When Steve dropped by yesterday and mentioned the broad subject of the so called “science” behind the Slabberator (also known as an Enterprise Risk Management dashboard) being a target rich environment he was not kidding. Stick with me on this post folks – some of the concepts get complex but I think if we can connect enough dots, especially in light of the recent financial meltdown, it will make some sense to even the most financially challenged among us.
By now we have produced enough links to verify that the existence of this contraption dubbed the Slabberator by Editilla is a given including it’s use in risk management at the enterprise level.
To me this begs some questions about whether the logic behind the dashboard holds water, especially in light of the current financial meltdown and whether or not it could explain some of the kookier recent P&C industry moves such Allstate’s unsuccessful rate and document battle with noncaptured insurance commissioner Kevin McCarty and State Farm’s subsequent request for 47% rate increase in Florida despite it’s withdrawal from coastal markets there. IMHO the science behind the slabberator is bogus and ultimately results in ordinary people being slabbed through denial of claims or very high insurance rates. Let’s start with the link left by Steve last night here on slabbed and a bit more from the article on ERM itself.
In 2004, the company “officially went live” with ERM and committed to this quantitative approach to risk management. Allstate appointed a CRO who oversees both ERM and Six Sigma. Allstate believes that Six Sigma is an excellent set of tools to help reduce defects in processes and to help develop future leaders, thereby reducing the level of operational risk. The two processes share a common analytical and quantitative skill set and provide a comprehensive view of both bottom-up process control and top-down enterprise risk. Today, the ERC holds discussions about the amount of risk various areas of business are willing to take on (e.g., catastrophes, interest rate, equity and mortality) and the accompanying risk/reward tradeoffs. This information is used to help set a capital plan for the businesses and then develop a corporate financial plan. Perhaps most important, Allstate’s senior leaders can now more effectively function as stewards of the entire portfolio of risks across the enterprise, rather than as owners of the particular risks managed within their respective functional areas.
The article continues:
Because Allstate has developed its own sophisticated internal risk models and has integrated them into a comprehensive risk framework to drive the company, it has been able to build a solid argument for using its own metrics — rather than relying solely on rating agency and RBC metrics — to determine overall capital needs as well as capital allocations and levels within the various businesses. While Allstate remains very mindful of the rating agencies’ capital formulas, EC has now been incorporated into the capital management dialogue with the rating agencies. Allstate initially held a series of discussions and demonstrated its EC and ERM analytics with rating agencies in the summer of 2005. Then, in late August, Hurricane Katrina hit the Gulf States, where Allstate had significant exposure. Because Allstate had been in discussions with the rating agencies before Katrina and was able to demonstrate the outcomes of its modeling, the groundwork had been laid for frank discussions about the company’s capital management options following Katrina, Rita and Wilma. As a result of these discussions and strong underlying performance, even in the face of substantial losses, Allstate was able to demonstrate an effective post-hurricane capital management strategy and maintain its ratings.
One question that naturally follows for me is did the hard line taken by insurers against policyholders here post Katrina derive from an effort to make the results fit the modeling? Even more important is whether it is possible for a “right tail” catastrophic event like Hurricane Katrina to be properly priced in any such modeling? More on that later. There is more in the Towers Perrin article that helps bring things into sharper focus:
Much of Allstate’s initial ERM work was for Allstate Protection, its property/casualty business, but there was activity in other areas as well. Allstate Financial, the life, savings and retirement business, had invested in its own ERM efforts, primarily through ALM modeling of annuities, life insurance and other products. In addition, Allstate was doing sophisticated modeling of assets on the investment side. For example, in 2005, the Allstate Investment group partnered with the ERM team to execute strategic asset allocation through dynamic modeling of assets and liabilities.
On the last page of the Towers and Perrin document is some Q&A with Tom Wilson, Allstate’s CEO:
Tom Wilson: At the outset, we didn’t even call it ERM. The insurance business allocates capital against risks every day, one policy at a time. The typical practice is to allocate capital “in bulk” within an individual line of business. Other industries, like manufacturing, use different economic models to calculate expected returns on investment. We recognized that we needed to have a comparable process for the insurance business, with a set of common analytic constructs that people could understand and use in their decision making. We saw an opportunity to use more sophisticated analytics to look at risks individually and across businesses simultaneously, so that we could answer questions such as: How much capital do we need to write a personal auto policy? What is the trade-off between writing an auto policy in one state versus the other? How about in comparison to writing homeowners? Or in comparison to increasing risk in our investment portfolio? We knew that, to answer these questions, we needed to better leverage and expand the way we manage, analyze and utilize our data, so that’s what we set out to do.
