[At the beginning we finished the previous session on Selection on Observables]

This part is about situations, where observed information alone does not make the treatment "as good as random". We will explore how an exogenous variation in another variable, an instrument, could be helpful. Such approach led to many successful analyses, yet it has many limitations.

We decided on these two dates for exams: 18.12.2023 and 17.1.2024.

Assignment 2 (deadline 28.11.2023)

R code with suggested solutions.