Several studies have considered associations between neighborhood food environments and dietary intake, particularly consumption of fruits and vegetables (1). Studies have used various methods to measure food environments, including objective measures (eg, measured distance to food stores [2,3]) and self-reported measures (eg, perceptions of food-store accessibility [4,5]). Few studies have used both objective and self-reported measures (6–9), however, even though both types of measures may be important to fruit-and-vegetable consumption. Self-reported measures may provide the clearest picture of people’s perceptions of their environment; objective measures may provide concrete targets for environmental interventions to promote greater fruit-and-vegetable intake.

Beyond distinctions between self-reported and objective measures of food environments, it may also be important to consider broader individual and environmental characteristics that could contribute to perceptions of fruit-and-vegetable accessibility and to levels of purchasing and consumption. For instance, having a high density of food stores that sell fruits and vegetables may matter little if transportation is not adequate to access those stores, if neighborhood crime makes shopping at those stores unsafe, or if the produce available in those stores does not meet individual preferences.

Several studies on fruit-and-vegetable intake have examined individual and neighborhood characteristics simultaneously using multilevel models (5–13). Few of these studies, however, have considered both objective and self-reported measures of food environments (6–9). Those that have generally have not given attention to some potentially relevant individual characteristics (eg, overall health status, body mass index [BMI]), and none have considered broader neighborhood issues (eg, public transportation, neighborhood crime).

Our study builds on prior research, considering and adjusting for several relevant individual, household, and neighborhood factors, to assess 1) how individuals’ self-reported perceptions compare with objective measures of neighborhood environments, and 2) how self-reported and objective neighborhood measures each relate to reported fruit-and-vegetable consumption.

The University of Pennsylvania institutional review board approved this study.

For the survey sample of 4,399 respondents poststratified to the 2010 US Census, the median age was 45 years, and 54.6% were women (Table 1). The population included almost as many non-Hispanic whites (39.5%) as non-Hispanic blacks (41.0%); 86.3% were born in the United States. About half (49.3%) had no schooling beyond high school and rated their health as less than very good (51.7%). A third (33.0%) were overweight; nearly another third (30.9%) were obese. About a quarter (23.8%) did not live with other adults; two-thirds (66.0%) lived without children. Almost half (48.8%) had annual household incomes of $50,000 or less. Nearly a quarter (22.2%) lived below the federal poverty level, 15.2% were food insecure, and more than one-third (36.3%) received public assistance. More than one-quarter (26.5%) reported living in neighborhoods with low levels of social capital. A great majority (89.8%) reported consuming fewer than 5 servings of fruits and vegetables per day. For food-environment perceptions, 29.5% reported a need to travel outside their neighborhood to get to a supermarket, 8.3% reported produce was hard or very hard to find in their neighborhood, and 23.6% reported that grocery quality was less than good.

Demographics (including vehicle ownership) varied by neighborhood, as did crime rates (Table 2). Nearly 21% of neighborhoods had a subway or trolley stop (compared with 40.1% of extended neighborhoods). Conventional supermarkets were available in about one-third fewer neighborhoods than limited-assortment supermarkets (13.7% vs 21.6%), and in about one-quarter fewer extended neighborhoods (46.4% vs 63.1%).

Table 3 shows results from multilevel models (in which, importantly, variable categorizations differ in some cases from those shown in Table 1). Perceived supermarket accessibility was directly related to perceived produce availability (OR, 1.76), perceived grocery quality (OR, 1.97), and having either a conventional (OR, 2.04) or limited-assortment supermarket (OR, 1.28) in the extended neighborhood; inverse associations were found with age (OR, 0.99), food insecurity (OR, 0.55), public assistance (OR, 0.74), and mean percentage of households in the neighborhood having no vehicle (OR, 0.99). Perceived produce availability was directly related to perceived supermarket accessibility (OR, 1.90), perceived grocery quality (OR, 2.35), health status (OR, 1.18), social capital (OR, 1.25), and rates of neighborhood crime (OR, 1.07); inverse associations were found with household poverty (OR, 0.85), food insecurity (OR, 0.71), and neighborhood poverty (OR, 0.98). Perceived grocery quality was directly associated with perceived supermarket accessibility (OR, 2.41), perceived produce availability (OR, 2.54), age (OR, 1.02), health status (OR, 1.16), BMI (OR, 1.11), and social capital (OR, 1.27); inverse associations included “other” race (OR, 0.57) and the mean percentage of racial/ethnic minorities in the neighborhood (OR, 0.99). Direct associations with reported fruit-and-vegetable consumption included female sex (OR, 1.73), level of education (OR, 1.30), health status (OR, 1.19), and neighborhood social capital (OR, 1.36); inverse associations were found with non-Hispanic black race/ethnicity (OR, 0.65), household poverty (OR, 0.87), and having a subway or trolley stop in the extended neighborhood (OR, 0.80).

