Advanced Econometrics Take Home Exam 2, WS 2022/23
Hello, dear friend, you can consult us at any time if you have any questions, add WeChat: daixieit
Advanced Econometrics
Take Home Exam 2, WS 2022/23
Estimating Returns to Schooling by a Control Function Ap- proach
Consider the data set “th2.mat” . It contains a subset of the data used by Jochmann and Pohlmeier (2004) to estimate returns to schooling. The variables are described in Table 1.
Variable name |
Description |
LNLOHN |
log wages |
BILZEIT |
Educational attainment in years |
ALTER |
Age in years |
KHGROSS |
Dummy for the region of residence during childhood (1=big, 0=otherwise) |
KHMITTEL |
Dummy for the region of residence during childhood (1=middle, 0=otherwise) |
KHKLEIN |
Dummy for the region of residence during childhood (1=small, 0=otherwise) |
ANZBR |
Number of brothers |
ANZSCH |
Number of sisters |
FLAECHEGYM |
Commuting area of upper secondary schools |
ALQ |
Regional unemployment rate at time of graduation |
Table 1: Description of the Variables
You want to estimate returns to schooling based on the following equation:
a) Estimate model (1) by OLS. Report estimated parameters and OLS standard errors. 0.5 P
b) Estimate standard errors for the OLS estimates from a nonparametric bootstrap (do
5000 bootstrap replications). Report and compare them with the standard errors of
part a). Also report 95% confidence bands using the percentile method. 1 P
Jochmann and Pohlmeier (2004) use the following four sets of instruments to estimate returns to schooling:
DENS Commuting area of upper secondary schools interacted with the dummies for the re- gion of residence during childhood. Instruments: FLGR = FLAECHEGYM*KHGROSS, FLMIT = FLAECHEGYM*KHMITTEL, FLKL = FLAECHEGYM*KHKLEIN, FLAECH- EGYM.
URG Unemployment rate at graduation interacted with age. Instruments: ALQ, ALQAL-
TER = ALQ*ALTER.
SIB Number of brothers and sisters. Instruments: ANZBR, ANZSCH.
ALL All 8 instruments from the three sets above.
c) To overcome the endogeneity problem estimate the model using a control function (CF) approach for all four sets of instruments. Report the estimated parameters and compare the estimates of β 1 of the control function approaches with the OLS estimate of β 1 . 2.5 P
d) Estimate the standard errors of the parameter estimates for all four CF approaches using a nonparametric bootstrap (do 5000 bootstrap replications). Also compute 95%
confidence bands for all parameters using the percentile method. Report your results. 2.5 P
e) Test the exogeneity of the variable BILZEITi for all four sets of instruments by a t-test
using the bootstrap standard errors from part e). What do you conclude? 1.5 P
f) Report the correlation between BILZEITi and the residuals from each of the four reduced form equations. Interpret your findings. What problem do you face with a correlation very close to one? Suggest a potential remedy for this issue. 1.5 P
g) Assume that the instruments are weak. What can you say about the quality of the exogeneity tests, given the presence of weak instruments? 0.5 P
Reference
Jochmann, M. and W. Pohlmeier (2004). “Der Kausaleffeekt von Bildungsinvestitionen: Empirische Evidenz f¨ur Deutschland.” In W. Franz, H. Ramser and M. Stadler (eds.), Bil- dungs¨okonomik. Wirtschaftswissenschaftliches Seminar Ottobeuren 33: Mohr Siebeck, 1-24.
2023-01-29