The Efficient Market Hypothesis in relation to bad news world events

(1)

The Efficient Market Hypothesis

in relation to

bad news world events

University of Groningen

Faculty of Economics and Business

Business Administration, Msc Finance

Marloes Bijen

Supervisors:

A.J. Meesters

L. Dam

(2)

(3)

Preface

The last six months I have been writing this thesis as the last step of completing my Master Business Administration Finance at the University of Groningen. The thesis does not only complete my Master, but also the fantastic time I had while studying in Groningen. The idea of this thesis arose during the summer holiday of 2009. I had finished the course Behavioral Finance and was interested in the behavior of investors. The subject of my research was not directly clear, but I would like to combine large world events and investor behavior. The general assumptions in finance theories about efficient markets and rational investors induce me to the topic: efficient markets during bad news world events.

There are a number of people I like to thank for their support in writing this thesis. First of all, I like to thank my supervisor Aljar Meesters for his useful comments and advice. He has always given me fresh ideas on how to best approach the problem. In addition, he was always prepared to help me and motivated me till the last moment to finish my thesis in a good way.

Furthermore, I would like to thank two people especially, my boyfriend Job and friend Dicky, for their help and improvement of my thesis. In addition, I would like to thank my parents, family and friends for always supporting me.

M.A.M. Bijen Flemingstraat 12 7555 BN Hengelo

(4)

Abstract

This paper examines the efficient market hypothesis in relation to unexpected world events, which are seen as bad news for the period January 1994 to December 2008. In this study bad news world events are defined as financial crises, terroristic attacks and natural disasters. These events are tested on the national stock market and on three control stock indices. These control markets are the S&P 500, FTSE Eurofirst 300 and the MSCI Emerging Market Index and represent the stock markets of the United States, Europe and emerging countries. The efficient market hypothesis is tested with the event study methodology. Evidence is found for significant abnormal returns during the event day and during arbitrary days in the event window. Financial crisis events show more frequent and higher significant abnormal returns than natural disaster events. These findings could be due to uncertainty in the stock market during bad news world events. Furthermore, this study is the first to find evidence that price volatility is an important factor in determining abnormal returns during unexpected bad news world events.

(5)

1. Introduction

On Tuesday 11 September 2001, the terroristic attack in the centre of the world’s financial community resulted in turmoil in financial markets. The stock exchanges in the United States cancelled trading and closed for a week until Monday 17 September. By the end of the week, the Dow Jones Industrial Average stock market index had fallen by -14.3%, the largest one-week drop in history.

More recently, on 15 September 2008, the bankruptcy of Lehman Brothers led to a decline of - 4.4% of the Dow Jones market index. This was the largest drop in points on a single day since the drop after the terroristic attack on 11 September 2001. These two major world events are examples of events that led to large stock price declines in stock markets all over the world. The major objective of this study is to examine if there exist a broad relation between world events and stock prices movements. More specific, this paper tests if the efficient market hypothesis holds when world events, seen as bad news, occur. There has been little research performed on the influence of major world events on stock prices, although the efficient market hypothesis is an important concept in the economical literature. The hypothesis states that stock prices react efficient to information and incorporate news quickly and accurately. Therefore, investors should not be able to predict future returns and make abnormal profits, i.e. profits that are not a compensation for risk (Reilly and Drzycimski, 1973).

However, many academic studies have challenged the notion of informational efficient equity markets by showing that past prices can predict future movements in prices and that investment strategies based on historical returns can generate subsequent risk-adjusted abnormal returns. De Bondt and Thaler (1985) were the first to provide evidence of market efficiency anomalies. They found that due to bad (good) news a stock decreases (increases) too far in price but returns to its intrinsic value as investors realize that they had overreacted. They call this effect the price overreaction hypothesis. The overreaction effect contradicts to the efficient market hypothesis since it is possible to predict future movements and profit from arbitrage opportunities. De Bondt and Thaler (1985) used the research methodology to form portfolios of winner (i.e. positive) and loser (i.e. negative) stocks based on historical returns. They found that loser portfolios outperform the market by an average of 19.6% after 36 months while the winner portfolios underperform the market by an average of 5% after 36 months.

(8)

Their results provided evidence of the price overreaction effect for informed and uninformed events. These two research methods are often used to investigate the efficient market hypothesis. However, a serious shortcoming of these research methods is the lack of information that induces the large abnormal returns. As a result, these methods make it difficult for investors to cater on the arbitrage opportunities, since only afterwards the extreme returns could be observed. This study differs from much of the existing research since it examines the efficient market hypothesis using well-known world events. The research method enables to investigate and compare reactions of different stock indices for the same types of bad news events. Therefore, it is possible to draw some general conclusions about the influence of world events on the stock market. In addition, bad news world events lead often, as can be seen from the terroristic attack of 11 September 2001, to large drops of the stock market. Firms and also individual investors experience large financial losses during these world events. Therefore, it is important for investors to cater on these events, to reduce the financial losses and profit from arbitrage opportunities.

Furthermore, the last decade’s non-U.S. equity markets, particularly in the developing countries of Asia and Latin America, have grown considerably to the U.S. stock markets. This is the result of the above average economic growth in these countries as well as the increasing flow of funds from developed countries towards emerging equity markets. In addition, the expansion of the internet and developments in communication technology mean that information nowadays can be more quickly and more cheaply diffused than ever before. The rise in capital across stock markets around the world has been stimulated by a decline of trading barriers. As a result, stock market investments are no longer constrained to financial professionals and to investments in the own country. More individuals have begun to participate actively in stock markets across the world. Finally, more savings for retirement have been invested in stock markets through pension funds. Therefore, a better understanding of the relation between stock markets and world events could help financial professionals and individual investors to formulate investment strategies.

(9)

on three control stock indices. These indices, the S&P 500, FTSE Eurofirst 300 index and the MSCI emerging market index, act as control stock markets to represent a benchmark for stock markets of the United States, Europe and emerging countries. The control indices are used to investigate differences in reactions between stock markets on world events.

While using a setting in which the efficient market hypothesis has not previously been tested, a new phenomenon is found. During bad news world events like financial crises, terroristic attacks and natural disasters significant positive and negative abnormal returns are found during the event day, but also on arbitrarily days in the event window. These findings suggest that uncertainty in stock markets increases during unexpected bad news world events. Changing price volatility results in significant abnormal returns during these days. These results are new, since previous studies used other research methods with randomly selected large returns, instead of new

information, as events. The significant positive and negative abnormal returns found in this paper after an event, are in other studies selected as event. Therefore, researchers were not able to find increasing uncertainty in the stock market. The results of this paper show nevertheless, that it is not possible as investor to profit from these abnormal returns since they are found in an

unstructured way.

