Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.
3 Studying Food Deserts Through Different Lenses The food environment is complex, as are peopleâs decisions about where to live and shop, and what to buy and eat. As discussed in Chapter 2, the availability of healthy foods affects food choice, but supply alone does not guarantee that healthier foods will be purchased, especially by price- conscious shoppers. Presenters in the next session looked at food deserts with some complementary methodological approachesâepidemiology, geography, and economics and urban planningâand discussed what they can and cannot tell us about the link between food access and health. EPIDEMIOLOGICAL APPROACH As researchers investigate ways to improve peopleâs health, they have turned to the local food environment. Ana Diez Roux, of the University of Michigan, described aspects of the food environment as âfeatures of the local physical environment that facilitate the consumption of certain types of foods and detract from the consumption of others.â She suggested sev- eral health-related reasons for this approach: â¢ ocusing exclusively on individuals, without taking account of F their surrounding environment, has led to disappointing results; â¢ eighborhoods serve as the context for physical and social expo- N sures that may be related to health; â¢ eighborhoods may have causal effects and/or constraints on N prevention efforts; and 27
28 THE PUBLIC HEALTH EFFECTS OF FOOD DESERTS â¢ eighborhood differences, including race, ethnicity, and socioeco- N nomic factors, may contribute to health inequalities and have pub- lic health and policy relevance. Space as a Health Determinant Spaceâas defined by place, neighborhood, or environmentâis a key dimension across which health is patterned. For example, there is a five- fold difference in diabetes prevalence rates across New York City neigh- borhoods, according to calculations by the New York City Department of Health & Mental Hygiene, the U.S. Centers for Disease Control and Prevention, and the World Health Organization. These differences could occur because residents are segregated by factors that research has shown are health-related (such as income, race, or ethnic group), but the features of the places themselves may contribute to the problem. However, Diez Roux emphasized that environmental constraints and reinforcements, such as local food availability and affordability, are just some of many factors that affect health. As discussed by the previous panel, researchers have used many different databases and instruments to understand local food environ- ments. Two important points emerged: (1) the local food environment is patterned by areas of socioeconomic, race, and ethnic composition; and (2) features of the local food environment have been cross-sectionally associated with the diet of residents and with related health outcomes. In an ancillary study of the Multiethnic Study of Atherosclerosis (MESA), in which Diez Roux was the principal investigator, supermarkets were less common in low-income areas, and liquor stores and small grocery stores were more common (Moore and Diez Roux, 2006). The smaller stores did not stock many healthy foods. Moreover, even supermarket offerings can vary by location. A comparison of two supermarketsâone in pre- dominantly African-American Baltimore City and one in predominantly Caucasian Baltimore Countyâshowed that the inner-city store offered far fewer healthy options (see Table 3-1). A Gap: Longitudinal Studies and Better Understanding of Causality How do these differences in food availability relate to health out- comes? Studies show that respondents living in neighborhoods with the lowest availability of healthy food, as indicated by surveys of residents or lower density of supermarkets, were 32 to 55 percent less likely to have a good-quality diet than those with greater availability (Moore et al., 2008). In addition, measures of the availability of healthy food in stores are also related to the diet of residents (Franco et al., 2009). Diez Roux stressed that
STUDYING FOOD DESERTS THROUGH DIFFERENT LENSES 29 TABLE 3-1â Healthy Food Availability Index, Comparing Two Supermarkets Location: Baltimore City Baltimore County Racial composition: 97% African American 93% Caucasian Median household income: $20,833 $57,391 Foods Availability Points Availability Points Skim milk Yes â2 Yes ââ 3 Fruits 17 â2 59 ââ 4 Vegetables 38 â3 74 ââ 4 Lean meat No â â2 Yes ââ 3 Frozen foods No â0 Yes ââ 3 Low-sodium foods No â0 Yes ââ 2 100% whole wheat bread Yes â2 Yes ââ 4 Low-sugar cereals Yes â2 Yes ââ 2 Modified NEMS-S (0-27) 18 â 25 NOTE: NEMS-S = Nutrition Environment Measures Survey in stores. SOURCE: Franco et al., 2008. research has focused mostly on cross-sectional data, comparing different people living in different neighborhoods. A great need is longitudinal evidence relating changes in healthy food availability to changes in diet over time. A few longitudinal studies have taken place, even though causality is not yet clear. For example, using MESA data, people who live in neigh- borhoods with higher healthy food availability scores had a 45 percent reduced incidence of diabetes over a five-year period (Auchincloss et al., unpublished). Another study (Sturm and Datar, 2005, 