Is there a Halloween effect in the European stock market

Is there a Halloween effect in the European stock market

Amsterdam Business School

Name Catriona de Goede

Number 10280375

BSc in Business Economics Specialization Finance

Supervisor Ilko Naaborg Completion 2/2/2015

Abstract

Previous research shows that stock returns are higher in the months November until April than in the rest of the year. This remarkable phenomenon in stock markets is known as the Halloween effect or the Halloween indicator. It got its origin of an old market saying “Sell in May and go away”. In contrast to what other research reported, this research cannot find an empirical significant Halloween effect in the EURO STOXX 50 index between 2008-2013. Which is noteworthy because recent research has shown a strongly significant Halloween effect in European countries.

Keywords: Halloween effect, EURO STOXX 50, stock returns, efficient markets, market anomalies,

Table of Contents 1. Introduction 2 2. Literature review 3 2.1 Efficient Market 3 2.2 Market Anomalies 4 2.3 Halloween Indicator 5 2.3.1 Existing literature 5

2.3.2 Debate and Empirical evidence 5

2.3.2 Hypothesis 6

3. Methodology and Data 7

3.1. Methodology 7

3.2. Data and descriptive statistics 8

4. Analysis 9

4.1. Empirical Results 10

4.2 Robustness check 12

5. Conclusion and discussion 13

1. Introduction

This research is about the anomaly the Halloween effect (Halloween Indicator), also known as “Sell in May, and go away”, at the specified European market EURO STOXX 50. This Halloween effect is reported for the first time in 1998 (Jacobsen & Visaltanachoti, 2009). This anomaly indicates that stock returns are higher in the months between November and April than the returns between May and October. Assuming market efficiency, it is doubtful whether or not there could be any truth in a simple and inherited market saying such as “Sell in May and go away.” Apart from the January effect, there are no reasons to assume that market returns in the period May to October would be significantly different from the remainder of the year (Bouman & Jacobsen, 2002).

Although this anomaly is not as famous as the January effect or the Monday effect, there is already some research about this subject. In general, these previous studies focus on a Halloween effect in different countries around the world. This research differs from previous research as it investigates the Halloween effect on the European stock market, based on the EURO STOXX 50. This index contains the fifty biggest stocks from 12 Eurozone countries (STOXX.com). Beside the different area that will be investigated, also the time horizon is different. In previous studies (Bouman & Jacobsen, 2002; Jacobsen & Visaltanachoti, 2009; Haggerd & Witte, 2009) they took a time horizon varying from 18 years to sub periods of 26 years. This research considers a time horizon of 5 years, specifically 2008 to 2013.

The main purpose of this study is to test whether there is a Halloween effect in the European stock market between 2008 and 2013.

This is an interesting question because the Halloween effect does seem to disappear or reverse itself after discovery. It continues to exist even though investors may have become aware of it (Bouman & Jacobsen, 2002). As Schwert (2003) points out in his article about anomalies and market efficiency, many well-known anomalies do not hold in different random periods and this makes the Halloween effect

interesting. Bouman and Jacobsen (2002) also report a highly significant effect in European countries, and also proves that it is robust over time. This is why the EURO STOXX 50 is used to test this Halloween effect.

The research question will be answered with the use of four different models and the monthly returns of fifty companies. To test the four models, the simple OLS regression method is used. These models will be extensively discussed later in this study.

The remainder of this research is organized as follows. Section 2 is a literature review in which main theories and the hypothesis are discussed. The next section is about the methodology and data. Section 4 discusses the empirical results and the robustness check. The last section contains the main conclusions, some discussion about the limitations of this research and suggestions for further research.

2. Literature review

This section will discuss the main theories in the existing literature. First the theories behind efficient markets will be discussed, followed by the research about market anomalies and finally the theories behind the Halloween effect, the relatedness with this research and the hypothesis that will be tested.

2.1 Efficient Markets

An efficient market means that the current price of a security “fully reflects” all available information. Conditions for a capital efficient market are (i) no transaction costs in trading securities, (ii) all available information is costless available to all market participants, and (iii) all agree on the implications of current information for the current price and distributions of future prices of each security (Fama, 1970). Unfortunately all three conditions do not hold in the real world markets. Potential sources are transaction costs, information that is not freely available to all

participants, and disagreement among investors about the implications of given information.

Fama (1970) mentions that it is best to use the random walk model as an extension of the general expected return or “fair game” efficient markets model in the sense of making a more detailed statement about the economic environment. This random walk arises when the environment is such that the evolution of investors’ tastes and the process of generating new information are combined to produce equilibria in which return distributions repeat themselves through time.

In another paper of Fama (1991) is reported that market efficiency is not per se testable and that it must be jointly tested with another model of equilibrium, an asset-pricing model. When it is jointly tested with an asset-asset-pricing model and there are found some (seasonal) anomalies, the way it should be split between market

inefficiency or a bad model of market equilibrium is ambiguous. Keim (1988) reviews this study. He argues that seasonal anomalies in returns are anomalies in the sense that asset-pricing models do not predict them, but they are not necessarily barriers for market efficiency.

