Content uploaded by Daniel Meierrieks
Author content
All content in this area was uploaded by Daniel Meierrieks on Oct 28, 2017
Content may be subject to copyright.
Hit and (They Will) Run:
The Impact of Terrorism on Migration
Forthcoming in: Economics Letters
Axel Dreher∗Tim Krieger†Daniel Meierrieks‡
April 7, 2011
Abstract
We analyze the influence of terrorism on migration for 152 countries during 1976-2000. We
find robust evidence that terrorism is among the ‘push factors’ of skilled migration, whereas it
is not robustly associated with average migration.
Keywords: terrorism, skilled migration, brain drain, panel estimation
JEL Classification: D74, F22
∗Corresponding Author. Heidelberg University, Alfred-Weber-Institut for Economics, Bergheimer Str. 58, 69115
Heidelberg, Germany, KOF Swiss Economic Institute, Switzerland, University of Goettingen, IZA and CESifo,
Germany. Ph.: +49 (0)6221 54 2921, fax: +49 (0)6221 54 3578, e-mail: mail@axel-dreher.de.
†University of Paderborn, Department of Economics, Warburger Straße 100, 33098 Paderborn, Germany. Ph.:
+49-(0)5251-60-2117, fax: +49-(0)5251-60-5005, e-mail: tim.krieger@notes.uni-paderborn.de.
‡University of Paderborn, Department of Economics, Warburger Straße 100, 33098 Paderborn, Germany. Ph.:
+49-(0)5251-60-2115, fax: +49-(0)5251-60-3540, e-mail: daniel.meierrieks@notes.uni-paderborn.de.
1
1 Introduction
Several recent studies suggest that terrorism constrains economic development by, e.g., impeding
investment and tourism and diverting the international flow of goods and capital (e.g., Neumayer,
2004; Abadie and Gardeazabal, 2008; Gaibulloev and Sandler, 2008; Sandler and Enders, 2008).
Beyond its negative economic consequences, terrorism also produces social costs that are reflected
in, e.g., reduced life satisfaction (Frey et al., 2009) and political costs in the form of, e.g., government
instability (Gassebner et al., 2007, 2011).
These negative effects of terrorist activity – in addition to the direct threat to one’s life – tend to
worsen individual living and working conditions, so that they ought to impact individual migration
decisions. Here, the desire and possibility to emigrate are expected to differ depending on the indi-
vidual levels of education, given that the (direct and opportunity) costs of education are substantial.
In times of terror, the returns to education decrease by, e.g., increasing socioeconomic insecurity
and constraining entrepreneurial activity. From a skilled individual’s perspective the remaining skill
premium may be considered too low to recoup the costs of a previous high-level education. Given
that human-capital investment is irreversible, we therefore expect skilled workers to be particularly
keen to emigrate in order to protect their human capital from devaluation and to yield a sufficient
return to education, so that a ‘brain drain’ may indeed occur. This outcome is reinforced by the
fact that potential host countries increasingly resort to quality-selective immigration policies and
prefer skilled over medium and low skilled immigrants (e.g., Docquier et al. 2007), thus making
it relatively easy only for skilled workers to leave their terror-ridden home countries. By contrast,
for individuals with average or low levels of human capital terrorist activity in their home country
may result in additional costs and restrictions to emigration, e.g., due to increasing travel costs or
the introduction of specific restrictions for immigrants from terror-rich countries in the destination
country.
Based on these considerations, our hypothesis is that terrorism is among the drivers of skilled
migration, while its effect on average migration may be less clear. As discussed in Eggert et al.
(2010), previous studies on the ‘brain drain’ tend to emphasize the role of ‘pull factors’ of individual
migration and education decisions, especially with respect to income differentials between home
and target countries. By contrast, our approach explicitly focuses on a potential ‘push factor’. In
this sense, our analysis complements Docquier et al. (2007) who find that political instability is
positively related to skilled migration. We provide panel evidence on the determinants of skilled and
average migration, also properly taking account of serial correlation, heterogeneity and endogeneity.
Given that political instability and terror are closely related (Campos and Gassebner 2009), we also
control for instability. To preview our findings, independent of the statistical methods we detect
a robust positive relationship between terrorism and skilled migration, controlling for a variety of
variables. By contrast, terrorism is not robustly associated with average migration, indicating that
the effect of terrorism on migration depends on individual levels of education.
In the next section, we introduce our methodology and the data. Section 3 provides our empirical
results, while Section 4 concludes.
