Presentation is loading. Please wait.

Presentation is loading. Please wait.

ECON 3039 Labor Economics 2015-16 By Elliott Fan Economics, NTU Elliott Fan: Labor 2015 Fall Lecture 41.

Similar presentations


Presentation on theme: "ECON 3039 Labor Economics 2015-16 By Elliott Fan Economics, NTU Elliott Fan: Labor 2015 Fall Lecture 41."— Presentation transcript:

1 ECON 3039 Labor Economics 2015-16 By Elliott Fan Economics, NTU Elliott Fan: Labor 2015 Fall Lecture 41

2 Difference-in-difference 2Elliott Fan: Labor 2015 fall Lecture 4 1929 macroeconomic shock in the US: Banks closed, mostly due to confidence crisis Federal Reserve had two policies: 1.Lend easy money to help bank 2.Stay neutral (to prevent moral hazard)

3 Difference-in-difference 3Elliott Fan: Labor 2015 fall Lecture 4 Bank districts in the US: The U.S. Federal Reserve System is organized into 12 districts, each run by a regional Federal Reserve Bank. Depression-era heads of the regional Feds had considerable policy independence. The Atlanta Fed, running the Sixth District, favored lending to troubled banks. By contrast, the St. Louis Fed ran the Eighth District according to a restrict credit in a recession.

4 Difference-in-difference 4Elliott Fan: Labor 2015 fall Lecture 4 The question: Did liquidity easing help prevent deterioration of bank crisis? If so, what’s the extent of such monetary policy? A natural starting point is to make a comparison of D6 and D8 in the following graph:

5 Difference-in-difference 5Elliott Fan: Labor 2015 fall Lecture 4 Terminologies: Treatment group: D6 Control group: D8 Treatment: monetary easing Outcomes: number of banks; wholesales

6 Difference-in-difference 6Elliott Fan: Labor 2015 fall Lecture 4

7 Difference-in-difference 7Elliott Fan: Labor 2015 fall Lecture 4 I believe your eyes will be focused on the change in the number of banks in D8 from 1930 to 1931, relative to the change in D6

8 Difference-in-difference 8Elliott Fan: Labor 2015 fall Lecture 4 Using mathematics:

9 Difference-in-difference 9Elliott Fan: Labor 2015 fall Lecture 4 The next questions: Can we rely on a comparison using only 4 points? How do we know if the difference-in-difference estimate is statistically significant?

10 Difference-in-difference 10Elliott Fan: Labor 2015 fall Lecture 4 Common trend assumption To answer the first question, we need to know whether the ‘control group’ returns the counterfactual outcome for the treatment group One way to verify this is to compare the trends before and after the policy intervention

11 Difference-in-difference 11Elliott Fan: Labor 2015 fall Lecture 4

12 Difference-in-difference 12Elliott Fan: Labor 2015 fall Lecture 4 Regression Y dt : number of banks in district d in year t TREAT d : a dummy variable indicating D6 (TREAT d = 1) or D8 (TREAT d =0) POST t : a dummy variable indicating years after 1930 (POST t = 1) or years before or at 1930 (POST t = 0)

13 Difference-in-difference 13Elliott Fan: Labor 2015 fall Lecture 4 Regression

14 Difference-in-difference 14Elliott Fan: Labor 2015 fall Lecture 4 Results:

15 Difference-in-difference 15Elliott Fan: Labor 2015 fall Lecture 4 Another example: Minimum Legal Drinking Age (MLDA)Minimum Legal Drinking Age (MLDA) In 1970s, some states in the US reduced MLDA from 21 to 18, 19, or 20, while others remained unchanged Many concerned about adverse effects on youth We can structure a DD approach to estimate the effects

16 Difference-in-difference 16Elliott Fan: Labor 2015 fall Lecture 4 A baseline version of the model: Consider only 2 states: Alabama lowered MLDA to 19 in 1975, while Arkansas remained unchanged We are intended to know whether reducing MLDA leads to more deaths for those aged 19-20.

17 Difference-in-difference 17Elliott Fan: Labor 2015 fall Lecture 4 A baseline version of the model: Y st : Youth deaths in state s in year t TREAT s : a dummy variable indicating D6 (TREAT s = 1) or D8 (TREAT s = 0) POST t : a dummy variable indicating years after 1930 (POST t = 1) or years before or at 1930 (POST t = 0)

18 Difference-in-difference 18Elliott Fan: Labor 2015 fall Lecture 4 A baseline version of the model:

19 Difference-in-difference 19Elliott Fan: Labor 2015 fall Lecture 4 Why is common trend so important?

20 Difference-in-difference 20Elliott Fan: Labor 2015 fall Lecture 4 Why is common trend so important?

21 Difference-in-difference 21Elliott Fan: Labor 2015 fall Lecture 4 Now we extend the model to 3 states: Arkansas remained unchanged; Alabama lowered MLDA to 19 in 1975; Tennessee reduced MLDA to 18 in 1974 Three concerns: 1.There are multiple states 2.States amended law in different years 3.They reduced to different ages

22 Difference-in-difference 22Elliott Fan: Labor 2015 fall Lecture 4 How to deal with multiple states? We simply consider Alabama and Tennessee as two different treatment groups We assign one dummy variable to represent each of the two states in the regressions

23 Difference-in-difference 23Elliott Fan: Labor 2015 fall Lecture 4 States amended law in different years The variable POST t can be tailored to be different for states amending the law in different years We can view the data spreadsheet to comprehend this

24 Difference-in-difference 24Elliott Fan: Labor 2015 fall Lecture 4 They reduced to different ages: This should be considered as different degrees of the treatment Tennessee from 21 to 18: TREATs = 1 Alabama from 21 to 19: TREATs = 2/3

25 Difference-in-difference 25Elliott Fan: Labor 2015 fall Lecture 4 The model: LEAGALst = TREAT s x POST t Thus, we can extend the model to accommodate more states.

26 Elliott Fan: Labor 2015 Fall Lecture 426


Download ppt "ECON 3039 Labor Economics 2015-16 By Elliott Fan Economics, NTU Elliott Fan: Labor 2015 Fall Lecture 41."

Similar presentations


Ads by Google