CREATE Research Archive

Published Articles & Papers

1-1-2010

The Adverse Effect of Transnational and DomesticTerrorism on Growth in AfricaKhusrav GaibulloevUniversity of Texas at Dallas

Todd SandlerUniversity of Texas at Dallas, tsandler@utdallas.edu

Follow this and additional works at: http://research.create.usc.edu/published_papers

This Article is brought to you for free and open access by CREATE Research Archive. It has been accepted for inclusion in Published Articles & Papersby an authorized administrator of CREATE Research Archive. For more information, please contact gribben@usc.edu.

Recommended CitationGaibulloev, Khusrav and Sandler, Todd, "The Adverse Effect of Transnational and Domestic Terrorism on Growth in Africa" (2010).Published Articles & Papers. Paper 152.http://research.create.usc.edu/published_papers/152

The Adverse Effect of Transnational and Domestic Terrorism on Growth in Africa

by

Khusrav Gaibulloev kng031000@utdallas.edu

and

Todd Sandler*

School of Economic, Political & Policy Sciences University of Texas at Dallas

800 W. Campbell Road Richardson, TX 75080-3021 USA

tsandler@utdallas.edu Phone 1-972-883-6725 Fax 1-972-883-6486

May 2010

Running Title: Terrorism Impact on African Growth Word Count: 8,349 *This study was funded by the US Department of Homeland Security through the Center for Risk and Economic Analysis of Terrorism Events (CREATE) at the University of Southern California, grant number 2007-ST-061-000001. However, any opinions, findings, and conclusions or recommendations are solely those of the authors and do not necessarily reflect the view of the Department of Homeland Security or CREATE. Gaibulloev is a Postdoctoral Researcher and Sandler is the Vibhooti Shukla Professor of Economics and Political Economy.

The Adverse Effect of Transnational and Domestic Terrorism on Growth in Africa

Abstract

With panel estimates, this paper investigates the neoclassical determinants of income per capita

growth for 51 African countries for 1970–2007, while accounting for cross-sectional (spatial)

dependence and conflict (i.e., terrorism, internal conflicts, and external wars). For the entire

sample, fixed-effects panel estimates find that transnational terrorism has a significant, but

modest, marginal impact on income per capita growth. These results hold for two different

terrorism event data sets. However, domestic terrorist events do not affect income per capita

growth. This suggests that an earlier growth study, which did not include domestic terrorist

events for a different sample and time period, provided an accurate picture for Africa. The paper

contains a host of robustness checks that find virtually identical results. Alternative terrorist

variables are also used, with little qualitative change in the findings. The absence of a domestic

terrorism impact is surprising because there were generally many more domestic than

transnational terrorist incidents in Africa. To promote growth, host and donor countries must

direct scarce counterterrorism resources to protect against transnational terrorism in particular.

Keywords: growth in Africa, transnational terrorism, domestic terrorism, conflict, fixed-effects

panel

The Adverse Effect of Transnational and Domestic Terrorism on Growth in Africa

Introduction

In their bid to force governments to concede to their demands, terrorists plan attacks that have

adverse consequences on targeted countries’ economies. Thus, Euskadi ta Askatasuna (ETA)

targeted tourist sites and commerce centers in Spain, while Jemaah Islamiyah bombed a popular

nightclub in Bali and a tourist hotel in Jakarta. Modern-day terrorists have damaged

infrastructure – e.g., train stations, bus stations, airports, and stock exchanges – not only to create

anxiety in a targeted audience, but also to disrupt the economy. Terrorists hope that economic

costs when combined with human losses from economic-damaging attacks will pressure

besieged governments to concede to their political demands. In Africa, terrorist groups have also

sought out economic targets – e.g., the Islamic Group staged the Luxor massacre of tourists on

17 November 1997 at Hatshepsut’s Temple in Egypt. This armed attack murdered 62 and

injured 24 (Mickolus & Simmons, 2002). The car bombing of the Israeli-owned Paradise Hotel

in Mombasa, Kenya on 28 November 2002 by an al-Qaida affiliated Somali group killed 16

(including 3 suicide terrorists) and injured 80 (Mickolus & Simmons, 2006). Two surface-to-air-

missiles on this same day narrowly missed hitting an Israeli-chartered airline taking off from

Mombasa airport.

Terrorism can negatively influence a targeted country’s economic growth through a

number of channels.1 First, terrorist attacks may enhance uncertainty which limits investments

and diverts foreign direct investment to safer venues, as documented by a number of studies

(Abadie & Gardeazabal, 2008; Enders, Sachsida & Sandler, 2006; Enders & Sandler, 1996).

Second, augmented security outlays by a targeted government may crowd out productive public

and private investment (Blomberg, Hess & Orphanides, 2004; Gaibulloev & Sandler, 2008,

2009). Third, a terrorist campaign raises the costs of doing business through higher wages,

2

larger insurance premiums, and greater security expenditures, which, in turn, decrease profits,

productivity, and growth. Fourth, terrorist attacks may dampen growth by destroying or

degrading social overhead capital that facilitates commerce and daily routines. Disruption to

transportation, communication, and electricity infrastructure may have short-term dire economic

consequences. Fifth, terrorism impacts specific industries – e.g., airlines and tourism (Drakos,

2004; Drakos & Kutan, 2003; Enders, Sandler & Parise, 1992; Ito & Lee, 2005) – which, in turn,

may limit growth. This may be especially true when terrorists target export-sector assets in an

export-led-growth economy. If such attacks on, say, a country’s mineral wealth make export of

resources unreliable, then importing countries will turn to more reliable sources of supply when

available. Sixth, terrorism may cause donor countries to curtail foreign assistance owing to

stability concerns. On a smaller scale, terrorism adversely affects economic growth for many of

the same reasons – e.g., capital flight, increased uncertainty, destroyed infrastructure, and

increased security spending – that internal conflicts or civil wars impede economic growth.

The primary purpose of this paper is to present panel estimates of the neoclassical

determinants of income per capita growth for 51 African countries for 1970–2007, while

accounting for cross-sectional (spatial) dependence and conflict (i.e., terrorism, internal conflicts,

and external wars). We are particularly interested in the impact of terrorism on growth, because

other studies focused on the impact of internal wars on growth (e.g., Collier & Hoeffler, 2002;

Murdoch & Sandler, 2002a). Unlike Blomberg, Hess & Orphanides (2004), our study

distinguishes between domestic and transnational terrorist incidents and does not rely solely on

transnational attacks. To accomplish this distinction, we use Enders, Sandler & Gaibulloev’s

(2010) division of the Global Terrorism Database (GTD) into the two types of terrorism. Our

study also accounts for cross-sectional dependence. The analysis includes six post-9/11 years,

during which transnational terrorists sought refuge in weak or failing African states. A

3

secondary purpose is to investigate the robustness of our estimates by adding a host of

macroeconomic variables (e.g., trade openness and government spending) to the baseline growth

models. These robustness checks are bolstered by alternative specifications of the terrorist

variable.

