Hull & White - Wrong Way Risk
Hull & White - Wrong Way Risk
Hull & White - Wrong Way Risk
CVA AND WRONG WAY RISK
1. Introduction
It has for some time been standard practice for derivatives dealers to adjust the reported value of
their derivatives transactions with a counterparty to reflect the possibility of losses being
incurred because of a default by the counterparty. The adjustment is known as the credit value
adjustment or CVA.
A derivatives dealer has one CVA for each counterparty. These CVAs are themselves derivatives
and must be managed similarly to other derivatives. CVAs are particularly complex derivatives.
In fact, the CVA for a counterparty is more complex, and more difficult to value, than any of the
transactions between the dealer and the counterparty. This is because the CVA for a counterparty
is, as we will see, contingent on the net value of the portfolio of derivatives outstanding with that
counterparty.
Calculating CVAs is a very computationally intensive exercise for a derivatives dealer. Gregory
(2009) provides an excellent discussion of the issues. Statistics published about Lehman Brothers
give a sense of the scale of the problem. Reuters (2008) reported that, at the time of its failure
Lehman, had about 1.5 million derivatives transactions outstanding with 8,000 different
counterparties. This means that it had to calculate 8,000 different CVAs and the number of
derivative transactions on which each CVA was dependent averaged about 200.
Another measure, more controversial than CVA, is debit value adjustment or DVA.
1
The DVA
calculated by the dealer for a counterparty is an estimate of the costs to the counterparty of the
possibility that the dealer might default. The possibility that it might default is in theory a benefit
to the dealer and accounting standards require the book value of the derivatives outstanding with
a counterparty to be calculated as their no-default value minus the CVA plus the DVA.
2
The
1
DVA is sometimes also called debt value adjustment.
2
Thereisacomplicationhereinthatifthedealerdefaultssubsequentdefaultsbythecounterpartyareirrelevant
andviceversa.SeeHullandWhite(2001)foradiscussionofthispointinthecontextofcreditdefaultswaps.
3
reason why DVA is controversial is that there is no way a dealer can monetize DVA without
actually defaulting.
It is interesting to note that when the credit spread of a derivatives dealer increases, DVA
increases. If the dealer uses DVA accounting, this in turn leads to an increase in the reported
value of the derivatives on the books of the dealer and a corresponding increase in its profits.
Some banks reported several billion dollars of profits from this source in 2008. Not surprisingly,
DVA gains and losses have now been excluded from the definition of common equity in
determining regulatory capital. In this paper we will focus on CVA, but many of the points we
make are equally applicable to DVA.
Market variables that affect the no-default value of a dealers outstanding transactions with a
counterparty also affect the dealers CVA for that counterparty. We will refer to these market
variables as the underlying market variables. In addition, CVA is affected by the
counterpartys term structure of credit spreads (the counterparty credit spreads). CVA
therefore gives rise to two types of exposures. One arises from potential movements in the
underlying market variables; the other from potential movements in counterparty credit spreads.
In December 2010, the Basel Committee on Banking Supervision published a new regulatory
framework for banks known as Basel III.
3
It requires a dealers CVA risk arising from changes
in a counterparty credit spreads to be identified and included in the calculation of capital for
market risk. However, the dealers CVA risk arising from underlying market variables are not
included in this calculation. Some dealers have developed their own sophisticated systems for
managing both types of risk. These dealers feel that Basel III proposals are inadequate because,
if a dealer hedges against the underlying market variables, the hedging trades will lead to an
increase, not a decrease, in its required capital.
4
Since the crisis, governments have moved to require standardized over-the counter derivatives
to be cleared through central clearing parties. This paper focuses on how counterparty credit risk
3
See Basel Committee on Banking Supervision (2010)
4
See, for example, Pengelly (2011)
4
5
How derivatives dealers will assess their credit exposure to arising from transactions that are cleared through a
CCP remains to be seen. The dealer is exposed to a) a default by any of the other clearing house members and b) a
default by the clearing house itself.
5
counterparty tends to be high (low) when the dealers exposure to the counterparty is low (high)
is referred to right-way risk.
