Click here for PDF Version of this article: whysea.pdf
The following was written by MRCI's Editor Jerry Toepke for Dr. Van K. Tharp's book, Trade Your Way to Financial Freedom, published by McGraw-Hill in late 1998. We hope it helps explain the concept behind seasonal research and how it is derived, some of its strengths and weaknesses, and how it can be used and/or incorporated into various styles of commodity trading.
The seasonal approach to markets is designed to anticipate future price movement rather than constantly react to an endless stream of often contradictory news. Although numerous factors affect the markets, certain conditions and events recur at annual intervals. Perhaps the most obvious is the annual cycle of weather from warm to cold and back to warm.
However, the calendar also marks the annual passing of important events, such as the due date for U.S. income taxes every April 15th. Such annual events create yearly cycles in supply and demand. Enormous supplies of grain at harvest dwindle throughout the year. Demand for heating oil typically rises as cold weather approaches but subsides as inventory is filled. Monetary liquidity may decline as taxes are paid but rise as the Federal Reserve recirculates funds.
Natural Market Rhythms
These annual cycles in supply and demand give rise to seasonal price phenomena --- to greater or lesser degree and in more or less timely manner. An annual pattern of changing conditions, then, may cause a more or less well-defined annual pattern of price responses. Thus, seasonality may be defined as a market's natural rhythm, an established tendency for prices to move in the same direction at a similar time every year. As such, it becomes a valid principle subject to objective analysis in any market.
In a market strongly influenced by annual cycles, seasonal price movement may become more than just an effect of seasonal cause. It can become so ingrained as to become nearly a fundamental condition in its own right --- almost as if the market had a memory of its own. Why? Once consumer and producers fall into a pattern, they tend to rely on it, almost to the point of becoming dependent on it. Vested interests then maintain it.
|Annual Price Cycles|
|Seasonal Patterns |
|Bottom||Seasonal Low||Greatest Supply/
|Ascent||Seasonal Rally||Increasing Demand/
|Peak||Seasonal High||Greatest Demand/
|Descent||Seasonal Decline||Decreasing Demand/
"Pattern" implies a degree of predictability. Future prices move when anticipating change and adjust when that change is realized. When those changes are annual in nature, a recurring cycle of anticipation/realization evolves. This recurring phenomenon is intrinsic to the seasonal approach in trading, for it is designed to anticipate, enter, and capture recurrent trends as they emerge and exit as they are realized.
The first step, of course, is to find a market's seasonal price pattern. In the past, weekly or monthly high and low prices were used to construct relatively crude studies. Such analysis might suggest, for instance, that cattle prices in April were higher than in March 67% of the time and higher than in May 80% of the time. Computers, however, can now derive a daily seasonal pattern of price behavior from a composite of daily price activity over several years. Properly constructed, such a pattern provides historical perspective on the market's annual price cycle.
Basic Pattern Dynamics
Consider the following seasonal pattern that evolved for January Heating Oil. Demand, and therefore prices, are typically low during July --- often the hottest month of the year. As the industry begins anticipating cooler weather, the market finds increasing demand for future inventory --- exerting upward pressure on prices. Finally, the rally in prices tends to climax even before the onset of the coldest weather as anticipated demand is realized, refineries gear up to meet that demand, and the market begins to focus instead on inventory liquidation.
The other primary petroleum product encounters a different, albeit still weather-driven, cycle of demand as exhibited in the seasonal pattern for August Gasoline. Prices tend to be lower during the poorer driving conditions of winter. However, as the industry begins to anticipate the summer driving season, demand for future inventory increases and exerts upward pressure on prices. By the official opening of the driving season (Memorial Day) refineries then have enough incentive to meet that demand.
Seasonal patterns derived from daily prices rarely appear as perfect cycles. Even in patterns with distinct seasonal highs and lows, seasonal trends in between are sometimes subject to various, even conflicting forces before they are fully realized. A seasonal decline may typically be punctuated by brief rallies. For example, even though cattle prices have usually declined from March/April into June/July, they have exhibited a strong tendency to rally in early May as retail grocery outlets inventory beef for Memorial Day barbecues. Soybean prices tend to decline from June/July into October's harvest, but by Labor Day the market has usually anticipated a frost scare.
Conversely, a seasonal rally may typically be punctuated by brief dips. For example, uptrends are regularly interrupted by bouts of artificial selling pressure associated with First Notice Day for nearby contracts. Such liquidation to avoid delivery can offer opportunities to take profits and/or to enter or reestablish positions.
Therefore, a seasonal pattern constructed from daily prices can depict not only the four major components of seasonal price movement, but also especially reliable segments of larger seasonal trends. Recognizing fundamental events that coincide with these punctuations can provide even greater confidence in the pattern.
