Consumers can improve the sustainability of their diets in all of the following ways except:

The committee examined evidence for the impact of various environmental factors on the feasibility of defining the adequacy of SNAP allotments. The range of available evidence included specific environmental factors affecting food choices: food prices; access to food outlets offering a wide range of healthy foods; and disparities in access, particularly in transportation. The committee notes that relevant data collection is ongoing but the data are not yet available through the ERS National Household Food Acquisition and Purchase Survey. When completed, this survey will provide detailed information on household food purchases and acquisitions, including foods purchased for consumption at and away from home and foods acquired through public and private food and nutrition assistance programs. This dataset will be useful for a broad range of economic analyses of food choices and for understanding the implications of food choices for diet quality.

Access to food varies substantially across households because of the various factors affecting food prices. These factors include a number of environmental dimensions, such as geographic region of the country; urban versus rural setting; types of stores available (e.g., supermarkets, convenience stores, mass merchandisers, warehouse club stores); and types of foods available, such as healthier versus less healthy and degree of processing (e.g., raw ingredients, processed ingredients, processed foods, fully prepared foods). Over time, food prices are influenced by changes in costs, due largely to inflationary factors, for farm-level inputs and production, transportation at each stage of production, food processing, and food distribution, which may result in greater or lesser changes than in overall prices for all goods and services. Considering food prices as a component of the evidence needed to define the adequacy of SNAP allotments is important because the influence of food prices on the likelihood of food insecurity is both positive and significant; an increase of one standard deviation in the price of a food basket based on the TFP results in increases in food insecurity of 2.4 percentage points for adults and 3.7 percentage points for children (Gregory and Coleman-Jensen, 2011). In the following discussion, the committee describes the differences in food prices across several dimensions and changes in food prices over time in the context of how these patterns affect food access as a factor in defining the adequacy of SNAP allotments.

Food prices vary across geographic regions of the United States because of differences in the cost of living and other market conditions (Todd et al., 2010). The ERS Quarterly Food at Home Price Database is based on 2006 Nielsen Homescan data and can be used to examine differences in food prices across regions. It is preferred to the Bureau of Labor Statistics' (BLS's) Consumer Price Index (CPI) because the CPI is available only for a limited number of food items and four broad regional markets (Todd et al., 2010). Using 2006 Nielsen Homescan data from the ERS Quarterly Food at Home Price Database, Todd and colleagues (2010) show that the average prices per 100 grams across 50 food categories and 35 market regions differed by 125 percent to 217 percent in the highest-cost versus the lowest-cost region (see Table 4-1).3 The smallest price difference identified was for canned soups, sauces, and prepared foods, while the largest price difference was for low-fat cheese. In general, grain-based foods, prepared foods, snack foods, and carbonated beverages appeared to have smaller price differences than fruits and vegetables, dairy foods, and meat and poultry. For particular types of foods, the ratio of the average price to the national average varied substantially across market regions. For example, as shown in Table 4-2 for low-fat milk, the average price per 100 grams was 73 percent of the national average for the full sample in Salt Lake City but 129 percent of the national average in urban New York. For purchases made by low-income households in the Nielsen Homescan sample (income below 185 percent of the poverty level), the corresponding ratios ranged from 66 percent to 133 percent across the same market regions. Across the sample of other food categories examined in the report (canned fruit, packaged whole grains, eggs, and carbonated beverages), the lowest relative prices were generally 70 to 90 percent of the national average, while the highest relative prices were generally 120 percent to 140 percent of the national average. Overall, the ranges appeared to be similar for the full Homescan sample and the low-income portion of the sample, with generally similar rankings by market group. However, some of the differences in average prices across regions and across income levels could be due to differences in the level of quality of foods purchased.

