Socio-Economic Planning Sciences xxx (xxxx) xxx–xxx

Contents lists available at ScienceDirect

Socio-Economic Planning Sciences

journal homepage: www.elsevier.com/locate/seps

Youth unemployment and terrorism in the MENAP (Middle East, North

Africa, Afghanistan, and Pakistan) region
Aniruddha Bagchi, Jomon A. Paul∗
Department of Economics, Finance, & Quantitative Analysis, Kennesaw State University, 560 Parliament Garden Way, Kennesaw, GA 30144, USA

A R T I C L E I N F O A B S T R A C T

Keywords: This paper builds on existing research investigating the root causes of terrorism by considering youth un-
Youth unemployment employment as a measure of economic deprivation. In particular, the study focuses on terrorism in Middle
Domestic and transnational terrorism Eastern and North African (MENAP) countries and features terrorism incident count data for the period
Military expenditure 1998–2012 using negative binomial regression models. In our exogenous model, we find that while youth un-
Foreign direct investment
employment tends to increase domestic terrorism, it does not have any significant effect on transnational ter-
Under-five mortality rate
rorism. Given concerns about endogeneity of youth unemployment in these models, we use two kinds of cor-
rections-instrumental variables and lagged variables. We control for endogeneity by using military expenditure,
under-five mortality rate and foreign direct investment as instruments. We are not able to reject the null hy-
pothesis that youth unemployment is exogenous. Using lagged variables, we find a similar result as noted in the
exogenous specifications with regard to the effect of youth unemployment on domestic and transnational ter-
rorism. We also find evidence that domestic terrorism tends to have a positive relationship with press freedom,
religious and linguistic fractionalization, and area of the country. Transnational terrorism has a positive asso-
ciation with the quality of democracy and a negative association with political stability and regulatory quality.
The amount of natural resources tends to be negatively associated with domestic terrorism and positively as-
sociated with transnational terrorism.

1. Introduction
This article examines the role of youth unemployment as a de-
terminant of terrorism in the MENAP (Middle East, North Africa,
Afghanistan and Pakistan) region. The list of countries in this region is
defined in the World Economic Outlook database by International
Monetary Fund [1] and is presented in Table 1.
There has been a lot of research on the determinants of terrorism,
and it has been summarized in Gassebner and Luechinger [2] and
Krieger and Meierrieks [3]. Studies investigating the role of the youth
population on terrorism are somewhat limited. Our motivation for
considering the effect of youth unemployment specifically comes from
the role of three determinants of terrorism highlighted in the literature:
youth bulge, unemployment and inequality.
First, consider the role of youth bulge as a determinant of terrorism.
It is hypothesized that a country or a region is susceptible to violence
such as terrorism when there is an increase in the size of the youth
cohort relative to its adult population. Urdal [4] points out two reasons
behind this hypothesis. If a country experiences a youth bulge but the
labor market fails to absorb them, then this would lead to a sense of
grievance among the youth. Further, this adds to the ease with which
terrorist groups can recruit these individuals resulting in increased in-
cidence of terrorism [4].
Second, unemployment has also been considered as a possible de-
terminant of terrorism in Feldmann and Perala [5] and Piazza [6]. As
Piazza [6] p. 166 states, “…the average national unemployment rate for
each country would be expected to bear a significant positive re-
lationship with terrorism, as unemployment precipitates the stress of
idle workers who might suffer from unmet economic expectations and
therefore turn to political violence.”
Third, inequality is also a possible determinant of terrorism. Gurr
[7] explains that political violence is a consequence of collective frus-
tration created due to a sense of relative deprivation. In his words,
“'Relative deprivation' is the term... used to denote the tension that
develops from a discrepancy between the “ought” and the “is” of col-
lective value satisfaction, and that disposes men to violence.” The dis-
crepancy between an individual's expected and actual welfare due to
lack of employment opportunities generates collective discontent, and
ultimately to terrorism. Using data from the World Values Survey and
the EuroBarometer Survey Series, MacCulloch [8] finds that more

∗
Corresponding author.
E-mail addresses: abagchi@kennesaw.edu (A. Bagchi), jpaul17@kennesaw.edu (J.A. Paul).

https://doi.org/10.1016/j.seps.2017.12.003

0038-0121/ © 2018 Elsevier Ltd. All rights reserved.

Received 9 May 2017; Received in revised form 7 December 2017; Accepted 27 December 2017

