Title: Correcting estimation bias in regime switching dynamic term structure models
Authors: Liu Liu - The University of Manchester (United Kingdom) [presenting]
Sungjun Cho - Alliance Manchester Business School (United Kingdom)
Abstract: The small sample bias in the estimation of a regime switching dynamic term structure model is assessed and corrected for. Using the dataset from 1971 to 2009, there are two regimes driven by the conditional volatility of bond yields and risk factors. In both regimes, the process of bond yields is highly persistent, which is the source of estimation bias when the sample size is small. After bias correction, the inference about expectations of future policy rates and long-maturity term premia changed dramatically in two high-volatility periods: the famous 1979--1982 episode and the recent financial crisis. Empirical findings are supported by Monte Carlo simulation, which shows that correcting small sample bias leads to more accurate inference about expectations of future policy rates and term premia compared to before bias correction.