Like Todd and colleagues (2010), Leibtag and Kumcu (2011), using Nielsen Homescan data, found substantial price variation by region of the country when examining more disaggregated data for fresh fruits and vegetables for 2004–2006. As shown in Table 4-3, the average prices for fruits and vegetables differed by 12 percent (green beans) to 140 percent (watermelons) between the lowest- and highest-cost regions. The minimum prices were at least 11 percent below the national average price, while the maximum prices were up to 47 percent above the national average price. Aggregated across the 20 categories of fresh fruits and vegetables, prices were lowest in the Metro South 2 region, comprising Nashville, Birmingham, Memphis, and Louisville, and highest in San Francisco (see Table 4-4). According to Leibtag and Kumcu's (2011) analysis, these differences in prices across regions have substantial implications for the purchasing power of benefits provided by nutrition assistance programs.

The data in Table 4-2 also provide evidence of differences in prices by urban versus rural areas. Higher costs in urban areas may reflect the higher costs of operating retail establishments in those areas and the fact that urban communities are often served by smaller stores with higher prices (Stewart and Dong, 2011). For example, prices for low-fat milk in non-metropolitan East North Central (80.3 percent relative to the national mean) are lower than those in the corresponding metropolitan areas in the same region: metropolitan Ohio (81.5 percent); metropolitan Midwest comprising Indianapolis, Detroit, Milwaukee, and Grand Rapids (82.4 percent); and Chicago (87.3 percent). Prices for low-fat milk in non-metropolitan New England (107.6 percent relative to the national mean) are lower than those in the corresponding metropolitan area of Hartford (115.5 percent) but similar to those in the metropolitan area of Boston (107.1 percent). Generally, patterns of lower prices in non-metropolitan areas were found across the other food categories examined (canned fruit, packaged whole grains, eggs, and carbonated beverages). Stewart and Dong (2011) found similar results using data from the Nielsen Homescan panel for 2006, which showed that prices paid by households in urban areas for fresh vegetables and salty snacks were significantly higher than those in non-urban areas.

Households acquire food for use at home from a broad variety of stores and outlets (see Box 4-2), including traditional supermarkets and grocery stores; convenience and combination grocery stores (e.g., drug stores), mass merchandisers or supercenters and warehouse club stores, farmers' markets, and specialty and gourmet food stores. Over time, there has been a trend for a larger portion of food purchases to be made in stores other than traditional supermarkets (Leibtag, 2005).

Types of Food Stores and Outlets.

Prices vary across types of stores, and thus the types of stores accessible to households affect their total food budgets and ability to acquire healthy foods. In particular, greater access to large grocery stores in suburban than in inner-city and rural areas may result in the poor paying higher prices for food (Andreyeva et al., 2008). In an extensive comparison of food prices at nontraditional discount stores4 and traditional food stores, Leibtag and colleagues (2010), using 2004–2006 Nielsen Homescan data, found that national average unit prices (total price divided by product weight) were significantly lower in nontraditional stores for 86 percent of food products at the broadest level of comparison. At the Universal Price Code (UPC) comparison level, 82 percent of products had significantly lower prices in nontraditional stores, with an average price discount of 7.5 percent. Meat products had the largest price discounts at nontraditional stores. Furthermore, all canned products had significantly lower prices in nontraditional stores on average. With respect to specific markets, those with more nontraditional stores had smaller differences in prices between those stores and traditional stores, which could be due to increased competition or the exit of traditional stores in markets with a large number of nontraditional stores (Leibtag et al., 2010).

In studies focused on specific areas, food prices were also found to differ substantially across store types. In a study of six neighborhoods in New Haven, Connecticut, Andreyeva and colleagues (2008) found that prices in 2007 were 51 percent higher on average in small neighborhood stores than in supermarkets across 622 food items. In another analysis, Andreyeva and colleagues (2008) constructed a representative basket of food items from 75 stores to compare prices across types of stores (convenience versus grocery, including small neighborhood grocery stores) and neighborhoods (low-income versus high-income). Results indicated that the average price of the market basket was about 4 percent higher in higher-income neighborhoods and also about 4 percent higher in convenience stores than in grocery stores; thus, the authors conclude that price differences across neighborhoods and store types are relatively modest. In a study of food prices in 77 stores in a rural South Carolina county, Liese and colleagues (2007) found that prices of selected foods were substantially higher in convenience stores than in supermarkets (greater than $2 million in annual sales) and grocery stores (less than $2 million in annual sales). The differences were statistically significant for apples, packaged bacon, packaged smoked turkey, canned salmon, canned tuna, low-fiber breakfast cereals, whole milk, and low-fat/nonfat milk.

