journal of technical analysis (jota). issue 23 (1986, february)

62
MARKET TECHNICIANS ASSOCIATION JO UFINAL lssue 23 February 1986

Upload: beniamin-paylevanyan

Post on 06-Aug-2015

95 views

Category:

Economy & Finance


6 download

TRANSCRIPT

Page 1: Journal of Technical Analysis (JOTA). Issue 23 (1986, February)

MARKET TECHNICIANS ASSOCIATION

JO U FINAL lssue 23 February 1986

Page 2: Journal of Technical Analysis (JOTA). Issue 23 (1986, February)
Page 3: Journal of Technical Analysis (JOTA). Issue 23 (1986, February)

MARKEX'TI!CHNIC~AssocIATIONJ~ Issue 23 February 1986

Editor : Henry 0. Pruden, Ph.D. Adjunct Professor Golden Gate University San Francisco, CA 94105

Manuscript Ikviebers: Arthur T. Dietz, Ph.D. Professor of Finance Graduate School of Business Administration, EXxy University Atlanta, Georgia

Frederick Dickson Portfolio Manager Millburn Corporation New York, New York

Richard Orr, Ph.D. Vice President for Research John Gutman Investment Corporation New Britian, Connecticut

David Upshaw, C.F.A. Director of Portfolio Strategy Research Waddell dnd Reed Investment Management Kansas City, Missouri

Anthony W. Tabell Technical Analyst Delafield, Harvey, Tabell Princeton, New Jersey

Robert T. wood, Ph.D. Associate Professor of Finance Pennsylvania State University State College, Pennsylvania

printer: Golden Gate University 536 Mission Street San Francisco, CA 94105

Publisher: Market Technicians Association 70 Pine Street New York, New York 10005

ran Journal/February 1986

Page 4: Journal of Technical Analysis (JOTA). Issue 23 (1986, February)

TABLEOFCONTENTS

FROM THE EDITOR: A WELCOME TO ACADEME ........

MTAOFFICERS ANDCOMMITTEECHAIRPERSONS. .......

MEMBERSHIP

sTyL;E SHEElT

AND SUBSCRIBER INFORMATION. ........

FOR SUBMISSION OF ARTICLES ........

THE ART OF TECHNICAL ANALYSIS HarryW.Laubscher................

RUMINATIONS ON '86 David Upshaw, C.F.A.. . . . . . . . . . . . . . .

AVERYVOLATILEYEAR Stan Weinstein. . . . . . . . . . . . . . . . . .

THEVALUELINEMYTH Don Dillistone. . . . . . . . . . . . . . . . . .

S'IWZHASTICS??? Arthur Merrill. . . . . . . . . . . . . . . . . .

STOCK MARKET TIMING: AN EMPIRICAL EVALUATION Michael J. Flanagan, Ph.D.. . . . . . . . . . . .

7

11

16

21

31

33

2 M'A Jourrd/E'eb~ 1986

Page 5: Journal of Technical Analysis (JOTA). Issue 23 (1986, February)

AWELCOMETOACADEME

The MARKET TECHNICIANS ASSOCIATION JOURNAL is dedicated to advancing the theory and the practice of market timing. This Editor views the advance- ment of technical theory and practice as a concommitant process: theory enhances the technician's capacity to act with skill, while the tech- nician's practical experience enlightens and leavens our technical theories.

. Technical market analysis is often referred to as the oldest of practices and the youngest of sciences. The MTA JOURNAL has benefitted greatly from the contributions of its professional members, who over the years have shared their thoughts and experiences with their colleagues through the disciplined mediumof an academic-type journal. Since the MarketTech- nicians Association is above all a professional organization, first prior- ity shallcontinuetobegiven tothepublication of top-flight articles from practitioners in the field.

The MTA JOURNAL also has a duty to perform in terms of scientific disci- pline. It welcomes critical appraisals of technical analysis, rigorous empirical tests of technical hypotheses, and studies in new conceptual directions for technical analysis. 'Ib fulfill this scientific obligation, the MTA JOURNAL welcomes the participation by scholars 'from all realms of academe. The Editor and The Reviewers of the MTA JOURNAL are neither tied to a single discipline, nor to a traditional method of analysis nor to a popular academic theme. The Journal welcomes conceptual models, critical appraisals, historical case studies, psychological or sociologicalan- alyses, computer science applications, comparative market analysis, visual artistry, empirical tests of technical hypotheses or contributions from any academic discipline which might further the theory and practice of tech- nical analysis.

We are confident that you shall be pleased and honored to see your work published in the MTA JOURNAL. Please direct your manuscripts or your inquiries to the Editor, MTAJOURNAL,Box 1348,Ross,CA 94957,0r tele- phone (415) 459-1319.

Thank you,

Editor, fiPrZJOURNAL

MIX JournalDeb- 1986

Page 6: Journal of Technical Analysis (JOTA). Issue 23 (1986, February)

PRESIDENT Gail Dudack Pershing/Div. DLJ 212-312-3322

1985-86 MARI(ETTECHNICLANsAssocIATION

OFFICERS

vIcxPRsII3ENT Cheryl Stafford Wellington Managerwnt 617/227-9500

John Murphy JJM Technical Advisors 212/724-6982

David Krell New York Stock Exchange

VICEPRESIDBW (Seminar) Robert Simpkins Delafield, Harvey, Tabell 609/924-9660

COMMITI!EECHAIF@EEGONS

PRIGEGWS EIXKSb~~Cl?ELATI~ Robert Colby Tony Tab&l 212/399-6002 609/924-9660

Robert Prechter 404/536-0309

John Brooks 404/266-6262

JOURNAL Henry Pruden 415/459-1319

HxKaTIoN Fred Dickson 212/398-8489

XXRl3X'I2iTION Charles Comer & John Brooks 212/825-4367 404/266-6262

cor%wmRswcIAL~~ John McGinley 203/762-0229

BP Phil Roth 212/742-6535

EuIuREs--GROUP William Byers 212/925-6651

Ralph Acaqora 212/510-3750

4 MIZI Journal/Ekbruary 1986

Page 7: Journal of Technical Analysis (JOTA). Issue 23 (1986, February)

ELIGIBILITY: REGULAR MEMBERSHIP is available to applicants "whose total professional efforts are spent practicing financial technical analysis which results in an identifiable research product that is either made available to the investing public or becomes a primary input into an active portfolio management process." (From revised Constitution)

ASSOCIATE MEMBER status is "reserved for professional users of technical analysis (i.e. money managers, traders,brokers, floor specialists, etc.) who are not engaged primarily in technical research, but for whom technical analysis is the basis of their decision-making process." (From revised Constitution)

SUBSCRIBERcategory is available to individuals who are interested in keeping abreast of the field of technical analysis, but who don't fully meet the requirements for regular or associate membership. Privileges are noted below.

Applications Fees: A one-time application fee of $10.00 should accompany all applications for regular and associate members, but not for sub- scribers.

Dues: Dues for regular members, associate members and subscribers are $100.00 per year and are payable upon receipt of dues notice in September each year. -------------------------e-----e- ----___--______---__-~~---~~--~---~~-------~

Invitation to i%nthly MIA Educational Meetings

Receive Monthly MTA Newsletter

Receive Tri-Annual MTA Journal (Nov-Feb-May)

Use of MTA Library

Participate on Various Committees

Eligible to Chair a Conunittee

Eligible to Vote

Fee Discount - MTA Annual Seminar (MaY)

Regular Members

Yes

Yes

Yes

Yes

Yes

Yes

Yes

Yes

Associate Members Subscribers

Yes Yes

Yes Yes

Yes Yes

Yes Yes

Yes Yes (IZxeptional membership)

No No

No No

Yes Yes

Annual Subscription to the MTA Journal ONLY -- $35.00 per three issues. Single Issue of MTA Journal (including bmssues) -- $15.00 each.

IWA Joumal/E'ebnmry 1986

Page 8: Journal of Technical Analysis (JOTA). Issue 23 (1986, February)

MTA Editorial Policy

The MARKEXTECHNICIANS AS$OCIATIONJOURNAL is published by the Market Technicians Association, 70 Pine Street, New York, New York 10005 to promote the investigation and analysis of price and volume activities of the world's financial markets. The MTA Journal is distributed to individuals (both academic and practitioner) and libraries in the United States, Canada, Europe and several other countries. The Journal is copyrighted by the Market Technicians Association and registered with the Library of Congress. All rights are registered with the Library of Congress. All rights are reserved. Publication dates are February, May, and November.

Style for the IWA Jaxnal

All papers submitted to the MTA Journal are requested to have the follow- ing items as prerequisites to consideration for publication:

1. Short (one paragraph) biographical presentation for inclusion at the end of the accepted article upon publication. Name and affiliation will be shown under the title.

2. All charts should be provided in camera-ready form and be properly labeled for text reference.

3. Paper should be submitted typewritten, double-spaced in completed form on 8 l/2 by 11 inch paper. If both sides are used, care should be taken to use sufficiently heavy paper to avoid reverse side images. Footnotes and references should be put at the end of the article.

4. Greek characters should be avoided in the text and in all formulae.

5. Two submission copies are necessary.

Manuscripts of any style will be received and examined, but upon acceptance, they should be prepared in accordance with the above policies.

MI!A Joumal~ebmary 1986

Page 9: Journal of Technical Analysis (JOTA). Issue 23 (1986, February)

By Harry W. Laubscher

Having spent almost twenty-nine of the last thirty years working in the stock market's highways and byways, I hope that I have learned something. We all manage to learn a great deal, regardless of what field we work in, butalltoo often many of us tend to forget many of the things that were learned and which should not'have been forgotten. It has been often said that the stock market, along with drink and women, is one of the great levellers of our time. Many of the important pieces of market lore that we learned along the way, and then in later years tended to forget, nodoubt could have saved many of us from experiencing many of the mistakes that,all of us make. I am reminded of this lately as I see a great rush on the part of inexperienced brokers and traders to be "in the crowd" regardless of where that crowd is headed. For some strange reason, the more the stock market rises, the more bullish many of us tend to become, finally resulting in a great rush to own shares right at the top of the market. On the other hand, it usually works out that the lower the market goes in bear markets, the more bearish more people tend to become. It has always been so, and as long as people are the driving force behind all market movements, up or down, it will always be so.

We have all heard some of the sayings for which Wall Street is so well known, such as "sell on the good news and buy on the bad news." I've found that this does, indeed, work out to one's benefit more often than not. During the Union Carbide fiasco in India when the shares of the company dropped sharply to near 33, the wise people were there buying all they could get, knowing full well' that the lemming instinct had once again taken things too far. The recovery in price of those shares since then is in the record for all to read and great profits have been made in what "everyone knew" was going to be a disaster for the company. More recently, we have the situation of Texaco and Pennzoil. The rather silly awarding to Penn- zoil of several billions of dollars was recognized by many savvy traders as an opportunity to acquire an historically "good" company at what appeared to be bargain priced levels. As of this writing, the shares of Texaco are still floundering near the 30-31leve1, and although my point and figure work suggests a potential downside count to approximately the 29 level, I am advising investors with some patience to start acquiring Texaco shares in the 30-31area. In time, this should work out tobe a good buy. I use it as another example of the unsophisticated atmosphere that appears to be so prevalent today.

