P2.T7.

Operational & Integrated Risk Management

Bionic Turtle FRM Practice Questions

Michael Crouhy, et al., The Essentials of Risk

Management
By David Harper, CFA FRM CIPM
www.bionicturtle.com
Crouhy, Chapter 17: Risk Capital Attribution and Risk-Adjusted Performance
Measurement
P2.T7.606. ECONOMIC CAPITAL AND RISK-ADJUSTED RETURN ON ECONOMIC CAPITAL (RAROC) . 3
P2.T7.607. ADJUSTED RAROC AND RAROC IN PRACTICE ........................................................ 9

Appendix (OPTIONAL)

Crouhy, Chapter 14: Capital Allocation and Performance Measurement

P2.T7.1. RISK-ADJUSTED RETURN ON CAPITAL, RAROC...........................................................14
P2.T7.2. ALLOCATING ECONOMIC (RAROC) CAPITAL TO CREDIT, MARKET & OPERATIONAL RISK .17
P2.T7.3. LOAN EQUIVALENCE AND ADJUSTED RAROC (ARAROC) ...........................................20

Crouhy, Chapter 15: Model Risk

P2.T7.604. MODEL ERROR AND MODEL IMPLEMENTATION RISK..................................................24
P2.T7.605. MITIGATION OF MODEL RISK ...................................................................................29

2
Crouhy, Chapter 17: Risk Capital Attribution and Risk-
Adjusted Performance Measurement
P2.T7.606. Economic capital and risk-adjusted return on economic capital (RAROC)
P2.T7.607. Adjusted RAROC and RAROC in practice

P2.T7.606. Economic capital and risk-adjusted return on economic

capital (RAROC)
Learning Objectives: Define, compare, and contrast risk capital, economic capital and
regulatory capital, and explain the motivations for using economic capital. Describe the
RAROC (risk-adjusted return on capital) methodology and its benefits. Compute and
interpret the RAROC for a project, loan, or loan portfolio, and use RAROC to compare
business unit performance.

606.1. Peter Smith works at Bland Bancorp, a large international bank and financial services
firm. The bank has recently made several significant investments but their success and
profitability are highly uncertain; further, the bank has also made several acquisitions. At a
board committee meeting, Peter is asked to explain the difference between risk capital,
economic capital and regulatory capital. Which of the following is the TRUE statement that is
part of his response?

a) "Economic capital is less than risk capital because risk capital is the capital required to
guarantee shareholders with 100.0% confidence that it holds enough risk capital to ride
out any eventuality"
b) "Regulatory capital is almost certainly higher than economic capital in our securitization
business, therefore we do not need to know its economic capital; in other words,
whenever regulatory capital exceeds economic capital, there is no reason or need to
know economic capital"
c) "We need enough risk capital to absorb unexpected losses and remain solvent over a
time horizon with a selected confidence level, but with respect to risk capital, (unlike
regulatory capital) the horizon and confidence belong to us as policy parameters."
d) "For capital budgeting purposes, we should use ex ante RAROC with economic capital
in the numerator; for performance evaluation we should use ex post RAROC with
regulatory capital in the numerator"

3
606.2. With respect to risk-adjusted performance metrics, let "ratio consistent" refer to a ratio
that adjusts both the numerator and denominator for risk. Among the risk-adjusted performance
measures discussed in Crouhy, each of the following is true statement about the defined
metric EXCEPT which statement is not true?

a) RAROC (risk-adjusted return on economic capital) risk-adjusts the numerator's return by

subtracting expected loss from expected revenue and risk-adjusts the denominator by
substituting economic capital for accounting capital. RAROC is therefore ratio-
consistent.
b) RORAC (return on risk-adjusted capital; aka, ROCaR or ROC) only adjusts the
denominator, in practical applications using VaR in the denominator. Because its
numerator is "net income," RARAC is ratio-inconsistent.
c) RAROA (risk-adjusted return on risk-adjusted assets) has a denominator in common
with RORAA (return on risk-adjusted assets) but RAROA employs "risk-adjusted
expected net income" in the numerator while RORAA employs "net income" in the
numerator. Therefore, RAROA is ratio-consistent but RORAA is ratio-inconsistent.
d) NPV (net present value) the discounted value of expected cash flows but its glaring
weakness is that it is not compatible with the capital asset pricing model (CAPM) and,
related, NPV has no convenient method for incorporating risk

606.3. Assume a $1.0 billion corporate loan portfolio offers a return of 5.0% per annum. The
bank (the lender) has a direct operating cost of $6.0 million per annum and an effective tax rate
of 25.0%. The portfolio is funded by $1.0 billion in retail deposits with a transfer-priced interest
rate charge of 1.40%. Risk analysis of the unexpected losses associated with the portfolio tell us
we need to set aside economic capital of $80.0 million against the portfolio; i.e., 8.0% of the
loan amount. The bank's economic capital must be invested in risk-free securities and the risk-
free rate on government securities is only 1.0%. The expected loss on the portfolio is assumed
to be 1.0% per annum; i.e., $10 million. Which is NEAREST to the risk-adjusted return on
economic capital (RAROC)?

a) 8.75%
b) 13.00%
c) 19.50%
d) 25.25%

4
Answers:

606.1. C. TRUE: "We need enough risk capital to absorb unexpected losses and remain
solvent over a time horizon with a selected confidence level, but horizon and confidence
belong to us as policy parameters."

In regard to false (A): The firm cannot realistically achieve 100.0% confidence in its own
solvency

In regard to false (B): "The new regulatory capital requirements imposed by Basel III make it
likely that for some activities, such as securitization, regulatory capital may end up much higher
than economic capital. Still, economic capital calculation is essential for senior management as
a benchmark to assess the economic viability of the activity for the financial institution. When
regulatory capital is much larger than economic capital, then it is likely that over time the activity
will migrate to the shadow banking sector, which can price the transactions at a more attractive
level. "

In regard to false (D): "RAROC for Performance Measurement: We should emphasize at this
point that RAROC was first suggested as a tool for capital allocation on an anticipatory or ex
ante basis. Hence, expected revenues and losses should be plugged into the numerator of the
RAROC equation for capital budgeting purpose. When RAROC is used for ex post, or after the
fact, performance evaluation, we can use realized revenues and realized losses, rather than
expected revenues and losses, in our calculation. "

Please note the definitions by Crouhy (emphasis ours):

 Risk capital: "Risk capital is the cushion that provides protection against the various
risks inherent in the business of a corporation so that the firm can maintain its financial
integrity and remain a going concern even in the event of a near-catastrophic worst-case
scenario. Risk capital gives essential confidence to the corporation’s stakeholders, such
as suppliers, clients, and lenders (for an industrial firm), or claimholders, such as
depositors and counterparties in financial transactions (for a financial institution). Risk
capital is often called economic capital, and in most instances the generally accepted
convention is that risk capital and economic capital are identical (although later in this
chapter we introduce a slight wrinkle by defining economic capital as risk capital plus
strategic capital) ... Risk capital measurement is based on the same concepts as the
value-at-risk (VaR) calculation methodology that we discuss in [Chapter 7: Measuring
Market Risk: Value at Risk ...]. Indeed, risk capital numbers are often derived from, or
supported by, sophisticated internal VaR models, augmented in recent years by stress
testing. However, the choice of the confidence level and time horizon when using
VaR to calculate risk capital are key policy parameters that should be set by
senior management (or the senior risk management committee). Usually, these
decisions should be endorsed by the board. Risk capital should be calculated in such
a way that the institution can absorb unexpected losses up to a level of confidence in
line with the requirements of the firm’s various stakeholders. No firm can offer its
stakeholders a 100% guarantee (or confidence level) that it holds enough risk capital to
ride out any eventuality. Instead, risk capital is calculated at a confidence level set at
less than 100%— say, 99.90% for a firm with conservative stakeholders.

