Matching

Question 1

How does matching approximate the counterfactual?

Answer 1

Identifies control group that is as similar as possible to the treatment group (in terms of observables)

Question 2

Similarities and differences between multiple regression and matching

Answer 2

Both rely on CIA, but weaker form for matching (both assume that once you condition/match on observed characteristics you controlled for all possible differences)
Both only account for observable differences

Matching requires common support but regression does not
Matching can only deal with binary treatment variables, regression also continuous treatment variables

Question 3

Sibling fixed effects

Answer 3

Deals with family specific unobserved variables: siblings share similarities such as family background

Question 4

Twin fixed effects

Answer 4

Deals with family and genetic unobserved variables

Question 5

Limitations twin fixed effects

Answer 5

1. Twins should differ in their treatment status, otherwise the pair cancels out
2. Are twins representable?
3. Vulnerable to measurement error
4. Within-twins pair treatment variation may be correlated with unobserved within-twins factors that directly affect the outcome