Food insecurity (FI) is the outcome of a household being unable to acquire, or being uncertain of having, enough food to meet the needs of all its members. Although limited access is a widely recognized barrier to the acquisition of adequate amounts of nutritious food, and despite the widespread recognition that increasing access can improve food security, no empirical study has so far assessed and quantified the relationship between access to different types of food stores and the likelihood of a household to experience FI. As existing programs both at the national and local levels aim to improve food security through food access, understanding which food stores could be more effective in mitigating FI could prove helpful to policymakers.
This study aims to measure the impact of the presence of different food stores on a household’s adult FI status, combining two years (2004 and 2005) of household-level Current Population Survey–Food Security Supplement (CPS-FSS) observations with metropolitan statistical area (MSA)-level data on the number of food store establishments, retrieved from the County Business Patterns Database of the Bureau of Labor Statistics, and Thomas J. Holmes’s Wal-Mart location database.
MSA-level food access (FA) measures are defined as follows:
- Number of North American Industrial Classification System (NAICS) 445110 stores > = 50 employees / 100,000 people (GROC: medium-large size “traditional” food stores, i.e., supermarket and grocery stores);
- Number of NAICS 445120 + NAICS 445110 stores < 50 employees / 10,000 people (SMALL: convenience stores and small grocery stores; easy-to-access proximity store);
- Number of NAICS 447110 Stores / 100,000 people (GSCNV: convenience stores attached to gas stations: small food stores, which may be hard to access);
- Number of Wal-Mart Supercenters / 1,000,000 people (WMSC: proxy for non-traditional, low-priced stores); where they are characterized in function of size and type, the latter defined according to the NAICS.
Household’s adult FI status is characterized by a binary indicator obtained combining High and Marginal Food Security (FI = 0), Low Food Security, and Very Low Food Security (FI = 1), from the CPS-FSS 12-month adult food insecurity summary status. The probability of observing an FI status is assumed to be a function of FA, household (and household head) characteristics, a proxy of the price levels (general Consumer Price Index (CPI)), and State-level fixed effects. After matching the different data sources and discarding observations with missing values, 36,887 observations were retained. The sample was further divided to obtain subsamples of low-income households (income below 185 percent of the current poverty threshold) and of households with and without children.
Estimating the effect of FA on the probability of observing adult FI is complicated by spurious correlation, which can be expected since the FA measures are likely to be correlated with unobservable factors affecting FI. Assuming that the conditional probability of observing adult FI is normally distributed, a two-step instrumental variables probit estimator was used, adopting store-type-specific identification strategies that rely on isolating market-level drivers of store location, unlikely to be correlated with an individual household’s FI status. Because recovering marginal effects from the two-step IV-probit estimates is challenging, they were obtained from the coefficients obtained using a two-stage residual inclusion approach.
The empirical results show that food access affects the likelihood of experiencing adult FI, although the effects differ across samples and FA measures. In most cases the results support the necessity of correcting for endogeneity of food stores’ location (the support is weak in the case of WMSC).
Specifically, food access has little to no impact on adult FI among households without children, a result that appears robust across FA measures. Households without children seem to have fewer incentives to explore the resources available to them to diminish their risk of being food insecure. On the other hand, some food outlets seem to have a considerable mitigating impact on the probability of experiencing adult FI, in particular for low-income households and households with children. Therefore, the food environment seems to play a large role in shaping FI for those households facing hardships due to lack of resources (low-income levels) or having to provide sustenance for their children.
The results show a substantial FI mitigating effect of both medium-large food stores and proximity stores, particularly among low-income households. A marginal increase in GROC (i.e., an increase of one store per 100,000 individuals) causes a decrease in the probability of being afflicted by adult FI in a range between -1.17 percent to -4.74 percent. A marginal increase of one proximity store (small store or convenience store) per 10,000 individuals leads to a reduction of the probability of experiencing adult FI by -1.67 percent to -11.27 percent. The results also indicate that the presence of convenience stores attached to gas stations may worsen food security, due perhaps to a combined effect of inconvenient access, limited assortment, and higher prices.
The results illustrating the effect of Wal-Mart on adult FI were mostly inconclusive due to the poor performance of the identification strategy adopted. However, some results point to the company having a modest direct adult FI-easing effect, and a detrimental indirect effect (via a negative impact on the number of other food stores helping reduce FI), suggesting a null net effect of WMSC on adult FI.