2 Data and Method
We compile data for 152 countries for the 1976-2000 period. Our main dependent variable (skilled
migration) is defined as the (estimated) ratio of the number of skilled emigrants that are 25 or
2
older to six major receiving countries (USA, UK, Germany, France, Canada and Australia) to the
total number of skilled natives aged 25 or older.1To assess whether the effects of terrorism depend
on education levels, we use an alternative dependent variable (average migration) which is defined
as the ratio of the total number of emigrants aged 25 or older to these six countries to the total
number of natives aged 25 or older. The data are drawn from Defoort (2008).2
Raw data for the construction of our main explanatory variable (terror ) is from the Global Terrorism
Database (LaFree and Dugan, 2007). We use information on the total number of terrorist attacks
in the country of interest and the victims (i.e., the number of the killed and wounded) of these
attacks to construct a population-weighted terrorism index.3In some specifications, we use further
indicators of political instability to assess the robustness of the effect of terrorism on skilled and
average migration.4
As control variables, in our baseline specification we consider the effect of per capita income (gdp),
population size (pop) and trade openness (trade), where all series are logged and drawn from the
PENN World Table (Heston et al., 2009). We also take into account the sending country’s level of
political development (democracy) and its square (democracy2 ) from the Polity4 Project (Marshall
and Jaggers, 2008). We expect a high level of socioeconomic development to be negatively related
to migration because it reflects an adequate return to education that makes migration less likely.
Population size is also anticipated to be negatively associated with migration, given that internal
migration is likely to increase with country size, making international migration less attractive.
Trade openness is anticipated to be positively related to migration, indicating, e.g., a country’s
travel restrictions and its international socioeconomic integration. Finally, we consider a non-linear
relationship between a country’s level of political development and skilled migration. Authoritar-
ian governments may easily impose migration restrictions (impeding migration), while repression
may also foster politically motivated flights (increasing migration). For democratic countries the
situation is expected to be the other way around, so that the link between political openness and
migration may be non-monotonic. In some specifications, we also control for the effect of (logged)
government size (gov ) and the rate of economic growth (growth) on migration. Both data series
are drawn from the PENN World Table.5
Initial tests indicate that serial correlation and heteroscedasticity may bias our statistical analysis.
We thus run a series of feasible generalized least squares (FGLS) regressions with a common AR(1)
process, heteroskedasticity-robust standard errors and the inclusion of country-specific effects to
1Skilled emigrants are those with a post-secondary certificate (Defoort, 2008).
2The dependent variables are available for five points in time (1980, 1985, 1990, 1995, 2000). We construct
five-year averages of our explanatory variables to estimate their effect on migration.
3We adjust the index for population size to consider potential scale effects, where terrorism is expected to be
more threatening for countries with smaller populations. Formally, the index for country iin year tis defined as:
terrori,t = ln(e+attacksi,t
populationi,t +victimsi,t
populationi,t ).
4Specifically, we assess the independent influence of adverse regime changes (regime change ) and
genocides (genocide) on migration, where data are provided by the Political Instability Task Force
(http://globalpolicy.gmu.edu/pitf/index.htm). We also use the (logged) number of battle deaths in civil wars (civil
war) to indicate incidences of civil war, either using a low (at least 25 battle deaths per year) or high (at least 1000
battle deaths per year) threshold. The civil war data are from Lacina and Gleditsch (2005).
5Note that we generally include country-fixed effects in our empirical analysis to account for certain time-invariant
factors that may also be considered as potential determinants (‘pull factors’) of migration. When we exclude the
fixed effects in our estimations to consider these time-invariant factors, we find that a variety of them (distance,
common languages, colonial ties, landlocked, resource endowments and religious fractionalization) are associated
with migration. Our main empirical findings are not affected by the exclusion of the country-specific effects.
3
analyze the effect of terrorism on migration. We also acknowledge that reverse causation may
be an issue. For instance, skilled migration may exacerbate socioeconomic and political crises in
sending countries and consequently amplify terrorist activity. Thus, we furthermore run a series of
system-GMM estimations that account for potential endogeneity to provide more robust evidence.
3 Empirical Results
The results from the FGLS estimations of the effect of terrorism on skilled migration are reported
in Table 1. They indicate that skilled migration is more common in small, poor and semi-open
countries. With respect to our main variable of interest, we find that terrorist activity is indeed
robustly and positively associated with skilled migration. This finding survives the inclusion of
further controls, in particular those indicating other forms of political instability.6Our analysis thus
indicates that terrorism makes it less attractive for the highly skilled to stay in their home countries,
potentially due to diminishing returns to education. This may be a consequence of the constraining
effect of terrorism on socioeconomic activity, opportunities and security. Quantitatively, doubling
the incidence of terror according to our index increases the share of skilled emigrants in total
migration by about 0.01. This amounts to an elasticity of about 0.08. While there are no empirical
studies that provide results that can be directly compared to ours, our findings are qualitatively in
line with earlier studies arguing that political instability and violent conflict are among the drivers
of (skilled) international migration (e.g., Hatton and Williamson, 2003; Docquier et al., 2007).
