Applied Geography 31 (2011) 1224e1231
Contents lists available at ScienceDirect
Applied Geography journal homepage: www.elsevier.com/locate/apgeog
Assessing the proximity of healthy food options and food deserts in a rural area in Maine Teresa A. Hubley* University of Southern Maine, 45 Commerce Drive, Suite 11, Augusta, ME 04330, United States
a b s t r a c t Keywords: Public health Food deserts Nutrition
The purpose of the project described in this paper was to assess and describe the food environment facing public assistance clients in a rural county in Maine. Using the concept of a “food desert” and an objective tool for rating participating food outlets, the research team developed a spatial model of client access to healthy foods. The final map shows that most rural residents are within acceptable distances of well-rated stores, though these may not be supermarkets. Ó 2010 Elsevier Ltd. All rights reserved.
Introduction The presence of food deserts as a contributor to hunger and poor nutrition in the United States has recently become a cause for concern, having spawned many studies (McKinnon, Reedy, Morisette, Lytle, & Yaroch, 2009) and having even merited mention by the First Lady in a campaign to address healthy weight and eating (Obama, 2010). A “food desert” is defined as a populated area with deficient access to the most well-stocked outlets, the large stores or supermarkets that usually provide abundant, good quality, lowpriced food choices (Forsyth, Lytle, & Van Riper, 2010; USDA, 2009). Supermarkets are considered desirable because they can, through economies of scale, provide lower prices and greater variety, thus mitigating some of the common factors that may prevent consumers from making healthy food choices (Giskes, Van Lenthe, Brug, Mackenbach, & Turrell, 2007; Kaufman, 1999; Krukowski, West, Harvey-Berino, & Prewitt, 2010; USDA, 2009). Low access to supermarkets in the United States has been linked with poor quality diets (Frank et al., 2006; Sharkey, 2009; Sharkey & Horel, 2008). The United States’ Food, Conservation, and Energy Act of 2008 refined the population of concern in food desert studies to “predominately lower income neighborhoods and communities” in recognition of the additional burden faced by poor households in obtaining food (USDA, 2009). Urban-based studies, which dominate the body of work on this subject, define “deficient access” as any greater than a “reasonably walkable” distance, variously cited as 500 or 1000 m, 1 km, or half a mile (Apparicio, Cloutier, & Shearmur, 2007; Bustillos, Sharkey, Anding, & McIntosh, 2009; Larsen & Gilliand, 2008; McEntree &
* Tel.: þ1 207 626 5292; fax: þ1 207 626 5210. E-mail address:
[email protected]. 0143-6228/$ e see front matter Ó 2010 Elsevier Ltd. All rights reserved. doi:10.1016/j.apgeog.2010.09.004
Argyeman, 2010; Sharkey, 2009; USDA, 2009; Wrigley, Warm, & Margetts, 2002). For rural studies in the United States, ten miles is a common measure of acceptable distance to a supermarket (McEntree & Argyeman, 2010; Morton & Blanchard, 2007; USDA, 2009). This definitional framework assumes that supermarkets are the most consistent means to meet the criteria of maximal food choices and that there can be a proxy distance that denotes “reasonable access,” outside of which consumers must be located in food deserts. The project described in this paper tests this proposition by gathering data on quality, availability, and price of food choices across store types and measuring distances to the most favorable performers, regardless of type. The concept of “access” The first step in defining access to retail locations using spatial terms is to map all the demographic points of interest. The most basic of these are store locations and population centers. One method of connecting these points is to measure the distances from groups of persons, using summary points of reference such as the population-weighted centers of census blocks (Apparicio et al., 2007; Pearce, Witten, Hiscock, & Blakeley, 2007; Sharkey & Horel, 2008). Another method is to measure the distance from actual households to retail establishments (Hermstad, Swan, Kegler, Barnette, & Glanz, 2010; McEntree & Argyeman, 2010). By contrast, “food desert” studies begin with retail outlet locations, rather than population locations, and measure outward from these points to define reasonable travel distances, beyond which consumers are said to reside in a “desert,” a place of scarcity having low access to food. While the general concept of “areas with low access to food” is broadly understood, “low access” is an imprecise term that remains in flux, due to measurement challenges and