Please cite this article as: Bagchi, A., Socio-Economic Planning Sciences (2018), https://doi.org/10.1016/j.seps.2017.12.003
A. Bagchi, J.A. Paul
people prefer a revolt in their country when inequality is high.
In the literature, the support for each of these variables in terms of
its impact on terrorism is rather weak. For example, Gassebner and
Luechinger [2] conclude that none of these three variables are robust
correlates of terrorism. Therefore, one reasonable question to ask is if
these variables together can cause terrorism. Youth unemployment is
related with the three variables mentioned above and therefore can
combine elements of each of these variables. Therefore, we examine the
role of youth unemployment as a determinant of terrorism.
Recently, there has been a surge in terrorist attacks in the MENAP
countries. In these countries, the magnitude of youth unemployment is
much higher than the magnitude of adult unemployment. Also, youth
unemployment is much higher in MENAP countries than in developed
economies. Therefore, it makes sense to examine if higher youth un-
employment in the MENAP region is a root cause of the recent surge in
terrorist attacks in this region.
Schomaker [9] examines the cause of domestic terrorism in the
Middle East. She postulates that this problem is the result of several
features in the Middle East such as youth bulge, lack of political par-
ticipation and lack of migration opportunities. Previously, we discussed
the role of youth bulge in causing terrorism. Lack of political partici-
pation matters because it allows citizens some degree of power to
change their conditions. Migration opportunities matter because it al-
lows individuals to exit a country if they are not satisfied with their
living conditions. In our study, we are more concerned with youth
bulge in causing all kinds of terrorism (domestic and transnational) but
we also control for some features, such as quality of institutions and
quality of democracy.
We first consider an exogenous model and find that the effect of
youth unemployment is statistically significant in the domestic ter-
rorism model. This finding differs from the literature which generally
concludes that unemployment does not matter for terrorism. Further,
there are concerns about endogeneity in this model. We correct for
endogeneity by using instrumental variables and lagged variables. We
consider three instruments viz. military expenditure, lagged under-five
mortality rate and foreign direct investment. We discuss the rationale
for their choice in detail in Section 5. Our analysis indicates that we
cannot reject the null hypothesis that youth unemployment is exo-
genous.
In addition to instrumental variables, we also consider lagged
variables. Here we find evidence that an increase in youth unemploy-
ment in the past leads to an increase in the number of domestic terrorist
attacks in the current period. This effect is also statistically significant.
This result does not hold for transnational attacks.
Other findings are as follows: Linguistic fractionalization tends to
increase the frequency of both (domestic and transnational) types of
terrorist attacks. However, in general, the determinants of domestic
terrorist attacks are different from the determinants of transnational
attacks. The frequency of domestic terrorist attacks has a positive re-
lationship with regulatory quality and press freedom. Regarding
transnational terrorism, we find that it depends negatively on political
stability. That is, unstable countries are generally associated with more
transnational terrorism.
The rest of this article is organized as follows: Section 2 presents our
literature review. Section 3 describes our data. Section 4 describes our
findings in an exogenous framework, and Section 5 extends our results
to the endogenous and lagged variables framework. Section 6 sum-
marizes our findings and provides conclusions and directions for future
research.
2. Literature review
As mentioned in the Introduction, there are several reasons to be-
lieve that youth unemployment has a positive causal relationship with
terrorism. Terrorism is a form of crime, often with a political objective.
Therefore, the motivation for terrorism is partly rooted in the
2
Socio-Economic Planning Sciences xxx (xxxx) xxx–xxx
motivation for crime. It follows from Becker [10] that an unemployed
person has a higher incentive to commit crime because the opportunity
costs of such acts are low. The low opportunity cost of crime can also
lead to terrorism as long as there are other factors present, such as a
political grievances. Therefore, unemployment increases the incentive
to join a terrorist movement. It also makes it easy for terrorist organi-
zations to recruit volunteers [3].
There are several reasons to believe that youth unemployment
matters more to terrorism than aggregate unemployment. First, the
magnitude of youth unemployment is much larger than that of ag-
gregate unemployment. For example, according to the KILM (Key
Indicators of the Labour Market) database of the International Labour
Organization (ILO), the employment-to-population ratio in the Middle
East in 2012 was only 21.8% for the youth while the corresponding
ratio for adults (aged 25 years or more) was 50.9%. In that same year,
the youth employment-to-population ratio for developed economies
and the European Union was 38.8% and the corresponding ratio for
adults was 57.6% [11]. Therefore, the magnitude of the problem is
much worse for the youth than it is for adults. Further, there are some
other well-known harmful effects of unemployment such as skill de-
gradation. For the youth, the cost of skill degradation is much more
pronounced because it translates into low wages for most of their
earning lives. These financial hardships stemming from reduced liveli-
hood could lead to other serious issues. For instance, Cincotta et al. [12]
and Urdal [13] find that youth unemployment is a key cause of civil
war. We explore if youth unemployment has any relationship to ter-
rorism or not.
In the MENAP region, the extent of youth unemployment is quite
high when one compares it with the OECD countries. Dhillon and
Yousef [14] summarize several reasons behind this. In many countries
in the MENAP region, the government generally provides for free
education. However this system has come under strain in recent times
because of the youth bulge. According to the World Bank [15] p. 2), this
region “…has not capitalized fully on past investments in education, let
alone developed education systems capable of meeting new challenges.
The education systems did not produce what the markets needed, and
the markets were not sufficiently developed to absorb the educated
labor force into the most efficient uses.”
Scores on TIMSS (Trends in International Math and Science Study)
or PISA (Programme for International Student Assessment) are com-
monly accepted to be an internationally comparable measure of edu-
cation quality. It follows from the same report [15] p. 19 that the
average score for Middle Eastern countries (compiled over several years
between 1995 and 2003) was 399 in comparison to an international
average of 489. The poor quality of education makes it difficult for the
youth to find meaningful employment. Another factor that makes it
hard for the youth to gain employment is excessive reliance on oil ex-
ports and the resulting lack of diversification of the economy [16] p.
324. The consequence of poor job prospects is that it makes it chal-
lenging for the youth to marry or secure housing. All of these factors
breed resentment and can act as a catalyst for terrorism. As Berrebi [17]
notes, “highly educated individuals would be particularly frustrated by
the loss of economic opportunities and the alternative economic cost of
their risking arrest or worse would be lower.”
There is an additional channel that links youth unemployment with
terrorism. This line of research considers the role of social identity in
explaining the motivation for terrorist attacks. Social identity theory
was developed in psychology by Tajfel and Turner [18,19]. This theory
stresses that individuals identify with a social group (such as an ethnic
group) and their sense of self-worth depends on the status of the group.
Therefore, an individual tends to favor members of their own group and
discriminate against those who belong to other groups. The concept of
social identity has been formalized by Shayo [20] and has been further
extended by Sambanis and Shayo [21] to explain conflict.
Let us consider the role of social identity in the context of this paper.
Consider a country in which citizens have a strong sense of social
A. Bagchi, J.A. Paul
Fig. 1. Transnational incident data distribution (MENAP: 1998–2012).
identity. If there is a high degree of unemployment amongst members
of a social group (such as an ethnic group), then this will make many
persons in that group resentful, and this will include both employed as
well as unemployed persons. Such grievances may lead to more ter-
rorism. It is easy to see why unemployed persons will have the moti-
vation to join terrorist movements. However, social identity theory
implies that mass unemployment will motivate even employed persons
to join such movements because they will view widespread misery
among their social group as a personal affront.
In summary, youth unemployment affects terrorism through both
through the “demand side” as well as “supply side” factors. Youth un-
employment makes it easy for terrorist groups to recruit volunteers and
this is an example of a supply side factor. Additionally widespread
youth unemployment creates resentment and generates the demand for
terrorism. Given the discussion above, it is reasonable to explore if
youth unemployment is a key determinant of terrorism in the MENAP
region. Therefore, our primary research question is as follows.
Hypothesis 1. Increased youth unemployment increases incidents of both
domestic and transnational terror in MENAP countries.
Our work is also related to the literature (primarily empirical) on
the root causes of terrorism. These papers differ from each other in
several aspects such as the sample of countries, period of study etc. In
spite of these differences, there are some common findings and we
summarize them below. In general, there are two broad classes of re-
sults on the root causes of terrorism. One group of papers has the
message that economic factors such as poverty and inequality reduce
terrorism. Li and Schaub [22] find a negative relationship between
economic development of a country and greater trade openness with
the number of terrorist incidents inside the country. Berrebi [17] notes
the positive relationship between lack of employment opportunities and
the likelihood of the youth joining terrorist groups. Blomberg and Hess
[23] find a positive relationship between economic development and
transnational terrorism in high income countries and a negative re-
lationship in low income countries.
The second group on the other hand posits that political and in-
stitutional factors are the major determinants of terrorism. For example,
Eyerman [24] finds that established democracies are less likely to face
terrorism than non-democracies and that newly formed democracies
are more vulnerable to terrorism than other types of states. Li [25] finds
that democratic environment increases satisfaction and political effi-
cacy of citizens. It helps reduce their grievances and thereby deter
terrorist recruitment and increases public tolerance toward counter-
terrorism policies. Testas [26] shows that repression decreases ter-
rorism in the short run but not in the long run. This suggests that a
democratic environment would help better curb the risk of terrorism.
3
Socio-Economic Planning Sciences xxx (xxxx) xxx–xxx
In the same spirit, there are several papers that reject the role of
economic variables as determinants of terrorism. For example, Krueger
and Laitin [27] find that GDP per capita and GDP growth rate are not
reliable predictors of transnational terrorism. Instead, they find that
politically free countries are likely to be targeted by terrorists from
politically oppressed countries. Krueger and Malečková [28] examine
the case of Palestinian terrorism and find that wealthier and more
educated individuals are more likely to be terrorists than poorer in-
dividuals. Abadie [29] finds that GDP per capita has little to do with
terrorism and that lack of political rights is the most important de-
terminant of such acts. Piazza [6] suggests that unemployment is not a
strong indicator of terrorism. In fact, he does not find any relationship
between economic-development variables such as economic growth,
inflation, unemployment, inequality, etc. and terrorism. Krueger and
Laitin [27] suggest that political repression and not economic status is a
better determinant for the national origins of terrorism. Burgoon [30]
finds that social welfare, measured by social spending as a percent of
GDP, reduces terrorism.
The relationship between unemployment and terrorism is an under-
researched area so far. There is some research on the relationship be-
tween aggregate unemployment and terrorism (such as [6] and the
conclusion is that aggregate unemployment does not cause terrorism. In
contrast, research on the role of youth unemployment as a cause of
terrorism tends to find evidence of a relationship. For example, Oyefusi
[31] examines the reasons for youth participation in political violence
in Nigeria's Delta and finds that education in combination with un-
employment induces the youth to participate in political violence.
Caruso and Schneider [32] find evidence of a positive association be-
tween youth unemployment and terrorism in Western Europe. Caruso
and Gavrilova [33] find a positive relationship between the growth rate
of youth unemployment and terrorism in Palestine. We extend this
literature and discuss the relationship between both domestic as well as
transnational terrorism in the MENAP region. It is also possible that
youth unemployment affects terrorism endogenously. Therefore, we
consider an exogenous as well as an endogenous framework in our
paper.
3. Description of data
In this paper, we use data from the Global Terrorism Database
(GTD). The database lists terrorist incidents all over the world 1970
onwards. We extract information about the number of terrorist attacks
that took place in the MENAP countries (listed in Table 1) for the years
1998 through 2012. Therefore we are interested in examining if a high
rate of youth unemployment makes a country more vulnerable to ter-
rorist attacks. Sometimes, it is difficult to separate criminal activities or
political violence from terrorist attacks. If there is any doubt about
whether or not an attack is an act of terrorism, then the database ex-
plicitly mentions that. We drop these doubtful observations from our
sample.
