Portland State University PDXScholar

Urban Studies and Planning Faculty Nohad A. Toulan School of Urban Studies and Publications and Presentations Planning

1-1-2017 Culturing Food Deserts: Recognizing the Power of Community-Based Solutions

Catherine Brinkley University of California, Davis

Subhashni Raj SUNY University at Buffalo

Megan Horst Portland State University, [email protected]

Follow this and additional works at: https://pdxscholar.library.pdx.edu/usp_fac Let us know how access to this document benefits ou.y

Citation Details Horst, M., McClintock, N., & Hoey, L. (2017). The intersection of planning, , and food justice: a review of the literature. Journal of the American Planning Association, 83(3), 277-295.

This Article is brought to you for free and open access. It has been accepted for inclusion in Urban Studies and Planning Faculty Publications and Presentations by an authorized administrator of PDXScholar. Please contact us if we can make this document more accessible: [email protected]. CULTURING FOOD DESERTS: RECOGNIZING THE POWER OF COMMUNITY-BASED SOLUTIONS Culturing Food Deserts: Recognizing the Power of Community- Based Solutions

CATHERINE BRINKLEY, SUBHASHNI RAJ and MEGAN HORST

Food deserts, places where residents lack nearby , have received a ention from the media, academics, policy-makers, and activists. The popular policy response is to establish a new . Yet, communities who live in food deserts may already have their own well-adapted strategies to access healthy food. In this article, we argue that policy-makers all too often overlook in situ opportunities, and may even disrupt low-cost healthy food access options with supermarket interventions. We use evidence to make two main points. First, we demonstrate the limitations of focusing on food deserts when interpreting -related disparities. By conducting a US national county-level multi-variable spatial regression analysis of socio-economic status, built environment and food environment factors, we determine that diet-related health outcomes do not clearly correlate with supermarket access. Instead, health outcomes are most strongly associated with income and race. This suggests that interventions to improve healthy eating should begin with a focus on anti-poverty. Second, we identify alternative paths, beyond supermarkets, to healthy food access through a literature review. We ground-truth our  ndings with interviews from and Planning Department o cials in seventeen counties with a high percentage of the population living in a food desert, but low levels of diet-related disease. Our research suggests new avenues for research and  nancing centred around existing community-based practices of established food banks and farm-to-market opportunities.

This research shifts the conversation on communities did not have the purchasing food desert classi cation and supermarket power to sustain a supermarket business intervention to supporting already existing, model. A review of forty-nine food desert prevalent, and well-adapted community studies providesprovides evidenceevidence wherethat lacklack ofof a e orts. The starting point for this conver- supermarket1 compounds individual dis- sation is a critique of the term, associated advantages in the ability to procure a healthy methodology, and various policy responses diet (Beaulac et al., 2009). Food is generally to ‘food deserts’. The term ‘food desert’ more ‘expensive and relatively unavailable’ originated in the early 1990s to describe (Cummins and Macintyre, 2002, p. 2115) lack of nearby a ordable, healthy food retail leading to fewer healthy food options (Cummins and Macintyre, 2002, p. 2115). (Beaulac et al., 2009) and quanti ed negative The term successfully drew a ention to the diet-related health outcomes, including spatial injustice of the food system, where and (Larson et al., 2009; more sparsely populated or low-income Morland et al., 2006). More recently, scholars

328BUILT ENVIRONMENTVOL43NO3 BUILT ENVIRONMENT VOL 43 NO369 3 URBAN AND REGIONAL FOOD SYSTEMS and public agencies have a empted to may have their own well-adapted systems operationalize the food desert term through of street food (Ammerman, 2012), smaller mapping and measurement. An example supermarkets (Raja et al., 2008), farmers’ exists with the Food Atlas project of the markets (Larsen andand Gilliland,Gilliand, 2009) and United States Department of Agriculture urban gardens (Wang et al., 2014). By focusing (USDA), which de nes a food desert as an myopically only on supermarkets in the food area without a supermarket within 1 or 10 environment, the term misses other important miles for urban and rural areas respectively sources of food in communities including (USDA, 2015). According to the USDA, 23.5 unhealthy food options, such as fast food ven- million Americans live in food deserts. dors. To this end, a recent review of sixteen We argue that the food desert term is studies finds that fast food consumption imprecise. First, the research and professional correlates withwith obesityobesity (Rosencheck,(Rosenheck, 2008). community often casually conflates spatial Another review of forty articles finds that fast and other, perhaps more important, factors food restaurants are more prevalent in low- of access. For example, Drewnowski et al. income and ethnic neighbourhoods (Fleisch- (2014) find that in Seattle and Paris, two cities hacker et al., 2011). with different built and food environments, Practice mirrors the above flaws in research. physical proximity to supermarkets was not For example, the Robert Wood Johnson predictive of obesity, whereas the following Foundation award winning County Health factors were: income levels, surrounding Rankings & Roadmaps tool includes a variable property values, and preferences for shop- ‘healthy food access’, which measures spatial ping at lower-cost stores. The focus on geo- proximity to a supermarket. While subtle, use graphic proximity to supermarkets also reduces of ‘food deserts’ as a healthy food indicator attention to explanatory variables related to arguably perverts policy-oriented results power structures, such as race and income, by overlooking spatial proximity to other which are rarely considered in models. healthy food sources as well as economic Indeed, a growing number of studies find barriers to access. Because the data for diet- socio-economic factors to be significant related health are primarily framed around predictors of health conditions compared to supermarket access, it is no wonder that the factors measuring built and food environ- chief policy response lies in publically-funded ments (Drenowski et al., 2014; Raj et al., supermarket financing. Chrisinger (2016) 2016). The result is that work on food deserts reviewed 126 new supermarket financing too often ‘bounds health problems within initiatives over the past decade, finding that low-income communities’ (Bedore, 2010), 80 per cent use federal public funding. A disregards the agency of such communities broad range of literature has documented to generate alternative responses that deserve that although new supermarkets can help support, and ignores the policies and actors alleviate food insecurity in some contexts which shape disparities in food systems and (Cummins et al., 2005), more often than not, accessibility (Shannon, 2014). new supermarkets alone are insufficient for Second, the food desert term is far too health behaviour change (Morland et al., 2006; narrow in its focus on supermarket retail. Wang et al., 2007; Larson et al., 2009; Walker The term ‘desert’ often implies a lack of life, et al., 2010; Sadler et al., 2013; Cummins et though biologists are apt to point out that the al., 2014; Elbel et al., 2015; Fuller et al., 2015; desert biome contains thriving ecosystems if Dubowitz et al., 2015). Complicating matters, you know how to look for them. Emphasis on many people do not shop at their closest measuring proximity to supermarkets ignores stores (Drewnowski et al., 2012; Cannuscio et other sources of food in neighbourhoods al., 2014). The failure of supermarket inter- and people’s resourcefulness. Food deserts ventions is compounded by several docu-