Whether healthy foods are found to be more expensive than less healthy foods may depend on how both “healthy” and the units of measure are defined (Carlson and Frazao, 2012; Lipsky, 2009). Numerous studies have examined whether healthy foods are more or less expensive than less healthy foods, using either secondary data available across a broad range of foods or data collected from specific stores for a more limited set of foods. While some argue that nutritious diets of whole grains, lean meats, and fresh vegetables and fruits are affordable, others believe that energy-dense foods (i.e., with more calories per serving) are less expensive (Drewnowski, 2010; Drewnowski and Eichelsdoerfer, 2010; Drewnowski and Specter, 2004; Monsivais et al., 2010). For example, Monsivais and colleagues (2010) linked longitudinal retail price data for 378 foods and beverages in Seattle for 2004–2008 with energy density (kcal/g) and two measures of nutrient density—the Naturally Nutrient Rich (NNR) score, which is the sum of the percent Daily Values (DV) per 100 kcal for 16 nutrients, and the Nutrient Rich Food Index, which is based on the levels of nine positive and three negative nutrients relative to calories. They found that the mean cost of the most nutrient-dense foods (those with high positive nutrients and low negative nutrients relative to calories) were substantially higher and increasing more rapidly than the mean costs of the least nutrient-dense foods.

A recent analysis by Carlson and Frazao (2012) defines a healthy food as one that contains at least half the portion size defined by the 2010 DGA in at least one of the major food groups and has only a moderate amount of saturated fats, added sugars, and sodium. Their analysis examines the price per calorie, per edible gram, and per average portion consumed using the CNPP food prices database based on 2003–2004 Nielsen Homescan data as the source of prices. The authors found that foods high in calories tend to have a lower price per calorie than foods lower in calories. As shown in Figure 4-1, while vegetables have the highest price per 100 calories, they have the lowest price on the basis of price per 100 edible grams or per average portion. Dairy products have the lowest price per 100 calories but one of the highest prices on a 100 edible grams basis. “Moderation” foods, which are defined primarily as those high in sodium, added sugars (including sugar-sweetened beverages), or saturated fat, have a relatively low price per 100 calories but a relatively high price on an average portion basis. Thus, Figure 4-1 shows that comparisons of the costs of healthy versus unhealthy foods can be misleading if expressed on the basis of 100 calories (and 100 edible grams) because many healthy foods have fewer calories per serving (and per 100 grams) than unhealthy foods. Most important, on a per-serving basis, grains, dairy products, fruits, and vegetables, cost less than moderation foods.

Todd and colleagues (2011) examined differences in prices of a selected set of healthy foods relative to their less healthy counterparts by geographic region using the ERS Quarterly Food at Home Price Database for 1998 through 2006 and found mixed results by food type and region. For some foods, such as whole grains compared with refined grains and fresh and frozen dark green vegetables compared with starchy vegetables, the healthier version was more expensive in all geographic regions. For other foods, such as orange vegetables compared with starchy vegetables and skim and 1 percent milk compared with 2 percent and whole milk, the healthier version was less expensive than the less healthy version in some geographic areas. The magnitudes of the price differences for healthier versions of foods varied substantially across the country.

In addition to the studies discussed above, other studies have examined the prices of healthier food alternatives based on the prices of foods available in stores in specific areas. For example, Jetter and Cassady (2006) collected price data from 25 stores in Los Angeles and Sacramento for September 2003 through June 2004 and calculated the average cost of a standard market basket based on the TFP versus the average cost of a healthier market basket. Because of the higher costs of healthier foods such as whole grains, lean ground beef, and skinless poultry, the average cost of the healthier market basket was $230, compared with $194 for the TFP market basket. Using similar data for a longer time period, Cassady and colleagues (2007) estimated that a low-income family would need to devote 43 percent to 70 percent of its food budget to fruits and vegetables to meet the 2005 DGA, thus indicating the high price of fruits and vegetables relative to other foods. Because of their relatively high costs, discounts on fruits and vegetables might result in fairly substantial increases in consumption (Dong and Lin, 2009). In a separate study of food prices in 77 stores in a rural South Carolina county in 2004, Liese and colleagues (2007) found that the prices of more healthful versions of foods were higher than those of the less healthful versions, with the exception of milk. For example, high-fiber bread was more expensive than low-fiber bread, low-fat beef was more expensive than high-fat beef, and chicken breasts were more expensive than chicken drumsticks in convenience stores and supermarkets. The results of these more narrowly focused price studies demonstrate that substantial differences in the prices of healthier foods can occur in specific geographic areas, although studies using larger secondary data sources show a broader pattern of mixed results depending on the level of aggregation of the data by food type and by geographic region.