— Table 1 here —
The results from the system-GMM estimations that account for potential reverse causation are
reported in Table 2. Again, we find that skilled migration is associated with small country size
and low levels of socioeconomic development. While we now also find that higher levels of trade
openness lead to more migration, we find no evidence of a non-linear relationship between political
development and the brain drain. Rather, more politically-open countries are found to experience
stronger skilled migration. Considering the impact of terrorism, we again find that stronger terrorist
activity leads to an increase in skilled migration. Once again, this finding is robust to the inclusion
of other indicators of political instability and time-invariant factors. The system-GMM findings thus
reinforce those from the FGLS estimations and provide additional support for our main hypothesis.
— Table 2 here —
Finally, in Table 3 we focus on the effect of terrorism on average migration and present assorted
FGLS and system-GMM findings. With respect to the controls, we find that average migration is
more common in small, less developed and democratic countries. As concerns our main variable of
interest, we find that terrorism is not robustly associated with average emigration.7Our empirical
6According to Campos and Gassebner (2009) there may also be indirect linkages at work, e.g., from terrorism
via general political instability to skilled migration. However, the qualitative and quantitative effect of terrorism
remains similar when we control for other types of instability. Our findings are also qualitatively unchanged when
we run pooled OLS estimations or regressions with Newey-West standard errors.
7The FGLS results provide some evidence that average migration and terrorism are negatively related. Potentially,
this may indicate that terrorism increases the costs of migration by, e.g., making travelling more difficult, so that
the wealthiest (i.e., the skilled) are most likely to migrate. However, the system-GMM results provide no evidence
of a robust association between terrorism and average migration, so that the FGLS results may be driven by, e.g.,
reverse causation.
4
findings thus indicate that the effect of terrorism on migration depends on individual levels of
education. In particular the highly skilled have the incentives and means to migrate, while they
are also preferred to the less skilled by their host countries.
— Table 3 here —
4 Conclusion
We empirically assessed the influence of terrorism on skilled migration for 152 countries over the
1976-2000 period. We found robust evidence that terrorism increases skilled emigration, suggesting
that terrorism affects the cost-benefit considerations of the highly educated in ways that make
emigration more attractive. We found no robust evidence that average emigration is related to
terrorism, which indicates that the effect of terrorism on migration depends on the level of education.
References
Abadie, A., Gardeazabal, J., 2008. Terrorism and the world economy. Eur. Econ. Rev. 52, 1–27.
Campos, N., Gassebner, M., 2009. International terrorism, political instability and the escalation
effect. CEPR Discussion Paper No. 7226.
Defoort, C., 2008. Long-term trends in international migration: An analysis of the six main
receiving countries. Population 63, 285–317.
Docquier, F., Lohest, O., Marfouk, A., 2007. Brain drain in developing countries. World Bank
Econ. Rev. 21, 193–218.
Eggert, W., Krieger, T., Meier, V., 2010. Education, unemployment and migration. J. Publ.
Econ. 94, 354–362.
Frey, B.S., Luechinger, S., Stutzer, A., 2009. The life satisfaction approach to valuing public
goods: The case of terrorism. Publ. Choice 138, 317–345.
Gaibulloev, K., Sandler, T., 2008. Growth consequences of terrorism in Western Europe. Kyklos
61, 411–424.
Gassebner, M., Jong-A-Pin, R., Mierau, J.O., 2007. Terrorism and electoral accountability: One
strike, you’re out! Econ. Lett. 100, 126–129.
Gassebner, M., Jong-A-Pin, R., Mierau, J.O., 2011. Terrorism and cabinet duration: An empirical
analysis. Int. Econ. Rev., forthcoming.
Hatton, T.J., Williamson, J.G., 2003. Demographic and economic pressure on emigration out of
Africa. Scand. J. Econ. 105, 465-485.
Heston, A., Summers, R., Aten, B., 2009. Penn World Table version 6.3. Center for International
Comparisons of Production, Income and Prices at the University of Pennsylvania.
Lacina, B., Gleditsch, N.P., 2005. Monitoring trends in global combat: A new dataset of battle
deaths. Eur. J. Popul. 21, 145–166.
5
LaFree, G., Dugan, L., 2007. Introducing the Global Terrorism Database. Terrorism and Polit.
Violence 19, 181–204.
