BUSINESS SCHOOL

RESEARCH METHODS

CMSE11086

Exam Date: March 4th, 2017

From and To: 10.00 - 11.00

Please read full instructions before commencing writing

Please make sure you read carefully and understand the instructions below and that you comply with
these fully.

1. Use only a HB pencil (provided).

2. Write your matriculation number. Using the HB pencil provided, write your name, the name of
the Examination and the date in the shaded box on the multiple choice answer sheet.

3. Each question contains five possible answers, only one of which is correct. Place a firm horizontal
pencil mark in the appropriate gap, e.g., if the answer to question 5 was C.

Q5 [A] [B] [C] [D] [E]

4. Faint lines will not be read! Very thin marks made with a sharp, hard pencil may not be read. If a
soft pencil is used there should be no need to press heavily. This should make rubbing out easy.
Erasures must be completed clean without smudging. An unclean or incomplete erasure may be
read as an answer!

5. Marks are given for correct answers. No marks will be subtracted for a wrong answer. No mark is
either given or subtracted for:
 no attempt
 if a line is placed through more than one option
 The multiple choice answer sheets are machine marked and that mark is final.
Special instructions
 NONE

Special items
 NONE

Convenor of Board: Roberto Rossi

Answer ALL the questions below. Questions are equally weighted.

1. In a study relating college grade point average (GPA) to time spent in various activities,
you distribute a survey to several students. The students are asked how many hours
they spend each week in four activities: studying, sleeping, working, and leisure. Any
activity is put into one of the four categories, so that for each student, the sum of
hours in the four activities must be 168. In order to examine the effect of these
activities on GPA, what is the correct way to specify the model?

A. 𝐺𝑃𝐴𝑖 = 𝛽0 + 𝛽1 × 𝑠𝑡𝑢𝑑𝑦𝑖 + 𝛽2 × 𝑠𝑙𝑒𝑒𝑝𝑖 + 𝛽3 × 𝑤𝑜𝑟𝑘𝑖 + 𝛽4 × 𝑙𝑒𝑖𝑠𝑢𝑟𝑒𝑖 + 𝑢𝑖

B. 𝐺𝑃𝐴𝑖 = 𝛽1 × 𝑠𝑡𝑢𝑑𝑦𝑖 + 𝛽2 × 𝑠𝑙𝑒𝑒𝑝𝑖 + 𝛽3 × 𝑤𝑜𝑟𝑘𝑖 + 𝛽4 × 𝑙𝑒𝑖𝑠𝑢𝑟𝑒𝑖 + 𝑢𝑖
C. 𝐺𝑃𝐴𝑖 = 𝛽0 + 𝛽1 × 𝑠𝑡𝑢𝑑𝑦𝑖 + 𝛽2 × 𝑠𝑙𝑒𝑒𝑝𝑖 + 𝛽3 × 𝑤𝑜𝑟𝑘𝑖 + 𝑢𝑖
D. A and C.
E. B and C.

2. A researcher is interested in the effect of university degree on wages and runs the
following regression:

𝑤𝑎𝑔𝑒𝑠 = 𝛽1 × 𝑢𝑛𝑖 + 𝛽2 × 𝑛𝑜𝑡𝑢𝑛𝑖 + 𝑢𝑖

where uni takes value 1 for those with a university degree and 0 otherwise and notuni
takes value 1 for those without a university degree and 0 otherwise.
What can you say about this regression?

A. ̂1 and 𝛽
𝛽 ̂2 are the average wages for university graduates and non-graduates,
respectively.
B. It suffers from perfect multicollinearity and therefore 𝛽1 and 𝛽2 cannot be
estimated.
C. ̂1 − 𝛽
(𝛽 ̂2 ) is the effect of university degree on wages.
D. ̂1 is the effect of university degree on wages.
𝛽
E. Both answers A and C.
3. You perform the following regression

𝑤𝑎𝑔𝑒𝑠 = 𝛽0 + 𝛽1 × 𝑒𝑑𝑢𝑐 + 𝛽2 × 𝑓𝑒𝑚𝑎𝑙𝑒 + 𝛽3 × 𝑒𝑑𝑢𝑐 × 𝑓𝑒𝑚𝑎𝑙𝑒 + 𝑢

where educ are years of education and female is a dummy that takes value 1 for
females and 0 for males.
How would you test that the regression result for women is the same as the
regression result for men?

A. Performing the test H0: 𝛽3 = 0.

B. Performing a joint test with H0: 𝛽1 = 𝛽2 = 𝛽3 = 0
C. Performing a regression for men; i.e. 𝑊𝑎𝑔𝑒𝑠 = 𝛽0𝑀 + 𝛽1𝑀 × 𝑒𝑑𝑢𝑐 + 𝑢 and a
regression for women; i.e. 𝑊𝑎𝑔𝑒𝑠 = 𝛽0𝐹 + 𝛽1𝐹 × 𝑒𝑑𝑢𝑐 + 𝑢
and testing H0 : 𝛽1𝐹 = 𝛽1𝑀
D. Performing a joint test with H0 : 𝛽2 = 𝛽3 = 0
E. Both answers C and D.

