by Martin Zweig
In the far-right column of table 30, you’ll see our update of Tabell’s calculations for the actual percentage changes on the S&P during the various months. On this basis, July comes out ahead, having risen an average of 1.85%. January was next at 1.62%, followed by December at 1.44%, August at 1.20%, and June at 0.99%. As with the previous study, September was the worst month, with only 29 gains versus 40 losses and a total average return of -1.32%.
Once again, human emotions probably account for the bulk of the patterns of these returns. Indeed, the days before and after Christmas and the day before New Year’s account for almost all of December’s 1.44% average gain. January’s advance is attributable to the end of tax selling, as is explained in the last section of this chapter. A portion of July’s excellent performance is accounted for by the pre–July 4th holiday pattern. The rest of the solid results for the July–August period are, in my book, attributable to the fact that, on balance, people generally feel better in the summer vacation periods than they do at other times of the year. That’s in sharp contrast to the dreary short days of winter when, in many sections of the country, you are apt to feel miserable fighting snowbound traffic.
TABLE 29
MONTHLY TENDENCIES OF DOW JONES INDUSTRIAL AVERAGE: 1897 to 1996
Month
% of cases in which Dow rose in the month
January 65.6
February 49.5
March 59.6
April 54.5
May 52.5
June 51.5
July 61.6
August 64.6
September 40.4
October 54.5
November 59.6
December
71.7
Average all month over 99 years: 57.0%
Source: Arthur A Merrill; Merrill Analysis Inc., Box 228, Chappaqua, NY 10514
TABLE 30
MONTHLY SEASONAL TENDENCIES: S&P 500 INDEX 1/1928 to 3/30/96
Direction of S&P 500 Index
Month
Advances
Declines
Average % Change in S&P 500 Index
January 45 23 +1.62
February 37 32 -0.02
March 43 27 +0.28
April 24 28 +1.00
May 39 30 -0.20
June 37 32 +0.99
July 41 28 +1.85
August 41 29 +1.20
September 29 40 -1.32
October 40 30 +0.09
November 38 32 +0.34
December
52
18
+1.44
Total 482 349 +0.61
Source: Anthony W. Tabell; Delafield, Harvey, Tabell, 600 Alexander Road, Princeton, NJ 08540
Both Tabell and Merrill found that September was the worst month of the year, which is not surprising if you monitor mood swings. September’s that time when vacation is over and people begin to think about the winter that lies ahead. It’s also a time for those back-to-reality blahs, when days off are fewer and the work load increases.
For trading strategies involving months of the year, the best would be to buy stocks at the end of May, hold them throughout June, July, and August, and sell after the Labor Day holiday. This would produce an average gain of 3.5% on the S&P through August, plus another fraction in the first days of September prior to the Labor Day holiday. Even that decent return in the summer, though, might be swamped by the more traditional monetary, sentiment, and momentum indicators. But if these measures are bullish and summer pops up on the calendar, you might buy a bit more aggressively than you otherwise would.
END-OF-MONTH PATTERNS
In still another study, Art Merrill found that the market tended to be stronger than normal in the last three days of a month and the first six days of the subsequent month, while the rest of the time the market performed below its average. In a more comprehensive study a decade or so ago, analyst Norman Fosback (Institute of Econometric Research, Ft. Lauderdale, Florida) found a slightly tighter month-end tendency in which the market performed abnormally well in the last trading day if the month and the first four trading days of the following month, a total span of five “seasonal” days.
Some of these good returns are accounted for by the pre–July 4th holiday and the pre–New Year’s holiday, which always fall within this five-day pattern, and the pre–Labor Day trading day, which usually falls within that period. Occasionally, Easter and Memorial Day might contribute a seasonally strong day as well to this five-day pattern. However, the upward bias of this five-day seasonal trend goes well beyond what can be explained by a few holidays.
Fosback studied 568 such month-end periods from 1928 to April 1975. He noted that a $10,000 investment in the S&P 500 only on those five trading days at month’s end would have grown to $569,135, a gain of .71% per trading period, or an annualized profit of 43.8%. By contrast, the “nonseasonal” days—those from the fifth day of the month up through and including the next-to-last day of the month—would have withered a $10,000 investment to only $844, an annualized loss of 6.6%.
I have updated Fosback’s idea from April 1975 through June 1985. The upward bias when tested against the S&P 500 has not been as strong in the most recent decade, but there still is such a tendency. A $10,000 investment over the month-end seasonal period in the past decade would have appreciated to $13,322, a gain of .24% per period, or 12.8% a year, roughly only one-third the returns that Fosback found in the prior thirty-seven years.
I next went back to Fosback’s starting date of 1928 and tested the month-end tendencies against my own Zweig Unweighted Price Index through June 1985. The results were somewhat more consistent, although once again not quite as good in more recent years. For the entire 57½ years from 1928 through mid-1985, a $10,000 investment in the Zweig Unweighted Price Index during the month-end seasonal periods would have appreciated to $495,300, a gain of .57% per period, or an annualized pace of 33.4%. That breaks down, by the way, to a gain of just over 7% in each calendar year, given that one has invested in only sixty trading days in that year.
