| Abstract: | This paper investigates alternative forecasting methods when few observations are available. An illustration is provided by Spanish monthly maize prices after Spanish accession into the EC. Sophisticated multiple-equation models are difficult to specify in situations of limited data, and simpler models have to be considered. In this paper, several individual and composite forecasting methods are compared, based on 24 one-period-ahead forecasts generated from these models. Results based on different quantitative and qualitative measures show that composite forecasting methods are more accurate. In situations where severe multicollinearity exists, forecasting performance is improved by modelling this problem explicitly. |
