No Good Deed Goes Unpunished: Refugees, Humanitarian Aid, and Terrorism

Choi, S. W., & Salehyan, I. (2013). No Good Deed Goes Unpunished: Refugees, Humanitarian Aid, and Terrorism. Conflict Management and Peace Science, 30(1), 53-75.

Keywords: humanitarian aid, refugees, domestic terrorism, international terrorism

Regression methods: negative binomial ML, logit, fixed effects, GEE

The authors look at the effects of hosting refugees on incidence of terrorism in the receiving countries. They base their considerations on the premise that although refugees are likely to be victims of violence themselves, they may also contribute to a spread of violence and instability in their host countries. A number of possible reasons why providing shelter to refugees may lead to terrorism are given. They include attempts to attack refugee camps by militants, ease of recruitment from within the refugee camps for insurgencies and terrorist campaigns, rise in anti-immigrant movements in the receiving country, increased aid resources may invite looters, and aid workers are easy targets of kidnappings for ransom.

The authors’ empirical strategy is based on a negative binomial model with the number of all, only domestic or only international terrorist attacks (or casualties) as a dependent variable. They take their data from what is considered to be one of the best data sources on terrorism – the Global Terrorism Database (GTD) – and its decomposition into domestic and international incidents by Enders, Sandler and Gaibulloev (2011). The key independent variable is a number of refugees that a country receives from other states. They add a bunch of the usual controls and the lagged dependent variable which adds a dynamic dimension to the model. To deal with the excessive zeros problem, they also employ a form of a logit model.

The paper presents a number of alternative model specifications and all of them indicate that counties which receive more refugees are more likely to suffer from both international and domestic terrorism. The results also hold when ITERATE data is used instead of GTD. In their robustness check, the authors introduce a terrorism hot spot variable which singles out countries that are located in a neighbourhood of states that experience over-average intensity of terrorism. The reasoning behind this move is that being in a bad neighbourhood can increase the number of refugees as well as country’s exposure to political violence through “non-refugee” shocks. Although being located in a terrorist hot spot turns out to contribute to the frequency of terrorism, it doesn’t remove the significance of the refugee variable. The results appear to be somewhat weaker in the developed OECD countries, but as the authors notice the distribution of refugees and structure of aid in those countries is likely to be different.

The penultimate section of the article is the weakest element of the study. The authors make an effort to analyze the link between humanitarian assistance and terrorism. However, they encounter data limitations and are able to illustrate only a relatively small number of cases for the post-1998 period. No formal analysis is applied and therefore the conclusion that “a large number of attacks are against aid workers and humanitarian supplies” is not backed by hard evidence.

The article is one of the early papers on the relationships between refugees and terrorism, it gives results that are immune to various model specifications and contributes to the literature on the causes of terrorism. It comes with results that are relevant for policy making and should be considered by governments when dealing with humanitarian crises, both in developing and developed countries. The issue of receiving refugees, and immigrants in a wider sense, is likely to be climbing up the governments’ agendas in developed countries. As the recent riots in the suburbs of Stockholm and earlier in Paris show authorities will have to come up with a more effective way of protecting and integrating new-comers with native societies.