(10)

2. Literature review

In this section, literature and theories about the efficient market hypothesis in relation to bad news world events are discussed. The section starts with the definition and features of the efficient market hypothesis and the price overreaction hypothesis. The second subsection discusses

empirical evidence that supports the efficient market hypothesis. In addition, theories which reject the efficient market hypothesis are discussed. In the third subsection factors which try to explain market efficiency anomalies are considered. In subsection four the impact of major world events on stock markets is analysed. Finally, subsection five describes the impact of the efficient market hypothesis on international stock markets with the focus on European and emerging markets. In the last subsection is the research question formulated.

2.1 The Efficient Market Hypothesis

The efficient market hypothesis is an important concept that has become widely accepted in the economic literature since the late 1950s. The efficient market hypothesis is based on the assumption that all new relevant information is incorporated and reflected in stock prices. Fama (1970) distinguishes three categories of market efficiency: the weak, semi-strong and strong form. The weak form of the efficient market hypothesis suggests that all historical prices of a stock are reflected in the current price. The semi-strong form of market efficiency states that stock prices quickly reflect all publicly available information. Studies of Fama (1970, 1991) support this form of market efficiency. The strong form of market efficiency supposes that stock prices reflect all publically as well as private information. Fama (1998) assumes that the efficient market hypothesis still holds when new information enters the market. Because he expects that

overreaction to this information occurs at the same time as underreaction, causing the sum effect to reflect the stock price at its intrinsic value (Fama, 1998). As a result, the efficient market hypothesis supposes that there will be no return reversals observed on the day following the particular events and implies that that there are no arbitrage opportunities. In other words, it would not be possible to predict future returns or make abnormal profits (i.e. profit without extra risk).

(11)

on how representative an event is in comparison with their own beliefs. The representative bias results in the effect that investors exaggerate the economic implications of major events by overvaluing stocks in reaction to good news and undervaluing stocks in reaction to bad news. The price overreaction effect supposes that stock prices divert away from their intrinsic value on the moment new information enters the market and converts back to the intrinsic value when

investors realize that they overreact (i.e. a negative price shock should be followed by an increase in price the following day). The testable aspects of the overreaction hypothesis includes that a price reaction followed by a reversal can be taken as evidence of overreaction. Furthermore, the more extreme the initial movement, the greater will be the subsequent adjustment (Mun, Vasconcellos and Kish, 1999).

In contrast to the overreaction hypothesis, the underreaction hypothesis supposes that investors do not respond strong enough to unfavourable and favourable information. Specifically, on the release of unfavourable (favourable) news, investors temporarily price securities above (below) their new intrinsic values (Spyrou, Kassimatis and Galariotis, 2007).

Brown, Harlow and Tinic (1988) explain anomalies of the efficient market hypothesis with the uncertain information hypothesis. They find that with the arrival of unexpected information, investors become uncertain which leads to increased price volatility. The response of investors is setting stock prices below their expected values, at favourable as well as unfavourable events. Reducing the uncertainty results in positive price changes, regardless of whether the initial event was perceived as positive or negative.

(12)

In conclusion, markets react not always efficient as assumed in economic theories. Therefore, the following subsection describes the varying methods used to test the efficient market hypothesis. In addition, empirical evidence is highlighted. After that, in subsection 2.3 different factors are discussed which try to explain anomalies of the efficient market hypothesis.

2.2 Empirical evidence of inefficient markets

Most studies find reversal patterns of stocks for large price changes and for different time periods. They provide this as evidence of the stock market overreaction effect and indirectly as evidence of inefficient markets.

The first and most prominent research in the area of long-term stock market overreaction is the study by De Bondt and Thaler (1985). Using historical information, they argue that going long in a portfolio consisting of badly performing stocks in the past, i.e. loser stocks, and going short in a portfolio consisting of good performing stocks in the past, i.e. winner stocks, will produce abnormal returns for the long term. De Bondt and Thaler (1985) used monthly stock returns from the New York Stock Exchange for the period from January 1926 until December 1982 and constructed portfolios based on winner and loser stocks. Winner (loser) portfolios consist of the 35 best (worst) performing stocks over the past 36 months. They find that loser portfolios outperform the market by an average of 19.6% after 36 months and winner portfolios

underperform the market by an average of 5%, even after controlling for risk differences and size. Furthermore, they find significant reversal patterns after large changes in stock prices, which they highlight as evidence of the price overreaction effect.

(13)

Another method, in which the overreaction hypothesis is studied, is the analysis of stock returns following large one-day stock price changes. This method uses a predetermined return (i.e. 10%) to select the sample firms. Subsequently, they looked for reversal patterns in the following days of the large price movement. The following three studies are only a small selection of the broad collection of literature in which this method is used (e.g. Cox and Peterson, 1994; Larson and Madura, 2003).

Howe (1986) investigates the overreaction hypothesis for stocks traded on the AMEX and the NYSE from 1963 to 1981 using weekly returns. He defined an event as a stock price change of 50% in either direction within one week. Results of his study show that stocks with bad news (i.e. negative price changes) experience significantly higher returns than the market during a period of 20 weeks after the event and stocks with good news (i.e. positive price changes) experience 30% lower returns then the market during the 50-week period following the event. In a similar study, Bremer and Sweeney (1991) investigate all Fortune 500 companies where there is a one-day price decline of 10% or greater in the period July 1962 until July 1986. They found that the average rebound after one-day is 1.77% and the rebound after a two-day period is 2.2%. They notice that extreme large negative 10-day rates of return are followed by larger than expected positive rates of return over the following days and that this price adjustment lasts approximately two days. They conclude that a slow recovery period is inconsistent with the notion that market prices quickly reflect all relevant information. More recently, Spyrou, Kassimatis and Galariotis (2007) examine short-term investor reactions to extreme events in the United Kingdom stock market for the period 1989 to 2004. A difference of two standard deviations above or below the average daily index return is defined as an extreme event. Their results indicate a significant

underreaction effect to both positive and negative extreme events.

(14)

They find significant abnormal returns for both NASDAQ winner and loser portfolios, indicating the existence of an overreaction effect. Furthermore, their results show a stronger overreaction effect for NASDAQ losers than for winners. These findings contradict to the efficient market hypothesis, where it is not possible to predict future returns.

All previous mentioned studies select large returns to examine the overreaction effect. They find price reversals and use this as evidence of the price overreaction effect. Furthermore, they argue that the overreaction effect is an indication of fundamental market inefficiency. However, other studies (Atkins and Dyl, 1990; Zarowin 1990; Chopra, Lakonishok and Ritter, 1992) suggest that the obviously overreaction hypothesis is due to other factors like size, risk, calendar effects or the bid-ask spread. The following section discusses these factors in more detail.

