CANCELLED! CHG-seminar: "Taking an Extra Moment to Consider Treatment Effects on Distributions"

Image

Dennis Petrie conducts research in health economics with a focus on drugs, alcohol, disabilities, and cancer. He analyzes large datasets to improve healthcare decisions and health policy, and has published in prominent journals such as the Journal of Health Economics and JAMA.
 

Abstract
In this paper, we present a new, flexible method for estimating the impact of a social policy on parameters of the distribution that can be expressed as raw moments, such as the variance (E[Y^2] − E[Y]^2) of outcome Y.

Our approach involves predicting the observed and counterfactual values for each raw moment (E(Y^2), E(Y)) using standard causal methods. We then combine these predictions to estimate a policy’s impact on the parameters of interest. Our new approach is as good as existing approaches under selection on observables but also extends to any method that can credibly identify observed and counterfactual raw moments. In a difference-in-differences set-up, assuming parallel trends in the first four raw moments, we estimate the impact of a major Swedish comprehensive school reform on the mean, variance, skewness and kurtosis of years of schooling and log income, as well as their covariance.

The results show meaningful reductions in inequality in both years of education and log income, consistent with the overarching aim of the reform. The method we propose provides new avenues for causal research into inequality impacts of social policies.