5
 In theory, this means that there is a probability of around 1/ 10 of 1.0% that actual losses
will exceed the amount of risk capital set aside by the firm over the given time horizon
(generally one year). The exact choice of confidence level is typically associated with
some target credit rating from a rating agency such as Moody’s, Standard & Poor’s, and
Fitch as these ratings are themselves explicitly associated with a probability of default. It
should also be in line with the firm’s stated risk appetite
 Strategic risk capital = goodwill + burned-out capital: "Risk capital is the capital cushion
that the bank must set aside to cover the worst-case loss (minus the expected loss) from
market, credit, operational, and other risks, such as business risk and reputation risk, at
the required confidence threshold; e.g., 99%. Risk capital is directly related to the value-
at-risk calculation at the one-year time horizon and at the institution’s required
confidence level. Strategic risk capital refers to the risk of significant investments about
whose success and profitability there is high uncertainty. If the venture is not successful,
then the firm will usually face a major write-off, and its reputation will be
damaged. Current practice is to measure strategic risk capital as the sum of
burned-out capital and goodwill. Burned-out capital refers to the idea that capital is
spent on, say, the initial stages of starting up a business but the business may ultimately
not be kicked off due to projected inferior risk-adjusted returns. It should be viewed as an
allocation of capital to account for the risk of strategic failure of recent acquisitions or
other strategic initiatives built organically. This capital is amortized over time as the risk
of strategic failure dissipates. The goodwill element corresponds to the investment
premium— i.e., the amount paid above the replacement value of the net assets (assets
– liabilities) when acquiring a company. (Usually, the acquiring company is prepared to
pay a premium above the fair value of the net assets because it places a high value on
intangible assets that are not recorded on the target’s balance sheet.) Goodwill is also
depreciated over time. "
 Economic capital is the sum of risk capital plus strategic risk capital. Economic capital
is the denominator in the RAROC ratio.

6
606.2. D. FALSE. This is false because NPV is compatible with CAPM and does have a
basic risk-adjustment mechanism in the selection of the discount rate. Riskier cash flows
are discounted at higher rates.

In regard to (A), (B) and (C), each is TRUE.

Crouhy's definitions include:

 RAROC (risk-adjusted return on capital) = risk-adjusted expected net income/ economic
capital. RAROC makes the risk adjustment to the numerator by subtracting a risk factor
from the return— e.g., expected loss. RAROC also makes the risk adjustment to the
denominator by substituting economic capital for accounting capital.
 RORAC (return on risk-adjusted capital) = net income/ economic capital. RORAC makes
the risk adjustment solely to the denominator. In practical applications, RORAC =
P&L/VaR
 ROC (return on capital) = RORAC. It is also called ROCAR (return on capital at risk).
 RORAA (return on risk-adjusted assets) = net income/ risk-adjusted assets.
 RAROA (risk-adjusted return on risk-adjusted assets) = risk-adjusted expected net
income/ risk-adjusted assets.
 S (Sharpe ratio) = (expected return – risk-free rate)/ volatility. The ex post Sharpe ratio—
i.e., that based on actual returns rather than expected returns— can be shown to be a
multiple of ROC.
 NPV (net present value) = discounted value of future expected cash flows, using a risk-
adjusted expected rate of return based on the beta derived from the CAPM, where risk is
defined in terms of the covariance of changes in the market value of the business with
changes in the value of the market portfolio (see Chapter 5). In the CAPM, the definition
of risk is restricted to the systematic component of risk that cannot be diversified away.
For RAROC calculations, the risk measure captures the full volatility of earnings,
systematic and specific. NPV is particularly well suited for ventures in which the
expected cash flows over the life of the project can be easily identified.
 EVA (economic value added), or NIACC (net income after capital charge), is the after-
tax adjusted net income less a capital charge equal to the amount of economic capital
attributed to the activity, times the after-tax cost of equity capital. The activity is deemed
to add shareholder value, or is said to be EVA positive, when its NIACC is positive (and
vice versa). 2 An activity whose RAROC is above the hurdle rate is also EVA positive.

7
606.3. C. 19.50%

RAROC = (after-tax expected risk-adjusted net income)/(economic capital). In this case:

 Expected revenue = $1.0 billion loan portfolio * 5.0% = $50.0 million
 Expected losses = $1.0 billion loan portfolio * 1.0% = $10.0 million
 Interest expense = $1.0 billion borrowed funds * 1.40% = $14.0 million
 Operating cost = $6.0 million (given as an assumption)
 Return on economic capital (EC) = $80.00 EC * 1.0% = $0.80 million
 Tax rate = 0.25 (given as assumption)
Such that RAROC = [($50.0 - 10.0 - 14.0 - 6.0 + 0.80)*(1.0 - 0.25 tax rate)] / 80.0 = 19.50%

Discuss in forum here: https://www.bionicturtle.com/forum/threads/p2-t7-606-economic-

capital-and-risk-adjusted-return-on-economic-capital-raroc-crouhy.9726/

8
P2.T7.607. Adjusted RAROC and RAROC in practice
Learning Objectives: Explain the impact of changing assumptions used in calculating
economic capital, including choosing a time horizon, measuring default probability, and
choosing a confidence level. Calculate the hurdle rate and apply this rate in making
business decisions using RAROC. Compute the adjusted RAROC for a project to
determine its viability. Explain challenges in modeling diversification benefits, including
aggregating a firm’s risk capital and allocating economic capital to different business
lines. Explain best practices in implementing a RAROC approach.

607.1. Assume a bank's $2.0 billion corporate loan portfolio offers a return of 6.0% per annum.
The expected loss on the portfolio is estimated to be 1.5% per annum; i.e., $30 million. The
portfolio is funded by $2.0 billion in retail deposits with a transfer-priced interest rate charge of
2.00%. The bank (the lender) has a direct operating cost of $16.0 million per annum and an
effective tax rate of 25.0%. Risk analysis of the unexpected losses associated with the portfolio
tell us we need to set aside economic capital of $200.0 million against the portfolio; i.e., 10.0%
of the loan amount. The bank's economic capital must be invested in risk-free securities and,
unfortunately in the regime of ultra low interest rates, the risk-free rate on government securities
is only 1.0%.

Although the loan portfolio's risk-adjusted return on capital (RAROC) is positive and seemingly
high, the bank wants to adjust the traditional RAROC calculation to obtain a RARO measures
that takes into account the systemic riskiness of the expected returns. If the risk-free rate is
1.0% (as above), and the expected rate of return on the market portfolio is 8.0% such that the
equity risk premium is 7.0%, and the beta of the firm's equity is 1.60, which of the following is
the correct adjusted RAROC and is the project advisable?

a) RAROC is 6.25% but no, the project is bad because ARAROC is below the risk-free rate
b) RAROC is 8.00% but no, the project is bad because ARAROC is below the risk-free rate
c) RAROC is 9.80% and yes, the project is good because ARAROC is above the risk-free
rate
d) RAROC is 13.50% and yes, the project is good because ARAROC is above the risk-free
rate

9
607.2. Consider a business unit (BU) which consists of two activities, X and Y, for which the
firm's risk staff has calculated three different measures of risk capital:

Please note:
 Stand-alone capital is the capital used by an activity taken independently of the other
activities in the same business unit; ie, risk capital calculated without any diversification
benefits. Here the stand-alone capital for X is $ 70 and for Y it is $ 80.
 Fully diversified capital is the capital attributed to each activity X and Y, taking into
account all diversification benefits from combining them under the same leadership.
Here the overall portfolio effect is $30 = $70 + 80 - 120. The firm's analysts allocated the
portfolio effect pro rata with the stand-alone risk capital: $30 × 70/150 = $14 for X and
$30 × 80/150 = $16 for Y, so that the fully diversified risk capital becomes $56 for X and
$64 for Y.
 Incremental capital (which is called marginal capital by Crouhy) is the additional capital
required by an incremental deal, activity, or business. It takes into account the full benefit
of diversification. Here the marginal risk capital for X (assuming that Y already exists) is
$40 = $120 - 80, and the marginal risk capital for Y (assuming that X already exists) is
$50 = $120 - 70. The summation of the marginal risk capital, $90 in this example, is less
than the full risk capital of the BU.

