Journal Editor & Reviewers

The Quantification Predicament
Timothy W. Hayes, CMT

Seasonality in Canadian Equity Prices
Don Vialoux, CMT

Patterns of Seasonal Variation in Canadian Fixed-Income Markets
R. Alain Rivet

The High-low Index as a Tool to Enhance Returns
Harold 6. Parker, Jr., CMT

Answering the Bell of Sentiment Indicators
Brent L. Leonard

Using the Z-Trend Oscillator for Long-Term Bond Market Timing
Robert T. Zukowski, CMT

A Study in Volume and Price Alerts
David Bryan

Editor

Henry 0. Pruden, Ph.D.
Golden Gate University
San Francisco, California

Associate Editor

George A. Schade, Jr., CMT
Scottsdale, Arizona

“This indicator has always produced huge profits! In fact, you could have doubled your money in just six months!”

Such a claim could be a sales-pitch. It could also be an analyst’s enthusiasm about some workjust completed. But in either case, such claims appear to be meeting increasing skepticism, perhaps because enough have proven to be based more on fiction than quantifiable fact, perhaps because enough investors have been burned bv indicators that have failed to pa11 out lvhen put to real-time use, or perhaps because the combination of ever-strengthening computing power and ever-increasing program complexitv has made excessive optimization as easier and more dangerous than ever.

In any case, the need to quantify accurately and thoroughly is greater than ever. Honest and reliable quantification methods, used in the correct wa!; are needed for increased research credibility. The!- are needed to impart objectivity. Thev are needed for effective analvsis and for the somid backing of research findings. The alternative is the purely subjective approach that uses trendlines and chart patterns alone, making no attempt to quantifv historical activity. But when the quantification process fails to deliver, instead producing misleading messages, the subjective approach is no worse an alternative - a misguided quantification effort can be worse than none at all. The predicament, then, is how to trulv add value through quantification.

The Concerns

The major reason for quantif;\ing results is to assess the reliabilitl\ and value of a current or potential indicator, and the major reason we have indicators is to help us interpret the historical data. The more effective the interpretation of historical market activity, the more accurate the projection about a market’s future course. An indicator can be a useful source of input for developing a market outlook if quantitative methods back its reliability.

But for several reasons, quantification must be handled with care. The initial concern is the data used to develop an indicator. If it’s inaccurate, incomplete, or subject to revision, it can do more harm than good, issuing misleading messages about the market that’s wider analysis. The data should be clean and should contain as much history as possible. M”nen it comes to data, more is better - the greater the data history the more numerous the like occurrences, and the greater the number of market cycles mlder study.

This leads to the second quantification concern, and that’s sample size. The data may be extensive and clean, and the analvsis may yield an indicator that foretold the market’s direction with 100% accuracy. But if, for example, the record was based on just three cases, the results would lack statistical significance and predictive \-alue. In contrast, there would be few questions regarding the statistical Validity of results based on more than 30 observations.

The third consideration is the benchmark, or the standard for comparison. The test of an indicator is not whether it would have produced a profit, but whether the profit would have been any better than a random approach, or no approach at all. \Yithout a benchmark, “random walk” suspicions may haunt the results.’

The fourth general concern is the indicator’s robustness, or fitness - the consistence of the results of indicators with similar formulas. If, for example, the analysis would lead to an indicator that used a 30-lveek moving average to produce signals with an excellent hypothetical track record, how different would the results be using moving averages of 28,29,31, or 32 weeks? If the answer was “dramaticallv worse”, then the indicator’s robustness would be thrown into question, raising the possibilitv that the historical result was an exception to the rule rather than a good example of the rule. A1n indicator can be considered “fit” if various alterations of the formula would produce similar results.

Moreover, the non-robust indicator may be a symptom of the fifth concern, the optimization process. In recent years, much has been \vritten about the dangers of excessive curve-fitting and o\-er-optimization, often the result of miharnessed computing power. ;\s analytical programs have become increasingly complex and able to crunch through an ever-expanding multitude of iterations, it has become easy to over-optimize. The risk is that, armed with numerous variables to test rvith minuscule increments, a program may be able to pick out an impressive result that ma? in fact be attributable to little more than chance. The accuracy rate and gain per annum columns of Figure 1 compare results that include an impressive-looking indicator that stands in isolation (top) with indicators that look less impressive but have similar formulas (bottom). One could have far more confidence using an indicator from the latter group, even though none of them could match the results using the impressive-looking indicator from the top group.

What follows from these five concerns is the final general concern of lvhether the indicator will hold up on a real-time basis. One approach is to build the indicator and then let it operate for a period of time as a real-time test. At the end of the test period, its effectiveness would be assessed. To increase the chances that it will hold up on a real-time basis, the alternatives include out-of-sample-testing and blind simulation. An out-of-sample approach might, for example, require optimization over the first half of the date range and then a real-time simulation over the second half. The results from the two halves would then be compared. X blind-simulation approach might include optimization over one period followed by several tests of the indicator over different periods.

Whatever the approach, real-time results are likely to be less impressive than the: were during an optimization period. The reality of an! indicator developed through optimization is that, as history never repeats itself exactly, it is unlikely that any optimized indicator will do as well in the real-time future. The indicator’s creator and user must decide how much deterioration can be lived with, which will help determine ivhether to keep the indicator or go back to the drawing board.

Trade-Signal Analysis

With the general concerns in mind, the various quantification methods can be put to use. The first, and perhaps most widely used, is the approach that relies on by and sell signals, as shown in Figure 2.? M%en the indicator meets the condition that it deems to be bullish for the market in question, it flashes a buy signal, and that signal remains in effect until the indicator meets the condition that it deems to be bearish. X sell signal is then generated and remains in effect until the next buy signal. Since a buy signal is always followed by a sell signal, and since a sell signal is always followed by a buy signal, the approach lends itself to quantification as though the indicator was a trading system, with a long position assumed on a buy signal and closed out on a sell signal, at which point a short position would be held until the next buv signal.

The method’s greatest benefit is that it clearly reveals the indicator’s accuracy rate, a statistic that’s appealing for its simplicity - all else being equal, an indicator that had generated hypothetical profits on 30 of 30 trades would be more appealing than an indicator that had produced hypothetical profits on 15 of 40 trades. ;Uso, the simulated trading system can be used for comparing a number of other statistics, such as the hypothetical per annum return that would have been produced bp using the indicator. The per annum return can then be compared to the gain per annum of the benchmark index.

But the method’s greatest benefit may also be its biggest drawback. No single indicator should ever be used as a mechanical trading svstem - as stated earlier, indicators should instead be used as tools for interpreting market activity. Yet, the hypothetical and actual can be easily confused. Although the signal-based method specifies hot a market has done between the periods from one signal to the next, they are not actual records of real-time trading performance. If thev were, the results would have to account for the transaction costs per trade, with a negative effect on trading results. Figure 3 summarizes the indicator’s hypothetical trade results before and after the inclusion of a quarter-percent transaction cost, illustrating the impact that transaction costs can have on results. The more numerous the signals, the greater the impact.

Also, as noted in the results, another concern is the maximum drawdown, or the maximum loss betsveen any consecutive signals. But again, as long as it is clear that the indicator is for perspective and not for dictating precise trading actions, indicators with trading signals can provide useful input when determining good periods for entering and exiting the market in question.

Zone Analysis

In contrast to indicators based on trading signals, indicators based on zone analysis leave little room for doubt about their purpose - th& don’t even hare by and sell signals. Rather, zone analysis recognizes black, white and one or more shades of gray It quantifies the market’s performance with the indicator in various zones, which can be given such labels as “bullish”, “bearish” or “neutral,” depending upon the market’s per annum performance during all of the periods in each zone. Each period in a zone spans from the first time the indicator enters the zone to the next observation outside of the zone. Unlike the signal-based approach, the indicator can move from a bullish zone to a neutral zone and back to a bullish zone. An intervening move into a bearish zone is not required.

Zone analysis is therefore appealing for its ability to provide useful perspective without a simulated trading system. The results simply indicate how the market has done with the indicator in each zone. But this type of analysis has land mines of its own. In determining the appropriate levels, the most statistically-preferable approach would be to identi$ the levels that would keep the indicator in each zone for roughly an equal amount of time. In man) cases, however, the greatest gains and losses will occur in extreme zones visited for a small percentage of time, which can be problematic for several reasons:

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

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

3. the indicator’s usefLllness must be questioned if it’s neutral for the vast majority of time.

A good compromise between optimal hypothetical returns and statistical relevance would be an indicator that spends about 30% of its time in the high and low zones, like the indicator in Figure 4. For an indicator with more than four Tears of data, that would ensure at least a year’s worth of time in the high and low zones and would make a deficiencv of observations less likely. In effect, the time-in-zone limit prevents excessive optimization bv excluding zone-level possibilities that would look the most impressive based on per annum gain alone.

Another consideration is that in some cases, a closer examination of the zone performance reveals that the bullish-zone gains and bearish-zone losses occurred with the indicator moving in particular directions. In those cases, the bullish or bearish messages suggested by the per annum results would be misleading for a good portion of the time, as the market might actuall!- have had a consistent tendency, for example, to fall after the indicator’s first move into the bullish zone and to rise after its first move into the bearish zone.

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

Alternatively, “rising” and “falling” can be defined using percentage reversals from extremes, in effect using zones and trading signals to confirm one another. In Figure 6, for example, the CRB Index indicator is “rising” and on a sell signal once the indicator has risen from a trough, whereas it’s “falling” and on a buy signal after the indicator has declined from a peak. Even though the reversal requirements resulted from optimization, the indicator includes a few poorly-timed signals and lvould be risky to use on its own. But the signals could be used to provide confirmation with the indicator in its bullish or bearish zone, in this case the same zones as those used in Figure 4. For example, in late 19i2 and early 1973 the indicator would have been rising and in the upper zone, a confirmed bearish message. The indicator would then have peaked and started to lose upside momentum, generating a “falling” signal and losing the confirmation. That signal would not be confirmed mitil the indicator’s subsequent drop into its lower zone.

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

Subsequent-Performance Analysis

In addition to using signals and zones, results can be quantified by gauging market performance over various periods following a specified condition. In contrast to the trade-signal and zone-based quantification methods, a system based on subsequent performance calculates market performance after different specified time periods have elapsed. Once the longest of the time periods passes, the quantification process becomes inactive, remaining dormant until the indicator generates a new signal. In contrast, the other two approaches are alwars active, calculating market performance with every data update.

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

In its most basic form, the results might list performance over the next five trading day, 10 trading days, etc., summarizing those results with the average gain for each period. However, the results can be misleading if several other questions are not addressed. First of all, how is the average determined? If the mean and the median are close, as they are in Figure 7, then the mean is an acceptable measure. But if the mean is skelced in one direction by one or a few extreme observations, then the median is usuallv preferable. In both cases, the more observations the better.

Secondly, what’s the benchmark? While the zone approach uses relative performance to quantif) results, trade- signal analysis includes a comparison of per annum gains with the buy-hold statistic. LikeFvise, the subsequent-performance approach can use an all-period gain statistic as a benchmark. In Figure 7, for instance, the average loday gain in the Dow Industrials has been 2% follo\+.ing a signal, nearly seven times the 0.370 mean gain for all loday periods. This indicates that the market has tended to perform better than normal follov+ig signals. That could not be said if the lo-da! gain was 0.4% following signals.

