Application of Non Linear Models in the Determination of the Behaviour of Interest Rates in Kenya

Abstract

Empirical literature shows that stock returns could be non linear. However, studies on the non linear behavior of interest rates in developing economies are limited. This study aims at filling this knowledge gap by comparing linear and non linear models in predicting interest rates. The study compared the Random Walk Model, Moving averages Models, Autoregressive Models, Autoregressive Moving Average Models, Autoregressive Conditional Heteroskedasticity Models. The main variable for the study was the Treasury bills interest rate series. In Kenya, this is the Central Bank of Kenya three month Treasury bill rate. The study applied the monthly averages of the 91-day Treasury bill rate for the period between August 1991 and December 2011 which were obtained from the Central Bank of Kenya. The Akaike Information Criterion (AIC) and the Bayesian Information Criterion (BIC) also called Schwarz-Bayesian information criterion (SBC) were used to select the best fitting model from each type of models. The results indicate that non linear GARCH (2, 1) performs better than any other models. This is because it has the lowest AIC and BIC values among all the models tested. Therefore this study concluded that non linear models are better than linear models in predicting interest rates in Kenya. Thus interest rates are non linear