This article uses linear and nonlinear diffusion index models to forecast, one step ahead, the quarterly growth rate of Brazilian Agricultural GDP. These models are composed by common factor which allow a significant reduction in the number of the original explanatory variables. After comparing the forecasts of these two models between themselves and with the ones generated by an AR model, used as benchmark, one comes to the conclusion that among the diffusion index models, the nonlinear model with a threshold effect presents a small improvement, in terms of predictive efficiency, in relation to the linear and the AR models.
agricultural GDP; Diffusion Index Model; forecast; nonlinearities
