In Brazil and the world, much of the investment portfolios of insurers and complementary pension funds consists of securities issued by governments. Thus, the returns of these papers account for the relevant part of the profitability of these portfolios. This work aims at analyzing the impact of the incorporation of economic data (interest rate, output gap and inflation rate) in the forecast of the Term Structure of Interest Rates in four countries: USA, UK, Brazil and Chile. Different models of vector autoregression (VAR) are tested and compared with other models such as random walk and autoregressive models (AR), in terms of performance of out-of-sample forecasts.