Measuring household food security in a low income country: A comparative analysis of self-reported and objective indicators

Measuring food security in an accurate and cost effective way is important for targeted food relief and for designing anti-poverty programs. A number of studies have established different food security indicators as alternatives to calorie intake but none has conducted a comparative study of multiple indicators. We present a comparative analysis of 3 alternative indicators, (a) dietary diversity score (DDS) (b) self-reported food security and (c) land holding, as alternatives to calorie intake for prediction of household food security. We assess the reliability of the 3 indicators through their relation with household calorie intake, food and non-food expenditure, and nutritional status. We use a nationally representative cross sectional data consisting of 4,423 households from Bangladesh. We find no systematic difference in association with access to food (as measured by household food and non-food expenditure) among the alternative indicators. We also find that the land indicator predicts nutritional status better than DDS and self-reported indicator. We compare the ability of different indicators to predict calorie intake using a Vuong closeness test and find that DDS or any indexes of multiple indicators consisting of DDS can predict calorie intake better than others and more closely resemble the true model of calorie intake. We find the similar conclusion from out-of-sample prediction ability, lowest mean square error and mean absolute error, of alternative indicator. Finally, we try to find a discontinuous break point in DDS at different calorie intake threshold points which we can use as a cut-off points to identify food insecure household. We find no such structural points in DDS distribution