Economic Modelling 35 (2013) 21–29

Contents lists available at SciVerse ScienceDirect

Economic Modelling

j ourna l homepage: www.e lsev ie r .com/ locate /ecmod

An analysis of a causal relationship between economic growth andterrorism in Pakistan

Muhammad Shahbaz a,⁎, Muhammad Shahbaz Shabbir b, Muhammad Nasir Malik c, Mark Edward Wolters d

a COMSATS Institute of Information Technology, M. A. Jinnah Campus, Defence Road, Off Raiwind Road, Lahore, Pakistanb University of Illinois at Urbana-Champaign, 504 E Armory, Champaign, IL 61820, USAc Huazhong University of Science and Technology, Wuhan, Chinad University of Illinois at Urbana-Champaign, 1206 South Sixth Street, Champaign, IL 61820, USA

⁎ Corresponding author.E-mail address: shahbazmohd@live.com (M. Shahba

0264-9993/$ – see front matter © 2013 Elsevier B.V. Allhttp://dx.doi.org/10.1016/j.econmod.2013.06.031

a b s t r a c t

a r t i c l e i n f o

Article history:Accepted 18 June 2013Available online xxxx

Keywords:TerrorismEconomic growthCointegration and causality

The present study investigates the causal relationship between terrorism and economic growth in Pakistanby incorporating capital and trade openness in production function. The study covers the time period of1973–2010. The ARDL bound testing approach has been applied to cointegration to examine the long-run re-lationship between terrorism and economic growth. The VECM Granger causality approach is used to test thedirection of causality between terrorism and economic growth. Our empirical results confirm the existence oflong-run relationship between terrorism and economic growth. The Granger causality analysis indicates thatterrorism is Granger cause of economic growth. The feedback effect is found between terrorism and tradeopenness. The relationship between terrorism and capital is bidirectional.

© 2013 Elsevier B.V. All rights reserved.

1. Introduction

The question of the economic consequences of war and internalconflicts has historically received much attention by academics andresearchers. However, a closely related significantly different formof disruption, terrorism and its impact on economic growth has notbeen as heavily analyzed or deliberated upon in existing economicliterature. Academic literature focusing on thedevelopment of terrorismhas mainly focused on and dealt with the effects of political and institu-tional factors on terrorism (Gassebner and Luechinger, 2011; Kriegerand Meierriecks, 2011). These mostly transnational studies haveshown that political and institutional factors have played an importantand influential role on the growth of terrorism than economic factors.However, previous articles tended to focus on a more transnational ap-proach while this study takes a country specific approach by focusingspecifically on Pakistan. Sanchez-Cuenca and Calle (2009) posited thatstudies focusing on transnational terrorism may bring about incorrectassumptions on domestic or country specific acts of terrorism. Thereforean economic approach to the development of terrorism in Pakistan isvital in developing a country specific model and better understandingcountry specific effects of terrorism development in Pakistan, insteadof the traditional transnational approach as seen in Piazza (2006),Krieger and Meierriecks (2011) and Gassebner and Luechinger (2011).

Now focusing on economic effects of terrorism development inPakistan there should theoretically be a negative relationship betweenterrorism events and economic growth. Terrorism has the potential

z).

rights reserved.

to impede economic activity via its multipronged affects including,but not limited to, the redirection of government expenditures fromgrowth-enhancing investment activities to less productive expendi-tures such as defense related activities. Also the reduction of foreigndirect investment (FDI) and portfolio investment (PI) can arise as a re-sult of the increase in the perceived political and country specific riskof the economy, and destruction of physical infrastructure. Further-more, with the increase in terrorist activities the probability of deathalso increases and individuals tend to associate less utility towardsfuture consumption. As a result, individuals may substitute savingsfor current consumptionwhich further weighs down the capital forma-tion process and thus hinders economic growth. Conversely, low levelsof economic development, unequal distribution of wealth, and high un-employment rates can reduce the opportunity costs of engaging in ter-rorist activities and therefore may increase terrorism.

This study focuses on the causal relationship between terrorism andeconomic growth in the case of Pakistan. Our findings show that eco-nomic growth is responsible for terrorists' activities asGranger causalityis running from economic growth to terrorism in the long run and thefeedback hypothesis exists in the short run. A feedback relationship isfound between terrorism and trade openness and the same inferencescan be drawn for terrorism and capital, as well as capital and tradeopenness. This study provides new directions for policy makers tocontrol terrorism by distributing the fruits of economic growth equallyto all segments of population. The rest of the study is organized asfollows: Section 1.1 reports terrorism events in Pakistan; Section 2highlights the review of related literature; Section 3 details the estima-tion strategy; Section 4 covers the results and discussion and Section 5concludes the study with policy implications.

22 M. Shahbaz et al. / Economic Modelling 35 (2013) 21–29

1.1. Terrorism in Pakistan

Pakistan shares its international borders with Afghanistan, China,Iran and India. Pakistan shares its longest border with India andIndo-Pakistani relations have been marked by decades of severe ad-versity including three wars and frequent minor cross-border militaryinfiltrations by both sides. The governments of both countries haveblamed their counterparts for funding and supporting separatist/terrorist organizations' activities in their territories. Pakistan sharesits second longest border with Afghanistan and Pakistani–Afghan re-lations have been affected by the issues of Pashtunistan, the Sovietwar, the advent of the Taliban, the current war in Afghanistan andAfghanistan's relations with India. The American–Afghan war in thepost 9/11 era has adversely affected the security situation in Pakistan.Retreating from Afghan territory, the Taliban were pushed into the bor-dering zones with Pakistan. At present Khyber Pakhtunkhwah (KPK)province, a border province with Afghanistan, is the source of almostall terrorist activities in Pakistan. There is a clear influx of terrorists acrossPakistani–Afghan border in the wake of military operations againstAl-Qaeda in Afghanistan led by U.S. forces. In fact, Pakistan has beenthe largest sufferer in terms of the number of terrorist activities in thepast decade. Table 1 below summarizes terrorism events in Pakistanover the seven-year period 2004–2010 by type of terrorist activity.

Despite the overall reduction in the number of terrorist attacks,the severity of the terrorist attacks in Pakistan has increased recently.According to the South Asia Terrorism Portal (SATP), a terrorismdatabase, 2654 civilians were killed in terrorist violence from January2010 to May 2011, as compared to around 1600 civilian deaths from2003 to 2006. The reasoning behind this is the willingness of terroriststo engage in suicide bomb attacks. According to SATP, there wereseventy-six suicide attacks in Pakistan in 2009 as compared to onlytwo in 2003. From the post-attack evidence, it has been observedthat terrorists are increasingly using younger children from financiallydeprived households to carry out their suicide attacks. The terroristactivities have grown in response to the antiterrorist military initia-tives of the Government in certain areas of the KPK. Though someof the attacks listed in Table 1 may have been relatively localizeddue to the limited destructive reach of many of the acts of terrorism,the overall combination of numerous smaller attacks combined withlarger attacks can be seen as an overall destruction of physical infra-structure in Pakistan.

2. Literature review

Terrorism is defined as the use of violence or the threat of violenceto induce psychological fear in noncombatant targeted audience(s)

Table 1Terrorism events in Pakistan.

