MSc Business Economics
Master Thesis
The impact of terrorism on stock markets: differentiation
based on attack characteristics
Student:
Stefan van den Born
Student ID:
10002546
Supervisor:
P.F.A. Tuijp MPhil
Date:
July 27
th, 2015
Abstract
The main focus of this research is to study the differential effect of terrorism on stock
markets based on different characteristics of terrorist attacks. I consider 699 terrorist
attacks that occurred in the period of 1974 – 2014 and targeted the U.S. and European
countries. I implement a non-‐parametric methodology to identify abnormal movements
in event-‐day returns. Next, I determine whether the probability associated with these
abnormal movements depends on the nature of terrorist events for different industries.
The main results suggest that the impact of terrorism on stock market returns varies per
attack characteristic and across industries. The effect on returns of the life insurance
industry shows the highest dependency on the nature of attacks, and this is least the
case for the airline industry. In addition, I show how the differentiation based on attack
characteristics can be used for hedging strategies against terrorism risk.
Statement of Originality
This document is written by Stefan van den Born who declares to take full responsibility for the contents of this document.
I declare that the text and the work presented in this document is original and that no sources other than those mentioned in the text and its references have been used in creating it.
The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.
Table of contents
1.
Introduction ... 52. Literature Review ... 6
3. Methodology ... 11
4. Description of data ... 14
5. Empirical analysis ... 17 5.1. Main results ... 17 5.1.1. Aero/Defense ... 18 5.1.2. Life Insurance ... 19 5.1.3. Non-‐Life Insurance ... 20 5.1.4. Travel/Tourism/Leisure ... 22 5.1.5. Airline ... 22
5.1.6. Construction & Materials ... 24
5.1.7. Relation to the first hypothesis ... 24
5.2. Robustness analysis ... 25 5.2.1. Aero/Defense ... 25 5.2.2. Life Insurance ... 27 5.2.3. Non-‐Life Insurance ... 27 5.2.4. Travel/Tourism/Leisure ... 28 5.2.5. Airline ... 28
5.2.6. Construction & Materials ... 28
5.2.7. Hedging analysis ... 30
5.2.8. Relation to the second hypothesis ... 32
5.3. Discussion ... 32 6. Conclusion ... 34 7. Reference list ... 36
1. Introduction
Over the last few decades the threat of terrorism has become a matter to which governments are increasingly forced to pay attention. The 9/11 attacks have had a major impact, not only nationally but also on a global scale. Conventional economic literature on terrorism mostly focuses on the economic and financial costs associated with terrorist attacks [see for examples Eldor & Melnick (2004), Kollias et al. (2013), and Arin et al. (2008)]. These existing studies argue that terrorist activities form a significant potential threat to companies, the stability of financial markets, and even to a nation’s overall economy
However, in the widely existing literature on this matter some interesting questions still remain unsolved. One of these thinly explored subtopics includes the differential effect that terrorist attacks with different characteristics may have on stock markets. Stock price reactions may be terrorist activity-‐specific, in the sense that certain types of attacks may generally affect different stocks than attacks with other characteristics. For example, attacks that involved armed assaults may affect stock markets differently than attacks in which bombings/explosions took place. Furthermore, another interesting topic in this field of research that may require more investigation is the existence of potential opportunities to hedge against different types of terrorism. Firms and investors, especially those that are financially vulnerable to terrorist attacks, might be able to hedge (part of) the risk that is associated with terrorism, by investing in stocks that are less negatively –or even positively– impacted by terrorist activities. In this way investors may be able to protect their portfolios from severe losses resulting from a terrorist attack.
The first aim of this paper is to investigate whether a distinction can be made in the effect of terrorist attacks on stock prices, based on distinctive characteristics of these attacks. The central question to this study is therefore: ‘To what extent are stock markets affected differently by terrorist activity, based on the distinctive characteristics of these attacks’? The next objective is to investigate if there may be potential hedging opportunities against terrorism risk, so that investors that are particularly exposed to this type/these different types of risk may be able to construct their investment portfolios in a way that is (to some extent) terrorism-‐ proof.