Each of the following statements is true EXCEPT which is not?

a) Given that the risk capital for the business unit is $120.0 (as shown), the implied
correlation between activities must be zero
b) Fully diversified capital ($56 for X and $64 for Y) should be used for assessing the
solvency of the firm and minimum risk pricing
c) Incremental capital (aka, Crouhy's marginal risk capital; $40 for X and $50 for Y) should
be used for active portfolio management or business mix decisions,
d) Stand-alone capital ($70 for X and $80 for Y) should be used for incentive
compensation; and fully-diversified capital ($56 for X and $64 for Y) can be used to
assess the extra performance generated by the diversification effects, such that
performance measurement involves both stand-alone and fully-diversified perspectives

10
607.3. According to Crouhy, Galai and Mark, best practice in the implementation of a RAROC
system should include each of the following elements EXCEPT which is incorrect?

a) Given the strategic nature of the decisions steered by a RAROC system, the marching
orders must come from the top management of the firm; specifically, the CEO and his or
her executive team should sponsor the implementation of a RAROC system and should
be active in its diffusion
b) In order to preserve the integrity and transparency of the RAROC system, there should
be an exclusively quantitative decision rule for activities: if the ex-ante RAROC does not
exceed the firm's hurdle rate (if the RAROC return is "low"), the firm should exit the
activity
c) The value at risk (VaR) methodologies for measuring market risk and credit risk that
underpin RAROC calculations are "generally well accepted" by business units (although
this is not yet true for operational risk); in practice, disagreements concern the setting of
the parameters that feed into these models which determine the size of economic capital
d) Balance sheet requests from the business units, such as economic capital, leverage
ratio, liquidity ratios, and risk-weighted assets, should be channeled to the RAROC
group every quarter; limits are then set for economic capital, leverage ratio, liquidity
ratios, and risk-weighted assets

11
Answers:

607.1. D. RAROC is 13.50% and yes, the project is good because ARAROC is above the
risk-free rate

RAROC = (after-tax expected risk-adjusted net income)/(economic capital). In this case:

 Expected revenue = $2.0 billion loan portfolio * 6.0% = $120.0 million
 Expected losses = $2.0 billion loan portfolio * 1.5% = $30.0 million
 Interest expense = $2.0 billion borrowed funds * 2.0% = $40.0 million
 Operating cost = $16.0 million (given as an assumption)
 Economic capital = $200.0 million = 10.0% * $2.0 billion (given as an assumption)
 Return on economic capital (EC) = $2.0 million = $200.0 EC * 1.0%
 Tax rate = 0.25 (given as assumption)

Such that RAROC = [($120.0 - 30.0 - 40.0 - 16.0 + 2.0)*(1.0 - 0.25 tax rate)] / 200.0 = 13.50%
Adjusted RAROC = RAROC - β(e)*[R(m) - Rf] = 13.50% - 1.60*[8.0% - 1.0%] = 2.30% and
2.30% is greater than the risk-free rate.

Alternatively, per Crouhy's Risk Management text, we can compare (RAROC - Rf)/β(e) to [R(m)
- Rf] for the same result; in this case, (13.50% - 1.%)/1.6 = 7.813% which is greater than the
7.0% ERP.

607.2. A. False. If the implied correlation is zero, then BU capital = SQRT(70^2 + 80^2) =
$106.3; BU risk capital of 120.0 implies correlation of about +0.2768.

In regard to (B), (C) and (D), each is TRUE.

Crouhy: "As [the] example shows, the choice of capital measure depends on the desired
objective. Fully diversified measures should be used for assessing the solvency of the firm and
minimum risk pricing. Active portfolio management or business mix decisions, on the other
hand, should be based on marginal [i.e., incremental] risk capital, taking into account the benefit
of full diversification. Finally, performance measurement should involve both perspectives:
stand-alone risk capital for incentive compensation, and fully diversified risk capital to assess
the extra performance generated by the diversification effects. However, we must be cautious
about how generous we are in attributing diversification benefits. Correlations between risk
factors drive the extent of the portfolio effect, and these correlations tend to vary over time.
During market crises, in particular, correlations sometimes shift dramatically toward either 1 or –
1, reducing or totally eliminating portfolio effects for a period of time. "

12
607.3. B. FALSE. The RAROC system should also incorporate the QUALITY of earnings,
in addition to the quantity of earning.

In regard to (A), (C) and (D), each is TRUE.

Crouhy's six recommendations for implementing a RAROC system are:

1. Senior management commitment
2. Communication and educatoin
3. Ongoing consultation
4. Maintaining the integrity of the process
5. Combine RAROC with qualitative factors, and
6. Put an active capital management process in place.
With respect to Recommendation #5 (Combine RAROC with qualitative factors) he writes,
"Earlier in this chapter, we described a simple decision rule for project selection and capital
attribution; i.e., accept projects where the RAROC is greater than the hurdle rate. In practice,
other qualitative factors should be taken into consideration. All the business units should be
assessed in the context of the two-dimensional strategic grid shown in Figure 17-3 (see below).
The horizontal axis of this figure corresponds to the RAROC return calculated on an ex ante
basis. The vertical axis is a qualitative assessment of the quality of the earnings produced by
the business units. This measure takes into consideration the strategic importance of the activity
for the firm, the growth potential of the business, the sustainability and volatility of the earnings
in the long run, and any synergies with other critical businesses in the firm. Priority in the
allocation of balance sheet resources should be given to the businesses in the upper right
quadrant. At the other extreme, the firm should try to exit, scale down, or fix the activities of
businesses that fall into the lower left quadrant. The businesses in the category Managed
Growth, in the lower right quadrant, are high-return activities that have low strategic importance
for the firm. In contrast, businesses in the category Investment, in the upper left quadrant, are
currently low-return activities that have high growth potential and high strategic value for the
firm."

Discuss in forum here: https://www.bionicturtle.com/forum/threads/p2-t7-607-adjusted-raroc-

and-raroc-in-practice-crouhy.9731/

13
Appendix (OPTIONAL)
This Appendix contains Practice Questions that were written for an earlier iteration of the FRM syllabus.
As such, they are decidedly optional and not necessary to your current preparation. They are included
merely as a supplemental resource in case you want to give additional practice to this topic. Why might
previous questions remain relevant? Because, in general, the FRM exam has a loose rather than tight
linkage to Learning Objectives (LOs); this has been argued as both a strength (i.e., less opportunity to
"game" the exam via rote memorization) or a weakness (i.e., less clarity on exactly what will be asked on
exam day) of the exam. GARP's approach is not necessarily identical to other certification exams whose
LOs might be a specific map to the actual questions. As the 2020 Learning Objectives document itself
advises, "The FRM is a comprehensive exam and you are expected to be familiar with a broad range of
risk management concepts and techniques. Key concepts appear in the Study Guide as bullet points at
the beginning of each section to help you identify the major themes and knowledge domains associated
with the readings listed under each section." Notice how the emphasis is on the broad range of concepts
and the broad knowledge points; in fact they are called "broad knowledge points." For this reason,
engaging with Practice Questions that we have relegated to the Appendix is certainly optional, but at the
same time is unlikely to be a waste of time if you want to drill down further on a particular topic.

Crouhy, Chapter 14: Capital Allocation and Performance

Measurement
P2.T7.1. RISK-ADJUSTED RETURN ON CAPITAL, RAROC
P2.T7.2. ALLOCATING ECONOMIC (RAROC) CAPITAL TO CREDIT, MARKET &
OPERATIONAL RISK
P2.T7.3. LOAN EQUIVALENCE AND ADJUSTED RAROC (ARAROC)

P2.T7.1. Risk-adjusted return on capital, RAROC

Learning Objectives: Describe the RAROC (risk-adjusted return on capital) methodology
and discuss some of the potential benefits of its use. Define, compare and contrast
economic and regulatory capital. Compute and interpret the RAROC for a loan or loan
portfolio, and use RAROC to compare business unit performance.

