School of Economics, Finance and Accounting, Coventry University, CV1 5FB, United Kingdom

Transcription

1 Effects of Terrorist Attacks on Stock and Foreign Exchange Markets: International Evidence Jin Suk Park and Mohammad Khaleq Newaz * School of Economics, Finance and Accounting, Coventry University, CV1 5FB, United Kingdom Abstract This study examines the effects of terrorist attacks on stock and foreign exchange markets by using the event studies, the mean-volatility and the quantile regression models on more than 10,000 terrorist attacks in 36 countries from 1996 to About 50% of the terrorist attacks significantly affect the market returns and 20% of them change the short-term volatility. However, the sign of the impacts is mixed unlike typical belief. Also, it varies across countries e.g. powerful countries such as the US, the UK, Russia and China are not negatively affected by the attacks. On the other hand, the different quantiles of returns are affected asymmetrically, e.g. stronger impacts on both tails of returns are found and negative impacts are clearer when the stock markets perform relatively well. Keywords: terrorism, terrorist attack, stock market, foreign exchange market, quantile regression JEL classification: C22, E44 *Corresponding author. addresses: jin.park@coventry.ac.uk (J.S. Park); mohammad.newaz@coventry.ac.uk (M.K. Newaz) 1

2 Highlights A comprehensive study of more than 10,000 terrorist attacks in 36 countries reveal that about 50% of the individual attacks affect the stock and the foreign exchange market returns. Contrast to typical belief, the attacks do not always have negative impact on market returns nor increase volatility. Negative effect of terrorist attacks is not shown in powerful countries e.g. USA, UK, Russia and China. Strong negative impacts on stock markets are evidenced only when the stock markets perform relatively well. 2

3 1. Introduction Research on terrorism in the context of financial markets has not received much attention from academics and policy makers until recently despite the common belief that terrorist attacks can adversely affect the markets and create future uncertainty which increases investment risk and reduces investors confidence. So far, the empirical research on terrorist attacks on financial markets is limited, with mixed results. For instance, most of the studies concluded that terror attacks have a negative impact on financial markets e.g. (Arin, Ciferri, & Spagnolo, 2008; Chesney, Reshetar, & Karaman, 2011; Eldor & Melnick, 2004; Karolyi, 2006; Mansoor, Mustofa, & Ugur, 2017). In contrast, Apergis & Apergis (2016) reported positive effect of the 2015 Paris terrorist attacks on the defensive companies stock return and Aslam & Kang (2013) concluded significant positive return one day after the attack in Pakistan. By using the event study of 14 attacks, Chen & Siems (2004) did not find any evidence that a terrorist attack has a negative impact on stock markets. They conclude that US financial markets are more flexible, liquid and recover quickly than other global financial markets; hence the reaction of the terror attack is less severe. The terrorist attacks of 9/11 changed the scope of the terrorism as a geo-political risk that negatively affects the global economy and financial markets. Subsequently, the existing literature mainly focus on very limited terrorist events and particularly focus on 9/11, 7/7 London attack, 11/13 Paris attack etc. to examine its impacts on global stock markets by using popular event study methodology. However, comprehensive research on terror attacks and their impacts on stock and foreign exchange markets are still infancy. This study makes two contributions: first, we explore the 10,834 individual terrorist attacks across 36 markets between 1996 and 2015 and show how they affect the market and their effects vary across the markets. Second, this is the first study to our knowledge that employs the quantile regression method to investigate the effect of terror attack on the different quantiles of stock and foreign exchange market returns. We apply the forecasting-based event study approach, the time-series mean-volatility models and the quantile regression models. The event study approach shows that about 50% of individual terror attacks affects the returns in the stock and foreign exchange markets while 20% of them affect the volatilities. The mean-volatility models used on each country reveal that around 30% of the countries are significantly exposed to the attacks. The signs of their impacts are mixed in both returns and volatilities. The mean-volatility models also show that powerful nations, e.g. USA, UK, Russia and China, are not influenced by terrorist attacks. Contrast to typical belief, the terrorist attacks do not always have negative impact on market returns nor increase volatility. However, the quantile regression reveals that the impact of terrorism is more dominant in the tails of conditional distribution of returns. For examples, negative impacts on stock markets is evidenced when market performs well. The remainder of the study is organised as follows. Section 2 presents the data, Section 3 describes methodology and Section 4 reports the results. Section 5 concludes. 2. Data The stock market and the foreign exchange market are represented by the main stock market index and the real effective exchange rate (REER) of a country. The daily observations of the stock indices and the REER from 36 countries plus MSCI World Index are collected over the period from 1/1/1996 to 31/12/2015 from the Datastream (Table 1). A total of 10,834 daily terrorist attacks during the same period are collected from the Global Terrorism Database (GTD). The GTD is the most extensive dataset of terrorism attacks maintained by the National Consortium for the Study of Terrorism and Responses to Terrorism at the University of Maryland, USA. The daily terrorism index is calculated following Cukierman (2004), Eckstein and Tsiddon (2004) and Arin, Ciferri, & Spagnolo (2008). 3

