Final thoughts on causal inference

Wrapping up

In this module, we learned about various quasi-experimental methods: regression discontinuity design (RDD), the synthetic control method, and difference-in-differences (DiD).

The central theme across these methods was this. Answering causal questions typically require a randomized setting to remedy confounding bias, but in the absence of enforced randomization, researchers go on treasure hunts in search of quasi-experimental settings where:

• Either the treatment assignment is naturally randomized (like in RDD)
• Or the treatment assignment is not random but we can find an equivalent control group for the treated group (like in synthetic control and DiD)

In any case, natural experiments shouldn’t be confused with fully-randomized experiments. Sometimes a quasi-experimental study is persuasive, other times it’s a hard sell and you’ll get the sense that the researchers are grasping at straws. Treatment assignment as-if randomized is usually less of an issue in natural experiments based on lotteries and very random natural events such as earthquakes.

When you’re assessing a quasi-experimental study or embarking on your own, it’s more important to think long and hard about the assumptions than it is to dwell on the number crunching. Think about the eligibility criteria: who qualified for the treatment and who didn’t. Do subjects self-select into treatment and control groups in ways that are unobserved? Are variables that are potentially relevant for selection into the treatment group potential confounders?

The end

This may sound rather abrupt, but you’ve just reached the end of this course on causal inference 🎉🎊🥂

You may have learned many things about causal inference and you may have already forgotten some others. Hopefully, though, you’ve learned enough to challenge your non-causal friends and family next time they make or cite a dubious causal claim.

We have intentionally left out many other tools and variations of the methods we discussed in this course. We did that not to overwhelm you in your first exposure to causal inference. Hopefully, you are encouraged to continue your journey on your own and if so, don’t forget to tell us how your journey evolves and let us know if you have any questions.

You can also take other courses offered by cauzl. Hopefully, those will keep you busy and add to your understanding of empirical research.

Practice

How do you feel about what you learned in this course? Hint: there is a correct answer!

🙂

🙁