T.A. Hubley / Applied Geography 31 (2011) 1224e1231
1225
Table 1. RFANA Model Of access. Concept
Data need
Supply of retail outlets [includes simple presence of the outlet, ability of the outlet to accept SNAP funds in payment] Availability of acceptable products [includes simple presence of the product, reasonable quality of products, knowledge and ability of the consumer to utilize the products] Proximity [includes simple distance, quality of egress (e.g. passable roads, walkability)] Available income [acknowledging low income as a barrier] Price [impact affected by income and ability to use SNAP]
List of stores that accept SNAP
the complexity of elements affecting spatial access (Morton & Blanchard, 2007). Furthermore, many seminal food desert studies have derived their concepts of “reasonable distance” from work in a narrow range of spatial contexts. A recent report to the U.S. Congress by the Economic Research Service of the United States Department of Agriculture (USDA, 2009), which includes an exhaustive overview of the literature, points out that most such studies have been conducted in areas of high population density. Nevertheless, some common definitions of “reasonable distance” can be found. Food desert studies focused on access to supermarkets in urban areas, where “walkability” is a well-established measure of reasonable distance from a supermarket to a customer, generally present this distance as roughly one mile (USDA, 2009). Such studies begin by mapping locations of supermarkets and then establishing a buffer zone, consisting of the established reasonable distance. The resulting map would then be layered over a census map of the city to show pockets of population outside the walkable zone. Those areas outside the walkable buffers would be identified as “food deserts” (Apparicio et al., 2007; Frank et al., 2006; Larsen & Gilliand, 2008; USDA, 2009). For rural areas, the most “reasonable” distances are less well-established than for urban areas, with some researchers defining “highly accessible” as “within ten miles of the population (McEntree & Argyeman, 2010; USDA, 2009)”. Beyond ten miles, the literature is split over whether driving distances of up to twenty miles might be considered “medium access (USDA, 2009)”. This variability may be related both to the lesser numbers of rural studies and to the high variability of terrain and weather conditions in rural areas, as well as “reasonable routes” reported by residents themselves. Aside from which distances are most relevant, food environment researchers question the ascendancy of supermarkets as the only relevant measure of food security (McKinnon et al., 2009, Raja, Ma, & Yadav, 2008, Shaw, 2006). Other venues, such as small ethnic grocery stores found in urban areas, have been shown to provide quality, pricing, and availability on a par with larger chain grocery stores and to attract a loyal customer base (Short, Guthman, & Raskin, 2007). These smaller stores do have positive impacts on vegetable and fruit consumption in neighborhoods (Bodor, Rose, Farley, Swalm, & Scott, 2007). The challenge of measuring reasonable distance follows the challenge of defining this distance and defining the relevant points of interest from which to measure. Besides miles or meters, another method of measuring “acceptable travel” is to express the distance in time, such as that area which can be covered in 5 min, computed from an average walking speed (Raja et al., 2008). Both distances and times can be measured as Euclidian (“as the crow flies”) or through networks, such as travel time and distance across streets and roads (McKinnon et al., 2009). These measures may not have much relation, however, to the real behaviors and barriers that pertain to consumers, who find themselves navigating construction
Availability and quality of healthy foods as measured by survey, qualitative data on distances travelled by consumers, qualitative data on consumer food preparation knowledge and processes Distances between stores and major roads; Placement of stores relative to the population; Distance to large or super outlets SNAP and poverty rates Price of specific food items as measured by survey; Relative price of food items as determined through scoring and testing Data elements used in this study
sites, slow trains, physical disabilities, and weather delays (to name a few obstacles) that are not part of the equation for a remote researcher using GIS data to measure distance. Thus “reasonable distances” and “low access” will defy summary in a manner that can be used uniformly across all studies.