BUILT370 ENVIRONMENT VOL 43 NO 3 BUILTENVIRONMENTVOL43NO3293 CULTURING FOOD DESERTS: RECOGNIZING THE POWER OF COMMUNITY-BASED SOLUTIONS URBAN AND REGIONAL FOOD SYSTEMS mented cases where communities resist the publicinterviewed o cials public in outlierofficials counties. in outlier The counties. case Table 1.County-level descriptive statistics. The Supermarket proximity variable (PROX) uses the USDA use of eminent domain to acquire land and studiesThe case allow studies us toallow triangulate us to triangulate our regression our de nition in measuring the percentage of the population that is further than one mile from a supermarket in urban counties and more than 10 miles (16 km) from a supermarket in rural counties. public funds to finance supermarket inter- regression ndings, identify findings, potential identify omi potential ed variables omitted ventions (Sibilla, 2013; Horst et al., 2016). New ofvariables importance, of importance, and explore and explore sorting sorting or eco- or Category Variable Short Units Descriptive Data Data methods of promoting healthy food access logicaleco-logical fallacies fallacies (Robinson, (Robinson, 1950) where1950) wheregroup Name Variable Statistics Year Source Name are needed. orgroup spatial or spatial correlations correlations of the of average the average are mis- are Instead of focusing on supermarkets, we interpretedmisinterpreted at the at level the oflevel individuals. of individuals. In sum, Health Obesity OB % of adults that 31.02 ± 4.52 2013 CDC County argue that it would be more empowering to thisIn sum,mixed thismethods mixed approach methods highlights a proach how Outcome report BMI >= 30 level estimates engage communities in a conversation about communitieshighlights how achieve communities be er thanachieve expected better Diabetes DB % of people 11.24 ± 2.49 2013 diagnosed with what they are already doing that works, and diet-relatedthan expected health diet-related in the food health desert in thecontext, food diabetes be open to a wide range of possibilities to allowingdesert context, us to allowing consider us scaling to consider up and scaling out scale such ventures. By overlooking residents’ suchup and practices out such as alternativepractices asinterv alternativeentions Socio- Median MHHI US$ 43144.87 ± 10742.29 2010 other means of sourcing healthy food, policy- to supermarket  nancing. Our methods are economic household interventions to supermarket financing. Our variables income makers may unintentionally cause further discussed in greater detail below. methods are discussed in greater detail below. White WH Percentage 78.29 ± 19.89 2010 injustices in the local food system or miss an opportunity to augment an already existing African AA Percentage 8.75 ± 14.42 2010 Multi-Variable Regression Analysis American low-cost locally-based option, such as healthy No vehicle TRP Percentage 3.15 ± 3.21 2010 street food vending (Brinkley et al., 2013). We We employ a cross-sectional research design Lack of PROX Percentage 23.56 ± 20.25 2010 aim to expose such options for health with to study the e ect of ‘food desert’ conditions, supermarket the research below. socio-economic and built environment factors proximity on obesity and diabetes (health outcomes) in Convenience CONV No. of 0.59 ± 0.31 2012 USDA Food 3,142 counties in the United States. County- stores convenience Environment Methods level secondary data from the Centers for stores per Atlas 1,000 persons The goal of our research methods are two- Disease Control and Prevention (CDC), the Specialty SPC No. of specialty 0.05 ± 0.07 2012 fold. First, we build from the above literature USDA Food Environment Atlas and USDA’s stores* stores per 1,000 review from the public health and planning Economic Research Service2 provide esti- persons disciplines by conducting a quantitative mates for the prevalence of diagnosed adult analysis of diet-related health outcomes in diabetes and obesity. The Rural variable is Food Fast food FFR No. of fast food 0.58 ± 0.30 2012 environment restaurants restaurants per relation to socio-economic indicators and bivariate and denotes counties that are more 1,000 persons the percentage of the population living in urban (1–3) or rural (4–9) in nature, based Full service FSR No. of full 0.79 ± 0.59 2012 food desert conditions. The multivariate on and adjacency to a metro restaurants service statistical regression analysis is intended area represented in Rural-Urban Continuum restaurants per to highlight the importance of explicitly Codes (RUCC; Butler and Beale, 1990). Obesity 1,000 persons considering power dynamics that manifest is measured as the percentage of adults in a Farmers FRM No. of farmers 0.05 ± 0.09 2013 markets markets per in the built environment and food systems, county who report BMI >= 30, while diabetes 1,000 persons and thereby impact individual health. As we is the percentage of people diagnosed with suggest in the literature review, we suspect diabetes in a county. Explanatory variables Built environment Recreational RC No. of that proximity to a supermarket is not a encompass socio-economic characteristics facilities recreational very powerful indicator of poor diet-related along with the food and built environment facilities per health outcomes, and that socio-economic (table 1). Descriptive statistics are reported in 1,000 persons 0.07 ± 0.07 2012 characteristics are more tightly linked. The table 1. Not all explanatory variables tested Metropolitan analysis emphasizes that areas that are food were used in the regression analysis. For or non- metropolitan MT Dichotomous 0.37 ± 0.48 2013 USDA deserts but have low rates of poverty are just example, we considered childhood poverty variable Economic as likely to have positive health outcomes. and educational levels, but these variables Research Second, the regression analysis allowed us correlate with median household income Service RUCC to identify outlier case studies. To be er (MHHI), and they increase multicollinearity codes understand why food desertsdeserts areare notnot consist related- in the model. Note:* Specialty stores include retail primarily engaged in selling specialized lines of food, such as bakeries, meat and entlyto poor related health to outcomes,poor health we outcomes, interviewed we Nationally, about 30 per cent of adults are seafood markets, dairy stores, and produce markets.

330BUILT ENVIRONMENTVOL43NO3 BUILT ENVIRONMENT VOL 43 NO371 3 372 BUILTENVIRONMENTVOL43NO3 URBAN AND REGIONAL FOOD SYSTEMS

Table 1.County-level descriptive statistics. The Supermarket proximity variable (PROX) uses the USDA de nition in measuring the percentage of the population that is further than one mile from a supermarket in urban counties and more than 10 miles (16 km) from a supermarket in rural counties.

Category Variable Short Units Descriptive Data Data Name Variable Statistics Year Source Name Health Obesity OB % of adults that 31.02 ± 4.52 2013 CDC County Outcome report BMI >= 30 level estimates Diabetes DB % of people 11.24 ± 2.49 2013 diagnosed with diabetes

Socio- Median MHHI US$ 43144.87 ± 10742.29 2010 economic household variables income White WH Percentage 78.29 ± 19.89 2010 African AA Percentage 8.75 ± 14.42 2010 American No vehicle TRP Percentage 3.15 ± 3.21 2010 Lack of PROX Percentage 23.56 ± 20.25 2010 supermarket proximity Convenience CONV No. of 0.59 ± 0.31 2012 USDA Food stores convenience Environment stores per Atlas 1,000 persons Specialty SPC No. of specialty 0.05 ± 0.07 2012 stores* stores per 1,000 persons

Food Fast food FFR No. of fast food 0.58 ± 0.30 2012 environment restaurants restaurants per 1,000 persons Full service FSR No. of full 0.79 ± 0.59 2012 restaurants service restaurants per 1,000 persons Farmers FRM No. of farmers 0.05 ± 0.09 2013 markets markets per 1,000 persons

Built environment Recreational RC No. of facilities recreational facilities per 1,000 persons 0.07 ± 0.07 2012 Metropolitan or non- metropolitan MT Dichotomous 0.37 ± 0.48 2013 USDA variable Economic Research Service RUCC codes

Note:* Specialty stores include retail primarily engaged in selling specialized lines of food, such as bakeries, meat and seafood markets, dairy stores, and produce markets.

