Scribd
Upload
43 views1 page

Final 2014

This document contains instructions for a final exam with two questions. Question 1 asks to compare two regression models using an estimate of the variance of the error terms to test which model fits the data better. Question 2 asks to analyze a 2SLS estimator obtained from a misspecified IV regression that is missing some variables, including finding the asymptotic distribution of the estimator.

Uploaded by

Adalberto Calsin Sanchez
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
43 views1 page

Final 2014

This document contains instructions for a final exam with two questions. Question 1 asks to compare two regression models using an estimate of the variance of the error terms to test which model fits the data better. Question 2 asks to analyze a 2SLS estimator obtained from a misspecified IV regression that is missing some variables, including finding the asymptotic distribution of the estimator.

Uploaded by

Adalberto Calsin Sanchez
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 1

Econometrics 710

Final Exam, Spring 2014


1. You have n iid observations (yi ; x1i ; x2i ); and consider two alternative regression models

yi = x01i 1 + e1i (1)


E (x1i e1i ) = 0

yi = x02i 2 + e2i (2)


E (x2i e2i ) = 0

where x1i and x2i have at least some dierent regressors. (For example, (1) is a wage regression
on geographic variables and (2) is a wage regression on personal appearance measurements.)
You want to know if model (1) or model (2) ts the data better. Dene 21 = E e21i and
2 = E e2 . You decide that the model with the smaller variance t (e.g., model (1)
2 2i
ts better if 21 < 22 .) You decide to test for this by testing the hypothesis of equal t
H0 : 21 = 22 against the alternative of unequal t H1 : 21 6= 22 . For simplicity, suppose that
e1i and e2i are observed.

(a) Construct an estimate b of = 2


1
2:
2
p
(b) Find the asymptotic distribution of n b as n ! 1:
(c) Find an estimator of the asymptotic variance of b.
(d) Propose a test of asymptotic size of H0 against H1 :
(e) Suppose the test accepts H0 . Briey, what is your interpretation?

2. Take the linear instrumental variables equation

yi = x01i 1 + x02i 2 + ei
E (zi ei ) = 0

where x1i is k1 1, x2i is k2 1, and zi is ` 1, with ` k = k1 + k2 : The sample size is n.


Assume that Qzz = Ezi zi0 > 0 and Qzx = Ezi x0i has full rank k:
Suppose that only (yi ; x1i ; zi ) are available, and x2i is missing from the dataset.
Consider the 2SLS estimator b 1 of 1 obtained from the misspecied IV regression, by re-
gressing yi on x1i only, using zi as an instrument for x1i .

(a) Find a stochastic decomposition b 1 = 1 + b1n + r1n where r1n depends on the error ei ,
and b1n does not depend on the error ei :
(b) Show that r1n !p 0 as n ! 1:
(c) Find the probability limit of b1n and b 1 as n ! 1.
(d) Does b suer from omitted variables bias? Explain. Under what conditions is there
1
no omitted variables bias?
(e) Find the asymptotic distribution as n ! 1 of
p
n b1 1 b1n :

You might also like