School Holidays and Stock Market Seasonality

• Published date: 01 March 2018
• Author: Lily Fang,Yuping Shao,Chunmei Lin
• Date: 01 March 2018
• DOI: http://doi.org/10.1111/fima.12182

School Holidays and Stock Market

Seasonality

Lily Fang, Chunmei Lin, and Yuping Shao∗

Using school holiday data from 47 countries, we find a strong link between school holidays and

market returns. Stock market returns in the month after major school holidays are 0.6% to 1%

lower than in other months. This explains, but is not limited to, the “September effect.” In the

United States, September is the only month that exhibits a negative average return over the past

century. The postschool holiday effect remains even with monthly fixed effects. We explore the

explanation that the effect is due to investor inattention during school holidays, which slows the

incorporation of (negative) information in security prices.

In this paper, we document a novel asset pricing pattern. Globally, stock market returns are

0.6% to 1% lower in the months after major school holidays than at other times. We present

evidence that this seasonality in stock returns is attributable to investor inattention during school

holidays resulting in the slow incorporation of news, especially negative news.

It is widely known among Wall Street traders that September, the month after the summer

school holiday, has historically been the worst performing month. This “September effect” is

striking in magnitude. For example, an article in the Wall Street Journal (WSJ) entitled “How to

Play the September Effect” (September 6, 2013) reports that since 1896, the Dow Jones average

return for the month of September has been −1.09%, while that of all other months has been

+0.75%. A more recent article on the same topic illustrates the persistence of the September

effect. It is the only month that has a negativeaverage return for 20, 50, and 100 years.1However,

the origin of the September effect is not well understood. The aforementioned 2013 WSJ article is

accompanied by the following subtitle “it’s wellknown that September has been the worst month,

on average, for stocks. But no one has come up with a plausible explanation of why that is.”

We contribute to the understanding of the September effect, a remarkable yet hitherto not

well-explained phenomenon in the academic literature, in two ways. First, we conjecture and

present evidence that the September effect reflects a broader “postschool holiday effect”whereby

market returns after major (long duration) school holidays are significantly lower than other

times. In addition, we propose an explanation for the postschool holiday effect and also the

September effect that has long puzzled traders. We hypothesize that returns after major school

holidays are low because during school holidays,the market is collectively less attentive to news.

As a result, information is incorporated into prices slowly. This effect is particularly strong for

negative information as taking advantage of negative news is more difficult and requires more

Wethank Bing Han (Editor), an anonymous referee, seminar participants at INSEAD, the HarvardBusiness School, MIT

Sloan, the Yale School of Management, Boston College, Brandeis University, Erasmus University Rotterdam, and VU

University Amsterdam for their comments and suggestions. All errorsand omissions are our own.

∗Lily Fang is an Associate Professor of Finance at INSEAD in Singapore. Chunmei Lin is an Assistant Professor of

Finance at Erasmus School of Economics and the Tinbergen Institute in Rotterdam, Netherlands. YupingShao is a PhD

student in Finance at the National University of Singaporein Singapore.

1“Some Stock Strategists Brace for September Swoon,” The WallStreet Journal, September 1, 2014.

Financial Management •Spring 2018 •pages 131 – 157

132 Financial Management rSpring 2018

attention, a scarce cognitive resource (Kahneman, 1973), than positive news.2Miller (1977) and

Diamond and Verrecchia (1987) provide a theoretical foundation that suggests constraints on

short selling would impede the incorporation of negative information into stock prices. In this

paper, we present evidence that inattention among professional investors curtails their arbitrage

activities, compounding the effects of short sale constraints.

Prior literature demonstrates a preholiday return effect. Lakonishok and Smidt (1988) and Ariel

(1990) examine the daily returns around public holidays and find significant positive returns on

the days immediately preceding public holidays.The postschool holiday low return effect we find

here differs from the preholiday effect in two ways. First, the nature of the holiday we examine

is different. Our school holiday periods are longer (they are at least a week in duration) than

the public, statutory holidays studied previously. Moreover, during school holidays, the markets

remain open as opposed to public and statutory holidays when the market is closed. Since markets

remain open during the school holidays we examine, any reduced trading cannot be explainedby

traders’ inability to trade. It is attributed to traders being less attentive and less likely to trade.

The second difference is that while we find low postschool holidayretur ns, the previousliterature

finds high preholiday returns. Our effect is robust to controlling for preholiday returns.

Our empirical analysis consists of two main parts. First, we establish the empirical fact that

there is a broader postschool holiday effect that encompasses the September effect.To this end, we

collect school holiday data using school calendars from 47 countries around the world. We find

a striking pattern. Returns are, on average, 1% lower in the months after major school holidays.

More importantly, this lower return is not driven by September alone, eventhough the September

effect is pervasive in the northern hemisphere. When September is excluded,there is still a retur n

gap of 0.6% between postschool holiday months and other times, and the difference remains

highly significant.

To make sure that our findings are indeed a postschool holiday effect rather than a spurious

result, we conduct a number of validation tests that exploit exogenous variations in school

calendars. For example, in the United States, while the school yearbegins in September for many

states, some states in the south have school years that begin in early August. For example, in

Georgia, Indiana, Nevada, Oklahoma, Tennessee, and Hawaii, the school holiday ends in early

August. For stocks headquartered in these states, we find that it is August, rather than September,

that exhibits particularly low returns. In France, whilesome major school holidays have common,

nationwide dates, others are staggered by region in order to reduce tourist congestion. We expect

and find that the postschool holiday effect is stronger after nationwide school holidays than

regional ones. Moreover, we determine that one region’s postholiday dates do not predict low

returns for another region (a placebo test) unless the predictor is the last region that goes on

holiday. These tests make the implicit assumption that investors exhibit local bias. Local stocks

are heavilyheld by local investors. Thus, we are jointlytesting local bias and the postschool holiday

effect. Local bias among professional investorshas been documented in the prior literature. Coval

and Moskowitz (1999) find that US investment managers exhibit a strong preference for locally

headquartered firms. Bok, Kang, and Kim (2010) conf irm that local institutional investors have

an information advantage and, as such, their trading plays an important role in incorporating the

information of local firms.

2Forexample, Hong, Lim, and Stein (2000) provide evidence that firm-specif ic information, especially negativeinfor ma-

tion, diffuses slowly. Stambaugh, Yu,and Yuan (2012) find that the short legs of long-short strategies are more profitable

following strong market-wide sentiment indicating the slower incorporation of negative information. Fang and Yasuda

(2014) confirm that negative recommendations consistently earn larger alphas that dissipate more slowly than positive

recommendations.