1.1. Consider the following statements about economic capital (EC):

I. Economic capital must equal or exceed regulatory capital

II. Economic capital should exceed equity capital
III. Economic capital should equal to the discounted stream of future excess return on
capital
IV. Economic capital absorbs expected losses (EL) and should equal the firm’s reserves
V. Economic capital absorbs unexpected losses (UL) up to a certain confidence level and
over a specified time horizon

Which of the above statements is (are) TRUE?

a) I. only
b) V. only
c) I., II. and III.
d) All are true

14
1.2. Each of the following is true about the PURPOSE of economic capital at a bank (EC)
EXCEPT:

a) Economic capital, if properly calibrated, guarantees the bank will never default
b) Economic capital is designed to keep a bank SOLVENT; i.e., EC is a solvency measure
c) Economic capital provides a means for the capital allocation within the bank
d) Economic capital is a metric (a risk currency) that can aggregate across and up/down
the organization

1.3. A loan (asset) portfolio with a principal of $1.0 billion pays an average annual rate of 5.0%
is funded by $1.0 billion in deposits (liabilities) that pay depositors an average interest rate of
2.0%; i.e., the gross spread is 3.0%. The bank's economic capital, which happens to equal its
equity capital, is 8.0% of assets; this $80 million in equity (economic capital) is invested in
Treasury bills that pay 1.0%. The expected losses (EL) on the loan portfolio is 1.0%. Finally, the
operating costs allocated to this loan are $6.0 million. What is the risk-adjusted return on capital,
RAROC? (please note that consistent with revised Crouhy we assume loan/assets are equal to
deposits/liabilities. We do NOT assume that deposits = assets minus economic capital.)

a) 8.67%
b) 11.13%
c) 18.50%
d) 20.50%

1.4. $10 billion in deposits paying 2.0% funds a $10.0 billion loan portfolio that earns 4.0% with
an expected loss (EL) of 1.0%. Economic capital (equity) equals 12.0% of the loan assets and
invested safely to earn 2.0%. Operating cost is $22.0 million. What is the risk-adjusted return on
capital, RAROC, for this loan?

a) 4.25%
b) 8.50%
c) 9.75%
d) 11.15%

1.5. Consider the following statements about risk-adjusted return on capital (RAROC):

I. If a project's RAROC is greater than the firm's cost of equity capital, then the project
should be accepted
II. As the denominator of RAROC is economic capital only, the numerator must deduct
expected or realized losses
III. RAROC can be used for capital allocation, on an ex ante basis
IV. RAROC can be used for performance evaluation, on an ex post basis

Which of the statements is (are) TRUE?

a) I. and III. only

b) II. and IV. only
c) III. and IV. only
d) All four

15
Answers:

1.1. B. V. only: Economic capital absorbs unexpected losses (UL) up to a certain

confidence level and over a specified time horizon. EC is a VaR-type measure of risk (a risk
currency).

Each of the other statements is FALSE.

1.2. A. EC cannot calibrate to 100% confidence, it is too costly (and shareholders do not
seek a risk-free investment with their equity)
 In regard to (B), (C), and (D), each is TRUE about EC.

1.3. C. 18.50%
Numerator:
Revenue = $1,000 loan * 5% = $50 million,
Cost of capital (COC) = $1,000 deposits * 2% = $20 million,
Return on EC (REC) = $1,000 * 8% * 1% = $0.8 million
EL = $1,000 * 1% = $10 million
OpEx = $6 million

Denominator = Economic capital = 8% * $1,000 = $80 million, such that:

RAROC = (Revenue - COC + REC - EL - OpEx)/EC = (50 - 20 + 0.8 - 10 - 6) / 80 = 18.50%

1.4. B. 8.50%
Numerator:
Revenue = $10,000 loan * 4% = $400 million,
Cost of capital (COC) = $10,000 deposits * 2% = $200 million,
Return on EC (REC) = $10,000 * 12% * 2% = $24 million
EL = $10,000 * 1% = $100 million
OpEx = $22 million

Denominator = Economic capital = 12% * $10,000 = $1.2 billion, such that:

RAROC = (Revenue - COC + REC - EL - OpEx)/EC = (400 - 200 + 24 - 100 - 22) / 1200 =
102/1200 = 8.5%

1.5. D. All four are true.

 In regard to I., note that this is true even for ARAROC.
 In regard to II., the denominator of RAROC is EC which covers unexpected losses (UL)
not expected losses (EL); for ratio consistency, then, EL should be deducted from the
numerator.

Discuss here in forum: https://www.bionicturtle.com/forum/threads/l2-t7-1-risk-adjusted-return-

on-capital-raroc.4674/

16
P2.T7.2. Allocating economic (RAROC) capital to credit, market &
operational risk
Learning objectives: Explain how capital is attributed to market, credit, and operational
risk. Calculate the capital charge for market risk and credit risk. Explain the difficulties
encountered in attributing economic capital to operational risk.

2.1. A bank has a daily 99% value at risk (VaR) limit of $1 million. Per Crouhy’s approach to
economic (not regulatory) capital, the bank employs three multipliers to determine the economic
capital charge for market risk: F(1) = 2.0 and multiplies VaR to account for the day-to-day event
risk not captured by the bank’s VaR model; F(2) = 0.15 and is a small charge for the unused
portion of the VaR limit; F(3) = 3.0 and is penalty multiplier for exceeding the VaR limit. If the
bank’s 99% VaR is realized to be $1.5 million (exceeding the limit), what is the economic capital
charge for market risk?

a) $1.5 million
b) $3.5 million
c) $4.5 million
d) $6.0 million

2.2. A bank has a daily 99% VaR limit of $1 million and employs the following multipliers to
allocate RAROC (economic) capital to market risk as a function of its VaR limit: F(1) VaR
multiplier = 1.5; F(2) multiplier for unused portion = 0.10; and F(3) penalty multiplier for excess
over the VaR limit = 5.0. If the bank’s 99% VaR is $800,000 (i.e., under the limit), what is market
risk charge to RAROC capital?

a) $20,000
b) $820,000
c) $1.22 million
d) $2.20 million

17
2.3. A bank maps external credit ratings to internal risk ratings (RR) as follows: A (S&P) = 3
(internal), BBB = 4, BB = 5. The bank attributes the economic (RAROC) capital charge for credit
risk by applying capital factors to the market value of positions. The capital factors are given by
a two-dimension matrix: tenor (maturity) and risk rating. The middle portion of the matrix looks
as follows. For internal rating = 3: 1.8%, 1.9% and 2.0% (tenors 4, 5, and 6). For internal rating
= 4, 2.3%, 2.5%, and 2.7% (tenors 4, 5, and 6). For internal rating = 5, 8.0%, 8.4%, and 8.8%
(tenors 4, 5, and 6).

Which of the following bonds has the highest economic capital charge for credit risk?

a) $10 million bond with 4-year maturity and rated BB

b) $10 million bond with 6-year maturity and rated BBB
c) $20 million bond with 5-year maturity and rated A
d) $20 million bond with 4-year maturity and rated BBB

2.4. According to Crouhy, each of the following is a difficulty encountered in attributing economic
capital to operational risk EXCEPT:

a) Operational risk is less developed because “credit and market risk are much bigger risks
to them”
b) There are few internal data points available to build operational loss distributions
c) We have to be careful not to double-count operational losses; i.e., operational losses
tend to defy mutually exclusive categories
d) Measurement of operational risk is least reliable at the catastrophic end (extreme tail) of
the operational risk spectrum

18
Answers:

2.1. C. $4.5 million

$1.5 MM VaR * 2.0 [F(1) multiplier] + 0 for no unused portion + $500,000 excess over limit * 3.0
[F(3) penalty multiplier] = $4.5 million

2.2. C. $1.22 million

$800,000 * 1.5 F(1) + $200,000 unused * 0.01 (F2) + 0 for excess = $1.22 million

2.3. A. $10 million bond with 4-year maturity and rated BB

$10 million bond with 4-year maturity and rated BB: 8% factor * $10 MM = $800,000
$10 million bond with 6-year maturity and rated BBB: 2.7% factor * $10 MM = $270,000
$20 million bond with 5-year maturity and rated A: 1.9% factor * $20 MM = $380,000
$20 million bond with 4-year maturity and rated BBB: 2.3% factor * 20 MM = $460,000

2.4. A. False: “Banks measure credit and market risk because they can, not because they
are the biggest risks that they face. Operational risk is larger, more dangerous, and no
one knows exactly what to do about it.”