A third question is how much risk has there been following a buy-signal system. or reward following a sell-signal system? Using a buv-signal svstem as an example. one \cay to address the question would be to list the percentage of cases in lshich the market was higher over the subsequent period, and to then compare that with the percentage of cases in which the market was higher over an; period of the same length. Again using the IO-da! span in Figure 7 as an example, the market has been higher after 75% of the signals, yet the market has been up in only 3% of all 1Odav periods, supporting the significance of signals. Additional risk information could be provided by determining the average drawdown per signal - i.e., the mean maximum loss from high to low following sig- nals. The mean for the lo-day period, for example, was a maximum loss of 0.7% per signal, suggesting that at some point during the lo-day span, a decline of 0.7%’ could be considered normal. The opposite approaches could be used with sell-signal indicators, with the results reflecting the chances for the market to follow sell signals by rising, and to what extent.

Along with those questions, the potential for double-counting must be recognized. If, for example, a signal is generated in January and a second signal is generated in February, the four-month performance follo\ving the January signal would be the same as the three-month perfor- mance following the February signal. This raises the question of lvhether the three-month return reflects the impact of the first signal or the second one. Moreover, such signal clusters give heavier tveight to particular periods of market performance, making the summar? statistics more difficult to interpret. Problems related to double-counting can be reduced or eliminated bv adding a time re quirement. For the signals in Figure’ 7, for instance, the condition must be met for the first time in .30 davs - if the ratio reaches 1.92, drops to 1.90, and then returns to 1.92 two days later, only the first day will have a signal. The time requirement eliminates the potential for doublecounting in anv of the periods of less than 30 davs, though the longer peiiods still contain some overlap in this example.

Another application of subsequent-performance analy sis is shown in Figure 8, which is not prone to any double-counting. The signals require that three conditions are met, all for the first time in a given year - the Dow Industrials much reach its highest level in a year, another index performance analysis for both buy signals anb sell signals can be used together in an indicator. For each time span, the chart’s box lists the market’s performance after buy signals, after sell signals, and for all periods.

Reversal-Probability Analysis

Finally the subsequent performance approach is useful for assessing the chances of a market reversal. In Figure 10, the “signal” is the market’s year-to-year change at the end of the year, with the siglials (years) categorized by the amount of change - )-ears with anr- amount of change, those with gains of more than S%, etc. In this case, the subsequent-perforcance analvsis is limited to the year after the various one-year gains. But the analysis takes an additional step m assessing the chances for a bull market peak tvithin the one- and nvo-year periods after the years with market gains, or a bear market bottom Athin the one- and two-year periods after the years Gth market declines.

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

Using this percentage as a benchmark, it can then be determined whether there’s been a significant increase in the chances for a peak or trough in the year after a one-year gain or loss of at least a certain amount. The chart’s boxes show the peak chances following up years and the trough chances following down years, dividing the number of cases by the number of peaks or troughs. For example, prior to 1995, there had been 31 years with gains in excess of 15% startingwith 1899. After those Tears, there was a 52% chance for a bull market peak in the subsequent year (16 following-years with peaks / 31 years with gains of more than 15%). The chances for a peak within two years increased to 74%, which can be compared to the benchmark chance for at least one peak in 61% of the two-year periods (since several two-year periods contained more than one top, this is not the exact double of the chances for a peak in any given year).

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

Conclusion

Each one of these methods can help in the effort to assess a market’s upside and downside potential, with the method selected having a lot to do with the nature of the indicator, the time frame, and the frequency of occurrences. The different analytical methods could be used to confirm one another, the confirmation building as the green lights appeared. An alternative would be a common-denominator approach in which several of the approaches would be applied to an indicator using a common parameter (i.e., a buy signal at 100). Although the parameter would most likely be less than optimal for an) of the individual methods, excessive optimization would be held in check. But whatever approaches are used, it needs to be stressed that each one of them has its own means of deception. By better understanding the potential pitfalls of each approach, indicator development can be enhanced, indicator attributes and drawbacks can be better assessed, and the indicator messages can be better interpreted.

The process of developing a market outlook must be based entirely on research, not sales. The goal of research is to determine if something works. The goal of sales is to show that it does work. Yet in market analysis, the lines can blur if the analyst decides how the mark& is supposed to perform, then sells himself on this view by focusing only on the evidence that supports it. What’s worse is the potential to sell oneself on the value of an indicator by focusing only on those statistics that support one’s view, regardless of their statistical validity. ;Is sholvn by the various hazards associated with the methods described in this paper, such self-deception is not difficult to do.

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

1. Reference to Burton Malkiel’s A Random itTalk Down Wall Street, which argues that stock prices move randomly and thus can& be forecasted through technical means.

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

Introduction

“Buy them when it snows, sell them when it goes!” That’s the expression used by well known equity market strategists in Canada. The expression refers to the strategy of buying Canadian stocks when the snow starts to fall in November and taking profits when the snow melts in March. Canadian stocks tend to be stronger during this period each year than at other times of the year.

The evidence of seasonal strength provided by these strategists has been mainly anecdotal. They point to statistics measuring the low point for the Toronto Stock Exchange 300 Composite Index (TSEC) in November to the high point in March of the following year. The statistics, when calculated this way, indeed show that the TSEC exhibits strong seasonahty. As indicated in Exhibit 1, the average return on investment (excluding dividends) during the twelve selected periods picked from November 1982 to November 1994 was an amazing 12.2%. In contrast, as indicated in Exhibit 2, the average annual gain by the TSEC during the same 12-year period using the end of November as a base date each year was only 6.1%. The implication is that the investor can optimize his investment returns by purchasing Canadian stocks at their lows in November and by going short when Canadian stocks reach their highs in March.

This anecdotal evidence may be impressive but is clearly flawed and statistically incorrect. It implies that the investor knows when the lows will be made in November and the highs will be reached in March. As indicated later in this report, the evidence also leads to a misleading strategy. ‘Yet, the evidence suggests that a statistically correct study might provide an interesting insight on the seasonality of Canadian stock prices. This report uses a simple statistical method (i.e. arithmetic or mean averages) to examine the seasonality of Canadian stock prices using the end of November and the end of March each year as a base. The period of examination was from the end of November 1982 to the end of November 1994. In addition, this report examines the fourteen industry subindexes that make up the TSEC to determine the industry groups that tend to outperform and underperform the TSEC during the November to March period. Next, the major factors causing seasonal strength during this period are examined. Finally, the report looks at the employment of investment strategies using the findings of this report.

The following study indicates that Canadian stock prices show seasonal strength from the end of November to the end of March. The Toronto Stock Exchange 300 Composite Index was examined during the twelve periods from the end of November to the end of March starting in November 1982 and ending November 1994. The average gain each year during this period (excluding dividends) was 6.4%. Gains were realized in ten of the twelve periods.

In contrast, the TSEC displayed in Exhibit 2 rose only 6.1% per year on average during the 1 S-year period from November 1982 to November 1994. In addition, the TSEC showed gains during only seven of the twelve years.

A Study of Seasonal Price Performance of the TSEC

Indeed, as the next table illustrates, investors who held Canadian stocks during the past 12 periods from the end of March to end of November lost money.

Higher returns from the end of November to the end of March are not a new phenomenon. They have existed since at least the start of taxation of capital gains in Canada in December 1971, as illustrated in the following table:

Canadian stock prices from the end of November to the end of March rose in 18 of 22 years. Average gain during each of the 22 periods was 5.2%. Median gain (identified as the return for the end of November 1985 to end of March 1986 period) was 6.7%.

In contrast, seasonal strength did not appear consistently in the end of March to the end of November periods during the past 22 years. As the next table illustrates, Canadian stocks prices recorded an average gain of only 0.5% during the 22 periods. They rose in only 7 of the 22 periods. Median return (identified as the return for the end of March to the end of November 1984 periods) was -0.6%.

In conclusion, investing in Canadian stocks during the four-month period from the end of November to the end of March provides a slightly higher return with only four months of stock market risk than employing a buy/hold strategy with 12 months of stock market risk. Indeed, investing from the end of November to the end of March avoids an eight month period of stock market risk when Canadian stock prices tend to record little or no return.

Seasonal Strength in Industry Groups that Make Up the TSECS

Some industry groups exhibited more seasonal strength than others during the November to March period. A similar analysis of the 14 industry subindexes that make up the TSEC was completed from the end of Novemberb 1982 to the end of November 1994. Results of the analysis were as follows:

The results show that economically sensitive stock groups such as paper and forest products, base metals, communications and transportation tend to outperform the interest sensitive groups including pipelines, financial services and utilities.

Industry groups that recorded the greatest seasonal strength from November to March also tended to exhibit the greatest seasonal weakness from March to November.

Results show that holding economically-sensitive stock groups from the end of March to the end of November is an inferior strategy. Indeed, holding short positions in industry groups such as Paper and Forest Products, Conglomerates, Transportation and Metals and Minerals can be a profitable strategy.

Reasons For Seasonality

Seasonality occurs for at least three reasons:

1) Seasonal strength in U.S. equity markets during the end of November to end of March period has an influence on Canadian equity prices. As the following table suggests, the S&P 500 Index from November, 1982 to November 1994 showed seasonal strength during the end of November to end of March periods in U.S. equity markets (although not as strong as seasonal strength in Canadian markets). From November 1982 to November 1994, the S&P 500 index rose an average of 9.4% per year (excluding dividends). During the twelve test periods, the S&P 500 index rose an average of 6.9% (i.e., 73% of the move made the by S&P 500 occurred during one third of the year).

Influences on Canadian equity prices occur indirectly because of the close relationship between the Canadian and American economies. They also occur directly through inter-listed trading activity in stocks that make up the TSE 300 Composite Index. A study in October 1994 indicated that 51 percent of the weighting of the TSE 300 Composite Index is based on securities inter-listed on U.S. exchanges and 18.7 percent of the value of trading in TSE 300 Composite Index stocks occurred in U.S. markets.

2) Although Canadian and American tax laws differ, surges in money flows and investment decisions occur in Canada near year end for similar reasons that they occur in the United States. Stock prices tend to rebound later when selling for tax purposes has abated.

3) Contributions to individual retirement plans in Canada, (known as Registered Retirement Savings Plans) usually concentrated in February, tend to have a similar but proportionally greater impact on Canadian equities prices because of more liberal contribution rules and “Canadian content” regulations.

A Practical Method to Take Advantage of Seasonal Strength in the TSECA

According to this report, seasonality among the 14 TSEC subindexes is strongest with Canadian forest product stocks. A portfolio of forest product stocks seasonally invested beginning in November 1982 and finally liquidated in March 1994 should provide favourable results. The implication is that an investor who continues to buy Canadian forest product stocks at the end of November and sells them at the end of March greatly enhances prospects for an above average return on investment.

A study was completed to examine the profitability of the strategy using an initial investment of $100,000. Funds were allocated by the TSEC weighting in forest product stocks when the investment initially was made. The weightings were kept constant throughout the period of investment (despite the continuous change in the weightings over time). Eight of the nine forest product stocks in the TSEC in November 1982 were included in the study. AI1 of the eight companies remained Canadian based companies throughout the period of investment, although many were involved with mergers, takeovers and restructurings. When a company was merged or taken over, prices for the surviving company were taken. The ninth company was excluded (Consolidated Bathurst) because it did not survive throughout the investment period as a Canadian forest product company. (It was acquired by Stone Container). Other Canadian forest product stocks subsequently added to the TSE Forest Product Index also were excluded because their prices were unavailable throughout the study period. Price data were adjusted for stock splits and reverse stock splits that occurred during the period of investment. Each stock was acquired at the price on the last trade date in November and liquidated at the price on the last trade date in March.