And yet, I stop to wonder if I ever really did meet anyone at all who could accurately be described as sophisticated in the stock market. Being sophisticated in the stock market probably is as out of place as being logical. And we all know that in order to be successful in the world of investing, logic has tobe left outside the door. This brings me around to the inevitable question: "Is understanding the stock market now becoming more of a science and less of an art?" No doubt, it is a question that has troubled the minds of many marketeers for many years. I know, as

7 MI24 Journal/February 1986

Page 10: Journal of Technical Analysis (JOTA). Issue 23 (1986, February)

a result of my recent trip to Japan, that the Japanese believe that scientific applications can be applied to the stock market and they have gone to great lengths in employing those applications. More than any group I know, the Japanese are attempting to make it more of a science than it has been. And yet, I know that whenever you have to deal with something that involves people to any great extent , science can only be carried so far before art has to take over. Thanks to many of the new inventions that have come along over the years, such as the price quoting machines, information is much more readily available, making the formerly onerous job of keeping up to date much less so. may, a great deal of information is quickly available -- perhaps too much so -- and thus the odds in decision making have increased on the side of error. Now I know that that last sentence doesn't seem right somehow, but then you are still thinking _ logically, aren't you? And that doesn't work in regard to the market. Too much information, too easily obtained leads one into too many possible byways, and therefore, increases the chances for error. Too many people believe that the more you know about something, the better off you are apt to be. I thoroughly agree, except in the stock market. In this arena, one often can lose sight of the forest for all the trees that are available and it often helps touse less data and a bit more gut feel. And, very often, who you know is just as important, if not more so, than what you know. How else do you explain the success levels of those who tend to make it in the market?

And this brings metooneof my favorite sayings about the market. It is one in which I thoroughly believe and have seen the workings of it spread far and wide, among all types of marketeers. "The stock market is one of the easiest places in the world to get rich.“ All you have to do to make it so is to avoid what most of the others are doing. For example, I long ago gave up buying The Wall 'Street Journal. It has too much information of too little worth and not enough of the really valuable stuff. Barrons is somewhat better in that respect, but the new newspaper Investors Daily has it allover both of the Dow Jones papers. Once you start getting really good news on what is going on in the market, the path to wealth is soon beneath your feet. It helps also to look around you, ask the man in the street what he thinks about the economy, or whatever, and when you have determined what the general drift of conventional wisdom is, go the oppo- site way. One of the biggest obstacles to obtaining wealth in the stock and bond markets is to fall prey to the enticements of quick profits. Of course, they are grand to have but more often than not it pays to "let your profits run, while making all endeavors to cut losses short." Tbo often, technically-oriented traders and investors see more in the chart than really is there to be seen. False breakouts, upor down, make us nervous and we jump, only to find out later on that there was no alarm except in our own minds.

I also believe that it is a bit wise to be skeptical of almost everything. At times you will have to depend on what appears to be the wisdom of those around you, those in whom you have faith to do the jobs with which they are involved. But one also should take the time to hear what others have to say,,and then go and do a bit of "checking itout." It can't hurt. I also believe that too many investors fall prey to the belief that information on

8 IWA Joumalfiebruary 1986

Page 11: Journal of Technical Analysis (JOTA). Issue 23 (1986, February)

revenues, management, contracts, industry items, sales and earnings are what makes stocks move up and down. They seldom stop to think that all that kind of information only has todo with the company itself, NOT the shares of the company in question. The only thing that makes shares move up or down in price is buying or selling pressure outweighing one another. If nobody sells, then all the good information on dividends- and earnings isn't going to move shares upward. If all the bad news makes buyers disappear, then shares can't move downward. So trying to gauge what the buying pressure is probably is the most important thing that anyone can do in the search for profitability. At Paine Webber, every week we publish a relative strength analysis of over 4000 issues that when taken in conjunction with some other information, a ffords a very good indication of whether or not buying, or selling, pressure is rising or_ declining. Once you have that tool in your hands, the game becomes a lot easier. Over the last ten-eleven years, every single issue recommended in my TRENDS & OPPORTUNITIES MARKET LEXTER has been based on my reading of the buying or selling pressures. That has helped us achieve a 95% success ratio in those recommendations, 250 profits, nine losses and three unchanged since 1974, regardless of whether a bull market or abear market was in the driver's seat. Get to know what direction the pressure is moving in and you are halfway to your objectives. And that goes just as importantly for short- selling.

While we are on the subject of short-selling, technical analysis can be of great help in helping clients make money on the short side. Once you find a chart pattern that is descriptive of distribution, move on to find out if the selling pressure has been increasing, or if the buying pressures are ebbing. If both suggest you should be shorting the stock, go a step further and check out the short-interest. If it is high, so much the better, since most of the shorting is still done by professionals. And don't fall for that old saw about stocks with high short interest holding up well, because there is a buyers' floor under the price. It is quite truethatthose who sell short must sooner or later buy back in again in order to take their profit, or their loss. But a check of past bear markets will show that stocks that had the highest levels of short-interest usually sold off quite nicely, enabling those who sold short to repurchase shares AT LOWER IXVELS. A floor under stocks with high short-interest is about as fleeting as support levels in a bear market. I always try and remind brokers who ask me about support levels in a bear market that support is only a seven letter word and usually doesn't afford the support sought. Support, on the other hand, is much more important, technically, in a rising market. The same goes for resistance levels. In bull markets, those resistance levels usually provide only fleeting roadblocks to ad- vances. In bear markets, upside resistance takes on much more power, on average. There always are exceptions, of course.

I guess if I had only one tool to select from all those that are available among the various charts and chart services, I would come down on the side of a good weekly bar line service. Something like Mansfield that provides the relative strength indicator graphically presented, various moving aver- ages, and then throws in upside volume and downside volume to make it a bit easier. If you like having the fundamentals, those are provided as well.

9 m Jotim-d/February 1986

Page 12: Journal of Technical Analysis (JOTA). Issue 23 (1986, February)

They once used to give earnings' estimates, but not anymore. Too bad! It helped to gauge things better if you knew what the "street" was expecting. Then, when the winds of winter were blowing and I was all snug by my fireside, I'd take out my barline book of charts and go through it every week, looking for those seven cardinal patterns that indicate either accum- ulation or distribution. You all know what they are. You don't need me taking up valuable space to repeat them again. Once I was able to cor- rectly identify some of those patterns, I would put some of my funds to work. I guess that when push comes to shove, those important patterns of accumulation or distribution are the most important things in our world of Technical Analysis. Without the knowledge of them, we're always back at square one.

Now I certainly don't mean to knock point and figure analysis. I have found it tobe too helpful over the years to give it a position below the salt. It is an invaluable toolin trying to gauge just how far a move is going to carry ONCE THAT MOVE HAS STARTHD. But, if pressed, I still would have to say that a bar line chartwill tell you when the move is going to start. Thenyouwould moveon to aP&F chart. If anyone out there wants to make a lot of money in this business, I would suggest that they start a point and figure weekly chart service of the 500 most actively traded listed bonds. As far as I know, there is no such service available. Since we are in the still early stages of a super cycle bull market in bonds, their price performance will become increasingly important as the next five to seven years roll by. And their volatility also will increase, making a point and figure analysis far more valuable. I've thought of doing it myself, but am just too tired to take on another chore.

The twenty-nine years will soon be rolling into a nice round thirty. I came to Wall Street intending to stay only twenty years, but the work was so interesting and the people with whom I worked were so pleasant and helpful, that I stayed on and on and on. This has been the most fascina- ting of my three careers and I am wondering if the fourth and next career will be as rewarding.

Harry W. Laubscher is a member of the MTA and a Technical Analyst at Tucker Anthony, New York, New York.

10 m!A JournalDeb- 1986

Page 13: Journal of Technical Analysis (JOTA). Issue 23 (1986, February)

RuM1NAmm ON '86

By David Upshaw, C.F.A.

Another Icok At The Four-Year Market Pattern Gives 1986 Low odds Of Being An Up Year

Unless the SPII collapses between now and December 31, the year 1985 will go into the history books as the fourth year of consecutive price gain for the index, as follows:

1982 +15.0% 1983 +18.2% 1984 .l% 1985 Jl9.0% as of 11/20

This fact may be a useful piece of information to know as we assess the 1986 market outlook. From 1920 to date, the SPII [Standard & Poor's Indus- trial Index] gained for four consecutive years in only five instances. They were the four-year periods ended in 1927, 1928, 1945, 1952, and 1961. Here's how the SPII performed in the years following those four-year gain periods:

1928 +40.7% 1929 -18.5% 1946 -12.2% 1953 - 7.5% 1962 -12.8%

In the 1920-1985 period, the SPII wentupin five consecutive years only once, from 1924 through 1928:

1924 +16.2% 1925 +26.7% 1926 + 6.6% 1927 +36.8% 1928 +40.7%

Then the market went down for four consecutive years for the only time in the 1920-1985 period:

1929 -18.5% 1930 -30.0% 1931 -46.9% 1932 -18.0%

For this study, I looked at more than the sequences just described. The period 1920-1985 was broken down into overlapping four-year spans: 1920- 1923, 1921-1924, and so on, upto1981-1984. I noted the number of years that the SPII gained in each of those 61 four-year periods. For example, there were 7 four-year periods in which th,e SPII gained in 2 out of 4 years. There were 31 four-year periods in which the SPII gained in 3 years

11 MI'A Jourrd/E'ebm 1986

.

Page 14: Journal of Technical Analysis (JOTA). Issue 23 (1986, February)

out of 4.

Next, I looked at what the market did in each of the 61 years that followed those 61 four-year spans. I found that the market rose 24 times and declined 7 times in the years that followed the four-year spans in which the market rose 3 out of 4 years. The market rose in only 3 of the years that followed spans in which the SPII rose only 1 year out of 4. The record of all the years that followed the four-year spans is shown here:

# of Times % of Years Market Gained Market Gained

Market Gained # of Times the Following the Following Year

0 out of 4 years 1 1 100% 1 out of 4 years 7 3 43% 2 out of 4 years 17 11 65% 3 out of 4 years 31 24 77% 4 out of 4 years 5 1 20%

TOtdlS 61 40 66%

Historically, by far the best outlook for a given market year has been when the market has gained in any three of the previous four years. The SPII has gained in 77% of the years following such a pattern.

On the other hand, the SPII has only gained in one year following a string of four consecutive gaining years. You may argue that1984, with its .l% gain, was not really a gain year. Well, it wasn't a loss year, and for the purpose of this study, a gain is a gain is a gain. The year 1985 had a 77% chance of being a winner according to this history, and it has been. (The year 1985 followed a four-year span that had three gain years.) The year 1986 has a 20% chance, according to this study.

Surely, something is different about every cycle, and history is only a rough guide. Butdon't ignore the immortal words of Irene Peter: "Just because everything is different doesn't mean anything has changed."

"When We're Out Together Dancin' Peak to Peak"

The chart on the reverse side shows how much the S&P 500 gained from the peak of one bull market to thepeak of the nextbullmarket, beginning in 1962. I see the chart as showing five bull markets. The peak-to-peak gains are as follows:

1962-66 +32% 1966-68 +15% 1968-73 +11% 1973-80 +17%

12 ran J-/E'ebruary 1986

Page 15: Journal of Technical Analysis (JOTA). Issue 23 (1986, February)

1962-80 Mean +19% Standard deviation 9 percentage points

1980-86 +52%

1962-85 Mean +24% Standard deviation 14 percentage points

The current peak-to-peak measurement of +52% surpasses the 1962-66 gain of 32%, and is over twice the +19% mean of the 1962-80 period.

I measured only complete bull markets , which is why I did not use the 1983 high of 172.65 shown on the chart. The 1983-84 decline was the intervening corrective leg of the current two-legged bull market, in my opinion.

To go back to the great 1962-66 bull market (32% peak-to-peak): the peaks were four years apart, almost to the day. We are now more than five years past the 1980 peak, and the gain from that peak is 44%.

Henry Ford said, "History is bunk."

Shakespeare penned, "What's past is prologue."

Upshaw once wrote, "The past never repeats exactly. If it did, history professors would make great money managers." Even so, these historical figures are worth writing -- and thinking -- about.