Why study the impact of terrorism on African growth? Previous studies on terrorism and

growth have either focused on the world (e.g., Blomberg, Hess & Orphanides, 2004; Tavares,

2004) or on Europe or Asia (Gaibulloev & Sandler, 2008, 2009, respectively). When examining

the underlying determinants of growth, Artadi & Sala-i-Martin (2003) showed that these

determinants differ greatly for Africa compared to the rest of the world. In particular, Africa

suffers from low openness, low primary school enrollment, high public spending as a share of

GDP, high population growth, and low investment as a share of GDP. These factors may make

African growth particularly prone to react adversely to violence in the form of terrorism and

wars. Worldwide panel growth estimates present an average picture that will not reflect Africa’s

growth response to terrorism if Africa reacts differently than other regions. Moreover, an

African dummy variable in a worldwide panel study does not quantify how Africa responds to

each of the determinants of growth, including the diverse forms of terrorism and conflicts.

In an earlier study, Blomberg, Hess & Orphanides (2004) showed that African growth,

indeed, reacted differently than other regions and the world at large to transnational terrorism

and other growth determinants. For example, Africa displayed the smallest response to

investment shares. These authors also found that Africa had a much larger growth reaction to

terrorism than the global or other regional samples. This large response for Africa seems out of

character for their 1968–2000 sample period, because Africa had relatively little terrorism

compared with other regions that showed less of a growth response to transnational terrorism

(see Blomberg, Hess & Orphanides, 2004, Tables 4 and 5). Africa has been plagued by terrorism

4

and internal conflicts over the last few decades; thus, it is of interest to discern the economic

ramifications of these hostile influences. Another justification for our African focus is that many

transnational terrorist groups (e.g., al-Qaida, Fatah, al-Jihad, Abdullah Azzam Brigades, Armed

Islamic Group, Islamic Union, Army for the Liberation of Rwanda, and Revolutionary United

Front) have active cells or infrastructure in Africa during the sample period. In particular, al-

Qaida has operated out of Sudan, Somalia, Liberia, Kenya, Tanzania, and Comoros (Lyman &

Morrison, 2004). Spectacular terrorist events in Africa include the “Black Hawk Down” incident

against US peacekeeping troops in Mogadishu, Somalia on 3–4 October 1993; the near-

simultaneous bombings of the US embassies in Nairobi, Kenya and Dar es Salaam, Tanzania on

7 August 1998; and the bombings of tourist sites in Sharm el-Sheikh, Egypt on 23 July 2005. In

recent years, failed and weak states, mineral wealth, and emerging Islamic extremism attracted

transnational terrorist groups that sought bases in Africa. Thus, the inclusion of post-9/11 years

is particularly attractive for our study. Africa is an important source of strategic resources

including oil. Given recent discoveries of rich offshore oil reserves in West Africa, Africa’s oil-

supplier status will grow (Lyman & Morrison, 2004). Consequently, terrorist attacks in Africa

have the potential to disrupt crucial supply lines to the industrial world. Our study assesses how

sensitive African economies are to terrorist-induced economic stress. A final ground for

focusing on Africa is that it contains many developing countries that receive foreign aid

(Hoeffler, 2002). Such countries’ economies are particularly sensitive to the harmful influence

of terrorism (Keefer & Loayza, 2008). It is essential to quantify the impact of terrorism on

growth if rich countries and multilateral institutions are to allocate aid properly and to attribute

reduced growth to its proper cause.

The paper is rich in findings. For the African sample, transnational terrorism (measured

in terms of incidents per million people) has a significant, but modest, negative consequence on

5

income per capita growth for 1970–2007. By contrast, domestic terrorism does not have a

significant effect on African income per capita growth. By separating the two forms of

terrorism, we establish that counterterrorism resources must be directed to curbing transnational

terrorism to promote African growth. These findings are robust to the inclusion of

macroeconomic and political variables. They are also robust to alternative forms of the terrorism

variables. As in other studies, internal and external conflicts have a negative influence on

African growth.

The remainder of the paper has four sections. The next section contains preliminaries

including definitions and literature review. The empirical methodology is indicated in the third

section, where the empirical specification and data are presented. Estimation and results are

given in the fourth section, followed by concluding remarks in the final section.

Preliminaries

Terrorism is the premeditated use or threat to use violence against noncombatants by subnational

groups or individuals in order to obtain a political or social objective through the intimidation of

a large audience beyond that of the immediate victims. Thus, bombings and armed attacks are

intended by terrorists to raise the anxiety level of citizens (i.e., the audience), so that they

pressure their government to grant the terrorists’ demands. This definition rules out state

terrorism but not state-sponsored terrorism, where a government clandestinely aids a terrorist

group through intelligence, funding, training, safe havens, or other means (Mickolus, 1989). A

key element in any terrorism definition is the motive for political or social change. Without this

motive, violent acts of bombings, kidnappings, or armed attacks are merely criminal activities for

extortion or sociopathic reasons. Our focus on noncombatant victims is to distinguish terrorism

from internal and external conflict, where guerrillas target government military forces.

6

A crucial distinction is between domestic and transnational terrorism. Through its

perpetrators, victims, audience, and influence, domestic terrorism solely involves the venue

country. The assassination of local government officials by a domestic terrorist group is a

domestic terrorist event. Since 1992, the Armed Islamic Group (GIA) has conducted a terrorist

campaign against Algerian civilian and government workers. Many, but not all, GIA attacks

have been domestic terrorism. In Cape Town, the People Against Gangsterism and Drugs

(PAGAD) has planted bombs primarily against domestic targets in their 1986–2000 terrorist

campaign. Generally, domestic terrorist attacks far outnumber transnational terrorist attacks (see

Appendix).

Transnational terrorist incidents have ramifications that extend beyond the venue country.

Terrorist events that start in one country and concludes in another (e.g., a skyjacking of a plane

in Egypt that is made to fly to Algeria) are transnational terrorist attacks. If the victims or

perpetrators in a terrorist incident are from countries other than the host or venue country, then

the terrorist attack is transnational. The Luxor massacre in November 1997 involved foreign

nationals from Japan, Switzerland, Germany, the United Kingdom, France, and elsewhere; thus,

this armed attack is a transnational terrorist incident.

