Académique Documents
Professionnel Documents
Culture Documents
Derivatives
Lecture 04:
Value-at-Risk 2
Unit Content
Week 1: Introduction to Risk, Risk Management and Derivatives
Week 2: Financial Statistics
Week 3: Value-at-Risk 1
Week 4: No Classes Ekka Public Holiday
Week 5: Value-at-Risk 2
Week 6: Forwards and Futures 1
Week 7: Forwards and Futures 2
Week 8: Mid-Semester Exam and Reflective Practices
Week 9: Swaps: interest rate
Week 10: Options: introduction and pricing with the binomial
model
Week 11: Options: pricing with the Black-Scholes model
Week 12: Options: trading strategies and risk management
Week 13: Derivative Disasters
2
Week 14: Revision
Lecture Outline
Review of VaR
Portfolio VaR
Individual Stock Return Statistics and VaR
Portfolio Statistics and VaR
Time-varying Correlations
VaR Horizons
Return and Standard Deviation Adjustments
Readings:
RiskMetrics (1996) Technical Document, pp. 5-9, 45-56, 64-72,
77-88 and 93-101. Skim pp.21-30. Note: use the lecture notes
as a guide to what to focus on.
Hull et al. (2013) Fundamentals of Futures and Options, Ch. 20,
pp. 419-431 and 436-442. Mathematical Appendix, pp. 531-538.
3
Value-at-Risk
Defined
0.0151
Value-at-Risk
Normal VaR with and without time varying volatility
Value-at-Risk
Better
Portfolio VaR
Portfolio VaR
Portfolio Theory
Calculating Portfolio Returns
where:
is the proportion of wealth invested in
asset i
because of the budget constraint
N = 2:
N = 3:
e that this is the portfolio log return approximation. RiskMetrics (1996, pp.48-49
justifies the use of this approximation.
8
Portfolio Theory
Calculating Portfolio Expected Returns
where:
is the proportion of wealth invested in
asset i
because of the budget constraint
N = 2:
N = 3:
Portfolio Theory
Calculating Portfolio Return Variance
N = 2:
N=3
10
Portfolio VaR
Portfolio example
N = 3:
CBA, TLS and BHP
Download:yahoofinance.com
Merge Data: Date and Adjusted Close Price.
Clean Data: Remove any day where 1 or more stocks do
not trade.
For example, we will remove a day when
CBA and TLS
trade but BHP does not.
Returns:
Daily Log Returns
Statistics:
Unconditional Mean, Standard Deviations
and
Correlations
VaR:
Individual stock and portfolio (weights detailed
later)
11
Portfolio VaR
Plot of BHP closing price and associated log
Returns.
Whats happened in the middle of 2001?
On 29 June 2001 BHP merged with Billiton:
Read this announcement
This artefact must be removed from the
data as the change in price does not reflect
a change in value as shareholders were
entitled to 1.06 bonus shares (refer article).
What to do?
Portfolio VaR
13
Portfolio VaR
orrect: Create a new date row that starts at the first date and then includes all d
uding weekends trust me) until the last observation. Then use the excel functio
ookup(lookup values, data, column, FALSE)
14
Portfolio VaR
peat for CBA and TLS. Then remove any date where there is a #N/A. This can be
multiplying the 3 prices on a given day together as #N/A is returned if a #N/A
15 will f
resent on that day. TIP: copy, paste special dates as values otherwise filter
Portfolio VaR
Portfolio VaR
17
Portfolio VaR
Exceedances for all three stocks follow results from last weeks lecture.
That is, too few exceedances at the 95% level and too many at the 99% level.
18
Portfolio VaR
Portfolio Statistics
he portfolio will consist of 30% BHP, 60% CBA and the remainder in TLS (thats
1
19
Portfolio VaR
20
Portfolio VaR
n at only
21
Portfolio VaR
portfolio VaR is less than the weighted average of the individual VaRs because
he diversification benefit that arises from investing across stocks that are not pe
itively correlated. Whats the benefit of the reduced VaR?
22
Portfolio VaR
der that has a daily limit of VaR$ = 20,000 can increase the exposure to the port
e the amount of the individual assets. As a result, the annual dollar expected ret
500 higher than any other asset. This includes the asset with the highest expecte
n. Given the weights of the portfolio and the portfolio size, the individual VaRs fo
portfolio are:
$25,503.34
23
Portfolio VaR
nce again, it is evident that volatility is not constant. The VaR could be updated t
eflect the changes in volatility.
24
Portfolio VaR
25
Portfolio VaR
26
Portfolio VaR
27
Time-Varying Correlations
Given that variance of asset returns are time varying,
it is likely that the correlation of asset returns are also
time varying.
Capturing the time-varying nature of correlation is
important because it is correlation that drives
diversification benefit. This benefit directly flows
through to VaR by reducing the portfolios return
standard deviation.
where
Time-Varying Correlations
To measure time-varying correlations, we simply
Time-Varying Correlations
Some detail
30
Time-Varying Correlations
Some detail
31
Time-Varying Correlations
VaR Horizons
All of the VaRs considered thus far have had a horizon of
1-day. But theres no reasons, and in fact there are some
very good reasons, why VaR cannot be calculated over
longer horizons. For example, the Market Risk Charge
applied to a banks trading book uses a 10-day VaR.
How do we adjust for longer horizons?
1.
Use returns over the longer horizons. This either requires more
data or creates issues with overlapping returns.
2.
VaR Horizons
Returns over longer horizon:
e: log return are additive but simple (relative) returns are not. Refer RiskMetrics
34
VaR Horizons
Variance Adjustment:
1-day Variance:
2-day Variance:
k-day Variance:
ote: we are assuming that the returns (not the squared returns) are uncorrelated
rough time. This is evident by the lack of any covariance terms above.
35
VaR Horizons
Portfolio Normal VaR at different horizons
1-day N VaR:
2-day N VaR:
k-day N VaR:
Note: this also applies to Normal VaR for individual assets.
36