From the start we were mindful that risk analytics can easily lead to risk reduction exclusively.Using the tools we’ve developed for ERM, we can increase exposures to risk where we see benefit and opportunity and economic return.
Tillinghast: How does Allstate’s ERM process create value?
Tom Wilson: It has helped us optimally structure our investment portfolio. We do sound asset/liability modeling now, but I believe there are opportunities for us to identify ways of intelligently taking on more risk. Most insurance companies are good at managing risks inside the shape of a normal distribution of outcomes. The differentiator is how well they manage risk for low frequency, high-severity events. One example worth mentioning: Though we don’t insure for earthquakes in California, ERM helped us identify tail risks associated with fire following earthquake exposure. We were able to take measures to protect against this type of event through a combination of reinsurance and rating/policy changes, and our analysis also led to insights that had a bearing on our underwriting criteria for insuring property in California. (emphasis Sop)
The implications of Mr Wilson’s statements are staggering and bear out much of what we have said over the past year on slabbed in capitalizing on an artificially fragmented marketplace to allocate capital for instance. More needles in Nowdy’s haystack of needles. In addition I highlighted a tacit admission from Mr Wilson that Allstate manages under the assumption that right tail cat events, aka, low frequency, high-severity events can be modeled even going so far as to identify a success of the program. AS CEO he is paid to pump his company so it is natural he would not mention any failures. Along those lines I am told Mr Wilson’s predecessor and now AIG CEO Ed Liddy went on CNBC on August 29th 2005 mentioning reinsurance for Katrina losses no doubt also the result of applied ERM. Again we’ll visit that later so we can accumulate more evidence so the blind will see.
Next up is Mr CLS on Yahoo! Allstate with a particularly insightful post, the last part of which is salient to the Slaberrator, and right tail ERM modeling:
WSJ October 8, 2005, pC1-C4; “STATE FARM Ins. Co. (part of Zurich Group) Cat Bond issued this Cat Bond(Kamp Re). A “PARTIAL” loss payment was made on December 14, 2007 (that’s TWO years after Katrina, a long time to wait for a first party policyholder to get his claim payment, but then again when State Farm has their contractor engineering firms deliberately changing and falsifying engineering reports to “flood” instead of “wind” could explain a “partial payment”) and it is EXPECTED that investors will experience a TOTAL loss of principal.” God only knows when the “final” loss payment will be made, if the “money” can be found.
The SPV (special purpose vehicle) is set up offshore in a foreign domicile which allows the SPV to be treated as tax-exempt. (NO U.S. tax payments from them, no tax dollars going into the U.S. treasury),……..(that’s you & me, GAR, like Ms. Helmsley said, “only the little average people pay taxes).
Anyway, a “swap arrangement” is made between the TRUST and a “counterparty” to allow the trust to invest in high-grade bonds, which are held for the benefit of both the investors and the ceding insurer, while also allowing for the LIBOR base interest rate to be paid to the investors. The counterparty agrees to pay LIBOR in return for receiving the interest earned by the TRUST from the bonds in its portfolio.
Foreign investors could therefore rest assured that their income from these bonds would be treated as portfolio interest, which is exempt from U.S. federal income tax withholding, rather than as dividends from equity, which are not exempt from U.S. federal income tax withholding.
Now with the collapse of LEH, the RICO insurance lawsuit; the Rigsby’s false claims (Qui tam); there is some “tail-risk” coverage associated with cat bonds which would allow time to research, discover, and pay all claims relating to a major event like HURRICANE Katrina.
Again we hear the term “tail-risk” and it’s new fangled association with handling reinsurance through securitization, the Catastrophe Bond. We have written about Cat Bonds extensively here on slabbed. Type it in our search box for those posts. On the Slabberator aka the ERM dashboard we note note this area is addressed via the left middle controls under “reinsurance logic” where the risk meter is set to the desired level and the reinsurance treaty cost is input. This has to work correct? The math, while complicated, is also straightforward. Not so fast.
In fact the lessons of economic history, especially recent events show exactly the opposite. Tail risk can not be priced by the private marketplace! Why? Because math and statistics are empirical while human behavior is unpredictable. The assumptions always leave out a variable or misstate their magnitude.
is it any wonder the insurance industry as exhibited by Allstate was willing to endure huge contempt citations in Missouri and New Mexico claiming these documents including the dashboard were so called trade secrets. That’s right ladies and gents, we were leaked the dashboard by an interested party wondering it’s meaning. What we find is it is a bogus rationalization concocted by 7 figure consultants who don’t know their ass from a hole in the ground that ultimately resulted in rape of coastal wind insurance consumers. In short our Katrina experience is the intersection and end result of McKinsey, ERM and Six Sigma. No wonder our collective Katrina experience with private insurers post storm is best described by the term cluster fuck.