BUILT372 ENVIRONMENT VOL 43 NO 3 BUILTENVIRONMENTVOL43NO3313 CULTURING FOOD DESERTS: RECOGNIZING THE POWER OF COMMUNITY-BASED SOLUTIONS URBAN AND REGIONAL FOOD SYSTEMS considered obese, and 11 per cent of the popu- selected counties were contacted for semi- University of California, Davis Internal Review lation has been diagnosed with diabetes (table structured interviews. Board, interviews were solicited by email 1). The average median household income is and phone with County Public Health and $43,144. About 78 per cent of the population Planning Departments from twenty-two Interviews with Representatives identifies as white, with 9 per cent identifying counties (see appendix for interview solicita- in Select Counties as African American. On average, 3 per cent tion and format). Interviews were completed of the population lacks a car and lives more The interviews serve three purposes. First, for eight counties and only one representative than 1 mile (1.6 km) from a supermarket. interviewees helped verify the quantitative interview was obtained for nine other Another 24 per cent of the population lives  ndings by explaining whether the census counties (table 2, figure 1). Fifteen-minute far away from a supermarket, defined for data matched local understanding of the phone interviews with semi-structured inter- urban populations as living more than one andare the dependent vari- county. Second, interviewees provided insight view questions were recorded, transcribed mile from a supermarket, and for rural areas ables obesity and diabetes, respectively, for into the spatial arrangement of the built environ- and coded for seventeen counties. We trian- as more than 10 miles (16 km) from a super- counties nested in states. The random e ects ment, food environment and population gulated interview material by soliciting market. There are on average more con- portionofthemodelcontains,the characteristics of the counties to be er interviews from multiple county officials and venience stores (0.59) than specialty stores random intercept;, the random understand why counties lack supermarkets. verifying findings with secondary literature (0.05) and farmers markets (0.05) per 1,000 slope; and is the county level error term Finally, interviewees identi ed alternative through document review and web searches. people in the county. Similarly, there are more for the model. Fixed e ect terms are found pathways to healthy food in their counties. We full service restaurants (0.79) than fast food in table 1. speci cally asked about food retail options, Limitations restaurants (0.58) or recreational facilities policy interventions, and the presence and (0.07) per 1,000 people. Thirty-seven per cent Case Selection activities of organizations such as food banks. The mixed-methods model seeks to trian- of counties are metropolitan counties. Following a protocol approval from the gulate quantitative and qualitative  ndings, We ran a two-level generalized linear We focus our interviews on outlier counties mixed effects model with a Gaussian distri- to identify spatial and policy measures that bution and identity link to test the effect of may engender diet-related health in the food Table 2.Case study counties with a high proportion of the population living in a food desert (PROX), but socio-economic, food, and built environment desert context. Based on the literature, outlier be er-than-expected diet-related health outcomes. Rural-urban continuum codes (RUCC) denote counties measures on diabetes and obesity at the counties would be those where despite food that are more urban (1) or rural (9) in nature based on degree of urbanization and adjacency to a metro area. county level, accounting for state level cluster- desert conditions, be er than average health State County Obesity (%) Diabetes PROX (%) Population in Population RUCC Code ing using Stata/IC version 13.1 software. The outcomes (obesity and diabetes) are found. (%) Poverty (%) two-level generalized linear mixed effects These are interesting cases because they dis- model was fitted using the step-wise process prove the assumption that the presence of a Alaska Denali* 27.0 5.8 18.5 12.8 1,867 8 outlined by Tabachnick et al. (2001). Model food desert is linked with poor health out- Colorado Costilla 21.9 8.3 55.3 25 3,518 9 fit diagnostics (deviance, AIC and BIC) were comes. Some researchers hypothesize that rural Colorado Crowley* 21.6 6.0 30 28.3 5,322 8 estimated and examined. Multicollinearity populations have unique coping mechanisms Colorado Hinsdale* 20.5 6.6 26.8 7.5 813 9 was addressed by dropping interaction terms, to deal with this poor geographic access Colorado Lake* 17.5 5.4 26.5 12.2 7,306 6 such as the race variable ‘white’ (WH). (Smith and Morton, 2009). Such ‘uniqueness’ Colorado Mineral 18.1 6.8 22.5 6.8 721 9 Heteroscedasticity was controlled by employ- to cope has not been evaluated. Using the list Colorado Saguache 19.9 7 51.5 25.1 6,208 9 ing robust standard errors. We tested the function in STATA, we identi ed the counties Colorado San Juan* 19.3 6.7 19.4 16.5 692 9 covariance structures available in STATA that had a high prevalence of the population Colorado Washington 23.2 6.5 19.1 12.9 4,803 9 for the final model. The standard covariance in food desert areas and yet displayed low structure was the best suited for our models. levels of obesity and diabetes, based on the Idaho Boise* 22.4 9.6 30.6 15.9 6,795 2 Normality was tested by plotting the normal standard deviation from the means in those Kansas Riley 23.8 5.6 21.7 22.3 75,394 3 probability plots and histogram of the resi- variables (see table 2). The selection process Montana Daniels* 25.5 9.0 34.9 6.7 1,791 9 duals and looking at the scatter plots of the took into account the RUCC yielding seven Montana Granite 29.0 9.1 21.9 12.8 3,138 8 standardized residuals. Our final models are urban and  fteen rural cases. Our inquiry New Mexico Dona Ana 23.6 7.4 20.9 27.8 213,460 3 specified below: calls a ention to outliers as a means of under- New Mexico Taos 17.4 7.7 21.1 23.7 33,035 7 standing planning and policy in places that Texas Brazos 28.4 7.6 18.1 29.3 203,164 3 at a macro level would be considered a food Virginia Rappahannock 27.2 11.7 18.7 10.7 7,478 1 desert and should have higher obesity and diabetes rates. Public o cials in each of the Note:* denotes only one interview.

332BUILT ENVIRONMENTVOL43NO3 BUILT ENVIRONMENT VOL 43 NO373 3 374 BUILTENVIRONMENTVOL43NO3 URBAN AND REGIONAL FOOD SYSTEMS selected counties were contacted for semi- University of California, Davis Internal Review structured interviews. Board, interviews were solicited by email and phone with County Public Health and Planning Departments from twenty-two Interviews with Representatives counties (see appendix for interview solicita- in Select Counties tion and format). Interviews were completed The interviews serve three purposes. First, for eight counties and only one representative interviewees helped verify the quantitative interview was obtained for nine other  ndings by explaining whether the census counties (table 2, figure 1). Fifteen-minute data matched local understanding of the phone interviews with semi-structured inter- county. Second, interviewees provided insight view questions were recorded, transcribed into the spatial arrangement of the built environ- and coded for seventeen counties. We trian- ment, food environment and population gulated interview material by soliciting characteristics of the counties to be er interviews from multiple county officials and understand why counties lack supermarkets. verifying findings with secondary literature Finally, interviewees identi ed alternative through document review and web searches. pathways to healthy food in their counties. We speci cally asked about food retail options, Limitations policy interventions, and the presence and activities of organizations such as food banks. The mixed-methods model seeks to trian- Following a protocol approval from the gulate quantitative and qualitative  ndings,

Table 2.Case study counties with a high proportion of the population living in a food desert (PROX), but be er-than-expected diet-related health outcomes. Rural-urban continuum codes (RUCC) denote counties that are more urban (1) or rural (9) in nature based on degree of urbanization and adjacency to a metro area.

State County Obesity (%) Diabetes PROX (%) Population in Population RUCC Code (%) Poverty (%)

Alaska Denali* 27.0 5.8 18.5 12.8 1,867 8 Colorado Costilla 21.9 8.3 55.3 25 3,518 9 Colorado Crowley* 21.6 6.0 30 28.3 5,322 8 Colorado Hinsdale* 20.5 6.6 26.8 7.5 813 9 Colorado Lake* 17.5 5.4 26.5 12.2 7,306 6 Colorado Mineral 18.1 6.8 22.5 6.8 721 9 Colorado Saguache 19.9 7 51.5 25.1 6,208 9 Colorado San Juan* 19.3 6.7 19.4 16.5 692 9 Colorado Washington 23.2 6.5 19.1 12.9 4,803 9 Idaho Boise* 22.4 9.6 30.6 15.9 6,795 2 Kansas Riley 23.8 5.6 21.7 22.3 75,394 3 Montana Daniels* 25.5 9.0 34.9 6.7 1,791 9 Montana Granite 29.0 9.1 21.9 12.8 3,138 8 New Mexico Dona Ana 23.6 7.4 20.9 27.8 213,460 3 New Mexico Taos 17.4 7.7 21.1 23.7 33,035 7 Texas Brazos 28.4 7.6 18.1 29.3 203,164 3 Virginia Rappahannock 27.2 11.7 18.7 10.7 7,478 1

Note:* denotes only one interview.