Food prices vary over time because of changes in the availability of supplies of raw commodities, changes in farm-level production costs, and changes in food processing costs (e.g., due to changes in land, capital, energy, and labor costs). In addition, some food prices, particularly for fresh fruits and vegetables, vary seasonally, being lower during the products' harvest seasons.

Rising food prices may reduce the purchasing power of benefits received through federal nutrition assistance programs, depending on how and whether the benefits are adjusted for inflation over time (Hanson and Andrews, 2008). As discussed in Chapter 2, there is a lag of up to 16 months between calculation of the TFP cost and adjustments to the maximum SNAP benefit to account for inflation. Food prices are particularly affected by changes in world supply and demand and may increase more or less than the overall price level. As shown in Figure 4-2, the percentage changes in the CPI for food generally track the changes in the CPI for all products, but the timing and magnitude of the changes are not always aligned. After a relatively stable period during the 1990s and early 2000s, increases in the CPI for food exceeded the overall CPI over many of the past few years. However, Hausman and Leibtag (2007) show that the methodology used by BLS to calculate the CPI may overstate the price of food because it does not fully capture lower prices in supercenters and other nontraditional retail outlets that sell food.

In addition to differences between changes in overall food price levels and changes in the price levels for all products, overall food price levels have been shown to differ from the price levels for the TFP (Hanson and Andrews, 2008). Figure 4-3 shows that the TFP price index, which is calculated using CNPP's monthly updates of the costs of the TFP, is rising more rapidly than the CPI for food consumed at home. The differences are likely due to larger shares of fresh fruits and vegetables and eggs, which have the most volatile prices, in the TFP price index than in the CPI for food consumed at home (Hanson and Andrews, 2008). Because SNAP benefits are adjusted annually in October using the prior year TFP price index, the food purchasing power of the benefits may decline to the extent that the adjustments do not fully account for the potential monthly rise in the cost of the TFP. However, because the TFP is a theoretical construct, actual food purchase patterns likely differ from the foods represented in the TFP. Thus, the extent to which the difference between the TFP price index and the CPI represents true higher costs to SNAP recipients is uncertain. The 16-month lag in adjusting the maximum benefit according to the CPI is a more certain contributor to a larger discrepancy between SNAP benefits and the actual cost of the TFP. The temporary increase in SNAP benefits under the American Recovery and Reinvestment Act of 20095 likely reduced this discrepancy, but the extent of reduction is currently unknown.

As described above, choosing foods that make up a diet consistent with the recommendations of the 2010 DGA, such as increased consumption of fruits and vegetables and whole-grain-rich foods and decreased consumption of solid fats and added sugars (USDA and HHS, 2010), can be challenging for populations with limited resources as a result of factors affecting food prices both regionally and locally. In light of this evidence, the committee examined additional evidence for an impact of the ability of low-income populations to access affordable healthy foods on the purchasing power of SNAP allotments under the assumptions of the TFP.

The committee identified a number of observational studies showing correlations between various means of access to food outlets and purchasing behavior. Personal transportation was previously discussed as an individual factor. The following discussion focuses on other barriers to access to healthy foods.

Urban or rural locale In a survey of the availability of fruits and vegetables in urban and rural areas of upstate New York, Hosler and colleagues (2008) identified one urban minority neighborhood among those surveyed that, in this respect, was the most disadvantaged site within an urban locale, as measured by the population density of stores selling these products. This community was found to be lacking not in the number of food stores but in an accessible high-impact super produce store. By contrast, such stores were available in a higher-income urban mixed neighborhood in the same locale, illustrating that disparity in access to fresh produce was associated with a single disadvantaged area within a larger locale.