2.2 Market Anomalies

In recent years there has been a proliferation of empirical studies documenting unexpected or anomalous regularities in security rates of return. These anomalous regularities in average stock return are not explained by the capital asset pricing model (CAPM) (Fama & French, 1996). The findings present a potentially serious challenge to classical models of market equilibrium and have stimulated the

development of new theories that can account for them (Lakonishok & Smidt, 1988). There are different kind of anomalies. An anomaly like the Halloween effect is called a calendar anomaly or seasonal anomaly. This is because this effect appears to be related to the calendar or to the different seasons. Some other well-known calendar or seasonal anomalies are the January effect, the Monday effect, the Weekend effect and the Turn-of-the-Month effect.

The fact that these anomalous regularities are not explained by the capital asset pricing model means that there is some need for some expansion of this model. This is what Fama and French did with their three-factor model (1993). In this model they include (i) the excess return on a broad market portfolio (RM – Rf); (ii) the difference between the return on a portfolio of small stock and the return on a portfolio of large stock (SMB, small minus big); and (iii) the difference between the return on a portfolio of high-book-to-market stocks and the return on a portfolio of low-book-to-market stocks (HML, high minus low) (Fama & French, 1996).

However, it is possible that these anomalies are the product of sampling error and/or data mining. For this reason it is important to test for the existence of these regularities in data samples that are different from those in which they were originally discovered (Lakonishok & Smidt, 1988).

2.3 Halloween indicator 2.3.1 Existing literature

Since it was first reported in 1998, the Halloween effect is still present. The Halloween effect is based on the old market saying “Sell in May and go away” (Bouman & Jacobsen, 2002). This old market saying is also known as the Halloween indicator. Bouman and Jacobsen find in 2002 that there is a substantial difference between returns in the period May-October and the remainder of the year. They also find that there is a positive and significant relation between their three proxies for the length and timing of summer vacations, and the impact of vacation on trading activity and the Sell in May effect.

Jacobsen and Visaltanachoti mention in their research about the Halloween effect in U.S. sectors (2009) that there are three explanations for the changes in investor behavior causing the Halloween effect. The first explanation is, find by Bouman and Jacobsen (2002), the changes in risk aversion, or changes in liquidity, due to vacation behavior of investors. The second explanation is changes in risk aversion due to SAD, Seasonal Affective Disorder. This second explanation is

reported by Kamstra, Kramer and Levi (2003) who explain the similar pattern in stock returns as a SAD effect in stock returns. SAD refers to a seasonal affective disorder, where the decreasing hours of daylight during fall makes investors depressed. The last explanation is mood changes due to temperature changes. This last explanation is reported by Coa and Wei (2005) who test for a relation between temperature and stock returns and find that stock returns are significantly negatively related to temperature.

Haggerd and Witte (2009) report that it is important to take outliers into consideration when investigating the Halloween effect. Outliers are an important aspect that researchers have to investigate as a possible source of the Halloween effect. Other research where this is taken into account, like Lucey and Zhoa (2008), specified that the Halloween effect, when it does appear, might simply be the January effect in disguise (Haggerd & Witte, 2009).

2.3.2 Debate and Empirical results

In line with the main theories, Bouman and Jacobsen (2002) find that based on the old market saying “Sell in May and go away” (or the Halloween indicator) there is a statistically significant Halloween effect present at the 10-percent level in 20 of the 37

countries. So it can be said that the mean returns are significantly higher during November to April than during May to October. Bouman and Jacobsen (2002) face after their research still a problem: history and practice tells us that the old saying is right, while stock market logic tells us it is wrong. In the past, academic research offers a series of possible explanations for this type of finding, such as the lack of economic significance, data mining, or risk differences.

Jacobsen and Marquering (2008) shows that many things are correlated with seasonal effects so that it is difficult to distinguish between the explanations that are given in earlier research and are currently available.

2.3.3 Hypothesis

In this research there is only taken some specific market into account instead of a nation or specific different sectors. There is chosen for the EURO STOXX 50 index. The methodology used in this research is in the base very simple, but is broadened with different control variables in different models.

As many studies report, the effect tends to be particularly strong and highly significant in European countries. The Halloween effect is expected to result in low returns between May and October and high returns between November and April (Bouman & Jacobsen, 2002).

In total four different models will be used to test if there is a Halloween effect in the EURO STOXX 50. The hypothesis, based on existing literature, is that there is a Halloween effect in the European stock market. This means that the coefficient of the used dummy variable for the Halloween effect should be significant. If this is not the case, it can be concluded that there is no Halloween effect. As the theory expects, the coefficient should be positive. When the coefficient is positive, this means an excess return on the average monthly return. When the coefficient is negative, this means that the Halloween effect can be defined as smaller returns in November to April and larger returns in May to October. Negative returns are difficult to reconcile with time varying risk premiums or any equilibrium asset pricing model (Jacobsen & Visaltanachoti, 2009).

