Skip to main content
NCBI home page
NIHPA Author Manuscripts logoLink to NIHPA Author Manuscripts
. Author manuscript; available in PMC: 2017 May 1.
Published in final edited form as: J Adolesc Health. 2016 Mar 16;58(5):551–557. doi: 10.1016/j.jadohealth.2016.01.004

Temporal Effects of Child and Adolescent Exposure to Neighborhood Disadvantage on Black/White Disparities in Young Adult Obesity

Nicole Kravitz-Wirtz a,1
PMCID: PMC4842347  NIHMSID: NIHMS756647  PMID: 26995292

Abstract

Purpose

This study investigates the effects of duration and timing of exposure to neighborhood disadvantage from birth through age 17 on obesity incidence in early adulthood, as well as black/white disparities therein.

Methods

Individual- and household-level data from the 1970 to 2011 waves of the Panel Study of Income Dynamics are merged with census data on respondents’ neighborhoods (n=1,498). Marginal structural models with inverse-probability-of-treatment and censoring weights are used to quantify the probability of being obese at least once between ages 18 and 30 as a function of average exposure to neighborhood disadvantage throughout childhood and adolescence or during each of three developmental stages therein.

Results

Longer-term exposure to neighborhood disadvantage from ages zero to 17 is more common among blacks than whites and is associated with significantly greater odds of being obese at least once in early adulthood. Exposure to neighborhood-level deprivation during adolescence (ages 10-17) appears more consequential for future (young adult) obesity than exposure that occurs earlier in childhood.

Conclusions

The duration and timing of exposure to neighborhood disadvantage during childhood and adolescence are associated with obesity incidence in early adulthood for both blacks and whites. However, given inequalities in the likelihood and persistence of experiencing neighborhood disadvantage as children and youth, such adverse effects are likely to be more concentrated among black versus white young adults.

The increased prevalence and associated health consequences of obesity have been called among the most burdensome public health issues facing the nation today [1]. Although interventions targeting individual dietary and exercise habits retain popular appeal, there is growing consensus among public health stakeholders that understanding and addressing obesity, as well as racial disparities therein, requires attention to factors in the broader environment [2, 3]. Prior evidence suggests, for example, that residents of under-resourced neighborhoods characterized by the relative absence of healthy food stores, a preponderance of fast food and alcohol outlets, and systemic constraints on physical activity and social interaction tend to have higher body mass index (BMI) [4-11]. Numerous studies further document that exposure to disadvantaged neighborhoods is unequally distributed both in the population and across the lifecourse. African Americans, in particular, are not only more likely than statistically comparable whites to ever reside in areas characterized by high levels of social and structural adversity, but also more likely to do so for prolonged periods of time [12-16].

These findings point to the importance in neighborhood effects research of characterizing if, as well as when and for how long, residential exposures occur. For instance, persistent exposure may be necessary for young people to learn the skills associated with and to internalize preferences for healthier food and more active lifestyles [17]. In addition, there is research to suggest that the increasing desire for autonomy and expanding social interactions during the adolescent stage of the lifecourse may make it a sensitive period for the development of obesity [18-21], as well as for the effects of neighborhoods on health-related outcomes more generally [22]. Until recently, however, the majority of scholarship in this area measured neighborhood characteristics only once or over just a short window of observation, conflating persons who were recently exposed with those who have experienced residential adversity for sustained or during developmentally sensitive periods [23].

Such a conceptualization is inconsistent with most theories of neighborhood effects, which tend to specify mechanisms that are affected by the duration and timing of exposure [14, 24, 25], as well as with a lifecourse perspective in which experiences earlier in life are posited to have formative and enduring impacts on future outcomes, even when controlling for more contemporaneous determinants [26]. The reliance on largely cross-sectional data has made it particularly difficult to account for the evolving and interrelated nature of individual and neighborhood characteristics over time, as well as for the movement of individuals who already have physically active lifestyles and healthy diets into neighborhoods with characteristics that support such behaviors. The small but developing body of longitudinal research in this area suggests that children and youth, particularly girls and young women of color, who experience more disadvantaged circumstances not only have higher BMI at baseline, but also gain body mass at a faster rate over time compared to their more affluent counterparts [21, 27-30]. While these studies typically assess respondents’ weights and heights at several points over the life course, they tend to measure neighborhood characteristics along with individual-level control variables at only one time point, usually the start of data collection or respondents’ year of birth, or ignore time-varying confounding when such variables are measured repeatedly. As a result, the development and implementation of associated health promotion interventions continues to be hampered by challenges to causal inference and most large-scale efforts still focus on the characteristics of people rather than the characteristics of places.