In this study I analyze a sample of 699 events in the period 1974 – 2014 and identify which events resulted in abnormal negative movements in event-‐day returns for six different industries that are found to be relevant for this particular research topic by previous studies. I use a non-‐parametric approach, which is a relatively new methodology in this research field. Previous related studies have mainly analyzed the effect of terrorism on stock markets by using more conventional approaches such as the event-‐study and the GARCH-‐approach (Cam, 2008;
Nguyen and Enomoto, 2009; and Bashir et al., 2013). After I determined whether terrorist attacks in the sample led to abnormal negative event-‐day returns, I classify the attacks by their nature and examine whether the effect on different stock markets is different for attacks with varying characteristics. The results of this study may then be useful in understanding which type of attacks a particular industry is exposed to. This study differentiates the effect of terrorist attacks, whereas existing literature typically viewed terrorist activity as a homogenous type of exogenous shock. In this way this paper may be an interesting contribution.
In this study I also aim to identify a strategy to hedge against different types of attacks, for industries that seem particularly vulnerable. In a simplistic way I construct a portfolio of stocks that experience a less pronounced negative impact of terrorism, and compare the returns with other stocks over different time windows. With the results I intend to give a direction of where investors and industries should look for strategies to minimize their exposure to terrorism risk. The negative effect of terrorism on stock returns is widely found in existing literature, as highlighted in studies, among others, by Glaser and Weber, 2005; Karolyi and Martel, 2010; and Charles and Darné, 2006.
The results of this research suggest that the extent to which industries are exposed to terrorism risk depends on the characteristics of terrorist attacks. Previous studies have found that particular industries that are significantly impacted by terrorism include the airline, insurance, tourism, and defense industries. Among others, these industries are taken into consideration in this study as well. Across industries I find no evidence of oversensitivity to terrorism of one sector relative to the others. The effect on the life insurance industry in found to be especially dependent on attack characteristics, which is least the case for the airline industry. The hedging analysis suggests that risk imposed by terrorism can be reduced on average by investing in a portfolio that constitutes and index for the aero/defense industry and the S&P 500.
This paper is organized as follows. Section 2 provides an overview of the existing literature. Section 3 gives a description of the non-‐parametric methodology and model used in this study. Section 4 describes the data. Section 5 provides the results of the empirical analysis of the differential impact of terrorism on stock markets, including robustness checks and a discussion on the analysis. Section 6 concludes.
2. Literature review
The existing literature on the effect of terrorism on financial markets is extensive. Abadie and Gareazabal (2008) look into the effect of terrorism on a macro-‐economic scale. They provide evidence that terrorist activities may induce large capital movements across international
markets, and that they have a detrimental effect on net foreign investment positions. They argue that mobility of capital in an open economy is a key determinant in the short-‐ and long-‐run impact of terrorism. With their model they show that terrorist activity increases financial uncertainty and that it may have a detrimental effect on the expected return to investment. Abadie and Gardeazabal (2003) conduct a case study on the terrorist events that occurred in the Basque Country and were caused by the terrorist organization ETA since the late 1960’s through the end of the twentieth century. Over this period they find an average decrease of 10 percent in the Basque per capita GDP as compared to a similar synthetic region, that doesn’t suffer from terrorist activity. Melnick and Eldor (2010) found evidence that media coverage of terrorist attacks influences the extent to which terrorism causes damage to an economy. They argue that there is a trade-‐off between the role of free press to provide the public in its right to know, and the averse effect this free media coverage on terrorism may have from an economic perspective. The view that terrorism adversely affects stock markets is widely accepted and supported by existing literature. Chesney et al. (2011) find that terrorist events significantly impact stock markets, as well as bonds and commodities. The overall effect is found to be negative and comparable to that of a natural disaster or a financial crash. By conducting three different methods (an event-‐study, a filtered GARCH-‐EVT and a non-‐parametric approach) they investigate the effect of 77 terrorist events on financial markets. They find that specific sectors such as the insurance sector and the airline industry especially have negative financial exposure to terrorist events. The impact on other sectors is found to be inconclusive. Finally the findings show a common positive effect on the aero/defense, pharma/biotech and gas/oil stocks industry, which indicates the existence of potential hedging strategies.
Arin et al. (2008) find evidence that terrorism has a significant impact on stock market return as well as stock market volatility. The overall effect is found to be negative and larger in emerging markets. Stock and bond market volatility are the main focus in a study by Kollias et al. (2013). They particularly concentrate on how the relationship between the two markets is affected by exogenous shocks caused by terrorist attacks in Spain, Great Britain, France and Germany. They find evidence of a flight-‐to-‐quality/safety. In particular, their findings indicate that investors in the four sample countries tend to move the allocation of their funds away from stock markets, towards bond markets that show relatively higher volatility persistence. They also uncover the different impact between domestic and transnational terrorist attacks, finding that financial market’s stability was more prone to domestic terrorist attacks than to conflicts that took place in other countries.