As indicated earlier, income per capita growth can also be influenced by internal and

external conflicts. We use data on these conflicts drawn from a dataset, Major Episodes of

Political Violence (MEPV), 1946–2008, maintained by Marshall (2009) at the Center for

Systemic Peace, George Mason University. Intrastate or internal conflict (defined later) occurs

when a country is directly affected by civil violence, civil war, ethnic violence, or ethnic war at

home, orchestrated by an organized group. External conflicts are international violence or war

involving two or more countries.

We next turn to two essential neoclassical determinants of income per capita growth

7

(growth). The initial level of income per capita (y) is typically viewed as a positive influence on

growth owing to the notion of convergence, in which the rise in income per capita of a poorer

country outpaces that of a richer country (Barro, 1991; Barro & Sala-i-Martin, 1992).

Diminishing returns is behind convergence, in which countries have an easier time in adding

output where there are less initial inputs and output. Countries are assumed to possess identical

production functions and transition equations, but differ in their starting income per capita. This

latter assumption is more appropriate for a cohort of countries at similar stages of development

where production technologies are similar. This is more likely to hold for our African sample

than for a diverse global sample. A second key neoclassical influence on growth is investment

share (I/GDP). Larger investment shares result in greater capital accumulation, which fosters

growth through capital and embodied technological change.

At a later point, we also introduce some macroeconomic variables that may affect

growth. Often, openness – the ratio of the sum of exports and imports to GDP – is viewed as a

positive determinant on growth (Blomberg, Hess & Orphanides, 2004). Openness may foster

growth as greater exports increase aggregate demand and more imports enhance resources and

technology transfers, particularly in the case of developing countries.

Political violence in various forms are anticipated to limit income per capita growth

(Barro, 1991; Blomberg, Hess & Orphanides, 2004; Murdoch and Sandler, 2002b, 2004). In the

introduction, we indicated numerous grounds why transnational and domestic terrorism are

anticipated to reduce growth. There are grounds for anticipating that transnational terrorism will

be more detrimental than domestic terrorism on growth. When a terrorist event kills a foreigner

or destroys foreign property, there is a greater chance that this will negatively impact foreign

assistance and/or foreign direct investment (a key source of savings), which then reduces growth.

Terrorists deliberately target foreign assets to affect the venue country’s economy adversely.

8

Often, foreign targets include those of donor and investor nations. Attacks against foreign

tourists may have particularly large economic impacts (Enders, Sandler & Parise, 1992). Past

studies have sustained these hypothesized decreases. Blomberg, Hess & Orphanides (2004) and

Tavares (2004) found that each year of transnational terrorism limited economic growth for a

global sample by 0.048% and 0.038%, respectively. Gaibulloev & Sandler (2009) found in a

study of Asia that an additional transnational terrorist incident per million persons reduced

economic growth by about 1.4 percentage points annually. In Asia, the mean number of

incidents per million persons was just over 0.05, so that the average anticipated reduction in

growth is slightly greater than the amounts displayed by Blomberg, Hess & Orphanides (2004)

and Tavares (2004) for their global samples. By splitting the Asian sample into developed and

developing countries, Gaibulloev & Sandler (2009) showed that the developed countries did not

display any adverse growth consequences from terrorism. Only the developing Asian economies

were negatively impacted by terrorism. Since Africa contains mostly developing countries, this

finding bodes badly for Africa.

Blomberg, Hess & Orphanides (2004), Murdoch & Sandler (2002b, 2004), and others

have shown that internal and external conflicts resulted in reduced income per capita growth.

These conflicts harmed growth by diverting funds to security spending, raising uncertainty,

increasing diseases, reducing trade, eliminating human capital, and causing capital flight. For

Asia, Gaibulloev & Sandler (2009) established that internal and external conflicts had a much

greater harmful impact on growth then transnational terrorism. Blomberg, Hess & Orphanides

(2004) also uncovered that conflicts decreased growth more than transnational terrorism for their

worldwide sample.

Methodology

9

Empirical specifications

The baseline model is a modification of the neoclassical growth model with the addition of

terrorism and conflict variables:

( )0 1 1 2 3 4 51ln /it it it it itit

growth y I GDP terror external internalβ β β β β β− −= + + + + +

i t itα η ε+ + + , (1)

where subscript i = 1, …, N indicates the country and subscript t = 1, …, T indexes the time

period. The convergence term in county i is captured by the natural logarithm of lagged income

per capita, 1ln ity − . Initially, the explanatory variable terror is the number of terrorist attacks per

million persons; alternative measures for terror are introduced later. The internal and external

terms are dummy variables for internal and external conflicts, respectively. The βs are

regression coefficients; iα is the country-specific fixed effect; tη is the time-specific effect; and

itε is the random error term.

Our empirical strategy consists of the following steps. First, we implement the fixed-

effects estimator on the baseline model, Equation (1), for the sample of African countries. We

use two separate measures of terrorism: transnational terrorist attacks per million persons and

domestic terrorist attacks per million persons. Second, we check the robustness of the baseline

results by including a host of possible determinants of growth mentioned in the literature. Third,

we investigate the sensitivity of our results to various specifications of the terrorism variables.

One of the concerns in cross-country panel analysis is that the error terms are likely to be

correlated across countries. This may arise from observable or unobservable common factors

(e.g., shocks), omitted from the model. Cross-sectional dependence (CD) results in biased

standard errors, which influence statistical inference. Inclusion of time dummy variables would

address the issue only if the spatial correlations are equal for every pair of countries, which is a

10

strong assumption. We implement Driscoll & Kraay’s (1998) nonparametric estimator of the

standard error, consistent with the general form of spatial dependence, autocorrelation, and

heteroskedasticity. Pesaran’s (2004) CD test is applied to search for cross-sectional

dependence.2 We also check the results using Huber/White robust standard errors, and robust

standard errors for multi-way (two-way in our case) clustering proposed by Cameron, Gelbach &

Miller (2009).

Data

We construct an unbalanced panel dataset of 51 African countries for 1970–2007. Table IA in

the Appendix summarizes raw data and sources. Our terrorism data are drawn from two sources.

Data on transnational terrorist incidents for 1970–2007 are taken from International Terrorism:

Attributes of Terrorist Events (ITERATE) (Mickolus et al., 2008). ITERATE relies on print and

electronic media for observations of terrorist events, including the event’s date and country

location. For each sample African country, we know the annual number of transnational terrorist

incidents.3 This yearly total is divided by the country’s population to give our continuous

measure of transnational terrorist incidents per million persons ( )_ _ ittrans terror iter . We favor

a continuous measure of terrorism that is normalized by the population, because more populous

countries appear better able to absorb a terrorist attack without displaying economic

ramifications (Blomberg, Hess & Orphanides, 2004; Gaibulloev & Sandler, 2009). We examine

alternative measures for robustness check.

