By Glen Fest and Holly Sraeel
After $1 trillion in writedowns, the financial industry must come to grips with a harsh reality: Risk models cannot account for unknown variables, which can deplete liquidity and send institutions into a death spiral, unless they are continuously calibrated with forward-looking elements.
The ability to distinguish between risk and uncertainty - the latter being unknown unknowns, or "unk-unks," as they are known in the aerospace industry - is critical if financial institutions are to recover from the economic meltdown that has hit capital reserves, dried up securitization markets and caused 25 banks to fail.
The latest bank to indicate that it may succumb to insufficient capital reserves? Coral Gables, FL-based BankUnited, the $14.2 billion-asset company whose losses from optional ARMs prompted the Office of Thrift Supervision in September to order the bank to boost capital levels by December 31.
As the global financial system becomes more interconnected and risk dispersed across a broader set of players and markets, experts say it is possible for institutions to manage probabilities associated with market, credit, operational and liquidity risks to one's advantage - if they turn unknown variables into manageable risks.
Risk models of late did not account for future variables and probabilities that were changing through mechanisms institutions did not understand, says Fair Isaac research fellow Larry Rosenberger. "Tools that work very well under stable conditions need to be recalibrated during turbulent times," he says.
Case in point: All parties in the mortgage lending value chain acted as though the underlying consumer risk would not react to changing product features and underwriting standards. "The error," he says, "was that incorrect assumptions were made."
What Rosenberger advocates is risk modeling that relies on predictive analytics providing context and knowledge, including future elements as well as historical data, to turn unk-unk variables into risks that can be managed.
That means identifying probabilities based on micro- and macro-economic scenarios. "People have done a thorough job in risk modeling under two assumptions: that markets remain reasonably tame and liquidity exists, and that risk takers do not take extreme positions that game the models," says Till Guldimann, vice chair of SunGard Data Systems. "If you have highly asymmetric positions, forget most of the models."
The current focus is, of course, on credit risk modeling. Understanding credit risk scenarios can help foresee the potential impact of events on an institution's loan portfolio which can, in turn, affect capital reserves. While the Fed wants banks to stress test their capital reserves better, investors are demanding that banks stress test their credit card, auto, and loan portfolios, prompting banks to refocus on credit stress testing. "One of the things we're beginning to see is not what the impact of 10 percent unemployment will be [on the] portfolio, but what [the] types of parameters are that would break your portfolio," says Fair Isaac svp and chief research officer Andrew Jennings, whose firm introduced a new stress test tool in its risk management suite in early 2008.
In a speech last May, Federal Reserve Chairman Ben Bernanke endorsed the concept of greater credit stress testing, but not until the financial crisis unleashed its full fury later in the year did his message hit home for many bankers. "You're going to see much more [testing]," says Dennis Moroney, research director for bankcards at TowerGroup. Asset securitization investors "are not going to come back unless they feel confident that the vehicles they are putting their money in are not only going to be able to generate returns, but also whether they'd be able to get their money out in the end."
What scenarios are banks stress testing? Anything from the potential demise of General Motors and Chrysler to the gray circumstances around whether the U.S. Treasury's $200 billion Term Asset-Backed Securities Loan Facility can jumpstart consumer and small-business lending. Moroney thinks not, opining in December that TALF doesn't address lenders' "anxiety" about source funding and portfolio management flexibility.
While the need for credit stress testing is clear, it's not an easy exercise. Metavante's John Hurlock, director of integrated risk management, argues that traditional credit-risk indicators and statistical models aren't built to measure outcomes that were never experienced. "The opportunity to misclassify a loan as a good loan, one that will default, or vice-versa, has never been greater," Hurlock wrote in a recent paper.
The value-at-risk model, for instance, reflects daily market price behavior, but not exceptional circumstances that make a market event like the 1987 Black Friday crash a "once in x billion" years possibility, Hurlock says.
The challenge around building new models is daunting. Consultant Joe Ellis, president and chairman of New York-based CEIS Review, says a major stumbling block for banks is having the right data on hand. "You can interface with the loan accounting system and pull a lot of hard data out, but a lot of the credit data you need is sitting in the [paper] credit file," he says. A serious data-management project to merely launch a stress-testing program can be costly - from $75,000 to $300,000 "right off the bat," he says.
The trouble with gathering data is matched by the difficulty in deciding how to turn outcomes into profitable business decisions. "The insight today that comes from stress testing is how [to] help banks decide what their provisioning should be, and in a more accurate way then they had done prior," says Fair Isaac's
The answer to that will build confidence among regulators and investors seeking greater transparency. "There needs to be risk taking, certainly; risk taking generates high returns, but also [can trigger] high losses," says Moroney. The latter outcome, he says, is what banks need to address.
Originally published in Bank Technology News.
Register or login for access to this item and much more
All Information Management content is archived after seven days.
Community members receive:
- All recent and archived articles
- Conference offers and updates
- A full menu of enewsletter options
- Web seminars, white papers, ebooks
Already have an account? Log In
Don't have an account? Register for Free Unlimited Access