In-class exercise#
Note
Please respect the time limits. Of course you do not need to set a timer, but it is important that you make it to the very end.
Be prepared to present/discuss each of the following questions in class. You should designate group members for each of the subtasks (e.g., task 1 / example 1, task 2 / question 1, etc.) such that everybody is prepared to take a lead on at least one of these subtasks.
Task 1: Structural vs. MTE (30min)#
Discuss your answers to the quiz question “Discuss how a fully structural approach might have looked like in the Moffitt & Zahn (2022) paper. Stay in a static setting, i.e., assume that agents do not take future periods into account.”
Task 2: Treatment of heterogeneity (20min)#
Discuss your answers to the quiz question “How does Chan (2017) introduce individual-level heterogeneity? I.e., why could outcomes of families in the same treatment arm differ?”
Task 3: Protocols#
The three tasks are pretty different this week and people will benefit from summaries of the other groups’ results.
Please choose one person to make notes of the results of your group, updating them during the plenary discussion.
After the lecture, please share them via Zulip.
Task 3.1: Identification (60min)#
Describe intuitively how discounting behaviour is identified in Chan (2017). Are any functional form assumptions / additive separability required?
Task 3.2: Results (60min)#
Read sections 5.1 and 5.2 of Chan (2017), skipping comparisons to prior literature.
Describe the results in your own words, paying particular attention to heterogeneity and different predictions of different discounting models. In particular, assuming that the preferred specification is the “correct” one, what would we get wrong if we used the simpler models?
Task 3.3: Comparison to treatment evaluation approaches (60min)#
In a nutshell, the treatment evaluation literature compares (mean) outcomes across various treatment arms, thereby obtaining (average) causal effects of a particular treatment.
Skim through sections 4 and parts of 5 Chan (2017) and answer the following questions.
What would a classical treatment evaluation approach have done to evaluate the experiment?
Do we learn anything from Chan’s approach that we could not learn from that? Why or why not? If your answer is yes, what do we learn in addition?