Improvements in Neighborhood Socioeconomic Conditions May Improve Resident Diet

Abstract

Neighborhood socioeconomic conditions (NSECs) are associated with resident diet, but most research has been cross-sectional. We capitalized on a natural experiment in Pittsburgh, Pennsylvania, in which 1 neighborhood experienced substantial investments and a sociodemographically similar neighborhood that did not, to examine pathways from neighborhood investments to changed NSECs and changed dietary behavior. We examined differences between renters and homeowners. Data were from a random sample of households (n = 831) in each of these low-income Pittsburgh neighborhoods that were surveyed in 2011 and 2014. Structural equation modeling tested direct and indirect pathways from neighborhood to resident dietary quality, adjusting for individual-level sociodemographics, with multigroup testing by homeowners versus renters. Neighborhood investments were directly associated with improved dietary quality for renters (β = 0.27, 95% confidence interval (CI): 0.05, 0.50) and homeowners (β = 0.51, 95% CI: 0.10, 0.92). Among renters, investments also were associated with dietary quality through a positive association with commercial prices (β = 0.34, 95% CI: 0.15, 0.54) and a negative association with residential prices (β = −0.30, 95% CI: −0.59, −0.004). Among homeowners, we did not observe any indirect pathways from investments to dietary quality through tested mediators. Investing in neighborhoods may support resident diet through improvements in neighborhood commercial environments for renters, but mechanisms appear to differ for homeowners.

Abbreviations

CI

confidence interval

HEI

Healthy Eating Index

NSEC

neighborhood socioeconomic condition

PHRESH

Pittsburgh Hill/Homewood Research on Shopping and Health

Neighborhood disadvantage has been tied to cardiometabolic health disparities (1). Yet, mixed findings regarding relationships between neighborhood socioeconomic conditions and dietary behaviors and cardiometabolic health (e.g., obesity) (2–12) have limited our understanding of whether improving neighborhoods would improve resident health. Inconsistent findings may be due to the lack of longitudinal data and consideration for whether changes in neighborhood socioeconomic conditions (NSECs) influence health, because much research on neighborhood health effects (13) has been based on cross-sectional studies. Authors of a recent review found that 86% of 432 studies used cross-sectional designs (12). Such approaches limit our ability to discern whether it is NSECs or other types of neighborhood conditions related to NSECs that influence residents’ health, as well as whether investing in neighborhoods might improve dietary behaviors.

In addition, few studies have directly addressed the question of how shifts in NSECs might operate to affect health. Consideration of the pathways through which NSECs may affect resident dietary behaviors is needed to identify the modifiable steps along the causal path that lead to health disparities. Among low-income populations, stress has been associated with nutrient-poor foods, fat intake (14), unhealthy dietary behaviors (15), and uncontrolled and emotional eating (16). Living in disadvantaged neighborhoods can lead to chronic stress, which may lead to a greater preference for fat, sugar, and nutrient-dense foods (17, 18), decreased self-regulation (19), or altered reward and appetite pathways (20–22). Housing instability (e.g., difficulty paying rent, frequent moves) (23) can also be a significant source of stress and can affect the amount households can spend on food (24, 25). Renters may be especially vulnerable to adverse impacts of housing cost on diet because they may have limited funds to pay for healthy foods that are typically more costly than nutrient-poor, energy-dense foods (26–28). Finally, NSECs may affect diet by increasing neighborhood satisfaction, which appears to encourage health behaviors (29, 30).

The Pittsburgh Hill/Homewood Research on Shopping and Health (PHRESH) Study followed randomly sampled households from 2 neighborhoods between 2011 and 2014 to assess the impact of opening a full-service supermarket in 1 neighborhood on residents’ diet. Relative improvements in resident intakes of kilocalories, added sugar, and the combination of solid fats, alcohol, and added sugar were observed among residents of that neighborhood, relative to residents of a sociodemographically matched comparison neighborhood (1 without a supermarket) (29). Although dietary outcomes of residents improved in the intervention neighborhood relative to the comparison, the improvements were not associated with use of the new supermarket (29). Therefore, it remained unclear how changes within the neighborhood might have contributed to residents’ improved dietary behaviors.

