Relationships between food insecurity and negative health outcomes for children have been widely documented. Previous research, spanning numerous academic studies, has found that children in households suffering from food insecurity are more likely to suffer from diminished psychosocial functioning, frequent stomachaches and headaches, worse health outcomes, increased odds of being hospitalized, higher levels of hyperactivity, greater propensities to have seen a psychologist, behavior problems, worse developmental outcomes, and higher levels of iron deficiency with anemia. These consistently negative health findings emerge from a variety of data sources, employ a variety of statistical techniques, and appear to be robust to different measures of food insecurity.
The previous work has made the implicit assumption that food insecurity is measured without error. In reality, however, this variable is imperfectly measured and patterns of measurement error are unlikely to conform with the classical assumptions. This study examines what can be learned about relationships between health outcomes and food insecurity status when the latter is subject to nonrandom classification error. In particular, the nonparametric framework allows the study of partial identification under corrupt samples given minimal assumptions on the error-generating process. Within this environment, the study estimates sharp worst-case bounds on conditional health outcomes that exploit all available information under the maintained assumptions.
To isolate the identification problem associated with potentially misreported food insecurity status, the study begins by assuming that food insecurity is reported without error. This is the implicit assumption that is imposed in all previous work on this topic. The study then estimates sharp bounds that impose no assumptions on the patterns of classification errors in a binary conditioning variable. Next, it estimates narrower sets of bounds for certain benchmark cases that impose structure on the reporting error process. In one case, the study considers the identifying power of an assumption that food insecurity misreporting arises independently of true food insecurity status. In another case, it considers the identifying power of an assumption that food insecurity is potentially underreported but households do not falsely claim to be food insecure.
For these analyses, the study uses data from the 1999-2002 National Health and Nutrition Examination Survey (NHANES), a survey designed to assess the health and nutritional status of adults and children in the United States. In contradistinction to other surveys that rely exclusively on interviews, the NHANES contains both interviews and physical examinations. Since food insecurity is rare among households above 200 percent of the poverty line, the study limits the sample to households with incomes below this threshold.
The examination of differences in the health status between food-secure and food-insecure children under the assumption of fully accurate reporting of food security status yields results similar to those found previously in the literature. For the examination of what occurs when food insecurity is potentially misreported, the study concentrates on two health outcomes—childhood overweight (children in food-insecure households are less likely to be overweight than those in food-secure households) and reports of very good or excellent health versus reports of good, fair, or poor health (children in food-insecure households are much less likely to be in the former category compared with children in food-secure households).
The study finds that claims regarding these statistically significant associations between food insecurity and negative health outcomes for children rely heavily on assumptions about the accurate reporting of food insecurity status. In the case of childhood overweight, the study finds that if about 2 percent of households potentially misreport food insecurity status, one can no longer assume that children in food-insecure households are less likely to be overweight. In the case of general health, the breakdown is not as sudden. But, even under the least restrictive assumption of no false positives, if more than 10 percent of households misreport food insecurity status, one can no longer say that children in food-insecure households are more likely to be in poor health.
Direct inquiries about this study to the Project Contact listed above.