BUILT374 ENVIRONMENT VOL 43 NO 3 BUILTENVIRONMENTVOL43NO3333 CULTURING FOOD DESERTS: RECOGNIZING THE POWER OF COMMUNITY-BASED SOLUTIONS

Figure 1.Counties included in the case study. but is limited in available national-level spatial variations in socio-economic group- data, the county unit of analysis, practitioner ings, the built environment or the food environ- understandings of their counties, and limited ment. Moreover, obesity and diabetes health case studies. As the case studies revealed, outcomes are multi-causal. Thus, the modest the model omits important demographic goal of this regression analysis is to explore variables, such as length of residence in the broad usefulness of socio-economic, built thecounty, county, as aswell well as as food food environment environment vari- and food environment variables as predictive ables, such the prevalence of roadside stands, in comparison to one another. In particular, and built environment variables, such as recent literature reviewed in the introduction hiking trails. Indeed, diet-related health out- indicates that research and practice often comes are the result of many cumulative omit explanatory variables related to power decisions at the individual-level. County- structures, such as race and income. level data may be too large a scale to pick up Interviews in study counties seek to

334BUILT ENVIRONMENTVOL43NO3 BUILT ENVIRONMENT VOL 43 NO375 3 URBAN AND REGIONAL FOOD SYSTEMS correct for the model’s lack of longitudinal due to land-use policies, such as redlining, data, omitted variables and sorting effects as in which mortgages are denied in certain populations move. Yet, a larger sample of case locations based on skin colour. The legacy studies is needed to further explain and verify of explicitly racist policies has left long- quantitative findings. The cases identified term and on-going disparities across food tend to be skewed regionally (figure 1). and built environments (Eisenhauer, 2001). For example, the case selection yields more Logically, the quality and quantity of food rural than urban counties because food and built environment features in a county deserts are more likely to be located in rural may be mediated by income levels, where areas. According to the USDA, food deserts higher income communities cultivate more are distributed among 448 counties in the opportunities to support healthy lifestyles. United States, of which 98 per cent are rural. This may explain the impact of recreational Despite using an indicator that highlights the facilities. For an additional recreational lack of supermarkets in rural areas, there is facility per 1,000 persons there is 4.19 per cent a puzzling over emphasis on urban areas in less obesity and 1.08 per cent less diabetes. food desert research and policy responses The descriptive data indicate the availability (McEntee and Agyeman, 2010; Walker et of recreational facilities to be sparse and al., 2010). Since rural counties are under- highly variable among the counties. The per- represented in food systems research the centage of the population with lack of car skewness in the case selection is important access and the rural nature of the county to ensure rural county considerations are were not predictive. factored in the qualitative analysis. Though In contrast, the effects of food and built this study focuses across urban and rural environment variables are less predictive counties, future research may be needed across both diet-related health outcomes. to drill down to smaller geographic levels. Proximity to a supermarket, while statistic- Companion pieces outside of the US can ally significant, is not practically meaningful determine external validity internationally. given the small coefficient sizes. A percentage increase in the number of people who do not live in close proximity to a supermarket is Findings and Discussion associated withwith 0.010.01 per per cent cent less less in obesity Across models, socio-economic predictors and a0.00 0.00 per per cent cent decrease decrease in in diabetes.diabetes. While showed similar signs for predicting diet- a 1 per cent lower incidence of obesity related health outcomes when compared strongly correlates with two more full to built and food environment variables. service restaurants and one farmers’ market Every $10,000 more in median household per 500 residents, the effects of these food income is associated with 1 per cent less environment variables are insignificant on obesity and 0.71 per cent less diabetes. Every diabetes outcomes. Instead, fewer convenience percentage increase in the proportion of stores and more fast food restaurants are the African American population correlates associated with lower diabetes incidences. with greater prevalence of obesity (0.08 For example, for every additional fast food per cent) and diabetes (0.03 per cent). Our restaurant per 1,000 people, counties have a  ndings are supported by recent literature 0.65 per cent less diabetes. Every convenience that emphasizes the importance of including store per 1,000 people correlations with 0.75 income and race variables in understanding per cent more obesity and 0.79 per cent more health (Drenowski et al., 2014; Raj et al., diabetes (table 3). 2016). It is important to note that in many Because built environment obstacles, such places in the US, race and income are highly as car access, were not predictive and food spatially correlated (Downey, 1998) in part environment predictors were divided, we

BUILT376 ENVIRONMENT VOL 43 NO 3 BUILTENVIRONMENTVOL43NO3353 CULTURING FOOD DESERTS: RECOGNIZING THE POWER OF COMMUNITY-BASED SOLUTIONS

Table 3.E ects of food and built environment on diet-related health outcomes in the United States.

Obesity Diabetes Fixed E ects Coe cient SE p-value Coe cient SE p-value MHHI –1.00 0.096 0 –0.71 0.079 0 African American 0.08 0.011 0 0.03 0.004 0 No vehicle 0.07 0.049 0.155 0.05 0.052 0.316 Supermarket proximity –0.01 0.003 0 0 0.002 0.017 Convenience stores 0.75 0.264 0.005 0.79 0.142 0 Specialty stores –2.03 0.864 0.019 –0.37 0.359 0.299 Full service restaurants –1.2 0.235 0 –0.09 0.071 0.177 Fast food restaurants –0.26 0.291 0.377 –0.65 0.129 0 Recreational facilities –4.19 1.147 0 –1.08 0.391 0.006 Farmers’ markets –2.17 0.476 0 0.1 0.298 0.729 Metro –0.41 0.195 0.037 –0.25 0.073 0.001 Intercept 35.55 0.767 0 13.86 0.499 0

Random E ects Level 2: Intercept 9.32 2.471 5.32 1.954 MHHI 0.24 0.092 0.1 0.045 Level 1: Residuals 7.04 0.363 1.79 0.089

Deviance Statistics Deviance 15281.3 11037.7 AIC 15311.3 11067.7 BIC 15402.1 11158.4

ascertain that socio-economic variables are diabetic) andand ruralrural (35.5 counties per cent(35.5 obese, per cent13.9 more informative in understanding health. obese,per cent) 13.9 counties, per cent) our counties, regression our results regression also These results indicate that to provide resultsshow that also peopleshow that experience people experience their localized their enduring systemicsystemic changeschanges thatto address localizedfood systems food unevenly systems depending unevenly ondepending income, disparate health outcomes, policies need to onrace income,and the opportunitiesrace and the (farmers’ opportunities markets addressexplicitly explicitly address economic access and race. (farmers’and recreational markets facilities) and recreational available facilities) in their In support, the level two intercepts for both availablecommunities. in their There communities. is also some There variation is also models indicate significant between-state someremaining variation between remaining counties, between indicating counties, our variation in median household income. State indicatingmodel does our not model fully does explain not fullythe complexity explain the political structures structures confer confer different different values social complexityof diet-related of diet-related health outcomes. health outcomes. which,values which,in turn, in turn, affect affect how how resources resources are are directed and distributed among communities. Interview Findings Indeed, income may moderate values and, in turn, influence state-level variation in The interviews of representatives from seven- regional food and health policies which teen counties allowed us to further examine alter social support, infrastructure and food place-based and population-based character- policies broadly. While results (see table 3) istics that create food desert conditions and indicate similarsimilar diet-related diet-related disease prevalence prevalence for in uence health outcomes. We identi ed the urbanfor urban (35.14 (35.14 per per cent cent obese, obese, 13.61 13.61 perper cent following key themes as important to be er