The next step is to classify each terrorist attack as domestic or
transnational. In principle, a terrorist attack can be classified as trans-
national based upon any of the following three criteria: (i) the terrorist
group's nationality is different from the location of the attack (that is, if
the terrorist group crosses an international border in carrying out an
attack), (ii) the terrorist group and the victim have different national-
ities, and (iii) the victim's nationality is different from the country in
which the attack occurs.1 The GTD database reports a variable called
INT_ANY that takes a value 1 if an attack has any of the above men-
tioned three elements present, 0 if none of these elements are present
and −9 otherwise. We classify any attack as transnational if the
1
The GTD database assigns a nationality to each group. This is not explicitly reported
but can be inferred in many cases. No information on nationality is provided in the case of
lone-wolf attacks. Also, there is no information on nationality of the actual perpetrator.
A. Bagchi, J.A. Paul
Fig. 2. Military expenditure in MENAP (1998–2012).
(Source: World Bank).
variable INT_ANY takes a value of 1. All other incidents are classified as
domestic.
There are some issues about transnational terrorism that we would
like to clarify. In the literature, there are papers (such as [28,34] etc.)
that examine the factors that cause a nation to be the source of ter-
rorism. However, there are also other papers (such as [25,35] etc.) that
examine the factors that make a nation a target of terrorism. Finally,
there is a third strand of research such as Kurrild-Klitgaard et al. [36]
that look at both kinds of countries-the originator as well as the victim.
Krieger and Meierrieks [3] provide an excellent overview on the
determinants of transnational terrorism. They conclude that everything
else remaining constant, a country is more likely to be a target of
transnational terrorism if it is economically successful, politically open
etc. We extend this strand of the literature and examine if a high level of
youth unemployment can make a nation a target of both domestic as
well as transnational terrorism. The reason for looking at the link be-
tween domestic terrorism and youth unemployment follows easily from
our discussion of the primary hypothesis. Below, we explain why youth
unemployment in a country may also make it more vulnerable to
transnational terrorism.
There are two kinds of transnational attacks we consider in this
paper. The first kind is attacks against foreign nationals in a country.
There are several reasons why terrorist groups in a country can attack
foreigners. One reason is that in several MENAP countries, foreign
militaries have been used either to stabilize a country or to assist its
government. National governments such as in Afghanistan depend on
these foreign forces for their survival. Therefore, terrorist groups attack
foreign militaries in order to force them to leave. In addition to mili-
taries, international aid agencies and foreign businesses are also tar-
geted by terrorists.
The second kind of transnational attacks are attacks by foreign
terrorist groups. The motivation for these attacks is as follows: Consider
an ethnic group X that resides in two countries- 1 and 2.2 Now suppose
there is a terrorist group based principally in country 1 but that claims
to represent the interests of the whole ethnic group X. Further, assume
that there are grievances (such as mass unemployment) amongst
members of X in country 2. In this case, country 2 will be vulnerable to
terrorist attacks from country 1 because of two reasons. First, it follows
from social identity theory that the terrorist group would consider
fighting the government of country 2 as a legitimate objective. The
second reason is that terrorist attacks often depend upon internal help
(such as accurate information) for its success. Unemployed youth (of
2
The same argument applies to religious and linguistic groups also.
4
Socio-Economic Planning Sciences xxx (xxxx) xxx–xxx
ethnic group X) residing in country 2 will have a motive to provide such
help to the terrorist group.
As a real world example, consider the case of the Kashmir region
that is divided between India and Pakistan. Several terrorist groups
based in Pakistan such as Lashkar-e-Taiba, Jaish-e-Mohammed etc.
have been responsible for attacks inside Indian Kashmir in recent times.
According to our definition, these would qualify as transnational at-
tacks. However, the motivation for these attacks is the grievance of
Kashmiris residing in India against the Indian government. Since in-
ternational borders in the MENAP region do not always follow ethnic,
religious or linguistic contours, therefore it makes sense to explore if
youth unemployment in MENAP countries makes them more vulnerable
to transnational terrorist attacks.3
We present the annual numbers of total, domestic and transnational
terrorist incidents for the MENAP countries in Table 2. Even with our
broad definition of transnational terrorism, we find that domestic ter-
rorism makes up about 90 percent of the total incidents of terrorism in
our sample. Therefore domestic terrorism poses a much bigger chal-
lenge for developing countries than transnational terrorism.
Our variable descriptions and their type i.e. dependent, independent
or instrumental are included in Appendix Table A1. We now briefly
discuss the rationale behind choice of variables and what they embody.
First, we consider youth unemployment, which is the primary
variable of interest. There are several ways of measuring the extent of
youth unemployment. In this paper, we use the employment-to-popu-
lation ratio of the youth cohort (also known as the youth employment
rate). We measure youth employment rate by the employment-to-po-
pulation ratio of the youth population. This is defined as
Total Number of Employed persons aged between 15 − 24 years 4
Population aged between 15 − 24 years
. This ratio measures the
probability that a randomly selected person aged between 15 and 24
years is employed and is an index of the ease of finding employment for
the youth. For our regressions, we estimate the coefficients of this ratio.
While interpreting results, it is often convenient to refer to youth un-
employment and this is negatively related with the youth employment
rate. As we mention above, in the MENAP countries, youth un-
employment and the frequency of terrorist attacks is higher than in
developed countries.
Another commonly used measure of labor market slack is the un-
employment rate. The unemployment rate is the proportion of the labor
force that does not have a job and is actively looking for work [11]. If
the labor market is depressed for a long time, then it may induce many
individuals to drop out of the labor force. This effect decreases both the
numerator as well as the denominator in the expression for the un-
employment rate. As a result, it is possible for the unemployment rate to
decrease in a depressed labor market. Quite naturally, a critique of the
unemployment rate is that the unemployment rate understates labor
market slack [37]. In the MENAP region, the youth face significant
barriers in the labor market, and consequently have a high chance of
dropping out of the labor force. The unemployment rate therefore un-
derstates the barriers that the youth face in the labor market.
According to the International Labour Organization [11] p. 33, the
“… employment-to-population ratio provides information on the ability
of an economy to create employment; for many countries the indicator
is often more insightful than the unemployment rate.” The
3
The purpose of this paper is to examine if youth unemployment in a country can make
it a target (location) of terrorist attacks. Given this objective, there is a stronger rationale
for counting attacks against foreign nationals as transnational attacks in a country. If
citizens of a country believe that unemployment in a country is a result of the policies of
certain foreign countries, then they might want to target military personnel, aid workers
etc. of those countries. Therefore, as a robustness check, we also check our results on
transnational terrorism by excluding attacks by foreign terrorist groups. In our dataset,
there are relatively few attacks committed by foreign terrorist groups. Therefore, the
results stay the same even when we exclude these attacks. These results are available in
the extended appendix section (see Table EA.1).
4
We use data from World Bank. Comparable numbers are also available at the KILM
database of the International Labour Organization (ILO).
A. Bagchi, J.A. Paul
Fig. 3. Military expenditure (% of GDP): MENAP vs. OECD (1998–2012).
(Source: World Bank).
employment-to-population ratio is not affected by people dropping out
of the labor force and therefore we use the employment-to-population
ratio of the youth population.5
Below, we introduce our independent variables.
Politics: One set of controls is the nature of the politics. This is
measured by three variables: (i) Lack of Political Rights, (ii) Polity and
(iii) Press Freedom.
Lack of Political Rights measures the extent of political rights, such
as pluralism. The role of political rights have been explored in Testas
[26], Abadie [29] and Bandyopadhyay and Younas [38] etc. These
papers find that this variable has a statistically significant non-linear
relationship with the number of terrorist attacks. Widespread un-
employment results in terrorism in the presence of other contributory
factors, such as political repression. Hence, we control for this variable.
The variable Polity measures the quality of democracy and takes
values between −10 and +10, with higher values representing a
higher quality of democracy.6 The effect of the quality of democracy on
terrorism can go either way. On one hand, in a democratic society, the
government is usually more responsive to the demands of the citizens.
However, on the other hand, governments in democratic countries have
to be more restrained and this makes it easy for terrorist groups to
operate.
The third variable that captures the nature of politics is freedom of
the press. This has been controlled in Li [25]. There are two reasons
why more press freedom can lead to more terrorist attacks. First, ter-
rorist attacks are more likely to be reported if the target country has a
free press. This may create an appearance that open societies are more
vulnerable to terrorism. Second, a free press offers terrorist groups a
forum to attract attention. Since gaining publicity is one of the major
motivations of terrorist attacks, therefore a free press has the perverse
effect of encouraging terrorists [25,39]. Indeed, the literature (such as
[25] has found a statistically significant positive relationship (i.e. as
press freedom goes up, it tends to increase incidents of transnational
terrorism) with transnational terrorism.
Governance Indicators: The extent of terrorism also depends upon
institutional factors. To control for the quality of institutions, we use
the World Governance Indicators. The governance indicators that are
relevant are: (i) Voice and Accountability, (ii) Political Stability, (iii)
Government Effectiveness, (iv) Regulatory Quality, (v) Rule of Law and
vi) Control of Corruption. Some of these institutional factors have been
controlled for in the literature. Kis-Katos et al. [34] find that failing
5
As a robustness check, we also consider the youth employment-to-population ratio
among males and females to check if there are any different effects across gender and if
the results hold qualitatively. These results are available upon request from authors.
6
In three kinds of situations, Polity is coded differently. These are: (i) Foreign
Interruption (coded as −66), (ii) Interregnum or Anarchy (coded as −77), and (iii)
Transition (coded as −88). We view these three situations as reflecting a poor quality of
the democracy and use a value of −10 in each case.
5
Socio-Economic Planning Sciences xxx (xxxx) xxx–xxx
states are associated with more terrorist attacks. An unstable state may
not be able to police its entire territory effectively and this allows ter-
rorist groups to operate easily in those areas. We include political sta-
bility as one of our independent variables to control for this effect.
Another governance indicator that is of interest to us is rule of law.
Bandyopadhyay and Younas [38] explicitly control for this and find
that better enforcement of the law reduces terrorism. Our view is that
the overall quality of governance affects the extent of terrorism. Hence,
in addition to these two variables, we also included a few more gov-
ernance indicators. One problem with using governance indicators is
that these are based on perceptions. However, this problem is present in
other variables also, particularly those which control the nature of
politics, and is an unavoidable issue in this line of research. Given
concerns that the governance variables are highly correlated, we de-
veloped a correlation matrix (see Table 3) and only retained those
variables that would not cause multicollinearity issues. For our decision
purposes, we used an absolute correlation value of 0.75. Because of this
step, we retained political stability, voice and accountability and reg-
ulatory quality to represent governance indicators in our modeling
specifications.
Degree of fractionalization: The extent of diversity in a country
can lead to friction between different social groups and this may ulti-
mately result in terrorism. To control for the extent of diversity, we
include indices of ethnic, linguistic, and religious fractionalization from
Alesina et al. [40]. A higher value reflects the probability that two
randomly chosen individuals from the same country belong to two
different groups. These measures have been used in Tavares [35],
Abadie [29] and Bandyopadhyay and Younas [38].
Geographical Variables: Everything else remaining constant, a
country will be more vulnerable to terrorist attacks if it has the char-
acteristics of a suitable target. Some of these characteristics in turn
depend upon geography. Therefore, we consider geographical variables
such as (i) land area, (ii) elevation, and (iii) fraction of the country in
tropics. A large country is more vulnerable to attacks because it pro-
vides more targets. A mountainous country is harder to police and
hence is more vulnerable to terrorist attacks [29]. Gaibulloev and
Sandler [41] find that terrorist groups are more likely to survive if they
are based in a tropical country.7
Additionally, we also control for the rate of urbanization. This has
been done previously by Tavares [35] and Urdal [4]. Urdal [4] cor-
rectly points out that urban areas are an attractive target for terrorists
because of their higher population densities. Noting that natural re-
source and its criminal exploitation is one of the root causes of ter-
