Thursday, May 29, 2008

Defending Libor

(Felix Salmon) Carrick Mollenkamp and Mark Whitehouse got some pretty heavyweight backing for their Libor investigation today: before running it on the front page of the WSJ, they got sign-offs from Darrell Duffie of Stanford, Mikhail Chernov of London Business School, and David Juran of Columbia. In the blogosphere, their findings have been received uncritically by Ryan Chittum ("Lying about Libor" is his headline), Angus Robertson, Paul Jackson, and others.

But Alea is not convinced at all, and neither is JP Morgan, and neither am I.

One thing I find extremely suspicious is the fact that the WSJ's interactive graphic shows implied rates only back to August 2007, thereby only showing what's happened since the credit crunch hit. If they went back further, their methodology might be exposed:

Pre-credit crunch, pre-August 9th 2007, when OIS-Libor spread was below 10 bp, the Journal calculation would have resulted in a "risk-free Libor" below the OIS fixed, a proxy for a risk-free rate.

What's more, there are lots of places where banks actually borrow real cash, like the commercial paper market. Why would the WSJ try to use credit default swaps to gauge what cash borrowing rates should be, when they can look to something like the CP market instead? Clearly, I think, the answer is that the CP market wouldn't give them the answer that they're looking for.

Alea does a good job of explaining the theoretical weaknesses behind the WSJ's methodology. But my gut reaction that the methodology is flawed was based on none of those. Rather, I mistrust any calculation which assumes that since last summer there has been a clean and predictable and precise relationship between cash credit products, on the one hand, and credit default swaps, on the other. Yes, Libor is a borrowing rate, and yes, there is some kind of credit spread baked in to it. But to assume that Libor equals a risk-free borrowing rate plus a default-risk premium is silly and simplistic - especially when you don't back-test your model to a time when things were much less volatile.

It's worth remembering that the interbank markets are based on long-standing relationships which are necessary to any smoothly-functioning financial system. Yes, Citigroup's credit default swaps might be pricing in a relatively high probability of default, but that doesn't mean that Citi's counterparties will charge it a similar premium to insure against default risk, as the WSJ seems to think. Maybe they trust Citi more than the rest of the market does, or maybe they realise that any possible world in which Citi defaults is a possible world in which they've got much bigger things to worry about than their interbank lines.

Do I think that Libor is perfect? No. In this world, no spread measure is going to be perfect, especially at tenors of longer than a couple of weeks. But Libor is not nearly as flawed as the WSJ makes it out to be.

What the WSJ has done is come up with a marginally interesting intellectual conundrum: why is there a disconnect between CDS premia, on the one hand, and Libor spreads, on the other? But the way that the WSJ is reporting its findings they seem to think they're uncovering a major scandal. They're not.

JPM on WSJ's dodgy LIBOR analysis (Alea):

See previous post for background info.

The article uses spreads in the CDS market to conclude that Libor fixings are too low. We find the article to be deeply flawed and highlight below some of the main shortcomings of the analysis.

First, the methodology is based on too high a risk free rate which produces a large upward bias in the Journal’s measure of bank borrowing costs. For example, on April 16th, the WSJ calculates a risk-free rate of 2.47%. This is 50 bp higher than 3-month OIS on that date and 80 bp higher than 3-month GC. By adding bank CDS to this artificially high risk free rate, they generate a Libor rate of 2.97%, or 25 bp higher than the actual reported Libor of 2.73%. If instead, bank CDS spreads were added to OIS, GC repo or another more appropriate risk-free rate, the Journal article might have concluded that BBA Libor is fixing 25 or 50 bp too high rather than too low relative to the actual rates banks fund themselves at.

Second, the methodology inappropriately combines 1-year CDS with 3-month Libor in assessing the accuracy of Libor. Besides the obvious flaws in mixing rates and spreads of different terms, 1-year CDS should be expected to have much greater dispersion than 3-month Libor across banks because of the increased uncertainty associated with longer term credit risk. This increased dispersion produces unrealistically high 3-month borrowing rates using the WSJ methodology for some of the banks in the panel.

Finally, the implied rates calculated by the WSJ using 1-year CDS are well above the rate in virtually every short term financing market used by banks. Three-month commercial paper issued by financials has averaged 11 bp below BBA Libor fixings over the last year. As discussed in our recent research note, Eurodollar deposit data has averaged slightly above BBA Libor but the differences can easily be explained by a funding advantage that exists for the largest banks that make up the Libor panel.

As discussed in our recent research note, while Libor has a number of shortcomings, we don’t find evidence that it is biased too high or too low. Moreover, we expect any changes that the BBA makes to the definition of Libor will have virtually no effect on where it fixes relative to OIS over the long run. Instead Central Bank efforts to inject liquidity are likely to cause Libor to narrow sharply relative to the funds rate over the next several months.

JPM [thanks S... ?....]

No comments: