Technical Analysis Explained

by Martin J Pring

The strongest period in the cycle is centered on years ending in a five and extends until the early part of the seventh year. There is a distinct upward bias beginning at the tail end of the seventh or the middle of the eighth year, which runs through to the third quarter of the ninth. Indeed, Table 24.1 shows that 9 years rallied 10 times from a total of 13, only being eclipsed by the super 5 years, which have a 12-1 rally advantage. That exception developed in 2005, but would not be an exception had we used the S&P Composite in our calculation. The weakest years are those ending in a seven or a zero.

  TABLE 24.1 The 10-Year Stock Market Cycle

  Chart 24.9 compares the average cycle with the S&P Composite in the opening decade of the current century. By and large, the trajectories were quite similar until the 2008 financial crisis began to unfold and the market sold off sharply instead of rising. This discrepancy once again offers a timely warning that however promising a historical pattern or relationship may be, it is mandatory to form a weight-of-the-evidence conclusion because in the world of financial forecasting nothing is predetermined except the recurring cycle of fear and greed! Moreover when using these types of cyclical patterns and the cycles themselves for analysis greater emphasis should be placed on direction rather than magnitude.

  CHART 24.9 S&P Composite versus the Decennial Pattern 2000–2009

  The 41-Month (4-Year) Cycle

  The so-called 4-year cycle is a 40.68-month (41-month) cycle. It has been observed to operate in stock prices since 1871. Around 1923, Professor Joseph Kitchin was also able to show a cycle of 41 months in bank clearings, wholesale prices, and interest rates in the United States and United Kingdom. This cycle has since carried his name.

  The Kitchin cycle as applied to stock prices is illustrated in Charts 24.10a and b. Between 1871 and 1946 it has occurred 22 times with almost uncanny consistency. Then in 1946, as Edward Dewey describes it, “Almost as if some giant hand had reached down and pushed it, the cycle stumbled, and by the time it had regained its equilibrium, it was marching completely out of step from the ideal cadence it had maintained for so many years.”4

  CHART 24.10a The 41-Month Rhythm in Stock Prices, 1868–1945

  CHART 24.10b The 41-Month Rhythm in Stock Prices Reversed, 1946–1968

  The 4-year cycle can also be observed by looking for a major buying opportunity every 4 years, and in this way is arguably the most reliable of the cycles described here. Charts 24.11 and 24.12 show that this usually develops after a decline, such as in 1962, 1966, 1970, 1974, 1978, 1982, 1990, 1994, 1998, and 2002. Sometimes, as in 1986 and 2006, the market is very strong and the buying opportunity develops after a sideways consolidation. The 4-year cycle year of 2010 also developed a buying opportunity, but this time it came after an intermediate decline in the middle of the year. The charts center the cycle in 1921 and 1974, respectively, so that the vertical lines represent the idealized lows. The important thing to notice is that the actual buying opportunity lows develop close to either side of these points. Many of them closely intersect the vertical lines as well. The most notable failure developed in 1930 about halfway down the greatest bear market in history. Once again, this anomaly indicates the fact that a particular indicator or cycle that has operated successfully in the past is no guarantee that it will continue to do so in the future.

