Terrorism and the Stock
Performance of Publicly Listed
Firms 2003-2007
An empirical analysis to test for the terrorism attack effects on listed companies worldwide.
Maria Kyriakou
June 2008
MASTER THESIS
UNIVERSITY OF GRONINGEN
Faculty of Economics and Business
MSc BA Finance (Risk and Portfolio Management)
Maria Kyriakou 1
Terrorism and the Stock
Performance of Publicly Listed
Firms 2003-2007
An empirical analysis to test for the terrorism attack effects on listed companies worldwide.
Abstract
The aim of this thesis is to study terrorist events and their effects on stock price movements of publicly-trading affected firms, through 2003-2007. Data with respect to terrorist attacks have been collected from two official databases that run data for this purpose. 50 attacks are studied with the use of standard event study methodology. Both AARs and CAARs over the (-5, +5) event window are found statistically insignificant, as well as abnormal returns on the event date itself (day 0). We can, thus, assume that terrorist events do not convey any information to the market place. Results with respect to the volatility of AARs indicate an increased uncertainty of investors about experiencing losses of terrorist attacks in the future. In addition, I uncover a statistically significant 1.89% positive response on day-3 following the event for attacks with casualties involved.
JEL Classification: G12, G14.
Key Words: terrorism, stocks, event study, efficient markets, CAARs, AARs.
Maria Kyriakou Moesstraat 16 C-04 9717 JW Groningen
Maria Kyriakou 2
PREFACE
Throughout the academic year 2007-2008 I have been studying the MSc BA Finance at the University of Groningen. During the last term of this Master, students are required to write a thesis on a topic, which has to be chosen by them. The subject of my Master Thesis is the effect of terrorist attacks on listed companies worldwide during the period 2003 through 2008. During the process of writing my thesis many people helped me a lot to whom I will always be grateful. First of all, my supervisor, Dr. Peter Smid, who was always willing to advise me and give me practical help.
Maria Kyriakou 3
TABLE OF CONTENTS
PREFACE ... 2
TABLE OF CONTENTS ... 3
1. INTRODUCTION... 5
2. THEORYANDEMPIRICS... 8
2.1 Theory... 8
2.1 1 What constitutes terrorism? ... 8
2.1 2 Which parties are involved?... 8
2.1 3 What motivates terrorism? ... 9
2.1 4 Economic and Social Consequences of Terrorism ... 9
2.1 5 Consequences for the Stock Market & Signalling Effects ... 13
2.1 6 Security projects: A way to mitigate terrorist risk ... 15
2.1 7 Behavioral Explanations ... 17
2.1 8 Summary... 18
2.2 Empirics... 19
2.2 1 Introduction ... 19
2.2 2 Global stock markets and recent terrorist attacks... 19
3. METHODOLOGY... 25
3.1 Introduction... 25
3.2 Definition of an Event Study... 25
3.3 Models for Modelling Normal Stock Returns in an Event Study... 26
3.2 1 Constant Mean Return Model ... 26
3.2 2 Market Model ... 27
3.4 Measuring Abnormal Stock Returns in an Event Study... 28
3.5 Aggregation of Abnormal Returns... 29
3.5 1 Across Securities: Average Abnormal Returns(AARs) ... 29
3.5 2 Over the Event Window: Cumulative Abnormal Returns(CARs) ... 30
3.5 3 Across Securities and Over the Event Window: Cumulative Average Abnormal Returns(CAARs) ... 30
3.7 Hypotheses... 30
4. DATA... 33
4.1 Terrorist Events 2003-2007... 33
4.2 Nature of affected firms... 34
Maria Kyriakou 4
5. EMPIRICALRESULTS... 36
5.1 Cross-sectional Analysis... 39
6. CONCLUSIONSANDFUTUREDIRECTIONS... 44
7. REFERENCES... 45
Maria Kyriakou 5
1. INTRODUCTION
A suicide attack claimed by the Taliban killed seven people, including an American and a Norwegian journalist, on Monday 14th January 2008 at the five-star “Serena” hotel in Kabul where the Norwegian foreign minister was staying1. Serena Hotel trades under the
symbol “TPS” (Serena) Ltd. on Nairobi Stock Exchange. Its stock price was 33 the day when the incident occurred and it close to 31 on Tuesday 15th January 2008, a loss of more than 6% in a single trading day2.
The moment the information about a terrorist attack reaches financial markets, market participants react, usually, strongly. Any attack reduces investors’ confidence and makes them less willing to spend. Further on, business conditions change rapidly because there is pessimism about the future. Terrorism has certain characteristics: it can affect anyone, and entails a large degree of uncertainty in the sense that no one knows when and where terrorists will strike. This uncertain nature of the terrorist threat is what makes it the most difficult to measure and manage. However, the uncertainty about terrorism must not be left outside corporate decision making. Much has been written and said about the cost of terrorism-related attacks on human lives, property and in terms of macroeconomic factors such as GDP, commodity prices, oil prices, aggregate consumption, and investors’ risk appetite. The bulk of the literature has especially increased after the attacks to the World Trade Centre in the United States on 11th September (thereafter, 9/11). According to the International Monetary Fund [2001] the possible long term consequences of the 9/11 attack are among others: higher operating costs for firms, higher level of inventories, higher risk premia, shift of resources from civilian labour force to military services and shift away from globalization. Indeed, it is not pure luck that the literature about the events of 9/11 is so extensive. The terrorist attacks themselves cost the lives of 2,976 people, from which financial services employees accounted for 74%. To deal with the post-event crisis, authorities decided to suspend the operations of the American Stock Market for a week. When operations resumed, Dow Jones fell around 1,000 points during the days that followed.
Despite the bulk of the literature that deals with the 9/11 attacks, only little has been written about the effects of less known terrorist attacks on financial markets and, especially, on the individual firms’ stock price. As a consequence of this lack of literature, knowledge is scarce on how investors make their decisions under the high degree of uncertainty that arises from a crisis situation like a terrorist event. One of the reasons for this is that there is no exact
1
Source: www.reuters.com (Access date: 15/1/2008).
2
Maria Kyriakou 6 measure of turbulence on stock markets. If such measure was invented, it would be beneficial for both regulators and firms. In particular, firms want to quantify the risk associated with terrorist attacks so as to manage and minimize it. Terrorist hazard affects firms, in terms of damaging their financial assets, their human capital, their know-how and their reputation. Nonetheless, progress in the field of measurement has been made. For instance, Maillet and Michel [2005] attempt to measure stock market turbulence after a crisis and introduce the
Index of Market Shocks. When applying this measure to assess extreme events that took place
in the American and French Stock markets, they find that the 9/11 attack was the worst after the historical financial crisis of 1987.
The thesis aims to extend the literature that documents how firms and investors make their decisions. It offers the opportunity to study how a large, unexpected event like a terrorist attack affects investor’s forecasts of expected returns and volatility. The goal is to see whether terrorist events provide information to the marketplace. If terrorist events convey information, one would expect bigger terrorist attacks in terms of casualties, to have a larger impact on the valuation of the affected firm’s equity value. The results of the thesis have important implications for asset managers and regulators who want the terrorist risk quantified so as to manage it. Finally, the answers have broader implication for the society as a whole, if one bears in mind the increasing frequency of terrorist events in western societies. Daily stock price data can be used to analyze how a terrorist attack alters the affected firm’s stock price during the first five days following the attack. The attacks are identified with respect to the type of target which must be a publicly listed firm and the analysis is done with the use of event study methodology. The strength of the event study methodology arises from the fact that abnormal returns of a specific firm at a certain time can be aggregated across many firms that experienced similar events at different times so as generalized results can be drawn. Further, from the attacks that affected publicly listed firms, I form subsets of the sample across various dimensions. In particular, I distinguish between affected firms trading in emerging markets and firms that were involved in an attack with victims. The data are used to address the following research questions:
Research Question 1: “Terrorist attacks reduce stock prices and increase risk of the affected firm in the days immediately following the attacks”.
Maria Kyriakou 7 efficient market hypothesis [Fama, 1970]. This hypothesis states that as new publicly available information becomes available, investors immediately reassess individual firms’ discounted present and future value. This reassessment is reflected in new stock prices. Event study methodology attempts to capture the change in stock prices and is the methodology employed here.
Research Question 2: “News about a terrorist attack is considered unanticipated, thus, price movements before the event should not be observed”.
It is obvious that a terrorist attack is an exogenous event, thus, no one can anticipate it and profitably trade on this information. Hence, I expect to find zero abnormal returns on the five days before such an event occurred.
Research Question 3: “Higher number of casualties in an attack will result in lower stock prices”.
Research Question 4:“If the attacked firm’s stock trades in an emerging stock market, the drop in stock price will be higher than in a developed market”.
When an attack takes place, it is interesting to know which events result in more pronounced stock price responses. I expect to find that higher number of injuries or fatalities in an attack, will lead to more pronounced effects. This is derived from the fact that higher number of casualties magnifies the impact of the attack and, thus, intensifies its consequences. With respect to the public nature of the affected firm, it may be that emerging market trading stocks experience higher declines of security prices. This assumption arises from the generalized economic and political instability of such markets. Research questions 3 and 4 will be answered with the use of cross-sectional regression analysis.
Maria Kyriakou 8
2. THEORY AND EMPIRICS
2.1 Theory
2.1 1 What constitutes terrorism?
Before proceeding to any in depth analysis, it is important to give a definition to what terrorism is. According to Boaz Ganor, the Executive Director of “International Institute for Counter-Terrorism (ICT)”:
“Terrorism is the deliberate use of violence against civilians in order to achieve political ends. Political objectives are always a factor in terrorism; it is a highly effective tactic and modus operandi to achieving political goals. I also decline the use of any adjectives to describe civilians, such as ‘innocent’ or ‘non-combative,’ since these distinctions are manipulated by different parties according to their not unbiased perspectives.” (ICT’s 7th International Conference Highlights 2008, p. 12)
It is worth noting here that there is no official definition of what constitutes terrorism, making it more difficult for regulators and firms to beat. However, an acceptable definition is the one given by B. Ganor. Emphasis should be put on the term political ends; this is what differentiates terrorism from other criminal activities. More specifically, the most frequent targets of terrorist attacks are not individuals3 or a whole country, otherwise we would talk about a murder or a war, respectively. The most frequent targets are firms, though. Furthermore, what terrorist groups want to achieve is not the maximum number of casualties but to disrupt the economic activity or production process of a firm or a country. Nonetheless, publicity does also play a crucial role when wanting to assess the criticality of an attack. Through the world press coverage, a terrorist group achieves publicity and its threats are spread throughout the world. Hence, when evaluating the costs of terrorism one must not only consider the direct losses in terms of casualties, material costs, etc. but also the indirect costs, those that cannot be seen at first sight.
2.1 2 Which parties are involved?
According to Trajtenberg [2003] there are three parties involved in the fight against terrorism: (1) Terrorists; the utility they derive from an attack is a positive function of the number of casualties, (2) Firms, which have to operate under the threat of terrorism that results in decreased earnings, and (3) The Government, that decides on the funds to be invested in
3 Sometimes famous people may be victims of terrorist attacks, though. For example the murder of Kennedy
Maria Kyriakou 9 counterterrorism projects. Firms and governments employ different strategic plans to counter terrorist risk. Contrary to the government which is fighting terrorism at its source, firms by deciding how much to invest in securing their assets, use a strategy which entails deploying resources to protect likely targets in their homeland. However, co-operations between firms and the government exist, since both of them want to mitigate terrorist risk and alleviate its consequences. In the case of the 9/11 attacks, shortly after the events had occurred, on the one hand, regulatory measures were taken to boost the American economy. These came with the form of cuts in interest rates, increases in overnight repos and liquidity injections. On the other hand, individual firms, inspired by a patriotic behaviour, did also want to contribute to stabilize markets and help the country. This was, also, driven by the fact that shutting down the stock market for a prolonged period would signal that terrorists had won. Indeed, one of the incentives of the 9/11 terrorist attacks was to disrupt the US financial system.
2.1 3 What motivates terrorism?
Terrorist organizations need support from the society surrounding them to meet their political ends. They aim to attract public opinion because this way the message they want to pass will be spread all over the world. Religion, most of the times, serves as a platform due to the moral explanations it provides to terrorist actions which otherwise would seem immoral. Furthermore, it provides the massive support necessary to the terrorist groups. However, culture does, also, play its role in providing a collective identity and massive support. Frayman [2006] discusses that terrorism is motivated by a blend of religious and cultural reasons which are not easy to separate one from the other. He suggests that if one wants to comprehend the motivations behind terrorism they have to integrate those two terms. Despite the fact that the motives of terrorists, being them religious or cultural, are not standard, terrorists, most of the times, follow certain routes for their actions. Just to name a few of them, these comprise assassinations, suicide attacks, bombings, kidnappings and airplane hijackings.
