Causal Inference with Matching Methods

Suppose you want to know the effect of enrollment in a Massive Open Online Course (MOOC) on income. You are worried that people with more traditional education-e.g., someone with a bachelor's degree-will be more likely to enroll in a MOOC and also more likely to make more income, compared to someone with less traditional education-e.g., someone with only a high school diploma. So you make the unconfoundedness assumption that, holding traditional education fixed, people randomly enroll in MOOCs. You then use matching to estimate the ATE of MOOC enrollment, and find that it is positive. What policy might you recommend from this, if any?

Possible Answers

No recommendation-correlation does not equal causation and so we cannot conclude what the effect of MOOC enrollment is.
MOOC enrollment should be promoted as a means of improving people's wages.
The costs of enrolling in MOOCs should qualify for federal student loans, just as traditional educational approaches do.
MOOCs should be banned so as to promote traditional educational methods.

Exercise

Causal Inference with Matching Methods

Instructions

Possible Answers