Stock Market Cycles, Part 4: Primary Cycles

 |  Includes: DIA, QQQ, SPY
by: John Lounsbury

See parts 1, 2 and 3.

In Table 1 we show the 32 complete primary cycles that have occurred since 1900.

Table 1. Primary Market Cycles (Part 1)

Table 1.  Primary Market Cycles (Part 2)


We have had 32 completed primary cycles since 1900, the latest being completed with the primary top of 10/9/2007. In the two preceding parts of this series, we have reviewed the 32 primary bull markets of the past 107+ years (Part 2) and the 33 primary bear markets (Part 3). We have defined (Part 1) a cycle to be a contiguous (in time) pair of one bull market and one bear market. We chose to look at cycles comprised of one market top (through the next bottom) to the next market top. 

Some of the primary tops and bottoms in Table 1 are also secular tops and bottoms and are so identified by italicized blue labels. Secular market cycles will be discussed in the next article in this series (Part 5).

Note that the returns listed in the table do not include dividends. This is important because, through much of the last century, dividends for DJIA stocks were an important component of total return.

Just as we did for primary bull and bear markets individually, we have looked at distributions of durations and returns over complete primary cycles. We are dividing the primary market cycles into two groups: those with a gain over the cycle (17 positive cycles) and those with a loss over the cycle (15 negative cycles).

Positive Primary Cycles

Since the statistics for all 17 positive primary market cycle durations are quite dispersed (poor “tightness”), the durations have divided into four groups. One group (duration <200 days) has only one sample. These duration groups are displayed in Table 2.

Table 2. Primary Market Cycle Durations (Positive Cycles)

Seven of the 17 had an average duration less than 2 ½ years; the remaining 10 had durations averaging more than six years. The three longest cycles have an average duration just over 13 years.

The gains over the positive primary market cycles are displayed in Table 3. The attempt to partition this data into sub-groups met with questionable success, from a statistical point of view. One statement that can be made: Over 1/3 of the positive cycles had only small returns (<12%) and nearly 1/3 more had only modest returns (12-40%).

Table 3. Primary Market Cycle Gains (Positive Cycles)

In Table 4, the distribution over time of positive primary market cycle durations is displayed.

Table 4. Positive Primary Market Cycle Occurrences Over Time (Durations)

Positive primary market cycles have been quite uniformly distributed over time. Exceptions: The 1930’s had the most cycles (3) and only the 1950’s had no cycles completed.

In Table 5, the distribution over time of positive primary market cycle gains is displayed.

Table 5. Positive Primary Market Cycle Occurrences Over Time (Gains)

Three of the four large gain cycles have ended in the last 42 years. The fourth ended in 1929. Small and very small gain cycles have occurred in all decades except the 1920’s, the 1940’s and the 1950’s. It is noteworthy that there have been only two medium gain cycles.

The graph below shows that there is fair correlation (R-squared = 0.66) with a straight line for gain plotted against duration for positive primary market cycles. The trend is greater gain for longer duration positive primary cycles. An attempt to fit the data to a quadratic yields the same equation. One other problem, in addition to the R-squared value of 0.66, is the fact that the y-intercept indicates a loss of 18.3% when duration is zero. For zero duration, there is no market cycle, and the gain should be zero. In the second following graph, the data point (0,0) is added to the 17 data points. This appears to be nearly as good a correlation as the graph with the y-intercept of -18.3%, because the R-squared value is only slightly reduced to 0.64.

Click to enlarge

Click to enlarge

Negative Primary Cycles

Since the statistics for all 15 negative primary market cycle durations are somewhat dispersed (fair “tightness”), the durations have divided into three groups. Table 6 displays the statistics for negative primary cycle durations.

Table 6. Primary Market Cycle Durations (Negative Cycles)

Not only are the negative primary market cycles of much shorter duration than for positive cycles, the sub-group statistics are better behaved (tighter distributions).

Table 7 shows analysis of the losses for negative primary cycles.

Table 7. Primary Market Cycle Losses (Negative Cycles)

The statistics behave very well for the large loss group (more than 20% loss), even when it is enlarged to include the two samples with losses between 10% and 20%. The distribution is disperse (poor “tightness”) for the eight samples with losses between 0 and 10%.

