MODELLING RISK FACTORSDR NGUYEN THI THUY LINH
Agenda1. Measuring returns2. Time aggregation 3. Portfolio aggregation4. Moving average5. Exponentially weighted moving average (EWMA)6. Mapping a portfolio7. Variance Covariance approach8. Principal Component Analysis9. Comparisons of approaches
DR NGUYEN THI THUY LINH2
Measuring returnsThe relative rate of change in the spot price
Alternatively, we could construct the logarithm of the price ratio
This is also
Since ln(1+x) is close to x if x is small, Rt should be closed to rtprovided the return is small
DR NGUYEN THI THUY LINH3
Time aggregation To translate parameters over a given horizon to another horizon.
For example: we have data for daily return which we compute a daily volatility, we want to extend to a monthly volatility Returns can be easily aggregated when we use the log of the price ratio
DR NGUYEN THI THUY LINH4
Time aggregation (cont.)The expected return
The variance
Assuming returns are uncorrelated and have identical distributions across days, we have
More generally, define T as the number of steps. The multiple-period expected return and volatility are
DR NGUYEN THI THUY LINH5
Time aggregation (cont.)Assume now that the distribution is stable under addition (stays the same whether over one period or over multiple periods). The multiple-period VaR is
If returns are not independent, we may be able to characterized longer-term risks:
The variance of two-day returns
is the autocorrelation coefficient or the serial autocorrelation coefficient
DR NGUYEN THI THUY LINH6
Microsoft Example
We have a position worth $10 million in Microsoft shares The volatility of Microsoft is 2% per day (about 32% per year) We use N=10 and X=99
DR. NGUYEN THI THUY LINH7
Microsoft Example continued
The standard deviation of the change in the portfolio in 1 day is $200,000 The standard deviation of the change in 10 days is
200 000 10 456, $632,
DR. NGUYEN THI THUY LINH8
Microsoft Example continued
We assume that the expected change in the value of the portfolio is zero (This is OK for short time periods) We assume that the change in the value of the portfolio is normally distributed Since N(2.33)=0.01, the VaR is
2 33 632 456 473 621. , $1, ,
DR. NGUYEN THI THUY LINH9
Portfolio aggregation
DR NGUYEN THI THUY LINH10
Portfolio
Now consider a portfolio consisting of both Microsoft and AT&T Assume that the returns of AT&T and Microsoft are bivariate normal Suppose that the correlation between the returns is 0.3
DR. NGUYEN THI THUY LINH11
AT&T Example
Consider a position of $5 million in AT&T The daily volatility of AT&T is 1% (approx. 16% per year) The S.D per 10 days is
The VaR is50 000 10 144, $158,
158 114 2 33 405, . $368,
DR. NGUYEN THI THUY LINH12
S.D. of Portfolio
A standard result in statistics states that
In this case X = 200,000 and Y = 50,000 and = 0.3. The standard deviation of the change in the portfolio value in one day is therefore 220,227
YXYXYX 222
DR. NGUYEN THI THUY LINH13
VaR for Portfolio
The 10-day 99% VaR for the portfolio is
The benefits of diversification are(1,473,621+368,405)1,622,657=$219,369
657,622,1$33.210220,227
DR. NGUYEN THI THUY LINH14
Moving average
DR NGUYEN THI THUY LINH15
DR. NGUYEN THI THUY LINH16
