An Explanation for the Halloween Effect (Sell in May, go away)
On this page:
We provide the first empirical explanation for the existence of the Halloween effect, commonly known as "Sell in May and go away." This phenomenon suggests that stocks perform better between November and April than they do between May and October, and is observable in equity markets of 36 countries, from at least 1970 to present (Bouman and Jacobsen, 2002; Zhang & Jacobsen, 2021). According to Bouman and Jacobsen (2002), the mention of "Sell in May and go away" can be traced back to a Financial Times article from 1964, and possibly even earlier.
Various explanations for the anomaly have been proposed over the decades, yet none have been fully supported by data or further analysis, nor have they been widely accepted by the academic community. These explanations include:
- Negative news sentiment (Bouman & Jacobsen, 2002)
- Summer vacation (Bouman & Jacobsen, 2002)
- January effect (Lucey & Zhao, 2008)
- Weather (Jacobsen & Marquering, 2008)
- Political climate (Powell et al., 2009)
- Self-fulfilling prophecy (Lloyd et al., 2017)
- And many more
We provide an explanation that wasn't considered in previous research: the significant reduction in regulatory information volume during the May-October period. We analyze the disclosure dates of all SEC filings from the last three decades (1994 to 2023), which are the primary source of information for shareholders, investors, and news outlets, and find a significant decrease in information volume during the May-October period. We observe a similar, though less pronounced, temporal information asymmetry in the press releases published by companies.
Information is the primary driver of informed decision-making and stock prices. During the May-October period, the reduced volume of new information may result in fewer decisions by investors. This decrease in decision-making leads to less market activity, reduced trading volume, and lower volatility. Consequently, the market may stagnate or even decline due to the lack of activity of their participants. Therefore, the reduction in disclosure volume during the summer months may provide a plausible explanation for the Halloween effect.
This summary outlines our key findings prior to the publication of our academic paper.
Main Findings and Results
Our research reveals a significant temporal pattern in disclosure activities, with a 23.4% decrease in monthly SEC filing volume during the May-October period, aligning with the Halloween effect. In other words, approximately one quarter less information flows into the market during this period compared to the November-April period. Additionally, insider transactions drop by 32.3%.
Figure 1 shows the mean number of SEC filings published per month from 2004 to 2023, with the 2.5% and 97.5% percentiles highlighted and the high and low months marked accordingly. We find that the most SEC filings are consistently published in February each year for the last three decades.
We also observe that September is the month with the least information disclosed, followed by October. In Figure 2, when filtering for the top 2% most common disclosure form types (which represent 80% of the total annual information volume), we find that May marks the month of highest filing volume, with February being the second highest. June and July consistently show a sharp decline in information volume, with August being an exception due to SEC-mandated 10-Q, 13F and N-PORTP filings, which peak in August.
The average total number of filings per month can be categorized by their EDGAR form types, with insider trading and material event disclosures representing between 40-50% of the entire annual information volume. Both filing types are unscheduled and are not disclosed on a repetitive or cyclical timeline.
Figure 3 illustrates the temporal pattern in insider trading activity reported on Form 4. February and May consistently see the most insider trading, followed by a sharp decline from June to September, with September marking the month with the lowest disclosures.
As shown in Figure 4, a similar pattern is observable in material event disclosures, which peak in May and decline over the summer months, reaching their lowest volume in September. It is important to note that some events in 8-K filings are disclosed on a repetitive schedule. For instance, Item 2.02 "Results of Operations" is often used by companies to disclose preliminary quarterly or annual results before filing their 10-K or 10-Q reports. Therefore, the August volume spike can be attributed to the quarterly results published in August, aligning with the 10-Q pattern shown further below.
Finally, we decompose the overall temporal pattern into their constituent parts. We group form types into the following categories:
- Clear pattern not caused by regulatory deadlines: These are event-driven and mostly unscheduled filings.
- Occurring once per year: Examples include annual reports on 10-Ks, shareholder proxy statements (DEF 14A), and mutual fund annual reports (N-CEN).