2004 2005 2006 2007 2008 2009 2010 Total

Armed attack 50 166 106 315 779 850 500 2766Arson/firebombing 1 3 4 12 75 64 44 203Assault 8 12 43 82 70 30 245Assassination 3 1 4Barricade/hostage 1 7 9 22 8 47Bombing 97 246 247 460 673 687 415 2825Hijacking 2 18 7 2 29Kidnapping 2 15 19 96 312 284 112 840Near miss/non-attackincident

1 2 1 7 18 23 18 70

Other 3 2 5 1 1 12Suicide 1 1 5 41 58 84 40 230Theft 2 12 12 18 2 46Threat 1 15 1 1 18Unknown 8 2 13 49 102 18 192Vandalism 1 3 2 6Total 155 451 402 1029 2091 2215 1190 7533

Source: Worldwide Incidents Tracking System.

typically by an illicit and usually clandestine political, religious, ideo-logical, revolutionist or separatist organization in order to inducepolitical and/or economic disruptions as a short-term objective andthus to achieve other medium-to-long-run objectives through thisshort-term goal. The economic literature does not provide a conclusiveanswer regarding bidirectional causal linkage between terrorism andeconomic growth. In hindsight, an increase in economic growth rateshould lead to a decline in terrorism by increasing the opportunitycosts of engaging in terrorist activities, however on the other handif the benefits of economic growth are not widespread and there is anunequal distribution of wealth, geographically or otherwise, it maycause domestic terrorism to rise. Alternatively, an increase in terroristactivities may lead to a decline in economic growth. It is also possiblethat causality exists from both sides or there may be no causality at allbetween economic growth and terrorism.

Additionally, Krieger and Meierriecks (2010) analyzed the effectof income inequality on terrorist activities in their study covering65 countries from 1975 to 1999. They found a positive and significantcorrelation between income inequality and terrorist activities. Theyalso noted that income inequality is a highly significant predictor ofthe number of terrorist attacks. This becomes rather relevant whenlooking at Pakistan in particular. Yasmeen et al. (2011) exposed thatthe national poverty rate in Pakistan is 34% and over 80% of the popu-lation lives on less than two US dollars a day.

This would be in contradiction to Gassebner and Luechinger's(2011) assessment of 70 plus previous terrorism studies. Using extremebound testing they used infantmortality rates as their proxy for povertyand saw that it had a robust and negative relationship with terrorismactivities. The study reviewed numerous previous studies and compiledan overall broad transnational model for different factors relating toterrorist activities. They also found that economic development had aminor effect on terrorist activities.

Inter alia, Collier (1999), Frey et al. (2007), Enders and Sandler(2008), Eckstein and Tsiddon (2004) and Mirza and Verdier (2008)have discussed the theoretical framework regarding channels throughwhich terrorism impedes economic growth. The potential costs of ter-rorism borne by an economy, in terms of hampered economic growth,can be classified as direct and indirect costs. Collier (1999) identifiedthe most obvious and direct perils of civil wars, of which terrorismcan be considered a related phenomenon, as destruction of physicalcapital including the destruction of public infrastructure and the lossof human capital. Simultaneously, transaction costs are amplified asa result of the reduced security and the effectiveness of governmentinstitutions is compromised. A key factor affecting economic growthis the share of GDP directed toward investment spending. Blomberget al. (2004) and Gaibulloev and Sandler (2008) pointed out that ter-rorism diverges economic activities away from investment spendingto government spending mainly for instituting non-productive de-fense mechanisms against terrorist activities.

Knight et al. (1996) quantified the impact of military spendingon gross domestic product (GDP) and showed that an additional2.2% of GDP spent on the military, sustained over seven years (thelength of a typical conflict) would lead to a permanent loss of around2% of GDP. Abadie and Gardeazabal (2008) showed that significantreductions exist in net foreign investment positions in a countrydue to terrorist risk. Enders and Sandler (1996) investigated theimpact of terrorism on the net foreign direct investment (NFDI) inSpain and Greece using VAR analysis. They found that terrorism re-duced the NFDI by 13.5% and 11.9% in these countries respectivelyas investors sought out less violence-prone countries; however theimpact is expected to be smaller for larger diversified economies.In addition, Coe and Helpman (1995) identified that foreign directinvestment (FDI) plays a vital role in technology transfer which en-hances total factor productivity. Moreover, terrorism can adversely dis-rupt financial markets, thereby decreasing investment flows (Abadieand Gardeazabal, 2003).

1 Lütkepohl (1982) argued that omission of important variables would risk provid-ing potentially biased and inappropriate results. No causal relation is found in the bi-variate system due to these neglected variables.

23M. Shahbaz et al. / Economic Modelling 35 (2013) 21–29

Collier et al. (2002) estimated that the share of private wealth heldabroad increased from 9% to 20% for countries experiencing sustainedperiods of internal conflict. In addition to capital flight, the phenome-na of human capital flight or brain drain, population displacement,destruction of social capital and psychological effects including de-pression and posttraumatic stress disorders are also associated withterrorism and internal conflicts. Eckstein and Tsiddon (2004) andNaor (2006) argued that terrorism increases the perceived probabili-ty of an untimely death and therefore prompts people to substitutefuture savings with current consumption in order to enhance utilityin the present at the expense of the future which is another causefor decline in economic activity. Araz-Takay et al. (2009) investigatedthe macroeconomic effects of terrorism by controlling for the possiblenon-linear and endogenous relationship between political conflictand economic activity. They confirmed that terrorism has a large sig-nificant negative impact on economic activity and the impact is moresevere during expansionary periods, and that the impact of economicactivity on terrorism is significant only in recessionary periods. Interalia, Mirza and Verdier (2008) and Nitsch and Schumacher (2004)documented the negative impact of terrorism on bilateral interna-tional trade. In the case of Pakistan, a potential cost is the loss ofrevenue that could be generated from serving as a trade route be-tween India, China, Iran and other Middle Eastern states which is notcurrently possible due to massive terrorist activities in the borderingareas of Pakistan.

Terrorism can also impede economic growth through its detri-mental effect on tourism, unarguably one of the largest industries inthe world when taken into consideration with the allied businessessuch as airlines, hotels, transportation and the products and servicesconsumed by tourists. Enders et al. (1992) and Drakos and Kutan(2003) suggested that terrorismdoes have a significant negative impacton tourism. Other channels through which terrorism may impact theeconomy include increased unemployment and increased future costsof disabilities, as well as physical and mental injuries. Other studies in-cluding Enders and Sandler (1996), Abadie and Gardeazabal (2003),Tavares (2004), Chen and Siems (2004) and, Gaibulloev and Sandler(2008) have shown that economic growth is negatively affected by ter-rorism albeit to different levels in different economies.

On the contrary, Sandler and Enders (2004) and Freytag et al.(2009) posited that terrorists are rational individuals and base theirdecisions on a cost–benefit analysis of terrorist activities. Therefore,lower levels of economic activity, associated with lower opportunitycosts of terrorism, incentivize terrorist activities while high levels ofeconomic activity are associated with higher opportunity costs of ter-rorism and reduced terrorist activity. Gries et al. (2011) investigatedthe growth–terrorism causality by using data of Western Europeancountries and revealed a profound impact of economic activity on ter-rorism for only three out of seven countries. Shahbaz (2013) discussedthe feedback effect between terrorism and inflation suggesting thatterrorism widens the demand–supply gap through the destruction ofpublic infrastructure that leads to inflation which further increasesterrorist activities. Bravo and Dias (2006) have also shown that be-tween 1997 and 2004 the largest number of terrorist attacks tookplace in less developed economies with low dependence on interna-tional trade as a confirmation of the ‘deprivation’ approach to thecauses of terrorism.