  CHART 24.11 S&P Composite, 1910&1958, 4-Year Cycle

  CHART 24.12 S&P Composite, 1959–2018, 4-Year Cycle

  Presidential Cycle

  The presidential cycle is closely allied to its 4-year counterpart but is more of a pattern than an actual cycle with two lows separated by a cyclic high. The presidential cycle can be split into 4 separate years like the presidential term. In the first year, presidents like to do some economic housecleaning to prepare for better times as the subsequent election approaches. Wars also tend to develop in the first half of the term. The most recent example was the Iraq invasion, which began in late 2003, although strictly speaking, that took place very early in the second half of the term. The Afghanistan surge by President Obama was advertised in December 2009 less than 1 year into his presidency. The point here is that the first year of the cycle involves economic pain and stress and is a drag on equity prices. In Stock Market Cycles (Wiley, 2002), Jeffrey Hirsch lays out the fact that between 1833 and 2011 the market gained 724 percent in the last 2 years of the presidential cycle compared to 273.1 percent for the first two. By way of contrast to the austerity of the first 2 years, the preelection and election years are characterized by pump priming. The percentage return for each of the 4 years in the cycle is shown in Chart 24.13. Returning to the first year we find that a number of bear markets begin at such times. Examples include 1929, 1937, 1957, 1969, 1973, 1981, and 2001, although in the case of the latter the actual high was registered in 2000.

  Chart 24.13 also shows that the third year of the cycle is by far the best. Indeed, there has not been a down year in the third year of a presidential election year since the outbreak of World War II in 1939, when the Dow Jones Industrial Average (DJIA) lost 3 percent.

  CHART 24.13 Presidential Cycle, 1833–2011, Annual Returns

  Chart 24.14 lays out the average performance for each year of the cycle. Here it is evident that the period that starts at the end of the mid-term year and ends toward the final quarter of the preelection year is statistically the strongest. When combined with the strongest year in the decennial cycle (1915, 1935, 1955, 1975, and 1995), an explosive rally typically develops every 20 years. The manuscript for this book is being prepared in early 2013, so it will be interesting to observe how the next fifth decennial third presidential combination due in 2015 will behave.

  CHART 24.14 Composite Presidential Cycle, 1900–2010

  Seasonal Patterns

  There is a distinct seasonal pattern of stock prices that tends to repeat year after year. Equities seem to have a spring rise, a late-second-quarter decline, a summer rally, and a fall decline. Apart from seasonal changes in the weather that affect economic activity and investor psychology, there are also some seasonal patterns in financial activities. For example, July and January are heavy months for dividend disbursement, retail trade around the year’s end (Christmas) period is the strongest of the year, and so on.

  Stocks purchased in October have a high probability of appreciating if held for a 3- or 6-month period, as that month ends the worst-performing 6 months of the year. It has also posted more bear market lows than any other month. Not surprisingly, the November–January period offers the best-performing consecutive months of the year (S&P average gain of 4.3 percent since 1950). Failure to do well in this part of the year is often a sign of trouble. The late Edson Gould observed that “if the market does not rally, as it should during bullish seasonal periods, it is a sign that other forces are stronger and that when the seasonal period ends those forces will really have their say.”

  Close to the center of this 3-month bullish seasonal is the period surrounding Christmas marked by the last five trading days of the old year and embracing the first two of the new year. This is the so-called Santa Claus rally, which according to Jeffrey Hirsch in Stock Market Cycles, has averaged 1.5 percent since 1953. Yale Hirsch, Jeffrey’s father, coined the expression “If Santa Claus should fail to call, bears may come to Broad and Wall.” Examples include 1968, 1981, 2000, and 2008, all of which were followed by bear markets. This, of course is a great example of Edson Gould’s failed seasonal principle outlined above.

  Chart 24.15 represents the seasonal tendency of the stock market to rise in any given month. The probabilities were calculated over the twentieth century by Tim Hayes at Ned Davis Research (see Table 24.2). All movements are relative, since a month with a strong tendency will be accentuated in a bull market, and vice versa. It is also important to note that the direction of the trend is more important than the level.

  CHART 24.15 The Seasonal Pattern in the Stock Market

  TABLE 24.2 DJIA Monthly Performance Since 1900

  The January Barom
eter was originally devised by Yale Hirsch in 1972. Simply stated, the indicator adheres to the maxim that “as goes the S&P in January, so does the whole year” (Jeffrey Hirsch, Stock Market Cycles, p. 143). According to Hirsch, the barometer has an 88.7 percent accuracy.

