Spillover as Movement Agenda Setting: Using Computational and Network Techniques for Improved Rare Event Identification

Abstract

The increasing availability of data, along with sophisticated computational methods for analyzing them, presents researchers with new opportunities and challenges. In this article, we address both by describing computational and network methods that can be used to identify cases of rare phenomena. We evaluate each method’s relative utility in the identification of a specific rare phenomenon of interest to social movement researchers: the spillover of social movement claims from one movement to another. We identify and test five different approaches to detecting cases of spillover in the largest data set of protest events currently available, finding that an ensemble approach that combines clique and correspondence analysis and an ensemble approach combining all methods perform considerably better than others. Our approach is preferable to other ways of analyzing such cases; compared to qualitative approaches, our computational process identifies many more cases of spillover—some of which are surprising and would likely not be otherwise investigated. At the same time, compared to crude quantitative measures, our approach substantially reduces the “noise,” or identification of false-positive cases, of movement spillover. We argue that this technique, which can be adapted to other research topics, is a good illustration of how the thoughtful implementation of computational methods can allow for the efficient identification of rare events and also bridge deductive and inductive approaches to scientific inquiry.