2.1 4 Economic and Social Consequences of Terrorism
IMF [2001], Blomberg et al. [2004], Enders and Sandler [2005] and Frey et al. [2007], among others, distinguish among various economic and social consequences of terrorism. In the current section, I expand their classification augmenting it with further references and consequences.
Maria Kyriakou 10 increase in terrorism in the other country. Drakos and Kutan [2003] study the effect of terrorism in Greece, Italy and Turkey from 1990 to 2000 and find significant substitution effects. Some terrorist attacks aim to harm specific industries and most of the times tourism is among them, as with the case of the 9/11 attacks.
b) Foreign Direct Investments. Foreign firms will choose to reduce their investments in a country that suffers from terrorist threat because of the generalized uncertainty. Furthermore, terrorism causes a redirection away from investment spending towards government spending, (Blomberg et al., 2004).
c) Savings and Consumption. The relation is not clear because terrorism increases risks associated with savings but at the same time induces investors to place their money in safer instruments (such as durable goods) instead of consuming it. Moreover, according to IMF [2001] investors are more willing to save money instead of investing, after an attack has occurred, because the investor’s perception of business risk is altered after an event. d) Investments. Both the level and the composition of investments are affected. Investments
in a terror stricken country are perceived as riskier. Furthermore, terrorism raises the costs of running a business in a country in terms of higher insurance premia, increased security costs and higher salaries to exposed employees.
e) Stock Markets. The effect that terrorist attacks have on stock markets will be extensively elaborated on later in this section as this is the main research interest of this thesis. With hindsight, the channel through which stock markets are affected is double. First, the perceived riskiness of equity is increased. Secondly, investors become more risk averse and, thus, less willing to bear equity risk. As a consequence, risky assets fall in value together as their demand drops. The effect of contagion to other stock markets may also arise, since markets are said to be highly interrelated. Emerging markets might also be critically affected, since their recovery is, most of the times, less vigorous.
f) Other Financial & Commodity Markets. Derivative markets can provide useful information after an attack. Moreover, the affected country may be viewed as riskier which can affect the foreign exchange rate of the attacked country negatively. Also, commodity markets (especially oil) are considered to show lower demand and a sharp fall in prices.
Maria Kyriakou 11 the importing, developed country may be reduced. Hence, the impact of poverty might be, further, intensified in the developing country, [see, (IMF 2001)].
h) Urban Economy. The relation is nor straightforward. It may be that large cities take advantage of economies of scale when countering terrorism, but at the same time, they constitute the main targets of terrorist attacks because of the high population density. This can be translated to increased expenditures on citizen protection. Expenditures include higher transportation costs, longer waiting time in airports, loss of privacy and more funds allocated to military services. Indeed, people suffer from these measures. Hence another subsidiary cost arises: funds must be allocated in educating people to support the security measures.
i) National Income and Growth. Literature suggests that developing countries suffer more than developed ones. This may be a consequence of the fact that a developed economy is better diversified in terms of economic activities compared to a developing country that may be tied exclusively to tourism, fishing, etc. Moreover, a developed country possesses more effective ways to recover from an attack4. So for a developed economy the most notable effect may be a drop in demand mainly caused by investors’ loss of confidence. However, for a developing economy a slowing external demand has been noticed mainly in terms of tourism. Furthermore, the quality of financial markets may deteriorate as in the case of Latin America, which results in an increase of borrowing costs. Also, terrorism may have a negative impact on growth (Blomberg et al., 2004), which is an important factor for positive stock returns. If there is no growth in an economy, firms cannot increase their profits which is the vehicle for positive stock returns.
j) Human Life and Human Capital. Terrorism causes insecurity; it incurs the loss of human life and shortens life span (see, Eckstein and Tsiddon, 2004). It affects politics (Berrebi and Klor, 2006) and human capital (Tsiddon, 1995). In addition, people suffer from a reduction in civil liberties that arise from all kind of counterterrorism programmes. It is worth noting that research suggests that individuals are very willing to sacrifice their liberties for the sake of counterterrorism programmes. For instance, Viscusi and Zeckhauser [2003]study whether individuals are willing to sacrifice their civil liberties in order to contribute to the post- 9/11 terror risk reduction policies and find that they are, indeed, willing to do so.
k) Investors. Investors’ risk attitude is altered because their confidence is harmed. Confidence is an important variable in reducing activity in a developed economy. Investor’s outlook about the future gets bearish since there is a high degree of uncertainty.
4
Maria Kyriakou 12 The remedy for investors would be to invest in familiar firms and markets and seek professional advice so as to get better diversified portfolios.
l) Insurance Industry. In the case of large scale events, the insurance industry can be dramatically damaged through the resulting bulk of claims. This was the case in the events of 9/11, when the insurance industry suffered severe losses that are estimated to be as high as $40 billion. As a result of this, many insurers quickly tried to formalize the exclusion of coverage of terrorist events. Thus, the terror risk would be shifted back to policy holders which would make them more exposed to risk and losses. US government took action against this phenomenon by signing into law the Terrorism Risk Insurance
Act of 2002, which nullified terrorist events exclusion by insurers. Research by Griffith et al. [2003] studies the abnormal returns of the insurance companies during the days that
information about the progress of passing the Act was released. They find negative abnormal returns during the days that the Congress was more probable to pass the Act. Their results indicate that investors reassessed the market value of insurance firms. Similar findings are provided by Cummins and Lewis [2003] who find that insurance firms’ stocks decreased sharply after the 9/11 attack but those with good ratings rebounded after one week. Another tactic that insurance companies are forced to follow, after a large scale disaster, is the need to raise premia. After a large scale event there is, usually, a high degree of parameter uncertainty and insurers find difficulty in setting credible premia. To illustrate this, insurers set their prices using past records about events. New events, especially rare ones, lead to new estimates about premia. Evidence by IMF [2001] regarding the 9/11 attacks, suggests that higher premia are here to stay. What is more, the market assessment is very positive towards them because equity prices for insurers have increased around 10% relatively to the US market.
m) Airlines. IMF [2001] provides evidence that after the 9/11 attacks the US airline industry suffered some 20% loss of its, relative to the US market, value, while the equivalent figure for Japan and Europe was 15%. A significant disruption of this sector may have a negative effect on short term GDP. However, the relationship is not clear because it may be that the decrease in air travel demand may be offset by other means of transport such as the railroad.
