Is “Go Away in May” a Good Portfolio Play?

A Few More Pieces to the Halloween Effect Puzzle

It is quite common for financial news sources to continually offer us new ways to engage financial markets as we attempt to invest our capital. Occasionally these sources will repeat stories on methods and trading rules that seem to provide reliable profits. For example, the “Halloween effect” is a trading rule that seems to receive attention every spring and summer. Earlier this year, Forbes, CNBC, Barron’s, CBS News, and FOX Business News have all run stories on the effect.[1] [2] [3] [4] [5] [6]

The Halloween Effect

The Halloween effect is a simple trading rule that puts an investor into positions in the equities market from November through April (i.e. winter) and into positions in interest-bearing securities from May through September (i.e. summer). Among the many calendar-based stock market anomalies (e.g. day-of-the-week effect, day-of-the-month effect, the holiday effect and the January effect), the Halloween effect seems particularly compelling because of its seemingly large potential payoffs and the endless attention it receives in the media.

With the backdrop of the academic contention of efficient markets, is this trading rule (aka sell in May and go away) really as easily executed and as profitable as the mantra asserts? Or are there other overlooked factors that may influence the investment risk and rewards? For example, some argue that spurious correlations (i.e. some catalyst unrelated to the calendar that has little to do with stock prices) and chance are driving this unusual result. To address these questions, we consider published empirical results on portfolio formation and outcomes, as well as offer a potential application for individual investors—with an associated word of caution.

For the Record

Many, if not most, calendar anomalies fail to deliver outsized gains to investors once they are identified and published.[7] [8] [9] A number of studies have documented the Halloween effect, and tried to explain its existence and persistence. For example Bouman and Jacobson[10] examine the seasonal effect in 37 different countries over the period of 1970–1998. In 35 of 37 markets the returns in the November/April periods are higher than those in the May/September period. In 20 of these markets there is a statistically significant advantage to following the Halloween Effect trading rule. This advantage ranges from 1.5 percent to 8.9 percent annually. The advantage remains in 14 of these markets even after controlling for the January Effect (another calendar anomaly that suggests January is a preferred month for going long in the equity markets and earning abnormally high returns).

Bouman and Jacobson conclude that data mining, the January effect, volume changes or transitory risk premia in interest rates do not drive their results. They speculate that summer vacations are the most likely reason for the anomaly, especially because July and August tend to be poor-performing months in almost all countries in their sample. This conclusion is consistent with the fact that many of the markets that manifest this seasonal pattern are European.

Since the publication of the Bouman and Jacobson studies, several follow-up studies have been published. One of these studies (Jacobson and Visaltanachoti)[11] shows that the Halloween effect is likely confined to specific sectors and industries. They test 17 U.S. sectors and 49 industries. Their results indicate that for successful November/April investing, the most attractive sectors are Consumer Durables, Chemicals, Construction, Fabricated products and Machines. The four most attractive May/October industries are Drugs, Soap, Tobacco, and Perfumes.[12]

They also test a sector rotation strategy over the period 1926–2006 and over a second period from 1990–2006. The strategy requires investing in five consumer sector industries (Beer, Drugs, Food, Tobacco, and Utilities) in the summer and then rolling the portfolio over to five production sectors (Automobiles, Coal, Construction, Machinery, and Textiles) during the winter. The results are summarized in Table 1 below. The average monthly returns, as well as the risk-adjusted return metrics (as measured by the Sharpe Ratio and Jensen’s Alpha) suggest that the rotation strategy may have some merit.

Andrade, Chhaochharia, and Fuerst[13] perform an out-of-sample test of these results over the period 1998–2012. While their results are not as prevalent and pronounced as those in previous studies, they still confirm the Halloween effect in many European countries and in the U.S. Additionally, not all European stock markets manifest the anomaly and only some of the European results exceed those in the U.S., essentially weakening the argument that extended European vacation times are a preferred explanation of the Halloween effect.

