Estimating the Term Structure of Interest Rate Volatility in Extreme Values

This paper proposes an extreme value approach to estimating the term structure of interest rate volatility, and shows that the volatility of interest rate changes is overestimated by the standard approach that uses the thin-tailed normal distribution. The volatility of maximal and minimal changes in three-, six-, and twelve-month T-bill rates is estimated over the late 1950s through the end of 1999. The empirical results indicate that the volatility of daily changes in short rates obtained from the fat-tailed generalized error distribution is almost the same as the volatility of the extremes obtained from the generalized Pareto distribution