Doubts on LIBOR

Yesterday’s Wall Street Journal has a story about bankers questioning the reliability of Libor. Last month, the BIS Quarterly Review discussed the issue at length with lots of data and some amount of econometrics:

• Interbank rate fixings during the recent turmoil by Jacob Gyntelberg and Philip Wooldridge, and
• What drives interbank rates? Evidence from the Libor panel by Francois-Louis Michaud and Christian Upper)

Gyntelberg and Wooldridge find that: “The US dollar market stands out for being the one market where Libor rose by substantially less than similar fixings during the stress period. The average spread between Sibor and Libor widened from about zero in the normal period to 2 basis points in the stress period, and the spread between H.15 and Libor widened from -1 to 7 basis points.” Of these, Libor is the only one that is used as a reference rate for swaps and other derivatives while H.15 (named after the table number in which the Federal Reserve publishes the data) is the only one which is based on actual transactions.

Since H.15 was 7 basis points above Libor, it does confirm that banks were actually paying more than what the Libor panel was quoting during the Libor fixing. Though 7 basis points is not a trivial difference when trillions of dollars of debt is referenced to this rate, it is much less than the 30 basis points being mentioned in the WSJ article. Moreover, there is a different way of looking at whether Libor is too high or too low and that is by comparing it to the Overnight Index Swap based on overnight rates. Under the expectations hypothesis, the OIS and Libor must be equal. Michaud and Upper find that during the crisis, Libor exceeded the OIS by 50 to 90 basis points. From this perspective, the problem is that Libor was too high, not that it was too low.

Gyntelberg and Wooldridge re-estimated Libor using a bootstrap technique instead of the trimmed mean used by the BBA and found that the bootstrapped estimate is not significantly different from Libor. “Moreover, the 95% confidence interval around the bootstrapped mean loosely corresponds to the interquartile range in the Libor panel ... In other words, the bootstrap technique indicates that 19 days out of 20, the design of the Libor fixing produces an estimate that is close to the true interbank rate. This is the case even during the stress period.”

They also argue that “many of the banks on the US dollar Libor panel are also on the euro Libor panel, and there are no signs that signalling distorted the latter fixing.” I do not find this argument convincing because Chapter 2 of the same issue of the BIS Quarterly Review provides a chart on page 21 which highlights how lopsided the US dollar interbank market has become. The data suggest that US banks have raised more dollar deposits from non banks than they have lent to non banks while the position is the reverse for European banks. The result is that European banks have probably borrowed about half a trillion dollars from US banks in the short term inter bank market. In times of heightened concerns about counter party risk, positions of this size become difficult to roll over and poses huge systemic risk. The incentives for strategic quoting are much higher in the dollar market than in the euro market.

All this has a bearing on the common assumption made in recent years in the credit derivative market (both in the theoretical literature and in the practicing world) that the correct risk free rate is the swap rate (essentially Libor) and that the TED spread is essentially a liquidity premium and not a credit spread. Since the crisis of 2007 and 2008 is simultaneously about liquidity and about counterparty risk in the inter bank market, all the turmoil fails to throw light on this hugely important issue.

Posted at 1:10 pm IST on Thu, 17 Apr 2008         permanent link

Comments