Hugues Pirotte (v2 - June 2004)

RM - Exercise 1

VaR – Implementing the variance-covariance approach (with Excel)

You have the following portfolio of assets (we won’t explain here how you did get there):
Asset Description Market unit Currency
price
1 7 shares of a fund indexed on the 1500.00 EUR
EuroStoxx50
2 2 shares of a fund indexed on the 10000.00 EUR
Dow Jones
3 10 10-years US Treasury 0-coupon bonds 650.00 USD
(face value: 1000.00 USD)

The exchange rate USD/EUR (dollars per euro) is currently trading at 1.25. Your
reference currency is the Euro, i.e. your objective is to consume Euros in fine.
a) Identify the risk sources of this portfolio.
b) Compute exposures of each position to the risk sources mentioned under a).
c) Standard deviations of annual returns and cross-correlations are proposed here
below. Compute the variance-covariance matrix.

Standard Correlations
deviations
EuroStoxx 50 DJ USD US 10
years
EuroStoxx 50 30.00% 1.00 0.49 0.64 -0.28
DJ 20.00% 0.49 1.00 0.80 -0.37
USD 10.00% 0.64 0.80 1.00 -0.43
US 10 years 9.00% -0.28 -0.37 -0.43 1.00

d) Compute the weekly VaR of this portfolio following the var-covar methodology.
e) What is the contribution of every risk source to the overall VaR?

SOLUTION
a) Risk sources are :
o an exposure to the risk (market, stock) of the evolution of the EuroStoxx50.
o an exposure to the risk (market, stock) of the evolution of the Dow Jones.
o a currency risk exposure to the USD/EUR.
o a risk exposure to interest-rate fluctuations, which will induce a fluctuation in
bond prices.
b) The act of splitting exposures of every position and allocating them to the four
risk sources mentioned under a) is called « mapping ».
If we multiply shares in every position to their unit market values and express
them in Euros, we will obtain the following amounts:
Position Description Computation Value in Euros
1 Shares of EuroStoxx 50 : 7 × 1500 = 10500 (33.12%)
2 Shares of Dow Jones : (2 × 10000) / 1.25 = 16000 (50.47%)
3 US Bonds: (10 × 650) / 1.25 = 5200 (16.40%)

Now, we should allocate these values to the four risk exposures. Position 1 is
exposed to « EuroStoxx 50 » risk, while position 2 shows an exposure to the
« Dow Jones » but also to « USD currency ». For the third position, it is a « US
10y bonds » risk and also a « USD currency » risk. The amount of the exposures
is simply based on the value of the position in the reference currency, i.e. Euro.
The following table or matrix provides results by position and by risk source
(« mapping ») while calculating (for each column) the total exposure to each
risk source :

EuroStoxx 50 Dow Jones $/€ US 10y bonds

Shares EuroStoxx 50 : 10500
Shares Dow Jones : 16000 16000
US bonds: 5200 5200
Total 10500 16000 21200 5200

We end up with total exposures for the four exposures of the 3-positions
portfolio.

c) A variance or covariance is obtained by the multiplication of standard deviations

of all possible pairs of risk sources, times their correlation. Thus, any variance
(or covariance) is obtained from the following multiplication :
σ ij = σ i × σ j × ρ i , j
Applying this to all possible pairs, we end up with the following « variance-
covariance » matrix :

Variances-covariances
EuroStoxx 50 DJ USD US 10y
EuroStoxx 50 0.09000 0.02940 0.01920 -0.00756
DJ 0.02940 0.04000 0.01600 -0.00666
USD 0.01920 0.01600 0.01000 -0.00387
US 10y -0.00756 -0.00666 -0.00387 0.00810

d) Based on the relative VaR formula, we should now multiply the total exposures
once by each column of variance-covariances to obtain a net sum of these
multiplications per column.
In summary, the vector of total exposures (4 values) should be first multiplied
by the variance-covariance matrix, which means to sum all line-by-line
multiplications. Then, we repeat this column by column. Thus, for the first risk
source, we have that :
 EuroStoxx 50
10500 × 0.09000 = 945.00
16000 × 0.02940 = 470.40
21200 × 0.01920 = 407.04
5200 × -0.00756 = -39.312
Total 1783.13

Once done for all columns, we obtain the following results :

EuroStoxx 50 DJ USD US 10y

1783.13 1253.27 649.8 -225.86

Then we re-multiply again this vector by the vector of total exposures (the
result is that we have computed all combinations twice):

EuroStoxx 50
1783.13 × 10500 18722844.0
1253.27 × 16000 20052288.0
649.48 × 21200 13768891.2
-225.86 × 5200 -1174493.8
Total 51369530.4

This total variance in Euros is annual.

To convert it to a weekly one, we need to divide it by 52 (weeks per year). To
end up with a standard deviation, we still need to compute the square root of
this result. Overall, we have :
51369530.4
= 993.92
52
Knowing that the critical value (zc ) for a 95% confidence degree is 1.644853,
then we have still to multiply 993.92 by that number to find out the desired
weekly VaR for a 95% confidence degree, i.e. :
993.92 × 1.644853 = 1634.85
For any time-horizon change, it suffices to multiply this number by the square
root of the desired order of magnitude ( × 52 to obtain back the annual VaR, i.e.
11789.07). And, for any change in the confidence degree, we just need to divide
it by 1.644853 and re-multiply it by the respective value of zc.

e) A priori, using the first vector of results would be a good way to compute the
contribution of every source of risk to the overall VaR. If we take the total value
of that vector and we estimate the weight of each sum components on the total,
then we will have the required result :

EuroStoxx 50 DJ USD US 10y Total

1783.13 1253.27 649.8 -225.86 3460.01
51.54% 36.22% 18.77% -6.,53% 100.00%

We would therefore see that the bonds position brings in an important

diversification within the portfolio reducing the total VaR (negative value). But,
this computation is not correct. Looking at
 δVaR ΣW ΣW
= zc T = z 2c T
δW '
W ΣW
' VaR
we obtain that
 Σ W  m  ΣW  m
VaR = W '  zc T  = VaR
 ∑ ∑
= w j zc T
 W '
Σ W  W ' Σ W  j j =1
j
  j =1 
In the present case, replacing values, the Delta Var (see first equation here
above) gives:
Delta VaR
EuroStoxx 50 0.4092
DJ 0.2876
USD 0.1491
US 10 an -0.0518

Applying this to each “weight mapping” by risk source and by position, we do

obtain:
EuroStoxx 50 DJ USD US 10 an Total
1 4296.81 0.00 0.00 0.00 4297.81
2 0.00 4601.91 2384.83 0.00 6988.74
3 0.00 0.00 775.07 -269.54 508.53
4296.81 4601.91 3159.90 -269.54 11789.08