Bueno de Mesquita, (2005) analyzed the interaction between ter-rorist organizations, political factors and potential terrorists. He foundthat though the pool of potential terrorist activists resides in thelower end of the socio-economic classes, the actual terrorist activiststended to be the higher educated of the lower socio-economic classes.Additionally, the author noted that government action does limit theability of terrorist organizations to put on large scale attacks, thoughthe government action does have negative economic connotations forthe local economy, as noted in other studies. Krieger and Meierrieks(2011) analyzed various origins and targets of transnational terrorism.

Their study on determinants of the origin of transnational terrorismfound that economic deprivation is not as influential in transnationalterrorism as political or institutional factors. They noted that previousresearch is conflicted in their overall assessment of links between eco-nomic development and terrorism. Additionally, they stated that thetargets of transnational terrorist activities are associated with locationswith higher economic development, higher levels of population, andpolitical instability, while rebellion and civil war is the modus operandiin less economically developed regions.

Piazza (2006) evaluated the deprivation hypothesis that poverty,inequality, poor economic development, and unemployment are theprime causes of terrorism. However, the results did not indicate anycausality between economic growth and terrorism. Instead the struc-ture of party politics was found to be the most significant predictor ofterrorism. Similarly, Pinar (2011) scrutinized the causes of separatistterrorism in South-Eastern parts of Turkey where government poli-cies are geared toward improving economic conditions in pursuanceof the widely accepted hypothesis that poverty is the main drivingforce behind separatist terrorism. However, there was no causal rela-tionship found between economic development and separatist terror-ism in South-Eastern Turkey. Recently Nasir et al. (2008) investigatedthe direction of causal relationship between economic growth and ter-rorism and found no causality running either from economic growth toterrorism or from terrorism to economic growth.1 Inter alia, Blomberget al. (2004) and Enders and Sandler (2006) have found that the ad-verse economic effects of terrorism are not statistically significant forOECD countries and mature economies. Gries et al. (2011) found thatin bivariate settings, the impact of economic performance on domesticterrorism is very strong but in trivariate settings the impact of economicgrowth on terrorism diminishes. Also, terrorism is almost never foundto affect economic growth in bivariate or trivariate specifications. Simi-larly, Meierrieks and Gries (2012) reported that economic growthGranger causes terrorism in Latin American countries. Brockhoff et al.(2012) collected data of 133 countries to explore the impact of educa-tion on domestic terrorism over the period of 1984–2007. Their resultsindicated that lower education promotes terrorism where socioeco-nomic, political and demographic conditions are critical while highereducation is linked to low terrorism in those clusters where such condi-tions are favorable. Similarly, Azam and Thelen (2008) indicated thathigher education (secondary school enrolment) is helpful in reducingdomestic terrorism and Kurrild-Klitgaard et al. (2006) found insignifi-cant impact of education on terrorism although it is positive. Sulan(2013) exposed that terrorism has eaten up 33.02% of its real macro-economy every year.

3. Estimation strategy

We have incorporated capital and trade as a potential variable toinvestigate the causal relationship between terrorism and economicgrowth by augmenting the empirical model developed by Gries et al.(2011). Terrorism lowers economic activities by destructing the phys-ical and human capital stock which are considered as important fac-tors of production. The loss of physical and human capital is linkedwith the decline in domestic production and hence economic growthis inversely affected. Terrorism affects economic growth indirectly byaffecting the allocation of domestic resources, savings, and investmentdecisions. These factors decline domestic production as well as ex-ports capacity of the country which disrupts the international marketsand in return, it declines economic growth.

24 M. Shahbaz et al. / Economic Modelling 35 (2013) 21–29

TheAutoregressiveDistributed LagModel or simply theARDL boundtesting approach to cointegration developed by Pesaran et al. (2001)has been used to conduct cointegration analysis between terrorism,economic growth, capital and trade openness in the case of Pakistan.The ARDL bound testing approach to cointegration is preferred over tra-ditional cointegration approaches due to its merits. For instance, theARDL bounds testing can be applied regardless of whether the variablesare integrated of order I(0) or integrated of order I(1). The ARDL boundtesting approach to cointegration has better properties for smalldata sample. Haug (2002) has shown that the ARDL approach tocointegration provides better results for a small sample data set as com-pared to traditional approaches such as the Engle and Granger (1987);the Johansen and Juselius (1990); and Phillips and Hansen (1990)methods. These approaches require that all the series should haveunique order of integration. In addition, unrestricted error correctionmodel (UECM) is derived from the ARDL model using simple linearspecification (Banerjee and Newman, 1993) which integrates bothlong-run as well as short-run dynamics. The UECM model does notseem to lose information about long-run relation. Another advantageof the ARDL bound testing approach is that the unrestricted model ofECM accommodates lags that capture the data generating process in ageneral-to-specific framework (Laurenceson and Chai, 2003). Theunrestricted error correction model (UECM) of the ARDL bound testingapproach to cointegration version is as follows:

Δ lnGt ¼ α∘ þ αTT þ αG lnGt−1 þ αTA lnTAt−1 þ αK lnKt−1

þαTR lnTRt−1 þXp

i¼1

αiΔ lnGt−i þXq

j¼0

αjΔ lnTAt−j

þXm

l¼0

αkΔ lnKt−l þXn

n¼0

αlΔ lnTRt−n þ μ t

ð1Þ

Δ lnTAt ¼ β∘ þ βTT þ βG lnGt−1 þ βTA lnTAt−1 þ βK lnKt−1

þβTR lnTRt−1 þXp

i¼1

βiΔ lnTAt−i þXq

j¼0

βjΔ lnGt−j

þXm

l¼0

βkΔ lnGKt−l þXn

n¼0

βlΔ lnTRt−n þ μ i

ð2Þ

Δ lnKt ¼ ϕ∘ þ ϕTT þ ϕG lnGt−1 þ ϕTA lnTAt−1 þ ϕK lnKt−1

þϕTR lnTRt−1 þXp

i¼1

ϕiΔ lnKt−i þXq

j¼0

ϕjΔ lnTAt−j

þXm

l¼0

ϕkΔ lnGt−l þXn

n¼0

ϕlΔ lnTRt−n þ μ i

ð3Þ

Δ lnTRt ¼ φ∘ þ φTT þ φG lnGt−1 þ φTA lnTAt−1 þ φK lnKt−1

þφTR lnTRt−1 þXp

i¼1

φiΔ lnTRt−i þXq

j¼0

φjΔ lnTAt−j

þXm

l¼0

φkΔ lnKt−l þXn

n¼0

φlΔ lnGt−n þ μ i

ð4Þ

Where T is trend variable, G is for economic growth proxies by realGDP per capita, terrorism is indicated by TA and proxies by number ofterrorist attacks, K indicates real capital stock per capita. We havecomputed capital stock using the perpetual inventory method withan annual depreciation of 4%. TR is for trade openness that is proxiedby real trade per capita (real exports + real imports/population). α∘,β∘, ϕ∘, φ∘ and αT, βT, ϕT, φT are the drift components and time trendsrespectively while μi is assumed to be white noise error processes.In order to ensure that serial correlation does not exist, the AkaikeInformation Criteria (AIC) is used to select the optimal lag structureof first differenced regression. Pesaran et al. (2001) determined theupper and lower critical bounds to conclude that either cointegration

for long-run relationship exists or not among the running variables.The null hypotheses of no cointegration are:

