Forecasting the growth rate of Iranian agricultural sector (a comparison of univariate and multivariate methods)

The policy makers and economic strategists are trying to model the factors affecting the agricultural sector growth and to use them in the growth forecasting process. Today forecasting is regarded as an important instrument for economic policymakers. There are different methods used to forecast the economic variables. In this paper, the growth rate of Iranian agricultural sector is forecasted and the forecasting accuracies of univariate and multivariate methods are compared. The methods used in this paper include single exponential smoothing with trend, double exponential smoothing with trend, Holt-Winters additive algorithm, Holt-Winters multiplicative algorithm, auto-regressive integrated moving average process, vector auto-regressive approach and artificial neural networks. For univariate models, it was found that the artificial neural networks model, single exponential smoothing with trend and double exponential with trend have marginally better forecasting performance than those of the other methods in this group. Furthermore, for multivariate models the artificial neural networks forecast is more accurate than vector auto-regressive model.