Of course my rather strong opinion does not derive from the bar rantings of pissed off insurance consumers. Financial/economics history is replete with such folly. Before we visit with Nassim Taleb let’s take a mathmetical mystery tour down the road to insolvency beginning in 1998 with the collapse of Long Term Capital Management, a hedge fund founded by the best and brightest financial minds of that time:
Long-Term Capital Management (LTCM) was a U.S. hedge fund which used trading strategies such as fixed income arbitrage, statistical arbitrage, and pairs trading, combined with high leverage. It failed spectacularly in the late 1990s, leading to a massive bailout by other major banks and investment houses, which was supervised by the Federal Reserve.
LTCM was founded in 1994 by John Meriwether, the former vice-chairman and head of bond trading at Salomon Brothers. Board of directors members included Myron Scholes and Robert C. Merton, who shared the 1997 Nobel Memorial Prize in Economic Sciences. Initially enormously successful with annualized returns of over 40% (after fees) in its first years, in 1998 it lost $4.6 billion in less than four months following the Russian financial crisis and became a prominent example of the risk potential in the hedge fund industry. The fund folded in early 2000.
The story behind the story with LTCM speaks to the folly inherent in Enterprise Risk Management:
The idea behind LTCM was quite simple to articulate but not necessarily that easy to implement. LTCM was to look for arbitrage opportunities in markets using computers, massive databases and the insights of top level theorists. These opportunities arose when markets deviated from normal patterns and was likely to re-adjust to the normal patterns. By creating hedged portfolios the risks could be reduced to low levels. According to the model developed by Merton the risk could be reduced to zero, but in practice some of the crucial assumptions of Merton’s model did not hold so the risk of the hedged portfolios was not really zero, as subsequent events proved.
Myron Scholes stated the objective of LTCM in a striking image. He said LTCM would function like a giant vacuum cleaner sucking up nickles that everyone else had overlooked.
Then we fast forward to subprime and the securitization of subprime mortgages into Mortgage Backed Securities and their derivative friend the Credit Default Swap. We’ve written about the financial crisis extensively too and in a comment to a post Nowdy did several weeks ago Russell left us a link that has surfaced more times since in the blogosphere than I can count in Michael Lewis’s End of Wall Street piece at Portfolio.com that sums up many of the criticisms from an insider’s perspective with the incredible story of Meredith Whitney and Steve Eisman:
Meredith Whitney didn’t sink Wall Street……This woman wasn’t saying that Wall Street bankers were corrupt. She was saying they were stupid. These people whose job it was to allocate capital apparently didn’t even know how to manage their own.
Finally we go back to the future with Nassim Taleb on theory behind this subject. First off this recent addition to his website added evidently in hopes people would refrain from slaughtering his Black Swan concept:
For the last 12 years, I have been telling anyone who would listen to me that we are taking huge risks and massive exposure to rare events. I isolated some areas in which people make bogus claims –epistemologically unsound. The Black Swan is a philosophy book (epistemology, philosophy of history & philosophy of science), but I used banks as a particularly worrisome case of epistemic arrogance –and the use of “science” to measure the risk of rare events, making society dependent on very spurious measurements. To me a banking crisis –worse than what we have ever seen — was unavoidable and NOT A BLACK SWAN, just as a drunk and incompetent pilot would eventually crash the plane. And I kept receiving insults for 12 years!
His well informed rant continues:
I just came back from Paris where I listened to people defend mathematical risk management, frustrated at my attempts to conceal my anger. Then I was emailed a million versions of Greenspan’s testimony –the sight of whom makes me fly into rage. I am now convinced that an (advanced) economics degree lowers one’s ability to understand the difference between absence of evidence and evidence of absence. Some people need to be locked up, and locked up quickly.
I end this lengthy post with its mountain of evidence again with Taleb early on in his 12 year crusade against what was then called “financial engineering”:
Derivatives Strategy: What problems do you have with financial engineering?
Nassim Taleb: I disagree with such an approach in financial risk management. Some people looked at the literature and saw differential equations and said, “Gee, it’s like engineering.”
Engineering relies on models because you can capture the relationships in the physical world very well. Models in the social sciences serve a different purpose. They make strong assumptions. Economists have known for a long time that math in their profession has a different meaning. It’s just a tool, a way to express yourself.