Value at Risk (VaR) Models
Risk
Maximum Potential Loss ... 1. ... with a predetermined confidence level2. ... within a given time horizon
VaR = Market Value x Sensitivity x Volatility
Three main approaches:1. Variance-covariance (parametric)2. Historical Simulations3. Monte Carlo Simulations
DR. NGUYEN THI THUY LINH17
10 yrs Treasury BondMarket Value: 10 mlnHolding period: 1 month YTM volatility: 30 b.p. (0,30%)Worst case: 60 b.p.Modified Duration: 6
VaR = 10m x 6 x 0.6% = 360,000
The probability of losing more than 360,000 in
the next month, investing 10 mln in a 10 yrs
Treasury bond, is lower than 2.5%
VaR models: an example
DR. NGUYEN THI THUY LINH18
VaR models: an exampleVaR = 10 mln x 6 x (2*0.3%) = 360,000 Euro
Market Value
(Mark to Market)
A proxy of the sensitivity of the bond price to
changes in its yield to
maturity (for a stock it
would be the beta)
An estimate of the future
variability of interest
rates (for a stock it would
be the volatility of the
equity market)
A scaling factor needed to obtain the
desired confidence level under the
assumption of a normal distribution
of market factors returns
DR. NGUYEN THI THUY LINH19
Estimating Volatility of Market Factors Returns
Historical Volatility
Backward looking
Implied Volatility
Option prices: forward looking
Three main alternative criteria
Garch models (econometric)
Volatility changes over time autoregressive
DR. NGUYEN THI THUY LINH20
Estimating Volatility of Market Factors Returns
1
)(1
2
n
RRn
ti
t
01/10/96 6,74% 01/10/97 6,87%01/11/96 -5,38% 01/11/97 -3,20%01/12/96 6,92% 01/12/97 4,05%01/01/97 0,89% 01/01/98 7,68%01/02/97 14,42% 01/02/98 11,27%01/03/97 -3,76% 01/03/98 4,84%01/04/97 -1,93% 01/04/98 20,14%01/05/97 5,34% 01/05/98 -7,65%01/06/97 -1,47% 01/06/98 1,86%01/07/97 10,66% 01/07/98 1,33%01/08/97 7,76% 01/08/98 3,07%01/09/97 -2,37% 01/09/98 -16,69%
Standard Deviation = 7,77%
Historical Volatility: monthly changes of the Morgan Stanley Italian
equity index (10/96-10/98)
DR. NGUYEN THI THUY LINH21
Estimating Volatility of Market Factors Returns
Most VaR models use historical volatility It is available for every market factor Implied vol. is itself derived from historical Which historical sample? Long (i.e. 1 year) high information content, does not reflect
current market conditions Short (1 month) poor information content Solution: long but more weight to recent data (exponentially
weighted moving average)
DR. NGUYEN THI THUY LINH 22
Example of simple moving averages
17/05/2015
DR. NGUYEN THI THUY LINH 23
Example of simple moving averages
17/05/2015
DR. NGUYEN THI THUY LINH 24
Example of simple moving averages
Figure 3 The Echo Effect Problem
0,0%
0,4%
0,8%
1,2%
1,6%
2,0%
7
/
0
2
/
2
0
0
1
7
/
1
6
/
2
0
0
1
7
/
3
0
/
2
0
0
1
8
/
1
3
/
2
0
0
1
8
/
2
7
/
2
0
0
1
9
/
1
0
/
2
0
0
1
9
/
2
4
/
2
0
0
1
1
0
/
0
8
/
2
0
0
1
1
0
/
2
2
/
2
0
0
1
1
1
/
0
5
/
2
0
0
1
1
1
/
1
9
/
2
0
0
1
1
2
/
0
3
/
2
0
0
1
1
2
/
1
7
/
2
0
0
1
1
2
/
3
1
/
2
0
0
1
-8,0%
-4,0%
0,0%
4,0%
8,0%
12,0%
Daily returns (right hand scale)23-days moving standard deviation (left hand scale)
17/05/2015
Exponentially weighted moving average (EWMA)
DR NGUYEN THI THUY LINH25
DR. NGUYEN THI THUY LINH 26
01 2
23
34
1
1 2 3 1
xt xt xt xtn xt n
n
...
...