It’s unfortunate that the 33.4% annualized rate cannot be earned over a complete calendar year since there are only sixty days in which you can trade with the month-end seasonal tendency. Nonetheless, that 7% per annum is gained in less than one-quarter of all trading days, and it leaves you free with cash to invest at prevailing interest rates over the rest of the time. Moreover, you would not want to be in stocks during the nonseasonal days if you could avoid all transaction costs. In the ten years from April 1975 to June 1985, a $10,000 investment in the ZUPI at month-end periods would have grown to $17,659, a rate of .46% per month’s end, or an annualized pace of 26.3%.
Once again, if you trade at month’s end you have the problem of transaction costs. However, here the strategy of buying no-load funds just prior to the last trading day of the next month, and selling them after the fourth trading day of the next month, is somewhat more palatable since it involves only twelve switches a year with mutual funds. Some mutual funds are willing to accept that level of trading. You could also combine the effects of holiday periods and month-end periods, or even of the strong Friday tendency.
For example, if the fifth trading day of the month is a pre-holiday period—say the Thursday before Easter—you could hold stocks for one more day. Likewise, if the next-to-last trading day of the month is a holiday or a Friday, you could buy one day earlier before the month-end period. In the long run, this would enhance returns and enable you to stay in the market a bit longer to generate more profits.
Outside of those months such as December, when the holiday effect coincides with the month-end seasonal effect, I can offer no reasonable explanation for this month-end behavior. It’s doubtful that it derives from any economic origin, and I cannot explain it on emotional or “mood” grounds. I doubt that people feel more “up” at month’s end or month’s beginning than they do at other times. So, the causes of this effect remain a mystery. But what is fact is that this market pattern has persisted f
or decades, even though the impact in the most recent decade has not been quite as strong as in earlier spans.
THE PRESIDENTIAL ELECTION CYCLE
There is a theory that presidential elections make a difference to more than just the candidates involved; that they also greatly affect the stock market. This belief is based on the premise that the party in power will attempt to do whatever it can economically to stay in power. The implication is that the incumbents will take positive economic action in the year or two before an election, which normally might translate into better-than-average results for stocks. Of course, the piper must eventually be paid, and this would lead to rather poor stock market returns in the year or two following the election. Let’s check the actual results to see if there’s any validity to this theory.
I’ve gone back and matched stock market performance against the election cycle since 1872. Since the election itself comes at the beginning of November, I measured a year’s performance on an October-to-October basis using average prices for the Octobers. In other words, for the year 1984, the market’s performance is measured on the basis of the average price of October 1984 relative to the average price of October 1983. After 1926 the S&P 500 Index was used. Prior to that, I used the most appropriate available stock index, such as that compiled by the Cowles Commission.
Table 31 basically backs up the presidential theory. The preelection years, each of which starts two years before an election and ends one year before an election, show the greatest gains, with average returns of 7.0% per annum. The market rose 70% of the time, climbing 21 times and falling 9. An investor would have increased his initial stake seven-fold in those 30 pre-election years.
TABLE 31
THE STOCK MARKED AND THE ELECTION CYCLE: 1872 to 1992
Direction of Market:
Year
No. of Cases
$10,000 Investment
Annualized Return (%)
Up
Down
% of Years Market Was Up
Pre-Election 30 76,545
+7.0 21 9
70.0
Election 31 42,969
+4.8 22 9
71
Post-Election 30 26,345
+3.3 16 14
53.3
Midterm
30
10,976
+0.3
16
14
53.3
All years 121 943,355
+3.8% 75 46
62.0%
Note: Market is measured by October-to-October change, with average S&P or Cowles Commission prices used for the Octobers.
The next-best returns were made in the election years themselves, that is, in the years ending on the eve of the election. The market rose 22 times and fell 9, for a 71% success rate. The annualized rate of gain was 4.8%, which was slightly better than the market’s overall return of 3.8% for the 120 years through 1992.
Once the elections were over, the stock market did, in fact, do a bit worse than was normally the case. In the post-election years, stocks appreciated by only 3.3% per year, more than a half percent worse than buy-and-hold. Year-by-year results were almost a standoff, with 16 winners and 14 losers.
Finally, in midterm years, beginning one year after the end of the election and ending two years after the election, the market on average gained slightly, rising at an annualized rate of .3%. Stocks rose 16 times and declined 14 times, a 53.3% success average.
Thus, the results show that there is something to the election cycle, but clearly not enough to dominate strategy. For example, the pre-election year is the best of the four-year cycle, but on occasion the results have been far less than satisfactory. In 1903 stocks plunged 26.9%, in 1907 they fell 33.4%, and in 1931 the market was blasted for a 42.8% loss. In the election year itself, another supposedly strong period, the market suffered losses of 16.8% in 1920, 30.5% in 1932, and 16.8% in 1940. More recently, the market dropped a modest 1.9% in the election year of 1984. Obviously, one should not bet the deed to the ranch on the stock market’s going up in pre-election or election years, even though there would be small odds in your favor.