In regard to (B), (C), and (D), each are true.

Crouhy (emphasis mine): “Operational risk measurement is very much a “work in progress.”
One approach is to define an operational value-at-risk as the loss that is likely to arise from an
operational failure over a certain time period, with a particular probability of occurrence. A
primary difficulty with estimating operational risk in this way is that there are very few internal
data points available to build the loss distribution. Nevertheless, one can look to external
information. Some of the information to build a loss database may be gathered from court
records (e.g., extensive information is available on retail and wholesale money transfer
operational losses). A database is relatively easy to develop for those losses that are incurred
regularly, such as credit-card fraud.

One must be careful not to double-count operational failures. For example, operational risk
may be reflected in loan losses (e.g., improperly assigned collateral), and may already be
captured in the RAROC credit capital model.

The measurement of operational risks is least reliable at the catastrophic end of the
operational risk spectrum—those risks that occur very rarely even at an industry level but
which might destroy the bank. Here, judgment is required. One can expect that advanced
analytic tools, over time, will be developed to analyze catastrophic operational risk capital more
satisfactorily.”

Discuss in forum here: http://www.bionicturtle.com/forum/threads/l2-t7-2-allocating-economic-

raroc-capital-to-credit-market-operational-risk.4679/

19
P2.T7.3. Loan equivalence and adjusted RAROC (ARAROC)
Learning objectives: Describe the Loan Equivalent Approach and use it to calculate
RAROC capital. Explain how the second-generation RAROC approaches improve
economic capital allocation decisions. Compute the adjusted RAROC for a project to
determine its viability.

3.1. Each of the following is true about the Loan Equivalent Approach, as it is used to calculate
RAROC/economic capital, EXCEPT for:

a) Bank guarantees and letters of credit tend to have 100% loan equivalent factors
b) The loan equivalent approach is generally not appropriate for derivatives because they
are bilateral contracts which can have negative market-to-market values
c) Undrawn credit commitments typically have a nonzero loan equivalent
d) The loan equivalent approach is compatible with an approach that estimates future credit
loss distributions as function of three variables: exposure, probability of default, and
recovery

3.2. Which counterparty credit exposure metric is typically used as a basis to determine the loan
equivalent amount, i.e., even if multiplier might be applied to this basis?

a) Current exposure (CE)

b) Expected exposure (EE); aka, expected credit exposure (ECE)
c) Potential future exposure (PFE); aka, worst credit exposure (WCE)
d) Average expected credit exposure (AECE); aka, expected positive exposure (EPE)

3.3. The “loan equivalent exposure” for a certain portfolio is $480 million with an expected loss
of 1.0% and a worst case expected loss of 3.0%. “Expected loss,” per Crouhy’s usage refers to
probability of default net of recoveries: EL = PD * LGD. “Worst case expected loss,” then, refers
to expected loss with 99% confidence. (For convenience, these are also exactly the numbers
Crouhy uses in his example) Consider the following statements:

I. The credit risk capital (economic capital attributed to credit risk) is $9.6 million
II. To determine the loan equivalent exposure (the credit exposure) of $480 million, this
approach uses potential future exposure (PFE or WCE) rather than average expected
credit exposure (AECE or EPE), because it employs a confidence level
III. This approach is considered quite precise (rather than approximate) because it
generates the full distribution of losses

Which of the above statements is (are) TRUE?

a) I. only
b) I. and III.
c) II. and III.
d) All three

20
3.4. According to Crouhy, which of the following best summarizes the advantage of adjusted
RAROC (ARAROC) over RAROC?

a) RAROC is computed based on book values, but ARAROC employs more useful market
values
b) RAROC is essentially a single-period metric, but ARAROC attempts to makes an
adjustment for stochastic changes to risk over time
c) Unlike RAROC which is not fully risk-adjusted, ARAROC is not sensitive to higher
returns earned due to higher risk by way of higher volatility and/or correlation
d) Unlike RAROC which can neither be compared to a hurdle rate “nor is compatible with
the loan equivalence approach,” ARAROC overcomes both limitations

3.5. The riskfree rate (Rf) is 2.0% and the expected rate of return on the market (Rm) is 8.0%. A
bank figures that the risk-adjusted return on capital (RAROC) of a risky project with a beta of
1.60 (i.e., beta of the project with respect to the market) is 11.0%. Is the project advisable if the
bank uses adjusted RAROC (ARAROC)?

a) Not advisable, ARAROC is negative at 0.600%

b) Not advisable, ARAROC is only 5.625%
c) Advisable, ARAROC of 5.625% is greater than 3.750%
d) Advisable, ARAROC is positive at 0.600%

21
Answers:

3.1. B. The loan equivalent approach is largely designed to overcome the problem of
measuring credit exposure in a derivative contract
 In regard to (A), (C) and (D), each is true.

3.2. D. Average expected credit exposure (AECE); aka, expected positive exposure (EPE)

Crouhy: “The crucial problem in developing an accurate loan equivalent measure of credit risk
for derivatives for the purpose of an RAROC analysis is to quantify properly the future credit risk
exposure. This is a complex problem because it is the outcome of multiple variables, including
the structure of the financial instrument that is under scrutiny, and changes in the values of the
underlying variables. The amount of money that one can reasonably expect to lose as a result
of default over a given period is normally called the ‘‘expected credit risk exposure.” The
expected credit exposure is an exposure at a particular point in time, while the “cumulative
average expected credit exposure” is an average of the expected credit exposures over a given
period of time. The average expected credit exposure is typically used as the loan equivalent
exposure for the purpose of RAROC analysis.

The maximum amount of money that could be lost as a result of default within a given
confidence interval is called the “worst case credit risk exposure” (sometimes called the
“maximum potential credit risk exposure”). The worst case credit risk exposure is an exposure at
a particular point in time, while a related measure, the average worst case credit risk exposure,
is an average of the worst case exposures over a given period of time. If one wishes to control
the amount that could be at risk to a given counterparty, then the worst case exposure is
particularly important in terms of allocating credit risk limits. One can use either the worst case
or average worst case credit risk exposure as a measure when setting limits to credit risk
exposure—one simply needs to be consistent. If one uses the worst case credit risk exposure to
measure credit risk, then limits should obviously be set in terms of worst case credit risk
exposures (in contrast to the average worst case credit risk exposures).”

3.3. A. I. only

The credit risk capital = economic capital = unexpected loss = $480 MM * (3% - 1%) = $96
million; 1% is the expected loss and is covered by a provision.

 In regard to II., this is false: AECE or EPE determines the exposure (analogous to
EAD), which is an INPUT then into the OpRisk VaR.
 In regard to III., this is false, the full distribution is not being simulated: “This loan
equivalent approach to calculating the average expected exposure is a proxy for more
sophisticated approaches; it has the virtue of facilitating comparison to a more
conventional loan product. Another approach would be to generate—using analytical,
empirical, or simulation techniques—the full distribution of losses, and then to select the
appropriate confidence interval percentile.”