0 1
1 11
i xt ii
Estimating Volatility of Market Factors ReturnsExponentially weighted moving average (EWMA)
= return of day t = decay factor (higher , higher persistence, lower decay)
tx
17/05/2015
DR. NGUYEN THI THUY LINH 27
Figure 4 An Example of Volatility Estimation Based Upon an Exponential Moving Average
0,0%
0,4%
0,8%
1,2%
1,6%
2,0%
2,4%
7
/
0
2
/
2
0
0
1
7
/
1
6
/
2
0
0
1
7
/
3
0
/
2
0
0
1
8
/
1
3
/
2
0
0
1
8
/
2
7
/
2
0
0
1
9
/
1
0
/
2
0
0
1
9
/
2
4
/
2
0
0
1
1
0
/
0
8
/
2
0
0
1
1
0
/
2
2
/
2
0
0
1
1
1
/
0
5
/
2
0
0
1
1
1
/
1
9
/
2
0
0
1
1
2
/
0
3
/
2
0
0
1
1
2
/
1
7
/
2
0
0
1
1
2
/
3
1
/
2
0
0
1
-8,0%
-4,0%
0,0%
4,0%
8,0%
12,0%
Daily returns (right hand scale)23-days simple moving standard deviation (left hand scale)23-days exp. weighted moving standard deviation (left hand scale)
17/05/2015
DR. NGUYEN THI THUY LINH 28
Figure 5 An Example of Historical Volatility Estimation Based Upon Different Decay Factors
S&P 500 equally-weighted index daily returnsMoving standard deviations based on different decay factors
0,4%
0,5%
0,6%
0,7%
0,8%
0,9%
1
0
/
0
1
/
2
0
0
4
1
0
/
0
8
/
2
0
0
4
1
0
/
1
5
/
2
0
0
4
1
0
/
2
2
/
2
0
0
4
1
0
/
2
9
/
2
0
0
4
1
1
/
0
5
/
2
0
0
4
1
1
/
1
2
/
2
0
0
4
1
1
/
1
9
/
2
0
0
4
1
1
/
2
6
/
2
0
0
4
1
2
/
0
3
/
2
0
0
4
1
2
/
1
0
/
2
0
0
4
1
2
/
1
7
/
2
0
0
4
1
2
/
2
4
/
2
0
0
4
1
2
/
3
1
/
2
0
0
4
-2,0%
0,0%
2,0%
4,0%
6,0%
8,0%
Daily returns (right hand scale)23-days exp. weighted moving standard deviation (l =0,94)23-days exp. weighted moving standard deviation (l =0,90)23-days exp. weighted moving standard deviation (l =0,99)
17/05/2015
Diversification & correlations
VaR must be estimated for every single position and forthe portfolio as a whole
This requires to aggregate positions together to get arisk measure for the portfolio
This can be done by:mapping each position to its market factors;estimating correlations between market factors returns;measuring portfolio risk through standard portfolio theory.
DR. NGUYEN THI THUY LINH29
DR. NGUYEN THI THUY LINH30
An example
CurrencyPosition (
mln)
Worst case
(1.65*std.dev.)VaR (Euro)
USD -50 0.92% 460.000
Yen 50 1.76% 880.000
Sum of VaRs: 1,340,000
821,74054.0880)460(2880460
2
22
$,$22
$
mmmm
VaRVaRVaRVaRVaR YenYenYenTot
Diversification & correlations
If correl. /$-/Yen = 0.54
DR. NGUYEN THI THUY LINH31
Diversification & correlations
Three main issues1) A 2 positions portfolio VaR may be lower than the more risky position VaR natural hedge2) Correlations tend to shoot up when market shocks/crises occur day-to-day RM is different from stress-testing/crises mgmt3) A relatively simple portfolio has approx.ly 250 market factors large matrices computationally complex an assumption of independence between different types of market factors is often made
222EquityIRFXTot VaRVaRVaRVaR
DR. NGUYEN THI THUY LINH 32
Mapping Estimating VaR requires that each individual position gets associated to its relevant market factors
17/05/2015
DR. NGUYEN THI THUY LINH33
MMapping stock portfolio Equity positions can be mapped to their stock index through their beta coefficient In this case beta represents a sensitivity coefficient to the return of the market index
Individual stock VaR
Portfolio VaR
jiii VMVaR
j
N
iiij VMVaR
1
DR. NGUYEN THI THUY LINH34
MMapping of a stock portfolioExample
817,7326,207,0481
%99,
j
N
iiiP VMVaR
Stock A Stock B Stock C PortfolioMarket Value (EUR m) 10 15 20 45Beta 1,4 1,2 0,8Position in the Market Portfolio (EUR m) 14 18 16Volatility 15% 12% 10%Correlation with A 1 0,5 0,8Correlation with B 0,5 1 0Correlation with C 0,8 0 1
Mapping of equity positions
DR. NGUYEN THI THUY LINH35
MMapping of a stock portfolioExample with individual stocks volatilities and correlations
589,9222 ,,,222
%99, CBCBCACABABACBAP VaRVaRVaRVaRVaRVaRVaRVaRVaRVaR
Stock A Stock B Stock C PortfolioMarket Value (EUR m) 10 15 20 45Beta 1,4 1,2 0,8Position in the Market Portfolio (EUR m) 14 18 16Volatility 15% 12% 10%Correlation with A 1 0,5 0,8Correlation with B 0,5 1 0Correlation with C 0,8 0 1
Mapping of equity positions
Stock A Stock B Stock C MappingVolatilities & Correlations
VaR(99%) 3.490 4.187 4.653 7.817 9.589
VaR of an equity portfolio
DR. NGUYEN THI THUY LINH36
Mapping of a stock portfolio
Mapping to betas: assumption of no specific risk the systematic risk is adequately captured by a CAPM type model it only works for well diversified portfolios
Stock A Stock B Stock C MappingVolatilities & Correlations
VaR(99%) 3.490 4.187 4.653 7.817 9.589
VaR of an equity portfolio
DR. NGUYEN THI THUY LINH37
Figure 6 Main Characteristics of the Parametric Approach
2. Portfolio: 3. Risk measures:
stocks
rates
commodities
fx
1. Risk factors:Are defined either as price changes (assetnormal) or as changesin market variables(delta normal) theirdistribution is thensupposed to benormal.