A working definition of “access” The Rural Food Access Needs Assessment (RFANA) of Maine seeks to contribute to the working knowledge of how access affects healthy eating behaviors and to apply this knowledge directly, through tailored nutrition education programs. Based on work such as that cited above, the RFANA team believes the availability of well-priced foods of acceptable quality may affect the ability of participants in Maine’s segment of the federal Supplemental Nutrition Assistance Program [SNAP] (formerly known as “Food Stamps”) to adhere to the Dietary Guidelines promoted by the educational portion of SNAP, widely known as SNAP-Ed. One objective of RFANA is to spatially describe disparity in access to healthy foods in a targeted study area. The concept of access is enhanced by using survey data in order to objectively portray one facet of “availability,” namely how many retail sources are present that provide acceptable foods. For the purposes of RFANA, especially without access to target population addresses, the “food desert” approach lends itself more readily to characterizing access to healthy foods in the study area. The RFANA team began their study with the locations of food outlets and measured outwards in this initial phase of exploring the spatial dimensions of the food environment for SNAP recipients. Because rural Maine is subject to weather extremes that can affect travel and many less travelled roads are in poor condition, the ten-mile distance was chosen as the cut-off for “food deserts”. The RFANA model of “access,” which includes reference to potential economic barriers as well as spatial issues, is summarized on Table 1. In response to the framework of the conceptual model, the team used GIS to select an initial study area based on poverty rate, SNAP participation rate, and population density, and then to locate both supermarkets and their ten-mile radii with respect to the location of the general population. The team then used an adapted version of a well known tool (described below) for collecting store data to assess and score all the regular food stores in the study area that accept SNAP funds in payment for food. The inclusion of stores that are not supermarkets takes this assessment beyond the standard food desert concept to consider the possibility that consumers may have greater access to healthy foods than might be supposed by virtue of the lack of supermarkets in their area. The project described below closed the loop on this concept by generating a second map that showed 10 mile buffers around stores that scored well on the survey, regardless of whether they were supermarkets.
1226
T.A. Hubley / Applied Geography 31 (2011) 1224e1231
Fig. 1. State of Maine with food insecurity risk rates. This illustrates the location of major cities and the target county.
Materials and methods Developing and using a food survey tool The RFANA team went beyond measuring distances to stores in characterizing the food environment in rural Maine by gathering empirical data about food available in these stores. The well-known Nutrition Environment Measure Survey (NEMS) tool (Glanz, Sallis, Saelens, & Frank, 2007) was chosen as a model for developing a method of measuring the food environment on the ground. The RFANA team received formal training from Emory University staff in the use of NEMS and then adapted the tool and its scoring mechanism to create the Maine Nutrition Environment Measure Survey (ME-NEMS). The reconstituted survey focuses on the “healthy” options only, eliminating the comparison between “healthy” and
“less healthy” choices found in the original tool. The final survey consists of five master food groups: fruits, vegetables, meats, grains, and dairy. The team adjusted the list of fruits and vegetables surveyed for seasonal availability in Maine, projecting a survey start date in late spring. In order to verify the validity and reliability of the survey, the team conducted a pilot test in the immediate vicinity of their Augusta, Maine, office. Three team members surveyed 6 sample stores, repeating observations in two of their assigned stores again on another day and surveying two other stores besides. This pattern resulted in 18 observations, which could be used to generate interrater reliability and testeretest analyses. The results of this pilot showed overall agreement across most items, with a need to adjust some items on the survey and some of the written instructions accompanying the survey. For example, surveyors did not make
T.A. Hubley / Applied Geography 31 (2011) 1224e1231
1227
Fig. 2. Somerset County with census block populations and 10 mile buffers around supermarkets. Areas outside the buffers can be “food deserts.” Note the above 2500 towns, which are classified as “urban”.
