Blog Week in Risk (ending August 28th)

David Harper CFA FRM

David Harper CFA FRM
Subscriber
FRM Forum Q&A (selected)
Low interest rates
Banks
  • America’s Biggest Banks Have a New Name for Their Venmo-Killer: Zelle http://www.wsj.com/articles/america...-name-for-their-venmo-killer-zelle-1472047872
  • Thought Volcker Rule Went Too Far? There’s More Coming for Banks http://www.bloomberg.com/news/artic...le-went-too-far-there-s-more-coming-for-banks “The Federal Reserve and other agencies are poised to issue a long-overdue report required by the law that lays out recommendations beyond the Volcker Rule to prevent financial firms from triggering an economic crisis … The document will include plans for restricting banks’ investments in copper and hard-to-value assets, said one of the people.”
  • Implementation of Basel standards - A report to G20 Leaders on implementation of the Basel III regulatory reforms http://www.bis.org/bcbs/publ/d377.htm From the summary: “The Basel III capital and liquidity standards have generally been transposed into domestic regulations within the time frame set by the Basel Committee. Further progress has been made towards implementing the Basel III framework since last year’s report. Its key components, including the risk-based capital standards and the Liquidity Coverage Ratio (LCR), are now enforced by all member jurisdictions. Also, all member jurisdictions that are home jurisdictions to global systemically important banks (G-SIBs) have the G-SIB framework in force. Further, member jurisdictions continue their efforts to adopt other Basel III standards, including the leverage ratio and the Net Stable Funding Ratio (NSFR). Non-Basel Committee jurisdictions also report substantial progress in adopting the framework’s core elements.”
  • Opaque Assets at Europe Investment Banks Fuel Capital Doubts http://www.bloomberg.com/news/artic...ent-banks-opaque-assets-fuel-capital-concerns “Banks split assets into three categories: Level 1 for those with transparent prices, like stocks; Level 2 for assets where some external data is available, including over-the-counter derivatives such as interest-rate swaps; and Level 3 for the most illiquid. Banks value these using their own models based on historical data and risk assumptions.
Risk
  • Deloitte’s C-Level, Client-Facing Risk Executive (CRO) http://www.garp.org/#!/risk-intelli...deloitte-c-level-client-facing-risk-executive “Has the stature of risk management changed since the global economic crisis? It’s very different today than it was in 2007. It is much more prevalent today for a CRO to have a direct reporting line to the CEO. And you see the rise of risk officers and board committees across industry sectors. I believe the financial crisis drove these changes. Senior leaders realized that risk wasn’t positioned in the right way, with a seat at the executive table to help make strategic decisions.”
  • Benchmarking Operational Risk Models (Federal Reserve Banks) http://papers.ssrn.com/sol3/papers.cfm?abstract_id=2829267
  • Risk Parity isn’t the Problem, it’s the Solution http://gestaltu.com/2016/08/risk-parity-isnt-problem-solution.html/ “Risk parity is characterized by three primary features. First of all, Risk Parity implementations almost always invest in a diverse basket of asset classes which react in different ways to various economic environments … Second, Risk Parity is about balancing risk. To balance risk, high risk assets like equities must necessarily receive a smaller allocation in portfolios than lower risk assets like bonds … The third distinguishing feature of Risk Parity strategies is that they are usually managed to a target volatility, so that to the greatest degree possible investors receive the experience they signed up for, even during the most hostile market conditions.”
  • Riskalyze.com case study for United Planners Financial Services https://www.riskalyze.com/casestudies/up Here is some information on their questionnaire which apparently produces a single risk number http://kb.riskalyze.com/category/24-risk-questionnaire
Pensions
Natural Disasters (including Weather and Climate Change)
Regulations
Forecasting
Other
  • Visualizing Residuals https://drsimonj.svbtle.com/visualising-residuals
  • The Sinister Side of Cash http://www.wsj.com/articles/the-sinister-side-of-cash-1472137692 “There is little debate among law-enforcement agencies that paper currency, especially large notes such as the U.S. $100 bill, facilitates crime: racketeering, extortion, money laundering, drug and human trafficking, the corruption of public officials, not to mention terrorism … Cash is also deeply implicated in tax evasion, which costs the federal government some $500 billion a year in revenue.”
  • PricewaterhouseCoopers Settles $5.5 Billion Crisis Era Lawsuit (Bankruptcy trustee sued accounting firm over failure to catch fraud at Taylor Bean) http://www.wsj.com/articles/pricewaterhousecoopers-settles-5-5-billion-crisis-era-lawsuit-1472226383 “At issue in the case, one of the few allegations of wrongdoing during the financial crisis that has reached a courtroom, was a fundamental question in accounting: How much responsibility do auditors have for catching fraud? … The collapse of Colonial, which had $25 billion in assets and $20 billion in deposits, was the biggest bank failure of 2009. The FDIC estimates Colonial’s collapse will cost its insurance fund $5 billion, making it one of the nation’s most expensive bank failures.”
  • Real Estate Strikes Out on Its Own in the Stock Indexes www.nytimes.com/2016/08/27/your-money/real-estate-strikes-out-on-its-own-in-the-stock-indexes.html Real Estate is moving out of the Financial sector and will be added as its own new sector (the 11th) in the Global Industry Classification (GICS) structure. Here is the revised GICS classification map (Excel spreadsheet) http://trtl.bz/GICS-map
  • Suits join the hoodies with blockchain push (Cost savings and low returns drive lenders into working with system underpinning bitcoin) https://www.ft.com/content/be30b74c-6a01-11e6-ae5b-a7cc5dd5a28c “The latest example of big banks organizing themselves to exploit the potential of blockchain technology came this week with the announcement that four big lenders have teamed up to develop a utility settlement coin — a new form of digital cash. The four banks — UBS, Santander, Deutsche Bank and BNY Mellon — stress that they are not creating a new cryptocurrency. Instead, the system they are developing uses blockchain technology to create different coins that are each directly convertible into existing currencies deposited at central banks. In essence, it is a way of putting dollars, euros and pounds on the blockchain.” Fintech 2.0 Report here http://trtl.bz/fintech20-santander
  • How leading institutions are changing the rules on portfolio construction http://trtl.bz/mckinsey-rules-portfolio-construction “By far the most important change, however, is coming to the 80/20 alpha/beta management approach. Institutions plan to change those proportions by focusing on building portfolio-construction capabilities, given that these drive the vast majority of long-term returns. The most striking finding from our research is that almost 80 percent of institutions plan to reinforce their central portfolio-construction team, with most expecting to add three to five people. In interviews, leaders also said they expect a more dynamic decision-making process structured around top-down economic scenarios, which they hope will provoke more debate and move them away from a rote approval of strategic asset allocation by the executive committee and board.”
 
