The default free short term interest rate is one of the most commonly researched economic variables. It directly influences the short end of the term structure and, thus, has implications for valuing fixed income securities and derivatives. Furthermore, it is a general reference point for asset pricing on the basis that expected equilibrium returns on risky assets are expressed in terms of excess returns relative to the risk free rate.
From a macroeconomic perspective, the short rate serves as an important input for business cycle analysis through the cost of credit, and its dynamics are to some degree governed by the stance of monetary policy and inflationary expectations. Given the vital role played by short term interest rates in both the financial market and the economy, an enormous amount of work has been directed towards modelling and estimation of the short rate dynamics in the past three decades. Be that as it may, little consensus exists amongst financial practitioners about the appropriate choice of a short rate model both from a theoretical perspective and empirical application.
There are many contending short term interest rate models which have been developed in the literature. The leading theoretical models specify continuous time processes for the interest rate following the seminal work of Merton (1973) on the arithmetic Brownian motion representation. This specification, however, has been criticised for allowing negative interest rates and provides only, at best, a rough approximation to the actual process. The negative interest rate problem is overcome by Vasicek (1977) who imposes a mean reversion (or stationarity) condition in the short rate model.
Motivated by the observation that the discrete interest rate data display strong heteroskedasticity, Cox, Ingersoll and Ross (1985) (CIR) develop the square root model of short rates which allows short rate volatility to peak with interest rate levels, the so called mlevel effectm. Both the Vasicek and CIR models provide closed form solutions and have been widely applied to discrete time data on short term interest rates. Nevertheless, more flexible empirical specifications have been sought with the aim of obtaining an adequate characterisation of the actual short rate process. This has led Chan, Karolyi, Longstaff and Sanders (1992) (CKLS) to consider estimating the exponent parameter measuring the degree of level dependence in short rate volatility.
Download
Predicting Short Term Interest Rates: Does Bayesian Model Averaging Provide Forecast Improvement?