ConfidentialReportfor theCompanysC.E.O.
ConfidentialReportfor theCompanysC.E.O.
ConfidentialReportfor theCompanysC.E.O.
Risk factors are mapped to individualpositions based on virtual componentsand linear coefficients(deltas). Portfolio riskis estimated based on the correlation matrix
VaR is quicklygenerated as a multiple () of the standard deviation.
0%
2%
4%
6%
8%
10%
12%
14%
16%
-
6
0
7
-
5
4
3
-
4
7
9
-
4
1
5
-
3
5
1
-
2
8
8
-
2
2
4
-
1
6
0
-
9
6
-
3
2
3
2
9
6
1
6
0
2
2
4
2
8
8
3
5
1
4
1
5
4
7
9
5
4
3
6
0
7
Variazioni di valore del portafoglio (euro, valore centrale)
%
d
i
c
a
s
i
DR. NGUYEN THI THUY LINH 38
Mapping FX forwardA long position in a USD forward 6 month contract is equivalent to: A long position in USD spot A short deposit (liability) in EUR with maturity 6 m A long deposit (asset) in USD with maturity 6 m
titiSF
f
dt
1
1
17/05/2015
DR. NGUYEN THI THUY LINH39
MMapping FX forwardExample: Buy USD 1m 6 months forward
FX and interest rates
099.990$5,002,01
000.000.1
USDI
119.118.12,1099.990 EURD1. Debt in EUR
2. Buy USD spot
3. USD investment
2,1099.990 spotUSD
EUR/USD Spot 1,206 m EUR interest rate 3,50%6 m USD interest rate 2,00%EUR/USD 6 m forward 1,209
DR. NGUYEN THI THUY LINH40
MMapping FX forward
849.18483,0326,2%5,1119.118.16 EURVaR miEUR
259.16549.13490,0326,2%2,1099.9906 EURVaR miUSD
919.82099.69326,2%3099.990 EURVaRUSDspot
Market factor Volatility EUR/USD EUR 6 m IR USD 6 m IREUR/USD Spot 3% 1 -0,2 0,4EUR 6 m IR 1,50% -0,2 1 0,6USD 6m IR 1,20% 0,4 0,6 1
Correlation withVolatilities and correlations - Forward position market factors
DR. NGUYEN THI THUY LINH41
MMapping FX forward
USDspotmiUSDUSDspotmiUSDUSDspotmiEURUSDspotmiEUR
iUSDiEURmiUSDmiEURUSDspotmiUSDmiEURmUSD VaRVaRVaRVaR
VaRVaRVaRVaRVaRVaR
,66,66
,6622
62
6
622
2
646.834,0919.82)259.16(2)2,0(919.82849.182
6,0)259.16(849.182919.82259.16849.18 222
Total VaR of the USD 6 m forward position
DR. NGUYEN THI THUY LINH 42
Mapping of a FRA An FRA is an agreement locking in the interest rate on an investment (or on a debt) running for a pre-determined A FRA is a notional contract no exchange of principal at the expiry date; the value of the contract (based on the difference between the pre-determined rate and the current spot rates) is settled in cash at the start of the FRA period. A FRA can be seen as an investment/debt taking place in the future: e.g. a 3m 1 m Euro FRA effective in 3 months time can be seen as an agreement binding a party to pay in three months time a sum of 1 million Euros to the other party, which undertakes to return it, three months later, increased by interest at the forward rate agreed upon
17/05/2015
DR. NGUYEN THI THUY LINH43
MMapping of a FRA
1-11-2000 1-2-20011-8-2000
investment
1m
1,013m1m
mf
1.013m
Example: 1st August 2000, FRA rate 5.136% Investment from 1st November to 1st February 2001 with
delivery: 1,000,000 *(1 + 0.05136 * 92/360) = 1,013,125 Euros. Equivalent to:
a three-month debt with final principal and interest of one million Euros;
A six-month investment of the principal obtained from the transaction as per 1.