Marshall, M.G., Jaggers, K., 2008. Polity IV Project: Political regime characteristics and transi-
tions, 1800–2008. Center for Systemic Peace and George Mason University.
Neumayer, E., 2004. The impact of political violence on tourism: Dynamic cross-sectional esti-
mation. J. Conflict Resolut. 48, 259–281.
Sandler, T., Enders, W., 2008. Economic consequences of terrorism in developed and developing
countries: An overview, in: Keefer, P., Loayza, N. (Eds.), Terrorism, economic development,
and political openness. Cambridge University Press, New York, pp. 17–47.
6
Tables
Table 1: FGLS Panel Estimates of the Effect of Terrorism on Skilled Migration
(1) (2) (3) (4) (5) (6) (7)
terror 0.011 0.011 0.009 0.009 0.010 0.011 0.013
(3.05)∗∗∗ (2.76)∗∗∗ (2.53)∗∗ (2.52)∗∗ (2.78)∗∗∗ (3.09)∗∗∗ (3.25)∗∗∗
pop -0.044 -0.039 -0.041 -0.040 -0.042 -0.041 -0.043
(7.76)∗∗∗ (7.23)∗∗∗ (7.29)∗∗∗ (6.96)∗∗∗ (7.33)∗∗∗ (6.98)∗∗∗ (7.35)∗∗∗
gdp -0.037 -0.031 -0.039 -0.038 -0.038 -0.036 -0.037
(11.64)∗∗∗ (9.48)∗∗∗ (11.45)∗∗∗ (11.31)∗∗∗ (11.78)∗∗∗ (11.08)∗∗∗ (11.54)∗∗∗
democracy 0.000 0.001 0.000 0.000 0.000 0.000 0.001
(3.38)∗∗∗ (3.62)∗∗∗ (3.18)∗∗∗ (3.18)∗∗∗ (3.29)∗∗∗ (3.09)∗∗∗ (3.99)∗∗∗
democracy2 -0.000 -0.000 -0.000 -0.000 -0.000 -0.000 -0.000
(6.03)∗∗∗ (5.72)∗∗∗ (5.90)∗∗∗ (5.84)∗∗∗ (6.17)∗∗∗ (5.41)∗∗∗ (6.20)∗∗∗
trade -0.000 0.003 0.003 0.003 0.001 -0.001 0.003
(0.07) (1.04) (0.98) (1.22) (0.55) (0.42) (1.24)
regime change 0.012
(3.39)∗∗∗
civil war (low) 0.000
(6.87)∗∗∗
civil war (high) 0.000
(1.70)∗
genocide 0.003
(0.96)
gov 0.005
(1.97)∗∗
growth 0.000
(3.42)∗∗∗
Mean VIF 1.40 1.42 1.39 1.39 1.39 1.40 1.38
No of Observations 692 692 692 692 692 692 692
No. of Countries 152 152 152 152 152 152 152
Notes: Results from FGLS models with AR(1) disturbance, fixed effects and skilled migration as dependent
variable. Absolute, robust t-values in parentheses. Period dummies included in all specifications.
∗p < 0.1; ∗∗p < 0.05;∗∗∗ p < 0.01.
7
Table 2: Dynamic Panel Estimates of the Effect of Terrorism on Skilled Migration
(1) (2) (3) (4) (5) (6) (7) (8) (9) (10)
terror 0.062 0.040 0.041 0.060 0.062 0.040 0.050 0.057 0.057 0.062
(3.43)∗∗∗ (1.97)∗∗ (1.71)∗(2.48)∗∗ (4.23)∗∗∗ (1.56) (2.12)∗∗ (3.17)∗∗∗ (3.29)∗∗∗ (3.62)∗∗∗
pop -0.016 -0.017 -0.020 -0.020 -0.019 -0.019 -0.017 -0.011 -0.018 -0.019
(1.67)∗(1.61) (1.78)∗(2.21)∗∗ (2.02)∗∗ (1.83)∗(1.73) (1.20) (1.89)∗(1.92)∗
gdp -0.043 -0.043 -0.040 -0.040 -0.040 -0.035 -0.042 -0.035 -0.040 -0.052
(4.59)∗∗∗ (4.76)∗∗∗ (4.49)∗∗∗ (4.57)∗∗∗ (4.61)∗∗∗ (2.86)∗∗∗ (4.56)∗∗∗ (3.68)∗∗∗ (4.31)∗∗∗ (5.27)∗∗∗
democracy 0.004 0.003 0.004 0.004 0.003 0.003 0.003 0.004 0.004 0.004
(2.37)∗∗ (2.02)∗∗ (2.58)∗∗∗ (2.55)∗∗ (2.66)∗∗∗ (1.81)∗(1.77)∗(3.08)∗∗∗ (2.61)∗∗∗ (2.51)∗∗
democracy2 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000
(0.78) (0.96) (0.94) (1.08) (1.29) (0.19) (1.20) (0.80) (0.73) (0.81)
trade 0.064 0.055 0.038 0.060 0.051 0.040 0.034 0.064 0.060 0.063
