LIU Xiuli, XIANG Xin, QIN Minghui, DOU Yuxing
This paper first discusses the strategic and scientific significance of forecasting medium- and long-term grain demand. Then it summarizes the current status of domestic and foreign research from four aspects, including the characteristics and trends of grain consumption in China, the main impacting factors of grain consumption, the forecasting methods for the total grain demand, and categorical grain demand. Then the paper analyzes the reasons that lead to the apparent differences among the existing prediction results and unsatisfied forecasting accuracy. First, it is difficult to grasp the changing patterns of residents' dietary structure in the medium and long term in different regions with various income levels. And the highly aggregated grain (food) classifications cover up the critical trends of the changing dietary structure of residents. These affected the prediction accuracy. Second, the main impacting factors of grain demand were not comprehensively considered and quantitatively analyzed in the models, such as the age and gender structure of the population, dietary structure, income level, food production transition, food waste, etc. Third, the differences in designed income level, price elasticity, the conversion rate from feed grains to animal foods, food waste rate at the consumption stage, and the neglect or simple estimation of residents' eating out had caused significant differences in prediction results. Finally, this paper puts forward the following outlook on medium and long-term grain demand forecasting research. 1) Some interdisciplinary and cross-sectoral frameworks should be established. More attention should be paid to new trends, impacting factors of grain demand, and integrating key factors of interdisciplinary and cross-sectoral frameworks. 2) Emphasis should be placed on combining text mining, machine learning, and other data processing methods for statistics and calibration of relevant data, and giving full play to the role of multisource data in forecasting. 3) Based on the forecasting model for grain demand, we should simulate the impact of changes in population policies, grain prices, taxation and subsidy policies, etc. on medium- and long-term grain demand and grain security to provide a policy basis for formulating appropriate medium- and long-term grain supply plans.