Conversely, don’t bet against the market just because an election was recently over. In 1933 stocks soared 31.4% on the heels of an election; in 1945 they moved up 27.8%, in 1961 they rose 26.6%, and in 1989 they gained 29.3%. Likewise, in the midterm year stocks on average do their worst, but there have been many cases of exceptionally large gains, including 1950, up 25%; 1954, with a 34.3% gain; and 1958, with a nice 23.5% appreciation. Stocks didn’t do all that badly in two recent midterm years either, showing a 5.8% gain in 1978 and an 8.9% profit in 1982.
YEAR-END TAX SELLING
Late in the year, investors holding depressed stocks frequently sell them to take advantage of the tax loss. By establishing the loss before the calendar is over, it can be used to offset capital gains that the investor might have on other stocks, real estate, or any other asset. Thus, in the first few weeks of December, the market is often artificially depressed by tax selling, at least in those stocks that haven’t done well during the year. Such selling is minimal in stocks that have acted well, since few investors have losses in them. But stocks that, in the month of December, are wallowing near their lows for the year are prime targets for tax selling. However, by Christmas time or so, these stocks tend to bounce.
Ben Branch (Journal of Business, April 1977) has examined the eleven year-end periods from 1965 through and including 1975. His objective was to purchase stocks making new lows during each year’s last full week of trading. The stocks were arbitrarily sold four weeks later, in late January. He found that, on average, these depressed stocks had rebounded a full 9.0% in just one month, whereas the New York Stock Exchange Composite Index—a weighted measure of some 1700 Big Board stocks— had climbed only 2.6%.
A similar study in a doctoral dissertation by Robert McEnally (University of North Carolina, 1969) examined the tax-selling effects from 1946 through 1959. Starting with a sample of 650 stocks, he planned to buy the 10% that had performed the worst during the calendar year. He bought them at year’s end and held those 65 worst stocks for one month, selling them at the end of January. On average, the artificially depressed stocks gained 5.9% in the following month versus a rise of only 2.8% for the sample of all stocks.
Other studies in more recent years have confirmed the tendency for stocks at or near their lows in December to outperform the market over the next several weeks. This approach, by the way, works best in years where the market as a whole is relatively near its own low in December, implying that many more stocks are making individual lows. In very strong market years, such as 1976 or 1982, when stocks finished the year virtually at their highs, you’ll find few individual issues at or near their lows. In those years, the tax-selling game is not worth playing—especially since the handful of stocks that are down-and-out is likely to have some very real problems.
If you want to play this tax-selling game, review Barron’s or one of the daily newspapers in the month of December and search for stocks that have made new lows sometime around late December. As long as there are at least several dozen stocks from which to choose, I would advise buying equal dollars’ worth of at least several issues. Buy them and hold them for a few weeks. The sell date is quite arbitrary, but if the market acts well in January, I would continue to hold the stocks until late in that month. If by the second week of January the stock market as a whole is no longer acting well, I would jettison them.
CHAPTER 10
Major Bull and Bear Markets—How to Spot Them Early
The big money is made or lost in stocks during the most violent bull and bear markets. The bad news for those who crave action is that the market does not behave dynamically all that often. Even within the great bull market advances, there are periods of lull. I would estimate that stocks spend only about 20% of the time in the most active phases of the bull trend and only about 10% in the severe downward periods of major bear
markets. Roughly 70% of the time stocks either meander in a neutral trading range or undergo minor rallies or declines within their various bull and bear cycles. During that 70% span—let’s call it the neutral area—your overall market strategy doesn’t matter all that much. You could be fully invested, partially invested, or all in cash. If you have a broadly diversified portfolio while the market hovers in these neutral ranges, you are not likely to make or lose a lot of money.
The other 30% of the time, when dynamic phases of major bull and bear markets are in progress, is another story—and it’s worthwhile to examine the chief characteristics of these periods. The good news is that you can watch for a few key signals that offer exceedingly high probabilities of catching the best portion of the great bull markets while avoiding the devastation of the worst phases of major bear markets. The first part of this chapter will focus on the bull market signals, and the last part on the bearish indications.
TWO BASIC INGREDIENTS FOR BULL MARKETS
There is no official way to define or measure the greatest bull markets in history. To keep things simple, I have arbitrarily selected the eleven times since 1926 when the Standard & Poor’s 500 Index recorded the greatest maximum percentage gains within an eighteen-month period. I’m using 1926 as the base date because that is when the S&P Index began.
In many cases stock prices continued to rise after eighteen months, but I figured a year and a half was appropriate for determining a truly robust bull market—especially since the greatest percentage gains come in the early months of such markets. Occasionally the S&P peaked well before eighteen months. For example, if the market was, say, 100% higher in twelve months and only 50% higher eighteen months after the bull cycle began, I used the 100% gain as the yardstick.
