BankThink

Would the Stress Tests Have Caught WaMu?

  • Last week, we witnessed the largest initial public offering of a technology company in U.S. history — a company that offers its main service for free, lives entirely on the internet, and is run by a 28-year-old. The euro fell into deeper jeopardy, not merely because of news from the debt markets and at the ballot box, but because Greek and Spanish borrowers reacted to that news in real time, withdrawing much of their euro-denominated savings from domestic banks. And a $2 billion bank loss, based on computer-driven and model-based decision making, almost instantaneously impacted policy discussions and market movements, even before facts became clear.

    May 23
  • Federal banking regulators on Monday provided fresh guidance on how the largest institutions should move ahead with their own routine stress testing exercises.

    May 14

Society has a deep interest in preventing bank failures.  Their occurrence invariably triggers widespread economic stagnation and unpopular calls on the public purse for bailouts to prevent such fallout.  In instituting the Comprehensive Capital Analysis and Review, also known as stress tests, the Federal Reserve is trying to prevent or lessen the odds of such events. 

Now that more is known about the CCAR, we can critically assess how effectively the Fed is going about its crucial task.  Indeed, we can validate the Fed's approach to stress testing by asking whether the failure of Washington Mutual would have been prevented had the CCAR existed from 2005 to 2008. 

Our main thesis is that WaMu failed because of collective, industry level macroeconomic forces and actions.  If WaMu had been the only bank engaged in subprime mortgages, we feel that it would have achieved CEO Kerry Killinger's stated aim of being the Walmart or Starbucks of the banking industry.  The housing boom and bust would have been very mild, the recession would have been caused by some other factor (perhaps a European sovereign debt crisis?), WaMu would exist today and be highly profitable, and the bank would be seen as having a strong social conscience, willing to offer affordable mortgage loans to more people. 

Under the reality we are all enduring, of course, there were many WaMus, busily writing poor quality mortgage loans. These actions caused house prices to boom and then crash. The resultant recession was and is deep and long. And WaMu, among many others, failed spectacularly.

By contrast, the 1981 recession, now considered to be the second worst postwar recession, caused only a small number of banking failures and none of a similar consequence as WaMu.  Put simply, recessions do not cause bank failures.  It is banks causing booms that cause recessions that cause banks to fail.

There are four key insights here, none of which is accounted for by the CCAR process.  The first is that collective actions by many banks can dramatically increase the odds of failure of any individual bank. The second is that the economy is endogenous in the context of widespread credit losses sufficient to cause bank failure.  In other words, not only does the economy affect the banking sector; the reverse is also true. The third is that to create a scenario where bank failure is most probable, a boom is required, followed by a deep recession.  The fourth is that it is insufficient to look only at existing loans in assessing the future credit losses of banks; planned future loan underwriting, especially during the hypothetical boom, also needs to be taken into account. 

The structure of the CCAR—taking a generic, deep, exogenous recession scenario together with a set of microeconomic models of credit loss in various categories—takes no account of collective behavior and is therefore unlikely to provide an indication of future bank failure. 

An injection of macroeconomic credit modeling could make these four features of bank failure explicit components of the CCAR process and thus dramatically improve its chances of forestalling the next WaMu.

The next CCAR should include the following features:

  • A Supervisory Stress Scenario (SSS) that encompasses a short boom followed by a severe recession.  The narrative behind the SSS should be made explicit and public.
  • Consistent with the SSS, industry-level projections of total outstanding credit and related credit losses (probability of default, exposure at default and loss given default) should be provided in each product category (home equity lines of credit, commercial real estate, etc.).  The development of such projections should be done in concert with the economic drivers of the SSS to ensure appropriate feedback mechanisms between credit outcomes and economic performance.  Industry-level credit loss projections should implicitly include losses from new loans originated during the hypothetical boom.
  • Existing microeconomic credit models should be used to assess relative risk appetite of banks and thus to allocate industry-level lending activity and future credit loss outcomes to individual banks.  Bank credit losses should be calibrated to industry-level loss projections stated publicly in the SSS.
  • If future bank plans suggest aggressively seeking market share in a given product line or category, a higher proportion of industry level credit losses derived from new loans should be apportioned to that bank (and less to other, less aggressive institutions).  A higher share of profits from such activities should also be apportioned.

Had such a CCAR existed back in 2005, it would have had an outside chance of identifying problems at WaMu. Fed researchers back then would have needed a time machine to design a scenario as conducive to bank failure as that which actually unfolded.  They could, however, have posited a broad nationwide boom and bust in house prices leading to a bad recession.  They could have also designed a scenario involving a boom and bust in commercial real estate, corporate lending, margin lending or credit card lending, or all of the above simultaneously.  None of these events was without precedent in 2005 and it was well known that these types of booms and busts could cause recessions.  It is conceivable that Fed researchers could have built a "lots of bad new mortgages and house price declines lead to a recession" scenario in the context of CCAR 2005.
In early 2005, though, WaMu had just announced its push into subprime home mortgages but had not had time to stock the tinderbox.  A "recession right now" scenario in 2005 should not have produced red flags since it would mimic the situation of banks on the eve of the 1981 recession.  Such a scenario would have kept the tinderbox bare at WaMu.  If the Fed had assumed a successful ramp-up of WaMu's publicly stated business goal during the boom and allocated a higher proportion of elevated industry losses to WaMu (and Countrywide, and New Century), it would have produced loud warning bells.

The existing CCAR structure may well have identified problems at WaMu in 2007 or 2008, well after the seeds of failure had sprouted.  The Fed should instead be aiming for a process that identifies future failure while the bank can still be saved. The macro concepts described here are crucial if this is to be achieved.   

Tony Hughes is a senior director for consumer credit at Moody's Analytics.

For reprint and licensing requests for this article, click here.
Law and regulation
MORE FROM AMERICAN BANKER