'First Five Days' Indicator Bearish on 1986: Don't Laugh

The SPIC [Standard & Poor's-Composite Index] closed at 207.97 today, the fifth trading day of the new year. The loss from the close on December 31 is 1.6%; thus this rather arcane indicator has predicted that 1986 will be a down year.

How good has this indicator been? Well, darn good, whenitwas bullish, and pretty good, when it was bearish. Here is the record for the 55-year period 1931-1985:

The SPIC advanced in 34 of the 55 years, or in 62% of the years. A forecaster like, say, Barton Biggs, who was bullish all the time would have been right 62% of the time.

The "first five days" indicator (FFDI) gave 39 "up market" signals. The market gained on the year in 29 years, or 74% of the time.

The FFDI gave 16 "down market" signals, which were followed by 11 down years; so when the FFDI was bearish, it was correct 69% of the time.

In its Bull and Bear modes combined, the FFDI has been correct on 40 out of 55 forecasts, for an accuracy record of 73%. That's much better than the 62% accuracy record of the eternal bull. It's odd, kooky, and baffling, but there it is.

13 MJX Joumal/Febmary 1986

Page 16: Journal of Technical Analysis (JOTA). Issue 23 (1986, February)

The most notable failure of the FFDI was its 1985 forecast, which was bearish. The SPIC gained 26.3% in price during the year just ended. Even so, if one invested in the SPIC using only the FFDI as a timing guide since 1950, one would have been invested inonly 23 years ofthepast36 years, and would have had an annual compound rate of return of 11.8% (not includ- ing dividends) for those 23 years, in spite of being out of the market in 1985. By contrast, a buy-hold strategy would have produced a compound rate of return over the 36 years of 7.3%. (I have ignored taxes, and I have not computed interest income earned during the 13 years one would have been out of the market.)

David Upshaw is a member of the MTA and Director of Portfolio Strategy Research, Waddell and Reed Investment Management, Kansas City, Missouri.

14 MI’A Jou.n-d/‘l?ebrua~~ 1986

Page 17: Journal of Technical Analysis (JOTA). Issue 23 (1986, February)

00 dd

Q (13 . h s;

zF 2861

z k

. 186I .

l aa6I i

f

6L61

El461

f A? T

-K- 72

-zL --- = - >

US1

9L61

I SLSI

U61

IL61

BL61

6961

am

L961

9961

S361

b96I

E361

t3

T- Z361

m Journal/E’ebruary 1986

15

Page 18: Journal of Technical Analysis (JOTA). Issue 23 (1986, February)

By Stan Weinstein

While the major trend of the market is still bullish, I do not expect 1986 to be an "instant replay" of 1985! I know that many analystsare talking about a non-stop continuation of this rally that will take the market to 2000-perhaps even closer to 3000, but I feel that the probabilities favor a very different type of year. I expect a very choppy market that is likely to make a high within the first 90 days of 1986 (on this side of 1700). After that I feel that a decline toward the 1300-1400 area is a real possibility.

There are several reasons for my projected scenario, so let's go through them one at a time to see why I feel exactly as I do. (1) First of all, there's the reality of the 4 year cycle which is no longer getting very much press (and as a contrarian I feel that's significant). As you know, there was an important bottom made in the late 1957-58 period, then again in 1962, 1966, 1970, 1974, 1978, 1982 and we'll see about 1986, Obviously, if this pattern holds true to form, an important decline must first precede an important low , so that's point number 1 to consider. But there's much more involved in my viewing this as a late bull move than cycles, so let's look at the additional evidence.

(2) Another compelling reason for advocating a selective and cautious approach to the market at this point in time is the action of my proprietary NYSE Survey. I feel that any stock, commodity, or market average can be classifiedas being in one of four stages (see Chart 1). State 1 is the basing phase that takes placebefore an important advance gets underway. Once the base is completed and the stock breaks out above its resistance level, as well as its moving average, it enters.State 2 (pt. A) which is the Advancing phase. As long as all rallies move to new highs y and reactions hold above prior lows (as well as the 30 wk. moving average), I consider the stock to be in Stage 2 and feel that investors should stay with the position. Eventually, a pattern starts to unfold which is just the opposite of what took place in the Stage lbase. Now it trends side- ways as it nears the moving average and rallies no longer move to new highs. A Stage 3 potential top formation is now unfolding and once it breaks below its support level, as wellasthe moving average (pt.B), it enters State 4 which is the declining phase.

Each week I calculate the percentage of technically healthy NYSE stocks (in Stages 1 and 2) and chart that percentage against the DJ Industrial Average (Chart 2). Usually the two move pretty muchin gear, but when this gauge starts to diverge (by refusing to confirm downside Dow movement which is favorable), the odds are strong that a change in market trend will soon unfold. Note how in late 1974, after the worst crash in a generation, this gauge moved higher while the Dow continued to decline (pt. A). This action foreshadowed the 1975-76 bull market. Then in late 1976 as the DJI topped out near 1020 (pt. B), notice how this gauge had already shown weakness for several months as it once again correctly forecast trouble ahead of the

16 KI'A Journal~&ruary 1986

Page 19: Journal of Technical Analysis (JOTA). Issue 23 (1986, February)

1977 bear market. The next clear-cut signal was the sharp reversal from very low levels to far healthier readings in the late 1978 (pt. C), this time pretty much coincident with the turn in the averages. Then in early 1980 (pt. D), as the Bunker Hunt panic hit the street, this gauge held far above its 1978 low even though the Dow average dropped below its 1978 low which was a very positive divergence and a new 14-month advance was soon underway. By mid-1981 (pt. E), the market was near its high for the past 5 years, but the NYSE Survey wasn't able to better its 1980 high. Soon thereafter, the June 1981-August 1982 bear market was a reality. This indicator really did its thing as the 1982 bottom was forming. For several months (pts. F, G, and H), the Dow dropped to a series of new lows while this fine technical tool not only refused to confirm, but actually started trending higher. By the time the Dow hit its final August1982 low near 772, the NYSE Survey had improved to the point where 50% of all NYSE stocks were rated healthy (versus 25% several months earlier). This extreme divergence was then followed bythepowerhouse August1982-January 1984 bull market. As the market neared 1300 in January 1984 (pt. I) and bullish sugarplums were dancing in traders' and investors' heads, the Survey went into a tailspin.

After reaching a peak reading of over 90% a year earlier and then declining throughout all of 1983, it broke below the 50% level and the January-July 1984 mini bear market unfolded as the Dow lost over 200 points. Then in July (pt. J) of that year as the market bottomed near 1080, this gauge once again registered a positive divergence as it moved up to the 50% level after having been below 30% just a few months earlier. advance from 1079 to close to 1600 was soon underway.

The July 1984-??? Now let's look at

the recent action of this gauge which is once again giving us a loud and clear message. In late July,(pt. K), the Dow moved to yet another new high at 1372, but this indicator only moved to a position slightly above 80% (versus its reading of 95% in early 1983). That was strike one. It then deteriorated quickly during the July-September intermediate term decline. Then when the new rally got underway, another new high (pt. L).

the Dow once again moved to yet However, once this gauge failed to better its

prior peak and there are now 3 clear declining peaks (pts.M,N,O). That's strike two. Strike three will occur when renewed weakness once again drops this gauge below the 50% level.

(3) The next reason to be on guard againstan important topin the coming months is the fact that our Secondary Offerings guage (which is simply a 10 wk. total of the number of such offerings that are listed in the Market Laboratory section of Barrons each week) is weakening (see Chart 3). While this is not a pinpoint indicator, history shows that when there are very few such offerings, the overall market is usually close to an important bottom (such as was the case in early 1980, the summer of 1982 and the summer of 1984). Conversely, when a relatively high level of secondary offerings come to the market (40 or more on a 10 wk. basis) showing that big money feels that it's time to start bailing out of stocks, the market is usually approaching a top area (second quarter of 1981 and summer of 1983). It's therefore cause for concern that this indicator recently moved into sell territory in total agreement with the warning being flashed by our NYSE Survey.

Km Journal/February 1986 17

Page 20: Journal of Technical Analysis (JOTA). Issue 23 (1986, February)

(4) Finally, there's the reality of the lagging number of stocks hitting new highs (see Chart 4). In the late 1980-81period,this phenomena went on for quite a while but eventually took its toll as the June 1981-August 1982 bear market unfolded. Again throughout1983, this pattern could be observed and once again there were problems as the secondary market topm out in June of 1983 and the DJI hit its peak in early January 1984 and then both sectors sold off until they made their low in July. Once again, this pattern is making itself felt as the latest peak was lower than the Decem- ber 1985 peak which was slightly lower than the February 1985 peak (which were all lower than the 1982 peak) despite the fact that the popular averages continue to move higher.

Therefore, while there is likely to be further upside action over the next few months, the market is obviously becoming far more selective, and it is vital that we concentrate new buying in only the technically strong sectors (such as Homebuilding, Home Furnishings, Mobile Homes, Papers, Savings & Loans, Technology and Truckers) and in addition, that we keep a very close watch for signs of an important top in the not-too-distant future!

Stan Weinstein is Editor & Publisher of The Professional Tape Reader.

18 MICA Jou.rrd/Februa.Ky 1986

Page 21: Journal of Technical Analysis (JOTA). Issue 23 (1986, February)

Chart 1

.

“Ideal” STAGE i

Chart 2

STAGE 1 and STAGE 2 STOCKS ,

Chan 3

MTA Journal/February 1986 19

Page 22: Journal of Technical Analysis (JOTA). Issue 23 (1986, February)

Chart 4

550

500

450

400

350

300

250

200

150

100

050

000

950

900

850

800

+600 +400 +200

D.J.I.

NYSE HIGH-LOW DIFFERENTIAL

(COMMON STOCX ONLY-WEOCLY)

0 -200

-400 i82 J 1’83 I 1’84 .I A’85 .h J’86

N Y a ‘il k Y N L ia

20

MB Jou.rnal/E’ebrUary 1986

Page 23: Journal of Technical Analysis (JOTA). Issue 23 (1986, February)

By Don Dillistone

At the Hilton Head Conference of the Market Technicians Association, one of the speakers made an off-the-cuff comment that he wouldn't use the Value Line Composite Index in his technical work because of its so-called down- ward bias. This myth has been around for some time, but I found it a little surprising that it still is not realized throughout the technical communitythattheideathat a geometric average has a downward bias is indeed just that, a myth.

The Value Line is one of several developed over the years to provide an unweighted average or index to track the overall stock market. An un- weighted average treats each and every stock exactly the same, so that a one per cent move in stock A has exactly the same weight on the result as a one per cent move in stock B, no matter what the underlying prices of the two stocks happen to be. This is in contrast with the such averages as the Dow Jones Industrials, which simply adds up all the prices and divides the sum by a particular divisor. There are 30 stocks in the Dow Jones Industrials, but the current divisor, rather than 30, is 1.116, reflecting adjustments over the years to take into account stock splits, additions and deletions. But no matter what the divisor, the calculation means that a one per cent move in a hundred dollar stock has twice the weight on the result as a similar one per cent move in a 50 dollar stock. It was to avoid this that the concept of unweighted indexes or averages was born. Simply put, the idea is to zero in on percentage changes rather than dollar changes.

I have no intention to denigrate the work of Norman G. Fosback, but it is only in his otherwise excellent book, Stock Market Logic (l), that I have found a fully documented attack on geometric averages in general and the Value Line Composite Index in particular. So I have referred to his arguments so as not to be accused of setting up a "straw man" in order to knock it down.

An arithmetic average is calculated by adding up n items and dividing the sum by n to obtain the arithmetic average. On the other hand, a geometric average is calculated by multiplying n items together and then taking the nth root of the result to obtain the geometric average. In his book, Mr. Fosback states that "a geometric stock market average will always (sic) be below a simple average" (2), and because of this, argues that the best unweighted stock market average should be calculated by taking the arithmetic average of the per cent changes of every stock price over its previous day's price and that the new level of the stock market average should be adjusted by a similar percentage. This is in contrast with the Value Line method of taking the geometric average of the same percentage changes and using the result to calculate the current day's index. (Value Line's average has been indexed).