(3.03)∗∗∗ (1.84)∗(1.39) (3.88)∗∗∗ (2.97)∗∗∗ (1.69) (1.43) (2.97)∗∗∗ (2.88)∗∗∗ (3.01)∗∗∗
regime change 0.017
(0.55)
civil war (low) -0.000
(0.46)
civil war (high) 0.000
(0.46)
genocide -0.011
(0.87)
gov 0.16
(0.46)
growth -0.000
(0.16)
common language 0.070
(2.78)∗∗∗
religious fractionalization 0.160
(3.20)∗∗∗
landlocked -0.066
(1.72)∗
Mean VIF 1.40 1.42 1.39 1.39 1.39 1.40 1.38 1.42 1.39 1.42
No. of Instruments 23 28 28 28 28 28 28 24 24 24
M1/M2 test (Pr>z) 0.00/0.42 0.03/0.28 0.00/0.30 0.00/0.44 0.00/0.42 0.00/0.32 0.00/0.42 0.00/0.37 0.00/0.37 0.00/0.22
Hansen test (Pr>z) 0.53 0.15 0.31 0.41 0.50 0.15 0.02 0.53 0.48 0.54
No of Observations 692 692 692 692 692 692 692 692 692 692
No. of Countries 152 152 152 152 152 152 152 152 152 152
Notes: Results from two-step system-GMM models. Skil led migration as dependent variable. Absolute, robust z-values in parentheses. GMM type:
terror, democracy, democracy2, trade (further controls). IV type: gdp, pop. Period dummies included in all specifications. ∗p < 0.1;∗∗ p < 0.05;∗∗∗ p < 0.01.
8
Table 3: FGLS and System-GMM Panel Estimates of the Effect of Terrorism on Average Migration
(1) (2) (3) (4) (5) (6) (7) (8)
terror -0.001 -0.001 -0.001 -0.001 -0.019 -0.013 -0.014 -0.012
(2.10)** (2.08)** (1.30) (1.83)* (1.69)* (1.04) (1.20) (0.84)
pop -0.016 -0.017 -0.015 -0.014 -0.008 -0.007 -0.008 -0.006
(12.61)*** (13.51)*** (11.01)*** (9.59)*** (2.07)** (2.28)** (1.96)** (1.96)**
gdp -0.006 -0.008 -0.006 -0.006 0.003 0.003 0.003 0.004
(14.90)*** (19.68)*** (12.37)*** (13.52)*** (0.80) (1.06) (0.85) (1.18)
democracy 0.000 0.000 0.000 0.000 0.001 0.001 0.001 0.001
(2.38)** (2.27)** (2.48)** (1.67)* (1.27) (1.31) (1.29) (0.97)
democracy2 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000
(3.03)*** (1.92)* (2.47)** (1.15) (2.26)** (2.49)** (2.16)** (2.19)**
trade 0.001 0.000 0.001 -0.000 -0.06 -0.013 -0.016 -0.010
(2.07)** (0.66) (2.47)** (0.32) (1.92)* (1.84)* (1.85)* (1.73)*
regime change -0.001 -0.002
(1.88)* (0.49)
civil war (high) -0.000 -0.000
(5.69)*** (0.72)
gov 0.001 -0.005
(2.28)** (0.73)
Mean VIF 1.40 1.42 1.39 1.40 1.40 1.42 1.39 1.40
No. of Instruments 27 32 32 32
AB test (Pr>z) 0.01/0.03 0.01/0.02 0.01/0.02 0.01/0.03
Hansen test (Pr>z) 0.66 0.71 0.88 0.73
No of Observations 692 692 692 692 692 692 692 692
No. of Countries 152 152 152 152 152 152 152 152
Notes: Average migration as dependent variable. Results from FGLS models with AR(1) disturbance, fixed effects
in specifications (1) to (4). Absolute, robust t-values in parentheses. Results from two-step System-GMM models in
specifications (5) to (8). Absolute, robust and Windmeijer-corrected z-values in parentheses. GMM type: terror,
democracy, democracy2, trade (further controls). IV type: pop, gdp, period dummies. Separate instruments
for each period until collapsed. Period dummies included in all specifications. ∗p < 0.1; ∗∗ p < 0.05;∗∗∗ p < 0.01.
9