One explanation for these results may be improvement in NSECs. During the period the supermarket was constructed and opened, there were additional investments in the intervention neighborhood, including housing redevelopment and the opening of a nonprofit center focused on sustainable energy and workforce development. We estimated that the intervention neighborhood underwent more than $295.7 million in total development between 2011 and 2014—25 times the estimated investments ($11.5 million) in the comparison neighborhood over the same period (31). Prior PHRESH analyses showed concomitant increased commercial and residential sales prices in the intervention neighborhood (31). To our knowledge, no prior analysis has tested whether these investments and associated NSEC changes might account for the differential dietary improvement across neighborhoods.

In the present analysis, we explored this question, testing whether differential neighborhood socioeconomic shifts explain the previously observed differential dietary improvements across neighborhoods. As a secondary question, we examined 2 psychological pathways by which NSEC shifts may contribute to changes in residents’ diet: stress and neighborhood satisfaction. We hypothesized that 1) improved NSECs explain differentially improved dietary behaviors associated with neighborhood investments, and 2) increased neighborhood satisfaction and reduced perceived stress further explain this relationship, mediating the pathway between improved NSECs and improved dietary behaviors. Finally, we examined whether associations differed between residents who owned homes and those who were renters, hypothesizing that homeowners benefit more from neighborhood investments. Figure 1 depicts hypothesized pathways.

Figure 1.

Hypothesized relationships in Pittsburgh Hill/Homewood Research on Shopping and Health Study, 2011–2014.

METHODS

Study design and participants

The PHRESH Study enrolled adults residing at addresses randomly selected from 2 racially isolated, low-income Pittsburgh neighborhoods: the Hill District (the intervention neighborhood receiving investments) and Homewood (the comparison neighborhood with few investments). The 2 neighborhoods were sociodemographically matched for race/ethnicity, average household income, and similar factors. Trained data collectors went door to door May through December 2011 to enroll participants and administer surveys; a follow-up data collection was conducted May through December 2014. Sampling approach, recruitment, and eligibility details are described elsewhere (29, 32). At baseline, the primary food shoppers from 1,372 households were enrolled (Hill District, n = 897; Homewood, n = 475), and 831 were re-interviewed in 2014. All study protocols were approved by RAND’s Human Subjects Protection Committee.

Measures

Dietary behavior

We assessed resident diet using self-administered 24-hour dietary recalls, administered during the in-person surveys and repeated 7–12 days later by phone (33). We derived Healthy Eating Index (HEI)-2010 (34) scores to measure overall dietary quality based on compliance with the US Dietary Guidelines for Americans, calculating a single per-person score for each wave on the basis of the average of the 2 recalls (35). Scores range from 0 to 100; a score greater than 80 indicates good dietary quality, a score of 51–80 reflects a need for improvement, and a score less than 51 indicates poor diet. Recent estimates show an average score of 57.2 for the US population and 55.0 among Non-Hispanic Blacks (36).

Neighborhood and homeowner status

We assigned participants to the neighborhood in which they resided in 2011. Residents reported whether they rented or owned their home in 2011.

Neighborhood socioeconomic conditions

Commercial andresidential property sales data. Hill District and Homewood comprise 9 sub-neighborhoods. We purchased data from RealStats, Inc., a firm that collects information on all recorded real estate transactions throughout the Pittsburgh metropolitan region; the data included 40,300 property sale transactions that occurred between 1990 and 2015 in these 9 sub-neighborhoods, as described elsewhere (31). To account for differences in type of property that influenced sales price, we derived value-added measures, estimating the sales price as a function of lot size, number of bedrooms, frontage footage, number of baths, square footage of building, number of stories, year built, and transaction type, as well as indicators for which of the 9 sub-neighborhoods the sale was in (which captured additional value in price from being in a given sub-neighborhood, all else being equal). To account for annual fluctuations, we estimated average value-added across 3 years of data around baseline (2010–2012) and follow-up (2013–2015). We used log-transformed value-added estimates in analyses to account for skewed distributions. At baseline and follow-up, each participant was assigned the sales value added for the sub-neighborhood in which they lived at the time.

Neighborhood population living below poverty

We used 5-year tract-level estimates from the American Community Survey (37) of the percentage of the population living with income below the federal poverty level for baseline (2006–2010) and follow-up (2012–2016). Thirteen census tracts spanned the intervention (n = 6) and the comparison (n = 7) neighborhoods. We assigned estimates for the percentage of the population living below poverty to each participant at baseline and at follow-up on the basis of the census tract in which they resided at the time.

Psychological factors

Neighborhood satisfaction.

