Sep 3, 2018 · 1 min read
Hi Chris! Thanks for your kind response. Yes, the big issue in regression is the causal interpretation. What I did basically just controlling for observable confounders to derive the causal explanation. This approach is well known in causal inference methodology. Fortunately, I managed to control for quite good confounders in my model. Should be noted here, I intended to see the similarity between the two challengers’ supporters. Not precisely measuring the association of religion with the vote share as the primary variable of interest.