3. Methodology and Data

In this section the methodology and the data will be discussed. First the four models that are used in this study will be presented. Thereafter, the origin of the data will be discussed and descriptive statistics will be provided.

3.1. Methodology

In this study, the Halloween effect is measured with the monthly return of the fifty most important stocks of Europe. These most important companies of Europe are ranged together in the EURO STOXX 50 index. For answering the research question “Is there a Halloween effect in the European stock market?” this study makes use of the monthly returns of each company for the years 2008 to 2013. The monthly returns are calculated as the end of the month stock price minus the closing price of the month before, divided by the closing price of the month before, (Pt – Pt-1)/Pt-1. In total

there will be tested at four different linear models. The first two models are also used in the research of Bouman and Jacobsen in 2002 and the last two models are used in the research of Jacobsen and Visaltanachoti in 2009.

The first simple linear regression model to test if there is a Halloween effect is looking like

𝑅_𝑡 = 𝛼 + 𝛽₁ 𝐻𝐴𝐿_𝑡+ 𝜀_𝑡

where Rt is the return of the EURO STOXX 50 per month, HALt is a dummy variable

with HALt = 1 if the return is measured in the months November to April and zero

otherwise. α, 𝛽1 and εt are the models’ intercept, regression coefficient and error

term, respectively. The advantage of this simple regression model is that you can easily include other variables. As already mentioned, this study will use four different models which are an extension of this simple model.

The second model is also a linear regression model in which one more variable is included, namely

𝑅_𝑡 = 𝛼 + 𝛽₁ 𝐻𝐴𝐿_𝑡+ 𝛽₂ 𝐽𝐴𝑁_𝑡+ 𝜀_𝑡

where the variable JANt, which states for the January effect, is added. Also this

otherwise. α, 𝛽1 and εt are the models’ intercept, regression coefficient and error

term, respectively. This control variable is added to the model because some literature suggest that the Halloween effect is an broader version of the January effect. To correct for this, the variable JANt is added.

The third linear regression model includes two more control variables and will look like

𝑅_𝑡− 𝑅_𝑓,𝑡 = 𝛼 + 𝛽₁ 𝐻𝐴𝐿_𝑡+ 𝛽₂ 𝐽𝐴𝑁_𝑡+ 𝛽₃ (𝑅_{𝑀𝑘𝑡,𝑡}− 𝑅_𝑓,𝑡) + 𝜀_𝑡

where variable (RMkt – Rf) is added to second linear regression model. Also here α, 𝛽₁

and εt are the models’ intercept, regression coefficient and error term, respectively.

This model is the same as the one before only this model includes a variable that corrects for excess return on the market.

The last linear regression model that will be taken into account includes the four Fama and French factors and will look like

𝑅𝑡− 𝑅𝑓,𝑡 = 𝛼 + 𝛽1 𝐻𝐴𝐿𝑡+ 𝛽2 (𝑅𝑀𝑘𝑡,𝑡− 𝑅𝑓,𝑡) + 𝛽3 𝑆𝑀𝐵𝑡+ 𝛽4 𝐻𝑀𝐿𝑡

+ 𝛽5 𝑊𝑀𝐿𝑡+ 𝜀𝑡

where RMkt is the market return, Rf the risk-free rate of 10-year central government

bonds in the Eurozone, SMBt is the equal-weight average of the returns on the three

small stock portfolios for Europe minus the average of the returns on the three big stock portfolios, HMLt is the equal-weight average of the returns for the two high

book to market portfolios minus the average of the returns for the two low book to market portfolios and WMLt is the equal-weight average of the returns for the two

winner portfolios minus the average of the returns for the two loser portfolios. α, 𝛽₁ and εt are the models’ intercept, regression coefficient and error term, respectively.

Because of the different companies in the EURO STOXX 50 it can be helpful to correct for these differences with the popular Fama and French four factors.

3.2. Data and descriptive statistics

The data that is used in this research has been collected from DataStream. From all the fifty companies in the EURO STOXX 50 the monthly closing stock prices for the year 2008 to 2013 are collected. EURO STOXX 50 was formed at 28th _{of February}

1998. This index covers fifty stocks from 12 Eurozone countries: Austria, Belgium, Finland, France, Germany, Greece, Ireland, Italy, Luxembourg, the Netherlands, Portugal and Spain. In the Appendix you can find the list of these companies.

The market return is a collection of the closing prices of the EURO STOXX 50 and the return is calculated at the same way as is done with the closing prices of the fifty companies. The risk-free rate is collected from the DNB, De Nederlandsche Bank, from the 10-year Central Government Bond rates. The three other Fama and French factors are coming from their official data library (Fama & French, Data library). In this data library, they offer the monthly factors for several regions. For this study only the European values are used.