22
3.4. C. Unlike RAROC which is not fully risk-adjusted, ARAROC is not sensitive to higher
returns earned due to higher risk by way of higher volatility and/or correlation

Crouhy: “Four points are worth noting. First, RAROC is sensitive to the level of the standard
deviation of the risky asset (Table 14.6). So RAROC may indicate that a project achieves the
required hurdle rate, given a high enough volatility (sA), even when the net present value of the
project is negative. Second, RAROC is sensitive to the correlation of the return on the
underlying asset and the market portfolio. Third, the ARAROC measure is insensitive to
changes in volatility and correlation. Fourth, if a fixed hurdle rate is used in conjunction with
RAROC, high-volatility and high-correlation projects will tend to be selected.”

3.5. B. Not advisable, ARAROC is only 5.625%, which is less than the excess market
return of 6.0%

The market’s excess return = 8% - 2% = 6.0%.

ARAROC = (11% - 2%)/1.6 = 5.625%, which is lower than 6% so the project is not accepted.
Another way to view this is simply like a Jensen’s alpha,
Compare project return to return implied by CAPM:
11% project versus 11.6% = 2% + 1.6*6%

Discuss in forum here: http://www.bionicturtle.com/forum/threads/l2-t7-3-loan-equivalence-

and-adjusted-raroc-araroc.4688/

On the difference between RAROC and ARAROC:

http://www.bionicturtle.com/forum/threads/difference-between-raroc-and-araroc.6669/

23
Crouhy, Chapter 15: Model Risk
P2.T7.604. Model error and model implementation risk
P2.T7.605. Mitigation of model risk

P2.T7.604. Model error and model implementation risk

Learning Objectives: Identify and explain errors in modeling assumptions that can
introduce model risk. Explain how model risk can arise in the implementation of a model.

604.1. According to Crouhy, Galai & Mark, JPMorgan's London Whale incident "showed that
model risk has no respect for the size or standing of an institution." According to the US
Senate's Subcommittee Report, despite portraying itself as an expert in risk management, the
bank's Chief Investment Office (CIO) which was charted with managing excess deposits "placed
a massive bet on a complex set of synthetic credit derivatives that, in 2012, lost at least $6.2
billion."

Which of the following is TRUE as risk management failure that contributed to the loss at
JPMorgan's CIO?

e) Soon after breaching the bank' and CIO's VaR limit, a new VaR model was adopted (and
approved) that reduced the SCP VaR by 50%, enabling the CIO to end its breach
f) The CIO switched from its historical practice of marking credit derivative positions at or
near the midpoint price in the daily range to assigning the favorable price within the daily
price range
g) SCP trades routinely breached the limits on all five key metrics used by CIO (ie, VaR,
CS01, CSW10%, stress loss, and stop loss), and the breaches were reported to
management, but the breaches were largely ignored
h) All of the above are true, according to the Senate Subcommittee: the loss was caused
by failures in operational risk, model risk, and corporate governance

24
604.2. Crouhy, Galai & Mark explain that the main cause of model risk are either (i) model error
or (ii) implementation. Model error is when "the model might contain mathematical errors or,
more likely, be based on simplifying assumptions that are misleading or inappropriate."
Implementation is when "the model might be implemented wrongly, either by accident or as part
of a deliberate fraud. "

Each of the following is a classic example of how model error can be introduced EXCEPT which
is the LEAST likely assumption, by itself, to create model error risk?

a) To assume an asset's distribution is stationary over time in order to maintain or improve

the tractability of the model
b) To assume a delta-neutral hedging strategy is risk-free and can be maintained because
active re-balancing is unrealistic
c) To assume asset returns, follow an empirical distribution simply because the historical
data happens to be easily available
d) To assume the forward rates--i.e., that are used to value fixed-income instruments--are
log normal although interest rates have shifted into a long-term regime of negative
territory

604.3. In regard to model implementation, Crouhy says "even if a model is correct and is being
used to tackle an appropriate problem, there remains the danger that it will be wrongly
implemented. With complicated models that require extensive programming, there is always a
chance that a programming “bug” may affect the output of the model. Some implementations
rely on numerical techniques that exhibit inherent approximation errors and limited ranges of
validity. Many programs that seem error-free have been tested only under normal conditions
and so may be error-prone in extreme cases and conditions. "

With respect to model implementation, which of the following is the BEST pieces of advice?

a) Ensure responsibility for data accuracy is clearly assigned

b) Remove outliers in all cases because outliers distort skew and kurtosis of the distribution
c) Volatility and correlation should be directly observed rather than forecast; if these two
inputs cannot be observed, seek an alternative approach
d) Seek the maximum length of the sampling period in order to improve the power of
statistical tests and reduce estimation errors

25
Answers:

604.1. D. All of the above are TRUE, according to the Senate Subcommittee: the loss was
caused by failures in operational risk, model risk, and corporate governance

In regard to true (A) which is model risk (emphasis ours): “… CIO traders, risk personnel,
and quantitative analysts frequently attacked the accuracy of the risk metrics, downplaying the
riskiness of credit derivatives and proposing risk measurement and model changes to lower risk
results for the SCP [Synthetic Credit Portfolio]. In the case of the CIO VaR, after analysts
concluded the existing model was too conservative and overstated risk, an alternative CIO
model was hurriedly adopted in late January 2012, while the CIO was in breach of its own and
the bankwide VaR limit. The bank did not obtain OCC approval as it should have to use the
model for the SCP. The CIO’s new model immediately lowered the SCP’s VaR by 50%,
enabling the CIO not only to end its breach, but to engage in substantially more risky derivatives
trading. Months later, the bank determined that the model was improperly implemented,
requiring error-prone manual data entry and incorporating formula and calculation errors. On
May 10, the bank backtracked, revoking the new VaR model due to its inaccuracy in portraying
risk, and reinstating the prior model.”

In regard to true (B) which is an operational risk failure (emphasis ours): "To minimize its
reported losses, the CIO began to deviate from the valuation practices it had used in the past to
price credit derivatives. In early January, the CIO had typically established the daily value of a
credit derivative by marking it at or near the midpoint price in the daily range of prices (bid-ask
spread) offered in the market place. Using midpoint prices had enabled the CIO to comply with
the requirement that it value its derivatives using prices that were the most representative of fair
value. But later in the first quarter of 2012, instead of marking near the midpoint, the CIO began
to assign more favorable prices within the daily price range (bid-ask spread) to its credit
derivatives. The more favorable prices enabled the CIO to report smaller losses in the daily
profit/ loss (P& L) reports that the SCP filed internally within the bank.”

In regard to true (C), which is a case of poor corporate governance: “In contrast to
JPMorgan Chase’s reputation for best-in-class risk management, the whale trades exposed a
bank culture in which risk limit breaches were routinely disregarded, risk metrics were frequently
criticized or downplayed, and risk evaluation models were targeted by bank personnel seeking
to produce artificially lower capital requirements. The CIO used five key metrics and limits to
gauge and control the risks associated with its trading activities, including the Value-at-Risk
(VaR) limit, Credit Spread Widening 01 (CS01) limit, Credit Spread Widening 10% (CSW10%)
limit, stress loss limits, and stop loss advisories. During the first three months of 2012, as the
CIO traders added billions of dollars in complex credit derivatives to the SCP, the SCP trades
breached the limits on all five risk metrics. In fact, from January 1 through April 30, 2012, CIO
risk limits and advisories were breached more than 330 times. … The SCP’s many breaches
were routinely reported to JPMorgan Chase and CIO management, risk personnel, and traders.
The breaches did not, however, spark an in-depth review of the SCP or require immediate
remedial actions to lower risk. Instead, the breaches were largely ignored or ended by raising
the relevant risk limit.”

26
604.2. C. FALSE: To assume asset returns follow an empirical distribution (simply
because the historical data happens to be easily available) is considered an advantage
over "theoretical" (ie, parametric) distributions.

In regard to (A), (B) and (D), each is cited as an example of model error that can create model
risk.

