The Financial Impact of Terrorism on West European Markets

School of Economics and Management

The financial impact of terrorism on West European markets

Master Thesis

Student name: Koen van Uden Student number: 1266258 Supervisor: dr. K.K. Nazliben Second reader: dr. F. Castiglionesi Submission date: August 2018

Abstract

The main focus of this paper is to study the financial impact of terrorism on West European markets. The dataset examined includes 69 terrorist attacks that targeted West European countries over an 18-year time period. An event study approach is used to measure the event day en post-event day reaction of national indexes and industries to terrorism. The results show that the average attack negatively affects the West European market by -0.2%, this impact is only transitory since the results show a significant positive 10-day cumulative abnormal returns. Terrorist attacks with more than 100 casualties have an even more negative impact on the West European markets and are more likely to have a permanent impact. The attacked country suffers the most financial damage and this damage is permanent. The results show that all industries experience negative abnormal returns on the day of the event, two industries show significant results. The oil, gas and water industry is struck the most and is the only one to be permanently damaged. Furthermore, plausible explanations for the results are discussed from a behavioural finance perspective.

Keywords: Terrorism; Financial markets; Event study methodology; Reversal effect; Investor sentiment

Table of Contents 1. Introduction ...... 4 2. Literature Review ...... 7 3. Hypotheses ...... 18 3.1 Hypothesis – Event day ...... 18 3.2 Hypothesis – Post-event ...... 19 3.3 Hypothesis – Struck country ...... 20 3.4 Hypothesis – Industry analysis ...... 21 4. Data and Methodology ...... 22 4.1 Data ...... 22 4.1.1 Terrorism ...... 22 4.1.2 Exchange Markets ...... 23 4.2 Methodology ...... 25 4.2.1 Market reaction – Event day ...... 25 4.2.2 Market reaction – Post-event ...... 28 4.2.3 Market reaction – Struck country ...... 29 4.2.3 Market reaction – Industry analysis ...... 30 5. Empirical Research Findings and Discussion ...... 31 5.1 Results market reaction – Event day ...... 31 5.2 Results market reaction – Post-event ...... 37 5.3 Results market reaction – Struck country ...... 41 5.4 Results market reaction – Industry analysis ...... 44 6. Conclusion ...... 47 6.1 Research objectives: Summary of findings and conclusions ...... 47 6.2 Contribution and Recommendations ...... 49 6.3 Limitations...... 50 7. References ...... 52 8. Appendix ...... 56

1. Introduction

Terrorism is one of the biggest threats to humanity, property and the economy. As this article is written, recent terrorist attacks include the 2017 Barcelona attack, where 16 people were killed and 152 injured, the 2017 London Bridge attacks in which eight people were killed and 48 injured, the 2017 Manchester Arena bombing in which 22 people died and 250 were injured. Terrorism has a significant presence in the life of our society, which cannot be eliminated completely. It creates fear and it evokes negative feelings. The greater the fear the more the uncertainty in the stock market, which negatively affects stock prices. In addition, terrorism negatively influences the mood of an investor, which negatively affects the investing behaviour of an investor. The National Consortium for the Study of Terrorism and Responses to Terrorism (START, 2017) states that terroristic activity in West Europe has increased over the last decade, which makes terrorism a very topical field to study.

Given that it is easy for investors to buy and sell stocks, the revelation of new information will results in a high sensitivity to their prices. These price changes are observable when unanticipated events occur, especially adverse shocks such as large scale macroterrorism, which are terrorism incidents causing more than $1 billion of loss or 500 deaths (Woo, 2003). For instance, on the day of the September 11 attacks, the MSCI World Index lost 1.98% of its value, even though the U.S. stock market was closed. There has been much written about the direct costs of the September 11 attacks. The International Monetary Fund estimated the total direct cost from short-term effects following the September 11 attacks to be $21.4 billion (Richman, Santos, & Barkoulas, 2005). Lenain, Bonturi and Koen (2002) estimate the combined direct costs of the September 11 attacks for the private sector, the government enterprises and the rescue operations to amount to 27.2 billion USD. However, much less is known about the indirect costs of terrorism. Some have argued that indirect economic costs are driven by a decrease in consumer/investor confidence or other macroeconomic factors (Buesa et al. 2007). Others suggest that it may be driven by fear, which negatively affects investor sentiment and drives down stock prices (Lerner et al. 2003). Ultimately, an understanding of the financial impact of terrorism and the nature of its effects – whether the financial impact is temporary or permanent – is needed for investors to arm themselves against terrorism risk.

The objective of this paper is to study the financial impact of terrorist attacks on West European markets. I examine the investors’ reaction to 69 terrorist attacks that targeted Belgium, France, Germany, Italy, Netherlands, Spain and United Kingdom that took place in

4 the period 2000 up to and including 2017. This will be done by calculating the abnormal returns of the national indexes of the selected West European countries, using daily data. I expect that the stock price changes will reflect the financial impact of terrorism, since the efficient market hypothesis states that a market in which prices fully reflect available information, prices will change only when new information arrives, such as a terrorist attack. Even if the market does not fully reflect the available information – as suggested by behavioural finance – a terrorist attack might cause a shift in investor sentiment which has as consequence that the stock prices change regardless.

This paper is not the first study of the impact of terrorism on stock markets, the majority of previous work focussed on terrorist attacks with a large scale, such as the September 11th attacks (Abadie & Gardeazabal, 2008; Ito & Lee, 2004; Johnston & Nedelescu, 2006). This paper does not only focus on large scale event, but uses multiple terrorist attack from 2000 up to and including 2017. Furthermore, this study does not focus on a single nation or industry as is mainly done in the past (Drakos, 2004; Araña & León, 2008; Enders & Sandler, 1991). This study measures the event day and post-event day market reaction for both the national indexes and the industries. It uses a simplified empirical method to measure the impact, which makes it easy to replicate in another setting. This paper is also one of the few to study the link between terrorism and the behaviour of stock markets. This study is unique in three important dimension. Firstly, this study is unique since it is the first to examine the driver “targeted country”, it examines whether the impact of a terrorist attack in a certain West European country will have an increased impact on that country’s stock market when compared to the other West European stock markets. For example: does the metro bombing in Spain, 2004, has an increased impact on the IBEX-35 when compared to the FTSE 100. Secondly, this study is one of the few to examine the impact of terrorism on industries. Thirdly, I examine the reaction for both the national indexes and the industries to seven terrorist attacks with more than 100 casualties. It is expected that an attack with more than 100 casualties will show the impact on national indexes more clearly. This is due to the fact that people perceive the loss of humankind losses the worst (Karolyi & Martell, 2006). With this new information on this topic, investors can diversify their portfolio better against terrorism risk. Furthermore, this study uses an updated database, which automatically updates the current existing literature. This is valuable since the total number of terrorist attacks has increased, and the stock market changes over time, so there might be a different impact of terrorist attacks on the stock market than a few years ago.

My results indicate that terrorism has a significant financial impact on West European markets, this is supported by previous research within this field. On the day of the event, the average terrorist attack affects the West European market significantly negative by -0.2% which corresponds to an average loss of €42.3 million of market value per company per attack. For countries separately, terrorism has a significant negative impact on the national index of: the United Kingdom, Spain, France and Italy. Terrorist attacks with more than 100 casualties have an even more negative impact on the West European national markets. I also uncover significant positive 10-day cumulative abnormal returns, this indicates that there is a short-term reversal effect. This suggests that investors in West European markets overreact to terrorism. In other words, the impact of terrorism on the stock market is only transitory. All West European indexes show positive 10-day CAARs, four out of the seven listed are significant impacted, including Spain, France Belgium and Italy. The financial impact of terrorist attacks with more than 100 casualties is more likely to be permanent. Furthermore, the index that only includes the market reaction of the targeted country per attack has more negative returns than the average West European country and does not show a short-term reversal effect. This suggest the attacked country suffers more financial damage and this damage is permanent. Striking is the fact that the targeted country suffers less financial damage from terrorist attacks with a large number of casualties. Lastly, the market reaction of the industries is tested. The results show that all five industries experience negative abnormal returns on the day of the event, two industries show significant results. The oil, gas and water industry is struck the most and is the only one to be permanently damaged. The other industries show positive 10-day CAARs and thus show a short term reversal effect, which indicates an overreaction by investors. At last, it can be stated that the differences in financial impact of terrorist attacks across West European markets is affected by investor/market sentiment.

To study the financial impact of terrorism on West European markets, first the existing literature on the impact of terrorism on countries, industries and economies will be reviewed. This literature review will be the base for creating the hypothesis in this paper. Then each hypothesis is explained and linked with the existing literature. Thereafter the data will be discussed, followed by the methodology that is used. Chapter 5 will discuss the empirical research findings and compare them to the existing literature. Finally the conclusion will be given, this includes a summary of the findings and a conclusion per research objective. It will also contain the contribution to the existing literature, its recommendations to investors and the limitations of this study.

2. Literature Review

This literature review will examine the drivers of the economic costs surrounding a terrorist attack. A distinction can be made between these economic costs, namely direct costs and indirect costs. Direct economic costs of an attack include the costs that are immediately visible, such as the consequences of the loss of property and the loss of human lives. Indirect economic costs include the costs that arise through fear which decreases consumer and investor confidence, it also includes macroeconomic impacts. Another part of the indirect economic costs is the fact that investors sell their shares if a terrorist attack is perceived as bad news, which drives down stock prices, since investors update their beliefs towards new information. This literature review will focus on the indirect costs surrounding a terrorist attack, since the main area of interest in this paper is the investors’ reaction to terrorism. I will also explore the drivers behind the financial costs of a terrorist attack using a behavioural finance perspective. The following part firstly reviews the direct costs involved with terrorist attacks. Then the focus is narrowed down to the indirect costs of terrorism, these costs will be discussed on industry level and country level. Finally the behavioural finance perspective of the negative impact of terrorism on the stock market will be reviewed, in particular the changes in investor sentiment. The last subsection of this literature review extracts the main messages of the studies presented and the main lessons we can learn from my study.

Buesa et al. (2007) examine the direct economic costs of the Madrid train bombings in 2004 with regards to the economy of Madrid. The total loss of this terrorist attacks to the Madrilenian economy amounts to 212 million euros. This total loss should be considered a minimum, since Buesa et al. use a conservative way of measuring the costs. This paper does not take into consideration the loss in operating revenue due to the damage to the trains, even though Buesa et al. define direct costs as “costs that are an immediate consequence of the attacks and … losses due to pauses or delays in the economic activities affected directly or indirectly by the attacks”. Buesa et al. state that the direct economic costs are only a small proportion of the costs involved with the attack, however they did not measure the total costs. Brück & Wickström (2004) agree with Buesa et al., they state that the direct economic costs of terrorism are the most visible costs, but the fear and indirect costs cause the most damage to the economy in the long term. This is especially true for the costs of the September 11, 2001 attacks. Lenain, Bonturi and Koen (2002) estimate the combined direct costs of the September 11 attacks for the private sector, the government enterprises and the rescue and clean-up operations to amount to 27.2 billion USD.

Lenain, Bonturi and Koen end their study with the remark that the direct costs involved with the September 11 attacks are not even close to the indirect costs. Chen and Siems (2004) confirm this remark, they investigate the impact of the September 11 attacks on global capital markets. With the use of an event study methodology they conclude that the S&P 500 had negative abnormal returns on the day of the event of -4.84%, significant at the 1% level. This percentage can be expressed in numbers, if we multiply the total market value of all S&P 500 companies in 2001 – around 10.000 billion USD – with the abnormal returns on the event day, we get a loss of 484 billion USD. Their analysis shows that the investors, consumers and firms that are not direct victims of the terrorist attacks still lose from terrorism. It also confirms the expectation of Lenain, Bonturi and Koen (2002) and Buesa et al. (2007) that the direct costs are only a small proportion of the total costs.

To finalize the findings on the direct economic costs of an attack, I will discuss the studies by Karolyi and Martell (2006) and Berrebi and Klor (2005). Karolyi and Martel study the stock price impact of terrorist attacks that target publicly traded companies. I consider these attacks to have a direct impact to the attacked firm, since these attack cause physical damage and directly affects the operating performance of the targeted firm. Karolyi and Martell find a drop in stock price for the targeted firms of -0.83% on the day of the attack, this corresponds to an average loss in market capitalization of 401 million USD. The mean abnormal returns for the competitors are also measured, the results show that the stock price of the largest competitor also drops on the day of the event. This shows that competitor of the targeted company also faces economic costs, even though they are not directly impacted by the terrorist attack. A drawback of this study is that Karolyi and Martell do not measure the stock price reactions of non-related companies. This could have shown the real difference in financial impact between the targeted company – that faces both direct and indirect costs – and a company that is only indirectly impacted. The abnormal returns of the competitor do not show the total indirect costs, since these companies might be positively affected by the direct losses of their competitors. Another finding of Karolyi and Martell is that human capital losses are associated with larger negative abnormal returns than the loss of physical assets. This might suggest that fear is an important driver behind the economic costs of terrorist attacks. Berrebi and Klor (2005) also study the financial impact of terrorism that targeted firms, but during the 1998-2000 Palestinian- Israel conflict. They find on the day of the event a negative reaction of the targeted company of -0.77%. This drop in price is close to findings by Karolyi and Martell, but is not even close if we exclude the defence-related companies. Berrebi and Klor find a drop in price of -4.58% for

8 non-defence-related companies, this implies that defence firms profit from an attack, were other firms suffer. Berrebi and Klor only include terrorist attacks in which someone other than the terrorist died. This could exclude important events, since Karolyi and Martell’s results show that kidnappings of company executives are associated with large negative abnormal returns.

The economic cost of fear is investigated by Becker and Rubinstein (2004). They conclude that fear negatively impacts the economy in the long run, since fear affects a persons’ utility in each state of nature. Becker and Rubinstein state that terrorism takes advantage of the fact that people are human and rational. Terrorism generates fear, even if the attack is small scaled and if the probability of another attack is small, it still has a substantial impact on the economy. It can be concluded that fear drives the indirect costs for the economy. These indirect costs vary in their distribution across countries, industries and time. Some industries are more sensitive to terrorist attacks than others and suffer more damage from these attack. Terrorism can also cause indirect costs through changes in demand. The transport industry and tourism industry show the largest reductions in demand (Drakos, 2004; Ito & Lee, 2004; Araña & León, 2008).