rorism [42], we included rents from natural resources as a control. We
obtained this data from World Bank [68]. We also included a dummy
variable for Kuwait, Oman, Qatar and Saudi Arabia as these countries
are likely to be outliers due to their high levels of economic develop-
ment when compared to remaining countries in the MENAP region.
Other possible independent variables: The above independent
variables were all motivated by the literature on determinants of ter-
rorism. Apart from these, we also include some other variables that can
be confounding factors.
It is interesting to examine if human capital development has a
significant effect on terrorism or not. In order to answer this question,
we control the educational level of a country using the UNDP Education
Index. This index is measured by using information on the mean of
years of schooling for adults aged 25 years and expected years of
schooling for children of school entering age. A higher value of this
index represents a higher average level of education. We also
7
For domestic terrorism, the base of operations of the terrorist group and the target are
both located in the same country and hence there is a clear case for controlling for the
area in tropics. For transnational terrorism, this variable matters more for the base of
operations. Since we are interested in examining the characteristics of the victim country
for transnational attacks, therefore we exclude this variable in our regressions for
transnational attacks.
A. Bagchi, J.A. Paul
Fig. 4. Military expenditure vs transnational terrorism in MENAP.
(Source: World Bank).
considered variables that could potentially control for the effect of
external influence such as foreign aid. Ultimately, these variables were
dropped from our main regressions because of excessive missing values.
For instance, the sample size dropped to under 90 observations going
from almost 186–200 sample size that our original models are built on.
While youth unemployment can cause terrorism, it is also plausible
that the reverse is true (that is, terrorism may also cause youth un-
employment). To control for endogeneity we use two approaches. One
approach is to use instrumental variables and the instruments we use
are military expenditure, foreign direct investment and the under-five
mortality rate. We discuss the underlying rationale that motivated their
selection in detail in section 6.
The other approach we employ to address concerns of reverse
causality is to use lagged values of independent variables. It is rea-
sonable to assume that disaffected youth may resort to terrorism only
with a lag. There are several reasons for the lag. First, once a person
joins a terrorist group, it is difficult to come back to a normal life. It is
quite natural for a person to think carefully before deciding whether or
not join a terrorist group. Additionally, it takes time to plan and execute
an attack. For all of these reasons, it is plausible that youth un-
employment today will affect terrorism tomorrow, and this idea is also
discussed in our paper.
To the best of our knowledge, there is no single data set that in-
cludes all of the variables required for our analysis. Therefore, it was
required that the information on these variables be collected from a
variety of sources as indicated in Appendix Table A2. There are 23
countries in the MENAP region. Out of these, we had to drop 6 countries
(mentioned in Table 1) due to lack of information. Further, data on
World Governance Indicators are not available for two years- 1999 and
2001 and we had to drop these two years also. Ultimately, our sample
comprises of 17 countries over 13 years, and consequently our max-
imum sample size is 221.
We lost observations due to missing data on variables used in our
study including press freedom (loss of 27 observations), control of
corruption (loss of 4 observations), military expenditure (loss of 10
observations), FDI (loss of 2 observations), religious fractionalization
(loss of 10 observations), etc. Various combination of other variables
resulted in an additional loss of 8 observations. The surviving sample
with no missing values for any of the variables has 160 observations
with no obvious pattern or bias in the observations removed to raise
concerns about the resulting sample. Appendix Table A3 provides the
summary statistics for this sample. In our regressions, the sample size is
6
Socio-Economic Planning Sciences xxx (xxxx) xxx–xxx
more than 160 because we do not use all of the variables at a time
especially as we endeavor to determine the preferred specifications.8
4. Results: exogenous model
In this paper, our dependent variable is a count variable, therefore
we consider using either Poisson Regression or Negative Binomial
Regression. However, we find substantial over-dispersion in our data
(see Fig. 1), and therefore Negative Binomial Regression seems more
appropriate.
Also, as Fig. 1 shows, the number of terrorist incidents is 0 on many
occasions. Hence, there is also a need to consider between Negative
Binomial Regression or Zero-inflated Negative Binomial Regression.
Zero-inflated Binomial Regression is the appropriate method if there are
reasons to believe that there are two groups of countries - in the first
group, terrorist incidents never take place, and in the other group,
terrorist incidents can take place but there may be years with 0 in-
cidents. In the MENAP region, there are some countries such as Qatar or
Oman that are remarkably terror free, while there are other countries
such as Afghanistan, Pakistan or Iraq that account for most of the ter-
rorist attacks. Therefore, it is plausible that there are two types of
countries, and the large number of 0s occurs partly because one group is
generally terror free. Using the Vuong test, we find that Negative Bi-
nomial Regression is the more appropriate method in this case. These
results can be seen in Table 4. This being said, our primary findings
even with a zero-inflated negative binomial setup do not change qua-
litatively. These results are available from the authors upon request.
We separately estimate the regressions of total, domestic and
transnational terrorist incidents using the following specification [43]:
α exp(Xi′β1 + Youth employment rate β2) ⎞
= ∑ ⎧yi ln ⎛
n
⎨ + α exp(Xi′β1 + Youth employment rate β2) ⎠
⎜ ⎟
i=1 ⎩ ⎝
1
− ln(1 + α exp(Xi′β1 + Youth employment rate β2))
1
α
+ ln Γ ⎛yi + ⎞ − ln Γ(yi + 1) − ln Γ ⎛ ⎞ ⎫
1 1
⎝ α⎠ ⎝ α ⎠⎬
⎭ (1)
The above equation is expressed as log-likelihood function, a typical
representation in a count model. In the above equation, yi represents
the response variable measured by count of terrorism incidents. The
vector X includes other potential predictors of terrorism such as mea-
sures of press freedom, religious, linguistic and ethnic fractionalization,
country geography, etc. These variables are listed in Table 4 and are
motivated by the literature. Table 4 columns (1)–(2) and) (3)–(4) pre-
sent results for domestic and transnational terrorism incidents respec-
tively. We discuss the results pertaining to domestic terrorism first, and
then turn to transnational terrorism.
We specifically elaborate on the preferred models (see columns (2)
and (4)). While we ran several models before arriving at the preferred
model, for the sake of brevity and to avoid confusion, we only present
the full and preferred ones. The preferred model was selected using
metrics such as Akaike Information Criterion (AIC) and Bayesian
Information Criterion (BIC). As noted earlier, another aspect that was
instrumental in our choice of the models we refer to as “preferred” is
our observation that Government effectiveness, Rule of Law, Control of
Corruption and Regulatory Quality were highly correlated (correla-
tion ≥ 0.75 and statistical significance <0.01) with VIF greater than
10. Thus including both in the same specification would result in ser-
ious multicollinearity issues. We chose Regulatory Quality because it
generates a better fit of the models in terms of AIC and BIC. We use the
preferred specification as the baseline model in the paper.
8
For details on the preferred specifications, refer to section 5.
A. Bagchi, J.A. Paul
Table 1
List of countries in the MENAP region.
ID Countries
1 Afghanistan
2 Algeria
3 Bahrain*
4 Djibouti*
5 Egypt
6 Iran
7 Iraq
8 Israel
9 Jordan
10 Kuwait
11 Lebanon
12 Libya
13 Malta*
14 Morocco
15 Oman
16 Pakistan
17 Qatar
18 Saudi Arabia
19 Syria
20 Tunisia
21 United Arab Emirates*
22 West Bank and Gaza*
23 Yemen*
Due to data limitations, we had to drop six countries (in
asterisks) from our sample.
4.1. Domestic terrorism
In our preferred specification, we find that for the case of domestic
terrorism, the coefficient of the youth employment rate has a negative
sign and is statistically significant when modeled exogenously. This
finding differs from papers such as Piazza [6] which conclude that
unemployment does not matter for terrorism. However, this finding is
consistent with papers (such as [31–33] that use youth unemployment
as a control variable instead of aggregate unemployment. In terms of
magnitude, the regression indicates that everything else remaining
constant, a one unit increase in the youth employment rate will reduce
the number of domestic terror attacks by approximately 10.7%. When
evaluated at the average number of attacks, this is equivalent to 7.1 less
attacks per country year.
Regarding the other independent variables, we find that domestic
terrorism increases with the degree of press freedom and this effect is
statistically significant. One of the aims of terrorist groups is to gain
publicity and a free press helps them in this regard. A similar re-
lationship has been found in the literature (such as [25,39]).
We also find a positive relationship between domestic terrorism and
religious and linguistic fractionalization. This is consistent with the
concept of identity and adds support to Rummel's hypothesis [66] that
the likelihood of collective violence is higher in states in which power is
centralized and is dependent on one's social membership such as re-
ligion or language. This is a characteristic that is synonymous with
MENAP countries and therefore increase in religious or linguistic di-
versity can result in increased incidents of terror.
Additionally, we find that the magnitude of domestic terrorism is
higher in larger countries. This is because larger countries offer more
targets. Finally, domestic terrorism tends to have a negative association
with the amount of natural resources.
4.2. Transnational terrorism
It follows from Table 4 (column (4)) that transnational terrorism has
a statistically insignificant relationship with youth employment rate
when modeled using an exogenous specification. In terms of magnitude,
the exogenous model indicates that everything else remaining constant,
a one unit increase in the youth employment rate will reduce the
7
Socio-Economic Planning Sciences xxx (xxxx) xxx–xxx
number of transnational terror attacks by 0.9%. When evaluated at the
average number of attacks, this is equivalent to 0.062 less attacks per
country year.
We find a negative relationship between transnational terror and (i)
regulatory quality and (ii) political stability. Regulatory Quality is a
governance indicator that captures a government's ability to implement
policies for the development of the private sector. Our results show that
creating a rules-based system for the private sector reduces transna-
tional terrorism. Regarding political stability, a regime is perceived to
be unstable if there is a high likelihood of it being overthrown. Such a
perception leads to more incidents of transnational terrorism. Several
other papers (such as [34,44] have a similar finding. The reason is that
it is easy for the terrorist groups to operate or to recruit in failing or
failed states.
Additionally, we find a positive relationship between transnational
terrorism and (a) amount of natural resources in a country and (ii)
polity. A natural resource rich country in the MENAP region tends to
attract foreigners and this presents more opportunities for terrorist at-
tacks.
Polity measures the quality of democracy. We argued above that the
coefficient of Polity can go either way. If there is an improvement in the
quality of democracy, then on the one hand there is a reduction in the
incentive to participate in terrorist movements because non-violent
methods of protest are feasible, while on the other hand terrorism can
go up because the government will be forced to be restrained in
countering terrorist groups. In the case of the MENAP countries, we find
that democratic reforms tend to increase transnational terrorism.
5. Results: endogenous model
It is possible that in the regressions discussed above, youth em-
ployment rate is an endogenous variable. To deal with potential en-
dogeneity issues, we introduce three instruments: military expenditure,
under-five mortality rate and foreign direct investment.
5.1. Military expenditure and youth employment
Below, we examine if there is any plausible reason to believe that
military expenditure in the MENAP region is associated with youth
employment. First, let us consider the data. It follows from Fig. 2 that
there was a surge in military expenditures in MENAP countries during
our period of analysis. It also follows from Table 2 that this trend co-
incides with the surge in terrorism. Below, we first argue why high
levels of military expenditure will result in youth unemployment. Then
we argue that because of the special context of the MENAP region, there
are reasons to believe that military expenditure has an indirect re-
lationship with transnational terrorism through its effect on youth un-
employment.
As Fig. 3 shows, the MENAP countries on an average spend a larger
fraction of their GDP on the military compared with OECD countries.
This can affect growth in several ways. Landau [45] highlights a non-
linear relationship between military expenditure and economic growth
in less developed countries with it being positive as expenditure rises
and then becoming negative as this burden becomes too big.
Aizenman and Glick [46] find that military expenditure reduces
growth if its primary motivation is rent-seeking and corruption. In the
MENAP countries, this seems to be the case. Not surprisingly research
featuring countries from the MENAP region has shown that increased
military expenditure has a negative impact on economic growth [47].
Hence, excessive military expenditures may decrease youth employ-
ment by dampening economic growth.
There is another reason why military expenditures in the MENAP
region would be expected to have a negative relationship with youth
employment. Most of the MENAP countries are autocratic regimes.
Suppose youth employment decreases in such a country because of poor
economic management by the government. This would lead to unrest
A. Bagchi, J.A. Paul Socio-Economic Planning Sciences xxx (xxxx) xxx–xxx