This study uses the 1970 to 2011 waves of the Panel Study of Income Dynamics (PSID) merged with census data on respondents’ neighborhoods to investigate the effects of duration and timing of exposure to neighborhood disadvantage from birth through age 17 on obesity incidence in early adulthood, as well as black/white disparities therein. It employs a statistical approach that explicitly accounts for time-varying phenomena, allowing individual- and household-level characteristics to moderate the relationship between neighborhoods and BMI while also adjusting for potential confounding due to neighborhood selection bias at each wave of data collection. Findings, therefore, provide among the strongest evidence to date for the adverse effects on young adult obesity of more prolonged exposure to neighborhood disadvantage throughout the child and adolescent lifecourse, as well as for the sensitivity of such effects to the developmental timing of exposure.

Methods

The PSID is a longitudinal survey of US residents and their families conducted annually between 1968 and 1997 and every two years thereafter. The analytic sample for this study consists of the 4,334 black and white individuals born into PSID family units between 1970 and 1980. Respondents were dropped if they were not continuously present for every year from ages zero to 17 or if they did not respond to any questions about weight and height in young adulthood, leaving 1,498 individuals. Final sample members were more likely to be white, female, to be born to a married, slightly older mother, and into a household in which the head had attended at least some college. Adjustment for nonrandom attrition using censoring weights is discussed below.

Dependent Variable

Obesity is determined using self-reported weight and height to calculate BMI, the ratio of weight in kilograms (kg) to height in meters squared (m2). Previous research shows a strong correlation between self-reported and directly measured weight and height and no significant differences in reporting across gender or race/ethnicity [31, 32]. Given this study’s focus on obesity incidence in young adulthood, the outcome of interest is any report of weight and height amounting to a BMI greater than or equal to 30 kg/m2 between ages 18 and 30.1

Independent Variable

Similar to most research in this area, census tracts are used to approximate neighborhood boundaries. Despite limitations to this operational definition [cf. 33], there is broad consensus that census data at the tract level not only provide convenient access to considerable information over extensive time periods, but also serve as a reasonable proxy for, or are at least highly correlated with, the causally relevant definition of a neighborhood [34, 35]. The measure of neighborhood disadvantage is based on the following census tract items derived from the Neighborhood Change Database (NCDB) [36] over the period 1970 to 20002: (1) proportion of residents below the poverty line; (2) proportion of residents in the civilian labor force and unemployed; (3) proportion of households with public assistance income; (4) proportion of households with children that are female-headed; (5) proportion of residents with less than a high school diploma; (6) proportion of residents with a bachelors or graduate/professional degree; and (7) proportion of residents employed in managerial or professional/technical occupations.

Using principal components analysis (PCA), these seven items are combined to generate a composite index of neighborhood disadvantage. Specifically, the results of PCA produce a set of factor loadings for each item for each principal component. Using the factor loadings from the first principal component as weights, a neighborhood disadvantage score is then constructed for each census tract (or neighborhood) at every year between 1970 and 2000. The neighborhood disadvantage scores are then divided into quintiles ranging from the least (level 1) to the most (level 5) disadvantaged based on the distribution of all tract-year scores derived from the NCDB between 1970 and 2000, and subsequently merged with individual-level data on PSID respondents using the PSID’s supplemental, restricted-use Geocode Match Files. Procedures were approved by the University of Washington’s Institutional Review Board and through contractual agreement with the University of Michigan.

For the duration component of the analysis, a cumulative measure of exposure to neighborhood disadvantage throughout childhood and adolescence is calculated as the average of all the neighborhood disadvantage quintiles to which respondents were exposed from ages one through 17. Likewise, for the timing component, three separate measures of timing-specific exposure to neighborhood disadvantage during early childhood, late childhood, and adolescence are calculated as the average of all neighborhood disadvantage quintiles experienced by respondents from ages one to five, ages six to 11, and ages 12 to 17, respectively. Neighborhood disadvantage at birth is not included in these calculations but rather as part of a vector of time-invariant covariates.

