Lecture 4 - RDD

Question 1

What are 2 reasons for not making a randomization and what can we do instead?

Answer 1

1) it’s to expensive
2) Nobody wants to participate

in this case, we can do a Regression Discontinuity Design in which we rank our observations in a order.

Question 2

What are 4 advantages of the RDD?

Answer 2

1) RDD has the advantage of looking like randomizing without actually doing it
2) By ranking them, we remove the option of making a choice that affects the results –> No bias.
3) Those groups close to the cut-off group will have very similar characteristics –> good groups to divide into treatment and control.
4) The only difference between the groups are my treatment

Question 3

What are 1 disadvantage of the RDD?

Answer 3

1) The outcome of the analysis will only look at those next to the cut-off. Doesn’t examine other numbers or values as well.

Question 4

What are 2 important conditions when using RDD?

Answer 4

1) They shouldn’t be able to change their groups.

2) The cut-off is externally imposed.

Question 5

If we have 80% power, but we want to make clusters, what should we do to the sample size?

Answer 5

Probably increase it in order to be able to compare our clusters with each other.