In our analysis of food environments, objective and self-reported measures were not entirely concordant with each other; for example, an objective measure of having a supermarket in the extended neighborhood was associated with better perceived supermarket accessibility but not produce availability or grocery quality. No neighborhood perceptions were related to fruit-and-vegetable consumption. There were nonintuitive associations between some variables in models. Sociodemographic differences in fruit-and-vegetable consumption were evident even after adjusting for individual, household, and neighborhood factors.

Research shows that perceptions of neighborhood food environments are strongly and directly related to each other (4) but may not be strongly or directly related to more objective measures of neighborhood environments. For instance, whereas Moore et al showed a direct association between the perceived availability of healthful foods and store density (26), Caspi et al showed a 31.5% mismatch between measured distance to the nearest supermarket and perceived accessibility (6), Williams et al showed poor correlation between several perceived and objective aspects of neighborhood food environments (ie, accessibility, variety, and cost) (27), and Gustafson et al showed an inverse correlation between the perceived availability of healthful foods and the number of stores selling them (8). The critical point may be that all food stores are not equal. Our analyses showed that supermarket presence (objectively measured) was related to perceived accessibility but not produce availability or grocery quality, suggesting that even if supermarkets are present in a neighborhood, they may not satisfy individual preferences or prompt individuals to shop there.

Positive perceptions of food environments may relate to greater self-efficacy for consuming fruits and vegetables (28). Whether such perceptions relate to reported fruit-and-vegetable consumption is unclear, but earlier multilevel modeling by a member of our team suggested no strong relationships (5). Other multilevel modeling, in different settings using different measures, showed that only a few tested perceptions (of affordability, selection, and accessibility) were directly associated with produce consumption, each with varying degrees of significance for fruits versus vegetables (10,12). A study considering fruits and vegetables together showed perceptions of produce selection and quality — but not affordability — were directly associated with fruit-and-vegetable consumption (13), whereas another study found that none of these perceptions was associated with consumption (29). All of these studies were limited by not considering objective measures of the food environment.

Objective measures of the food environment, along with self-reported measures, were considered in a few studies using multilevel models (6–9). Caspi et al showed that objective distance to a supermarket was not associated with fruit-and-vegetable intake (6), consistent with our findings. Although the Caspi study did show that perceiving a supermarket within walking distance of home was associated with greater reported fruit-and-vegetable intake, the effect size was small (0.5 servings/day) (6). Caldwell et al showed that greater perceived and objective availability were both associated with increased fruit-and-vegetable consumption (7). Although our perception measure was the same as in the Caldwell study, we considered the neighborhood (supermarket presence) as opposed to the store (produce display space) for our objective measure, and we adjusted for multiple factors excluded from the Caldwell models (models that included only sex, age, and dietary intake) (7). Gustafson et al showed that strongly agreeing that a neighborhood has many healthful foods was not associated with fruit and vegetable intake and that living in neighborhoods with a supercenter and convenience store was associated with lower intake (8). These findings suggest that store type matters, consistent with differences we found related to conventional versus limited-assortment supermarkets. Zenk et al showed that neither satisfaction with neighborhood produce nor measured distance to the nearest supermarket was associated with fruit-and-vegetable consumption (9). The Zenk findings are most consistent with our own; however, the Zenk study also showed that when only large grocery stores were considered, having such a store in the neighborhood was associated with consuming 0.69 more daily servings of fruits and vegetables (9). Our analyses did not show a difference in fruit-and-vegetable consumption by store type but did show that larger (conventional) supermarkets were associated with better perceptions of supermarket accessibility than smaller (limited-assortment) supermarkets.