Drakos (2004) shows in his paper that the 9/11 attacks negatively affected airline stocks listed on various international stock markets. He found that on these particular stock markets systematic as well as idiosyncratic risk had significantly increased as a result of the attacks.
Specifically, by decomposing total risk into its systematic and idiosyncratic parts, he finds evidence that systematic risk as measured by beta has more than doubled on average in the airline industries that are investigated in this study. Also, his findings indicate that the systematic part of total risk has increased significantly in ratio to diversifiable risk. The evidence in this paper marks the importance for the airline sector, or for investors that heavily invest in this industry to construct their portfolios in such a way that they are able to offset the increased volatility and uncertainty they may have to deal with due to terrorism. Research that helps develop such portfolio strategies is interesting to this group of investors that is particularly exposed to terrorism risk.
Haque and Varela (2001) apply a so-‐called safety-‐first portfolio theory in a setting that is exposed to financial risk caused by terrorism. As in many studies in this research field, their main focus are the 9/11 attacks. They define safety-‐first as the concept that is largely described in the literature on portfolio theory, that aims to bound the risks of unfavorable outcomes. Their findings include that optimal ex ante safety-‐first portfolios on 9/11 have high U.S. weights, and on july 7, 2006 (thus, ex post) low U.S. weights. They find that investor’s wealth is retained even without the ex post optimal portfolios. The practical implication is that these safety-‐first portfolios can provide coverage against loss of wealth caused by unlikely but extreme events such as the 9/11 attacks.
The study by Brounrn and Derwall (2010) examines the effects of terrorist attacks on international stock markets. They focus on stock markets of the most economically significant countries in the world, and address a sample of 31 attacks. By taking an event-‐study approach they find that these terrorist attacks result in moderately negative price effects. They compare the price reactions to those induced by another type of exogenous shocks, namely earthquakes, and find that the impact of terrorism is stronger. The price reactions are found to be more pronounced in markets and industries that are directly affected by the attacks. After adjusting for idiosyncratic risk, the effects on stock prices are found to be prevalent only on the short-‐run, and generally last no longer than the event-‐day itself. They find an exception for the impact of the 9/11 attacks, that led to a statistically significant increase in systematic market risk on the long-‐run.
Another paper that finds a significant negative effect of terrorism on financial markets is the study of Zussman and Zussman (2006), which in addition analyzes the effect of the Israeli government’s anti-‐terrorism policy of assassinating high-‐ranked terrorist leaders, on financial stocks. They find evidence that a policy of assassinating highly ranked political leaders has a counterproductive effect on Israeli stocks’ returns, but that targeting high military leaders has an overall positive effect on these returns. The paper gives an example of how government decisions may have influence on investors’ investment allocations.
Eldor and Melnick (2004) analyze the reaction of the Israeli stock market and foreign exchange markets to terrorist activity. The data they use in their study distinguishes several event characteristics including location, type of attack and target and the number of casualties. The number of 639 events analyzed in their study is typically large compared to other studies. They find that suicide attacks and the number of victims have a permanent attack on both the stock and foreign exchange market. Location of a terror attack, being one of the main cities in Israel, had no significant effect. Furthermore, their findings indicate that markets function in an efficient way with regard to the incorporation of news on terror events. The Israeli stock market is considered as a developed market, which may provide reasons to believe that the findings in their paper can be generalized to other developed financial markets in western society.
