What's new

# Week in RiskWeek in risk (May 12th)

#### David Harper CFA FRM

##### David Harper CFA FRM
Staff member
Subscriber
New Practice Questions
• P1.T4.918. Internal versus external credit ratings (de Servigny Ch.2) https://trtl.bz/2vQQWE1
• P2.T6.909. Credit exposure metrics continued (expected positive exposure and effective exposure) (Gregory Ch.7) https://trtl.bz/2vZj4ER
In the Forum

1. Correlation versus dependence: A few of us had a good discussion about correlation and dependence here https://trtl.bz/2PZ0JRH. In the FRM, it's important to know (per Dowd and Meissner) that correlation typically refers to linear (aka, Pearson's) correlation, but dependence is a much broader set of associations. Make sure you know the test for statistical independence. If variables are not independent, they are dependent. Dependence includes linear correlation and non-linear associations. We have a wonderful excuse to practice some logic here. Start with the true conditional: If variables are correlated, then they are dependent. The contrapositive must also be true: If variables are independent, then they are uncorrelated. But we cannot assert either the inverse or the converse: Zero correlation does not imply independence; and dependence does not imply correlation. Not unrelated, Meissner's characterizations about correlation versus volatility give candidates confusion every year because he's mixing empirical observations with mathematical truisms. See https://trtl.bz/2Q1Ur3A Here is an example of a mathematical truism: because ρ(x,y) = σ(xy)/[σ(x)σ(y)], higher volatility implies lower correlation. But an empirical observation does not need to follow such a clinical rule!

2. Potential future exposure is a long-dated value at risk (VaR) but for unrealized gains: @RushilChulani asks a good question about potential future exposure (PFE) here https://trtl.bz/2PZ1pXf and I just want to remind you that PFE is only different from value at risk (VaR) in its application and context. Both PFE and VaR are merely statistical quantiles. Specifying the distribution is the hard part, but then both PFE and VaR ask a one-sided question: on the only side of the distribution that we care about (i.e., the loss for VaR, the unrealized gain for PFE), where do we locate the 5.0% or 1.0% or 0.1% tail?

3. Instructive practice paper errors (subadditivity and implied volatility): As I mentioned last week, GARP's prior practice papers have errors so we don't recommend using them prior to the current 2019. They continue to be instructive only if you are willing to learn from mistakes. This week's discussion include an interested fallacy with respect to VaR's lack of subadditivity https://trtl.bz/2Q2KkeX and a misinterpretation of the implied volatility skew for in/out of the money options https://trtl.bz/2Q13EZW. Here is a question from a third-party (i.e., neither GARP nor us, but posted on our forum) which does not use the proper foreign exchange (FX) convention: https://trtl.bz/2Q13EZW

4. Other (forward rate notation, portfolio VaR): I was glad to answer this question about proper forward rate notation, if only because I've had years of practice reading and writing the different styles used: https://trtl.bz/2PWr3vC. And @gprisby asked a good question: in regard to portfolio VaR, is component VaR always greater than incremental VaR? https://trtl.bz/2Q2KWkL

External

1. Cyber risk grows in prominence, but necessarily skills are short and non-traditional: Norman Marks shared two New Reports on the Cost and Incidence of Cyber Breaches https://trtl.bz/2PZod9m. The Financial Stability Board (FSB) published Cyber Security: Finding Responses to Global Threads https://trtl.bz/2VjacnM. Marsh published its 2019 edition of the MMC Cyber Handbook https://trtl.bz/2VoAgOI. Cyber Risks to Exceed Natural Disasters for Insurers (says Scor CEO): https://trtl.bz/2Q2rZi2. Cybersecurity Jobs Abound. No Experience Required (WSJ) https://trtl.bz/2PZp2Pu “It’s really difficult to determine just by someone’s academic track record if they could be great at the kind of problem-solving we do, but it’s relatively easy in a simulation game,” [IBM's head of offensive-security services] said.

2. Chief Risk Officer (CRO): An Increasingly Vital Role in Effective Risk Oversight (by Carol Williams) https://www.erminsightsbycarol.com/chief-risk-officer-vital-role/ who points to the 2019 The State of Risk Oversight: An Overview of Enterprise Risk Management Practices (10th Edition) https://trtl.bz/2Q2t4q6. Also: Rising Through the Regions: John Turpen’s Road to CRO (GARP) https://trtl.bz/2PXdfBh.