4 Table 1. Descriptive Statistics ID Country Price index Start Obs Stock Forex Days Num Wnd Kld T index AT Austria ATX 11/04/ AU Australia ASX200 11/04/ BL Belgium BEL20 11/04/ BR Brazil BOVESPA 02/01/ BU Bulgaria SOFIX 14/04/ CA Canada TSX Comp 11/04/ CL Chile IGPA General 29/03/ CN China Shanghai A 03/06/ CR Croatia CROBEX 30/01/ CZ Czech PX 10/04/ DN Denmark OMXC20 11/04/ FI Finland OMXH 11/04/ FR France FAC40 11/04/ GE Germany DAX30 11/04/ GR Greece Athex Comp. 11/04/ HK Hong Kong Hang Seng 29/10/ HU Hungary BUX 19/01/ ID Indonesia IDX Comp. 14/05/ IN India NIFTY500 11/04/ IR Ireland ISEQ Overall 11/04/ IS Israel TA125 09/04/ JP Japan NIKKEI225 11/04/ ML Malaysia KLCI 31/10/ MX Mexico IPC 31/07/ NL Netherland AEX 11/04/ NO Norway Oslo All 11/04/ PH Philippines PSEi 16/01/ RS Russia RTS 31/03/ SA South Africa JSE All 11/04/ SP Spain IBEX35 11/04/ SZ Switzerland SMI 11/04/ TK Turkey BIST100 27/01/ TL Thailand Bangkok SET 15/09/ TW Taiwan TAIEX 16/04/ UK UK FTSE100 11/04/ US US S&P500 11/04/ SUM/MEAN Note: All data end on 31 Dec 'Stock' and Forex represent the average of log returns of the stock index and the real effective exchange rate, respectively. 'Num' is the total number of terrorism attacks. 'Wnd' and 'Kld' are the total number of victims, wounded and killed, respectively. 'T Index' is the average value of terrorism index that is calculated as ln(e+num+wnd+kld). 3. Methodology This study employs three methods. The event study approach is used as the analysis of individual attacks. The mean-volatility model is for the country-specific main analysis based on the terrorism index. The quantile regression model is for the analysis of different quantiles of returns. First, the forecasting-based event study approach estimates a mean-return model on the estimation window prior to individual terrorist attacks and obtain forecast errors within the event window including the attack day. Forecast errors are sorted as abnormal returns and then cumulative abnormal returns (CAR) are tested for the significance of attacks. The mean-return model utilizes the previousday stock and foreign exchange market returns and the returns to the MSCI World Index that controls for external factors. The estimation and the event windows are [-100,-1] and [0,5], respectively 1. For market volatility, the test for the equality of variance between two windows is used. Second, the mean-volatility model analyses each country separately over the entire sample period to examine the impact of the terrorist attacks, represented by the terrorism index, on both market returns 1 Two more event windows of [0,2] and [0,3] and two different mean-return models (the constant mean-return and AR(5)) are also tested and the overall results were similar. Note that attacks within the estimation window are included as their impacts are utilised in subsequent forecasting. The technical details of the event study are discounted as it is one of the most commonly-used methods (Chesney et al., 2011). 4