consistent decisions on whether to use bagged or loose vegetables when recording prices, thus leading to a new instruction to prioritize bagged vegetables. As per significant differences in units across fresh produce, we used information from USDA to develop and update all the final data to a standard unit for price comparison. In its final form, the ME-NEMS addresses 62 food items for a total of 370 fields. Creating and applying site selection criteria The RFANA team selected the target area for study using 3 characteristics that indicate potential income barriers to effective
nutrition management: 1) low population density, 2) prevalence of low income, and 3) high participation in SNAP. ArcEditor was used to create a state map of “total risk rate” for all Maine communities representing all three criteria combined, sorted into three natural breaks (see Fig. 1). Based on the relatively high risk rates for towns in its boundaries and in consideration of driving time for the survey team, Somerset County, with a population 51,658 as of the 2007 census estimate, was chosen as a target. As further incentive to choose this area, staff identified this county as the site of an active and engaged local SNAP-Ed program that could be engaged to tailor programs based on findings.
1228
T.A. Hubley / Applied Geography 31 (2011) 1224e1231
As of the 2007 census, the 35 Somerset County towns with population data available range from 1 person per square mile to 145 persons per square mile, with a median value of 23 persons per square mile. The poverty rate (as per the 2000 census) for the county is 17.2%, whereas the state poverty level is 12.2%. Somerset County presents a mixture of high and low density settings, with six towns in the highest level of the three population breaks (i.e. above 2499), and thus designated as an “Urban” (i.e. high density) community. The largest town in Somerset County is Skowhegan, with a 2007 population of 8758, following by Fairfield at 6734. The majority of the county’s population is concentrated in the southern half, clustered around Skowhegan and adjacent towns. The county is accessed by means of three major roads (in addition to at least three state roads intersecting these); the NortheSouth state route 201, the EasteWest route 2, and a small stretch of interstate 95. There is little to no public transportation available anywhere in rural Maine. According to the 2000 census, Somerset County as a whole has slightly more vehicles on average per household than the rest of Maine. Once Somerset County was chosen as a target area, the team requested a list of all retail food outlets that accept SNAP benefits from the local USDA administration. The USDA provided a list of stores categorized by self-identified types (such as “convenience” and “supermarket”), of which 62 were located in the target county. While it is likely that SNAP recipients also access food from other outlets using cash, the RFANA focus upon only SNAP outlets was maintained in order to assess the food environment for the poorest of the poor in our study area. The team used a web-based geocoding service to estimate the approximate location of each store and mapped the list using ArcEditor. Locations were reviewed in GoogleMaps and cross-checked against address lists for accuracy. In order to assess the extent of food deserts in the study area, the team produced a map showing the distribution of population at the census block level, the lowest level available, and the location of the six “Urban” towns. Major roads were also displayed on this map. Every SNAP-participating large and super store in the state was then given a ten-mile buffer, overlaying this on our base map. The result (see Fig. 2) is a map that illustrates the relative clustering of the population of Somerset County in its southern half. Food deserts appear on the edges of the main population cluster and at one community in the northern half.
Table 2. ME-NEMS scores (bold) with 95% CI.
Analyzing store data
Results
The RFANA field team gathered data from most of the SNAP outlets in Somerset County. This raw data on simple availability, price, and quality were compared across different store types with Chi Square, one-way ANOVA, and Tukey Post-Hoc tests. The team also used an adapted version of the NEMS scoring scheme tailored to the ME-NEMS to generate summary scores representing the overall acceptability of the food offerings in the outlets using Microsoft Excel. The scores serve to summarize the performances of the individual stores compared to their peers across the county. For example, the scores for price reflect the percentage of items in each store that are priced below the 60th percentile for the population, assigning a higher score for lower prices. Scores for availability and quality are based on percent of items present and percent of present items that meet pre-defined quality standards. The team SPSS to analyze differences among scores associated with different groups of stores, via one-way ANOVA and Tukey Post-Hoc tests. ArcEditor was used to examine the location of the various types of stores by distance from each other, distance to the major road system, presence of food deserts, and relation to population clusters and centers. The goal of this exercise was to determine whether the stores appeared to occur in groupings that may differ significantly.
