Stock Market Cycles, Part 5: Secular Cycles

See parts 1, 2, 3 and 4.

The have been 6 secular bull markets and 5 secular bear markets completed since 1900. As in previous parts of this series, we have used the Dow Jones Industrial Average (DJIA) historical data.

Secular Bull Markets

Table 1. Secular Bull Markets

There have been five defined and confirmed secular bull markets in the past 107+ years. We have assumed (for this analysis) the market top of 10/9/2007 is the top of the sixth secular bull market. This has yet to be confirmed with a subsequent primary bull market top below 12,748 (more than 10% below the secular top on 10/9/2007). We do not analyze the bull market which topped 1/19/1906 because the previous primary bear market bottom can not be identified in the data starting 1/1/1900 (i.e., the bottom may have occurred before 1900). Therefore, this article analyzes five secular bull markets.

In Table 1 it is seen that there have been two shorter secular bull markets and three longer ones. The average and standard deviation values show only poor to fair “tightness” and have little meaning. Calculating any statistics on the two groups separately has no meaning because the sample sizes are too small. The best that can be said is that the five secular bear markets that have occurred since 1900 appear to fall into two groups. Whether these represent members of two separate normal distributions or are simply extreme points for a larger single normal distribution remains to be seen with longer history (more samples).

There is remarkable correlation of the five secular bull markets with a quadratic equation (R-squared very close to 1). It remains to be seen if this correlation continues when future secular markets are added. One significant problem with the quadratic shown is that it does not pass through the origin. Obviously, zero year duration requires a zero gain; the quadratic shows a gain of 273% for zero duration, which is nonsense. An alternative graph is shown below:

A much better y-intercept is obtained (32.9%) when the data is fitted to an exponential equation, as shown above. The trade-off is a lower R-squared value (.896), which is still very good but less than the .96 value with the quadratic.

A more satisfactory result is obtained if the point (0,0) is added to the five experimental points in the quadratic equation. This is shown in the graph below, where a still excellent R-squared value of 0.94 is obtained. Until more data is available, the graph below should be taken as the best representation of the relationship between gain and duration for secular bull markets.

It will be interesting to see if future secular cycles define a basis for distinguishing between the three relationships we have considered here.

Secular Bear Markets

The table below shows the secular bear markets since 1900 (based on daily closing prices):

Table 2. Secular Bear Markets

The statistics for secular bear markets show tighter distributions than we saw for secular bull markets, but none are excellent. One notable outlier is secular bear number 2 (ending 7/8/1932) with an 89% loss. If this were not included the statistics would be improved.

Attempts to plot a graph of loss versus duration for secular bear markets has been unsuccessful. The graph below shows what happens when a straight line fit is attempted. Similar results have been obtained trying to fit other functions. Not only is the R-squared value close to zero, the plot goes no where near the origin, which is a required reality. The data points are nearly randomly scattered.

Secular Market Cycles

As with analysis of primary market cycles, secular cycles are analyzed from one top to the next. In Table 3 are shown the five secular market cycles completed since 1900.

Table 3. Secular Market Cycles

Statistical analysis, at least in the way attempted for secular bull markets and bear markets, does not make sense with the above data. Some observations are appropriate, though. Three of the cycles are similar in duration and much longer than the other two, which are quite close to each other, as well. The average annualized gain the three long lasting cycles is 10% and the average for the two short cycles is -3%.

In the graph below, a remarkably good correlation (R-squared = 0.97) is obtained for a quadratic equation fit. The graph does have a major flaw: The value of gain for zero duration is 416%.

If the equation is fitted to six data points (the five experimental points plus the required (0,0) point), the graph below is obtained. The correlation with the data is still very good, with R-squared = 0.92. A quadratic relationship appears to exist between duration and gain (loss) for secular market cycles.

Summary observations for secular bull markets are:

1. There is a wide variety of durations for secular bull markets, ranging from as short as just over 2 ½ years to 32 ¾ years. Over the past 107+ years, short and long duration bull markets have alternated. This may be coincidence.
2. The sample size is too small and the dispersion is too great to make any use of mean and standard deviation possible for secular bull market cycles. 
3. Secular bull markets show a strong quadratic correlation for increase in gain with increased duration.
4. The two short secular bull markets have had much larger average annual gains (22% and 39%) than have the three longer lived bulls (10% to 14%).