Table 8 shows the distribution of the negative primary market cycles over the 107+ years of this study, broken into different durations.

Table 8. Negative Primary Market Cycle Occurrences Over Time (Durations)

In contrast to positive primary market cycles, which were rather evenly distributed over the years, negative cycles have been experienced much less uniform spread over time. More than ½ occurred during the 1930’s. There were four decades (1920’s, 1950’s, 1980’s and 1990’s) that saw no negative primary cycles. Further indication of the isolation of negative primary bear markets, 73% occurred before 1941 (38% of the time span), only 13% occurred between 1941 and 1976 (33% of the time), none have occurred between 1977 and 2001 (24% of the time), 13% have been completed in 2001 and 2002 (0.8% of the time) and none have yet been completed since 2002 (5% of the time).

Let’s look at the distribution of negative primary bear cycles differently: Three time periods comprising 54% of the time since 1900 have seen no negative bear cycles end (1913-1929, 1941-1968, and 1976-2000).

In Table 9 are shown the distributions of negative primary market cycle losses since 1900. The only time we see large losses per cycle is the 1930’s, when these events occurred five times. Of the remaining ten negative cycles, eight are identified as small losses (<10%). Only two negative cycles fall in the medium loss category and, curiously, they occurred very close together. One of these ended in1938 and the other in 1940. When the years 1930 through 1940 are excluded, only bear market cycle losses of less than 10% have occurred.

Table 9. Negative Primary Market Cycle Occurrences Over Time (Losses)

The following graph shows the relationship between loss and duration for the 15 negative primary market cycles.

Click to enlarge

The quadratic formula is displayed because it has the largest R-squared value of all the formulas tested, including a straight line. Even so, the R-squared value of 0.41 indicates only fair (not good) correlation of the equation to the data.  The general relationship is that longer durations appear to have smaller losses.  Stated differently, primary cycles with large losses tend to be of short duration. Note that one can not extend the mathematical relationship to zero duration because, obviously, there can be no loss for a market cycle that doesn’t exist (zero duration). However, a zero return can occur for a non-zero duration.

In the graph below the data for all 32 primary market cycles (positive and negative) is fitted to a linear equation with fair correlation (R-squared = 0.73). It can also be fitted to a quadratic with almost identical R-squared. The y-intercept for the combined data differs from that obtained in the previous graph for positive cycles only (approximately -30% here vs. -18% earlier). These differences are not significant considering the scatter in the data. 

Click to enlarge

When we add the null point (0,0) to the 32 primary cycle data set we obtain the graph below. The correlation is somewhat poorer, as R-squared is reduced to 0.65, compared to 0.73 above. The difference between the two graphs is most evident at the extremes (very short and very long durations. In effect, the line is pivoted about a focal point near 2000 days duration, with the line (in the graph below) showing higher gains for very short durations and lower gains for very long durations, but little change near 2000 days duration.

Click to enlarge

Some summary observations are:
  1. There have been almost as many (47%) complete primary market cycles with a loss (negative cycle) as with a gain (positive cycle).
  2. Positive cycles show a more even distribution over the 107+ years since 1900 than do the negative cycles.
  3. The average gain for positive cycles (73.5%) is much bigger than the average loss for negative cycles (-12.7%).
  4. Positive cycles last much longer (average duration 5.4 years) than negative cycles (1.5 years).
  5. The trend for positive cycles is that gain increases with greater longevity.
  6. The trend for negative cycles is that loss is greatest for short cycles and declines as one moves to longer lived cycles. This might appear to be counter to the behavior of positive cycles (see 4. above). However, when one looks at the graph for all primary market cycles, positive and negative combined, the data clearly is consistent with a simple straight line relationship.
  7. Most of the negative primary market cycles (87%) have occurred in only 39% of the time since 1900.
  8. There have been three periods of time (more than 54% of the total time span) where no negative primary market cycles have occurred.
  9. If we leave out the Great Depression, all negative primary market cycles have lost less than 10%.