Data for both domestic and transnational terrorist attacks are available in the Global

Terrorism Dataset (GTD), compiled by the National Consortium for the Study of Terrorism and

Responses to Terrorism (START) (2009) at the University of Maryland. However, domestic and

transnational terrorist events are not distinguished within GTD. Hence, our second source for

11

terrorism data is Enders, Sandler & Gaibulloev (2010), who engineered a procedure to

dichotomize GTD data into transnational and domestic events.

Figure 1 displays the annual number of transnational terrorist events in Africa: the solid

time series depicts this number for ITERATE, while the broken time series depicts this number

for GTD. The data for 1993 are missing from GTD; thus, we interpolate 1993 values as the

average of the 1992 and 1994 values.4 For 1970–1979 and 1981–1988, the number of ITERATE

events consistently exceeds the number of transnational events in GTD. The average numbers of

incidents in ITERATE are about 15 and 41 for the two periods, while the average numbers of

transnational incidents in GTD are about 5.4 and 16 for the two periods, respectively. Similarly,

for 1990 and 1999–2000, the number of terrorist incidents in ITERATE are much greater than

the number of transnational terrorist incidents in GTD. After 2006, GTD reports a sharp increase

in the number of transnational terrorist incidents. For other years, the two series generally track

each other well.

[Figure 1 near here]

More detailed information on terrorist events by country and by African regions is

reported in Table IIA in the Appendix. For example, based on the total number of transnational

terrorist incidents over the sample period, Algeria, Angola, Ethiopia, and Somalia are among the

top five countries, based on ITERATE and GTD data. In terms of domestic terrorist events,

Algeria, Angola, Burundi, Egypt, and South Africa are the most terrorism-plagued countries in

Africa (Table IIA).

Given that ITERATE uses a consistent coding approach, Enders, Sandler & Gaibulloev

(2010) devised a calibration procedure for the GTD data. In short, these authors first calibrated

transnational terrorist events in GTD using these consistently coded events in ITERATE. The

calibration was based on a ratio of means. Next, they calibrated GTD’s domestic terrorist

12

incidents by the same ratios of means for the requisite periods. For the world sample, they

inflated the number of pre-1977:2 GTD incidents (domestic and transnational) by a factor of 2.06

and deflated the number of GTD incidents for 1991:2–1997:4 by a scaling factor of 0.52. (The

numbers after the colon denote quarters.) However, we cannot use these scaling factors and

periods for Africa, because Figure 1 indicates that the terrorism data for the Africa sample

behave differently than the data for the world sample. Following the approach suggested by

Enders, Sandler & Gaibulloev (2010), we calibrate GTD data using scale factors specific to

Africa. In particular, we scale up the number of GTD incidents prior to 1980 by a factor of 2.80

(= 15/5.4), the ratio of the mean number of ITERATE incidents to the mean number of GTD

transnational incidents for this period. Similarly, we inflate the number of GTD incidents for

1981–1988 by a scale factor of 2.52; for 1990 by a scale factor of 1.67; and for 1999–2000 by a

scale factor of 2.63. We use these scaling factors to modify GTD domestic incidents as well.

[Figure 2 near here]

Figure 2 shows that the calibrated GTD transnational incidents (broken line) track

ITERATE incidents (solid line) reasonably well. The dotted series is a plot of GTD data,

adjusted by the scaling factors suggested by Enders, Sandler & Gaibulloev (2010). As expected,

this modified data does not track ITERATE well. Once again, we divide the annual number of

transnational and domestic terrorist events in each country by its population to come up with the

number of transnational incidents per million persons ( )_ _ ittrans terror gtd and the number of

domestic incidents per million persons ( )_ _ itdom terror gtd .

Information on conflicts is obtained from Major Episodes of Political Violence (MEPV)

database (Marshall, 2009). MEPV defines major episodes as “…the systematic and sustained

use of lethal violence by organized groups that result in at least 500 directly-related deaths over

13

the course of the episode.” Episodes fulfilling the casualty threshold are classified into seven

armed conflict types: international violence, international war, international independence war,

civil violence, civil war, ethnic violence, and ethnic war. Our internal conflict variable

( )itinternal is a dummy, which is 1 if a country is directly affected by civil violence, civil war,

ethnic violence, or ethnic war during a given year, and is 0 otherwise. The external conflict

dummy ( )itexternal is 1 if a sample country is directly engaged in either an international

violence or international war, and is 0 otherwise. There are many more countries with internal

conflicts (more than ten countries) each year than those with external conflicts. Except for a

couple of periods, such as the second half of the 1970s and the second half of the 1990s, there

are few to no external conflicts recorded. The polity variable ( itpolity ) comes from Polity IV

dataset (Marshall & Jaggers, 2009). The variable reflects three key inter-related elements:

opportunities for political participation, constraints on executive power, and government support

and protection of civil liberties. This index ranges between –10 (strongly autocratic) and +10

(strongly democratic).

The main source of information for gross domestic product (GDP) is World Bank

(2009).5 We obtain GDP per capita at purchasing power parity (PPP) in current international

dollars for 2000 and extend the series forwards and backwards using the real growth rate of GDP

per capita in local currency units (LCU). The resulted series are GDP per capita based on PPP at

constant 2000 prices. Our variable 1ln ity − is the log of PPP-based GDP per capita at constant

2000 prices, lagged by one year. The economic growth ( )itgrowth variable is the growth rate of

PPP-based GDP per capita in constant 2000 prices.

Data on population is drawn from World Bank (2009). For each country-year, we

compute the growth of population, itpopulation growth , as our variable. We obtain data on gross

14

capital formation as a percentage of GDP, government consumption as a percentage of GDP, and

export and import as a percentage of GDP from United Nations Statistics Division (UNSD)

(2008). Using these data, we compute the investment as a share of GDP lagged by one year, the

government consumption as a share of GDP lagged by one year, and the trade (export plus

import) as a share of GDP lagged by one year. We denote these variables as 1/ itI GDP − ,

1/ itG GDP − , and 1/ itTRADE GDP − , respectively.