In Mr. Fosback's book, he compares the difference using as an example an

21 ME4 Journal/J?ebruary 1986

Page 24: Journal of Technical Analysis (JOTA). Issue 23 (1986, February)

equal dollar amount in each of two stocks, one of which subsequently moves up by 25 per cent while the second one drops by 25 per cent. Using the arithmetic method of averaging these changes, since one stock moved up .25 per cent while the other moved down 25 per cent, the average change would be zero. On the other hand, the geometric calculation shows the first stock up by 1.25 and the second by 0.75. The geometric average of these two, the square root of (the product of 125 divided by 100 times 75 divided by 100, or 1.25 times 0.75), is 0.96824584. Thus, the geometric average would show a decline of more than 3 per cent even though the total value of the two stocks remained unchanged. From this, he develops the proposition that the geometric average has a downward bias.

This proposition is then promoted to the rank of a premise to illustrate that the Value Line has a downward bias. Unfortunately, the proposition is built upon sandy soil. It is always dangerous to argue by analogy unless the analogy is rigorously pushed to extremes, i.e., reductio ad absurdum to show that it remains true.

To illustrate this, let us take as being self-evident that if a given method provides a downward bias for 1,700 stocks, itwillalso provide a downward bias for two. We can then assume a universe of two stocks, A and B, each one trading on day one at $100, an unweighted average calculated by using the average arithmetic changes and a geometric index, both at 100 on day one, along with a mythical portfolio manager who owns a single share of each stock.

On day two, after A had appreciated to 125 and B depreciated to 75, Mr. Fosback would point out (quite correctly) that the total value of the portfolio had not changed. This would be reflected in the arithmetic unweighted average remaining unchanged at 100. But the geometric average would show a loss to 96.82. Butifthe same two stocks on day three were to return to 100, the arithmetic unweighted average would show Stock A down by 20 per cent but stock B upby 33.3 per cent- the result being that the average arithmetic change would have been plus 6.67 per cent and the new average value would be 106.67. Cm the other hand, the geometric average of the same change would show that the change would have been the square root of (100 divided by 75) times the square root of (100 divided by 125), or 1.0327956. Multiplying that factor by the previous, unrounded value of 96.824584 results in a new value of precisely 100.

Stretching the analogy this far leaves our portfolio manager with a share in two stocks, eachof whichistrading at 100. The geometric average of the two-stock market universe shows that the market average has returned to 100, but the arithmetically unweighted average shows that the market has gone up to 106.67. The analogy has clearly brokendown: in one instance it seems to show that the arithmetically calculated unweighted average better reflects the action of a mythical portfolio, in a second instance, the one calculated geometrically appears superior.

But now the odds that a geometric averageof changes has a downward bias are becoming very long indeed. If the Value Line had set its index at 100 on day one, it would be back to 100 on day three. Furthermore, no matter

22 PII% Journal/February 1986

Page 25: Journal of Technical Analysis (JOTA). Issue 23 (1986, February)

how many times werepeatedthe above exercise with stocks Aand B, every time they simultaneously returned to 100, so would thevalue Line'Index. But using the impeccable logic based on his premise that the Value Line has a downward bias, Mr. Fosback points out that this bias would have shown up every day since its inception, leading to the obvious conclusion that after hundreds of daily calculations "the cumulative consequence is that the Value Line Average is so biased downward that its current level, placed in the perspective of its own history, is utterly without meaning".

In reality, the mathematics used in calculating the Value Line Index cannot possibly lead to a cumulative error.

The geometric method involves, throughout, commutative calculations. That is, it is a matter of complete indifference as to the order in which the various calculations are carried out. (A numerator that calls for the multiplying of 4 times 3 and a denominator calling for the result to be divided by 6 is always going to result in the answer 2 no matter in which sequence we carry out any of the called-for operations. It doesn't matter at what point we divide by 6). The arithmetic averaging of per cent changes in non-commutative. (The price of stock A must ALWAYS be divided by yesterday's price of stock A and the result added to the price of stock B ALWAYS divided by yesterday's price for stock B).

That doesn't make one method true and the other one not, but it does give us a powerful tool in simplifying the geometrical method.

To start with, and I hate to belabour this, but one other algebraic truth should perhaps also be mentioned. If we have a series of commutative calculations and each element is raised to the same exponent, then the calculations can be carried out first and then the result raised to the same exponent. (5 times 5 times 7 times 7 provides the same result as 5 times 7 times 5 times 7, or more apropos to the following discussion, the square root of 4 times the square root of 9 gives the same result at taking the square root of 36 would).

The calculation that I outlined above for the Value Line called for the dividing of today's prices by yesterday's prices by the day before's prices and so on. The fact that the Value Line calculation is commutative, though, means that it can bedone in any order. The significance of this is that I don't have to divide today's price of a stock by yesterday's price, as long as some time in my calculation I do indeed divide the calculation by yesterday's price. That means that the time sequence of price changes is irrelevant. Furthermore, even the level of yesterday's index becomes irrelevant when the entire calculation is simplified.

In the example to calculate a Value Line Index (I) for the two stocks, A and B, I used the following general equation:

MI!A Journal/February 1986 23

Page 26: Journal of Technical Analysis (JOTA). Issue 23 (1986, February)

For simplicity's sake, I started my index at 100, but to that in a general way, I would have to make use of a constant (K) so that:

100 K=-

(AlBl) l/2

-

Thus my index on day one, besides being 100, could also be expressed as:

I1 = K (AlBl) l/2

But I can also use a new constant (C) such that:

C = K2

Then

I1 = (C Al Bl) u2 . . . . . (2)

Now if I use the general equation (1) to calculate the have:

12 = 11

But substituting for I2 from (2)

Similarly

13 = I2 A3B3 0 v

A2B2

Substituting for I2 from (3) .

I3 =

24

A2B2 A3B3

>

w

P.V

AlBl A2B2 . . . . .(4)

Index for day 2, I

~124 Jourrd/February 1986

Page 27: Journal of Technical Analysis (JOTA). Issue 23 (1986, February)

If we build our general equation (1) from day one, then, we would have:

A2B2 A3B3 AN-PN-1 IN = CAIBl.-.-....

AIBl A2B2 AN-2%-2

But the above equations can be simplified such that:

l/2 .

I2 = K A2 B2) .

I3 = (C A3 B3) w ‘N-1 l = (c Ati-1 s-1) 1’2 IN = (C AN s) J-i2 . . . . . (5)

From this, it is evident that no cumulative error can arise from the method of calculating theValue Line Index. Italsobegs thequestion as to why Value Line bothers to use the previous day's prices. There are probably two reasons for this:

The first has to do with "significant figures". A figure calculated by multiplying the prices of 1700 stocks with each other (or using logarithms) is enormously larger than that for 1700 rates of change (virtually all of them between something like 1.05 and 0.95 and most between 1.01 and 0.99).

The second may well be a case of mathematical fastidiousness. Assuming there are 1,700 stocks in the Index, a three-for-one stock split in one of them, if accounted for by -a change in the constant, would affect that constant by a factor equivalent to the 1700th root of 3, whatever that might be. It could, in fact, be safely ignored, as it could in any unweighted averageof1,700 stocks. But Value Line may well have viewed that idea with distaste, along with an equal distaste in having to calcu- late the constant to enough significant figures to reflect the change. But by using rates of change for their stocks , it becomes a simple matter to change thepreviousday's price so as to reflect the split and go on with the calculation.

Whatever the reason, I understand that Value Line uses equation (1) rather than equation (5). But as we have seen, they would both give exactly the same answer. In any case, a double top at 209, for example, even though the two events might be over two years apart, would still be a double top.

That still leaves the question unanswered as to whether the 209 is a good measure of the market's level. Unfortunately, the answer is a subjective one and trying to remember how to do high school algebra doesn't help. But a computer simulation does.

In this simulation, I have used the same universe of two stocks (A and B) and the same mythical portfolio. Stock A and B each start out at $100. A was allowed to increase by a steady, arithmetic increment and B to decrease

25 MI24 Joumal/E'ebruary 1986

Page 28: Journal of Technical Analysis (JOTA). Issue 23 (1986, February)

were allowed. to return by the same increment to $100. The portfolio was unweighted after each change. That is, if at the start, A moved from $100 to $110, then enough A was sold and B bought so that each represented an investment of $105. In each "trading day", there was one transaction in A and one in B, involving two calculations to unweight the portfolio. At the end of each day, an arithmetically unweighted average (AUA) was calculated using the closing prices of our universe of two stocks. Once $125 and $75 for A and B had been reached and again after their return to $100, the AUA was again calculated and also a Value Line Index (VLI). (Since both stocks started out at $100, the indexing factor was 1). In addition, an index (P) for the value of the portfolio was also calculated for those two points, also initialized at 100. P could be regarded as a weighted index, weighted by the shares held in each of A and B, but since the calculation calls for equal dollar amounts in each stock then it actually is also an unweighted index.

The calculation was done on a Commodore 64. Truncation and possibly the fact that A was always calculated ahead of B led to some suspicion about my own "significant numbers". This particularly showed up after A and B had retuned to 100, when logic dictates that the holding should have been exactly one share in each. But this does not reflect on the practicality of the results. The increment chosen was one-tenth of one cent.

PoF?IFoLIO

S'IWKA . S'I0CKB mAL

SHAE7EsowNED 0.77459985 1.29099084

PRICE! 124.999678 , .75.0003219 193.649456

P 96.8247281

VLI 96.8246668

AUA 96.8247955

MARKEZINDICAT0RS

DISTANCE FROM P

.0000613

.0000674

26 KCAJ oumal/E'ebruary 1986

Page 29: Journal of Technical Analysis (JOTA). Issue 23 (1986, February)

PORTFOLIO

SHARES OWNED

PRICE

STCCKA S'IOCKB

1.00000126 1.00000126

100 100

mAL

200.000252

MARKET 1NDrcAmRS

DISTANCE FROM P

P 100.000126

VLI 100 .000126

AUA 99.996712 .003414

The perceived problem with the VLI apparently arose from using the analogy that if a mythical portfolio of equal amounts of dollars in each of two stocks saw one appreciate by 25 per cent and the second decline by 25 per cent, then a Value Line Index showing that a market made upby a universe of those two stocks showing that the market had declined by over 3 per cent had a downward bias. This Was because both P and AUA would have shown no change. But a more accurate calculation of a truly unweighted portfolio (Table One) shows that when the two stocks did (approximately) that, a decline of over 3 per cent would be what an unweighted index should show. In the analogy, had P and AUA remained constantly unweighted throughout the pair of 25 per cent moves,' they too would have shown a decline of over 3 per cent.

If our mythical portfolio had truly been instantaneously and always unweighted, (ie an infinite numberofcalculations ratherthanlOO,OOO), then it would have ended up worth exactly the same $200 it started out at. Instead (Table Two), it was worth a quarter of one cent more than that which meant that P was out by a factor of one-half of that. Not bad, but it should be noted that the Value Line wasn't even out that much. In fact, it was dead right! For that matter, the AUA was also virtually the same as the VLI.

For practical purposes, all three ended up at 100. This should dispel any notion that the VLI has a downward bias, unless at the same time P and AUA also have a downward bias. This raises the question as to whether all unweighted indexes might have a downward bias, but that is beyond the scope of this paper.

