The Mood Factor in the Dutch Stock Exchange: Evidence 1998-2008

The Mood Factor in the Dutch Stock Exchange:

Evidence 1998-2008

An empirical analysis of the effects of weather, lunar circle, and Dutch National football team results, on the returns of the Dutch Stock Exchange.

Panagiotis Liontos

June 2008

UNIVERSITY OF GRONINGEN

The Mood Factor in the Dutch Stock Exchange: Evidence

1998-2008

An empirical analysis of the effects of weather, lunar circle, and Dutch National football team results, on the returns of the Dutch Stock Exchange.

Panagiotis Liontos1

ABSTRACT

This paper investigates, empirically, the relationship between five mood proxy variables (based on weather, lunar circle and Dutch national football team results) and daily Dutch equity returns, for the period 1998-2008. This study is motivated by psychological evidence and casual intuition, that people’s decisions are influenced by their feelings, especially when the decision involves risk and uncertainty. It appears that weather results are a matter of local ownership of companies, since negative correlation between temperature, good

weather2 and equity returns is found only on small-cap index. As far as moon phases are

concerned, negative returns, of about 0,05%, around the six days of the full moon (three before and three after the full moon) are identified, only after controlling for economic changes in a European index. I also find another surprising result which is positive returns (of 0,2%) after losses of the Dutch National football team, but only for the large-cap index.

JEL classification: G12; G14

Keywords: Weather effect, lunar effect, football effect, mood, Dutch Stock Exchange

…It is one of the secrets of Nature in its mood of mockery that fine weather lays heavier weight on the mind and hearts of the depressed and the inwardly tormented than does a really bad day with dark rain sniveling continuously and sympathetically from a dirty sky.

Muriel Spark, Territorial Rights, 1979.

1

MSc BA Finance Student, University of Groningen.

2

…There is something haunting in the light of the moon; it has all the dispassionateness of a disembodied soul, and something of its inconceivable mystery.

Joseph Conrad, Lord Jim, 1900. …Some people think football is a matter of life and death. I assure you, it’s much more serious than that.

Bill Shankly3, 1955

3

PREFACE

Throughout the academic year 2007-2008 I have been studying the MSc BA Finance at the University of Groningen. During the last term of this Master, students are required to write a thesis on a topic, which has to be chosen by them. The subject of my Master Thesis is the mood factor in the Dutch stock exchange, for the period 1998-2008.

During the process of writing my thesis many people helped me a lot to whom I will always be grateful. First of all, my supervisor, Drs M. M. Kramer, who was always willing to advise me and give me practical help. I would also like to thank my family for their mental and financial support. Finally, I thank Maria.

TABLE OF CONTENTS

PREFACE...4

1. Introduction...6

1.1 Mood, Judgement and Decisions...7

1.2 Mood and Investor Decision Making...8

1.2.1 Weather, Mood...9

1.2.2 Weather, Stock Market...10

1.2.3 Lunar Phases, Mood...12

1.2.4 Lunar Phases, Stock Market...12

1.2.5 Football, Mood...13

1.2.6 Football, Stock Market...14

2. Data ...15 3. Methodology ...17 3.1 Weather...18 3.2 Lunar Circle...20 3.3 Football...21 3.4 Joint Tests...22 4. Results ...23 4.1 Weather Results...23

4.1.1 Weather Results for the Large-Cap Index...24

4.1.2 Weather Results for the Small-Cap Index...25

4.2 Lunar Phases...27

4.2.1 Lunar Phases Results for Large-cap Index...27

4.2.2 Lunar Phases Results for the Small-Cap Index...28

4.3 Football Results...29

4.3.1 Football Results for the Large-Cap Index...30

4.3.2 Football Results for Small-Cap Index...30

4.4 Joint Tests...31

4.4.1 Joint Tests for the Large-Cap Index...31

4.4.2 Joint Tests for the Small-Cap Index...31

5. Conclusions...32

6. References ...33

1. Introduction

THE EXISTENCE OF PREDICTABLE BEHAVIOUR in stock returns may lead to

profitable trading strategies and abnormal returns. Predictable behaviour, itself, rejects the hypotheses for efficient capital markets as suggested by Fama (1970). But, can we actually predict stock returns? Investors tend to be ruled by their psychological and behavioural biases, which influence their decisions. They are ordinary people that have feelings and emotions which many times make them take irrational decisions. Research by Harlow and Brown (1990) and Odean (1998 & 1999) shows that when making investment decisions, investors face biases such as loss aversion, overconfidence and mood fluctuations. Concerning the latter, a relatively new and interesting kind of studies, tests the weather, the lunar phases and football results which are found to have major impact on human and investor mood and, therefore, on asset prices. But if stock prices are influenced upon the mood of investors and not on their reasoning, two things are probably occurring. First, that mood can, actually, influence investors’ decisions and secondly that mood has an impact on the marginal investor4. It is striking that the efficient market concept is in doubt, since the valuation by rational investors doesn’t compensate for the irrationality of others. If we consider that the marginal investor’s mood is influenced by weather, lunar circle and football results, then clearly the markets are not efficient.

The purpose of this paper is to investigate whether there is a causal relationship between Dutch equity returns and the following mood proxy variables: cloud cover, humidity, temperature, lunar phases, and football results of the Dutch national team, for the period January 1998-January 2008, and consequently, to test for the efficient market hypothesis.

My paper contributes to the investor mood literature by (1) testing for the existence of the mood proxy variables mentioned before, and (2) testing for interactions between them. In addition, it contributes to the existent literature, as these variables haven’t been tested for the Dutch stock exchange since 2004 and also have never before measured together for the specific stock market.

4

1.1 Mood, Judgement and Decisions

There is extended literature in psychology which considers how emotions and mood influence human decision making. Wright and Bower (1992), Bagozzi et al. (1999), claim that individuals in good mood have positive evaluations of many kinds, such as life satisfaction, past events, people and consumer products. On the other hand, people that are in a bad mood, according to Isen et al. (1978) and Forgas and Bower (1987), find negative prospects more available or salient. Concerning the previous findings, of great interest are the findings of Clore et al. (1994) and Forgas (1995), who state that mood mainly affects decisions and judgements for which people lack concrete information. This argument can be the case in investors, who most of the time lack or misinterpret information. In addition, Loewenstein et al. (2001), state that decisions under the influence of feelings are not rational ones.

Other studies about the effects of mood on human behaviour and decision making, find that good mood leads people to less critical information processing, both in weak and strong arguments, while bad mood has exactly the opposite effect (Schwarz (1990), Petty et al. (1991), and Sinclair and Mark (1995)).

A very important concept, which is also relevant to my study, is risk and how this can be influenced by mood. However, as Isen (2000) states, this assessment is rather complex and totally depends on the task and situation. Nonetheless, Isen (2000) proves that being in a good mood can have many benefits for the individual. For example, a person in a good mood generates more unusual associations, performs better than average in problem-solving and generally has greater mental flexibility.

Another important element of both emotions and mood, a more superficial one, as stated by Frijda (1998) and Schwharz (1990), provides individuals with useful information about the environment. For instance, many people claim that they “feel good or bad vibes” or act according to the saying “do whatever your heart tells you”. In addition, considerable research is conducted on affect. One of the most important ones, and in the same direction with what has already been mentioned, is the “affect heuristic” introduced by Slovic et al. (2002)5.

