history judge the latest two sig-nals? The April 1998 signal has already marked an almost 3-year top in thedaily advance/decline line and the Value Line Composite (Geometric) In-dex (Geometric). Only history will tell us whether the September 2000signal will mark a secular market top of historic duration. Based on theprior history of the “Sign of the Bear,” there appears to be an excellentchance it will.

Chart 3

This chart shows all the “Sign of the Bear” signals since the late 1920s.



IDENTIFYING BEAR MARKET BOTTOMS AND NEW BULL MARKETS

Paul F. Desmond

Ask one hundred investors whether this is a bull market or a bear mar-ket, and you are likely to find their opinions split evenly down the middle.No one is really certain that the September 2001 low marked the end of thebear market and the start of a new bull market. But, this uncertainty isnothing new. As long as stock exchanges have existed, analysts and inves-tors have always placed heavy emphasis on the difficult task of identifyingthe primary trend of the stock market. Everyone’s ideal market strategy is,at least in theory, to avoid the ravages of each bear market, and then tomove aggressively into stocks after each important market bottom. Tofurther maximize the benefits of a new bull market, time is of the essence.An investor should buy as close to the final low as possible. This is the‘sweet spot’ for investors – the first few months of a new bull market inwhich so many stocks rise so dramatically. But, theory and reality, espe-cially in the stock market, are often entirely different matters. To bringthis theoretical investment strategy to reality, an investor would need atime-tested method of identifying major market bottoms – as opposed tominor market bottoms – and would have to apply this method quickly, tocapture as much of the bull market as possible. Traditional methods ofspotting major turning points in the market often leave a great deal to bedesired. The financial news typically remains negative for months after anew bull market has begun. The economic indicators offer little help since,historically, the economy does not begin to improve until about six to ninemonths after the stock market has already turned up from its low. Evensome widely accepted technical indicators, such as 200-day moving aver-ages or long-term trendlines, can sometimes take several months to iden-tify a major turning point in the market. To spot an important market bot-tom, almost as it is happening, requires a close examination of the forcesof supply and demand – the buying and selling that takes place during thedecline to the market low, as well as during the subsequent reversal point.

Important market bottoms are preceded by, and result from, importantmarket declines. And, important market declines are, for the most part, astudy in the extremes of human emotion. The intensity of their emotionscan be statistically measured through their purchases and sales. To clarify,as prices initially begin to weaken, investor psychology slowly shifts fromcomplacency to concern, resulting in increased selling and an accelerationof the decline. As prices drop more quickly, and the news becomes morenegative, the psychology shifts from concern to fear. Sooner or later, fearturns to panic, driving prices sharply lower, as investors strive to get out ofthe market at any price. It is this panic stage that drives prices down toextreme discounts – often well below book values – that is needed to setthe stage for the next bull market. Thus, if an investor had a method foridentifying and measuring panic selling, at least half the job of spottingmajor market bottoms would be at hand.

Over the years, a number of market analysts have attempted to definepanic selling (often referred to as a selling climax, or capitulation) in termsof extreme activity, such as unusually active volume, a massive number ofdeclining stocks, or a large number of new lows. But, those definitions donot stand up under critical examination, because panic selling must bemeasured in terms of intensity, rather than just activity. To formulate ourdefinition of panic selling, we reviewed the daily history of both the pricechanges and the volume of trading for every stock traded on the New YorkStock Exchange over a period of 69 years, from 1933 to present. We brokethe volume of trading down into two parts – Upside (buyers) Volume and

Downside (sellers) Volume. We also compiled the full and fractional dol-lars of price change for all NYSE-listed stocks that advanced each day(Points Gained), as well as the full and fractional dollars of price changefor all NYSE-listed stocks that declined each day (Points Lost). Thesefour daily totals – Upside Volume and Points Gained, Downside Volumeand Points Lost – represent the basic components of Demand and Supply,and have been an integral part of the Lowry Analysis since 1938. (Note:an industrious statistician can compile these totals from the NYSE stocktables in each day’s Wall Street Journal.)