We measured individual-level neighborhood satisfaction with the survey question, “All things considered, would you say you are very satisfied, satisfied, dissatisfied, very dissatisfied, or neutral—neither satisfied nor dissatisfied with your neighborhood as a place to live?” (38) Responses were assigned values 1 to 5, where higher values indicate greater satisfaction.

Perceived stress.

We used the well-validated Perceived Stress Score-4 to measure perceived stress (39, 40). Residents were asked 4 questions with a past-month reference period: “1) how often have you felt that you were unable to control the important things in your life?; 2) how often have you felt confident about your ability to handle your personal problems?; 3) how often have you felt that things were going your way?; and 4) how often have you felt difficulties were piling up so high that you could not overcome them?” Response categories of never, almost never, sometimes, fairly often, and very often were assigned values of 0 through 4. Questions 2 and 3 were reverse coded and values across the 4 questions were summed so that higher values indicate more stress (for years 2011 and 2014, Chronbach α = 0.66 and 0.71, respectively).

Covariates

Body mass index.

We calculated body mass index as weight without shoes (in kilograms) divided by height (in meters squared); the interviewer conducted the measurements. Interviewers measured height to the nearest eighth inch (3 mm) using a carpenter’s square (triangle) and an 8-foot (2.4 m) folding, wooden ruler. Weight was measured to the nearest tenth of a pound (0.04 kg) using the SECA Robusta 813 digital scale.

Food security.

Residents responded to the US Department of Agriculture 10-item Adult Food Security Survey Module (41, 42) that incorporates questions about conditions and behaviors that characterize households when they are having difficulty meeting basic food needs. On the basis of standard coding procedures, we classified household food-security status as high, marginal, low, or very low.

Sociodemographics.

We included data on participants’ age, sex, household income, marital status, educational attainment, children in the household, participation in Supplemental Nutrition Assistance Program, and number of years lived in the neighborhood.

Residential relocation.

We created a variable to indicate the 109 residents who moved across sub-neighborhoods but remained within their larger neighborhoods between 2011 and 2014.

Statistical analyses

We calculated means and standard deviations for continuous variables and percentages for categorical variables for the study population and by neighborhood using Stata 15.0 (StataCorp, College Station, Texas). We constructed a structural equation model to examine longitudinal pathways from neighborhood to dietary outcomes, including direct and indirect pathways through NSECs, neighborhood satisfaction, and stress. (43) We used Mplus, version 7.11,⁴³ with robust maximum likelihood estimation. Modeling specification and weighting details are presented in the Web Appendix.

We tested for moderation of associations by homeowner status with multigroup models using likelihood ratio tests to compare 1 model in which we constrained path coefficients to be equal across homeowners and renters and another model allowing the coefficients to differ between homeowners and renters. A nonstatistically significant difference (P > 0.05) would indicate that parameters were similar between groups (43).

Finally, to account for sample attrition between baseline and follow-up, we used weights so results could be generalized to the baseline sample. Attrition weights were calculated using the inverse predicted probability of response at follow-up, based on a logistic regression model that included all the sociodemographic and additional baseline individual characteristics as predictors of response at follow-up. We also estimated an unweighted multigroup model to assess robustness of findings.

RESULTS

Among 831 participants who completed baseline (2011) and follow-up (2014) surveys, we excluded those for whom we had incomplete property sales data at follow-up (n = 62), who moved from the Hill District to Homewood or vice versa during the study period (n = 6), were missing dietary recalls (n = 15), or reported extreme dietary intakes (<500 kcal/day or > 4,000 kcal/day) (n = 9). The analytic sample included 739 adults.

We compared all covariate and outcome variables between excluded and included individuals using t tests and χ² statistics. Compared with the 92 excluded persons, those in our analysis were older (mean age, 55 vs. 48 years, 95% confidence interval (CI): 3.3, 10.2); more likely to be homeowners (35% vs. 10%, 95% CI: 14.2, 34.7); more likely to be food secure (69% vs. 46%, 95% CI: 12.9, 33.7); less likely to participate in Supplemental Nutrition Assistance Program (49% vs. 69%, 95% CI: 3.1, 8.8); had a higher mean annual household income (in thousands) ($13.8 vs. $10.3, 95% CI: 0.5, 6.5); lived longer in their neighborhood (mean 29.6 vs. 21.6 years, 95% CI: 3.0, 13.1 years); and reported higher neighborhood satisfaction (mean 3.6 vs. 3.1, 95% CI: 0.2, 0.7).