Each transaction was examined to determine its feasibility (i.e. the ability to complete the transaction on the given date at its indicated size.)

The calculations were purposefully completed to provide a conservative return on investment.

• A commission of 1.0% of value was charged on all purchases and sales (current commission charges now are substantially less. However, in the early 1980s in Canada before commissions became negotiable, a 1 .O% rate was a fair estimate of cost).

• Dividends were not included.

• Interest from cash balances held from the end of March to the end of November each year was not included.

As one would expect, knowing how well the forest products in general have done during the end of November to end of March periods, results from the portfolio were impressive. The $100,000 portfolio appreciated in value to $406,773 during the 12 periods of investment.

Conclusion

Seasonality analysis shows that Canadian stock prices have been significantly strong during the four month period from the end of November to the end of March during the 12-year period ending November 1994. Three main factors probably influencing Canadian stock prices during this period were year end transactions for tax purposes, RRSP contributions and seasonal influences by U.S. equity markets. These three factors are expected to continue to influence the seasonality of Canadian equity prices in the future.

Investors can continue to look for opportunities to take advantage of seasonal strength in Canadian stock prices. Each year as November approaches, they can examine Canadian stock groups such as forest products and base metal stocks that tend to outperform the TSEC during the next four months. An examination of technical patterns and fundamental outlooks for individual stocks within these groups could help to identify potentially profitable investment opportunities. Investors subsequently should liquidate positions by the end of March and, when appropriate, consider short positions in these groups.

A. Introduction

The literature of technical analysis and investing has over the years contained many references to “seasonal effects” and the resultant consequences on investors’ rates of return. This paper will examine apparent seasonal effects in the Canadian tixed income markets for evidence that such effects are statistically significant.

B. Background

1. CYCLE ANALYSIS

Cycle analysis has been used in many different contexts and applied to many different time frames, ranging from several decades, as in the works of Nikolai Kondratieff and his adherents, to more contemporary comments applied to day trading or weekly trading in the commodity futures and financial futures markets.’ In this paper the objects of study will be the month end closes of two Canadian fixed income indices, the Scotia McLeod Long Term Bond Price Index and the Scotia McLeod Mid Term Bond Price Index, for the period 1948 through 1994, and 1980 through 1994, respectively. Initially, patterns of month to month variation within each year will be examined; sub sequent sections will look at possible relationships between seasonal variations and multi-year trends of bond yields and bond prices.

2. CANADIAN FIXED INCOME MARKETS

The beginnings of the fixed income markets in Canada can be traced back to the early 1870’s; the first issues of Government of Canada marketable bonds came out at that time to refinance existing provincial obligations.* Debt outstanding grew rapidly during World War I, and during World War II. The next period of rapid growth in debt outstanding occurred during the 1975-1990 period.3 As at December 1994, the value of unmatured marketable bonds issued by the government of Canada had reached Cdn. $234 billion. These bonds are held by a wide variety of investors, both foreign and domestic, individuals as well as institutional investors such as banks, pension funds and mutual funds. As an example, the value of assets held by Canadian bond-oriented mutual funds was Cdn. $13.9 billion as of August, 1994.4 The fixed income markets in Canada therefore enjoy significant size, a wide variety of participants, and for most issues of government of Canada and provincial government bonds, very good liquidity.

3. MAJOR TRENDS IN CANADIAN BOND PRICES

The value of the Long Term Bond Price Index has ranged from a low of 59.34 (September 1981) to a high of 216.21 (January 1948) (Table 1 (d)). The extent of this variation is not out of line with the variations observed in the yields of U.S. treasury bonds in the years since 1960; the latter went from 4% in 1960 to the inflation-induced levels of 14% in the early eighties, dropped back to 5.9% in the fall of 1993, and returned to the 8% level in November 1994.s

An examination of the yearly averages of the Long Term Bond Price Index provides a very good overview of the major trends in Canadian bond prices from 1948 to 1994. Prices declined fairly steadily from 1948 through 1953, rebounded briefly, and declined again until 1958. Prices remained relatively stable from 1959 through 1965. The price decline from 1966 through 1981 was briefly interrupted by short intervals of relative stability during the 1970-1972 and the 1974-1978 periods. (Graph 2)

A clear multi-year bear market trend can be identified as having existed from 1966 through 1981; similarly a clear bull market trend can be identified from the 1981-1994 period, with 1981 being the changeover year. (Table 1 (d), “Average” column and Graph 2).

C. Methodology

1. DATA SOURCES

For the purposes of this study, two monthly data series were examined: (1) the Scotia McLeod Long Term Bond Price Index for the period 1948-1994; and (2) the Scotia McLeod Mid Term Bond Price Index for the period 1980-1994. These two indices are the best-known bond indices in Canada. Along with government of Canada bonds, these indices comprise four different groups of bonds: 10 utilities, 10 municipals, 10 provincials and 10 industrials. The first data series, was started in 1947 and has been updated monthly since then. These two series were picked for this study because they represent a source of continuous, internally consistent and easily available data. As of November 1995, the sector weighting of the Long Term Bond Index was as follows: Government of Canada, 59%; Provincials, 25%, Corporates 13%, municipals, 3%. The Mid-Term Bond Price Index had a similar weighting. Both the long term and mid-term bond price indices were recalculated in 1985, with a new base of 1985 = 1OO.6

2. ANALYSIS OF DATA

A more formal statistical method will be used to test for seasonality, relating each monthly datum not only to the average value of the index for that year, but also to the index value for the month immediately preceding it. Tests of statistical significance will be tabulated not only for the raw ihdex values, but also for each of the derived data series.

The null hypothesis can be stated as follows: none of the data series examined contains seasonal variations that display statistical significance. In proceeding with the statistical treatment of the monthly index values, the author is starting out with one basic premise, namely, any seasonal difference is worth examining only if the average price in one period differs significantly from its price in another period. This can be accomplished by calculating for the period under study the index average and standard deviation by month, and from it deriving the “p-value”, or confidence interval for each month.7 The criterion in the economic literature of “p” value of .lO or less will be used. This is the criterion mentioned in August 1992 issue of Technical Analvsis of Stocks and Commodities, in an article by Dr. Lewis C. Mokrasch. The method used here is an adaptation of the algorithms developed by Dr. Mokrasch. a For this study, the month-end closes for the Scotia McLeod Long-Term Bond Price Index and the Scotia McLeod Mid-Term Bond Price Index were used. This study examines the average monthly values for both indices, and also includes tests of statistical significance for those values, as well as for two other sets of monthly averages derived from the raw index numbers. These are, respectively, the average index value for each month expressed as a percentage of the year’s average value, and, the month-to-month change in each index expressed as a percentage of the previous month’s datum. The sequence of the calculations can be seen by consulting Tables 1 (d) through 3(f), which are included in the body of this paper. In addition to the raw index values for each data series, two sets of detrended data were used.

First, the raw index values for a calendar year were expressed as a percentage of the 12 month average. For each calendar year, the average 12 month index value was calculated; the average index value for each month was divided by the 12 month average for the same year and expressed as a percentage. In the subsequent set, each datum was compared to the index value of the month immediately preceding, with the difference expressed as a percentage. The same series of calculations was performed as in the other tables. The reader should also note that the Long Term Bond Price Index series has been broken out into one subseries, for the 1980-1994 period, in order to compare the results of Long Term index and the Mid-Term index for the same time frame.

In order to determine the statistical significance of each month’s variation, the following formulae were used:

(Click To Enlarge)

(Click To Enlarge)

Raw Index Values

For the 1948-1994 period, the month displaying the largest standard deviation was July, the smallest, December. The high for the year occurred in January, the low for the year came in September. As none of the “p” statistics for the raw index average monthly values is near .lO, it can be concluded that there is no statistically significant variation. (Table 1 (a))

Monthly Average Values

When the index values are re-expressed as a percentage of the year’s average, a slightly different pattern emerges. The largest standard deviation appears for January. There are five months that display “p” values smaller that .lO: September (.03); J anuary (.05); February (.06); July (.06) ; August (.07). These results suggest that from a seasonal point of view, bonds tend to be more of a sale in January, and more of a buy in September.

Month-to-Month Average Price Changes

An analysis of the average monthly percentage price changes also suggest some possible trading strategies. For the data series as a whole, the average month to month price change is quite small, -.l 1. This is as could be expected, i.e. short term price fluctuations tend to cancel one another out. The largest average positive price change occurred in October (t.90); the largest negative price change was in July (-.59). The “p” value for October computes as .03; this suggests that in terms of short-term trading, October is a good month for taking profits. (Table l(c))

(Click To Enlarge)

(Click To Enlarge)

RAW INDEX VALUES

For this data series, the December value displayed the average high for the year, as could be expected in a secular bull market. The lows for the year occurred in April and July. None of the “p” value for the raw index values are significant. (Table 2 (a))

MONTHLY AVERAGE VALUES

For the detrended data, the largest standard deviation OccurredinJanuary. ThemonthsofApril (.lO);July (.lO); and December displayed statistically significant values. This suggests that the optimal trading strategy would be to buy at the April lows and take profits at year-end in December. It is worth nothing that the statistically significant low point for the year of the monthly average values is April, rather than January or February. (Table 2(b))

MONTH-TO-MONTH AVERAGE PRICE CHANGES

For this data series, the average month-to-month price change is quite small, with an upward bias (t .19%). The largest positive average price change took place in October (+2.73%); the largest negative price change occurred in March (-1.02%). The only month with a significant “p” value was October (.02). This would suggest that October month end is an optimum time for profit-taking for short-term trading, as the average monthly value in creased from 98.54 in September to 101.16 in October.

(Click To Enlarge)

>

(Click To Enlarge)

RAW INDEX VALUES

There is only one set of data available for the Mid-Term Bond Index, for the years 1980-1994, with the general trend being a fairly steady advance in prices from 1981 through 1993, with only a brief interruption for the years 1987,1988 and 1989.

For the raw data, the average high for the year occurred in December, while the average low was registered in September. The month with the largest standard deviation was July. None of the “p” values for the raw data aresignant. (Table 3(a))

MONTHLY AVERAGE VALUES

For the monthly values expressed as a percentage of the year’s average, the month with the smallest standard deviation was April. The months of April and December both displayed “p” values of .08. This would suggest a trading strategy of taking advantage of the April dip in prices to buy, and taking profits at year-end. (Table 3(b))

MONTH-TO-MONTH AVERAGE PRICE CHANGES

Concerning the average month-to-month price change, the value for this particular series was .OS%, i.e. a small change, but with an upward bias. The largest average positive change occurred during October, (2.17%) while the largest average negative change took place during February (-.98%). The month showing the largest standard deviation was July with a value of 3.14. The month with a statistically significant “p” value was October (.Ol ). This would suggest the following strategy for Mid-Term bonds: make use of April price weakness to buy, as noted in the previous section and take advantage of the price strength during October to sell.

Conclusion

The analysis of seasonal variations in the Canadian fixed-income markets, making use of detrended data, allows us to reject the null hypothesis that seasonal variations in the Canadian fixed-income markets are devoid of statistical significance. The study of the patterns of seasonal variations can provide trading strategies that can be tailored to long-term versus mid-term bonds. These strategies can also be customised to be better applied to month-to-month price changes versus changes in bond prices over a twelve-month trading cycle. The study of seasonal variations is best used as a confirming indicator for those with a time frame longer than a year. Seasonal patterns do provide valuable signals for shorter-term trading.