H∘ : αG ¼ αTA ¼ αK ¼ αTR ¼ 0 H∘ : βG ¼ βTA ¼ βK ¼ βTR ¼ 0H∘ : ϕG ¼ ϕTA ¼ ϕK ¼ ϕTR ¼ 0 H∘ : φG ¼ φTA ¼ φK ¼ φTR ¼ 0

The alternate hypotheses of cointegration are:

H1 : αG≠αTA≠αK ≠αTR≠0; H1 : βG≠βTA≠βK ≠βTR≠0H1 : ϕG≠ϕTA≠ϕK ≠ϕTR≠0; H1 : φG≠φTA≠φK ≠φTR≠0

The calculated F-statistics have been compared with the lowercritical bound (LCB) and upper critical bound (UCB) computed byPesaran et al. (2001) as follows:

F−statistic NUCB ¼N cointegration exists;F−statisticb LCB ¼N no cointegration existsLCB bF−statisticsbUCB ¼N inconclusive results

We have used critical bounds generated by Narayan (2005) ratherthan Pesaran et al. (2001) to test cointegration. The critical boundsgenerated by Pesaran et al. (2001) are suitable large sample size(T = 500 to T = 40,000). It is pointed out by Narayan and Narayan,(2005) that the critical values computed by Pesaran et al. (2001)may provide biased decision regarding cointegration between theseries. The critical bounds by Pesaran et al. (2001) are significantlydownwards (Narayan and Narayan, 2005). The upper and lower criticalbounds computed by Narayan (2005) are more appropriate for smallsample ranges from T = 30 to T = 80.

The direction of causal relationship between terrorism, economicgrowth, capital, and trade openness has been determined by meansof a standard Granger causality test augmented with a lagged error-correction term. According to Granger representation theorem if thevariables are integrated of order I(1) and cointegration exists amongthe variables then at least unidirectional Granger causality should exist.

Engle and Granger (1987) further elaborated that Granger causal-ity can produce misleading results if cointegrated variables are testedat first difference through vector auto regression (VAR). However, theaddition of another variable, error-correction term can help to capturethe long-run relationships. Therefore, an error-correction term is in-cluded in the augmented version of Granger causality test and the re-sult is a bivariate pth order vector error-correction model (VECM)which is as follows:

Δ lnGt ¼ α∘1 þXl

i¼1

α11Δ lnGt−i þXm

j¼1

α22Δ lnTAt−j þXn

k¼1

α33Δ lnKt−k

þXo

r¼1

α44Δ lnTRt−r þ η1ECMt−1 þ μ1i

ð5Þ

Δ lnTA ¼ β∘1 þXl

i¼1

β11Δ lnTAt−i þXm

j¼1

β22Δ lnGt−j þXn

k¼1

β33Δ lnKt−k

þXo

r¼1

β44Δ lnTRt−r þ η2ECMt−1 þ μ2i

ð6Þ

Δ lnKt ¼ ϕ∘1 þXl

i¼1

ϕ11Δ lnKt−i þXm

j¼1

ϕ22Δ lnGt−j þXn

k¼1

ϕ33Δ lnTAt−k

þXo

r¼1

ϕ44Δ lnTRt−r þ η3ECMt−1 þ μ3i

ð7Þ

Δ lnTR ¼ φ∘1 þXl

i¼1

φ11Δ lnTRt−i þXm

j¼1

φ22Δ lnGt−j þXn

k¼1

φ33Δ lnTAt−k

þXo

r¼1

φ44Δ lnKt−r þ η4ECMt−1 þ μ4i

ð8Þ

Table 2Descriptive statistics and correlation matrix.

Variables ln Gt ln TAt ln TRt ln Kt

Mean 4.3651 1.3453 8.9649 8.3414Median 4.4010 1.6120 8.9481 8.3885Maximum 4.5402 2.1825 9.4231 8.7272Minimum 4.1629 −6.73E-07 8.5452 7.9465Std. dev. 0.1180 0.6176 0.2424 0.1970Skewness −0.3699 −0.8881 0.0193 −0.3433Kurtosis 1.9434 2.5078 2.1780 2.3764Jarque–Bera 2.6341 5.3788 1.0719 1.3623Probability 0.2679 0.0679 0.5851 0.5060ln Gt 1.0000ln TAt −0.0664 1.0000ln TRt 0.1758 0.1975 1.0000ln Kt 0.0353 0.0538 −0.1558 1.0000

Table 3Clemente–Montanes–Reyes unit root test with one structural break.

Variable Innovative outliers Additive outlier

t-Statistic TB1 Decision t-Statistic TB1 Decision

ln TAt −2.941 1982 Unit root exists −3.678** 1992 Stationaryln Gt −1.848 1992 Unit root exists −5.405* 1989 Stationaryln Kt −3.806 1981 Unit root exists −4.361* 1990 Stationaryln TRt −3.458 1984 Unit root exists −5.438* 2006 Stationary

Note: * indicates significance at 1% level of significance.

25M. Shahbaz et al. / Economic Modelling 35 (2013) 21–29

Where difference operator is indicated by Δ; lagged of residualterm generated from long-run equation i.e. ECMt − 1 and μ1i, μ2i, μ3iand μ4i are error terms assumed to be normally distributed withzero mean and finite covariance matrix. The existence of a short-runcausal relation is indicated by significance of t-values of 1st differencedvariables and significance of t-values relating to error-correction termconfirms a long-run causal relationship.

For example, α11,i ≠ 0 ∀ i indicates that causality is running fromterrorism to economic growth in the short run. The joint short-runand long-run Granger causality is investigated by the significanceof joint χ2-statistic on the lagged error-correction term and firstdifference lagged concerned independent variable. However, the cau-sality should be interpreted in strict Granger causality sense i.e. it isonly predictive and not deterministic.

The data on terrorism (terrorist incidents) is collected from SouthAsian Terrorism Portal (SATP), maintained by the Institute of ConflictManagement, India.2 The data on real gross fixed capital formation,trade openness (exports + imports) and real GDP has been obtainedfrom world development indicators (CD-ROM, 2011). We also useddata on population to convert all the series into per capita. All theseries have been transformed into logarithms to attain consistentand efficient results. The study covers the time period of 1973–2010.

4. Results and their discussions

Table 2 details the descriptive statistics and correlation matrix.The results reveal that all the series are normally distributed andshows the authenticity of data to be used in this analysis as confirmedby Jarque–Bera statistics. The correlation analysis indicates the nega-tive correlation between terrorism and economic growth. A positivecorrelation is found between trade openness and economic growthand the same inference is drawn for capital and economic growth.Terrorism is also positively correlated with trade openness andcapital. A negative correlation also exists between capital and tradeopenness.