In reviewing these numbers, we found that almost all periods of sig-nificant market decline in the past 69 years have contained at least one,and usually more than one, day of panic selling in which Downside Vol-ume equaled 90.0% or more of the total of Upside Volume plus DownsideVolume, and Points Lost equaled 90.0% or more of the total of Points Gainedplus Points Lost. For example, April 3, 2001 qualified as a valid 90%Downside Day. To clarify, the following table was shown in Lowry’s Daily

Market Trend Analysis Report of April 4, 2001:

Daily Upside Downside Points Points +Vol +PointsTotals Volume Volume Gained Lost % %

March 30 964,227,570 508,158,900 1,116 296 65.5 79.0

April 02 383,754,900 1,004,545,180 298 933 27.6 24.2

April 03 146,576,520 1,439,436,850 148 1,447 9.2 9.3

On April 3rd, Downside Volume equaled 90.8% of the sum of Upside plus Downside Volume:

1,439,436,850 / (146,576,520 + 1,439,436,850) x 100 = 90.8%

AND, Points Lost equaled 90.7% of the sum of Points Gained plus Points Lost:

1447 / 148 + 1447) x 100 = 90.7%

The historical record shows that 90% Downside Days do not usuallyoccur as a single incident on the bottom day of an important market de-cline, but typically occur on a number of occasions throughout a majordecline, often spread apart by as much as thirty trading days. For example,there were seven such days during the 1962 decline, six during 1970, four-teen during the 1973-74 bear market, two before the bottom in 1987, seventhroughout the 1990 decline, and three before the lows of 1998. These90% Downside Days are a key part of an eventual market bottom, sincethey show that prices are being deeply discounted, perhaps far beyond ra-tional valuations, and that the desire to sell is being exhausted.

But, there is a second key ingredient to every major market bottom. Itis essential to recognize that days of panic selling cannot, by themselves,produce a market reversal, any more than simply lowering the sale priceon a house will suddenly produce an enthusiastic buyer. As the Law ofSupply and Demand would emphasize, it takes strong Demand, not just areduction in Supply, to cause prices to rise substantially. It does not matterhow much prices are discounted; if investors are not attracted to buy, evenat deeply depressed levels, sellers will eventually be forced to discountprices further still, until Demand is eventually rejuvenated. Thus, our 69-year record shows that declines containing two or more 90% DownsideDays usually persist, on a trend basis, until investors eventually come rush-ing back in to snap up what they perceive to be the bargains of the decadeand, in the process, produce a 90% Upside Day (in which Points Gainedequal 90.0% or more of the sum of Points Gained plus Points Lost, and onwhich Upside Volume equals 90.0% or more of the sum of Upside plus


Downside Volume). These two events – panic selling (one or more 90%Downside Days) and panic buying (a 90% Upside Day, or on rare occa-sions, two back-to-back 80% Upside Days) – produce very powerful prob-abilities that a major trend reversal has begun, and that the market’s SweetSpot is ready to be savored.

Not all of these combination patterns – 90% Down and 90% Up – haveoccurred at major market bottoms. But, by observing the occurrence of90% Days, investors have (1) been able to avoid buying too soon in arapidly declining market, and (2) been able to identify many major turningpoints in their very early stages – usually far faster than with other formsof fundamental or technical trend analysis. Before reviewing the histori-cal record, a number of general observations regarding 90% Days mighthelp to clarify some of the finer appraisal points associated with this veryvaluable reversal indicator:

1. A single, isolated 90% Downside Day does not, by itself, have any longterm trend implications, since they often occur at the end of short termcorrections. But, because they show that investors are in a mood topanic, even an isolated 90% Downside Day should be viewed as animportant warning that more could follow.

2. It usually takes time, and significantly lower prices, for investor psy-chology to reach the panic stage. Therefore, a 90% Downside Day thatoccurs quickly after a market high is most commonly associated with ashort term market correction, although there are some notable excep-tions in the record. This is also true for a single 90% Downside Day(not part of a series) that is triggered by a surprise news announcement.

3. Market declines containing two or more 90% Downside Days oftengenerate a series of additional 90% Downside Days, often spread apartby as much as 30 trading days. Therefore, it should not be assumedthat an investor can successfully ride out such a decline without takingdefensive measures.