Evaluating the amount of wrong-way or right-way risk in transactions with a counterparty
requires a good knowledge of the counterpartys business, in particular the nature of the risks
facing the business. It also requires knowledge of the transactions the counterparty has entered
into with other dealers.
6
The latter is difficult to know precisely but the extra transparency
provided by post-crisis legislation may help.
One situation in which wrong-way risk tends to occur is when a counterparty is selling credit
protection to the dealer. (AIG and monolines are obvious examples here.) This is because credit
spreads are correlated. When credit spreads are high (low) the value of the protection to the
dealer is large (small) and as a result the dealer has a large (small) exposure to the counterparty.
At the same time the credit spreads of the counterparty are likely to be high (low) indicating a
higher (lower) probability of default by the counterparty. Similarly, right-way risk tends to occur
when a counterparty is buying credit protection from the dealer. A situation in which a
counterparty is speculating by entering into many similar trades with one or more dealers is
likely to lead to wrong-way risk for those dealers. Transactions with a dealer that are concerned
with a strategy by the counterparty to partially hedging an existing exposure should give rise to
right-way risk because the counterpartys overall risk is reduced when the transactions have
negative value to the counterparty and positive value to the dealer.
7
This paper explains the way CVA is calculated and the advanced approach for calculating the
CVA risk capital under Basel III. The paper then proposes a model for incorporating wrong-way
risk into CVA calculations. Other authors who have attempted to tackle the wrong-way risk
problem include Cespedes et al (2010) and Sokol (2010). These authors propose an exposure
sampling approach where the process followed by the exposure is approximated with a one-
factor Markov process and a Gaussian (or other) copula is used to model the dependence
between this process and the time to default.
6
A complication is that sometimes a company will trade with many different market participants in an attempt to
conceal its true exposure.
7
An exception could be when the counterparty is liable to run into liquidity problems. Although the assets being
hedged have increased in value, the dealer might be unable to post collateral when required. An example here is
Ashanti Goldfields in September 1999. See for example Hull (2010, page 395) for a description of what happened.
6
Our model is simpler than the exposure sampling approach and involves a relatively small
adjustment to the method used to calculate CVA when the usual assumption of no dependence
between exposure and probability of default is made. We specify a deterministic model relating
the hazard rate of the counterparty to the value of the dealers transactions with the counterparty.
Numerical results from the implementation of the model are presented.
2. Calculating CVA
Suppose that T is the longest maturity derivative outstanding between a derivatives dealer and
one of its counterparties. Define v(t) as the value of a derivative that pays off the dealers net
exposure to the counterparty at time t and R as the recovery rate (assumed to be deterministic and
constant through time). If we assume that the probability of a counterparty default at time t is
independent of v(t) then, as discussed by, for example, Hull and White (1995), Canabarro and
Duffie (2003), and Picault (2005),
( ) ( ) ( )
0
CVA 1
T
t
R u t v t dt
=
=
}
(1)
where u(t) is the probability density function of the risk-neutral time to default for the
counterparty.
If no collateral is posted the net exposure, E
NC
(t), at time t is given by
( ) ( ) ( )
max , 0
NC
E t w t = (2)
where w(t) is the value to the dealer of the derivatives portfolio at time t. We assume that
collateral is posted continuously and that there is no minimum transfer amount.
8
If transactions
are collateralized with zero threshold and an early termination is declared as soon as the
counterparty fails to post the required collateral, then in the event of default at time t the
available collateral is w(t). If there is a threshold, K, the available collateral is max(w(t)K, 0). In
practice, a period of time elapses between the time when the counterparty ceases to post
8
Calculationscanbeadjustedtorelaxtheseassumptions.
