COVERS
EVERY LEARNING
OBJECTIVE ON
THE EXAM
FRM
EXAM REVIEW
STUDY GUIDE:
RT
BITE-SIZED
LESSON
FORMAT
WILEY
Wiley FRM Exam Review Study Guide 2017
Fart II
Wiley FRM Exam Review Study Guide 2017
Part II
Market Risk Measurement and Management,
Credit Risk Measurement and Management,
Operational and Integrated Risk Management,
Risk Management and Investment Management,
Current Issues in Financial Markets
Christian H. Cooper, CFA, FRM
W il
ey
Cover image: Loewy Design
Cover design: Loewy Design
Copyright © 2017 by John Wiley & Sons, Inc. All rights reserved.
Published by John Wiley & Sons, Inc., Hoboken, New Jersey.
Published simultaneously in Canada.
No part of this publication may be reproduced, stored in a retrieval system, or transmitted in any
form or by any means, electronic, mechanical, photocopying, recording, scanning, or otherwise,
except as permitted under Section 107 or 108 of the 1976 United States Copyright Act, without
either the prior written permission of the Publisher, or authorization through payment of the
appropriate per-copy fee to the Copyright Clearance Center, Inc., 222 Rosewood Drive, Danvers,
MA 01923, (978) 750-8400, fax (978) 646-8600, or on the Web at www.copyright.com. Requests
to the Publisher for permission should be addressed to the Permissions Department, John Wiley &
Sons, Inc., I l l River Street, Hoboken, NJ 07030, (201) 748-6011, fax (201) 748-6008, or online
at />Limit of Liability/Disclaimer of Warranty: While the publisher and author have used their best
efforts in preparing this book, they make no representations or warranties with respect to the
accuracy or completeness of the contents of this book and specifically disclaim any implied
warranties of merchantability or fitness for a particular purpose. No warranty may be created or
extended by sales representatives or written sales materials. The advice and strategies contained
herein may not be suitable for your situation. You should consult with a professional where
appropriate. Neither the publisher nor author shall be liable for any loss of profit or any other
commercial damages, including but not limited to special, incidental, consequential, or other
damages.
For general information on our other products and services or for technical support, please contact
our Customer Care Department within the United States at (800) 762-2974, outside the United
States at (317) 572-3993 or fax (317) 572-4002.
Wiley publishes in a variety of print and electronic formats and by print-on-demand. Some
material included with standard print versions of this book may not be included in e-books or in
print-on-demand. If this book refers to media such as a CD or DVD that is not included in the
version you purchased, you may download this material at . For
more information about Wiley products, visit www.wiley.com.
Library o f Congress Cataloging-in-Publication Data:
ISBN 978-1-119-38577-6
Printed in the United States of America
10 9 8 7 6 5 4 3 2 1
Contents
How to Study for the Exam
xi
About the Author
xii
Market Risk Measurement and Management
Lesson: Kevin Dowd, Measuring Market Risk, 2nd Edition (West Sussex, England:
John Wiley & Sons, 2005). Chapter 3. Estimating Market Risk Measures:
An Introduction and Overview
3
Lesson: Kevin Dowd, Measuring Market Risk, 2nd Edition (West Sussex, England:
John Wiley & Sons, 2005). Chapter 4. Non-Parametric Approaches
11
Lesson: Philippe Jorion, Value-at-Risk: The New Benchmark for Managing Financial Risk,
3rd Edition (New York: McGraw-Hill, 2007). Chapter 6. Backtesting VaR
15
Lesson: Philippe Jorion, Vaiue-at-Risk: The New Benchmark for Managing Financial Risk,
3rd Edition (New York: McGraw-Hill, 2007). Chapter 11. VaR Mapping
19
Lesson: "Messages from the Academic Literature on Risk Measurement for the Trading Book,"
Basel Committee on Banking Supervision, Working Paper No. 19, January 2011.
25
Lesson: Gunter Meissner, Correlation Risk Modeling and Management
(Hoboken, NJ: John Wiley & Sons, 2014). Chapter 1. Some Correlation Basics:
Properties, Motivation, Terminology
31
Lesson: Gunter Meissner, Correlation Risk Modeling and Management (Hoboken, NJ:
John Wiley & Sons, 2014). Chapter 2. Empirical Properties of Correlation:
How Do Correlations Behave in the Real World?
37
Lesson: Gunter Meissner, Correlation Risk Modeling and Management (New York:
John Wiley & Sons, 2014). Chapter 3. Statistical Correlation Models— Can We Apply
Them to Finance?
41
Lesson: Gunter Meissner, Correlation Risk Modeling and Management (Hoboken, NJ:
John Wiley & Sons, 2014). Chapter 4. Financial Correlation Modeling— Bottom-Up Approaches
(Sections 4.3.0 (intro), 4.3.1, and 4.3.2 only)
43
©2017 Wiley
CONTENTS
Lesson: M c e T u c k m n , Fixed Income Securities, M Edition (Hoboken, NJ: John Wiley & Sons, 2011).
