Exam EcoI PZ2 WS2122

Prof. Dr.
Kai Carstensen
Prof. Dr. Uwe Jensen
Chair of Econometrics
Examination in Econometrics I
(Winter Term 2021/22)
Examination regulation
March 21, 2022, 12:00
Preliminary remarks:
1. Please read these instructions carefully!
2. Write your name and enrollment (matriculation) number on every sheet of paper!
3. Don’t use a pencil!
4. The exam problems are listed on 3 pages. Check your exam for completeness!
5. Round your solutions to 4 decimal places.
6. For all tests use a significance level of 5%, if nothing else is specified.
7. You have 60 minutes in total to answer the exam questions.
Good luck!
Question 1 (23 points)
Consider the following model developed for analyzing individual emotional stability and
its determinants.
EmoStab = Emotional stability (0 = low, ..., 7 = high)
Illness = Number of doctor visits in 3 months
Illness2 = Illness2 /1000
Age = Age (in years)
Age2 = Age2 /1000
Lif eSat = Life satisfaction (0 = low, ..., 10 = high)
N otEmpl = 1 for not employed
LHhInc = Log of household income in euros
LLabInc = Log of labor income in euros
Height = Height (in cm)
It is known that the variation of emotional stability increases with increasing illness and
decreases with increasing age.
Based on an individual cross-section data set, a random sample of size N = 1274, a LS
estimation has led to the following results:
Robust
Variable Coeff. std. err.
Const 1.1961 0.1003
Illness -0.0382 0.0040
Illness2 0.5108 0.0940
Age 0.0213 0.0034
Age2 -0.1588 0.0325
Lif eSat 0.1793 0.0060
N otEmpl -0.1478 0.0251
1. Test the significance of the illness parameters separately.
2. Give reasons for using heteroskedasticity-robust standard errors in this example.
3. Shortly explain why measuring illness by number of doctor visits is a very rough
approximation and what the resulting error has probably done with the illness pa-
rameters, as precisely as possible (sign of the effect).
4. Adding the variable LHhInc, another LS estimation with the same data has led to
the following results:
Robust
Variable Coeff. std. err.
Const 1.1163 0.1058
Illness -0.0382 0.0040
Illness2 0.5111 0.0940
Age 0.0204 0.0034
Age2 -0.1417 0.0333
Lif eSat 0.1791 0.0060
N otEmpl -0.1114 0.0295
LHhInc 0.0079 0.0034
1
Shortly explain, as precisely as possible (sign of the effect), the change in the
N otEmpl parameter from the first to the second table.
5. Using the estimated relation
EmoStab = . . . + 0.0204 Age − 0.0001417Age2 + . . . ,
derive mathematically whether the relation between age and emotional stability is
monotonous for the individuals in the data set. Interpret your result.
6. Under which conditions can we interpret the age parameters as causal effect on
emotional stability? Shortly explain why.
7. Can we interpret the Lif eSat parameter as causal effect on emotional stability?
Shortly explain why (not).
8. A colleague is planning to use instrumental variables as remedies for the problems

she has detected in the previous items. She proposes to use LLabInc and/or Height
as instruments for Lif eSat. Shortly discuss the advantages and disadvantages of this
idea. Deal with the exogeneity and relevance of the instruments and the variance of
the IV estimator.
Height probably exogenous but not highly correlated with Lif eSat, i.e. not very
relevant. High variance of IV estimator. Requires larger sample size than we have
here for a satisfying result. (5 P.)
1. Consider a linear regression model
y i = x i β + εi
where the explanatory variables xi are assumed to be orthogonal to the error term.
(a) Set up the population orthogonality condition and the sample moment condi-
tion for GMM estimation with g(wi , θ0 ) = x0i εi .
(b) Show that in case of exact identification the GMM estimator is equivalent to
the OLS estimator.
(c) Find the asymptotic variance matrix V of the GMM estimator.
2. Given the population model y = β0 + β1 x1 x2 + β2 x22 , assume that E(x1 |x2 ) =

E(x2 |x1 ) = 0 and that x1 and x2 are independent. Derive the linear projection
L(y|1, x2 ).
2
Consider the linear regression model with scalar regressor xi and parameters θ = (β, σ 2 )0
iid
yi = xi β + ui , ui ∼ N (0, σ 2 ).
Suppose you have a random sample of i = 1, ..., N observations with N 2

P
PN i=1 xi = 40 and
i=1 yi xi = 28. You want to test the hypothesis H0 : β = 0 vs. H1 : β = 1.
1. Write down the log likelihood function for observation i and for the full sample.
2. Derive the score with respect to θ = (β, σ 2 ) for observation i.
3. Derive the Hessian Hi (θ) and its conditional expectation A(xi , θ).
4. Derive the maximum likelihood estimators β̂ and σ̂ 2 .
5. Calculate the likelihood-ratio test statistic.
6. What is your test decision for the likelihood-ratio test? You can assume σ 2 = 1.

Exam EcoI PZ2 WS2122

Uploaded by

Copyright:

Available Formats

Exam EcoI PZ2 WS2122

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Exam EcoI PZ2 WS2122

Uploaded by

Copyright:

Available Formats

Prof. Dr.

(Winter Term 2021/22)

March 21, 2022, 12:00

1. Please read these instructions carefully!

3. Don’t use a pencil!

5. Round your solutions to 4 decimal places.

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

5. Using the estimated relation

EmoStab = . . . + 0.0204 Age − 0.0001417Age2 + . . . ,

8. A colleague is planning to use instrumental variables as remedies for the problems

Question 2 (20 points)

1. Consider a linear regression model

2. Given the population model y = β0 + β1 x1 x2 + β2 x22 , assume that E(x1 |x2 ) =

Suppose you have a random sample of i = 1, ..., N observations with N 2

2. Derive the score with respect to θ = (β, σ 2 ) for observation i.

4. Derive the maximum likelihood estimators β̂ and σ̂ 2 .

5. Calculate the likelihood-ratio test statistic.

You might also like

Pfad - The Proxy pFad of © 2024 Garber Painting. All rights reserved.