DS: So real engineering could lead to a bridge that you could reliably drive cars across. But modeling in financial engineering isn’t certain enough to run a portfolio …
NT: Exactly. In finance, you are not as confident about the parameters. The more you expand your model by adding parameters, the more you become trapped in an inextricable apparatus of relationships. It is called overfitting.
DS: What do you think of value-at-risk?
NT: VAR has made us replace about 2,500 years of market experience with a co-variance matrix that is still in its infancy. We made a tabula rasa of years of market lore that was picked up from trader to trader and crammed everything into a co-variance matrix. Why? So a management consultant or an unemployed electrical engineer can understand financial market risks.
To me, VAR is charlatanism because it tries to estimate something that is not scientifically possible to estimate, namely the risks of rare events. It gives people misleading precision that could lead to the buildup of positions by hedgers. It lulls people to sleep. All that because there are financial stakes involved.
To know the VAR you need the probabilities of events. To get the probabilities right you need to forecast volatility and correlations. I spent close to a decade and a half trying to guess volatility, the volatility of volatility, and correlations, and I sometimes shiver at the mere remembrance of my past miscalculations. Wounds from correlation matrices are still sore.
DS: Proponents of VAR will argue that it has its shortcomings but it’s better than what you had before.
NT: That’s completely wrong. It’s not better than what you had because you are relying on something with false confidence and running larger positions than you would have otherwise. You’re worse off relying on misleading information than on not having any information at all. If you give a pilot an altimeter that is sometimes defective he will crash the plane. Give him nothing and he will look out the window. Technology is only safe if it is flawless.
A lot of people reduce their anxiety when they see numbers. They want a triple-digit delta or gamma, for example, not taking into account that it is foolish to be precise with deltas when you don’t even know the parameters.
Before VAR, we looked at positions and understood them using what I call a nonparametric method. After VAR, all we see is numbers, numbers that depend on strong assumptions. I’d much rather see the details of the position itself rather than some numbers that are supposed to reflect its risks.
Clearing firms understood that very well. Ironically, the stock market crash coincided with the discovery of this so-called parametric system used to run the risks of option traders. In the old days the clearing firms looked at how many calls you were short and how many you were long, and if you sold a lot of calls they would get nervous and call you up and ask you to liquidate some of them. After they went to parametric monitoring of option positions using second-rate statistical methods, the options traders started building up massive short put positions that, along with portfolio insurance, helped to accelerate the crash. Now they’re coming back to square one with their nonparametric methods, particularly with the puts.
DS: Do you think the whole idea of trying to use statistics to model a particular distribution is fraudulent? Or is it possible to come up with something approximating the truth?
NT: The problem we have with statistics is that although we know something about distributions, we know very little about processes. A process is a distribution that has time in it, and things change with time. People look at fat tails and say, “We can simulate distributions with fat tails.” But the reason distributions have fat tails may be because these distributions don’t have stable properties over time.
DS: VAR proponents will also admit that VAR doesn’t work as well on something with an asymmetric payoff.
NT: Yes, but any dynamic trading strategy by a leveraged investor that has a stop loss in it has an asymmetric payoff and needs to be treated like an option. If I trade deutsche mark or bond futures with a stop loss, the frequency of my losses will be greater than the frequency of my profits but the magnitude of my losses will be smaller to compensate. It will look like a payoff of an option, and that’s not captured by VAR. The VAR assumes than traders are stuffed animals between two reports.
DS: Are you saying VAR can’t be used to measure risks on a trading desk?
NT: The risks of common events perhaps, those that do not matter, but not the risks of rare events. Moreover traders will find the smallest crack in the VAR models and try to find a way to take the largest position they can while showing the smallest amount of risk. Traders have incentives to go for the maximum bang because of the free option they’re granted.
In the end someone turned the dials and the goals were set. Word made it’s way down the chain to claims where the scheme was executed. And back to Mssrs Liddy and Wilson and their success story with ERM. What the Tower Perrin piece did not mention was subsequent write downs of their subprime exposure or the fact Allstate was running naked for reinsurance for Louisiana Katrina/Rita losses. The output of the dashboard is only as good as the assumptions it uses. Meantime this country’s taxpayers were raped right along with Gulf Coast Katrina/Rita victims. Under this new fangled way of shedding risks and getting federal backstops the bailout for insurers is continuous and under the radar….another needle in a haystack of needles.
Introduction: The Scheme: you lived the movie
Chapter 1: The Scheme: first there were just word games