Dealer polling in the presence of possibly noisy reporting

The value of a vast array of financial assets are functions of rates or prices determined in OTC, interbank, or other off-exchange markets. In order to price such derivative assets, underlying rate and price indexes are routinely sampled and estimated. To guard against misreporting, whether unintentional or for market manipulation, many standard contracts utilize a technique known as trimmed-means. This paper points out that this polling problem falls within the statistical framework of robust estimation. Intuitive criteria for choosing among robust valuation procedures are discussed. In particular, the approach taken is to minimize the worst-case scenario arising from a false report. The finite sample performance of the procedures that qualify, the trimmed-mean and the Huber-estimator, are examined in a set of simulation experiments.