1.2 Mood and Investor Decision Making

According to Nofsinger (2008), mood can, actually, influence investors’ financial decisions. He names this, the misattribution bias. People tend to misattribute their mood in their investment decisions. For instance, if someone is in a good mood, he or she is rather likely to be optimistic about a particular investment, and exactly the opposite may occur if the individual is in a bad mood. In addition, Nofsinger claims that optimism leads investors to less critical analysis and often makes them ignore negative information.

In this section I analyze what impact major mood factors, such as weather, lunar circle and football results have on human mood, and therefore, their impact on stock markets. These factors are chosen because their study is relative new and not many authors have attempted to explore their influence on stock returns.

5

1.2.1 Weather, Mood

There is extensive research concerning the weather influences on human behaviour, in the field of psychology. Sunshine and cloudiness and their influence on human mood have drawn the attention of many researchers. In general, all of the practitioners agree that lack of sunlight leads to depression and bad mood, while when there is sunlight people feel good and are, generally, optimistic about the future. For example, Rind (1996) claims that sunshine is linked to optimism and, on the other hand, that lack of sunshine leads to depression and even suicide (Eagles(1994)) and Titjen and Kripke (1994)). According to Schwarz and Clore (1983), on sunny days people tend to rate their life satisfaction higher than they do on cloudy and rainy days, although their well-being doesn’t change on a day-by-day basis, as weather does. In addition, Rind (1996), besides the findings mentioned before, in the same research did an experiment very important for my study. He actually proved that human beings get influenced by sunshine, even when they don’t get exposed to it. His experiment takes place in a large hotel, where individuals’ mood, whose rooms didn’t have any windows, is tested. He measures the mood of these people by measuring the tips that they gave to the server of their breakfast, who always mentioned to them the weather outside. The results are astonishing. His tip was on average 18.8%6 on rainy days, 24.4% on cloudy days, 26.4% on partially sunny days and 29.4% on sunny days. Finally, sunlight from another point of view has been examined by Kamstra et al. (2003). In their research, they state that seasonal affective disorder (SAD)7, causes distress to individuals.

Concerning humidity, Howarth and Hoffman (1984) in their study find that high humidity levels are negatively correlated to positive human behaviour.

Another weather factor widely tested is temperature. Howarth and Hoffman (1984), find that under cold temperature8, people tend to be more aggressive. Other studies, by Allen and Fisher (1978) and Wyndham (1969), find that task-performing activities are impaired when people are exposed to extremely high and low temperatures. In addition, Pilcher et al. (2002), find that high temperatures cause to

6

Percentage on the bill.

7

It concerns the period of the year that the daylight lasts less than the rest of the year.

8

people hysteria and apathy. Aggression could result in more risk-taking, while apathy could impede risk-taking.

1.2.2 Weather, Stock Market

There is a significant number of authors that have proven that weather conditions (mainly sunshine/cloudiness) can actually have an impact on stock returns. Saunders (1993), in the most influential paper in the weather effect literature, examines the weather impact on the New York Stock Exchange (for the periods 1927-1962 and 1927-1962-1989), and finds that when the weather is cloudy most indices give negative returns, on average. To be more specific, he proves that when cloud cover is more that 85% (almost full cloudiness), returns are significantly below average, while cloud cover is less than 20% (almost full sunshine), returns are found significantly above average.

Hirshleifer and Shumway (2003), test for sunshine, precipitation and snowiness effects in 26 stock markets around the world (for the period 1982-1997), among which is the Amsterdam stock exchange. They find negative relation between returns and high cloud coverage in 25 out of 26 stock markets. However, only 9 are statistically significant, among which is not Amsterdam. They actually find that cloud cover and rain have a negative but insignificant effect on returns. Finally, precipitation and snowiness are not significant for any stock market. Contrary to the studies of Saunders (1993) and Hirshleifer and Shumway (2003), Loughran and Schultz (2004) give a different but very interesting approach, when checking for weather effects on 25 stocks listed on Nasdaq. Because of the fact that returns are not only influenced by investors located in New York, which is the city where the stock market is positioned, they also take under consideration the weather in the cities that companies listed on Nasdaq, are located9. By doing so, they try to include as many investors around the USA as possible. Their belief that people tend to invest more money to local firms has to do with the familiarity bias, as this is illustrated by many authors, among which are Coval and Moskowitz (1999). They find that the weather in the area near each company’s headquarters has no significant effect on the stock

9

returns, but they do find negative relationship between the cloudiness in New York and stock returns.

In another effort to replicate Saunders’ (1993) findings, Kramer and Runde (1997), check for weather effects in the Frankfurt stock exchange, for the period 1960-1990. Specifically, they check for cloud cover, humidity and rain effects, but they claim that results are totally dependent on the way that the null hypothesis is phrased and that generally no relationship between weather and stock returns exists. In addition, Goetzmann and Zhu (2005), give another very interesting aspect of weather effects. When checking for cloud cover effects for the NYSE, they come to the conclusion that market makers’ and not individual investors’ behaviour is responsible for the relation between returns and weather.

Nonetheless, there are researchers that test smaller markets that the US one. For instance, Dowling and Lucey (2005) test for weather effects on the Irish stock market for the period 1988-2001. They state a statistically negative relationship between stock market returns, cloudiness and rain. In another paper, for the weather effects (sunshine and humidity) in the Spanish stock exchange for the period 1981-2000, Pardo and Valor (2003), in an effort to check for weather effects in an open outcry trading system and a screen trading environment, they find no effects in any of the two systems.

Cao and Wei (2005) in their study, test for temperature effects in eight major stock markets10 around the world (for the period 1962-2001). They find an overall negative correlation between temperature and stock markets returns.

Finally, Jacobsen and Marquering (2008) present the view that there is no need for daily weather data so as to prove a correlation between weather and stock market returns, but only dummy variables for summer and winter. What they actually do is an effort to “defend” the “Sell in May”11 effect, illustrated by Jacobsen and Bouman (2002). In order to prove their saying, they test the paper by Kamstra et al. (2003), which is about the SAD effect, and the Cao and Wei (2005) that tests for temperature effects. Their view is that the use of dummy variables for winter/summer makes the variables SAD and temperature redundant.

10

These are: US, Canada, Britain, Germany, Sweden, Australia, Japan and Taiwan.

11

1.2.3 Lunar Phases, Mood

Lunar circle and the impact that this has on humans, has long being questioned by people, even from the ancient times. Moon and its circles around the earth have been said of having various effects on humans and their lives. For instance, there are beliefs that link the moon cycle with events such as epilepsy, crimes, suicides, disasters, fertility, all of which are believed to take place around the full moon. Another evidence, is the existence of the word “lunacy” that is used to describe mental disorders. Finally, many religious ceremonies, of various religions, are linked with the moon cycle. For example, Easter and Passover, which are ceremonies that don’t have a specific date, are placed every year according to a specific full moon.

To prove some of the above, there are scientific researches that give evidence about the relation between moon and humans. Law (1986) finds a relationship between the women menstruation and lunar circle. The effect on fertility is researched by Criss and Marcum (1981), who state that births vary during the lunar circle, with the greatest numbers occurring at the third lunar quarter. In addition, Liber (1978), reports a significantly high number of crimes that take place around the full moon. As far as the human mood is concerned, Nofsinger (2008) states the general belief of psychologists, that full moon causes depression.

However, there are studies that doubt the effect of moon phases on human behaviour. Rotton and Kelley (1985), examine 37 studies and come to the conclusion that lunar phases have no influence on humans. Furthermore, Kelley, Rotton and Culver (1996), in a continuation of their previous research, claim that existing studies about moon are unreliable.