Covariates

Individual- and household-level variables are included in analyses as either time-invariant or time-varying. Time-invariant covariates include a respondent’s race, gender, birthweight, mother’s age at birth, mother’s marital status at birth, and household head’s educational attainment at birth. Time-varying covariates, measured for every respondent at each wave between ages zero and 17, include household head’s marital status, employment status, and work hours, as well as family size, homeownership, public assistance receipt, and total household income, standardized using the Consumer Price Index to 1985 dollars.

Statistical Analysis

Analyses employ marginal structural logistic regression models for the probability of being obese at least once in early adulthood as a function of (1) average exposure to neighborhood disadvantage throughout childhood and adolescence or during each of three developmental stages therein, and (2) time-invariant covariates. Adjustment for time-varying covariates is achieved through the use of inverse-probability-of-treatment (IPT) weights. This approach and its utility for longitudinal neighborhood effects research have been described in detail elsewhere [24, 37]. In brief, the IPT weights are derived by using a respondent’s neighborhood conditions in the prior year, time-invariant covariates, and both prior year and concurrent time-varying covariates to predict their probability of exposure to their actual (observed) level of neighborhood disadvantage in subsequent years. The inverse of this predicted probability is used to weight each respondents’ contribution to a pseudo-population in which time-varying covariates are balanced in expectation across the five levels of neighborhood disadvantage.3 Obesity incidence in young adulthood can then be regressed on either cumulative or timing-specific measures of exposure to neighborhood disadvantage using a conventional logistic regression model that does not include the time-varying covariates as controls.

Missing Data and Sample Attrition

Missing data on all independent variables are multiply imputed using the two-fold fully conditional specification algorithm [38]. The two-fold approach imputes missing values using chained equations at each time point conditional on information at the same time point plus or minus user-specified adjacent time points. This study employs the two years before and after any missing values for the imputation based on the assumption that measurements taken outside this window are unlikely to provide substantial additional information and may result in over-fitting. Due to computational constraints, only the neighborhood disadvantage scores rather than the individual census tract items are imputed.

To minimize the effects of biasing attrition, censoring weights for each respondent at each year are generated in the same manner as the IPT weights described above, except now the weights model the probability of remaining in the study for each respondent at each year, conditional on the same covariates as before. The pseudo-population to which the final logistic regression models are fit is actually constructed based on the product of the IPT and censoring weights for each respondent.

Results

Tables 1 and 2 present descriptive statistics for the time-invariant and time-varying sample characteristics, respectively, for the total sample as well as separately for black and white respondents.

Table 1.

Time-invariant sample characteristics

Total
 Black
 White
(n=1,498) (n=500) (n=998)
Ever obese in early adulthood, percent
 BMI ≥ 30 kg/m2 31.04 41.40 25.85
 BMI < 30 kg/m2 68.96 58.60 74.15
Average BMI across early adulthood, mean (SD) 26.75 (5.62) 28.30 (6.08) 25.97 (5.20)
Gender, percent
 Female 53.47 58.80 50.80
 Male 46.53 41.20 49.20
Birthweight, percent
 Less than 88 ounces 7.01 10.20 5.41
 88 ounces or more 92.99 89.80 94.59
Mother’s marital status at birth, percent
 Unmarried 20.16 48.40 6.01
 Married 79.84 51.60 93.99
Household head’s education at birth, percent
 Less than high school 28.37 53.60 15.73
 High school graduate 36.92 33.60 38.58
 At least some college 34.71 12.80 45.69
Mother’s age at birth, mean (SD) 24.59 (5.06) 22.75 (5.09) 25.51 (4.80)
Open in a new tab

Note: Statistics reported for respondents not lost to follow-up before age 18 and who answered at least one set of questions about height and weight in early adulthood (first of 10 imputation datasets); BMI=Body mass index

Table 2.