5 and volatility. The study adopts the identical explanatory variables used in the event study but adds both of the current and the previous-day terrorism indices which account for the size and the timing of attacks within a day e.g. within and out of trading hours. The mean-return (r t) model is accompanied by one of the popular asymmetric volatility models, EGARCH-in-mean, as follows. r it = α i + β 1i r it 1 + β 2i r jt 1 + β 3i r Wt 1 + β 4i σ it 1 + β 5i T it + β 6i T it 1 + ε it (1) ln(σ 2 it ) = ω i + γ 1i ε t 1 ε + γ t 1 σ 2i + γ t 1 σ 3i ln(σ it 1 ) + γ 4i T it + γ 5i T it 1 (2) t 1 2 where r it is the return to the stock (or foreign exchange) market at time t, r j is the return to the foreign exchange (or stock) market in the same country, r W is the return to the global market index, T is the terrorism index and ε it ~N(0, σ 2 it ). The impact of terrorist attack is tested by H 0 : β 5 = 0 or β 6 = 0 in the mean-return equation in (1) and H 0 : γ 4 = 0 or γ 5 = 0 in the volatility equation in (2). Last, the quantile regression model (Koenker & Bassett, 1978) finds the relationship between a specific quantile of market returns, instead of their conditional mean, and the set of explanatory variables above. The regression estimators for quantile θ is the solution to the following minimization problem: β θ = arg β min ( t:r θ r t x t >x t β t β + t:r (1 θ) r t x t <x t β t β ) (3) i.e., β s that minimise a weighted sum of the absolute errors where y is spilt at proportions θ below and (1- θ) above. x t is a vector of the explanatory variables. Three quantiles namely, 0.1, 0.5 and 0.9 are used and two slope equality tests (0.1 vs 0.5 and 0.5 vs. 0.9) examine the robustness of the findings. 4. Results Table 2 reports the percentage of the individual terrorist attacks that significantly affect market returns or volatility. 48.5% and 50.1% of the attacks affect the returns in the stock markets and the foreign exchange markets, respectively. The lower percentage, 22.8% and 21.2%, influence the volatility in those markets. Despite the fact that the impact of individual terrorist attacks is clear, the mixed sign contrast to the common belief of decreasing returns and increasing volatility, deserve more investigation. Additionally, the year-by-year analysis does not show a clear pattern in the change in their impact over time (Table A1 in Appendix). 5

6 Table 2. Results - Event Studies Stock Markets Foreign Exchange Markets ID Days R+ R- ΣR V+ V- ΣV R+ R- ΣR V+ V- ΣV AT AU BL BR BU CA CL CN CR CZ DN FI FR GE GR HK HU ID IN IR IS JP ML MX NL NO PH RS SA SP SZ TK TL TW UK US Total Note: This table shows the percentage of terrorist attacks that positively (+) or negatively (-) affect the returns (R) or the variances (V) in the stock markets or the foreign exchange markets in each country. Σ indicates the sum of positive and negative impacts. Table 3 presents the summary results from the mean-volatility models for each country, which adopt the terrorism index to control for the scale of the attacks. 40% and 28.57% of the countries are significantly affected by the attacks in terms of the returns in the stock and the foreign exchange markets, respectively. However, clearly the impacts of terror attack differ across countries. For examples, more powerful countries like US, UK, Russia and China does not experience negative effects. However, Brazil and India have negative impact, possibly because of military, security investment or financial markets themselves are not as strong as those powerful countries. Similar to the results with individual attacks in Table 2, the impact on the market volatility is only observed in the smaller proportion of the markets. The 3-year moving-window analysis is presented in Table A2 in Appendix, where no obvious pattern is seen. 6