1.2.4 Lunar Phases, Stock Market

Quite impressive findings are provided by Dichev and Janes (2003). They examine all major US indices for 100 years and 24 international ones for 30 years. They find that in the 15 days around new moon, returns are twice as much than the 15 days around full moon. For the international stock exchanges the returns around the new moon are even more than double. For the Netherlands they find an annualized return of approximately 5% for the days around the full moon and 15% for the days around the new moon.

Nonetheless, Herbst (2007) tests for lunar effects on the Dow Jones Industrial Average (for the period 1980-2004) and finds no significant results. Finally, Dowling and Lucey (2005) investigate moon effects on the Irish stock exchange (for the period 1988-2001). They find a significant negative effect of new moon, but the most interesting is that they also combine cloud cover with full moon. In this attempt they find negative returns during full moon and low cloud cover, and neutral returns for full moon on days with high cloud cover. The latter research has to do with the belief that moon light is the one that has the effect on humans, and not the moon’s appearance by itself.

1.2.5 Football, Mood

The greatest difference of football results with weather and lunar circles is that the latter ones can be predicted accurately (in a great extend) beforehand, while football results can only be forecasted. This is what makes the studying of this specific mood factor, quite interesting. In addition, the popularity of this sport around the world and in individual countries makes sure that the outcomes of important games influence a great amount of the population of these countries.

Schwarz et al. (1987) document another very important finding. They claim that the outcomes of just two games of the national football team of Germany, in the World Cup of 1992, were enough for German subjects’ to change the view of their own well-being and their view about national issues.

In general, the work of Loewenstein (2000), who introduces the visceral factors12, gives the fundamental reason for specific actions of individuals. When individuals are under the influence of these factors, in our case the fans after a loss of their team, their actions and, therefore, their decision making, are seriously altered and they act in a totally irrational way. The author even claims that visceral factors are more basic to daily functioning than the higher level cognitive processes that are believed to underlie decision making.

1.2.6 Football, Stock Market

Although football is so popular and it is said to have such a great impact, as mentioned before, on fans’ mood, there is only one study that has examined football’s effect on stock exchanges. Edmans et al. (2007) test 39 national teams’ results around the world, for the period 1973-2004. The day following a football loss the stock returns are found to decline, on average, by 0.21%. It is worth mentioning that the decline for World Cup losses is 0.49%. Another useful finding is that they identify no positive effects after wins. The reasons behind this, as they claim, besides the one given before (reference point-win) by Nofsinger (2008), is that a win means that the team continues, while a loss means that the team is eliminated. It is obvious that a loss looms much more than a win13. Finally, they state that for the other sports tested, there is a decline for losses of about 0.19% and 0.21% for cricket and basketball respectively.

In this paper I test the hypotheses that cloudy weather, high humidity, high temperatures, days around the full moon and Dutch national football team’s loses, separately and in combination, correlate negatively to the returns of the Dutch stock exchange. The remainder of the paper is organized as follows. In section 2, I introduce the datasets. In section 3, the methodology is discussed and the construction

12

They refer to a wide range of negative emotions such as anger and fear, drive states such as hunger, thirst and sexual desire and feeling states such as pain and loss.

13

of the variables used is explained. I document in section 4 the results and section 5 concludes.

2. Data

Since I want to check for correlation between stock market returns, weather, lunar circles and football results, I need data for all of them.

To determine stock market returns, I select the Morgan Stanley Capital International (MSCI)14 Netherlands index, the MSCI Netherlands Small-Cap and the MSCI Europe excluding Netherlands (as a control variable). The returns are collected from DataStream, and they cover the period from the 2nd of January 1998 to the 31st of December 2007, which includes 2,553 daily observations (for additional information, check summary statistics at Table I, Table II for correlation matrix and also Figures I, II & III of the Appendixes).

I have to stress the reason for checking for mood effects on the MSCI Netherlands Small-Cap index. The large-cap index includes big firms; therefore, I have to test possible effects of size or ownership, which mainly concern the location of investors. If there are effects, then the weather data from Amsterdam are not enough, since stock holders may be in other parts of the world where the weather is totally different than the Dutch one. Another important issue is that many of the firms included in the MSCI Netherlands index, are also listed in other stock markets around the world. Therefore, it may be the case that specific stock movements are either due to correlation among stocks listed in different countries or because of efforts for arbitrage between markets. Hence, my research becomes more accurate if I check for mood effects on small cap stocks, whose ownership is mainly Dutch and are, most of the times, traded only in the Dutch stock market.

Furthermore, I use the MSCI Europe excluding The Netherlands to adjust for European stock markets’ fluctuations and, therefore, to make sure that possible findings are only due to weather, lunar, football effects.

Based on price levels at closing, I compute the daily return for day t by using the following formula:

14

(Price levelt – Price levelt-1)/Price levelt-1.

Weather data are obtained from the online database15 of the Dutch Meteorological Institute (Koninklijk Nederlands Meteorologisch Instituut). The weather observations are daily and are taken from the meteorological station that is located in the Schipol Airport of Amsterdam that is the nearest observation point to the Dutch stock market.

Lunar data are taken from the website of NASA16. The website gives the full moon, new moon dates for the period I want to test (for further general information about the moon phases check Panel A on the Appendixes). I then use the formula provided by Yuan et al. (2006), so as to give a value to each day, based on how close it is to a full moon:

Cosine (2πd/29.53)

Where d is the number of days since the previous full moon and 29.53 is the number of days from a full moon to the next one, for the year 2000.

Table I: Summary Statistics

I provide information, such as mean, minimum, maximum, standard deviation and total prices, for the period tested, on MSCI Netherlands, MSCI Neth.Small-Cap MSCI Europe excluding Netherlands, Cloud cover, humidity, temperature, cloudlee2, cloudmore7, hum38-70, hum90-100, good weather, bad weather, lunfull6,

lunnew6, wins and losses.

Mean Minimum Maximum Std.Dev. Total

Return MSCINeth. 0,0002 -0,074 0,081 0,014 18,76%

Return MSCI Small-C 0,0002 -0,064 0,045 0.011 30,98%

Return MSCIexNeth. 0,0003 -0,062 0,056 0,012 58,77% Cloud Cover 5,32 0 9 0,022 -- Humidity 80,67% 38% 100% 0,092 -- Temperature 11,01 -6,1 26,7 0,059 -- Cloudless2 -- -- -- -- 343 Cloudmore7 -- -- -- -- 935 Hum38-70 -- -- -- -- 280 Hum90-100 -- -- -- -- 520 GoodWeather -- -- -- -- 337 Badweather -- -- -- -- 349 Lunfull6 -- -- -- -- 601 Lunnew6 -- -- -- -- 608 Wins -- -- -- -- 11 Losses -- -- -- -- 5

Table II: Correlation Matrix

I provide correlations between the variables: MSCI Neth-Large & Small-Cap, MSCI Europe Excluding The Neth., Temperature, Cloudiness, Humidity and Lunar.