In regard to true (A), Crouhy et al say "the most frequent error in model building is to assume
that the distribution of the underlying asset is stationary (i.e., unchanging) when in fact it
changes over time. The case of volatility is particularly striking. "

In regard to true (B), Crouhy et al say "As a practical example, most derivative pricing models
are based on the assumption that a delta-neutral hedging strategy can be put in place for the
instruments in question— i.e., that the risk of holding a derivative can be continually offset by
holding the underlying asset in an appropriate proportion (hedge ratio). In practice, a delta-
neutral hedge of an option against its underlying asset is far from being completely risk-free,
and keeping such a position delta-neutral over time often requires a very active rebalancing
strategy. Banks rarely attempt the continuous rebalancing that pricing models assume. For one
thing, the theoretical strategy implies the execution of an enormous number of transactions, and
trading costs are too large for this to be feasible. Nor is continuous trading possible, even
disregarding transactions costs: markets close at night, on national holidays, and on weekends."

In regard to true (D), Crouhy et al say "A model can be found to be mathematically correct and
generally useful and yet be misapplied to a given situation. For example, some term structure
models that are widely used to value fixed-income instruments depend upon the assumption
that forward rates are log normal— that is, that their rates of change are normally distributed.
This model seems to perform relatively well when applied to most of the world’s markets— with
the exception of Japan for the last 10 years and the United States and Europe in the immediate
post-crisis years (because central banks implemented quantitative easing monetary policies and
flooded the markets with huge amount of liquidity). Post-crisis markets were characterized by
very low interest rates, and Japan sometimes exhibited negative rates; in these conditions,
different statistical tools (e.g., Gaussian and square root models) for interest rates work much
better. "

27
604.3. A. TRUE: Ensure responsibility for data accuracy is clearly assigned. It's easy to
underestimate the importance of data governance, but this seemingly mundane piece of
advice is important: if nobody has responsibility, then some of the other rules and
guidelines won't be nearly as relevant.

In regard to false (B), Crouhy et al say "All statistical estimators are subject to estimation errors
involving the inputs to the pricing model. A major problem in the estimation procedure is the
treatment of outliers, or extreme observations. Are the outliers really outliers, in the sense that
they do not reflect the true distribution? Or are they important observations that should not be
dismissed? The results of the estimation procedure will be vastly different depending on how
such observations are treated. Each bank, or even each trading desk within a bank, may use a
different estimation procedure to estimate the model parameters. "

In regard to false (C), Crouhy says "Volatilities and correlations are the hardest input
parameters to judge accurately. For example, an option’s strike price and maturity are fixed, and
asset prices and interest rates can easily be observed directly in the market— but volatilities
and correlations must be forecast ... Throughout the history of the derivatives markets, the fact
that model parameters such as volatility and correlation cannot be observed directly has given
rise to many opportunities for both genuine mistakes and deliberate tampering that can be
countered only through robust control procedures and independent vetting."

Finally, Crouhy also writes:

"The most frequent problems in estimating values, on the one hand, and assessing the potential
errors in valuation, on the other, are:
 Inaccurate data. Most financial institutions use internal data sources as well as external
databases. The responsibility for data accuracy is often not clearly assigned. It is
therefore very common to find data errors that can significantly affect the estimated
parameters.
 Inappropriate length of sampling period. Adding more observations improves the
power of statistical tests and tends to reduce the estimation errors. But the longer the
sampling period, the more weight is given to potentially stale and obsolete information.
Especially in dynamically changing financial markets, “old” data can become irrelevant
and may introduce noise into the estimation process.
 Problems with liquidity and the bid/ ask spread. In some markets, a robust market
price does not exist. The gap between the bid and ask prices may be large enough to
complicate the process of finding a single value. Choices made about the price data at
the time of data selection can have a major impact on the output of the model."

Discuss in forum here: https://www.bionicturtle.com/forum/threads/p2-t7-604-model-error-and-

model-implementation-risk-crouhy-galai-mark.9704/

28
P2.T7.605. Mitigation of model risk
Learning objectives: Explain methods and procedures risk managers can use to mitigate
model risk. Explain the impact of model risk and poor risk governance in the 2012
London Whale trading loss and the 1998 collapse of Long Term Capital Management
(LTCM).

605.1. In response to the question, "How Can We Mitigate Model Risk?" Crouhy, Galai & Mark
offer this advice: "One important way to mitigate model risk is to invest in research to improve
models and to develop better statistical tools, either internally at the bank or externally at a
university (or at an analytically oriented consulting organization). An even more vital way of
reducing model risk is to establish a process for the independent vetting of how models are both
selected and constructed. This should be complemented by independent oversight of the profit
and loss (P& L) calculation."

Each of the following is true about the phases of this model vetting EXCEPT which is false?

a) The vetting team should ask for full documentation of the model
b) The middle office must have independent access to an independent market risk
management financial rates database
c) The benchmark model should have the same implementation as the proposed model,
but it should have different parameters
d) The model must be stressed tested, including looking at limit scenario(s) in order to
identify the range of parameter values for which the model is accurate

605.2. According to Crouhy et al, "The failure of the hedge fund Long Term Capital
Management (LTCM) in September 1998 provides the classic example of model risk in the
financial industry. " Each of the following is true about the collapse of Long Term Capital
Management EXCEPT which is false?

a) Before the collapse, their relative-value and market-neutral strategies were viewed by
lenders as relatively safe ("low-risk") despite the high leverage (debt-to-equity of 25)
b) LTCM should have raised red flags due to blatant governance problems including a
board that lacked economic credentials and executives without seasoning and actual
fixed income trading experience
c) The "flight to quality" triggered by the Russian default abruptly decoupled the historical
patterns of spread convergence (between safe and risky bonds) and the previously
negative correlation between stock returns and interest rates
d) Weakness of LTCM's value at risk (VaR) model included: 10 days was too short a time
horizon; liquidity risk was not sufficiently modeled as a factor; and stress testing was
insufficient especially with respect to correlation and volatility risks

29
605.3. According to Crouhy, Galai & Mark, which of the following statements about model risk
is TRUE?

a) Finance requires the use of models and models necessarily imply model risk, therefore
model risk is inevitable
b) Model risk is unique because it is predominantly the manifestation of a technical issue:
given sufficient and qualified quantitative and programing expertise, model risk should
be virtually eliminated in a firm
c) The "Tinkerbell phenomenon" refers to a model which is too small in its capability, so to
speak; for example, the model can only accommodate a certain range of inputs so it
cannot realistically model the full range of financial transactions
d) Given identical confidence levels and time horizons for value at risk (VaR) and expected
shortfall (ES), research demonstrates that different third-party vendors and different
groups within an institution will generate "remarkably similar" VaR and ES estimates

30
Answers:

605.1. C. False: The benchmark model should have the same implementation as the
proposed model, but it should have different parameters. The opposite is the case: The
parameters should be the same, but the implementation benefits from being different.
See #4 below for details.

In regard to (A), (B) and (D), each is TRUE.

Crouhy et al: "The role of vetting is to offer assurance to the firm’s management that any model
for the valuation of a given security proposed by, say, a trading desk is reasonable. In other
words, it provides assurance that the model offers a reasonable representation of how the
market itself values the instrument, and that the model has been implemented correctly. Vetting
should consist of the following phases:

7. Documentation. The vetting team should ask for full documentation of the model,
including both the assumptions underlying the model and its mathematical expression.
This should be independent of any particular implementation, such as a spreadsheet, R
(a statistical programming language), or a C + + computer code, and should include: The
term sheet or, equivalently, a complete description of the transaction; A mathematical
statement of the model (which should include: an explicit statement of all the
components of the model; the payoff function and/ or any pricing algorithm for complex
structured deals; the calibration procedure for the model parameters; the hedge ratios/
sensitivities); Implementation features— i.e., inputs, outputs, and numerical methods
employed; A working version of the implementation
8. Soundness of model. An independent model vetter needs to verify that the
mathematical model is a reasonable representation of the instrument that is being
valued. For example, the manager might reasonably accept the use of a particular model
(e.g., the Black model) for valuing a short-term option on a long-maturity bond but reject
(without even looking at the computer code) the use of the same model to value a two-
year option on a three-year bond. At this stage, the risk manager should concentrate on
the finance aspects and not become overly focused on the mathematics.
9. Independent access to financial rates. The model vetter should check that the middle
office has independent access to an independent market risk management financial
rates database (to facilitate independent parameter estimation).
10. Benchmark modeling. The model vetter should develop a benchmark model based on
the assumptions that are being made and on the specifications of the deal. Here the
model vetter may use a different implementation from the implementation that is being
proposed. A proposed analytical model can be tested against a numerical approximation
technique or against a simulation approach. (For example, if the model to be vetted is
based on a “tree” implementation, one may instead rely on the partial differential
equation approach and use the finite-element technique to derive the numerical results.)
Compare the results of the benchmark test with those of the proposed model.