4. A regression of average weekly earnings (AWE) on 𝑎𝑔𝑒, 𝑎𝑔𝑒 2 and 𝑎𝑔𝑒 3 produces the
following results

̂ = 410 + 1 × 𝑎𝑔𝑒 − 0.01 × 𝑎𝑔𝑒 2 + 0 × 𝑎𝑔𝑒 3

𝐴𝑊𝐸
R-square=0.08 and n = 350.

What is the difference in earnings for a worker who is 21 years old compared to a
worker who is 20 years old?

A. 0.99.
B. 1.41.
C. 0.59.
D. 0.41.
E. None of the above.
5. A researcher who is interested in the gender wage gap runs a regression of annual
salary (AS) on a female dummy variable (1 = female; 0 = male), a university degree
dummy variable (1 = respondent has a university degree; 0 = respondent does not
have a university degree), and their interaction. This regression, where annual salary is
measured in thousands of pounds, produces the following results:

̂ = 20 − 5 × 𝑓𝑒𝑚𝑎𝑙𝑒 + 20 × 𝑑𝑒𝑔𝑟𝑒𝑒 + 3 × 𝑓𝑒𝑚𝑎𝑙𝑒 × 𝑑𝑒𝑔𝑟𝑒𝑒

𝐴𝑆
R-square= 0.08 and n = 350

What is average annual salary for men without a university degree, women without a
university degree, men with a university degree, and women with a university degree?

A. 20; 15; 40; 38.

B. 20; 15; 20; 23.
C. 20; 15; 40; 43.
D. 40; 35; 60; 43.
E. None of the above.

6. Choose the correct answer (s)

A. logs have a convenient percentage/elasticity interpretation

B. Slope coefficients of logged variables are invariant to rescalings
C. Taking logs often eliminates/mitigates problems with outliers
D. Taking logs often helps to secure normality and homoscedasticity
E. All of the above

7. Suppose the following model:

𝑦 = 𝛽0 + 𝛽1 𝑥1 + 𝛽2 𝑥2 + 𝛽3 𝑥1 𝑥2 + 𝑢

How can you test whether 𝑥2 is statistically significant when 𝑥1 is at its 75th percentile
(𝑞1,75 )?

A. Running 𝑦 = 𝛿0 + 𝛿1 𝑥1 + 𝛿2 𝑥2 + 𝛽3 (𝑥1 − 𝑞1,75 ) × (𝑥2 − 𝑞2,75 ) + 𝑢 and testing

H0 ∶ 𝛽3 = 0. Note that 𝑞2,75 is the 75th percentile of 𝑥2 .
B. Running 𝑦 = 𝛿0 + 𝛿1 𝑥1 + 𝛿2 𝑥2 + 𝛽3 (𝑥1 − 𝑞1,75 ) × (𝑥2 − 𝑞2,75 ) + 𝑢 and testing
H0 ∶ 𝛿2 = 0. Note that 𝑞2,75 is the 75th percentile of 𝑥2 .
C. Running 𝑦 = 𝛿0 + 𝛿1 𝑥1 + 𝛿2 𝑥2 + 𝛽3 (𝑥1 − 𝑞1,75 ) × (𝑥2 − 𝑞2,75 ) + 𝑢 and testing
H0 ∶ 𝛿2 = 𝛽3 = 0. Note that 𝑞2,75 is the 75th percentile of 𝑥2 .
D. Running 𝑦 = 𝛿0 + 𝛿1 𝑥1 + 𝛿2 𝑥2 + 𝛽3 (𝑥1 − 𝑞1,75 ) × (𝑥2 − 𝑞2,75 ) + 𝑢 and testing
H0 ∶ 𝛿0 = 𝛿2 = 𝛽3 = 0. Note that 𝑞2,75 is the 75th percentile of 𝑥2 .
E. None of the above.
9. Consequences of measurement error in the dependent variable are:

A. Estimates will be less precise because the variance will be higher

B. OLS will be unbiased as long as the measurement error is unrelated to the values
of the explanatory variables
C. Sample selection
D. A and B
E. A, B and C

10. Suppose you run the following model

𝑌𝑡 = 𝛽0 + 𝛽1 × 𝑌𝑡−1 + 𝛽2 × 𝑌𝑡−2 + 𝛽3 × 𝑌𝑡−3 + 𝑢𝑡

But the “true model“ includes an additional lag of Y as an explanatory variable; i.e.
𝑌𝑡−4 . What are the consequences of running the three-lags model?