MSCI-Large MSCI-Small MSCI

EurExNeth Temperature Cloud Humidity Lunar MSCI-Large 1 0,676 0,893 -0,023 -0,022 -0,009 -0,012 MSCI-Small 0,676 1 0,672 -0,057 -0,014 0,012 -0,018 MSCI EurExNeth 0,893 0,672 1 -0,036 -0,016 0,002 -0,006 Temperature -0,023 -0,057 -0,036 1 -0,181 -0,389 0,034 Cloud -0,022 -0,014 -0,016 -0,181 1 0,580 -0,014 Humidity -0,009 0,012 0,002 -0,389 0,580 1 -0,065 Lunar -0,012 -0,018 -0,006 0,034 -0,014 -0,065 1 3. Methodology

Netherlands), for both indices. In my first attempt, they are tested in the form they appear in the text below; in the second I add the control variable. I use this variable to remove effects of European economic changes from the Dutch index movements.

The null and alternative hypotheses, for all of the following regressions, are:

H0: There are no weather, lunar, sports effects on the returns of the Dutch Stock

Exchange.

H1: There are weather, lunar, sports effects on the returns of the Dutch Stock

Exchange.

3.1 Weather

Firstly, I check for a simple linear relationship between cloud cover and Dutch equity returns. The regression for the cloud/sunshine effects, similar to those presented by Downing and Lucey (2005) is:

r

=a

+a

CLOUD

+e

(1)

Where,

rt is the return at time t, a1 the constant, a2 is the coefficient of CLOUD, CLOUD is a

variable that varies between 0 (no cloud cover) and 9 (100% cloud cover) octants and et the error term.

Another regression for the cloudy weather, so as to check for extreme value effects of cloudiness on stock returns, is:

rt= a

+a

D

CLOUDMORE7+a

D

CLOUDLESS2+e

(2)

Where,

rt is the return at time t, a2 the coefficient of DtCLOUDMORE7, CLOUDMORE7 is a

dummy variable that equals 1 if cloud cover is more than 717, a3 the coefficient of

17

DtCLOUDLESS2, DtCLOUDLESS2 is a dummy variable that equals 1 if cloud cover

is less than 218 and et the error term.

To test for humidity effects on the stock exchange, I use the following regression:

r

=a

+a

HUMID

+e

(3)

Where,

rt is the return at time t, a1 is the constant, a2 the coefficient of HUMIDt, HUMIDt

value based on percentage humidity for day t, et the error term.

To make it more specific, I put dummy variables, representing high and low humidity. The average humidity in this decade of data is found to be, approximately, 80%. Therefore, I choose to check for humidity levels that are at least 10% away from the average (either higher or lower). The regression is:

r

=a

+a

D

HUMOVER90+a

D

HUM38–70+e

(4)

Where,

rt is the return at day t, a1 the constant, a3 the coefficient of DtHUMOVER90, Dt

HUMOVER90 is a dummy variable that equals 1 if relative humidity is over 90%, a3

is the coefficient of Dt HUM38–70, Dt HUM38–70 is a dummy variable that equals 1

if relative humidity is between 38% and 70%, et is the error term.

For the effects of temperature, the regression used is the one introduced by Cao and Wei (2005):

R

= a

+a

TEMPT

+

ε

(5)

Where,

Rt is the daily return, a1 is the constant, a2 is the coefficient of temperature, TEMPTt is

the daily temperature (in degrees Celsius) at time t and εt is the error term.

18

Furthermore, Downing and Lucey (2005), introduce the good/bad weather concept. This quite interesting concept is regressed as follows:

r

=a

+a

D

GOODWEATHER+a

D

BADWEATHER+e

(6)

Where,

rt is the daily return at time t, a1 the constant, a2 the coefficient of

DtGOODWEATHER, DtGOODWEATHER is a dummy variable that equals 1 if cloud

cover is less than 2 and humidity is not above 90%, α3 is the coefficient of

DtBADWEATHER, DtBADWEATHER is a dummy variable that equals 1 if cloud

cover is more than 7 and humidity is above 90%, and et the error term.

3.2 Lunar Circle

I use three regressions to check for lunar effects on the Dutch stock exchange. The regression for measuring lunar effects provided by Downing and Lucey (2005) is:

r

=a

+a

LUNAR

+e

(7)

Where,

rt is the return on day t, a1 the constant, a2 the coefficient of LUNARt, LUNARt is the

variable for lunar phases, whose price ranges from 1 (full moon) to -1(new moon), and et the error term.

To compare for the returns around full and new moon days, I take the following regression that actually focuses on these days:

19

As mentioned in the data section: LUNAR is a variable that varies between 1 (full moon) and -1 (new moon). Each day has a value based on how close it is to a full moon. The formula used to calculate this is:

r

=a

+a

D

LUNFULL6+a

D

LUNNEW6+e

(8)

Where,

rt is the return at day t, a1 is the constant, a2 the coefficient of DtLUNFULL6,

DtLUNFULL6 is a dummy variable that takes a value of 1 if the day is either a full

moon day or within three calendar days (before and after) of a full moon, a3 the

coefficient of Dt LUNNEW6, DtLUNNEW6 is a dummy variable that takes a value of

1 if the day is either a new moon day or within 3 days (before and after) of a new moon and et is the error term.

To combine full moon and cloudy days, so as to check for moon light effects, I use the regression:

r

LUNFULL6=a

+a

DtCLOUDLESS2+a

DtCLOUDMORE7+e

(9)

Where,

rtLUNFULL6 is the return on 6 days around a full moon, α1 the constant,

CLOUDLESS2 is a dummy variable that equals 1 if cloud cover is between 0 and 2, representing low cloud cover. CLOUDMORE7 is a dummy variable that equals 1 if cloud cover is between 7 and 9, representing high cloud cover, and et is the error term.

3.3 Football

The regression proposed by Edmans et al. (2007)20 is:

r

=

β

+

β

W

+

β

L

+

β

MSCI+u

(10)

Where,

rt is the daily return at time t, β0 is the constant, βw is the coefficient of wins, Wt is the

dummy variable that takes the price of 1 for wins, βl the coefficient of losses, Lt the

dummy variable that takes the price of 1 in case of losses, βE is the coefficient of

20

MSCI Europe without The Netherlands, MSCI is the variable for the returns of MSCI Europe without The Netherlands and ut is the error term.

It worth’s mentioning that all equal results are excluded from my research21.

3.4 Joint Tests

In order to identify the power of each mood factor in the presence of all the others, I run two more regressions:

r

=a

+a

CLOUD

+a

HUMIDt+a

TEMPT

+a

LUNAR

+e

(11)

Where,

rt is the return on day t, a1 is the constant, a2 is coefficient of CLOUDt, CLOUDt the

variable of percentage of clouds, a3 the coefficient of HUMIDt, HUMIDt the variable

of the percentage of humidity, a4 the coefficient of TEMPTt, TEMPTt the variable of

temperature, a5 the coefficient of LUNARt, LUNARt the variable of lunar phase and et

the error term.

My second overall regression focuses on some of the variables mentioned up to now:

rt=a1+a2DtGOODWEATH+a3DtBADWEATH+a4LUNARt+ a5Wt+a6Lt +et (12)

Where,

rt is the return at time t, a1 is the constant, a2 is the coefficient of DtGOODWEATH,

DtGOODWEATH is the dummy variable that takes the price of 1 in a day with good

weather, a3 the coefficient of DtBADWEATH, DtBADWEATH the dummy variable

that has the price 1 in a day with bad weather, a4 the coefficient of LUNARt, LUNARt

is the variable of lunar phase, α5 the coefficient of the Wt, Wt the dummy variable for

wins, a6 the coefficient of Lt, Lt is the dummy variable for losses and et is the error

term.

