C6. The Conditional Independence Assumption (CIA)

Question 1

What is the Conditional Indpendence Assumption (CIA)?

Answer 1

Question 2

CIA and the Fundamental Equation fo Causal Inference

Answer 2

Question 3

How does conditional independence impact out equation?

Answer 3

Question 4

How can we get the ATE(x) for each sub-population?

Answer 4

Divide the data into cells and take sample averages of outcomes of treated and untreated observations within each cell to get the ATE(x) for each sub-pop.

Question 5

How can we get overall ATE and ATT?

Answer 5

Question 6

Three problems with conditional independence

Answer 6

1. Credibility of CIA
2. The common support problem/curse of dimensionality
3. “Bad controls”

Question 7

How is credibility a problem for the CIA?

Answer 7

Conditional independence strategy based on an assumption - all sources of non-random selection are controlled for.

In order to believe the CIA (in the Angrist example), you need to think that things which lead a qualified applicant to drop out last minute were independent of their potential earnings. In fact, could be flakiness which is plausible tied with them.

Question 8

How is common support a problem?

Answer 8

Question 9

What are “bad controls”?

Answer 9

Bad controls are variables that are themselves outcome variables in the notional experiment at hand.