You are on page 1of 7

Econ 301 - Econometrics I

Fall 2021 - BUT Exam

Name & Surname: OzU ID:

The Part I has 10 True/False & Fill in the Blanks questions. Each question is
worth 3 pts.

For Part II, use the space below to write your answers.

Computers, phones, smart watches are NOT permitted.

Lecture notes, cheat sheets, other course materials and a calculator are allowed.

Total Time : 2 hours


Part I - True/False & Fill in the Blanks (30 Points)
1) If you add an omitted variable into a regression model, the coefficients of the remaining variables
can change, yet they cannot lose significance.

2) Compared with the unrestricted regression, estimation of a least squares regression under a
restriction (say β2 = β3 = 0) will result in a higher R2 if the constraint is true.

3) I run a regression of y on x and I save the residuals  = y − ŷ. If I find that Cov(x, e) = 0, I
have right to conclude that the variable x was exogenous in my regression.

4) Since x3 is an exact function of x, we will be faced with the exact multicollinearity if we attempt
to use both x and x3 as regressors.

5) If I estimate the regression model without intercept, then it is not likely that I fall into dummy
variable trap.

6) Suppose that you are trying to find the parameters of a Cobb-Douglas production function

Qi = β0 Lβi 1 Kiβ3 .

Because the functional form is multiplicative and nonlinear, there is no way to estimate the model
using the linear regression toolbox.

7) Overfitting a model leads to loss of efficiency of the estimators.

8) To test the joint hypothesis that the population coefficients are both zero, the error terms should
be homoskedastic.

9) A good regression model will always have an adjusted-R2 above 0.5.

10) offers a plausible way eliminate the bias from simultaneous causal-
ity, error-in-variables, omitted unobserved variables.

Page 2
PART II - Questions (70 Points)
Question 1 (20 Points) You are trying to understand the impact of education on the fertility,
you have a data set that includes information for a sample of women in Botswana: their education
(variable educ), their age (variable age), and the number of children they have (variable children).
You run three regression models and found the output below.

Table 1: Regression Output


Dependent variable: Children
(1) (2) (3)
Education −0.210∗∗∗ −0.090∗∗∗ −0.091∗∗∗
(0.008) (0.006) (0.006)

Age 0.175∗∗∗ 0.332∗∗∗


(0.003) (0.017)

Age2 −0.003∗∗∗
(0.0003)

Constant 3.496∗∗∗ −1.997∗∗∗ −4.138∗∗∗


(0.056) (0.094) (0.241)

Observations 4,361 4,361 4,361


R2 0.137 0.560 0.569
Adjusted R2 0.137 0.559 0.568
Residual Std. Error 2.064 (df = 4359) 1.475 (df = 4358) 1.460 (df = 4357)
Note: ∗ p<0.1; ∗∗ p<0.05; ∗∗∗ p<0.01

a) Interpret the models. Do you really need to estimate the second and third model? (5 points)

Page 3
b) Do you think the absence of Age2 in Model (2) causes omitted variable bias? (5 points)

c) How Age affects the fertility of a woman? Provide an explanation using the results from Model
(3)? (5 points)

d) Construct a 95% confidence interval for the intercept of the first model. (round the numbers to
make the calculation easier) (5 points)

Page 4
Question 2 (50 Points) Consider a regression model to explain the salaries of CEOs in terms of
annual firm sales:

ln(salary) = β0 + β1 ln(sales) + β2 roe + β3 negros + ε

where salary is CEO’s salary in thousands of TL, sales is the firm’s sales in millions of TL, roe
is firm’s return on equity and negros is a dummy variable which is equal to 1 if return on firm’s
stock is negative.

a) State whether each of salary, sales, roe, negros, β0 , β1 , β2 and β3 is parameter (P), dependent
variable (DV) or independent variable (IDV): salary : , sales : , roe : , negros : ,
β0 : , β1 : , β2 : , β3 : . (5 points)

b) Interpret the coefficients. What is the impact of sales on salary? What is the impact of
roe on salary? (10 points)

c) Do you think that the regression model could possibly suffer from outlier problem? If it
does, what would be your solution? (5 points)

Page 5
d) Your friend claims that roe and negros are likely to be correlated. What happens if the
correlation among those variables is high? What happens if the absolute value of the correlation
coefficient is 1? What happens if the absolute value of the correlation coefficient is 0? (10 points)

e) How would you detect the multicollinearity? Discuss the alternative detection methods? (5
points)

Page 6
f ) Do you think the variance of the error terms (of your regression model) are constant across
the values of the independent variables? Does it create a problem if the variance of error terms
varies? If it does, how would you solve that problem? (10 points)

g) Why should/should not I include posros, a dummy variable which is equal to 1 if return on
firm’s stock is positive to the regression model? (5 points)

Page 7

You might also like