Causal Linkages Between Domestic Terrorism and Economic Growth

Abstract

We use the Hsiao–Granger method to test for terrorism–growth causality for seven Western European countries. In bivariate settings, the impact of economic performance on domestic terrorism is very strong. In trivariate settings, the impact of performance on terrorism diminishes. In general, we find that economic performance leads terrorist violence in robust ways only for three out of seven countries. Terrorism is almost never found to causally influence growth in bivariate and trivariate specifications. Our findings indicate that the role of economic performance in determining terrorist violence appears to have been important for some countries, whereas all attacked economies have been successful in adjusting to the threat of terrorism.

Keywords:

• Domestic terrorism
• Economic growth
• Hsiao–Granger causality
• Western Europe

JEL Codes:

• D74
• N4
• 052

ACKNOWLEDGEMENTS

An earlier version of this paper was presented at the 9th Annual Meeting of the European Public Choice Society in Athens (April 2009). We thank the conference participants for their helpful comments. We also thank two anonymous referees for their very valuable comments and suggestions. Moreover, we are indebted to Wolfgang Eggert, Manfred Kraft, Judit Kapas and Ekkart Zimmermann for further comments, support and encouragement. The usual disclaimer applies.

Notes

¹ For our analysis we focus on domestic terrorism, i.e. terrorism involving only citizens, groups or the territory of one country. In contrast, transnational terrorism means terrorism involving citizens, groups or the territory of more than one country. Here, either the sources or targets of transnational terrorism can be analyzed.

² The TWEED dataset provides information on domestic terrorism for 18 Western European countries.

³ Frey et al. (2007) and Enders and Sandler (2008) bring forward similar theoretical channels of influence of terrorism on the economy. These studies also provide additional evidence on the macroeconomic consequences of terrorism that is not reported here.

⁴ The empirical evidence reported in the following almost always refers to the phenomenon of transnational terrorism, given the lack of data on domestic terrorism in the past (Enders and Sandler, 2008). Although we do not believe that transnational and domestic terrorism interact symmetrically with the economy, we report fitting evidence to show (1) that terrorism may damage the economy, and (2) economic factors may determine terrorist activity.

⁵ Evidence on the targets of transnational terrorism either finds that short-run and long-run economic conditions do not matter strongly for terrorists' attack decisions (Piazza, 2006; Drakos and Gofas, 2006; Kurrild-Klitgaard et al., 2006), or indicates that economically successful countries are more prone to terrorism (Tavares, 2004; Blomberg et al., 2004b). Attackers from abroad may not be deterred by increased economic opportunities in the country they attack. Instead, economic success may increase the payoffs from terrorist strikes. That is, the cost-benefit considerations of transnational attackers may differ from those of domestic terrorists.

⁶ One can argue that other factors potentially driving growth as well as terrorism should also be controlled for, such as the quality of economic institutions or political instability. We opt for trade openness because it matches our analysis on theoretical grounds and because data is available for all countries and time periods.

⁷ Mirza and Verdier (2008) provide an overview of the terrorism–trade literature, focusing on transnational terrorism. It is reasonable to believe that at least some related findings can be transferred to the relationship between domestic terrorism and trade. Note that in the broader sense trade openness may also reflect the general quality of institutions (Rodrik, 2002). Better institutional quality should coincide with less terrorist activity.

⁸ When we use count data on terrorist attacks, we cannot possibly evaluate the ferocity of such attacks. For instance, both a minor, politically motivated damage to property and a severe bombing with multiple casualties count as one attack. Here, it appears reasonable to assume that there is a much stronger interaction between the bombing and economic factors.

⁹ We use the natural logarithm to better account for outliers. We add unity to the observations in order to compute the natural logarithm also in those years when there were no victims from terrorist attacks. Note that by adequate data processing (as described later) we circumvent potential problems associated with ‘excess zeros’ (i.e., when there are many observations without victims from terrorism). By using estimated rather than observed values for the terrorism series, ‘excess zero’ problems do not arise.

¹⁰ Note that when two or more series used simultaneously in the following causality analysis are found to be I(1), further tests for cointegration would be required in order to account for long-run equilibrium relationships between the series. As two of the three series analyzed here are always I(0), there is no need to discuss the issue of cointegration any further.

¹¹ The use of additional unit root tests by Clemente et al. (1998) and Bai and Perron (1998) generally confirms our results on stationarity and structural breaks.

¹² The trivariate VAR is not presented due to space constraints.

¹³ We are well aware that terrorism may substantially damage certain sectors of the economy. We hinted at corresponding evidence in our literature review. Similarly, terrorist violence may constrain economic growth in regions most affected by conflict. For our sample this may apply to the Basque country and Northern Ireland, as corresponding evidence by Abadie and Gardeazabal (2003) and Fielding (2003) suggests. Our analysis simply provides little evidence that such economic damages feed through to overall growth at national levels.

¹⁴ Triacca (1998) provides proof for the following argument: If a variable Y₃ does not cause Y₁ in a bivariate system, but in a trivariate one where the variable Y₂ is also included, then Y₂ must cause Y₁ in the bivariate and trivariate system. We apply his proof to our causality evidence for Germany and Spain, where Y₁ is economic growth, Y₂ is trade openness and Y₃ is terrorist violence.

¹⁵ See Burgoon (2006) for an in-depth discussion of the potential links between terrorism and welfare policies. See Krieger and Meierrieks (2010) for an explicit empirical analysis of the effect of welfare policies on terrorism in Western Europe which finds that such policies may work as tools for scaling down terrorism.

¹⁶ However, clearly differentiating between domestic and transnational terrorism (as in our analysis) may be helpful to keep apart potentially different terrorists calculi. As discussed before, transnational terrorists may be driven by other cost-benefit considerations than domestic terrorists. For instance, while transnational terrorists may find it attractive to attack rich countries because of the increased benefits of such attacks, domestic terrorists may reduce attacks at the same time due to the increased opportunity costs of violence.

¹⁷ For instance, Barros and Proença (2005) analyze which characteristics are associated with Islamic terrorist attacks in North America and Europe. Barros et al. (2007) scrutinize which factors influence the likelihood of US citizens to fall victim to a terrorist attack in Europe. That is, these studies investigate terrorist behavior with respect to the ideological affiliation and choice of target. Such analytical approaches may, amongst others, be useful to expand the empirical literature, also with respect to the terrorism–growth nexus.