3. Suppose you want to evaluate impacts of job-training program. Unfortunately, the program was already implemented, and you could not use RCT. Followings are the weekly income of the participants and non-participants, before the training.

1

2

3

4

5

6

Participants

12

15

16

13

10

14

Non-participants

10

8

10

12

11

5

3-1. Suppose you want to conduct difference-in-differences to estimate the impact of the training program. In an ideal situation, what kind of data do you need to use diff-in-diff?

3-2. (10%) Suppose you have an ideal data, that you discussed in 3-1. You got following income distribution after the training. What is the estimated impact of the training, using the idea of difference-in-differences?

1

2

3

4

5

6

Participants

15

14

22

17

12

17

Non-participants

12

6

13

14

14

6

(Note: when we rigorously implement diff-in-diff, we need to think about statistical inference. We do not have to consider statistical features in this problem, as the main aim of this problem set is tobecome familiarwith the methodology)