4. Impressive, big-volume “snap-back” rallies lasting from two to sevendays commonly follow quickly after 90% Downside Days, and can bevery advantageous for nimble traders. But, as a general rule, longer-term investors should not be in a hurry to buy back into a market con-taining multiple 90% Downside Days, and should probably view snap-back rallies as opportunities to move to a more defensive position.

5. On occasion, back-to-back 80% Upside Days (such as August 1 andAugust 2, 1996) have occurred instead of a single 90% Upside Day tosignal the completion of the major reversal pattern. Back-to-back 80%Upside Days are relatively rare except for these reversals from a majormarket low.

6. In approximately half the cases in the past 69 years, the 90% UpsideDay, or the back-to-back 80% Upside Days, which signaled a majormarket reversal, occurred within five trading days or less of the marketlow. There are, however, a few notable exceptions, such as January 2,1975 or August 2, 1996. As a general rule, the longer it takes for buy-ers to enthusiastically rush in after the market low, the more investorsshould look for other confirmatory evidence of a market reversal.

7. Investors should be wary of upside days on which only one component(Upside Volume or Points Gained) reaches the 90.0% or more level,while the other component falls short of the 90% level. Such rallies areoften short-lived.

8. Back-to-back 90% Upside Days (such as May 31 and June 1, 1988) area relatively rare development, and have usually been registered nearthe beginning of important intermediate and longer term trend rallies.

A detailed Appendix is attached, showing each 90% Day (or back-to-back 80% Upside Days) over the past 40 years, since January 1, 1960.

But, several examples may make it easier to visualize the concepts pre-sented here. The charts to follow show the Dow Jones Industrial Averagein the months before and after a number of major market bottoms. Anoscillator of the intensity of each day’s trading, in terms of both Price andVolume, is also shown on each chart (for simplicity’s sake, the Price andVolume percentages have been combined into a single indicator). The 90%Days, both Downside and Upside, are highlighted with an arrow. The back-to-back 80% Upside Days are highlighted with a dot.

Exhibit 1

During 1962, seven 90% Downside Days were recorded during Mayand June, the last one occurring two days before the final low. The 90%Upside Day was recorded on June 28, just three days after the low in theDow Jones Industrial Average.

Exhibit 2

LOWRY'S 90% DAYS - 1962 (DAILY)

MAR62 APR62 MAY62 JUN62 JUL62 AUG62 AUG 31

520

542

564

586

608

630

652

674

696

718

740

0

90

-90

1962

Dow Jones Industrial Average

Lowry Intensity Oscillator


APR70 MAY70 JUN70 JUL70 AUG70 SEP70 SEP 30

620

640

660

680

700

720

740

760

780

800

820

0

90

-90

1970




Five 90% Downside Days were recorded during the final months of the1969-1970 bear market, the last one occurring one day before the low. The90% Upside Day occurred on May 27, one day after the market low.

Exhibit 3

The final months of the 1973-1974 bear market contained four 90%Downside Days (a total of 14 occurred throughout 1973 and 1974), the lastoccurring on December 2, four days before the final low in the Dow JonesIndustrial Average. Back to back 80% Upside Days occurred on Decem-ber 31, 1994 and January 2, 1995 – an unusually long sixteen days after the1974 market low. Another 90% Upside Day, a superfluous confirmation ofthe new bull market, occurred on January 27, 1975, thirty-three days afterthe bottom day.

Exhibit 4

Three 90% Downside Days were recorded during the final months ofthe 1980 decline. The 90% Upside Day occurred on March 28, one dayafter the market low. Another superfluous 90% Upside Day occurred onApril 22, after a successful test of the lows.

Exhibit 5

In 1987, 90% Downside Days occurred on October 16 and on “BlackMonday,” October 19. The 90% Upside Day occurred two days later, onOctober 21. Then, like aftershocks following a major earthquake, two more90% Downside Days occurred on the first successful test of the lows inlate October, followed by a 90% Upside Day on October 29. The after-shocks continued in December and January, each followed by an equiva-lent 90% Upside reversal.