In 2011, study participants were, on average, 55 years old, predominantly (95%) non-Hispanic Black, and female (78%) (Table 1). More Hill District (ie, the investment neighborhood) participants were renters (71%) compared with Homewood, where approximately half (52%) of the participants were renters. Several characteristics differed between renters and homeowners in both neighborhoods. Renters reported lower incomes, less food security, lower educational attainment, were less likely to be married, and lived fewer years in their neighborhood than did the homeowners. Renters were also more likely to participate in the Supplemental Nutrition Assistance Program and to live with children, compared with homeowners. Because homeowners and renters often lived in different sub-neighborhoods, it was possible for property sales prices to vary by homeowner status as well as neighborhood (Table 2). For example, residential property sales value added at baseline was highest for Hill District homeowners: $16,800 higher in these residents’ sub-neighborhoods than the average value across both neighborhoods.

Table 1.

Study Population Characteristics in 2011, by Neighborhood (Hill District and Homewood, Pittsburgh, Pennsylvania) and Renter Versus Homeowner Status, Pittsburgh Hill/Homewood Research on Shopping and Health Study, 2011–2014

Baseline Characteristic	Hill District (Intervention), %		Homewood (Comparison), %
	Renter (n = 363)	Homeowner (n = 145)	Renter (n = 119)	Homeowner (n = 112)	Overall (n = 739), %
Black race/ethnicity	95.0	93.8	95.0	95.5	94.9
Age, yearsa	53.8 (16.5), (20–87)	59.1 (13.0), (27–100)	55.3 (16.2), (21–88)	55.0 (13.9), (19–91)	55.3 (15.7), (19–100)
Annual per capita household income (in thousands)b	5.0 (5.0–15.0)	15.0 (11.3–33.8)	5.0 (5.0–15.0)	12.5 (7.5–25.0)	8.8 (5.0–15.0)
Low or very low food security	36.6	19.3	32.8	24.1	30.7
Highest educational attainment
Less than high school	13.8	4.8	14.3	5.4	10.8
High school diploma	49.9	29.7	47.1	23.2	41.4
Some college/technical school	28.1	40.0	30.3	45.5	33.4
College degree	8.3	25.5	8.4	25.9	14.3
Any children in household	28.1	20.7	29.4	24.1	26.3
Female sex	82.6	76.6	71.4	72.3	78.1
Married/living with partner	10.5	28.5	16.8	26.1	17.4
Years lived in the neighborhoodc	31.9 (24.1)	41.4 (20.7)	12.1 (14.7)	25.9 (17.5)	29.6
Household participated in SNAP in past year	63.4	22.1	64.7	21.4	49.1

Abbreviation: SNAP, Supplemental Nutrition Assistance Program.

^a Values are expressed as mean (standard deviation), range.

^b Values are expressed as mean (interquartile range).

^c Values are expressed as mean (standard deviation).

Table 2.

Baseline Neighborhood Socioeconomic Conditions, Perceptions, and Diet by Year (2011), Neighborhood (Hill District and Homewood, Pittsburgh, Pennsylvania), and Renter Versus Homeowner Status, Pittsburgh Hill/Homewood Research on Shopping and Health Study, 2011–2014

Baseline Characteristic	Hill District (Intervention), mean (SD)		Homewood (Comparison), mean (SD)
	Renter (n = 363)	Homeowner (n = 145)	Renter (n = 119)	Homeowner (n = 112)
Exposure: residents’ neighborhood socioeconomics
Percent population with income below poverty	45.1 (11.3)	33.7 (14.2)	29.0 (8.0)	30.7 (7.9)
Residential sales, value addeda	3.9 (3.0)	16.8 (13.2)	−19.1 (7.4)	−18.0 (8.4)
Commercial sales, value addeda	−51.5 (77.7)	2.0 (33.5)	−25.7 (28.6)	−16.7 (22.8)
Mediators: residents’ perceptions
Neighborhood satisfactionb	3.7 (1.1)	3.7 (1.2)	3.3 (1.1)	3.2 (1.1)
Perceived stressc	4.9 (3.2)	3.8 (2.9)	5.2 (3.6)	5.1 (3.3)
Outcome: residents’ dietary intake
Dietary quality (Healthy Eating Index 2010)	47.7 (12.6)	52.5 (13.3)	48.7 (11.6)	51.5 (14.7)

Abbreviation: SD, standard deviation.

