7SSMM700 Quantitative Methods for Finance

and Data Analytics

Solution to class exercise

Marina Dolfin

King’s Business School

King’s College London

Week 3

M.Dolfin (King’s Business School King’s College London)

M.Sc. Banking and Finance Week 3 1 / 22
 Question 1

Consider the following model:

it = α + β 1 inflationt + β 2 outputgapt + ε t

where:
it is the targeted interest rate;
inflationt is the inflation gap (difference between the targeted and the
actual);
outputgapt is the difference between the actual and the potential
output;
ε t is a random error term.

You estimate the model via OLS using quarterly data from 2000 to the
end of 2020.

M.Dolfin (King’s Business School King’s College London)

M.Sc. Banking and Finance Week 3 2 / 22
 Question 1 (contn’d)

M.Dolfin (King’s Business School King’s College London)

M.Sc. Banking and Finance Week 3 3 / 22
 Question 1 (contn’d)

You decide to test the hypothesis H0 : β 1 = 0.5.

Which test would you use?

Perform the test and reach a conclusion.

M.Dolfin (King’s Business School King’s College London)

M.Sc. Banking and Finance Week 3 4 / 22
 Solution

We have a single hypothesis involving one coefficient, so both the t-test

and the F -test can be in principle used. In this example use the t −test.

Decide the null. If no other info about the parameter is given then set:

H0 : β 1 = 0.5 vs HA : β 1 6= 0.5

where HA is the alternative hypothesis.

Calculate the statistic:

( β̂ 1 − β H0 )
t − statistic = =
SE ( β̂ 1 )
0.15451 − 0.5
= = −5.81339
0.05943

M.Dolfin (King’s Business School King’s College London)

M.Sc. Banking and Finance Week 3 5 / 22
 Solution

Decide a level of significance of the test, say 5% given the sample size.

The degrees of freedom is T − k = 84 − 3, hence the critical value is 1.99.

Rejection rule: the null hypothesis is rejected since

|t-statistic| > critical value

Conclude that the coefficient is statistically different from 0.5.

Also note that the conclusion would be the same with other standard levels
of significance, however less likely to reject as the significance level reduces.

M.Dolfin (King’s Business School King’s College London)

M.Sc. Banking and Finance Week 3 6 / 22
 Question 2

Assume you add the first lag of asset price volatility, S (−1), to the set of
regressors.

Results are given in Table 2 below.

Why is the residual sum of squares lower than in Table 1?

Use this information to perform a test on the null hypothesis that this
variable is statistically significant.

M.Dolfin (King’s Business School King’s College London)

M.Sc. Banking and Finance Week 3 7 / 22
 Question 2 (contn’d)

M.Dolfin (King’s Business School King’s College London)

M.Sc. Banking and Finance Week 3 8 / 22
 Solution

The RSS is lower as we are estimating the model with one less restriction.

Another way of saying the same is that we have one regression more; this
reduces the RSS by definition.

M.Dolfin (King’s Business School King’s College London)

M.Sc. Banking and Finance Week 3 9 / 22
 Solution
The RSS can be used to build the F -test.

The F test is:

H0 : β 3 = 0 vs HA : β 3 6= 0

(RRSS − URSS ) DF
F -statistic = × =
URSS g
(33.16178 − 31.42886) 80
= × = 4.41101
31.42886 1
where:
RRSS = Restricted Residual Sum of Squares
URSS = Unestricted Residual Sum of Squares
DF = Degrees of freedom
g =Number of restrictions
M.Dolfin (King’s Business School King’s College London)
M.Sc. Banking and Finance Week 3 10 / 22
 Solution

Decide a level of significance of the test, say 5% given the sample size.

Rejection rule: the null hypothesis is rejected since

F-statistic > critical value

Conclude that the coefficient is statistically different from 0 at 5% s.l.

We choose to use model in table 2 where the first lag of asset price
volatility must be added to the set of regressors. Note it is the same
conclusion as using the t-test.

Also note that the conclusion would be the same with 10% s.l. but might
not be the same at 1% s.l.

M.Dolfin (King’s Business School King’s College London)

M.Sc. Banking and Finance Week 3 11 / 22
 Question 3

Given the results of the two models that follow:

M.Dolfin (King’s Business School King’s College London)

M.Sc. Banking and Finance Week 3 12 / 22
 Question 3 (contn’d)

M.Dolfin (King’s Business School King’s College London)

M.Sc. Banking and Finance Week 3 13 / 22
 Question 3 (contn’d)

M.Dolfin (King’s Business School King’s College London)

M.Sc. Banking and Finance Week 3 14 / 22
 Question 3 (contn’d)

Is the market using all available information? Use the F-test

M.Dolfin (King’s Business School King’s College London)

M.Sc. Banking and Finance Week 3 15 / 22
 Solution

The unrestricted regression as the one in which the coefficients are

freely determined by the data:

rj,t − rrf ,t = αj + β j (rm,t − rrf ,t ) + γj Debtj,t + ej,t , (1)

and the restricted regression is the one in which the coefficients are
restricted, i.e. the restrictions are imposed on some βs:

rj,t − rrf ,t = αj + β j (rm,t − rrf ,t ) + ej,t , (2)

The RSS from model (2) will be higher than the RSS from model (1)
because a restiction has been imposed!

M.Dolfin (King’s Business School King’s College London)

M.Sc. Banking and Finance Week 3 16 / 22
 Calculating the F statistic

The test statistic is given by

RRSS − URSS T −k
×
URSS g
where
URSS = RSS from unrestricted regression
RRSS = RSS from restricted regression
g = number of restrictions
In our case:
121.576 − 117.977 120 − 3
× = 3.569195691
117.977 1

M.Dolfin (King’s Business School King’s College London)

M.Sc. Banking and Finance Week 3 17 / 22
 Is gamma equal to zero?

Decide the significance level of the test:

This depends on the number of observations: The higher the number of

observations the more the evidence you want against the null hypothesis
to reject it, so use 5% (be more confident whether reject/fail to reject).

M.Dolfin (King’s Business School King’s College London)

M.Sc. Banking and Finance Week 3 18 / 22
M.Dolfin (King’s Business School King’s College London)
M.Sc. Banking and Finance Week 3 19 / 22
 Solution

Decide
Do not reject at the 5% and 1% s.l.
Reject at the 10% s.l.

M.Dolfin (King’s Business School King’s College London)

M.Sc. Banking and Finance Week 3 20 / 22
 Single linear restrictions

It is easy to show that, when g = 1,

(tn )2 = Fn .

Finally, in applied work it is very important to specify the alternative

hypothesis appropriately.

Taking into account the one-sided nature of the alternative hypothesis

results in a power improvement of our tests, i.e. increases the probability
of rejecting H0 when it is false.

M.Dolfin (King’s Business School King’s College London)

M.Sc. Banking and Finance Week 3 21 / 22
 Is the market using all available information?

Note that when g = 1,

(tn )2 = Fn .
Indeed:
(−1.889389615)2 ' 3.569195691

Hence the decision must be the same

M.Dolfin (King’s Business School King’s College London)

M.Sc. Banking and Finance Week 3 22 / 22