2.3 Factors inducing the price overreaction effect

Some studies question the strong findings of the price overreaction effect on grounds of size differences between winner and loser stocks. If it can be proved that loser stocks are from smaller than average firms whereas winner stocks are not, the price overreaction effect of losers could simply be a reaction of the small-firm effect.

Brown, Harlow and Tinic (1988) give as alternative explanation of price reversals after large price returns the uncertain information hypothesis, which is also mentioned in subsection 2.1. The hypothesis predicts that both the risk and expected return of a firm increase systematically after extreme stock price changes. Investors incorporate a risk premium into stock prices when confronted with new information and as a result, prices increase and negative returns show a reversal pattern.

Correspondingly, if returns to companies primarily occur in January, the price overreaction effect could be a reaction of the January effect. Investors choose to sell some of their stocks before the end of the year in order to claim a capital loss for tax purposes. In January, the same investors quickly reinvest their money in the market, causing stock prices to rise.

(15)

possible to make profit from excess returns. In the following paragraphs, several studies which control for these factors will discussed.

Chopra, Lakonishok and Ritter (1992) re-examine the findings of De Bondt and Thaler (1985) with a multiple regression model and include prior returns, size and betas. Findings indicate that loser portfolios formed on prior five-year returns outperform winner portfolios by 5 to 10% per year during the following five years. In addition, the study shows evidence for larger arbitrage portfolio returns during the month January and for smaller firm size. The study of Chopra, Lakonishok and Ritter (1992) is in contrast with the study of Fama and French (1986), who compare the returns of De Bondt and Thaler (1985) with returns of size-matched portfolios and conclude that the size-effect can only explain a fraction of the price overreaction effect. This statement is supported by Dissanaike (2002), who reports only a size-effect within the sample of large FT500 companies. However, Zarowin (1990) finds the tendency for losers to be smaller-sized firms than winners. When losers are compared to winners of equal size, there is little evidence of any return discrepancy. In periods when winners are smaller than losers, winners outperform losers. He concludes that a widely regarded efficient market anomaly is subsuming by the size phenomena. Moreover, Ma, Tang and Hasan (2005) examine the relation between the announcement-period stock returns and the subsequent price reversal after controlling for size. They find that the overreaction effect last only about two days for NASDAQ stocks and the magnitude of the reversal effect is smaller for larger NASDAQ firms. From these studies can be concluded that the size effect has a large influence on winner-loser portfolios. When portfolios control for size-effects only small evidence of the price overreaction hypothesis is proved. Just like size, risk also has a significant influence on the price overreaction hypothesis. While Zarowin (1990) and Chopra, Lakonishok and Ritter (1992) find significant abnormal profits after controlling for risk, Chan (1988) came to another conclusion. He constructs portfolios for the period of December 1926 till December 1985 and found that the risk in portfolios correlates with the level of the expected market-risk premium. Accordingly, after controlling for systematic risk, only small abnormal returns are found for the contrarian investment strategy. Therefore, he concludes that it is not possible to take abnormal profits after controlling for risk.

Atkins and Dyl (1990) investigate the short-term overreaction hypothesis for NYSE stocks and control for seasonality effects and for the bid-ask spread. They select 300 randomly trading days to eliminate any biases resulting from day-of-the week or month-of the-year effects. Their study indicates that arbitrage portfolios earn positive returns after eliminating seasonality effects. In addition, they offer the bid-ask spread as explanation for reversal patterns but found no

(16)

evidence that investors could not earn abnormal profits from price reversals. They conclude that the market is efficient when transaction costs are considered. The results support both the overreaction and efficient market hypothesis since price reversals are observed, but there are no arbitrage opportunities. Similar results are found by Akhigbe, Gosnell and Harikumar (1998) who re-examined the work of Atkins and Dyl (1990) using contemporaneous measures of the bid-ask spread. Nevertheless, price reversals are not always primarily induced by bid-ask spreads (Cox and Peterson, 1994; Otchere and Chan, 2003).

Size, risk, seasonality effects and bid-ask spreads are puzzling factors in explaining efficient market anomalies. Whereas these factors have a significant influence on stock markets it is not always clearly perceptible if stock markets are efficient when controlled for these factors. The striking results between studies are attributing to different aspects including varying tested stock markets and the methodology used to test the efficient market hypothesis. This paper will test the efficient market hypothesis for similar events with stock indices instead of individual firms (e.g. Reilly and Drzycimski 1973; Ferri and Chung-ki, 1996; Schnusenberg and Madura, 2001). Stock indices are composed of large and small firms. Therefore, price reversals could not be mainly induced by the size effect. Moreover, potential biases due to bid-ask spreads and cross-sectional differences of individual stocks are avoided by the use of indices because portfolios of stocks (i.e. indices) are used. Finally, this paper controls for calendar effects since the events are not

randomly chosen but selected on basis of their news impact in the world. This paper will not control for size, calendar and bid-ask spread effects, since it use a different research method. However, because not only the methodology but also the stock market affects results, this study will test several international stock indices. Subsection 2.5 discusses the efficient market hypothesis for international stock markets. Before analysing the international stock markets, the research method and the influence of major world events on the stock market are discussed first.

2.4 The influence of world events on stock markets

(17)

events selected on base of their information and not on ground of their returns in the stock market.

Only little research is performed to the relation between efficient markets and world events. Reilly and Drzycimski (1973) test the efficient market hypothesis by examining abnormal returns during periods following widely known events. They select seven major world events which are assumed to have a large impact on the equity market. Their study conducts a test on the scale, direction, consistency and speed of stock market reaction to major world events. They found that over half of all significant stock price changes took place before the stock market opened the following day. Moreover, two-third of all significant price changes on the first full day are not followed by significant price changes during the second full day. Nevertheless, four out of seven events generated small net profit opportunities for the large samples. The authors conclude that the efficient market hypothesis is more frequently supported by more important and unexpected events. In a similar study, Niederhoffer (1981) examines the relationship between world events and movements in stock prices. World events are defined as the occurring of five-to-eight column headlines in the New York Times. Results show that larger changes are substantially more likely following world events than on randomly selected days. Furthermore, extreme bad news world events show a reversal pattern of 1.1% after a two to five day period. In addition, 25% of the negative price declines show a rise on the second day. These findings confirm the market’s strong tendencies to overreact on extremely bad news world events. Results show that national events have a larger impact on the stock market than regional events. In addition, clusters of world events, especially those occurring on a crisis, show larger changes than isolated world events. The two studies about world events show varying results when they tested the efficient market hypothesis. Reilly and Drzycimski (1973) conclude that more unexpected and important events support the efficient market hypothesis, while Niederhoffer (1981) ascertain that stock markets have a stronger tendency to overreact on extreme large clustered world events than on regional isolated events. These are the only two studies who investigate major world events in the field of efficient markets. Therefore, another study is analysed which first observes large abnormal returns and then examines if the abnormal returns are influenced by world events.

