The basic business of banking is lending money to companies and people. That business is risky but that is the job: If you are good at banking you make profitable loans, if you are bad at banking you make unprofitable loans, etc., or that is the idea anyway.
Regulators, meanwhile, are in the business of trying to prevent you from blowing up your bank, which they do largely by capital regulation: Rather than decide which loans you should make, which is your job, regulators decide how much capital you need to have as a cushion against losses on your loans.
In the abstract, you should have more capital to cushion against losses in risky loans than in safe loans. But who decides which loans are safe and which are risky? There are no satisfying answers; the main unsatisfying answers are:
- No one (which would encourage banks to make risky loans);
- Ratings agencies (often the pre-crisis answer, but everyone hates them now and Congress has required that they be written out of capital regulation);
- Regulators (why should they be better at assessing risk than bankers?); or
- Bankers (ooh I bet you can write this parenthesis yourself).1
The answer for U.S. regulation is more or less "bankers, with some oversight by regulators": Big banks will be able to use what is called the "Advanced Internal Ratings-Based approach" to assess how risky their loans are, and that assessment feeds into how much capital they need to hold against their loans.
So I suppose let's be cynical about that shall we? Here is an interesting Federal Reserve discussion paper that looks at how nine large U.S. banks are doing, specifically as to how they assess the risk of their corporate loan books for Advanced Internal Ratings-Based risk capital. The headline I guess is that (1) banks have a wide dispersal in the loss-given-default (LGD) that they assign to loans, that is, the amount of money they expect to lose if a loan defaults, (2) that seems to be systematic -- that is, some banks tend to have lower LGDs on all their loans than other banks, and (3) it matters for capital:2
Our findings imply that, if all banks had the same portfolio of loans, then the bank that sets the highest LGDs would be required to hold approximately twice a much capital for regulatory purposes as the bank that sets the lowest LGDs. Thus, the differences in LGDs estimated in the previous section could have a strong impact on Basel II minimum regulatory capital.
A similar international exercise earlier this year, looking not at loans but at trading portfolios, found that the banks with the highest risk estimates would have to hold 10 times as much capital against the same portfolio as the banks with the lowest estimates, so I guess this 100 percent difference doesn't look too bad by comparison.3
I guess a boring thing to think here is "Bank 1 (the bank with the lowest estimate of LGD -- the Fed doesn't give names) is super lax and risky and should have to raise its estimates to be in line with other banks." But that is a really boring thing to think! One possibility is that Bank 1 thinks corporate loans are pretty safe, relative to other opportunities, while Bank 8 (with the highest LGD estimate) thinks they're really risky. So Bank 1 is doing a lot of corporate lending and less, I dunno, mortgage lending, while Bank 8 has cut back on corporate lending to spend more time warehousing aluminum or whatever.
Or not, I mean, maybe Bank 1 is trying to get away with calling everything safe and holding as little capital as possible, that totally might be what's going on here.
The interesting thing is that the banks seem to be making economic decisions based on their risk assessments. The Fed authors look at loan share -- that is, the percentage of a syndicated loan that a bank will take up and hold on its books. And it turns out that loan share is negatively correlated with high LGD: The riskier a bank (idiosyncratically) thinks a loan will be, the less of that loan it will hold.
Which, duh, but it's not necessarily the result you'd expect if you thought that banks' risk-based capital models were about cynical gamesmanship.4 The alternative conclusion is that banks' risk-based capital models are about making good lending decisions:5 That banks really want to make more safe loans and fewer risky loans.
It's just that they have different ideas about how safe or risky loans are. That's a good thing! A world with a monoculture of risk assessment -- where everyone buys stuff because uniform capital regulation says it's safe -- is a lot like the 2008 world of risk-free mortgage-backed securities, or the more recent world of risk-free European government debt. Those worlds actually turned out to be pretty risky. Giving banks some leeway to do their job of risk assessment might be a lot safer, even if their actual assessments end up disagreeing.
1 There's also an answer that's like "the market" -- like, use credit default swap spreads or interest rates or whatever to measure risk -- which has its own more complicated issues. But that's hard to use with non-traded loans, where judging risk based on interest rates is sort of the same as saying the bankers decide.
2 Technically it doesn't, because U.S. banks' "minimum regulatory capital is still determined by Basel I rules," which are more mechanical, and they can't "start using their internal estimates of risk parameters to calculate capital requirements until regulators approve them, and none of these banks have yet been approved." But eventually. Presumably this paper is part of the pondering about approving those models.
3 Err. So actually that study found a 2.5x disparity between the highest and lowest banks for one simple portfolio, an equity index future. You might think to yourself that an equity index really should have more volatility -- more dispersion of outcomes -- than an investment-grade corporate loan. (Cf.) So you might find it weird that the variation in equity-index risk measures is only a little bit bigger than the variation in loan risk measures. But, meh, there's a lot more statistical data on equity indices than there is on like each individual bank loan.
4 Oh I mean sure maybe it is:
Our findings in this section are consistent with banks participating more heavily in loans that they consider less risky. Thus, these findings suggest that the systematic differences in LGD that we found in Section 3 actually represent differences in the losses that banks expect to incur if their borrowers default on their loans. Still, we recognize that there are at least two alternative explanations for this negative relation between LGD and loan share. First, banks may also be responding to regulatory incentives that require them to hold less capital for such loans. For instance, if a bank convinces its regulators that it should assign to a loan a lower LGD than this bank assigns to similar loans, then this bank has an incentive to hold a larger share of this loan, even if this bank believes that this loan is not safer than others in its portfolio. Second, there may be reverse causation between LGD and loan share, as banks have a stronger incentive to reduce the LGDs of the largest loans in their portfolios. This happens because the amount of capital that banks must hold for loans is proportional to LGD and to the exposure at default (EAD).
More strictly, "that banks' risk-based capital models are connected to banks' actual risk models that they use to try to make good lending decisions," or whatever. Probably.