Chapter 6. Empirical Approaches to Risk Metrics and Hedges
49
Lesson: Bruce Tuckman, Fixed Income Securities, 3rd Edition (Hoboken, NJ: John Wiley & Sons, 2011).
Chapter 7. The Science of Term Structure Models
53
Lesson: BruceTuckman, Fixed Income Securities, 3rd Edition (Hoboken, NJ: John Wiley & Sons, 2011).
Chapter 8. The Evolution of Short Rates and the Shape of the Term Structure
63
Lesson: BruceTuckman, Fixed Income Securities, 3rd Edition (Hoboken, NJ: John Wiley & Sons, 2011).
Chapter 9. The Art of Term Structure Models: Drift
67
Lesson: BruceTuckman, Fixed Income Securities, 3rd Edition (Hoboken, NJ: John Wiley & Sons, 2011).
Chapter 10. The Art of Term Structure Models: Volatility and Distribution
77
Lesson: John Hull, Options, Futures, and Other Derivatives, 9th Edition (New York:
Pearson Prentice Hall, 2014). Chapter 9. OIS Discounting, Credit Issues, and Funding Costs
81
Lesson: John Hull, Options, Futures, and Other Derivatives, 9th Edition (New York:
Pearson Prentice Hall, 2014). Chapter 20. Volatility Smiles
85
Credit Risk Measurement and Management
Vi
Lesson: Jonathan Golin and Philippe Delhaise, The Bank Credit Analysis Handbook
(Hoboken, NJ: John Wiley & Sons, 2013). Chapter 1. The Credit Decision
91
Lesson: Jonathan Golin and Philippe Delhaise, The Bank Credit Analysis Handbook
(Hoboken, NJ: John Wiley & Sons, 2013). Chapter 2. The Credit Analyst
95
Lesson: Giacomo De Laurentis, Renato Maino, and Luca Molteni, Developing,
Validating and Using Internal Ratings (West Sussex, United Kingdom: John Wiley & Sons, 2010).
Chapter 2. Classifications and Key Concepts of Credit Risk
97
Lesson: Giacomo De Laurentis, Renato Maino, and Luca Molteni, Developing,
Validating and Using Internal Ratings (West Sussex, United Kingdom: John Wiley & Sons, 2010).
Chapter 3. Ratings Assignment Methodologies
105
Lesson: Rene Stulz, Risk Management & Derivatives (Florence, KY: Thomson South-Western, 2002).
Chapter 18. Credit Risks and Credit Derivatives
119
Lesson: Allan Malz, Financial Risk Management: Models, History, and Institutions
(Hoboken, NJ: John Wiley & Sons, 2011). Chapter 7. Spread Risk and Default Intensity Models
127
Lesson: Allan Malz, Financial Risk Management: Models, History, and Institutions (Hoboken, NJ:
John Wiley & Sons, 2011). Chapter 8. Portfolio Credit Risk (Sections 8.1,8.2,8.3 only)
135
Lesson: Allan Malz, Financial Risk Management: Models, History, and Institutions (Hoboken, NJ:
John Wiley & Sons, 2011). Chapter 9. Structured Credit Risk
139
©2017 Wiley
Lesson: Jon Gregory, Counterparty Credit Risk and Credit Value Adjustment: A Continuing Challenge
for Global Financial Markets, 2nd Edition (West Sussex, UK: John Wiley & Sons, 2012).
Chapter 3. Defining Counterparty Credit Risk
147
Lesson: Jon Gregory, Counterparty Credit Risk and Credit Value Adjustment: A Continuing Challenge
for Global Financial Markets, 2nd Edition (West Sussex, UK: John Wiley & Sons, 2012).
Chapter 4. Netting, Compression, Resets, and Termination Features
151
Lesson: Jon Gregory, Counterparty Credit Risk and Credit Value Adjustment: A Continuing
Challenge for Global Financial Markets, 2nd Edition (West Sussex, UK: John Wiley & Sons, 2012).
Chapter 5. Collateral
153
Lesson: Jon Gregory, Counterparty Credit Risk and Credit Value Adjustment: A Continuing
Challenge for Global Financial Markets, 2nd Edition (West Sussex, UK: John Wiley & Sons, 2012).
Chapter 7. Central Counterparties
159
Lesson: Jon Gregory, Counterparty Credit Risk and Credit Value Adjustment: A Continuing
Challenge for Global Financial Markets, 2nd Edition (West Sussex, UK: John Wiley & Sons, 2012).
Chapter 8. Credit Exposure
165
Lesson: Jon Gregory, Counterparty Credit Risk and Credit Value Adjustment: A Continuing
Challenge for Global Financial Markets, 2nd Edition (West Sussex, UK: John Wiley & Sons, 2012).