In table 1 below, the summary statistics are shown. It shows that there is a large difference in the minimum and maximum returns of all companies in the EURO STOXX 50. So is the highest return measured for one of the companies 70,75% and the lowest return -51,54%. This is notable in comparison with the mean return. This could indicate that there were large fluctuations in returns in the years between 2008 and 2013. The risk of all companies together are relatively low. It has a volatility of 9,21%. Also the market has a low volatility, 5,98%. Table 1 also shows that the volatility of the free is the lowest. This is in line with the theory behind the risk-free rate.

Table 1 Summary Statistics

Variable Obs Mean S.D. Min Max

Rt 3600 0,0021 0,0921 -0,5154 0,7075 RMkt 3600 -0,0030 0,0598 -0,1469 0,1469 SMBt 3600 0,0865 2,0202 -4,5100 4,8500 HMLt 3600 -0,1444 2,7198 -4,6000 7,4500 WMLt 3600 0,6224 4,9569 -25,9600 9,8700 Rf,t 3600 0,0376 0,0064 0,0210 0,0482

Note: Rt is the return of all companies together, RMkt is the market return, SMB, HML and WML are the

Fama and French factors, and Rf is the risk-free rate based on 10 year central government bonds of the

Euro area.

4. Analysis

In this section the main results will be presented. First only the empirical results and their meanings are discussed. Later on the robustness check will be discussed.

4.1. Empirical Results

This research investigates whether the monthly returns of the fifty companies of the EURO STOXX 50 significantly differ in the months November to April from the remainder of the year. The returns of each company are summarized in figure 1 and 2. Figure 1 shows the difference of the first 25 companies and figure 2 show the other 25 companies. The full company list is provided in Appendix 1.

Fig. 1 Monthly average returns of company 1 till 25 for the years 2008 to 2013.

Note: Summer includes the returns of the months between May and October and winter includes the returns of the month between November and April.

Fig. 2 Monthly average returns of company 26 till 50 for the years 2008 to 2013.

Note: Summer includes the returns of the months between May and October and winter includes the returns of the month between November and April.

-3,00% -2,00% -1,00% 0,00% 1,00% 2,00% 3,00% 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 Ret u rn Company Summer Winter -3,00% -2,00% -1,00% 0,00% 1,00% 2,00% 3,00% 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 Ret u rn Company Summer Winter

These two tables do not directly show a Halloween effect. Only 29 of the fifty companies show a possible Halloween effect. The other 21 companies have a higher return in the summer month than in the winter months, which does not corresponds with the Halloween effect theory. Looking at the business sector of these companies, it shows that the companies with higher returns in the summer months are often part of the banking and utilities sector. These sectors can be seen as part of the

consumption section, which tends to outperform in the summer months (Jacobsen & Visaltanachoti, 2009). So this economic results are in line with the theory.

Table 2 provides the results from all four regression analyses. From regression 1, which only take the dummy for Halloween effect into account, it can be concluded that in the months November to April the returns are 0,40% higher than in the other half of the year. This is not a very high percentage of excess return in comparison to the summer months, but also this result is not significant at a 5% level. Even at a 10% level this result is not significant so it cannot be concluded that there is a Halloween effect.

Table 2 Summary of regression 1 till 4.

Note: These regressions are based on a simple OLS regression. Under the coefficient the corresponded p-value is mentioned and the models are tested at a 5% significance level.

Independent Variables Dependent Variable Rt Dependent Variable Rt-Rf

(1) (2) (3) (4)

Halloween effect (HAL) 0,00403 0,00855 -0,00133 -0,00056

(0,186) (0,008) (0,583) (0,812)

January effect (JAN) -0,02712 0,00481

(0,000) (0,277) RMarket-Rf 1,01823 0,97336 (0,000) (0,000) SMB 0,00032 (0,602) HML 0,00073 (0,230) WML -0,00054 (0,070) Intercept 0,00008 0,00008 0,00609 0,00469 (0,969) (0,969) (0,001) (0,019) Summary Statistics R-squared 0,0005 0,0065 0,4510 0,4520 N 3600 3600 3600 3600

In the second regression, a control variable for the January effect is included. These results show a 0,86% excess return in the winter months. This result is

significant at a 5% level. In this case it could concluded that correcting for the January effect, there is a Halloween effect. One remarkable note in this result is the negative significant coefficient for the January effect. This suggest that in the month January the returns are 1,86% lower than in the other months. Lower returns in January are in contrast with the theory about the January effect. At the turn of the year, certain types of securities tend to produce abnormal returns (Haugen & Jorion, 1996).

The third regression controls for the January effect and includes a correction for excess returns in the market. Unless the higher R-squared of 0,4510 compared to the other models, the coefficient of the Halloween indicator is negative and

insignificant. This negative coefficient could mean that the returns in the month November till April are 0,13% lower than in the other months of the year. Only the coefficient for the market correcting variable is significant at a 5% level and has a value of 1,02 which means that the companies in the EURO STOXX 50 are reacting in line with the market.