Estimation and results

The fixed-effects estimates of our baseline model are presented in Table I. Model 1 consists of

lagged per capita income, investment share in GDP, number of ITERATE transnational terrorist

incidents per million persons, and internal and external conflict dummy variables. Models 2 and

3 repeat the analysis of Model 1 by replacing the ITERATE terrorism variable with the number

of GTD transnational terrorist incidents per million persons and the number of GTD domestic

terrorist incidents per million persons, respectively. Finally, Models 4 includes both GTD

terrorism variables (i.e., transnational and domestic).

[Table I near here]

The Pesaran’s CD test for cross-sectional dependence is marginally significant (except

for Model 3) when the time dummy variables are included; without these time dummies, the CD

test is highly significant. This suggests that the time dummy variables eased the cross-sectional

(spatial) correlation issue, but concerns remain. Therefore, we estimate our models using

Driscoll and Kraay’s standard errors, robust to general forms of heteroskedasticity,

autocorrelation, and cross-sectional dependence. We note that the results generally hold if we

replace Driscoll and Kraay’s standard errors with Huber/White robust standard errors.

15

Lagged GDP per capita has a negative impact, consistent with convergence, while

investment share has a positive influence on the growth of per capita income, consistent with

growth theory. Both coefficients are robust to alternative specifications. With regard to political

violence variables, all coefficients are negative and statistically significant except for the

domestic terrorism variable, which is not statistically significant. The estimated impact of

transnational terrorism is about –0.01 using ITERATE and –0.02 using GTD, which means that,

on average, an extra transnational terrorist incident per million persons reduces economic growth

by about 1 to 2% in a given year.6 At first, these growth effects seem large if one considers the

small and, sometimes, negative annual growth rate of African economies during the study period.

The impacts are, however, in terms of per million persons: a country with a population of 50

million would have to experience 50 transnational terrorist incidents in a given year to have its

income per capita growth rate decline by 1%. For our sample countries, the average annual

number of transnational terrorist events per million persons is about 0.1. Thus, on average, the

negative marginal impact of transnational terrorism in our African sample is about 0.1

percentage point of income per capita growth, which is small and consistent with, but somewhat

larger than, that of the literature (Blomberg, Hess & Orphanides, 2004; Tavares, 2004). The

coefficients of internal and external conflicts are about –0.02, indicating that a conflict reduces

income per capita growth by approximately 2% in a year.

[Table II near here]

There are a number of possible or additional determinants of economic growth, discussed

in literature. For example, Artadi & Sala-i-Martin (2003) showed that, among the other factors,

the degree of economic openness, government spending, and political and economic institutions

are important determinants of economic growth in Africa. We investigate the robustness of the

above terrorism results by including a host of additional macroeconomic and political variables

16

to the growth model in Table II. The CD test statistics suggest the presence of cross-sectional

dependence, which we take into account by estimating Driscoll and Kraay’s standard errors. As

seen in Table II, the terrorism variables are robust to the inclusion of additional variables in

terms of their statistical significance, sign, and magnitude; the impact of transnational terrorism

is negative and statistically significant, whereas the negative effect of domestic terrorism is not

significant. Over the sample period, there are 1,005 transnational terrorist incidents and 4,610

domestic incidents recorded in Africa based on GTD data. Transnational terrorist attacks killed

around 2,956 people or around 3 persons per incident and wounded 6,222 people or about 6

persons per incident, while domestic attacks killed 18,742 people or about 4 persons per incident

and wounded 15,375 people or about 3.3 persons per incident. In terms of human life, domestic

terrorism is as deadly or damaging as transnational terrorism. However, transnational and

domestic terrorist incidents differ, among other things, because the former targets foreign

citizens, foreign businesses (personnel and assets), and international institutions. Consequently,

transnational terrorist attacks can adversely influence foreign direct investment, aid flows, and

trade, which may partly explain the different impact on growth between two types of incidents.

Internal and external conflicts remain negative and statistically significant in all models.

With regard to new variables, consistent with Artadi & Sala-i-Martin (2003), the share of

government expenditure in GDP and openness (measured by the share of trade in GDP) are

associated with negative and positive growth of GDP per capita, respectively. The negative

influence of the government expenditure share is consistent with crowding out of more

productive uses of these resources, while the positive influence of openness is consistent with

trade-induced technological transfers and the stimulation of expenditures. The impacts of

population growth and degree of democracy (polity) are not statistically significant.

The results in Table II generally hold if we replace Driscoll and Kraay’s standard errors

17

with Huber/White robust standard errors. Cameron, Gelbach & Miller (2009) proposed a general

estimator of standard errors that allows for multi-way clustering. Using their estimator, we

cluster at the country and year level to allow for both cross-sectional and time series dependence.

Based on the cluster-robust standard errors, most of the results of Table II hold, except that the

ITERATE transnational terrorism and external conflict variables are not significant.7

[Table III near here]

Next, we implement a number of additional robustness checks – see Table III. To save

space, we only report estimates of the terrorism variables. As discussed in the previous section,

our findings are based on calibrated GTD data. First, we check our results by re-estimating the

models of Table II using the original (not adjusted) GTD data. Country-level statistics of

terrorist events for 1993 are obtained from Appendix I of the GTD Codebook. Our results, save

for a change in magnitude of the GTD transnational terrorism coefficient, are remarkably robust.

Second, there is no satisfactory answer as how to measure terrorism, given available data. We

re-estimate our models using two alternative measures of terrorism from the literature: number of

terrorist incidents (not normalized by population) and a dummy variable for terrorism (i.e.,

whether or not there was terrorism in a given year for a country). For both cases, our results

generally hold in terms of sign and statistical significance. The only notable difference is that

domestic terrorism is significant in Model 3 when we apply a dummy. Finally, we use one-year

lagged value of the number of terrorist incidents per million persons. The estimate of ITERATE

transnational terrorist incidents is not significant, while the estimate of GTD transnational

terrorist incidents remains negative and statistically significant.

Concluding remarks

Both terrorism and conflict (internal and external) adversely affect income per capita growth in

18

Africa. Past studies have shown that Africa responds differently than the rest of the world to

growth factors (Artadi & Sala-i-Martin, 2003; Blomberg, Hess & Orphanides, 2004); hence,

there are grounds for our analysis to quantify these differences for an up-to-date sample that

accounts for terrorism and other growth factors in Africa. By distinguishing between the two

types of terrorism, we show that only transnational terrorism has a significant negative influence

on growth in Africa for our panel estimates. Domestic terrorism does not have a significant

adverse impact on African growth. Our results are robust not only to the inclusion of other

economic and political variables (e.g., government spending share, trade openness, democracy

index, and population growth), but also to alternative specifications of the terrorism variable.