Store by type and location
Score areas and methods>>>>>>
Grand total (of 79)
Quality (of 12)
Availability (of 24)
Price (of 43)
Sum of all score areas
Based on % items acceptable by type area
Based on number of items present by type area
Based on % prices < 60th percentile for each item
Analysis level
N Score 95%CI Score 95%CI Score 95%CI Score 95%CI
All cases Super Grocery Convenience
50 8 8 34
36.48 74.13 43.00 26.09
Dense Sparse Sparse super Sparse grocery Sparse Convenience Dense super Dense grocery Dense convenience
24 26 3 6 17 5 2 17
34.21 10.60 5.67 2.30 38.58 8.99 6.35 1.20 72.33 3.79 12.00 0.00 42.83 22.62 7.17 4.22 31.12 9.62 5.06 1.78 75.20 3.21 12.00 0.00 43.50 285.89 7.00 63.53 21.06 5.66 3.65 1.48
11.13 4.20 17.54 7.10 12.81 1.36 19.46 4.08 22.67 1.43 38.00 4.30 13.83 5.92 21.83 13.69 10.71 2.68 15.35 5.60 23.00 0.00 40.20 3.21 12.00 76.24 24.50 146.12 7.53 1.80 10.06 3.48
Urban Rural Rural super Rural grocery Rural convenience Urban super Urban grocery Urban convenience
28 22 3 4 15 5 4 19
31.43 42.91 72.33 51.00 34.87 75.20 35.00 19.16
10.68 13.68 22.67 15.50 11.40 23.00 11.25 7.32
6.69 6.02 2.16 12.00 18.25 7.13 5.57 4.35
9.18 5.25 9.85 7.00 3.79 12.00 34.10 7.75 10.63 5.80 3.21 12.00 34.93 6.50 3.89 3.21
1.19 12.00 0.00 22.88 3.61 13.38 1.12 9.12
3.44 0.05 0.00 6.15 1.87 0.00 8.37 1.27
1.90 0.30 4.85 1.63
5.82 0.38 1.43 9.23 3.01 0.00 9.13 1.46
18.54 39.38 22.50 12.71
15.54 22.36 38.00 27.75 17.80 40.20 17.25 8.68
3.81 2.04 10.60 3.25
12.50 1.67 4.30 19.02 6.35 3.21 20.06 2.06
Stores were then classified by whether they were found in towns with more than 2500 persons (“Urban”) or in areas with fewer persons (“Rural”) in the first instance, as this distinction was part of the site selection process. In an effort to fine-tune the distinction between “rural” and “urban”, ArcEditor was used to rank census block groups by their person per square mile density (as per the 2000 census), emerging with four levels. The top two levels were collapsed into the new category “Dense” and the lower two into the category “Sparse”. Scores from the ME-NEMS were reanalyzed using SPSS to compare scores and item availability across the stores in both of these categories.
The team arranged to survey all the SNAP outlets in the county. Access was denied to one large chain, eliminating one store. During the survey, some stores were found to have been closed for business or to have ceased accepting SNAP, while others had just begun to accept SNAP. The team purged the list by eliminating defunct entries and specialty food stores, such as seafood stores, and adding new SNAP outlets. The final list consisted of 50 stores. In order to streamline analysis and reporting, similar selections among the categories provided by USDA were collapsed. For example, “Large Grocery,” “Superstore” and “Supermarket” merged to become “Super”. As some of the types assigned appeared inconsistent with observations of the surveyors, the team conducted a statistical analysis of the correlation between numbers of registers and assigned category. The results of this study show a strong relationship. Subsequent studies of the trends in the data by store type show a consistent clustering by type. Therefore, the assigned types were retained for analysis. The total count by store type, as well as type and location, appears on the Table 2, along with a comparison of total scores
T.A. Hubley / Applied Geography 31 (2011) 1224e1231
1229
Fig. 3. Somerset County with census block group populations and 10 mile buffers around top tier stores. The top two most populous sets of block groups are the “dense” areas, while the others are the “sparse” areas. Note how the “food deserts” are smaller in this depiction.