7 Table 3. Results - Mean-Volatility Models Stock Markets Foreign Exchange Markets ID Return Volatility Return Volatility AT ** *** AU * BL * BR *** * BU CA CL *** CN * *** *** CR ** *** *** CZ *** DN * *** *** FI FR ** GE * * GR ** * HK *** HU ** *** ID *** *** IN *** *** IR * IS * JP *** * ML *** MX * * NL NO *** *** PH * *** *** RS *** SA SP *** SZ TK TL * *** ** TW *** *** *** UK US ** Sig 40.00% 28.57% 28.57% 34.29% % 11.43% 17.14% 25.71% % 17.14% 11.43% 8.57% Note: This table shows the estimates and the significance of the terrorism index in the mean-volatility models. The numbers are the sum of terrorism index at t=0 and -1. ***, ** and * indicates the significance of the terrorism index in affecting the markets in the attack or the following day, at 1%, 5% and 10% level, respectively. 'Sig' indicates the percentage of the countries that are significantly affected by terrorist attacks. '+' and '-' are the percentage of either significantly positively or negatively affected countries, respectively. Table 4 presents how the impact of terrorism on market returns differ in three quantiles for each country. Compared with Table 3, the impacts become more dominant on in both tails of return distribution (0.1 and 0.9) in the larger number of countries. In other words, the markets with extreme positive and negative returns are more strongly affected by terrorist attacks in many countries. Particularly in the stock markets, it is more likely that terrorism attacks negatively affect the market when it performs well than not (47.2% vs. 25%), but the same difference is not shown in the foreign exchange markets. The 3-year moving window analysis (Table A3 in Appendix) shows similar associations. 7

8 Table 4. Results - Quantile Regression Models Stock Markets Foreign Exchange Markets θ AT *** *** *** *** * *** *** * AU *** *** BL *** *** *** *** *** *** *** *** BR *** *** *** *** * *** BU *** *** *** *** *** *** *** * CA ** *** CL *** *** *** ** ** *** CN *** *** ** CR *** *** *** *** *** *** *** *** CZ *** *** *** *** ** DN *** *** *** *** *** *** *** *** FI *** *** *** *** *** ** *** *** * FR *** ** *** ** * GE ** ** GR * *** * ** *** * HK *** *** * *** *** *** *** *** *** *** HU *** *** *** *** *** *** *** *** * *** ID * *** IN ** ** ** ** IR ** IS *** * *** * JP *** ** * ML *** *** * *** MX *** ** *** *** * NL *** *** *** *** *** *** NO *** *** *** *** *** *** *** *** PH * *** ** RS *** * ** *** *** SA *** ** SP * SZ *** *** *** *** * *** *** TK * *** TL *** ** *** *** *** *** *** TW *** *** ** *** ** *** * *** *** UK *** ** US ** Sig Note: This table shows the significance of current- or previous-day terrorism attacks in the quantiles of 0.1, 0.5 and 0.9 in the quantile regression models. The number are the sum of estimated coefficients of current- or previous-day terrorism attacks. '15' indicates the results of the slope equality test of the current-day coefficients between 0.1 and 0.5 quantiles. '59' indicates that between 0.5 and 0.9 quantiles. ***, ** and * indicates the statistical significance at 1%, 5% and 10% levels, respectively. In summary, the event study reveals that about 50% of the individual terrorist attacks significantly affect the returns in the stock and the foreign exchange markets while around 20% of them affect the market volatility. This impact varies widely across different markets. The mean-volatility model with the terrorist index shows that powerful countries such as the US, the US, Russia and China are not negatively affected by the attacks while 30-40% of the countries shows the significant impacts. Overall, the terrorist attacks do not always decrease returns nor increase volatility. Specifically, commonly-expected negative impact of the attacks are only dominant when stock market performs well according to the quantile regression analysis. It also shows that the impact of terrorist attacks is stronger in both tails of return distribution in both markets. 8

