Doctoral Dissertation Research: Discrete Time-Series Cross-Section Models of Political Economy

Washington University, Saint Louis MO

Investigators

Abstract

Doctoral Dissertation Research: Discrete Time-Series Cross-Section Models of Political Economy Discrete time-series cross-sectional (TSCS) data are a very important type of data in a variety of disciplines, including political science and economics. TSCS data have both a time and spatial dimension and are thus rich in structure for analysis. However, this correlated data structure invites multiple possible sources of error and raises several important methodological challenges. This research will develop a new method for analyzing TSCS data: a Bayesian generalized linear multilevel model with two-dimensional random effects and p-th order autoregressive errors for analyzing the temporal and spatial dependence of discrete TSCS data. The model empowers substantive researchers to better understand the dynamic process under investigation and to conduct comparative studies. Estimating the parameters of the proposed model and performing computations to facilitate model comparison are difficult and computationally intense. This project will solve these technical difficulties by designing a hybrid Markov Chain Monte Carlo algorithm with a special focus on simulation efficiency. To facilitate model choice, this research will also develop a relatively easy means of computing Bayes factors. A well-designed, user-friendly, and open source R package will also be provided to the scientific community that implements the proposed model and methods. This project will apply the model to several important questions in the field of international political economy. Substantively, the model will be applied to studying the "democratic advantage" theory in political science.