across sub-scores and geographic areas. All but one of these stores, a supermarket, is located in the more populous southern half of the county. All but 7 of the stores are within 2 miles of a major road. One of the 50 stores, a convenience store, is located in a “food desert.” The mean grand total score differences were compared for significance at the 0.05 level for all of the assigned groupings, using t-tests for the three pairs of level categories and ANOVA’s with Tukey post-hoc tests for the three types and then the three groups of six sub-types in each level. For the three basic types alone (super, grocery, and convenience), all differences observed were significant.
Differences between scores corresponding to both type and location compared within the two working spatial categories (i.e. rural/urban and sparse/dense) varied in significance. Most prominently, convenience stores in rural areas did not significantly differ from grocery stores in either rural or urban areas. Meanwhile, convenience stores in more heavily populated areas consistently scored lower than any other type of store across all grouping. This same pattern was observed in comparing convenience stores in “sparse” areas to grocery stores elsewhere. The overall strong performance of convenience stores in less populated areas in comparison to convenience stores and even grocery stores in more
1230
T.A. Hubley / Applied Geography 31 (2011) 1224e1231
populated areas suggests they should perhaps be re-classified with a distinctive type (such as “general store”). Aside from the striking difference between convenience stores in more remote locations and those in more central locations, the store types showed no significant differences when paired by type across location. For example, super stores in rural and urban areas perform just as well as each other, the same being true of grocery stores. This may be a feature of the small numbers found in the super and grocery categories, as together they make up one-third of the store inventory, compared to the far more numerous convenience stores. Availability Within the five main groups, specific items strongly influenced the scores. In particular, fruit juice, canned vegetables, whole wheat bread, tuna, milk, and cheese are widely available, thus driving up the scores and results in their respective categories. Fresh foods (produce, skinless chicken, lean ground beef) and frozen fruits and vegetables are more restricted in distribution, meaning that while their categories may be broadly available, variety may be a problem for consumers. Convenience stores in less densely populated areas once again show a tendency to score better than their more urban counterparts and even on a par with urban groceries. Quality The survey measured quality for fresh produce (fruits and vegetables), fresh meats, and milk. Quality was most likely to be acceptable across all the stores. Studies of trends in quality found no relationship between quality and store type or location. However, this result should be evaluated in light of the relative availability of the items for which quality is measured (i.e. only milk is almost universally available). Price As expected, super stores in general score highest on all elements of price, meaning their prices are more consistently lower than the median. Grocery and convenience stores score close together in all areas, except for vegetable prices, where convenience stores score lower, most likely as a consequence of greater vegetable availability in more rural settings. Convenience stores in low density settings generally score as well as grocery stores in high density settings on price. Proximity to “top tier” stores When the scored survey data were assessed for trends in grand totals, 18 stores were found to have scored above the confidence interval around the mean for total score (36 out of a possible 79). These stores were designated as “Top Tier” stores, stores that perform well across several dimensions, including the five master food areas and the three major concept areas of quality, price, and availability. Of the 18 stores found to be “Top Tier,” 11 scored significantly above the mean in all these areas. All three types of stores are represented among the “Top Tier,” though the 6 convenience stores are all located in less densely populated areas. All 8 of the super stores are included in the “Top Tier.” A new map (see Fig. 3) showing the population relative to a tenmile radius around each “Top Tier” stores changes the perception created by mapping only the “food deserts” as defined above. Most of the areas found to be deserts in the original analysis are within ten miles of a “Top Tier” store. However, all the “Top Tier” stores are within two miles of a major road.