Finally, the ARDL cointegration approach can only be used if thevariables are stationary either at I(0) or I(1) or mutually cointegrated.In the case where variables are integrated at I(2), calculated F-statisticcannot be used to determine the long-run relationship. In order toverify whether any variable is integrated at I(2), ADF unit root testby Dickey and Fuller (1979), DF-GLS unit root test by Elliot et al.(1996) and Ng–Perron unit root test by Ng and Perron (2001) were

2 SATP compiles terrorist attacks in Pakistan in the form of descriptive news ar-ranged chronologically, derived from various news sources, separating suicide attacksprovides a unique dataset, to study pure effect of terrorism as opposed to effect ofothers forms of conflict as studies, typically, clump together insurgencies and acts ofwarfare and crime under the umbrella of terrorism. Furthermore, as mentioned abovesuicide incidents do not suffer from the same degree of reporting bias as compared toother terrorist incidents, due to their inherent spectacular nature.

applied.3 Baum (2004) contested that ADF, DF-GLS and Ng–Perronunit root tests do not provide information about structural breaksin the series and their results may be biased. To resolve this issue,we used a Clemente et al. (1998) detrended structural break unitroot test with one and two structural breaks occurring in series.Clemente–Montanes–Reyes unit root test provides information abouttwo possible structural break points in the series through (1) an addi-tive outlier (AO) model that points out sudden changes in the meanof a series and (2) an innovational outlier (IO) model that indicatesgradual shifts in the mean of the series. As a result, the additive outliermodel is more appropriate for series having sudden structural changesas compared to gradual shifts. The results of Clemente–Montanes–Reyes unit root test with one structural break are reported in Table 3while the results for this test with two structural breaks are reportedin Table 4.

The results of Clemente–Montanes–Reyes unit root test show thatterrorism, economic growth, capital and trade openness have unitroot problem at I(0) while the variables become stationary at I(1).The results of Clemente–Montanes–Reyes unit root test lead us to in-vestigate the long-run relationship between the series by applyingthe ARDL bound testing approach to cointegration. The ARDL boundstesting approach requires the selection of appropriate lag length asthe F-statistic is very sensitive to lag order of the variables (Feridunand Shahbaz, 2010). We followed AIC criterion to choose the appro-priate lag length that provides appropriate information regardinglag order selection. Lag length is shown in the third line of Table 5.Table 5 provides the results of the ARDL bounds testing approach tocointegration. The calculated F-statistics are 9.896, 6.862 and 6.775greater than upper critical bounds generated by Narayan (2005) at1% and 5% levels of significance when terrorism, capital and tradeopenness are treated as dependent variables. The critical boundsdeveloped by Pesaran et al. (2001) are not suitable for small sampledata. Our analysis indicates that there are three cointegrating vectorswhich validate the existence of a long-run relationship between eco-nomic growth, terrorism, capital, and trade openness in the case ofPakistan for the period of 1973–2010.

At the 5% significance level, all diagnostic tests do not exhibit anyevidence of violation of the classical linear regression model (CLRM)assumptions. Specifically, Jarque–Bera (J–B) normality test cannot re-ject the null hypothesis, meaning that the estimated residuals are nor-mally distributed and the standard statistical inferences (i.e. t-statistic,F-statistic, and R-squares) are valid. At the same level of significance,both Breusch–Godfrey LM test and ARCH LM test consistently revealthat the residuals are not serially correlated, and are also free fromheteroskedasticity problems. There is no specification problem withthe models. It is indicated that all series such as economic growth,terrorism, capital, and trade openness have unit root problems attheir level form while they are found to be stationary at 1st difference.It implies that the variables are integrated at I(1). This unique levelof integration leads us to use Johansen multivariate approach tocointegration for robustness of long-run relationships. The findingsshow that there are two cointegration vectors between economic

3 Results of these tests are available upon request from authors.

Table 5The results of ARDL cointegration test.

Bounds testing to cointegrationDependent variable Gt = f (TAt, Kt, TRt) TAt = f (Gt, Kt, TRt) Kt = f (Gt, TAt, TRt) TRt = f (Gt, Kt, TAt)Optimal lag length (2, 2, 2, 1) (1, 1, 2, 2) (2, 1, 1, 2) (2, 2, 2, 1)F-statistics 1.222 9.896* 6.862** 6.775**

Critical values (T = 38)

Lower bounds I(0) Upper bounds I(1)

1% level 6.053 7.4585% level 4.450 5.56010% level 3.740 4.780

Diagnostic testsR2 0.5887 0.6823 0.7214 0.7642F-statistics 1.6101 (0.1645) 2.8638 (0.0147) 3.280 (0.0081) 3.6475 (0.0049)J–B normality test 0.5555 (0.7545) 0.9671 (0.6166) 1.0746 (0.5842) 0.3394 (0.8438)Breusch–Godfrey LM test 2.4172 (0.1210) 2.2035 (0.1393) 1.5722 (0.2363) 0.8113 (0.4334)ARCH LM test 0.2037 (0.6548) 1.5427 (0.2230) 0.2222 (0.6405) 0.8730 (0.3571)W. Heteroskedasticity test 0.9319 (0.5520) 0.9572 (0.5263) 1.6571 (0.1485) 0.9905 (0.5039)Ramsey RESET 0.0318 (0.8604) 0.0008 (0.9774) 0.5020 (0.4877) 0.0561 (0.8144)

Note: *, ** and*** indicate significance at 1%, 5% and 10% levels respectively.

Table 4Clemente–Montanes–Reyes unit root test with two structural breaks.

Variable Innovative outliers Additive outlier

t-Statistic TB1 TB2 Decision t-Statistic TB1 TB2 Decision

ln TAt −4.256 1982 1986 Unit root exists −6.582* 1992 2002 Stationaryln Gt −2.155 1992 1992 Unit root exists −6.020* 1989 2001 Stationaryln Kt −4.743 1981 1986 Unit root exists −6.087* 1990 2004 Stationaryln TRt −3.746 1984 1989 Unit root exists −5.570** 2004 2006 Stationary

Note: * indicates significance at 1% level of significance.

26 M. Shahbaz et al. / Economic Modelling 35 (2013) 21–29

growth, terrorism, capital, and trade openness in the case of Pakistan forthe period of 1973–2010 which confirm the robustness of the long-runrelation (Table 6).

The next step is to investigate the direction of causality betweeneconomic growth, terrorism, capital, and trade openness after findingevidence of cointegration. The VECM Granger causality approachshould be conducted when variables are cointegrated. The VECMGranger causality approach provides short-run and long-run causalrelationships between economic growth, terrorism, capital, andtrade openness. The statistical significance of lagged residual termi.e. ECMt − 1 indicates long-run Granger causality while the joint sig-nificance of the lagged explanatory variables shows the short-runcausal relationship between the variables. The results of the Grangercausality test are reported in Table 7.

The results point out that there is bidirectional causal relationbetween terrorism and capital, trade openness and terrorism and,capital and trade openness in the long run. The feedback effect be-tween terrorism and capital reveals that terrorism activities lead toan increase in public capital loss by destroying public infrastructuresuch as roads, schools, hospitals, telecommunications and banks, etc.which leads to a decline in production and increases the gap betweendemand and supply. Though many of the acts of terrorism may be

Table 6Results of test of cointegration.