It is also instructive to make certain comparisons between the Long Term Index and Mid-Term Index. Regarding the raw index values, the standard deviation for the long-term index is much larger (41.23 and 13.50 vs. 9.86). This is to be expected, since longer-term bonds are more volatile than near-term bonds. This may also be partially explained by the difference in the number of observations (564 for the Long Term Index compared to 180 for the Mid-Term index). When the detrended data for the same time periods are examined (Tables 2 (b) and 3(b)), the difference in standard deviations is much smaller (5.12 versus 3.75), but it is the Long-Term index that carries the higher value. Within the calendar vear, therefore, it is the Long-Term index that has the higher variability. In terms of month-to-month price changes, the Long Term Index shows a standard deviation of 2.27 for the 1948-1994 period, and 3.50 for the 1980-1994 period, compared to 2.58 for the Mid-Term Index. Within the calendar year, the month-to-month price changes of the Long Term Index display the greater variability. For the 1980-1994 period, the April lows and December highs are statistically significant for both the Long-Term and the Mid-Term index. This clearly indicates April month-end as a favourable buy point and December as a good time for profit-taking.

The best indicator is provided by the analysis of the monthly average values of the Long Term Bond Index. It is the best indicator for the following reasons:

1. It incorporates the largest sample: forty seven years of data, or a total of five hundred and sixty-four pieces of data.

2. It covers periods of both declining and rising bond prices, thereby providing the conclusions with the greatest applicability.

If we examine the monthly average values of the Long Term Bond Index just for the years 1980-1994, the only month that shows any significance is December (p=O.O7), with the value for April at the limit (p=.lO). However, when the longer time frame 1948-1994 is examined, we find September (p=O.O3) to be a statistically significant time to look at buying, whereas January (p=O.O5) and Feb ruary (p=O.O6) can be considered statistically significant times to conduct selling.

In terms of month-to-month price changes, all three data series display both statistical significance for the month of October and a positive average price change for that same month, suggesting that it is also a good time of the year to be a seller of Canadian bonds. The study of seasonal patterns in the Canadian fixed-income markets appears to be of value to those seeking investment opportunities in this arena.

End Notes

1. John J. Murphy, Technical Analvsis of Futures Markets, N.YI.F., 1988, pp 414455

2. J. E. Hatch, Robert F. White, Canadian Stocks, Bonds, Bills and Inflation, 1950-1987, The Research Foundation of the Institute of Chartered Financial Analysts, 1988, p. 27

3. Ibid., see also Bank of Canada Review, various issues 1988 1994

4. Albert S. Thompson, The Canadian Mutual Fund Industry, Moss, Lawson & Co. (Research Report), July 1994

5. Charles Kirkpatrick II, Charles Dow Looks at the Long w, Barron’s, June 1994

6. Note: The Scotia McLeod Mid-Term Bond Price Index and the Scotia McLeod Long Term Bond Price Index are copyright Scotia McLeod Inc. The monthly index values are courtesy Scotia McLeod Inc. The monthly index values are courtesy Scotia McLeod Inc. The figures for the raw index values are from Scotia IMcLeod’s Handbook of Canadian Debt Market Indices, Toronto, 1994

7. Dr. Lewis C. Mokrasch, Detecting: Seasonalitv, Technical Analysis of Stocks & Commodities, August 1992

8. Ibid.; see also from the same author, Looking at lo-Year Stock Price Patterns, Technical Analvsis of Stocks& Commodities, April 1991

9. Ibid.; see also A.N. Beals, Statistics for Economics and an M. Abramowitz and E. Stagun, Handbook of Mathematical Functions

The author gratefully acknowledges the assistance of Dr. L.C. Mokrasch and Mr. Harvey Chan for this study.

(Click To Enlarge)

(Click To Enlarge)

(Click To Enlarge)

(Click To Enlarge)

(Click To Enlarge)

(Click To Enlarge)

(Click To Enlarge)

Introduction

The analysis of new 52-week high and low data on the NYSE can provide valuable insight into the near- to intermediate-term treud of the market. These data have been used in a wide variety of ways over the years, but most commouly they are used as a ratio (highs divided by lows) or as a difference (highs minus lows) .’ The interpretation of the data has generally fallen iuto two general categories as well. Oue method relies 011 the data to confirm or diverge from the popular indexes such as the Dow Joues ludustrials or the S&P 500. The logic behind this method is that a healthy market which is making new highs should be accompanied by a large aud/or rising number of individual stocks making new highs as well. Sew highs ill the indexes without IKW highs iu the high-low indicator are considered suspect. The reverse would be true of bottoms. The second general use of the high-low indicator has beeu as au overbought-oversold indicator. The logic behiud this method is that extremes in the indicator point to unsustainable extremes iu the market. Most forecasts usiug these two types of iuterpretatious of high-low data are rather subjective.

The shortcoming of subjectivity of interpreting the high-low index was addressed by Abe Cohen with the Chartcraft High-Low Index. He displayed his High-Low Index 011 a point aud figure chart. The index is cou- strutted by dividing the uumber of uew daily 52-week highs on the SYSE by the sum of new highs and new lows. .A simple lo-day moving average of Llie rcsultiiig percentage data (((Highs/ (Highs t LOWS) ) / 10) is theu plotted 011 a point and figure chart using a box size of 2% and a three-box reversal arid bounded on the top by 100 aud the bottom by 0. The result is a chart that is elegant in its simplicitv and objective because it filters out the “noise” ok small ((6%) reversals and shows reversal points clcarlv (the reversal from X’s to O’s or vice versa is unequivocal). Cohen considered levels below 10% to be oversold aud ihose above 90% 10 be overbought aud a reversal from those extremes to be buy (from oversold) and sell (from overbought) signals.

Cohen’s logic seems to go oue step beyond that of previous indicators. Using the decision rules in the paragraph above, this indicator gives iiot only ali easilv determiued indicatioil of extreme overbought and oversold levels, but also provides au objective method for determining when these conditions are reversing. This siguificantly enhances the usefulness of high-low data because it allows them to be used as a timing tool for intermediate-term moves. Unfortunately, the market rarely gets to the extreme 10% or 900/o levels before reversing, aud this can strand the investor in a reversing market without gettiug a signal from the indicator. Cohen’s rules resulted in only two completed signals in the lo-year period studied. When invested, the Cohen method returned a respectable compounded annualized rate of return of 18.9%. However, it was in the market only 16% of the time aud captured only a little over a quarter of the total up move. For more useful entry aud exit signals, some alteration of the decision rules seems in order.

Method

Testing was done using Cohen’s calculatiou and plot- tiug methods, described above. Only long positions were taken using the following decision rules:

Buy: Indicator reaches 40% or less alld reverses up by 6 percentage points.

Exit: Indicator reaches 70%’ 01‘ greater aud reverses up by 6 percentage points.

Sell Stop: Indicator reverses prior to the sell signal above aud declines to a level below the low of the column of O’s preccdiug the buy signal. The sell stop would be triggered, for instance, if the indicator falls to 38 and reverses up to 44 or greater (crcatiug a bottom at 38) and the has lo 36 prior 10 risiiig to 70.

The indicator is illustrated in Figure 1.

These decision rules were established empiric&
based on the author’s experience. The levels of 405
and 705 looked as if thev would give a good balauce
of profitable signals VS. “whipsaws” and would achieve
the primary objective of developing a tool to assist an
equity investor in achieving results superior to a buv
and hold strategy Therefore, the investor is either
long or out of the market for the purposes of the test.
X secondary objective as to have drawdowns due to
market fluctuations that were significarltlv less than a
buy alld hold strategy.

The testing was doue over the lo-year period from
19851994. This period 1va.s chosen because Grst, the
method for reporting the underlying data ivas cousis-
tent, and second, it produced enough signals to give
reliable results.’

Results

The results for the studv oeriod indicate that the high-low iudki can be verv useful for timing entry into the equity market. The buys and sells are listed and summarized in Table 1 and illustrated irl Figure 2.

The high-low index aud the decision rules described above resulted in the investor being in the market for a total of 3.76 wars out of the 9.74 years between tde start of the first signal and the end of the last signal. The compounded annual rate of return’ \vhilc illvested was 23.57%, excluding dividends. The maximum adverse excursion was 7.26%. (:Ilnximum n&r.sr: txxusion is the maximum percentage dccliue ill equity due to market puctuation of the equal placed into each /ewebxtc. This differs from the term drtudown, which is the I maximum percentage dccliue irl equity due to market fluctuation from the previous peak Icvcl ofequitv.) The equity curve for using the high-low index during . the study period is conta:led in Figure 3. The starting point for the curve is the S&P 500 level at the begiuuiug of the test period.

If one had instead bought the S&P 500 at 167.16 on December 19, 1984 and sold at 467.91 on November 3, 1994, OIIC ivould have achieved au 11.14%) compounded annual rate of return. excluding dividellds. for the 9.i4 year period. Using the buy a11d hold method. the maximum drawdown would have been 34.23%.

Discussion and Conclusion

The High-Low indicator produced reliable signals and performed well versus a buy and hold stratcg) during a period when INN and hold tvorked extremely well. The trading signals’ for the studs period had the 36 following fworable characteristics:

• 70% were profitable (20% with no ad\Terse excursion)

• High annualized rate of return while invested (23.57%)

• Low adverse excursion from entry (7.26% maximum and 2.07Y0 average)

• Ratio of % gain to M loss was very favorable (7.31% avg. gain vs. 2.78% avg. loss)

• Total Gaiu/Maximum Drawdown ratio was 9.46 vs. a ratio of 2.67 for S&P 500 buy and hold

• A Student’s T test of the results indicates that they are highly significant, with only a 1% probability that’the) were achieved by chance

In addition, the indicator has the advantages of being objective aud easilvmaintaincd from readily available data.

The indicator gate an al’eragc of only two completed signals per year and the exit rules had an in\.estor out of the market for two-thirds of the time studied. These characteristics may be considered to be either a positive OI Ilegativc dcpcnding 011 one’s i1lvestment objectives; however, thcv dicl result in lolvcr-than-market risk. The most significant disad\autaqe to the indicator rvould seem to be that it tends to bc iarly with its exit signals. Examiuarion of Figure 2 reveals that one would haye forcgo1le siguificant upside movemeut iu both 1986 aud 1989 by using this indicator as an exit tool. This would suggest that other indicators might be useful adjuncts to more precisely time market exits. It might also be useful to examine SOIW other decision hcnchmarks in the future. Nevertheless, this illdicator, with the current benchmarks, resulted iI1 the investor capturing two-third of the points iu the up I/ewebOYC iu the market during the study period while being inwtcd for odv one-third of the time.

Bibliography

Cohen, A.1V.. How to Use the Three-Point Reversal Method of Point 8: Ficrure Stock Market Trading, Sew Rochelle, Sli, Chartcraft, Inc., 1987

Colby Robert M’. and Thomas A. Sleyers, The Enclclopedia of Technical hlarket Indicators, Homcwood, IL, DOW Jones-Irwin, 1988

Fosback, Sorman G., Stock Market Lozic, Fort Lauderdale, FL, The Institute for Econometric Research Inc.. 1986

Pring, hlartin J., Technical Analvsis Esnlnincrl, McGraw Hill, New York. 1985

The purpose of this review paper is to list, explain, and evaluate several well-known stock market setltitnettt indicators over many periods of time. These indicators inelude Option put/call ratios, advisorv letters, short interest, mutual futtd cash, and other contrary against-the-crowd statistics.

