Practice exam

Exercise I. (Total mark: 30 points + 10 Bonus points).

The worksheet DATAEXI contains data on some constituents of the S&P 500 index. For each
company, we have the following information:
- DY: the Dividend Yield based on the last dividend payout.
- SECTOR: the sector in which the company operates.
- BETA: the (CAPM) beta of the company.
- EPS: the earnings per share for the company based on the last report.
- EM: the company’s Enterprise Multiple (value to EBITDA ratio).
- G: last quarter’s (Year on Year) growth.

Given these data, address the following questions in the worksheet SOLUTIONEXI:

a) Run the following regression: 𝑫𝒀𝒊 = 𝜶 + 𝜷𝟏 𝑩𝑬𝑻𝑨𝒊 + 𝜷𝟐 𝑬𝑷𝑺𝒊 + 𝜷𝟑 𝑬𝑴𝒊 + 𝜷𝟒 𝑮𝒊 + 𝜺𝒊 . And

address the following questions:
• run the regression through the data analysis tool

a.1) (4 points) Construct 95% confidence intervals for each of the parameters involved and
test the null hypothesis that each of them is individually equal to zero.
Note: Confidence intervals generated automatically by the Data Analysis add-in will not be
considered as valid.

•
• Take for every parameter the coeff and se from the regression summary and fill it in the
formula for the confidence interval.

•
 • Look at the p-value of every parameter to conclude if this parameter is significantly

different from zero. they are all smaller than

the p-value of 0.05.

a.2) (2 points) What percentage of the variation in dividend yield is captured by this model?

• Take the R square from the summary output from the regression.

a.3) (6 points) Intuitively, we would expect stocks that have experienced larger growth over
the last quarter to payout larger dividends. Test H0: β4 =0 against H1: β4 >0 at the 1%
significance level. Holding all else constant, do stocks with higher growth tend to more pay
more dividends?

• T-statistic calculated through so from the summary output of the

regression, divide the estimated coefficient by the standard error of the parameter.
• the critical value is just calculated through =T.INV(0.99, total number of df from
regression summary-2)
b) It is rumored that stocks in the ENERGY sector typically payout larger dividends, and those
in the LIFESCIENCES typically payout smaller dividends. To formally assess differences in
dividend payout across sectors, generate two dummy variables, named EN and LS, taking on
value equal to one if the underlying observation corresponds to a company in the ENERGY
sector and a company in the LIFESCIENCES sector, respectively, and zero otherwise, and run
the following regression:
DYi =α+β1BETAi +β2EPSi +β3EMi +β4Gi +δ1ENi +δ2LSi +εi

• first add the two columns for the dummy’s

• then, run the new regression including these dummy’s

b.1) (10 points) Test the joint hypothesis H0: δ1 = δ2 = 0 using a χ-test and the 10% significance
level, estimating the restricted and unrestricted models. Can we reject the null hypothesis
that the average dividend yield is the same for all sectors?

• calculate the Chi-square statistic through the R square’s of the summary output’s from
the regressions.

• the chi statistic is way bigger than the critical value ➔ reject H0
b.2) (8 points) What is the expected dividend yield for a company that has BETA = 1, EPS = 5,
EM = 10 and G = 3 if this company is in the ENERGY sector? What would be the expected
dividend yield if it was in LIFE SCIENCES? What if it was a company in any of the remaining
sectors?

• Fill in the values for the regression model, and just play around with the values for the
dummy’s in the model when needing to calculate for every sector.

b.3) (Bonus: 10 points) Given the general characterization of multiple linear hypothesis H 0: Rβ
= c for joint testing, determine the matrix R, the vector c and the number of restrictions q
associated with the hypothesis H0: α = 0; β1 = 1; δ1 + δ2 = 1. If willing to test this combined
hypothesis against a two-sided alternative with a 1% significance level using a Wald test,
which critical value would you consider?
Exercise II. (Total mark: 20 points).

The worksheet DATAEXII-III contains daily simple returns of six funds (exchange traded
funds, ETF’s) over the period 05/01/2015-26/11/2021. Out of these six funds, you are
requested to select a specific subset and perform all operations on this selection. Your
analysis must be based on the returns of the assets ETF1, ETF2, ETF3, ETF4 and ETF5.
Please pay attention to carry out this selection as specifically indicated because otherwise
Exercises II and III, which build on these data, will be consider invalid. Select the required
data, copy-paste them into the worksheet MYDATA and delete the unnecessary columns in
the latter. Then address the following questions in the worksheet SOLUTIONEXII. Note:
Procedures involving macros (such as Kstat or Data Analysis) are not allowed in the resolution
of this exercise.

a) (2 points) Generate the daily simple returns of an equally weighted portfolio on the
selected ETFs (you can copy the selected data in MYDATA in this worksheet if this is more
convenient to you). Hint. Recall that the simple return of a portfolio on k assets is given by
∑𝒌𝒊=𝟏 𝒘𝒊 𝒓𝒊𝒕 , with 𝒓𝒊𝒕 and 𝒘𝒊 denoting the individual simple return and the weight of the i-th
asset.

• Make a new table with the weights for the portfolio, which are all 0.2 as they need to be
equally weighted.

• Then use =SUMPRODUCT(B2:F2,$J$2:$N$2) in which the first values are the returns of
the etfs and the fixed row is the weights.

b) (4 points) Using the corresponding Excel functions, report the annualized mean return
(×250), annualized volatility (× √𝟐𝟓𝟎), min, max, number of observations, skewness, excess
kurtosis, median, and quartiles.