Hypothesis Trace teststatistic

5% CV Hypothesis Maximumeigenvalue

5%CV

R = 0 90.9337* 47.8561 R = 0 50.8150* 27.5843R ≤ 1 40.1187* 29.7970 R = 1 26.8379* 21.1316R ≤ 2 13.2807 15.4947 R = 2 10.9689 14.2646R ≤ 3 2.31176 3.8414 R = 3 2.3117 3.8414

Note: * indicates significance at 1% level of significance.

considered small due to their low destructive nature, the overall effectcan be seen as an increase in public capital loss.

Additionally, the aforementioned increase in the gap between de-mand and supply can lead to a rise in inflation which is resulting fromincreased terrorist activities (Shahbaz, 2013). From the other side,rising inflation increases poverty that further promotes terrorism. Thebidirectional causal relationship between terrorism and trade opennessindicates that a rise in terrorism Granger causes international capitaland trade flows by lowering foreign direct investment aswell as domes-tic output and increases capital outflow (Shahbaz et al., 2010). Thisleads to a lower share of exports in international markets. The threatof terrorism not only declines public investment but also lowers foreigndirect investment in the host country. This leads to an increase in unem-ployment which in turn increases terrorist activities.

The unidirectional causality is found running from economic growthto terrorism. The rise in per capita income (economic growth) contrib-utes to terrorism. The main reason is unequal distribution of incomeindicating that economic growth benefits the elite class as comparedto the bottom 20% segment of the population leading to high levelsof poverty4 that is resulting from increased terrorism in Pakistan. Thehigh income inequality and poverty also stimulate misery in the coun-try which also increases terrorist's activities.

In the short run, terrorism and economic growthGranger cause eachother and unidirectional causality is found running from terrorism totrade openness. In addition to that, the significance of ECMt − 1 also ex-hibits that if the system exposes to shock it will converge to long-runequilibrium at a relatively high speed for terrorism (−0.9149), andtrade openness (−0.6172) the VECMs as compared to the convergencespeed for capital (−0.3948) where the numbers in parentheses are theVECMs (vector error correction terms).

4 The recent wave of inflation is hitting the poor segments of population significantlyand more than 40% of the population of Pakistan is living below the poverty line.

Table 7The VECM Granger causality analysis.

Dependent variable Type of Granger causality

Short-run Long-run Joint (short- and long-run)

Δln Gt Δ ln TAt Δ ln Kt Δ ln TRt ECMt − 1 Δ ln Gt, ECMt − 1 Δ ln TAt, ECMt − 1 Δ ln Kt, ECMt − 1 Δ ln TRt, ECMt − 1

F-statistics [p-values] [t-Statistics] F-statistics [p-values]

Δ ln Gt – 3.3439**[0.0499]

0.1059[0.8998]

0.7608[0.4767]

−0.0345[0.6916]

– 2.2293[0.1068]

0.2007[0.8950]

0.5278[0.6668]

Δ ln TAt 3.4095**[0.0473]

– 0.32980.7218]

1.6949[0.2019]

−0.9149*[−4.8043]

8.2315*[0.0004]

– 5.7403*[0.0034]

7.2372*[0.0010]

Δ ln Kt 1.3120[0.2853]

0.2623[0.7711]

– 0.3040[0.7402]

−0.3948*[−3.0578]

4.8695*[0.0075]

3.4437**[0.0300]

– 3.8744**[0.0195]

Δ ln TRt 2.3616[0.1128]

3.3626**[0.0491]

0.1062[0.8996]

– −0.6172*[−3.2310]

7.3363*[0.0009]

5.1608*[0.0058]

3.9605**[0.0180]

–

Note: The asterisks ***, ** and * denote the significance at the 1, 5 and 10% levels, respectively.

27M. Shahbaz et al. / Economic Modelling 35 (2013) 21–29

The main drawback of causality tests that was pointed out byWolde-Rufael (2009) is that Granger causality tests do not seem to de-termine the relative strength of causality effects beyond the selectedtime period. In such circumstances, causality tests are inappropriatebecause these tests are unable to indicate how much feedback hasexisted from one variable to another. To examine the feedback fromone variable to another and to check the relative effectiveness ofcausality effects ahead of sample period, we have applied variance de-composition to examine the direction of causality between economicgrowth, terrorism, capital, and trade openness following Wolde-Rufael(2009). It is noted that variance decomposition is applied to investigatethe response of the dependent variable to shocks stemming from inde-pendent variables. The variance decomposition method is an alternateof impulse response function. This process explains how much of thepredicted error variance for any variable is described by innovationsgenerated throughout each independent variable in a system overvarious time horizons. The results reported in Table 8 show that

Table 8Variance decomposition approach.

Period S.E. ln Gt ln TAt ln Kt ln TRt

Variance decomposition of ln Gt:1 0.0171 100.000 0.0000 0.0000 0.00002 0.0253 95.7342 2.8060 0.0692 1.39053 0.0330 96.7350 1.6677 0.7440 0.85304 0.0407 96.1298 2.7252 0.5642 0.58065 0.0473 95.1751 3.7285 0.4189 0.67736 0.0525 93.3076 5.2813 0.7203 0.69067 0.0569 91.0538 6.4447 1.8698 0.63158 0.0608 88.1688 7.1042 4.0138 0.71319 0.0640 85.1269 7.2361 6.4328 1.204010 0.0666 82.1263 7.4313 8.2954 2.146711 0.0687 79.5296 7.8269 9.5986 3.044612 0.0704 77.4896 8.2190 10.7047 3.586513 0.0719 75.9137 8.3708 11.9178 3.797514 0.0733 74.6402 8.3041 13.1586 3.896915 0.0746 73.6039 8.1660 14.1595 4.0704

Variance decomposition of ln TAt:1 0.6900 5.4220 94.5779 0.0000 0.00002 0.7019 7.1714 91.4084 1.2269 0.19323 0.7483 10.7323 80.7862 4.8748 3.60654 0.8125 17.8724 69.0423 7.6839 5.40125 0.8258 20.0889 66.8357 7.7768 5.29856 0.8477 21.3290 64.5516 7.8932 6.22607 0.8799 23.0665 61.6154 7.8159 7.50208 0.9005 24.8782 58.8754 9.0799 7.16639 0.9129 24.7805 57.3249 9.8869 8.007510 0.9199 24.4275 56.7778 9.7738 9.020711 0.9280 24.0202 57.0035 9.6041 9.372012 0.9307 23.9243 57.0034 9.7536 9.318513 0.9358 23.7635 56.3904 10.4764 9.369514 0.9419 23.6115 55.7122 11.3785 9.297615 0.9461 23.5472 55.2757 11.7250 9.4518

economic growth is explained predominantly by its own innovativeshocks (73.60%) while terrorism, capital and trade openness explaineconomic growth through their innovative shocks accounting for8.16%, 14.15%, and 4.07% respectively.

On the other hand, empirical evidence indicates that economicgrowth explains a substantial portion of terrorism by its innovativeshocks i.e. 23.54% while 55.27% of terrorism is due to its own innovativeshocks. This shows that 55.27% is the impact of factors outside the empir-ical model such as political and institutional factors. Capital and tradeopenness also contribute to terrorism through their shocks but their im-pact isminimal i.e. 11.72% and 9.45% respectively. This implies a unidirec-tional causal relationship running from economic growth to terrorism.This finding is consistent with the VECM Granger causality analysis.

