The Amazing Story of Stock Market Seasonality
by Arthur J. Minton, Ph.D.
Conventional wisdom states that the stock market is random, that is, there are no patterns or mechanisms which
can be understood and used to consistently predict market movements up or down. The theory of market
seasonality states just the opposite. This theory holds that there are real factors operating on the stock market
which cause consistently recurring patterns over decades of market history. These factors can be used as the
basis for an improved, less risky, investment method.
For our purposes, I will use the word “seasonality” to define a number of seasonal trends that seem to exist in the
stock market. A “seasonal trend” is a recurrent time period when the stock market has a statistically high tendency
to either rise or fall.
Seasonality is not a magic formula. In what follows, you will see that following certain seasonal trends have
generated impressive results over the long term. Seasonal investing relies on probabilities, not certainties, and
therefore is subject to periods where it underperforms more conventional investment methods. In addition, the
investment discipline that emerges from a study of seasonality tends to be very quantitative, taking little notice of
the usual indicators of market health, such as earnings growth, recessions, war, and the day-to-day news that
surrounds the stock market. As you will see, investments have a definite buy and sell date, regardless of the
investment world’s “noise”. This type of quantitative approach may not appeal to many investors, in spite of its
impressive track record.
One of the first researchers to document seasonal trends was Norman Fosback in his ground-breaking book Stock
Market Logic (1976). In this book Fosback introduced what he called “econometrics” to the study of stock market
behavior. In a wide ranging analysis, Fosback documented the effect of interest rates, Fed policy, PE ratios and
many other fundamental and technical factors at play in the stock market. For our purposes, the most interesting
chapter of the book was called “The Seasonality Indicator”. In this chapter, Fosback argued that all of the long-term
returns of the stock market (from 1928 to 1975) were contained in two recurring periods – the last day and first four
days of each month, plus two days prior to market holidays – about 25% of all available trading days.
According to Fosback, a hypothetical investor holding the S&P 500 index on these seasonally favorable days
(paying no commissions for trading, receiving no dividends, and no interest on cash) grew a $10,000 investment on
December 31, 1927 to $1,440,716 by December 31, 1975. An investor holding the S&P 500 constantly during the
same time frame (again, appreciation only) would have ended up with $51,441. Finally, an investor holding the
S&P 500 during the seasonally unfavorable days, or about 75% of the time, would have seen his investment
decline to just $337, a loss of over 90%.
Over the entire 48-year period, the seasonal strategy was down just seven years, and losses were small enough to
be recouped within one or two years. In contrast, the seasonally unfavorable investment was down in 28 of the 48
years, even though it was invested for three times as many days. The long-term rate of return for Fosback’s
seasonal system was about 11% annually, a 50% improvement over the appreciation rate of the S&P 500 over the
same period. Of course, this did not include dividends, interest earned when in cash, nor expenses.
The problem for Fosback’s research in 1976 was that there was no way to implement his findings. No load, no
transaction fee index funds did not exist. That problem was solved in the mid-90’s when Rydex Funds introduced
the first index funds which could be traded as often as daily without transaction fees.
The Best Days of the Month
Fosback’s research highlights two forms of seasonality – monthly and pre-holiday. Let’s take a look at monthly
seasonality in detail.
The theory of monthly seasonality holds that the turn-of-the-month period, essentially the last few days and first few
days of each month, capture the lion’s share of monthly returns. The cause of this effect has been debated for
years, but there’s no denying its reality.
First, let’s take a look at the beginning of the month.
Day of Month Cumulative % Gain + (-)
Trading Day 1 223
Trading Day 2 524
Trading Day 3 303
Trading Day 4 65
Trading Day 5 31
The table above shows the cumulative gain for each of the first five days of the month from 1900 to 2008, using
the Dow Jones Industrial Average. For example, if an investor had held the Dow for just the first day of each
month, a $1,000 investment would have grown to $3,232, a gain of 223%. A $1,000 investment in the Dow for all
five days over the same period would have grown to over $180,000.
The final two days of the month have also produced outsized returns over the past 107 years. (The last day of the
month is -1, the next to the last is -2, and so on.)
Day of Month Cumulative % Gain + (-)
Trading Day -2 128
Trading Day -1 279
A $1,000 investment in the Dow on the two final days of each month would have grown to about $8,500.
Now let’s look at a trading system comprising all six days. We will start on December 31, 1933, using the Dow as
our trading vehicle. A $1,000 investment on these six contiguous days grew to $202,335 by 2008. The annualized
rate of return was 29.9% and the average daily gain during the six days was 72 times greater than the average
daily gain on all other days of the month (0.4%).
The Mid-Month Bulge
Beginning in the early 1980’s, a new trend began to emerge – the mid-month bulge. Trading days 9-12 begin to
exert a strong upward influence on stock prices based upon (in theory) the rising use of IRA accounts and 401K
plans, both of which allow bi-monthly contributions. A $1,000 investment in these four mid-month days (using the
Dow) from 1980 through 2007 would have grown to over $2,500.
The Worst Days of the Month
Just as dramatic as the best days of the month, the pattern of worst days has been remarkably consistent over the
past 108 years.
