An Evaluation of the USDA Food Security Measure with Generalized Linear Mixed Models

Over the last decade, new information has been developed and collected to measure the extent of food insecurity and hunger in the United States. Common measurement of the phenomenon of hunger and food insecurity has become possible through efforts of the U.S. Department of Agriculture (USDA) to develop a set of survey questions that can be used to obtain estimates of the prevalence and severity of food insecurity. This paper takes a closer look at the measurement of food insecurity and the effect of household variables on measured food insecurity. The effects of demographic and survey-specific variables on the food insecurity/hunger scale are evaluated using a generalized linear model with mixed effects. Data come from the 1995, 1997, and 1999 Food Security Module of the Current Population Survey. The results generally validate the model currently used by the U.S. Department of Agriculture. In addition, our approach makes it possible to consider the effect of demographics and several survey design variables on food security among measurably food-insecure households. The analysis of the expanded model with the 1995 data finds results similar to those reported based on the Rasch model used by the USDA. Even though the sample size was reduced and a number of screening and questionnaire changes were introduced in 1997 and 1999, the results for those years appear mostly unchanged and confirm the robustness of the scale in measuring food insecurity. There is some evidence that interpretation of questions may vary among different demographic groups.