MI!A Journal/R&~ 1986 27

Page 30: Journal of Technical Analysis (JOTA). Issue 23 (1986, February)

The reason that both P and AUA more accurately reflect (in Table One) the more than 3 per cent decline in the market, was due to a factor peculiar to each. By the time A and B had reached (approximately) $125 and-$75, the relative shareholdings in the portfolio of A and B had been adjusted 50,000 times. Each one of these involved reducing the holding in a rising stock (A) and increasing the holding in a declining stock (B). Thus, P, which reflected the total market value of the portfolio, both declined and at the same time more accurately reflected an unweighted portfolio. In the origi- nal analogy, it reflected a portfoliothathapppened to be unweighted at the start, but was a weighted one at the end.

The increased accuracy of AUA was not due to the number of its calculations (25,000), but to a reduction of an error in the calculation itself. In the analogy, when stock A had gone up to $125, that represented a factor of 1.25 (100 x 1.25). But when stock B declined to 75, that represented a factor of the inverse of 1.333 (100/1.333). The calculation for AUA though, treated the two as if they were exactly the same (25 per cent either way). It thus did not reflect the greater proportional downward move. In other words, the error introduced an upward bias. In a move from $100 to $75, the error is significant. In a move from $lOOto $99.999,it is trivial, even though it is cumulative. Thus, as shown in Table One, AUA provides a reasonably accurate answer.

If this thesis is true, it should follow that the greater the number of calculations and the smaller the incremental changes, the greater the accuracy of both P and AUA.

INCREMENTS

-logy 100 100

$25.00 98.437500 100

5.00 97.1477534 97.4582374

1.00 96.8891540 96.9510392

.25 96.8407227 96.8561826

.125 96.8326527 96.8403820

.OOl 96.8246668 96.8247955

A - 125 B - 75

VLI - 96.8245837

P AUA

A - 100 B - 100

VLI - 100

P AUA

103.359375 106.666667

100.668651 101.313152

100.133420 100.261376

100.033339 100.065281

100.016668 100.032635

100.000126 99.996712

28 m'A Jourrdfiebruary 1986

Page 31: Journal of Technical Analysis (JOTA). Issue 23 (1986, February)

The final result for AUA (99.996712) may well have been caused by the truncation operation of the computer. If not, it negates the thesis that AUA has an upward bias.

But what the results do show is confirmation that the more accurately P and AUA are calculated, the closer they approach VLI. Nevertheless, with the exception noted, both P and AUA were consistently above VLI. But since in Table Three it is evident that they can be higher than they should be when meaningful increments are used, the difference is more likely due to errors in P and AUA than in the calculation for the Value Line. Furthermore, if there were a downward bias in the calculation for the Value Line, this should show up in a cumulative, growing error. This does not occur.

CONCLUSIONS

1) The Value Line Index cannot have a cumulative error arising from the means of its calculation.

2) The Value Line Index does not suffer from any downward bias other than some it might share with many other unweighted indexes.

3) For practical purposes, an arithmetically unweighted average (AUA) can be used, but the Value Line Index is more accurate.

1. Fosback, Norman G., Stock Market I&c, (Fort Lauderdale, Florida: The Institute for Econometric Research, second printing, 1977).

2. Ibid: All references are found on pages 293-296.

Don Dillistone is a Technical Analyst with Richardson Greenshields of Winnepeg, Canada. He is a CFA, a member of the MTA and of the Canadian Society of Technical Analysts.

ML24 Jourrd/Ekbruary 1986 29

Page 32: Journal of Technical Analysis (JOTA). Issue 23 (1986, February)

1 REM H IS 2 REtI L IS 3 I331 ‘fl-i IS 4 PEP1 ‘r’L I :Z 5 REP1 id IS rJ REM t-1 IS 7 t?EM D IS 8 REM E IS 9 REP1 ‘y’ 15 18 REM H IS 11 RENT IS 12 REM U If-3 13 REtI !s 1 :s

-

MlTA Joumal/E%?bmary 1986

Page 33: Journal of Technical Analysis (JOTA). Issue 23 (1986, February)

BY Arthur A. Merrill

The definition of stochastics in Webster's New International Dictionarv:

-cs ???

Stochastic: adj. Conjectural; given to , or skillful in conjecturing.

The definition of conjecture:

Conjecture: v.t. To form opinions concerning, on grounds confessedly insufficient to certain conclusion; to indulge in surmise.

Under this definition, most of our work is stochastic.

In econometrics, the term "stochastic" is used to signify formulas that take account of the imprecise nature of economic data. For example, the formula for a definite straight line regression is Y=A + BX. In econometrics, this formula is modified to Y=A + BX + e , where e is an error term (the deviation of Y from the straight line). The modified formula is stochastic; it recognizes the fact that in economic data the figures never lie on a perfect straight line.

The "stochastic& that we see produced in current computer programs is something entirely different. I'm not too happy about the label, since it might be confused with the econometric "stochastic&'. But it seems to be firmly established, so I expect I'll have to learn to like it.

Anyway, the technical "stochastics" is a term that has been applied to the closing price analysis developed by Larry Williams and George Lane. The general idea is that it is a bullish sign when closing prices crowd the upper end of a price range; it's a bearish.sign when prices close low. Larry Williams considers the distance from the low of a price range up to the closing price as a measure of buying pressure; the distance down from the top of the range to the close is a measure of selling pressure.

Here are some details:

Larry Williams uses a very short term range, and smooths with moving averages. The top of his range is the current day's high, or the previous day's low or the previous day's close, whichever is lower. He calculates buying pressure (as measured by the distance from the low of the range up to the close) as a percent of the range. He calculates 7 day, 14 day and 28 day moving averages, then averages the three averages to get his "ultimate oscillator."

31 M!m Journal/February 1986

Page 34: Journal of Technical Analysis (JOTA). Issue 23 (1986, February)

George Lane's "stochastic" uses the range based on the last five days. His formulas:

%K=lOO(C-L)/(H-L) where C= closing price

H= highest price in last five days L= lowest price in last five days

%D= 3 day moving average of %K * slow %D= 3 day moving average of %D

* - this is essentially the case, although George's calculation is slightly

different from a straight moving average. He totals 3 days of the numer- ator of %K and then divides by 3 days of the denominator.

Scientific Market Systems uses a range based on the highs and lows in the last 20 days. They use formulas similar to George Lane's for %K and %D.

Gerald Appel likes to use a shorter term range with 13 units instead of 20.

This is just a brief outline topromote understanding of the term "sto- chastic&'. For detailed suggestions for interpretation of the figures, you can consult the experts:

Larry Williams, P.O. Box 1781, Kalispell, MT 55901

George C.Lane, Investment Educators, P.O.Box 2354, Des Plaines, IL 60016

Scientific Investment Systems Inc., 62 Wellesley Street, Suite 503, Toronto,- Canada M5S 2x3'

Gerald Appel, Signalert Corp., 150 Great Neck Road, Great Neck, New York 11021

-

Arthur Merrill is a member of the MTA and the head of Merrill Analysis, Inc., Chappagua, New York.

32 MI24 Joumal/E'ebruary 1986

Page 35: Journal of Technical Analysis (JOTA). Issue 23 (1986, February)

IYWRKETTIMING: AN EMPIRICAL l5!smmTIoN

Eiy Michael J. Flanagan, Ph.D.

SYNOPSIS: Several quantitative technical analysis models were used to determine if returns superior to a buy and hold strategy were obtainable with aggressive growth mutual funds using a transaction timing strategy. The results of various empirical tests of transaction timing using technical trading models indicated that the major advantage of transaction timing lies in avoiding losses during bear market declines and participating in the early stages of bull market rallies.

This article is based upon the author's doctoral dissertation, submitted to Golden Gate University, San Francisco,California. The major purpose of the dissertation study was to critically evaluate the applications of several quantitative technical analysis models to the problem of stock transaction timing. The study tested the hypothesis that past price behavior can be used to obtain returns superior to a buy-and-hold policy. Concurrently, the study tested the null hypothesis that the "weak form" of the efficient market hypothesis, also known as the random walk, was true. The study also sought to evaluate if a personal computer program could be designed to optimize the parameters of trading strategies based on technical analysis of historical price data, and if such trading strategies that were optimized after-the-fact could be applied successfully to future price data.

Several major findings of this research can be spotlighted.

1. The investment approach of buy-and-hold in bull markets and holding cash in bear markets was explored using the top performing aggressive growth funds during the 1971-80 period. A transaction-timing strategy approach as compared tobuy-and-hold was impressive and ranged from 135 percent for the nonvolatile Janus Fund to 450 percent for the highly volatile 44 Wall Street Fund.

2. Trading strategies incorporating trend following and character of Market Models (trend/countertrend) surfaced as superior while conducting the empirical evaluation. Successful character-of-market model signals will usually generate good returns , even though they do not possess the inherent robustness of trend-following models. If a signal is missed, the trend-following model always serves as a backup and ensures that the trader is not on the wrong side of the market during strong price trends.

3. Trend-following systems that are based on a single parameter moving average should use a smoothing factor value in the 0.01 - 0.04 range to minimize trading whipsaws (equivalent to a moving average length of 100 to 400 days). Abullish outlook should be assumed if a market index such as the Standard and Poor's 500 exceeded its long-term (smoothed) trend. A bearish outlook ought to be assumed if the index was less than the long- term trend.

33 MI'A Journal/February 1986

Page 36: Journal of Technical Analysis (JOTA). Issue 23 (1986, February)

4. The performances of four investment strategies in four different environments were computed. The following strategies were used: (a) buy-and-hold policy; (b) buy-and-hold policy except during bear markets when all equity holdings are liquidated to cash or equivalents; (c) buy- and-hold policy during bull markets and transaction timing during bear markets; and (d) unrestricted transaction timing.

The following investment environments were assumed: (a) investment vehicles with notransactioncosts (e.g., no-load mutual funds) in a tax- sheltered environment (e.g., pension plans); (b) investment vehicles with no transaction costs in a non-tax-sheltered environment; (c) investment vehicles with transaction costs (e.g., common stocks, options, stock index futures, commodities, etc.) in a tax-sheltered environment; and (d) investment vehicles with transaction costs in a non-tax-sheltered environment.

-

The returns obtained with each strategy/environment combination showed that unrestricted transaction timing is the most profitable strategy with no- load mutual funds in a tax-sheltered environment, while restricted transaction timing (buy-and-hold during bull markets and transaction timing during bear markets) is the most profitable strategy.in the other three environments.

5. A personal computer program can be successfully designed to automatically optimize the parameters of stock market trading strategies based on technical analysis and historical price data.

-. 6. Trading strategies that are optimized ex post (after the

fact) on historical price data can be applied to fu=re price data to produce returns in excess of'those obtainable with a buy-and-hold policy.

7. The empirical results obtained in this study suggest that the random walk hypothesis does not adequately describe the behavior of auction market prices. The test data showed significantly superior returns with transaction timing models compared to a buy-and-hold policy which does not support the random walk hypothesis.

Technical analysis is based solely on the past behavior of companies or of the market, using such indicators as price,volume, odd-lot trading, ad- vance-decline ratios, and the like. Technical analysis rests on the as- sumption that the forces of supply and demand are the sole determinants of the price of a stock. Changes in the balance between supply and demand produce trends which can be best detected by the various technicalindi- caters.

Trading strategies that employ technical analysis belong to one or more of the following five classifications: trend following, character-of-market analysis, price chart analysis, indicator analysis and structural theories, such as wave principles and time cycles. The trading strategies developed

34 PEA Joumal/i?ebruary 1986

Page 37: Journal of Technical Analysis (JOTA). Issue 23 (1986, February)

for this study were based on objective technical analysis techniques that belong to the trend-following and/or character-of-market analysis classifications. These strategies provide nonambiguous trading signals and can be processed automatically on a digital computer.

