ECONOMETRICS III Final Exam DNI / NIE: NIA:

22/03/2023
Name: Surname 1: Surname 2:

INSTRUCTIONS
• Fill in the data that identifies you, both on this instructions sheet and on the answer sheet.
• You are allowed a maximum of 60 minutes to answer the test. If you finish earlier, you can hand in the test and start
with the second part of the exam. In total, the maximum exam time allowed is 120 minutes.

• In each question there is only one correct answer. Correct answers are worth one point, wrong answers detract 1/2
points and questions left unanswered are worth 0.

1. To model the choice of different tomato varieties by consumers of a supermarket, we estimate a conditional
Logit model. The options to choose from are Cherry tomato, Canary tomato and Kumato, so that there
are J = 3 alternatives. If our model includes consumers’ household income (EUR) and the product price
(EUR/Kg) among the independent variables, then:
(a) . . . for each of the 3 tomato varieties, we will obtain a different coefficient estimate for the household
income variable
▶ (b) None of the other options is correct
( c ) . . . for each of the 3 tomato varieties, we will obtain a different estimate of the impact of increasing the
product price by 1 EUR/Kg on the utility of that alternative

2. We estimate a Poisson model, where the number of car accidents reported by each driver is the dependent
variable and age (in years) and income (in thousands of Euros) are the independent variables. The estimated
coefficients for age and income are −0.01 and −0.35, respectively. Which of the following statements is
correct?
(a) When age increases by one year, the fitted value of the average number of accidents reported decreases
by approximately 1%
(b) When age increases by one year, the fitted value of the variance of accidents reported is multiplied by
e−0.01
▶ ( c ) Both of the other options are correct

3. We have estimated a model for the expenditure on food, where one of the explanatory variables is the education
level, which has 3 categories (N 1, N 2, N 3, ordered from lowest to highest level). In the model, apart from
the constant term, we have introduced two dummy variables: D1 which is equal to 1 if the individual has
education level N 1 and equal to 0 otherwise, and D3 which is equal to 1 if the individual has education level
N 3 and equal to 0 otherwise. The coefficients estimated for the variables D1 and D3 are −0.2 and 0.15,
respectively. If the dependent variable is in logarithms, which of the following statements is NOT correct?
▶ (a) Keeping constant the rest of the variables, the expenditure on food of an individual with education level
N 1 is about 35% lower than the one of an individual with level N 3
(b) Keeping constant the rest of the variables, the expenditure on food of an individual with education level
N 2 is about 15% lower than the expenditure of an individual with level N 3
( c ) Keeping constant the rest of the variables, the expenditure on food of an individual with education level
N 3 is equal to the one of an individual with education level N 1 multiplied by e0.35

Remember that we can only do the interpretation in % if the associated betas are in between -0.2
and 0.2, so that A would be correct if said: multiplied by e^-0.35

Permuta n. 1 – Soluzioni – Pagina 1

 4. We have estimated the following wage equation:

Y = 0.2 + 0.07X − 0.11D − 0.002(D ∗ X) + û

where Y is the logarithm of wage, X is age in years and D is a dummy variable that takes value 1 for females
and 0 for males. Which of the following statements is correct?
(a) For any individual between 20 and 50 years old, one additional year of age implies an increase in wage
▶ (b) For a male and a female 45 years old, an additional year of age implies changes in the wage that have
different signs, depending on the gender If age>35, the effect of age is positive for a male
( c ) For any female, one additional year of age always impliesand negative
a decrease for a female
in wage

5. Consider a Logit model. In which of the following situations will the marginal effect of a change of a variable
be the same for two different individuals?
Because the marginal effect deppends on the density.
(a) Never If Pr(y=1\mid x_1'\beta) is 0.2 and for the other 0.8, as the normal is symmetric
this means that -x_1'\beta=x_2'\beta. Therefore the density is the
same
(b) If the values of all explanatory variables, except the one for which the marginal effect is being calculated,
are the same for the two individuals

▶ ( c ) When the probability of choosing alternative 1 is 0.2 for the first individual and 0.8 for the second
individual

6. We estimate a conditional Logit model, with the choice of holiday type (relaxing, cultural, adventurous,
foodie) as a dependent variable and the price of each holiday plan among the explanatory variables. The
price is measured in Euro and introduced with a quadratic profile: β1 price + β2 price2 . Which of the following
statements is correct?
(a) If price increases by 1 Euro for all four alternatives, the fitted probabilities of each alternative will not
change Note that in here we use a quadratic profile, so that the effect deppends
on the price level
▶ (b) The change in fitted probabilities of the relaxing alternative when its price increases by 1 Euro (and all
others remain constant) may be different from the change in fitted probabilities of the cultural alternative
when its price increases by 1 Euro (and all others remain constant)
( c ) If price increases by 1% for all four alternatives, the fitted probabilities of each alternative will not change