Because Africa contains so many developing countries and weak democracies, insights

from the literature suggest that African growth may be especially hard hit by terrorism (see, e.g.,

Gaibulloev & Sandler, 2009; Keefer & Loayza, 2008). This worry is compounded by the

increasing presence of fundamentalist terrorists in Africa (Lyman & Morrison, 2004) and the

transference of terrorist attacks to Africa over the last decade or so (Enders & Sandler, 2006).

Though the effect of terrorism on growth is significant, it is surprisingly modest when one

considers the average population levels. This is an important consideration because our primary

terrorism variable is the number of incidents per million persons. Most sample countries have

relatively few terrorist incidents per year and large populations. The average annual number of

transnational terrorist incidents per million persons is approximately just 0.1 for our sample.

Hence, an average sample country sustains a very small annual reduction in its income per capita

growth of 0.1%, because it must experience an additional incident per million persons for growth

to fall by 1% or 2% in the case of transnational terrorism. This is encouraging news. The

absence of a significant domestic terrorism impact on African growth is also good news insofar

as Africa is plagued by much more domestic than transnational terrorism. When trying to limit

19

negative growth effects of terrorism, this study shows that counterterrorism resources should be

directed to transnational, rather than domestic, terrorism.

The transference of transnational terrorist attacks to Africa in recent years is largely due

to enhanced border security in the United States, Europe, and other wealthy countries. Often,

US, Israel, and European assets are targeted in Africa-based transnational terrorist attacks

(Enders & Sandler, 2006). Thus, these rich countries have an obligation to help African

countries bolster their counterterrorism responses if asked. Since the same terrorist groups – e.g.,

al-Qaida – may target more than one African country, there is a need for a coordinated African

response in such instances.

When judging the effectiveness of their aid for development, donor countries must take

into account that terrorism will curb recipient countries’ growth. This reduced performance

should not always be attributed to other nefarious causes such as corruption or bad policies. The

results of this paper allow for a calculation of these generally modest growth losses.

20

Footnotes

1. On the effects of terrorism on economies, see Keefer & Loayza (2008) and Sandler &

Enders (2008).

2. The Driscoll and Kraay’s standard errors are computed using the Stata program of

Hoechle (2007), which can handle unbalanced panels. Pesaran’s CD test is performed using the

Stata program of De Hoyos & Sarafidis (2008).

3. In ITERATE, both starting location and ending location of an incident are reported,

which causes a problem if the starting and ending locations differ (e.g., letter bombing or

skyjacking). We had 137 such events in our sample. After reading the history of these events,

we excluded 2 events, and assigned 93 events to location start and 42 events to location end.

4. Country-level estimates of terrorist events for 1993 are available in Appendix I of the

GTD Codebook, but these estimates likely underestimate the number of incidents. For example,

the number of GTD transnational incidents in Africa declines sharply from 105 in 1992 to 34 in

1993, and then increases rapidly to 150 incidents in 1994. By contrast, the number of ITERATE

transnational incidents in Africa increases steadily; 47 in 1992; 98 in 1993; and 129 in 1994. We

check our results using GTD values from Appendix 1 (see next section).

5. World Bank (2009) does not have PPP-based GDP data for Zimbabwe, nor does it

have any GDP data for Somalia. Information on GDP at constant LCU is also missing for some

country-years. We use GDP at constant LCU from the United Nations Statistics Division

(UNSD) (2008) to fill in the missing observations. For Zimbabwe, PPP-based GDP per capita at

current prices for 2000 is obtained from IMF (2009). For Somalia, the real GDP per capita at

PPP is obtained from Penn World Table Version 6.2 (Heston, Summers & Aten, 2006) for 1970–

2004 and extrapolated to 2005–2007 using per capita GDP growth at LCU from UNSD (2008).

The correlation between the growth rates of real GDP per capita at LCU from World Bank and

21

UNSD is about 0.9. The correlation between GDP per capita (PPP) from World Bank and IMF

is above 0.9, while this correlation is about 0.64 for Penn World Table and World Bank data.

6. Gaibulloev & Sandler (2008) found that the effect of transnational terrorist events on

growth is larger than that of aggregate terrorist events for Western Europe. In addition, they

showed that the influence of transnational terrorism on income per capita growth is greater than

that of domestic terrorism.

7. These results are available from the authors upon request.

22

References

Abadie, Alberto & Javier Gardeazabal (2008) Terrorism and the world economy. European

Economic Review 52(1): 1–27.

Artadi, Elsa V. & Xavier Sala-i-Martin (2003) The economic tragedy of the XXth century:

growth in Africa. NBER Working Paper Series, National Bureau of Economic Research,

Cambridge, MA.

Barro, Robert J (1991) Economic growth in a cross section of countries. Quarterly Journal of

Economics 100(2): 223–251.

Barro, Robert J. & Xavier Sala-i-Martin (1992) Convergence. Journal of Political Economy

100(2): 223–251.

Blomberg, S. Brock; Gregory D. Hess & Athanasios Orphanides (2004) The macroeconomic

consequences of terrorism. Journal of Monetary Economics 51(5): 1007–1032.

Cameron, A. Colin; Jonah B. Gelbach & Douglas L. Miller (2009) Robust inference with multi-

way clustering. Working Paper #09–9, Department of Economics, University of

California at Davis.

Collier, Paul & Anke Hoeffler (2002) On the incidence of civil war in Africa. Journal of

Conflict Resolution 46 (1): 13–28.

De Hoyos, Rafael E. & Vasilis Sarafidis (2006) Testing for cross-sectional dependence in panel-

data models. Stata Journal 6(4): 482–496.

Drakos, Konstantinos (2004) Terrorism-induced structural shifts in financial risk: Airline stocks

in the aftermath of the September 11th terror attacks. European Journal of Political

Economy 20(2): 436–446.

Drakos, Konstantinos & Ali M. Kutan (2003) Regional effects of terrorism on tourism in three

Mediterranean countries. Journal of Conflict Resolution 47(5): 621–641.

23

Driscoll, John C. & Aart C. Kraay (1998) Consistent covariance matrix estimation with spatially

dependent panel data. Review of Economics and Statistics 80(4): 549–560.

Enders, Walter; Adolfo Sachsida & Todd Sandler (2006) The impact of transnational terrorism

on US foreign direct investment. Political Research Quarterly 59(4): 517–531.

Enders, Walter & Todd Sandler (1996) Terrorism and foreign direct investment in Spain and

Greece. Kyklos 49(3): 331–352.

Enders, Walter & Todd Sandler (2006) Distribution of transnational terrorism among countries

by income class and geography after 9/11. International Studies Quarterly 50(2): 367–

393.