Maria Kyriakou 13 firms or industries that are seen as safer and investors may have optimistic prospects about them. These businesses will increase in value after a terrorist attack. According to Money Marketing [2005, p.26], after the 7 July 2005 terrorist attacks in London’s public transport system, defence stocks, tobacco, utilities and some financials have been benefited.
o) Higher Transaction Costs. This consequence can be further divided into increased operating costs for firms (mainly through the higher insurance premia that firms have to pay), higher level of inventories (for instance, air travel is less reliable and production may be disrupted if firms do not have the necessary components for their production) and higher risk premia (investor’s risk aversion increases which may lead to higher required rates of return).
As a last word in this sub-section, Walker [2007] discusses that terrorists will gain more power and terrorist attacks will be more frequent in the days to come. All the aforementioned negative consequences and many others cannot be fully mitigated. However, individual investors and firms should get on with life. Portfolios and business strategies must be designed in such a way so as to deal with potential terrorist crises, risk, volatility, inflation and longevity. Finally, globalization is what is needed in the fight against terrorism.
2.1 5 Consequences for the Stock Market & Signalling Effects
Maria Kyriakou 14 before experiencing such a shock, may be inclined to reconstruct their portfolios after a terrorist attack, so as to get protected from similar shocks in the future.
The stock market itself can provide useful information which can help regulators detect potential terrorist attacks or suspects behind terrorist attacks. It serves as a barometer for assessing the impact of a terrorist event on the financial markets. Negative abnormal returns shortly after an attack can reveal new sources of risks which are translated in huge losses from firms. It is worth noting here that if the risk gets more thoroughly examined and comprehended the impact of future attacks might be less pronounced, assuming rational markets and investors. Finally, a thorough understanding of the nature of terrorism contributes in designing successful strategies to alleviate its costs or even to prevent terror. Radlaur [2001] provides useful evidence on that regarding the events of 9/11. He discusses how authorities investigated suspicious transactions that happened before the 9/11 attacks and had to do with the harmed companies, mainly airline ones. He provides evidence on suspicious transactions consisting of stocks, puts and calls that justify the belief that there were certain investors that had prior knowledge about the attacks. Undoubtedly, whoever possessed this kind of advanced information must be connected with the planning and organization of the attacks themselves. Suspicious transactions are based on criteria like: (1) Good timing, for instance someone who shorts stocks just a few days before the attacks and, thus, before the stock experiences a major drop in its price (2) Nature of transactions, which has to be very specific, for example someone buys puts on the 9/11 affected airlines and not on others; here investigators can come to the conclusion that the buyer knew in advance which the targets of the attack would be, (3) Large transactions, this constitutes the most reliable indicator for investigators; if someone who usually makes small transactions suddenly starts trading large amounts, his trades are consider abnormal and may be evidence of suspicious inside knowledge, and (4) Too speculative transactions, which turned out to become valuable immediately after the 9/11 attacks. Radlaur, however, discusses some possible limitations if someone wishes to trace the transactions. He claims that anyone associated with the securities transactions would have been clever enough to cover his identity behind false names and corporations. Hence, authorities will also have to deal with this burden.
Maria Kyriakou 15 author, also, claims that some of these banks provide their clients with more privileges such as offering them an anonymous cash card that allows withdrawals all around the world. Of course authorities have tried to take action and have tried to implement a more transparent banking system. But most of their attempts have been in vague because of the lack of cooperation among countries. As stated previously, one of the main problems to this lack of cooperation is that there is no standard definition of what constitutes terrorism. As a result each country has its own interpretation. This in turn results in different suggestions on how to deal with terrorism.
2.1 6 Security projects: A way to mitigate terrorist risk
A terrorist attack can have negative consequences on a significant number of financial assets and market participants, as discussed in sub-section 2.1 4. Stock prices of directly affected firms would clearly be harmed by an attack, which is translated to direct asset losses. In turn, a fall in asset prices increases the cost of borrowing, since the firm is regarded as a riskier borrower. To mitigate these negative effects and to insure their business continuity, firms must construct investment and financial strategies to deal with potential terrorist attacks. Most of the times, firms employ strategies that entail increases of security measures. However, they do not, always, have incentives to invest in security projects, especially when the latter produce positive externalities to other firms. Of course externalities could be internalized to the firm but this cannot be done automatically. Usually governors and regulators have to take measures to protect the firm. One solution to this is for governors to implement fines to those firms that do not invest in terror protection. Another solution is to let governors assess the frequency of terrorist attacks and, in effect, inform firms.
Literature provides evidence on the benefits of security protection measures with respect to terrorist attacks. Karolyi [2007] investigates the risk and return tradeoff of terrorism- related investment strategies in the U.S. during 1994-2006. He constructs two different portfolios: one which has to do with stocks that are located in countries with high level of terrorist activities on the basis of a terrorism-related risk scores and a second
terror-free portfolio, which screens out individual holdings of firms that are considered to operate in
countries that sponsor terrorism5. He finds a positive 0.16% risk premium per month for the former and -0.016% for the latter; both of them are statistically insignificant, though. He comes to the conclusion that terrorist related investments do not add value bearing in mind that the risk borne for both portfolios is almost of the same magnitude. Krug and Reinmoeller [2003] claim that if a firm invests in terrorist protection, its asset value will increase.
5
Maria Kyriakou 16 However, not all firms face the same vulnerability in being attacked, so different measures apply to different firms. No matter what the vulnerability of each firm is, it is worth repeating that terror is not a natural hazard that firms cannot mitigate. All firms can take measures to protect themselves. According to Krug and Reinmoeller there are three characteristics which, best, describe a highly vulnerable firm: (1) Exposure, the better known a firm, the higher is the risk. So for a firm to get protected it could be wise to stay in obscurity or hide behind local joint venture partners, (2) Geographic Spread, terrorists do not have to travel any more if they want to attack a certain company; they can instead attack its subsidiary. A possible protective strategy could be to relocate production plants in countries with low level of terrorism, (3)
Organizational form, certain kinds of organizations such as the lean governance structure can
make a firm vulnerable. The remedy would be an organizational slack. The aforementioned, possible protection measures describe the degree to which a firm is willing to bear terrorist risk.
Maria Kyriakou 17 electronic attacks. But, why do firms not invest in security projects? Heal and Kunreuther [2003] provide an answer to this when they claim that the incentives for a firm to carry on security projects are an increasing function of how many other firms have done so. If a critical number of firms invest in such projects, then all firms will want to act similarly.