Kollias et al. (2011) address the questions whether stock markets’ price reactions to terrorist incidents have changed over time, and whether market size and maturity as well as type of perpetrators and target are determinants of the size of these reactions. In their event-‐ study they make a comparison between the London stock exchange and the Athens stock exchange, which they consider as large and small capitalization markets respectively. They find no conclusive evidence in support of a changing price reaction through time in both markets, with regard to abnormal returns. However, their findings do indicate that the number of injuries and fatal casualties have a significant effect on stock market volatility, with the effect being greater on the small capitalization market than on the large one. The finding that points to a difference in vulnerability between small and large stock markets is line with Arin et al. (2008). Some papers analyze the role of the insurance sector in counteracting the risk that is imposed by terrorism. Kunreuther & Michel-‐Kerjan (2004) discuss in their paper the U.S. government’s role to co-‐operate with the private sector to construct a sustainable terrorism risk insurance industry. Ibragimov et al. (2009) discuss that even though the market capacity for terrorism risk insurance is large enough, the existence of heavy left tails in risk distributions – a non-‐negligible possibility of extreme losses – lead to so-‐called nondiversification traps, in which case the value of diversification of this type of risk is decreased. High welfare loss may be the result of this trap, and they discuss the role of a central agency to coordinate the (re)insurance market to make the transference of catastrophic risk (including terrorism risk) possible. As Bourioux and Scott (2004) discuss in their paper about terrorism risk coverage by capital markets, that after the 9/11 disaster insurance companies have largely limited or excluded terrorism risk coverage from their polices. The paper gives suggestions for combined public and private initiatives to provide coverage for terrorism risk. Capital markets may have the possibility to substitute for the insurance market in counteracting the financial risk caused by terrorism. The findings of Chen and Siems (2004) indeed demonstrate that U.S. capital markets have become more resilient to terrorist attacks in recent years (at the beginning of this
decennium) as compared to the past. They surveyed the effect of 14 terrorist and military attacks in a period ranging from the beginning of the twentieth century up until the 9/11 attacks. In particular, they find empirical evidence that U.S. capital markets have become more able to absorb the negative impact of attacks due to an increased ability of the banking and financial sector to provide liquidity, which helps to stabilize capital market and dampens panic reactions.
Various financial scholars have found indications of potential hedging opportunities in a situation of terrorist activity. Mueller and Stewart (2014) evaluate the counterterrorism policy conducted by the U.S. government and the FBI. They estimate that domestic counterterrorism expenditures have increased by some $75 billion per year over the decade following the 9/11 attacks. One could expect that such increased spending as a response on terrorist events gives a boost to stock returns in the homeland security and defense sector (Lenain et al., 2002). In the existing literature there is some empirical evidence in favor of this prognosis [see for examples Chesney et al. (2011) and Berrebi and Klor (2010)]. Mueller and Stewart conclude that is complex to determine the appropriate level of counterterrorism spending. They provide a framework that may support governments in evaluating the extent to which their marginal increases in counterterrorism spending is justified.
However, only very little research is done on how the effect of terrorism on stock markets may vary according to the distinctive characteristics of the attack. Moreover, only little is known about strategies that in a specific way describe how to cover investment portfolio returns from terrorism risk. This thesis aims to contribute to existing literature by exploring the differential effects of terrorist attacks as well as hedging strategies against terrorism risk. In contrast to conventional literature that is concentrated on the negative relation between terrorism and stock markets, this study mainly focuses on the ways to effectively counteract against the negative impact that terrorist activity may have on certain stocks.
Previous studies have mostly lacked the distinction regarding the nature of terror attacks. Generally, in the existing literature the impact of terrorism on stock markets is generalized, in the sense that different types of attacks are considered as being equal in terms of their characteristics. However, terror attacks may affect stock markets differently based on their distinctive characteristics. Moreover, related studies have focused on the impact of terrorist acts on defense and homeland security related stocks relative to overall or very specific stock markets such as the S&P 500, the Israeli/Tel Aviv stocks market, and airline industry stocks.1
Also, related studies only analyze a relatively small number of events –typically no more than 50-‐100– or focus specifically on the impact of major acts such as the 9/11 attacks. Another
1 See for examples Berrebi and Klor (2010), Zussman and Zussman (2006), Melnick and Eldor (2010), and
Eldor and Melnick (2004).
potential shortcoming that is generally observed in these studies may be the use of one-‐sided approaches such as VAR-‐ and GARCH-‐methods and an event study approach. Existing literature on the impact of terrorism on financial markets shows criticism to the use of the event study methodology, and it is reasonable to explore different methods to investigate this relation (Bashir et al. 2013). This paper aims to contribute to existing research by distinguishing between terrorist events based on their nature. The focus of this paper is on terrorism’s impact on a variety of stock markets, by analyzing a substantially larger number of events than typical related studies as well as by using a relatively new approach.