Drakos (2004) studies the impact of terrorism on the airline industry. He uses data from the period 2000-2002 to measure the impact of the September 11 attacks on 13 airline stocks across the world. The results show that the average airline stock lost 30% of its value. The September 11 attacks reduced the short-term demand for air travel with 27%, it also had as consequence that insurance companies reassessed the probability of another attack using airplanes which also lowered stock price valuations. It took airline stocks on average 84 trading day to recover to the pre-attack price level. The US airline stocks were struck harder than non- US airline stocks. This suggests that the struck country suffers more from a terrorist attack than other countries. Drakos also states that the systemic risk of airline stocks doubled and that the idiosyncratic risk also increased. This implies that the costs of raising capital for airlines increases.

The demand for transport is linked with the demand for tourism, therefore you could expect the tourism industry to be also negatively impacted by the September 11 attacks. Araña and León (2008) investigate the impact of terrorism on tourism demand. They use the same event as Drakos (2004), namely the September 11 attacks. Araña and León use two methods to measure tourism demand. The first being a calculation of the loss in revenue in the tourism industry. The results show that the Mediterranean tourism industry has 27% lower revenues in the months October, November and December of 2001 than the same months in 2000, even

9 though there was an upward trend in revenue in the tourism industry. The second method is a survey to measure the amount willing to pay extra for a safer holiday destination. The survey was conducted in the same month as the attacks. The results show that households were willing to pay up to 230 USD more for a safer holiday destination. Even though the survey was anonymous, I expect the results to be biased since the respondents only could make two choices, an expensive safe destination and a cheap risky destination. From this we can conclude that the September 11 attacks caused a decrease in tourism demand, especially more risky destinations suffered more damage. The main drawback of the papers of Drakos (2004) and Araña and León (2008) is the fact that the focus of these studies is narrowed down to one of the largest historical terrorist attack and therefore lacks the power to be generalized.

Enders and Sandler (1991) study the relationship of terrorism and tourism in Spain. They use multiple terrorist attacks across the time period 1970-1988. They estimate that on average over 140 thousand tourists are scared away by each terrorist attack in Spain. Enders, Sandler and Parise (1992) use an econometric analysis to measure the impact of terrorism on tourism. They use a sample of European countries to calculate the present value of the loss in tourism revenue. Between 1974 and 1988 Italy lost over 1.1 billion dollars, Austria lost over 4.5 Billion dollars in the same period. The loss of revenue due to terrorist attack in continental Europe as a whole amounted to 16.1 billion dollars. Even though these numbers are dated, they still show that there is causality between terrorism and tourism.

For investors it is useful to know whether an industries is sensitive to terrorism, so they can minimize terrorism risk by diversifying across different industries. Chesney, Reshetar and Karaman (2011) study the impact of 77 terrorist attacks on global stock market and on different industries, and therefore this study can be used to develop a diversification strategy for minimizing terrorism risk. They also analyse the impact of natural catastrophes and financial crashes in the same time period as the terrorist attacks, namely 1994-2005. The results show that 68% of all terrorist attacks has a negative impact on at least one stock market. The European stock markets are more sensitive to terrorist attacks than US markets. The industries that suffers the most damage from an attack are the insurance industry and the airline industry, the banking industry suffers the least amount of damage. This result is in contrast to financial crashes where the banking industry is struck the most. The other industries: aerospace, defence, pharmaceutical and oil/gas industries are both positively as negatively impacted by terrorist attacks and natural disasters. From this can be concluded that the banking industry can be used to diversify the terrorism risk, but this industry exhibits significant negative returns with regards

10 to financial meltdowns. The safest choice for diversifying is the US Treasury bond, followed by the commodity market. The main drawback of the results by Chesney Reshetar and Karaman is the fact that they include military bombings as terrorist attacks. Most papers including mine exclude these attacks since they do not deliberately target civilians or non-combatants, which is considered a main requirement for an event to be classified as a terrorist attack.

Orbaneja, Iyer and Simkins (2017) and kollias et al. (2013) both investigate the relation between terrorism and oil markets. They both do not back up the statement of Chesney, Reshetar and Karaman (2011) that terrorist attacks have a small impact on the oil industry. Kollias et al. (2013) investigate the effects of terrorism on relationship oil prices and stock indexes. Their results show that in some cases the relationship between oil prices and stock markets stay neutral and in some cases, correlations become negative. Kollias et al. conclude that terrorism causes shocks in oil prices which negatively impacts the industry. These shocks can be seen as another indirect economic costs of terrorism, since these shocks increase the price of oil, which lead to higher costs for different industries, because oil is the main energy source for the world (International Energy Agency, 2015). Orbaneja, Iyer and Simkins investigate the impact of terrorism on oil markets, by focussing on the risk factor distance between the attack and the nearest oil refinery in the Middle East and on the risk factor size. They find that both distance and size of the attack negatively impacts the oil returns by respectively 0.40% and 0.23%, both significant at the 1% level. The fact that the risk factor distance of an attack affects the impact of terrorism on the oil market might explain why Chesney, Reshetar and Karaman (2011) only find a small impact on the oil market. They use the FTSE Europe oil/gas index to measure the impact of terrorism on the oil industry, this index is affected less by the risk factor distance, since the Middle East holds about half of the world oil reserves (Orbaneja, Iyer and Simkins, 2017).

As indicated before, the impact of terrorism differs across industries, which makes it possible to diversify the terrorism risk. The following part is about the indirect economic costs of terrorism across national markets. Eldor and Melnick (2004) analyse the impact of Palestinian terrorist attacks on the stock market of Israel. They use data that consists of 639 terrorist attacks that happened in the time period 1990 till 2003. Eldor and Melnick find that the stock market of Israel is negatively impacted by -2.3% through terrorist attacks. Eldor and Melnick also investigate whether a terrorist attacks has permanent or transitory impact on the Israel stock market. They conclude that the terrorist attacks after September 27, 2000 had a permanent effect

11 on the stock market. Their results also show that the higher the number of people killed and injured the more likely the impact of terror is permanent on the stock market. A flaw in the study by Eldor and Melnick is the fact that they use the day after the attack as the event day, this means that they measure the impact of the attack not on the same day as the terrorist attack. They justify this decision with the fact that it takes time for the market to show a reaction to an event. I agree with the fact that the market needs some time to reflect the news, however the market has enough time to react on the same day as the attack if the terrorist attack takes place in the morning. Therefore it is possible that the findings of Eldor and Melnick do not represent the real impact of terrorism on the Tel Aviv 100 Index. The main finding of Eldor and Melnick is that markets did not become desensitized to terrorist attacks, which implies that the impact does not weaken with more attacks. Enders and Sandler (2000) findings show similar results. They investigate whether the impact of terrorist attacks depends on seasons. They conclude that the impact of terrorism is not sensitive to seasons but that the impact depends on the number of incidents in the very recent past. Enders and Sandler results show that the financial impact of terrorism increases with the number of incidents.

In contrast to the previous papers that investigate the impact of terrorism on a national level, Arin, Ciferri and Spagnolo (2008) do not focus on only one country, they measure the impact of terrorism on six different financial markets. They use daily date from the UK, Thailand, Turkey, Spain, Israel and Indonesia. The terrorism data consists of a daily terror index, this index is not specified in this paper and makes it therefore hard to replicate. Nevertheless, Arin, Ciferri and Spagnolo find statistically significant causality effects between terrorist attacks and all six countries. The results show that the European countries are less sensitive to these events. This implies that the investor in Spain and the UK is more resilient to terrorist attacks.

The previous mentioned papers all agree on the fact that terrorism negatively affects stock markets, including national indexes and industry indexes. Why these indexes show negative returns after an attack has not fully been explained. The supply and demand of a single stock determines its price. The fact that a company’s stock price decreases after an attack even though it is not directly affected by the attack is a weird phenomenon. The logic of demand and supply states that this price drop can be explained by the fact that supply must have increased or the demand must have decreased. Karolyi (2006) state that terrorist attacks create fear, the greater the fear the more the uncertainty in the stock market, this has as result that investors are more

12 likely to flee to different financial instruments such as bonds. This implies that fear causes a negative demand shock which lowers the price of a stock. The following section will explain the behavioural finance perspective of the negative impact of terrorism on the stock market, in particular the changes in investor sentiment.

Baker and Wurgler (2007) investigate the impact of investor sentiment on the stock market. They define investor sentiment as: “a belief about future cash flows and investment risks that is not justified by the facts at hand”. Investors make decision based on economic effects and on sentimental effects. Economic effects are influences from economic sources, such as an annual report or an oil discovery. An oil discovery can drop the price of oil which positively affects the airline industry, since it will have a lower cost of fuel which lowers the costs of flying. Investors can react rationally upon the news of the oil discovery by buying airline companies shares. Not all decisions of investors are based on economic effects, some are based on investors’ emotions. A terrorist attack is an example of a sentiment effect. A terrorist attack is not an economic based event but still impacts the financial market, since this event influences the emotions of investors. The fact that terrorism negatively impacts the stock market – as stated in the previous mentioned papers – can thus be explained by investor sentiment. The beliefs by investors about the future cash flows and investment risks, which affects the stock price, are thus not solely based on economic facts. Goldberg and Leonard (2003) state that many of the changes in the financial market are unrelated to economic announcements and past movements of stock prices. Economic news is only one source of information for financial markets. This suggests that investor sentiment is a driver for the changes in the financial market.

Lerner et al. (2003) investigate the influence of terrorist attacks on investor sentiment. They use the aftermath of the September 11 attacks as a field study to understand the responses of investors to risk. With the use of surveys, Lerner et al conclude that investors show the emotions fear, anger, anxiety and sadness after the attacks. The results also indicate that 60 percent felt depressed because of the attacks. Furthermore the attacks have as consequence that U.S citizens feel less save, because of fear for another attack. A drawback of this paper is that the depression level and emotions are self-reported instead of determined by a psychologist. The main drawback is that the study does not include a pre-attack survey, which would have showed the changes in emotions and in depression level. Even though the study has some drawbacks, it can be concluded that terrorism negatively influences the mood of an investor.

Solomon, Greenberg and Pyszczynski (1991) state that changes in the mood of an investor have an effect on the investing behaviour of the investor.

Glaser and Weber (2005) use the September 11 attacks to analyse the influence of terrorism on the expected returns and volatility forecasts of individual investors. They use the results of a questionnaire before the attacks and after attacks to come to a conclusion. They find that the return expectations of investors increased after the attacks, this suggests a belief in mean reversion. The results of the questionnaire show that investors are aware of the fact that the negative shock of the stock market is caused by an overreaction, and is therefore mainly interpret as temporary rather than permanent. Glaser and Weber also report an increase in volatility forecast by investors, which suggests a higher future risk. The response to the questionnaire is of a homogenous group of investors, which might bias the results. Burch, Emery and Fuerst (2003) also investigate the September 11 attacks, but test the hypothesis that closed-end mutual fund discounts reflect small investor sentiment. Burch, Emery and Fuerst document a significant increase in discounts after the terrorist attacks, this suggests a negative shift in investor sentiment. The previous literature does not come to a conclusion whether closed-end mutual fund discounts reflect small investor sentiment. Elton, Gruber and Busse (1998) do not find evidence for this hypothesis. But this can be explained by the fact that they did not get to use the events of September 11 in their study, since this event was an unforeseen, negative and exogenous shock to the financial markets which avoids many of the issues of extant studies.

Market sentiment is the attitude of investors towards a financial market, this attitude is affected by emotions. In other words, stock prices do not reflect solely the business performance, since performance is not affected by a terrorist attack. Nikkinen and Vähämaa (2010) investigate this phenomena, they study the effects of terrorism on stock market sentiment. They use option prices to analyse these effects. Option prices contain information about investors’ expectations regarding the future asset value, which makes it a great instrument to measure market sentiment. The results show that terrorism has a strong adverse effect on the stock market sentiment. Their findings also suggests that terrorism has a transitory impact on stock market sentiment, even though market participants impose higher probabilities for further negative returns of the FTSE 100 index. All in all, we can conclude that terrorism has a negative impact on market/investor sentiment.

Elster (1998) investigate the influence of emotions on economics, he suggests that there is a link between rationality and emotions in decision making. Trading involves brain functions such as numerical computation and logical reasoning, these are often tempered by emotional responses, such as fear. Elster state that investors exhibit irrational biases that are attributed to psychological factors. Blasco and Ferreruela (2008) investigate another element of the behaviour finance that affects investors’ decisions, namely herd behaviour. This arises when investors copy the observed decisions of other participants in the markets. Blasco and Ferreruela examine the intentional herd behaviour of investors within international markets. From the seven investigated stock markets, the IBEX 35 (Spain) is the only market to exhibit a significant herding effect. In another study by Blasco (2005) he examines the reaction of the IBEX 35 to news. The results show that the Spanish market tends to react more quickly to general information – such as bad news – rather than to firm-specific information. This suggests that IBEX 35 investors tend to use more emotions in their decisions.

The mood and emotions of investors are also affected by seasons, and therefore seasons could affect investor sentiment. The effect of seasons on investor sentiment has not been elaborately discussed in the existing literature. Chan, Khanthavit and Thomas (1996) and Kramer and Weber (2012) are one of the few to investigate this. Chan, Khanthavit and Thomas use daily returns to identify seasonality. They find that stock prices are equal across calendar months, but excess returns are generated around the days of cultural important holidays. Kramer and Weber (2012) find that investors strongly prefer safe choices in the winter. They state that during the winter period depression rates increase, this is caused by less daylight. Enders and Sandler (2000) examine whether the impact of terrorism on the stock market is depending on seasonality. They conclude that the impact of terrorism is not sensitive to seasons but that the impact depends on the number of incidents in the very recent past.

