When and how do disasters affect civil war risks? A longitudinal analysis on the effects of natural disasters

The relationship between natural disasters and violent conflicts is an interdisciplinary topic of conflict research that becomes increasingly important in the context of climate change. However, empirical studies that establish the connection of climate change and variability on the different phases of conflicts are scarce. We contribute to this debate with a specially designed dataset that merges the EM-DAT: International Disaster Database that covers natural disasters, the UCDP/PRIO Armed Conflict Dataset that covers civil wars, the Quality of Government Institute Standard dataset (QoG), and the Varieties of Democracy dataset (V-Dem). Thus, we obtain extensive information on internal conflicts, national political and economic data, as well as natural disasters from 1946-2018. We focus on inter-annual variation in weather conditions, such as storms, earthquakes or weather shocks and how this triggers a) the onset, b) the duration of violent conflicts, and c) the ending of violent conflicts. We test several competing arguments and mechanism in our paper. First, it may be argued that disasters increase the risk for the onset of civil wars or for existing ones to be continued due to increasingly scarce resources. This may be especially the case when state capabilities are already low. I this line of reasoning disasters trigger and exaggerate existing grievances and problems. Yet, a competing argument may also be developed, namely that disasters lower the likelihood to start civil wars and raise the probability to end them since (potentially) conflicting parties may unite in order to combat the effects of a disaster. We argue that for these effects to occur an important distinction within disasters should be made, viz., between disasters that have immediate consequences, e.g., floods or earthquakes, and disasters that develop more slowly, e.g. droughts. To simultaneously estimate and disentangle effects between countries, e.g. ethnic fractionalization, stable and instable democracies, and within countries, e.g. (fast and slow) occurrence of disasters, we use a mixture of fixed-effects and random-effects models, so-called hybrid models. We demonstrate that natural disasters cannot explain the onset of civil war and do not lead to more wars. Instead, natural disasters increase the probability that existing wars will continue. However, this effect is conditional on the national political and economic conditions and on the specific type of disaster, i.e. whether a disaster has immediate or delayed consequences. Specifically, disasters with immediate consequences show stronger effects on all phases of war than disaster with slower consequences.