21

4. Results

In this section the statistical findings of my research are presented. As mentioned in the methodology part, every of the 12 regressions (except for the 10th)22, is tested with and without the control variable, for each of the two indices.

For all regressions I use the OLS (Ordinary Least Squares) method. Then, I check all datasets for normality, by using the test introduced by Jarque-Bera (hereafter J-B) (1981). All regressions reject the null hypothesis of normality, at the 5% level of significance (check Tables A, B and F, G in the appendixes). To continue, I test the regressions for heteroskedasticity23, by conducting the test illustrated by White (1980). From Tables B and G in the appendixes, it is obvious that the second set of regressions (those that include the MSCI ex. The Netherlands variable) appear to be heteroskedastic, since the null hypothesis of heteroskedasticity is rejected at the 5% significance level. The next step is to try to identify possible ARCH (autoregressive conditionally heteroskedastic) effects for these regressions (Engle (1982)). The results (check Table C and H in the appendixes) reject the null hypothesis of no ARCH effects24. Since I don’t know the source of heteroskedasticity, I estimate the regression by using the Newey-West estimator (Newey and West (1987)) (check Table J in the appendixes). After this I conduct again White and ARCH tests, but heteroskedasticity appears again (the tests give exactly the same results as before the Newey-West model). Therefore, it is necessary to transform my model to GARCH (Generalized-ARCH) (Engle (2001)). In the following subsections results of each category (weather, lunar, football results and joint tests) are further analyzed for each of the two indices.

4.1 Weather Results

I have data for three weather variables: cloud cover, humidity, temperature. With these data I check whether negative weather is associated with below average

22

Due to the fact that there are not that many observations and there is great chance of having biased results.

23

When the errors don’t have constant variance, they said to have heteroskedasticity.

24

equity returns and vice versa. To be more specific, I expect high cloud cover, high humidity and high temperature to cause lower returns than average.

4.1.1 Weather Results for the Large-Cap Index

In Table III the results of all weather regressions, are presented (check also Tables A & B of the Appendixes). Contrary to some of the existent literature, no weather effect is identified for the large-cap index. All weather variables tested and even their extreme values, are proved to be statistically insignificant in every level of significance. Even by adding the control variable, MSCI Europe excluding The Netherlands, no significant result appears. In addition, due to the fact that heteroskedasticity appears in my second sample (the one with the control variable, I turn the model to GARCH, but again no result is identified).

Concerning cloud cover, my results of no relation between cloud cover and stock returns are contrary to Saunders (1993) who finds negative cloud cover effects in NYSE. In addition, my findings are partly different than those of Hirshleifer and Shumway (2003) who present cloud effects in 9 out of the 26 stock markets that they test. However, they find no cloud effects on the Amsterdam stock exchange. Nonetheless, Loughran and Schlutz (2004), Goetzmann and Zhu (2005), Dowling and Lucey (2005), present different cloud results than the ones I find, for different markets and periods, though.

On the other hand, my findings are in accordance to those given by Pardo and Valor (2003) and Kramer and Runde (1997), for the Spanish and German stock exchange, respectively. Especially those for the German stock exchange may be of great importance to my study, since Germany is geographically close to The Netherlands and one can consider more or less similar weather conditions.

As far as humidity results are concerned, I find no humidity effects on the Dutch equity returns. My results are in accordance to those provided by Kramer and Runde (1997), who find no humidity effects in the German stock exchange. However, they are different than those by Dowling and Lucey (2005), who state humidity effects on the Irish equity returns.

(2005), who actually identify temperature effects on eight major stock markets around the world.

4.1.2 Weather Results for the Small-Cap Index

In Table III the weather results for the small-cap index are presented (check also Tables F & G in the Appendixes). Contrary to my previous findings for the large-cap index, a temperature effect appears in the small-large-cap index. Specifically, there is a negative correlation between temperature and small-cap equity returns. This finding is in accordance to both psychological and economic researches. From a psychological point of view, lower temperature leads people to aggression, while higher temperature to apathy. There is a connection between aggression and risk-taking and at the same time apathy impedes risk-taking. Therefore, lower temperature causes higher stock returns and vice versa. From an economic perspective, the only temperature research is the one by Cao and Wei (2005) who find exactly the same results with the ones I provide.

Furthermore, when I transform my second model (the one with the control variable) to GARCH a negative relationship between small-cap returns and GOODWEATHER25 variable are identified. It is interesting that I find significant results when I combine humidity with cloud cover and not when I test each variable independently. It appears that the combination of the two has, actually, effects on Dutch investors. On the other hand, my result is contrary to the ones that psychologist’s state. Low cloud cover and humidity are indicators of good mood and, therefore, of positive returns. It may be the case though, that low cloud cover and humidity result to high temperature, for which the negative relation with the equity returns has been explained before. In addition, a more convincing explanation is that humidity has effects on humans only under extreme temperature, which is not the case for The Netherlands. This may be the case for Southern countries, such as Spain, but Pardo and Valor (2003) in their research for humidity effects in the Spanish stock exchange, give no findings.

Finally, I can assume that the no-findings in the large-cap index are partially due to factors of size and ownership of firms included in this index.

25

Table III: Weather Results

MSCI Large-Cap

MSCI

Small-Cap

Regr Variable OLS

OLS with

Control GARCH OLS

BADWEATH. 0,001 0,0003 0,00008 0,00001 -0,0002 0,0001 (0,297) (0,389) (0,753) (0,857) (0,639) (0,662)

MSCIexNeth -- 1,085 1,046 -- 0,619 0,642

-- (0,000) (0,000) -- (0,000) (0,000)

P-values in parentheses.

*Statistically significant at the 10% level of significance. **Statistically significant at the 5% level of significance.

Note: Columns from left to right present: regression, variable, OLS, OLS with control variable, GARCH, for large and small-cap indices, respectively.

4.2 Lunar Phases

I use regressions 7-9 so as to check for moon effects on the Dutch equity returns. Again I test the regressions twice (with and without the control variable), for both indices. I use the variables: LUNAR that takes a specific value according to the distance of each day to the previous full moon (as described in the data and methodology sections), LUNFULL6 and LUNNEW6 which are dummy variables for the six days around the full and new moon, respectively. Finally, I check for the belief that moon light has the effect on humans and not the mere existence of the moon, by combining cloud cover and returns on full moon days.

I expect to find negative returns during full moon and positive ones in the days of the new moon and also positive returns on the days of full moon with high cloud cover.

4.2.1 Lunar Phases Results for Large-cap Index

Results of the three regressions are presented in Table IV (for additional information check Table A and B in the Appendixes). A small but significant negative result (0,05%) is found around the six days of the full moon, but only after adjusting for European stock markets’ moves.

find significant negative returns in the 15 days around the full moon in US and international indices.

My results are contrary to those by Herbst (2007), who states no significant findings in his research for lunar effects on the Dow Jones and those by Dowling and Lucey (2005) that find negative returns around the new moon days in the Irish stock exchange.

The most striking is that I find significant results after adjusting for European economic changes. A possible explanation could be that Dutch are more influenced by the full moon than the majority of the European investors. However, there is no research supporting this kind of an argument; therefore, I can only assume that, plainly, there is no explanation of this phenomenon.

4.2.2 Lunar Phases Results for the Small-Cap Index

In Table IV the results of lunar circle in the small-cap index are presented (additional information is provided in Tables F & G of the Appendixes). The same negative effect of full moon is identified, as in the large-cap sample. The fact that it is exactly the same as before is not surprising. Lunar circles take place at the same dates around the world, which is not the case for the weather that differs from one country to the other.