2.1 7 Behavioral Explanations
Sunstein [2003] elaborates on the relationship between terrorism and probability neglect. The term “probability neglect” is defined as the overreaction of people to low probability events. One example that the author uses is the case of the anthrax scare of October 2001 which cost the life to a few people. However, people turned excessively pessimistic and anxious about the probability of getting infected. They focused on the consequences of getting ill rather than on the probability of the event. They got over-concerned about low probability events and less concerned about every day problems or events that are more probable to occur. Sunstein, also, claims that terrorists exploit this behavioral bias of investors by spreading fear about imminent terrorist attacks which in fact outrun statistical reality. His study has implications for governors on how to react and what measures to take to dampen concern of the public about the threat. He suggests that they inform and educate people rather than take regulatory measures. A study that is contradicting Sunstein [2003] is the one by Fischhoff et al. [2003]. They analyze the findings from a survey that was conducted in November 2001 in the US and aimed to sketch how people perceived terror risk during the following year. Results indicate that they consider themselves as facing less risk than the average person of the group to which they belong. This finding reveals emotional or cognitive biases. In general, subjects that lived outside the area of the 9/11 attacks, perceived themselves as less probable to suffer risk from terror compared to the ones that lived close to the attacked area. The heterogeneity of the results must be taken into account if regulators want to follow an effective strategy to confront the risk of terrorist attacks.
Maria Kyriakou 18 analyzed this cognitive bias. They argue that salience is one parameter that affects investor’s judgement under uncertainty. They use the following example: If someone sees a house burning their subjective probability of such an event taking place is greater than the impact of reading about a fire in a newspaper. Furthermore, the subjective probability increases if an event occurred recently. For imperfectly informed agents, the use of the availability heuristic may or may not be consistent with the Bayesian use of information. In either case it produces significantly exaggerated judgments about the probability of getting harmed. Hindsight bias is also an outcome of the availability heuristic. It explains the tendency of people to believe that they knew about a risk even though the risk was completely random and unanticipated. Finally, the embeddedness effect may also have a place here. According to this, whether people are asked about the perceived riskiness of category A and about the riskiness of a larger category B in which A is embedded, the same result is got. So, when investors think about all the risks they are likely to face and at the same time they assess the probability of the terrorist threat they give the same result. The problem is that terrorist attacks force individuals to completely alter their risk beliefs. Before a terrorist event they do not even consider it among their possible outcomes and their outcome set is rather incomplete. But, after experiencing an attack, they anticipate such events with high probability. Finally, it has frequently been stated that investors tend to extrapolate when examining past stock returns [see, for example, Nofsinger, 2008, p.5]. This implies that if a stock performed badly in the recent past, investors regard it as a loser and presume that it will continue to perform badly in the future. The opposite applies for good performing stocks. As regards to terrorist events, an attacked firm that sees its stock performing badly for the period following the attack will continue to deliver negative returns, if investors extrapolate when examining past performance.
2.1 8 Summary
The main findings of the literature review can be summarized as follows:
• The parties involved in the fight against terrorism are terrorists, firms and governments. • The channel through which stock markets are affected is double. First, the perceived
riskiness of equity is increased. Secondly, investors are more risk averse and, thus, less willing to bear equity risk.
• Terrorist risk entails uncertainty; this uncertainty must not result in letting terrorist risk outside corporate decision making.
Maria Kyriakou 19
2.2 Empirics
2.2 1 Introduction
This section discusses the up to date empirical literature on stock market reactions to terrorist attacks.
2.2 2 Global stock markets and recent terrorist attacks
Walker et al. [2005] examine the impact of 138 aviation disasters that took place between 1962-2003 on the short and long term performance of airline and airplane manufacturers’ stocks. Among the events studied, there were the 9/11 terrorist attacks, as well as other terrorist attacks. They find that airline stocks experience a negative cumulative abnormal return in of 2.8% within the first trading day after the 9/11 disaster, followed by an additional 1.08% decline in the week to follow. What is more, the cumulative abnormal returns persist until six months after the disaster. The long term performance of airline stocks is not affected, though. If compared to airlines, airplane manufacturers experience an equivalent drop of 0.8% on the event day and a 1% decline during the first seven trading days following the event. However, they recover rather earlier than airlines. Walker et al. measure the impact of the disaster with respect to the number of fatalities it causes or if it is a criminal activity such as the 9/11 disaster. They find that the bigger the impact of the event, the larger the stock price decrease. They suggest a clear violation of the efficient market hypothesis that states that stock prices should immediately adjust to an announcement and that delayed price declines should not be observed. They conclude that the stock price of both airlines and airplane manufacturers is expected to gradually drop after an event. Hence, investors can exploit this drop by selling short the airline’s stock on the announcement day and buying it back about one week later.
Chen and Siems [2004] have studied the link between various financial markets reactions and 14 terrorist and military attacks as from 1915. They apply an event-study methodology to calculate the number of days needed for the US and several global stock markets to recover after an attack. They find that 1 to 3 days is, usually, needed for the market value to recover in most of the 33 capital markets studied. However, regarding the 9/11 incident, the stock market recovered in 40 days. They, thus, conclude that the more important a terrorist attack is the more days a market needs to recover.They also find that foreign stock markets were more adversely affected by “9/11” and the Iraqi invasion than were the US markets. Finally, they suggest that U. S. markets are more resilient than they used to be in the past and they recover sooner from terrorist attacks than other markets do.
Maria Kyriakou 20 buybacks that were announced shortly after the 9/11 attack caused any positive or negative effects on the financial markets. They claim that buybacks in general signal negative information about a firm because they reduce its cash and equity. But as an exception, in the case of the 9/11 terrorist attacks, buybacks contributed to the stabilization of the stock market. Buyback announcements that occurred within two weeks after the terrorist attack led to positive returns of the firm’s stock. They argue that firms that were eager to show “patriotism” were rewarded by “patriotic” investors, which means that investors perceived the buyback announcement as good news. Cummins and Lewis [2003] focus on the stock performance of insurance firms following the 9/11 attacks and find strong negative stock responses, as well as, an increased volatility of abnormal returns. Karolyi and Martell [2006] examine, with the use of the event study methodology, the stock price impact of 75 terrorist attacks on international firms during 1995-2002. They uncover a statistically significant
0.83% stock price drop on the day of the attack which is equivalent to a market capitalization
loss of $401 million. They also find that the impact of the attack depends on the country where the event took place. More specifically, the wealthier a country is, the larger the stock price responds. Finally, they provide evidence that terrorist attacks such as kidnappings of firm’s executives, are associated with larger drops in the firm’s stock price as compared to attacks that comprise physical losses.
