I’ve been doing stress tests for a long time; first at a cell-phone company, on their networks. Later, moving over to finance, with research for capital reserving confidence intervals, and later with other things.
The FDIC released it’s stress testing numbers, and everyone is puzzled at how low they are. Include me in that. The “More Adverse” stress case scenario should represent a 10% likelihood. They have unemployment at 10.3% under this 10% likelihood scenario. That strikes me, and everyone I’ve talked to, as very low. I’d back-of-envelope the 10% scenario more at 12-13%, or at least 11%.
There are two ways people often approach these stress tests – (1) is to take a bad case and see how bad the results it generates are afterwards. “If unemployment is 12%, we can’t make our capital reserving needs by $50m.” (2) is to control the other way – take a “how bad” aftermath, and then see what case it takes to get there. “If unemployment is 11.1%, we can make our capital reserving needs by $50K; if it is 11.2%, we are under $100K.” Not to sound conspiratorial, but I get the sense that these numbers have that “this is the worst the number can be without breaking it” trickery thing engineers often do to appease management. They take the worst (2) that still passes, and then back that number into (1). The 10.3% reeks of that hand-waiving magic.
I worry that this unemployment number, as well as the rest, has been cherry-picked to not break too much. Maybe bottom-out citi, but leave all the rest in place. We shall see….