336BUILT ENVIRONMENTVOL43NO3 BUILT ENVIRONMENT VOL 43 NO377 3 URBAN AND REGIONAL FOOD SYSTEMS understanding food deserts and healthy brand-name supermarkets from locating in food access: varied underlying explanatory the county to protect smaller-scale, locally models for the lack of supermarkets; the managed stores from competition. Thus, the overlooked role of sorting; and the wide built environment is mediated through social variety of strategies employed by residents and economic control, a finding supported by and community leaders in so-called food the regression models. deserts to access healthy food. Second, interviewees attributed better First, interview data suggest that both built than expected health outcomes to the effects environment and socio-economic factors con- of sporting and physical activity levels over tribute to a lack of supermarkets in the case the food environment. Physical labour and study counties. In half the counties, existing outdoor lifestyles were cited as reasons for large lot land uses, such as a military base, better-than-expected health outcomes in ten prison, state park or university, prohibited out of the thirteen counties that responded to supermarkets from locating nearby due this question, with the physical environment to the lack of sufficient open land and low contributing recreational opportunities customer densities. Population characteristics in half of the counties. In three counties, were given as explanations for food desert interviewees believed that the presence conditions in two-thirds of the counties, of universities or military bases gave the with explanations ranging from a lack of pur- population a more favourable physical chasing power due to poverty (seven counties), fitness average average than than it might might have otherwisehave otherwise itinerant populations engaged in seasonal been the case. Also notable is that the state work (three counties), inaccurate census data of Colorado is overrepresented amongst the for areas where there are undocumented pop- outlier case counties. Colorado has ranked ulations (two counties), and more people per as the fittest state in the union for the past household in multi-generational housing decade according to the National Centers for leading to under-estimates of consumer base Disease Control and Prevention, and has the (one county). Two counties with universities lowest rates of obesity and adult diabetes indicated that students do not purchase at and second-lowest rate of supermarkets, but prefer street food, fast food (2016). Interviewees from Colorado counties and restaurant venues for the convenience of indicated that the better than expected diet- ready-made food. This sentiment is echoed related health outcomes could be due to in a recent Atlantic article titled, ‘Why Do self-selection, as health-motivated residents Millennials HateHate Groceries?’Groceries?’ (Thompson, (Thomson, 2016). choose to live near recreational opportunities Further, these results suggest that a afforded in the mountains, such as hiking nuanced understanding of the factors leading and skiing, or work directly with resort facil- to a lack of supermarkets is necessary. In ities. Due to a lack of health services, many some cases, land use and purchasing power residents with health conditions choose to rationales ran in tandem with sparsely pop- relocate to lower altitudes, further skewing ulated low-income communities unable to the place-based health outcomes with resi- support the conventional supermarket busi- dents self-sorting as fitter, younger residents ness model. In other cases, the built environ- moving in and older, less healthy residents ment and social networks worked to opposite move away to be closer to health services. ends. In one largely rural, low-income county, Respondents believed these factors were several wealthy residents heavily influence more influential than the food environment land-use decisions to keep the county rural on positive health outcomes. and historic in character while also support- Third, residents living in food deserts use ing a Michelin star restaurant. The wealthier a variety of strategies to access food. Inter- community intentionally prevented large viewees from all the counties reported

BUILT378 ENVIRONMENT VOL 43 NO 3 BUILTENVIRONMENTVOL43NO3373 CULTURING FOOD DESERTS: RECOGNIZING THE POWER OF COMMUNITY-BASED SOLUTIONS bulk shopping at large stores outside the conditions represents a largely overlooked community as the primary mode of food mode of intervening in healthy food policy. procurement. Residents also used smaller- Many food banks were heavily embedded scale, local supermarkets in the county for in publicly-run programmes. For example, staple items. Only two of the counties had public agencies run the federal food sub- public transit, and none had formal ride- sistence programmes (Women Infant and sharing programmes. Several counties used Children and Supplemental Nutritional social media networks, like facebook, to Assistance) while coordinating free breakfasts solicit ride-sharing and to pick up groceries and lunches through public schools in or medications for community members. combination with food banks. In one largely Because supermarket shopping is the primary rural, low-income county, several wealthy form ofof obtainingobtaining food, food, many many scholars food scholars would residents ‘single-handedly’ finance healthy wouldsay that say a focus that onfocus food on deserts food desertsand lack and of food options, such as the school lunch lacksupermarkets of such stores is warranted.is warranted. Our Our results, and weekend backpack of healthy food however, point to other highly prevalent food programme run through the local food bank. acquisition practices when such convenience The counties were engaged in a variety of is absent. About half the counties had interventions to increase access to healthy established farmers’ markets or roadside food and to enhance healthy eating. Counties farm stands, and all but two had seasonal reported involvement from public, private, farm-to-market opportunities. For example, and non-profit agencies across community- the largely secluded ethno-religious Hutterite led efforts. University extension offered farming community comes to the main town master gardener, 4H, and food preservation in Daniels County, Montana weekly during courses to help stretch home-grown sup- the growing season to sell produce, baked plies through the winter months for self- goods, eggs and milk from the back of provisioning. Private companies, like Alison’s their truck. Amish farmers in other parts of Pantry and local grocery co-ops, sold bulk the country are similarly engaged in direct meals directly to homes in two Colorado marketing practices (Kraybill et al., 2010; counties. An example of the community-led Brinkley, 2017). All but one county had some efforts which combine government, non- form of road-side vending, with the majority profit and business institutions is exemplified being seasonal, fresh produce trucks that in San Juan County, Colorado, where a local source from local farms or produce terminals restaurant coordinates with the food pantry and park temporarily for sales. Gardening run by the Catholic Church and the Area was a prevalent practice in all but three Agency for Aging to run a weekly senior counties. In locales where growing seasons dinner. were short and altitudes prevented vegetable and fruit growing, backyard poultry were Conclusion popular. One-third of the counties reported hunting and gathering as important methods This research asks researchers and policy- of supplementing their diets. makers to move beyond the narrow focus All but two counties had food banks, with on supermarket proximity and acknowledge varying degrees of involvement in fresh food the forces shaping the spatial inequities in purchasing. In comparison, none of the counties healthy, a ordable food access, namely, race reported a food policy council. Food bank and poverty. As the regression model shows, management varied: some were run through purchasing power a ributes, represented by non-profits, others through churches, public median household income, more strongly agencies, or combinations of the above. The correlate with diet-related outcomes than other central role of the food bank in food desert factors, including supermarket proximity.