Table 2 adding support to our argument.

Terrorism incident count for MENAP countries (1998–2012).

Year Total Incidents Transnational Domestic 5.2. Lagged under-five mortality rate and youth employment

1998 205 35 170

In the literature, the current child mortality rate has often been
1999 124 35 89
2000 207 35 172 considered as an index of development [4] p. 616. It has been used to
2001 243 56 187 examine if the level of development has any relationship with ter-
2002 266 85 181 rorism. We however use the lagged value of a child mortality measure
2003 272 113 159 called the under-five mortality rate as a measure of the youth bulge.
2004 470 153 317
2005 751 117 634
The under-five mortality rate is the probability that a newborn baby
2006 1291 136 1155 would die before its fifth birthday.
2007 1499 122 1377 During the last fifty or so years, the population of the MENAP region
2008 2064 191 1873 increased 3.7 times, more than any other major world region [49]. With
2009 2372 84 2288
increased access to modern medical services and public health inter-
2010 2458 112 2346
2011 2659 134 2525 ventions, such as antibiotics, immunization, and sanitation, mortality
2012 4026 242 3784 rates have dropped dramatically as well.10 As a result, in the MENAP
region, the under-five mortality rate dropped from close to 200 deaths
per 1000 live births in the early 1950s to fewer than 50 deaths per 1000
among the citizenry. It is rational for the regime to support the military live births at the turn of the 21st century. These multiple factors have
to protect it in the case of an anticipated uprising (e.g., Syria). If the led to an increase in the youth population entering the labor market
primary reason for such unrest is poor economic management by the that failed to absorb so many entrants. On top of that, many of these
regime, then one would expect to find a negative relationship between entrants are not properly trained for the available jobs. According to
military expenditure and youth employment. Barro and Lee [50], the average years of total schooling is only 3.85 for
This leads to the second research question we focus on in this study. Afghanistan, 5.02 for Pakistan and 7.16 for Iraq. The corresponding
Hypothesis 2. Military expenditure is negatively related with youth numbers are also quite low for most other MENAP countries. In com-
employment. parison, this number is 13.18 for USA and 12.24 for UK.
This lack of employment opportunities often leads to increased
Now let us consider the direct relationship between military ex- frustration among young people which can manifest in the form of
penditure and transnational terrorism. As we discussed before, there are violent acts including domestic and transnational terrorism. Therefore,
two kinds of transnational attacks that we consider in this paper-attacks lagged under-five mortality rate has an indirect effect on incidences of
against foreigners and attacks by foreign terrorist groups. In our data, terrorism through its effect on youth employment. This leads to the
the former is much more common. The targets of these attacks are third research question in this study.
mainly western military personnel. It is natural for the target countries
(such as USA) to expend resources for the security of their personnel. Hypothesis 3. Lagged under-five mortality rate is positively related with
However, the military expenditure data we use is the expenditure by youth employment.
the host country and this is not primarily used to protect foreigners.9 Lagged under-five mortality rate is unlikely to have any strong di-
Therefore, it is plausible that given the special context of the MENAP rect effect on terrorism of either variety. Therefore, we use it as an
region (in which foreign militaries are present in many countries), instrument. Our econometric tests below confirm that this is indeed a
military expenditure will not have a direct effect on transnational ter- valid instrument.
rorism. It is however plausible that military expenditure may have a
direct effect on domestic terrorism. Therefore we use military ex-
penditure as an instrument in the case of transnational terrorism but not 5.3. Foreign direct investment and youth employment
in the case of domestic terrorism. Later on, we confirm econometrically
that this is a valid instrument. Below, we consider the role of Foreign Direct Investment (FDI). We
There is another reason for the use of military expenditure as an argue below why FDI has an indirect impact on terrorism through its
instrument in the case of transnational terrorism. The military ex- effect on youth employment. Later on, we show why the direct effect of
penditure of a country is primarily determined by military budgets of FDI on terrorism is likely to be small in the MENAP countries.
principal adversaries. For example, Pakistan's budget depends to a large Javorcik [51] examines the empirical evidence about the relation-
extent on India's budget. According to Abdelfattah et al. [48], Egypt's ship between FDI and “good” jobs. Good jobs pay higher wages to
military expenditure increases with Israel's military expenditure. In workers, have high productivity and have positive spillover effects. The
general, the number of transnational attacks is quite few and is unlikely empirical evidence suggests that FDI creates good jobs from the per-
to result in a major change in a country's military budget. The lack of a spective of workers because such jobs are likely to pay more than in-
direct relationship between military expenditure and transnational digenous firms and provide more training than local firms. From the
terrorism is well depicted in Fig. 4. Further, the coefficient of variation perspective of the economy, such jobs are also good jobs because FDI
for transnational terrorism (0.64) and military expenditure (0.23) data flows increase the productivity of the industry and create extensive
for the years we study (depicted in Fig. 4) are significantly different knowledge spillovers. These knowledge spillovers accelerate economic
growth which aids in job creation.
In the context of the MENAP region, Dhillon et al. [52] point out
9
that traditionally the greatest source of employment in this region was
According to Belasco [69], the U.S. military expenditure in 2011 was $107 bn. in
the public sector. However the public sector also has a tendency not to
Afghanistan and $47 bn. in Iraq (see Fig. 3 of the report). This can be compared with the
military expenditures of the host government. The primary source of data on military fire workers. This tends to favor older workers (who already have the
expenditures is Stockholm International Peace Research Institute (SIPRI). According to jobs) over the younger workers. Therefore one plausible way of
them, the military expenditure in current $ in 2011 was approximately only $0.3 bn. in
Afghanistan and $6 bn. in Iraq. The upshot is that military expenditure by a country's’
10
government in the MENAP region does not equate to the total amount spent on security in Given that there would be a lag between the generation of population bulge and
the country because a substantial part of these expenditures are incurred by foreign youth unemployment, therefore we consider the lagged value of the under-five mortality
militaries. rate.

8
A. Bagchi, J.A. Paul Socio-Economic Planning Sciences xxx (xxxx) xxx–xxx

Table 3
Correlation table: Governance indicators.