A recent example of associations between access to healthy food and consumption is a cross-sectional community survey, conducted in 2002–2003, which was used as a data source for analyzing associations between neighborhood availability and consumption of dark green and orange vegetables in an ethnically diverse low- to moderate-income population in Detroit (Izumi et al., 2011). Data derived from the survey included the frequency of consumption of these vegetables and their availability in all food stores in the communities studied. The mean intake of dark green and orange vegetables among all participants was found to be 0.61 servings per day. The lowest intake was among participants living in neighborhoods where no store carried five or more varieties of such vegetables; residents in those neighborhoods consumed 0.17 fewer servings per day than those in neighborhoods where at least two stores provided more variety. The results of this study suggest a direct relationship between the availability of vegetables and consumption patterns within a locale.

Fisher and Strogatz (1999) conducted a telephone survey to (1) determine whether there is an association between the availability and consumption of low-fat milk, and (2) assess whether the availability of low-fat milk is associated with household income and racial composition. The study was carried out in three different geographic settings in New York: large metropolitan, midsize urban, and rural. Household interviews were conducted in each setting to determine the type of milk consumed. In each store surveyed within a corresponding zip code, containers of whole, 2 percent, 1 percent, and skim milk were counted for each container size (quart, half gallon, and gallon). The percentage of low-fat milk in the store and the average percentage across all stores in a zip code were then determined. A direct correlation was found among the percentage of low-fat milk in stores, consumption of low-fat milk in the household, and income level by zip code. In particular, low-fat milk tended to be less common in the stores located in rural or low-income areas and areas in which the majority of the population was nonwhite. The authors note, however, that only 51 percent of survey respondents reported usually purchasing milk within their residential zip code.

Cheadle and colleagues (1991) conducted a similar study using a survey of healthful food choices in different grocery store environments to assess the relationship between individual dietary choices and the grocery store environment. They carried out a telephone survey in 12 different communities, including the corresponding larger zip code area, to obtain self-reported dietary intake data on low-fat and high-fiber foods, as well as the availability of health information in the community stores. They found a significant correlation between the availability of healthful foods in the community and zip code area stores and the self-reported healthfulness of participants' diets. In a review of cross-sectional studies on associations between food environment and consumption, Rose and colleagues (2010) identified a number of studies that combined in-store measures with mapping of store location and found significant direct associations between neighborhood food environment and measures of consumption.

These studies are important because they suggest a link between purchasing power and access to food. A general conclusion that can be drawn from this work is that although associated with income, access to food outlets and healthy foods needs to be considered in the context of how certain factors within the food environment affect the cost of healthier food options. Overall, the evidence suggests that limited access to healthy food may influence food shopping and spending behavior by reducing choices.

Farm-to-consumer venues Farm-to-consumer venues show promise in improving dietary intake among all people in the United States, including low-income groups (Blanck et al., 2011). However there are few such venues, especially in low-income communities (FNS, 2011). In addition, many farmers' markets do not accept SNAP. Although USDA figures indicate that the number of farmers' markets accepting SNAP has increased by 16 percent since 2010, more needs to be done to increase the number of these venues authorized as retailers by the program (FNS, 2011). The lack of awareness of farm-to-consumer venues, the lack of farmers' markets and farm stands close to home, the lack of transportation to these venues, inconvenient hours, and affordability concerns are additional barriers to use of farm-to-consumer venues among those receiving federal food assistance (Briggs et al., 2010).

According to a 2009 USDA report, 23.5 million people lack access to a supermarket within a mile of their home (Ver Ploeg et al., 2009). Limited access to food stores is not unique to urban areas; about 20 percent of rural counties across the United States (418 counties) also have been identified as areas where half the population lives more than 10 miles from a large food store (Morton and Blanchard, 2007). The disparate distribution of grocery stores and supermarkets in low-income neighborhoods or geographic areas is especially notable in light of the distribution of racial/ethnic groups within these tracts. For example, Mantovani and Welsh (1996) found that “a large majority of low-income households are in close proximity to a full-line grocery store or supermarket” (p. iv), but minority households in rural areas live farther from these types of food stores than nonminority households.

