We had hypothesized that being exposed to higher neighborhood density of unhealthy food establishments may lead to a stronger relationship between SNAP participation and obesity since respondents would have a greater chance of using their SNAP benefits in these outlets. Our findings did not support this hypothesis. Contrarily, a recent study has found that the regional density of SNAP-authorized stores is positively associated with obesity in metropolitan areas in the U.S. [
19]. We did not have information about SNAP certification of the food outlets included in our analysis, however, and the presence of these outlets in the neighborhood is not equivalent to usage by SNAP beneficiaries. Moreover, we did not have the actual addresses of respondents and, therefore, were not able to estimate the distance of respondents’ homes to these food establishments. SNAP participants are likely to go outside their neighborhoods for SNAP purchases, especially if local food outlets do not accept SNAP benefits. Previous research shows that SNAP participants often carpool to farther but cheaper stores and sometimes visit multiple stores throughout the month to get better deals [
20]. However, a study based on L.A. FANS data report that 34% of respondents shop for groceries at stores within a 15-min walk of their home and over 50% do their grocery shopping either in their own census tract or the neighboring one [
21].
Han et al. [
9] found that a large number of supermarkets and grocery stores in one’s zip code reduced the strength of the association between SNAP participation and BMI in a US representative sample. We used food environment indicators in the respondents’ census tract, which is a smaller geography than zip codes and therefore may better reflect neighborhood availability. In addition, we looked at a different dimension of the food environment than Han et al. [
9], focusing on the density of unhealthy food outlets instead of on supermarkets and grocery stores. Even though the majority of SNAP-participating households report using supermarkets as their main type of food store [
22], most also report redeeming SNAP benefits in other stores [
23,
24], with 42% using convenience stores [
24]. Moreover, Rigby et al. [
7] found that the majority of SNAP-authorized stores in low-income and black and mixed-race neighborhoods in Leon County, Florida were not supermarkets or grocery stores but convenience and other stores, accounting for 69–75% of the SNAP-authorized stores available. Similarly, Shannon [
8] found that in low-income, minority neighborhoods in Minnesota, 46% of SNAP redemptions take place in convenience stores. In addition, Laska et al. [
25] found that SNAP-authorized small- to mid-size retailers in Minneapolis- St. Paul, Minnesota, did not stock a variety of healthy foods, particularly fresh foods and DeWeese et al. [
26] found that SNAP-authorized corner stores in New Brunswick, New Jersey stock less healthy foods compared to non-SNAP vendors. Even though these geographical areas are not directly comparable to Los Angeles, we anticipate SNAP participants living in low-income and minority neighborhoods to have similar food accessibility issues as well as similar redemption and food purchasing patterns across the US. Research shows that the price of both healthy [
9] and unhealthy foods [
27] may modify the association between SNAP participation and obesity. Zhang et al. [
27] found that higher prices of unhealthy foods (e.g. fast food, sodas) attenuates the association between SNAP participation and obesity (since SNAP participants would be less likely to purchase these expensive foods), whereas Han et al. [
9] found that lowering the price of fruits and vegetables would have the same attenuating effect. Food price, store location, and transportation options are all likely to interact in SNAP participants’ decisions of where to make food purchases. Interestingly, a recent study found that proximity to food retailers did not modify the association between receiving a fruit and vegetable incentive and the purchase of such foods among SNAP participants in Hampden County, Massachusetts [
28], implying that price incentives would benefit SNAP participants regardless of stores location. Los Angeles is heavily sprawled, however, with store location and availability of transportation likely playing a larger role in Los Angeles compared to other places.
One of the strengths of this paper is the extensive amount of income-related data available in L.A. FANS, which allowed us to identify a more accurate eligible non-participant group than in previous research linking SNAP participation and obesity, the vast majority of which classifies individuals as SNAP-eligible following simplistic income cut-off points (<130–185% federal poverty line) (e.g. [
4,
29]). Additionally, predictors of SNAP participation in this population were identified in a previous study [
16], so we were able to include these predictors as covariates in our analyses and reduce the impact of self-selection into SNAP in our results. Moreover, having access to yearly food environment data through NETS allowed us to get the counts of food establishments in the census tracts in the specific years where L.A.FANS respondents lived at the time of the survey. As for the limitations, the cross-sectional nature of this study makes it difficult to ascertain the temporality of the relationship between SNAP participation and obesity. Both BMI and SNAP participation are self-reported and subject to bias; however, the self-report of weights and heights by L.A. FANS respondents has been found to be valid [
30]. Moreover, the data used in this study is relatively old (2000–2002), with data available at the census tract-level only. We cannot discard the possibility of different results if we had access to more current data, especially since some improvements in the food environment have taken place in Los Angeles since 2000–2002 [
31], and/or data on smaller geographies. Furthermore, we did not account for population density, which may have confounded our results. Finally, the SNAP participation – obesity relationship has been found most consistently among women [
2]. Given the small number of men in our sample (
n = 278), we were unable to stratify our analysis by gender. However, our results remain the same if the sample is restricted to women only (data not shown).