Discussion Implications The result of this assessment project so far shows greater spatial access to healthy food choices than predicted using the strictest “food desert” paradigm, i.e. reasonable access for low-income consumers to supermarkets (USDA, 2009). This result suggests that rural areas defy easy summary with regard to the food environments they present to residents, a finding in line with studies such as Short et al. (2007). Data from our study show that some healthy foods are widely available to SNAP recipients across their remote rural area, though some gaps remain. Where SNAP recipients do have access to fresh foods, their quality is reliably acceptable. Most of the residents of the study area live within 10 miles of a store that performed well across the majority of the elements found in the survey. These findings suggest components of a tailored educational approach for the SNAP-Ed projects now conducted in the area. For example, the nearly universal availability of certain items (such as canned tuna) could form the basis for recipe development that can meet the needs of families receiving SNAP. Limitations The main shortcoming of this analysis is that it represents comparisons between relatively small subsets of a small population, thus weakening the conclusions that can be drawn from the variations observed. In particular, only 2 of the 8 grocery stores were in densely populated areas while 3 of the 8 supermarkets were located in sparsely populated areas. The sample is also incomplete. While the majority of stores accepting SNAP in Somerset County were included, some may have been missed, due to the static nature of the list used to identify them. There was also one super store that refused the surveyor access. Lastly, the tool used gathered data on select representative items, passing over other potential measures of available healthy food. For example, legumes have been noted as conspicuously absent. The quality measure, unlike price and availability, only applied to fresh produce and meat and to milk. Lastly, this survey was conducted in early spring before the appearance of many types of produce and wider participation in SNAP by Farmers’ Markets. Results of surveys conducted at other times of the year could potentially shift perception of the food environment and demonstrate other needs not accounted for in this initial study. The data presented here could benefit from the inclusion of qualitative data on SNAP-recipients’ shopping practices and their perceptions of their level of access to good quality food choices. As Shaw (2006) suggested, barriers to healthy eating are not only spatial in nature but also include complex beliefs and behaviors at the consumer level. Maine’s SNAP-Ed program acknowledges this connection by including cooking demonstrations and recipe distribution as an effort to expand the knowledge base of SNAP recipients and facilitate healthy eating choices. In recognition of the need to add more data to the model, the RFANA team has conducted intercept surveys at large grocery stores and is also designing a new round of data gathering that would include more objective assessments of use patterns through the analysis of receipts. The SNAP-Ed program’s understanding of the challenges and needs of SNAP recipients in rural areas as they attempt to meet their families’ nutritional needs will continue to evolve as new data are added to the model. Acknowledgements The original work for this project was carried out under the direction of Jigna Dharod, Ph.D., currently on faculty at the
T.A. Hubley / Applied Geography 31 (2011) 1224e1231
University of North Carolina Greensboro. Also instrumental to the project were Matthew L’Italien, RD, and Emily Stiles, RD, both of whom assisted in redesigning and testing the ME-NEMS tool and gathering the data, and Kevin Scribner, team statistician. This project was funded through USDA’s SNAP-Ed program and the Maine Nutrition Network at the University of Southern Maine. References Apparicio, P., Cloutier, M.