(18)

public information reduces uncertainty, as uninformed winners experience overreaction and informed winners do not. These results correspond to the findings of Daniel and Titman (2001) which show that investors overreact to intangible information (i.e. information that cannot be observed from financial statements).

The above-mentioned studies show all significant influence of new information on the stock market, but their conclusions of the influence of world events on the stock market differ. Reilly and Drzycimski (1973) conclude that more unexpected and important events show more frequent efficient markets, whereas other studies (Niederhoffer, 1981; Larson and Madura, 2003) conclude that uninformed and intangible information stimulates an overreaction effect. Moreover, the overreaction effect is most likely to occur when unanticipated news enters the stock market (Niederhoffer, 1981; Howe, 1986). This statement is supported by Brown, Harlow and Tinic (1988). They explain this effect by the increase of risk and expected returns after news of extreme events. When an event with a negative impact on the stock markets occurs, investors become uncertain and risk-premiums increase. When the uncertainty decreases positive price reversals are observed. In addition, the tendency for a price reversal is expect to be stronger when the initial price change is larger (Bremer and Sweeney, 1991; Akhigbe, Gosnell and Harikumar, 1998). Whereas, the small attention in the literature to major world events, this study will fill in a gap in the literature. Furthermore, this study tries to clarify the varying results between the studies of Reilly and Drzycimski (1973) and Niederhoffer (1981). The focus in this paper is on the relation between negative major world events and stock markets. Different international indices are used to investigate the influence of similar events on different stock markets. The following subsection pays attention to the efficient market hypothesis in relation to international stock markets.

2.5 The efficient market hypothesis and international stock markets

Most studies test the efficient market hypothesis for equity markets in the United States (Howe 1986; Schnusenberg and Madura, 2001; Ma, Tang and Hasan, 2005; Michayluk and Neuhauser, 2006). However, research of the efficient market hypothesis is not limited to the United States, but expands to several international stock markets. This subsection focuses on stocks markets in emerging countries and in Europe.

(19)

December 1990. After controlling for risk, he provides evidence of long-term price overreaction effects for the Sao Paulo Stock Exchange. The loser (winner) portfolios outperform

(underperform) the market by a statistically significance of 17.63% (-20.25%) after a two-year period.

In a more widespread study, Harvey (1995) examines the predictability and risk of twenty emerging markets. Results show that the amount of predictability in emerging markets is larger than found in developed markets, which is in accordance with the research of Lasfer, Melnik and and Thomas (2003). In addition, the world market portfolio beta shows little influence on the expected returns of emerging markets. This indicates that local information has a larger influence than global information in predicting future returns of emerging markets. This conclusion is in contrast with the study of Niederhoffer (1981) who concludes that national information affects stock markets more than regional information. The different findings are due to various research methods. Where Harvey (1995) uses winner and loser portfolios, Niederhoffer (1981) uses major world events as test variable for the efficient market hypothesis.

Zamri and Hussain (2001) investigate the Kuala Lumpur Stock Exchange (KLSE) for a period of ten year. Even after controlling for size, time-varying risk and seasonal effects, they find

significant price reversals of both winner and loser portfolios for the following three year. Furthermore, abnormal returns are found by holding a contrarian investment strategy. In a similar study, Hammed and Ting (2000) provide evidence of short term overreaction effects of the Kuala Lumpur Stock Exchange. From these studies can be concluded that the Kuala Lumpur Stock Exchange shows overreaction effects for both the long and short term. Evidence of overreaction effects is also found for the Chinese stock markets. Wang, Burton and Power (2004) analyse a sample of 300 Chinese shares of the Shanghai and Shenzhen stock market in the period August 1994 to July 2000. Their results show that winner portfolios underperformed the market by -0.55% and loser portfolios outperformed the market by 0.52% after two weeks. When the

Chinese stock market is controlled for firm-size, risk and the bid-ask spread still abnormal profits for short term contrarian investment strategies are found (Kan, Liu and Ni, 2002).

(20)

The efficient market hypothesis is moreover investigated for other international stock markets, including Europe. Ajayi and Mehdian (1994) examine eight international stock market indices for the period April 1985 through July 1990. They find that loser portfolios outperform winner portfolios for five of the eight stock markets (e.g. Canada, Germany, Italy, Japan, and the United Kingdom). Furthermore, they provide evidence that variances are higher during negative news events than during positive news events, which is explained by the uncertain information hypothesis. During negative news events, investors become uncertain which result in increasing volatility. Moreover, when Rouwenhorst (1998) examine twelve European stock markets and controls for the factors risk and size, no overreaction effect is found. The international diversified portfolio of past winners outperformed the portfolio of past losers by about 1% per month for the 12 European stock markets. These results are supported by the study of Lasfer, Melnik, and Thomas (2003) which investigate 39 developed and emerging stock indices. Their research show that negative (positive) price changes are followed the next day by negative (positive) abnormal returns, therefore no price reversals are observed.

In contrast to previous results, several studies found even after controlling for risk, evidence of overreaction effects for European stock markets. Baytas and Cakici (1999) examine equity markets for several international stock markets. They find that the average return of loser portfolios outperform the average return of winner portfolios by 62.9% in France, 58.5% in the United Kingdom, 50.5% in Germany and 21.6% in Italy. The concluded that contrarian

investment strategies result in abnormal profits. Mun, Vasconcellos and Kish (1999) investigate the efficient market hypothesis further for the French and German stock markets with

nonparametric test statistics. Contrarian investments strategies result in abnormal profits even after controlling for risk and January effects but they decrease over time. This implies that investors overreact immediately when new information enters the market. Furthermore, for the Spanish stock market are short and long term contrarian investment strategies found (Alsonso and Rubio, 1990; Forner and Marhuenda, 2003). In addition, not only the Spanish stock market shows overreaction effects, the same reactions are found on Greek stock markets. The Athens Stock Exchange is investigated with weekly and daily data and both methods show short-term overreaction effects (Galariotis, 2004; Diacogiannis et al., 2005). Contrarian investments strategies produce significant profits which are neither full explained due to risk nor due to market frictions.

(21)

effects are observed. The effects of these factors are in more detail described in subsection 2.3. Furthermore, emerging stock markets show a low correlation with stock markets in developed countries. This result is favourable for investors which could reduce portfolio risk.