Voice and Accountability Political Stability Government Effectiveness Regulatory Quality Rule of Law Control of Corruption

Voice and Accountability 1

Political Stability 0.3073 1
Government Effectiveness 0.6898 0.6456 1
Regulatory Quality 0.7001 0.5856 0.9104 1
Rule of Law 0.6567 0.7499 0.9057 0.893 1
Control of Corruption 0.6234 0.7408 0.9158 0.8653 0.917 1

alleviating youth unemployment is to encourage the growth of the exogenous and therefore not valid.
private sector. Developing an indigenous private sector takes time and
attracting FDI therefore seems to be one of the viable solutions. Hence, 5.4. IV regressions
we hypothesize that greater FDI in the MENAP countries should in-
crease youth employment, and therefore should reduce terrorism in- We use a two stage instrumental variables approach to account for
directly. This leads us to our fourth research question. endogeneity (see Ref. [43], pp. 417–421). In the first stage, we model
Hypothesis 4. Foreign direct investment is positively related with youth youth employment as the endogenous predictor. This is because while
employment rate. youth employment can cause terrorism, it is also possible that terrorism
causes unemployment by dampening economic growth.
Now let us consider the relationship between FDI and terrorism (of Therefore it is not possible to rule out the fact that the youth em-
both kinds). FDI is unlikely to directly result in terrorism. One concern ployment rate might be correlated with an unobservable predictor or
is if FDI can lead to a reduction in terrorism. This is plausible, but is predictors that are latently included in the regression term.
likely to occur with a delay. We however consider contemporaneous Consequently, the two stage estimation is done using the following
values of FDI and therefore this reverse effect is not a significant issue in specifications:
our model.
Stage 1: Youth employment rate = θ0 + θ1 (Instrumental Variables ) + θ2 X ′ + ξ (2)
Enders and Sandler [53] consider this issue and find that terrorism
reduced FDI in some relatively weak European countries such as Spain
and Greece but do not find this for stronger countries such as France, Specifically, we first estimate the extent of the youth employment
Germany and Italy. They provide several possible reasons explaining rate (that is, the youth employment-to-population ratio) in a country as
why terrorism may not have a perceptible effect on FDI in stronger a function of all other exogenous variables (X) that are included in
⎧ ⎛ ⎞
n ⎪
⎪ ⎜ α exp(Xi′γ1 + Youth employment rate he above equationsnts·rsion parameater and count of terrorism incidents , the fuctions·ieve it is for the student population γ2) ⎟
Stage 1:  = ∑
ˆ
yi ln⎜ ⎟
⎨
i=1 ⎪ ⎜ 1 + α exp⎛⎜Xi′γ + Youth employment rate he above equationsnts· rsion parameater and count of terrorism incidents , the fuctions·ieve it is for the student population γ2⎞⎟ ⎟⎟
⎜
ˆˆ
⎪ ⎝ ⎠⎠
⎩ ⎝
1

⎫
⎪
1 ⎪
− ln⎛⎜1 + α exp⎛⎜Xi′γ1 + Youth employment rate he above equationsnts· rsion parameater and count of terrorism incidents , the fuctions· ieve it is for the student population γ2⎞⎟⎞⎟ + lnΓ⎛⎜yi + ⎞⎟ − lnΓ(yi + 1) − lnΓ⎛⎜ ⎞⎟
1 ˆˆˆ 1
α ⎝ ⎝ ⎠⎠ ⎝ α⎠ ⎝ α ⎠⎬
⎪
⎪
⎭
(3)

countries. First, terrorists tend to target businesses from certain coun-
tries only. A stronger country will have a better diversified pool of
foreign investors and this insulates a large part of FDI flows from these
attacks. Second, stronger countries can provide better security to for-
eign investors. Third, a strong country will have a larger volume of FDI
than a weak country. Hence, an outflow of the same size (in nominal
terms) will be less perceptible in the strong country. Clearly, terrorism
does not necessarily lead to a perceptible reduction in FDI in each and
every country.
In the MENAP countries, we do not find any strong and stable re-
lationship between FDI and terrorism of either kind. Ultimately, the
strength of the relationship is an empirical question. We find that FDI
satisfies the tests of a valid instrument.
Some other instruments along the lines of FDI and military ex-
penditure in terms of their job creation abilities for youth that are worth
considering include value addition from manufacturing, number of new
registered businesses during a year and government spending on edu-
cation as a proportion of total expenditure. We evaluated these in-
struments in our efforts to resolve endogeneity concerns. However,
while their inclusion as instruments similar to FDI and military ex-
penditure can be justified logically, our empirical analysis indicated
that they failed as valid instruments. Specifically, in a test of over-
identifying restrictions we were able to reject the null hypothesis (p-
value<.1) that the instruments are uncorrelated with the error term of
the dependent variable. This indicates that the instruments are not
equation (1) as well as the instrumental variables. For the case of do-
mestic terrorism, we use lagged under-five mortality rate and foreign
direct investment as instruments, while for the case of transnational
terrorism we also use military expenditure as a third instrument in
addition to these two.
Next we obtain the residuals or errors from the first stage Ordinary
Least Squares regression. The residuals are used to adjust for en-
dogeneity in the count model used in the second stage. In the second
stage, we employ negative binomial regression including the predicted
value of youth employment but excluding the instrumental variables.
Standard errors are based on robust estimators. The IV estimates are
presented in Table 4 columns (5) and (6) for domestic and transnational
terrorism respectively. The results from the first stage are included in
Appendix Table A4.
In a test of overidentifying restrictions we were not able to reject the
null hypothesis (p-value>.1) that the instruments are uncorrelated
with the error term of the dependent variable. This indicates that the
instruments are exogenous and therefore valid. Our post estimation test
of endogeneity indicated that the residual or error term from the first
stage regression is positive and statistically insignificant (p-value>.1)
and thus we cannot reject the null hypothesis that youth employment
rate is exogenous.
In the case of transnational terrorism, we find from the first stage of
the regression that military expenditure, under-five mortality rate and
foreign direct investment are statistically significant instruments (see

9
A. Bagchi, J.A. Paul Socio-Economic Planning Sciences xxx (xxxx) xxx–xxx

Appendix Table A4). Further, in the case of domestic terrorism, under-
five mortality rate is a statistically significant instrument (See Appendix
Table A4). Also we find from the first stage of the regression that
military expenditure has a negative relationship with youth employ-
ment whereas lagged under-five mortality rate and foreign direct in-
vestment each have a positive relationship with youth employment.
This validates our hypotheses regarding the above instruments. Overall,
the statistics noted in the endogenous specifications are consistent with
that expected in latent class endogenous models.
We find that in the endogenous model, domestic terrorism decreases
with the youth employment rate and this is statistically significant. The
magnitude implies that a one unit increase in the youth employment
rate will lead to 20.26 less domestic attacks per country year (when
evaluated at the average number of attacks). Regarding the other in-
dependent variables, the results remain qualitatively the same as in the
exogenous model.
We also find that youth employment rate tends to decrease trans-
national terrorism and this effect is statistically insignificant. The
magnitude implies that a one unit increase in the youth employment
rate will lead to 0.11 less transnational attacks per country year (when
evaluated at the average number of attacks).
5.5. Alternate specification: lagged variables
As mentioned above, the problem of reverse causality is a possible
concern. It means that while youth employment may reduce terrorism,

Table 4
Modeling specifications.

Dependent variable→ Domestic terrorism Domestic terrorism Transnational Transnational Domestic terrorism Transnational terrorism
incidents incidents terrorism incidents terrorism incidents incidents incidents

(Full Specification) (Preferred (Full Specification) (Preferred (IV) (IV)

Specification) Specification)

Independent variables↓ (1) (2) (3) (4) (5) (6)

Youth employment-to- −0.114*** −0.114*** −0.020 −0.009 −0.365** −0.016

population ratio (0.044) (0.043) (0.029) (0.020) (0.169) (0.068)
Lack of Political rights −0.045 0.077 0.047 −0.193
(0.253) (0.236) (0.027) (0.298)
Polity 0.023 0.002 0.286* 0.061*** 0.024 0.051*
(0.037) (0.033) (0.184) (0.020) (0.036) (0.028)
Voice accountability 0.642 0.896 1.827*** 1.738*** −0.971 1.311***
(0.958) (0.956) (0.628) (0.384) (1.512) (0.455)
Political stability −2.111*** −2.056*** −1.264*** −1.196*** −0.654 −1.238***
(0.293) (0.295) (0.232) (0.170) (1.057) (0.357)
Regulatory quality 0.788 0.623 −1.184*** −1.098*** 1.877** −1.092***
(0.472) (0.461) (0.343) (0.277) (0.875) (0.383)
Lack of Press freedom −0.044** −0.040** −0.023 −0.060**
(0.019) (0.020) (0.014) (0.022)
Religious fractionalization 6.349*** 5.874*** −0.285 0.014 9.889*** −0.632
(1.551) (1.500) (1.173) (0.892) (3.660) (1.587)
Ethnic fractionalization −0.514 0.062 1.256 0.552 2.867 0.960
(2.383) (2.182) (1.164) (0.756) (3.432) (1.365)
Linguistic fractionalization 3.464* 4.543*** 1.293 1.217** 8.811*** 1.602
(1.775) (1.550) (1.316) (0.619) (3.284) (1.302)
Ln (Country area) 1.245*** 0.954*** −0.520 −0.440** 0.793** −0.578**
(0.432) (0.367) (0.323) (0.188) (0.353) (0.243)
Urbanization rate −0.005 −0.010 −0.017* −0.022*** −0.065* −0.019
(0.015) (0.014) (0.010) (0.007) (0.038) (0.015)
Tropical area 6.753 8.874 0.084 13.816
(6.306) (5.850) (4.016) (8.603)
Ln (Elevation) 0.687 −0.163 −0.063 −0.320
(0.575) (0.410) (0.270) (0.395)
Natural Resources −0.029*** −0.033*** 0.016* 0.014** −0.038*** 0.017**
(0.011) (0.010) (0.009) (0.006) (0.012) (0.007)
Dummy-Kuwait 1.833 0.371 −2.182* −1.958** −1.244 −2.409***
(2.189) (1.745) (1.219) (0.892) (2.144) (1.099)
Dummy-Oman −23.290 −27.683 −17.551 −19.792 −38.689*** −15.528***
(3591.482) (8837.113) (8871.587) (24397.680) (9.443) (1.216)
Dummy-Qatar −7.647 −13.701 −2.843 −2.002 −11.380*** −2.977
(8969.756) (31592.510) (2.542) (1.661) (3.591) (2.282)
Dummy-Saudi Arabia −4.436** −4.624** 4.391*** 4.206*** −8.972** 4.293***
(1.974) (1.939) (1.390) (0.636) (3.830) (0.910)
Intercept −15.926 −8.713 8.788 7.604 2.876 10.318
(7.205) (4.064) (5.157) (3.459) (8.214) (5.305)
Alpha 0.747 0.769 0.415 0.445 0.752 0.356
LR (Alpha) 3541.84 3714.98 159.26 262.29 2751.676 213.5173
Vuong test [ZINB vs. NB]: 0.39 0.05 0.68 0.54 0.00 0.28
Number of observations 170 170 170 201 168 186
Log pseudo-likelihood −401.905 −402.609 −315.438 –387.985 −396.840 −355.596
AIC 4.975 4.972 3.958 4.040 4.974 4.028
BIC −190.403 −194.130 −117.234 −178.001 −186.606 −163.109