The reason that this is a Review article rather than Research is that there has been much written ott these indicaters by the experts of the industry (although very little recently, which I hope to update). Each indicator’s peaks and troughs will be juxtaposed with the appropriate index or average. I intend to first define and describe each Indicator and assess its efficacy; then, in a Discussion section, I place each on a Bell/Growth Curve model in its appropriate place itt time.

These littdittgs should be of use to attyotte who needs to ascertain markel direction attd reversals for trading.

Much has been written over the years about cotttrar) opinion; it has becotne widely accepted and clever to go against the crowd - “When everyone looks 011e way, look the other!” Although primarily a true concept, there are a few considerations I would like to bring to light. Most serious investors arc familiar with the South Seas Bubble and Tulip Bulb Mania from Mackay’s Extraordittay Popular Delusions and the Madness of Crowds. Although history repeats itself, it almost never does it exactly in the satne way. Try developing a tulip craze in Holland today, or, observe the Deutschbank’s tight stance against installation after the wheelbarrows of DMarks decades ago. 111 his talk at the 1994 attttual TSAR cotlferettce itt Satt Francisco, John Bollinger stressed how importattt it is to know against whom to be cotttrary. Should otte take a position against the world being round, or the SUN rising tomorrow! Rather, the successful ittvestor has to estab- lish, through introspection, att ittterttal tnottitor which will wart1 hitn when he has stopped doing his own analysis and has begutl relying on peers, tnedia itetns, or a guru for opiniotts, “tips,” and timing.

In his book, Humphrey Neill explains that contrary opinion is not necessarily cynical or negative, but sees both sides of att issue using ottc’s experiettce and logic to see reality. Just as some oscillators can be useful in the middle of a trend but wrong at the extremes, so are the majorit) often correct during a Bull or Bear market but mauically wrong when it reverses, especially when they are required to act, like buy or sell, rather than just observe. Examples of herd logic at these junctures are “This titne it is differ- ent,” or “What cat1 possibly go wrong.” Or at the nadir, “This company is doing cvervthing tvrotlg - it’s hopeless.” In the following pages I would like to illustrate which indicators are the most effective in forecasting markets, individually and in combination.

One category of setitimettt measuremettt is the surveys found in Barron’s and elsewhere ott advisors, letter writers and investors. Although the majority of these surveys only go back a few years, their roots can be found (according to Keill) itt an SEC poll before the Crash of ‘46 where advice from Brokers and Advisors showed a bearish per cent of only 4.1%.

Earl Hadady of the Bullish Consensus feels so strongl) about this ittdicaror that he feels (itt his excellent article itt the 1986 11T.A lourttal) that Polling is a third and most importattt tnethod of attalvsis, above Technical attd Fundamental. The basic qu&ott of why investors bought or sold (the public tteeds attswers, the media attempts to fill that need, either in honest attctnpts or in some cases intentionally misleading) is ttot important; rather what the public is really doing, as tnanifest itt the Technical signals of Price and l’olume over Time. Cttfortunately, just as the media and economists range widelv in their beliefs and advice, so do technically oriented ‘gurus and letter writers. As Hadady points out, extreme examples (70% or more) occur less than ottcc a year. If 80% are of otte tnittd, ottlv 1 of 5 traders (especially in zero-sutn Futures markets) hold a cotttra positiott - therefore they are the strong hands of Richard \\‘yckoff’s Composite Operator, or the Big Money that controls markets), itnpervious to tnargitt calls or scared motley and itt tto hurry to get out without a large profit when the majority is sated - as indicated whet1 favorable ttcws now has tto effect. It is at this point that shorts are covered, tnargitts arc full, and complacency is rampant.

In sumtnatiott then, by way of paraphrasing into an anagram, Edwin Lefevrt? in Reminiscences of a Stock Operator, the tnotto F.I.G.H.T. could represent Fear, Ignorance, Creed, and Hope over Titne exetnplifyittg the emotions which we need to control to be the ultimate, dispassionate Cotnposite Operator. or ideal Lradcr.

One way to analyze tnarkets by the notion that there is a “cotttrollittg factor” or IVyckoffian Composite Operator behind tnarket movements l\.as portrayed in a white paper lvritten by Dr. Henry “Hank” Prudett for a class at Golden Gate University He likens the market to a clothing Fashion Cycle wherein otte or tnore top designers in the haute couture world decides a new dress length, style, color is needed, it is thett created and diffused throughout the fashion elite, adopted and itnitated by the general public, until the last housel\ife in a fartn community in the Midwest has given in to the new look. Magazines. stores, media shows have “told” the public what to wear, driving existiug dresses, ties, and other clothing into premature obsolescence. Indeed, if print and television media cau “hype” or market athletic events, songs aud movies, why uot glamor stocks, mutual funds and other securities?

The Indicators

The odd-lot short ratio is derived by taking odd-lot purchases added to odd-lot sales, dividing by two (much like open interest in Futures is obtained), aud dividing that into odd-lot short sales. I did not find this iudicatot au effective contrary tool, especially in relatiori to its success before the current bull market, for the following reasons: only 2 major spikes above 13 occurred in this 12. year time frame (see chart 1). Although both preceded large upmoves, they were the result of a sideways trading range (1986) and a sharp selloff (1990). However, seven other smaller spikes above 10 did not render bull markets. Conversely low readings did riot indicate down moves in the market wtth three exceptions- 1987, 1990. and mid-1991 - versus several that preceded upmoves. Other reasons might include these: smart money was shorting in small odd-lots to avoid the uptick rule, now extant in over-the-counter stocks; some shorting was used in a derivative fashion to hedge arid box positions, more than in the past; mauv odd-lotters with scarce mouey moved to index and equity options over the past fifteen to twentv years.

NYSE SHORT RATIO; S&P 500

Merrill Lynch data 1965-94

Looking at monthly data ou NYSE short interest ratio aud its effect on the S&P 500 Index, historically this was an accurate measure of contrary opiniou. where the earlv adopters of trend xvere correct and profitable, and those at the manic end (see arrows 011 the left side of the bottom part of Chart 2) were 180% wrong. Sharp rallies, abetted by short covering, ensued in cyclic fashion. Once we euded the 17-18 year trading rauge cycle and started the curreut bull market of 1982, things noticeably changed: shorting became and remained excessive, again mostly due to derivative hedging wherein shorts do uot have to be covered aud strong hands do not have to meet margin calls. Another factor to cousider is that currently over 10% of the NOSE is Closed End funds, mostly bond and country types. Still, as the arrows continue to show, rising spikes seem to jibe with up moves 011 the S&P 500, with the one exceptiou.

SPECIALIST SHORT SALES VS. PUBLIC

Mernill Lynch data 1975-94

What appears to be a better indicator of shorting sentiment, although far from perfect, is the Specialist versus Public ratio, shower below (Chart 3). Specialists are the closest persons to buyers’ and sellers’ decisions, although there is a oue to two-week delav in finding their actions. We can observe that not only are the Buy aud Sell signals mostly accurate (B & S not mine), with an occasional misfire (O), but over the long haul, timing market trades would afford you better than 50% gain over buy-and-hold. The “middle clip” iu Charts 4 & 5 refers to the areas between the dotted lines, lower half.

MUTUAL FUND CASH RATIO

ICI - Ned Davis Resenrrh 1978-93

Chart 4 right illustrates how excessive cash can power markets upward while, at least iu a major Bull market, too little doesn’t always correlate to a major decline. One reason for this is that the pressure of short term performance, especiallv with “Money Management Consultants” demanding low cash ratios for clieuts, poses the threat of moving them to another monev manager who will “rotate” the cash into auother sector.

In addition to the fact that excess mutual fund cash does precede rallies, the reciprocal occurrence of mutual fuud buying climax (as depicted in the Ned Davis Chart 5) pre- cedes either substantial declines, or at least long, sidewavs trading rarlges. Obversely from the 1987 Crash until well into 1989, Mutual Fund redemptions exceeded sales throughout that up market, just in time to buy (A] into the next decline (B).

MARGIN DEBT 1967-93

As the long term chart iudicates (Chart 6 with the opposiug arrows) Margin Debt has historically beeu a correct indicator of major tops, especially iu 1973, just before 1982 and dramatically iu 1987. After the 1990 corrcctiotr caused the last Margin debt reconciliation or covering, the chart shows a straight up trend, reflectiug the investiug consumer’s, gownmcnt’s and even global appetite for spcudirrg 011 credit. Although accurate, like maw oscillators the trerrd cm stay iu its extreme mode seemirrgly iudefirritclv 0111/eweb warning of its imminent bursting.

As I mentioned earlier iu the odd-lot short paragraph, wheu the option market got popular, especially iu March 1983 with the advent of the OEX (S&P 100 Index), the least accurate of traders, the uuderfiuauced public, switched from odd lots of stock to optious OH stocks aud indices. At the prcseut time, more than 1500 stocks, OI 7.5% of the stock market capitalization, has equitv options. The uumbcr of sector indices has also burgeoued dramatically It has been commonl!; thought that when put volume heavily outnumbers call volume, this is a contrarv irrdicator that the market OI- uuderlriug entitv will rise. This is true for the short-term day trader; however, looking at the history of the OEX on a weekly basis (Chart 7)) the opposite seems to be true. Over the 12 years, usiug Reuters parameters of below .Z OEX put/call ratio as bearish arid over 1.50 being bullish, Ive cm see the high slumbers are almost alwavs at the top, proving the put buyers correct. Similarly the lower numbers cousistentlv occur at or near the bottoms, when the call buyers would benefit, especially in the mid-198586 span.

Curiously, from August 17, 1987 to October 16, 1987. the OEX put/call ratio was locked in a 60-100 range. actually risirig iuto the last few davs before the crash (theoreticalh bullish). The highest reading ever was in late 1983 - 9.28 - interestiugly just before the big decline of January 1984 of some 30 OEX points.

Being a veteran Option Specialist for the OEX’s largest trading firm and author of an article iu the 1993 MTX ,Journal (#41) 011 V.O.I.C.E., a treatment of OEX Volume aud Open Interest input into a TRIN formula with excellent results (as did,Jim M- . LI tm, Ray Hines, John Bollingcr, aud others in slightly different ways), I was quite surprised bv thcsc lindiugs. Obviously further study using moving averages, could daily data abstracts arc necessarv to verify this conundrum. Looking at current dailv data in the next chart. we do see a more positive correla;iou bctlveeu high put volume, both in the OEX and all-equity CBOE charts, and upward price movement. This is line for dav and short term trading, but I cannot use a high coefficient in my Intermediate, position-trading Master Indicator, for which I am currently collecting data and fine-tuning, possibly for a future paper.

VIX INDEX

CBOE 1983-91

Just a brief word about the L?X Index, Lvhich measures the Volatility of the OEX Iudcx (S&P 100) from mid 1985 011, as shown iu Chart 8 above. It actuallv depicts the Implied I’olatility of 8 OEX optiolk, iu aud out of the moue): uear mouths. Since it has only bceu around in a Bull market, its only volatile during moves of the major uptreud, [bit11 sharp up spike when the market declines, and Ilat or coiliug during trading ranges. Chart 11 shows the historical high of 150 in the Crash of 1987, and siuglc digit loins during 1993 and 1994, possiblv ai1 harbinger of things to come. Although the VIX is verv good for trading strategies (busing or selling options depending 011 the volatility), I find it less useful than the Option Premium Ratio, which combiues put/call sentiment with volatility (see next page).

