Time series for currency exchange rate of the Brazilian Real

Abstract

In our global village the currency exchange rate of a country is considered by international investors as an important yardstick for measuring the health of its economy. In the paper an analysis is made of the Brazilian Real using three different methodologies: the Box-Jenkins or SARIMA; exponential smoothing; and a backpropagation neural network trained by the Levenberg-Marquardt algorithm. Not surprisingly, our study indicates that given the same input data different paradigms yield different results. However, presumably due to the intervention of the Central Bank, the time series exhibits quasiperiodic behaviour. Extrapolations are made into the future. Possible implications are discussed. Our methodology can be applied to any currency extant.