2008) confirmed that higher prices of fruits and vegetables were linked to greater increases in childrenâs weight over time. In the Moving to Opportunity study of families who had moved from poor to non-poor neighborhoods (Kling et al., 2007), BMI was significantly reduced, although the reasons for this are not understood. Also, as discussed further in this chapter and in Chapter 5, ânatural experimentsâ occurred in Glasgow and Leeds in the United Kingdom, where supermarkets opened in public housing areas and the areas could be studied before and after. Challenges to understanding the causal links remain, including deter- mining which aspects of the local food environment (e.g., availability, price, convenience) are most relevant to health, how to measure and what to use as a proxy, the scale at which changes to the local food environ- ment are most effective, and reasonable time lags and critical periods in which to expect any effects to occur. In addition, dynamic processes are
30 THE PUBLIC HEALTH EFFECTS OF FOOD DESERTS involved because just as healthy food availability may impact dietary patterns, the diets of residents may impact the kinds of foods that are available. The complex and dynamic nature of these processes means that multiple kinds of evidence will be needed to identify the best strategies for intervention, said Diez Roux. These might include improved observa- tional studies and qualitative research, evaluation of natural experiments, and dynamic simulation approaches. Ultimately, it may be necessary to act based on the best available evidence and then rigorously evaluate the impact of these actions so that they can be improved or modified. Diez Roux concluded by stressing why it is worth focusing on loca- tions. Place-based and individual inequalities are mutually reinforcing, and neighborhood differences that result from specific polices are ame- nable to intervention. Ultimately, the goal is not just to understand causa- tion, but to facilitate change for health and non-health benefits. The new paradigm is an interdisciplinary one and integrates transportation, urban planning, food access, and community development policies as part of dealing with peopleâs health. GEOSPATIAL APPROACH Steven Cummins, of Queen Mary, University of London, brought a geographic perspective to the food desert discussion, given that space, place, and distance are features that may affect the food environment and thus become determinants of diet and health. Research into food access began in the 1960s, stemming from concern about social disparities in access to basic services, rather than from the perspective of health. In the 1990s, he said, residents in a deprived urban housing scheme in west Scotland coined the term âfood desertâ to describe the lack of access to a healthy, reasonably priced food supply. During this period, food stores were leaving urban centers for outlying areas, resulting in fewer, larger stores concentrated in edge-of-town sites (see Figure 3-1). A recent sys- tematic review of 48 studies from 1966 through 2007 (Beaulac et al., in press) shows equivocal findings about the existence of food deserts in many European countriesâbut clear evidence of disparities in food access in the United States by income and race. Natural Experiments The underlying conceptual model behind why food deserts affect health is that of âdeprivation amplificationâ: Residents of low-income neighborhoods are exposed to poor-quality local food environments that amplify their individual risk factors for poor health (Macintyre, 2007). Exposure to these environments may contribute to the development of
STUDYING FOOD DESERTS THROUGH DIFFERENT LENSES 31 FIGURE 3-1â Comparison of two urban food environments: Springburn and Shet- tleston, Glasgow, 1 mile apart. Figure 3-1.eps SOURCE: S. Cummins, 2009. bitmap image socioeconomic and spatial inequalities in diet-related diseases, such as obesity, diabetes, and heart disease. The opening of two large supermarkets in Glasgow and Leeds, both in deprived areas in the United Kingdom, provided the opportunity to study the effect of increasing access to food retail opportunities as a solu- tion (the Leeds study was discussed by Neil Wrigley during the second day of the workshop and is summarized in Chapter 5). In Glasgow, a large Tesco supermarket was opened in an area of multiple deprivation with high concentrations of public housing and very few food options. Cummins reported that the Glasgow Superstore Study (Cummins et al., 2005) did not show that the new store resulted in positive impacts on healthy food consumption. Although disappointed with the findings of a study that had obvious intuitive policy appeal, he has since sought to understand why these negative results occurred. Follow-up qualitative work revealed community behaviors that had not been revealed through surveys. For example, some residents purposefully chose not to shop at the new store out of concern that they would be tempted to spend too much. Additionally, what constitutes âlocalâ was clearly different for different people, whose spatial behaviors are affected by their daily rou- tines. For instance, some people had always preferred to shop in other neighborhoods, perhaps where they had grown up, previously lived, or worked, and thus continued to do so. Understanding what drives spatial behavior is of paramount impor- tance in strengthening causal inference. Instead of just focusing on the supply side, Cummins stated that researchers need also to focus on demand and the geographic choices that people make that shape their