Apart from the question of distance from a food store is that of where SNAP participants are more likely to shop. Ohls (1999) analyzed data from the National Food Stamp Program Survey, conducted between June 1996 and January 1997. The analysis examined the food shopping opportunities of low-income households, including SNAP participants and eligible nonparticipants. The study found that most low-income households shopped at supermarkets but tended to supplement their purchases by shopping at neighborhood grocery stores, convenience stores, bakeries, and produce markets. They also engaged in “careful” shopping practices, including making bargain purchases, taking advantage of special offers, and using shopping lists to extend their food dollars. Olander and colleagues (2006) and Castner and Henke (2011) also found that most SNAP participants redeemed their benefits at supermarkets, and their purchase patterns were similar to those identified by Cole (1997).

As noted by Mantovani and Welsh (1996), minority low-income groups may experience disparities in access that are not seen across the low-income population as a whole. In a study examining associations between local food environments and neighborhood racial/ethnic and socioeconomic composition, Moore and Diez Roux (2006) analyzed census tract demographics as well as food store characteristics in selected study areas in Maryland, New York, and North Carolina. Their comparison across study areas and across racial/ethnic composition revealed that the predominantly minority and racially mixed areas had at least twice as many grocery stores but fewer than half the number of supermarkets compared with predominantly white areas. The low-income and nonwhite areas also had fewer fruit and vegetable markets, bakeries, specialty stores, and natural food stores.

A cross-sectional survey in Michigan (Zenk et al., 2005) assessed the availability, quality, and price of fresh produce in various types of stores—large and small grocery stores, “mom and pop” stores, and convenience and specialty stores—in three Detroit communities and an adjacent suburb. The communities surveyed varied in racial/ethnic composition and socioeconomic characteristics and exhibited different health profiles for diet and obesity-related diseases. Among the findings was that produce quality was lower in low-income African American communities than in more affluent or racially mixed neighborhoods. Moreover, the low-income African American communities had more than four times more liquor stores and fewer grocery stores per 100,000 residents compared with the racially mixed communities.

Overall, this body of evidence suggests that supermarket access is poorer among low-income and minority populations, and that individuals without ready access to supermarkets have more difficulty finding fruits and vegetables in their neighborhood. In addition, individuals with supermarkets in their neighborhood are more likely than those lacking nearby supermarkets to eat more fruits and vegetables.

In a cross-sectional study of 25 stores in South San Diego County, California, Emond and colleagues (2012) examined associations between the availability, quality, and cost of healthy and unhealthy food items and store location—specifically, non-ethnically based supermarkets and Latino grocery stores (tiendas) in low-income areas. They found no difference in the availability of fresh produce by store type and quality differences for only one fruit item. Further, the price per pound for fresh produce was lower in the tiendas than in the supermarkets. However, the cost of skim milk was significantly higher in the tiendas and lean ground beef was significantly less available than in the supermarkets surveyed. Similarly, Andreyeva and colleagues (2008), conducted two studies examining changes in price differences between large grocery stores and small neighborhood markets over the past 35 years (study 1), and price and nutritional quality as a function of income and neighborhood (study 2) in New Haven, Connecticut. In assessing the results of both studies, they concluded that the availability of many healthful food items was lower and produce quality was worse in lower-income than in higher-income areas even though average prices were not significantly different between the two types of neighborhoods. In Baltimore, Maryland, Franco and colleagues (2007) conducted a small observational study to determine the availability and price of food in 240 stores in area neighborhoods. In the neighborhoods surveyed, 94.4 percent of residents were African American; 64.4 percent of family households were female headed; the unemployment rate of residents was 23.5 percent, with a median household income of $15,493; and only 53.6 percent of adults had completed high school. Of the 187 food stores located within the city, 17 were classified as supermarkets, 136 as grocery stores, and 34 as convenience stores. No fresh fruits or vegetables, whole-wheat bread, or skim milk was found in the city's grocery stores; other food items, such as whole milk, soda, chips, and canned foods, were typically available. Further, the price of whole milk, cereal, and white bread at a representative store was 20 percent higher than in the closest supermarket, 0.9 miles away. Overall, food outlets in lower-income and minority neighborhoods tend to stock lower-quality items than food outlets in predominantly higher-income, white neighborhoods.

