Disastrous Voting
Abstract
The central question of this study is ’Is there an observable general trend of electorates punishing incumbents for natural disasters across countries and elections?’ Many scholars have argued for the existence of such behavior, yet the empirical evidence rests mostly on single-country and even single-election studies. I look for a generalizable trend in two original data sets with country-election and country-constituency-election as the unit of analysis respectively. I test the punishment hypothesis by correlating the occurrence of natural disasters to the performance of incumbent parties in national lower house elections. Furthermore, I propose that the effect of natural disasters on incumbents’ electoral performance varies depending on patronage expectations of the electorate and influx of international humanitarian aid. The analysis uses linear mixed-effects models with lagged variables estimated via restricted maximum likelihood estimation. The data does not support the rejection of the null hypothesis of no generalizable trend. Also, the results for the conditioning effects of patronage and international humanitarian aid are too volatile to draw inference. This study finds itself in the company of a few studies which also used cross-sectional time-series approaches, yet on a smaller scale, and which also were unable to find a generalizable trend. Together, their results are of importance for a research agenda with renewed interest as they caution scholars to attribute too much external validity to existing studies with research designs which focus on single countries and elections. I find no evidence for a generalizable trend once effects specific to countries and elections are washed out. In return, one could interpret this as evidence of electorates not blindly punishing incumbents for random external shocks. Instead, an electorate’s reaction to a natural disaster may depend upon intervening factors which allow the electorate to make a rational punishment/reward decision.