The geography of violence during a presidential election: Evidence from Zimbabwe

Successful election monitoring depends on the ability to predict where violence will be used to intimidate voters. However, the strategy of a violent electoral candidate will depend on the particular characteristics of the election, including (i) whether other candidates are able to organize violence, (ii) whether the outcome depends on nationwide voting shares or shares in individual electoral districts, and (iii) how costly it is to transport the resources used to perpetrate violent acts. Although there is already ample theory and evidence for some combinations of these characteristics, analysis of other combinations is still lacking. This paper presents a theory designed to analyse strategy in an election in which the outcome depends on nationwide voting shares and only one candidate is able to organize violence, distinguishing between the case of costless transportation and the case in which transportation is prohibitively costly. The theory predicts that in the case of costly transportation with constant returns to scale, violence will be targeted at areas where support for the opposition is relatively low. Evidence suggests that the second round of the 2008 presidential election in Zimbabwe corresponds to such a case, and statistical analysis of the geographical distribution of violence around the time of the election indicates that areas of opposition weakness did indeed suffer more violence.