7
collateral and an early termination event being declared by the dealer. This is referred to the
cure period. Suppose that the length of the cure period is c. The collateral available if there is
termination event at time t is
( ) ( ) ( )
max , 0 C t w t c K = (3)
An independent amount, I, can be treated as a negative threshold so that
( ) 0 , ) ( max ) ( I c t w t C + =
The net exposure, E
NET
, at time t is therefore
( ) ( ) ( ) ( )
max , 0
NET NC
E t E t C t = (4)
To approximate the integral in equation (1) we can choose times t
i
(0 i n) with t
0
= 0, t
n
= T
and t
0
< t
1
< t
2
< < t
n
and set
( )
1
CVA 1
n
i i
i
R u v
=
=
(5)
where u
i
is the probability of default between times t
i1
and t
i
and
( )
*
i i
v v t = with
( )
*
1
0.5
i i i
t t t
= + . Note that u
i
is the unconditional risk-neutral probability of default between
times t
i1
and t
i
(as seen at time zero). It is not the probability of default conditional on no earlier
default.
9
The u
i
s are usually calculated from credit spreads. Sometimes a complete term structure of
credit spreads for the counterparty can be observed in the market; sometimes it has to be
estimated using credit spread data for other companies. If s
i
is the credit spread for a maturity of
t
i
, an estimate of the average risk-neutral hazard rate between times 0 and t
i
is ( ) 1
i
s R so that
the probability of no default between times 0 and t
i
is ( ) exp 1
i i
s t R (
. It follows that
9
A confusion concerning the use of conditional and unconditional default probabilities in calculating CVA was
pointed out by Rebonato et al (2010).
8
1 1
exp exp
1 1
i i i i
i
s t s t
u
R R
| | | |
=
| |
\ . \ .
(6)
Using a delta/gamma approximation, the impact on CVA of a small change As in all the s
i
s is
( )
( )
( )
1 1
1
1
2
2 2 1 1
1
1
CVA exp exp
1 1
1
exp exp
2 1 1 1
n
i i i i
i i i
i
n
i i i i
i i i
i
s t s t
t t v s
R R
s t s t
t t v s
R R R
=
(
| | | |
A = A
| | (
\ . \ .
( | | | |
+ A
| | (
\ . \ .
(7)
This equation enables the dependence of CVA on counterparty credit spreads to be included in
the banks model for calculating market risk capital. Equations (5), (6), and (7) correspond to the
equations used in the Basel III advanced approach for determining capital for CVA risk.
10
DVA can be handled similarly. If R is the recovery rate of the dealer, v(t) is the value of a
derivative that pays off the counterpartys exposure to the dealer at time t, and u(t) is the
probability density function of the dealers time to default, then the right hand side of equation
(1) gives the DVA. The counterpartys net exposure to the dealer, after taking collateral posted
by the dealer into account, can be calculated in an analogous way to that indicated above and the
approximations indicated in equations (5) and (6) can be used with u
i
defined as the dealers risk-
neutral probability of default between times t
i1
and t
i
, v
i
defined as the net exposure of the
counterparty to the dealer at time
*
i
t , and s
i
defined as the dealers credit spread for a maturity t
i
.
With this redefinition of variables, equation (7) gives the sensitivity of DVA to the dealers
credit spread (but this is not part of Basel III market risk capital calculations).
3. Monte Carlo Simulation
In practice, the v
i
s are almost invariably calculated using Monte Carlo simulation. The market
variables affecting the no-default value of a dealers derivatives with a counterparty are
10
Charging capital for the CVA risk associated with credit spreads is likely to encourage dealers to hedge spread
risk by buying credit default swaps. This might have the unintended consequence of increasing credit spreads.
9
11
When interest rates are stochastic a convenient numeraire is the value of a risk-free zero coupon bond maturing at
the next valuation time. The Monte Carlo simulation can then be implemented by assuming that a) the returns on
market variables between valuation dates and b) the discount rates between valuation dates equal the yield on the
numeraire bond.
12
Only a small amount of additional computation time is necessary to calculate DVA as well as CVA.
13
Note that the impact of a new transaction on CVA can be positive or negative.
10
exposures calculated at these times. An interesting point is that these estimates should in theory
be made using the real-world measure, not the risk-neutral measure.
4. CVA Exposures
Calculating the sensitivity of CVA to a small parallel shift in a counterpartys credit spread is
straightforward using equation (7).
Calculating the first and second partial derivatives of CVA with respect to the underlying market
variables is liable to be more time consuming. Consider a market variable x with initial value x
0
.