All previous studies that tested the efficient market hypothesis for emerging and European stock markets are performed with the same method. Portfolios of winner and loser stocks are formed and they are observed for some following period. This paper uses another method to investigate the efficient market hypothesis for international stock markets. It determines the influence of bad news world events on national stock markets. In addition, benchmark indices for the United States, Europe and emerging countries are used to test the influence of the same world events on different stock markets.

2.6 Summary of the discussed literature and hypothesis formulation

Extensive research has been performed to the efficient market hypothesis in the course of which various results are found. Some studies found strong evidence of overreaction effects while other studies explain the effect by factors like size, risk, seasonal effects or the bid-ask spread. Several stock markets and different research methods are used to investigate the efficient market

hypothesis. However, from the literature can be concluded that markets react not always as efficient as assumed in economic theories. This could result in various arbitrage opportunities for investors to profit from anomalies of the efficient market hypothesis.

This paper differs from previous studies in two ways. Firstly, another research method is used. The method most commonly used to test the efficient market hypothesis is characterized by forming portfolios based on past returns during an estimation period. If the loser portfolio outperforms the winner portfolio for some following period, future returns could be predicted, which support evidence opposed to the efficient market hypothesis. Another frequently used method for examining the efficient market hypothesis is uses extreme abnormal returns of the stock market. Portfolios from large abnormal returns are formed and reversing trends in the stock market are investigated.

(22)

and Drzycimski, 1973). However, these studies differ in several ways from this research. First, Niederhoffer (1981) examined the influence of world events on stock prices of one stock index, the S&P composite index, while this paper uses a number of stock indices to investigate world events. It tests first the impact of the world event on the national stock index of the country in which the event occurred. After that, it tests the influence of the world event on three control stock indices which represent stock markets of the United States, Europe and emerging countries. The control stock indices are used to compare the reactions of various stock indices during the same world events. Second, this study uses an event study methodology with twenty-six days in the event window, whereas the other two studies only examine the influence of world events on a shorter time period. Finally, the two studies, which also make use of world events, show

contrasting results. Niederhoffer (1981) concluded that national, clustered events and extreme negative events support the overreaction effect, while Reilly and Drzycimski (1973) concluded the opposite. Their results show that more unexpected and important events are, the stronger support they provide for the efficient market hypothesis. Therefore, this study will try to clarify these striking results.

The aim of this study is to test the efficient market hypothesis in relation to bad news world events. Bad news world events are defined as large events in the world with have an expected negative influence on the stock market. The research hypothesis is formulated as follow:

Does the efficient market hypothesis holds in case of bad news world events?

(23)

3. Events

The previous section discussed some literature of the efficient market hypothesis and defined the research question. This section describes the selection criteria for the events. In addition, the selected events will be separately per type of event discussed.

3.1 Event selection criteria

There is no unified way to select major world events for testing the efficient market hypothesis. Therefore, selection criteria are formulated to construct a framework on which the events should be selected. Previous studies used selection criteria like stock markets which are divided in winner and loser portfolios or, they defined an extreme abnormal return (i.e. 10%) and select the corresponding firms to form portfolios (Howe, 1986; Bremer and Sweeney, 1991; Cox and Peterson, 1994). Other methods to select events use various key words that highlight

unanticipated events and screen databases using these words (Brooks, Patel and Su, 2003). Major world events are defined by Niederhoffer (1981) as the occurring of five-to eight-column

headlines in the New York Times. This study makes no use of newspapers but formulates four main criteria to select world events in a well-considered way.

Firstly, changes in stock prices arise from new unexpected information. This means that the events are experienced as large economic, political or geographical shocks in the world. Larson and Madura (2003) found evidence of the price overreaction effect for uninformed events and not for informed events. So, unexpected events led more frequently to the price overreaction effect. Secondly, Niederhoffer (1981) describes in his article that national events are more likely to affect the stock market than regional events. Using this knowledge, it is plausible to assume that world events have an even larger impact on the stock market than national events. Therefore, only events which are assumed to have a large impact in the world are selected.

Thirdly, this paper investigates the efficient market hypothesis for different national and

international stock markets. To examine differences between reactions in stock markets on world events, it is important to select events occurring across varying countries in the world.

Furthermore, world events have contagion effects, effects which influence the general process of shock transmissions across countries. Hon, Strauss and Yong (2004) found this effect for the terroristic attack on September, 11, 2001 in the United States. They found that European stock markets responded more similar to the United States stock market shocks for about three months. However, most of the studies found only contagion effects for stock markets in emerging

(24)

on other stock markets, but also the other way around, the influence of events in emerging countries on stock markets of the United States. Therefore, events are selected which occurred across the whole world.

Finally, the number of events chosen in this study should be limited to allow a detailed analysis of the efficient market hypothesis. Therefore, events are selected between the period of January 1994 and December 2008. In addition, only world events which have an expected negative influence on stock markets are selected because negative abnormal returns led more frequently to overreaction effects than positive returns (Atkins and Dyl, 1990; Ma, Tang and Hasan 2005). Moreover, one day is chosen as the event day, even as the event occurred during a financial crisis. The reason for this selection criterion is that one day price declines show more often reversal patterns (Brown, Harlow and Tinic, 1988; Bremer and Sweeney, 1991).

3.2 Event selection

The selected events should satisfy the four main criteria mentioned in subsection 3.1. Moreover, to examine the reaction of stock markets on different world news, various types of events are selected. The selected events could be classified in three categories; financial crises, terroristic attacks and natural disasters, which are described in more detail in the following paragraphs. 3.2.1 Financial crises

The contagion effect, (i.e. a shock in a country leads to the transmission of shocks in other countries) was a prominent feature of the financial crisis that has flooded emerging market economies in recent years (Caramazza, Ricci and Salgado, 2000). In December 1994, the Mexican government let the peso float. The peso fell by more than 12% and Mexico ended up in a financial crisis.

(25)

Lehman Brothers occurred during the financial crisis in the United States, it can be seen as an important day, because the event led to the expanding of the instability of global markets. In addition, strong contagion effects in international stock markets are observed after this event. These four financial crisis events are selected because of their large impact on the world economy in the period between January 1994 and December 2008. In addition, the events occurred across different geographical and economical areas. The price overreaction effect is investigated during the Asian financial crisis by Otchere and Chan (2003) for the Hong Kong stock market and by Michayluk and Neuhauser (2006) for the United States stock market. Both studies found a short-term overreaction effect. However, they investigated the stock markets during the Asian financial crisis and selected events on base of extreme negative returns. This research method varies with the research method used in this paper. Events are selected on base of their information and not on base of their extreme returns. Table 1 presents the four selected financial crises.