Table 9 reveals that a substantial portion (32.14%) of capital isexplained by shocks in economic growth while 37.30% is due to itsown innovative shocks. Trade openness and terrorism explain capitalby 22.43% and 8.11% respectively. This implies that economic growth

Table 9Variance decomposition approach.

Period S.E. ln Gt ln TAt ln Kt ln TRt

Variance decomposition of ln Kt:1 0.0202 0.0574 0.4537 99.4887 0.00002 0.0268 1.3251 0.6134 95.8849 2.176413 0.0333 10.3992 2.0542 73.2559 14.29054 0.0395 20.6573 1.8938 53.9265 23.52225 0.0434 30.4922 2.1242 44.9182 22.46536 0.0457 36.2967 1.9242 40.9874 20.79157 0.0474 36.7957 2.5334 39.5615 21.10938 0.0489 35.9262 5.6525 38.3846 20.03669 0.0500 35.6094 7.7441 36.7781 19.868210 0.0511 35.2184 7.5349 37.1354 20.111111 0.0522 34.3994 7.3861 38.8720 19.342312 0.0529 33.6161 7.2977 39.1623 19.923713 0.0536 32.7548 7.2059 38.1137 21.925414 0.0541 32.2947 7.6842 37.5023 22.518615 0.0543 32.1494 8.1173 37.3022 22.4309

Variance decomposition of ln TRt:1 0.0303 0.0571 6.1017 5.3815 88.45952 0.0364 14.8940 6.0544 9.7008 69.35073 0.0427 24.7292 16.9535 7.3253 50.99184 0.0469 26.2910 16.6550 14.5289 42.52495 0.0530 27.3962 18.2140 20.7428 33.64696 0.0578 33.4392 18.0210 19.3849 29.15477 0.0608 38.5589 16.4629 17.6675 27.31058 0.0629 40.7007 15.3884 17.6156 26.29529 0.0643 42.0356 14.9876 17.8063 25.170310 0.0656 41.7727 14.7717 17.5682 25.887311 0.0666 40.8587 14.9653 17.1698 27.006012 0.0670 40.3330 15.8227 17.0508 26.793313 0.0676 39.8850 16.0814 17.5435 26.489914 0.0684 39.3574 15.7107 18.9732 25.958515 0.0691 38.8366 15.5478 20.0502 25.5652

28 M. Shahbaz et al. / Economic Modelling 35 (2013) 21–29

and trade openness cause capital. Finally, 25.56% of trade openness isexplained by its own innovative shocks while 38.83%, 15.54% and20.05% is due to economic growth, terrorism and capital. There is uni-directional causality that runs from economic growth to trade open-ness validating growth-led-trade hypothesis in Pakistan confirmedby Shahbaz (2012).

5. Conclusions and future research

This paper aims to investigate the causal relationship betweenterrorism and economic growth by incorporating capital and tradeopenness as potential variables in the period of 1973–2010. TheARDL bounds testing approach to cointegration and the VECMGrangercausality approaches have been applied to test the long-run relation-ship and direction of causality between the variables. Our empiricalresults confirm a long-run relationship between economic growth,terrorism, capital, and trade openness in the case of Pakistan. Thebidirectional causality is found between terrorism and capital, tradeopenness and capital, and terrorism and trade openness. The unidirec-tional causality is running from economic growth to terrorism. In suchcircumstances, the government must pay her attention to distributethe fruits of economic growth on equal basis to control terrorism.The trickle-down effect of economic growth will lower income in-equality and poverty which may be helpful in controlling terrorists'activities.

Our study has potential to incorporate other factors affectingterrorism in the case of Pakistan. For example, we can include incomeinequality (poverty), capital flight, and foreign direct investmentfollowing Krieger and Meierriecks (2010), Gassebner et al. (2008)and Ali (2011) to reinvestigate the relationship between terrorismand economic growth in Pakistan. Our empirical model has potentialfor political instability following Krieger and Meierrieks (2011) whileinvestigating causality between economic growth and terrorism. Arapid change of political governments may provide a vacuum toterrorists to peruse their agendas. So, weak or failed states may con-tribute to international terrorism due to domestic political instability(Krieger and Meierrieks, 2011).

References

Abadie, A., Gardeazabal, J., 2003. The economic costs of conflict: a case study of theBasque Country. American Economic Review 93, 113–132.

Abadie, A., Gardeazabal, J., 2008. Terrorism and the world economy. European EconomicReview 52, 1–27.

Ali, A., 2011. Economic Cost of Terrorism: A Case Study of Pakistan. http://www.issi.org.pk/publication-files/1299569657_66503137.pdf.

Araz-Takay, B., Arin, K.P., Omay, T., 2009. The endogenous and non-linear relationshipbetween terrorism and economic performance: Turkish evidence. Defence andPeace Economics 20 (1), 1–10.

Azam, J.-P., Thelen, V., 2008. The roles of foreign aid and education in the war on terror.Public Choice 135, 375–397.

Banerjee, A., Newman, A., 1993. Occupational choice and the process of development.Journal of Political Economy 101 (2), 274–298.

Baum, C.F., 2004. A review of Stata 8.1 and its time series capabilities. InternationalJournal of Forecasting 20 (1), 151–161.

Blomberg, S.B., Hess, G.D., Athanasios, O., 2004. The macroeconomic consequences ofterrorism. Journal of Monetary Economics 51 (2), 1007–1032.

Bravo, A.B.S., Dias, C.M., 2006. An empirical analysis of terrorism: Islamism and geopo-litical factors. Defence and Peace Economics 17 (4), 329–341.

Brockhoff, S., Krieger, T., Meierrieks, D., 2012. Great expectations and hard times — the(nontrivial) impact of education on domestic terrorism. CEB Working PaperNo. 12/004, Solvay Brussels School of Economics and Management, Centre EmileBernheim.

Bueno de Mesquita, E., 2005. The quality of terror. American Journal of Political Science49 (3), 515–530.

Chen, A.H., Siems, T.F., 2004. The effects of terrorism on global capital markets. EuropeanJournal of Political Economy 20 (2), 349–366.

Clemente, J., Montanes, A., Reyes, M., 1998. Testing for a unit root in variables with adouble change in the mean. Economics Letters 59 (2), 175–182.

Coe, D.T., Helpman, E., 1995. International R & D spillovers. European Economic Review39 (5), 859–887.

Collier, P., 1999. On the economic consequences of civil war. Oxford Economic Papers51 (1), 168–183.

Collier, P., Hoeffler, A., Pattillo, C., 2002. Africa's Exodus: Capital Flight and the BrainDrain as Portfolio Decisions. World Bank, Washington, D.C. (Processed).

Dickey, D., Fuller, W.A., 1979. Distribution of the estimates for autoregressive timeseries with unit root. Journal of the American Statistical Association 74 (366),427–431.

Drakos, K., Kutan, A.M., 2003. Regional effects of terrorism and tourism in threeMediterranean countries. Journal of Conflict Resolution 47 (5), 621–641.

Eckstein, Z., Tsiddon, D., 2004. Macroeconomic consequences of terror: theory and thecase of Israel. Journal of Monetary Economics 51 (1), 971–1002.