Time-varying sample characteristics

Total (n=1,498)
 Black (n=500)
 White (n=998)
Age 3 Age 9 Age 15 Age 3 Age 9 Age 15 Age 3 Age 9 Age 15
Neighborhood disadvantage quintile, percent
 1st quintile (least disadvantaged) 11.62 13.95 17.36 2.00 2.40 4.80 16.43 19.74 23.65
 2nd quintile 14.09 13.42 15.89 2.60 3.60 5.40 19.84 18.34 21.14
 3rd quintile 17.82 18.89 20.16 7.20 5.80 8.80 23.15 25.45 25.85
 4th quintile 23.03 21.63 20.29 16.40 16.60 19.00 26.35 24.15 20.94
 5th quintile (most disadvantaged) 33.44 32.11 26.30 71.80 71.60 62.00 14.23 12.32 8.42
Household head’s marital status, percent
 Unmarried 16.96 22.16 25.90 37.20 44.40 49.60 6.81 11.02 14.03
 Married 83.04 77.84 74.10 62.80 55.60 50.40 93.19 88.98 85.97
Household head’s employment status, percent
 Unemployed 16.09 16.49 15.29 31.80 33.00 30.20 8.22 8.22 7.82
 Employed 83.91 83.51 84.71 68.20 67.00 69.80 91.78 91.78 92.18
Public assistance receipt, percent
 Received public assistance 10.41 11.48 7.14 23.60 27.60 18.60 3.81 3.41 1.40
 Did not receive public assistance 89.59 88.52 92.86 76.40 72.40 81.40 96.19 96.59 98.60
Homeownership, percent
 Does not own home 43.32 35.38 30.17 67.80 61.00 52.40 31.06 22.55 19.04
 Owns home 56.68 64.62 69.83 32.20 39.00 47.60 68.94 77.45 80.96
Household income in (1985) $1000, mean (SD) 19.34
(15.84)		 22.79
(25.45)		 25.86
(28.89)		 15.13
(10.92)		 15.96
(13.22)		 17.47
(15.00)		 21.45
(17.43)		 25.79
(25.45)		 30.07
(32.98)
Household head’s work hours/week, mean (SD) 37.43
(17.22)		 36.66
(18.12)		 37.52
(17.54)		 28.22
(17.88)		 27.05
(19.56)		 28.54
(19.44)		 42.04
(14.89)		 41.48
(15.24)		 42.02
(14.57)
Family size, mean (SD) 4.28
(1.58)		 4.55
(1.31)		 4.04
(1.26)		 4.74
(2.25)		 4.70
(1.75)		 4.62
(1.52)		 4.06
(1.03)		 4.47
(1.01)		 4.30
(1.09)
Open in a new tab

Note: Statistics reported for respondents not lost to follow-up before age 18 and who answered at least one set of questions about height and weight in early adulthood (first of 10 imputation datasets).

Temporal Patterns of Exposure to Neighborhood Disadvantage

Figure 1 illustrates overall and race-specific patterns of exposure to neighborhood disadvantage throughout the entire child and adolescent lifecourse, as well as separately for early childhood, late childhood, and adolescence. Higher values indicate more prolonged exposure to neighborhood disadvantage. The mean level of neighborhood disadvantage to which respondents, irrespective of race, are exposed throughout childhood and adolescence is 3.40, on average – that is, children and youth tend to be exposed to neighborhoods that fall between the third and fourth quintiles of the disadvantage distribution. This value is in the same general range during each of the three developmental stages, although it is slightly smaller during adolescence (mean=3.26).

Figure 1.

Figure 1

Open in a new tab

Mean level of cumulative (ages 1-17) and timing-specific (ages 1-5, 6-11 and 12-17) exposure to neighborhood disadvantage for all respondents and for the black and white subsamples; NH=Neighborhood

When the total sample is stratified by race, the mean level of exposure to neighborhood disadvantage from ages one through 17 is 4.45 among black respondents compared to just 2.88 among white respondents. This suggests that while a considerable share of black respondents reside, on average, in the most disadvantaged neighborhoods, their white counterparts tend to experience neighborhoods characterized by significantly less deprivation. Again, the same general pattern of inequality in exposure to neighborhood disadvantage emerges during all three developmental stages for black versus white respondents.

Cumulative and Timing-Specific Neighborhood Effects on Obesity

Table 3 displays coefficients from an unadjusted logistic regression model (Model 1) examining the effect of race on obesity incidence, as well as from marginal structural models using IPT and censoring weights to estimate the effects of duration (Model 2) and timing (Models 3-6) of exposure to neighborhood disadvantage. Model 1 merely quantifies the statistically significant disparity in the probability of being obese at least once between ages 18 and 30 among black versus white respondents, without considering prior neighborhood exposures or individual- and household-level covariates. This unadjusted estimate indicates that black respondents have over two times greater odds of being obese at least once in early adulthood compared to white respondents.

Table 3.

