The persistent need for food security pushed the development of dedicated agricultural forecasting systems that not only serve as guide for production, but also as a cautionary measure in case of threats to food supply. The Early Warning System (EWS), a project initiated by the Bureau of Agricultural Statistics (BAS) in collaboration with the Statistical Research and Training Center (SRTC), is basically a group of structured models that could generate short run (6-month) forecasts for agricultural commodities. The EWS will serve as a basis for policy formulation to ensure adequate food supply, proper market planning, and effective food distribution program. Nine agricultural commodities were initially selected for the development of the EWS. Such agricultural products included rice, corn, coconut, mango, banana, onion, broiler, egg, and swine. The researchers implemented three general approaches in constructing the EWS: Survey-Based Method, Statistical Modelling, and Expert Opinion Survey. The survey-based method involved collecting direct or derived forecasts from BAS surveys. The accuracy of survey-based methodology was evaluated by comparing average deviation of forecast from final estimates for a time series of crop production. For the statistical modelling approach, specific techniques were applied such as exponential smoothing, Univariate Box Jenkins ARIMA model, and Regression with Autocorrelated Errors. Each statistical approach was evaluated using forecast errors. Lastly, the Expert’s Opinion Survey engaged the outlook of industry experts in determining short term forecasts for agricultural goods. Across commodities, survey-based and statistical modelling techniques proved to efficient methods for forecasting palay and corn. On the other hand, both the survey method and statistical model resulted in unacceptable forecast deviations for crops such as onion, mango, and banana.

Keywords:

  Request for Download