Table IV: Lunar Results

MSCI Large-Cap

MSCI Small-Cap

Regr Variable OLS

OLS with

Control GARCH OLS

OLS with Control GARCH 7 Constant 0,0002 -0,0001 -0,0007 0,0002 0,00005 0,00004 (0,540) (0,360) (0,450) (0,444) (0,973) (0,079) LUNAR -0,0002 -0,0001 -0,0002 -0,0003 -0,0002 -0,0002 (0,590) (0,360) (0,220) (0,377) (0,349) (0,231) MSCIexNeth -- 1,085 1,045 -- 0,619 0,643 -- (0,000) (0,000) -- (0,000) (0,000) 8 Constant 0,0003 0,00001 0,00009 0,0003 0,00002 0,0005 (0,430) (0,569) (0,432) (0,303) (0,363) (0,003) LUNFULL6 -0,001 -0,0006 -0,0005 -0,0007 -0,0007 -0,0005 (0,335) (0,051)* (0,032)** (0,157) (0,063)* (0,096)* LUNNEW6 0,0001 -0,0003 -0,0002 0,0002 -0,00008 -0,0002 (0,875) (0,344) (0,3490 (0,762) (0,827) (0,515) MSCIexNeth. -- 1,085 1,045 -- 0,619 0,643 -- (0,000) (0,000) -- (0,000) (0,000) 9 Constant -0,00001 -0,0005 -0,0003 0,00006 -0,0002 -0,0002 (0,988) (0,171) (0,299) (0,917) (0,609) (0,575) CLOUDLESS2 -0,0004 0,0008 0,0002 -0,001 -0,0007 -0,0009 (0,807) (0,296) (0,681) (0,243) (0,462) (0,258) CLOUDMORE7 -0,001 -0,0004 -0,0003 -0,0007 -0,0005 -0,0002 (0,507) (0,512) (0,681) (0,423) (0,490) (0,771) MSCIexNeth -- 1,095 1,06 -- 0,62 0,603 -- (0,000) (0,000) -- (0,000) (0,000) P-values in parentheses.

*Statistically significant at the 10% level of significance ** Statistically significant at the 5% level of significance

Note: Columns from left to right present: regression, variable, OLS, OLS with control variable, GARCH, for large and small-cap indices, respectively

4.3 Football Results

In this subsection the Dutch football teams’ effects on the Dutch Stock Market are given. It worth’s mentioning that this regression is the only one tested once (with the existence of the control variable), for each of the two indices, due to the small amount of football results (check Table I) during the 10-year period checked.

4.3.1 Football Results for the Large-Cap Index

Table V (and also Table B of the Appendixes) gives the results of the 10th regression. The results are surprising and contrary to the only paper published on football results, by Edmans et al. (2007). I find positive returns, of about 0,2%, after losses of the Dutch football team.

A possible explanation is that losses are most of the time followed by the termination of the Dutch team from the World or European championship that takes place. A rational explanation might be that losses lead Dutch investors “back to work”, since there is no football to watch anymore. However, the small sample of losses (5 observations) may be the reason for my finding.

4.3.2 Football Results for Small-Cap Index

In Table V the results for the small-cap index are presented (check also Table F in the Appendixes). As opposed to the large-cap findings, no statistically significant effect is identified in the small-cap index. It appears that small-cap investors are rather indifferent for the Dutch teams, results or that the latter do not affect their investing behaviour. Table V: Football Results MSCI Large-Cap MSCI Small-Cap

Regr Variable OLS

OLS with

Control GARCH OLS

OLS with Control GARCH 10 Constant -- -0,0001 -0,00009 -- 0,00001 0,0003 -- (0,311) (0,325) -- (0,937) (0,011) Win -- 0,002 0,002 -- 0,0009 0,003 -- (0,176) (0,437) -- (0,573) (0,217) Loss -- -0,0003 0,002 -- -0,003 -0,002 -- (0,864) (0,025)** -- (0,175) (0,579) MSCIexNeth -- 1,085 1,046 -- 0,619 0,642 -- (0,000) (0,000) -- (0,000) (0,000) P-values in parentheses.

Note: Columns from left to right present: regression, variable, OLS, OLS with control variable, GARCH, for large and small-cap indices, respectively.

4.4 Joint Tests

The results of the two joint tests are presented in this subsection. I conduct these tests so as to check for the strength of mood proxy variables in the presence of all the others. In the first regression I test all the weather variables with the LUNAR variable. In the second I use the GOOD-BADWEATHER concepts, as representative of weather, LUNAR and Win-Loss variables. Again the regressions are run twice (with and without the control variable), for both indices.

4.4.1 Joint Tests for the Large-Cap Index

Table VI gives the results of the joint tests. For the large-cap index, as expected, the only significant results is the positive returns after losses of the Dutch football team, exactly of the same magnitude, as before (0,2%).

4.4.2 Joint Tests for the Small-Cap Index

Table VI: Joint Tests Results

MSCI Large-Cap MSCI Small-Cap

Regr Variable OLS

OLS with

Control GARCH OLS

OLS with Control GARCH 11 Constant 0,003 0,0006 0,001 0,002 0,0006 0,002 (0,423) (0,645) (0,334) (0,465) (0,732) (0,141) CLOUD -0,0001 -0,00001 -0,00001 -0,0001 -0,00006 -0,00007 (0,352) (0,872) (0,982) (0,263) (0,495) (0,279) HUMID -0,00009 -0,00001 -0,00001 0,00004 0,00004 -0,00001 (0,806) (0,562) (0,332) (0,892) (0,831) (0,384) TEMPT -0,00006 0,00001 -0,00003 -0,0001 -0,00006 -0,00007 (0,179) (0,518) (0,846) (0,005)** (0,036)** (0,002)** LUNAR -0,00002 -0,0001 -0,0001 -0,0002 -0,0002 -0,0002 (0,599) (0,413) (0,198) (0,431) (0,394) (0,226) MSCIexNeth -- 1,085 1,046 -- 0,618 0,642 -- (0,000) (0,000) -- (0,000) (0,000) 12 Constant -0,00002 -0,0002 -0,0001 0,00008 0,0001 0,0004 (0,936) (0,209) (0,356) (0,746) (0,517) (0,008) GOODWEATH. 0,0005 0,00008 -0,00001 0,0002 -0,0006 -0,0007 (0,549) (0,814) (0,947) (0,726) (0,204) (0,074)* BADWEATH. 0,0008 0,0003 0,00009 0,0003 -0,0002 0,0002 (0,342) (0,377) (0,722) (0,566) (0,641) (0,628) Win 0,001 0,002 0,002 0,0007 0,001 0,003 (0,726) (0,172) (0,431) (0,757) (0,566) (0,202) Loss 0,0008 -0,0003 0,002* -0,001 -0,003 -0,001 (0,842) (0,879) (0,022) (0,649) (0,173) (0,599) LUNAR -0,0002 -0,0001 -0,0002 -0,0002 -0,0002 -0,0002 (0,529) (0,441) (0,199) (0,368) (0,379) (0,254) MSCIexNeth -- 1,085 1,046 -- 0,618 0,642 -- (0,000) (0,000) -- (0,000) (0,000) P-values in parentheses.

*Statistically significant at the 10% level of significance. ** Statistically significant at the 5% level of significance.

Note: Columns from left to right present: regression, variable, OLS, OLS with control variable, GARCH, for large and small-cap indices, respectively.