It is necessary to calculate the effect on the paths sampled of changing x
0
to x
0
+c and x
0
c for a
small c when all random number streams are kept the same. When the variable follows geometric
Brownian motion this is not too difficult. A small percentage change at time zero leads to the
same small percentage change at all future times on all simulation trials. (This is true both when
the volatility is deterministic and when the volatility is stochastic.) For other variables such as
those following mean reverting processes, the impact of a change at time zero on the change at
future times is liable to depend on the path followed by the market variable.
Suppose that
+
i
v and
i
v are the values calculated for v
i
when the initial value of the market
variable is x
0
+c and x
0
c, respectively. From equation (5)
( )
1
CVA 1
2
n
i i i
i
R
u v v
x
+
=
c
=
c c
( )
2
2 2
1
CVA 1
2
n
i i i i
i
R
u v v v
x
+
=
c
= +
c c
These equations enable CVA risks relating to the underlying market variables to be assessed and
hedged. As already mentioned, under Basel III CVA exposures arising from the underlying
market variables are not included in the calculation of market risk capital.
11
simulations, up to any time T matches that calculated from credit spreads. When b > 0, h(t) is an
increasing function of w(t), which corresponds to wrong-way risk; when b < 0, h(t) is a
decreasing function of w(t), which corresponds to right-way risk. In the model given by equation
(8), h(t) increases exponentially as w(t) takes more extreme values.
14
In the model in equation (9)
h(t) increases linearly as w(t) takes on more extreme values.
15
For moderate values of w(t) the
two models are very similar. In the results we present we use the model in equation (9), which is
illustrated in Figure 1.
The parameter b requires some judgment about the sensitivity of the counterpartys credit spread
to the value of the portfolio. The type of estimate required can be illustrated with an example.
Suppose that the current value to the dealer of the portfolio, w(0), is $3 million and the
counterpartys five-year credit spread is 300 basis points. Assuming a recovery rate of 40%, this
means that the average five-year hazard rate is 5% per year. Also, suppose that it is estimated
that, if w(0) increased to $20 million, the spread would rise to 600 basis points, corresponding to
an average five-year hazard rate of 10%. Assuming the term structure of hazard rates is flat,
16
when w(0) = 3 million, h(0) = 5% and when w(0) = 20 million, h(0) = 10%. Solving a pair of
simultaneous equations for a(0) and b, we find that at time zero, for the model in equation (8), b
is 0.0408 per million dollars while, for the model in equation (9), b is 0.0423 per million dollars.
The implementation of the model requires a small change to the procedure for calculating CVA
described in Section 3. The values of
( ) ( )
*
for 1
i
a t i n s s must be determined so that the average
survival probability, across all simulations, up to time
*
i
t equals the survival probability
calculated from the term structure of credit spreads. This means that we require
1 1
1
exp( ) exp for 1
1
m k
k k
ij
j i
s t
h t k n
m R
= =
( | |
A = s s
| (
\ .
where h
ij
and w
ij
are the values of ) (
*
i
t h and ) (
*
i
t w on the jth simulation trial and m is the number
of simulation trials.
14
When b > 0 the relevant extreme values are high; when b < 0 they are low.
15
Either model can be modified to incorporate a cap on h(t).
16
Given the imprecision of any attempt to quantify wrong-way risk this is a reasonable assumption.
13
17
The results from using the model in equation (8) are similar.
14
magnitude of the effect is difficult to predict. Indeed, in some cases, even the direction of effect
can be difficult to predict. As shown by the tables, sometimes the direction of the effect depends
on the collateral arrangements.
In general, the impact of wrong-way and right-way risk on CVA depends in a complex way on
CVA itself and the collateral arrangements. To illustrate this, we randomly generated 250
portfolios. Each portfolio consists of 25 options on one of five different assets. The asset prices
are assumed to follow geometric Brownian motion with pairwise correlations of 0.36. Each
option has the following properties.
(a) It is equally likely to be long or short
(b) It is equally likely to be a call or a put
(c) The underlying is equally likely to any one of the five assets
(d) All maturities between one and five years are equally likely
(e) All strike prices within 30% of the current asset price are equally likely
(f) The underlying principal is $25 million.
