October 3, 2005
It's well known that in post-war data, the stock market has shown seasonally strong returns, on average, during the period from November through April (about 18% annualized since 1945), and more modest results from May through October (about 7% annualized). This pattern is known as the Hirsch Cycle, though various analysts have modified it by adjusting it here and there. Still, since the average return during the May-October period has generally been higher than the Treasury bill yield, a simple buy-and-hold strategy on the S&P 500 would have generated a higher return than a timing strategy based on this pattern.
While historical seasonal patterns don't influence our investment positions to any significant extent, I don't entirely ignore them either. With October here, I thought it would be a good time to look at some of these. Several years ago, Nelson Freeburg – who is also a member of our Board of Trustees – published a series of reports on the subject in Formula Research. Nelson's work on this is far more extensive than mine, so the results presented here are drawn from and fully attributable to Nelson's insightful research.
Probably the first caveat when talking about seasonal patterns is that they don't always persist. Nelson notes that “between 1886 and 1950, August was by far the most profitable month for stocks. Over that time the DJIA rose an average of 2.62% in August. In the second best month, January, stocks rose just 0.80% on average. After World War II August did an about-face. Since 1950 the average August return dropped to 0.03%, little more than breakeven.”
Still, the patterns are interesting, and at least some have been successfully followed in real-time. Norm Fosback used to publish a seasonal trading strategy in his newsletter, Market Logic, with good results. That strategy was invested in S&P 500 futures only during the two days before a market holiday, the last two trading days of each month, and the first four trading days of the next month (the “pre-holiday” and “month-end” seasonals).
Nelson notes that the annualized return isolated only to those days works out to about 34% since 1928. Including the days when the strategy was out of the market, presumably earning the T-bill yield, the annualized gain from 1952 through 2001 would have been 13.6% before transaction costs, compared with 12.8% for a buy-and-hold. That's not a bad comparison given that it assumes being invested only one-third of the time. Of course, trading costs, taxes and slippage would impair the results from actually trying to trade that strategy, given the hundreds of entry and exit points, but it's actually a pattern worth keeping in mind if you're relatively indifferent between executing a trade one day or a few days later, and one of those seasonals lies in-between.
Yet another seasonal pattern is the Electoral Cycle – stocks have historically performed well during the year before an election and during the election year itself (18.8% and 9.3% on average since 1950), and have lagged during the post-election year and the mid-term year (3.1% and 7.6%, respectively).
With those patterns in mind, Nelson examined the interaction between the Election Cycle and the Hirsch Cycle. Specifically, the Election Cycle is bullish during the pre-election and election year and bearish otherwise. The Hirsch Cycle is bullish from November through April and bearish otherwise. When both have been bullish, the S&P 500 has appreciated at an annualized rate of 28%. When both have been bearish (as they are now), the average annualized gain has been only 4%. Mixed combinations averaged about 10% for Election bullish, Hirsch bearish, and 11% for Election bearish, Hirsch bullish.
Though that would seem to be a negative for this month, it turns out that the month of October has historically displayed a very idiosyncratic pattern. October has typically been very strong during the negative phase of the electoral cycle and very bearish in the positive phase of the cycle. So depending on how much one risks over-refining the data, the seasonal implications might be positive for this month rather than negative (one can go too far, of course, since October 19th has a very, very, bad seasonal record due to a single data point in 1987...)
Is seasonality worth using in practice?
All of which brings us to a finer point that has been a subject of my own research. Having examined the evidence on seasonality, and also considering the risk that seasonal patterns might break down in real-time, I wondered several years ago how much to rely on these patterns. On one hand, some of them actually had statistical significance that persisted even after accounting for limited sample size. On the other hand, with no solid explanations except those grounded in pure speculation, these patterns had the feel of superstition.