7. In a panel data consumption model for a sample of households interviewed 5 different times, we have introduced
as an explanatory variable the yearly aggregate consumer price index (CPI). Which of the following methods
allows us to estimate the coefficient of the CPI?
(a) Within-groups estimator
(b) Between-groups estimator
▶ ( c ) Both of the other options are correct

′
8. We are estimating a panel data model of the type Yit = Xit β +αi +εit , where Y is the grade point average and
the vector X contains a number of individual characteristics of the student. If we suspect that the student’s
unobserved individual characteristics, αi , are important determinants of the grade point average:
(a) A within-group model will allow us to obtain consistent estimates of the β vector, and a between-group
model will do so too, if the αi terms have an impact on Yit but are independent between students
▶ (b) A within-group model will allow us to obtain consistent estimates of the β vector, and a between-group
′
model will do so too, if αi has an impact on Yit but is not correlated with the vector Xit

( c ) Only a within-group model will allow us to obtain consistent estimates of the β vector, if αi has an impact
on Yit , while a between-group model will not be able to deliver consistent estimates

The within group provides consistent estimates, but the between one will not be able to
eliminate those individual characteristics. For its estimates to be consistent we need
the individual effects to be incorrelated with the explanatory variables

Permuta n. 1 – Soluzioni – Pagina 2

 9. We have estimated a Tobit model and the coefficient of the explanatory variable X is 0.5. Which of the
following statements on the marginal effect of the dependent variable (Y ) with respect to the variable X
could be correct, for a specific individual? They should have the same sign
(a) The part of the marginal effect that is produced through changes in the probability that Y > 0 is 0.3
and the part that is produced through the effect on E(Y |Y > 0) is −0.1

(b) The part of the marginal effect that is produced through changes in the probability that Y > 0 is 0.2
and the part that is produced through the effect on E(Y |Y > 0) is 0.4
▶ ( c ) None of the other options is correct The overall effect will be beta*normal cdf.
The maximum value this can take is 0.5 and the minimum is 0.
10. Which of the following statements about the Probit model is correct?
(a) The value of the likelihood function of a Probit is always inferior to the value of the likelihood function
of a Logit, when the variables in the two models are the same
(b) For all individuals, the sum of the fitted probabilities of choosing each alternative is equal to the number
of individuals in the sample who chose alternative 1
▶ ( c ) The value of the likelihood function always lies between 0 i 1

11. The negative binomial of Type I and the negative binomial of Type II model differ in that:
(a) The negative binomial of Type I is based on the assumption of equidispersion, while the negative binomial
of Type II is not
▶ (b) None of the other two options is correct
( c ) The negative binomial of Type I models the variance of the dependent variable yi as a constant, while
the negative binomial of Type II models it as a random variable

12. We have estimated a Tobit model in which the dependent variable is the expense in alcoholic drinks and the
only explanatory variable, apart from the constant term, is the logarithm of income. Which of the following
instructions allows us to obtain marginal effects as the elasticity of the expense in alcoholic drinks with respect
to income?
(a) margins, eyex(*)
Note that X is already in logs, so we do not need
eyex
(b) margins, eyex(*) predict(ystar(0,.))
▶ ( c ) margins, eydx(*) predict(ystar(0,.))

13. We have estimated a model explaining the number of doctor visits through Poisson, Negative Binomial Type
I and Negative Binomial Type II. If the value of the logarithm of the likelihood function of the Poisson model
is −150 and the one of the Negative Binomial Type I is −130, which of the following values for the value of
the log of the likelihood function of the Negative Binomial Type II is not possible?
▶ (a) −155
(b) −140
( c ) Both of the other options are possible

14. A researcher is interested in estimating the effect of the birth of an additional child on the mother’s supply
of hours in the labor market. The dataset used consists of a panel of 1,000 working women, each of which is
observed over 10 years. Denote X̄ the overall average of hours, supplied by the 1,000 women over the 10 years
and X̄i the average of hours supplied over the 10 years by each woman i. Which of the following options best
describes the concept of within-group variation, in this application?
(a) The variation in X̄i respect to X̄
▶ (b) The variation in yearly hours of labor supplied by each woman with respect to her own average, X̄i
( c ) The variation in yearly hours of labor supplied by each woman with respect to X̄
The within group estimation is constructed with a demeaned variable, computed as the average over tim
for each individual i

Permuta n. 1 – Soluzioni – Pagina 3

15. Suppose you are interested in the effect of a change in income (Xi ) on the probability of going on holidays
abroad (Yi = 1 if the individual went on holidays abroad that year and 0 otherwise). The specification of
your Probit model is Yi = α + β1 Xi + β2 Xi · Fi + β3 Fi + εi , where Fi is a dummy = 1 if Female and 0 if Male.
If βˆ2 is estimated to be very close to 0, that means:
(a) . . . that for females, the impact of a change in income on the probability of going on holidays abroad is
negligible

▶ (b) None of the other options is correct

( c ) . . . that for a female earning no income, the probability of going on holidays abroad is approximately
and on average the same as for a male earning no income

It means that the effect of additional income is the same for males and females

Permuta n. 1 – Soluzioni – Pagina 4