Zooming in Too Far? Micro Data and the Study of Political Violence

The increasing availability of micro data has been heralded as a huge boon in the field of political science. The quantitative study of topics such as the effects of political violence in developing countries have historically been nearly impossible due to limited availability of data. Now this and other phenomena can be quantified and analyzed at a much finer level. While we applaud the use of new data to explore these topics, in this paper we find that micro data is not always better when studying political violence. Using survey data from more than 4,000 Indonesian villages in the province of Aceh, we analyze the effect of political violence on population changes. Our findings suggest that highly-localized violence (at the village level) may not necessarily be a good predictor of village-level population changes, but that general violence at the sub district level is correlated with changes in the village-level population. We argue that this supports the use of a theoretically motivated level of aggregation in the study of political violence.