In the fourth regression the Fama and French four factors are taken into account. Also this regression has a higher R-squared, but the coefficients are again insignificant at a 10% significance level. The Halloween indicator is again negative and indicate a 0,56% lower return in the winter month. The Fama and French four factors are taken into account because of the different companies in the EURO STOXX 50, but it seems that it is not a valuable model to test the Halloween effect in this European market.

4.2 Robustness check

In the research of Haggerd and Witte (2009) it is shown that the Halloween effect is robust considering outliers, the January effect and transaction costs. To check whether the results of this research are robust to alternative specifications, the same

regressions are done but with the correction of robust standard errors. This regression analysis shows that the coefficients do not change, as expected. Only the p-values changed, but the empirical results and conclusions will stay the same. This means that the significance of the coefficients did not change. In table 3 the results of these regressions are summarized.

Table 3 Regression results with robust SE

Independent Variables Dependent Variable Rt Dependent Variable Rt-Rf

1 2 3 4

Halloween effect (HAL) 0,00403 0,00855 -0,00133 -0,00056 (0,189) (0,007) (0,579) (0,803)

January effect (JAN) -0,02712 0,00481

(0,000) (0,316) RMarket-Rf 1,01823 0,97336 (0,000) (0,000) SMB 0,00032 (0,661) HML 0,00073 (0,222) WML -0,00054 (0,216) Intercept 0,00008 0,00008 0,00609 0,00469 (0,970) (0,970) (0,001) (0,013) Summary Statistics R-squared 0,0005 0,0065 0,4510 0,4520 N 3600 3600 3600 3600

Note: These regressions are based on a simple OLS regression with robust standard errors. Under the coefficient the corresponded p-value is mentioned and the models are tested at a 5% significance level.

Only the p-value of the variable WML in the last regression changed from a 10% significance to an insignificant result in comparison to the results in table 2. The reason why p-values are changed is the correction of standard errors for possible heteroskedasticity. This means that the errors do not have a constant variance across all observations because it depends on the independent variables (Stock & Watson, 2011). Without correcting for heteroskedasticity, standards errors are not a reliable basis for hypothesis tests and confidence intervals.

5. Conclusion and discussion

The main purpose of this research is to test whether there is a Halloween effect in the European stock market between 2008 and 2013. This is tested with the use of the stock returns of the fifty companies in the EURO STOXX 50. The Halloween effect is tested with the use four different regression models. The first model included only a variable for the Halloween effect, the second model controls for the January effect, the third model controls for the January effect and excess return on the market, and the fourth model includes the four Fama and French factors. It can be concluded that

only model two represents a significant result, which means that the null hypothesis can be rejected. For the other three models this null hypothesis cannot be rejected, which means that it cannot be concluded that there is a Halloween effect based on the results of these models. Comparing these empirical results with the economic results from figure 1 and 2, it was expected that when there is a Halloween effect, it would not be very significant. Only 29 companies represent better returns in the winter months November till April.

Previous research mentioned that there is a highly significant Halloween effect detectable in European countries, but looking at the EURO STOXX 50, with the fifty greatest stocks of Europe, there is not found a significant Halloween effect.

The reason for this insignificant results could be the shorter time period of five years compared to the time horizon of previous research. As it was mentioned, other studies used a time horizon varying from 18 to 26 years. A larger data set could make the estimation results more reliable. Furthermore, it could be possible that there are more factors that have an effect on the stock market returns and which are not included in the models used in this study.

Because of the limitations of this research, it would be interesting to investigate more about the Halloween effect. The effect at a broader market in Europe could be studied, which include more companies in more different sectors, or some specific countries in Europe, like PIIGS (Portugal, Ireland, Italy, Greece and Spain) which have a high public debt. It may also be interesting to test the difference between the years before the crisis in 2008 and the years thereafter. Has the extent of the Halloween effect changed over the years? Is there still a Halloween effect

detectable after the crisis? These are questions that may be interesting for further research about the Halloween effect.

References

Bouman, S. & Jacobsen, B. (2002) The Halloween Indicator, “sell in May and Go Away”: Another Puzzle, The American Economic Review, 92(5), 1618-1635. Cao, M. & Wei, J. (2005) Stock market returns: A note on temperature anomaly,

Journal of Banking and Finance, 29, 1559-1573.

DNB, De Nederlandsche Bank, Interest rates,

http://www.statistics.dnb.nl/en/financial-markets/interest-rates/index.jsp, 19 January 2015.

STOXX, EURO STOXX 50 Index Information,

http://www.stoxx.com/indices/index_information.html?symbol=SX5E, 16 January

2015.

Fama, E.F. (1970) Efficient Capital Markets: A Review of Theory and Empirical work, The Journal of Finance, 25(2), 383-417.