Trend-following models, such as moving averages, perform best in strongly trended markets but tend to be whipsawed in trading range markets. In contrast, character-of-market models (momentum, osscillator, velocity- acceleration, excessive swing , and channel reversal models) perform well in tranding ranage markets and poorly in strongly trended markets. Whereas trend-following models confirm price reversals , character-of-market models attempt to anticipate such reversals. As the name suggests, trend/no trend models attempt to take advantage of the strengths of trend-following and - character-of-market models while avoiding most of the weaknesses of each. The development of a successful trend/no trend trading model is central to obtaining a robust, automatic indicator that signals the transition from a trading range to a trend-following market and vice versa.

All trading strategy development work was conducted on the Pennsylvania Mutual Fund price data for the period January 1973 through December 1977. This phase was used to test the program logic of each of the trading strategies and determine practical parameter range and incremental values for each strategy. The following strategies were included in the evaluation:

OSMA- Oversold Moving Average

ESMA - Excessive Swing Moving Average

FEMA- Filtered Exponential Moving Average

T/CT - Trend/Countertrend

EMA - Exponential Moving Average

FLTR - Filter Rule

CENA(l)Krossing Multiple Moving Average

m(2)-Confirming mltiple Moving Average

The second phase repeated the first phase parameter range tests on 44 Wall Street Fund data for the period January 1981 through December 1982. The performances of the eight trading strategies selected for evaluation are summarized in Table 1. As indicate,d three of the strategies (OSMA, ESMA, and T/CT) outperformed the other five. Because OSMA was a subset of ESMA, T/CT and ESMA were retained for use in all subsequent empirical tests.

MI!A Joumal/E'ebruary 1986 35

Page 38: Journal of Technical Analysis (JOTA). Issue 23 (1986, February)

TABIEl

EX POST PERFORMANCE SWNARY--ALL STRATEGIES FOR THE YEAR 1982: 44WAILSTREFTFUNJJ

Test Period--01/03/81 to 12/23/82 Signal-Transaction Belay Period : la

Trading Strategyb

OSMA

Best Parameters'

#l/O.03 #2/1.20 #3/0.06

l?E?turn 6)

75.79

ESMA #l/O.03 75.79 #2/1.20 #3/0.06

#l/O.08 50.96 #2/4.50

T/t= #l/O.03 88.86 #2/0.20 #3/0.05

EMA’ #l/O.02 42.82

FLTR #l/O.80 68.82 #2/6.00

m(1) #l/O.15 37.67 #2/0.05

m(2) #l/O.05 46.22 #2/0.05

BUY/HOLD -10.29

aAll trading strategies are optimized to produce the best investment return "after the fact,"

bRefer to the text for a description of the strategy acronyms.

=#l, #2, etc. refer to the parameter number. The adjacent number is the value selected by the optimi- zation program that produced the best return.

36

Page 39: Journal of Technical Analysis (JOTA). Issue 23 (1986, February)

The Excessive Swing, Moving Average Model (FSMA)

The excessive swing, moving average model (ESMA) uses three trading system ccmqonents:

1. A trend-following model using a single exponential moving average with a price-velocity-based deadband on the price/moving average crossover.

2. An excessive swing, moving average model to detect oversold and overbought conditions and antici- pate a price reversal.

3. Smoothed price velocity and acceleration to enable and disable the excessive swing model.

A typical trading pattern of the ESMA Model is shown in Figure 1.

ESMA includes all of the features and decision rules of the oversold, moving average model [OSMA] and provides the additional feature of generating trading signals based on the detection of overbought conditions. With the exception of appropriate changes in sign, the operation of the overbought and oversold subsystems is identical.

P~II Jcxirnd/Febr~ary 1986 37

Page 40: Journal of Technical Analysis (JOTA). Issue 23 (1986, February)

FIGuRE 1

A TYPICAL EXCESSIVF, SWING/MOVING AVERAGE TRADING PATTERN [FSMA]

In some instances during an oversold market condition, a buy signal will not be given by the excessive swing model because one or more of the signal requirements are not met. In this case, the trend-following model will serve as a backup, although with a loss of potential gain. It should be noted that if a price uptrend persists, a buy signal will always be rendered by the trend-following model.

38 MI'A Joumal/F&ruary 1986

Page 41: Journal of Technical Analysis (JOTA). Issue 23 (1986, February)

The Trend/Countertrend Model

The trend/countertrend (T/CT) model uses three trading system components:

1. A trend-following model using a single exponential moving average.

2. A smoothed price velocity &el.

3. A minimum/maximum price model that monitors smoothed price minimums and maximums.

A typical trading pattern is shown in Figure 2.

mx Jou.rnal/E'eb~ 1986 39

Page 42: Journal of Technical Analysis (JOTA). Issue 23 (1986, February)

FIGURE 2

ATYPICALTREND/COWEKEEND TRADING PATTERN [T/CT]

In some instances during an oversold market condition, a buy signal will not be given by the minimum/maximum price model because one or more of the signal requirements are not met. In this case, the trend-following model will serve as a backup, although with a loss of potential gain. It should be noted that if a price uptrend persists, a buy signal will always be rendered by the trend-following model.

40 MllA JournalDeb- 1986

Page 43: Journal of Technical Analysis (JOTA). Issue 23 (1986, February)

TRANSACTION TIMING USING TEXXNICAL TRADING MCDELS

This section presents the results of various empirical tests of transaction timing using technical trading models. Several precautions were taken to make the test results valid. One was a procedure to separate the development/optimization of trading strategies from the actual empirical testing of the trading strategies. This precaution, coupled with the constraints of personal computer usage, led to a focus upon an exhaustive simulation of the trading of a single highly volatile mutual fund (44 Wall Street) money market fund pair, followed by a more restrictive test of a second fund (Janus/Money Market fund pair). These mutual fund data were carefully selected and processed on an Apple II computer. A further precautionary step was to establish criteria and statistical tests in advance for evaluating the comparisons.

A reader wishing to study the research methodology more closely or anyone who wishes to replicate this study is invited to refer to the original dissertation entitled "Stock Market Transaction Timing: AnEmpirical Eval- uation" (Golden Gate University, 1983).

Test Results: Ex Post Determination of Trading Strategy Parameters, Using the 44 Wall Street Fund

A total of eighteen twelve-month test subperiods and one six-month subperiod were used to test the two trading strategies (ESMA and T/CT) selected in the phase l/2 test. The test subperiods, the strategy para- meters, and the investment performance of each strategy for each test subpericd are given in Table 2. The best performing strategy in each test period is marked with an asterisk. This strategy and its associated para- meters were used in the ex ante tests on contiguous twelve-month test periods, as described in the following paragraph.

41 m Journal/E&nX=y 1986

Page 44: Journal of Technical Analysis (JOTA). Issue 23 (1986, February)

TABLE2

EX POST DEX'ElWINATIONOF PV:44wALlL

TRADING STRAm SrzzEFTFuND

Test Subpericd

ID# Dates Strategy

NElIlE Parameters

Trading Strategy Perform ("a)

A1A Ju1'69-Dec'69 ESMA .03 / 1.7 / .Ol -7.80

Al Ju1'69-Jun'70

A2 Jan-tkc '70

A3 Jan-Dee '71

A4 Ju1'71-Jun'72

A5 Jan-Dee '72

A6 Jan-Dee '73

A7 Ju1'73-Jun'74

A8 Jan-Dee '74

A9 Jan-Dee '75

A10 Ju1'75-Jun'76

All Jan-Dee '76

Al2 Jan-Dee '77

A13 Ju1'77-Jun'78

Al4 Jan-Dec'78

Al5 Jan-Dec'79

Al6 Ju1'79-Jun'80

A17 Jan-Dec'80

Al8 Jan-Dec'81

A19 Ju1'81-Jun'82

42

*T/CT .03 /.245 / .04 2.45 ESMA .03 / 1.7 / l Ol -11.08

*T/CT .07 /.275 / .05 -3.47 "ESMA .03 / 1.6 / .07 88.74 T/a .04 /.165 / .04 74.35

*EsMA .03 / 1.5 / .07 47.26 T/@ .03/ .09 / .05 55.63

FSMA "T/CT ESMA

*T/CT ESMA

"T/CT "ESMA

T/m *E&IA

T/m ESMA

"T/CT ESMA

*T/CT *EsMA

T/CT ES

3 *T/ *EsMA T/m ESMA

*T/CT ESMA

"T/CT

*T/CT ESMA

*T/CT ESMA

*T/Cl? ESMA

.05 / .9 / .09 48.65 .07 /.09 / .03 61.60 .05 / 1.5 /.09 5.88 .03 /.15 / .02 4.83 .07 / 1.7 /.Ol 33.27 .07 /.245 /.05 38.83 .05 / .7 / .07 30.00 .07 /.18 / .05 24.32 .04 /1.2 / .06 3.81 .09 /.125/ .03 3.76 .03 / .7 / .08 108.51 .08 /.135/ .03 114.51 .05 / l.l/ .08 37.11 .03 /.135 /.03 39.62 .06 / 1.7 /.Ol 22.56 .02 /.135 /.04 20.56 .02 / 1.5 /.08 6.78 .25 /.105/ .07 7.47 .06 / 1.7 /.Ol 37.73 .25 /.105 /.06 27.39 .03 / .9 /.07 59.55 .60 /.105 /.07 78.48 .02 / 1.1 /.Ol 42.96 .25 /.105 /.08 41.31 .05 / 1.0 /.07 44.28 .40 / .12 /.04 57.74 .05 / 1.1 /.07 66.96 .40 /.105 /.04 67.14 .04 / 1.0 /.06 11.54 .40 / .21 /.05 24.10 .lO / 1.7 /.Ol 2.06

~!I24 Jourrd/February 1986

Page 45: Journal of Technical Analysis (JOTA). Issue 23 (1986, February)

TABLE 2, mnt'd.

EX POST DEIERMINATION OF TRADING STRATEGY PV: 44wALL.ziTFamFLJND

Test Subperiod

ID# Dates

A20 Jan-Dec'82

Trading Strategy Strategy

NZU-IX? Parameters Perform (%)

*ESMA .06 / .7 /.09 72.46 T/m .30 / .x05/ .05 60.56

aT/CT is not selected because the number of transactions is excessive. This is caused by the low value of parameter #2 and provides a warning that the lower-limit value should be increased.

bA modification was made to the T/CT decision rules to reduce whipsaws during oversold markets conditions. This was done by optimizing the moving average smoothing factor used in the detection of price maximums and minimums.

*Best performing strategy.

Test Results: Ex Ante Simulated Trading Performance, Using the 44 Wall Street Fund

The results of the ex ante trading simulations using twelve-month test subperiods are shown in Tables 3. Table 3 shows that one of the major benefits from transaction timing is obtained by avoiding the price declines that occur during bear markets (see 1973,1974, and 1981). The following observations can be made from the results in Table 3.

1. A transaction timing strategy is generally superior to a buy- and-hold policy during long-term price downtrends (bear market years such as 1973, 1974 and 1981) and during the early stages of a long-term price uptrend (bull market years such as 1970 and 1982).

