We use principal components to extract a time series for the Stochastic Discount Factor based on returns on Brazilian mutual funds that invest in fixed-income securities and foreign currency. This factor is then used to model the Brazilian Real/American Dolar Exchange rate using a Bivariate Generalized Autoregressive Conditional Heteroskedasticity-in-Mean (GARCH-M). Our empirical exercise based on asset pricing theory follows the methodology used in Chong, Chung and Ahmad (2002) and Da Costa et alii (2010). Our foreign currency depreciation forecasting error is 5,27% and we are able to say the correct variation sign in 57,5% of the time, during the January 2000 -December 2009 period. These results suggest to the literature that one should not omit the time varying second order conditional moments.