The main findings of the literature review can be summarized as follows:

 The direct economic costs of terrorism are the most visible costs, however they are only a small proportion of the total economic costs involved with terrorism. The indirect economic costs of terrorism outweigh the direct costs for both small and large attacks. This is also confirmed by Karolyi and Martell (2006), since they state that human capital losses are associated with larger negative abnormal returns than the loss of physical assets.  Fear is the main driver behind the economic costs of terrorism. Economic costs vary in their distribution across industries, the transport industry, insurance industry and tourism

industry are the most sensitive to terrorism (Drakos, 2004; Ito & Lee, 2004; Araña & León, 2008). The banking industry is the least sensitive to terrorism, the defence industry is affected positively by terrorism (Chesney, Reshetar & Karaman, 2011; Berrebi & Klor, 2005).  The aforementioned literature has mainly focussed on single events, such as September 11, and on single countries (Buesa et al., 2007; Drakos, 2004; Eldor and Melnick, 2004; Araña and León, 2008). But overall, the literature states that the impact of terrorist attacks differs across nations. More advance economies have a smaller reaction to terrorist attacks than countries where the capital market is more inefficient (Chen & Siems, 2004). Arin, Ciferri and Spagnolo (2008) confirm this, they state that European countries are less sensitive to terrorism.  The previous literature does not come to a consensus on whether terrorist attacks have a permanent or transitory impact on the stock market. Eldor and Melnick (2004) state that the target type and attack type determines whether the effect is permanent or transitory. Their results show that terrorist attacks after September 27, 2000 had a permanent effect on the stock market. Glaser and Weber (2005) state that the negative shock of the stock market is caused by an overreaction, and is therefore mainly interpret as temporary rather than permanent. In contrast to Glaser and weber, the results of Karolyi and Martell (2006) do not show a short-term reversal. They conclude that the market did not overreact in their sample.  The negative impact of terrorism on the stock market can be partly explained by a negative shift in investor sentiment. Investor sentiment is a driver for the changes in the financial market (Goldberg & Leonard, 2003; Solomon, Greenberg, & Pyszczynski, 1991). Lerner et al. (2003) state that terrorism negatively influences the mood of an investor, which affects the investing behaviour of an investor. Burch, Emery and Fuerst (2003) and Nikkinen and Vähämaa (2010) also document a negative shift in investor sentiment. Most previous literature state that the seasonality effect does not affect investor sentiment or the impact of terrorism (Enders & Sandler, 2000; Kramer & Weber, 2012).

The knowledge about the economic consequences of terrorism is still relatively limited, despite the fact that terrorism will not decline in the foreseeable future. In other words, there is a continuing need for research on the impact of terrorism on the stock market, and the aforementioned review of literature supports this claim. This study will contribute to the

16 literature mentioned above in the way that it provides investors with an in-depth view on the economic impact of terrorism on West European stock markets. This study will not focus on a single event such as the September 11 attacks, but on multiple terrorist attacks from 2000 up to and including 2017. This study will measure the negative impact of terrorism on the national indexes of West European countries on the event day. I will also examine the post-event window, to get a clear picture on whether terrorist attacks have a permanent or transitory impact on the stock market. To dive deeper in the impact of terrorist attacks on stock markets, I will study whether a terrorist attack has an increased impact on the stock market of the attacked country. Furthermore, this study will examine the reaction of industries to terrorist attacks, which has not been done in a West European setting. The next stage of this research will consists of the hypotheses and their link to the existing literature.

3. Hypotheses

To answer the main research question, what is the financial impact of terrorism on West European markets, we need to know how the market reacts to an attack, how it recovers and what the drivers behind the financial impact are. Therefore I have developed a set of hypotheses to help us understand the main question.

3.1 Hypothesis – Event day There are several papers that study the effects of terrorism on financial markets. Abadie and Gardeazabal (2003) study the financial impact of the terrorist conflict in the Basque Country that started in the 1960’s. They measure the financial impact in terms of costs regarding the economic growth and in terms of lost market capitalization of public firms. They report that the conflict results in a 10 percent decrease in GDP per capita. Berrebi and Klor (2005) also study the financial impact of a conflict, but the 1998-2000 Palestinian-Israel conflict. They find on the day of the event a negative reaction of the market of -0.77% overall, but even a stronger reaction of -4.58% for non-defence-related companies. Chen and Siems (2004) use an event study to measure the effects of fourteen major terrorist attacks on global capital markets. They report negative returns on the day of the event ranging from -4.93% around the invasion of France, -2.65% around the Kent State shootings, to as large as -7.14% around the September 11 terrorist attacks. Similar negative returns are found for the UK, Turkey, Spain, Israel and Thailand in a study by Arin, Ciferri and Spagnolo (2008). Finally, Karolyi and Martell (2006) study the impact of terrorism on companies of the United States using attacks that target publicly traded companies. They use an event study analysis and conclude that there is a significant negative stock price reaction of -0.83 percent, this results in an average loss of 401 million dollars in the company’s market capitalization.

Most previous research conclude the same, that terrorist attacks have a significant negative impact on financial markets. However, there has been some disagreement regarding the magnitude of the impact. Although, previous studies are unambiguous about the impact of terrorist attacks on financial markets, this paper will take a natural starting point regarding the impact. Therefore, the first hypothesis (H1) tests whether terrorist attacks have a negative impact on the West European markets.

푯ퟏ: Terrorist attacks will have a negative impact on the national index on the day of the event.

3.2 Hypothesis – Post-event In the previous part I discussed different effects that may cause negative abnormal returns regarding the event day, those different effects may lead to different patterns in stock prices in the days following the event. Traditional finance theory states that all investors act rational and that all available information is reflected in the stock prices, and only new unexpected information will lead to shifts in prices. If an event reveals new information regarding a company, investors will update their beliefs accordingly and prices will change. If this is the case, prices will stay on that level even after the attack, unless new information is revealed. Behavioural finance states that investors are not fully rational. Which suggests that prices might not correctly reflect the available information. This implies that the market value of a company can be different from its fundamental value. With this in mind, a terrorist attack might cause an overreaction in the market leading to abnormal returns on the event day. The overreaction might be caused by fear among investors with an increase in uncertainty as result. If the anticipated abnormal returns on the day of the event are caused by an overreaction, I expect the prices to show a reversal effect in the days after the event. When it takes the market index a maximum of ten market days to get to its pre-attack level, we can speak about a short-term reversal effect.

In Chen and Siems’ study (2004) they find a large difference in time needed for various indexes to recover from terrorist attacks. They conclude that the U.S. capital market recovered in the past faster than most global capital markets. This is explained by the differences in strength of the banking and finance sector in the country which can provide liquidity and minimize panic. Karolyi and Martell (2006) expected to find a short-term reversal effect following a terrorist attack, but their results did not show that. They conclude that the market did not overreact in their sample.

Even though the post-event day hypothesis is not elaborately enough discussed in previous literature to come to a consensus, I expect that the negative abnormal returns show a reversal effect in the days after the event. Because I expect that the abnormal returns on the day of the event are caused by an overreaction of the investors rather than by newly available information regarding the companies in the index. With this in mind the second hypothesis (H2) is tested.

푯ퟐ: The national index will show a reversal effect within ten trading days following the terrorist attack.

3.3 Hypothesis – Struck country To dive deeper in the impact of terrorist attacks on stock markets, I will study whether a terrorist attack has an increased impact on the stock market of the attacked country. It is expected that the attacked country’s stock market is struck the hardest, since the attack is highlighted more in the attacked country. This is because the news about the attack is spread more and faster in the attacked country than in other countries, so the fear regarding the investors is greater. The greater the fear the more the uncertainty in the stock market this can have as result that investors are more likely to flee to different financial instruments such as bonds (Karolyi, 2006). It is expected that this adverse shock will lead to increased abnormal returns. Suleman (2012) studies the stock market reaction to news about terrorism, in his study he finds that news has a negative impact on stock return and increases the volatility. Another finding of Suleman is that the financial sector is affected the most by news on terrorism when compared to other sectors. Another study regarding news and terrorism is written by Melnick and Eldor (2010). They conclude that media coverage is the main channel through which terrorism produces economic damage. They state that the economic damage caused increases monotonically with the amount media coverage. With this given, the struck country’s stock market should be impacted more heavily since the media coverage will be higher.

Another effect that will cause the attacked country’s stock market to react more negatively is the perceived probability that there will be another attack in the previous struck country. This is because people asses the probability of a second attack higher for a previous struck country than for a different country. This is based on the observation that perceptions of the future are based on the recent past (Goodwin, Willson, & Gaines, 2005). Because of the increased perceived probability of another attack, I expect that the struck country’s stock market will show more negative abnormal returns than non-struck countries’ stock markets. In line with the increased perceived probability that there will be another attack, is the negative impact on the tourism sector. Enders and Sandler (1991) study the relationship of terrorism and tourism. They estimate that on average over 140 thousand tourists are scared away each year by a terrorist attack in Spain. In another study of Enders (1992) he calculates the loss of revenue due to terrorist attack in continental Europe. He estimated that terrorism costs continental Europe amounts 16.1 billion dollars. From this we can conclude that fear regarding terrorist attacks will increased negative impact on the attacked country’s stock market.

Looking at a broader index, rather than returns for a specific company, I expect that the stock market of the attacked country will suffer more damage than other stock markets in West

Europe. Therefore, the third hypothesis (H3) is tested.

푯ퟑ: The attacked country’s national index will experience larger negative returns following an attack than other West European national indexes.

3.4 Hypothesis – Industry analysis For investors, it is useful to know whether the financial impact of terrorist attacks differs across industries, so the investors can adjusts their trading strategy accordingly. Therefore, this study will analyse the impact of terrorism on different industries in West Europe. The following industries are analysed: financial service; Oil, gas and water; insurance; travel and leisure; aerospace and defence. Firstly, the industry reaction on the day of the event is analysed. Thereafter, the post-event window is analysed, since it is useful to know whether the impact is permanent or transitory. It is expected that the insurance industry and the travel and leisure industry are struck the most. Terrorist attacks often cause property damage and casualties, these attacks increase the costs for insurance companies, therefore it is expected that the insurance industry shows a negative reaction to terrorist attacks. Terrorism also causes fear, which leads to less tourism, which affects the travel and leisure industry negatively.

Enders and Sandler (1991) study the impact of terrorism on the tourism of Spain. They conclude that tourism industry of Spain would have had 1.5 times the revenue in 1988 if it was not for terrorist attacks. Enders, Sandler and Parise (1992) state that the loss of revenue due to terrorist attacks amounts to 16.1 billion dollars for continental Europe in the period 1974-1988. Chesney, Reshetar and Karaman (2011) investigate the impact of terrorist attacks across different industries. They conclude that 61 of the 77 attacks lead to significant negative results in at least one industry index. The insurance industry and airline industry are impacted the most, while the banking industry is impacted the least.

To conclude which industries are negatively impacted, the fourth hypothesis (H4) is tested.

푯ퟒ: Terrorist attacks have a significant effect on the financial service industry; oil, gas and water industry; insurance industry; travel and leisure industry; aerospace and defence industry.

4. Data and Methodology

4.1 Data The data used in this paper consists of two types of data, the terrorist attacks and daily stock prices of the national indexes. The data on the terrorist attacks is retrieved from the Global Terrorism Database (GTD), which is produced by the National Consortium for the Study of Terrorism and Responses to Terrorism (START, 2017). It includes data on worldwide terrorist attacks since 1970 and includes information on over 170 000 attacks. Most studies on terrorism use this database such as: Arin, Ciferri and Spagnolo (2008), Drakos (2010), Orbaneja, Iyer and Simkins (2017) and Khalil and Akhtar (2017). The daily data on stock prices from the main indexes of seven West European countries is obtained from Datastream.

4.1.1 Terrorism Firstly we need to define terrorism, the GTD defines terrorism as: the threatened or actual use of illegal force and violence by a non-state actor to attain a political, economic, religious, or social goal through fear, coercion, or intimidation. To qualify as a terrorist attack in the Global Terrorism Database the attack needs to fulfil the following criteria: the act must be aimed at attaining an economic, religious, social or political goal. There must be evidence of an intention to convey, intimidate, or coerce some other message to a larger audience (or audiences) than the immediate victims. The action must be outside the context of legitimate warfare activities, i.e. the act must be outside the parameters permitted by international humanitarian law (particularly the admonition against deliberately targeting civilians or non-combatants) (START, Global Terrorism Database, 2017). The terrorist attacks that are used in this paper are narrowed down to attacks that target developed economies, namely West European countries. This paper uses data from 2000 until 2017. Only terrorist attacks with five or more casualties are included, including both injuries and fatalities, since these attacks are expected to show negative abnormal returns. Table 1 shows a summary of the data retrieved from the GTD (its full form can be found in Appendix table A.1). There are in total 69 attacks, 28 of those targeted locations in Spain, 13 in both France and Germany, 10 in the United Kingdom, 3 in Belgium and 1 in both the Netherlands and Italy. Furthermore, the average attack causes 9 fatalities and 83 injuries. In case of more than one attack on the same day in the same country, the number of casualties of all the attacks on the same day are summed and count only as one attack. For example the November 13th attacks in Paris, there were seven different attacks on the same day,

22 for the purpose of this study I count this only as one attack. The attacks are also divided into quarters, the first quarter consists of attacks that happened between January and March, the second quarter consists of attacks that happened between April and June, the third quarter consists of attacks that happened between July and September and the fourth quarter consists of attacks that happened between October and December. Terrorist attacks that happened on a day the market is closed or that happened after half past four PM, are assigned to the next trading day. The reason being that the market can only reflect investors’ thoughts and information when the investors are informed and when the market is open.

Table 1: Descriptive statistics of the terrorist attacks

Country Number of attacks Total amount of fatalities Total amount of injured Belgium 3 36 361 France 13 240 939 Germany 13 25 181 Italy 1 0 50 Netherlands 1 7 12 Spain 28 226 2681 United Kingdom 10 88 1554 Total 69 622 5778 Table 1 shows the number of attacks, the total amount of fatalities and injuries per country.

4.1.2 Exchange Markets The daily prices of the national indexes of Belgium, France, Germany, Italy, Netherlands, Spain and United Kingdom with their index ticker respectively: BEL 20, France CAC 40, DAX 30, FTSE MIB, AEX, IBEX 35 and FTSE 100 are retrieved from Datastream. These indexes are the biggest national indexes per country based on market capitalization. The market returns are represented by the MSCI World index. The same time span as the terrorist attacks is used for the indexes, namely 2000-2017. The indexes are an average based on market capitalization, so if a terrorist attack negatively impacts the AEX, it should negatively impact the individual companies of the Netherlands. The industry indexes are constructed using the companies of the seven national indexes, appendix table A.2 shows the companies included in each industry index. Some companies do not have historical stock prices from the starting date 12-12-1999, if that is the case, these companies are excluded from the index till they are included in the index.

