Short and Long Memory in Equilibrium Interest Rate Dynamics

This paper analyzes a large class of processes for the short-term interest rate that are derived in a discrete-time equilibrium framework. The dynamics of interest rates and yields are driven by the dynamics of the conditional volatility of the state variable. Under appropriate parameter restrictions, interest rates derived in this framework are nonnegative. We study Markovian interest rate processes as well as more general non-Markovian processes that display short and long memory. These processes also display heteroskedasticity patterns that are more general than those of existing equilibrium models. We find that deviations from the Markovian structure significantly improve the empirical performance of the model and that the data support the presence of long memory. We also find that the data support heteroskedasticity patterns that are different from the ones present in existing equilibrium models.
[ - ]
[ + ]