Comparing Household Survey-Based Measures of Food Insecurity Across Countries: Case Studies in India, Uganda, and Bangladesh

Ongoing research is exploring the validity and reliability of household food security assessment through surveys that ask respondents about behaviors and experiences known to characterize households having difficulty meeting their food needs. The Food Security Measurement Project in the United States of America has developed and tested a food security survey module for domestic use. Statistical methods based on the Rasch measurement model (a non-linear factor analytic model) are used to combine responses to questions in the module into a scale that measures the severity of food insecurity. The food security module and scale are in regular use in household surveys in the U.S. both for research and monitoring purposes. The U.S. Food Security Survey Module is also being adapted for use in low-income populations in several other countries. This paper examines data collected in three such adaptations: A study of child labor in Orissa, India; a general household survey in Kampala, Uganda; and a study of participants in an income-generation program in Bangladesh. Rasch-model-based methods are used to assess the suitability of the food security items in each study for scale construction and to assess the internal validity of scales based on the selected items. One of the strengths of the Rasch model is the ability to make scales comparable across surveys based on the presence of several equivalent items in the surveys. The feasibility of making such comparisons between surveys conducted in different countries, cultures, and languages is explored by attempting to adjust each of the three test scales to be comparable to the U.S. scale. The paper demonstrates the methods used to assess items, construct a scale, select appropriate thresholds for categorizing households as to food security status, and adjust the metric of the scale in order to make the scale equivalent in meaning to that in another country.