Abstract
INTRODUCTION: Studies show that natural disasters influence voters' perception of incumbent politicians. To investigate whether voters are prone to punish politicians for events that are out of their control, this study was conducted in the previously unstudied context of Croatia, and by considering some of the methodological issues of previous studies. METHOD: Matching method technique was used, which ensures that affected and non-affected areas are matched on several control variables. The cases of natural disaster in the present study were floods that affected Croatia in 2014 and 2015. RESULTS: Main results showed that, prior to matching, floods had an impact on voting behaviour in the 2014 and 2015 elections. Voters from flooded areas decreased their support for the incumbent government and president in the elections following the floods. However, once we accounted for differences in control variables between flooded and non-flooded areas, the flood effect disappeared. Furthermore, results showed that neither the presence nor the amount of the government's relief spending had an impact on voting behaviour. Discussion: Presented results imply that floods did not have an impact on the election outcome. Results are interpreted in light of the retrospective voter model.