Modelos de tasas de interes en Chile: una revision

In this work we revise Levels Models of the Interest Rates in Chile. In addition to the traditional Level Models by Chan, Karoly, Longstaff and Sanders (1992) in the USA, and Parisi (1998) in Chile, by Maximun Likelihood method we allow that the conditional volatility also include the unexpected information processes (GARCH model) and also that the volatility be function of the level of the interest rate (TVP-LEVEL model) as in Brenner, Harjes and Kroner (1996). For this we use market yields from the Bonos de Reconocimiento instead the monthly average yields from PDBC auctions, and enlarging the size and the frequency from the sample to 4 weekly yields with different terms to maturity: 1 year, 5 years, 10 years and 15 years. The main results from the study can be summarized in that the volatility of the unexpected changes in the rates depends positively on the level of the rates, especially in the TVP-LEVEL model. We obtain mean reversion evidence, such that the increments in the interest rates were not independent, contrary to that obtained by Brenner et al. in the USA. The LEVELS models are not able to adjust appropiately the volatility in comparison to an GARCH(1,1) model, and finally, the TVP-LEVEL model does not overcome the results from the GARCH(1,1) model.