The assets do not provide any income. They have an initial price of $25 and a volatility of 25%.
The risk-free rate is 5%, the credit spread of the counterparty (all maturities) is 125 basis points,
and the recovery rate is 40%.
Figure 2 provides a scatter plot of the relationship between the dollar change in CVA caused
increasing b from 0 to 0.01 per million dollars and the CVA for b=0 for the 250 portfolios when
there is no collateralization. It can be seen that the change in CVA tends to increase as CVA
increases. The reason is that as CVA increases there is a tendency for both the mean and standard
deviation of the value of the portfolio at future times to increase. The gap between the exposures
on high-w(t)-paths and low-w(t)-paths increases. Wrong-way risk causes the hazard rate to
increase dramatically on the high-w(t) paths and decrease (modestly) on the low-w(t) paths.
Figure 3 provides a similar plot to Figure 2 for the situation where the threshold is $10 million
and the cure period is 15 days. The average relationship between the change in CVA and CVA is
15
in this case is one where the change first increases and then decreases. To understand the reason
for this, we ignore the cure period and focus on the threshold. The impact of the threshold is to
restrict the net exposure to less than $10 million. For low w(t)s this restriction has little effect.
However, as w(t) increases, the net exposure is increasingly impacted by the $10 million
restriction. The gap between the exposure on high-w(t) paths and low-w(t) paths is much less
than in the no collateral case. The highest theoretical CVA is achieved when the value of w(t) is
certain to be above $10 million at all times. In this case, the net exposure is the same on all w(t)-
paths and wrong-way risk has no effect. This explains the pattern observed in Figure 3. The cure
period does have an effect on the results, but its effect is less than the effect of the threshold.
Figure 4 provides a similar plot to Figures 2 and 3 for the situation where the threshold is zero
and the cure period is 15 days. In this case the net exposure is entirely as a result of the cure
period. When w(t) is positive, the net exposure increases as the standard deviation of the change
in w(t) during the cure period increases. High standard deviations of w(t) tend to be associated
with high w(t)s. High w(t)s are in turn associated with increases in the hazard rate when there is
wrong-way risk. This leads to the pattern shown in Figure 4.
Figure 5 provides a plot for the situation where there is an independent amount equal to $5
million and a cure period of 15 days. As in the case of Figure 4 the impact of the cure period on
the exposure at time t for a particular w(t) depends on the standard deviation of the change in
w(t) during the cure period. Indeed, for positive exposures to be generated, the standard deviation
has to be sufficiently high that there is a reasonable chance of a $5 million increase in w(t) during
the cure period. The reasons for the pattern observed are similar to those for Figure 4.
The impact of right-way risk (where b is changed from 0 to 0.01 instead of from 0 to +0.01) is
similar to that shown in Figures 2 to 5, except that the change in CVA is negative instead of
positive. We have carried out other experiments involving portfolios of interest rate swaps and
portfolios. The results are similar to those in Figures 2 to 5.
7. Conclusions
We have proposed a simple model for handling wrong-way/right-way risk in the calculation of
CVA. The model can be implemented by making a small change to usual method for calculating
16
CVA. It is necessary for the dealer to make a single estimate describing the sensitivity of the
counterpartys credit spread to the value of its portfolio with the counterparty.
Tests of the model show that wrong-way and right-way risk have a significant effect on the
Greek letters of CVA as well as on CVA itself. Because CVA is such a complex derivative, it is
difficult to estimate these effects without a model. Indeed, even the sign of the effect can be
counterintuitive.
Further tests involving the random generation of portfolios indicate that when there is no
collateral or when collateral is posted with zero threshold or when collateral is posted with an
independent amount, the dollar impact of wrong-way risk on CVA tends to increase as CVA
increases. The situation where the threshold is materially positive is more interesting. As CVA
increases, the impact of wrong way risk tends to first increase and then decrease.
17
REFERENCES
Basel Committee on Banking Supervision (2010), Basel III: A Global Regulatory Framework
for More Resilient Banks and Banking Systems, http://www.bis.org/publ/bcbs189_dec2010.pdf,
December 2010.
