This is the other problem with Libor. The only way to protect against a charge of favouritism is for the analyst to use the unvarnished sample mean: exclude nobody’s view and calculate the simple average. Using its relative size rather than its likelihood a priori of being correct as the basis for discarding any view violates the criterion of structural view independence, which requires impartiality to be built into the method for analysing human opinions.
The distorting effort on Libor of excluding outliers was revealed after Bloomberg published an article in September 2007 identifying Barclays’ Libor submissions as exceptionally high and questioning Barclays’ credit worthiness as a result. In fact, high inter-bank rates were actually realistic in the stressed financial conditions of late 2007, but such views were excluded from Libor as outliers, almost certainly depressing its value in consequence. Then after the Bloomberg article, Barclays lowered its Libor submissions to fit in.
Unfortunately, the Wheatley Review decided to stick with stated preferences and retain the practice of discarding half the sample through double-sided 25 per cent trimming, even though this provided no protection against the previous Libor manipulation, and may, indeed, have contributed to it.
But the use of stated preferences to estimate important financial parameters like Libors is problematical, especially when the banks had, and presumably have, incentives to falsify their opinions. The hunt must surely be on to find a safer, revealed preference technique for estimating Libors without the risk of rigging in the future.
Philip Thomas is professor of engineering development at City University