Enders, Walter; Todd Sandler & Khusrav Gaibulloev, (2010) Domestic versus transnational

terrorism: Data, decomposition, and dynamics. Unpublished manuscript, Center for

Global Collective Action, University of Texas at Dallas.

Enders, Walter; Todd Sandler & Gerald F. Parise (1992) An econometric analysis of the impact

of terrorism on tourism. Kyklos 45(4): 531–554.

Gaibulloev, Khusrav & Todd Sandler (2008) Growth consequences of terrorism in Western

Europe. Kyklos 61(3): 411–424.

Gaibulloev, Khusrav & Todd Sandler (2009) The impact of terrorism and conflicts on growth in

Asia. Economics and Politics 21(3): 359–383.

Heston, Alan; Robert Summers & Bettina Aten (2006) Penn World Table Version 6.2.

Philadelphia, PA: Center for International Comparisons of Production, Income and Prices

at the University of Pennsylvania.

Hoechle, Daniel (2007) Robust standard errors for panel regressions with cross-sectional

dependence. Stata Journal 7(3): 281–312.

Hoeffler, Anke (2002) The augmented Solow model and the African growth. Oxford Bulletin of

24

Economics and Statistics 64(2): 135–158.

International Monetary Fund (IMF) (2009) World Economic Outlook (WEO) Database, April

(http://www.imf.org). Accessed 1 July 2009.

Ito, Harumi & Darin Lee (2005) Assessing the impact of the September 11th terrorist attacks on

US airline demand. Journal of Economics and Business 57(1): 75–95.

Keefer, Philip & Norman Loayza (eds.) (2008) Terrorism, Economic Development, and Political

Openness. Cambridge: Cambridge University Press.

Lyman, Princeton N. & J. Stephen Morrison (2004) The terrorist threat in Africa. Foreign

Affairs 83(1): 75–86.

Marshall, Monty G., 2009. Major Episodes of Political Violence (MEPV), 1946–2008. Fairfax,

VA: Center for Systemic Peace, George Mason University

(http://www.systemicpeace.org). Accessed 1 July 2009.

Marshall, Monty G. & Keith Jaggers, 2009. Polity IV Dataset Version 2007 and Dataset Users’

Manual. Fairfax, VA: Center for Systemic Peace and the Center for Global Policy,

George Mason University (http://www.systemicpeace.org). Accessed 1 July 2009.

Mickolus, Edward F (1989) What constitutes state support to terrorists? Terrorism and Political

Violence 1(3): 287–293.

Mickolus, Edward F.; Todd Sandler, Jean M. Murdock & Peter Flemming (2008) International

Terrorism Attributes of Terrorism Events (ITERATE), 1968–2007. Dunn Loring, VA:

Vinyard Software.

Mickolus, Edward F. & Susan L. Simmons (2002) Terrorism, 1996–2001: A Chronology, 2 vols

Westport, CT: Greenwood Press.

Mickolus, Edward F. & Susan L. Simmons (2006) Terrorism, 2002–2004: A Chronology.

Westport, Ct: Praeger Security International.

25

Murdoch, James C. & Todd Sandler (2002a) Civil wars and economic growth: A regional

comparison. Defence and Peace Economics 13(6): 451–464.

Murdoch, James C. & Todd Sandler (2002b) Economic growth, civil wars, and spatial

spillovers. Journal of Conflict Resolution 46(1): 91–110.

Murdoch, James C. & Todd Sandler (2004) Civil wars and economic growth: Spatial spillovers.

American Journal of Political Science 48(1): 138–151.

National Consortium for the Study of Terrorism and Responses to Terrorism (START) (2009)

Global Terrorism Database, CD-Rom. College Park, MD: University of Maryland.

Pesaran, M. Hashem (2004) General diagnostic tests for cross section dependence in panels.

Cambridge Working Papers in Economics No. 0435, Faculty of Economics, University of

Cambridge.

Tavares, Jose (2004) The open society assesses its enemies: Shocks, disasters, and terrorist

attacks. Journal of Monetary Economics 51(5): 1039–1070.

United Nations Statistics Division (UNSD) (2008) National Accounts Main Aggregates

Database (http://unstats.un.org/unsd/snaama/Introduction.asp). Accessed 1 July 2009.

World Bank (2009) World Development Indicators (WDI) (http://www.worldbank.org/data).

Accessed 1 July 2009.

0

20

40

60

80

100

120

140

160

1970 1974 1978 1982 1986 1990 1994 1998 2002 2006

Figure 1. Annual Number of Transnational Terrorist Incidents, 1970-2007

Num

ber

of in

cide

nts

ITERATE

GTD_trans

0

20

40

60

80

100

120

140

160

1970 1974 1978 1982 1986 1990 1994 1998 2002 2006

Figure 2. Annual Number of Transnational Terrorist Incidents Using Adjusted GTD Data

Num

ber

of in

cide

nts

ITERATE

GTD_trans1

GTD_trans2

Table I. Fixed-effects estimation of growth model with cross-sectional dependence, 1970–2007

Model 1 Model 2 Model 3 Model 4 ln yit–1 –0.042*** –0.040*** –0.041*** –0.040*** (0.013) (0.013) (0.013) (0.013) I/GDPit–1 0.161*** 0.161*** 0.158*** 0.161*** (0.032) (0.031) (0.031) (0.031) trans_terror_iterit –0.013** (0.006) trans_terror_gtdit –0.020*** –0.020*** (0.005) (0.005) dom_terror_gtdit –0.002 –0.0004 (0.002) (0.0013) internalit –0.016*** –0.013*** –0.016*** –0.013*** (0.003) (0.003) (0.003) (0.003) externalit –0.018** –0.019** –0.018** –0.019** (0.008) (0.008) (0.008) (0.008) Sample size 1751 1751 1751 1751 R-squared 0.14 0.15 0.13 0.15 CD test –1.72 –1.75 –1.43 –1.76 p-value 0. 085 0.080 0.154 0.078 Notes: CD test is Pesaran’s (2004) test of cross-sectional independence and p-value is the probability value of the null. Constant and time dummies are suppressed. Driscoll and Kraay’s standard errors are in parentheses. These standard errors are robust to general forms of heteroskedasticity, autocorrelation, and cross-sectional (spatial) dependence. R-squared (within) is computed after removing country effects. Significance levels: *** is .01, ** is .05, and * is .10.

