In 2001, the Bureau of Agricultural Statistics (BAS) and the Statistical Research and Training Center (SRTC) conducted a project to develop an Early Warning System (EWS) that could accurately predict the volume of agricultural commodities in the short run. Following this initial study on EWS, a supplemental study was conducted which employed more advanced methods especially in statistical modelling. EWS is primarily a set of well-defined methods that are capable of making short-run forecasts of agricultural commodities. Short-term numerical estimates were generated to aid in policy making and production planning. In this study, three general methodologies were applied: Survey-Based Method, Statistical Modelling, and Delphi Technique. Each method was consistently used in forecasting the agricultural goods selected for the study. Survey Method intended to gather direct forecasts from surveys while statistical modelling involved applying regression models and other useful techniques in forecasting commodity supply. More complex models such as the Univariate Box Jenkins (ARIMA) technique, Exponential Smoothing, and Life Cycle Model were tested and analyzed. On the other hand, the Delphi technique served as more of a supplement to the two aforementioned methods. This technique involved getting expert opinions on the volume of production for the next two quarters (6 months). These techniques were applied on nine selected commodities which will also comprise the EWS. Quality and accuracy of the techniques were compared using forecast evaluation methods (Average Deviation, Root Mean Square Error, and Mean Absolute Percentage Error). The study concluded that regression models produced best forecasts for palay and corn. Conversely, exponential smoothing generated more accurate forecasts for pork.

Keywords:

  Request for Download