3. Methodology
To investigate whether the effect of terrorist activities on stock markets differs based on characteristics, different types of attacks are categorized into classes as defined by the Global Terrorism Database (GTD). I distinguish the events’ attack type by the following categories: armed assault, bombing explosion, facility/infrastructure attack and ‘others’. I categorize target types by the following types: related to business, governmental, police, military, private citizen and property, transport and ‘others’. For attack type as well as target type, the ‘others’-‐ categories include observations with characteristics that don’t fall into the any of the specified categories. I also control for the cases in which more than 5 people were killed and property damage was larger than $1 million. This study investigates how different types of attacks may differently affect stocks. The main hypothesis of this thesis is: the effect of terrorism on stock markets differs based on the characteristics of the attacks.
Next, in order to investigate whether investors can protect their returns against terrorism risk, this thesis aims to search for a particular portfolio that show a common positive price reaction to the different kinds of terrorist activity. Based on the findings of Chesney et al. (2011), Berrebi & Klor (2010) and Mueller & Stewart (2014), who observe a positive relation between terrorism and the defense and homeland security industry, the aim is to identify stocks related to the defense industry, that show a less severe negative or positive price reaction to terrorist acts. Another hypothesis of this study is as follows: investors and industries can hedge against terrorism risk.
Following the methodology of Chesney et al. (2011), this thesis makes use of non-‐ parametric estimation, which is a statistical method that allows a functional form of a fit to data to be obtained without imposing any parametric assumptions. Non-‐parametric estimation lets the data speak for itself and overcomes a disadvantage of parametric econometrics when inconsistency between data and a particular parametric specification would result in non-‐ robustness. This method does not heavily depend on assumptions about distributions.
estimation, a local polynomial regression (LPR) is applied to time series data to get a non-‐ parametric conditional distribution of stock returns.2 This distribution is conditioned on the
average of a sample of returns before the event, computed as 𝑅!!!= ! !𝑅!!! ! !!! , (1)
where Rj is the return on the stock index being analyzed at time j and n is the number of
observations in the sample of the returns equal to 200. The value of the probability of the event-‐ day return, conditional on the average of the returns, is analyzed. If this probability is smaller than 10% and 5% the event-‐day return is interpreted as abnormal and extreme respectively, following Chesney et al (2011).
To give a more comprehensive description of the methodology, consider the following expression: !!!! 𝑌!− 𝛽!− 𝛽!(𝑋!− 𝑥!) !𝐾! 𝑋!− 𝑥! , (2) where: 𝑌 ! = 𝟏!!!!! = 1 if 𝑅! ≤ 𝑟! 0 if 𝑅! > 𝑟! , (3)
where Ri is the return on day i, and rt is the return on the day of the terrorist attack.
Furthermore, in equation (2) I take 𝑋! = 𝑅!!!, the return one day before day i, 𝑥!= 𝑅!!!, the
average index return of a sample of 200 observations preceding the day of the terrorist attack, and 𝐾! is a Kernel-‐function, which in fact determines the weight that is given to the 𝑋!− 𝑥! -‐ term. An appropriate Kernel-‐function and a bandwidth have to be chosen. As in the study of which this methodology is derived, an Epanechnikov kernel function is used, with a bandwidth of 2.34σs-‐1/5, where σs is the standard deviation of the sample of returns.
By applying an LPR I minimize expression (2) to get point estimates 𝛽! and 𝛽!. A point
estimate 𝛽! then corresponds to a conditional probability of an index return being less or equal to the (terrorist) event-‐day return. Where this conditional probability is less than 10% and 5%, we consider the event-‐day return as abnormal or extreme respectively.
After determining whether event-‐day returns were abnormal/extreme conditional on the average of index returns preceding a particular terrorist attack, the following probit regression is done to examine whether the nature (or characteristics) of terrorist events determine(s) how the stock indices are affected:
ℙ (𝐵!" = 1 𝐷!!, … , 𝐷!"; 𝛽!!, … , 𝛽!" = Φ(𝛽!!+ 𝛽!"𝐷!" ! !!! ), (4) where:
1. Φ ∙ is the cumulative distribution function of the standard normal distribution. 2. Bit is a dummy variable indicating whether index i’s event-‐day return was abnormal
on event-‐day t.
3. Dkt is a dummy variable indicating whether the terrorist attack on event-‐day t was of
type k, being one of the K attack characteristics described in the beginning of this section.