Chapter 10. Default Probability, Credit Spreads, and Credit Derivatives
171
Lesson: Jon Gregory, Counterparty Credit Risk and Credit Value Adjustment: A Continuing
Challenge for Global Financial Markets, 2nd Edition (West Sussex, UK: John Wiley & Sons, 2012).
Chapter 12. Credit Value Adjustment
177
Lesson: Jon Gregory, Counterparty Credit Risk and Credit Value Adjustment: A Continuing
Challenge for Global Financial Markets, 2nd Edition (West Sussex, UK: John Wiley & Sons, 2012).
Chapter 15. Wrong-Way Risk
181
Stress Testing: Approaches, Methods, and Applications, Edited by Akhtar Siddique and
Iftekhar Hasan (London: Risk Books, 2013). Chapter 4. The Evolution of Stress Testing
Counterparty Exposures
183
Lesson: Michel Crouhy, Dan Galai, and Robert Mark, The Essentials o f Risk Management, 2nd Edition
(New York: McGraw-Hill, 2014). Chapter 9. Credit Scoring and Retail Credit Risk Management
191
Lesson: Michel Crouhy, Dan Galai, and Robert Mark, The Essentials o f Risk Management, 2nd Edition
(New York: McGraw-Hill, 2014). Chapter 12. The Credit Transfer Markets— and Their Implications
197
Lesson: Moorad Choudhry, Structured Credit Products: Credit Derivatives & Synthetic Securitization,
2nd Edition (Hoboken, NJ: John Wiley & Sons, 2010). Chapter 12. An Introduction to Securitization
205
Lesson: Adam Ashcraft and Til Schuermann/'Understanding the Securitization of Subprime
Mortgage Credit," Federal Reserve Bank of New York Staff Reports, No. 318 (March 2008)
215
©2017 Wiley
CONTENTS
Operational and Integrated Risk Management
Lesson: "Principles for the Sound Management of Operational Risk" (Basel Committee on
Banking Supervision Publication, June 2011)
221
Lesson: Brian Noccoand Rene Stulz,"Enterprise Risk Management: Theory and Practice,
Journal o f Applied Corporate Finance 18, no. 4 (2006): 8 -2 0
229
Lesson: "Observations on Developments in Risk Appetite Frameworks and IT Infrastructure,"
Senior Supervisors Group, December 2010
233
Lesson: Anthony Tarantino and Deborah Cernauskas, Risk Management in Finance: Six Sigma
and Other Next Generation Techniques (Hoboken, NJ: John Wiley & Sons, 2009).
Chapter 3. Information Risk and Data Quality Management
239
Lesson: Marcelo G. Cruz, Gareth W. Peters, and Pavel V. Shevchenko, Fundamental Aspects
o f Operational Risk and Insurance Analytics: A Handbook o f Operational Risk (Hoboken,
NJ: John Wiley & Sons, 2015). Chapter 2. OpRisk Data and Governance
241
Lesson: Philippa X. Girling, Operational Risk Management: A Complete Guide to a
Successful Operational Risk Framework (Hoboken, NJ: John Wiley & Sons, 2013). Chapter 8.
External Loss Data
247
Lesson: Philippa X. Girling, Operational Risk Management: A Complete Guide to a
Successful Operational Risk Framework (Hoboken, NJ: John Wiley & Sons, 2013).
Chapter 12. Capital Modeling
251
Lesson: "Standardised Measurement Approach for Operational Risk— Consultative
Document" (Basel Committee on Banking Supervision Publication, March 2016).
255
Lesson: Kevin Dowd, Measuring Market Risk, 2nd Edition (West Sussex, UK: John
Wiley & Sons, 2005). Chapter 7. Parametric Approaches (II): Extreme Value
259
Lesson: Giacomo De Laurentis, Renato Maino, and Luca Molteni, Developing,
Validating, and Using Internal Ratings (Hoboken, NJ: John Wiley & Sons, 2010.