• First, make a table for the entire data with the averages. This will consist of expected
return daily variance and SD per ETF. The formulas are =average, =var.s, =stdev().
•
• Then we will annualize this table. By following the instructions. We want the following
table after:

• You know all the formula’s but note e(R) annual is e(R)* 250 and same for sd with the
square, said in the question.
• Note use =QUARTILE.INC(B2:B1734,1) not EXC
c) (4 points) Given the values in b), compute the JB test statistic and test normality of the
portfolio returns at the 5% significance level. Comment on results.

• Then we perform the Jarque bera test: =count*((skewness^2/6)+(excess kurtosis^2/24))

• Critical value: =CHISQ.INV(0.95,2) this is with 95% with two degrees of freedom.
• Decision: =IF(JB > CV, "reject", "accept")
d) (10 points) Make a histogram (30 bins) of the return of the equally weighted portfolio.
Note: The histogram must be constructed as explained in class. Native graphical options of
Excel to plot directly the histogram are not allowed.

• Then just select the two columns with percentages and number of observations per bin,
go to insert and insert a 2D column. Adjust the histogram after.
Exercise III. (Total mark: 44 points).

Copy the data of MYDATA in the worksheet SOLUTIONEXIII and address the following
questions.

a (8 points) For each of the assets in MYDATA, compute the individual daily mean in a column
vector, say 𝝁𝒅. Multiply the entries of this vector by 250 to obtain the annualized expected
return vector, 𝝁.
Then, use the macro-Covariance in the Data Analysis add-in to compute the (k × k) daily
covariance matrix of returns, say 𝛀𝒅 . Since the covariance matrix is symmetric (i, e., the (i,j)-
th element of 𝛀𝒅 is the same as the (j,i)-th element), Excel only reports the lower part. Fill up
missing values manually, copy-pasting the corresponding elements.
Finally, determine the annualized variance-covariance matrix, say 𝛀, multiplying each
element of 𝛀𝒅 by 250, i.e., Ω = 250×𝛀𝒅 . Address all questions below using µ, Ω and Solver,
assuming that short selling is allowed and that there is no risk-free asset unless otherwise
indicated.

• the first part, we already did in the previous exercise.

• or calculate manually like on the left or do it through the add in of data analysis. With
the add-in, you must copy and paste and transpose etc. to make it a full matrix instead
of half.

• make a new matrix for the covariance, by multiplying it with 250.

b) (8 points) Determine the optimal composition of two portfolios: b.1) one with 15% annual
expected return (i.e. target=0.15), and b.2) one with 7% annual volatility (i.e., target=0.07).

• first, we transpose the annualized returns calculated in the previous exercise. Then, we
set the weights just equal to the equally weighted portfolio (0.20 weight for each ETF).
Then link the E(R) and st.dev of the portfolios to the weights and returns of the ETFs, to
be able to use the solver.

• the red matrix is the covariance matrix you calculated previously (the annualized one,
that you multiplied with 250)

• then calculate the sharpe ratio.

 • then perform the solver.
o As objective you set the Sharpe ratio to maximize.
o By changing: the weights of the portfolio
o Constraints:
▪ Sum of the weights = 1
▪ Depending on which portfolio, you set return = 0.15 or st.dev = 0.07
o Don’t tick the box of the negative values because we allow for short selling.

c) (4 points) Determine the composition of the global minimum

variance portfolio.
• First set the weights to equally weighted portfolio and the
links with the formulas like in the previous exercise.
• Then for the solver:
o Objective is the st.dev and min
o By changing the weights
o Only constraint should be that the weights = 1
d) (10 points) Generate and plot the efficient frontier. To this end, consider the efficient
portfolios in b.1) and c) above and make linear combinations using a weighting factor λ = 0,
0.01, 0.02, ..., 1.

• Do this for lambda’s up until 1, with the same matrices.

• The E(R) is then those weights multiplied with the annualized returns. And the st.dev
anslo same formula as before, just with different weights.

• To construct the EF, use the expected return and st.dev over all lambda’s.
o Series 1 ➔ x values = st.dev row, y values = e(r) row
e) (10 points) Assume that short selling is allowed and a risk-free asset with annual returns of
0.05% (0.0005). Determine the composition of the tangency portfolio (TP). Represent the new
efficient frontier making linear combinations between the risk-free asset and the TP using a
weighting factor λ = 0, 0.01, 0.02, ..., 1.70. Plot the resultant line on the same graph generated
in d) above.

• To get the TP, do the same thing with the solver, just now max the sharpe ratio and no
target constraint, just the weight equal to one.
• For the EF:
o First put up a table as before, with the lambda’s going from 0-1.7 etc.
o The first column, where lambda = 0 ➔ put all the weights of the ETF = 0 and of
the RF asset to 100% (this is the portfolio of the RF asset)

o then calculate the E(R) in the same way as before:

o then the st.dev for the first column just set to 0%
o then from the second column onwards:
▪ calculate the weights for the portfolio in the same way as before:
▪ calculating the E(R) for every portfolio as before:
▪ for the st.dev we only take into account the risky portfolio, as the risk free
asset does not have risk / volatility:

▪ then for the graph, exact same thing as before:

f) (4 points) Determine the composition of the TP in e) above assuming that short selling is
not allowed.
• Is the same way with the solver etc. as in previous exercise, just tick the box where the
values cannot be negative, to assume there is no short selling.