Over one-in-ten U.S. households are food insecure. That is, these households include members who do not always have access to enough food for active, healthy living because their household lacks money or other resources for food. Children (18.2%) are far more likely than adults (10.8%) to be in households suffering from food insecurity (Nord, Andrews & Carlson, 2005). At the same time, approximately 46 million Americans, or 16% of the population, are without health insurance (DeNavas-Walt, Proctor & Lee, 2006). Research suggests that spells without insurance are usually relatively short in duration, but can be relatively frequent (Nelson, 2003). Children are most likely to experience repeated spells without health insurance, whereas adults age 55-64 experience the most frequent insurance coverage transitions. In addition, health care costs continue to increase both in nominal amount and as a share of household expenditures (U.S. Bureau of Labor Statistics, 2007). One of the largest components of out of pocket medical expenditures, prescription drugs, is driving much of the increase in out of pocket expenditures (Kaiser Family Foundation, 2006; U.S. Bureau of Labor Statistics, 2007). Despite an extensive network of private and public food assistance programs and public health insurance options, these statistics demonstrate that it is difficult for millions of families to meet these basic needs. There is some evidence that family resource allocation decisions involve tradeoffs between basic needs (Long 2003; Sharpe, Fan & Hong, 2001), but there is little research that moves beyond cross-sectional estimates to examine family-level economic outcomes associated with the acquisition of food and medical care over time.
Data and Analyses
This study of the relationship between health insurance, medical expenditures and food insecurity examined panel data from the 2001 Survey of Income and Program Participation (SIPP). The SIPP is a nationally representative survey of the non-institutionalized United States population conducted by the U.S. Census Bureau. The three-year 2001 panel collects a "core" set of questions that are collected from respondents every four months. The SIPP also consists of "topical" modules for questions that are not asked each wave. The timing and frequency of the topical modules varies, as does the duration of the reference period to which the questions refer. The data used here were collected from January, 2003 to December, 2003 (waves 7, 8, and 9) and, depending on the rotation group of the respondent, refer to a continuous 12 months that began as early as October 2002 and ended as late as December, 2003. It was from this 12-month period that health insurance status, employment, sociodemographic information, and family composition data were drawn. The analytic sample included 49,989 people age 0 to 87 who were members of 16,236 families in 2003. When weighted, this sample represented 170.3 million people who were members of 70.8 million families.
The multivariate analyses focused on the relationship between food insecurity and medical out of pocket expenditures while explicitly controlling for the potential endogeneity of the two variables. Specifically, a two stage probit least squares estimation (2SPLS) that simultaneously fit the probit and least squares equations was used. This approach allowed us to account for the joint decision making made by households about food and medical expenditures. A 2SPLS estimation, rather than single-equation estimation methods, allowed food insecurity status to be included among the explanatory variables in the medical out of pocket expenditure equation, and medical out of pocket expenditures to help explain food insecurity.
Results and Discussion
The results from this nationally-representative sample of families found no evidence that food and medical expenditures crowd out one another. That is, when …