Chapter 5. Validating Rating Models
263
Lesson: Michel Crouhy, Dan Galai, and Robert Mark, The Essentials o f Risk Management,
2nd Edition (New York: McGraw-Hill, 2014). Chapter 15. Model Risk
267
Lesson: Michel Crouhy, Dan Galai, and Robert Mark, The Essentials o f Risk Management,
2nd Edition (New York: McGraw-Hill, 2014). Chapter 17. Risk Capital Attribution and
Risk-Adjusted Performance Measurement
271
Lesson: "Range of Practices and Issues in Economic Capital Frameworks" (Basel
Committee on Banking Supervision Publication, March 2009)
275
Lesson: "Capital Planning at Large Bank Holding Companies: Supervisory Expectations and
Range of Current Practice," Board of Governors of the Federal Reserve System, August 2013
285
©2017 Wiley
CONTENTS
Lesson: Bruce Tuckman and Angel Serrat, Fixed Income Securities: Tools for Today's
Markets, 3rd Edition (Hoboken, NJ: John Wiley & Sons, 2011). Chapter 12. Repurchase
Agreements and Financing
291
Lesson: Kevin Dowd, Measuring Market Risk, 2nd Edition (West Sussex,
UK: John Wiley & Sons, 2005). Chapter 14. Estimating Liquidity Risks
299
Lesson: Allan Malz, Financial Risk Management: Models, History, and Institutions
(Hoboken, NJ: John Wiley & Sons, 2011). Chapter 11. Section 11.1, Assessing the
Quality of Risk Measures
303
Lesson: Allan Malz, Financial Risk Management: Models, History, and Institutions
(Hoboken, NJ: John Wiley & Sons, 2011). Chapter 12 Liquidity and Leverage
307
Lesson: Darrell Duffie,"The Failure Mechanics of Dealer Banks,"
Journal o f Economic Perspectives 24, no. 1 (2010): 51-72
317
Lesson: Til Schuermann, "Stress Testing Banks "prepared for the Committee on
Capital Market Regulation, Wharton Financial Institutions Center, April 2012
321
Lesson: "Guidance on Managing Outsourcing Risk," Board of Governors of the
Federal Reserve System, December 2013
325
Lesson: John Hull, Risk Management and Financial Institutions, 4th Edition
(Hoboken, NJ: John Wiley & Sons, 2015). Chapter 15. Basel I, Basel II, and Solvency II
329
Lesson: John Hull, Risk Management and Financial Institutions, 4th Edition
(Hoboken, NJ: John Wiley & Sons, 2015). Chapter 16. Basel II.5, Basel III,
and Other Post-Crisis Changes
339
Lesson: John Hull, Risk Management and Financial Institutions, 4th Edition
(Hoboken, NJ: John Wiley & Sons, 2015). Chapter 17. Fundamental Review of the Trading Book
345
Regulatory Readings
349
Risk Management and Investment Management
Lesson: Andrew Ang, Asset Management: A Systematic Approach to Factor Investing
(New York: Oxford University Press, 2014). Chapter 6. Factor Theory
353
Lesson: Andrew Ang, Asset Management: A Systematic Approach to Factor Investing
(New York: Oxford University Press, 2014). Chapter 7. Factors
363
Lesson: Andrew Ang, Asset Management: A Systematic Approach to Factor Investing
(New York: Oxford University Press, 2014). Chapter 10. Alpha (and the Low-Risk Anomaly)
369
Lesson: Andrew Ang, Asset Management: A Systematic Approach to Factor Investing
(New York: Oxford University Press, 2014). Chapter 13. Illiquid Assets (excluding
section 13.5, Portfolio Choice with Illiquid Assets)
379
©2017 Wiley
©
CONTENTS
Lesson: Richard Grinold and Ronald Kahn, Active Portfolio Management: A
Quantitative Approach for Producing Superior Returns and Controlling Risk, 2nd Edition
(New York: McGraw-Hill, 2000). Chapter 14. Portfolio Construction
383
Lesson: Philippe Jorion, Value a t Risk: The New Benchmark for Managing Financial Risk,
3rd Edition (New York: McGraw-Hill, 2007). Chapter 7. Portfolio Risk: Analytical Methods
389
Lesson: Philippe Jorion, Value a t Risk: The New Benchmark for Managing Financial Risk,
3rd Edition (New York: McGraw-Hill, 2007). Chapter 17. VaR and Risk Budgeting in
Investment Management
395
Lesson: Robert Litterman and the Quantitative Resources Group, Modern Investment
Management: An Eguilibrium Approach (Hoboken, NJ: John Wiley & Sons, 2003).
Chapter 17. Risk Monitoring and Performance Measurement
399
Lesson: Zvi Bodie, Alex Kane, and Alan J. Marcus, Investments, 10th Edition
(New York: McGraw-Hill, 2013). Chapter 24. Portfolio Performance Evaluation
405
Lesson: G. Constantinides, M. Harris, and R. Stulz, eds., Handbook o f the Economics o f
Finance, Volume 2B (Oxford, UK: Elsevier, 2013). Chapter 17. Hedge Funds, by William
Fung and David Hsieh
415
Lesson: Kevin R. Mirabile, Hedge Fund Investing: A Practical Approach to Understanding
Investor Motivation, Manager Profits, and Fund Performance (Hoboken, NJ: Wiley
Finance, 2013). Chapter 11. Performing Due Diligence on Specific Managers and Funds
421
Current Issues in Financial Markets
Lesson: Rainer Bohme, Nicolas Christin, Benjamin Edelman, and Tyler Moor, "Bitcoin:
Economics, Technology, and Governance!'Journal o f Economic Perspectives,
Vol. 29, No. 2, Spring 2015
427
Lesson: Dudley, William C., "Market and Funding Liquidity— An Overview," Remarks at the
Federal Reserve Bank of Atlanta 2016 Financial Markets Conference, Fernandina Beach,
Florida, May 1,2016
431
Lesson: "Chapter 2: Market Liquidity - Resilient or Fleeting?" International Monetary
Fund, Global Financial Stability Report, October 2015
433
Lesson: "Algorithmic Trading Briefing Note," Federal Reserve Bank of New York, April 2015
437
Lesson: Morten Bech, Anamaria Hies, Ulf Lewrick, and Andreas Schrimpf,"Hanging Up
the Phone— Electronic Trading in Fixed Income Markets and Its Implications,''BIS Quarterly
Review, 2016
441
Lesson: Morten Linnemann Bech and Aytek Malkhozov,"How Have Central Banks
Implemented Negative Policy Rates?"/?/5 Quarterly Review, March 6,2016
445
Lesson: "Corporate Debt in Emerging Economies: A Threat to Financial Stability?"Committee
on International Economic Policy and Reform, Brookings Institution, September 2015
449
©2017 Wiley
How to Study for the Exam
The FRM Exam Part II curriculum covers the tools used to assess financial risk:
•
•
•
•
•
Market risk measurement and management—25 %
Credit risk measurement and management—25%
Operational and integrated risk management—25 %
Risk management and investment management— 15%
Current issues in financial risk management— 10%
It is important to focus only on the learning objectives as you are asked about them and
pay close attention to the percentages of each section. That is the core of my focus through
the text, the online lecture sessions, and with the practice questions. A study hour doesn’t
count unless you are laser focused on specifically how GARP asks a learning objective.