2. During bull market periods, a buy-and-hold policy is generally superior totransactiontiming (see1975, 1976, 1977, 1979 and 1980). Bull market corrections commonly occur with an initial sharp drop and are eventually reversed by a sudden uptrend. Technical analysts generally agree with the random walk theorists .that short-term price movements are essentially random. By the time the trader reacts to sudden changes in price, a substantial portion of the loss/gain has often been made.

m Joumal/Ekbruary 1986 43

Page 46: Journal of Technical Analysis (JOTA). Issue 23 (1986, February)

TABLE3

44

SIMULA~ TRADINGPERFORMANCEUSINGA?WELVE-MONTH EX POST OPTIMIZATION IN'lXRVAL AND A TWELVFrMONTH

EXANIETESTSUE3PEXIOD: 44WALLS~FUND

Test Subperiod

Annual Performance (%)

Assoc. Buy Strategy Ex Post Ex Ante Ex Post and

ID # Year Name Subperiod Trading Trading Hold

A2 1970 T/m

A3 1971 ESMA

A5 1972 ESMA

A6 1973 T/m

A8 1974 T/m

A9 1975 FSMA

All 1976 T/m

Al2 1977 ESMA

A14 1978 T/m

Al5 1979 T/a

Al7 1980 T/e

=a 1981 T/m

A20 1982 T/m

AlA 69.65 74.35 14.84

A2 41.38 47.26 57.95

A3 -.49 4.83 -12.30

A5 37.84 38.83 -38.72

A6 -14.86 3.76 -56.85

A8 103.10 108.51 106.63

A9 9.17 20.56 15.38

'All -6.54 6.78 13.27

A12 73.30 78.48 43.34

Al4 25.53 41.31 49.93

Al5 25.59 67.14 32.92

Al7 6.69 24.10 -16.58

=a 46.27 60.56 la.40

aThe large discrepancy between the ex ante and ex post performances could be reduced by performing ex post optimi- zation runs more frequently than once a year.

Page 47: Journal of Technical Analysis (JOTA). Issue 23 (1986, February)

Test Results : Ex Ante Simulated Trading Performance, Usina the Janus Fund

The ex post and ex ante performance results using the Janus Fund are given in Tables 4 and 5, respectively. As anticipated, a fund such as Janus with limited volatility is not a suitable candidate for trading. Table 5 shows that the buy-and-hold policy produced superior returns in two of the four years. By inspection, the best strategy would have been to follow a buy- and-hold policy except during the 1973-74 bear market.

EXPOSTDFX!EXMINATIONOF TRADING STFWIEGY PV: JANUSFUND

Test Subperiod

ID# Dates

Trading Strategy Strategy

NiXIT Parameters Perform (%)

Al Jan-Dee '71 Esm .02/ .9 /.04 33.86 T/m .06/.045/.03 40.36

A2 Jan-Dee '72 ESMA .03/ 1.2/.06 23.12 T/CT .03/.185/.03 22.02

A3 Jan-Dee '73 FSMA .057/1.55/.09 .62 T/CT .07/ .08/.02 5.23

A4 Jan-Dee '74 ESMA T/m

.l/ 1.5/.09 .08/ .22/.03

-1.62 -2.36

45 MI'24 Joumal/Febmary 1986

Page 48: Journal of Technical Analysis (JOTA). Issue 23 (1986, February)

Tzmx5

SIMULATHD TRADING PERFORMANCE USING A TWELVE-I0?I'H M POST OPTIMIZATION INTERVAL AND A 'IWHLVE- MONI'HMAN'IETESTSUBPERIOD: JANDSFUND

Test Subperiod

Performance (05)

Assoc. BUY Strategy Ex Post Ex Ante ExPost and

ID# Year Name Subperiod Trading Trading Hold

A2 1972 T/a Al 18.65 22.02 21.99

A3 1973 T/m A2 -2.48 5.23 -12.63

A4 1974 T/m A3 -4.10 -2.36 -10.77

A5 1975 ESMA A4 -4.70 - 12.18

- Test Results: Evaluation of a Buy-and-Hold Policy in

Bull Markets and Holding Gash During Bear Markets

The investment approach of buy-and-hold in bull markets and holding cash in bear markets using the twenty top-performing aggressive growth funds during the 1971-80 period is summarized in Table 6. The returns using.this ap- proach as compared to buy-and-hold are impressive and range from 135 percent for the nonvolatile Janus Fund to 450 percent for the highly volatile 44 Wall Street Fund.

46 PfRi Jourd./Febm 1986

Page 49: Journal of Technical Analysis (JOTA). Issue 23 (1986, February)

BUY-AND-HOID PERFORMANCEOFTHETWENTY TOP-PERFORMING FUNDS DURING 1971-80 IN A TAX-SHELTRRED RNVIRONMEXI' AND

HOLDING CASH DURING THE 1973-74 BEAR MARKE!T

Aggressive Growth No-Ioad

Percentage Change in NAV per Share with Distribution Added Back for 1971-80 Perioda

Buy-and-Hold Buy-and-Hold Percentage ~ Funds Policy Policy; Cash Improvementb

Held During Over Buy-and 1973-74 Hold Policy

Twentieth Century Growth

44 Wall Street Janus Weingarten Quity Hartwell Leverage Acorn Stein Roe Special Constellation Growth ma Sequoia Founders Special Nicholas Afuture Mathers St. Paul Special Naess and Thomas Rxplorer Hartwell Growth Columbia Growth Lexington Growth

935 1627 174 518 2331 450 390 526 135 379 808 213 378 662 175 347 689 199 321 680 212 312 568 182 292 514. 176 279 498 178 264 462 175 259 1037 400 257 792 308 257 674 262 251 460 183 241 422 175 235 594 253 234 587 251 210 433 206 209 602 288

Average 328 748 228'

aThese figures give an approximation of the funds' relative volatility. Note that 44 Wall Street and Janus have the highest and lowest figures respectively.

bE ven if round-trip transaction costs and taxes on long-term capital gains are included, the return is still superior to a buy-and-hold policy. Refer to Table 9, which shows a 299 percent improvement over a buy-and-hold policy when taxes and transaction costs are included and the equity position is liquidated at the end ,of the period.

IWA Journal/February 1986 47

Page 50: Journal of Technical Analysis (JOTA). Issue 23 (1986, February)

Significance Tests on Selected Trading Simulations

Significance tests were performed on selected trading simulations as follows:

(see Table';). Buy-and-hold versus transaction timing for the period 1970-82

2. Buy-and-hold versus limited transaction timing using the twenty top-performing funds for the 1971-80 period (see Table 8).

Statistically significant results were obtained in both cases, meaning that the null hypothesis that a buy-and-hold policy is superior to transaction timing during the test periods must be rejected. It should be noted that statistical conclusions cannot be drawn about the performance of transaction timing versus buy-and-hold during time periods not included in the tests.

SIGNIFICANCE TEST RESULTS FOR COMPARING A BUY-AND-HOLD POLICY AND A TRANSACTION TIMING STRATEGY:

44 WALL S- FUND DURING 1970-82

Test Subpericd

ID# Year

Annual Return Difference

Buy and Transaction Between Hold Timing Returns P(bh) p (W W

A2 1970

A3 1971

A5 1972

A6 1973

A8 1974

A9 1975

All 1976

Al2 1977

Al4 1978

A15 1979

Al7 1980

48

.1484 .6965 -.5481

.5795 .4138 .1657

-.1230 -.0049 -.1279

-.3872 .3784 -.7656

-.5685 -.1486 -.4199

1.0663 1.0310 .0353

.1538 .0917 .0621

.1327 -.0654 .1981

.4334 .7330 -.2996

.4993 .2553 .2440

.3292 .2559 .0733

M.I'A Journal/E'&ruary 1986

Page 51: Journal of Technical Analysis (JOTA). Issue 23 (1986, February)

'mFm?a 7, ccmt'd.

Test Annual Return Subperiod Difference

Buy and Transaction Between ID# Year Hold Timing Returns

p (bh) p (W (d)

A18 1981 -.1658 .0669 -.1327

A20 1982 -.2787

Mean Value

Variance .089

Standard Deviation .298

t-Statistic -1.670

m Jou.rrd/Febnx=y 1986 49

Page 52: Journal of Technical Analysis (JOTA). Issue 23 (1986, February)

SIGNIFICANCE TEST RESULTS FOR WARING A BUY-AND-HOLD POLICY AND A LIMITED TRANSACTION TIMING STRATEGY: THE TOP

TWENTY BEST-PERFORMING FUNDS DURING 1971-80

Aggressive Growth No-Load Funds

Change in NAV per Share with Distribution Added Back for 1971-80 Period

Buy-and-Hold Buy-and-Hold Difference Policy Policy; Cash Between

Held During Returns

Twentieth Century Growth

44 Wall Street Janus Weingarten Equity Hartwell Leverage Acorn Stein Roe Special Constellation Growth Q=ga Sequoia Founders Special Nicholas Afuture Mathers St. Paul Special Naess and Thorns Explorer Hartwell Growth Columbia Growth Lexington Growth

Mean

9.35 16.27 -6.92 5.18 23.31 -18.13 3.90 5.26 -1.36 3.79 8.08 -4.29 3.78 6.62 -2.84 3.47 6.89 -3.42 3.21 6.80 -3.59 3.12 5.68 -2.56 2.92 5.14 -2.22 2.79 4.98 -2.19 2.64 4.62 -1.98 2.59 10.37 -7.78 2.57 7.92 -5.35 2.57 6.74 -4.17 2.51 4.60 -2.09 2.41 4.22 -1.81 2.35 5.94 -3.59 2.34 5.87 -3.53 2.10 4.33 -2.23 2.09 6.02 -3.93

-4.12

Variance 12.85

Standard Deviation 3.59

t-Statistic -5.13

50 MA Joumal/Febm 1986

Page 53: Journal of Technical Analysis (JOTA). Issue 23 (1986, February)

Comparative Performance of Alternative Investment Strategies

Based on the data presented in Table 3, the performance of alternative strategies in different investment environments can now be compared. The following strategies were used:

1. Buy-and-hold policy.

2. Buy-and-hold policy except during bear markets when all equity holdings are liquidated to cash or equivalents.

3. Buy-and-hold policy during bull markets and transaction timing during bear markets.

4. Unrestricted transaction timing.

The following investment environments were assumed:

1. Investment vehicles with no transaction costs (e.g., no-load mutual funds) in a tax-sheltered environment (e.g., pension plans).

2. Investment vehicles with rx> transaction costs in a non-tax- sheltered environment.

3. Investment vehicles with transaction costs (e.g., common stocks, options , stock index futures, commodities, etc.) in a tax-sheltered environment.

4. Investment vehicles with transaction costs in a non-tax- sheltered environment.

Table 9 summarizes the returns of each strategy/environment combination and indicates that unrestricted transaction timing is the most profitable strategy with no-load mutual funds in a tax-sheltered-environment, while restricted transaction timing (buy-and-hold during bull markets and trans- action timing during bear markets) is the most profitable strategy in the other three environments.

51 m.zi Journal/February 1986

Page 54: Journal of Technical Analysis (JOTA). Issue 23 (1986, February)

COMPARATIVEPERFORMANCEOFINVFSTI%NTSTRAT%IESIN DIFFEREETINWSTMEWENVIRONMEME: 44WALLSTREETFUND

Performance Improvement Over the 1970-82 Period in Different Investment Environments(

Tax-Sheltered Non-Tax-Sheltered

Investment Strategy

Transaction Costs

No Yes

Transaction Costs

No Yes

Buy-and-hold 294 291 276 274

Restricted trans- action timing (buy-and-hold

during bull mar- kets; hold cash during bear markets 173-741

1086 1157 927 820

(369)b (398) (336) (299)

Restricted trans- action timing (buy-and-hold

during bull mar- kets; transaction timing during bear markets [73-741

1189 1062 1065 799

(404) (365) (386) (284)

s ~

Unrestricted transaction timing between a long position in equities and cashC

2190 762 537 273

(745) (262) (195) (100)

aOnly eleven months of every year were used in the simulated trading tests. The results presented here cannot be compared to the results in Table 6.

bperformance.improvement compared to buy-and-hold.

'If short selling is possible (e.g., with stocks, futures, etc.), the performance improvement is poten- tially greater.

52 MllA Journal/E'ebrua.ry 1986

Page 55: Journal of Technical Analysis (JOTA). Issue 23 (1986, February)

CONCLUSIONS AND RECOMMENUATIONS

The findings of this research study led to the rejection of the null hypothesis that transaction timing is no better than a buy-and-hold strategy. Hence, the conclusion is drawn that a buy-and-hold policy can be beaten using a transaction timing strategy based on technical analysis. Furthermore, the empirical test results do not support the random walk hypothesis.

