Projects
Academic projects
Machine Learning Based Estimation of Average Treatment Effects under Unconfoundedness
I analyze the estimation of causal effects using machine learning methods. I study regression estimators, matching on the propensity score, inverse probability weighting, and hybrid methods such as bias-corrected matching and doubly robust estimators. These estimators require the estimation of the conditional outcome means, the propensity score, or both. In empirical applications, these functions are often estimated with ordinary least squares or logit. In this project I additionally consider machine learning methods to estimate these functions.
To illustrate my findings, I created an interactive online appendix with Shiny. The corresponding R code can be found on my GitHub repo.
This project was part of my PhD thesis.
Identification and Estimation of Intensive Margin Effects by Difference-in-Difference Methods, Journal of Causal Inference (2020)
This paper discusses identification and estimation of causal intensive margin effects. The causal intensive margin effect is defined as the treatment effect on the outcome of individuals with a positive outcome irrespective of whether they are treated or not, and is of interest for outcomes with corner solutions (e.g. expenditures, working hours). The main issue is to deal with a potential selection problem that arises when conditioning on positive outcomes. We derive sufficient conditions under which the difference-in-difference estimator - conditional on positive outcomes - identifies the causal intensive margin effect.
This project (joint with Markus Hersche) was part of my PhD thesis.
Labor or Leisure? Labor Supply of Older Couples and the Role of Full Retirement Age
This paper estimates the labor supply response when the spouse reaches the full retirement age. We exploit the age difference within couples and changes in pension legislation in Switzerland to identify the causal effect. We estimate the effect not only on labor market participation (extensive margin), but also on working hours (intensive margin).
This project (joint with Markus Hersche) was part of my PhD thesis.
Forecasting Personal Consumption Expenditures Using Google Trends
In my master thesis, I use internet search data from Google Trends to now- and forecast US real personal consumption expenditures. The methods are based on forecast combination. Out-of-sample forecasting exercises are conducted to compare the forecasting accuracy of models including internet search data to benchmark models.
Thesis available upon request.
Health Care Reform and the Number of Doctor Visits: A Bayesian Replication
The goal of this seminar paper (joint with Damiano Pregaldini) is to understand and apply Bayesian econometrics in the context of a replication study. We replicate the results of the paper “Health Care Reform and the Number of Doctor Visits - an Econometric Analysis” (Rainer Winkelmann, JAE 2004). Applying Bayesian inference on a Poisson and Poisson Log-Normal model with flat priors, we are able to confirm the author’s findings.
Paper available upon request.