Canabarro, Eduardo and Darrell Duffie (2003), Measuring and Marking Counterparty Risk,
Chapter 9 in Asset/Liability Management for Financial Institutions, Edited by Leo Tilman, New
York: Institutional Investor books.
Cepedes, Juan Carlos Garcia, Juan Antonio de Juan Herrero, Dan Rosen, and David Saunders
(2010), Effective Modeling of Wrong Way Risk, Counterparty Credit Risk Capital and Alpha
in Basel II, Journal of Risk Model Validation, 4, 1, pp 71-98
Gregory, Jon (2009), Counterparty Credit Risk: The New Challenge for Financial Markets
John Wiley and Sons.
Hull, John (2010), Risk Management and Financial Institutions, Second Edition, Pearson.
Hull, John and Alan White (1995), "The impact of default risk on the prices of options and other
derivative securities,'' Journal of Banking and Finance, Vol 19, pp 299-322.
Hull, John and Alan White (2001), Valuing Credit Default Swaps II: Modeling Default
Correlations, Journal of Derivatives, 8, 3 (Spring 2001), pp12-22.
Pengelley, Mark (2011), CVA Melee Risk, 24, 2, pp 37-39.
Picault, Evan (2005), Calculating and Hedging Exposure, CVA, and Economic Capital for
Counterparty Credit Risk, in Counterparty Credit Risk Modelling edited by Michael Pykhtin,
London: Risk Books.
Rebonato, Riccardo, Mike Sherring, and Ronnie Barnes (2010), Credit risk, CVA, and the
equivalent bond, Risk, 23, 9, pp 118-121.
Reuters (2008), Update 1-Lehman Brothers Holdings is Focus of Grand Jury Probe, October
16.
Sokol, Alexander (2010), A Practical Guide to Monte Carlo CVA, Chapter 14 in Lessons
From the Crisis edited by Arthur Berd, London: Risk Books
18
Table 1: Impact of wrong-way risk on CVA for a long forward contract to buy 100 million units
of a foreign currency in one year. The current exchange rate is 1.0, the domestic and foreign risk-
free interest rates are both 5%, and the volatility of the exchange rate is 15%. The credit spread is
125 basis points for all maturities and the recovery rate is 40%. K is the threshold in $ millions,
c is the cure period in days, and b is the parameter in equation (9).
No
Collateral
K = 10 K = 0 K = 5
c = 15 c = 15 c = 15
CVA ($ millions) for b = 0 0.048 0.036 0.011 0.002
Impact of b = 0.03 per $mm on:
CVA 54.8% 41.7% 37.3% 53.5%
Delta wrt Exch Rate 32.0% 15.6% 12.8% 39.3%
Gamma wrt Exch Rate 2.6% 25.4% 17.7% 0.7%
Delta wrt Spread 53.8% 41.2% 36.8% 52.8%
Gamma wrt Spread 181.8% 124.3% 122.8% 184.3%
Table 2: Impact of wrong-way risk on CVA for a short forward contract to sell 100 million units
of a foreign currency in one year. The current exchange rate is 1.0, the domestic and foreign risk-
free interest rates are both 5%, and the volatility of the exchange rate is 15% . The credit spread
is 125 basis points for all maturities and the recovery rate is 40%. K is the threshold in $
millions, c is the cure period in days, and b is the parameter in equation (9).
No
Collateral
K = 10 K = 0 K = 5
c = 15 c = 15 c = 15
CVA ($ millions) for b = 0 0.048 0.039 0.011 0.001
Impact of b = 0.03 per $mm on:
CVA 40.5% 34.0% 27.6% 28.9%
Delta wrt Exch Rate 16.2% 7.7% 1.9% 341.9%
Gamma wrt Exch Rate 7.0% 21.4% 16.4% 26.5%
Delta wrt Spread 40.0% 33.7% 27.4% 28.8%
Gamma wrt Spread 114.8% 91.0% 77.0% 70.7%
19
Table 3: Impact of right-way risk on CVA for a long forward contract to buy 100 million units
of a foreign currency in one year. The current exchange rate is 1.0, the domestic and foreign risk-
free interest rates are both 5%, and the volatility of the exchange rate is 15%. The credit spread is
125 basis points for all maturities and the recovery rate is 40%. K is the threshold in $ millions,
c is the cure period in days, and b is the parameter in equation (9).