We also note that there are many who are skeptical regarding the power of the Halloween effect and its ability to offer investors a “free lunch.” For example, Maberly and Pierce[14] test the Halloween effect using stock market (1970–1998) and futures market data (April 1982–April 2003). They note that the Bouman and Jacobson study terminates just before the Russian Government’s announcement of a moratorium on debt repayment that sent Long-Term Capital Management and financial markets into temporary distress (August 1998). They also assert that outliers drive the winter returns: specifically the market crash in October 1987. Controlling for this outlier and the January effect, Maberly and Pierce show that the Halloween effects disappears in the stock market, and the buy and hold strategy in the futures markets exceeds the returns offered by two variations of rotation strategies.

Table 1 Ferraro

Table 1: Bouman and Jacobson Halloween effect study results

We note that while the Maberly and Pierce study casts doubt on the rotation strategy, we feel that it does not adequately test it for two reasons. First, Maberly and Pierce tested a time period (1970–1998) that is much shorter than those tested in the other studies. Moreover, while the crash in October of 1987 can certainly be considered an outlier that drives the results over a shorter time frame, the summer months in general tend to experience more negative returns, with June, August, and September leading the way.[15] Second, the test using futures data may not detect the Halloween effect if it is sector specific, which was indicated in the Jacobson and Visaltanachoti study in 2009. Therefore, we do not feel that the industry level Halloween effect issue is settled by the previously discussed studies.

Finally, we make an additional observation that is critical of all of the studies that have considered the Halloween effect. All of these studies use average returns to determine the presence or absence of the effect. However, to effectively test this strategy one needs to consider geometric rates of return (i.e. compounded) rather than average. To see why this is important, consider the following example. Suppose an investor put $1,000 to work in an investment two years ago. At the end of the first year, the investor found that her portfolio had increased to $2,000, indicating a 100 percent return. At the end of the second year, the portfolio’s value was once again $1,000, indicating a 50 percent decline in the value of the portfolio. If we are considering average returns, this investor did rather well, with a 25 percent average return (100% – 50%/2). However, the geometric or compounded return is 0 percent—the investor is right back where she started two years ago. Therefore, when we test the strategy we want to look at compounded returns, not average returns. Compounded returns will accurately reflect the actual returns and risk to the strategies for specific holding periods, whereas average returns, as previously illustrated, can fool us into concluding that a profitable trading rule exists when, in fact, it may not.

Executing the Strategy

Of course, even if the Halloween effect is real, it doesn’t matter to investors if they cannot take advantage of it. Most academic studies use data that is aggregated across markets, sectors, and industries to run their tests—and this is a primary reason that individual investors are unable to capture the rewards of many of the “anomalies” that are presented as evidence against an efficient market. In the studies measuring sector effects, the total firms in each defined industry ranges from 1 to 72. Creating a portfolio of 72 firms to represent an industry isn’t a practical approach for most investors. Alternative approaches available to investors include forming portfolios comprised of single stocks, small groups of industry stocks (e.g. 2 to 5), sector mutual funds, or sector ETFs. While constructing portfolios of individual stocks would result in portfolios that most closely replicate the results of academic studies, this approach would incur considerable time and transactions costs, and potential drifting away from what works in the published studies. Of the remaining approaches, the sector ETFs are much more cost effective due to their lower management and expense ratios, relative to mutual funds. Therefore, we use sector ETFs to test the sector rotation strategy of the Halloween effect.

SPDR ETFs are used to test the Halloween strategy.[16] The SPDR sector ETFs were available for public investing on December 22, 1998. Therefore, the first investable period for the switching strategy begins in May 1999. We run the strategy through April 2014. This results in 15 years of testable data. For the consumer industries the Consumer Staples (XLP), Healthcare (XLV), and Utilities (XLU) ETFs are used. To proxy the production industries the Consumer Discretionary (XLY) and the Industrial (XLI) sector SPDRs are used.