Residents in many urban areas have few transportation options to reach supermarkets. To examine whether access to transportation plays a role in risk factors for food insecurity and access to food outlets, Bjorn and colleagues (2008) developed and mapped a number of food insecurity index values, including income, ethnicity, employment, and education. Analysis of the indices identified a number of high-risk areas lacking food access in Seattle, King County, Washington. Many of the high-risk lower-income neighborhoods assessed were racially and ethnically diverse. For some of these areas, transportation access was a major barrier to food security. Households in the areas at risk of food insecurity were more vulnerable to economic and social as well as geographic barriers that may have made them dependent on local convenience stores and/or required long trips to distant grocery stores.

Inadequate transportation can also be a major challenge for rural residents, given the long distances to stores. Sharkey and colleagues (2009) examined associations between neighborhood needs, as measured by socioeconomic deprivation and vehicle availability, and two criteria for food environment access: distance to the nearest food store and fast-food restaurant, and number of food stores and fast-food restaurants within a specified network distance of neighborhood areas. The authors analyzed data from the 2006–2007 Colonias Food Environment Project and the decennial 2000 U.S. Census Summary File 3. They found that the rural neighborhoods studied had better access to convenience stores and fast-food restaurants in terms of both distance and shopping opportunity compared with access to supermarkets. Supermarkets provided greater proximity and coverage than traditional grocery stores, but when neighborhood deprivation was taken into account, the neighborhoods with higher deprivation had the least access to supermarkets and grocery stores but the greatest access to convenience stores. When transportation access was considered, limited availability of a vehicle was correlated with greater proximity to a supermarket as well as other store types, but higher deprivation was associated with greater distance to supermarkets as well as other store types. Collectively, these results indicate an association between high-deprivation neighborhoods and both low access and limited transportation to supermarkets in a rural area.

In response to a request from Congress, USDA conducted a comprehensive 1-year study to assess the impact of limited access to food on local populations and outline recommendations for addressing the problem (Ver Ploeg et al., 2009). The study included two conferences on food deserts and a set of commissioned studies carried out in cooperation with the National Poverty Center at the University of Michigan, as well as reviews of the existing literature, a national-level assessment of access to supermarkets and large grocery stores, analysis of the economic and public health effects of limited access, and a discussion of existing policy interventions. Table 4-5 shows the study findings on access to supermarkets according to individual factors of low-income and underserved population groups in the United States. Data in the table indicate that the median distance to a supermarket for low-income, minority, and elderly populations is comparable to that for higher-income populations. However, as data in Table 4-6 indicate, the percentage of households without a vehicle is higher in low-income areas. For example, 2.5 to 3.3 percent of urban and 7.4 percent of rural low-income households live more than a mile from a supermarket and lack access to a vehicle. Research has shown that inadequate transportation is a significant barrier to access to supermarkets for residents of economically disadvantaged African American neighborhoods (Zenk et al., 2005).

The committee identified a number of studies examining associations between disparities in access to healthy foods and food insecurity, obesity, and obesity-related chronic disease. On the whole, the evidence supports a positive relationship between food insecurity and risk for obesity that is strongest among women (Dinour et al., 2007; Frongillo et al., 1997; Jilcott et al., 2011; Jones and Frongillo, 2006; Larson and Story, 2011; Velasquez-Melendez et al., 2011) and stronger among African American and Hispanic groups than whites (Fitzgerald et al., 2011; Sharkey and Schoenberg, 2005). The evidence reviewed was inconsistent as to significant associations between food insecurity and obesity in children (IOM, 2011).