5. Conclusions

Dutch equity returns, for the period 1998-2008. I find no weather effects on the large-cap index, but negative temperature and good weather effects on the small-large-cap index. As far as moon phases are concerned, negative returns, of about 0,05%, around the six days of the full moon (three before and three after the full moon) are identified, only after adding the control variable, for both of the indices. I also find another surprising result which is positive returns (of 0,2%) after losses of the Dutch National football team, in the large-cap index. It appears that the efficient market concept is in doubt and that if investors are aware of their mood they can actually benefit. However, the transactions costs should be seriously considered, since trading on temperature and full moon days means quite frequent trading, while the returns provided are quite small.

The broadest message of my study is that if one wants to understand security price movements, may have to go beyond the fundamentals of pricing, and try to find what influences investor mood. In addition, knowing how individuals are influenced by their mood will help them to become aware, and even to overcome their biases on decision making, caused by mood fluctuations.

Finally, I have to stress out that my findings should be attended with care. I have only used data for one country (The Netherlands), two indices (MSCI The Netherlands Large and Small-cap) and for a period of one decade. It may be that the use of data from other countries, other indices or greater time periods, provide different results. This is left for future research.

6. References

Allen, M., Fisher, J., (1978): “Ambient Temperature Effects on Paired Associate Learning”, Ergonomics, 21, 95–101.

Bagozzi, R., Gopinath, M., and Nyer, P. (1999): “The Role of Emotions in Marketing”, Journal of the Academy of Marketing Science, 27, 184-206.

Bless, H., Schwarz, N., and Kemmelmeier, M. (1996): “Mood and Stereotyping: The Impact of Moods on the Use of General Knowledge Structures”, European Review of Social Psychology, 7, 63-93.

Bouman, S., Jacobsen, B., (2002). “The Halloween Indicator, Sell in May and Go Away.” The American Economic Review, 92, 1618-1635.

Cao, M., Wei, J., (2005): “Stock Market Returns: A Note on Temperature Anomaly”, Journal of Banking & Finance, 29, 1559–1573.

Clore, L., Schwarz, N., Conway, N., (1994): “Affective Causes and Consequences of Social Information Processing”, Handbook of Social Cognition, 2nd ed., Lawrence Erlbaum, Hillsdale, NJ.

Conlisk, J. (1996): “Why Bounded Rationality?”, Journal of Economic Literature, 34, 669–700.

Coval, J., Moskowitz, T., (1999): “Home Bias at Home: Local Equity Preference in Domestic Portfolios”, Journal of Finance, 54, 2045-2073.

Criss, B., Marcum, P., (1981): “A Lunar Effect on Fertility”, Social Biology, 28, 75– 80.

Dichev, I., Janes, T., (2003): “Lunar Cycle Effects in Stock Returns”, Journal of Private Equity, 6, 8 –29.

Dowling, M., Lucey, B., (2005): “Weather, Biorhythms, Beliefs and Stock Returns-Some Preliminary Irish Evidence”, International Review of Financial Analysis, 14, 337-355.

Eagles, M., (1994): “The Relationship Between Mood and Daily Hours of Sunlight in Rapid Cycling Bipolar Illness”, Biological Psychiatry, 36, 422-424.

Edmans, A., Garcia, D., and Oyvind, N., (2007): “Sports Sentiment and Stock Returns”, Journal of Finance, 62, 1967-1998.

Engle R., (1982): “Autoregressive Conditional Heteroskedasticity with Estimates of the Variance of United Kingdom Inflation”, Econometrica, 50, 987-1007.

Engle R., (2001): “GARCH 101: The Use of ARCH/GARCH Models in Applied Econometrics”, The Journal of Economic Perspectives, 15, 157-168.

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

Forgas, P., (1995): “Mood and Judgment: The Affect Infusion Model (AIM)”, Psychological Bulletin, 117, 39-66.

Forgas, P., Bower, G., (1987): “Mood Effects on Person-Perception Judgments”, Journal of Personality and social Psychology, 20, 497-513.

Frijda, H., (1988): “The Laws of Emotion”, Cognition and Emotion, 1, 235-258.

Goetzmann, W., Zhu, N., (2005): “Rain or Shine: Where is the Weather Effect?”, European Financial Management, 11, 559-578.

Hanoch, Y. (2002): “Neither an Angel Nor an Ant: Emotion as an Aid to Bounded Rationality”, Journal of Economic Psychology, 23, 1 –25.

Harlow, V., Brown C., (1990): “Understanding and Assessing Financial Risk Tolerance: A Biological Perspective”, Financial Analysts Journal, 46, 50-62.

Hirshleifer, D., Shumway, T., (2003): “Good Day Sunshine: Stock Returns and the Weather”, Journal of Finance, 58, 1009–1032.

Howarth, E., Hoffman, S., (1984): “A Multidimensional Approach to the Relationship Between Mood and Weather”, British Journal of Psychology, 75, 15–23.

Isen, A., (2000): “Positive Affect and Decision Making”, Handbook of Emotion, Guilford, New York.

Isen, A., Shalker, T., Clark, M., and Karp, L., (1978): “Affect, Accessibility of Material in Memory and Behavior: A Cognitive Loop?”, Journal of Personality and Social Psychology, 36, 1-12.

Jacobsen, B., Marquering, W., (2008): “Is it the Weather?”, Journal of Banking & Finance, 32, 526-540.

Johnson, J., Tversky, A. (1983): “Affect, Generalization, and the Perception of Risk”, Journal of Personality and Social Psychology, 45, 20– 31.

Kamstra, J., Kramer, A., and Levi, D., (2003): “Winter Blues: A SAD Stock Market Cycle”, American Economic Review, 93 (1), 324–333.

Kaufman, E. (1999): “Emotional Arousal as a Source of Bounded Rationality”, Journal of Economic Behavior and Organization, 38, 135– 144.

Kelly, I., James Rotton, J., and Culver, R., (1996): “The Moon Was Full and Nothing Happened: A Review of Studies on the Moon and Human Behaviour and Human Belief”, The Outer Edge, Committee for the Scientific Investigation of Claims of the Paranormal (CSICOP), Amherst, NY.

Law, P., (1986): “The Regulation of Menstrual Cycle and its Relationship to the Moon”, Acta Obstetricia et Gynecologica of Scandinavica, 65, 45– 48.

Liber, A., (1978): “Human Aggression and Lunar Synodic Cycle”, Journal of Clinical Psychiatry, 39, 385.

Loewenstein, G., Weber, U., Hsee, K., and Welch, N. (2001):“Risk as Feelings”, Psychological Bulletin, 127, 267– 286.

Loewenstein, G., (2000): “Emotions in Economic Theory and Economic Behavior”, The American Economic Review, 90, 426-432.

Loughran, T., Schultz, P., (2004): “Weather, Stock Returns, and the Impact of Localized Trading Behavior”, Journal of Financial and Quantitative Analysis, 39, 343-364.

Newey, W., West K., (1987): “A Simple Positive-Definite Heteroskedasticity and Autocorrelation Consistent Covariance Matrix”, Econometrica, 55, 703-708.

Nofsinger J. (2008): “The Psychology of Investing”, 3rd ed., Pearson Prentice Hall.

Odean, T., (1998): “Are Investors Reluctant to Realize Their Losses?”, Journal of Finance, 53, 1775-1778.