  One final comment on seasonality derives from the fact that the May to October period has the worst track record. “Sell in May and go away” has some statistical merit on its side. For instance, Jeffrey Hirsch calculates that a hypothetical $10,000 investment in the DJIA compounded to $674,074 for the November–April period, compared to a $1,024 loss over a 62-year period. That does not mean that every May–October period lost money, nor that every November–April one made money, but it does indicate a strongly positive bias toward the November–April period.

  My friend Sam Stovall at S&P Capital IQ takes it one step further. He points out that two good performing sectors in the negative May–November period are the defensive consumer staples and health care. Alternatively, the higher beta materials and industrials outperform in the bullish November–May period.

  End of the Month Bullish Seasonal

  The year-end effect of superior returns also seems to apply to the month’s end. Data covering the 89-year period ending in 1986 show that returns from the last trading day of a month (day 1 in Figure 24.5) to the end of the third trading day of the new month are consistently good. The rationale for this effect may well come from higher month-end cash flows, such as salaries, dividends, etc.

  FIGURE 24.5 Turn of the Month (Average Daily Returns)

  Indeed, these four trading days average 0.118 percent, versus 0.015 percent for all trading days. Turn-of-the-month returns can be said to account for all the positive capital gain returns generated by the market. In an article entitled “Calendar Anomalies,”5 Bruce Jacobs and Kenneth Levy point out that this effect was less prevalent in the 1980s, which goes to show that it is not a wise policy to follow one indicator exclusively. In the 2001 edition of the Stock Traders Almanac, a must-read for the seasonally oriented trader, Yale Hirsch points out that this seasonal indicator shifted between 1981 and 2000 to the last 4 and first 5 trading days of the new month. The most bullish day of all appears to be the first day of the month, according to the Almanac’s 2013 edition, as the editors point out that in the previous 15 years the DJIA gained more points on that day than all the others combined.

  However, it does make sense to integrate this reliable long-term seasonal effect with short-term oscillators. Clearly, the potential for the market to advance at this time will be much greater if it is oversold going into the last (presumably) bullish day of the month.

  Days of the Week

  The term “blue Monday” is very much justified. The influence of weak Mondays originated during the 1929–1932 crash. During the Depression, the market advanced, on average, every day of the week except Mondays. It could be said that the entire market decline took place over weekends during the periods from Saturdays to closings on Mondays.

  Figure 24.6 shows the average return for each day from 1928 to 1982. Monday is the only down day. Remembering that this takes into account “black Thursday” in 1929 but does not include the 500-point drop that occurred on “black Monday” in 1987, it just goes to emphasize the point.

  FIGURE 24.6 The Day-of-the-Week Effect (Average Daily Returns)

  There does not appear to be any acceptable rationale for this effect, which also occurs in non-U.S. equity markets, debt instruments, and even orange juice.

  Pre-Holiday Advances

  The day preceding holidays is statistically a bullish period. This is indicated in Figure 24.7, which covers the period between 1963 and 1982. With the exception of Presidents’ Day, all these (average) pre-holiday trading sessions handsomely beat the average day.

  FIGURE 24.7 The Holiday Effect (Average Pre-Holiday Returns)

  Time of Day

  Recent studies6 have indicated that there is a definite time-of-day effect, as shown in Figure 24.8. There is little difference in the activity from day to day, except for Monday mornings. All days, however, show an upward bias going into the last half-hour. The study showed that this rallying effect was emphasized even to the closing bell, with the average return of the last trade equal to 0.05 percent, or 0.6 cents per share. The nearer the return took place to the closing bell, the higher it was. Trades after 3:55 P.M. averaged 0.12 percent returns, or 1.75 cents per share. That upbeat note is a good place to close this chapter.

  FIGURE 24.8 Time-of-the-Day Effect

  1 From Smith, Tides and the Affairs of Men, p. 55ff.

  2 The cycle was also noted by Professor William Stanley Jevons, an English economist, in the second half of the nineteenth century.