Elliot, G., Rothenberg, T.J., Stock, J.H., 1996. Efficient tests for an autoregressive unitroot. Econometrica 64 (4), 813–836.

Enders, W., Sandler, T., 1996. Terrorism and foreign direct investment in Spain andGreece. Kyklos 49 (3), 331–352.

Enders, W., Sandler, T., 2006. The Political Economy of Terrorism. Cambridge UniversityPress, Cambridge.

Enders, W., Sandler, T., 2008. Economic consequences of terrorism in developedand developing countries: an overview. In: Keefer, P., Loyaza, N. (Eds.), Terrorism,Economic Development, and Political Openness. Cambridge University Press,New York, pp. 17–47.

Enders, W., Sandler, T., Parise, P., 1992. An economic analysis of the impact of terrorismon tourism. Kyklos 45 (4), 531–554.

Engle, R.F., Granger, C.W.J., 1987. Cointegration and error correction representation:estimation and testing. Econometrica 55 (2), 251–276.

Feridun, M., Shahbaz, M., 2010. Fighting terrorism: are military measureseffective? Empirical evidence from Turkey. Defence and Peace Economics 21 (2),193–205.

Frey, B.S., Luechinger, S., Stutzer, A., 2007. Calculating tragedy: assessing the costs ofterrorism. Journal of Economic Surveys 21 (1), 1–24.

Freytag, A., Jens, J.K., Meierrieks, D., Schneider, F., 2009. The Origins of Terrorism —

Cross-country Estimates on Socio-economic Determinants of Terrorism. WorkingPapers 19, University of Paderborn, CIE Center for International Economics.

Gaibulloev, K., Sandler, T., 2008. Growth consequences of terrorism in Western Europe.Kyklos 61 (3), 411–424.

Gassebner, M., Luechinger, S., 2011. Lock, stock, and barrel: a comprehensive assessmentof determinants of terror. Public Choice 149, 235–261.

Gassebner, M., Jong-A-Pin, R., Mierau, J.O., 2008. Terrorism and electoral accountabili-ty: one strike, you're out! Economics Letters 100, 126–129.

Gries, T., Krieger, T., Daniel, M., 2011. Causal linkages between domestic terrorism andeconomic growth. Defence and Peace Economics 22 (5), 493–508.

Haug, A., 2002. Temporal aggregation and the power of cointegration tests: a MonteCarlo study. Oxford Bulletin of Economics and Statistics 64, 399–412.

Johansen, S., Juselius, K., 1990. Maximum likelihood estimation and inference oncointegration with applications to the demand for money. Oxford Bulletin ofEconomics and Statistics 52, 169–210.

Knight, M., Loayza, N., Villanueva, D., 1996. The peace dividend: military spending cutsand economic growth. IMF Staff Papers 43 (1), 1–37.

Krieger, T., Meierriecks, D., 2010. Does income inequality lead to terrorism? Availableat SSRN: http://ssrn.com/abstract=1647178 (or http://dx.doi.org/10.2139/ssrn.1647178).

Krieger, T., Meierriecks, D., 2011. What causes terrorism? Public Choice 147, 3–27.Krieger, T., Meierrieks, D., 2011. Does Inequality Lead to Terrorism? SSRN: http://ssrn.

com/abstract=1647178.Kurrild-Klitgaard, P., Justesen, M.K., Klemmensen, P., 2006. The political economy of

freedom, democracy and transnational terrorism. Public Choice 128, 289–315.Laurenceson, J., Chai, J.C.H., 2003. Financial Reforms and Economic Development in

China. UK, Edward Elgar, Cheltenham.Lütkepohl, H., 1982. Non-causality due to omitted variables. Journal of Econometrics

19, 367–378.Meierrieks, D., Gries, T., 2012. Economic performance and terrorist activity in Latin

America. Defence and Peace Economics. http://dx.doi.org/10.1080/10242694.2012.656945.

Mirza, D., Verdier, T., 2008. International trade, security and transnational terrorism:theory and a survey of empirics. Journal of Comparative Economics 36 (2),179–194.

Naor, Z., 2006. Untimely death, the value of certain lifetime and macroeconomicdynamics. Defence and Peace Economics 17 (4), 343–359.

Narayan, P.K., 2005. The saving and investment nexus for China: evidence fromcointegration tests. Applied Economics 37 (17), 1979–1990.

Narayan, P.K., Narayan, S., 2005. Estimating income and price elasticities of imports forFiji in a cointegration framework. Economic Modelling 22 (3), 423–438.

Nasir, M., Arif, A., Rehman, F. Ur, Tariq, M.S., 2008. Terrorism and economic growth:a case study of Pakistan. GCU Economic Journal 41 (2), 1–22.

Ng, S., Perron, P., 2001. Lag length selection and the construction of unit root test withgood size and power. Econometrica 69 (6), 1519–1554.

Nitsch, V., Schumacher, D., 2004. Terrorism and international trade: an empirical inves-tigation. European Journal of Political Economy 20 (2), 423–433.

Pesaran, M.H., Shin, Y., Smith, R.J., 2001. Bounds testing approaches to the analysis oflevel relationships. Journal of Applied Econometrics 16 (3), 289–326.

Phillips, P.C.B., Hansen, B.E., 1990. Statistical inference in instrumental variables regres-sion with I(1) processes. Review of Economic Studies 57, 99–125.

Piazza, J.A., 2006. Rooted in poverty? Terrorism, poor economic development, and socialcleavages. Terrorism and Political Violence 18, 159–177.

Pinar, D.-G., 2011. Separatist terrorism and the economic conditions in South-EasternTurkey. Defence and Peace Economics 22 (4), 393–407.

Sanchez-Cuenca, I., Calle, L., 2009. Domestic terrorism: the hidden side of politicalviolence. Annual Review of Political Science 12, 31–49.

29M. Shahbaz et al. / Economic Modelling 35 (2013) 21–29

Sandler, T., Enders, W., 2004. An economic perspective on transnational terrorism.European Journal of Political Economy 20 (2), 301–316.

Shahbaz, M., 2012. Does trade openness affect long run growth? Cointegration, causalityand forecast error variance decomposition tests for Pakistan. Economic Modelling29, 2325–2339.

Shahbaz, M., 2013. Linkages between inflation, economic growth and terrorism inPakistan. Economic Modelling 32, 496–506.

Shahbaz, M., Ahmad, N., Wahid, A.N.M., 2010. Savings–investment correlationand capital outflow: the case of Pakistan. Transition Studies Review 17 (1), 80–97.

Sulan, M., 2013. Terrorism and the macroeconomy: evidence from Pakistan. Defenceand Peace Economics. http://dx.doi.org/10.1080/10242694.2013.793529.

Tavares, J., 2004. The open society assesses its enemies: shocks, disasters and terroristattacks. Journal of Monetary Economics 51 (5), 1039–1070.

Wolde-Rufael, Y., 2009. The defence spending-external debt nexus in Ethiopia. Defenceand Peace Economics 20 (5), 423–436.

Yasmeen, G., Begum, R., Mujtaba, B., 2011. Human development challenges and oppor-tunities in Pakistan: defying income inequality and poverty. Journal of BusinessStudies Quarterly 2, 1–12.