Odean, T., (1999): “Do Investors Trade Too Much?”, American Economic Review,

89, 1279-1298.

Pardo, A., Valor, E., (2003): “Spanish Stock Returns: Rational or Weather-Influenced?” European Financial Management, 9, 117–126.

Rind, B., (1996): “Effects of Beliefs About Weather Conditions on Tipping”, Journal of Applied Social Psychology, 26, 137–147.

Rotton, J., and Kelly, I., (1985): “Much Ado About the Full Moon: A Meta-analysis of Lunar-Lunacy Research”, Psychological Bulletin, 97, 286-306.

Saunders, E., (1993): “Stock Prices and Wall Street Weather”, American Economic Review, 83, 1337–1345.

Schwarz, N., Clore, L., (1983): “Mood, Misattribution, and Judgements of Well-being: Informative and Directive Functions of Affective States”, Journal of Personality and Social Psychology, 45, 513–523.

Schwarz, N., Strack F., Kommer, D., and Wagner, D., (1987): “Soccer, Rooms, and the Quality of Your Life: Mood Effects on Judgements of Satisfaction With Life in General and With Specific Domains”, European Journal of Social Psychology, 17, 69–79.

Schwarz, N., (1990): “Feelings as Information. Informational and Motivational Functions of Affective States”, Handbook of Motivation and Cognition, Guilford Press, New York.

Simon, H. A. (1955): “A Behavioral Model of Rational Choice”, Quarterly Journal of Economics, 69, 99–118.

Slovic, P., Finucan, M., Peters, E., and MacGregor, D., (2002): “The Affect Heuristic”, Intuitive Judgment: Heuristics and Biases, Cambridge University Press, Cambridge, UK.

Wann, D., Melnick, M., Russell G., and Pease, D., (2001): “Sport Fans: The Psychology and Social Impact of Spectators”, Routledge London.

White, H., (1980): “A Heteroskedasticity-Consistent Covariance Matrix Estimator and a Direct Test for Heteroskedasticity”, Econometrica, 48, 817-838.

Wright, F., Bower, H. (1992): “Mood Effects on Subjective Probability Assessment”, Organizational Behavior and Human Decision Process, 52, 276-291.

Wyndham, C., (1969): “Adaptation to Heat and Cold”, Environmental Research, 2, 442–469.

Yuan, K., Zheng, L., and Zhu, Q., (2006): “Are Investors Moonstruck? Lunar Phases and Stock Returns”, Journal of Empirical Finance, 13, 1-23.

7. Appendixes

Figure I:

MSCI The Netherlands 1998-2008 MSCI The Netherlands

Figure II:

MSCI Netherlands Small-Cap MSCI Small Cap

0 20 40 60 80 100 120 140 160 180 200 1 /1 /1 9 9 8 1 /7 /1 9 9 8 1 /1 /1 9 9 9 1 /7 /1 9 9 9 1 /1 /2 0 0 0 1 /7 /2 0 0 0 1 /1 /2 0 0 1 1 /7 /2 0 0 1 1 /1 /2 0 0 2 1 /7 /2 0 0 2 1 /1 /2 0 0 3 1 /7 /2 0 0 3 1 /1 /2 0 0 4 1 /7 /2 0 0 4 1 /1 /2 0 0 5 1 /7 /2 0 0 5 1 /1 /2 0 0 6 1 /7 /2 0 0 6 1 /1 /2 0 0 7 1 /7 /2 0 0 7 Date P ri c e Figure III:

MSCI Europe Excluding The Netherlands 1998-2008 MSCI Europe Excluding The Netherlands

Panel A:

Brief Review of the Terminology and the Mechanics of the Lunar Cycle

Table A: Descriptive Statistics for All Regressions (MSCI Netherlands)

I provide: mean, maximum, minimum, standard deviation, R-squared, Durbin-Watson (D-W), Skewness, Kurtosis, Jarque-Bera (J-B) test, White test, F-statistic and observations.

*All coefficients are significant at the 5% significance level.

Table B: Descriptive Statistics for All Regressions (With the Control Variable)

I provide: mean, maximum, minimum, standard deviation, R-squared, D-W, Skewness, Kurtosis, Jarque-Bera test, White test, F-statistic and observations.

1 2 3 4 5 6 7 8 9 10 11 12 Mean 0,0002 0,0002 0,0002 0,0002 0,0002 0,0002 0,0002 0,0002 -0,0004 0,0002 0,0002 0,0002 Maximum 0,027 0,027 0,027 0,027 0,027 0,027 0,027 0,027 0,014 0,027 0,027 0,027 Minimum -0,039 -0,039 -0,039 -0,039 -0,039 -0,039 -0,039 -0,039 -0,023 -0,039 -0,039 -0,039 Std. Dev. 0,014 0,014 0,014 0,014 0,014 0,014 0,014 0,014 0,014 0,014 0,014 0,014 R-Squared 0,797 0,797 0,797 0,797 0,797 0,797 0,797 0,797 0,799 0,797 0,797 0,797 D-W 2,102 2,100 2,101 2,100 2,100 2,099 2,101 2,103 2,131 2,100 2,102 2,100 Skewness -0,28 -0,281 -0,274 -0,272 -0,284 -0,283 -0,283 -0,286 -0,247 -0,28 -0,278 -0,281 Kurtosis 5,228 5,223 5,225 5,223 5,233 5,219 5,233 5,225 4,537 5,228 5,236 5,225 J-B 561,631 559,177 558,831 557,233 564,599 557,652 564,458 561,535 65,263 561,155 565,011 560,264 p-value 0,000 0,000 0,000 0,000 0,000 0,000 0,000 0,000 0,000 0,000 0,000 0,000 White-test 68,569 67,262 68,569 68,642 67,489 66,97 66,864 67,164 18,915 66,849 28,777 33,911 p-value 0,000 0,000 0,000 0,000 0,000 0,000 0,000 0,000 0,000 0,000 0,000 0,000 F-statistic 5012,694 3339,627 5014,21 3342,646 5013,58 3340,595 5012,375 3345,856 793,717 3342,424 2004,432 1670,365 p-value 0,000 0,000 0,000 0,000 0,000 0,000 0,000 0,000 0,000 0,000 0,000 0,000 Observations 2553 2553 2553 2553 2553 2553 2553 2553 601 2553 2553 2553

Table C: ARCH Effects Test

F-statistic and probability are provided

*All values are significant at the 5% significance level.

Table D: GARCH Model

*All values are significant at the 5% significance level.

Note: rows from up to down: R-squared, D-W, F-statistic, p-value.

Table E: Football Results

I provide: Date, opponent, result (first I put always the Dutch team)

Date Opponent Result

Table F: Descriptive Statistics for All Regressions (MSCI Netherlands-Small Cap)

I provide: mean, maximum, minimum, Std. Dev., R-squared, D-W, skewness, kurtosis, J-B, p-value, White-test, p-value, F-statistic, p-value, observations

Table G: Descriptive Statistics for All 12 Regressions (MSCI Netherlands-Small Cap), with Control Variable

I provide: mean, maximum, minimum, Std. Dev., R-squared, D-W, skewness, kurtosis, J-B, p-value, White-test, p-value, F-statistic, p-value, observations

Table H: ARCH Effects Test (small-cap)

F-statistic and probability are provided

*All coefficients are significant at the 5% significance level.

Table I: GARCH Model (small-cap)

*All coefficients are significant at the 5% significance level. Note: rows from up to down: R-squared, D-W, F-statistic, p-value.