Following the strength of the opening days of the month, there tends to be a lull in the market.
Day of Month Cumulative % Gain + (-)
Trading Day 6 (42)
Trading Day 7 (11)
Trading Day 8 (45)
An investor holding the Dow on these days from 1900 to 2008 would have experienced a loss of about 70%.
Also toward the end the month, another weak period occurs.
Day of Month Cumulative % Gain + (-)
Trading Day -7 (21)
Trading Day -6 (18)
Trading Day -5 (55)
Considering that the Dow advanced about 20,000% over the same time period, the fact that these days
systematically (but not all the time) declined, strongly suggests a long-term mechanism at work.
Combining the Favorable Days – The Monthly 10
If we combine our two favorable monthly periods (“power periods”) into a single trading system, we would:
1. Own the Dow during the last two and first four trading days of each month, beginning
December 31, 1933 to December 31, 2007.
2. Also own the Dow during the four mid-month days (9-12) from December 31, 1979 to
December 31, 2007.
The results of our 10-day trading system are:
A $1,000 investment grew to $562,500.
The annualized rate of return was 28.4%.
The annualized rate of return for all other days was -1.8%.
A $1,000 investment on all other days shrank to $236, a 76.4% loss.
Source: Jay Kaeppel, Seasonal Trends in the Stock Market (Wiley 2008)
A word of warning here: there are several lengthy stretches of time when performance was down, flat, or
marginally up. There were also long time segments when these power periods were routinely and robustly up.
Like all systems, this one also requires a degree of patience from anyone using it as a real-time investment
Fosback’s original research included pre-holiday seasonality, which he defined as the two days prior to market
holidays. For our purposes, we will exclude Martin Luther King, Jr. Day, as it is a relatively recent addition. That
leaves us with:
Major Stock Market Holidays
New Year’s Day
Regarding President’s Day, it should be noted that prior to 1952 there were two holidays in February, marking Lincoln’s and Washington’s
birthdays. After 1952, these were combined into one, now President’s Day.
For most of the twentieth century, the day before a holiday was especially potent. A $1,000 investment on that
single day would have grown to $3,147 from 1933 to 2008. That’s an annualized return of 62.3% vs. a 5.9%
average annualized return for all other trading days. If we add in the next-to-last day before a holiday we get
additional gains, but not as powerful. A $1,000 investment in the Dow from 1933 to 2008 on these two days grew
to $4,552. The annualized return for these two days was 37.5% vs. 6% for all other trading days. Source: Jay
Kaeppel, Seasonal Trends in the Stock Market (Wiley 2008)
While no one would complain about an investment that delivers an annualized return of 37%, the problem with
holidays is that there are just too few of them. In addition, holidays often overlap with turn-of-the-month periods
(i.e. Labor Day, Memorial Day, New Years), thus losing their power to function as an independent source of return.
Nevertheless, some holidays such as Christmas and New Years are embedded in close proximity to one another
and when combined with the year-end period, produce amazing results. The last seven days of the year, which
includes two days prior to Christmas, has been up 93% of the time since 1979, using the Russell 2000 small-cap
index as the trading vehicle (the index began in 1979) with an average return of 2.5%. If the stock market behaved
like this all year long it would be up over 100% a year on average.
Understanding Alpha’s Trading System
The data presented above represents a broad statistical brushstroke which treats every month, every holiday, and
every year the same. While Alpha’s trading system is built around turn-of-the-month, mid-month, and pre-holiday
seasonality, we believe that important differences exist between months and especially between years, and that
seasonal factors can be refined into a more productive system.
Let’s take the turn-of-the-month period from September to October, for example. Across all years, this period is
much weaker than its counterparts in other months. September is historically the weakest month of the year,
averaging a negative rate of return (S&P 500, 1958-2007). This is due, in part, we think, to institutional investment
managers preparing for the fourth quarter and year-end. Gains and losses are harvested in this period to offset for
tax-purposes. The resulting proceeds are fed back into the market later in October and November, which accounts
for the historical strength of this year-end period. The September/October turn-of-the-month period is up less than
50% of the time – clearly a bad bet.
Another factor impacting our trading system is the decision to use small and mid-cap indexes as our trading
vehicles. Our research shows that market seasonality affects these indexes about twice as much as the Dow and
the S&P 500. Therefore, the profitability of seasonal investing is substantially magnified by using these indexes,
which are more volatile than their large company counterparts.
We also take into account the Presidential Election Cycle when configuring our trading calendar. The four-year
election cycle has an important impact on the stock market. For example, the pre-election year (year three) has not
been down since 1931 (using the S&P 500 total return index), reflecting the actions of the political class to create a
positive economic environment during the election year. The returns during the pre-election year are much higher
than the post-election and mid-term years.
