Resumen:
The standard approach to select among volatility models incorporates an important limitation: the non-observability of the true volatility. In order to overcome that limitation, a new approach to select the best volatility model is proposed in this paper. The procedure performs a competitive comparison of the ability of different volatility measures to forecast interest rates and term premia into the Term Structure of Interest Rates. The best volatility model will allow the best interest rate forecast. Term premia are estimated in a VARMA context using weekly data of continuous þ compounded yield to maturity of the Spanish interbank money market. Several naive volatility models such as historical volatility or exponential smoothing are analyzed. Although there was some variation across terms, the results show that the simplest volatility measure allows the best interest rates forecast.