How much do these things overestimate usually?

sabretoothtigger

I read the government stress tested the banks in 2004 with a 'wargames' scenario. They found out that northern rock and Hbos would fail.

Their conclusion was to do nothing because this was all just a maybe

Reminds me of the cuban missile crisis, also a 'maybe' but with everyone in the white house smoking opium instead





http://link.ft.com/r/2SRI11/HCGN5/RUSJO/R4XWR/X9ZY0/PJ/h

2005 - Mervyn King, Bank governor, alluded to the war games in a 2005 interview with the FT, saying the Bank had looked at a situation in which “there could be a problem in a particular institution which isn’t terribly big, which may for completely unpredictable reasons turn out to pose a liquidity problem to a very big institution”.

tomterm8

lostinrates wrote: »

Presumably, a worst case scenario suggests it is an overegged omlette of a figure they use to stress test...so how much do FSA and similar bods generally over estimate the figures by? is there an average?


http://uk.reuters.com/article/businessNews/idUKTRE54R1SX20090528

I come from a computing background. In computing, a stress test is where you break a computer program in as many ways as you can possibly think of. Whereas, what you are describing is something very different... more akin to performance testing, where you set a acceptable tolerance and see if the program will still work.

If you look at this as a performance testing exercise, it doesn't make much sence since there is little or no historical data to tell you how bad this current mess could turn out to be. From a statistical viewpoint, what you would do is look at the historical data and go for something maybe a standard deviation worse than any previous recession. As it is, until an inflection point is reached performance testing is more or less meaningless.