Effects of duration and timing of exposure to neighborhood disadvantage from birth through age 17 on obesity incidence in early adulthood (n=1,498)

Race-Only
 Duration
 Timing
Model 1 Model 2 Model 3 Model 4 Model 5 Model 6
Average NH disadvantage quintile
 Ages 1 through 17 1.37 ***
 Ages 1 through 5 (early childhood) 1.10 0.92
 Ages 6 through 11 (late childhood) 1.24 ** 0.98
 Ages 12 through 17 (adolescence) 1.35 *** 1.40 ***
Race (white)
 Black 2.03 *** 1.25 1.49 ** 1.34 1.21 1.24
Gender (male)
 Female 1.15 1.14 1.15 1.13 1.12
Birthweight (88 ounces or more)
 Less than 88 ounces 0.88 0.92 0.89 0.87 0.87
Mother’s marital status at birth (married)
 Unmarried 1.00 1.03 1.02 0.98 0.98
HH’s education at birth (less than high school)
 High school graduate 0.83 0.80 0.83 0.84 0.84
 At least some college 0.55 *** 0.49 *** 0.54 *** 0.55 * 0.54 ***
Mother’s age at birth 1.00 1.00 1.00 1.00 1.00
NH disadvantage at birth 0.93 0.98 0.97 0.97 1.00
HH’s marital status at birth (married)
 Unmarried 1.18 1.13 1.16 1.22 1.22
HH’s employment status at birth (employed)
 Unemployed 1.11 1.12 1.11 1.11 1.12
Homeownership at birth (owns home)
 Does not own home 0.94 0.98 0.95 0.93 0.93
Family size at birth 0.94 0.94 0.94 0.94 0.95
Public assistance receipt at birth (none)
 Received public assistance 0.51 ** 0.53 * 0.52 ** 0.49 ** 0.49 **
Household income at birth 1.03 1.00 1.02 1.05 1.05
HH’s work hours per week at birth 0.99 0.99 0.99 0.99 0.99
Year born (1970-72)
 1973-1975 1.41 * 1.39 1.39 1.44 * 1.44 *
 1976-1978 1.38 1.35 1.36 1.41 * 1.41 *
 1979-1980 1.62 ** 1.58 ** 1.58 ** 1.70 ** 1.72 **
Open in a new tab

Notes: Statistics reported for respondents not lost to follow -up before age 18 and who answered at least one set of questions about height and weight in early adulthood; Coefficients are combined estimates from 10 multiple imputation datasets; NH=Neighborhood; HH=Household head;

*

p<0.10;

**

p<0.05;

***

p<0.01

Model 2 adds a measure of cumulative exposure to neighborhood disadvantage and uses IPT and censoring weights to account for the dynamic role of time-varying individual- and household-level covariates while also controlling for time-invariant characteristics. It is worth noting the statistical non-significance of the gender variable in this and all subsequent models. Whereas prior research has often found stronger effects of childhood disadvantage on future BMI for females versus males [28, 30], much of this work has examined family-rather than neighborhood-level disadvantage and, consistent with the current findings, socioeconomic indicators at the neighborhood level more often show effects for both females and males [28]. Moreover, this second model shows that race is no longer a statistically significant predictor of obesity incidence. Rather, each unit (or quintile) increase in cumulative exposure to neighborhood disadvantage from ages one through 17 is related to a statistically significant 37% increase in the odds of obesity incidence in early adulthood. By computational extension, respondents who were persistently exposed to neighborhoods in the most disadvantaged quintile from ages one through 17 had nearly 3.5 times higher odds of being obese at least once between ages 18 and 30 compared to respondents who resided, on average, in the least disadvantaged neighborhood quintile.4,5

Models 3 through 5 examine the effects of timing of exposure to neighborhood disadvantage during early childhood, late childhood, and adolescence, respectively, without regard for the other two developmental stages. There is no statistically significant association between young adult obesity and the average neighborhood disadvantage quintile to which respondents are exposed during early childhood (Model 3), while race remains a significant (although considerably attenuated) predictor. Exposures that occur during both late childhood (Model 4) and adolescence (Model 5), however, each have independent effects on future BMI. In particular, each unit (or quintile) increase in the average level of exposure to neighborhood disadvantage during late childhood and adolescence is related to a statistically significant 24% and 35% increase, respectively, in the odds of being obese at least once between ages 18 and 30. When all three developmental stages are included in the model simultaneously (Model 6), though, only the adolescent time period from ages 12 through 17 remains statistically significant.