Table 1: Selected financial crisis world events

Name Country Financial Crisis

Floating of the Mexican peso Mexico Mexican financial crisis

Floating of the baht Thailand Asian financial crisis

Expanding of the rubble/dollar trading band Russia Russia financial crisis Bankruptcy of Lehman Brothers United States World financial crisis

3.2.2 Terroristic attacks

Terroristic attacks have serious implications for stock markets. When information becomes available about a disastrous event, investors search for safer financial instruments and panic selling arises. However, financial markets were efficient in absorbing the shocks caused by terroristic attacks and continued to perform their functions in an effective way (Eldor and Melnick, 2004; Johnstone and Nedelescu, 2006). The impact of the 11 September attack was visible worldwide on major equity markets. All major stock markets (e.g. -5.7 % London Stock Exchange, -8.7 % German Stock Exchange, - 9.2% Brazilian Stock Exchange) showed large declines. The decline in the European stock market, which opened before United States stock markets were re-opened, was even greater after 17 September, because of spill-over effects, side effects which follow from the terroristic attack (Brooks, Patel and Su, 2003). Abadie and

(26)

performance at the end of the cease-fire. Previous studies examine these two terroristic attack events, because they were the only terroristic attacks where the Dow Jones showed significant negative abnormal returns. However this paper uses another research method to select events. First are world events selected and afterwards is the impact on stock markets determined.

Therefore, this selection criterion was useless for this study. In this study, the selection criteria for terroristic attacks are based on casualties and geographic location. The terroristic attacks are widespread across the United States, Europe and emerging markets. Table 2 shows the selected terroristic attack events.

Table 2: Selected terroristic attack world events

Name Place Country Fatalities

11 September Attack New York United States 3,025

Madrid train bombings Madrid Spain 191 (injured 1,400)

London Bombings London United Kingdom 56 (injured 700)

Mumbai Attack Mumbai India 173 (injured 350)

Source: Worldwide Incidents Tracking System

3.2.3 Natural disasters

In December 2004 the South-East Asia Tsunami caused heavy damage to the coastlines across South Asia. This powerful earthquake caused 230,000 deaths and left millions homeless. In addition, the earthquake caused heavy damage to the economy. Lee, Wu and Wang (2007) concluded that the Tsunami was a major event that shocks South Asia markets and had an impact on the economy in emerging and neighbouring markets in the Asia region. Of all natural

disasters, earthquakes probably cause the most serious damage to human life, property assets and the economy.

Therefore, three earthquakes are selected as events. These events are selected on the basis of the highest casualties in the world in the period between January 1994 and December 2008.

(27)

Table 3: Selected natural disaster world events

Name Location Casualties Natural Disaster Richter

magnitude

Tsunami Indian Ocean/Thailand 230,000 Earthquake 9.3

Kashmir Pakistan 100,000 Earthquake 7.8

Eastern Sichuan China 87,587 Earthquake 7.8

Source: United States Geological Survey

In total are eleven world events selected which are assumed to have a negative influence on stock markets. In addition, these events satisfy the four main criteria described in the previous

(28)

28 Table 4: Selected bad news world events listed in chronological order

Nr. Event Events Date Category event Country Stock Index

1 Fixed rate band increase to 15% 20-12-1994 Financial crisis Mexico Dow Jones Mexico

2 Thailand floated the baht 02-07-1997 Financial crisis Thailand Dow Jones Thailand

3 Expanding of the rubble/dollar

trading band 17-08-1998 Financial crisis Russia FTSE Russia

4 11 September attacks 11-09-2001 Terroristic attack United States Dow Jones Composite 65

5 Madrid train bombings 11-03-2004 Terroristic attack Spain Dow Jones Spain

6 Tsunami 24-12-2004 Natural disaster (earthquake) Thailand Dow Jones Thailand

7 London bombings 07-07-2005 Terroristic attack United Kingdom Dow Jones London

8 Kashmir 08-10-2005 Natural disaster (earthquake) Pakistan MSCI Pakistan

9 Eastern Sichuan 12-05-2008 Natural disaster (earthquake) China Dow Jones China

10 Bankruptcy of Lehman Brothers 15-09-2008 Financial crisis United States Dow Jones Composite 65

(29)

29

4. Data

In the previous part the eleven selected bad news world events are discussed. In this section the selected indices and data associated with the events are described, followed by descriptive statistics of the data.

4.1 Selected indices

To investigate the efficient market hypothesis for the eleven events, national stock market indices are used. Besides three control indices are used to examine the influence of world events on a broader scale.

The initial sample consists of nine national stock indices and three control stock indices. The Dow Jones Global Country indices are used for the countries Mexico, Thailand, Spain, United Kingdom, China and India. These indices targeting 95% market capitalization coverage and are constructed and weighted using free-floated market capitalization. A free-float method does only include shares which are tradable to anyone and does not include restricted stocks, such as those held by company insiders. Three countries are not represented by the Dow Jones Global Country Index, but by other indices because the DJ Global Country index was not available for these countries at January 1994. For the United States the Dow Jones Composite 65 is used as stock market. This is a composite, price-weighted stock index. This means that each stock influences the index in proportion to its price per share. For Pakistan the Morgan Stanley Capital International country index is used. This index is just like the DJ Global Country Index constructed using free-floated market capitalization. For Russia the Financial Times Stock Exchange (FTSE) country index is used. This was the only Russian index which started with daily total returns from June 1997. The FTSE Russia is a market-cap weighted index designed to measure the performance of the 15 biggest and most liquid Russian companies trading on the London Stock Exchange.

The three control stock indices should represent emerging, European and United States stock markets. To test the influence of different categories of events on the United States stock market, the Standard & Poor 500 index is selected. The S&P 500 index is a free-floated capitalization weighted index which consists of the 500 large-cap common stocks actively traded in the United Stated. The FTSE Eurofirst 300 Index is used to investigate the influence of bad news world events on Europe. The FTSE

(30)

30

4.2 Data description

The final sample includes eleven events which occurred in the period between January 1994 and December 2008. The bad news world events are described in detail in Table 4 of the previous section. To investigate the eleven world events, daily return data are used to test the efficient market

hypothesis. The reason behind the choice of daily data is that they provide more explanatory power than weekly or monthly intervals (Brown and Warner, 1985; MacKinlay, 1997). In addition, most studies which examine the efficient market hypothesis use daily data, because their methodology of event selection was based on extreme daily changes (Larson and Madura, 2003; Otchere and Chan, 2003; Ma, Tang and Hasan, 2005). Furthermore, by testing the efficient market hypothesis, the direct influence of the world event on the stock market is investigated and not a delayed influence. With the use of weekly or monthly data it could happen that no price overreaction effect was found while there was a strong declining of the stock market the following days.

