Information and Accuracy in Interest-Rate-Risk Simulation (WP 94-7)

Abstract

Accurate interest-rate-risk modeling requires detailed information, but information is costly. By assessing the rate-sensitivity estimates from models that require differing degrees of detail, we calculate the marginal improvements in accuracy that result from marginal increments of informational detail. We limit the analysis to the instruments most commonly held by financial intermediaries.

Surprisingly, for most of those instruments, we can discriminate between model estimates with great precision. As a corollary, a high degree of model precision appears to be possible, provided that sufficiently detailed data are used. However, this precision holds only if the models' underlying assumptions are granted, and it is not easy to discriminate among alternative underlying assumptions. Therefore, model precision does not imply model accuracy.

Author

Mike Carhill