Fortunately, after examining the data, there turned out to be no dilemma at all. As I noted a few years ago, it's very true that the market has typically performed better during the November-April period than during the May-October stretch. But this performance breaks into a much different profile when the status of market action is considered. During the seasonally favorable November-April period, the market historically generated an average annualized return of 27.8% during weeks when our measures of market action were also favorable (as of the prior Friday), while unfavorable market action produced an average gain of just 0.4% annualized. Likewise, during the seasonally unfavorable May-October period, favorable market action produced an average annualized gain of 17.6%, while unfavorable market action produced an average loss of -8.5% annualized.
In short, seasonality has historically had a measurable impact on market returns, particularly when the pattern has been aligned with the status and quality of market action, but it is not sufficient reverse the favorable or unfavorable implications from those measures. Similarly, unfavorable valuation creates only a modest drag on returns when the quality of market action is favorable, even in combination with unfavorable seasonality.
As a result, I never use or act on seasonal patterns in a way that contradicts my conclusions from the analysis of valuations and market action. However, I do allow seasonal considerations to modestly amplify, mute or affect the specific timing of investment actions that I plan to take in any event.
For example, valuations are clearly unfavorable here. If market action was also clearly unfavorable, even the fact that it is October in a post-election year would not be enough to justify a positive exposure to market risk. At the same time, if I had some sell orders to execute, I would be most comfortable executing them within the first few days of the month (due to the favorable end-of-month seasonal) provided that the selling prices were satisfactory, and particularly if I had the opportunity to make the sales into short-term strength.
That said, it's important to remember that a bird in the hand is worth two in the bush. Seasonal patterns rarely amount to an average outcome of more than a percent or two on a monthly basis. Those considerations almost never result in my deferring a trade for more than a few days. I just don't believe in seasonals enough to put off good, planned transactions in hopes that some particular tendency will play out.
So suppose I was looking to add to certain stock holdings. In the event that the market actually delivered that bird in the hand – in the form of an actual market decline of a percent or two, I would almost certainly disregard any potential seasonal tendency in favor of the actual opportunity that had already arrived.
Given that market action is still clinging to a tenuously favorable condition, the various favorable seasonal factors here are a reminder that stocks have at least some potential that to strengthen rather than weaken. While I haven't taken any significant exposure on that consideration, it does slightly factor into my evaluation of various buying and selling points, and also adds to my comfort with the small contingent call option position the Strategic Growth Fund holds for other reasons.
Again, as of last week, valuations remained unusually unfavorable, while market action still clings to a tenuously favorable condition. Meanwhile, market breadth has been choppy and leadership has started to occasionally reverse, with a greater number of new lows than new highs on several recent days. The fact that these numbers are reasonably significant (both daily new highs and new lows over about 2.5% of total issues traded) is a mark of growing internal dispersion. It's not a profound negative, but certainly worth monitoring closely. As noted above, the Strategic Growth Fund holds a modest number of call options here that slightly modify our hedge percentage to roughly 85%, leaving the Fund with a modest sensitivity, roughly 15% of portfolio value, to overall market fluctuations.
In bonds, the Strategic Total Return Fund continues to carry a limited duration of about 2 years, mostly in short-term bills (whose yields are competitive with longer term bonds due to the flat yield curve) and moderate maturity inflation protected securities. The primary source of day-to-day fluctuation will continue to be the Fund's exposure to precious metals shares, which after clipping off a further small position on strength, is slightly below 20% of Fund assets. Keep in mind that this exposure will induce a certain amount of day-to-day fluctuation that I do not intend to hedge away, as I continue to view the conditions for these shares as favorable.
New from Bill Hester: Rally Boosters
|Home | The Funds | Open an Account | Account Access | Research & Insight | Site Map|
For more information about investing in the Hussman Funds, please call us at
513-587-3440 outside the United States
Site and site contents © copyright 2003 Hussman Funds. Brief quotations including attribution and a direct link to this site (www.hussmanfunds.com) are authorized. All other rights reserved and actively enforced. Extensive or unattributed reproduction of text or research findings are violations of copyright law.
Site design by 1WebsiteDesigners.