4. 𝛽!! is a constant for index i.
5. 𝛽!" is the coefficient for regressor k and index i.
In addition to estimating the standard probit regression coefficients, I estimate the Average Marginal Effects (AMEs) to examine the effect of a discrete change in the regressors. The discrete change in a regressor 𝐷!", which is one of the K attack characteristic dummy variables
that takes the values {0,1}, is given by the following equation: ∆!!"ℙ(𝐵!" = 1 𝐷!!, … , 𝐷!"; 𝛽!!, … , 𝛽!" = Φ(𝛽!!+ 𝛽!"𝐷!" ! !!! ) −Φ(𝛽!!+ 𝛽!"𝐷!"+ 𝛽!"𝐷!" ! !!!!! !!! !!! ). (5)
Equation (5) denotes the difference between the cumulative distribution function (cdf) with all regressors included and 𝐷!" takes the value of 1, and the same cdf for which regressor 𝐷!", of which the AME is estimated, takes the value of 0. The estimation of the AME for each regressor (terrorist attack characteristic) is useful for interpreting the effect of a particular terrorist attack characteristic under a ceteris paribus assumption.
The non-‐parametric methodology used in this study is considered as more appropriate than conventional methods (such as GARCH and event-‐study approaches) by Chesney et al. (2011). The reason is that it doesn’t impose strong parametric restrictions. Also, it is relatively less computationally intensive compared to the GARCH method. Moreover, the approach is new in the investigation of terrorism’s impact on stock markets.
This study aims to investigate the effect of terrorist attacks on sectors that are particularly exposed to these perilous events. As earlier mentioned, these include the insurance
industry, airline industry and travel/tourism industry. In this paper indices that are taken into consideration for these industries are, Aero/Defense indices, Life and Non-‐Life Insurance indices, Travel/Leisure/Tourism indices, Airline indices, and Construction & Materials indices. All indices have an All World, a Europe, and a U.S. variant. Moreover, this study aims to assess the question which stocks structurally outperform terrorism sensitive sectors. The results of testing the hypotheses of this thesis are used to gain insights into (1) the way in which different types of terrorist attacks affect different stocks and (2) a potential hedging strategy for industries or investors that are heavily invested in these industries, which are vulnerable to particular types of terrorist attacks. Specifically, this study examines whether adding defense industry related stocks to a portfolio, that (mainly) consists of stocks of terrorism risk exposed sectors, is a wise thing to do for an investor. Ideally, this study may help both companies and investors that are exposed to terrorism risk to protect themselves from severe losses. The thesis aims to contribute to existing literature by providing some answers to questions that still remain largely unsolved in the research field of terrorism’s impact on stock markets.
4. Description of data
This study analyzes a sample of terrorist events aimed against the U.S. and European countries in the period 1970 – 2014. Data on terrorist attacks are obtained from the Global Terrorism Database (GTD), which provides detailed information on terrorist events around the world from 1970 through 2013. I select events by the extent of their impact, so that only events that caused 5 or more fatal casualties, or those that resulted in property damage larger than $1 million are selected. In this way, I select a sample of 699 terrorist events for this study, out of a total number of 24,067 attacks aimed against the U.S. and Europe in the aforementioned period. This number of 699 events is substantially larger than examined in previous studies, which typically analyzed between 20-‐100 events.
Data on daily index returns of U.S., European and world indices are obtained from Datastream. When a terrorist attack occurred on a non-‐trading day, I use the return of the following trading day for that particular event-‐day. Datastream provides data on self-‐constituted indices, which are formed by a list of stocks of companies that operate in a particular industry.3 I
used U.S., European, and All World versions of these self-‐constituted indices for the industries under consideration. Table 1 presents the number of events that had abnormal and extreme event-‐day returns per index.
Table 1. Descriptive Statistics
By making use of a non-‐parametric approach I determine whether a negative event-‐day return was abnormal or extreme. As can be seen in Table 1, the different indices show a similar number of events with abnormal and extreme event-‐day returns. The Aero/Defense indices don’t show a clearly higher resistance to a negative impact of terrorism as compared to the other indices.
Abnormal Extreme Mean Std. Dev Min. Max.