Consistency is also key. Making a regular weekly study time is going to be important
to staying on track. There is a reason only -50% of the candidates pass the exam every
year. It’s a tough exam. It also tests intuition, not just memorization. That is why I
attempt at every opportunity to connect the dots across readings and teach how changing
environments change both markets and the models we use to model them, as well as
helping you with the questions GARP specifically wants you to calculate an outcome for.
Calculator policy:
It is best to begin your study with one of the approved calculators. You will not be admitted
to the exam without one of these approved calculators!
•
•
•
•
•
Hewlett Packard 12C (including the HP 12C Platinum and the Anniversary Edition)
Hewlett Packard 10B II
Hewlett Packard 10B 11+
Hewlett Packard 20B
Texas Instruments BA II Plus (including the BA II Plus Professional)
Every year, candidates are turned away from the exam site because of wrong calculators.
Make sure you aren’t one of them.
©2017 Wiley
xi
ABOUT THE AUTHOR
Christian H. Cooper is an author and trader based in New York City. He initially created
the FRM program because, as a candidate, he was frustrated with the quality of study
programs available. Writing from a practitioner’s point of view, Christian drew on his
experience as a trader across fixed income and equity markets, most recently as head of
derivatives trading at a bank in New York, to create a program that is very focused on exam
results while connecting the dots across topics to increase intuition and understanding.
Christian is active with the Aspen Institute; he is a Truman National Security Fellow, and a
term member at the Council on Foreign Relations.
xii
©2017 Wiley
M a r k e t Ri s k M e a s u r e m e n t
M a n a g e m e n t (MR)
a nd
The broad areas of knowledge covered in readings related to Market Risk Measurement
and Management include the following:
•
•
•
•
•
VaR and other risk measures:
o Parametric and nonparametric methods of estimation
o VaR mapping
o Backtesting VaR
o Expected shortfall (ES) and other coherent risk measures
o Extreme value theory (EVT)
Modeling dependence: correlations and copulas
Term structure models of interest rates
Discount rate selection
Volatility: smiles and term structures
©2017 Wiley
©
Do w d , Ch
a pt e r
3
Kevin Dowd, Measuring Market Risk, 2nd Edition (West Sussex, UK:
John Wiley & Sons, 2005). Chapter 3. Estimating Market Risk Measures:
An Introduction and Overview
After completing this reading you should be able to:
•
•
•
•
•
•
•
Estimate VaR using a historical simulation approach.
Estimate VaR using a parametric approach for both normal and lognormal return
distributions.
Estimate the expected shortfall given P/L or return data.
Define coherent risk measures.
Estimate risk measures by estimating quantiles.
Evaluate estimators of risk measures by estimating their standard errors.
Interpret QQ plots to identify the characteristics of a distribution.
Reading notes: This is a very dense reading about 20 pages long with a lot of VaR
“estimate” questions. This is probably one of the most important readings in the Market
Risk Measurement and Management section of the exam. Make sure you can calculate
these answers and be able to qualitatively discuss these topics. However, don’t get lost
in the technical material that asks you to “describe.” The assigned readings go into deep
detail that isn’t required for the exam. Make sure you are paying attention to what you are
actually asked.
Learning objective: Estimate VaR using a historical simulation approach.
First recognize that as flawed as VaR is, using VaR based on historical data is even worse.
The idea here is to order the historical losses observed in a portfolio, ordered by the actual
amount of the loss. So, if we have 1,000 loss observations ordered from smallest to largest
and we are interested in the 97% confidence level, that implies that 30 observations will be
in the tail of the distribution that will define the VaR of this portfolio.
