Together with Jamie Wheeler, I am studying economic and political predictors of revolution. I am taking various measures of well-being and applying them within the framework of a model of revolution I developed during a previous semester. My hope is to better predict instability in developing countries, which would have applications in fields from diplomacy to finance.
I became interested in this project while studying at Trinity College in Oxford, during a tutorial in comparative politics. I studied both revolution and regime failure, and noticed that many theories were more explanatory than predictive. These studies often were based on case studies, and I believe a large-n study (even with a large error coefficient) can help guide future studies.
I am planning on pursuing a PhD in Political Science. Obviously, there is no better way to prepare for graduate school than bolstering my resume with successful research, and hopefully conference presentations or a paper publication. After grad school, I hope to either teach at a university, work in a think-tank, or work as a policy advisor. Any of these career paths would also be aided by having published works and research experience.
I spend most my research time poring over databases and looking for data that most closely resembles the variables within the theoretical model. I hold weekly meetings with my mentor, where we cover what I have learned or want to focus on for the next week. We often spend large portions of this time debating the scope and structure of the project. Coming from different academic fields, we often look to apply vastly different tools to solve the problems in front of us. These differences can lively discussions, but also allow us to take a unique approach to a large, oft-studied issue in political science.
This week, I discovered several statistical methods to deal with the real-world challenges of measuring well-being in states that are, by definition, politically unstable. While data is often incomplete or must be estimated in many cases, careful selection of sources can provide results that are precise, but inaccurate. However, the direction of these inaccuracies is often similar, allowing this variability to become baked into the error term of the model. While this reduces the predictive power of the model, it allows the model to hopefully apply to a larger number of cases.