STATISTICAL WEIGHTING BASED MEASUREMENT FOR FOOD QUALITY AND SAFETY DIMENSION OF FOOD SECURITY AND EFFICIENCY ASSESSMENT

Abstract

The appropriate application of statistical approaches on a data set brings powerful results and insights for solving food security problems for current and future generations. Moreover, it provides resilient integrated measurement methods that seek to be a reference for governments and policymakers. In this spirit, with the focus on the quality and safety of food indicators introduced by the Global Food Security Index (GFSI), this research applies Variable Importance in the Projection approach (VIP) to statistically assign the importance of multiple indicators on achieving a certain level of food security and comparing the results with the weights that are subjectivity assigned by a group of experts. Then the research studies the efficiency for 46 countries on achieving their certain level of food security using GFSI weighted DEA, VIP weighted DEA, and unweighted DEA models. The results showed that the weights assigned to the indicators using the variable importance in projection approach vary compared to the weights assigned by experts. Although this difference was observed, when using the same method for calculating the overall score, the Weighted Arithmetic Mean (WAM), the ranking slightly changes, and the changes do not exceed ±6 ranks. Moreover, the top countries remain to be Norway, United States, and Netherlands in all five years and despite the changes in the weights used. The results on the efficiency study using weighted DEA and unweighted DEA model showed that countries like Azerbaijan, The Czech Republic, and Slovakia has always been highly efficient countries despite the model of Data Envelopment Analysis (DEA) used. The efficiency scores have not been noticed to fall below 0.48 in all the models used. Comparing the variance between the models used, the VIP weighted DEA efficiency scores appear to be closer to the results of the unweighted DEA where the linear program assigns the weights based on the output, which was the prevalence of severe food insecurity in the population, in all the models in this research.