It's popular to say that financial markets and regulators have extremely short memories and so let's say it about these new Basel liquidity coverage ratio rule changes out today. But not in an annoying sneery way. I mean, in an annoying sneery way, but not the obvious one.
The story is that among the post-2008 Basel mechanisms for keeping banks out of trouble is the required "liquidity coverage ratio," which for each regulated bank:
- tots up how much cash is likely to go out the bank's doors in a crisis due to things like customers withdrawing deposits, derivatives counterparties terminating trades or demanding more collateral, corporate clients drawing down lines of credit, etc.; and
- requires the bank to hold liquid assets that it could sell quickly in a crisis to meet those demands on cash.1
Virtually everything there is a term of art, but "crisis" and "liquid assets" are particularly squishy. When the LCR was first proposed it had rather harsh ideas of what sort of crisis might affect liquidity, and a rather narrow conception of what assets might be liquid enough to be sold quickly and economically in a crisis. The news today is that Basel has relaxed that approach in a number of specific ways described here and listed here; the brief version is that the types of assets that can be counted toward "high quality liquid assets" has been dramatically expanded to include a lot of corporate and RMBS debt, the assumed outflows in a crisis have been reduced, and the LCR is now being phased in from 2015-2019 instead of going into effect all at once in 2015.
A lot of people think this is a good thing, as it will reduce the already significant demands on "safe assets" and make banks a little more willing to use balance sheet to lend and stuff. As is true of everything that banks like, you can also if you are so inclined easily find people who think it's a bad thing. There is no particularly Platonic right answer. Basically the exercise here is (1) imagine a bad situation and (2) see if the bank survives your imagined bad situation with a given mix of liquid assets; step (2) is a question of simple arithmetic while step (1) is determined entirely by the direction in which your imagination runs. There are good practical and social reasons for making your bad situation basically "2007-2008, but a little worse," and so most of the debate is over translating that notion into liquidity outflows and asset haircuts, but if you think that that notion is conceptually suspect I can't really prove you wrong. If aliens invaded France, SocGen's liquidity reserve would probably not be suited to the situation.
But whatever. The jarring thing for me was this first bit of the changes to the LCR announced today:
Expand the definition of HQLA ["high quality liquid assets"] by including Level 2B assets, subject to higher haircuts and a limit
- Corporate debt securities rated A+ to BBB– with a 50% haircut
- Certain unencumbered equities subject to a 50% haircut
- Certain residential mortgage-backed securities rated AA or higher with a 25% haircut
Aggregate of Level 2B assets, after haircuts, subject to a limit of 15% of total HQLA
Does anyone else remember when regulators were going to get credit rating agencies out of the business of officially passing on what securities banks can and cannot hold? Because, y'know, everyone thought the agencies were dumb and corrupt and stuff? The new LCR rules have at least three credit-rating-determined categories:
- AA- or better corporate bonds, with a 15% haircut2;
- AA or better RMBS, with a 25% haircut;
- A+ or worse corporate bonds, with a 50% haircut.
This should make investment grade (esp. BBB- through A+) bonds relatively more valuable, versus high yield bonds, than they were on Friday, say,3 based not on relative credit quality but purely on the interaction of regulatory and ratings-agency rules.
Isn't that strange? The argument against using ratings in setting bank capital is some combination of (A) "the rating agencies are dumb and corrupt" and (B) "credit ratings don't measure market risk"; the argument in favor is some combination of (C) "ratings measure the risk of default - i.e. permanent devaluation of bank assets - which is really what capital is designed to guard against" and (D) "well, do you have a better idea of how to measure that?" And so there is much fulminating against giving official power to ratings, and not so much done about actually stripping them of that power in capital regulation.
The argument against using ratings in determining bank liquidity requirements is even stronger: (A) the agencies are no less dumb or corrupt just because you use their ratings for something else, (B) ratings measure neither market risk nor liquidity - you can usually sell bonds from a $1bn high-yield issuance much quicker and at less of a discount to market prices than you can sell bonds from a $25mm investment-grade private placement, and (C) risk of default of the underlying bonds is not at all conceptually linked to the liquidity requirements: you care if the bank can come up with money today, not with whether its loans will pay off in five years. The high-yield corporate bond market, you may have heard, is quite robust now; it's true that it did worse during the last crisis than did the investment-grade corporate market,4 but then again it did better than the investment-grade CDO market so there's that. Even the ratings agencies will tell you that their ratings should be used only as predictions of default likelihood, not of market risk or liquidity, so it's weird that they're enshrined as official predictors of the latter. But as is so often the case, the main consideration is (D): do you have a better idea?5
1.The phrase "that it could sell quickly" gets at an important regulatory-conceptual point in this sentence of the new changes to the Liquidity Coverage Ratio formula: "Incorporate language related to the expectation that banks will use their pool of HQLA during periods of stress." The idea here is that a lot of Basel-y regulation works as follows:
- Notice that when Bad Thing X happens, banks might run out of Good Thing Y (capital, liquidity, etc.), and that if they did the consequences would be disastrous;
- Mandate that banks always have enough Good Thing Y (amount >= Amount Z) so that even if Bad Thing X happens, they won't run out of it;
- But then when Bad Thing X (or Bad Thing happens, banks end up with an amount of Good Thing Y that is less than the regulatorily required Amount Z, though greater than zero;
- So semi-disastrous consequences (bank shutdowns, death-spiral capital raises, fire sales, what have you) ensue in order to get them back to the required amount of Good Thing Y, meaning that the regulatory regime doesn't really prevent the bad consequences it's aimed at.
This is called, like, "procyclicality" or whatever. A rule that says "you have to have a pool of high quality liquid assets in case of stress, but you can actually sell them if the stress actually happens," avoids that problem, and is to be preferred.
2.See pages 13-14 here.
3.And lo, though it's kind of a wash:
4.Though more on the new-issuance side - irrelevant to our purposes - than in secondary trading. FWIW, per Bloomberg LQD <Equity> GV and HYG <Equity> GV, 20-day realized volatility of indexy junk bonds hit a high of 78% in October 2008, while the same vol for investment grade bonds was 65%. Different but not crazily different.
5.I mean sure, right? How hard can that be? Give every corporate bond or whatever a score from 1 to 10 based on objective quantitative measures like (1) amount traded per day over the last year, (2) size of issuance, (3) price volatility, (4) inclusion in indices, (5) number of dealers who quote it, (6) bid/offer spread on those quotes, (7) etc. Etc. can if you like include credit ratings and other stuff that historically correlates with liquidity - there is after all a reason that the investment-grade ETF I used in footnote 3 is called "LundefinedQ[ui]D"a - but ratings wouldn't be the primary factor. (To be fair, the actual LCR rules include non-rating liquidity factors in classifying bonds, but in sort of a vague and yes/no way rather than quantifying those factors.)
I assert without evidence that one could probably specify a formula with a better correlation to actual in-crisis liquidity than just credit ratings?Of course that would be computationally difficult, especially for small banks with weird holdings, so maybe you could contract it out. Like, ask Moody's and S&P to get into the liquidity-scoring business. Even if you think they're dumb and corrupt, you should still prefer to use their relevant liquidity rating instead of their irrelevant credit rating.
a.It's not but you know what I mean.