Composite indicators for famine early warning systems.

Traditional famine early warning systems use a host of indicators to predict food crisis situations, from rainfall and increased rate of marketing of household durables to the behavior of birds and animals. Although many of these indicators are valid in general, limited understanding of the sensitivity and specificity of the distress signals makes food crisis prediction a highly subjective exercise. In order to make the system more effective and credible, we need to identify a limited number of 'composite' indicators, which naturally summarize most relevant food-related information contained in the specific predictors of food crisis. Considering the chronology of the food production and consumption chain, three composite indicators specific to three different stages of the chain have been identified. The satellite data based Normalized Deviation of Vegetative Index (NDVI), prices of major food grains, and malnutrition rates are found to be correlated not only with the quality and quantity of inputs of this process but also with the final outcome. Both NDVI and price data are widely used as important predictors of food crisis by famine warning systems. What we have demonstrated is that improved sensitivity of the indicators is likely to be due to their inherent capability of summarizing information from various specific measures. Child malnutrition rates also summarize inputs and outputs of the food consumption process very effectively, and therefore should be able to predict community level food crisis in an efficient manner. The empirical results confirm this conjecture by showing that malnutrition rates can predict food crisis probability three months into the future with a high degree of specificity. The use of 'composite' indicators not only simplifies the problem of aggregation, but is also likely to yield forecasts that are highly specific and sensitive.