Group Concentration and Violence: Does Ethnic Segregation Affect Domestic Terrorism?

Abstract

This paper examines the link between ethnic segregation and domestic terrorism. The results show that ethnic segregation has a positive and significant effect on the incidence of domestic terrorism, which indicates that countries where ethnic groups are spatially concentrated face a higher risk of suffering this type of violence. This finding is not affected by the inclusion in the analysis of different covariates that may affect both ethnic segregation and domestic terrorism. The observed relationship between the degree of spatial concentration of ethnic groups and domestic terrorism is confirmed by various robustness tests. The results also suggest that the threat of secession is an important transmission channel linking ethnic segregation and domestic terrorism.

Keywords:

• Domestic terrorism
• ethnic segregation

AMS Subject Classifications:

• D74
• J15
• K4

Acknowledgements

The author is grateful to one anonymous referee for her/his useful comments and suggestions to an earlier version of the paper. I would also like to thank the National Consortium for the Study of Terrorism and Responses to Terrorism (START) for allowing me to use the Global Terrorism Database.

Notes

No potential conflict of interest was reported by the author.

1 For a review of this literature, see Gassebner and Luechinger (2011), Kis-Katos, Liebert, and Schulze (2011) or Krieger and Meierrieks (2011).

2 Table A1 in the Appendix provides some descriptive statistics for the main variables used in the analysis.

3 The GTD is an open-source database maintained by the National Consortium for the Study of Terrorism and Responses to Terrorism (START) at the University of Maryland, which includes information on more than 150,000 terrorist attacks around the world since 1970. For further details, see http://www.start.umd.edu/gtd/.

4 The Appendix provides detailed information on the definitions and sources of the different controls used in the paper.

5 Some authors alternatively include as a control a measure of the level of ethnic tensions (Basuchoudhary and Shughart 2010; Gassebner and Luechinger 2011). Nevertheless, this is not a good idea in our context, as the degree of ethnic tensions is likely to be itself an outcome of the level of ethnic diversity and the spatial distribution of ethnic groups within a country. Consequently, the inclusion of the degree of ethnic tensions as an additional control in our analysis would make it difficult to find out the causal effect of ethnic segregation on the incidence of domestic terrorism. For further details on this issue, see Angrist and Pischke (2009, 64–68).

6 The index of ethnic fractionalization can be expressed as follows:

$$\begin{array}{c}\hfill F_i=\sum _{m=1}^M{\mathit{\pi }}_m(1-{\mathit{\pi }}_m)\end{array}$$

7 The main results of the paper remain unaffected if we alternatively employ the average value of the political rights and civil liberties ratings, or a democracy index drawn from the Polity IV Project.

8 See Alesina and Zhuravskaya (2011, 1889–1893) for further details and examples on the construction of the instrument.

9 Reverse causality might also affect some of the control variables. In order to mitigate this concern, for all time-varying covariates we use in our analysis their mean values over the period 1995–2000.

10 In view of the findings of Freytag et al. (2011) and Enders and Hoover (2012), we also investigate the possibility that the effect of GDP per capita on domestic terrorism may be non-linear. To that end, we include in the analysis the square of GDP per capita as an additional control. The results show that the coefficient of the index of ethnic segregation remains positive and statistically significant, but the estimates do not support the hypothesis of a non-linear link between GDP per capita and domestic terrorism.

11 This count of terrorist attacks and that based on the data-set compiled by Enders, Sandler, and Gaibulloev (2011) is highly correlated in our sample ($\mathit{\rho }=0.944$), although there are relatively important discrepancies between the two measures for some specific countries.

12 This indicator is more appropriate in our context than standard measures of fiscal decentralization, such as the subnational share in total government expenditure or the subnational share in total government revenue. Despite the popularity of these measures in the literature, their employment has been heavily criticized because they provide no information about the degree of autonomy of subnational governments (Rodríguez-Pose and Ezcurra 2010; Ezcurra 2015). This limitation of this type of measures is particularly relevant for our analysis, as the various arguments discussed in Section ‘Why Should Ethnic Segregation Affect Domestic Terrorism?’ on the role played by decentralization in this context are based on the existence of political decision- and law-making power at subnational level.

13 We take the mean value of respondents who opted for that answer across the different waves of the WVS conducted over the 1981–2004 time horizon in order to maximize the number of countries with non-missing observations (Alesina and Zhuravskaya 2011).

Additional information

Funding

This research has benefited from the financial support of the Spanish Ministry of Economy and Competitiveness [project number ECO2015-64330-P].