To calculate abnormal returns in the event study, returns of indices are used instead of returns of individual firms of a stock market. This study investigates the effect of the efficient market hypothesis for different types of events and not for a specific firm. Besides that, the differences in results between stock markets and indices will be small, because a stock market is considered to consist out of all publically traded securities whereas an index is composed of a group of stocks that intends to reflect the performance of the entire stock market. Furthermore, potential biases in the results due to the bid-ask spread are reduced by the use of indices (Spyrou, Kassimatis and Galariotis, 2007). In addition, stock indices consist of large and small firms so the size effect could not play a significant role in this research. Several studies which tested the efficient market hypothesis used indices (Mun, Vasconcellos and Kishet, 1999; Ferri and Chung-ki, 1996; Schnusenberg and Madura, 2001; Spyrou, Kassimatis and Galariotis, 2007).

The indices used for this study are selected from Thomson Datastream using the daily total return index. The total return index is preferred over the price index because total return index data are adjusted for dividends. The total return index expresses the theoretical growth in value of a

shareholding over a specified period, assuming that dividends are re-invested to purchase additional units of the stock. The adjustment for dividends is important to determine the impact of an event. Otherwise a large decrease could be the result of a stock going ex-dividend and not because of the world event. Furthermore, the national indices are selected in their local currency, because otherwise exchange rates could influence the data.

4.3 Descriptive statistics

(31)

31

median, standard deviation as well as the maximum and minimum for the national indices and control indices of the events over the estimation period are showed. The next four columns presented the values of the skewness, kurtosis and Jarque-Bera and the p-value of the Jarque-Bera test.

The descriptive statistics of the national stock indices consist of 100 observations, because for this study is an estimation period of hundred trading days used. The descriptive statistics of the control indices composed of 1100 observations, because the estimation period of hundred trading days is joined for the eleven world events.

The Jarque-Bera test statistic shows if residuals are normally distributed and is composed using both the skewness and kurtosis statistics. It should below 5.99 in order for the residuals to be normally distributed (Brooks, 2002). The skewness analyses whether the distribution is symmetric around its mean value, this should be near 0 when data are normally distributed. The kurtosis determines how fat the tails of the distribution are and this value is for a normal distribution close to 3.

Table 5: Descriptive statistics of the actual returns per event for the country and control indices over the estimation period

Nr. Event

National

indices Mean Median Max. Min.

St.

Dev. Skewness Kurtosis JB. P-value

1 DJ Mexico 0.0006 0.0000 0.0320 -0.0476 0.0127 -0.2814 4.0474 5.8912 0.0526* 2 DJ Thailand -0.0043 -0.0053 0.0570 -0.0608 0.0196 0.3909 4.0124 6.8177 0.0331** 3 FTSE Russia -0.0099 -0.0114 0.1628 -0.1470 0.0545 0.5672 4.5706 15.6400 0.0004*** 4 DJ Composite 65 -0.0003 -0.0004 0.0286 -0.0251 0.0089 0.1776 3.7582 2.9210 0.2321 5 DJ Spain 0.0017 0.0017 0.0201 -0.0215 0.0067 -0.3057 4.1552 7.1174 0.0285** 6 DJ Thailand 0.0006 0.0012 0.0300 -0.0356 0.0125 -0.3614 3.5042 3.2361 0.1983 7 DJ U.K. 0.0004 0.0000 0.0117 -0.0141 0.0049 -0.2194 3.2688 1.1035 0.5759 8 MSCI Pakistan 0.0012 -0.0004 0.0478 -0.0483 0.0170 0.2979 3.8639 4.5887 0.1008 9 DJ China -0.0008 0.0004 0.1121 -0.1207 0.0347 -0.1516 4.5609 10.5347 0.0052*** 10 DJ Composite 65 -0.0009 0.0000 0.0325 -0.0347 0.0127 -0.0135 3.3158 0.4185 0.8112 11 DJ India -0.0052 -0.0052 0.0885 -0.1353 0.0362 -0.1956 3.9590 4.4697 0.1070 Control indices MSCI EM -0.0006 0.0004 0.0653 -0.0816 0.0111 -1.5842 14.3288 6919.0785 0.0000** FTSE EU 300 - 0.0001 0.0002 0.0965 -0.0807 0.0125 -0.3465 14.5542 6699.0195 0.0000** S&P 500 -0.0002 0.0003 0.1096 -0.0946 0.0133 -0.2112 18.0352 11311.8377 0.0000**

***, ** and * indicate significant at the 1%, 5% and 10% level respectively (two-tailed test).

The most indices show relative identical variations of their indices, ranging from a value of 0.0049 for the DJ United Kingdom to a standard deviation of 0.0196 for the DJ of Thailand during the financial crisis. However, the FTSE of Russia, China and India show higher standard deviations ranging from a value of 0.0347 for the DJ China to a value of 0.0545 for the FTSE Russia. These indices are composed of significantly fewer firms than the other indices. The FTSE Russia contains only 15 firms.

(32)

32

rejected. The actual returns during the world events numbers 1, 2, 3, 5 and 9 and the actual returns of the three control stock indices show probability values for the Jarque-Bera test smaller than 0.1. The assumption that the residuals of returns of these world event indices are normally distributed is rejected.

Table I of the Appendix presents descriptive statistics of the abnormal returns of the eleven bad news world events during their event period. Six stock indices show negative abnormal mean returns, which imply that the world event has an average negative impact on the stock index during the event period. Furthermore, the stock indices of three bad news world events (i.e. 11 September attack, train

bombings in Madrid and the bankruptcy of Lehman Brothers) and all of the control stock indices show non- normally distributed abnormal returns.

In general, if data are not normally distributed it is not suitable to perform parametric tests. Although Brown and Warner (1980, 1985) state that non-normality is not an issue in the event study

(33)

33

5. Methodology

This section elaborates the methodology used in this study. In the first part of this section alternative methods to test the efficient market hypothesis are discussed. Secondly, a detailed description of the event study, used in this research is given. After that, statistical tests are described with the focus on the parametric t-test and the non-parametric Corrado rank test.

5.1 Discussion of alternative methods

There are various methods for testing the efficient market hypothesis. One of the two most usual methods to test the efficient market hypothesis is by forming winner (e.g. stocks with positive returns) and loser (e.g. stocks with negative returns) portfolios during an estimation period. If loser portfolios outperform winner portfolios during a following period, even the weak form of market efficiency does not hold (De Bondt and Thaler, 1985; Zarowin, 1990). The weak form of market efficiency suggests that all historical stock prices are reflected in current stock prices. Therefore, it would not be possible to predict future returns or profit from arbitrage opportunities.