Fama, E.F. (1991) Efficient Capital Markets: II, The Journal of Finance, 46(5), 1575-1617.

Fama, E.F. & French, K.R. (1993) Common risk factors in the returns of stocks and bonds, Journal of Financial Economics, 33, 3-56.

Fama, E.F. & French, K.R. (1996) Multifactor Explanations of Asset Pricing Anomalies, The Journal of Finance, 51, 55-84.

Fama & French, Data library,

http://mba.tuck.dartmouth.edu/pages/faculty/ken.french/data_library.html, 16 January 2015.

Finance Yahoo, historical prices EURO STOXX 50,

http://finance.yahoo.com/q/hp?s=%5ESTOXX50E+Historical+Prices, 16 January

2015.

Haggerd, K.S. & Witte, H.D. (2009) The Halloween effect: Trick or treat?

International Review of Financial Anlysis, 19, 379-387.

Haugen, R.A. & Jorion, P. (1996) The January Effect: Still There after All These Years, Financial Analysts Journal, 52, 27-31.

Jacobsen, B. & Marquering, W. (2008) Is it the weather? Journal of Banking &

Finance, 32, 536-540.

Jacobsen, B. & Visaltanachoti, N. (2009) The Halloween Effect in U.S. Sectors, The

Financial Review, 44, 437-459.

Kamstra, M.J., Kramer, L.A. & Levi, M.D. (2003) Winter blues: a SAD stock market cycle, American Economic Review, 93, 324-343.

Keim, D.B. (1988) Stock market regularities: A synthesis of the evidence and explanations, E. Dimson, ed., Stock market anomalies (Cambridge University Press, Cambridge, UK).

Lakonishok, J. & Smidt, S. (1988) Are Seasonal Anomalies Real? A Ninety-Year Perspective, The Review of Financial Studies, 1(4), 403-425.

Lucey, B.M. & Zhoa, S. (2008) Halloween or January? Yet another puzzle,

Internationa Review of Financial Analysis, 17, 1055-1069.

Scwhwert, G.W. (2003) Anomalies and market efficiency, Handbook of the

Economics of Finance, 1(B), 939-974.

Stock, J.H. & Watson, M.W. (2011) Introduction to Econometrics, Edinburgh Gate: Harlow.

Appendix

Table 4 Number, name, sector and country of each company in the EURO STOXX 50.