The returns are calculated on a daily basis using closing prices of indexes using the following formula:

pi,tpi,t−1 ri,t = pi,t−1 Where r denotes the return for terrorist attack i for time t and p indicates daily closing price of an index. For the industry indexes I calculated the returns per company per industry and took the average for each trading day. The returns used in this study are divided into three categories: estimation window returns, event day returns and post-event day returns. The estimation window starts 30 days before the attack and ends on the day before (푇−30 − 푇−1). The event day returns are the returns of the indexes when the news of the attack hits the market (푇0). The post-event day returns are the returns of ten market days after the attack (푇0 − 푇9). Figure 1 shows an overview of the time line.

Figure 1: Timeline for comparison Estimation window Event day Post-event day

푇−30 − 푇−1 푇0 푇0 − 푇9

Table 2 shows the descriptive statistics of the estimation window returns, the event day returns and the post-event day returns. The estimation window mean return is as expected, it is positive and close to 0.01% per day. The event day mean return shows a negative shock of -0.4% which is as expected. The post-event day mean return shows a positive shock of 0.15%, which may show the reversal effect.

Table 2: Descriptive statistics of event returns

(1) (2) (3) Estimation window Event day Post-event day returns returns return Mean 0.011% -0.402% 0.061% Std. Dev 0.015 0.010 0.013 Min -12.481% -5.657% -8.751% Max 11.491% 3.605% 7.081% Median 0.024% -0.237% 0.098% Observations 14490 483 4830 Table 2 shows the descriptive statistics for the mean return in the estimation window, on the event day and post-even day. The returns of all West European countries were taken together to estimate these statistics.

4.2 Methodology

4.2.1 Market reaction – Event day

The first hypothesis (H1) for this study is: “Terrorist attacks will have a negative impact on the national index on the day of the event”, this hypothesis will test the financial impact of terrorism on West European markets. The methodology for this study is an event study. The data used for this event study is panel data, since the data consists of returns from more than one moment in time and for more than one country.

A terrorist attack can only impact the stock market when investors are informed of the attack. This is mainly channelled through the news, which can take several hours after the actual attack took place. A reason for this delay can be that the authorities need to know what exactly happened, and the media needs to be present before the news can be spread to the investors (Kaplanski & Levy, 2010). Therefore terrorist attacks are assigned to the next trading day when the attack took place after half past four PM or on a day the market was closed. This day will be referred as the event day. Kaplanski and Levy (2010) and Borenstein and Zimmerman (1988) use the same way of reasoning in their articles.

I will calculate the abnormal returns (AR) separately for each country. In this way the differences in financial impact can be seen between the countries. The AR per country will also be divided into the quarters, to see if the AR depend on seasonality. It is expected that the attacks that happened in the fourth quarter, October till December, show increased AR for countries that are more religious and more dependent on tourism, such as Italy and Spain. To come to an overall conclusion whether West European firms are impacted by terrorist attacks, I also added the average of all West European national indexes to the data. This average is based on the market capitalization of the twenty biggest companies included in each national index on 31/12/2017. This new index will be referred to as the “West European index”.

This paper will use the Brown & Warner (1985) standard event study methodology. This methodology is based on the efficient market hypothesis, which states that in a market where prices fully reflect all available information, prices will only change when new information arrives. A common concern is that the event of interest is rarely unexpected. Usually, news about a merger or acquisition is leaked before the offering, which leads to price change before the event. Since terrorist attacks are unexpected events, it is appropriate to use an event study, because of the immediate price changes of stocks regarding terrorist attacks. However, there is the possibility that idiosyncratic effects are present. In traditional event studies, these effects

25 fade away when the number of events with non-overlapping event windows becomes large so that the resulting event-specific abnormal return truly captures the economic impact of the event on stock prices (Karolyi & Martell, 2006). In this study, there are 69 events in eighteen calendar years with only six clustered in time to yield overlapping event windows, so the results should not be affected by idiosyncratic effects.

The normal returns are calculated using the estimation window (푇−30 − 푇−1), which is a 30- day trading period prior to the terrorist attack. If the events overlap, an adjusted estimation window is used. This adjusted estimation window will consist of a 30-day trading period, excluding the event day and post-event window, of the previous attack. The estimation window is adjusted for six days, since then the stock prices should be rebounded (Brounen & Derwall, 2010). The estimation window is used to calculate the normal returns. Normal return (NR): returns that are expected in normal circumstances without an event. The normal returns are calculated using the Market Model. To calculate the normal returns, the return of the national index is regressed on the market return (푅푚푡) for each event i:

푁푅푖,푡 = 훼̂푖 + 훽̂푖(푅푚푡)

The abnormal returns (AR) are calculated using the following formula:

퐴푅푖,푡 = 푅푖,푡 − 푁푅푖,푡

The return (R) is the return on the event day. The event windows is (푇0) whereas, -1 for return one day before the incident, 0 for return at event day and 1 for returns after one day of the event. The average abnormal returns (AAR) are calculated over all events (N).

To test whether terrorist attacks have a significant impact on the stock market, I will use the following t-statistic:

퐴퐴푅푡 푇푆1,푡 = √푁 푠푡 Where s is:

푁 1 푠 = √ ∑(퐴푅 − 퐴퐴푅 )2 푡 푁 − 1 푖,푡 푡 푖=1

Besides the calculation of the event day AR, I will calculate the average West European market reaction to the seven terrorist attacks with the most casualties. It is expected that an attack with a high number of casualties will show the impact on national indexes more clearly. This is due to the fact that people perceive the loss of humankind losses the worst (Karolyi & Martell, 2006). With the introduction of an attack specific result there will be shed more light on the impact of terrorist attack on the financial market.

The common feature of dealing with daily returns on consecutive days, is the presence of anomalies such as autocorrelation and heteroskedasticity. The presence of autocorrelation of the residuals would bias the t-test or F-test, even if the beta coefficient is precisely measured. To detect the presence of autocorrelation, I will use the Breusch-Godfrey test (Breusch & Godfrey, 1980). This test uses the null hypothesis: no autocorrelation. If the null hypothesis is rejected the autocorrelation is confirmed. Schwert (1990) studies the importance of autocorrelation and concludes that studies that do not test for this anomaly show a weak tendency of movements in the aggregate stock returns. Autocorrelation can be explained by the fact that transactions are not done on a frequent base, also called non-synchronous trading (Fisher, 1966). Frequent trading reflects all available information on the end of the day, but non-synchronous trading does not reflect all information. When the market closes and not all information is reflected in the price since the securities are not traded frequent, the prices will only be adjusted on the next trading day. The Breusch-Godfrey test for autocorrelation will show whether the transactions are traded on a frequent base.

Similarly, the residuals should be tested for the presence of heteroskedasticity. When heteroskedasticity is present the variance of the error term is not constant in each observations but dependent on unobserved effects. One of the assumptions of Gauss-Markov needs the variance of the error term to be constant in order to meet the ideal conditions for a good OLS estimate. If this condition is not met and heteroskedasticity is present, the coefficient are still correct, but it is necessary to adjust the t-statistics and F-statistics. Breusch and Pagan (1979) developed a suitable test for this problem, namely the Breusch-Pagan test for homoskedasticity (constant error variances). This test uses the null hypothesis: homoskedasticity, and the alternative hypothesis: heteroskedasticity. If all assumptions’ requirements are satisfied, the OLS estimator is the best linear unbiased one.

4.2.2 Market reaction – Post-event

The main event window under study is the event day itself (푇0). However, it is useful to know whether the impact of terrorist attacks on the West European indexes is permanent or temporary. Therefore the second hypothesis (H2) for this study is introduced: “The national index will show a reversal effect within ten trading days following the terrorist attack”.

For this hypothesis, two post-event windows are introduced to examine how quickly and how well the market digests the news about an attack. Chen and Siems (2004) state in their study that the initial uncertainties regarding an event sometimes persist, which keeps the stock prices low and volatile (permanent shock). At other times, these uncertainties are reduced by the availability of new information that eases the tension of the market or these uncertainties are reduced by policy actions that promote greater market stability (temporary shock). The two post-event windows that are introduced are from the event day until the fourth day (푇0 − 푇4) and from the event day until the ninth day (푇0 − 푇9). For these two post-event windows, the cumulative average abnormal returns (CAAR) are calculated. If the CAARs for these periods are positive, the impact is only temporary.

Firstly the cumulative abnormal returns (CAR) are calculated using the following formula:

푇9

퐶퐴푅푖 = ∑ 퐴푅푖,푡

푡=푇1

Where CAR is the sum of the abnormal returns in the post-event window. The cumulative average abnormal returns (CAAR) are calculated over all events (N).

The statistical significance of the post-event period abnormal returns are computed using the test statistics described by Brown and Warner (1985):

퐶퐴퐴푅 푇푆 = √푁 2 푠 Where s is:

푁 1 푠 = √ ∑(퐶퐴푅 − 퐶퐴퐴푅 )2 푁 − 1 푖,푡 푡 푖=1

I will calculate the average 5-day and 10-day CAAR for each country separately, these CAARs will also be divided into quarters. In this way the differences in financial impact can be seen between the countries. Another advantage of calculating the CAARS separately for each country, is the ability to compare my results with Chen and Siems’ study (2004) in which they find a large difference in time needed for various indexes to recover from terrorist attacks. Also, the two CAARs are calculated for the portfolio of the average of the West European indexes. In addition to the 5-day CAAR and the 10-day CAAR, I will calculate the number of trading days it takes that the market index is back to its pre-attack level. Besides the calculation of the CAARs and the recovery time using the average of each attack, I will calculate the average West European post-event market reaction to the seven terrorist attacks with the most casualties.

4.2.3 Market reaction – Struck country Previously, the reaction of the national indexes to a terrorist attack is measured. To get an in- depth view of the financial impact of terrorism we need to know the drivers behind the impact.

Therefore the third hypothesis (H3) for this study is introduced: “The attacked country’s national index will experience larger negative returns following an attack than other West European national indexes”. Investors may be incentivised more to diversify their domestic portfolio when the national index of the attacked country show significant more negative returns. With this hypothesis we can also conclude whether fear is a direct driver behind the abnormal returns, since the fear will be greater to the country who is the victim of a terrorist attack.

To test whether a terrorist attack has an increased impact on the attacked country’s stock market, I will compare the AR of the struck country to the AR of other West European countries. A new dataset is introduced for this hypothesis, a dataset including only the AR of the struck countries. So, for each terrorist attack is the relevant abnormal return included in the dataset. In other words, for the attacks that targeted Spain, the AR from Spain are used. Comparing this with a dataset excluding the attacks on the struck country, we can conclude whether the attacked country suffers more damage than the other stock markets in West Europe. Using this dataset and the same event study methodology as mentioned for the event day hypothesis, the AR are tested on significance. Next to the calculations of the event day AR, I will also calculate the post-event market reaction of the attacked country. With this we can conclude whether the

29 attacked country’s stock market suffers more and whether it suffers longer. Furthermore, the hypothesis will be tested with regards to the seven attacks with the most casualties.

4.2.3 Market reaction – Industry analysis

The final hypothesis that is tested is (H4): “Terrorist attacks have a significant effect on the financial service industry; oil, gas and water industry; insurance industry; travel and leisure industry; aerospace and defence industry”. With this hypothesis we can conclude which industry is the driver behind the abnormal returns of the national indexes.

This hypothesis will be a combination of the first and second hypothesis, it uses the same methodology and assumptions as the first and second hypothesis, namely the standard event-study by Brown & Warner (1985). To test the impact of terrorism on different West European industries, I will calculate the AR on basis of the returns of the constructed industry indexes. The AR will be calculated separately for each industry. In this way the differences in financial impact can be seen between the industries. Next to the calculations of the event day AR, the post-event industry reaction is calculated. This is done by calculating the average 5- day and 10-day CAAR for each industry separately. With this we can conclude which industry is permanently affected and which is affected only temporarily.

5. Empirical Research Findings and Discussion

This section presents the results based on the approach designed in section 4. The first inevitable step is to gain an unbiased model to test for autocorrelation. The results from the Breusch- Godfrey test (Breusch & Godfrey, 1980) show that there is no autocorrelation present (appendix A.3). Furthermore, to have a good OLS estimate I tested whether the variance of the error term was constant, using the Breusch-Pagan test for homoskedasticity. The p-value is small enough to reject the null hypothesis of homoscedasticity, so there is heteroskedasticity in the national indexes sample (appendix A.4). So, for all further results the error terms are corrected for heteroskedasticity and are thus robust.

5.1 Results market reaction – Event day Table 3 lists each national index of West Europe with its abnormal returns on the event day and the following five trading days. To indicate whether the index returns are different from their expected means, the table includes t-statistics. These statistics indicate how significant the AR are. One could think of the t-statistic as a measure for economic impact. It tests whether the terrorist attacks that happened in the period 2000-2017 have an impact on the national indexes of West Europe, measured by the deviation of the index returns from their average. If it is expected that a terrorist attack has negative consequences for an index, one would expect to see a significant return deviation. The same logic holds for further results, such as cumulative average abnormal returns and for a single terrorist attack.