- Occurring quarterly: Examples include 10-Qs, investment manager portfolio holdings (13F-HR), and mutual fund quarterly reports (NPORT-P).
- Constant (no pattern, more or less the same every month, noise): Examples include 424B2 and FWP prospectuses, and SEC comment letters.
Limitations of our Explanations
Our data series begins in 1994, meaning we cannot validate our explanation for the period from 1970 to 1994. Before the introduction of the EDGAR system and electronic filings, the SEC regulated the market following the 1929 stock market crash with the introduction of the Securities Exchange Act of 1934, with information being disclosed on paper.
The Sarbanes-Oxley Act (SOX) of 2002 significantly changed the filing landscape, making Form 4 and Form 8-K filings the dominant types. However, these specific form structures did not exist prior to 2004. It can be assumed that the information disclosed in these forms post-2004 was still reaching the market and investors through other means before their formal introduction, supporting the validity of our explanation.
For example, 8-K filings provide a structured vehicle for material event disclosure but did not alter the frequency of corporate events such as mergers, acquisitions, management changes, entering into debt obligations or other material agreements, shareholder voting, bankruptcies, or changes to accountants. These events occurred with the same regularity before the introduction of 8-K filings; SOX, among other things, merely introduced a more structured method for reporting such information.
Conclusion
We propose a potential explanation for the Halloween effect, commonly known as "Sell in May and go away." Our analysis suggests that the volume and timing of SEC filings, which are the primary source of information for shareholders, investors, news outlets, journalists, and analysts, play a significant role in this phenomenon. Press releases, product launches, mergers, acquisitions, and other corporate actions also provide crucial market information and are reported with the SEC as Form 8-K filings.
Our findings indicate a regulatory information drought during the summer months, with June, July, and September consistently having the lowest volume of SEC filings throughout the year. Insider trading activity drops by 32% during this period.
During the May-October period, the relative scarcity of new information potentially leads to fewer informed decisions by investors, resulting in reduced trading volumes and lower market activity. This lack of activity can cause the market to stagnate or even decline.
Conversely, the peak in filings in February and May — driven by unscheduled insider trades, material events like management changes, and scheduled filings by investment managers disclosing annual portfolio holdings and performances, as well as annual and quarterly reports — triggers increased trading activity as investors react to new information.
Following this peak, the reduced volume of filings leads to fewer market-moving events, resulting in a period of relative calm or stagnation. This decline in filings and subsequent increase later in the year aligns with the historical pattern of better stock performance from November to April, providing a plausible explanation for the Halloween effect.
References
- Bouman, S., & Jacobsen, B. (2002). The Halloween indicator, “Sell in May and go away”: Another puzzle. American Economic Review, 92(5), 1618–1635. https://doi.org/10.1257/000282802762024683
- Jacobsen, B., & Marquering, W. (2008). Is it the weather? Journal of Banking & Finance, 32(4), 526–540. https://doi.org/10.1016/j.jbankfin.2007.08.004
- Lucey, B. M., & Zhao, S. (2008). Halloween or January? Yet another puzzle. International Review of Financial Analysis (Online)/International Review of Financial Analysis, 17(5), 1055–1069. https://doi.org/10.1016/j.irfa.2006.03.003
- Powell, J. G., Shi, J., Smith, T., & Whaley, R. E. (2009). Political regimes, business cycles, seasonalities, and returns. Journal of Banking & Finance, 33(6), 1112–1128. https://doi.org/10.1016/j.jbankfin.2008.12.009
- Lloyd, R., Zhang, C., & Rydin, S. (2017). The Halloween Indicator is More a Treat than a Trick. Journal of Accounting and Finance, 17(6), 96–108. http://digitalcommons.www.na-businesspress.com/JAF/JAF17-6/LloydR_abstract.html
- Zhang, C. Y., & Jacobsen, B. (2021). The Halloween indicator, “Sell in May and Go Away”: Everywhere and all the time. Journal of International Money and Finance, 110, 102268. https://doi.org/10.1016/j.jimonfin.2020.102268