Professional Documents
Culture Documents
Home Assignment 4
Home Assignment 4
Econometrics-II 2021
Home assignment 4
Due 10:00 April, 20 2021
where Cov(ci , xit ) 6= 0, uit are iid with E[uit |xi,1 ...xiT , ci ] = 0, V ar(uit ) = σ 2 .
a) To obtain a xed eects (within) estimator β̂F E we rst apply a xed-eect
transformation. Calculate the autocorrelation coecient for the transformed errors
Corr[üit , üi,t−1 ], where üit = uit − ūi .
b) To obtain a rst-dierence estimator β̂F D we rst apply a rst-dierence transformation.
Calculate the autocorrelation coecient for the transformed errors Corr[∆uit , ∆ui,t−1 ],
where ∆uit = uit − ui,t−1 .
Now suppose the errors in the original model were autoregressive, i.e.
where E[ui,0 |xi,1 ...xiT , ci ] = E[eit |xi,1 ...xiT , ci ] = 0, eit are iid, ui,0 is independent of all the
other random variables, V ar(ui,0 ) = 1−ρ σ2
2 , V ar(eit ) = σ ?
2
2. (WA, problem 11.14) Use the data in ezunem.dta. Consider a model for unemployment claims
in a city i in a year t:
ln(uclmsit ) = θt + ci + git t + δ1 ezit + uit ,
where ezit is an indicator that an enterprise zone regime was in operation.1
1 Russian equivalent is "Osobaya ekonomicheskaya zona a territory with the special tax regime that provides tax
credits and subsidies to rms that invest into the territory.
1
a) Apply a rst-dierence transformation and estimate the transformed model via xed
eects. What is the estimated eect of the enterprize zones?
b) Now let the enterprize zones aect not only level, but also the growth of unemployment
claims.
ln(uclmsit ) = θt + ci + git t + δ1 ezit + δ2 ezit t + uit
Reestimate the model.
c) Reestimate the model setting δ1 = 0. Does this model t better than the one from a)?
d) Let wi be an unobserved time-constant variable, and suppose we add β1 wi + β2 wi t to the
random growth model. Can either β1 or β2 be estimated? Explain.