Note: Robust absolute z-statistics, adjusted over countries, are shown in parentheses.
* Significance level at 10%.
** Significance level at 5%.
*** Significance level at 1%.

10
A. Bagchi, J.A. Paul Socio-Economic Planning Sciences xxx (xxxx) xxx–xxx

Table 5 measures of press freedom, religious, linguistic and ethnic fractionali-

Domestic and transnational terrorism and country characteristics in MENAP countries zation, country geography, etc., included in Table 5. Table 5 shows
(Using lagged variables).
results for domestic and transnational terrorism incidents. In this case
Dependent variable→ Domestic terrorism Transnational terrorism also, youth unemployment tends to increase the number of domestic
Independent variables↓ incidents incidents terrorist incidents and this is statistically significant. Consistent with
the literature, we find that current domestic terrorist attacks tend to
Domestic Incidents 0.003***
t-1
depend positively upon the number of past attacks. However, past at-
(0.001)
Transnational Incidents 0.012 tacks do not have a statistically significant impact on transnational
t-1
(0.009) terrorism. Hence, the incidence of domestic terrorism is strongly history
Youth employment-to- −0.101*** −0.027 dependent.
population ratio t-1 (0.036) (0.026)
Lack of Political rights t-1 0.243
(0.206)
Polity t-1 0.017 0.065** 6. Discussion and concluding remarks
(0.025) (0.025)
Voice accountability t-1 0.222 0.830* One major area of debate is whether or not economic factors have
(0.764) (0.493)
any role to play in determining the extent of terrorism. We show that
Political stability t-1 −1.376*** −0.988***
(0.266) (0.220) youth unemployment can explain incidences of domestic terrorism (but
Regulatory quality t-1 0.679 −0.553 not transnational terrorism) in the MENAP countries in the recent past.
(0.416) (0.351) There can be a concern that youth unemployment is an endogenous
Lack of Press freedom t-1 −0.015 variable. In order to address this, we instrument for youth unemploy-
(0.017)
ment using (i) Military Expenditure, (ii) FDI, and (iii) Lagged Under-
Religious fractionalization t-1 3.702** 0.451
(1.604) (1.066) five Mortality Rate. As hypothesized, we find that foreign direct in-
Ethnic fractionalization t-1 0.797 1.227 vestment and under-five mortality rate does indeed reduce youth un-
(1.463) (0.852) employment whereas military expenditure increases it.
Linguistic fractionalization 6.748*** 1.456**
t-1
Our main finding is that youth unemployment increases domestic
(1.514) (0.693)
Ln (Country area) −0.111 −0.439**
terrorism and we cannot reject the null hypothesis that youth un-
t-1
(0.426) (0.210) employment is exogenous. Additionally, we find some other determi-
Urbanization rate t-1 −0.025** −0.021** nants of terrorism. Our results indicate that domestic terrorism in-
(0.011) (0.009) creases with religious and linguistic fractionalization, freedom of the
Tropical area t-1 15.441**
press and area. A free press serves as an advertising platform to terrorist
(4.733)
Ln (Elevation) t-1 −0.108 groups and therefore perversely tends to increase domestic terrorism.
(0.334) Regarding transnational terrorism, we find that it tends to decrease
Natural Resources t-1 −0.008 0.020** with political stability and regulatory quality. We also find that trans-
(0.011) (0.008)
national attacks increase with the quality of the democracy.
Dummy-Kuwait −1.663 −2.792***
(1.188) (0.949)
The amount of natural resources is also an important factor related
Dummy-Oman −34.389 −16.466 to terrorism in the MENAP countries. We find that the amount of nat-
(8942.187) (3172.573) ural resources tends to be negatively associated with domestic terrorism
Dummy-Qatar −18.156 −2.515 but positively associated with transnational terrorism.
(16827.160) (1.835)
Our empirical findings have several policy implications. The posi-
Dummy-Saudi Arabia −5.822*** 3.414***
(1.676) (0.793) tive relationship between youth unemployment and domestic terrorism
Intercept 1.925 7.499 supports studies that highlight the fact that discontent caused due to
(4.584) (3.785) relative deprivation stemming from lack of employment opportunities
Alpha 0.627 0.648 can serve as a motivation for violent activities including acts of terror.
LR (Alpha) 1243.18 327.31
Vuong test [ZINB vs. NB]: 0.47 0.00
This provides motivation for governments of the MENAP nations to take
Number of observations 170 187 steps that enhance economic growth, build markets that create jobs to
Log pseudo-likelihood −426.404 −373.064 absorb the educated force and use them most efficiently and this reduce
AIC 5.264 4.193 likelihood of skill degradation, improve education systems so youth
BIC −209.358 −129.793
have requisite skills to take on the meaningful employment that has a
positive impact on their personal lives and families. These in turn will
help reduce grievances in the society, the root cause of terrorism. The
it is also possible that terrorism may impact youth employment. In negative relationship between transnational terrorism and political
order to address this issue, one possible way out is to use lagged values stability provides support to studies that have demonstrated that stable
of the independent variables. Such an approach has been used in governments play a pivotal role in combating the problem of transna-
Tavares [35], Freytag et al. [54] and Kis-Katos and Schulze [34]. tional terrorism.
Therefore, in our alternate specification, we use lagged values of the In the future extensions of this research, we aim to investigate the
independent variables. impact of role of modernization such as female political empowerment,
It has also been pointed out in Freytag et al. [54] and Kis-Katos and female education, etc., on domestic and transnational terrorism. There
Schulze [34] that the number of incidents of terror in a country in a is always a scope for debate regarding the suitability of instruments. We
given year is history-dependent. Therefore, we also include the number have provided logical arguments as well as statistical tests in support of
of terror incidents in the immediate past as an additional explanatory our instruments. In particular, we describe why our instruments are
variable. We apply the following model specification for our estimation. valid given the characteristics and recent history of the MENAP coun-

Terrorism Incidentst = exp(β0 + β1 Terrorism Incidentst − 1

+ β2 Youth Employmentt − 1 + β3 Xt − 1′ + ϑ)
The vector X includes potential predictors of terrorism such as
tries. These instruments may not necessarily work in other contexts,
such as in a study of terrorism in Europe. Therefore, there is a need to
(4) explore other possible instruments of youth unemployment that may be
valid more generally. We plan to follow up on this aspect in future
work.

