Quantitative Methods for Finance and Data Analytics

Solution to class exercise

Marina Dolfin

King’s Business School

King’s College London

Week 4

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

M.Sc. Banking and Finance Week 4 1 / 15
 Question 1

You estimate the following model:

excess returnt = α + βexcess return on the markett + et

where excess returnt is the difference between return on asset A and the
risk-free rate, excess return on the markett is the difference between
return on the market portfolio and the risk-free rate, and et is a random
error term. Data are for 174 months. You estimate the model via OLS;
results are reported in Table 1.

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

M.Sc. Banking and Finance Week 4 2 / 15
 Question 1 (cont’nd)

You want to evaluate whether your model is affected by a problem of first

order serial correlation. Which statistics do you take into consideration
among those in the estimation output? Comment.
M.Dolfin (King’s Business School King’s College London)
M.Sc. Banking and Finance Week 4 3 / 15
 Solution

The statistic taken into consideration is the Durbin-Watson statistics. The

DW test statistic is part of the standard regression output and is a test for
the existence of a first-order serial correlation versus no correlation. Notice
that the test has the null hypothesis that residuals are not correlated.

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

M.Sc. Banking and Finance Week 4 4 / 15
 Solution

The statistic can be written as DW ≈ 2(1 − ρ̂), where ρ̂ is the estimated

correlation coefficient for the residuals.
In fact:
T T T T
∑ (ût − ût −1 )2 = ∑ ût2 + ∑ ût2−1 − 2 ∑ ût ût −1 ' (1)
t =2 t =2 t =2 t =2

the first two terms of the RHS differ only in the first and last terms, so (as
T → ∞)
T T
2 ∑ ût2 − 2 ∑ ût ût −1 (2)
t =2 t =2

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

M.Sc. Banking and Finance Week 4 5 / 15
 Solution

2 ∑T 2 T
t =2 ût − 2 ∑t =2 ût ût −1

∑T
t =2 ût ût −1

DW ' = 2 1 − = (3)
∑T 2
t =2 ût ∑T t =2 ût
2

 (T − 1)cov (ût , ût −1 ) 

2 1− = 2(1 − ρ̂) (4)
(T − 1)var (ût )
with ρ̂ = corr (ût , ût −1 ).

When the DW statistic is equal to 2 means there is no autocorrelations

into residuals, when it is 0 it means there is positive autocorrelation and
when it is 4 there is negative autocorrelation. In this case DW stat is equal
to 2.036330 so we can conclude that we have evidence of no
autocorrelation.

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

M.Sc. Banking and Finance Week 4 6 / 15
 Question 2

A researcher estimates the following model for stock market returns, but
thinks that there may be a problem with it. By calculating the t-ratios,
and considering their significance and by examining the value of R 2 or
otherwise, suggest what the problem might be

ŷt = 0.638 + 0.402 x2t − 0.891 x3t

(0.436) (0.291) (0.763)

R 2 = 0.96, R̄ 2 = 0.89

How might you go about solving the perceived problem?

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

M.Sc. Banking and Finance Week 4 7 / 15
 Solution

The t-ratios for the coefficients in this model are given in the third row
after the standard errors. They are calculated by dividing the individual
coefficients by their standard errors.

ŷt = 0.638 + 0.402 x2t − 0.891 x3t

(0.436) (0.291) (0.763)
1.46 1.38 −1.17

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

M.Sc. Banking and Finance Week 4 8 / 15
 Solution

The problem appears to be that the regression parameters are all

individually insignificant (i.e. not significantly different from zero),
although the value of R2 and its adjusted version are both very high, so
that the regression taken as a whole seems to indicate a good fit.

This looks like a classic example of what we term near multicollinearity.

This is where the individual regressors are very closely related, so that it
becomes difficult to disentangle the effect of each individual variable upon
the dependent variable.

The solution to near multicollinearity that is usually suggested is that since

the problem is really one of insufficient information in the sample to
determine each of the coefficients, then one should go out and get more
data.

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

M.Sc. Banking and Finance Week 4 9 / 15
 Solution

Drop one of the collinear variables - so that the problem disappears.

However, this may be unacceptable to the researcher if there were strong a
priori theoretical reasons for including both variables in the model. Also, if
the removed variable was relevant in the data generating process for y, an
omitted variable bias would result.

Transform the highly correlated variables into a ratio and include only the
ratio and no the individual variables in the regression. Again, this may be
unacceptable if financial theory suggests that changes in the dependent
variable should occur following changes in the individual explanatory
variables, and not a ratio of them.

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

M.Sc. Banking and Finance Week 4 10 / 15
 Question 3

A financial econometrician thinks that the stock market crash of October

1987 fundamentally changed the risk-return relationship given by the
CAPM equation. He decides to test this hypothesis using a Chow test.
The model is estimated using monthly data from January 1981-December
1995, and then two separate regressions are run for the sub-periods
corresponding to data before and after the crash. The model is

rt = α + βRMt + ut ,

so that the excess return on a security at time t is regressed upon the

excess return on a proxy for the market portfolio at time t.

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

M.Sc. Banking and Finance Week 4 11 / 15
 Question 3 (cont’nd)
The results for the three models estimated for shares in British Airways are
as follows

1981M1 - 1995M12:
rˆt = 0.0215 + 1.491RMt , T = 180, RSS = 0.189

1981M1 - 1987M10:
rˆt = 0.0163 + 1.308RMt , T = 82, RSS = 0.079

1987M11 - 1995M12:
rˆt = 0.0360 + 1.613RMt , T = 98, RSS = 0.082

What are the null and the alternative hypothesis that are being tested
here, in terms of α and β?

Discuss parameter stability, and Perform the test. What is your conclusion?
M.Dolfin (King’s Business School King’s College London)
M.Sc. Banking and Finance Week 4 12 / 15
 Solution

Parameter structural stability refers to whether the coefficient estimates

for a regression equation are stable over time.

If the regression is not structurally stable, it implies that the coefficient

estimates would be different for some sub-samples of the data compared
to others.

This is clearly not what we want to find since when we estimate a

regression, we are implicitly assuming that the regression parameters are
constant over the entire sample period under consideration.

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

M.Sc. Banking and Finance Week 4 13 / 15
 Solution

If we define the coefficient estimates for the first and second halves of the
sample as α1 and β 1 , and α2 and β 2 respectively, then the null and
alternative hypotheses are

H0 : α1 = α2 and β 1 = β 2
and

H1 : α1 6= α2 and β 1 6= β 2

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

M.Sc. Banking and Finance Week 4 14 / 15
 Solution

The test statistic is calculated as

RSS − (RSS1 + RSS2 ) (T − 2k )

∗ =
RSS1 + RSS2 k
0.189 − (0.079 + 0.082) 180 − 4
∗ = 15.304
0.079 + 0.082 2

This follows an F distribution with (k; T - 2k) degrees of freedom. F(2,

176) = 3.05 at the 5% level. Clearly we reject the null hypothesis that the
coefficients are equal in the two sub-periods.

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

M.Sc. Banking and Finance Week 4 15 / 15