32 THE PUBLIC HEALTH EFFECTS OF FOOD DESERTS health behaviors. Learning what does not work also provides valuable lessons. Systems approaches may further understanding of food deserts and public health. The spatial patterning of health as an outcome could be con- ceived as an emergent property of a complex system incorporating both demand- and supply-side behaviors. Individuals affect and are affected by the environment around them in a complex dynamic system. Spatial microsimulation tools can help model and predict the effect of system change, rather than just describing it, if they are based on a solid theoreti- cal framework. For example, Cummins has been involved in developing a spatial microsimulation model to predict the spatial patterning of dia- betes changes due to age in Leeds. The next step is to try to predict what would happen to future diabetes prevalence if plausible, policy-relevant, contextual factors were modified. Better theories are required to inform better empirical research to elucidate causal processes and predict the public health effects of food deserts. Multiple approaches and methods, including better-quality basic theory and data, qualitative methods, natural experiments, and simula- tions, can help triangulate the evidence base and provide a fuller picture. Better understanding of spatial behaviors harnessed to advanced spatial methods will allow the development of possible levers for environmental interventions. ECONOMIC APPROACH One critique of cross-sectional studies linking food access to health outcomes is that those studies do not account for access as a factor; resi- dents choose where they live and are not randomly assigned to neigh- borhoods. Yan Song, of the University of North Carolina at Chapel Hill, presented perspectives from economics and urban planning as a third way to look at the food environment. Urban economics looks at issues of selectivity in how residents choose where to live, and could help explain links of food access to consumption and health outcomes. Urban planning has established principles about the mix between retail and commercial space, which can include the food environment. Song focused on how retail food outlets affect choices of where residents locate. Since the 1960s, economists have used hedonic price models and discrete choice models to explain residential location choice, focusing on characteristics related to housing and the surrounding community. Little research has been done on how food access may enter into peopleâs choices, although new urbanism or smart growth, in which a mixture of land uses are located in the same neighborhood within walking distance, would include food outlets.
STUDYING FOOD DESERTS THROUGH DIFFERENT LENSES 33 Hedonic Price Model A hedonic model examines the individual value-added factors in the total price of an item, such as the convenience of a residence to stores or work locations. This approach enables a researcher to identify the mar- ginal price of any given feature, potentially including the location or size of food establishments. Research into how people value mixed-use devel- opment shows a positive price premium for having a neighborhood cafÃ© and a walkable network of stores: about $6,500 in one study in Portland, Oregon (Song and Knapp, 2003, 2004; Song and Sohn, 2007). A negative premium was attached to commercial uses not in scale with the rest of the neighborhood, including big box stores. Discrete Choice Model In the discrete choice model, people are presumed to make a choice from a fixed set of alternatives. They decide where to live based on their own householdâs characteristics and the characteristics of potential dwell- ings. Is the food environment one of these characteristics? At this point, Song is not aware of any published studies on food environment as a factor in residential choice. Urban Planning and Food Environments Song explained that planners distinguish between basic, revenue- generating land use and nonbasic, service-related land use. The current curriculum at urban planning schools favors planning small-scale food stores in mixed-use development for easy access by local households. Economies of scale, consumer preferences, and existing zoning ordi- nances, however, can make this goal unrealistic. In summary, the research on food environment and residential selec- tion activity shows evidence of a price premium associated with healthy neighborhood stores, but these premiums have been observed only in high-income neighborhoods. No study has explicitly looked at how food retailers affect residential location choice. More refined surveys and more data, including natural experiments, may provide some answers. DISCUSSION: DIFFERENT APPROACHES Jill Reedy, of the National Cancer Institute at the National Institutes of Health, moderated the discussion that followed this panel. Many of the questions and comments related to the complexity of causes, mak- ing it difficult to separate the impact of the food environment from other variables.