It is remembered that technical analysis considers the action of the market only and is not based on any economic theory. The empirical evidence presented herein and elsewhere shows that quantitative technical analysis is a viable technique. Because quantitative technical analysis is based on time series analysis and not on the economic laws that explain price movements, the trader must periodically subject his trading system to a testing regime that will permit his system to profitably mimic the unknown economic laws.

This dissertation was based on the premise that better than average returns can be obtained by the few investors who pursue a disciplined and rational approach to stock market investment. The following general recommendations can be offered to investors interested in transaction timing.

1. Trend following systems that are based on a single parameter moving average should use a smoothing factor value in the 0.01 - 0.04 range to minimize trading whipsaws.

2. Investors should' not maintain any long equity positions during bear markets.

3. .If the investment environment is restricted to the consideration of transaction costs and the tax sheltering of transactions, then four combinations emerge as shown in Table 10. Based on the study results, four investment strategies are rated for profitability in each of the four environments. As indicated in Table 10, the strategies are ar- ranged in order of increasing complexity. A buy-and-hold policy is the simplest strategy because it involves only two timing decisions: when to invest in the market and when to liquidate market assets. On the other end of the spectrum, unrestricted transaction timing requires the trader's constant attention in order to achieve superior returns. For this reason, the average investor should probably pursue a strategy of investing long during bull markets and holding cash or equivalents during bear markets. Investors who are willing to devote above average time to their investment portfolio should consider transaction timing during the final stages of a bear market so as to participate in the initialuptrend of the eventual bull market, a period that typically yields extraordinary returns. Finally, unrestricted transaction timing is recommended for the serious investor, but only for tax-sheltered, transaction-free investment vehicles.

4. Trading strategies that are designed to detect and profit from short-term market swings (say 10 percent) should be periodically

53

m Journal/February 1986

Page 56: Journal of Technical Analysis (JOTA). Issue 23 (1986, February)

optimized for adaption to the current market climate.

5. The main disadvantage of using only historical price data in a trading strategy is that such a strategy is constrained to following rather than explaining price behavior.

The increasing use of personal computers should encourage a wide variety of research into transaction timing techniques. In particular, the

use of technical indicators with a proven forecasting record (e.g., The Odd

Lot Short-Sales Ratio) coupled with more sophisticated trading strategies (e.g., expert system-based), has potentialforimprovingtradingperfor- mance.

54 Ml7i JournalDeb- 1986

Page 57: Journal of Technical Analysis (JOTA). Issue 23 (1986, February)

-ED TRADING STRATEGIES IN INWSMEXWENVIRO~~S

DIFFERENT

Investment Investment Rnvironmnts Strategy (arranged in order of increasing complexity)

Tax-Sheltered Non-Tax-Sheltered

Transaction Costs Transaction Costs

No Yes No Yes

Buy-and-hold No No No No

Restricted trans- action timing (buy-and-hold during bull mar- kets; hold cash during bear markets

Yes

*

Yes

***

Yes

**

Yes

***

:.. Restricted trans- action timing (buy-and-hold during bull mar- kets; transaction timing during bear markets

Yes **

Yes ***

.

Yes Yes *** ***

Unrestricted transaction timing between a long position in equities and cash

Yes ***

No No No

Unrestricted transaction timing between long and short equity positions

Not tested in study

m: *--Good; **--Better; ***--Rest.

PfR4 Journal/February 1986 55

Page 58: Journal of Technical Analysis (JOTA). Issue 23 (1986, February)

Books

Alexander. Sidnev S. "Price Movements in Speculative Markets: Trends or Random Walks, No. 2." In The Random Character of Stock Market Prices, pp. 199-218. Edited by Paul H. Cootner. Cambridge, Mass.: M.I.T. Press, 1964.

Appel, Gerald. No Load Switch Fund Scalpers Manual. New York: Windsor Books, 1979.

Barnes, Robert M. Commodity Profits Through Trend Trading. New York: John Wiley & Sons, 1982.

Taming the Pits: A Technical Approach to Commodity Trading. New-York: Ronald Press, 1979.

Bronowski, Jacob. The Common Sense of Science. New York: Vintage Books, circa 1975.

Brown, Robert G. Smoothing, Forecasting, and Prediction of Discrete Time Series. Englewcod Cliffs, N.J.: Prentice-Hall, 1962.

Cleary, James P. and Hans Levenbach. The Professional Forecaster. Belmont, Calif.: Lifetime Learning Publications, 1982.

Cobleigh, Ira U. and Bruce K. Dorfman. The Dowbeaters. New York: Macmillan Publishing Co., 1979

Cootner, Paul H. "Stock Prices: Random vs. Systematic Changes." In The Random Character of Stock Market Prices, pp. 231-52. Edited by Paul H. Cootner. Cambridge, Mass.: M.I.T. Press, 1974

, ed. The Random Character of Stock Market Prices. Cambridge, Mass. : M.I.T. Press, 1964.

Editors of the No-Ioad Fund Investor. Handbook for No-Ioad Fund Investors. 1981 and 1982 eds.

Edwards, Robert D. and John Magee. Technical Analysis of Stock Trends. Springfield, Mass.: John Magee, 1966.

Elton, Edwin J. and Martin J. Gruber. Modern Portfolio Theory and Invest- ment Analysis. New York: John Wiley & Sons, 1981.

Fama, Eugene F. "Efficient Capital Markets: A Review of Theory and Empir- ical Work." In Modern Developments in Investment Management, pp. 109- 53. Edited by James Corie and Richard Brealey. New York: Praeger Publishers, i976.

56 HI'A Journal/E'ebruary 1986

Page 59: Journal of Technical Analysis (JOTA). Issue 23 (1986, February)

Fischer, Robert. Stocks or Options? Programs for Profits. New York: John Wiley & Sons, 1980.

Fosback, Norman G. Stock Market Logic. Fort Lauderdale, Fla.: Institute for Econometric Research, 1976.

Granger, Clive. "Empirical Studies of Capital Markets.: In Mathematical Models in Investment and Finance, pp. 201-25. Edited by H. Szego and H. Shell. New York: North-Holland, 1972.

Granger, Clive W.J. and Oskar Morgenstem. Predictability of Stock Market Prices. Lexington, Mass.: D.C. Heath & Co., 1970.

Granville, Joseph E. New Key to Stock Market Profits. Englewood Cliffs, N.J.: Prentice-Hall, 1963.

Hagin, Robert L. The Dow Jones-Irwin Guide to Modern Portfolio Theory. Homewood, Ill.: Dow Jones-Irwin, 1979.

Hayes, Michael. The Dow Jones-Irwin Guide to Stock Market Cycles. Home- wood, Ill.: Dow Jones-Irwin, 1977.

Kaufman, P.J. Commodity Trading Systems and Methods. New York: Wiley- Interscience, 1978.

19i9. Technical Analysis in Commodities. New York: Ronald Press,

Levy, Robert R "Conceptual Foundations of Technical Analysis." In Read- ings and Issues in Investments, pp. 207-13. Edited by Frank K. Reilly. Hinsdale, Ill.: Dryden Press, 1975.

Lorie, James and Richard Brealey, eds. Modern Developments in Investment Management. New York: Praeger Publishers, 1976.

MacDonough, Edward P. Trading Commodities by Microcomputer. Nashville, Term. : Trading Technology, 1980.

Malkiel, Burton G. A Random Walk Down Wall Street. New York: W.W. Norton & Co., 1981.

Markowitz, Harry M. Portfolio Selection: Efficient Diversification of In- vestments. Cowles Foundation Monograph no. 16. New Haven: Yale University Press, 1959.

Merrill, Arthur A. Filtered Waves--Basic Theory. New York: Analysis Press, 1977.

Montgomery, Douglas C. and Lynwood A. Johnson. Forecasting and Time Series Analysis. New York: McGraw-Hill Book Co., 1976.

57 MI!A Jaxnal/F&ruary 1986

Page 60: Journal of Technical Analysis (JOTA). Issue 23 (1986, February)

Pring, Martin J. Technical Analysis Explained. New York: McGraw-Hill Book Co., 1980.

Reilly, Frank K., ed. Readings and Issues in Investments. Hinsdale, Ill.: Dryden Press, 1975.

Sullivan, William G. and W. Wayne Claycombe. Fundamentals of Forecasting. Reston, Va.: Reston Publishing Co., 1977.

Teweles, Richard J.; Charles V. Harlow; and Herbert L. Stone. The Com- modity Futures Game. New York: McGraw-Hill Book Co., 1977.

Wilder, J. Welles, Jr. New Concepts in Technical Trading Systems. Greens- boro, N.C.: Trend Research, 1978.

Periodicals and Unpublished Dissertations

Barnfather, Maurice. "Small Is Beautiful." Forbes, February 1, 1982, pp. 95-98.

Bishop, E.L. III and J.R. Rollins. "Lowry's Reports: A Denial of Market Efficiency?" Journal of Portfolio Management, Fall 1977, pp. 21-27.

Black, Fischer. "Implications of the Random Walk Hypothesis for Portfolio Management" Financial Analysts Journal 27 (March-April 1971):16-22.

Fama, Eugene F. "The Behavior of Stock Market Prices." Journal of Busi- ness 38 (January 1965):34-105.

. 'Mandebrot and the Stable Paretian Hypothesis." Journal of Busi- ness 36 (October 1963):420-29.

. "Random Walks in Stock Market Prices." In Readings and Issues in Investments,pp.214-18. Edited by Frank K.Reilly. Hinsdale, Ill.: Dryden Press, 1975.

King, Benjamin F., Jr. "The Latent Statistical Structure of Security Price Changes." Ph.D. dissertation, University of Chicago, 1964.

Lapin, Lawrence L. "Simulation of Technical Stock-Market Trading Rules and Implications Regarding the Random-Walk and Fair-Game Hypotheses." Ph.D. dissertation, University of Santa Clara, 1972.

Levy, Robert A. "AnEvaluation of Selected Applications of Stock Market Timing Techniques, Trading Tactics and Trend Analysis." Ph.D. disser- tation, American University, 1966.

Pariseau, Richard Roland. "Stock Market States: A New Approach to Investment Timing for the Average Investor." DBA dissertation, George Washington University, 1981.

58 MA Journal/E'ebruary 1986

Page 61: Journal of Technical Analysis (JOTA). Issue 23 (1986, February)

Swerdlow, Skip. "The Profitability of Technical Common Stock Trading Strategies Employing Price and Volume Data: A Simulation Test of the Efficient Capital Markets Hypothesis." DBA dissertation, Arizona State University, 1974.

Newsletters, Investment Services, and Other Sources

Commodity System Reports. Santa Clara, Calif. Published monthly.

Computrac. The Technical Analysis Group, New Orleans, La.

44 Wall Street Fund, Inc. Annual Report. June 30, 1982.

Growth Fund Guide. Yreka, Calif. Published monthly.

Janus E'und Record. Undated.

Market Log' . The Institute for Fconomic Research, Fort Lauderdale, Fla. PubliiEed monthly.

No-Load Fund * X. DAL Investment Co., San Francisco, Calif. Published monthly.

No-Load Fund Investor. Hastings-on-the-Hudson, N.Y. Published quarterly.

Systems and Forecasts. Signalert Corporation, Great Neck, N.Y. Published bi-weekly.

Telephone Switch Newsletter. Richard Fabian, Huntington Beach, Calif. Published monthly.

Dr. Michael J. Flanagan is head of Flanagan Associates, Consulting Engineers, Oakland, California.

59 K.L!A Jouxmal/F&ruary 1986

Page 62: Journal of Technical Analysis (JOTA). Issue 23 (1986, February)