In this study, we described the processes involved in developing the CFD, a dataset containing information on food pantries in 12 US states. Descriptive findings indicate approximately three quarters of food pantries are located in urban areas, and almost two thirds were considered to have a faith affiliation, which were also more common in urban versus rural areas. Among pantries with hours of operation posted, 78.1% were open at least 1 day per week, and non-faith-based and urban food pantries were more likely to be open more often. This dataset can be linked via FIPS to a number of publicly available datasets, such as the USDA Food Access Research Atlas, the USDA Food Environment Atlas, and the U.S. Religion Census Data. Through linkage of this CFD with other datasets, a number of research questions can be examined.
Food insecurity affected 10.5% of households in the United States in 2020, and is more common among households with children, and Black or non-Hispanic householders . Given the prevalence of food insecurity, efforts to mitigate food insecurity have the capacity to greatly improve population health at multiple levels – national, state, county, household, and individual. The role of food assistance programs has increased as a result of the decline of non-food social programs. The largest publicly-funded food assistance program in the US is SNAP, which provided ‘food stamp’ benefits to more than 44 million people in 2016 . However, many eligible people do not participate in the program, and among those who do, approximately half of them continue to report being food insecure . Charitable food assistance programs or food banks, which were initially established to provide emergency food supplies, are now considered to supplement the governmental programs in their effort to address food insecurity. In fact, 26.5% of food insecure households and 4.8% of all US households used a food pantry in 2016, representing a 40 and 68% increase from 2001, respectively . For these reasons, it has become increasingly important to consider the effects of growing charitable food programs on food security and health .
Charitable food organizations, and other community-level initiatives, have the potential to improve individual health through emergency food provision . However, most health research, media attention, and governmental policy action is disproportionately focused on individual health and exposures, which limits our ability to understand structural drivers of inequality [43,44,45]. The USDA’s Food Access Research Atlas and the Food Environment Atlas provide data on food access and environmental indicators at the census tract- and county-levels. By assigning census tract numbers to each food pantry, the CFD is able to link with both atlases, providing an opportunity to explore the structural drivers of health inequality as it relates to food pantries, ‘food deserts’, federal food programs, food insecurity, and health. The CFD is also able to link with US Religion Census data , which contains data on congregations, members, adherents, and attendees, or the population purported to sustain the charitable food sector.
The CFD consists of more food pantries located in urban census tracts compared to rural tracts, reflecting the higher proportion of urban census tracts in the US. The proportion of urban census tracts that had at least one food pantry was 26% higher compared to rural census tracts, and the number of food pantries per urban population was higher than the number per rural population by approximately 40%. This is inconsistent with previously reported data using county-level information from the Map the Meal Gap project and the Hunger in America 2014 survey, which showed that the number of charitable food locations per 1000 people was highest in counties that were considered completely rural according to urban-rural continuum codes . However, the difference in defining urban and rural areas (i.e., county versus census tract) makes it difficult to compare the findings from the two studies. Given that food insecurity is more prevalent in more populated metropolitan areas compared to nonmetropolitan areas, this may indicate that the food pantries in our dataset are located in areas of greatest need . However, further research is needed to provide estimates of food insecurity at the census-tract level in order to determine if, in fact, the food pantries are concentrated in the areas that would benefit most from their service.
Charitable food organizations rely heavily on food and monetary donations, and volunteers for their operation. In this way, faith-based or religious organizations, with their ability to engage their communities and which often work for social justice and against inequality, are set up well to provide such services . In addition, volunteerism in food banks and pantries is often motivated by faith and has an important role in building community . This may explain the higher proportion of food pantries identified as faith-based in the CFD. The relationship between religion and population health has been extensively explored, and through its ability to provide social capital to communities, especially the most vulnerable communities, illustrate religions’ importance as a social determinant of health [49, 50]. However, the variability in the hours of operation of food pantries reflects the volunteer nature of food banking, and is a legitimate concern among clients given that many rely on prolonged use of food pantries [4, 51]. This may illustrate the limits of volunteerism in addressing food insecurity, which may be exacerbated as participation in faith-based communities is declining .
The large size and diversity of the CFD is a strength, which provides a foundation for future research exploring the relationship between charitable food and social- and health- related outcomes from a systems perspective. However, there are several limitations of this dataset, and the present study, that must be considered. First, the completeness of the dataset is uncertain. The foodpantries.org directory only contains information regarding updates, corrections, or new food pantries that they receive manually through an online submission form. Some food pantries may not be in the directory, while others may still be included despite closure. We documented 146 directory entries that were considered duplicates because they shared either the same name or address with another entry, which illustrates the limitation in the maintenance of the directory. Furthermore, Feeding America advertises 200 food banks and 60,000 food programs as part of its network; however, foodpantries.org documented only 15,494 food pantries in total in 2018, with 3777 included in the present study. While it is unclear how Feeding America defines a ‘food program’ or whether the 60,000 food programs are a cumulative or a point prevalence, there is clearly a large discrepancy. However, our documented totals for food pantries in Detroit, Michigan are very similar to previous research conducted 4 to 5 years earlier; in addition to foodpantries.org, authors also utilized Local Harvest and several local sources to identify food pantries .
Second, faith-based affiliations were subjectively determined using a collection of common Judeo-Christian terms, which may have led to some misclassification. This approach identified 63.2% of food pantries as Christian faith-based. In her 1998 book, Poppendieck states that “more than 70 percent of the pantries and kitchens affiliated with the Second Harvest Network are sponsored by churches or other religious organizations” and that this is likely an underestimate of “the prevalence of religious orientation”  (p188–89).
Third, while many food pantries are open at least once a week, the quantity of food available per family, their form (pre-packaged food boxes or grocery store style), and the quality of the food provided is unknown. Research suggests that the quality of food available at food pantries does not meet recommendations put in place by health professionals . Furthermore, we are missing data on days of operation for nearly a third of food pantries.
Fourth, food pantries that are only open a few times a year (i.e., one to four times) are also included in the foodpantries.org directory. These food pantries likely operate only during specific holidays (i.e., Christmas and Thanksgiving); while they can address immediate hunger, they will have limited impact on individual or population-level food insecurity.
Fifth, the proportion of food pantries per population used the most recent population estimates from the US Census that could be stratified by urban and rural census areas, which were estimated 8 years prior to the date that the number of food pantries was determined. The populations increase slightly each year, therefore, the proportions are likely over-estimated. Lastly, the inclusion of only 12 states may limit the generalizability of the data to the United States as a whole, though it is also unlikely that the 12 states selected are completely unique to the country.
To validate the completeness of the dataset, extensive ground truthing exercises and/or comparison to other existing local datasets collected through other means could be completed. This may mitigate some of the limitations previously described. This dataset could be updated through identical methods, and corresponding validation procedures. Ideally, all countries with charitable food systems, particularly those receiving public funds, should be keeping public records or datasets of food pantries to track the distribution of charitable food. Ensuring accurate and complete data is critical to informing policy related to food security.