Wednesday, March 4, 2009

Stress Testing Banks

About 18 months ago we were showing our then newly minted Economic Capital and RAROC computations to one of the banking regulators. Calibrate in your head what the world was like back then. It was one of those big phone calls with lots of analysts. Objections were raised due to the fact that the IRA method arbitrarily set the cutoff point for EC’s analysis at the B or better rating bond equivalent, that’s roughly 1,000 basis points (bp) maximum allowable loss. Looking at the output, the following ensued,

Unknown voice: “Dennis if your numbers are correct it would mean that there’s no way some of these large banks could possibly stay in business.”

My response, “Yes it does indeed question their viability.”

Next question, “So will you change your numbers?”

Response, “No we will not. I believe the numbers are correct.”

I went on to explain that we had in fact “parametrically stressed” the Economic Capital factors as an exercise to locate the point where one would again get analytical indications of business viability for some of these larger vulnerable institutions. Our findings were that the allowable loss margin threshold needed to be raised to around 3,000 bp to make that happen. I reminded everyone that the cutoff between investment grade and junk was around 2,800 bp, the term toxic was not yet in vogue. This was simply not “real world viable” as a modeling ground rule. Doing so would imply that IRA would be buying into sanctioning operational and strategic policies saying a banking industry averaging at least as risky as a junk bond was "normal". That’s just far too much rope to give to banking institutions that are supposed to operate safely and soundly. The notion fails the common sense test.

Silence. That pretty much ended that phone call.

Now some stress testing numbers to keep in mind.

  • There are rougly 885 publicly traded banks.

  • There are around 5,000 regulated bank holding companies.

  • There are 8,500'ish individual FDIC reporting bank units.

  • IRA publishes a detailed Bank Stress Rating for every one of the above for every quarter going back to 1995.

  • Each bank is tested using identical criteria.

  • The most recent reports cover the operating period ended December 31, 2008. Anyone can buy one for $50.00 on our website.

  • Our indexing methodology uses an analytical census of all active institutions in order to locate accurate behavior distributions.


Why bring this up?

The U.S. government is working on creating stress tests for a group of supposedly "mission critical" banks. There's a rule in analysis that when in doubt people do what they know. You have to watch out for that. Given the people working on this there's a good chance that the risk-stress methodology is likely unique per bank akin to the construction of an elaborate (that usually means opaque) boutique multi-asset class security with many of the risk inputs supplied by statistics provided by the bank being examined. Basel II called this type of thing the Internal Research Based (IRB) approach.

Some reminders are in order. IRB was meant for use by healthy institutions able to prove the empirical validity of their inputs and the stability and accuracy of their methods. Also IRB's are supposed to be accompanied by external research benchmarks and include an explanation of the differences, if any. The process is meant to promote counterparty confidence.

No comments:

Post a Comment