31
 11. Health check and stress test the model. Also, make sure that the model possesses
the basic properties that all derivatives models should possess, such as put/ call parity
and other nonarbitrage conditions. Finally, the vetter should stress test the model. The
model can be stress tested by looking at some limit scenario in order to identify the
range of parameter values for which the model provides accurate pricing. This is
especially important for implementations that rely on numerical techniques.
12. Build a formal treatment of model risk into the overall risk management
procedures, and periodically reevaluate models. Also, re-estimate parameters using
best-practice statistical procedures. Experience shows that simple but robust models
tend to work better than more ambitious but fragile models. It is essential to monitor and
control model performance over time."

605.2. B. FALSE: LTCM wash absolutely awash in perceived expertise and reputational
capital. The board included Myron Scholes and Robert Merton (Nobel Prize winners for the
Black-Scholes model!) and LTCM was founded by John Meriwether who was previously the
head of Solomon Brothers' bond arbitrage desk. Note also that in the three years prior to the
collapse the fund returned, respectively, 21%, 43% and 41%.

In regard to (A), (C) and (D), each is TRUE.

In summary, Crouhy et al write "LTCM failed because both its trading models and its risk
management models failed to anticipate the vicious circle of losses during an extreme crisis
when volatilities rose dramatically, correlations between markets and instruments became
closer to 1, and liquidity dried up. "
 In regard to true (A), "LTCM, like other hedge funds in early 1998, had positioned its
portfolios on the basis of particular bets, albeit bets that seemed pretty safe at first sight.
For example, LTCM bet that the spreads between corporate bonds and government
Treasuries in different countries, such as the United States and the United Kingdom,
were too large and would eventually return to their normal range (as they had always
done before). Such strategies are based on intensive empirical research and advanced
financial modeling. A trade to capture the relative-value opportunities uncovered by such
modeling might consist of buying corporate bonds and selling the relevant government
bonds short. Other positions involved betting on convergence in the key European bond
markets by selling German government bonds against the sovereign debt of other
countries, such as Spain and Italy, which were due to sign up for European economic
and monetary union (EMU). When the spread in yields narrows, such positions make
money whether the price level goes up or down. The return on such apparently low-risk
strategies tends to be quite small, and it becomes smaller and smaller as more players
enter the market to take advantage of the opportunity. As a result, hedge funds are
obliged to use leverage aggressively to boost their absolute performance. LTCM, for
example, was trying to earn a 1 percent return on its assets, leveraged 25 times, which
would yield a 25 percent return. LTCM was able to obtain huge loans, collateralized by
the bonds that it had invested in, because its strategy was widely viewed as safe by the
institutions that were its lenders. "

32
  In regard to true (C), "Most of the losses incurred by LTCM were the consequence of
the breakdown of the correlation and volatility patterns that had been observed in the
past. Several mechanisms came into play during this market turmoil as a consequence
of the flight to quality and the disappearance of liquidity: 1. Interest rates on Treasuries
and stock prices fell in tandem, because investors deserted the stock market and
purchased U.S. government bonds in a flight to quality. In normal markets, stock returns
and interest rates are negatively correlated— i.e., when interest rates fall, stock prices
rise. 2. When liquidity dries up in many markets simultaneously, it becomes impossible
to unwind positions. Portfolios that seem to be well diversified across markets start to
behave as if they were highly concentrated in a single market, and market-neutral
positions become directionally exposed (usually to the wrong side of the market). "
 In regard to true (D), "Risk control at LTCM relied on a VaR model ... According to
LTCM, the fund was structured so that the risk of investing in it should have been no
greater than that of investing in the S& P 500 ... However, some assumptions that are
usual in regulatory VaR calculations are not realistic for a hedge fund: 1. The time
horizon for economic capital should be the time it takes to raise new capital or the period
of time over which a crisis scenario will unfold. Based on the experience of LTCM, 10
days is clearly too short a time horizon for the derivation of hedge fund VaR. 2. Liquidity
risk is not factored into traditional static VaR models. VaR models assume that normal
market conditions prevail and that these exhibit perfect liquidity. 3. Correlation and
volatility risks can be captured only through stress testing. This was probably the
weakest point of LTCM’s VaR system. "

605.3. A. TRUE: Finance requires the use of models and models necessarily imply model
risk, therefore model risk is inevitable

Crouhy et al: "Models are an inevitable feature of modern finance, and model risk is inherent in
the use of models. Although our examples in this chapter have focused on market risk, many of
the principles we have discussed can be applied to make model risk transparent in other
spheres, including credit risk and asset/ liability management. "
 In regard to false (B) and false (C), Crouhy et al say in their Conclusion, "In this
chapter, we’ve stressed the technical elements of model risk, but we should also be
wary of the human factor in model risk losses. Large trading profits tend to lead to large
bonuses for senior managers, and this creates an incentive for these managers to
believe the traders who are reporting the profits (rather than the risk managers or other
critics who might be questioning the reported profits). Often, traders use their expertise
in formal pricing models to confound any internal critics, or they may claim to have some
sort of informal but profound insight into how markets behave. The psychology of this
behavior is such that we are tempted to call it the “Tinkerbell” phenomenon, after the
scene in Peter Pan in which the children in the audience shout 'I believe, I believe' in
order to revive the poisoned fairy Tinkerbell (see Box 15-6). The antidote is for senior
managers to approach any model that seems to record or deliver above-market returns
with a healthy skepticism, to insist that models be made transparent, and to make sure
that all models are independently vetted. "

33
  In regard to false (D), Crouhy writes: "How Widespread a Problem Is Model Risk? The
short answer is that in a modern financial system, model risk is everywhere— a fact that
has been recognized for some decades. Back in 1997, the Bank of England conducted a
survey highlighting the variation in models that existed among 40 major derivative
trading firms based in London. Vanilla foreign exchange instruments showed a relatively
low level of variation in both value and sensitivities, but some exotic derivatives
displayed large variations not only in value but also in sensitivity measures: 10 to 20
percent for swaptions and up to 60 percent for exotic foreign exchange instruments.
Another study in the same year showed that the different models available to calculate
VaR sometimes gave very different answers when applied to the same portfolio.
[Footnote 2] (The authors of this book know from experience that different groups within
the same financial institution can come up with significantly different valuations for
similar instruments.) "
Footnote 2: "Researchers presented an identical asset portfolio to a number of commercial
vendors of software for value-at-risk (VaR) calculations. Each was asked to use the same
volatility inputs, obtained from JP Morgan’s Risk Metrics, and to report the aggregate VaR for
the entire portfolio and the VaR figure for each type of instrument (such as swaps, caps and
floors, and swaptions). The variation among vendors was striking, given that in this case they
were supposed to be analyzing the same position (in relatively simple instruments), using the
same methodology, and using the same market parameter values. For the whole portfolio, the
estimates ranged from $ 3.8 million to $ 6.1 million, and for the portion containing options, the
VaR estimates varied from $ 747,000 to $ 2,100,000. "

Crouhy, Michel; Galai, Dan; Mark, Robert. The Essentials of Risk Management, Second Edition
(Kindle Locations 8884-8889). McGraw-Hill Education. Kindle Edition.

Discuss here in forum: https://www.bionicturtle.com/forum/threads/p2-t7-605-mitigation-of-

model-risk-crouhy.9711/

34