Among others, Barkoulas et al. [2005] study the short and long term effects of the 9/11 attack in 33 industrial and emerging economies. Their results indicate a statistically significant short term negative market reaction in 28 of the economies. Furthermore they find an increase in the level of systematic risk as measured by beta in 10 of the 33 economies. They conclude that the financial crisis caused by the 9/11 attack had only transitory consequences. They attribute this conclusion to the regulatory measures taken by US and international governors and financial institutions. Another study that attempts to assess the increase of betas is the one by Choudhry [2005]. He measures the time varying beta of 20 US firms and wants to investigate whether the 9/11 attacks had any effects on the beta of these firms. He claims that time varying beta is affected by changes in general economic conditions, business conditions specific to the firm or expectations about potential terrorist attacks in the future. His results indicate that for most of the studied firms there was indeed a change in their time varying beta. For some of the firms it was even lower than before the 9/11 events, implying that their systematic risk had decreased.
Maria Kyriakou 21 negative returns. Zussman and Zussman [2006] evaluate the impact of Israeli assassinations of Palestinian military and political leaders on Israeli and Palestinian stock markets. They include terrorist attacks as one of their control variables and find a negative effect on Israeli and Palestinian stock markets. Eldor and Melnick [2004] study the effect of 639 terrorist attacks on the stock market of Tel Aviv in Israel from 1990 to 2003. They distinguish location, type of attack and target, number of casualties and number of attacks per day and find that only suicide attacks have a significant influence on stock market losses and this effect is found to be permanent. Indeed, they estimate that Tel Aviv Stock Market lost 30% of its value on 30th June 2003 due to a terrorist campaign. This loss if translated to decreasing expected profits of firms show that terrorist attacks have real economic costs. Finally, there is an experimental study by Glaser and Weber [2005]. They use questionnaires to assess the expected returns and volatility estimates of individual investors, before and after the 9/11 attacks. They find higher expected returns after the attacks compared to those before the attacks. They suggest that investors believe in mean reversion, which indicates that investors perceive the sharp drop in stock prices after the attacks as temporary. They also find that volatility estimates have increased after the events.
TABLE I
Studies on the effects of terrorism on stock markets
The table shows the authors of the empirical paper, the methodology employed in the study, the topic, the sample date and size used, the findings and the significance of results. If applicable, *, **, ***or a **** indicate whether the results are statistically significant at 10%, 5%, 1% and 0.1% level respectively.
Study Method Topic Sample Date Sample Size Findings Significance of results
Abadie and Gardeazabal [2003]
case study for Spain (Fama-French 3-factor model)
how beginnings and endings of cease-fires in the Basque region of Spain affected stock prices of firms with
business interests in that region 1998-2000 73 firms
cease-fires are correlated with positive returns while ends of cease-fires are associated with negative returns
Basque stocks outperform non-Basque stocks when cease-fires
are getting credible**
Barkoulas et al. [2005] ICAPM
international short and long term stock
market reactions of the 9/11 attacks 2001-2002
33 developed and emerging
markets
negative short term impact in 28 countries and an increase of beta in 10 countries; long-term impact is transitory
Return 28 countries*** Beta 10 countries***
Chen and Siems [2004] event studies
response of global capital markets to
terrorist and military attacks 1915-2001 14 attacks
terrorist and military attacks affect capital markets in the short run; usually 1-3 days is needed for a market to recover following an attack except for the case of 9/11 where 40 days were needed.
Event day negative AR*** at 31 of 33 markets, and AR* at 2 of 33
markets following 9/11 attacks;
Choudhry [2005]
bivariate MA-GARCH
9/11 effetcs on the time varying beta of
US firms 1991-2002 20 firms
the post 9/11 period experienced a rise in time varying beta in most firms which made them more risky
Statistically significant at *,** or *** depending on the firm.
Cummins and Lewis [2003] Event studies
9/11 attacks on the stock price performance of US property-casualty
insurers 2001 43 US insurers
Strong negative impact of insurers’ stock prices, increased volatility of abnormal returns after the attacks, insurers with strong financial ratings recovered within the first post-event week contrary to weaker insurers
Maria Kyriakou 23
TABLE I (cont.)
Studies on the effects of terrorism on stock markets
Study Method Topic Sample Date Sample Size Findings Significance of results
Eldor and Melnick [2004]
time series methods
influence of terrorist attacks on the
Israeli stock market 1990-2003 639 attacks
only suicide attacks have a
significant influence Suicide attacks**
Glaser and Weber [2005] experimental study
investors were asked to give forecasts concerning the stock market before
and after the 9/11 attacks 2001 86 respondents
higher expected returns after the attacks compared to those before
the attacks N/A
Gu and Schinski [2003]
Market adjusted returns
Stock repurchase announcements
during the two weeks following the 9/11 attacks and their effects on the
firm’s stock price. 2001 316 firms
Positive market response to repurchase announcements; the earlier the announcement, the more positive the market response
Significant positive 1-day and 5-day AARs** Cross-sectional analysis with respect to firm size and relative
value: Not significant
Karolyi and Martell [2006] event studies stock price impact of terrorist attacks 1995-2002 75 attacks
they uncover a -0.83% stock price reaction on the first day of the attack (9/11 excluded); the wealthier a country is, the larger the drop in stock price
Excluding 9/11 attacks:-0.83% 1-day AAR***; all other 1-days’ AAR
not statistically significant Including 9/11 attacks:-2.24%
1-day AAR**
Walker et al. [2005] event studies
airlines and airplane manufacturers stock performance following aviation
disasters 1962-2003 138 disasters
Airlines: a -2.8% CAR on the first day after the 9/11 attacks and an additional -1.08% on the week to follow; the equivalent for airplane manufacturers is 0.8% and 1%, respectively.
1-day CAR airlines**** 1-day CAR manufacturers***
7-day CAR airlines**** 7-day CAR manufacturers*
Zussman and Zussman [2006] event studies
impact of Israeli assassinations of Palestinian military and political leaders on Israeli and Palestinian stock
markets. 2000-2004
159 assasination attempts
markets does not react to assassinations of low-ranked members of Palestinian terrorist groups; they drop sharply when the assassination has to do with senior political or military leaders
Israeli & Palestenian market reaction to senior political leader
assasination*** Senior military target** (Israel)
3. METHODOLOGY
3.1 Introduction
This section will discuss the methodology employed for the thesis which, in effect is a standard market model event-study methodology.
