Statistical analysis of determinants of household food insecurity in post-conflict Southern Sudan.

Abstract

Hunger and food insecurity has remained an endemic problem in Southern Sudan for the last three decades. Lack of a “gold standard” measure for determining causes of household food insecurity is well documented in the Food Security literature and the chase is still on for universally agreed standards. However, the Comprehensive African Agriculture Development Programme (CAADP)1 Framework for African Food Security (FAFS) has outlined four categorical measures for structured monitoring of household food insecurity, which are yet to be rolled out for implementation by country-level Food Security programmes. Analysis of a national household survey dataset has not been done using robust logistic regression techniques for statistically determining the factors influencing food insecurity in Southern Sudan. If such attempts are made, national food security programmes and the government statistical agency are not formally made aware of the results or do not own them. Hence, the agency has continued to lack institutional capacity to adapt the tools and techniques. This project attempts to explore the use of robust statistical techniques featuring the Ordinal Logistic Regression procedures of SPSS for analysing the Sudan Household Health Survey (2006) dataset and determine the strengths and magnitude of relationships of nineteen independent variables in predicting categories of food consumption scores. Food Security experts and international organisations, have regarded Food Consumption Scores as a proxy measure of Food Insecurity. Twelve factors were found to statistically determine food consumption. It is, therefore, ascertained that if this form of analysis were carried out immediately after the survey was completed it would have enabled prediction of the outcome of food insecurity in Southern Sudan for at least the following year. Nevertheless, the study found out that the same statistical modelling procedures could be adopted in similar national surveys. Indeed the study provides a basis for creating an institutional memory for statistical agencies to carry out similar analysis and thereby reducing data processing time without due reliance on outsourced international expertise.