No. Name Company Super sector Country No. Name Company Super sector Country

1 AIR LIQUIDE Chemicals FR 26 IBERDROLA Utilities ES

2 AIRBUS GROUP Industrial Goods & Services FR 27 INDITEX Retail ES

3 ALLIANZ Insurance DE 28 ING GROEP Banks NL

4 ANHEUSER-BUSCH INBEV Food & Beverages BE 29 INTESA SANPAOLO Banks IT

5 ASML HOLDING Technology NL 30 L'OREAL Personal & Household Goods FR

6 ASSICURAZIONI GENERALI Insurance IT 31 LVMH Personal & Household Goods FR

7 AXA Insurance FR 32 MUENCHENER RUCK. Insurance DE

8 BASF Chemicals DE 33 NOKIA Technology FI

9 BAYER Chemicals DE 34 ORANGE Telecommunications FR

10 BBV.ARGENTARIA Banks ES 35 PHILIPS ELTN.KONINKLIJKE Industrial Goods Services NL

11 BANCO SANTANDER Banks ES 36 REPSOL YPF Oil & Gas ES

12 BMW Automobiles & Parts DE 37 RWE Utilities DE

13 BNP PARIBAS Banks FR 38 SAINT GOBAIN Construction & Materials FR

14 CARREFOUR Retail FR 39 SANOFI Healthcare FR

15 DAIMLER Automobiles & Parts DE 40 SAP Technology DE

16 DANONE Food & Beverages FR 41 SCHNEIDER ELECTRIC SE Industrial Goods & Services FR

17 DEUTSCHE BANK Banks DE 42 SIEMENS Industrial Goods & Services DE

18 DEUTSCHE POST Industrial Goods & Services DE 43 TELEFONICA Telecommunications ES

19 DEUTSCHE TELEKOM Telecommunications DE 44 TOTAL Oil & Gas FR

20 E ON Utilities DE 45 UNIBAIL-RODAMCO Real Estate FR

21 ENEL Utilities IT 46 UNICREDIT Banks IT

22 ENI Oil & Gas IT 47 UNILEVER CERTS. Personal & Household Goods NL

23 ESSILOR INTL. Healthcare FR 48 VINCI Construction & Materials FR

24 GDF SUEZ Utilities FR 49 VIVENDI Media FR

Table 5 Fama and French factors RMkt* SMB*** HML*** WML*** Rf** RMkt-Rf RMkt* SMB*** HML*** WML*** Rf** RMkt-Rf 200801 -13,79% 0,13 0,37 0,6 4,22% -18,01% 201101 5,76% -0,79 4,84 -5,56 3,94% 1,82% 200802 -1,80% 2,7 -0,52 1,75 4,14% -5,94% 201102 2,01% -1,13 0,81 -0,18 4,48% -2,47% 200803 -2,59% -0,54 2,23 0,2 4,07% -6,66% 201103 -3,39% 2,31 -1,87 2 4,49% -7,88% 200804 5,43% -2,48 -0,07 0,27 4,28% 1,15% 201104 3,45% -0,97 -0,64 2,44 4,66% -1,21% 200805 -1,23% 0,58 -1,28 4,6 4,42% -5,65% 201105 -4,96% 0,09 -2,26 0,55 4,37% -9,33% 200806 -11,25% 1,58 -0,83 9,66 4,81% -16,06% 201106 -0,47% -1,42 0,16 2,31 4,37% -4,84% 200807 0,45% -2,23 0,73 -2,13 4,82% -4,37% 201107 -6,25% 0,38 -3,76 0,86 4,59% -10,84% 200808 -0,07% 0,02 -0,17 -4,92 4,50% -4,57% 201108 -13,79% 0,87 -3,83 -0,49 4,21% -18,00% 200809 -9,73% -3,19 0,6 3,83 4,50% -14,23% 201109 -5,32% -1,35 -1,95 -1,44 4,04% -9,36% 200810 -14,69% -4,51 -2,61 9,87 4,42% -19,11% 201110 9,43% -3,2 -1,61 0,18 4,09% 5,34% 200811 -6,23% -1,17 -3,53 2,71 4,20% -10,43% 201111 -2,30% -2,72 -3,42 6,06 4,41% -6,71% 200812 0,71% -1,34 2,11 -0,85 3,89% -3,18% 201112 -0,60% -0,98 -1,23 2,85 4,11% -4,71% 200901 -8,61% 4,85 -3,97 3,41 4,10% -12,71% 201201 4,32% 2,84 0,91 -8,61 3,92% 0,40% 200902 -11,66% 4,24 -4,6 4,5 4,20% -15,86% 201202 3,95% 1,44 -0,37 -1,81 3,75% 0,20% 200903 4,80% -2,51 1,9 -10,03 4,14% 0,66% 201203 -1,39% 0,47 -1,7 3,5 3,29% -4,68% 200904 14,69% 2,19 5,6 -25,96 4,09% 10,60% 201204 -6,90% 1,38 -4,47 6,62 3,39% -10,29% 200905 3,20% 1,22 -1,18 -7,72 4,13% -0,93% 201205 -8,13% 0,2 -2,67 7,37 3,53% -11,66% 200906 -2,02% 1,34 -1,28 2,5 4,33% -6,35% 201206 6,88% -4,49 3,13 -3,94 3,41% 3,47% 200907 9,84% -3,86 3,66 -1,75 4,09% 5,75% 201207 2,69% -1,25 -2,59 4,47 3,25% -0,56% 200908 5,19% 1,82 7,45 -9,33 3,89% 1,30% 201208 4,94% -0,14 3,16 -3,52 3,01% 1,93% 200909 3,51% 1,89 1,04 -3,06 3,86% -0,35% 201209 0,56% 1,46 2,28 -0,98 2,43% -1,87% 200910 -4,50% 1,46 -3,42 3,66 3,80% -8,30% 201210 2,01% -0,61 1,74 0,5 2,31% -0,30% 200911 1,96% -2,07 -1,18 2 3,84% -1,88% 201211 2,86% -2,34 -0,45 2,11 2,25% 0,61% 200912 6,00% -1,88 -1,5 0,43 3,87% 2,13% 201212 2,36% 2,31 2,77 -0,99 2,10% 0,26% 201001 -6,35% 4,74 -1,86 0,54 4,10% -10,45% 201301 2,54% 0,56 4,28 -0,6 2,40% 0,14% 201002 -1,74% -0,6 -1,67 0,33 4,11% -5,85% 201302 -2,57% 1,94 -3,28 4,3 2,86% -5,43% 201003 7,43% -0,97 4,18 5,05 3,99% 3,44% 201303 -0,36% -0,78 -4,23 2,61 3,03% -3,39% 201004 -3,90% 3,4 -0,61 2,06 4,17% -8,07% 201304 3,35% -1,48 3,7 0,86 2,86% 0,49% 201005 -7,33% -0,29 -3,51 1,3 3,68% -11,01% 201305 2,13% 1,11 2,57 0,75 2,69% -0,56% 201006 -1,42% -0,08 -2,16 1,67 3,70% -5,12% 201306 -6,03% 2,22 -2,69 1,13 3,07% -9,10% 201007 6,56% -1,71 4,94 -1,68 3,62% 2,94% 201307 6,36% -1,61 2,59 3,53 3,10% 3,26% 201008 -4,35% -0,08 -2,83 3,51 3,44% -7,79% 201308 -1,69% 2,95 0,32 -2,18 3,10% -4,79% 201009 4,76% 1,09 0,21 5,04 3,50% 1,26% 201309 6,31% 0,08 1,08 2,44 3,41% 2,90% 201010 3,53% 0,29 0,14 0,04 3,34% 0,19% 201310 6,04% -0,13 4,24 2,02 3,16% 2,88% 201011 -6,82% 1,25 -3,18 7,82 3,73% -10,55% 201311 0,61% 0,81 -0,64 2,61 3,17% -2,56% 201012 5,35% 2,49 1,31 1,48 4,07% 1,28% 201312 0,72% 0,43 0,17 1,64 3,31% -2,59%