OPTION PREMIUM RATIO BY CHRISTOPHER CADBURY, 1986-94

A rather recent indicator that has established many valid instances, primarily due to extensive research aud several aiticles by Christopher Cadbury (to whom I owe much gratitude for endless data), is the Option Premium Ratio. This cau only he found in the Sentimcut \\‘indoiv, Chart Page of Investor’s Business Dailv, item #.5, and essentially combines Put/ Call Option seutimeut with Implied Volatility of the 11X, only it includes all equity options, not just the OEX Index. Based on data from 10 years, (although listed options have been around over trventy) di\idiug put premiums by call premiums has ranged from .03 to a hqh of 1.74. Cadbury established that values below .29 and above 1 .I8 indicate a continuation of the trends down and up respectivclv - like extreme levels of other oscillators. Couverscly, OPR’s from 30 to mid-60s generate buy signals and levels to 1.18, sell signals iu about 200 different combinations of occurrcnccs.

Most of these abstracts are proveu almost unanimouslv bv 10 to 20 test examples, such as. “F&k co/eweblseculive davs of gaius or uuchaugcd values for the 6PR starting from .32 to 51 have always produced siguificaut rallies in the stock market”. X fewz however, such as “Ideutical values for the OPR iI1 the range between .SO to .88 separated bv .i to 7 tlavs have alwavs prod&l siguificaut declines in the stock market” have iusufficicut testing aud border 011 the “\\,hencver I wear a red tie 011 Friday the market goes up for 3 days” categorv. Below is a data table of oue of the most hcavilv tested’“pattern recognition” examples: it includes the date the 5-7 day series began and the OPR values; the next four columns list the number of Dow Jones points aud days just before the event, and the uumbcr of points in the subsequeut rally with the numbcr of day or weeks to complete it. More will be heard from 011 this excellent indicator - I intend to include it in My Master Indicator.

SENTIMENT INDICATORS - OPINIONS

The following section discusses the derivation of the 4 major sentiment surveys from newsletters, along with charts which show buy and sell points, and their effectiveness, as shown by %Gains - again, this paper is to review, not research the gathering details. Most effective, I found, were the Market Vane, and AAII newsletters.

1. INVESTOR’S INTELLIGENCE 1966-95

Investor’s Intelligence is published by Mlichael Burke’s Chartcraft. and cxprcsses the opinions of over 100 advisory letters everv week on CKBC arid later in Barron’s. Since 1966, this has beeu ;1u excellent coutrarv indicator with its “trading range” giving its best signals from high 30s (Yo of Bulls) as a Buy sigual and mid-iOs as a Sell. Although the Buy signals have proven verv consistent, the Sell indications, which before 1989 were quite consistent although very early (sometimes several mouths), have been effective in signaling trading ranges as our strong market ellsues.

2. CONSENSUS,INC.1984-94, KANSAS CITY,MO

The Bullish Consensus, from Consensus, Inc. in Kansas City, X10, also uses opinions from advisory services, mostly investment advisors from major brokerages using house organs versus newsletters. These figures also appear on a 900 liuc arid Barron’s 011 Saturday. As Chart 12 shows, there were a few very minor price reversals on major Sell signals, especiall) in the coiling action of both the S&P 500 and the indicator the last 3 years. Still, profits would have bested the market as measured by the Buy-Hold strategy (see upper left corner of chart). As I write this paper, this indicator has reached a four year high of 67 (twice), versus a 71 in the first quarter of 1991.

3. MARKET VANE CORP. 1988-94,PASADENA,CA

An even better sentiment indicator is found in the Market Vane of Market Vane Corp., Pasadena, CA. Comprised of 100 of the top Investment Advisors from Brokers, aud obtained on Monday each week, information appears on a 900 phone number and in Barron’s 011 Saturday of that week. Chart 12 indicates a more precise correlation between reversals, although again the sell signals in a strong Bull market tend to be more of a reaccumulation trading range than SAR (stop aud reverse). Once more, the last several vears resemble coiling action (extension waves of lesser degrees) with lower highs and higher lows in the Indicator. The chart ends with the spring of 1994 correctiou as the O/P lille portends a large upmove in the uear future. During the writing of this paper, it rallied up to 62 for the first time since 1987. At this time, March 25, 1995, it is curiously near midrange, or 47 to 53 area, uot forecasthig the selloffs of the previous 3 indicators.

4. AMERICAN ASSOCIATION OFlNDlVlDUALlNVESTORS SURVEY -1987-95

The final Iudicator of the Barron’s group is the AAII, or American Association of Individual Investors of Chicago, IL, the true retail trade. With 25 postcards mailed out each day of each week, uearly 100 come back with each investor’s opinion of the market for the next six months. As might be expected, this indicator has an al- most perfect correlation exemplifying the aforementioned “crowd” svndrome. Gains Per Atmum show more than 3 to 1 improvement over buy aud hold.

Discussion Section

In assembling and aualyziug all of the above data, what becomes iucreasingly evident is the difference in the time factor of each. After working uearly a year on coustructing a Master Indicator from the most successful of these Sentiment Indicators, it is very apparent that each of them has a different time frame. For example, the timing of the Put/Call OEX ratio is much more short term thau Margin Debt or Mutual Fund Cash. Sot OI+ that, the optimum position ou the Bell/Growth Curve (taken from the work of Everett in 1970) 011 the uext page is quite different. It is only through a corroborating “nesting” of several Iudicators that we cau hope to validate the Master Indicator, which would be a great topic for a future paper. Using Table 2 as a guide, with help from data by Yale Hirsch in his book Don’t Sell Stocks 011 Moudav, I will try to place each Indicator OII the Curve 011 Chart 14 somewhere between A aud E. The graph is a \lodel illustrating a homogeneous population of Investors and sentiment indicators, and uot an actual frequency distribution. The Growth line represents a Price line and an accumulation of the aggregate Indicators, while the Bell Curve depicts Volume as well as the timing phases. Beneath the Bell and Growth Curves I have listed the indicators under studv

Odd-lot shorting would be the highest early in A, with Public entering in the C segment - they would have to cover by C, with the Specialists startmg to short at E.

Mutual Fund Cash would be large at A, fueling the run through D, when it would drop into the single digit percentage. Conversely, Margiu would bc at its low at A, becoming manic at C and D, where the rising slope is sharpest. After au iutensive study of the history of the OEX Index, I can onlv find it useful iu a contrary way 011 a very short term basis. Another look at Chart 7 shows that in almost all cases, except iu tops of 1986 and 1987, high numbers were found at tops, low at the bottoms, meaning traders were correct in the 101lg view. I must say that our current Bull market has had high uumbers from hedging and from those speculators trying to call the top of this market. Similarly, the VIX Index aud the Option Premium Ratio, derived from option premiums rather thau V’olmnc, are short term, a11d lvould therefore be difficult to place on the Chart.

Finally, Bearish Sentiment and gloom from Investment letters and media (magazine covers, financial newspapers and TV) respectively, would be pervasive coming into A; they would gradually mutate into complaccucy through C, aud outright euphoria aud certainty by E.

Conclusion

In conclusion, what I have learned in researching and writing this paper is that although the basic concepts of Sentiment and all of Technical Analysis are eternal, some things do change as markets change. For example, sentiment indicators such as Odd-lot Shorting were rendered less effective by other inexpensive derivatives, such as options.

Also, just as some Oscillators change parameters in Bull versus Bear markets, Sentiment indicators are less reliable in cases like the present, where the stock market dots virtually nothing but rise. with an occasional sideways trading range. Nonetheless, the most effective of the previously reviewed categories, newsletter polling results, mutual fund cash, specialist short selling, and even option put/call ratios, should be monitored for giving reversal signals at extreme excesses, in conjunction with other technical tools such as cycles, oscillators, and support/resistance.

Sentiment is as important as an: other technical tools used by Technical Analvsts. and will continue to be so as we enter’ the area of “Behavior Finance” employing Seural Networks to quantik the Psycholop of Investing.

Bibliography

• The Crowd br Gustave Lc Bon, 1982. Cherokee Publishing Cb.

• The Art Of Contrarv Thinking by Humphrey B. Keill, 1992, Caxton Printers

• Reminiscences Of A Stock Operator hv Edwin Lefevre, 19?3, Doran, Fraser Publish&s

• Don’t Sell Stocks On Slondav bv Yale Hirsch, 1986, Facts On File Publication;

• Sletastock Technician Odd-lot. 1982-94

• Trendlines Odd-lot Short Sales, 1991-95

• SE5E and Specialist Short Sales - XIerrili Lynch D;Gl

• Investment (:ompany Institute Mutual Funds

• .Ned Davis Mutual Fund Buying

• VIX Chart, CBOE (Chicago Board of Options Exchange)

• Option Premium Ratio by Christopher Cadbury

• Merrill Lynch charts on Inwstor’s Intelligence, Consensus, Inc., Market \‘ane Corp., and American Association of Individual Investors

• OEX put/call ratio data, Bloombcrg Sew

• OEX charts - Reuters/Quotron Advantage AE

Overview

This paper examines the concept of modifying the Coppock Curve to better identify major tops attd bottoms in the bond futures market for lottg-term positioning. The modified versiott of the Coppock Curve is referred to as the Z-Trend Oscillator. Most oscillators are used for trading periods of price consolidation, but the Z-Trend Oscillator is specifically used for trading all market cottditiotts from accumulation, to trending, to distribution.

Introduction to Rate of Change and the Coppock Curve

Otte of the older, sitnpler tcchttical ittdicators to understand is the rate of chattge or ROC for short. ROC catt confirm tnarket trends and forewarn of market reversals. The ROC tneasures the pace at which price is chattgittg for any titne period under study. For example, a lo-da) ROC is calculated by subtracting the price today from the price 10 days ago. The result is thett plotted as a continuous series that oscillates above and below att equilibrium level that is usually set at 0. The closittg price is getterally used whett calculatittg the ROC. However, the ROC cat1 be altered to isolate volume attd other ittdicators such as moving averages. Trettdlitte analysis and indicator/price divergettccs arc other aspects of the ROC that catt be used to enhance reliability. With this kind of versatility, traders can mattipulatc the ROC itt mattv useful ways.

Edwitt S. Coppock, best kttowtt for the developtnettt of the Coppock Curve, used ROC as the basis for his work. First introduced in Barrott’s itt 1962, the Coppock Curve was ettdorsed arouttd the world as a long-term ittdicator used to forecast foreign and domestic equity markets.’ The goal of Coppock’s tnometttum rvork is to smooth a price series in such a way as to make the peaks attd troughs in ROC data significant. Smoothed mometttum (referred to as the Coppock Curve) looks attd acts much like a sitte curve or an overbought/oversold oscillator as it moves from positive (overbought) territory to negative (oversold) territory attd back again. Coppock hypothesized that the market’s emotional state could be determined by addittg up the percetttage price chattges for the time period under study to get a settse of tnarket tnometttutn. The result is a long-term curve that effectively measures tnarket momentum and filters out short-term attd intermediate-term tnarket swings.

Because it reflects mass psychology, the Coppock Curve is labeled by tnost technicians attd traders as a setttimettt indicator. As a result, the curve siguals market tops and bottoms quite well and proves to be a valuable addition to atty trader’s tool kit. Coppock combined art 1 l- and 14 mouth ROC, smoothed over by a lo-tnottth weighted moving average, which cat1 be explaitted by the yearly titne cycles frequent in tnost tnarket indices. A buy signal occurs when the curve turns up or becotnes positively sloped while below the zero litte. A sell sigttal occurs when the curve turns down or becomes negatively sloped while above the zero line.

