Faculty of Business and Law

Academic year: 2021/2022

End of Teaching Block or Date: Mock

Module Leader: Reinhold Heinlein

Module Code: UMED8M-15-2
Module Title: Introductory Econometrics
Duration: 2 hours

ON CAMPUS EXAM
Standard materials required:

Examination Answer Booklet Yes

Multiple Choice Answer Sheet No

Type of Graph Paper None

Number of graph paper sheets per student -

Additional materials required for this examination:

To be supplied by UWE Bristol
Please list any additional material and whether or not they need to be collected:

• Formulae Sheet & Statistical Tables (attached)

To be supplied by the student

Please list any additional material and whether or not they need to be collected:

• N/A

University approved calculator (non programmable) Yes

Students permitted to keep this examination paper No

Students are not permitted to turn this page over until the examination starts
Instructions to Candidates:
Candidates must answer ALL questions from Section A (50 marks),

TWO questions from Section B (50 marks).

Section A (50 marks)

1. A training manager wondered whether the length of time his trainees

revised for an examination had any effect on the marks they scored in the
examination. Before the exam, he asked a random sample of 10 of them to
estimate honestly how long, to the nearest hour, they had spent revising.
After the examination he investigated the relationship between the two
variables. The sample data provide observations on the following:
Yi = exam mark for individual i (%)
Xi = revision time by individual i (hours)
The manager believes that the relationship between revision hours and
exam mark takes the form:
Yi = B1 + B2Xi + ui
where B1 and B2 are unknown parameters. The stochastic disturbance
terms, ui, are assumed to be normally and independently distributed with
zero mean and constant variance 2.
Preliminary analysis of the sample data produces the following sample
information:

n=10

Lower case letters indicate that the variables are measured as deviations
from their respective sample means i.e xi = Xi - X .
 Use the above sample information to answer all the following questions.
Show explicitly all calculations.

a) Calculate estimates for the unknown coefficients, b1 and b2. [6 marks]

b) Present the resulting regression equation and interpret the meaning of the
estimated coefficients in this problem. [6 marks]

c) Calculate the estimated standard errors of the estimated coefficients. [5

marks]

d) Carefully explaining the procedure, test whether the slope coefficient, B 2, is

significantly greater than zero. Comment. [8 marks]

e) Generate a 95% confidence interval around the intercept. Comment. [5

marks]

f) Compute the value of R2 for the estimated regression equation and briefly
explain what the calculated value of R2 means. [5 marks]

g) Use R2 to test for evidence of a relationship between revision time and

marks. [8 marks]

h) Predict the mark for an individual who revised for 15hours. Calculate a
99% prediction interval for the predicted mark. [7 marks]
 Section B (50 marks)
Candidates should attempt to answer TWO questions.

2.
Television commercials can vary in length. The objective of a commercial
is to have as many viewers as possible to remember, and buy, the product.
An experiment was conducted to determine how the length of a
commercial is related to people’s memory of the product. 60 randomly
selected individuals watched a TV show that lasted for an hour. A brand
of toothpaste was advertised during the show: different lengths of time for
different individuals. The commercials also differed in their type: musical,
humorous, serious. After the show the viewers took a test that measured
how much they remembered of the advertised product on a scale of 1-30.
Recorded data was as follows:
testi = test score (on a 30-point test);
lengthi = length of the commercials (seconds);
musici = 1 if commercial was musical, otherwise 0
humouri = 1 if commercial was humorous, otherwise 0
seriousi = 1 if commercial was serious, otherwise 0

You initially explore the importance of length of time for explaining an

individual’s recollection of the product, ignoring the type of commercial.

testi = 2.871 + 0.233lengthi

(0.863) (0.021)
R2 = 0.6729 F(1,58) = 119.29 N =60

where the figures in parentheses denote standard errors.

a) Discuss and evaluate the results in this model. [5 marks]

 In order to explore the impact of commercial type you include the effect of
the dummy variables, humour and music. Model A is the initial regression.

testi Model A Model B Model C

2.871 1.665 2.876
constant
(3.33) (4.170 (5.09)
0.233 0.189 0.152
length
(11.1) (18.79) (9.430
4.956 2.801
music
(14.64) (3.070
3.688 1.777
humour
(11.030 (1.95)
0.060
music*length
(2.61)
0.055
humour*length
(2.34)
R2 0.6729 0.9357 0.9443
ESS 621.78 864.72 872.67

TSS 924.1 924.1 924.1

Note: t values in parenthesis in the table
b) Discuss the use of the dummy variable and the interpretation in Models
B and C. Do the results indicate that the type of commercial affects the
test score? [8 marks]

c) Perform an F test to see if the additional variables containing the type

of commercial are jointly significant in model C. [6 marks]

d) Compare and discuss which of the models is best using any test
statistics you feel appropriate, including any you may have included in
your answers to parts b) and c). [6 marks]
3. A researcher is interested in the relationship between public expenditure on
education per capita (EDUC; in thousands of US$) and gross domestic product
per capita (GDP; in thousands of US$) for a cross-section of 38 countries. She
estimates the following equation, with t-values in parentheses:
EDUCi = -0.067 + 0.058 GDPi + ei
t-value (-1.45) (16.02) R2 = 0.88

a) Heteroscedasticity is a typical problem in this kind of studies. What are the

consequences of heteroscedasticity when estimating a model with OLS? [6
marks]

b) The researcher plots residuals and squared residuals against GDP. What can
we learn about heteroscedasticity from these graphics? Explain your answer. [4
marks]
.5
.5