The Broad Market Switching Strategy

We begin by presenting the results of the Halloween effect strategy using the broad market (i.e. S&P 500) as the timing instrument. The results of the switching strategy and the buy & hold strategy are presented in Table 2. The unadjusted switching strategy results are presented in column (1). The results for the switching strategy after transactions costs (0.10% for T-bill buy/sell and 0.25% for ETF buy/sell) and taxes (25%) are presented in column (2). Column (3) contains the results for the buy & hold strategy.

Several interesting outcomes are obtained. First, when comparing average (and median) returns, which the previous studies do, the buy & hold strategy seems to dominate—especially after factoring in transactions costs and taxes. However, the standard deviation of the buy & hold strategy (0.1743) is materially higher than that of either of the other two strategies. That is, the buy and & hold strategy has much more volatile returns than the switching strategy. A simple difference in means test indicates that the strategy mean returns are not statistically different from one another. Moreover, the Sharpe Ratio, which standardizes the strategy’s excess returns by the volatility (i.e. standard deviation), indicates that the switching strategy provides higher risk-adjusted returns than the buy & hold strategy. This is true on a pre-tax and after-tax basis.

Table 2 Ferraro

Table 2: Returns to the Market Switching and Buy & Hold Strategies

We also note that the holding period returns are much higher for the switching strategy (116.3%) than they are for the buy & hold strategy (93.8%). Put differently, one dollar invested in the pre-tax switching strategy becomes $2.16 while the same dollar invested in the buy & hold strategy becomes $1.93. However, after factoring in transactions costs and taxes, the holding period return for the buy & hold strategy is materially higher than that for the switching strategy (66.4%). This suggests that the switching strategy will not work in taxable accounts, but appears to work in tax deferred accounts.

Finally, the cumulative returns for each strategy are presented in Figure 1. The volatility of the buy & hold strategy can be seen clearly in holding period returns, with two notable large draw-downs. The first occurs in the 3rd and 4th periods coinciding with the recession in 2001–2002. The second, and largest, draw down is associated with the beginning of the great recession in 2008. Similar to the observations made by Maberly and Pierce, the advantage to the switching strategy seems to stem from large negative outcomes during the summer period—especially the 2008 event. However, when this “outlier” is eliminated from the sample, the difference in means outcome remains unchanged due to the still-relatively-high standard deviation of the buy & hold strategy sample.

As investors, we are left with uncertainty regarding this switching strategy. If we are willing to stick to the strategy over long-periods of time (i.e. years), it appears that it may be worthwhile in tax-deferred accounts. However, because of the impact of taxes, the switching strategy underperforms the simple buy & hold strategy. We next turn our attention to employing the strategy by limiting ourselves to specific industries rather than the broader market.

Figure 1 Ferraro

Figure 1: Market Switching Strategy Holding Period Returns

The Switching Strategy with Specific Industry Applications

In this version of the switching strategy the rules are slightly different. In summer, the portfolio is 100 percent invested in the ETFs representing the consumer sector; in the winter, the portfolio in 100 percent invested in the ETFs representing the production sector. The descriptive statistics from this strategy (i.e. ETF sector rotation) and the buy & hold strategy are presented in Table 3.

Similar to the results reported in the previous section, a difference in means test indicates that the differences in the average returns are not statistically significant. Again the relatively high volatility and small sample size drive this result. However, this switching strategy has much larger annual increases and drawdowns, and higher volatility than the market switching strategy. Nevertheless, we note that compounded annual returns (7.98% vs. 5.28%) and the holding period returns (216.5% vs. 116.3%) are much larger than those in the market switching strategy.

Table 3 Ferraro

Table 3: Returns to the Sector Switching and Buy & Hold Strategies

Another important outcome of this switching strategy relates to the after-tax results relative to the buy & hold strategy. Similar to the previous test, the after-tax results of this strategy provide better risk-return outcomes than the buy & hold strategy, as measured by the Sharpe Ratio. However, this strategy also provides higher compounded returns and holding period returns than the buy & hold strategy, suggesting that the strategy is appropriate for taxable accounts as well as tax-deferred accounts. A dollar invested in the pre-tax and post-tax switching strategies becomes $3.16 and $2.18, respectively. However, a dollar invested in the buy & hold strategy becomes $1.94. The cumulative returns related to this strategy, and relative to the buy & hold strategy, are provided in Figure 2.