The Problem - Indicator Consistency

When the Coppock Curve is applied to the monthly continuation chart of U.S. Treasury Bottd future prices (UST’s) , traders are confronted with the probletn of indicator consistency. This tneatts that the cottfidettce level for each future buy/sell signal is significantly reduced because the ittdicator’s extremes or overbought/oversold levels vary frotn sigttal to signal. Notice how well chattges in the curve coincide with each major top attd bottom in UST’s (see chart 1). Again, the problem is that the ittdicator does ttot offer a high level of consistency for future buy/sell signals. In other words, traders are not sure how low or high the indicator will go before a signal is given.

That could lead to the wrong position in terms of timing, (see chart 2). 111 chart 2, two examples of false signals that resulted in big losses can be seen in June 1980 and January 1992. Basically, the curve failed to keep traders in a long-term position during these periods of price action.

The Solution - Modifying the Coppock Curve

By modifying the Coppock Curve, traders can isolate overbought/oversold conditions and buy/sell signals more effectively. The added value is a high performance momentum oscillator with fixed buy/sell zones. The Z-Trend Oscillator uses the same basic calculation as the Coppock (;urve. but has a few added dimensions. (1) The indicator is optimized wily OI/ewebC time and then back tested to lind optimum KOC and smoothing periods. Interestingly, the ROC part of the indicator when optimized coincided \\ith the 2 l-week cycle, and the weighted moving average portion coincided with the 40-week cycle, much like the yearly cycles Coppock found in most equity markets. Both the 21-week and 40-week cycles were popularized by Jim E. Tillman, CMT, of Interstate/Johnson Lane, and are frequently used in forecasting turning points in UST’s. (2) The indicator uses the concept behind J. Welles Wilder’s Relative Strength Index (RSI) to identify overbought/oversold conditions. III other words, the Z-trend Oscillator is a11 RSI study of the Coppock Curve. First, the raw numbers of the Coppock Curve are substituted for the usual closing price within the RSI calculation to normalize it on a scale of 0 to 100. Second, this modified version of the RSI is multiplied bv itself and then subtracted from 100 to make it oscillate above and below 0 and between defined Lanes. A&l example of defined zones would be between -70 and 70. This will identifv proxies of overbought/oversold. Once that is accompli&ed, buy/sell signals become more visible. This is where the Z-Trend Oscillator becomes useful. It maintains a long-term position during a sideways trend by decelerating as the market’s price ranges become more narrow. (3) The buy/sell equation is written by combining the slope of the indicator with overbought/oversold extremes to determine if a profitable trade exists (see table 1 for calculations).

There are four advantages to using the Z-Trend Oscillator over the Coppock Curve: (1) It is smoother and less volatile, (see chart 3 for a comparison). (2) The amplitude is controlled through the modified version of the RSI formula. (3) Major buy/sell signals become more visible. (4) Overbought/oversold zones are identified. The only disadvantage noticed is: (1) It is iueffective when used over shorter time periods such as weekly, daily arid intra-da):

Using and Customizing the Z-Trend Oscillator

A buy signal occurs when the Z-Trend oscillator (this month) is greater than it was (last month) and is less than -40. A sell signal occurs when the Z-Trend Oscillator (this month) is less thau it was (last month) arid is greater than .iO. Lead time is significantly increased over using Coppock’s original buy-sell strategy (see chart 4). X more couscrvative buy/ sell approach would be to wait until the indicator crossed above or below the 0 line. However, the Lero lille trade reduces profit and iucreases risk because siguals occur well after a top or bottom has beeri complete.

A histogram is used to display the indicator for clarity but is just as accurate /ewebvherl displayed as a liue chart. A 5moutil simple moving average of the indicator cau be used to help cuulirm market direction. Since this moving acragc is ouly used as a coufirmation tool and not part of the buviscll equation, it was uot optimized. If the indicator is above lhc 5-month simple moviug average, there is buviug coulirmatiou. 011 the other hand, if the indicator is below the 5-mouth simple moving average, there is sclling co1lfirmatioIl. AII expolleutial moving average could also be used in place of the simple moving average. Traders are eucouragcd to expel-iment with other moving select-ages and time periods because trading styles vdrv.

Applying the Z-Trend Oscillator to Market Conditions

From late 1977 to early 1996, the Z-Treud Oscillator generated a total of 0 ollt of 9 wimling trades. (see chart 5). From September 1982 to Jlarch 1983, the UST market was considered extremely overbought, which was evident by an indicator reading of greater than 70. That was a waruulg sign suggesting traders should start looking for a new sell signal. In fact, the signal was giveu in April 1983 when the indicator began to decelerate while still above the trigger level set at 50. The result was a 1Gmonth trade Ccldiug 12.69 points. From September 1987 to November 1987, the Z-Treud Oscillator rcachcd a rcadiug of less thau -70, which warned of a potential market reversal from down to up. This oversold rcadiug bcgau to unwind in December 1987 when the indicator touched the bul trigger level set at -40. The result \va?s a 24-mouth trade for 10.25 points.

Other examples of the Z-Trend Oscillator iI1 action can be sew during 1983, 198.5, 1988, 1991 alld 1994. These periods were major cougestiou zones, but uoticc that the indicator kept each position active by not getting too overbought or too oversold during each of these periods. 111 fact, the indicator hovered closer to the zero liue when price rauges became more uarrolv. 01lce price ranges begau to widen and the market resumed its original directiou. the iudicator accelerated. This is an important aspect \vhcu dealing rvith long-term tiuiiug iudicators because false buy/sell signals tend to occur during a nontrending (consolidation) period. [Note: During a sidcwavs price trend, the Z-Trend Oscillator’s maximum draw dowl (licgative open profit/loss) cau kcomc greater. This is caused by slippage/ commission aud price volatility Hoivever. these losses are kept to a miuinlum and arc Iiltcrcd out ouce the price trend resumes iI1 its original direction.]

Testing The Z-Trend Oscillator

Since UST futures oulv started trading in late-19Ti. the best way to show that the Z-Trend Oscillator is uot a “by chance” indicator is to conduct a number of tests (see chart 6 for the indicator and table 2 for the results). In the first test, the Coppock Curve was applied to UST’s using Coppock’s ll- and 14month ROC, smoothed over by a lo-month weighted moving average. Also used rvas Coppock’s buy/sell strategy in an attempt to show that this indicator needs to be modified in order to work properly when applied to UST’s. Conclusion: Due to the number of false signals and poor results, the only solution was to modifv the curve. In the second test, the Z-Trend Oscillator was used, but the ll-, 14 and lo-month parameters were maintained.

Buy and sell signals were also modified, (see chart 7 for the indicator and table 3 for the results). Conclusion: The results were a little better, but false signals were still present. Therefore, the onlv solution was to optimize the ROC and smoothing peiiods.

In the third test, the %-Trwd Oscillator was optimized IO liud optimum ROC alld smoothing periods (see chart 8 for the indicator and table 4 for the rcsultsj. The optimization process resulted iu a fi- and H-month ROC. smoothed over bv a lOmonth lveightcd moving average. Conclusion: The’optimizatioll process enhanced the e6 ICM fectlveness of the Z-Trend Oscillator, which is reflected in the results. However, to show that the 6, S-and lo-month parameters were valid, the Dow Jones-20 Bond Average (DJ-20 Bond Avg), a proxy of LIST’s going back to 1915, was tested.

The first test from 1915 to 1946 was a partial success. The &, S-and lO-month parameters worked extremely well. Howewr. the indicator failed to react to the different market conditions. The second test from 1947 to 1977 leas also a partial success. The 6-. 8- aud lOmouth parameters generated excellent results, but the indicator failed to react as it should have. given the different market conditions. The third test from 19% to 1996 was a complete success. The 6, 8- and lOmonth parameters worked extremely /ewebvell, aud the indicator operated properly given the different market conditions.

Conclusion: Though the Z-Trend Oscillator seems to work better on UST’s than on the DJ-20 Bond Avg, the results are very encouraging. Since the Z-Trend Oscillator managed to signal every major top and bottom over an 80 year period in the DJ-20 Bond Avg, the results revealed that buy/sell signals did not occur “by chance.” In an attempt to show that the Z-Trend Oscillator can significantly increase profit potential, one last test was conducted that revealed using the Z-Trend Oscillator is more profitable than using a simple buy-and-hold strategy, (see table 8 for the results).

Conclusions

Based on the results of all tests, it was well worth the time and effort to construct, customize, optimize, back test, and update the Z-Trend Oscillator. The tests show that a complete and effective indicator can be used to signal every major top and bottom in UST’s. Traders now have a long-term indicator within their technical arsenal that can (1) Consistently identify an overbought/oversold condition within the market; (2) Locate buy/sell signals with a much faster lead time; (3) Identify the long-term trend of the market. Traders who follow other markets could try this indicator on those markets. Note: The optimized parameters would most likely be different in other markets and overbought/oversold conditions could also vary. For example, instead of being overbought at 70 and oversold at -70; a market could become overbought at 50 and oversold at -50. Traders should consider isolating dominant time cycles within the market and using those cycles in place of the ROC and smoothing periods.

Long-term trend analysis is a very important aspect of technical analysis, and if done correctly, there is no reason why traders shouldn’t be on the right side of the trend. During the research, a few areas that warrant further attention were discovered: (1) Using indicator and moving average crossovers as the basis for the buy/sell equation. (2) Fitting the indicator to weekly, daily and intraday time periods. (3) Applying the indicator to commodity markets, currency markets and mutual funds.

Despite these minor troubling aspects, the Z-Trend Oscillator can do what was once thought impractical: consistently signal major tops and bottoms, and identify the trend for long-term positioning.

Bibliography

• Colby, Robert W. & Meyers, Thomas A., The Encvclonedia of Technical Market Indicators , 1988, Business One, Irwin, p. 414.

• Faber, Bruce R., “The Rate of Change Indicator,” Technical Analvsis of Stocks & Commodities, Volume 12; October 1992, p. 13.

• Hayes, Tim, “The Coppock Guide,” Technical Analvsis of Stocks & Commodities, Volume 11; March 1993, p. 50.

• Kemplin, Raymond, “The Coppock Curve: A Famous Indicator Flashes a Long-Term Buy Signal,” Barron’s, November 22, 1982, p. 10.

• Middleton, Elliott, “The Coppock Curve,” Technical Analvsis of Stocks & Commodities, Volume 12; November 1994, p. 59.

• Pring, MartinJ., Martin Prine on Market Momentum, 1993, Probus Publishing, p. 52.

• Wilder, Welles J., New Concerts in Technical Trading Systems , 1978 , Trend Research, p, 112.

From my earliest readings on technical analysis at the genesis of my investment career, to my present day more experienced view of the markets, I have learned that the role of volume in security analysis plays an important role in predicting the future path of stocks. Despite the belief that this role of volume analysis was true, somewhere in the stubborn Missouri - like recesses of my mind persistent doubt existed as to the actual validity of this concept. Yes, we have all seen stocks blast off with high volume and continue to advance. But as a group did they perform any better than the averages? From my readings there existed no recollection of a specific study correlating the subsequent price action of a stock after an unusual volume occurrence. Many authors have stated that price follows volume, but is this an educated opinion or is it a fact? Joseph Granville, in his book Granville’s New Stratew of Dailv Stock Market Timing for Maximum Profit’, states that “stocks do not rise in price unless demand exceeds supply. Demand is measured in volume and thus volume must precede price.” Although Mr. Granville was writing about his technique of “on balanced volume,” also known as OBV, which is an accumulation of positive or negative volume over a certain period, accumulation and distribution volume patterns point to probable changes in price. It is not the purpose of this paper to prove or disprove the usefulness of OBV, per se. The reference to Granville’s OBV is to only lay the foundation for the basic belief that most technicians adhere to the concept that volume precedes price. However, in this paper a tack is taken that is different from Granville’s approach. The primarv goal of this paper is to offer evidence that a sample of iust one tradine dav of unusual volume can predict subseauent price action.