* Here RMkt is defined by the monthly stock return of the EURO STOXX 50 from January 2008 to December 2013. ** Here Rf is the interest rate of the 10 ear central government bonds from the Eruo area.

Table 6 Linear simple regression per company

Company No. of Obs α β1 t-value p-value Company No. of Obs α β1 t-value p-value

AIR LIQUIDE 72 0,0022 0,0053 0,47 0,642 IBERDROLA 72 -0,0028 -0,0032 -0,16 0,870

AIRBUS GROUP 72 0,0165 0,0048 0,19 0,851 INDITEX 72 0,0247 -0,0158 -1,01 -0,316

ALLIANZ 72 -0,0123 0,0313 1,38 0,173 ING GROEP 72 -0,0012 0,0101 0,26 0,794

ANHEUSER-BUSCH INBEV 72 0,0071 0,0151 0,74 0,459 INTESA SANPAOLO 72 -0,0049 -0,0045 -0,16 0,873

ASML HOLDING 72 0,0111 0,0182 0,80 0,425 L'OREAL 72 0,0018 0,0070 0,52 0,602

ASSICURAZIONI GENERALI 72 0,0019 -0,0116 -0,52 0,602 LVMH 72 0,0089 0,0016 0,09 0,932

AXA 72 0,0020 0,0051 0,17 0,867 MUENCHENER RUCK. 72 -0,0003 0,0088 0,69 0,491

BASF 72 -0,0034 0,0264 1,28 0,206 NOKIA 72 -0,0048 -0,0119 -0,34 0,734

BAYER 72 0,0098 -0,0016 -0,11 0,917 ORANGE 72 -0,0013 -0,0216 -1,56 0,123

BBV.ARGENTARIA 72 0,0055 -0,0123 -0,47 0,638 PHILIPS ELTN.KONINKLIJKE 72 -0,0010 0,0099 0,48 0,630

BANCO SANTANDER 72 0,0069 -0,0155 -0,61 0,546 REPSOL YPF 72 0,0017 -0,0005 -0,03 0,978

BMW 72 0,0046 0,0196 0,85 0,397 RWE 72 -0,0097 -0,0096 -0,51 0,612

BNP PARIBAS 72 0,0081 -0,0102 -0,38 0,703 SAINT GOBAIN 72 -0,0134 0,0274 1,11 0,272

CARREFOUR 72 -0,0050 0,0030 0,16 0,873 SANOFI 72 0,0061 -0,0034 -0,25 0,801

DAIMLER 72 -0,0023 0,0157 0,59 0,560 SAP 72 0,0038 0,0127 0,79 0,431

DANONE 72 -0,0044 0,0087 0,70 0,486 SCHNEIDER ELECTRIC SE 72 -0,0066 0,0281 1,53 0,131

DEUTSCHE BANK 72 -0,0230 0,0396 1,28 0,204 SIEMENS 72 0,0007 0,0042 0,21 0,832

DEUTSCHE POST 72 -0,0004 0,0149 0,64 0,525 TELEFONICA 72 0,0056 -0,0228 -1,37 0,175

DEUTSCHE TELEKOM 72 0,0065 -0,0141 -0,95 0,345 TOTAL 72 -0,0016 -0,0005 -0,04 0,967

E ON 72 -0,0196 0,0107 0,56 0,581 UNIBAIL-RODAMCO 72 0,0087 -0,0036 -0,23 0,817

ENEL 72 -0,0030 -0,0105 -0,57 0,570 UNICREDIT 72 -0,0160 0,0047 0,13 0,895

ENI 72 -0,0071 0,0083 0,56 0,581 UNILEVER CERTS. 72 0,0076 -0,0085 -0,73 0,467

ESSILOR INTL. 72 0,0004 0,0185 1,37 0,176 VINCI 72 -0,0124 0,0293 1,60 0,114

GDF SUEZ 72 -0,0047 -0,0093 -0,56 0,579 VIVENDI 72 -0,0053 0,0033 0,19 0,850

SOCIETE GENERALE 72 0,0011 -0,0019 -0,06 0,956 VOLKSWAGEN PREF. 72 0,0173 0,0022 0,07 0,942