34 THE PUBLIC HEALTH EFFECTS OF FOOD DESERTS Food Access from the Workplace The research on food deserts looks at access from where people live versus where they work. One workshop participant said opposition by institutional review boards makes it difficult to collect data from work- places. In any event, Diez Roux cautioned against extrapolating too much from a current situation: people might buy food closer to work, for exam- ple, because there is no alternative closer to home. Some groups are com- bining data on a variety of aspects of the built environment, including food venues and their relationship to transportation routes, to see how they connect. Supermarkets as One Proxy Song was asked about the emphasis on large supermarkets from a planning perspective, given the emphasis on keeping buildings at a similar scale. She observed that spatial planning may not take household characteristics sufficiently into account. Cummins said that what planners want and what a local population wants might diverge. Often a success- ful local retail economy has a mix of different-sized stores. Diez Roux stressed the issue is access to healthy foods, not necessarily access to a supermarket. Environments have many features that interrelate, which implies thinking through the positive and adverse effects of a particular intervention. Role of Simulations Simulations are valuable, said Diez Roux, because they require think- ing through processes to create a valid model and may point out knowl- edge gaps that may have been overlooked by other research methods. Cummins noted that simulations can utilize existing observational data in a better way, perhaps linking together unconnected data sets. Song noted that in urban planning, simulations are used to build scenarios to observe the effect, holding everything else constant, of a specific policy intervention. Mixed Land Use Whether mixed land use promotes positive health effects is, according to Song, a debatable topic. It seems to depend on what the mixed uses actually entail. If they are appropriate and decrease automobile use, that would be healthy. Diez Roux said the literature is difficult to summarize because the measures have been so different. Proximity of destinations promotes walking, but the long-term health impacts are less known.
STUDYING FOOD DESERTS THROUGH DIFFERENT LENSES 35 Realistic Expectations An issue that came up several times during the workshop, including in this discussion, centered on realistic expectations from introducing a new supermarket into a food desert, in terms of changes in food intake and ultimately BMI or other health outcomes. The natural experiments with which Cummins has been involved led him to realize, he said, that robust underlying theoretical models and the time frames in which we might realistically see effects are still not fully known. One successful outcome could simply be increasing the number of food stores available, but a secondary outcome would be to see changes in health behaviors and then impacts on obesity or the prevalence of diabetes. Changes in important health behaviors and outcomes may take longer to ascertain than most current funding mechanisms allow. Diez Roux suggested look- ing at proximal outcomes in the short term, rather than trying to detect more distal effects. Cummins also suggested making more use of complementary activi- ties, such as mailings to residents or incentives, and evaluating the effect of these initiatives combined with changes in supply. One workshop participant questioned whether food desert health outcomes are really due to limited food access or perhaps more likely to limited healthcare access. Diez Roux agreed the issues are confounded because the real world is complex, and it is difficult to separate the causal effect of food access. Methodologically, researchers attempt to create boundaries through a variety of statistical controls. Cummins said spa- tial analytic approaches to measure access using GIS (geographic infor- mation systems) in longitudinal studies may help avoid the problem of using administrative boundaries, which may shift over time, as a proxy for neighborhoods. People have different perceptions of neighborhood boundaries. Using census tracts as a proxy, in his opinion, also has weak- nesses that qualitative research reveals. Questions remain about what is the most relevant and comparable spatial environment. Diez Roux agreed that a census tract is not ideal, but may serve as a useful although imper- fect proxy for the most relevant spatial context. Community and Interdisciplinary Initiatives Reedy summarized several questions from workshop participants related to work within communities. Partnering with community groups to conduct research is important in this kind of research, said Diez Roux, particularly in evaluating natural experiments and conducting qualitative studies. Studies have looked at various community benefits of addressing food issues. Urban agriculture is promoted in some cities to increase local
36 THE PUBLIC HEALTH EFFECTS OF FOOD DESERTS food production, as well as to increase physical activity. Other studies have looked at the effects of using local government subsidies to encour- age the opening of retail outlets that carry healthy foods: for example, if housing prices increase as a result, the tax base grows and the public investment has a positive fiscal return. Similarly, in the United Kingdom, retail leverage generation (planning gain) is considered a tool to improve the local economy through providing employment and upgrading public facilities such as sidewalks and other infrastructure. Despite concerns about the impact of a large store on smaller Glasgow retailers, the same number of small stores were in business 18 months later in the area that Cummins studied. To close the session, Diez Roux emphasized the need for interdisci- plinary research among epidemiologists, geographers, economists, and urban planners. Reedy expressed agreement on behalf of the other panel- ists and workshop participants.