I develop statistical methods for economic models. Unlike experimental data, economists do not have the privilege to conduct a lab-style experiment where confounding factors can be controlled. Some important questions are: Do countries become more democratic as they become richer, i.e., do income (GDP per capita) and democracy have a positive causal relationship? Or does government spending increase GDP? Yet these questions are difficult because the variables affect each other and one cannot experiment with the real economy. Therefore, economists develop an economic model and use a special statistical tool to reveal the true causal relationship between the variables.

Conventional statistical approaches build on the assumption that the underlying economic model is correctly specified. However, this assumption is often too strong. My research focuses on developing robust statistical methods against such violations of the correct model assumption. This robust method would provide a broader interpretation of economic models based on real data.