So “estimate” here really means “observe,” and in the example the 31st loss observation
would be the amount of the expected VaR of the portfolio based on the losses the portfolio
has historically experienced.
VaR is flawed when used incorrectly, and VaR based on historical observation is really
flawed. There is no insight into the risk the portfolio has, no idea how that risk may change
in the future, and so on. There are almost too many problems here to mention and even
more problems when used with a portfolio that has embedded options.
Learning objective: Estimate VaR using a parametric approach for both
normal and lognormal return distributions.
I will move forward on the assumption you are very comfortable with the lognormal and
normal distributions; take a break here if you need a review.
©2017 Wiley
©
MARKET RISK MEASUREMENT AND MANAGEMENT (MR)
Starting first with the assumption of a normal distribution, we have to define the mean and
standard deviation of the portfolio in order to calculate the VaR. Likely this will be given
to you on the exam, but in the real world this will be estimated parameters based on the
risk manager’s expectations.
Since we want to interpret VaR in terms of lost money, our formula for losses is:
VaR - - | i + oZ a
where negative mu is expected losses, sigma is the standard deviation of the loss, and Z is
the standard normal variate corresponding to the level of significance we are looking for.
Recall that the standard normal variate at the 95% level of significance is 1.645.
If profits and losses over some period are normally distributed with mean of 20 and
standard deviation of 30, the 95% VaR is calculated as:
VaR = -20 + 30(1.645) = -29.35
This is straightforward enough, but the learning objective is asking about a normal and
lognormal distribution of returns, not dollars gained or lost.
When we transition to returns, we have to add additional parameters to describe the
starting dollar value and ending dollar value so we can arrive at the return estimate.
Furthermore, we have to establish some critical value r* so we can compare our expected
probability to our critical value return where that probability is equal to our level of
confidence.
Taking the previous equation:
V a R - - j i + oZ a
I am going to modify it to:
r* = - f t + c Z a
This will be our return critical value that we need to establish a confidence interval around
the return of our portfolio at any given time, rt
Also, since any given return is the starting value relative to ending value, we can define
rt as:
Pt - Pt-i
Don’t let all the equations fool you; we are defining return as you normally would: My
return is my portfolio value at time t minus what it was before, divided by the starting
value.
4
©2017 Wiley
DOWD, CHAPTER 3
However, since we are thinking about VaR at risk, we are only concerned with the losses.
We can extend the relationship to include VaR and recover our critical value:
* _ fi - Pm _ VaR
1
Pm
Pm
Now, substitute into this equation:
r* = - |i + c Z a
to arrive at:
and last,
VaR = (-(I + oZ a )Pt_!
The only difference between the normal and lognormal version is going to be the critical
values.
Let’s learn a quick calculation. If I know returns are normally distributed at 12% with
standard deviation of 18% and I want the 95% level VaR:
-12% +18% (1.645) = 0.1761 (converting from percentages)
On the exam, you may be given estimates of profits and losses, which would mean you
don’t use the population parameters of mu and sigma, but the math and order are the same.
Lognormal Version
The real dividing issue between normal and lognormal that should be clear by now is
that a normal distribution is symmetric and assigns an equal probability of going up or
down in price. This implies that prices can be negative if you use the normal distribution.
Now for stocks, this doesn’t make sense since they have a zero floor. That is why when
we talk about financial assets we often refer to the distribution of returns being normally
distributed, not the asset prices themselves, which, of course, is the lognormal distribution.
This is why we need the lognormal.
So how do we calculate VaR when lognormally distributed? From the discussion of
normally distributed VaR, we know we are dealing with potential returns instead of dollar
values of potential loss (price moves) under the normal distribution, so we have the extra
step of converting our lognormally derived VaR back into dollar terms.
You do not have to memorize the derivation of this formula, but you will use the very last
line in the calculation.
©2017 Wiley
5
MARKET RISK MEASUREMENT AND MANAGEMENT (MR)
As in the normal, we are going to use this formula from before, but let’s call VaR a new
random variable that will correspond to a loss equal to our VaR. Ultimately, this is what we
want to know.
Our original equation:
VaR = - |i + oZ a
becomes:
X* = - p + a Z a
Since we are using geometric returns, we insert a term that describes the price at some
period right now (assuming a loss) relative to some time in the past to establish our return.
It sounds complicated, but this is the same as a $50 stock that goes to $45 so the loss is
10%. 45/50 = .90 & (1 - 0.90) = 0.10. We are going to do the exact same thing here.
X* = ln(Pricenow/Pricepast)
Reordering terms and using the properties of logarithms so the ratio of logarithms can be
expressed as the difference of two logarithms:
X* = ln(Pricenow) - ln(Pricepast)
.^.
0
0
0
0
0
Remember, X is our critical value that ultimately becomes our VaR as some certainty, alpha.
Reordering terms,
ln(Pricenow) = X* + ln(Pricepast)
Distribute the logarithm across: This goes back to early calculus, but don’t worry—you
don’t need to know it; just remember the last step.