.4
residuals squared
0

.3
Residuals

.2
-.5

.1
-1

0 10 20 30 40 0 10 20 30 40
GDP per capita GDP per capita

c) Now we want to perform a formal test. Outline the Goldfeld-Quandt test. [5

marks]

d) To test for heteroscedasticity, the researcher orders the data according to GDP
and regresses EDUC on GDP for the lowest 14 values of GDP (regression A)
and, separately, for the highest 14 values of GDP (regression B). The results are:
A. EDUCi = 0.003 + 0.031GDPi + e1i R2 = 0.29 n=14
(2.06) (4.49) RSS = 0.007

B. EDUCi = -0.165 + 0.063 GDPi + e2i R2 = 0.72 n=14

(-0.72) (5.61) RSS = 1.517
where RSS are the residual sum of squares.
Perform the Goldfeld-Quant test. [5 marks]
e) Explain an approach that could potentially deal with the heteroscedasticity. [5
marks]

4. An analyst investigates the correlation between a stock price index and GDP.
She estimates a regression model using annual data for 27 observations from
1990-2016. She estimates the following relationship:
NYSE t =−2015+0.77 GDPt +e t
(-6.58) (19.5)

R2 = 0.938 n = 27

where the figures in parentheses denotes t-ratios and the variables are:
NYSE t NYSE Composite Stock Price Index;
GDPt GDP ($, in billions).

a) Briefly explain the problem of serial correlation and the implications for the
coefficients of the estimated relationship if it occurs. [7 marks]

b) Perform a Breusch-Godfrey LM test for autocorrelation. Outline and perform

the test using the following auxiliary regression. [8 marks]
Source SS df MS Number of obs = 25
F(3, 21) = 27.89
Model 7150674.84 3 2383558.28 Prob > F = 0.0000
Residual 1794983.93 21 85475.4253 R-squared = 0.7993
Adj R-squared = 0.7707
Total 8945658.77 24 372735.782 Root MSE = 292.36

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

GDP .0118451 .020512 0.58 0.570 -.030812 .0545022

e
L1. 1.329317 .1590793 8.36 0.000 .9984937 1.660141
L2. -.6948482 .1564505 -4.44 0.000 -1.020205 -.3694915

_cons -84.86558 164.4483 -0.52 0.611 -426.8545 257.1234

c) As an attempt to overcome the autocorrelation problem, one can transform the

model so that the error terms in the new model are independent by using
Generalised Least Squares (GLS) method. Outline this procedure. [5 marks]

d) Outline alternative methods to overcome autocorrelation. [5 marks]

 END OF EXAMINATION

Formula Sheet

For the relationship (PRF) Yi = B1 + B2 Xi + ui

we have the estimated (SRF) Yi = b1 + b2 Xi+ ei

where: ui ~ N(0,σ2)

b 1=Y −b2 X (1)

(2)

b2

σ^ 2
=
∑ e 2i (3)

n−2
(4)
σ^ =√ σ^ 2
is the standard error of the regression (5)
∑ X 2i 2
σ^ =
∑ ∑ X 2i e 2i
n∑ xi 2
n (n−2 ) ∑ x 2i (6)
Var(b1) =

σ^ 2 ∑ e 2i
=
Var(b2) =
∑ x 2i (n−2 )∑ x 2i

Se(b ) =
√ var(b1 )
1 (7)

Se(b2) =
√ var(b2 )

x = X i− X and
where i
y i =Y i−Y

95% confidence interval for bi is given by:

b i±t 0.df 025 se( bi ) (8)
 2 ESS = 1 − RSS = 1 − ∑ e 2i
R =
TSS TSS ∑ y 2i (9)

OR

To test R2:

~ F(k-1,n-k)
(10)

adjusted R2 is: (11)

where ESS is the explained sum of squares, RSS is the residual sum of squares, n is the
number of observations in the sample, k is the number of coefficients estimated (including the
constant).

95% prediction interval for Y0 at X0 is:

[ ]
2 1/2
1 ( X 0 −X )
se( Y 0 ) = σ^ +
n ∑ x2 i (12)

√
2
1 ( X 0− X̄ )
Y 0 = b1 + b2 X 0 ±t 0. 025
σ^ +
df
n ∑ x2 (13)

Tests of joint significance for a restricted and unrestricted model:

( ESSu −ESSr ) / m
F=
RSS u / n−k (14)

Or (15)

Or (16)
F ~ F( m , n-k )
where m is the number of coefficients that are zero in the restricted model
UMED8M-15-2 Page 11 of 15
UMED8M-15-2 Page 12 of 15
UMED8M-15-2 Page 13 of 15
UMED8M-15-2 Page 14 of 15
UMED8M-15-2 Page 15 of 15