Table 3 reports that all national indexes show negative AR on the day of the event but only four out of the seven listed are significantly impacted by terrorist attacks. Both the United Kingdom and Spain as well as France show negative AR on the day of the event (or the immediate trading day), respectively -0.3%, -0.3% and -0.2, significant at the 5% level. The national index of Italy also shows a negative AR on the day of the event of -0.3%, but on a 10% significance level. The market capitalization based index of the seven West European indexes shows average AR of -0.2%, significant at the 10% level. This is in line with the first hypothesis. Furthermore, table 3 reports the AR of the five days following the attack. These post-event day AR show us whether a short-term reversal is present, as one would expect if the event day AR is caused by an overreaction. The average index, the West European index shows positive AR on the first day after the event of 0.2%, significant at a 10% level. This AR evens out the

Table 3: Mean abnormal returns Mean abnormal returns National index t = 0 t = 1 t = 2 t = 3 t = 4 t = 5 NLD -0.002 0.002 -0.001 0.002 -0.001 -0.001 (-1.405) (1.474) (-0.755) (1.209) (-0.659) (-0.333) UK -0.003** 0.001 -0.0004 0.0003 0.0002 0.001 (-2.281) (1.315) (-0.371) (0.276) (0.212) -0.614 ESP -0.003** 0.002 0.001 0.001 0.001 -0.001 (-1.963) (1.130) -0.349 (0.472) (0.623) (-0.629) BEL -0.001 0.001 -0.001 -0.00004 0.002* -0.001 (-0.620) (1.266) (-1.079) (-0.043) (1.814) (-0.697) FRA -0.002** 0.002 0.0003 0.001 -0.0002 0.0001 (-2.021) (0.997) -0.191 (0.517) (-0.123) -0.04 ITA -0.003* 0.004** -0.00001 0.001 0.001 -0.001 (-1.708) (2.314) (-0.007) (0.519) (0.567) (-0.589) GER -0.002 0.003* -0.0004 0.001 0.001 -0.001 (-1.544) (1.995) (-0.253) (0.729) (0.507) (-0.525) ALL -0.002* 0.002* -0.0003 0.001 0.0003 -0.0003 (-1.779) (1.643) (-0.195) (0.682) (0.221) (-0.254) Table 3 shows the mean abnormal returns for all West European countries in the sample and an average based on market capitalization (ALL). The robust standard errors are double clustered. ***, **, and *, represent coefficients that are statistically different from zero at the 1%, 5%, and 10% levels respectively. Day 0 is the day of the attack, or the first trading day immediately after the attack. The market return is the MSCI World return.

negative AR on the day of the event. In other words the market reaction on the day of the event is caused by an overreaction since the market shows a short-term reversal effect. Italy and Germany also show positive AR on the first trading day after the event, the AR amounts to 0.4% and 0.3% at a significance level of 0.5% and 10% respectively. Striking is the fact that the national index of Italy and Germany overcorrects the negative AR on the day of the event. Furthermore, only Belgium shows a significant AR in the post-event window excluding the first trading day.

To get more insight on the magnitude and importance of these numbers, I calculated the real economic value using the following calculation. The average market value of a company that is included in a West European index is €21.1 billion. Multiplying this times the average AR on the day of the terrorist attack, results in a loss of €42.3 million of market value per company per attack.

The seasonality effect is also measured, table A.5 of the appendix shows the results regarding this effect. From this table can be concluded that the fourth quarter, the months October, November and December show increased AR on the day of the event. Five out of the seven investigated countries show negative AR significant at the 5% level if the attack happened in the fourth quarter. In addition, these negative AR are twice as big as calculated in table 3, the West European index shows negative AR on the day of the event of -0.5% also significant at the 5% level. Spain and Belgium do not show significant AR in the fourth quarter, Spain shows negative AR of -0.6% significant at the 5 % level in the first quarter. Belgium does not show significant AR in any of the quarters.

Table 4: Average market reaction per attack Event day 5-day 10-day Days to Terrorist attack Event date AR CAAR CAAR rebound Madrid Train Bombing 11-3-2004 -0.019*** -0.028* -0.046** 25 (-2.614) (-1.699) (-2.061) London Bombing 7-7-2005 -0.014*** -0.010 -0.008 2 (-2.853) (-1.559) (-1.125) Paris Attack 13-11-2015 0.001 -0.001 0.006 2 (0.476) (-0.053) (0.179) Brussels Bombing 22-3-2016 0.002 -0.008 -0.017* 7 (0.518) (-1.365) (-1.673) Nice Attack 14-7-2016 -0.001* 0.005 0.021* 3 (1.723) (1.428) (1.833) Manchester Arena Bombing 22-5-2017 -0.008** -0.010* -0.020** 8 (-2.119) (-1.681) (-2.083) Barcelona Attack 17-8-2017 -0.006 0.002 -0.005 23 (-1.481) (0.256) (-0.401) Table 4 shows the West European index reaction to each of the seven attacks in the sample with more than 100 casualties. The starting date for the CAARs is the day of the event. The days to rebound represents the number of trading days it takes for the West European index to return to pre-attack level. The robust standard errors are double clustered. ***, **, and *, represent coefficients that are statistically different from zero at the 1%, 5%, and 10% levels respectively.

The results obtained from testing the market reaction to seven specific terrorist attacks are shown in table 4. These events are terrorist attacks with more than 100 casualties. For the purpose of the event day hypothesis, this section will only discuss the event day AR. These returns are those of the market capitalization based West European index. Four out of the seven attacks show negative AR on the day of the event. Both the Madrid bombing (2004) and the

London bombing (2005) cause the West European market to experience negative AR of respectively -1.9% and -1.4%, both significant at the 1% level. These two attacks cause the market to react almost ten times worse than the average terrorist attack. The Manchester arena bombings (2017) also negatively affects the market on the day of the event (or the immediate trading day) with -0.8% at a 5% significance level. The nice attack (2016) has as consequence that the average West European market shows AR of -0.1% at a 10 % significance level, which is only half of the impact that the average terrorist attack has.

The obtained results differ from the results from previous studies, such as Chen and Siems (2004). They investigated 14 terrorist/military attacks, 12 events experienced negative AR. The average attack causes -3.5% AR, which is much larger than the impact found in my study. The differences between the study done by Chen and Siems and the study done here are caused by differences in data. First of all, Chen and Siems only use 14 terrorist attacks with a big historic impact and an average of +1500 deaths, whereas my study uses all attacks that happened in West Europe with more than five casualties. Secondly, the time span of this study and Chen and Siems’ study differ a lot. They use the period 1915-2001, it is possible that terrorist attacks had more influence on the stock market in the past than nowadays. This could be explained by the fact that news only became available when the attack was very big, since smaller attacks were only covered by local media. Furthermore, the capital market is much more efficient nowadays, because of new technology. The communication is improved, which makes it easier to acquire information and implement these ideas into transaction. Therefore, it is possible that this study and Chen and Siems their study show different results. Chen and Siems also conclude that more advance economies have a smaller reaction to terrorist attacks than countries where the capital markets are more inefficient. This also might explain lower level of AR in this study, since this study only uses advanced economies in contrast to Chen and Siems.

In line with Chen and Siems’ study is Karolyi and Martell’s study (2006). Karolyi and Martell also examine the stock price impact of terrorist attacks, but focus only on attacks that targeted publicly traded firms. They find a statistically negative stock price reaction of -0.83% which is four times larger than the average result I found. Karolyi and Martell also calculate the mean AR in the days after the event, they did not find a significant result in the post-event day window, whereas I found a positive AR on the first day after the event. They also calculated the loss in millions of dollars in market value. Their calculations resulted in a loss of $484 million of market value per terrorist attack per firm, this is much larger than the €42.3 million

I found. These differences can be explained by the fact that Karolyi and Martell use a stricter definition for a terrorist attack, which causes the smaller attack to be excluded. They also use attacks in countries such as Nigeria and Colombia which might cause the market to overreact, if the statement of Chen and Siems: “advanced economies underreact to terrorist attacks” holds. The differences in results can also be explained by the fact that Karolyi and Martell only use terrorist attacks that target listed firms directly, which cause both direct and indirect costs. Even though, indirect costs of terrorism are significantly larger than the direct cost of terrorism. Each event in Karolyi and Martell’s study causes direct costs, whereas in my study the direct costs fade out, since I use indexes instead of single companies.

Arin, Ciferri and Spagnolo (2008) investigate the effects of terrorism on the following financial markets: Indonesia, Israel, Spain, Thailand, Turkey and the United Kingdom. The results show that the European countries are less sensitive to these events. This implies that the investor in Spain and the UK is more resilient to terrorist attacks. This might explain why terrorist attacks have less of an impact in this study than when compared to study by Chen and Siems and the study by Karolyi and Martell.

Furthermore, it is interesting to compare this study to a study that uses data of less efficient capital markets such as the study by Eldor and Melnick (2004). They analyse how the stock market of Israel reacts to terrorist attacks. They use data that consists of 639 terrorist attacks that happened in the time period 1990 till 2003. This is a lot more than the 69 terrorist attacks in West Europe in the time period 2000-2017. Eldor and Melnick (2004) find that the stock market of Israel is negatively impacted by -2.3% through terrorist attacks. They also conclude that the impact does not weaken with more attacks, which explains their higher level of impact. Another reason can be that the stock market is less efficient in Israel than in West Europe and therefore the impact might be bigger. The differences in impact found in this study and Eldor and Melnick their study could also be explained by the fact that both study use a different event day. Eldor and Melnick use the trading day after the attack as event day, whereas I use the day of the attack as event day if the attack happened before half past four PM.

Khalil and Akhtar (2017) recently studied the impact of terrorist attacks on the volatility of the Karachi Stock Exchange 100-index over the period 2004-2014. They conclude that terrorist attacks have a negative impact on the index, but the impact is highly dependent on target type and type of attack. Hostage takings and assassinations are the types of attacks that cause the impact on the KSE 100 to be the most negative, in this study these attacks are mostly

35 excluded, since Hostage takings and assassinations generally do not cause more than five casualties.

The seasonality effect on terrorism is measured by Enders and Sandler (2000). They conclude that the impact of terrorism is not sensitive to seasons but that the impact depends on the number of incidents in the very recent past. My study concludes the opposite, in the months October, November and December there is an increased negative impact. The financial consequences of terrorism is in the winter, when many religions have holy days, and tourism is up, greater than in the other seasons. The results stated in this study are in line with the behavioural finance. Kramer and Weber (2012) state that investors strongly prefer safe choices in the winter. In addition to this, Karolyi and Martell (2006) state that investors tend to seek for safer financial instruments after a terrorist attack. With this all in mind, it is no surprise that investors tend to overreact to an attack in the months October, November and December.

The results show that markets react differently to a terrorist attack. Spain, France, Italy and the UK all strongly react to such attacks, while their peers do not show a significant reaction. These differences can be explained by the fact that investors may exhibit irrational biases that are attributed to psychological factors. Elster (1998) suggests that there is a link between rationality and emotions in decision making. Trading involves brain functions such as numerical computation and logical reasoning, these are often tempered by emotional responses, such as fear. An element in this context is herd behaviour, this arises when investors copy the observed decisions of other participants in the markets. Blasco and Ferreruela (2008) examine the intentional herd behaviour of investors within international markets, the results show that the Spanish market exhibits a significant herding effect. They conclude that Spanish investment professionals are “satisfiers” not optimizers, these investors rely more on intuitive decisions. This is in line with my findings, since Spain is one of the four markets that overreacts to an attack. Blasco and Ferreruela also show that the German market tends to overreact to firm- specific information, this implies that the German investment professionals tend to use a more rational way of reasoning. This is also shown in my results, since Germany did not show a significant reaction to terrorist attacks. Blasco et al. (2005) examine the reaction of the IBEX 35 towards news. They state that in Spain, stocks tend to react more quickly to general information, such as bad news, rather than to firm-specific information. This implies that IBEX 35 investors tend to use more emotions in their decisions. These emotions, such as fear, are in line with the findings in my study, since the results in table 3 and 5 show an overreaction. Table

3 shows that the Spanish market heavily reacts to a terrorist attack and table 5 shows that this is an overreaction since the market shows a short-term reversal effect.

The negative AR on the day of the event show that stock prices do not solely reflect the business performance, since performance is not affected by a terrorist attack. This means that the stock prices are partly driven by emotions and thus not always represent the fundamental value of a stock. Nikkinen and Vähämaa (2010) examine this phenomena, they study the effects of terrorism on stock market sentiment. Market sentiment is the attitude of investors towards a financial market, this attitude is affected by emotions. Nikkinen and Vähämaa state that terrorism has a strong adverse effect on the market sentiment of the FTSE 100. Their findings also suggests that market participants impose higher probabilities for further negative returns of the FTSE 100 index. These findings may explain the significant negative AR for the UK and also the non-significant CAARs that are discussed in the next section. Overall we can conclude that the differences in financial impact of terrorist attacks across West European markets is affected by investor/market sentiment.

5.2 Results market reaction – Post-event While it is interesting to look at the immediate investor reaction to terrorist attacks, by analysing the abnormal returns on the event day, the cumulative average abnormal returns (5-day CAAR and 10-day CAAR) show the resilience of the indexes and the ability of the indexes to bounce back from the attacks. Table 5 shows the national indexes and their post-event reaction to the attacks. All West European indexes show positive 10-day CAARs, but only four out of the seven listed are significant impacted, this shows that there is a short-term reversal effect. Therefore, we cannot reject the hypothesis “The stock market will show a reversal effect within ten trading days following the terrorist attack”. Belgium shows the largest 10-day CAARs of 0.5% at a significance level of 10%. Both Spain and France have CAARS of 0.4% at a 10% significance level. Italy’s national index shows positive 10-day CAARs at a 5% significance level. The market capitalization based West Europe index also shows significant positive CAARs in the 10-day post-event window of 0.3%. I do not document significant CAARs in the 5-day window. The last column of table 5 shows the amount of trading days it takes for the national index to return to its pre-attack level. On average it takes the West European index five days to bounce back. The national index of the Netherlands and the United Kingdom both need

Table 5: Cumulative abnormal returns National index 5-day CAAR 10-day CAAR Days to rebound NLD -0.0002 0.002 7 (-0.576) (1.178) UK -0.001 0.002 7 (-0.921) (1.423) ESP 0.001 0.004* 5 (1.011) (1.766) BEL 0.001 0.005* 5 (1.063) (1.745) FRA 0.0002 0.004* 3 (0.618) (1.753) ITA 0.003 0.004** 2 (1.495) (1.971) GER 0.002 0.003 4 (1.346) (1.476) ALL 0.001 0.003* 5 (0.842) (1.771) Table 5 shows the 5-day and 10-day cumulative average abnormal returns for all West European countries in the sample and an average based on market capitalization (ALL). The starting date for the CAARs is the day of the event. The days to rebound represents the average number of trading days it takes for the market to return to pre-attack level (excluding the outliers). The robust standard errors are double clustered. ***, **, and *, represent coefficients that are statistically different from zero at the 1%, 5%, and 10% levels respectively.

on average seven trading days to recover from a terrorist attack. It is worth noting that these two countries both do not show significant positive CAARs, which show that the Netherlands and the United Kingdom are less resilience that the other markets. The national indexes with the significant positive CAARs all bounce back faster than the average national index. Italy is the market that needs the fewest amount of days to return to its pre-attack level, only two trading days. This leaves Germany as the odd one out, Germany is the only country with non-significant CAARs that needs fewer days to bounce back than average index.