3.2 Definition of an Event Study
Maria Kyriakou 26 immediately reassess the value of individual firms which results in a new stock price. The event study methodology attempts to capture this change in stock price after a specific event has occurred.
3.3 Models for Modelling Normal Stock Returns in an Event Study
Important in event studies is to identify the events, the selection criteria for a firm to be included in the study and the stock’s abnormal return. The latter is defined as the extra return the stock realized, given the occurrence of the event. It has to be contrasted to the normal
return which is the expected return that the stock would have earned had the event not taken
place. There are two main approaches for modelling normal returns (see, for example Brown and Warner [1985] and MacKinlay [1997]), the Constant Mean Return Model and the Market
Model6. The former utilizes the average price of the firm’s stock which is assumed to be constant, while the latter uses the market return. A brief overview of the Constant Mean Return Model is provided here, followed by a detailed elaboration of the Market Model, which is the one employed for this study.
3.2 1 Constant Mean Return Model
It constitutes the simplest method for modeling the normal return. One simply needs to subtract a stock’s
i
time series average over the estimation period from the actual return of the stock on the day of the eventt
. In mathematical representation this is equivalent to:, , i
i t i t
A R
=
R
−
R
(1)where ARi,t is the abnormal return of stock i in time t, Ri,t is the actual return of firm i in time t
and
R
iis the simple average return of firm i over an estimation period that is usually from -6 to -244 days relative to the event.
6 MacKinlay [1997] discusses other statistical models to measure normal returns which include factor models.
Maria Kyriakou 27
3.2 2 Market Model
The market model is the most popular way of predicting normal returns in an event study. In this case stock returns
R
i,t have to be regressed against a domestic market indexR
m,t over an “estimation period” t. Following Brown and Warner [1985] the estimation period will consist of a maximum number of 239 observations which will be from -6 to -244 days relative to the event. The event is designated as day “0”. The period -5 to +5 is designated the “eventperiod” τ . An illustrated overview of the current event study time line is given in Figure I.
FIGURE I
Time Line for Current Event Study
The Figure performs the time line used for the current study. The estimation window includes stock observations from day -244 to day -6 relative to a terrorist event that occurred at time 0. The period -5 to +5 is the event window.
MacKinlay [1997] claims that the event itself must not be included in the estimation period; otherwise the normal parameter estimates will be influenced. After regressing Ri,t on Rm,t one
derives the market parameters αi andβi for each of the affected firms’ stock prices. Regression
can be done by using standard Ordinary Least Square (OLS) procedures and software packages as EViews. The superiority of the market model to the constant mean return model has been well documented in the literature. MacKinlay [1997] discusses that the market model delivers a smaller variance of abnormal returns, implying a better ability for the model to detect abnormal returns. In addition, Binder [1998] argues that constant mean return models, contrary to market models, do not explicitly control for the stock’s risk or the market return over the estimation period. Finally, Brown and Warner [1985], claim that methodologies which are based on the OLS model and standard parametric tests are well specified under non-normally distributed daily observations. Taking into account all the aforementioned advantages, the market model will constitute the model employed for this thesis. Equation (2) performs it in mathematical representation.
Estimation window(t)
-244 -5 0 +5
Maria Kyriakou 28 , , , i t i i m t i t
R
=
α
+
β
R
+
ε
(2) E(εit = 0) var(εit) = σ2 εi Where:Ri,t = the actual return for stock i at time t,
Rm,t = is the domestic market return at time t, αi = the intercept term,
βi = the systematic risk of stock i,
εi,t = the error term,
σ2εi =the variance of the error term.
3.4 Measuring Abnormal Stock Returns in an Event Study
Equation (2) is used to measure normal returns relative to a domestic market index over the estimation period. The step that follows is to detect abnormal returns during the event period. This can be done by using Equation (3) as defined by Brown and Warner [1985], which subtracts the expected return from the actual return and predicts expected returns over the event period:
, , i i ,
i t i t m t
A R = R − α − β R (3)
where ARi,t is the abnormal return for stock i at time t,which actually is the error term
ε
i,t of the market model,α
i andβ
i are the OLS estimated parameters of the market model, Ri,t isthe actual return of security iat time t , and Rm,t is the domestic market return at time t
.
Following MacKinlay [1997] the variance of the abnormal returns of one day in the event period can be calculated with the use of Equation (4):2 , 2 , , 2 , 1 ( ) 1 v a r ( ) ( (1 ] ) ( ) m t M i t i T m t M t R R A R T R R ε σ = − = + + −
∑
(4)where
σ
2ε,i is the variance of residual term from the OLS estimated model given inEquation (2), T is the number of days in the estimation period ( a maximum number for T is 239), Rm,t is the domestic market return at time t
,
and RMis the mean market return over theMaria Kyriakou 29 From Equation (4) we notice that the variance of abnormal returns has two components: the variance of the error term and an additional variance due to the sampling error of
α
i andβ
i.From the second component we notice that the variance of abnormal returns depends on the length of the estimation period, such that longer estimation periods lead to smaller variances. According to MacKinlay if the estimation period is large enough (i.e.: it consists of more than 30 observations) then (4) converges to
σ
2ε,i . In our case, the estimation period consists of amaximum number of 239 observations, number sufficient to drop the second component. Thus, the variance of the abnormal returns of one day in the event period will be:
2
, ,
var(
AR
i t)
=
σ
εi (5)If assuming that a terrorist event has no impact on any stock’s abnormal return, then under H0
the abnormal return of any of our sample’s firm’s stock must be distributed as:
2
,
(0,
(
,))
i t i t
AR
N
σ
AR
(6)3.5 Aggregation of Abnormal Returns
3.5 1 Across Securities: Average Abnormal Returns(AARs)
Abnormal Returns can be aggregated across securities so as we obtain Average Abnormal
Returns over the event interval:
, 1 1 N t i t i A A R A R N = =
∑
(7)N is the number of securities and t is a day in the event interval.
We can get as many AARs as the number of days of the studied event period. In this case we can get AAR-5, AAR-4,…,AAR+5 , a total number of 11 AARs.