Figure 2 Ferraro

Figure 2: Sector Switching Strategy Holding Period Returns

We also note some additional differences in this switching strategy’s results. First, the after-tax version of this strategy seems to be reliably better than buy & hold. That is, over the 15-period test the buy & hold cumulative returns only exceed those of the after-tax switching strategy in one period. Second, over much of the test period the cumulative returns are similar for the buy & hold and the pre-tax and post-tax switching strategy. Finally, most of the advantage to the switching strategy seems to manifest in the final five years. These years represent the period 2009–2014, the most recent bull market.

Conclusions

Based on the empirical results of tests of the Halloween effect, it appears that there is a historical pattern of higher average stock returns in the winter months and lower average stock returns in the summer months. But is this historical pattern reliable enough to allow equity market investors to increase their returns to their portfolios while reducing risk? While the empirical tests suggest that a simple trading rule may allow investors to do just that, we offer the following observations and caveats:

  1. The empirical data derived from long-term studies are often presented and compared based on average returns—not compounded or geometric returns. For an investor executing a strategy repeatedly over a long investment horizon, the geometric risk-return characteristics are more important than average returns.
  2. When incorporating reasonable transactions costs and tax assumptions into the trading strategy, most, if not all, of the advantage disappears. Therefore, the strategy seems best suited for tax deferred investing.
  3. When you start matters. That is, the year you start the strategy can have a large impact on the cumulative returns. In simple tests involving 5-year and 10-year investment horizons (not provided in this study) it is evident that the advantage to the strategy doesn’t typically begin to manifest itself until around year 10. This means that for an investor to use this strategy, they must diligently execute it for a decade (or more) before they begin to see a clear benefit from using it. Investors have to ask themselves if they have the staying power to do this.
  4. Empirical results indicate that the Halloween effect is much more reliable when executing the trading strategy using industry level instruments rather than market level instruments. The results of the switching strategy using sector/industry ETFs seemed to be much more robust than those derived from the simple market index switching strategy. However, there was more risk involved in the former strategy. Perhaps further refinement of the strategy using individual stocks to represent industries will offer even more compelling evidence that this strategy works. But this is a question that can only be answered with additional research.
  5. The primary benefit of this strategy seems to be the ability to avoid the worst parts of market corrections—or portfolio drawdowns. The reason for this is not well understood. It has been hypothesized that major market corrections occur during certain parts of the year due to politics, phycology, and vacations—but no clear evidence has emerged to explain this outcome. Therefore, if an investor is willing to execute this strategy they must be willing to assume that all of the worst news for financial markets will continue to emerge during the summer interval. In our opinion there is no compelling reason to do this.

While we believe that the Halloween effect is an intriguing curiosity, there are many uncertainties standing in the way of investors and successful execution of the trading strategy developed to take advantage of the historical stock return pattern. To make this trading rule produce profits requires the discipline and diligence to rebalance a portfolio every six months for many years. But even then, there is no guarantee of favorable outcomes.

 

[1] Ferri, R., “Busting the Sell In May and Go Away Myth,” Forbes.com (2014). Retrieved on September 1, 2014 from http://www.forbes.com/sites/rickferri/2013/04/08/busting-the-sell-in-may-and-go-away-myth/.

[2] Gittler, M., “Sell in May and Go Away: Is it True?” CNBC.com (2014). Retrieved on September 1, 2014 from http://www.cnbc.com/id/101630987#.

[3] Hulbert, M., “Sell in May and Go Away?” Barrons.com (2014). Retrieved on September 1, 2014 from http://online.barrons.com/news/articles/SB50001424053111903409104579523753992188132.

[4] Swedroe, L., “Did You Sell in May?” CBSnews.com (2014). Retrieved on September 1, 2014 from http://www.cbsnews.com/news/did-you-sell-in-may/.