3. Other (Risk literacy, Experts, and the Tidyverse): 3.1. Risk Literacy And Interpreting The Probabilistic Risks Of Investing And Retirement Planning https://trtl.bz/2Q9ycsV. 3.2. The Peculiar Blindness of Experts https://trtl.bz/2PZz7vL. 3.3. A Beginner’s Guide to Tidyverse (The Most Powerful Collection of R Packages for Data Science) https://trtl.bz/2Q1yjGw

Last edited:

#### LisaCoy

##### New Member
New Practice Questions
• P1.T4.918. Internal versus external credit ratings (de Servigny Ch.2) https://trtl.bz/2vQQWE1
• P2.T6.909. Credit exposure metrics continued (expected positive exposure and effective exposure) (Gregory Ch.7) https://trtl.bz/2vZj4ER
In the Forum

1. Correlation versus dependence: A few of us had a good discussion about correlation and dependence here https://trtl.bz/2PZ0JRH. In the FRM, it's important to know (per Dowd and Meissner) that correlation typically refers to linear (aka, Pearson's) correlation, but dependence is a much broader set of associations. Make sure you know the test for statistical independence. If variables are not independent, they are dependent. Dependence includes linear correlation and non-linear associations. We have a wonderful excuse to practice some logic here. Start with the true conditional: If variables are correlated, then they are dependent. The contrapositive must also be true: If variables are independent, then they are uncorrelated. But we cannot assert either the inverse or the converse: Zero correlation does not imply independence; and dependence does not imply correlation. Not unrelated, Meissner's characterizations about correlation versus volatility give candidates confusion every year because he's mixing empirical observations with mathematical truisms. See https://trtl.bz/2Q1Ur3A Here is an example of a mathematical truism: because ρ(x,y) = σ(xy)/[σ(x)σ(y)], higher volatility implies lower correlation. But an empirical observation does not need to follow such a clinical rule!

2. Potential future exposure is a long-dated value at risk (VaR) but for unrealized gains: @RushilChulani asks a good question about potential future exposure (PFE) here https://trtl.bz/2PZ1pXf and I just want to remind you that PFE is only different from value at risk (VaR) in its application and context. Both PFE and VaR are merely statistical quantiles. Specifying the distribution is the hard part, but then both PFE and VaR ask a one-sided question: on the only side of the distribution that we care about (i.e., the loss for VaR, the unrealized gain for PFE), where do we locate the 5.0% or 1.0% or 0.1% tail?

3. Instructive practice paper errors (subadditivity and implied volatility): As I mentioned last week, GARP's prior practice papers have errors so we don't recommend using them prior to the current 2019. They continue to be instructive only if you are willing to learn from mistakes. This week's discussion include an interested fallacy with respect to VaR's lack of subadditivity https://trtl.bz/2Q2KkeX and a misinterpretation of the implied volatility skew for in/out of the money options https://trtl.bz/2Q13EZW. Here is a question from a third-party (i.e., neither GARP nor us, but posted on our forum) which does not use the proper foreign exchange (FX) convention: https://trtl.bz/2Q13EZW

4. Other (forward rate notation, portfolio VaR): I was glad to answer this question about proper forward rate notation, if only because I've had years of practice reading and writing the different styles used: https://trtl.bz/2PWr3vC. And @gprisby asked a good question: in regard to portfolio VaR, is component VaR always greater than incremental VaR? https://trtl.bz/2Q2KWkL

External

1. Cyber risk grows in prominence, but necessarily skills are short and non-traditional: Norman Marks shared two New Reports on the Cost and Incidence of Cyber Breaches https://trtl.bz/2PZod9m. The Financial Stability Board (FSB) published Cyber Security: Finding Responses to Global Threads https://trtl.bz/2VjacnM. Marsh published its 2019 edition of the MMC Cyber Handbook https://trtl.bz/2VoAgOI. Cyber Risks to Exceed Natural Disasters for Insurers (says Scor CEO): https://trtl.bz/2Q2rZi2. Cybersecurity Jobs Abound. No Experience Required (WSJ) https://trtl.bz/2PZp2Pu “It’s really difficult to determine just by someone’s academic track record if they could be great at the kind of problem-solving we do, but it’s relatively easy in a simulation game,” [IBM's head of offensive-security services] said.

2. Chief Risk Officer (CRO): An Increasingly Vital Role in Effective Risk Oversight (by Carol Williams) https://www.erminsightsbycarol.com/chief-risk-officer-vital-role/ who points to the 2019 The State of Risk Oversight: An Overview of Enterprise Risk Management Practices (10th Edition) https://trtl.bz/2Q2t4q6. Also: Rising Through the Regions: John Turpen’s Road to CRO (GARP) https://trtl.bz/2PXdfBh.

3. Other (Risk literacy, Experts, and the Tidyverse): 3.1. Risk Literacy And Interpreting The Probabilistic Risks Of Investing And Retirement Planning https://trtl.bz/2Q9ycsV. 3.2. The Peculiar Blindness of Experts https://trtl.bz/2PZz7vL. 3.3. A Beginner’s Guide to Tidyverse (The Most Powerful Collection of R Packages for Data Science) https://trtl.bz/2Q1yjGw
really looking forward to visiting all these on the descending order of chronology. something to keep me extremely busy and away from the depressing news out there.