Exam: Econometrics

Univ.‐Prof. Dr. Matthias Neuenkirch

2 March 2026

Q1: /30

Name: Q2: /20

Q3: /20

Matrikelnummer: Q4: /30

Q5: /20

• The exam has 5 larger questions. Each of these questions has subquestions.

• The exam has 15 pages, including this title page. Please make sure that you have a
complete version of the exam.

• Read the text of each ﴾sub‐﴿question carefully. Write legibly. You will receive no points
for illegible or unclear answers. You have to answer in English.

• The numbers in parentheses are the points for each ﴾sub‐﴿question. They indicate
how much time ﴾in minutes﴿ you should spend on each ﴾sub‐﴿question. The total
number of possible points is 120. In order to reach the total number of points you
have to answer each ﴾sub‐﴿question correctly. Use your time wisely. Try to answer
each ﴾sub‐﴿question, only then partial credit can be granted.

• The exam time is 120 minutes. You are responsible to hand over the exam properly.

• You are not to use any material other than a pen and a ruler. In addition, you can use a
small non‐programmable watch or alarm clock to pace yourself. You are not allowed
to use cell phones or calculators. Failure to comply with these rules will automatically
result in zero points.

• Please do not forget to fill out the information on top of this page and make sure that
this page is stapled to the rest of your exam. Exams without this information will not
be counted.

1
1 Multiple Choice ﴾30 points﴿
There is always one correct answer.

1. In the regression of y on x, the error term exhibits heteroskedasticity if

□ it has a constant variance.

□ V ar(y|x) is a function of x.

□ x is a function of y.

□ y is a function of x.

2. A theorem stating that the mean of a random sample converges in probability to the
population average is called

□ Law of Large Numbers.

□ Law of Iterated Expectations.

□ Large Sample Properties.

□ Central Limit Theorem.

3. If an explanatory variable is strictly exogenous it implies that

□ changes in the lag of the variable do not affect future values of the dependent
variable.
□ the variable is correlated with the error term in all future time periods.

□ the variable cannot react to what has happened to the dependent variable in
the past.

□ the conditional mean of the error term given the variable is zero.

4. Exclusion of a relevant variable from a multiple linear regression model leads to the
problem of

□ misspecification of the model.

□ multicollinearity.

□ perfect collinearity.

□ homoskedasticity.

2
5. The Gauss‐Markov theorem will not hold if

□ the error term has the same variance given any values of the explanatory vari‐
ables.
□ the error term has an expected value of zero given any values of the indepen‐
dent variables.
□ the independent variables have exact linear relationships among them.

□ the regression model relies on the method of random sampling for collection
of data.

6. A process is stationary if

□ any collection of random variables in a sequence is taken and shifted ahead

by h time periods; the joint probability distribution changes.

□ any collection of random variables in a sequence is taken and shifted ahead

by h time periods, the joint probability distribution remains unchanged.

□ there is serial correlation between the error terms of successive time periods
and the explanatory variables and the error terms have positive covariance.

□ there is no serial correlation between the error terms of successive time periods
and the explanatory variables and the error terms have positive covariance.

7. An explanatory variable is called endogenous if:

□ it is correlated with the error term.

□ it is uncorrelated with the error term.

□ its value is influenced by one or more of the independent variables within the
model.
□ none of the above

8. If the residual sum of squares in a regression analysis is 200 and the total sum of
squares is equal to 1000, what is the value of the coefficient of determination?

□ 0.25

□ 0.2

□ 0.8

□ 1.2

3
 9. The general approach to obtaining fully robust standard errors and test statistics in
the context of panel data is known as

□ confounding.

□ differencing.

□ clustering.

□ attenuating.

10. The violation of the homoskedasticity assumption in the linear probability model
leads to

□ the t and F tests to be still valid.

□ biased OLS estimators but valid standard errors.

□ no bias in the OLS estimators of the βj but generally invalid standard errors.

□ predicted probabilities to be larger than one or smaller than zero.

11. Zero correlation between two variables implies that

□ there is no relationship but it does not imply their independence.

□ there is no linear relationship but it does not imply their independence.

□ there is no linear relationship and they are independent from each other.

□ there is no relationship and they are independent from each other.

12. An equation in the simultaneous equations model satisfies the order condition for
identification if

□ the number of excluded endogenous variables from the equation is at most

as large as the number of right‐hand side exogenous variables.

□ the number of excluded endogenous variables from the equation is at least as

large as the number of right‐hand side exogenous variables.

□ the number of excluded exogenous variables from the equation is at most as

large as the number of right‐hand side endogenous variables.

□ the number of excluded exogenous variables from the equation is at least as

large as the number of right‐hand side endogenous variables.

4
13. Which of the following is true?

□ A variable has a causal effect on another variable if both variables increase or

decrease simultaneously.

□ The notion of “ceteris paribus” plays an important role in causal analysis.

□ Difficulty in inferring causality disappears when studying data at fairly high

levels of aggregation.

□ The problem of inferring causality arises if experimental data is used for anal‐
ysis.

14. Which of the following statements is true?

□ The condition Cov(z, u) = 0 can be tested statistically.

□ The condition Cov(z, x) ̸= 0 cannot be tested statistically.

□ The instrumental variables estimator is always biased if Cov(x, u) ̸= 0.

□ ̸ 0.
The OLS estimator is unbiased if Cov(x, u) =

15. A linear regression model has to be linear in

□ the residuals.

□ the variables.

□ the parameters.

□ the exogenous variables.