338BUILT ENVIRONMENTVOL43NO3 BUILT ENVIRONMENT VOL 43 NO379 3 URBAN AND REGIONAL FOOD SYSTEMS

As such, anti-poverty measures, such as Independent Cities (1 in Maryland, 1 in Missouri, minimum wage raises, and price support 1 in Nevada, and the remainder in Virginia), 1 Dis- for farmers’ markets (Young et al., 2013) trict – the Federal District or District of Columbia. We omit one US county (Kalawao, Hawaii) which are supported models for improving diets did not have median household income data. for low-income residents. Such policies will support low-income people regardless of how close they live to a supermarket. The case studies emphasize that food deserts REFERENCES di er vastly in their underlying physical, Ammerman, A.S. (2012) Accessing nutritious food social and policy characteristics, making in low-income neighborhoods. North Carolina  ndings from various studies on food deserts Medical Journal, 7(5), pp. 384–384. di cult to generalize beyond the need to give Beaulac, J., Kristjansson, E. and Cummins, S. (2009) communities the chance to innovate as suits A systematic review of food deserts, 1966–2007. Preventing Chronic Disease, 6(3), A105. their needs. In conclusion, it is imperative that future Bedore, M. (2010) Just urban food systems: a new direction for food access and urban social research explicitly considers variables related justice. Geography Compass, 4(9), pp. 1418–1432. to power structures in the food system, Brinkley, C. (2017) Visualizing the social and geo- namely income and race. These variables graphical embeddedness of local food systems. are all too often left out, and directly effect affect Journal of Rural Studies, 54, pp. 314–325. the research and policy recommendations Brinkley, C., Chrisinger, B. and Hillier, A. (2013) made. There are many pathways to healthy Tradition of healthy food access in low-income food access. Where supermarkets are lacking, neighborhoods: price and variety of curbside residents tend to travel long distances and produce vending compared to conventional retailers. Journal of Agriculture, Food Systems, and coordinate within their community for bulk Community Development, 4(1), 155–169. buying while relying on local smaller-scale Butler, M. A. and Beale, C.L. (1990) Rural-Urban Con- supermarkets for staple items. The com- tinuum Codes for Metro and Nonmetro Counties. monalities among the counties point to farm- No. 9028. Washington DC: US Department to-market, self-provisioning and food banks of Agriculture, Economic Research Service, as strategies for food acquisition. Future Agriculture and Rural Economy Division. research can help examine the potential of Cannuscio, C.C., Hillier, A., Karpyn, A. and such avenues to supplement food budgets, Glanz, K. (2014) The social dynamics of healthy food shopping and store choice in an urban and how such efforts can be supported or environment. Social Science & Medicine, 122, scaled-up to impact health outcomes. pp. 13–20. Chrisinger, B.W. (2016) Taking Stock of New Super- markets in Food Deserts: Pa erns in Development, NOTES Financing, and Health Promotion. Bedminster, NJ: The Investment Center. Available at: h p:// 1.We intentionally use the word supermarket, www.frbsf.org/community-development/ not grocery store for two main reasons. First, it is publications/working-papers/2016/august/new- the term used by the USDA in their de nition of supermarkets-in-food-deserts-development- food deserts. Second, as part of our argument in  nancing-health-promotion/. the article, the focus on supermarkets, which are Cummins, S. and Macintyre, S. (2002) A systematic larger in oor area and sales volume, can over- study of an urban foodscape: the price and look the presence of smaller stores with a variety availability of food in greater Glasgow. Urban of ownership models. Studies, 39(11), pp. 2115–2130. 2.There are 3,143 counties/equivalent counties Cummins, S., Findlay, A., Pe icrew, M. and in the United States (limited to the 50 states) – Sparks, L. (2005) Healthy cities: the impact of 3,007 counties, 16 Boroughs in Alaska, 11 Census food retail-led regeneration on food access, Areas in Alaska (for areas not organized into Bor- choice and retail structure. Built Environment, oughs by the State), 64 Parishes in Louisiana, 42 31(4), pp. 288–301.

BUILT380 ENVIRONMENT VOL 43 NO 3 BUILTENVIRONMENTVOL43NO3393 CULTURING FOOD DESERTS: RECOGNIZING THE POWER OF COMMUNITY-BASED SOLUTIONS

