Century. The modern bank is an amalgam of predecessors including the Union Bank of Switzerland and the Swiss Bank Corporation with brokerages such as PaineWebber and Dillon Read. In addition to its banking franchise, UBS is also the world’s largest money manager for wealthy individuals, but its troubles are legion. UBS has suffered $38 billion in write downs related to its subprime mortgage exposure; only Citigroup has lost more. UBS recently announced plans to exit the US municipal bond market, a business that, according to Bloomberg News, "it sought to dominate for almost a decade." Clients pulled out more than $12 billion in the first quarter and inflows of new money are down 82%. The bank is slashing its workforce and its market value has been cut in two. And if that weren’t enough, the Justice Department is investigating it for helping clients evade US taxes.
The Swiss Federal Banking Commission required UBS to prepare a detailed analysis on what went wrong. On April 18 the bank issued a document called "Shareholder Report on UBS’s Write-Downs" based on that regulatory requirement. The report tells the story of how a financial institution filled with highly intelligent, highly educated, and highly motivated people, screwed up. And not for the first time. In 1998, UBS lost more than $600 million in the Long Term Capital Management fiasco and had to contribute an additional $300 million to the subsequent bailout. To quote UBS’ own publicity (written before the current losses), "Far worse than its financial impact, the LTCM episode dealt a crushing blow to the firm’s image and confidence. An air of pessimism pervaded most divisions, leading to significant employee dissatisfaction and client losses." So UBS knew the stakes and the consequences, up close and personal.
Why then, did it happen again? The answer is deceptively simple. UBS believed its mathematical models. That’s it, the core mistake at the very heart of our current financial crisis. It’s not the only reason, but it is the single, indivisible particle from which everything else evolved, $323 billion of credit losses and write-downs worldwide, and counting.
We have to back up a bit to understand why this is so. In fact, we need to back up about 75 years. New Deal legislation created a legal framework and a set of national safety nets to encourage and protect home ownership. The Roosevelt program, however, was grounded on a familiar and practical reality. Mortgages were made to creditworthy borrowers by experienced home-town lenders who knew their customers and their communities. That was the ultimate safeguard. Local banking, however, is expensive and, over time, that expense made local lenders less competitive. Mortgages became less of a bank product than a product of financial markets. Risk was evaluated, not in terms of individual borrowers living in specific neighborhoods, but in terms of value at risk, Monte Carlo simulations, and two sigma events. models. They aren’t the real thing. When we see pictures in a magazine of a high fashion model wearing garish makeup, clothing that barely conceals anything, and an expression of unbridled lust, we do not expect to actually run into such a person at the supermarket. Those of us who were boys in the era before video games knew that the plastic models of jet planes we built couldn’t really fly at supersonic speeds. But the people who built the mathematical models for the behavior of subprime mortgages thought they were accurate representations of how those mortgages would perform in the real world.
If you don’t understand those terms, don’t worry. Most of the people who relied on the outputs of these statistical techniques didn’t understand them either. The problem of modern mathematical models is not that they produce flawed results. The people who write the models and who really do understand how they work are very, very smart. No, the problem with mathematical models is that they are
Paul Volcker, the Federal Reserve Chairman who preceded Alan Greenspan, stated the point succinctly in a speech last month. "Mathematical modeling, drawing strong inferences from the past has demonstrably failed to anticipate events of potentially systemic importance…Part of the problem, as I understand it, is that mathematical modeling simply cannot deal with markets, where it is not random or physically determined events, but human instincts that cause self-perpetuating waves of unwarranted optimism or pessimism."
Mathematical models can’t predict real human life because they lack the basic human traits of imagination and irrationality. Nevertheless, UBS relied on them implicitly. A couple of examples will suffice to demonstrate how bad it got. UBS’ risk management system downgraded its credit rating for Japan in 2003. As a result, the part of its investment apparatus that managed short-term funds stopped buying Japanese Government Bonds, the Japanese equivalent of US Treasury securities. Instead they bought triple-A rated classes of US asset-backed securities. These are pools of car loans, credit card receivables, and home equity lines of credit made to US borrowers. How much imagination does it take to understand that the Government of Japan is a better credit risk than the guy down the street? Apparently, it takes more imagination than UBS’ models. As a result of this decision, UBS’s cash equivalent trading area lost between $25-30 billion in the third quarter of 2007. This was about a 7% loss, which is an astonishing loss rate for short-term investments.
The second example is, if anything, more astonishing. Unfortunately, I can’t quote the most relevant paragraph in the UBS shareholder report because it is so full of jargon and acronyms as to be all but unintelligible without a glossary. (UBS thoughtfully provided its shareholders a six-page glossary along with its report, and I’ll have more to say on this later.) The gist is that UBS’s risk systems did not even collect, much less analyze, the relevant and available data on the mortgages making up the subprime and asset-backed pools it bought! Among other things, UBS ignored whether the loans were first or second mortgages, when they were made, and the credit scores of the borrowers. And UBS wasn’t alone. "Based on publicly available information, UBS believes that its approach to the risk measurement and valuation of structured credit products reflects issues which were not unique and that a number of other financial institutions with exposure to the US subprime market used similar approaches." Don’t blame us, everybody did it!
What they all did was build models with a few years of mortgage history, which showed continually rising home values and low delinquency rates. Since the future reflects the past (doesn’t it?) there was no warning of trouble ahead. But let’s suppose that some model builder had used her imagination and included a catastrophic loss scenario. Would anyone have believed it? The answer is, "No." Up until the present emergency, there had not been a prolonged nationwide housing slump since the Great Depression. Certainly there had been brief nationwide slumps and prolonged local slumps, but nothing like the present. It’s obvious in retrospect that home values can decline across the board, but even a year ago and certainly two years ago, it was unimaginable. And what’s unimaginable, even in the unlikely event that it can be modeled, will not be believed.
Besides, there were enormous pressures to believe what the models were saying. They were, after all, written by the smartest people in the room. Who would have the courage to doubt them, especially when everything was wrapped in jargon and in-crowd acronyms? Here’s a sample: "The business did not submit NBI requests for either the CDO structuring business or for the AMPS business and there were no overall notional portfolio limits established for CDOs or AMPS. CDO and AMPS deals were approved on a transaction-by-transaction basis, using the TRPA process."
UBS had a perfect risk management philosophy. Each business unit "owned" its risks, and independent controls were established to check and verify. Timely and full disclosure of risk was required. The main goals of risk management were the protection of earnings and the protection of UBS’ reputation. None of it mattered. UBS couldn’t protect itself against a risk it couldn’t imagine.
Some people, not at UBS, did imagine the risk. Some people, not at UBS, made enormous amounts of money betting that the housing boom would bust. But, even imagination isn’t enough. There were probably a lot more people who thought that housing would crash but who didn’t have the patience, the money, or the nerve to hold out through the long period when home values were soaring. And even imagination, patience, money, and nerve are insufficient without a healthy dose of luck. There is a cynical but sadly true saying that sums it all up, "The markets can stay irrational longer than you can stay solvent."