11
A. Bagchi, J.A. Paul
References
[1] International Monetary Fund. Changes to the database: world economic Outlook
database. http://www.imf.org/external/pubs/ft/weo/data/changes.htm; 2013.
[2] Gassebner M, Luechinger S. Lock, stock, and barrel: a comprehensive assessment of
the determinants of terror. Publ Choice 2011;149(3–4):235–61.
[3] Krieger T, Meierrieks D. What causes terrorism? Publ Choice 2011;147(1–2):3–27.
[4] Urdal H. A clash of generations? Youth bulges and political violence. Int Stud Q
2006;50(3):607–29.
[5] Feldmann AE, Perala M. Reassessing the causes of nongovernmental terrorism in
Latin America. Lat Am Polit Soc 2004;46(2):101–32.
[6] Piazza JA. Rooted in Poverty?: Terrorism, poor economic development, and social
cleavages 1. Terrorism Polit Violence 2006;18(1):159–77.
[7] Gurr TR. Why men rebel. Princeton, NJ: Center of International Studies, Princeton
UP; 1970.
[8] MacCulloch R. Income inequality and the taste for revolution. J Law Econ
2005;48(1):93–123.
[9] Schomaker R. Youth bulges, poor institutional quality and missing migration op-
portunities - triggers of and potential counter-measures for terrorism in MENA. Top
Middle East Afr Econ 2013;15:116–40.
[10] Becker G. Crime and punishment: an economic approach. J Polit Econ
1968;76(2):169–217.
[11] International Labour Organization. Key indicators of the labour market. Geneva:
International Labour Office; 2014.
[12] Cincotta R, Engelman R, Anastasion D. The security demographic: population and
civil conflict after the cold war. Washington: Population Action International; 2003.
[13] Urdal H. The devil in the demographics: the effect of youth bulges on domestic
armed conflict, 1950-2000. Social development papers, conflict prevention and
reconstruction. Washington: World Bank.US Energy Information Administration;
2004. No. 14.
[14] Dhillon N, Yousef T, editors. Generation in waiting: the unfulfilled promise of young
people in the Middle East. Washington DC: Brookings Institution Press; 2009.
[15] World Bank. The road not traveled: education reform in the Middle East and North
Africa. 2008. Washington D.C.
[16] Makdisi SA, Elbadawi I. Democracy in the Arab world: explaining the deficit. New
York: Routledge; 2011.
[17] Berrebi C. Evidence about the link between education, poverty and terrorism among
Palestinians. Peace Econ Peace Sci Publ Pol 2007;13(1):1–36.
[18] Tajfel H, Turner JC. An integrative theory of intergroup conflict. In: Worchel S,
Austin WG, editors. Psychology of intergroup relations. Monterey, CA: Brooks/Cole;
1979. p. 33–47.
[19] Tajfel H, Turner JC. The social identity theory of intergroup behaviour. In: Worchel
S, Austin WG, editors. Psychology of intergroup relations. Monterey, CA: Brooks/
Cole; 1986. p. 7–24.
[20] Shayo M. A model of social identity with an application to political economy: na-
tion, class, and redistribution. Am Polit Sci Rev 2009;103(2):147–74.
[21] Sambanis N, Shayo M. Social identification and ethnic conflict. Am Polit Sci Rev
2013;107(2):294–325.
[22] Li Q, Schaub D. Economic globalization and transnational terrorism: a pooled time-
series analysis. J Conflict Resolut 2004;48(2):230–58.
[23] Blomberg SB, Hess G. From (no) butter to guns? Understanding the economic role in
transnational terrorism. In: Keefer Philip, Loayza Norman, editors. Terrorism,
economic development, and political openness. New York: Cambridge UP; 2008. p.
83–115.
[24] Eyerman J. Terrorism and democratic states: soft targets or accessible systems. Int
Interact 1998;24(2):151–70.
[25] Li Q. Does democracy promote or reduce transnational terrorist incidents? J
Conflict Resolut 2005;49(2):278–97.
[26] Testas A. Determinants of terrorism in the Muslim world: an empirical cross-sec-
tional analysis. Terrorism Polit Violence 2004;16(2):253–73.
[27] Krueger A, Laitin D. Kto Kogo? A cross-country study of the origins and targets of
terrorism. In: Keefer Philip, Loayza Norman, editors. Terrorism, economic devel-
opment, and political openness. New York: Cambridge UP; 2008. p. 148–73.
[28] Krueger A, Malečková J. Education, poverty, and terrorism: is there a causal con-
nection? J Econ Perspect 2003;17(4):119–44.
[29] Abadie A. Poverty, political freedom, and the roots of terrorism. Am Econ Rev
2006;95:50–6.
[30] Burgoon B. On welfare and terror: social welfare policies and political-economic
roots of terrorism. J Conflict Resolut 2006;50(2):176–203.
[31] Oyefusi A. Oil, youths, and civil unrest in Nigeria's Delta. Conflict Manag Peace Sci
2010;27(4):326–46.
[32] Caruso R, Schneider F. The socio-economic determinants of terrorism and political
violence in Western Europe (1994-2007). Eur J Polit Econ 2011;27(Supplement
1):S37–49.
[33] Caruso R, Gavrilova E. Youth unemployment, terrorism and political Violence,
evidence from the Israeli/Palestinian conflict. Peace Econ Peace Sci Publ Pol
2012;18(2):1–35.
12
Socio-Economic Planning Sciences xxx (xxxx) xxx–xxx
[34] Kis-Katos K, Liebert H, Schulze GG. On the origin of domestic and international
terrorism. Eur J Polit Econ 2011;27(Supplement 1):S17–36.
[35] Tavares J. The open society assesses its enemies: shocks, disasters and terrorist
attacks. J Monetary Econ 2004;51(5):1039–70.
[36] Kurrild-Klitgaard P, Justesen MK, Klemmensen R. The political economy of
freedom, democracy and transnational terrorism. Publ Choice 2006;128(1/
2):289–315.
[37] Erceg C, Levin A. Labor force participation and monetary policy in the wake of the
great recession International Monetary Fund Working Paper 2013.
[38] Bandyopadhyay S, Younas J. Poverty, political freedom, and the roots of terrorism
in developing countries: an empirical assessment. Econ Lett 2011;112(2):171–5.
[39] Nacos B. Terrorism and the media. NY: Columbia University Press; 1994.
[40] Alesina A, Devleeschauwer A, Easterly W, Kurlat S, Wacziarg R. Fractionalization. J
Econ Growth 2003;8:155–94.
[41] Gaibulloev K, Sandler T. Determinants of the demise of terrorist organizations.
South Econ J 2013;79(4):774–92.
[42] OECD. Terrorism, corruption and the criminal exploitation of natural resources.
https://www.oecd.org/investment/Terrorism-corruption-criminal-exploitation-
natural-resources-2016.pdf; 2016, Accessed date: 5 September 2017.
[43] Hilbe Joseph. Negative binomial regression. second ed. Cambridge, UK: Cambridge
University Press; 2011.
[44] Piazza JA. Incubators of terror: do failed and failing states promote transnational
terrorism? Int Stud Q 2008;52(3):469–88.
[45] Landau D. The impact of military expenditures on economic growth in the less
developed countries. Defence Peace Econ 1994;5:205–20.
[46] Aizenman J, Glick R. Military expenditure, threats, and growth. J Int Trade Econ
Dev 2006;15(2):129–55.
[47] Abu-Bader S, Abu-Qarn AS. Government expenditures, military spending and eco-
nomic growth: causality evidence from Egypt, Israel, and Syria. J Pol Model
2003;25:567–83.
[48] Abdelfattah YM, Abu-Qarn AS, Dunne JP, Zaher S. The demand for military
spending in Egypt. Defence Peace Econ 2014;25(3):231–45.
[49] Population Reference Bureau. Population trends and challenges in Middle East and
North Africa. www.pbr.org; 2001, Accessed date: 27 August 2013.
[50] Barro R, Lee J. A new data set of educational attainment in the world, 1950-2010. J
Dev Econ 2010;104:184–98.
[51] Javorcik B. Does FDI bring good jobs to host countries? 2012 Background Paper for
the World Development Report 2013.
[52] Dhillon N, Salehi-Isfahani D, Dyer P, Yousef T, Fahmy A, Kraetsch M. Missed by the
boom, hurt by the bust: making markets work for young people in the Middle East.
Dubai School of Government & The Wolfensohn Center for Development at
Brookings; 2008.
[53] Enders W, Sandler T. Terrorism and foreign direct investment in Spain and Greece.
Kyklos 1996;49(3):331–52.
[54] Freytag A, Kruger JJ, Meierrieks D, Schneider F. The origins of terrorism: cross-
country estimates of socio-economic determinants of terrorism. Eur J Polit Econ
2011;27(Supplement 1):S5–16.
[57] Global Terrorism Database. National consortium for the study of terrorism and
responses to terrorism. http://www.start.umd.edu/gtd/; 2014.
[66] Rummel R. Is collective violence correlated with social pluralism? J Peace Res
1997;34(2):163–75.
[68] World Bank. Middle East and North Africa. 2013http://web.worldbank.org/
WBSITE/EXTERNAL/COUNTRIES/MENAEXT/0,
menuPK:247619∼pagePK:146748∼piPK:146812∼theSitePK:256299,00.html,
Accessed date: 15 January 2013.
[69] Belasco A. The cost of Iraq, Afghanistan, and other global war on terror operations
since 9/11. 2014. Retrieved August 21, 2015, from https://www.fas.org/sgp/crs/
natsec/RL33110.pdf.
Aniruddha Bagchi is Professor of Economics at the Coles College of Business, Kennesaw
State University, GA. He received his Ph.D. in Economics from Vanderbilt University.
Aniruddha's research is in two areas- Industrial Economics and Defense Economics. As an
Industrial Economist, he is interested in examining the processes and institutions that
encourage innovative activity. He has also written on the effect of regulation and eco-
nomic outcomes. His work has appeared in Canadian Journal of Economics, Southern
Economic Journal and International Journal of Production Economics. As a Defense
Economist, he is interested in determining optimal ways of preventing terrorist strikes.
Jomon A. Paul is Professor of Quantitative Analysis and Director of Research at Coles
College of Business, Kennesaw State University. He received his Ph.D. in Industrial
Engineering from University at Buffalo. His primary research interests include application
of operations research, simulation and statistical modeling to disaster planning, health-
care, and transportation domains. His research has appeared in Operations Research,
European Journal of Operational Research, International Journal of Production Economics,
Transportation Research Part E, Journal of Business Ethics, Journal of the Operational
Research Society, Annals of Operations Research, among several others. He is on the
Editorial Board for International Journal of Information Systems and Social Change.