-S., & Shearmur, R. (2007). The case of Montreal’s missing food deserts: evaluation of accessibility to food supermarkets. International Journal of Health Geographics, 6(4). Bodor, N., Rose, D., Farley, T., Swalm, C., & Scott, S. (2007). Neighbourhood fruit and vegetable availability and consumption: the role of small food stores in an urban environment. Public Health Nutrition, 11(4), 413e420. Bustillos, B., Sharkey, J., Anding, J., & McIntosh, A. (2009). Availability of more healthful food alternatives in traditional, convenience, and nontraditional types of food stores in two rural Texas counties. Journal of the American Dietetic Association, 109(5). Forsyth, A., Lytle, L., & Van Riper, D. (2010). Finding food: issues and challenges in using geographic information systems to measure food access. The Journal of Transport and Land Use, 3(1), 43e65. Frank, L., Glanz, K., McCarron, M., Sallis, J., Saelens, B., & Chapman, J. (2006). The spatial distribution of food outlet type and quality around schools in differing built environment and demographic contexts. Berkeley Planning Journal, 19(79e95). Giskes, K., Van Lenthe, F. J., Brug, J., Mackenbach, J. P., & Turrell, G. (2007). Socioeconomic inequalities in food purchasing: the contribution of respondentperceived and actual (objectively measured) price and availability of foods. Preventive Medicine, 45(2007), 41e48. Glanz, K., Sallis, J., Saelens, B., & Frank, L. (2007). Nutrition environment measures survey in stores (NEMS-S): development and evaluation. American Journal of Preventative Medicine, 32(4), 282e289. Hermstad, A., Swan, D., Kegler, M., Barnette, J. K., & Glanz, K. (2010). Individual and environmental correlates of dietary fat intake in rural communities: a structural equation model analysis. Social Science & Medicine, 71(2010), 93e101. Kaufman, P. R. (1999). Rural poor have less access to supermarkets, large grocery stores. Rural Development Perspectives, 13(3), 19e26, (Economic Research Service, USDA).
1231
Krukowski, R., West, D.-S., Harvey-Berino, J., & Prewitt, T. E. (2010). Neighborhood impact on healthy food availability and pricing in stores. Journal of Community Health, 35, 315e320. Larsen, K., & Gilliand, J. (2008). Mapping the evolution of ‘food deserts’ in a Canadian city: supermarket accessibility in London, Ontario, 1961e2005. International Journal of Health Geographics, 7(16). McEntree, J., & Argyeman, J. (2010). Towards the development of a GIS method for identifying rural food deserts: geographic access in Vermont, USA. Applied Geography, 30. McKinnon, R., Reedy, J., Morisette, M., Lytle, L., & Yaroch, A. (2009). Measures of the food environment: a compilation of the literature, 1990e2007. American Journal of Preventive Medicine, 36(4S). Morton, L., & Blanchard, T. (2007). Starved for access: life in rural America’s food deserts. Rural Realities, 1(4). Obama, M. (2010). First lady Michelle Obama launches let’s move: America’s move to raise a healthier generation of kids press release. Washington D.C.: White House, Office of the First Lady. February 9, 2010. Pearce, J., Witten, K., Hiscock, R., & Blakeley, T. (2007). Are socially disadvantaged neighborhoods deprived of health-related community resources? International Journal of Epidemiology, 36, 348e355. Raja, S., Ma, C., & Yadav, P. (2008). Beyond food deserts: measuring and mapping racial disparities in neighborhood food environments. Journal of Planning Education and Research, 27, 469e482. Sharkey, J. (2009). Measuring potential access to food stores and food-service places in rural areas in the U.S. American Journal of Preventive Medicine, 36(4S). Sharkey, J., & Horel, S. (2008). Neighborhood socioeconomic deprivation and minority composition are associated with better potential spatial access to the ground-truthed food environment in a large rural area. The Journal of Community and International Nutrition, 138, 620e627. Shaw, H. (2006). Food deserts: towards the development of a classification. Geografiska Annaler, 88(2), 231e247. Short, A., Guthman, J., & Raskin, S. (2007). Food deserts, oases, or mirages?: small markets and community food security in the San Francisco Bay area. Journal of Planning Education and Research, 26, 352e364. United State Department of Agriculture [USDA]. (June 2009). Access to affordable and nutritious food: Measuring and understanding food deserts and their consequences: A report to congress. Washington, DC: USDA. Wrigley, N., Warm, D., & Margetts, B. (2002). Deprivation, diet and food retail access: findings from the leeds “food deserts” study. Environment and Planning A, 34.