Summary observations for secular bear markets are:

1. There is more uniformity in duration and gain (loss) for secular bear markets than for secular bears.
2. The five secular bears have lasted from 1.9 to 8.9 years, with an average of 4.6 years.
3. The five secular bears have losses from 37% to 89%, with an average loss of 54%. If the 89% loss from the 1929 top is eliminated as an outlier, the remaining four bears had losses ranging from 37% to 52%, with an average of 46%.
4. The relationship between loss and duration for secular bear markets is random scatter.
5. The current secular bear market, which started from the 10/9/2007 top, has a loss of 42.3% as of 10/27/2008 (8175.77). This would be the next to smallest loss for a secular bear market. One is tempted to estimate that there is a 15-20% probability that we have reached the bottom for this secular bear market. Continuing with the same logic, the estimate would be a 60-65% probability that the bottom will occur between the current cycle low (8175.77) and 6770 (52% loss). These estimates are subject to high risk of error because they are based on a very small historical sample of five. 
6. If the 10/27/2008 low is the bottom of this secular bear market, it will be the shortest duration yet seen (1.05 years). If this secular bear duration equals the shortest previous value (1.9 years), the bottom would occur somewhere around September, 2009. If we equal the average of the four shortest bears (4.6years), this bear will bottom near June, 2012.

Summary observations for secular market cycles are:

1. The secular market cycles can be divided into two groups: The smaller group (2 samples) have durations very close to seven years, and the larger group (3 samples) with durations ranging from less than 24 years to nearly 35 years.
2. The relationship between duration and gain (loss) for secular market cycles has an excellent correlation to a quadratic equation.
3. If the past pattern of alternating long and short secular cycles is continued, the current cycle will be short rather than long.

Comments

[John Lounsbury:]

Stanlori -

I associate secular markets with major political and economic events. Since 1900, I would put these as the major events (in chronological order):

1907: Banking Crisis Shock
1914-1919: World War I Boom
1919-1921: Post War Bust (Strikes & Riots) Shock
1921-1929: The Roaring 20's Boom 3
1929: Stock Market Crash Shock
1930-1941: Great Depression Shock 1
1942-1945: World War II Boom
1946-1966: Post War Expansion Boom 1
1967-1982: Vietnam War, Oil Crisis & Watergate Shock 3
1982-2000: Technology Bubble Boom 2
2000-2003: Technology Deflation Shock
2003-2007: Housing & Credit Bubbles Boom
2007- ?: Deleveraging Shock 2

I have indicated my rank of economic impact. The first column of numbers ranks the worst three shocks and the second column ranks the most significant booms. It remains to be seen is my ranking of the current economic crisis as #2 will stand the test of time.

The four secular cycles since World War I have been associated with the major economic events of the past century:
1. Two secular bear markets (bottoms in 1932 and 1942) were associated with the Great Depression (Shock ranked #1).
2. The other two secular bear markets (bottoms in 1970 and 1974) were associated with the Vietnam War, an inflationary oil crisis and the Watergate era (Shock ranked #3).
3. Two secular bull markets (tops in 1929 and 1937) were associated with the Roaring 20's (Boom ranked #3) and a major rebound rally during the Graet Depression, which I would associate with the Roaring 20's, rather than anticipation of WW II and the post war boom.
4. Two more secular bull markets (tops in 1966 and 1973) are associated with the Post WW II expansion, which I have ranked as Boom #1. Just because the 1973 top occurred after the start of the next shock, I do not want to associate it with anticipation of the next boom, which didn't start for another nine years.
5. The latest secular bull market (top in 2007) is associated with technology boom (ranked as Boom #2). This secular bull includes in its final stages the credit and real estate bubble, but has the 18 year technolgy bubble as its foundation and major contributor.

The only general observation of this history that I want to make is that the four secular cycles since WW I have occurred in pairs. If this pattern continues, the secular bear market we are currently in (my assumption, it is not yet confirmed) will be followed over the next few years by a new secular top and a subsequent secular bottom before we embark on a longer lasting secular bull market.

A note of caution: there are very few samples in this analysis. The pairing of secular cycles closely in time may be pure coincindence.

2008 Nov 09 01:25 PM