Consider the seasonal price pattern that has evolved for September 30-Year Treasury Bonds. The U.S. Federal government's fiscal year begins October 1, perhaps increasing liquidity and easing borrowing demands somewhat (even if only for accounting purposes). Is it merely coincidental that the tendency for bond prices to rise from then tends to culminate with personal income tax liability for the calendar year?
Does the seasonal decline into May reflect a market anticipating tighter monetary liquidity as taxes are paid? Notice the final sharp decline beginning --- surprise! --- April 15th, the final date for payment of U.S. income taxes. Does liquidity tend to increase sharply after June 1 because the Federal Reserve finally recirculates funds?
Take a close look at the typical market activity surrounding December 1, March 1, June 1, and September 1 --- dates of first delivery against Chicago Board of Trade quarterly futures contracts on debt instruments. Finally, notice the distinct dips during the first and second week of the second month in each quarter --- November, February, May, and August. Bond traders know that prices tend to decline into the second day of a quarterly Treasury refunding --- at which time the market gets a better sense of the three-day auction's coverage.
Consider the pattern for November Soybeans as it has evolved in the 15 years since Brazil became a major producer with a crop cycle exactly opposite that in the Northern Hemisphere. Notice the tendency for prices to work sideways to lower into the "February Break" as U.S. producers market their recent harvest and Brazil's crop develops rapidly. By the time initial notices of delivery against March contracts are posted, the fundamental dynamics for a spring rally are in place - the Brazilian crop is "made" (realized), the pressure of U.S. producer selling has climaxed, the market anticipates the return of demand as cheaper river transportation becomes more available, and the market begins focusing attention on providing both an incentive for U.S. acreage and a premium for weather risks.
By mid-May, however, the amount of prime U.S. acreage available in the Midwest for soybeans is mostly determined and planting gets underway. At the same time, Brazil begins marketing its recent harvest. The availability of these supplies and the potential of the new U.S. crop typically combine to exert downward pressure on market prices. The minor peaks in late June and mid-July denote tendencies for crop scares to occur.
By mid-August, the new U.S. crop is "made" (realized), and futures can sometimes establish an early seasonal low. However, prices more often decline further into October's harvest low --- but only after rallying into September, perhaps on commercial demand for the first new-crop soybeans and/or concerns over an early crop-damaging frost. Notice also the minor punctuations (declines and rallies) associated with First Notice Day for July, August, September, and November contracts.
Such trading patterns do not repeat without fail, of course. The seasonal methodology, as does any other, has it own inherent limitations. Of immediate practical concern to traders may be issues of timing and contraseasonal price movement. Fundamentals, both daily and longer term, inevitably ebb and flow. For instance, some summers are hotter and dryer --- and at more critical times --- than others. Even trends of exceptional seasonal consistency are best traded with common sense, a simple technical indicator, and/or a basic familiarity with current fundamentals to enhance selectivity and timing of entry/exit.
How large must a valid statistical sample be? Generally, more is better. For some uses, however, "modern" history may be more practical. For example, Brazil's ascent as a major soybean producer in 1980 was a major factor in the nearly 180-degree reversal in that market's trading pattern from the 1970s. Conversely, relying solely on disinflationary patterns prevalent in 1981-1999 could be detrimental in a new inflationary environment.
During such historic transitions in underlying fundamentals, trading patterns will evolve. Analyzing cash markets can perhaps help neutralize such effects, but certain patterns specific to futures (such as those that are delivery- or expiration-driven) can get lost in translation. Thus, both sample size and the sample itself must be appropriate for its intended use. These may be determined arbitrarily, but best by a user fully cognizant of the consequences of that choice.
Related issues involve projecting into the future with statistics, which confirm the past but do not predict in and of themselves. The Super-Bowl winner/stock-market direction "phenomenon" is an example of statistical coincidence because there exists no cause-and-effect relationship. However, it does raise a valid issue: When computers mechanically sift only raw data, what discoveries are relevant? Does the simple, isolated fact that a pattern has repeated in 14 out of the last 15 years make it necessarily valid?
Certainly, patterns driven by known fundamentals inspire more confidence; but to know all relevant fundamentals in every market is impractical. Properly constructed seasonal patterns may typically help one find trends that have recurred in the same direction during the same period of time most years with a high degree of past reliability. Finding a "cluster" of such historically reliable trends, with similar entry and/or exit dates, not only reduces the odds of statistical aberration but also implies recurring fundamental conditions that presumably will exist again in the future and affect the market to one degree or another in a more or less timely manner.
A seasonal pattern merely depicts the well-worn path a market itself has tended to follow. It is a market's own consistency which provides the foundation for why seasonals work.