A review of 54 studies examining associations between neighborhood differences in access to healthy food and risk for obesity identified an association between better access to convenience stores and higher risk for obesity (Larson et al., 2009). An observational study that analyzed data collected from a telephone survey found increased odds of obesity associated with distance to a supermarket in metropolitan but not in nonmetropolitan areas (Michimi and Wimberly, 2010). In contrast, another analysis of data collected on more than 21,000 Special Supplemental Nutrition Program for Women, Infants, and Children (WIC) participants in Kansas failed to find an association between the availability of grocery stores and supermarkets within a census tract and body mass index (BMI) (Ford and Dzewaltowski, 2011). A recent ERS report (Ver Ploeg et al., 2009) suggests that access to a supermarket or large grocery store is a problem for a small percentage of households.

Other observational evidence supports a reduced risk for obesity and related conditions associated with better access to healthy foods. Morland and colleagues (2006) analyzed cross-sectional data collected from men and women participating in the third visit (1993–1995) of the Atherosclerosis Risk in Communities Study to determine whether characteristics of the local food environment were associated with the prevalence of cardiovascular disease risk factors. They found that people with access only to supermarkets or to supermarkets and grocery stores had the lowest rates of obesity and overweight, while those with access only to convenience stores had the highest rates. However, associations for diabetes, high serum cholesterol, and hypertension were not consistently observed.

To examine the association of retail food environments with obesity and diabetes, the California Center for Public Health Advocacy, PolicyLink, and the University of California, Los Angeles, Center for Health Policy Research combined individual-level demographic and health outcome data from the 2005 California Health Interview Survey (CHIS, 2007) with the locations of retail food outlets. Geographic information system software was used to calculate a Retail Food Environment Index for each adult CHIS respondent within a given radius around his/her home address. It was found that in California, rates of obesity and diabetes were 20 percent higher for those living in the least healthy food environments after controlling for household income, race/ethnicity, age, gender, and physical activity levels (Babey et al., 2008).

Evidence that access to food has a direct impact on pregnancy outcomes is limited. In a study examining associations between diet quality, measured by a Diet Quality Index, among pregnant women and distance from a supermarket, Laraia and colleagues (2004) found that women living more than 4 miles from a supermarket had a twofold increased risk for being in the lowest quartile of the DQI. In another study on access to food and birth outcomes, Lane and colleagues (2008) found that women living in food deserts without access to healthy foods had significantly more low-birth-weight infants than women who had access to supermarkets and a variety of foods.

Although diet is integral to the treatment of diabetes and maintenance of glycemic control, evidence now exists that foods recommended as part of a healthy diabetic diet are in short supply in low-income, nonwhite neighborhoods. To illustrate, Horowitz and colleagues (2004) documented and compared the availability and cost of foods recommended for people with diabetes in East Harlem and the adjacent more affluent and predominantly white Upper East Side neighborhood. They found that the East Harlem neighborhood had a shortage of food markets, and some stores did not carry foods needed for a healthy diabetic diet. Additionally, the neighborhood had few large stores with a variety of foods and fewer stores that carried recommended food items. Further, East Harlem had many more undesirable stores than the more affluent Upper East Side neighborhood. These disparities in availability of healthy foods many be a barrier to diabetes self-management for East Harlem residents.

As described above, limited access to healthy food can influence purchasing behavior. Therefore, it is possible that the availability of food outlets and costs of food items may impact the purchasing power of SNAP allotments for healthy foods under the assumptions of the TFP, which in turn may affect diet-related health outcomes for SNAP participants.

Environmental interventions to address challenges to food access show some promise. A recent study evaluating the impact of the first full-service supermarket to locate in Harlem in New York City found that the store allocated the same amount of space to fresh fruits, vegetables, fish, and meat as a typical suburban market, at similar prices (Lavin, 2005). The Pennsylvania Fresh Food Financing Initiative—a statewide financing program designed to increase supermarket development in underserved areas—has funded 78 fresh-food outlets in Pennsylvania, which have increased food access for 500,000 children and adults (Karpyn et al., 2010). More research is need to understand what changes might improve access to food outlets. Approaches at the environmental level might include transportation policies that address both affordability and routes, and incentive/financing programs to increase the number of and quality of supermarkets in low-income, minority communities.