Also the 5-day CAAR, 10-day CAAR and the days to rebound are calculated for the seven terrorist attacks in West Europe with more than 100 casualties. These results are shown in table 4. Whereas we see on average positive CAARs for the national indexes to all terrorist attacks combined, for the seven 100+ casualties attacks we see mostly negative CAARs. Two out of the seven attacks show significant negative CAARs, the Madrid train bombing and the Manchester Arena bombing. The Madrid train bombing has an AR on the event day of -1.9%,

38 a 5-day CAAR of -2.8% at a significance level of 10% and a 10-day CAAR of -4.6%, significant at a 5% level. This shows us that the backlash of the Madrid bombing holds for over ten days. In other words, the returns of the West European index keep on decreasing after an attack. It takes the national indexes on average 25 days to recover from the Madrid train bombings which explains the negative CAARs. The Manchester Arena bombing in 2017 causes AR of -0.8% on the day of the event, a 5-day CAAR of -1.0% significant at the 10% level and a 10-day CAAR of -2.0% significant at the 5% level. Whereas the same logic for this attack holds as for the Madrid 2004 bombings, it takes the market only eight trading days to return to pre-attack level after the Manchester bombing. Four out of the seven attacks have significant 10-day CAARs, where Madrid and Manchester are both already explained. Even though the 2016 Brussels bombing does not has a significant AR on the event day, it has a 10-day CAAR of -1.7% significant at the 10% level. This suggests the market did not show a negative reaction directly after the attack. The Nice 2016 attack is the only event out of the seven events that show positive CAARs. The Nice attacks have a negative financial impact on the day of the event, but this is corrected in the 10-day window. The 10-day CAAR for the Nice attack is 2.1%, this positive CAAR can be explained by the fact that it took the market only three days to bounce back to its pre-attack level. For the last column in table 4, the days to rebound column, we can conclude that the number of days it takes the market to rebound from an attack with more than 100 casualties, is larger than the number of days it takes the market to rebound from an average attack shown in table 5. Overall we can conclude that the financial impact of terrorist attacks with more than 100 casualties is more likely to be permanent, whether an average terrorist attack is more likely to have a transitory impact on the indexes. So for the 100+ casualties’ sample, the reversal hypothesis does not hold.

The seasonality effect is also measured for the post-event window, column 4 and 5 of table A.5 of the appendix shows the results regarding this effect. From this table can be concluded there is no reversal effect in the same quarter as the significant negative event-day AR. The only reversal effect that happens in the same quarter as the significant AR is the first quarter for Spain. Therefore it can be concluded that seasons do not affect the post-event market reaction.

Eldor and Melnick (2004) investigate whether terrorist attacks have a permanent or transitory impact on the Israel stock market. They conclude that terrorist attacks after September 27, 2000 have a permanent effect on the stock market, which contradicts with the conclusion in my paper. Even though, Eldor and Melnick state that their study has broader implication that extends to

39 western societies because of the democratic regime of Israel and their free markets, the differences in post-event reactions are significant. They also find that markets did not become desensitized to terrorist attacks, which implies that the impact does not weaken with more attacks. With this in mind and the fact that their study uses 639 terrorist attacks over a 13 year time period, it is possible to explain their findings on the permanent impact. This also could explain why I come to the conclusion that terrorist attacks have a transitory impact on the national indexes, since there only were 69 terrorist attacks across 7 countries and over an 18 year time period in West Europe that caused five or more casualties. Eldor and Melnick (2004) results also show that the higher the number of people killed and injured the more likely the impact of terror is permanent on the stock market. This is in line with my findings, since the seven investigated attacks with over 100 casualties have a more permanent character.

The paper on the impact of terrorist attacks on the Pakistan stock market also studies the market reaction in the post-event window (Khalil & Akhtar, 2017). Khalil and Akhtar find a statistically significant result that suggests that the national index of Pakistan current volatility is dependent of past volatility and on variance of past volatility. The results of Khalil and Akhtar’s study also suggests that the negative impact due to a terrorist attacks is absorbed in the short run, the Pakistan index recovers itself after a day. This is in line with the results in my study, table 3 shows that on average all West European indexes have positive AR returns on the first trading day after the event day, equal to the negative AR on the event day.

Chen and Siems (2004) conclude that out of the 14 events they investigate, 11 experience negative 6-day and 11-day CARs, the other 3 experience positive CARs. Three events experience significant negative CARs at the 1% level over the 6-day horizon, and two of these three events showed negative CARs over the 11-day event window. From this we can conclude that most terrorist attacks in their sample have a permanent effect on the stock market. One of the events that showed the significant negative CARs was the invasion of France by Hitler. These negative CARs are explained by the fact that during this time window, new information continued to have a negative effect on stock prices. The difference in the results of Chen and Siems’ study and my study, can be explained by the fact that they use periods of terrorist/ military attacks as one attack, which would lead to a chain of negative news, whereas my data only consist of terrorist attacks that happen on one day. Chen and Siems also show that the market becomes more resilient over time because of more efficient markets and more flexible and appropriate monetary and fiscal policies, which explains the more transitory nature of the market to attacks nowadays. Chen and Siems also compare the reaction of different global

40 stock markets to the invasion of Iraq into Kuwait. They conclude that the London stock market does not show significant 11-day CARs which is in line with my conclusion. But they also conclude that the Amsterdam and Frankfurt stock markets show negative CARs which contradicts with my non-significant CAARs for the Netherlands and Germany.

Glaser and Weber (2005) use the September 11 attacks to analyse the influence of terrorism on the expected returns and volatility forecasts of individual investors. They find that the return expectations of investors increased after the attacks, this suggests a belief in mean reversion by the investors. The results show that investors are aware of the fact that the negative shock of the stock market was caused by an overreaction. These results are in line with my study, since my results also show a temporary reaction to terrorism rather than permanent.

5.3 Results market reaction – Struck country The third hypothesis tests whether the struck country’s stock market is impacted more than the other West European stock markets. Table 6 shows the market reaction for the subset where 69 attacks are included with the returns of only the targeted countries, referred to as the “attacked index”. Table 6 also shows the market reaction of the seven countries excluding the terrorist attacks that targeted that country. The most important result from this table is the mean AR of the attacked index, this is -0.3% significant at the 5% level. From this we can conclude that the struck countries subset has more negative returns than the average West European country, which has AR of -0.2%, this is consistent with the third hypothesis. The results also show that the attacked index does not has significant 5-day and 10-day CAARs, which implies that terrorist attacks on the targeted countries have a permanent impact, since the market does not recover in the short run. It also takes the attacked index eight days to rebound to its pre-attack level, whereas it takes the average West European country five days. Furthermore, we can compare the results of the AR for all West European countries excluding the attacks on that specific country with those of table 3 that does include those attacks. The AR of the United Kingdom is -0.3% on the day of the event if we include all attacks, but -0.2% if we exclude the attacks that targeted the United Kingdom. Both Spain and France are negatively impacted by respectively -0.3% and -0.2% by all terrorist attacks on the day of the event, significant at the 5% level, but Spain and France are not significantly impacted if the attacks that targeted Spain and France are excluded. Striking is the fact that Germany does not show AR on the day of the event in table 3, where all attacks are included, but shows negative AR if we exclude the attacks

Table 6: Market reaction of attacked only versus attacks excluded National Event day 5-day Days to 10-day CAAR index AR CAAR rebound ATT -0.003** 0.003 0.007 8 (-2.016) (1.009) (1.338) NLDᵃ -0.002 -0.001 0.001 7 (-1.512) (-0.773) (0.786) UKᵃ -0.002* -0.001 0.003 4 (-1.861) (0.732) (1.494) ESPᵃ -0.002 -0.0001 0.003 7 (-1.097) (-0.562) (0.719) BELᵃ -0.001 0.002 0.005 3 (-0.442) (0.953) (1.398) FRAᵃ -0.002 0.0001 0.002 4 (-1.463) (0.046) (0.648) ITAᵃ -0.003* 0.003 0.005 4 (-1.650) (1.036) (1.054) GERᵃ -0.003** 0.0004 0.001 5 (-2.143) (0.631) (0.854) Table 6 shows the mean abnormal returns for the attacked only subset (ATT) and it shows the mean abnormal returns for all West European countries excluding the attacks on that specific country (ᵃ). It also shows the 5-day and 10-day cumulative average abnormal returns. The starting date for the CAARs is the day of the event. The days to rebound represents the number of trading days it takes for the market to return to pre-attack level. The robust standard errors are double clustered. ***, **, and *, represent coefficients that are statistically different from zero at the 1%, 5%, and 10% levels respectively.

that targeted Germany of -0.3% at 5% significance level. This can be explained by the fact that the terrorist attacks that targeted Germany have a smaller size, because the 13 attacks on German soil have an average of 14 casualties, whereas the average attack has 92 casualties. The Netherlands, Belgium and Italy all show no differences between table 3 and 6 regarding the AR on the day of the event. This can be explained by the fact that these three countries were only at most three times victim of a terrorist attack, which implies that these specific datasets are not that different. Furthermore, table 6 shows that there are no significant 5-day and 10-day CAARs. From table 6 we can conclude that the struck country’s stock market suffers more from a terrorist attack, because on the one hand the AR for the attacked index is higher than the West European index and on the other hand the negative impact is permanent. With this in mind you would expect to see terrorist attacks with more than 100 casualties, to cause more extreme AR for the attacked index. Table 7 shows that this logic does not hold.

Table 7 shows the market reaction of the national index of the country who is the victim of the attack. From the seven attacks that show significant negative returns in table 4 only the London bombing in 2005 still shows significant negative AR on the event day in table 7. This contradicts the previous stated conclusion: that struck country’s stock market suffers more from the terrorist attack. The Madrid train bombing in 2004 cause the IBEX 35 to show no AR, whereas the average West European country does show significant negative AR to this specific attack. The same goes for the Nice attack in 2016 and the Manchester arena bombing in 2017, both attacks do not cause the France CAC 40 and the FTSE 100 to show negative AR.

Table 7: Market reaction per attack (attacked countries only) Event day 10-day Days to Terrorist attack Event date 5-day CAAR AR CAAR rebound Madrid Train Bombing 11-3-2004 -0.008 -0.039 -0.048 21 (-0.977) (-1.669) (-1.349) London Bombing 7-7-2005 -0.013*** -0.014 -0.029* 4 (-2.693) (-1.353) (-1.870) Paris Attack 13-11-2015 0.002 -0.012 -0.01 2 (0.264) (-0.469) (-0.269) Brussels Bombing 22-3-2016 0.003 -0.008 -0.008 7 (0.401) (-0.851) (-0.535) Nice Attack 14-7-2016 0.001 -0.004 0.0139 2 (0.189) (-0.477) (1.027) Manchester Arena Bombing 22-5-2017 -0.007 -0.004 -0.013 7 (-1.402) (-0.426) (-0.989) Barcelona Attack 17-8-2017 0.003 -0.003 -0.006 23 (0.600) (-0.457) (-0.589) Table 7 shows the attacked country’s reaction to each of the seven attacks in the sample with more than 100 casualties. The starting date for the CAARs is the day of the event. The days to rebound represents the number of trading days it takes for the attacked country to return to pre-attack level. The robust standard errors are double clustered. ***, **, and *, represent coefficients that are statistically different from zero at the 1%, 5%, and 10% levels respectively.

Overall we can conclude that the only attack with more than 100 casualties that has a negative impact on the concerned stock market, the London bombing, has a permanent impact, since the 10-day CAAR is significantly negative. The other attacks with more than 100 casualties, do not have a significant impact on the concerned stock market. The answer to the third hypothesis needs to be specified in order to hold, namely: the struck country’s stock market suffers on

43 average more from a terrorist attack than the non-attacked country, but the investors of the struck country underreact to attacks with large number of casualties.

Furthermore, only Drakos (2004) addresses the fact that the struck country suffers more from a terrorist attack than other countries. He examines the impact of the September 11 attacks on 13 airline stocks across the world. The results show that the US airline stocks show more negative returns than non-US airline stocks on the day of the event, which is in line with my results.

5.4 Results market reaction – Industry analysis Previously, I analysed the impact of terrorism on the national indexes of West European countries. In this part I analyse the industries of those countries, to obtain an in-debt view of the financial impact of terrorist attacks. To gain an unbiased model I tested the industry data for autocorrelation. The results from the Breusch-Godfrey test (Breusch & Godfrey, 1980) show that there is no autocorrelation present (appendix A.6). Furthermore, I tested the presence of heteroskedasticity using the Breusch-Pagan test. Appendix A.7 shows that the p-value is small enough to reject the null hypothesis of homoscedasticity, so there is heteroskedasticity in the industry sample. So, for all further results the error terms are corrected for heteroskedasticity and are thus robust.

Table 7 shows that all five industries experience negative AR on the day of the event, two industries show significant results. The oil, gas and water industry is struck the most with -0.5% AR on the day of the event, significant at 1% level. This industry does not show a reversal effect, so the impact could be considered as permanent. The insurance industry is also struck significantly, with -0.1% at a significance level of 10%. The negative sign is as expected, since insurance costs increase with a terrorist attack. The impact of terrorism on the insurance industry is only temporary, since the 10-day CAAR is positive and significant at the 5% level. The positive 10-day CAAR outperforms the negative AR at the event day with 0.2%, this could be explained by the fact that the demand for insurance increases after an attack. Striking is the fact that the travel and leisure industry does not show significant AR at the day of the event. It was expected that this industry would suffer damage, since previous literature state that terrorism has a negative impact on tourism (Enders & Sandler, 1991; Baker, 2014; Enders, Sandler and Parise, 1992). The explanation for this could be that the constructed travel and leisure index only consists of six companies of which four are established after the year 2000.

Table 7: Market reaction of industries

Industry Event day AR 5-day CAAR 10-day CAAR Financial Service -0.002 0.003 0.007 (-0.999) (0.984) (1.629) Oil, Gas and Water -0.005*** -0.002 0.003 (-2.654) (-0.538) (0.542) Insurance -0.001* 0.0002 0.003** (-2.331) (1.391) (1.990) Travel and Leisure -0.002 0.002 0.010** (-0.763) (0.611) (2.053) Aerospace and Defence -0.0002 0.001 0.006** (-0.102) (0.247) (2.212) Table 7 shows the mean abnormal returns for five West European industries. It also shows the 5-day and 10-day cumulative average abnormal returns. The starting date for the CAARs is the day of the event. The robust standard errors are double clustered. ***, **, and *, represent coefficients that are statistically different from zero at the 1%, 5%, and 10% levels respectively.