A study by Ball et al considered only objective measures of the food environment and showed that store density did not relate to fruit-and vegetable consumption, nor did it mediate relationships with socioeconomic considerations (11). Our analyses likewise showed no relationship between objective measures of stores and fruit-and-vegetables consumption but did show racial and socioeconomic differences. These differences were not explained by different perceptions of food environments, as others have suggested (12); racial and educational disparities in fruit-and-vegetable consumption were evident in our analyses, even after adjusting for food-environment perceptions and several potentially confounding individual, household, and environmental factors.

Our analyses also found differences in neighborhood food-environment perceptions by sociodemographic characteristics. We found better perceptions of some aspects of neighborhood food environments with younger age; better reported health; not being poor, receiving public assistance, or being food insecure; and living in neighborhoods with greater social capital and less poverty. These issues may relate to people’s mobility, social support, material resources, and ability to travel and shop for food, all of which may constrain conceptualizations of “neighborhood.” For instance, wealthier, healthier people who own or have access to cars might have a more expansive concept of neighborhood such that having a supermarket in the “extended neighborhood” was directly associated with perceptions of produce accessibility, whereas having a supermarket in the census tract only was not. We found better perceptions of some aspects of neighborhood food environments among people who had a greater BMI, who were not a member of a racial/ethnic minority group, and who lived in neighborhoods with fewer minorities. These issues may relate to personal, social, and cultural ideas about eating. For example, even food that is available and high in quality might not be perceived well if the offerings do not satisfy cultural preferences.

Our study produced 2 counterintuitive findings. One was that perceptions of produce availability increased with increasing rates of neighborhood crime. The other was that having a subway or trolley stop in the neighborhood was associated with lower fruit-and-vegetable consumption. Both of these findings could have been due to chance. The former, being of small magnitude and marginal significance, is difficult to rationalize. The latter, however, might be explained by the fact that people may not use public transportation for food shopping (30). Neighborhoods with subway and trolley stops in our analyses had lower numbers of residents who owned a personal vehicle; thus subway and trolley stops might be a marker for lower rates of vehicle ownership and less ability to get to stores selling produce and high-quality groceries.

Our study had several strengths. First, it used population-based individual data from a large urban area and integrated neighborhood data from diverse sources. Second, analyses used rigorous statistical methods with appropriate survey weights and multiple imputation for missing data. Third, analyses modeled several individual, household, and neighborhood characteristics, including both self-reported and objective measures not considered in prior research. Additionally, the analyses examined 2 definitions of neighborhood and 2 definitions of supermarket. Finally, dependent variables included both food-environment perceptions and reported fruit-and-vegetable consumption to assess how both related to different individual, household, and neighborhood characteristics.

Despite methodological strengths, our analyses had limitations. First, the design was cross-sectional, precluding assessment of causality or directionality for the associations found. Second, because of multiple tested associations, nominal P values may not reflect the actual potential for false positive error. Third, the survey variables of interest were all derived from single-item questions, but, reassuringly, studies that used more detailed food-frequency questionnaires and screeners to assess dietary intake reported similar results (8,9,29), and a study testing similar measures of food-environment perceptions showed high reliability, particularly in urban areas (31). Although response rates for the SPHH survey used in our study were far from perfect, evidence increasingly challenges the presumed correlation between response rate and survey bias (14). Another limitation was that survey questions did not assess additional aspects of food environments, such as affordability or cost (10,12,15,16) or the competing accessibility of less healthful foods (15,16). Finally, geographic boundaries used to define neighborhoods may not have aligned precisely with survey respondents’ conceptualizations of their neighborhoods’ borders.

Our analyses found poor correspondence between objective and self-reported measures of neighborhoods and produced some nonintuitive findings on food environments and fruit-and-vegetable consumption. Key associations that emerged depended on definitions of neighborhoods and supermarkets and also on individual sociodemographic characteristics, suggesting that strategies to improve neighborhood food environments and individuals’ diets will require nuance. Although our analyses were cross-sectional, findings do not support certain proposed food-environment modifications to increase produce consumption. Adding supermarkets — even larger stores with wider selections — to urban neighborhoods might improve residents’ perceptions of supermarket accessibility but might not increase their fruit-and-vegetable consumption.