Table II. Robustness of estimates to inclusion of additional variables, 1970–2007

Model 1 Model 2 Model 3 Model 4 ln yit–1 –0.051*** –0.049*** –0.051*** –0.049*** (0.012) (0.012) (0.012) (0.012) I/GDPit–1 0.082** 0.084** 0.079** 0.084** (0.035) (0.033) (0.036) (0.033) G/GDPit–1 –0.105*** –0.101*** –0.103*** –0.101*** (0.036) (0.035) (0.036) (0.036) TRADE/GDPit–1 0.070*** 0.068*** 0.071*** 0.068*** (0.014) (0.013) (0.015) (0.013) trans_terror_iterit –0.013** (0.005) trans_terror_gtdit –0.019*** –0.019*** (0.004) (0.004) dom_terror_gtdit –0.002 –0.001 (0.002) (0.001) internalit –0.011*** –0.009*** –0.012*** –0.009*** (0.003) (0.003) (0.003) (0.003) externalit –0.015** –0.017** –0.016** –0.017** (0.007) (0.007) (0.007) (0.007) polityit –0.0002 –0.0002 –0.0002 –0.0002 (0.0003) (0.0004) (0.0003) (0.0003) population growthit 0.478 0.466 0.473 0.465 (0.357) (0.343) (0.372) (0.342) Sample size 1744 1744 1744 1744 R-squared 0.18 0.18 0.17 0.18 CD test –2.33 –2.31 –2.11 –2.32 p-value 0.020 0.021 0.035 0.021 Notes: CD test is Pesaran’s (2004) test of cross-sectional independence and p-value is the probability value of the null. Constant and time dummies are suppressed. Driscoll and Kraay’s standard errors are in parentheses. These standard errors are robust to general forms of heteroskedasticity, autocorrelation, and cross-sectional (spatial) dependence. R-squared (within) is computed after removing country effects. Significance levels: *** is .01, ** is .05, and * is .10.

Table III. Re-estimation of Table II by using unadjusted GTD data and alternative specifications of terrorism variables

Model 1 Model 2 Model 3 Model 4 Using unadjusted GTD data

trans_terror_iterit –0.013** (0.005) trans_terror_gtdit –0.033*** –0.032*** (0.005) (0.006) dom_terror_gtdit –0.004 –0.001 (0.003) (0.002)

Using number of terrorist incidents trans_terror_iterit –0.002** (0.001) trans_terror_gtdit –0.001*** –0.002*** (0.0005) (0.001) dom_terror_gtdit 0.00003 0.00008 (0.00003) (0.00005)

Using dummy variable for terrorism trans_terror_iterit –0.011** (0.005) trans_terror_gtdit –0.012** –0.011** (0.004) (0.005) dom_terror_gtdit –0.008** –0.005 (0.004) (0.004)

Using lagged value of the number of terrorist incidents per million population trans_terror_iterit-1 –0.002 (0.003) trans_terror_gtdit-1 –0.009** –0.009*** (0.004) (0.003) dom_terror_gtdit-1 –0.001 –0.001 (0.001) (0.001) Notes: See Models in Table II for specification. Driscoll and Kraay’s standard errors are in parentheses. These standard errors are robust to general forms of heteroskedasticity, autocorrelation, and cross-sectional (spatial) dependence. Significance levels: *** is .01, ** is .05, and * is .10.

APPENDIX Table IA. Raw data and sources Data Sources GDP per capita (PPP at current international dollars, LCU at constant prices)

World Bank (2009), IMF(2009), Heston, Summers & Aten (2006), UNSD (2008)

Investment share in GDP UNSD (2008) Government spending share in GDP UNSD (2008) Import share in GDP UNSD (2008) Export share in GDP UNSD (2008) Transnational terrorist incidents (ITERATE) Mickolus et al. (2008) Transnational terrorist incidents and domestic terrorist incidents (GTD)

Enders, Sandler & Gaibulloev (2010), START (2009)

Internal conflicts (MEPV) Marshall (2009) External conflicts (MEPV) Marshall (2009) Polity2 (Polity IV dataset) Marshall and Jaggers (2009) Population World Bank (2009)

Table IIA. Number of terrorist incidents by country and ranking of countries in terms of the number of terrorist incidents for 1970–2007

Country ITERATE GTD transnational GTD domestic Number Rank Number Rank Number Rank

East region 484 1 450 1 929 3 Burundi 21 18 28 12 184 5 Comoros 0 45 0 42 5 38 Djibouti 7 27 4 32 10 30 Eritrea 4 34 2 37 0 49 Ethiopia 82 4 66 4 47 16 Kenya 21 18 24 13 80 9 Madagascar 1 42 6 29 15 24 Malawi 0 45 0 42 5 38 Mauritius 0 45 0 42 1 47 Mozambique 70 7 40 10 141 7 Rwanda 7 27 33 11 80 9 Seychelles 1 42 0 42 0 49 Somalia 160 2 159 1 117 8 Tanzania 6 31 7 26 1 47 Uganda 34 11 53 6 166 6 Zambia 24 16 9 23 32 19 Zimbabwe 46 9 22 15 48 15

Middle region 133 4 141 4 401 4 Angola 82 4 84 3 310 4 Cameroon 3 37 4 32 12 27 Central African Rep. 3 37 5 30 5 38 Chad 11 22 13 19 15 24 Congo, Dem. Rep. 12 21 24 13 44 17 Congo, Rep. 15 20 12 22 11 28 Equatorial Guinea 2 40 0 42 2 46 Gabon 5 32 1 39 4 42

North region 389 2 262 2 1836 1 Algeria 93 3 142 2 1260 2 Egypt, Arab Rep. 166 1 50 7 500 3 Morocco 31 12 18 18 10 30 Sudan 76 6 45 9 60 12 Tunisia 23 17 8 25 7 35

South region 82 5 63 5 1489 2 Botswana 7 27 0 42 6 37 Lesotho 8 24 7 26 17 22 Namibia 10 23 4 32 14 26 South Africa 31 12 50 7 1445 1 Swaziland 26 14 2 37 9 33

West region 176 3 165 3 333 5 Benin 0 45 3 36 7 35 Burkina Faso 2 40 0 42 3 44 Cape Verde 0 45 0 42 0 49

Cote d'Ivoire 8 24 22 15 11 28 Gambia, The 0 45 0 42 4 42 Ghana 4 34 1 40 17 22 Guinea 3 37 1 40 9 33 Guinea-Bissau 0 45 0 42 5 38 Liberia 25 15 13 19 10 30 Mali 1 42 5 30 24 20 Mauritania 5 32 4 32 3 44 Niger 7 27 13 19 22 21 Nigeria 69 8 66 4 72 11 Senegal 4 34 7 26 57 13 Sierra Leone 40 10 22 15 41 18 Togo 8 24 9 23 51 14 Note: GTD does not have data for Cape Verde.