Two methodological approaches have dominated quantitative analysis in Political Science: model-based (i.e., observational study) and design-based (i.e., causal inference) approaches. Yet, it remains uncertain whether and when inferences from these approaches are different, which designs are more robust to specification errors, and how well diagnostic tests perform detecting model problems with data structures common to Political Science research.

In this project we address these problems by comparing the performance of a range of modeling approaches using Monte Carlo simulations. These include model-based approaches derived from traditional econometric approaches to time series data, design-based approaches, and hybrid approaches that combine elements of these two approaches.

Two methodological approaches have dominated quantitative analysis in Political Science: model-based (i.e., observational study) and design-based (i.e., causal inference) approaches. Yet, it remains uncertain whether and when inferences from these approaches are different, which designs are more robust to specification errors, and how well diagnostic tests perform detecting model problems with data structures common to Political Science research.

In this project we address these problems by comparing the performance of a range of modeling approaches using Monte Carlo simulations. These include model-based approaches derived from traditional econometric approaches to time series data, design-based approaches, and hybrid approaches that combine elements of these two approaches.