Food insecurity and childhood obesity: beyond categorical and linear representations

Previous work on the relationship between food insecurity and childhood overweight has lead to a wide array of answers – some have found a positive relationship, others no relationship, and still others a negative relationship. This previous work has shared one thing in common – all have used parametric models. In this paper we move beyond parametric models by using non-parametric models. With data from the 1999-2002 National Health and Nutrition Examination Survey (NHANES) and a wide array parametric methods, we find evidence across different samples of a positive relationship, no relationship, and a negative relationship between childhood overweight and food insecurity. When we turn to non-parametric methods, however, this ambiguity across samples is not as prevalent. Instead, across different samples, we find (a) increases in the probability of food insecurity in the middle of the BMI distribution, (b) increases in the probability at the very high end of the BMI, and (c) no relationship across the entire distribution. We present some parametric models that roughly mimic these relationships. Our results indicate that efforts to reduce food insecurity will either have no impact on childhood overweight or would lead to reductions in childhood overweight.