Another common method to test the efficient market hypothesis is by obtaining daily returns for all shares list on a stock market. If the returns satisfy the selection criterion of, for example 10%, the returns are defined as an event. Daily returns are then examined following the event date and

compared with the stock’s average return over the entire sample period (Bremer and Sweeney, 1991; Cox and Peterson, 1994). In addition, these studies search for reversal patterns to examine if it is possible to predict future returns.

Several studies have investigated the efficient market hypothesis. The definition of an event differs between studies. Nevertheless, most of these studies make use of the same research methodology, namely an event study. An event study examines the influence of an event (i.e. new information) on the stock price. This paper distinguishes itself from previous studies which tested the efficient market hypothesis. Although, it makes use of the event study methodology, events are defined as negative world news and not as randomness selected abnormal returns. With this method is it possible to compare the influence of different negative news on stock markets. This study selects three types of negative news, defined as financial crisis, terroristic attack and natural disaster. The selected events are described in Section 3. This section discusses the event study methodology.

5.2 Event study

The event study serves as important purpose in capital market research as a way of testing market efficiency. It is a statistical method which determined the impact of a specific event on the stock price. The efficient market hypothesis states that security prices at any time fully reflect all available

(34)

34

diverge from its normal return to reflect the influence of the event. The difference between the actual and estimated return is defined as the abnormal return. Systematically nonzero abnormal returns that persist after entering of new information into the stock market are inconsistent with market efficiency. To investigate if an event has caused abnormal returns it is first necessary to determine the exact event date and establish the estimation and event period. Second, abnormal returns in the event period are calculated following the Market and Risk Adjusted model of Brown and Warner (1980, 1985). Finally, statistical tests are applied to test whether the actual returns differ significantly from the estimated returns.

5.2.1 Event and estimation period

An important assumption in an event study is to correctly identify the event date t = 0. This should be the exact date when the new information entered the financial markets. New information has not only impact on the event date, but also on the days surrounding the event date. Therefore, event studies use an event period includes all days that are affected by the event. In this study the event period will be 26 trading days and starts 5 days before the event date and ends 20 days after the event date. The same number of post event trading days for the event period are used by Spyrou, Kassimatis and Galariotis (2007), Cox and Peterson (1994) and Larson and Madura (2003).

The estimation period is the period over which the expected normal returns of the stock prices will be estimated. The estimation window should normally not overlap the event period otherwise the event return could influence the estimated returns (MacKinlay, 1997). On the other hand, the estimation period should be close to the event date to make the best estimation for normal returns.

(35)

35

5.2.2 Abnormal return

The first step in calculating abnormal returns is to compute actual returns. The actual returns are calculated using the continuously compounded method (e.g. Newton da Costa, 1994; Mun,

Vasconcellos and Kish 1999; Otchere and Chan, 2003; Ma, Tang and Hasan, 2005). A problem with the arithmetic return is that it is not symmetric, which means that the normality assumption no longer holds for the distribution. To avoid this problem, the actual returnR_itis calculated with the

continuously compounded method:













=

−1

ln

it it it

I

R

(1)

In this formula I_it equals the total return index of index i at day t.

The second step in calculating abnormal returns is to calculate the estimated return. This return is the expected return of the stock prices of the index if the event did not take place. Brown and Warner (1980, 1985) provide three different models to calculate the estimated return, which are the Mean Adjusted, Market Adjusted and the Market and Risk Adjusted Return model. The Mean Adjusted Return model assumes that the expected returns are equal to the average return in the estimation period. The Mean Adjusted Return model is consistent with the Capital Asset Pricing Model under the assumption that a security has a constant systematic risk and that the efficient frontier (i.e. the lowest possible level of risk for its level of return) is stationary (Brown and Warner, 1980).

(36)

36

The following formula is used to calculate the estimated return E

( )

Rit with the Market and Risk

Adjusted Return model:

( )

Rit i iRmt jt

E =

α

+

β

+

ε

(2)

In this formula R_mt is the return of the FTSE World Index at day t,

ε

_jt is an error term and

α

_iand

β

_i

are Ordinary Least Square (OLS) parameters of the market model. These parameters are calculated with the statistical program Eviews 6.0 by regressing estimated returns against market returns over the estimation window of 100 trading days.

The FTSE World Index is used as benchmark market portfolio. However, a true benchmark of the market portfolio includes all assets, i.e. both traded and non-traded assets. Since this market portfolio is not available, a proxy for this portfolio is used. Other indices, like country indices could not be used as the most suitable benchmark for well diversified investors because country indices are not fully diversified and include unique (country) risk. The FTSE World is assumed to be a well-diversified portfolio and include only market risk and no unique risk. Rational investors are expected to be well diversified and the FTSE World therefore is the most suitable benchmark index.

The third step in computing abnormal return of stock indices it to calculate the difference between the actual and the estimated return of index i at day t. These abnormal returns are calculated for the event period, where t takes values between -5 and +20 trading days.

( )

it it

it R E R

A = − (3)

In this formulaA_it, RitandE

( )

Rit indicate respectively the abnormal, actual and estimated return of

index return i at day t.

5.3 Abnormal Return Aggregation

(37)

37

Abnormal returns must be aggregated over time for stock indices to make a statistical analysis over the event period. The cumulative abnormal return (CAR) is necessary in case of multiple period event windows. To figure out for which period new information has influence on the stock market, various CAR windows are used. CAR is calculated as the sum of the individual abnormal returns over various time intervals (

τ

₁_,

τ

₂) according to MacKinlay (1997):

∑

=

2 , 1 2 , 1 ) ( τ τ τ τ t it i

A

CAR

(5)

Once the CARs have been calculated for all securities in the sample it is not possible to aggregate through time and across securities. Therefore, the cumulative average abnormal return will be measured. Cumulative average abnormal returns (CAARs) are useful to compare the influence of different events for a variety of time periods. The cumulative average abnormal return (CAAR) is formulated by the following formula:

∑

=

N i

CAR

N

CAAR

1 ) ( ) (1,2 1, 2

1

τ τ τ τ (6) In this formula ₍ ₎ 2 , 1τ τ

CAAR

is the average of the cumulative abnormal returns over time period

τ

₁to

τ

₂, with N days in the sample.

5.4 Statistical tests

After calculating abnormal returns, statistical tests are necessary to determine if abnormal returns significantly differ from zero. The statistical significance of the abnormal returns will be tested using both the parametric t-test and the non-parametric Corrado rank test.

5.4.1 Parametric test

The null hypothesis tests if the abnormal returns during the event days are equal to zero. This parametric test is widely used in event studies and uses time series, taken into account any cross-sectional dependence in the security specific abnormal returns (Brown and Warner, 1980). The following test statistic is used to test the average abnormal return on day

The Efficient Market Hypothesis in relation to bad news world events