^a Property sales value added (in thousands USD) are expressed relative to the average across the 2 neighborhoods. A positive value indicates that the value is higher than the average across Hill District and Homewood.

^b Scale score range, 1–5.

^c Scale score range, 0–16.

The multigroup test by homeownership status was significant (χ² = 255.53 (34 df); P < 0.001), indicating that associations among the variables modeled may differ for these 2 groups of participants. In addition, the model that allowed for differing path coefficients for homeowners and renters had better fit root mean square error of approximation (0.048) and Comparative Fit Index (0.93) values than the model with path coefficients constrained to be equal for homeowners and renters (root mean square error of approximation = 0.054; Comparative Fit Index = 0.90), confirming the need to examine models separately for the 2 groups. Thus, we present estimated pathway coefficients and 95% confidence intervals for associations that were significant or near statistical significance (P < 0.10) among renters (Figure 2) and homeowners (Figure 3) separately.

Figure 2.

Structural equation model estimates for pathways to the Healthy Eating Index among renters, Pittsburgh Hill/Homewood Research on Shopping and Health Study, 2011–2014. Solid lines represent positive associations and dashed lines represent negative associations.

Figure 3.

Structural equation model estimates for pathways to the Healthy Eating Index among homeowners, Pittsburgh Hill/Homewood Research on Shopping and Health Study, 2011–2014. Solid lines represent positive associations and dashed lines represent negative associations.

Renters

Among renters, living in the investment neighborhood was associated with improved HEI both directly (β = 0.27, 95% CI: 0.05, 0.50) and indirectly through improvements in neighborhood socioeconomic conditions (Figure 2 and Web Table 1, available at https://doi.org/10.1093/aje/kwaa220). Hill District renters experienced increased commercial sales prices (β = 0.83, 95% CI: 0.12, 1.53), which were associated with higher HEI (β = 0.34, 95% CI: 0.15, 0.54). Increased residential sales prices associated with renting in the Hill District (β = 0.57, 95% CI: 0.13, 1.01) were marginally associated with renters’ lower HEI (β = −0.30, 95% CI: −0.59, 0.54). However, the net association of these 2 indirect paths was a positive association of increased sales prices with improved HEI (β = 0.11, 95% CI: −0.15, 0.37). We did not observe indirect pathways from NSECs to dietary quality through psychological mediators. However, improved neighborhood satisfaction was associated with renters’ improved HEI (β = 0.86, 95% CI: 0.21, 1.51). Web Table 1 presents all model estimates and Web Table 2 presents unweighted estimates that were nearly identical.

Homeowners

Among homeowners (Figure 3 and Web Table 1), living in the investment neighborhood was also directly associated with improved HEI (β = 0.51, 95% CI: 0.10, 0.92). Consistent with renter results, property sales value added (commercial and residential) increased in Hill District more so than Homewood, and the percentage of residents living below the poverty line also decreased. Neighborhood satisfaction increased, as well. However, none of these changes were associated with changes in HEI, indicating that changes in NSECs and psychological mediators did not play a role in the pathways from investments to dietary quality. In addition, homeowners’ increased stress was associated with worsening HEI (β = −0.43, 95% CI: −0.76, −0.11), but this was unrelated to neighborhood investments or changes in NSECs.

DISCUSSION

Prior work showed differential improvement in diet among residents in a neighborhood with a new full-service supermarket, compared with that of residents in a neighborhood without such a market, but dietary improvements were not associated with use of the supermarket. Building on this finding, we examined, using structural modeling, whether differential changes in NSECs may have explained the differential changes in dietary outcomes. This approach refines prior PHRESH Study difference-in-difference analyses by accounting for measurement error, modeling multiple hypothesized pathways simultaneously, and testing for heterogeneity across homeownership status. PHRESH Study researchers previously observed changes in multiple dietary outcomes (i.e., added sugars, kilocalories, solid fats, alcohol, and added sugar) and a trend in HEI-2010 (29). To simplify presentation and to reduce potential type I error, we focused here on HEI-2010 as a summary measure of dietary quality. Using this method, we found additional evidence that diet improved more in the investment neighborhood (i.e., the neighborhood with the new supermarket) compared with the comparison neighborhood.