DR. NGUYEN THI THUY LINH44
Variance-covariance approach
Assumptions and limits of the variance-covariance approach Normal distribution assumption of market factor
returns Stability of variance-covariance approach Assumption of serial indepence of market factor returns linear sensitivity of positions (linear payoff)
Markowitz Result for Variance oof Return on Portfolio
sinstrument th and th of returns between ncorrelatio is
portfolio in instrument th on return of variance is
portfolio in instrument th of weightis
Return Portfolio of Variance
2
1 1
ji
iiw
ww
ij
i
i
n
i
n
jjijiij
DR. NGUYEN THI THUY LINH45
VaR Result for Variance of PPortfolio Value (i = wiP)
day per value portfolio the in change the of SD the is return)daily of SD (i.e., instrument th of volatilitydaily the is
P
i
n
ijiji
jiijiiP
n
i
n
jjijiijP
n
iii
i
xP
1
222
1 1
2
1
2
DR. NGUYEN THI THUY LINH46
Covariance Matrix (vari = covii)
nnnn
n
n
n
C
varcovcovcov
covvarcovcov
covcovvarcov
covcovcovvar
321
333231
223221
113121
covij = ij i j where i and j are the SDs of the daily returns of variables i and j, and ij is the correlation between them
DR. NGUYEN THI THUY LINH47
Alternative Expressions for P2
transpose its is and is element th whosevector column the is where
T
T
i
P
j
n
i
n
jiijP
i
C
2
1 1
2 cov
DR. NGUYEN THI THUY LINH48
Alternatives for Handling IInterest Rates Duration approach: Linear relation between P and y but assumes parallel shifts) Cash flow mapping: Cash flows are mapped to standard maturities and variables are zero-coupon bond prices with the standard maturities Principal components analysis: 2 or 3 independent shifts with their own volatilities
DR. NGUYEN THI THUY LINH49
Handling Interest Rates: CCash Flow Mapping We choose as market variables bond prices with standard maturities (1mth, 3mth, 6mth, 1yr, 2yr, 5yr, 7yr, 10yr, 30yr) Suppose that the 5yr rate is 6% and the 7yr rate is 7% and we will receive a cash flow of $10,000 in 6.5 years. The volatilities per day of the 5yr and 7yr bonds are 0.50% and 0.58% respectively
DR. NGUYEN THI THUY LINH50
Example continued
We interpolate between the 5yr rate of 6% and the 7yr rate of 7% to get a 6.5yr rate of 6.75% The PV of the $10,000 cash flow is
540,60675.1
000,105.6
DR. NGUYEN THI THUY LINH51
Example continued
We interpolate between the 0.5% volatility for the 5yr bond price and the 0.58% volatility for the 7yr bond price to get 0.56% as the volatility for the 6.5yr bond We allocate of the PV to the 5yr bond and (1- ) of the PV to the 7yr bond
DR. NGUYEN THI THUY LINH52
Example continued
Suppose that the correlation between movement in the 5yr and 7yr bond prices is 0.6 To match variances
This gives =0.074)1(58.05.06.02)1(58.05.056.0 22222
DR. NGUYEN THI THUY LINH53
Example continuedThe value of 6,540 received in 6.5 years
in 5 years and by
in 7 years.This cash flow mapping preserves value and variance
484$074.0540,6
056,6$926.0540,6
DR. NGUYEN THI THUY LINH54
Principal Components AAnalysis for Interest Rates
DR. NGUYEN THI THUY LINH55
Principal Components AAnalysis for Interest Rates
The first factor is a roughly parallel shift (83.1% of variation explained) The second factor is a twist (10% of variation explained) The third factor is a bowing (2.8% of variation explained)
DR. NGUYEN THI THUY LINH56
Using PCA to calculate VaR Example: Sensitivity of portfolio to rates ($m)
Sensitivity to first factor is from Table 18.3:10h0.32 + 4h0.35 8h0.36 7 h0.36 +2 h0.36 = 0.08Similarly sensitivity to second factor = 4.40