Exhibit 6

LOWRY'S 90% DAYS - 1974 - 1975 (DAILY)

AUG74 SEP74 OCT74 NOV74 DEC74 JAN75 JAN 31

570

600

630

660

690

720

750

780

810

840

870

0

90

-90

1974 - 1975




FEB80 MAR80 APR80 MAY80 JUN80 JUL80 JUL 31

725

750

775

800

825

850

875

900

925

950

975

0

90

-90

1980



LOWRY'S 90% DAYS - 1987 - 1988 (DAILY)

AUG87 SEP87 OCT87 NOV87 DEC87 JAN88 FEB88 FEB 29

1600

1720

1840

1960

2080

2200

2320

2440

2560

2680

2800

0

90

-90

1987 - 1988




JUL90 AUG90 SEP90 OCT90 NOV90 DEC90 DEC 31

2300

2380

2460

2540

2620

2700

2780

2860

2940

3020

3100

0

90

-90

1990




Three 90% Downside Days were recorded during July and August 1990.As a demonstration that the record is not perfect, a 90% Upside Day wasrecorded on August 27. The Dow Jones Industrial Average moved side-ways for two weeks before dropping to new lows.

Two more 90% Downside Days were recorded during September andOctober before back-to-back 80% Upside Days were recorded on Friday,November 9 and Monday, November 12 – twenty days after the marketlow.

Exhibit 7

During 1998, three 90% Downside Days were registered during Au-gust. The 90% Upside Day occurred just five trading days after the marketlow, on September 8. Another, superfluous 90% Upside Day was regis-tered on October 15, five days after a successful test of the Septemberlows.

This review of 90% Days would not be complete without bringing therecord up-to-date. And, the recent history may hold a particularly impor-tant message for investors: There were no 90% Downside Days recordedduring 1999 or 2000. However, the sharp drop in the Dow Jones IndustrialAverage during the early months of 2001 generated two 90% DownsideDays, on March 12 and April 3. But, during the ensuing rally, investorbuying enthusiasm was not dynamic enough to generate a 90% UpsideDay, leaving the impression that the final lows had not been seen. Afterjust six weeks of rally to the May, 2001 rally peak, the market began toweaken again, eventually plunging to a three-year low in the midst of theSeptember, 2001 tragedy. But, as strange as it may seem, the selling dur-ing that decline never reached the panic proportions found near almost allmajor market bottoms in the past 69 years. Not even a single 90% Down-side Day was recorded from May through September. Thus, the probabili-ties drawn from past experience suggested that stock prices had not beendiscounted enough to attract a broad sustained buying interest. In short,the final market bottom had not been seen in September 2001. And, thehighly selective rally that ensued from the September 2001 low throughearly January 2002 was, once again, not strong enough to produce a 90%Upside Day, thus adding to the evidence that the final low for the DowJones Industrial Average has not yet been reached, and that a period ofinvestor panic, generating a series of 90% Downside Days, may still beahead.

It is important to recognize that the pattern of 90% Days is not a new,untried, backrecord discovery. The original research was conducted by theLowry staff twenty-seven years ago, in early 1975. The findings were firstreported to the investment community in 1982 at a Market TechniciansAssociation Seminar. Since that time, the history of 90% Days has beenrecorded day by day, and has proven repeatedly to be a very valuable toolin identifying the extremes of human psychology that occur near majormarket bottoms. Obviously, no prudent investment program should bebased solely on a single indicator. Other measurements of price, volume,breadth, and momentum are needed to monitor the strength of buying ver-sus selling on a continuous daily basis. But, we believe the 90% indicator,as outlined above, will be an enduring, important part of stock market analy-sis, since it, like the other facets of the Lowry Analysis, is derived directlyfrom the Law of Supply and Demand – the foundation of all macro-eco-nomic analysis.

APPENDIX

90 DN 90 UP DJIA COMMENTS

03-03-60 612.05 Isolated. 3 days before short term low.07-29-60 616.73 Isolated. Start of 3 week rally.

09-19-60 586.7610-24-60 571.93 1 day before the 1960 low.

