Combining term structure of interest rate forecasts : the Brazilian case

Issues like structural breaks and misspecification biases make it difficult to find a term structure of interest rates forecast model that dominates all competitors. Focusing on Brazilian data, this paper aims to identify the existence of combining methods that provide superior performance than individual models. Empirical results confirm that it is not possible to determine an individual model that consistently produces superior forecasts. Furthermore, the relative performance of these models may vary over time. The problems of using individual models may be reduced by applying forecast combining schemes. The empirical results show consistent forecast gains of combining schemes over time. In particular, the longer the forecast horizon, the greater the contribution of forecast combination.