9 5. Conclusion This paper examines the impact of terrorist attacks on the stock and the foreign exchange markets based on the wider range of data covering 36 countries and 10,834 terrorist attacks between 1996 and The findings of this comprehensive study support the significant impact of terrorist attacks on returns and volatilities in the financial markets, but the mixed sign of impact of the attacks, both individual and country-wide, may require further analysis using more factors or models. This study, however, provides evidences that the impacts are likely to be heterogeneous depending on the overall power of countries and markets and the quantiles of conditional return distribution, particularly two tails where the financial market performs relatively well or badly. References Apergis, E., & Apergis, N. (2016). The 11 / 13 Paris terrorist attacks and stock prices : The case of the international defense industry. Finance Research Letters, 17, Arin, K. P., Ciferri, D., & Spagnolo, N. (2008). The price of terror: The effects of terrorism on stock market returns and volatility. Economics Letters, 101(3), Aslam, F., & Kang, H.-G. (2013). How Different Terrorist Attacks Affect Stock Markets. Defence and Peace Economics, 26(6), Chen, A. H., & Siems, T. F. (2004). The effects of terrorism on global capital markets. European Journal of Political Economy, 20(2), Chesney, M., Reshetar, G., & Karaman, M. (2011). The impact of terrorism on financial markets: An empirical study. Journal of Banking and Finance, 35(2), Cukierman, A. (2004). Comment on: Macroeconomic consequences of terror: Theory and the case of Israel. Journal of Monetary Economics, 51(5), Eckstein, Z., Tsiddon, D. (2004). Macroeconomic consequences of terror: theory and the case of Israel. The Pinhas Sapir Center for Development, Tel-Aviv University, Discussion Paper No Eldor, R., & Melnick, R. (2004). Financial markets and terrorism. European Journal of Political Economy, 20(2), Karolyi, G. A. (2006). The Consequences of Terrorism for Financial Markets: What Do We Know?, (April). Koenker, R., & Bassett, G. (1978). Regression Quantiles. Econometrica, 46(1), Mansoor, M., Mustofa, J., & Ugur, G. (2017). The Impact of Terrorist Attacks on Foreign Exchange Rate: Case Study of Turkish Lira versus Pound Sterling. Economies, 5(1),

10 Appendix Table A1. Results - Event Studies: Year-by-Year Analysis Stock Markets Foreign Exchange Markets Year Days R+ R- ΣR V+ V- ΣV R+ R- ΣR V+ V- ΣV SUM Note: This table supplements Table 2 and shows the percentage of terrorist attacks that positively (+) or negatively (-) affect the returns (R) or the variances (V) in the stock markets or the foreign exchange markets in each year. Σ indicates the sum of positive and negative impacts. Table A2. Results Mean-Volatility Model: 3-Year Moving Window Analysis Stock Markets Foreign Exchange Markets Start End Countries R+ R- ΣR V+ V- ΣV R+ R- ΣR V+ V- ΣV 11/04/ /12/ /12/ /12/ /12/ /12/ /12/ /12/ /12/ /12/ /12/ /12/ /12/ /12/ /12/ /12/ /12/ /12/ /12/ /12/ /01/ /12/ /01/ /12/ /12/ /12/ /12/ /12/ /12/ /12/ /01/ /12/ /01/ /12/ Total/Avg Note: This tables supplements Table 3 and shows the percentage of the countries that are significantly positively (+) or negatively (-) affected by terrorist attacks in 3-year moving windows, in terms of the returns (R) and variances (V) in the stock and the foreign exchange markets. Σ indicates the sum of positive and negative impacts. 10

11 Table A3. Results - Quantile Regression Model: 3-Year Moving Window Analysis Stock Markets Foreign Exchange Markets Slope Slope Start End Countries R+ R- R+ R- R+ R R+ R- R+ R- R+ R /04/ /12/ /12/ /12/ /12/ /12/ /12/ /12/ /12/ /12/ /12/ /12/ /12/ /12/ /12/ /12/ /12/ /12/ /12/ /12/ /01/ /12/ /01/ /12/ /12/ /12/ /12/ /12/ /12/ /12/ /01/ /12/ /01/ /12/ Total/Avg R+ and R Note: This table supplements Table 4 and shows the percentage of countries that are significantly affected by current- or previous-day terrorist attacks in the quantiles of 0.1, 0.5 and 0.9. R+ and R- indicates whether the markets are positively or negatively affected by the attacks. '15' indicates the significance in the slope equality test of current-day terrorist attack between 0.1 and 0.5 quantiles. '59' indicates that between 0.5 and 0.9 quantiles. 11