Martin J. Pring, in his book Technical Analvsis Explained*, asserts the principle that volume goes with price. Most technicians willingly accept this principle at face value. Mr. Pring observes that a price rise accompanied by expanding volume is a normal market characteristic. He also writes that a breakout from a price pattern that occurs on heavy volume, especially on the downside, acts to confirm the price trend. Why do most technicians accept the concept that volume confirms the price trend? Or do they simply wish to believe it to be so? Indeed, what are the true probabilities of price following volume?

In this study, we make a major departure from the usual pattern of relating trend of volume over extended time periods to the study of a single occurrence of abnormal volume. Data were collected over the course of eighteen months in which stock prices, after an unusual day of volume and price behavior, were compared to the S&P 500 Index.

The data studied in this paper use either a combination of volume and price alerts or simply volume alone. The study does not incorborate any bersonal technical iudments. The study quantifies the action of stocks during the one year period of July 10, 1991 through July 9, 1992. The level of the S&P 500 was recorded along with the entry of each stock that qualified for the study.

Methodology

Before delving into the results of the study, a review of data sources and review methods shall be presented. We are cataloging securities whose shares demonstrate unusual volume characteristics and a combination of unusual volume and price characteristics. The study gathered data drawn from the stock listings in The Wall Street Tournal beginning with the letter “A” on the New York Stock Exchange. The “A” section, chosen for our sample, represents approximately 13% of all stocks listed on the NYSE. We deem this a meaningful sample for our exploratory considerations. The stock listings in The Wall Street Tournal underline the issues that are among the 40 largest percentage changes in trading volume, compared with average daily trading volume over the past 65 days. These issues are labeled “volume alert” stocks. Stock quotations in boldface have experienced price moves in excess of 5%. These issues are labeled “price alert” stocks. Issues printed in boldface type and underlined are labeled “volume& price alert” stocks, which signals that the stock had both a volume and price alert. The study breaks down into two specific areas of volume activity: first, on stocks exhibiting a volume alert (underlined), and second, on stocks exhibiting a price 8c volume alert (boldfaced and underlined). These two types of occurrences are the focus of study. Prices of securities and the S&P 500 were measured on the opening price the day after an alert was noted. The study excluded stocks under $7 l/2 and preferred issues. Measurement periods were one, three, and six months. The total number of occurrences compiled during the period studied totaled 375. The collection of alerts occurred for a period of one year and then an additional six months to collect the data for the longest comparison. The best and worst performing stocks in each group were disallowed in an effort to reduce any unusual distortions.

Each study covered stocks with up and down alerts. For example, an underlined security closing up for the day qualified as an up-volume alert. The two methods are:

1. Stocks demonstrating a volume alert, both positive and negative.

2. Stocks demonstrating avolume & price alert, both positive and negative.
   

Each of the foregoing study methods constitutes the remainder of the composition. If volume is believed to be an indicator of subsequent price action, then the sections labeled as down alerts represent periods in which the securities should perform inferior to the market. For example, in advancing markets the down alert stocks should lag behind the market. In declining markets, they should fall further than the market decline. In the up alert study, the securities’ performance should exceed the market’s performance. The S&P 500 is the measure of the markets performance. The percentage gains or losses present a net average of all stocks, as well as the S&P 500 for each category. Periods are broken down into categories of up or down alerts, not of a rising or declining market. In real time it is obvious that one knows only the direction of the alert.

Years ago I began trading securities that broke out to new highs from any of several different types of consolidation patterns. My actions, then and now, are no diffeent from those of many other technicians. Merrill Lynch Inc., for example, with whom I was an account executive for eleven years, publishes a report each morning called Dailv Market Observations3. One of the items covered in this report is a listing of stocks that demonstrate unusual volume characteristics, categorized by direction of price movement. The report, compiled by Philip L. Rettew, vice-president at Merrill Lynch, is an important source of volume data. In a later telephone conversation with Mr. Rettew regarding this report, I asked if a study had ever been undertaken to study the after- effects of positive and negative volume alerts. His answer was, “no, because that is not the objective of the report. It is published for the reason that its name implies. It is used for observations.” Mr. Rettew went on further to say that occurrences such as selling or buying climaxes, new offerings and programs trades might run counter to the perceived trend of the observed alert. These factors would make it all the more difficult to measure the performance. A high volume selling climax and the subsequent negative alert that it demonstrates should, and in most cases, actually be interpreted as a potential positive signal by the alert technician. Mr. Rettew says that “the report tries to identify or observe occurrences from a daily sea of data in which one should further explore.” Our study will attempt, despite the instances of buying and selling climaxes and other market noises, to determine if unusual activity in volume and price are worthy anticipators of subsequent price movements. In our case, we make observations without judgment.

Volume Alerts

The tables that follow are arranged so that each stock is measured against the corresponding movement in the S&P 500 index. For instance, in the first table the section for down alerts under the one month heading rose 1.2% compared to the S&P 500’s rise of .82%. We would wish, if volume is a useful indicator, to have the down alerts stocks decline more than the market. Following each period a 20% stop loss was also measured in an attempt to counter bad signals. It is a commonly held belief by most traders that the use of stop loss orders can effectively increase one’s odds of being successful in trading. This concept is utilized in this report simple as an easy addition to view the results with a small dose of money management.

Under the down alert category, the results were disappointing because the securities rose more than the market despite a negative volume alert. The three-month study was a particularly rough environment for the down alert securities because they actually increased in value during a declining market environment. The six-month period, although showing a smaller difference in percentage changes, also failed to perform as desired.

As an adjunct to the study, the 20% stop loss program was conducted, which significantly improved the desired results. Although massaging the results with a stop loss program does not necessarily add validity to the test results, it does add strength to the case for the use of sound money management principles. The employment of the stop loss narrowed the results for the one-month period to nearly neutral results. For the three-month period the results improved as the down alert stocks declined by 0.17% rather than rising by 1.86%. The six-month study showed the greatest improvement as the down alert stocks rose only 0.34% as compared to the nonstop loss gain of 4.55%. Although it might appear that the improvement of results by the use of a stop loss program might infer that the volume alert does not work, nothing could be further from the truth. The improvement of the results by the stop loss actually reinforces the significance of the results. Because a stock signals a report, perhaps in many cases a selling climax after a long decline, or even a buying climax after a long uptrend, it alerted us to a developing and changing situation. The stop loss simply prevented the continued loss of funds. The well used phrase to cut your losses and let your profits run still applies.

As illustrated in table #l, the up alerts produced desired results for all periods. The one and three-month studies managed to beat the averages by several percentage points. The three-month gain of 4.12% favorably compared to a market loss of 1.69%. The disparity represents a sizable difference not explained by chance. The gain from the three-month period to the end of the six-month period declined to only a 0.39% gain in favor of the studied stocks.

The use of the stop loss program again supplemented the desired results. The use of the stop loss improved the three-month results by roughly 20% from a gain of t4.12% to a gain +5.01%. The six-month results benefited by 44% as the gain increased from t3.94% to t5.76%

VOLUME ALERTS COMMENTS

From the data gathered in this study it is clear that the use of a positive volume alert can benefit the trader of securities. As to why the positive alerts performed better than the down alerts, one can only guess. Many of the negative alerts appeared climactic in nature and could easily explain part of the difference in less than anticipated results. As noted the use of the 20% stop loss did improve the results in most categories.

Price & Volume Alerts

The second part of the study combines the volume alert with the price alert. As previously noted, a price alert is demonstrated by a price move in excess of 5% in one trading session.

The following table outlines the results of the combined results.

In every period measured, except for the one-month up period, the alerts worked in their direction of signal. The down alerts worked in an exceptional manner by under-performing the S&P 500 by wide margins, even deQing the market’s general direction. For instance, the down alert stocks declined in the one and six-month periods despite an advancing market. During the six-month period the advantage to the trader was six percentage points. Had a trader shorted all of the stocks in the six-month study, he would have had a positive return of 3.88% during a period in which the market rose 2.14%. In each case, the stocks declined in value even though during the one and six-month studies the market rose. Evidently securities with a sharp rise in price accompanied by a sharp increase in volume performed as expected.

The evidence is also compelling for the up alerts, as they best the market by large margins. In the three-month period the alerts rose 2.16% during a market decline of 2.11%. During the six-month study, the alerts did nearly 33% better than the market.

The use of the stop loss increased the results in all cases. For instance, in the three-month up alert study, the results increased by 50% from a +2.16% to t3.08%.

PRICE & VOLUME ALERT COMMENTS

The results obtained when combining both volume alerts and price alerts in conjunction appear to be more than just random results. When the alerted securities go counter to the market by wide margins, then indeed the proof of a one-day event effecting future stock prices is hard core evidence of the theory’s validity.

General Thoughts & Conclusion

From the data shown in both categories of study, it is clearly evident that volume does indeed precede price in the sense that high volume alerts lead to improved performance of securities in relation to the market. It is more evident that a combination of unusual volume combined with a 5% or more price movement improves the random results of equity trading even further. The trader should carefully examine each alert to determine its best use. In many of the stocks measured, a signal occurs, which to the average technician represents a selling or buying climax. In reality, most technicians would probably not trade the signal, and some would have been tempted to go in the opposite direction of the alert. These contingencies in no way weaken the case for the use of using volume alerts for trading. It is not, nor was it ever contemplated, that this study should lead to a system of trading securities based solely on volume or price & volume alerts. However, if one wished to utilize the methods presented as a system, it is certainly plausible. The odds of success appear to be favorable.

The primary intention of this study is to examine the effects that a one-day volume event might cause to future price changes. The evidence presented makes a strong case that volume alerts, either singularly or with price, should not be ignored. It is a tool that can be used to alert the technician of possible impending change. Perhaps the alert is part of a larger technical pattern that is signaling the end or beginning of a technical pattern. It is a sign to investigate.

After collecting and analyzing the data over a period of l-1/2 years, and after trading in real time some years later, I find that I am constantly searching for unusual price and volume behavior in any stock. In my own particular trading methods, I look for several types of consolidation patterns for equity purchases. A volume or volume/price alert signals me to scrutinize the security. Often the action may be signaling the commencement of a new trenc or perhaps the death of an old trend. The individual trade must make that call. The investor will find that in addition to the previous day’s trading results, he or she wil find a useful technical tool that, if used consistently, car increase trading profits. Earlier in the paper, we state that the purpose of the study was to determine if the ex amination of our “alerts” would prove their utility as a tech nical tool. The data presented points to a positive conclusion.

The Final Word

A popular theory among technicians is that price follows volume. The evidence pathered in this study clearlv indicates that it would be we appropriate to state that brice follow an explosion oftice and volume. The initial kickoff of largl price moves, and higher than normal volume, leads tc continued outperformance in equities. It does not appear to be important if the volume signal was positive or negative.

Bibliography

1. Granville, Joseph Ensign, Granville’s New Strategy of Daily Stock Market Timing for Maximum Profit. Prentice-Hall, Englewood Cliffs, NJ, 1976

2. Pring, Martin J., Technical Analysis Explained. McCraw Hill, Inc., New York, 1980

Services

3 Daily Market Observations, Merrill Lynch Inc., One Liberty Plaza, New York, NY