The equivalent variance of each of the AARs is:
2 2, 1 1 var( ) N t i i AAR N =
σ
ε =∑
(8)Finally, to compute the significance of AARs, the following t-test is used: ( 0 , 1 ) v a r ( ) t t A A R t N A A R = (9)
Maria Kyriakou 30
3.5 2 Over the Event Window: Cumulative Abnormal Returns(CARs)
Abnormal returns can also be aggregated over the event window for each firm separately, so as we get the Cumulative Abnormal Return of security i over an event period (τ1, τ2) and its
equivalent variance: 1 2 1 2 , ( , ) i i t t CAR AR τ τ τ τ = =
∑
(10) ( , var( i ) var( i t) t CAR AR τ τ τ τ 2 1, 2) = 1 =∑
(11)3.5 3 Across Securities and Over the Event Window: Cumulative Average Abnormal Returns(CAARs)
To calculate the Cumulative Average Abnormal Return (CAAR) for a sample of N stocks and its equivalent variance one can utilize Equations (12) and (13) respectively:
1 , 2 ( ) ( 1 , 2 ) 1 1 N i i C A A R C A R N τ τ τ τ = =
∑
(12) ( 1, 2 ) 2 ( 1, 2 ) 1 1 var( ) var( ) N i i C A A R C A R N τ τ τ τ = =∑
(13)Finally, to test the null hypothesis that the CAAR on any given time interval during the event window is equal to zero, I use a Student’s t-test, which under the hypothesis of zero returns is of the form: ( 1 , 2 ) ( 1 , 2 ) ( , 1) v a r ( ) a d f N C A A R t t C A A R τ τ τ τ = − = (14)
Assuming that the return will be negative the t-test is a one-tailed one.
3.7 Hypotheses
Maria Kyriakou 31
Hypothesis 1
H0: The day-0 AAR should not be statistically different from zero.
H1: The day-0 AAR should be negative.
Assuming semi-strong form efficiency in the marketplace, investors must instantaneously absorb any new publicly available information that reaches the market. This information will be translated to new and correct stock prices in a way that no individual investor can earn abnormal returns following a disaster. The change in stock prices, among others, reflects the affected firms increased security costs or some updated market beliefs about its investment policy. In reality, there are frictions in the market and information does not get
instantaneously available to all market participants. However, I expect that information about
an attack will be available to all investors within the first trading day of the attack. That must be the time needed for stock prices to adjust to their new efficient levels. Thus, the day-0 AAR must be the most pronounced (compared to other-day AARs) and differ statistically from zero.
Hypothesis 2
H0: Terrorist attacks do not reduce stock prices on the 5 days immediately following the
attacks.
H1: Terrorist attacks do reduce stock prices on the 5 days immediately following the attacks.
The second testable hypothesis is a sequence of the first one. It indicates that assuming efficient markets, all post-event AARs and CAARs have to be zero. According to the efficient markets hypothesis, all stock market prices should instantaneously and correctly adjust to new information and subsequent stock price reactions should not be observed. However, I suspect that the findings will indicate negative AARs and CAARs.
Hypothesis 3
H0: News about a terrorist attack is considered unanticipated, thus, price movements before
the event should not be observed.
Maria Kyriakou 32 A terrorist attack is an exogenous event and no one can profitably trade on it before it happens7. Thus, AARs and CAARs must not statistically differ from zero before the occurrence of such an event.
Hypothesis 4
H0: Terrorist attacks do not increase risk of the affected firm in the days immediately
following the attacks.
H1: Terrorist attacks increase risk of the affected firm in the days immediately following the
attacks.
This hypothesis complements Hypotheses 1 and 2. If there is evidence of significant post event abnormal returns, there is a need to test if increased volatility can explain them. To draw results, each event’s day abnormal volatility can be compared with the volatility of the actual returns observed during the estimation period.
Hypothesis 5
H0: Higher number of casualties in an attack will not result in lower stock prices.
H1: Higher number of casualties will result in lower stock prices.
Hypothesis 6
H0: If the attacked firm’s stock trades in an emerging stock market, the drop in stock price
will not be higher.
H1: If the attacked firm’s stock trades in an emerging stock market, the drop in stock price
will be higher.
The last two sets of hypotheses test for any asymmetric affects on the basis of certain characteristics of the affected firm’s stock or the event itself. Rational and efficient markets suggest that firms that got more exposed after an attack8 will experience more pronounced stock price reactions than firms with less exposure. The number of casualties is used as a variable to assess the magnitude of the event. In the case of stocks that trade in emerging markets, more pronounced post attack effects can be explained by a generalized political instability and poverty of these markets.
7
Of course, as discussed in the literature review, anyone that is involved in a terrorist attack planning may possess superior information and trade profitably before the event by selling short the stocks of the affected firms. However, I assume that this is only the case in large scale attacks like the 9/11 ones.
8
Maria Kyriakou 33
4. DATA
The thesis studies terrorist events that occurred during the period 2003-2007 and affected publicly trading firms. We are, thus, interested in obtaining two different kinds of data: (1) information about terrorist events that occurred during the studied period and affected publicly trading firms and (2) historical stock and domestic market prices. A first raw sample of terrorist attacks includes all terrorist events that occurred worldwide during 2003-2007. From this sample it is essential to identify which of these terrorist events affect businesses. Further, the next step is to check which of the target firms are publicly listed in a stock market. In this section, I first, describe the process for obtaining information with respect to terrorist events. I, then, state how I acquired information regarding the nature of each of the affected firms, being it publicly or not. Finally, the current section finishes with the process followed to acquire historical stock prices of the affected firms and their equivalent domestic market index.
4.1 Terrorist Events 2003-2007
Data with respect to terrorist events have been retrieved from two on-line databases: (1)
Global Terrorism Database (online at: http://www.start.umd.edu/data/gtd/ ) (period
2003-2004) and (2) National Counterterrorism Center (online at: http://www.nctc.gov/) (period 2005-2007).
The Global Terrorism Database (GTD) contains data on terrorist events up to 2004. Data for the period 1/1/2003- 31/12/2004 are retrieved from this database (access date 1/5/2008). It is an open-source database and contains terrorist events as from 1970. It currently includes some 80,000 cases. The whole database is divided into two sub-categories. The first one is called GTD1 and contains cases from 1970 to 1997. The second one, which is the one used for this thesis, is called GTD2 and contains events that occurred during 1998-2004. The advantage of the GTD is that one can browse events by date, location, attack type, target type, perpetrator type, weapon type and number of casualties. Further on, in the case of target type, the database distinguishes among different types of targets including businesses. In the GTD2 sample there are 7,703 attacks from which business-related account for 9%9. When picking the 2003-2004 business-related events, I come up with a sample containing 114 raw observations. From the 114 observations when looking up for a specific business name in the event description, the sample reduces to 81 observations. The 33 remaining observations
9