(Pricenow) - (Pricepast)ex
Insert original equation for X*:
(Pricenow) = (Price past )e_tl+aZa
Skipping a few messy steps that are unnecessary to know for the exam, we arrive at
what you need to know for VaR under a lognormal distribution of returns, and that is this
relationship:
VaR = Ppast[l - exp(p - oZ)]
6
©2017 Wiley
DOWD, CHAPTER 3
To bring this all together, and since this is an “estimate” question, you may be asked to
calculate/evaluate this on exam day:
Let’s assume returns are normally distributed with mean 10% and standard deviation 18%,
and our portfolio is $100.
We can express the lognormal VaR at the 95% level as:
1 - exp(0.10 -0 .1 8 * 1.645) = 0.26 or $26
Learning objective: Estimate the expected shortfall given P/L or return data.
Risk as Shortfall
Another way to specify a risk objective is by means of the expected portfolio standard
deviation, which is the square root of the expected portfolio variance. For example, the
S&P 500 might have an annual portfolio standard deviation of 23%. Using this figure
and either a normal or lognormal distribution, we can quantify the probability that a
given portfolio will have a loss, or a return below a certain minimum requirement (this
probability is called a shortfall risk).
As might be expected, the focus of this criterion is minimizing the chances that a specified
minimum return (say RL) is not achieved. If the portfolio return (call it RP) is normally
distributed, then it is not very difficult to calculate the probability that RP will fall short
of Rl —in other words, P(RP < RL)—or to calculate the number of standard deviations RL
falls below the expected portfolio return, E(RP).
Note that the portfolio that maximizes E(RP) - RL will at the same time minimize P(RP < RL).
If we divide the expression E(RP) - RL by
the standard deviation of the portfolio, we
get the safety margin in units of portfolio standard deviation. This ratio is called Roy’s
safety-first criterion (SFRatio):
SFRatio = E
aP
Assuming returns are normally distributed, the safety-first ratio will be maximized by
choosing the optimal portfolio, using the three steps of Roy’s safety-first criterion:
1.
2.
3.
Calculate the SFRatio for the portfolio.
Calculate P(RP < Rl ) by looking up N(-SFRatio) = 1 - N(SFRatio).
Choose the portfolio that results in the lowest probability for step 2.
Example: You work for the investment advisor of a major educational institution
with a $2 billion endowment. The school wishes to be able to use $50 million of
the endowment’s investment income annually for operational expenses, but does
not wish to invade the endowment principal. Also, the school intends to place the
©2017 Wiley
©
MARKET RISK MEASUREMENT AND MANAGEMENT (MR)
endowment in one of three investment pools, which have the following return and risk
characteristics:
Expected return
Standard deviation of return
P o o ll
Pool 2
Pool 3
25%
35%
15%
25%
10%
15%
Find the shortfall level of return, RL. Next, according to Roy’s safety-first criterion, which
investment pool is best suited to the school’s objectives? Last, find the probability that the
optimal investment pool will fail to generate a return equal to or greater than RL. Assume
returns are normally distributed.
Solution: If the school wishes to spend $50,000,000 of the endowment investment
income annually, it would imply that the shortfall return level is equal to
$50,000,000/$2,000,000,000 = 2.5%.
Next, to figure out which investment pool is optimal for the school, we must calculate the
SFRatio for each pool.
Pool 1: SFRatio = (0.25 - 0.025)/0.35 = 0.6429
Pool 2: SFRatio = (0.15 - 0.025)/0.25 = 0.5000
Pool 3: SFRatio = (0.10 - 0.025)/0.15 = 0.5000
Pool 1 has the highest SFRatio, so if returns are normally distributed, then it is the optimal
portfolio, and it will minimize the probability that returns will fall short of RL.
The probability that Pool 1 will generate a return less than RL is found by looking up:
N(-SFRatio) = 1 - N(SFRatio) = 1 - N(0.6429) = 1 - 0.7389 = 0.2611 = 26.11%.
Regardless of which investment criterion comes first, risk and return objectives are
essential ingredients in setting a strategic asset allocation.
Learning objective: Define coherent risk measures.
A coherent risk measure satisfies the following five properties, all of which are highly
theoretical and some of which are hard to visualize.
First, a portfolio that has no holdings has no risk (normalized property); second, the lowerreturn portfolio has less risk (monotonicity property); third, risk doesn’t increase when
adding two portfolios together, also known as diversification (subaddivity property); fourth,
an identical portfolio of double the notional should have the potential for double the notional
loss (positive homogeneity property); and fifth, the addition of risk-free assets dilutes the
potential notional loss percentage (translation invariance property). This last one needs an
example: Pretend you have a $100 portfolio with a $1 loss potential for a 1% potential loss.
©2017 Wiley
DOWD, CHAPTER 3
Adding $100 of cash, pretending cash is risk free, increases the portfolio size to $200 with
the same potential for a $1 loss, so the loss percentage is reduced to 50 basis points.