Europe 83/699 50/699 -‐0.00003 0.014 -‐0.058 0.082
U.S. 97/699 53/699 -‐0.00054 0.014 -‐0.093 0.054
All World 95/699 56/699 -‐0.00023 0.011 -‐0.070 0.040
Abnormal Extreme Mean Std. Dev Min. Max.
Europe 85/699 59/699 0.00009 0.017 -‐0.083 0.117
U.S. 107/699 72/699 -‐0.00068 0.015 -‐0.084 0.072
All World 115/699 65/699 -‐0.00040 0.012 -‐0.051 0.062
Abnormal Extreme Mean Std. Dev Min. Max.
Europe 86/699 59/699 0.00003 0.015 -‐0.112 0.076
U.S. 80/699 49/699 -‐0.00025 0.011 -‐0.073 0.051
All World 101/699 63/699 -‐0.00012 0.010 -‐0.045 0.055
Abnormal Extreme Mean Std. Dev Min. Max.
Europe 96/699 59/699 -‐0.00040 0.013 -‐0.088 0.063
U.S. 105/699 63/699 -‐0.00130 0.017 -‐0.142 0.072
All World 103/699 55/699 -‐0.00061 0.011 -‐0.114 0.053
Abnormal Extreme Mean Std. Dev Min. Max.
Europe 96/699 64/699 -‐0.00053 0.015 -‐0.123 0.066
U.S. 95/699 53/699 -‐0.00179 0.027 -‐0.327 0.082
All World 90/699 53/699 -‐0.00044 0.013 -‐0.150 0.046
Abnormal Extreme Mean Std. Dev Min. Max.
Europe 91/699 59/699 -‐0.00009 0.011 -‐0.062 0.047
U.S. 115/699 66/699 -‐0.00145 0.015 -‐0.081 0.051
All World 95/699 58/699 -‐0.00004 0.011 -‐0.053 0.048
Construction & Materials (N=699) Number of events with abnormal and extreme negative event-‐day returns: A non-‐ parametric approach is used to determine whether negative movements in event-‐day returns were abnormal or extreme. Abnormality of the event-‐day returns corresponds to the conditional probability in the interval [0.05; 0.10]. Extreme index movements
correspond to the conditional probability in the interval [0.00; 0.05). Conditioning is done on a sample of 200 returns on the trading days preceding the event-‐day. Only events that resulted in 5 or more fatalities, or larger than $1 million property damage are taken into consideration. The four columns on the right side of the table display descriptive statitics of the returns. Aero/Defense (N=699) Life Insurance (N=699) Non-‐Life Insurance (N=699) Travel/Leisure/ Tourism (N=699) Airline (N=699)
For this study, GTD provides data on terrorist attack characteristics including descriptions on attack type as well as the target type of the concerning attacks. Table 2 presents the number of events sorted by category. Attacks whose primary objective is to cause physical harm or death directly to human beings are classified by armed assaults. The bombing/explosion category includes any attacks that involved some sort of explosives that cause physical damage to the surrounding environment. Acts (that exclude the characteristics of the aforementioned attack types) with the objective of causing damage to a non-‐human target are categorized as a facility/infrastructure attacks. For the target types, the business category is defined as entities (individuals or organizations) engaged in commercial activity, such as restaurants, stores and gas stations. Governmental, police-‐related and military targets can be human as well as non-‐human. The private citizens & property target type includes attacks on individuals (that don’t classify as the aforementioned target types), the public in general or attacks in public areas. Transport includes attacks on public transportation systems, excluding aviation. The total number of events under consideration increased over the three different decades in the period January 1974 – January 2014.
The majority of the attacks involved a bombing, explosion or an armed assault. As one expects, the proportion of events causing property damage that exceeds $1 million, is relatively large for attacks that involved an explosion or a facility or infrastructure attack (36% and 90% of the total number of attacks respectively). This ratio is also remarkably large for the business and governmental target type categories (67% and 48% of the total number of attacks respectively). I consider attacks with these aforementioned attack types –bombing/explosion and infrastructure– and target types –governmental and business– as attacks that have a primary aim to inflict property damage or economic damage to non-‐human objects. Attacks with a relatively high proportion of cases that led to more than five fatal casualties are assumed to have the goal of causing damage to human-‐life. The attack characteristics of such attacks include armed assaults and bombings or explosions for the attack type, as respectively 96% and 67% of the sample of events in these attack type categories resulted in more than five fatalities. Regarding target types, I assume that attacks aimed against targets other than businesses and governments have the primary aim to cause loss of life.
