The problem is what economists call the stigma of borrowing from the central bank. Imagine a financial institution that is facing a run but has good assets usable as collateral for a central bank loan. If all goes well, it will borrow, replacing the funding lost to the run; when the panic subsides, it can repay. However, if the financial institution believes that its borrowing from the central bank will become publicly known, it will be concerned about the inferences that its private-sector counterparties will draw. It may worry, for example, that its providers of funding will conclude that the firm is in danger of failing, and, consequently, that they will pull their funding even more quickly. Then borrowing from the central bank will be self-defeating, and firms facing runs will do all they can to avoid it. This is the stigma problem, and it affects everyone, not just the potential borrower. If financial institutions and other market participants are unwilling to borrow from the central bank, then the central bank will be unable to put into the system the liquidity necessary to stop the panic. Instead of borrowing, financial firms will hoard cash, cut back credit, refuse to make markets, and dump assets for what they can get, forcing down asset prices and putting financial pressure on other firms. The whole economy will feel the effects, not just the financial sector.
Pointer from Mark Thoma.
In effect, Bernanke is saying that you have to make firms that get into trouble want to be bailed out. If you make bailouts too painful for them, then “financial firms will hoard cash, cut back credit, refuse to make markets, and dump assets.”
I am not impressed by his reasoning. What he calls “stigma” is not a bug of the Bagehot principle. It is a feature.