1 yr 2 yr 3 yr 4 yr 5 yr+10 +4 -8 -7 +2
DR. NGUYEN THI THUY LINH57
Using PCA to calculate VaRcontinued
As an approximation
The f1 and f2 are independent The standard deviation of P (from Table 20.4) is
The 1 day 99% VaR is 26.66 h 2.33 = 62.12
21 40.408.0 ffP
66.2605.640.449.1708.0 2222
DR. NGUYEN THI THUY LINH58
When Linear Model Can be UUsed Portfolio of stocks Portfolio of bonds Forward contract on foreign currency Interest-rate swap
DR. NGUYEN THI THUY LINH59
The Linear Model and Options
Consider a portfolio of options dependent on a single stock price, S. Define
andSP
SSx
DR. NGUYEN THI THUY LINH60
Linear Model and Options ccontinued
As an approximation
Similarly when there are many underlying market variables
where i is the delta of the portfolio with respect to the ith asset
xSSP
iiii xSP
DR. NGUYEN THI THUY LINH61
Example
Consider an investment in options on Microsoft and AT&T. Suppose the stock prices are 120 and 30 respectively and the deltas of the portfolio with respect to the two stock prices are 1,000 and 20,000 respectively As an approximation
where x1 and x2 are the percentage changes in the two stock prices
21 000,2030000,1120 xxP
DR. NGUYEN THI THUY LINH62
But the distribution of the daily rreturn on an option is not normal
The linear model fails to capture skewness in the probability distribution of the portfolio value.
DR. NGUYEN THI THUY LINH63
Impact of gamma
Positive Gamma Negative Gamma
DR. NGUYEN THI THUY LINH64
Translation of Asset Price Change tto Price Change for Long Call
Long Call
Asset Price
DR. NGUYEN THI THUY LINH65
Translation of Asset Price Change tto Price Change for Short Call
Short Call
Asset Price
DR. NGUYEN THI THUY LINH66
Quadratic Model
For a portfolio dependent on a single stock price it is approximately true that
this becomes
2)(2
1 SSP
22 )(2
1 xSxSP
DR. NGUYEN THI THUY LINH67
Quadratic Model continued
With many market variables we get an expression of the form
where
This is not as easy to work with as the linear model
n
i
n
ijiijjiiii xxSSxSP
1 1 2
1
jiij
ii SS
PSP
2
DR. NGUYEN THI THUY LINH68
Monte Carlo Simulation
To calculate VaR using M.C. simulation we Value portfolio today Sample once from the multivariate distributions of the xi Use the xi to determine market variables at end of one day Revalue the portfolio at the end of day
DR. NGUYEN THI THUY LINH69
Monte Carlo Simulation
Calculate P Repeat many times to build up a probability distribution for P VaR is the appropriate fractile of the distribution times square root of N For example, with 1,000 trial the 1 percentile is the 10th worst case.
DR. NGUYEN THI THUY LINH70
Speeding Up Monte Carlo
Use the quadratic approximation to calculate P
DR. NGUYEN THI THUY LINH71
Comparison of Approaches
Model building approach assumes normal distributions for market variables. It tends to give poor results for low delta portfolios Historical simulation lets historical data determine distributions, but is computationally slower
DR. NGUYEN THI THUY LINH72
Stress Testing
This involves testing how well a portfolio performs under extreme but plausible market moves Scenarios can be generated using
Historical data Analyses carried out by economics group Senior management
DR. NGUYEN THI THUY LINH73
Back-Testing
Tests how well VaR estimates would have performed in the past We could ask the question: How often was the actual 10-day loss greater than the 99%/10 day VaR?
DR. NGUYEN THI THUY LINH74