  3 From Smith, Tides and the Affairs of Men.

  5 MTA Journal, winter 1989–1990.

  6 Harris, “How to Profit from Intradaily Stock Returns,” Journal of Portfolio Management, winter 1986.

  25 PRACTICAL IDENTIFICATION OF CYCLES

  This chapter discusses some of the basic principles of cyclic analysis and uses examples to illustrate some simple techniques that help in their identification.

  Cycles Defined

  A cycle is a recognizable price pattern or movement that occurs with some degree of regularity in a specific time period. A market, stock, or indicator that has a relatively consistent price low at 6-week intervals is said to have a 6-week cycle. That successive lows are higher or lower than their predecessor is of no importance in identifying the cycle. What is significant is that there is a clearly definable “low” point every 6 weeks, separated from its predecessor by a high point known as the cycle high. Figure 25.1 shows some possible examples.

  Figure 25.1 also shows that while cycle lows occur at approximately 6-week intervals, cycle highs can vary. Occasionally, they arrive early, as at point A; sometimes, they occur in the middle of the cycle, as at point B; but, they may also appear late, as at point C. Generally, when the cycle high develops shortly after the cycle low, the implications are that the upward part of the cycle is weak and that its overall strength lies on the downside. In this situation, each cycle low is normally below that of its predecessor. Similarly, a cycle high that is “late” in arriving, i.e., that arrives well after the halfway period, usually indicates a strong cycle, with the implication that the low will be above the low of the previous cycle. A number of different cycles can be observed for any market or stock, some long and some short in duration. The task of the technician is not to identify as many as possible, but to isolate the most dominant and reliable ones.

  FIGURE 25.1 Cycle Highs and Lows

  Principles

  There are several principles:

  1. The longer the cycle, the greater the amplitude in price is likely to be; e.g., a 10-week cycle will have far greater trading significance than a 10-hour cycle.

  2. It follows from item 1 that the larger the cycle, the greater the significance of the low.

  3. The larger the number of cycles reaching a low at around the same time, the stronger the ensuing price movement is likely to be.

  4. In a rising trend, the cycle high has a tendency to “translate to the right,” i.e., to occur after the halfway point of the cycle. The same principle holds in reverse for bear markets; i.e., there is a tendency for the cycle high to translate to the left.

  5. It is possible to observe cyclic highs that occur at regular time intervals.

  6. A projected cyclic high or low may develop in the opposite way to that anticipated. In such cases, the cycle is said to be “inverted.”

  Methods of Detection

  Many mathematical techniques have been used to identify cycles. Fourier analysis, for example, isolates the existence of various cycles by length, amplitude, phases, etc. Systematic reconnaissance is a technique that tests for periods requested. The result is a period gram that shows the most dominant cycles. Although such techniques can be useful, they tend to make technical analysis look as if it is an exact science, which it ve
ry definitely is not. This chapter will be confined to three methods of cycle identification: deviation from trend, momentum, and simple observation.

  Deviation from Trend

  This method takes a series of data and divides each item by a moving average (MA). The period under observation represents the deviation, and the MA represents the trend.

  Chapter 11 explained that since an MA is designed to reflect the underlying price trend, ideally it should be plotted halfway along its span. This is because the “average” price occurs halfway through the time span, e.g., in the seventh week for a 14-week MA. However, changes in direction of the MA usually occur far too late to offer timely signals for the purpose of identifying trend reversals. For this reason, technicians normally use an MA crossover for generating signals. Since only historical data are used in cycle identification, this disadvantage is not important. The MA deviation is, therefore, calculated by dividing the period in question by the midpoint of the MA. The price observation for February 27 is divided by a 13-week MA, as calculated on April 18; i.e., the MA is “moved” back 7 weeks. The result is then plotted as an oscillator, which isolates the cyclical high and low points.

  It is then a relatively simple task to see whether any consistent time periods separate these points. One method is to note down the time differences between all the cycle lows and highs in order to determine which ones come up most frequently. Since MAs smooth out all cycles within their time span, it is important to experiment with several averages in order to identify as many cycles as possible. The more reliable ones should then be used in the analysis.