To further explore this relationship and the disappearance of the late childhood effect from Models 5 to 6, additional analyses using different developmental cut-points were performed. These analyses revealed two rather than three distinct stages such that experiencing neighborhood disadvantage from ages 10 through 17 has a more consequential effect on obesity incidence in young adulthood than exposure that occurs between birth and age nine.

DISCUSSION

This study examined the effects of growing up in neighborhoods characterized by varying levels of disadvantage on obesity incidence in early adulthood utilizing the 1970 to 2011 waves of the PSID merged with census data on respondents’ neighborhoods. Consistent with previous research, findings show that neighborhood disadvantage, defined by the spatial clustering of poverty, unemployment, female-headed households, public assistance receipt, and educational and occupational marginalization, is associated with higher BMI. Whereas prior studies in this area often characterize the residential environment only once or over just a short window of observation, this study employs yearly measurements of respondents’ neighborhoods from birth through age 17, as well as statistical methods that account for dynamic individual- and household-level factors known to be predictive of future BMI but also related to the sorting of families into and out of neighborhoods over time. More specifically, this study is among the first to document that more prolonged exposure to neighborhood disadvantage throughout the entire child and adolescent lifecourse increases the risk of obesity in early adulthood, and in particular, that exposure to neighborhood disadvantage during adolescence is more consequential than exposure that occurs during earlier stages of development.

As noted previously, this latter finding is consistent with the hypothesis that adolescence may be a sensitive period for neighborhood-obesity effects. Adolescence is a time of rapid growth and physiological change. It is also marked by an increasing need for autonomy and expanding social interactions. The residential neighborhood, in turn, often constitutes both the physical and social space in which youth spend a large part of their mounting out-of-home time [39]. These physical conditions and social experiences not only structure opportunities to make healthy choices but also shape norms, values, attitudes, knowledge, and behavioral tendencies related to food and exercise that are likely to carry forward into adulthood [28, 40]. More broadly, although both the duration and timing of exposure to neighborhood disadvantage affected the incidence of young adult obesity across racial groups, a more nuanced interpretation suggests that such consequences are likely to be more concentrated among black versus white young people. Black children and youth were not only more likely than their white counterparts to be born into neighborhoods characterized by higher levels of disadvantage and thus fewer social and structural resources for healthy eating and physical activity, but also more likely to remain in similar types of health-compromising residential environments for the entirety of their pre-adult years. This suggests that the separate and unequal neighborhood environments in which black versus white children and youth tend to live, learn, and grow may play a critical role in producing and perpetuating black/white disparities in obesity in adulthood.

Limitations

Although this study uses panel data and unique statistical methods to address some of the most common challenges in neighborhood effects research, the results should be considered in the context of several remaining limitations. First, given the historical timing of this study, the sample is limited to black and white respondents. Future research examining disparities in obesity as a function of cumulative and timing-specific neighborhood effects among Asian and Latino populations is encouraged. Second, analyses relied on self-reported height and weight to determine BMI. Clinical assessment would increase the objectivity associated with measurement of these characteristics and reduce the potential for bias due to differential reporting across sociodemographic groups. Third, although there appears to be sufficient overlap in the race-specific distributions of cumulative and timing-specific exposure to neighborhood disadvantage to satisfy the positivity assumption on which the analyses are based, it would be enlightening to find a sample in which neighborhood stratification by race is less pronounced in order to further disentangle the mechanisms behind the relationships among race, place, and BMI. Finally and relatedly, this study did not assess the specific mechanisms thought to help explain how neighborhood-obesity effects transpire but rather the broader neighborhood conditions thought to engender such processes. Longitudinal research able to more explicitly measure these potential mechanisms including, for example, continuity and change in supermarket and fast food outlet density, perceptions of safety, collective efficacy, and institutional resources, as well as their dynamic associations with contextual characteristics such as population and development density would be valuable.

