Forecasting interest rates using shifting endpoints in small open economies: evidence from Canada and the UK

Abstract

This thesis examines the forecasting performance of widely used interest rate forecasting methods for small open economies such as Canada and the UK. In particular, I run a horse race between standard models and the models using shifting endpoints to see whether results for the US extends to small open economies. In this setup, three time-varying parameters, interpreted as factors corresponding to level, slope and curvature, are allowed to have shifting long-run means rather than a constant mean. The shifting endpoints are introduced by exponential smoothing of factors, and the connection of factors with certain macroeconomic variables and inflation expectations. In comparison to the random walk benchmark, allowing for shifting endpoints in yield curve factors offers significant gains in out-of-sample forecasting accuracy. Moreover, results suggest that there is a strong evidence that the US as a global economy has a spillover effect on the term structure of interest rates of Canada and the UK.