No
Collateral
K = 10 K = 0 K = 5
c = 15 c = 15 c = 15
CVA ($ millions) for b = 0 0.048 0.036 0.011 0.002
Impact of b = 0.03 per $mm on:
CVA 37.5% 32.7% 29.1% 35.7%
Delta wrt Exch Rate 26.7% 18.8% 14.8% 28.9%
Gamma wrt Exch Rate 8.2% 11.7% 16.0% 6.2%
Delta wrt Spread 37.2% 32.5% 28.9% 35.6%
Gamma wrt Spread 79.2% 74.5% 72.1% 77.3%
Table 4: Impact of right-way risk on CVA for a short forward contract to buy 100 million units
of a foreign currency in one year. The current exchange rate is 1.0, the domestic and foreign risk-
free interest rates are both 5%, and the volatility of the exchange rate is 15%. The credit spread is
125 basis points for all maturities and the recovery rate is 40%. K is the threshold in $ millions,
c is the cure period in days, and b is the parameter in equation (9).
No
Collateral
K = 10 K = 0 K = 5
c = 15 c = 15 c = 15
CVA ($ millions) for b = 0 0.048 0.039 0.011 0.001
Impact of b = 0.03 per $mm on:
CVA 33.9% 30.8% 25.9% 26.9%
Delta wrt Exch Rate 19.3% 13.6% 4.9% 209.1%
Gamma wrt Exch Rate 0.9% 14.4% 16.7% 37.5%
Delta wrt Spread 33.6% 30.6% 25.7% 26.7%
Gamma wrt Spread 78.8% 75.5% 71.3% 69.0%
20
Figure 1: The model in equation (9) when (i) a(t)= 4 and b=0.01 and (ii) a(t)= 4 and b=0.01
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
400 300 200 100 0 100 200 300 400
h(t)
w(t)
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
400 300 200 100 0 100 200 300 400
h(t)
w(t)
(i) (ii)
21
Figure 2: Impact of wrong-way risk for 250 portfolios of options when there is no collateral.
The horizontal axis shows CVA when b=0. The vertical axis shows the change in CVA when
b is increased from 0 to 0.01 per million.
0
0.2
0.4
0.6
0.8
1
1.2
0 0.5 1 1.5 2 2.5 3
ChangeinCVA
($millions)
CVA(nowrongwayrisk,$million
22
Figure 3: Impact of wrong-way risk for 250 portfolios of options when the threshold is $10
million and the cure period is 15 days. The horizontal axis shows CVA when b=0. The vertical
axis shows the change in CVA when b is increased from 0 to 0.01 per million
0
0.01
0.02
0.03
0.04
0.05
0.06
0.07
0.08
0.09
0.1
0 0.1 0.2 0.3 0.4 0.5 0.6
ChangeinCVA
($millions)
CVA(nowrongwayrisk,$million)
23
Figure 4: Impact of wrong-way risk for 250 portfolios of options when the threshold is zero and
the cure period is 15 days. The horizontal axis shows CVA when b=0. The vertical axis shows
the change in CVA when b is increased from 0 to 0.01.
0
0.01
0.02
0.03
0.04
0.05
0.06
0 0.05 0.1 0.15
ChangeinCVA
($millions)
CVA(nowrongwayrisk,$millions)
24
Figure 5: Impact of wrong-way risk for 250 portfolios of when there is an independent amount
of $5 million and a cure period of 15 days. The horizontal axis shows CVA when b=0. The
vertical axis shows the change in CVA when b is increased from 0 to 0.01 per million
0
0.005
0.01
0.015
0.02
0.025
0.03
0.035
0.04
0.045
0 0.01 0.02 0.03 0.04 0.05 0.06 0.07
ChangeinCVA
($millions)
CVA(nowrongwayrisk,$millions)