11-10-60 612.01 11 days after the 1960 low.04-24-61 672.66 Isolated. Bottom day of short term correction.05-22-62 636.3405-23-62 626.5205-28-62 576.9306-04-62 593.6806-12-62 580.9406-14-62 563.0006-21-62 550.49 2 days before the 1962 low.

06-28-62 551.35 3 days after the 1962 low.09-21-62 591.78 Cuban Missile Crisis.09-24-62 582.9110-19-62 573.29 3 days before the October low.

10-29-62 579.35 3 days after the October low.11-12-62 624.41 Superfluous confirmation.

12-10-62 645.08 Near bottom of short term correction.01-03-63 657.42 Important rally followed.

11-22-63 711.49 Kennedy Assassination11-26-63 743.52 Johnson Inauguration11-29-63 750.52 Successive 90% upside day. Confirmation.

06-14-65 868.7106-24-65 857.7606-28-65 840.59 1 day before the 1965 low.

06-30-65 868.03 1 day after the 1965 low.05-05-66 899.77

05-18-66 878.5007-25-66 852.8308-01-66 835.1808-26-66 780.5608-29-66 767.03

09-12-66 790.59 Premature.09-21-66 793.5910-03-66 757.96 4 days before the 1966 low.

10-12-66 778.17 2 days after the 1966 low.05-31-67 852.46 Mid-East War.06-05-67 847.77 Bottom day of war decline.

06-06-67 862.71 1 day after the bottom.02-08-68 850.3203-14-68 830.91 6 days before the 1968 low.

04-08-68 884.42 11 days after the 1968 low.07-28-69 806.23 Isolated. 2 month rally followed.

03-25-70 790.13 Isolated. Blow-off top.04-22-70 762.6104-27-70 735.1505-04-70 714.5605-20-70 676.5505-25-70 641.36 1 day before the 1970 low.

05-27-70 663.20 1 day after the 1970 low.11-30-70 794.09 Isolated. Important rally followed.

05-17-71 921.3006-18-71 889.16


APR98 MAY98 JUN98 JUL98 AUG98 SEP98 OCT98 NOV NOV 13

7140

7380

7620

7860

8100

8340

8580

8820

9060

9300

9540

0

90

-90

1998




08-03-71 850.0308-16-71 888.95 Nixon Price Controls.11-26-71 816.59 No prior 90% Down Day.11-29-71 829.73 Back-to-back 90% Upside Days.

05-09-72 925.12 Isolated. Short term low.05-14-73 909.69 Isolated.11-19-73 862.6611-26-73 824.9512-12-73 810.73

01-03-74 880.69 Bad signal. (See next series)01-09-74 834.7903-28-74 854.3504-23-74 846.0605-17-74 818.8407-08-74 770.5709-04-74 647.9209-12-74 641.7411-18-74 624.9212-02-74 603.02

01-02-75 632.04 Back-to-back 80% Upside Days on 12/31/74 & 1/2/7501-27-75 692.66 33 days after 1974 low. Start of bull market.

02-25-75 719.18 Isolated. Short term low.03-24-75 743.43 Isolated. Short term low.08-20-75 793.26

08-28-75 829.47 Within 2% of October 1975 low.09-19-75 829.7910-03-75 813.21 1 day after the 1975 low.

12-02-75 843.2012-03-75 825.49

01-05-76 877.83 Important rally followed.05-24-76 971.53 Isolated. 3 days before short term low.07-27-77 888.4310-12-77 823.98

11-10-77 832.55 Premature. Blow-off top.12-06-77 806.91

04-14-78 795.13 Back to back 80% Upside days on 4/13 & 4/14.Start of 1978 rally.

08-02-78 883.49 Isolated. Near blow-off top.10-16-78 875.1710-17-78 866.3410-20-78 838.1010-26-78 821.1210-27-78 806.05 1 day before the bottom.

11-01-78 827.79 1 day after the bottom.12-18-78 787.51 Test of November 1978 low.

01-03-79 817.39 Confirmed 11/1/78 signal.02-27-79 807.00 Isolated. Short term low.

03-08-79 844.85 Back to back 80% Upside days on 3/7 & 3/8.05-07-79 833.42 Isolated.10-09-79 857.5910-19-79 814.68

11-26-79 828.75 2 month rally followed.03-06-80 828.0703-17-80 788.6503-24-80 765.44 3 days before the 1980 low.