We also found that, among renters, these changes may have partially been the result of shifts in NSECs resulting from neighborhood investments. Increases in commercial and residential property values appeared to partially mediate associations between neighborhood investment and a summary measure of diet quality for renters only. And importantly, increased property values seemed to concurrently relate to both improved and worsened diet, depending on the type of property (i.e., commercial or residential). Specifically, increasing commercial values were associated with relatively improving dietary quality, whereas increasing residential values were associated with relatively worsening dietary quality, resulting in a small positive net mediated association.

Associations between increased commercial sales and improved renter dietary quality may reflect greater economic activity that infused vitality and perceived opportunity into the intervention neighborhood. We would expect positive associations for both renters and homeowners; however, the relatively small sample of homeowners (n = 145) may have limited our power to observe pathways from commercial sales value-added increases to dietary quality among this group. Negative associations between residential sales and dietary quality among renters may be due to correspondingly higher rental prices, resulting in the renters’ decreased ability to purchase and consume high-quality foods/beverages. We did not observe mediation by neighborhood poverty, nor did we observe mediation for any of the NSEC measures among homeowners.

Property value changes over time were different for renters than for homeowners because, in general, renters and homeowners lived in different geographic locations (sub-neighborhoods). That is, even within the larger neighborhood, renters and homeowners did not always live close to each other and homeowners lived in areas with greater property values than did renters. Although the homeowners nonetheless experienced relative increases in property values, just as renters did, we found no associations between these shifts and diet.

Notably, among both homeowners and renters, direct pathways between neighborhood investments and diet remained after accounting for changes in NSECs and potential psychological mediators. This finding suggests that other untested mediators may explain associations between investments and dietary improvement. One possibility is that neighborhood investments could have resulted in increased employment. However, in prior research, Hill District residents’ employment status did not improve more so than it did for Homewood residents (44). Another possibility is that investments improved resident hope and optimism that, in turn, may have driven dietary behaviors. Hopelessness has been linked to poor health behavior because individuals experiencing hopelessness perceive limited future opportunity for health improvement (45, 46). The investments also could have coincided with improvements in food environments beyond the new supermarket that would help residents support their diet quality. However, neighborhood audits previously found positive and negative changes in the neighborhood food-store availability of healthy foods in the investment neighborhood during this time (47). How restaurant food/beverage options may have changed is unknown.

Neighborhood disadvantage has been coupled with poor diet, increased rates of obesity, and cardiovascular disease–related morbidity and mortality (1, 48, 49). In most studies, researchers have not been able to examine changing neighborhood features or economic characteristics, perhaps explaining inconsistent findings (50, 51–56). For example, in the Moving to Opportunity for Fair Housing Demonstration Program, low-income families were randomly assigned to receive varying types of housing assistance and followed participants, including those who used the assistance to move into more socioeconomically advantaged neighborhoods (51). Moving to a low-poverty versus high-poverty neighborhood was associated with modest improvements in extreme obesity (51). However, in another Moving to Opportunity study (57), researchers found that child health problems may have limited families’ ability to move, suggesting selection bias. Resident health in association with neighborhood improvement or decline (vs. moving into a socioeconomically different neighborhood) (58) has been assessed in only a few randomized studies. In contrast, we were able to compare changes in NSECs resulting from major investments in 1 neighborhood with those occurring in a matched comparison neighborhood, we observed that these investments were associated with relatively improved diet both directly and (among renters) indirectly.

Our study has limitations. The research was set in 2 low-income, predominantly Black, urban neighborhoods; therefore, findings may not generalize to other neighborhoods. The investments that took place (31), though varied, may also not be generalizable to other neighborhood investments. In addition, we treat the neighborhood investments that occurred as exogenous (i.e., unrelated to resident characteristics), but we cannot confirm this assumption. Unmeasured confounding could bias our estimates away from or toward the null. In particular, the intervention neighborhood had higher property values at baseline than did the comparison neighborhood, which suggests an imperfectly matched control. Or, the observed shifts in residential and commercial property values could represent other results of investments (e.g., new roadways could affect traffic flow and influence home values). Last, we present 1 hypothesized causal model, but there may be other valid causal models that accurately reflect underlying relationships.

Despite these limitations, our study provides evidence that neighborhood investment may play a role in dietary behaviors, and that associated shifts in NSECs may operate differently for renters than for homeowners. Our findings shed light on some of the positive and negative downstream consequences of investing in neighborhoods, but their complexity indicates more needs to be learned about how and why investments may influence diet and which residents are most likely to benefit.