Dow Jones Industrials S&P 500 NASDAQ
Election Year (1949 – 2008) (1949 – 2008) (1971 – 2008)
Post-Election 3.2% 3.3% 4.0%
Mid-Term 7.3% 7.9% 8.5%
Pre-Election 17.7% 18.3% 34.2%
Election 6.9% 9.3% 9.9%
Source: Stock Trader’s Almanac, 2009
While the election year cycle is not carved in stone (there have been several periods where the weak years are the
strongest), it is statistically a long-term factor which we account for in our system. For example, in year two of the
cycle, we trade just 14 power periods, including mid-month and turn-of-the-month periods. The reason for this
austere market exposure is the fact that the mid-term year contains two quarters – Q2 and Q3 – which are routinely
down. Many of the biggest bear markets historically have concluded in year two in the second or third quarters.
The bear market of 2000-02, for example, concluded in October of 2002 – the mid-term year of the election cycle.
This propensity for negative returns causes us to avoid power periods which are far less robust than normal.
On the other hand, we invest 22 times in year three of the election cycle. As indicated above, the pre-election year
is hardly ever down and delivers extremely high returns on average. Statistically, it pays large dividends to have
more exposure during this very robust year.
Over a complete four-year cycle, our system makes 10.1 turn-of-the-month/pre-holiday trades and 9.3 mid-month
trades per year. Since 1979, using the Russell 2000 Index as the trading vehicle, our system has incorporated 358
turn-of-the-month/pre-holiday power periods, of which 303 were profitable – a win rate of 84.6%. Over the same
period, there were 334 mid-month trades, of which 278 were profitable – an 83.2% win rate.
An Added Control on Risk
In 2008, Alpha added a trading protocol designed to reduce risk during periods when seasonality is not working.
Like all investment disciplines, seasonal investing is subject to periods of underperformance and losses. In certain
market environments, the economic or political climate may be hostile to the stock market and the resulting flood of
bad news, which may continue for months, overwhelms the usual seasonally strong periods. It is during such
periods, which may last over a year, that investors can become discouraged and abandon a perfectly good long-
term strategy. It is a fact about human nature, that once a strategy or policy is abandoned, it is highly unlikely that it
will be embraced again. To do so requires the investor to realize that he or she made a mistake – not a trait found
in most human beings. Therefore, to prevent large drawdowns (top to bottom losses) and the emotional toll they
can have on clients, Alpha created a methodology designed to put our trading on the sidelines when events are
moving against us.
The system is very simple. We continuously track our last two trades.* If both were successful, we continue
trading as usual. But, if one is unprofitable, we reduce our exposure on the next trade to 50% of normal. If both
trades were unsuccessful, we pass on the next trade, remaining in cash until the ratio rises to 50% again, at which
point we resume trading at 50% of exposure. When both of the last two trades are profitable, we resume trading
This procedure has the effect of keeping us out of the market when a series of power periods are overwhelmed by
a “news driven” market.
Based on our historical research, we have found that there are periods, often lasting a year or more, when almost
all power periods produce gains time after time. Then, there are other times, which occur rarely, when seasonality
(and hence power periods) has a mixed record for many months. Over the long historical record (since 1900)
seasonality has always re-emerged from the doldrums and produced enviable returns with less risk.
*This is based on our long/short model which tracks both seasonal strength and seasonal weakness. Therefore, one of the trades counted
could be a short sale.
It is hard to be a long-term investor. When things aren’t working, even though we believe strongly in the underlying
discipline, the temptation to abandon the discipline in favor of what’s working at the time can be overwhelming.
The best illustration of this that I know of concerns Jean-Marie Eveillard and his mutual fund, First Eagle Global.
This fund is clearly a long-term winning phenomenon – having experienced just two losing years since it was
started in 1979. From its start to the end of 2008, the fund returned about 14% annually and rewarded every
shareholder who struck with it over the long term. In 1998, the fund suffered a small loss and in 1999 the fund was
modestly up. This happened during the technology/internet fad when that sector of the market was skyrocketing.
Eveillard kept to his discipline, avoiding faddish stocks and sticking with fundamentally undervalued and out of favor
companies. His investors fled in droves and by the end of 1999 he had lost 70% of his shareholders. From 1999 to
2008, his fund registered gains every year (avoiding the 2001-02 bear market entirely) averaging 15%.
The purpose of this material is to provide evidence that seasonal investing is a serious and successful long-term
strategy for controlling risk and achieving above average returns. By understanding the methodology and the long-
term evidence of its decade-by-decade consistency, investors can gain the confidence to persist when temporary
setbacks, which are inevitable, occur.
Seasonal investing has often been labeled “market timing”, but this is a misnomer. Seasonal investing is a long-
term strategy based on long-term factors which have nothing to do with short-term predictions about the direction of
the market. Seasonality reflects certain constants of human behavior which are triggered by the calendar, not the
latest “news” or the recent direction of the market.
The stock market is not random. Seasonal patterns which persist over decades do exist and the factors which
cause them are real. From time to time there are shifts in these patterns (such as the rise of mid-month power
periods in the 1980’s and thereafter), but they are slow to develop and our ongoing research will spot them and
incorporate them into our system eventually.
Our current system has undergone three major revisions since 2000. We introduced the election cycle differential
in 2004, mid-month trades in 2006, and risk-control in 2008. We continue to research ways to make our investment
method more profitable, consistent, and risk-averse.
Disclosure: Past performance is not a guarantee of future performance.