You are on page 1of 5

Prof. Dr.

Kai Carstensen
Chair of Econometrics

Examination in Econometrics I

(Winter Term 2014-2015)

New examination regulation: 5 credits!!!

April 2, 2015 , 15.30 - 16.30

Preliminary remarks:

1. Please read these instructions carefully!

2. At the beginning of the exam, fill in the cover sheet and hand in after the exam is finished!

3. You are permitted to use the following auxiliary tools:

(a) a non–programable pocket calculator,


(b) the formulary for Econometrics I without notes.

4. Write your name and enrolment (matriculation) number on every sheet of paper!

5. Don’t use a pencil!

6. The exam problems are printed on 4 pages. Check your copy for completeness!

7. You have 60 minutes in total to answer the exam questions.

Good luck!
Problem 1 (18 credits)
Assume the population model

E(y|x1 , x2 ) = β0 + β1 x1 + β2 x2 , (1)

where E(x1 ) = c1 and E(x2 ) = c2 and E(x2 |x1 ) = c2 .

1. Derive E(y).

2. Derive the linear projection L(y|1, x1 ) in terms of the population parameters.

3. Assume you run the regression yi = b0 + b1 xi,1 + ui using a sample i = 1, . . . , N , where


the population model is given in equation (1). Is the OLS estimator for b1 also a consistent
estimator for β1 ? Base your answer on the linear projection in part 2.

4. Is the OLS estimator for b0 also a consistent estimator for β0 ? Base your answer on the linear
projection in part 2.

5. Assume you are only interested in inference on β1 . Is is advisable to leave out x2 from the
regression? Explain your answer in one or two sentences.

1
Problem 2 (15 credits)
An analyst wants to check the impact of the weekly returns of the major German stock market
index (DAX) on the weekly returns of the Vienna Stock market index (W BI). She applies a
linear model

W BI = β1 + β2 DAX + u, with x = [1 DAX], E(x0 u) = 0, and rank E(x0 x) = 2, (2)

where following sample moments are given from a data set from the previous two years:
   
0 104 29.153 0 −1 0.0098 −0.0005
XX= , (X X) = ,
29.153 520.757 −0.0005 0.0020
   
0 −6.294 PN 2 0 160.848 −7.853
XY = , i=1 (ûi xi xi ) = , and y 0 y = 461.632.
390.641 −7.853 932.435

1. Estimate the model via OLS.

2. Calculate the R2 and interpret it.

3. Test the null hypothesis that DAX and WBI differ not systematically, i.e., only by a mean-
zero random variable, at the 5% level.

Table:
χ2 -distribution with ν degrees of freedom
(values correspond to the 95%-quantile)
ν 1 2 3 4 5 20 100
2
χ 3.842 5.992 7.815 9.488 11.07 31.41 124.3

2
Problem 3 (12 credits)
A training program is offered to a group of unemployed persons. Participation is voluntary. The
success of the programm shall be assessed. For this reason the following linear model is used:

jobi = α + xi β + γtri + ui , (3)

where job denotes whether the initially unemployed person is employed a year later (measure
for successful reintegration into the labor market). tr is a dummy denoting whether the person
participated or not and x represents further observable variables like age, age squared, sex and
education. i is the index for the persons in the group of people who were unemployed initially.

1. Given that the intrinsic motivation to get a job varies between persons and affects their
behavior, are there any problems in estimating the partial effect of the training programm
on jobi via OLS? Explain your answer breifly.

2. Someone argues that the number of letters of application sent off in the week before the
training has been offered is a proxy for the intrinsic motivation. It would be sufficient to
include this variable into the model and the OLS estimator would be a consistent estimator.
Do you agree? Explain your answer in one or two sentences.

3. Now assume that participation is mandatory for a randomly chosen part of the group, while
the others are not allowed to do the training. Does this affect your answer to question 1?

3
Problem 4 (15 credits)
Consider the linear model

y = β1 + β2 x2 + ... + βK xK + u, with E(x0 u) = 0, and rank E(x0 x) = K, (4)

with

E(u2 |x) = γ0 + γ1 x2K . (5)

1. What are the properties of the OLS estimator for β in this model?

2. Denote the matrix Ω that you need for the GLS estimation of β.

3. Assume γ0 and γ1 are unknown. How would you proceed to obtain an FGLS estimate for β?

4. Comment briefly on the following statement: “FGLS is the better method than OLS because
it provides regularly higher R2 .”

You might also like