Modelling Bounded Panel Global Food Insecurity Indicators with Correlated Random Effects
Francis Ayiah-Mensah
*
Department of Mathematics, Statistics and Actuarial Science, Takoradi Technical University, Sekondi-Takoradi, Ghana.
Emmanuel Harris
Department of Statistics and Actuarial Science, Kwame Nkrumah University of Science and Technology, Kumasi, Ghana.
Michael Asare Bediako
Department of Mathematics, Statistics and Actuarial Science, Takoradi Technical University, Sekondi-Takoradi, Ghana.
Vivian Nimoh
Department of Mathematics and Computer Studies, Holy Child College of Education, Takoradi, Ghana.
Emmanuel Ayitey
Department of Mathematics, Statistics and Actuarial Science, Takoradi Technical University, Sekondi-Takoradi, Ghana.
*Author to whom correspondence should be addressed.
Abstract
Food insecurity has become a major issue of concern in different parts of the world, but empirical analysis is frequently limited by the statistical analyses that do not consider the limited and heterogeneous character of prevalence statistics. This investigation seeks to revisit the world patterns in extreme food insecurity through a coherent statistical panel model that focuses on inference and policy application. In particular, the research is aimed at estimating global and country-specific time dynamics, measuring the unobserved heterogeneity and determining the existence of a homogenous global trend once methodological shortcomings in the previous studies are overcome. A beta mixed effects model of correlated random intercepts and slopes was estimated using a fractional response panel model with more than 29,000 observations on a global country-year panel of more than 20 years. The study findings indicate that there is no statistically significant global linear trend of serious food insecurity prevalence (β_year= 0.0029, p = 0.323) in contrast to the findings of previous linear panel studies. Significant between-country heterogeneity was observed, and there was a positive relationship between baseline prevalence and change across time (p = 0.36), which implies country-level heterogeneity through structural differentiation. The suitability of the proposed approach is supported by the model diagnostics and fit statistics (AIC = -33,488.6; BIC = -33,447.6). The novelty of this study is that it combines the statistical assumptions with limited prevalence information and directly models the heterogeneity of countries that are correlated. The research will help in enhancing the empirical basis of food insecurity monitoring in the world, and thereby a more plausible assessment of achievement of Sustainable Development Goal 2, as well as making evidence-based policy to eliminate severe food insecurity in the world.
Keywords: Severe food insecurity, beta mixed effects regression, global panel data, country-level heterogeneity, bounded outcome modelling, food security measurement