5
2 Econometric Concepts ﴾20 points﴿
2.1 What is a p‐value? Illustrate the derivation of the p‐value graphically.
﴾8 points﴿

6
2.2 Explain what the total sum of squares ﴾TSS﴿ measures. Describe
also the other two measures of variation that are connected
with the TSS alongside their relationship. ﴾6 points﴿

2.3 What is the difference between an unbiased and a consistent

estimator? ﴾6 points﴿

7
3 Time Series Analysis ﴾20 points﴿
3.1 Give an example of a highly persistent time series and explain
why it exhibits this feature. Calculate its variance. What is the
easiest way to detect high persistence statistically? ﴾10 points﴿

8
3.2 Interpret the following results at the 5% significance level. ﴾10
points﴿
You are interested in the interest rate setting behavior of the US Federal Reserve. Your
sample covers eight observations per year from August 1987 to December 2008. The
LHS variable is the central bank’s target rate ﴾in %﴿, Inf F orec is the four quarter ahead
expected inflation rate ﴾in %﴿, and U nempGapF orec is the four quarter ahead expected
unemployment gap ﴾in %﴿.

Estimate Std. Error t value Pr(>|t|)

(Intercept) 1.0560 0.2097 5.037 1.21e-06 ***
InfForec 1.7103 0.0798 21.433 < 2e-16 ***
UnempGapForec -1.4557 0.1044 -13.942 < 2e-16 ***
---
Signif. codes: 0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1

Residual standard error: 0.9341 on 169 degrees of freedom

Multiple R-squared: 0.8221, Adjusted R-squared: 0.82
F-statistic: 390.6 on 2 and 169 DF, p-value: < 2.2e-16

Breusch-Godfrey test for serial correlation of order up to 4

LM test = 113.43, df = 4, p-value < 2.2e-16

9
4 Panel Data Analysis ﴾30 points﴿
4.1 Formally derive the First Differences estimator for the follow‐
ing model. ﴾6 points﴿

yit = β0 + β1 x1,it + β2 x2,it + νit

νit = ai + uit

4.2 Verbally describe two different ways to derive the Difference‐

in‐Differences estimator. ﴾4 points﴿

10
4.3 Given perfect equivalence, which case yields a higher R2 , a Fixed
Effects regression or the one with a dummy variable for each
individual? Why? ﴾4 points﴿

4.4 Name the consequences of ﴾i﴿ a Fixed Effects model estimation

where Random Effects would have been appropriate and ﴾ii﴿ a
Random Effects model estimation where Fixed Effects would
have been appropriate. ﴾6 points﴿

11
4.5 Interpret the following results at the 5% significance level. Make
sure to interpret the marginal effect of trust. ﴾10 points﴿
You are interested in what drives cash in circulation ﴾CiC﴿. You estimate the following
econometric model:

yit = β0 + β1 x1,it + β2 x2,it + β3 x3,it + β4 x4,it + β5 x5,it + β6 x6,it + β7 x26,it + vit

Variable Description

yit cic_to_gdp_ln ln﴾CiC/GDP﴿, natural logarithm of the cash‐to‐GDP ratio

x1,it Interest Interest rate ﴾in %﴿
x2,it gdp_pcap_ln ln﴾GDP per capita﴿, Gross Domestic Product per capita
x3,it AgeDepRatioOld Old as % of the working‐age population
x4,it Internet Internet use as a percentage of the population
x5,it ControlCorr Control of corruption ﴾percentile rank﴿
x6,it Average_Trust Trust Index ﴾0‐1﴿
x7,it Average_Trust2 Trust squared

where i = 1, . . . , n denotes the individual country and t = 1, . . . , T denotes the time

period.

Pooling Model
Unbalanced Panel: n = 64, T = 1-18, N = 845

Coefficients:
Estimate Std. Error t value Pr(>|t|)
(Intercept) 3.962472172 1.785644432 2.2191 0.0267559 *
Interest -0.036272054 0.010863911 -3.3388 0.0008796 ***
gdp_pcap_ln -0.174850114 0.146092712 -1.1968 0.2317147
AgeDepRatioOld 0.032194145 0.012899787 2.4957 0.0127667 *
Internet -0.000049501 0.004365365 -0.0113 0.9909554
ControlCorr -0.008489485 0.004774463 -1.7781 0.0757589 .
Average_Trust 1.834783237 0.923140757 1.9875 0.0471956 *
I(Average_Trust^2) -3.124089761 1.163804784 -2.6844 0.0074137 **
---
Signif. codes: 0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1

Total Sum of Squares: 213.24

Residual Sum of Squares: 92.843
R-Squared: 0.48956
Adj. R-Squared: 0.47269
F-statistic: 29.0211 on 27 and 817 DF, p-value: < 0.000000000000000222

12
13
5 Instrumental Variables ﴾20 points﴿
5.1 Under what circumstances must an instrumental variables es‐
timation be used? What criteria must an instrument satisfy in
order to be considered a valid instrument? ﴾4 points﴿

5.2 Explain the objective of the Hausman test and the procedure
used to conduct this test. ﴾6 points﴿

14
5.3 Identification of a Structural Equation ﴾10 points﴿
Consider the following cross‐sectional model with y1 and y2 as endogenous variables
and z1 and z2 as exogenous variables. u1 and u2 satisfy the usual error term assumptions:

y1 = β10 + α1 y2 + β11 z1 + u1 ﴾1﴿

y2 = β20 + α2 y1 + β21 z1 + β22 z2 + u2 ﴾2﴿

You are initially interested in the estimator α1 .

a﴿ Why can this estimator not be consistently obtained by estimating equation ﴾1﴿ alone?

b﴿ Under which condition can you obtain a consistent estimator? Explain the procedure.

c﴿ Can you also use this approach for the estimator α2 ?

15