A. Strict Exogeneity will not hold.

B. Omitted Variable bias will follow.
C. Serial Correlation will follow
D. A, B and C
E. A and B
11. Using the following information and STATA commands answer the following question:

Krupp and Polard (1996) analysed the effects of antidumping filings by US chemical
industries on imports of various chemicals. This question focuses on one chemical,
barium chloride. There are 3 dummy variables; i.e. befile6 is equal to 1 during 6
months before filing, affile6 indicates the 6 months after filing, and aftdec6 denotes
six months after the positive decision. The dependent variable is the volume of
imports of barium chloride from China, chnimp. chempi is an index of chemical
production, gas is the volume of gasoline production and rtwex is and exchange rate
index.

reg lchnimp lchempi lgas lrtwex befile6 affile6 afdec6

Source SS df MS Number of obs = 131
F(6, 124) = 9.06
Model 19.4051607 6 3.23419346 Prob > F = 0.0000
Residual 44.2470875 124 .356831351 R-squared = 0.3049
Adj R-squared = 0.2712
Total 63.6522483 130 .489632679 Root MSE = .59735

lchnimp Coef. Std. Err. t P>|t| [95% Conf. Interval]

lchempi 3.117193 .4792021 6.50 0.000 2.168718 4.065668

lgas .1963504 .9066172 0.22 0.829 -1.598099 1.9908
lrtwex .9830183 .4001537 2.46 0.015 .1910022 1.775034
befile6 .0595739 .2609699 0.23 0.820 -.4569585 .5761064
affile6 -.0324064 .2642973 -0.12 0.903 -.5555249 .490712
afdec6 -.565245 .2858352 -1.98 0.050 -1.130993 .0005028
_cons -17.803 21.04537 -0.85 0.399 -59.45769 23.85169

predict residuals, residuals

predict fitted, xb
gen residuals2=residuals^2
gen fitted2=fitted^2
reg residuals2 fitted fitted2

Source SS df MS Number of obs = 131

F(2, 128) = 4.07
Model 2.62040837 2 1.31020418 Prob > F = 0.0193
Residual 41.1704635 128 .321644246 R-squared = 0.0598
Adj R-squared = 0.0451
Total 43.7908719 130 .33685286 Root MSE = .56714

residuals2 Coef. Std. Err. t P>|t| [95% Conf. Interval]

fitted -2.782466 4.415647 -0.63 0.530 -11.51958 5.954647

fitted2 .19655 .3582087 0.55 0.584 -.5122272 .9053272
_cons 9.995657 13.56665 0.74 0.463 -16.84828 36.83959
test fitted fitted2
( 1) fitted = 0
( 2) fitted2 = 0

F( 2, 128) = 4.07
Prob > F = 0.0193

A. I reject the null hypothesis of homoscedasticity at 5%. Thus there is evidence in

favour of heteroskedasctity.
B. I do not reject the null hypothesis of homoscedasticity at 5%. There is no
evidence of heteroskedasticity
C. I reject the null hypothesis of model misspecification at the 5%. There is
evidence in favour of the model being incorrectly specified.
D. I do not reject the null hypothesis of model misspecification at the 5%. There is
no evidence of model mispecification.
E. There is evidence of autocorrelation.
12. Choose the correct answer (s):

A. OLS estimators are still unbiased and consistent if errors are serially correlated.
B. Correctness of R-squared does not depend on serial correlation.
C. OLS standard errors and tests will be invalid if there is serial correlation.
D. OLS will not be efficient anymore if there is serial correlation.
E. All of the above

13. Answer True or False

A regression model can be useful for forecasting even if none of its coefficients have
causal interpretations.

A. True
B. False

14. The AR(p) model

A. is defined as 𝑌𝑡 = 𝛽0 + 𝛽𝑝 𝑌𝑡−𝑝 + 𝑢𝑡
B. is defined as: 𝑌𝑡 = 𝛽0 + 𝛽1 𝑋𝑡−𝑝 + 𝛽𝑝 𝑌𝑡−𝑝 + 𝑢𝑡
C. is defined as: 𝑌𝑡 = 𝛽0 + 𝛽𝑝 𝑌𝑡−𝑝 + 𝑢𝑡−𝑝
D. is defined as 𝑌𝑡 = 𝛽0 + 𝛽1 𝑌𝑡−1 + 𝑢𝑡−𝑝
E. None of the Above
15. Stationarity means that the

A. error terms are not correlated.

B. probability distribution of the time series variable does not change over time.
C. time series has a unit root.
D. forecasts remain within 1.96 standard deviation outside the sample period.
E. A and B

16. The random walk model is an example of a

A. deterministic trend model.

B. binomial model.
C. stochastic trend model.
D. stationary model.
E. a cross-sectional model.

17. Answer True or False for each one of the following statements:

i. The minimum value of the winsorised data is larger (or equal) to the minimum value
of the original series.
ii. A winsorisation of 10% results in 10% of the original observations being changed.
iii. The variance of winsorised data is smaller (or equal) to that of the original data.

A. (i) True, (ii) True, (iii) True

B. (i) True, (ii) False, (iii) False
C. (i) True, (ii) False, (iii) True
D. (i) False, (ii) True, (iii) False
E. (i) True, (ii) True, (iii) False

END OF THE EXAM PAPER