Cummins, S., Flint, E. and Ma hews, S.A. (2014). Amish enterprise: the collective power of ethnic New neighborhood grocery store increased entrepreneurship. Global Business and Economics awareness of food access but did not alter Review, 12(1/2), pp. 3–20. dietary habits or obesity. Health A airs, 33(2), Larsen, K. and Gilliland, J. (2009) A farmers’ pp. 283–291. market in a food desert: evaluating impacts on Downey, L.CULTURING (1998) Environmental FOOD DESERTS: RECOGNIZINGinjustice: is THE POWERthe OF price COMMUNITY-BASED and availability SOLUTIONS of healthy food. Health race or income a be er predictor? Social Science & Place, 15(4), pp.1158–1162. Cummins,Quarterly S.,, 79 Flint,(4), pp.766–778.E. and Ma hews, S.A. (2014). Larson,Amish N., enterprise: Story M.the collectiveand Nelson power M. of (2009)ethnic Drewnowski,New neighborhood A., Aggarwal, grocery A., storeHurvi increased , P.M., Neighborhoodentrepreneurship. environments: Global Business disparities and Economics in Monsivais,awarenessCULTURING P.of andfood Moudon, FOODaccess DESERTS: A.V.but (2012)did RECOGNIZING not Obesity alter THE POWERaccessReview OF COMMUNITY-BASED ,to 12 (1/2),healthy pp. foods 3–20. SOLUTIONS in the U.S. American anddietary supermarket habits or obesity.access: proximityHealth A airs or , price?33(2), Larsen,Journal K. of andPreventive Gilliland, Medicine J. (2009), 36(1), App.74–81. farmers’ pp. 283–291. Cummins,American S., Journal Flint, E.of andPublic Ma Health hews,, S.A.102(8), (2014). pp. McEntee,marketAmish J.enterprise:in and a food Agyeman, desert: the collective evaluatingJ. (2010) power Towards impacts of ethnic theon Downey,e74–e80.New neighborhoodL. (1998) Environmental grocery store injustice: increased is developmenttheentrepreneurship. price and availabilityof a GlobalGIS method Businessof healthy for and food.identifying Economics Health Drewnowski,raceawareness or income A.,of Moudon, fooda be eraccess predictor? A.V., but Jiao, did SocialJ., Aggarwal,not Science alter rural&Review Place ,food, 1215(1/2),(4), deserts: pp.1158–1162. pp. 3–20. geographic access in A.,Quarterlydietary Charreire, habits, 79(4), orH. pp.766–778. obesity.and Chaix, Health B. A(2014) airs, 33Food(2), Larson,Larsen,Vermont, K.N., andStoryUSA. Gilliland, M.Applied and J. Nelson(2009)Geography AM. farmers’, (2009)30(1), Drewnowski,environmentpp. 283–291. A., and Aggarwal, socioeconomic A., Hurvi status , in P.M., u- Neighborhoodpp.165–176.market in a food environments: desert: evaluating disparities impacts onin Downey,enceMonsivais, obesity L. (1998)P. rates and Moudon,Environmentalin Sea le A.V.and (2012)in injustice:Paris. Obesity Inter- is Morland,accessthe price toK., and healthy Diez availability Roux, foods A.V.,ofin healthy the Wing U.S. food. S.American (2006)Health nationalandrace orsupermarket income Journal aof be Obesityaccess: er predictor?, 38proximity(2), pp. Social 306–314. or Scienceprice? JournalSupermarkets,& Place ,of 15 Preventive(4), pp.1158–1162. other Medicine food stores,, 36(1), and pp.74–81. obesity: DubowiAmericanQuarterly , T., , Journal79 Ghosh-Dastidar,(4), pp.766–778. of Public Health M., Cohen,, 102(8), D.A., pp. McEntee,Larson,the atherosclerosis N., J. andStory Agyeman, M. risk and in J. communities(2010)Nelson Towards M. study.(2009) the Drewnowski,e74–e80.Beckman. R.,A., Steiner,Aggarwal, E.D., A., Hunter, Hurvi G.P., P.M., et developmentAmericanNeighborhood Journal of environments: aof GIS Preventive method Medicinefordisparities identifying, 30 (4),in Drewnowski,alMonsivais,. (2015) Diet A.,P. andMoudon, and Moudon, perceptions A.V., A.V. Jiao, (2012) changeJ., Aggarwal, Obesity with ruralpp.access 333–339. foodto healthy deserts: foods geographic in the U.S. access American in A.,supermarketand Charreire,supermarket introduction H. access:and Chaix, proximityin a foodB. (2014) desert, or price?Food but NationalVermont,Journal Centersof PreventiveUSA. for Applied DiseaseMedicine ,GeographyControl 36(1), pp.74–81. and, 30 Pre-(1), environmentnotAmerican because Journal of and supermarket ofsocioeconomic Public Health use. Health ,status 102(8), A in airspp. u-, McEntee,ventionpp.165–176. J. (2016) and Agyeman, Adult Obesity J. (2010) Maps . TowardsAvailable the at: ence34e74–e80.(11), obesity pp. 1858–1868. rates in Sea le and in Paris. Inter- Morland,hdevelopment ps://www.cdc.gov/obesity/data/prevalence- K., Diez of aRoux, GIS method A.V., Wing for identifying S. (2006) Eisenhauer,Drewnowski,national Journal E. A., (2001) Moudon, of Obesity In poor A.V.,, 38health:(2), Jiao, pp. supermarketJ., 306–314.Aggarwal, maps.html.Supermarkets,rural food deserts: other foodgeographic stores, and access obesity: in DubowiredliningA., Charreire, , T., and Ghosh-Dastidar, urban H. and . Chaix, M., GeoJournal B. Cohen,(2014), 53D.A.,Food(2), Raj,theVermont, S., atherosclerosisRaja, S. USA.and Dukes, Appliedrisk B.in (2016)communitiesGeography Bene, cial study.30 but(1), pp.Beckman.environment 125–133. R., andSteiner, socioeconomic E.D., Hunter, status G.P. in u-et constrained:Americanpp.165–176. Journal role ofof urban Preventive agriculture Medicine programs, 30(4), Elbel,alence. (2015)B., obesity Moran, Diet rates andA., in perceptionsDixon,Sea le andL.B., in change Kiszko,Paris. Inter-with K., Morland,inpp. supporting 333–339. K., Diez healthy Roux, eatingA.V., Wingamong S. youth.(2006) Cantor,supermarketnational JournalJ., Abrams, introduction of Obesity C. et ,al 38in. (2015)(2), a food pp. Assessment desert,306–314. but NationalJournalSupermarkets, Centersof Hunger otherfor &Disease foodEnvironmental stores, Control and Nutritionand obesity: Pre-, Dubowiofnot a becausegovernment-subsidized , T., Ghosh-Dastidar, of supermarket use.M.,supermarket HealthCohen, A D.A., airsin a, vention12the(3), atherosclerosis pp. (2016) 1–23. Adult risk Obesity in communities Maps. Available study. at: 34(11), pp. 1858–1868. high-needBeckman. areaR., Steiner,on household E.D., Hunter,food availability G.P. et Raja,hAmerican S.,ps://www.cdc.gov/obesity/data/prevalence- Ma, C.Journal and Yadav,of Preventive P. (2008) Medicine Beyond, 30 food(4), Eisenhauer,andal. (2015) children’s E. Diet (2001) anddietary In perceptionspoor intakes. health: Publicchangesupermarket Health with maps.html.desertspp. 333–339. measuring and mapping racial dis- Nutritionredliningsupermarket, and18(15), introductionurban pp. nutrition. 2881–2890. in a GeoJournal food desert,, 53 but(2), Raj,Nationalparities S., Raja, Centers in S. neighborhoodand forDukes, Disease B. (2016)food Control environments. Bene and cial Pre- but Gundersen,pp.not 125–133.because C., ofWaxman supermarket E., Engelhard, use. Health E., ASatoh, airs, constrained:Journalvention of (2016) Planning role Adult of Education urban Obesity agriculture Mapsand Research. Available programs, 27(4), at: Elbel,A.34(11), andB., pp.Moran,Chawla, 1858–1868. A.,N. (2013)Dixon, Map L.B., the Kiszko, meal gapK., pp.inh ps://www.cdc.gov/obesity/data/prevalence-supporting 469–482. healthy eating among youth. Eisenhauer,2013:Cantor, food J., E. Abrams, insecurity(2001) In C. poor estimateset al health:. (2015) at supermarket Assessmentthe county Robinson,Journalmaps.html. W.S.of Hunger (1950) &Ecological Environmental correlations Nutrition and, level.ofredlining a government-subsidized Feeding and urban America. nutrition. Available supermarket GeoJournal at: ,h 53 inp://(2), a Raj,the12 S.,(3), Raja,behavior pp. S.1–23. and of Dukes, individuals. B. (2016) American Bene cial Socio- but www.feedingamerica.org/hunger-in-america/high-needpp. 125–133. area on household food availability Raja,logicalconstrained: S., Ma, Review C. roleand, 15 ,ofYadav, 351–357/ urban P. agriculture (2008) Beyond programs food our-research/map-the-meal-gap/2013/map-the-and children’s dietary intakes. Public Health Elbel, B., Moran, A., Dixon, L.B., Kiszko, K., Rosenheck,desertsin supporting measuringR. (2008) healthy Fast and food eatingmapping consumption among racial youth. anddis- meal-gap-2013-exec-summ.pdf.Nutrition, 18(15), pp. 2881–2890. Cantor, J., Abrams, C. et al. (2015) Assessment increasedparitiesJournal ofin caloric Hungerneighborhood intake: & Environmental a systematicfood environments. Nutritionreview of, Gundersen,Fleischhacker,of a government-subsidized C., S.E.,Waxman Evenson, E., Engelhard, K.R.,supermarket Rodriguez, E., Satoh, in a aJournal12 trajectory(3), pp. of Planning1–23. towards Education weight and gain Research and obesity, 27(4), A.D.A.high-need and and Chawla, Ammerman, area on N. household (2013) A.S. (2011)Map food the A availability systematicmeal gap Raja,risk.pp. S., 469–482. ObesityMa, C. Reviewsand Yadav,, 9(6), P. pp. (2008) 535–547. Beyond food 2013:reviewand children’s food of insecurityfast dietaryfood estimatesaccess intakes. studies. atPublic the ObesitycountyHealth Sadler,Robinson,deserts R.C., W.S.measuring Gilliland, (1950) J.A. Ecologicaland and mapping Arku, correlations G. (2013)racial Com- anddis- level.ReviewsNutrition Feeding, , 18(5),(15), pp.America. pp. e460–e471. 2881–2890. Available at: h p:// 12 munitytheparities behavior developmentin neighborhood of individuals. and thefood inAmerican environments. uence of Socio- new www.feedingamerica.org/hunger-in-america/ Fuller,Gundersen, D., Engler-Stringer,C., Waxman E., Engelhard,R. and Muhajarine, E., Satoh, foodlogicalJournal retail Review,Review of Planning sources, 15,, pp.351–357/ Educationon 351–357. the price and and Research availability, 27(4), our-research/map-the-meal-gap/2013/map-the- N.A. (2015)and Chawla, Examining N. (2013) food purchasingMap the meal pa ernsgap Rosenheck,ofpp. nutritious 469–482. R. (2008) food. FastJournal food of Urbanconsumption A airs, 5and(4), meal-gap-2013-exec-summ.pdf. from2013: salesfood datainsecurity at a full-service estimates atgrocery the county store Robinson,pp.471–491.increased W.S. caloric (1950) intake: Ecological a systematic correlations review and of Fleischhacker,interventionlevel. Feeding inS.E., aAmerica. former Evenson, food Available desert.K.R., Rodriguez,Preventitiveat: h p:// Shannon,athe trajectory behavior J. (2014) towards of Food individuals. deserts:weight governinggain American and obesity Socio- D.A.Medicinewww.feedingamerica.org/hunger-in-america/ and ReportsAmmerman,, 2, pp.164–169. A.S. (2011) A systematic inrisk.logical the Obesity Reviewneoliberal Reviews, 15, city351–357/, 9. (6),Progress pp. 535–547. in Human Geo- reviewour-research/map-the-meal-gap/2013/map-the- of fast food access studies. Obesity Horst, M., Raj, S. and Brinkley, C. (2016) Ge ing Sadler,Rosenheck,graphy, R.C., 38 Gilliland,R.(2), (2008) pp. 248–266. J.A.Fast and food Arku, consumption G. (2013) Com- and Reviewsmeal-gap-2013-exec-summ.pdf., 12(5), pp. e460–e471. outside the supermarket box: alternatives to Sibilla,munityincreased N. (2013) development caloric Philadelphia intake: and a wantsthesystematic in to uence use review eminent of new of Fuller,Fleischhacker,‘food D., deserts’. Engler-Stringer, S.E., Progressive Evenson, PlanningR. K.R.,and , Muhajarine,207Rodriguez,(Spring), domainfooda trajectory retail to sources towardsturn onan weighttheartist’s price gain andstudio and availability obesityinto a N.pp.D.A. (2015) 9–12. and Ammerman,Examining food A.S. purchasing (2011) A systematic pa erns parkingofrisk. nutritious Obesity lot and Reviewsfood. supermarket. Journal, 9(6), ofpp. Urban 535–547.Forbes A Magazineairs, 5(4), fromreview sales of datafast atfood a full-service access studies. grocery Obesity store Kraybill, D.B., Nolt, S.M. and Wesner, E.J. (2010) Sadler,(online).pp.471–491. R.C., Gilliland,Available J.A. at: andh p://www.forbes.com/ Arku, G. (2013) Com- interventionReviews, (5), in pp.a former e460–e471. food desert. Preventitive 12 Shannon,munity J. development (2014) Food deserts: and the governing in uence obesityof new Medicine Reports, 2, pp.164–169. Fuller,BUILTENVIRONMENT D., Engler-Stringer,VOL43 NOR. 3and Muhajarine, infood the retail neoliberal sources city on. theProgress price andin Human availability Geo-381 Horst,N. (2015) M., Raj, Examining S. and Brinkley, food purchasing C. (2016) paGe erns ing graphy,of nutritious 38(2), food.pp. 248–266. Journal of Urban A airs, 5(4), outsidefrom sales the data supermarket at a full-service box: alternatives grocery store to Sibilla,pp.471–491. N. (2013) Philadelphia wants to use eminent ‘food deserts’. Progressive Planning, (Spring), intervention in a former food desert.207 Preventitive Shannon,domain J. (2014)to turn Food an deserts: artist’s governing studio obesityinto a pp. 9–12. Medicine Reports, 2, pp.164–169. parkingin the neoliberal lot and supermarket. city. Progress Forbesin Human Magazine Geo- Kraybill,Horst, M., D.B., Raj, Nolt,S. and S.M. Brinkley, and Wesner, C. (2016) E.J. Ge (2010) ing (online).graphy, 38 Available(2), pp. 248–266. at: h p://www.forbes.com/ outside the supermarket box: alternatives to Sibilla, N. (2013) Philadelphia wants to use eminent 340BUILT ‘food ENVIRONMENT deserts’. ProgressiveVOL43 NOPlanning3 , 207(Spring), domain to turnBUILT an ENVIRONMENT artist’s studio VOL 43into NO 381 a 3 pp. 9–12. parking lot and supermarket. Forbes Magazine Kraybill, D.B., Nolt, S.M. and Wesner, E.J. (2010) (online). Available at: h p://www.forbes.com/

