Thread by @dsquareddigest

I think the coronavirus crisis ought to have really undermined this kind of modelling. As we've seen, the bad thing about widespread bankruptcy isn't that banks might have enough capital - it's that everyone's bankrupt! voxeu.org/article/how-mu…

In a situation like the COVID shock, it's obviously appropriate for the government to try to ease the constraints and insulate the nonfinancial corporate sector. The government has to provide that sort of insurance - or at least it's massively more efficient for it to do so ...

(if you want bank sector capital to be the cushion that protects against shocks of that size, then a) you need to rewrite the bankruptcy code, b) 24 years out of 25, solvent borrowers are going to be paying a premium for this risk and c) cmon, b) isn't a credible equilibrium

Bank capital requirements fundamentally aren't there to absorb massive exogenous losses like those caused by pandemics or wars. They are for the most part there to try to mitigate *endogenously caused* business cycles, where banks lend too much.

That means that, beyond a certain level probably somewhere between the Basel 1 and 2 requirements, bank capital is simply a measure of the extent to which supervisors would really like to put lending restrictions in place but don't have the mojo to do so.

... which implies there's no tradeoff and therefore no optimum. the toy model in which capital is a buffer against random shocks drawn from a probability distribution doesn't describe how the industry works. To conclude with a proverb -

Banks don't fail because of high risk business. Everyone knows high risk business is high risk and takes care. Banks fail because of low risk business which ain't.

Envoi - this is a huge problem for @EBA_News this year, because they have to prepare a stress test which fulfills the dual goals of a) being a credible worst case pandemic scenario and b) being a useful basis on which to set capital requirements out to 2023. Good luck ...