This has as consequence that the index still has some idiosyncratic risks, which biasses the results. Table 7 shows that the travel and leisure industry has a significant 10-day CAAR of 1.0%, sigificant at 5%. The aerospace and defence industry does not show significant event- day AR, but does show a positive 10-day CAAR of 0.6%, significant at the 5% level. One of the explanations of the positive 10-day CAAR could be that terrorism induce an increase in government expenditures on defence, in order to decrease future terrorist attacks. The financial service industry does not show significant AR on the day of the event, which can be explained by the fact that the performance of the financial service industry is not affected by terrorist attacks. Overall, it can be concluded that industries react differently to terrorism. On the event day, the AR are mainly negative for all industries and the post-event window shows that the AR are caused by an overreaction. With regards to the fourth hypothesis, terrorist attacks have a significant negative effect on most industries but there is a short-term reversal effect.

The results support the conclusion of Chesney, Reshetar and Karaman (2011), they conclude that the insurance industry and airline industry are affected by the highest number of attacks, while the banking industry is struck the least. The oil/gas industry shows mostly negative returns. Overall, they find more negative reactions than positive for all indexes, which they explain by a decrease in consumer confidence and a fear for a possible economic slowdown. This is clearly shown in the transport industry, for example, a drop in demand for

45 flights. This can be linked with the drop in oil prices, which is partly dependent on the demand for air travel.

Orbaneja, Iyer and Simkins (2017) also find a negative relation between terrorism and oil markets. Kollias et al. (2013) find negative effects of terrorism on the oil price-stock index. They also find co-movement between CAC40, DAX and oil, which corresponds with my results, since all three indexes show a negative sign.

The study by Berrebi and Klor (2005) shows that defence related companies profit from terrorist attacks. My results do not show a positive abnormal return on the day of the event, but do show a significant positive 10-day CAAR which indicates a profit in the short run. The study by Berrebi and Klor only focuses on the financial impact of terrorism that targeted firms, which might explain the difference in results.

6. Conclusion

The overall aim of this study was to advance an understanding of the financial impact of terrorism on West European markets, particularly the reaction of the investor to terrorist attacks. The specific research objectives were:

 To measure the impact of terrorism on national indexes on the day of the event.  To identify a reversal effect, to conclude whether a terrorist attack has a permanent or transitory impact on the national indexes.  To identify a potential driver behind the impact of terrorism on national indexes, by measuring if the attacked country suffers the most financial damage from the attack.  To measure the impact of terrorism on West European industries on the day of the event and to analyse the post-event reaction of the industries.

This section will revisit the above mentioned research objectives, it will contain a summary of the findings and a resulting conclusion. The contribution of this study to the existing literature will be clarified. Recommendation will be made, based on the summary and on the conclusion. Additionally, the limitations of this study will be discussed.

6.1 Research objectives: Summary of findings and conclusions The first hypothesis measured the impact of terrorism on national indexes on the day of the event, the results showed that terrorism negatively affects the stock market. The market capitalization based index of the seven West European indexes (Netherlands, United Kingdom, Spain, Belgium, France, Italy and Germany) is significantly negative impacted by terrorist attacks. For countries separately, terrorism has a significant negative impact on the national index of: the United Kingdom, Spain, France and Italy. The months October, November and December show increased abnormal returns on the day of the event. This shows the seasonality effect and implies that investors tend to overreact to an attack during the winter, when many religions have holy days and tourism is up. Furthermore, it can be concluded that terrorist attacks with more than 100 casualties have a more extreme impact on the West European national markets, which is in line with the existing literature. The differences in financial impact of terrorist attacks across West European markets on the day of the event is affected by investor/market sentiment. Overall from my results and the previous literature can be conclude that terrorism negatively affects the stock market on the day of the event.

The second hypothesis investigated the impact of terrorism on national indexes in the post-event window, the results showed that there is a short-term reversal effect. The market capitalization based West Europe index shows significant positive CAARs in the 10-day post- event window. This indicates that investors in West European markets overreact to terrorism. In other words, the impact of terrorism on the stock market is only transitory. All West European indexes show positive 10-day CAARs, four out of the seven listed are significant impacted, including Spain, France Belgium and Italy. The results do not document significant CAARs in the 5-day window, even though it takes the average West European index five days to bounce back to its pre-attack level. For terrorist attacks with 100+ casualties the 5-day and 10-day CAARs are mainly negative, this indicates that the backlash holds on for over 10 days. Overall from my results and the previous literature can be concluded that the financial impact of terrorist attacks with more than 100 casualties is more likely to be permanent, whereas a small-scaled terrorist attack is more likely to have a transitory impact on the indexes.

The third hypothesis tested a driver behind the impact of terrorism on national indexes, it tested whether the attacked country suffers more financial damage. It can be concluded that that the attacked index – the index that only includes the market reaction of the targeted country per attack – has more negative returns than the West Europe index. The findings also suggest that there is no short-term reversal effect, since the attacked index does not has significant 5- day and 10-day CAARs. For countries separately, the United Kingdom, Spain and France are less negatively impacted if the attacks that targeted those countries are excluded. Striking is the fact that Germany is not significantly impacted if all attacks are included, but does show significant negative AR if we exclude the ones that targeted Germany. The other countries do not show any differences if we exclude the attacks that targeted that specific country. From the seven attacks that caused more than 100 casualties only the 2005 London Bombing still has a significant negative impact if the impact is measured on the targeted country. Overall can be concluded that the targeted country suffers the most financial damage from a terrorist attack, and the damage is permanent, but the targeted country suffers less financial damage from terrorist attacks with a large number of casualties.

The last hypothesis tested the market reaction of the industries; financial service; oil, gas and water; insurance; travel and leisure; aerospace and defence. The results show that all five industries experience negative AR on the day of the event, two industries show significant results. The oil, gas and water industry is struck the most and is the only one to be permanently damaged. The other industries show positive 10-day CAARs and thus show a short-term

48 reversal effect, which indicates an overreaction by investors. Overall can be concluded that the aerospace and defence industry and travel and leisure are the best terrorism risk diversifier, since these industries have near zero abnormal returns on the day of the event and a positive significant 10-day CAAR.

6.2 Contribution and Recommendations The literature review made clear that there are both direct and indirect costs involved with terrorism. The indirect costs are a large proportion of the total economic costs of terrorism and are mainly caused by negative market reaction on the day of the event. Regardless of the fact that the drop in stock prices is permanent or transitory, it is useful for investors to know which indexes are affected by terrorist attacks, so they can arm themselves against sudden drops. This empirical research is unique: no other researcher carried out a study of such depth within a West European setting. To date, the work produced by other researchers has concentrated on large attacks and on a single market, my study includes both small events and large events and includes multiple markets. This study measures the event day and post-event day market reaction for both the national indexes and the industries. It uses a simplified empirical method to measure the impact, which makes it easy to replicate in another setting. The findings of this study are useful as they reveal not only the indexes which may be affected by terrorist events, but also the direction of the impact (positive or negative). The results of this study can be used by investors to diversify terrorism risk across different countries and industries. In addition, this study offers an insight into a driver behind the financial impact of terrorism. This study also updates the existing literature which is valuable, since the financial environment is constantly changing and therefore the financial impact also changes. Finally, this study is unique since it links the financial impact of terrorism to findings in the behavioural finance.

I would recommend investors to diversify their portfolio across West European indexes to minimize exposure to the targeted country. The results show that the national index of Belgium is the safest choice to diversify, since this index is not significantly impacted on the day of the event as opposed to other West European countries, in addition this market does show positive abnormal returns in the post-event window and therefore would generate positive returns after the attack. I would also recommend investors to diversify their portfolio across industries to decrease the exposure to terrorism risk. A short-term investing strategy would be to buy stocks on the day of the attack that are: sensitive to terrorism and have a transitory

49 reaction to terrorism, such as insurance industry stocks, and sell them after the prices have rebounded to pre-attack level. Overall I would suggest to keep the financial consequences of terrorism in mind, but the results show that the financial damage by small-scaled terrorist attacks is only temporary, and therefore does not affect the returns of a portfolio in the long run.

Further research could consider to construct and test different portfolios diversification strategies based on the results in this study. One could calculate the difference in performance of portfolios, for example: the first investor holds 1/7 of their portfolio in each of the West European national indexes and the second investor takes the direction of the impact of terrorist attacks into consideration. Furthermore, an autocorrelation analysis can be used to construct a trading strategy based on terrorist attack. If the returns around the day of an attack exhibit autocorrelation, the stock could be characterized as a momentum stock, since its past returns influences its future returns. Further research could also consider to investigate the relation between negative returns on the day of an attack and the direct financial costs, since it is expected that higher direct costs are linked with higher total costs. This could be done by measuring the economic damage and observe if that drives the negative returns. Further research on the driver target would also be beneficial. It is interesting to see if the same logic holds for targeted industry as for targeted country. For example, one could check whether transport industry is struck more if this industry is targeted by the terrorist attack.

6.3 Limitations A limitation of this study is that it only uses the event study methodology to measure the financial impact of terrorism on the stock market. An improvement for the next research in this field is to use multiple methodologies. The GARCH method would give insight in the volatility changes after an attack. If the volatility increases around the day of the event, the risk for investing in the stock market increases. Another useful method would be the non-parametric approach, this methodology does not impose strong parametric restrictions as opposed to an event study and is less computationally intensive than the GARCH method. Furthermore, this study only uses attacks with more than five casualties which makes it less generalizable for all terrorist attacks. In addition, it only uses attacks that target West European countries which might cause the results to be more extreme, if distance is a driver behind the impact. The results are also less generalizable, since attacks on advanced economies show less of the financial impact. The main deficiency of this research is that I use a rather small sample. The researched

50 base would be better if I used more European countries and more attacks. A larger dataset would have made the results more generalizable. Moreover, this study gives investors only the tools to construct a terrorism risk diversification strategy, it does not measure the optimal portfolio and it does not show the real value to diversify terrorism risk.

7. References

Abadie, A., & Gardeazabal, J. (2003). The economic costs of conflict: A case study of the Basque country. American Economic Review, 93(1), 113-132. Abadie, A., & Gardeazabal, J. (2008). Terrorism and the world economy. European Economic Review 52(1), 1-27. Araña, J. E., & León, C. J. (2008). The impact of terrorism on tourism demand. Annals of Tourism Research, 35(2), 299-315. Arin, K., Ciferri, D., & Spagnolo, N. (2008). The price of terror: The effects of terrorism on stock market returns and volatility. Economics Letters, 101(3), 164-167. Baker, D. (2014). The effects of terrorism on the travel and tourism industry. International Journal of Religious Tourism and Pilgrimage, 2(1), 9-13. Baker, M., & Wurgler, J. (2007). Investor sentiment in the stock market. Journal of economic perspectives, 21(2), 129-152. Barros, C., & Gil-Alana, L. (2009). Stock market returns and terrorist violence: Evidence from the Basque country. Applied Economics Letters, 16(15), 1575-1579. Becker, S. G., & Rubinstein, Y. (2004). Fear and the response to terrorism: an economic analysis. University of Chicago mimeo. Berrebi, C., & Klor, E. (2005). The impact of terrorism across industries: An empirical study. Hebrew University of Jerusalem working paper. Blasco, N., & Ferreruela, S. (2008). Testing intentional herding in familiar stocks: An experiment in an international context. The Journal of Behavioral Finance, 9(2), 72-84.

Blasco, N., Corredor, P., Del Rio, C., & Santamarıá , R. (2005). Bad news and Dow Jones make the Spanish stocks go round. European Journal of Operational Research, 163(1), 253-275. Borenstein, S., & Zimmerman, M. B. (1988). Market incentives for safe commercial airline operation. The American Economic Review, 78(5), 913–935. Breusch, T. S., & Godfrey, L. G. (1980). A review of recent work on testing for autocorrelation in dynamic economic models. University of Southampton. Breusch, T. S., & Pagan, A. R. (1979). A simple test for heteroscedasticity and random coefficient variation. Econometrica: Journal of the Econometric Society, 1287-1294. Brounen, D., & Derwall, J. (2010). The impact of terrorist attacks on international stock markets. European Financial Management, 16(4), 585-598. Brown, J. R., Cummins, J. D., Lewis, C. M., & Wei, R. (2004). An empirical analysis of the economic impact of federal terrorism reinsurance. Journal of Monetary Economics, 51(5), 861-898. Brown, S. J., & Warner, J. B. (1985). Using daily stock returns: The case of event studies. Journal of Financial Economics, 4(1), 3-31. Brück, T., & Wickström, B. A. (2004). The economic consequences of terror: Guest editors' introduction.

Buesa, M., Valino, A., Heijs, J., Baumert, T., & Gomez, J. (2007). The economic cost of March 11: Measuring the direct economic cost of the terrorist attack on March 11, 2004 in Madrid. Terrorism and Political Violence, 19(4), 489-509. Burch, T. R., Emery, D. R., & Fuerst, M. E. (2003). What van “nine‐eleven” tell us about closed‐end fund discounts and investor sentiment? Financial Review, 38(4), 515-529. Chan, M. L., Khanthavit, A., & Thomas, H. (1996). Seasonality and cultural influences on four Asian stock markets. Asia Pacific Journal of Management, 13(2), 1-24. Chen, A., & Siems, T. (2004). The effects of terrorism on global capital markets. European Journal of Political Economy, 20(2), 349-366. Chesney, M., Reshetar, G., & Karaman, M. (2011). The impact of terrorism on financial markets: An empirical study. Journal of Banking & Finance, 35(2), 253-267. Drakos, K. (2004). Terrorism-induced structural shifts in financial risk: airline stocks in the aftermath of the September 11th terror attacks. European Journal of Political Economy, 20(2), 435-446. Drakos, K. (2010). Terrorism activity, investor sentiment, and stock returns. Review of Financial Economics, 19(3), 128-135. Drakos, K., & Kutat, M. A. (2003). Regional effects of terrorism on tourism in three Mediterranean countries. Journal of Conflict Resolution, 47(5), 621-641. Eldor, & Melnick. (2004). Financial markets and terrorism. European Journal of Political Economy, 20(2), 367–386. Elster, J. (1998). Emotions and economic theory. Journal of Economic Literature, 36(1), 47–74. Elton, E. J., Gruber, M. J., & Busse, J. A. (1998). Do investors care about sentiment? The Journal of Business, 71(4), 477-500. Enders, W., & Sandler, T. (1991). Causality between transnational terrorism and tourism: The case of Spain. Studies in Conflict & Terrorism, 14(1), 49-58. Enders, W., & Sandler, T. (2000). Is transnational terrorism becoming more threatening? A time-series investigation. Journal of Conflict Resolution, 44(3), 307-332. Enders, W., Sandler, T., & Parise, F. G. (1992). An econometric analysis of the impact of terrorism on tourism. Kyklos, 45(4), 531-554. Fisher, L. (1966). Some new stock-market indexes. The Journal of Business, 39(1), 191–225. Glaser, M., & Weber, M. (2005). September 11 and stock return expectations of individual investors. Review of Finance, 9(2), 243-279. Goldberg, L. S., & Leonard, D. (2003). What moves sovereign bond markets? The effects of economic news on US and German yields. Journal of Economics and Finance, 9(9), 87-99. Goodwin, R., Willson, M., & Gaines, S. (2005). Terror threat perception and its consequences in contemporary Britain. British Journal of Psychology, 96(4), 389-406. International Energy Agency. (2015). Key World Energy Statistics OECD/IEA 2015. Ito, H., & Lee, D. (2004). Assessing the impact of the September 11 terrorist attacks on US airline demand. Journal of Economics and Business, 57(1), 75-95.