BUILTENVIRONMENTVOL43NO3 381 URBAN AND REGIONAL FOOD SYSTEMS

sites/instituteforjustice/2013/12/03/philadelphia- Wang, H., Qiu, F. and Swallow, B. (2014) Can com- wants-to-use-eminent-domain-to-turn-an- munity gardens and farmers’ markets relieve artists-studio-into-a-parking-lot-and-super food desert problems? A study of Edmonton, market/#7c5995d28576. Canada. Applied Geography, 55, pp. 127–137. Smith, C. and Morton, L.W. (2009) Rural food Young, C.R., Aquilante, J.L., Solomon, S., Colby, L., deserts: low-income perspectives on food access Kawinzi, M.A., Uy, N. et al. (2013) Improving in Minnesota and Iowa. Journal of Nutrition fruit and vegetable consumption among low- Education and Behavior, 41(3), pp.176–187. income customers at farmers markets: Philly Tabachnick, B.G., Fidell, L.S. and Osterlind, S.J. Food Bucks, Philadelphia, Pennsylvania, 2011. (2001) Using Multivariate Statistics. Boston: Allyn Preventing Chronic Disease, 10, e34. and Bacon. Thompson, D. (2016) Why do millennials hate groceries? The Atlantic. 2 November. Available ACKNOWLEDGEMENTS at: h ps://www.theatlantic.com/business/arch We thank our colleagues who a ended the Food ive/2016/11/millennials-groceries/506180/. Desert panel at the October 2016 American USDA (2015) Food Access Research Atlas. Available Collegiate Schools of Planning conference in at: h ps://www.ers.usda.gov/data-products/ Houston, Texas. The panel consisted of food food-access-research-atlas/go-to-the-atlas/. planning scholars, practitioners and community Walker, R.E., Keane, C.R. and Burke, J.G. (2010) Dis- members; discussion consisted of forty individuals parities and access to healthy food in the United representing geographic areas in California, Mas- States: a review of food deserts literature. Health sachuse s, New York, Oregon, and Texas. The and Place, 16(5), pp. 876–884. session was recorded with permission of a endees, transcribed and coded. Researchers used the in- Wang, M.C., MacLeod, K.E., Steadman, C., Williams, sights from the expert panel to inform the L., Bowie, S.L., Herd, D. et al. (2007) Is the regression analysis as well as interview questions opening of a neighborhood full-service for case study counties. grocery store followed by a change in the food behavior of residents? Journal of Hunger and Environmental Nutrition, 2(1), pp. 3–18.

Appendix: Interview Solicitation and Interview Questions

Dear [Public Health Department O cial/ County Planning O ce O cial]

I am [name, title and a liation redacted to protect author’s identity]. We are conducting interviews with planning and public health departments in counties that have a high percentage of the population living in a food desert, and lower than expected levels of diet-related disease (diabetes and obesity). As you know, the United States Department of Agriculture (USDA) has de ned a food desert as a census tract with a substantial share of low-income residents who lack access to a supermarket or healthy, a ordable food retail outlet. Yet, ‘food deserts’ often have their own well-adapted systems of street food, urban gardens and ride-shares to grocery outlets. From this study, we hope to learn about pathways to health in the food desert context. Would you please let me know of a date and time within the next three weeks for a 15 minute, 5-question phone interview? Participation is voluntary, and we will send quotes to you for approval before we incorporate them into the  nal study and any resulting publications. I am including the phone interview questions for you to review below.

Thank you for your consideration. Author, [phone number]

Question 1.How would you characterize your county’s diet-related health?

After brie y explaining the spatial data we have on your county, you are invited to re ect if it is representative of the county as a whole and matches your understanding

BUILT382 ENVIRONMENT VOL 43 NO 3 BUILTENVIRONMENTVOL43NO3413 CULTURING FOOD DESERTS: RECOGNIZING THE POWER OF COMMUNITY-BASED SOLUTIONS

Question 2.Can you think of some examples of actions taken in the county that may be contributing to improved health outcomes; i.e. low diabetes and lower rates of obesity than might be expected?

E.g. sidewalks, walking trail, walk-to-school, WIC, county health fairs

Question 3.Tell us about the food planning and policy organizations and e orts in your county, if there are any.

Who leads them? What are their main activities?

Question 4.Where do most residents obtain their food?

Prompts: Markets (farmers’ market) Convenience stores/bodegas Garden private/Community garden Food bank Food/produce truck Buy directly from farms Roadside stands CSA Food delivery Small grocery stores Ride shares/carpool (public transit) Hunting/gathering – berry picking Large groceryGrocery store,store, brand-named brand-named Restaurants – fast food

Question 5.What are some explanatory reasons for resident health given the prevalence of food deserts in your county?

Ask interviewee to re ect on their answers to questions 1 and 3. Do those answers adequately re ect healthy food policies in the county?

Can he or she think of some examples of actions taken in the county that may be facilitating improved access to healthy foods in disadvantaged neighborhoods?

[Researcher summarizes response to see if it is correct after each question.]

ACKNOWLEDGEMENTS

The authors thank the numerous interviewees who graciously gave their time as well as Breana Inoshita, Abbie Schmidt and Tapinder Sidhu for their diligence in transcribing interviews.

342BUILT ENVIRONMENTVOL43NO3 BUILT ENVIRONMENT VOL 43 NO383 3