Empirical Methods for Causal Inference

This dissertation reflects the use of various methods of causal inference using observational data based on three articles in health economics and epidemiology. These case studies encompass analyses of the implementation of a quality improvement framework in UK primary care (using Differencein-differences), the analysis of the impact of bearing children on oral health (using Instrumental Variables) and the analysis of an implementation of altered provider incentives in the Danish dental care system (using Interrupted Time Series Analysis). All of these make use of methods utilizing quasi-experiments that allow post-hoc derivation of causal effects if certain identifying assumptions are met. Disregarding causality in empirical research altogether exposes results to the risk of enormous biases and false interpretations. This is arguably problematic as expressing the simple causal relationship between two variables X and Y statistically is not possible in an explicit manner - a problem known as the "fundamental problem of causality" where only one of multiple possible outcomes manifests in the real world. This thesis offers a remedy to this problem by showcasing how the causal frameworks of Pearl, Campbell and Rubin may contribute to causal argumentation and ultimately more robust research