Johnston, R. B., & Nedelescu, M. O. (2006). The impact of terrorism on financial markets. Journal of Financial Crime, 13(1), 7–25. Kaplanski, G., & Levy, H. (2010). Sentiment and stock prices: The case of aviation disasters. Journal of Financial Economics, 95(2). Karolyi, G. (2006). The consequences of terrorism for financial markets: What do we know? Canadian Investment Review. Karolyi, G., & Martell, R. (2006). Terrorism and the stock market. The Ohio State University. Khalil, F., & Akhtar, W. (2017). Impact of terrorist attacks on stock market performance: A case of Pakistan. Asian Journal of Economic Modelling, 5(2), 208-222. Kollias, C., Kyrtsou, C., & Papadamou, S. (2013). The effects of terrorism and war on the oil price– stock index relationship. Energy Economics, 40(1), 743-752. Kramer, L. A., & Weber, J. M. (2012). This is your portfolio on winter: Seasonal affective disorder and risk aversion in financial decision making. Social Psychological and Personality Science, 3(2), 193-199. Lee, C. M., Shleifer, A., & Thaler, R. H. (1991). Investor sentiment and the closed‐end fund puzzle. The Journal of Finance, 46(1), 75-109. Lenain, P., Bonturi, M., & Koen, V. (2002). The economic consequences of terrorism. Lerner, J. S., Gonzalez, R. M., Small, D. A., & Fischhoff, B. (2003). Effects of fear and anger on perceived risks of terrorism: A national field experiment. Psychological science, 14(2), 144- 150. Melnick, R., & Eldor, R. (2010). Small investment and large Returns: Terrorism, media and the economy. European Economic Review, 54(8), 963-973. Nikkinen, J., & Vähämaa, S. (2010). Terrorism and stock market sentiment. Financial Review, 45(2), 263-275. Orbaneja, J. R., Iyer, S. R., & Simkins, B. J. (2017). Terrorism and oil markets: A cross-sectional evaluation. Finance Research Letters, 1-7. Richman, V., Santos, M. R., & Barkoulas, J. T. (2005). Short-and long-term effects of the 9/11 event: The international evidence. International Journal of Theoretical and Applied Finance, 8(7), 947-958. Schwert, G. W. (1990). Stock volatility and the crash of ’87. Review of Financial Studies, 3(1), 77–102. Solomon, S., Greenberg, J., & Pyszczynski, T. (1991). A terror management theory of social behavior: The psychological functions of self-esteem and cultural worldviews. Journal of Social Psychology, 24(3), 93-159. START, N. C. (2016). Global Terrorism Database Paris 2015. Retrieved from Global Terrorism Database: https://www.start.umd.edu/gtd START, N. C. (2017). Global Terrorism Database. Retrieved from Global Terrorism Database: https://www.start.umd.edu/gtd Suleman, T. (2012). Stock market reaction to terrorist attacks: Empirical evidence from a front line state. Australasian Accounting, Business and Finance Journal, 6(1), 97-110.

United States State Deparment. (2016). Country Reports on Terrorism 2015. Bureau of Counterterrorism and Countering Violent Extremism. Woo, G. (2003). The evolution of terrorism risk modeling. Journal of Reinsurance, 10(3), 1-9.

8. Appendix

A. 1 Terrorist attacks This table lists terrorist attacks in West Europe in the period 2000-2017. The table is sorted by date and provides information on the location, the organization responsible for the terrorist attack, number of fatalities, number of injured, the type of the attack and the amount of property damage. Source: Global Terrorism Database.

Date Country (City) Fatalities Injured Attack type Property damage 21-1-2000 Spain (Madrid) 3 4 Bombing Minor 6-3-2000 Spain (Donostia-San Sebastian) 0 7 Bombing Minor 29-4-2000 Germany (Hamburg) 0 9 Bombing Minor 8-5-2000 Spain (Vigo) 2 4 Armed Assault Minor 5-7-2000 United Kingdom (Drumcree) 0 7 Unarmed Assault No 12-7-2000 Spain (Madrid) 0 8 Bombing Minor 27-7-2000 Germany (Dusseldorf) 0 10 Bombing Minor 8-8-2000 Spain (Madrid) 0 11 Bombing Minor 20-8-2000 Germany (Eisenhuttenstadt) 0 7 Bombing Minor 29-9-2000 Spain (Barcelona) 0 6 Bombing No 30-10-2000 Spain (Madrid) 3 30 Assassination Minor 11-11-2000 Spain (Donostia-San Sebastian) 0 11 Bombing Minor 22-2-2001 Spain (Donostia-San Sebastian) 2 5 Assassination No 28-6-2001 Spain (Madrid) 0 16 Bombing Major 10-7-2001 Spain (Madrid) 1 12 Bombing Minor 23-7-2001 France (Borgo) 0 14 Bombing Major 3-8-2001 United Kingdom (Belfast) 0 7 Bombing Major 18-8-2001 Spain (Salou) 0 13 Bombing Minor 12-10-2001 Spain (Madrid) 0 14 Bombing Major 6-11-2001 Spain (Madrid) 0 95 Assassination Minor 28-2-2002 Spain (Portugalete) 0 5 Bombing Minor 1-5-2002 Spain (Madrid) 0 17 Bombing Minor 4-8-2002 Spain (Alicante) 2 30 Bombing No 30-5-2003 Spain (Sanguesa) 2 6 Bombing Minor 4-6-2003 Belgium (Brussels) 0 20 Unarmed Assault No 19-7-2003 France (Nice) 0 8 Bombing Major 19-11-2003 Italy (Rome) 0 50 Unarmed Assault No 11-3-2004 Spain (Madrid) 193 2050 Bombing Major 9-6-2004 Germany (Mulheim) 0 22 Bombing Minor 8-10-2004 France (Paris) 0 10 Bombing Minor 9-2-2005 Spain (Madrid) 0 43 Bombing Major 25-5-2005 Spain (Madrid) 0 34 Bombing Minor 7-7-2005 United Kingdom (London) 52 784 Bombing Major 11-9-2005 United Kingdom (Belfast) 0 50 Bombing Minor 29-12-2006 Spain (Madrid) 2 12 Bombing Minor 17-4-2008 Spain (Bilbao) 0 7 Bombing Minor 21-9-2008 Spain (Ondarroa) 0 21 Bombing Major

30-10-2008 Spain (Lyon) 0 17 Bombing Minor 20-3-2009 France (Lyon) 0 10 Armed Assault Minor 1-5-2009 Netherlands (Apeldoorn) 7 12 Assassination Minor 29-7-2009 Spain (Burgos) 0 46 Bombing Major 19-3-2012 France (Toulouse) 4 1 Armed Assault No 10-8-2013 United Kingdom (Belfast) 0 56 Armed Assault Minor 24-5-2014 Belgium (Brussels) 4 1 Armed Assault No 21-12-2014 France (Dijon) 0 11 Unarmed Assault Minor 7-1-2015 France (Paris) 12 12 Armed Assault Minor 9-1-2015 France (Paris) 5 3 Armed Assault Minor 4-9-2015 Germany (Heppenheim) 0 5 Armed Assault Minor 17-10-2015 Germany (Cologne) 0 5 Assassination No 13-11-2015 France (Paris) 130 413 Bombing Major 7-12-2015 Germany (Altenburg) 0 10 Facility Attack Minor 24-12-2015 Germany (Wallerstein) 0 12 Facility Attack Minor 22-3-2016 Belgium (Zaventem) 32 340 Bombing Major 14-7-2016 France (Nice) 87 433 Armed Assault Minor 18-7-2016 Germany (Wurzburg) 1 5 Armed Assault No 22-7-2016 Germany (Munich) 10 27 Armed Assault No 24-7-2016 Germany (Ansbach) 1 15 Bombing Minor 15-10-2016 Spain (Alsasua) 0 5 Unarmed Assault No 20-11-2016 France (Paris) 0 15 Unarmed Assault No 19-12-2016 Germany (Berlin) 12 48 Unarmed Assault Minor 22-3-2017 United Kingdom (London) 5 50 Armed Assault Minor 20-4-2017 France (Paris) 2 3 Armed Assault No 22-5-2017 United Kingdom (Manchester) 22 512 Bombing Major 3-6-2017 United Kingdom (London) 8 48 Armed Assault Minor 19-6-2017 United Kingdom (London) 1 10 Unarmed Assault Minor 28-7-2017 Germany (Hamburg) 1 6 Armed Assault No 9-8-2017 France (Lev) 0 6 Unarmed Assault No 17-8-2017 Spain (Barcelona) 16 152 Armed Assault Minor 15-9-2017 United Kingdom (London) 0 30 Bombing Minor

A.2 Industries This table lists each company for each industry. All companies are listed on one of the seven West European indexes. The brackets show the country of listing of the corresponding company.

Financial service Oil, Gas and water Insurance Travel and leisure Aerospace and defence ING Group (NLD) Royal Dutch Shell (NLD) NN Group (NLD) Carnival Corporation (UK) BAE Systems (UK) HSBC (UK) BP (UK) Prudential (UK) Easy Jet (UK) Rolls-Royce Group (UK) Banco Santander (ESP) Repsol (ESP) Mapfre(ESP) GVC Holdings (UK) Safran (FRA) KBC (BEL) Total S.A. (FRA) Ageas (BEL) International Airlines (ESP) Airbus (FRA) BNP Paribas (FRA) ENI (ITA) AXA (FRA) Accor (FRA) Leonardo (ITA) Intesa Sanpaolo (ITA) Linde (GER) Generali (ITA) Deutsche Lufthansa (GER) Deutsche Bank (GER) Allianz (GER)

A. 3 Autocorrelation test national indexes

A. 4 Heteroskedasticity test national indexes

A. 5 Market reaction per quarter National Quarter Event-day AR 5-day CAAR 10-day CAAR index NLD (1) -0.002 (-0.862) -0.002 (-0.508) -0.003 (-0.512) (2) -0.0002 (-0.072) 0.006 (1.112) 0.007 (0.654) (3) -0.001 (-0.589) -0.003 (-0.715) 0.002 (0.281) (4) -0.006** (-2.142) -0.0001 (-0.027) 0.0003 (0.035) UK (1) -0.001 (-0.284) 0.001 (0.141) 0.003 (0.449) (2) -0.004 (-1.628) -0.002 (-0.368) -0.002 (-0.328) (3) -0.001 (-0.989) -0.0002 (-0.063) 0.006 (1.190) (4) -0.005** (-2.306) -0.015 (-0.713) 0.001 (0.174) ESP (1) -0.006* (-1.837) -0.007* (-1.853) 0.012* (1.642) (2) -0.003 (-1.213) 0.002 (0.393) -0.0004 (-0.043) (3) -0.002 (-0.772) -0.001 (-0.162) 0.009 (1.571) (4) -0.003 (-0.912) 0.010* (1.723) 0.013 (1.535) BEL (1) -0.002 (-0.809) -0.004 (-0.525) 0.005 (0.481) (2) -0.0003 (-0.121) 0.003 (0.713) 0.001 (0.102) (3) 0.001 (0.602) 0.001 (0.210) 0.006 (1.054) (4) -0.002 (-1.134) 0.005 (1.188) 0.007 (1.096) FRA (1) -0.002 (-0.535) -0.003 (-0.603) -0.001 (-0.123) (2) -0.002 (-0.821) 0.002 (0.384) 0.002 (0.215) (3) -0.001 (-0.637) 0.002 (0.456) 0.008* (1.890) (4) -0.004** (-2.328) -0.001 (-0.196) 0.003 (0.392) ITA (1) -0.004 (-1.000) -0.001 (-0.118) -0.007 (-0.767) (2) -0.002 (-0.717) 0.005 (0.798) 0.004 (0.421) (3) -0.001 (-0.271) 0.002 (0.689) 0.006 (1.166) (4) -0.005** (-1.889) 0.005 (0.712) 0.011 (1.128) GER (1) -0.001 (-0.256) -0.004 (-0.923) -0.009 (-1.254) (2) -0.002 (-0.556) 0.004 (0.931) 0.007 (0.959) (3) -0.001 (-0.465) 0.007* (1.672) 0.011* (1.762) (4) -0.006** (-2.282) -0.002 (-0.456) -0.001 (-0.167) ALL (1) -0.002 (-0.850) -0.003 (-0.805) -0.004 (-0.732) (2) -0.002 (-0.848) 0.003 (0.615) 0.003 (0.397) (3) -0.001 (-0.681) 0.001 (0.401) 0.001 (1.475) (4) -0.005** (-2.180) -0.002 (-0.145) 0.001 (0.440) Table A.5 shows the mean abnormal returns for all West European countries per quarter. It also shows the 5-day and 10-day cumulative average abnormal returns per country per quarter. The starting date for the CAARs is the day of the event. The first quarter consists of attacks that happened between January and March, the second quarter consists of attacks that happened between April and June, the third quarter consists of attacks that happened between July and September and the fourth quarter consists of attacks that happened between October and December. The robust standard errors are double clustered. ***, **, and *, represent coefficients that are statistically different from zero at the 1%, 5%, and 10% levels respectively.

A. 6 Autocorrelation test industries

A. 7 Heteroskedasticity test industries