Last edited:

David Harper CFA FRM

David Harper CFA FRM
Subscriber
I just wanted to share the following revision text (that I was working on for our notes) which seeks to synthesize the concepts around spectral risk measures (notice links to three forum discussions above). Over the years, we've talked about it a lot (and I contributed to the confusion by writing an incorrect question years ago), and I wanted to try and capture the essential logic.

==========================================
Why expected shortfall is spectral but VaR is not (yet both are general risk meaures)

Dowd introduces the following general risk measure (GRM):
t4_31_3_spectral_measure.png

The keyword here is general. It is so general that it is really sort of an empty container. This GRM assigns weights to quantiles, q(p), according to the weighting function, φ(p). This weighting function, φ(p), is the key feature of the GRM. This general risk function is so general that it includes value at risk (VaR) which is the special case where a single quantile like the 95th percentile is assigned the entire weight. Expected shortfall (ES) is also a special case: ES is where the quantiles in the loss tail (for example, the 5.0% tail) are each assigned an equal weight, and the rest are assigned zero. If we compare a 95% VaR to a 95% ES, in terms of the general risk function, they share in common that both assign zero weights to all quantiles below 95%. They are different in that VaR assigns all of the weight to the 95% quantile, but ES is a weighted average of all losses in the 5% tail (ie., greater than the 95% quantile). The general risk function is elegant because it is about the weighting function, and the weighting function quantifies risk aversion.

But the general risk function is too general to be useful. Dowd introduces three conditions on the general risk function that, if satisfied, render it a spectral risk measure. The important criteria is weakly-increasing, which quantifies risk-aversion. Intuitively, this is the idea that larger losses deserve greater weights in the weighting function (or, at least that they do not deserve less weight!). In this way, we can view a spectral risk measure as a "qualified" general risk measure, where qualified implies that it reflects rational risk-aversion. But, note that a spectral risk measure is still pretty general because there can be many different risk weighting functions. Finally, spectral risk measures are necessarily coherent by definition. We like coherence because coherent risk measures are sub-additive. Sub-additive risk measures do not penalize diversification: by combining positions, the aggregated position is not somehow riskier than the sum of the individual components.

Given this foundation, we can summarize the relationships as follows:
  • Both ES and VaR are instances (special cases) of the general risk measure (GRM) because they each implement the weighting function, albeit very differently. VaR weighs only the quantile associated with the confidence, ES assigns equal weight to all losses in the tail above the confidence.
  • A "qualified" version of the GBM is a spectral risk measure which imposes a risk-aversion condition on the GBM. The spectral risk measure is coherent (which, by definition, implies the spectral risk measure is sub-additive). A spectral risk measure is necessarily a GBM, but a GBM is not necessarily spectral.
  • ES is spectral (and therefore coherent, and therefore sub-additive), but VaR is not spectral (and therefore not necessarily coherent and therefore not sub-additive. VaR is not always coherent, so we say "VaR is not coherent." However VaR can be coherent, we just can't make that assumption without a closer look. Unfortunately, VaR tends to be incoherent at the worst time, when tails are heavy, and extreme losses are highly dependent)
 
Last edited:

David Harper CFA FRM

David Harper CFA FRM
Subscriber
I also wanted to share an update to my attempt to visualize the concept of concordant/discordant pairs. As mentioned above, and confirmed by the author himself, the Meissner text contains an error. For example, he writes that {(1,4), (3,3)} is neither, but the pair {(1,4), (3,3)} is actually discordant. Visually, concordance is easily revealed (see my new diagrams below): simply draw a line connecting the (Cartesian) points. If the slope is positive, the pairs are concordant. If the slope is negative, they are discordant. Easy, yes?

Or, equivalently, pairs are concordant if their relationship is captured in one of the green quadrants; they are discordant if their relationship falls into one of the red quadrants. In my diagrams, the X and Y axes are given by thick black lines. {(1,4), (4,7)} is concordant because (4,7) is "up and to the right" of (1,4). But {(1,4), (3,3)} is discordant because (3,3) is "down and to the right" of (1,4). {(1,4), (5,4)} is neither because they have a horizontal relationship; although not displayed, just for example, {(1,4), (1,7)} is also neither because they have a vertical relationship. I hope that helps solidify these concepts!

0907-concordant-pairs.png
 
Top