The course is conducted in English。

This graduate-level course explores the intersection of machine learning and econometrics, bridging predictive analytics with statistical inference and causal modeling. Since Varian (2014), the conventional distinction has been that machine learning focuses on using data to predict variables, while econometrics applies statistical methods for prediction, inference, and causal analysis in economic relationships. However, more recent perspectives suggest that machine learning represents a new wave of nonparametric statistical and econometric techniques (Chernozhukov, 2016). Moreover, machine learning has introduced exciting advancements in econometric applications, such as estimating conditional average treatment effects (Athey and Imbens, 2016) and serving as the first stage in two-stage estimation (Chernozhukov et al., 2018).

In this course, we—a group of well-trained econometricians—will explore machine learning together. On the estimation and prediction side, we will cover techniques like Lasso, neural networks, and random forests. On the inference and causal modeling side, we will discuss inference methods after applying machine learning, as well as approaches inspired by computational statistics, including post-model selection inference and large-scale hypothesis testing. More importantly, we will explore open research questions in inference, and you are encouraged to engage with these challenges as potential research projects.

One midterm presentation (April 8, 40%). One final presentation (May 27, 60%).