As stated before in this paper, existing literature shows a considerable amount of evidence that terrorist events have an overall negative impact on stock index returns. Therefore I expect to find a general negative effect, and more pronounced negative effects on the returns of insurance, travel/leisure/tourism, and airline indices for the cases in which human lives were the primary target. The reason for these expectations is that consumers’ risk perceptions may be adversely affected by terrorist activities that lead to personal victims, which may particularly result in lower demand for air travel and travel in general, and higher premiums for insurances
due to reconsiderations of risks by insurance companies (Drakos, 2004 and Ibragimov, 2009). Moreover, due to this adverse change in risk perception, military intervention by governments may be more likely to occur, which may reduce the likelihood of abnormal negative returns on aero/defense indices. Attacks that are characterized by the goal of causing damage to non-‐ human objects, such as business-‐related buildings and other properties, are expected to have a significant negative impact on the construction and materials and non-‐life insurance industries. The rationale behind this expectation is that terrorism may lead to increased insurance premiums and a decrease in terrorist coverage due to a reassessment of risk by the property insurance business, as well as increased uncertainty for lenders and investors. This in turn may result in a slow-‐down in the completion or initialization of construction projects (Kunreuther et al. 2003). Table 2 5. Empirical analysis 5.1. Main Results
This section provides the results by which I will answer the research question of this study. The regressions address the relation between the occurrence of abnormal negative event-‐day index
Total Victims killed ≥ 5 Property damage > $1mil. Attack type Armed assault 246 237 10
Bombing/Explosion 321 216 115
Facility/Infrastructure 80 8 72
Other 52 46 8
Attack target Business 123 41 82
Governmental 64 34 31
Police 90 86 5
Military 172 165 10
Private citizens & property115 91 25
Transport 53 45 13 Other 82 45 39 Jan. 1974 -‐ Jan. 2014 Jan. 1974 -‐ Jan. 1984 Jan. 1984 -‐ Jan. 1994 Jan. 1994 -‐ Jan. 2014 Victims killed ≥ 5 507 50 184 273 Property damage > $1mil 205 44 38 123 Number of attacks 699 93 221 385
Period
Number of events by categories: only events that caused 5 or more fatalities, or larger than $1 million property damage are taken into consideration. The events occured in the period of January 1974 – January 2014.
returns and the nature of the terrorist events. In this section I also examine whether the results are conform to the expectations based on existing empirical research.4
5.1.1. Aero/Defense
Columns 1, 2 and 3 in Table 3 present the results for the All World, U.S. and Europe Aero/Defense indices respectively. Column 1 shows that events with more than five fatal casualties have a statistically significant positive effect for the All World index. However, the average marginal effect (AME) of this attack characteristic on the probability of an abnormal negative event-‐day return is statistically insignificant.
Column 2 indicates that attacks that involved a bombing or explosion have a statistically significant positive relation to the occurrence of an abnormal negative event-‐day return for the U.S. index. However, the AME is statistically insignificant. Regarding the target type, attacks aimed against military units, private citizens and property, and transport have a statistically significant negative effect, and are less likely to result in abnormal negative event-‐day returns. Events being one of these three categories decrease the probability of a negative event-‐day return by 11.4, 10.0, and 10.3 percent points respectively. Attacks with these characteristics are assumed to have the main objective of causing human victims, and according to my expectation such attacks may have a higher likelihood of leading to military intervention by governments. This may positively affect Aero/Defense index returns, which translates to a decreased probability of abnormal negative returns.
Column 3 shows that events leading to more than five fatal casualties have a statistically significant positive effect for the Europe index. This relation is statistically insignificant with respect to the AME. The probability of an abnormal negative event-‐day return decreases by 8.4 percent points when the target type is in the business category. Such attacks are assumed to have the primary goal of inflicting property damage. Given the importance of business activities for economies in general, defense and homeland security spending may be increased in the case of business-‐related targets which translates in a reduces the likelihood of abnormal negative returns for the defense industry.
5.1.2. Life Insurance
Columns 4, 5 and 6 in Table 3 present the results for the All World, U.S. and Europe Life Insurance indices respectively. Column 4 indicates that attacks aimed against governmental targets increase the probability of the occurrence of an abnormal negative event-‐day return by 12.6 percent points for the All World index.
4