03-28-80 777.65 1 day after the 1980 low.04-22-80 789.85 Superfluous confirmation.

09-26-80 940.1009-29-80 921.93

11-12-80 964.93 Back to back 80% Upside Days on 11/11 & 11/12.Blow-off top.

12-08-80 933.7001-07-81 980.8902-02-81 932.2505-04-81 979.1108-24-81 900.1109-25-81 824.0101-11-82 850.46

01-28-82 864.25 Premature.02-08-82 833.43

03-22-82 819.54 2 month rally followed.08-17-82 831.24 No prior 90% Down Days.08-20-82 869.29 9 days after the 1982 low.10-06-82 944.26 Superfluous confirmation.

10-25-82 995.13 Isolated. One day from short term low.11-03-82 1065.49 Prelude to 1983 rally.

01-24-83 1030.17 Isolated. Bottom day of short term correction.

07-20-83 1227.86 Isolated. Blow-off top.08-02-84 1166.08 No prior 90% Down Days. 5 days after 1984 low.08-03-84 1202.08 Back-to-back 90% Upside Days.12-18-84 1211.57 Isolated. Start of major uptrend.

06-09-86 1840.1507-07-86 1839.0009-11-86 1792.8911-18-86 1817.21

01-02-87 1927.31 Late. Start of 1987 rally.01-05-87 1971.32 Back-to-back 90% Upside Days.

10-16-87 2246.7410-19-87 1738.41 “Black Monday”

10-21-87 2027.85 2 days after 1987 low.10-22-87 1950.4310-26-87 1793.93 First test of 1987 low.

10-29-87 1938.3311-30-87 1833.5512-03-87 1776.53 Second test of 1987 low.

01-04-88 2015.2501-08-88 1911.3101-20-88 1879.14

01-25-88 1946.45 Back to back 80% Upside days on 1/22 & 1/23.One day after 1988 low.

04-14-88 2005.6405-11-88 1965.85

05-31-88 2031.1206-01-88 2064.01 Back-to-back 90% Upside Days.06-08-88 2102.95 Superfluous confirmation.

08-10-88 2034.14 Isolated. Near short term low.09-02-88 2054.59 Important rally followed.

10-13-89 2569.2601-12-90 2689.2101-22-90 2600.45

05-11-90 2801.58 Unusually late. Start of 2 month rally.07-23-90 2904.7008-21-90 2603.9608-23-90 2483.42

08-27-90 2611.63 Premature.09-24-90 2452.9710-09-90 2445.54

11-12-90 2540.35 Back to back 80% Upside days on 11/11 &11/1202-11-91 2902.23 Superfluous confirmation.

08-19-91 2898.03 Russian coup attempt.08-21-91 3001.79

11-15-91 2943.2011-19-91 2931.57 Near short term low.02-16-93 3309.49 Isolated. Short term low.02-04-94 3871.4203-29-94 3699.02

04-05-94 3675.41 One day after 1994 low.03-08-96 5470.45 Isolated. Short term low.07-05-96 5588.1407-15-96 5349.51

08-02-96 5679.83 Back to back 80% Upside days on 8/1 & 8/2. Importantrally followed

04-11-97 6391.69 Isolated. Short term low.10-27-97 7161.15 Bottom day of 3 month decline.01-09-98 7580.42 Isolated. Short term low.04-27-98 8917.6408-04-98 8487.3108-27-98 8165.9908-31-98 7539.07

09-08-98 8020.78 5 days after the 1998 low in DJIA10-15-98 8299.36 5 days after the Oct. test of the lows.03-16-00 10630.60 Isolated.

04-14-00 10305.7703-12-01 10208.2504-03-01 9485.71

No further 90% days as of February 26, 2002

90 DN 90 UP DJIA COMMENTS 90 DN 90 UP DJIA COMMENTS

Journal of Technical Analysis (JOTA). Issue 57 (2002, Winter)

Economy & Finance

charles dow

dow award winners

dow award papers

cmt charles

scheinman charles

mta journal

eliades charles

useful journal