Learning objective: Estimate risk measures by estimating quantiles.
This portion of the reading gets very technical, but stay focused on what you are asked.
Any measure of risk is only as good as its precision. Consider shortfall risk: How sure are
we that we have estimated either the probability or the degree (amount) of loss? VaR is
fairly precise, but that isn’t the only measure of risk.
So when we are talking about quantiles (a quantile comes from statistics, and it is basically
just a regularly spaced interval that divides up a probability distribution into equal-sized
parts—5%, 10%, or half of a distribution), we are talking about setting up a confidence
indicator around the risk measure to indicate the precision of that particular risk measure.
So a quantile by definition is just a sample of a defined size (e.g., top 5%) of a given
distribution, and, just like any other sampling, we can define the standard error of the
sample (quantile).
Now the relationship you need to understand is that the standard error for any sample
size falls as the sample size gets larger. Also, the standard error rises as the estimator
gets farther into the tail. So the standard error of a one-in-a-million event is much wider
than the standard error of a one standard deviation estimation. Stated differently, the
more extreme the probability of an event, the less precise we can be about its estimation.
This has huge implications for risk management using VaR and is one of the key reasons
VaR creates a false sense of security among those who don’t understand the quantitative
limitations of the model itself.
So how do we apply this to expected shortfall (ES) and VaR?
When comparing VaR and ES, both have similar standard errors for normal distributions,
but expected shortfall has much bigger standard errors in distributions with heavy tails,
also known as reality. By definition, ES estimates are less accurate than VaR when
considering distributions with heavy tails.
Learning objective: Evaluate estimators of risk measures by estimating
their standard errors.
The method of estimating I think will most likely be covered on the exam is one that is
called “bootstrapping” and shouldn’t be confused with the method of yield curve creation
by the same name. In this case we take a large sample of estimators from a particular
distribution and then estimate the standard error of that sample. This is convoluted for
sure, but just know for the exam that there is this method called bootstrapping and it
means to create a standard error estimate for a sample of estimators from a particular
distribution.
©2017 Wiley
!
MARKET RISK MEASUREMENT AND MANAGEMENT (MR)
Learning objective: Interpret QQ plots to identify the characteristics of a
distribution.
Recall from the Part I material that a distribution is just a special equation that describes
the distribution of some data set. The probability density function is the name of the
unique equation that describes what we call the bell curve. However, we can define any
equation that fits a set of data as a density function as long as it fits the data and satisfies
a few requirements. A QQ plot is useful to find that special equation that describes a set
of data. QQ stands for quantile-quantile rank and plots the empirical (observed) data in a
particular quantile to what is predicted in the distribution we are testing.
What you need to know for the exam is if a QQ plot is linear, then the empirical data
matches the distribution we are testing, but if it is nonlinear, our empirical data does not
match the distribution we are testing.
Stated differently, if the plot is not linear at a 45-degree angle, the empirical data and the
distribution belong to two different families of distributions.
10
©2017 Wiley
Do w d , Ch
a pt e r
4
Kevin Dowd, Measuring Market Risk, 2nd Edition (West Sussex, England: John Wiley &
Sons, 2005). Chapter 4. Non-Parametric Approaches
After completing this reading you should be able to:
•
•
•
•
Apply the bootstrap historical simulation approach to estimate coherent risk
measures.
Describe historical simulation using nonparametric density estimation.
Compare and contrast the age-weighted, the volatility-weighted, the correlationweighted, and the filtered historical simulation approaches.
Identify advantages and disadvantages of nonparametric estimation methods.
Reading notes: This is a relatively long reading of 30 pages with no “calculate” questions.
You can waste a lot of time here if you aren’t careful. Most of the assigned reading covers
material not actually on the FRM exam. Recall that nonparametric means a way to model
data without actually using the parameters of a model— so this is using historic data or
other number crunching to measure market risk without the use of probability density
functions.
Learning objective: Apply the bootstrap historical simulation approach to
estimate coherent risk measures.
This is a little misleading: a coherent risk measure we defined in the first reading. Recall
that Var is not a coherent risk measure because of subadditivity, and all you need to know
here is how the bootstrap method actually works.
Bootstrapping is another term for resampling from our distribution with replacement. We
resample many times, replacing each time, and create a large number of samples.
We can chart each of these samples on a histogram and then look at any alpha, or degree
of certainty, we want, and that is our bootstrapped VaR. One thing to know for the exam is
that the bootstrap method won’t tell us much about how precise these sample estimates of
risk measures are, but this is how the method works.
Learning objective: Describe historical simulation using nonparametric
density estimation.
Remember, all density estimation refers to is an attempt to build a probability distribution
using past observed profit and loss data. It requires no model parameters since no model
is actually used. We are simply looking at the past and trying to explain the future from it.
Historical simulation has many, many problems, but it is a widely used method.
©2017 Wiley