[5] Tuchman, M., “Does Sell in May and Go Away Work?” Forbes.com (2013). Retrieved on September 1, 2014 from http://www.forbes.com/sites/mitchelltuchman/2013/04/30/does-sell-in-may-and-go-away-work/.

[6] Vasel, K. B., “Sell in May and Go Away”—Does it Still Work?” FOXbusiness.com (2014). Retrieved on September 1, 2014 from http://www.foxbusiness.com/2014/05/01/sell-in-may-and-go-away-still-applicable/.

[7] Malkiel, B., “The Efficient Market Hypothesis and Its Critics,” The Journal of Economic Perspectives, Vol. 17, 59-82, (2003).

[8] Thaler, R., “Anomalies—The January Effect,” Economic Perspectives, Vol. 1, No. 1, 197-201, (1987).

[9] Thaler, R., “Anomalies – Seasonal Movements in Security Prices II: Weekend, Holiday, Turn of the Month, and Intraday Effects,” Economic Perspectives, Vol. 1, No. 1, 169-177, (1987).

[10] Bouman, S., and Jacobson, B., “The Halloween Indicator, ‘Sell in May and Go Away’: Another Puzzle,” American Economic Review, Vol. 92, No. 5, 1618-1635 (2002).

[11] Jacobson, B., and Visaltanachoti, N., “The Halloween Effect in U.S. Sectors,” The Financial Review, Vol. 44, 437–459 (2009).

[12] We note that all of these industries fall into the Consumer durables sector.

[13] Andrade, S. C., Chhaochharia, V., and Fuerst, M. E., “Sell in May and Go Away’ Just Won’t Go Away,” Financial Analysts Journal, Vol. 69, No. 4, 94–105 (2013).

[14] Maberly, E. D., and Pierce, R. M., “Stock Market Efficiency Withstands another Challenge: Solving the “Sell in May/Buy after Halloween Puzzle,” Econ Journal Watch, Vol. 1, No. 1, 29–46 (2004).

[15] Browning, E.S., “Some Stock Strategies Brace for September Swoon,” WSJ.com (2014). Retrieved on September 1 from http://online.wsj.com/articles/some-stock-strategists-brace-for-september-swoon-abreast-of-the-market-1409592143.

[16] Currently there are many sector funds that are more concentrated in the specific strategy industries. However, most of them became available for public investing in 2006 and 2007. Therefore, there is insufficient data to adequately test the switching strategy using these ETFs.

 

Authors of the article
Steven R. Ferraro, CFA, PhD
Steven R. Ferraro, CFA, PhD,

, is an associate professor of finance at Pepperdine’s Graziadio School of Business and Management where he teaches corporate finance, valuation and corporate combinations, and investments. His current research interests include corporate restructuring, event-driven investing, and real estate investment trusts. Dr. Ferraro is managing director of the Center for Valuation Studies and principal of Ferraro Capital Management. He holds a PhD from Louisiana State University and is a Chartered Financial Analyst (CFA). He is also a recent recipient of the Howard A. White teaching award.

Richard Powell, JD, PhD, CPA
Richard Powell, JD, PhD, CPA,

, is an Associate Professor of Accounting at Pepperdine University. Courses taught include Financial Accounting, Managerial Accounting, and Taxation. Dr. Powell has over 20 years of consulting experience in the fields of Accounting, Taxation, Law, and Finance. As a practicing attorney, his experience has emphasized commercial law, real estate, litigation, and taxation. Dr. Powell holds a PhD in Accounting from the University of Arkansas, a JD in Law from the University of Illinois, an MBA from the University of Washington with concentrations in Accounting and Finance, and a BA from Carroll College. He is a licensed attorney, a member of the Washington State Bar Association, and a Certified Public Accountant.

More articles from 2014 Volume 17 Issue 3
Related Articles
GBR Special Issue: Spiritual Leadership

GBR Special Issue: Spiritual Leadership

Living in a complex world and moving into the smart machine age, the need for good leadership is even greater; spiritual leadership provides a compass to navigate through difficult decisions.