代做Empirical Applied Econometrics Problem Set 1代写Java编程

Empirical Applied Econometrics

Problem Set 1

(Due April 28 11:59pm)

This empirical problem set estimates return to schooling and examine the human capital externalities in  cities using  a  Korean  dataset,  Local  Area  Employment  Survey  (LAES,   지역별고용조사),  from Statistics Korea.

-     You can download the microdata version fromhttps://mdis.kostat.go.kr/index.do.

-     You must use the microdata version, not the aggregate version from KOSIS.

-     Intro.do and reg_IV.do will be helpful in doing this problem set, but exact details can be different. In that case, clearly write down your assumption(s).

-     You  should submit your word file  (pdf format) with nice looking tables and figures, as well as your do files.

-     If you worked with your friends, please write down her (his) name.

Part 1. Estimating return to schooling

1. Download the first 2019 survey of the LAES dataset (2019년 상반기, A형) and restrict samples to aged  16-65, not in  school, and wage-earners (use the variable monthly wage,  월평균임금).  How many observations do you have after the restriction? What is the mean monthly wage? Median? Show the histogram.

2. Transform monthly wage into “log weekly wage” . Transform education variables into “years of schooling” . Show the histogram of the two transformed variables. Show the scatter plot of the two variables, along with the linear fit.

3. Generate variables “years of experience” and “years of experience square” . Show the scatter plot between the log weekly wage and years of experience.

4. Regress the log weekly wage on years of schooling. Use three different versions of standard errors (simple, robust, and clustered standard error). For the clustered one, cluster by cities (행정구역).

- Interpret the coefficient on years of schooling. Is the estimate statistically significant?

- Which standard error is the smallest among the three? Explain why.

5. Now add control variables, including years of experience and its square term, to the regression from the question 4. You choose the control variables by yourself (Add at least three or more). Recall that some variables could be “bad controls” .

- How has the coefficient on years of schooling changed? Explain reasons of the change.

- Interpret the coefficients on the control variables you chose.

6. Run some sub-sample analysis, with full controls and clustered standard errors. Please run at least three different analyses. Report and explain the results.

7. Construct city-level share of college graduates (out of population) and use it as an instrument for the years of schooling (you have to merge the datasets). Run a 2SLS regression with full controls and clustered standard errors.

- How has the coefficient on years of schooling changed from the OLS estimate in the question 5?

- What assumption(s) do we need for “the share of college graduates” be avalid IV?

- What is the first stage F-statistics? Is the instrument strong?

Part 2. Human Capital Externalities in Cities

8. Summarize the share of college graduates variable. What city shows the highest (lowest) share of college graduates? Make Top10 and Bottom10 rankings table across cities.

9.  Construct  other  city-level  characteristics  (population,  manufacturing  employment,  agriculture employment, age 60 or more, age less than 30, and female population). The final variables, which will be used as control variable, are (log) population, share manufacturing, share agriculture, share age 60+, share age 30-, share and female. Summarize those.

10. Merge the city-level dataset with the individual-level dataset (that were used in part 1). Check whether they are merged correctly. Examine the correlation between individual level schooling and city-level share of college graduates. Are they positively or negatively correlated?

11. From the regression in the question 5, include the share of college graduates as a regressor.

- What are the coefficients on the individual- and the city-level return to schooling? Interpret.

- How has the individual-level coefficient changed?

- Recall that we use the share of college graduates as an instrument in question 7. Do you think it is a good instrument for the years of schooling? Explain.

12. Add other city-level controls. How has the coefficient on the share of college graduates changed?

13. Run the regression by two skill levels (college graduates or more and non-college graduates). How are the estimates on social returns (share of college graduates) different? Interpret.

14. (Optional) for final project, you may want to include other city-level characteristics and whether it affects wages of workers. You can examine preliminary correlation here.