Conclusions

Findings from the present study add to the growing body of evidence suggesting that place-based, developmentally-appropriate, and ongoing investments in the social, economic, institutional, and infrastructural aspects of under-resourced neighborhoods and neighborhoods of color can help reduce adult obesity incidence. This study further suggests that given practical and economic constraints, neighborhood-based efforts targeting the food and physical activity structures and behaviors of adolescents (rather than younger children) may be particularly successful. Whereas prior neighborhood effects research based on cross-sectional data may have concluded with similar general sentiments, the implementation of associated health practice and policy changes has often been hampered by challenges to causal inference. Using longitudinal data and a statistical method that attempts to model the full data distribution, this study provides among the strongest evidence to date for the consequences of both sustained and adolescent exposures to neighborhood disadvantage on future (young adult) obesity, especially among black children and youth for whom such exposures tend to be more common and more persistent.

Supplementary Material

1

Acknowledgements

This research was supported in part by a Shanahan Endowment Fellowship and a Eunice Kennedy Shriver National Institute of Child Health and Human Development training grant, T32 HD007543, to the Center for Studies in Demography and Ecology at the University of Washington. The data used in this study was collected with the support of the National Institutes of Health under grant number R01 HD069609 and the National Science Foundation under award number 1157698. The author thanks the members of her dissertation committee and three anonymous reviewers for helpful comments on earlier versions of the manuscript.

Footnotes

Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

Implications and Contribution

Experiencing neighborhood adversity throughout childhood and adolescence, and in particular from ages 10 to 17, increases obesity risk in early adulthood. Ongoing and developmentally-appropriate investments in the social, economic, institutional, and physical structures of under-resourced communities and communities of color can have long-term benefits for population health and health equity.

1

A combined measure of overweight/obesity (BMI>=25 kg/m2) was also examined with similar results, although statistical significance was often attenuated. Results examining BMI as a continuous measure (average BMI between ages 18-30) were consistent with those presented for obesity.

2

The NCDB ensures that census data for all four decades has been normalized to 2000 tract boundaries and can therefore be compared across years without having to adjust for potential changes in boundary definitions over time. Data for intercensal years are imputed using linear interpolation. If data in 1970 and/or 1980 was missing (likely due to as yet untracted areas) and linear interpolation was impossible, data from 1980, 1990, and/or 2000 was used to extrapolate values for up to five years prior to the most recent value (e.g., if data in 1970 was missing, the linear interpolation from 1980 to 1990 was extended to 1975).

3

In practice, IPT weights can be highly variable. To increase efficiency and obtain narrower confidence intervals around the subsequent neighborhood effect estimate, the IPT weights for each respondent at each year are stabilized by multiplying each one by the same probability as was used to generate it, except only baseline covariates and prior year neighborhood conditions are included as regressors (i.e., time-varying covariates are excluded).

4

Exp(5-1)×0.31=3.46, where 0.31 is the log odds for the neighborhood disadvantage variable.

5

These estimates exceed those produced using more conventional regression methods that condition on both time-invariant covariates as well as all time-varying covariates averaged over ages one to 17 (Appendix A), suggesting that prior research based on such techniques may over-control the indirect (or mediated) pathways through which prolonged neighborhood exposure impacts future BMI and black/white disparities therein.

REFERENCES

  • 1.U.S. Department of Health and Human Services . The Surgeon General’s call to action to prevent and decrease overweight and obesity. U.S. Department of Health and Human Services, Public Health Service, Office of the Surgeon General; Rockville, MD: 2001. [PubMed] [Google Scholar]
  • 2.Caballero B. Obesity prevention in children: opportunities and challenges. Int J Obes Relat Metab Disord. 2004;28(S3):S90–S95. doi: 10.1038/sj.ijo.0802797. [DOI] [PubMed] [Google Scholar]
  • 3.Connelly JB, Duaso MJ, Butler G. >A systematic review of controlled trials of interventions to prevent childhood obesity and overweight: a realistic synthesis of the evidence. Public Health. 2007;121(7):510–7. doi: 10.1016/j.puhe.2006.11.015. [DOI] [PubMed] [Google Scholar]
  • 4.Gordon-Larsen P, et al. Inequality in the Built Environment Underlies Key Health Disparities in Physical Activity and Obesity. Pediatrics. 2006;117(2):417–24. doi: 10.1542/peds.2005-0058. [DOI] [PubMed] [Google Scholar]
  • 5.Morland K, et al. Neighborhood Characteristics Associated with the Location of Food Stores and Food Service Places. American Journal of Preventive Medicine. 2002;22(1):23–30. doi: 10.1016/s0749-3797(01)00403-2. [DOI] [PubMed] [Google Scholar]
  • 6.Morland K, Roux A.V. Diez, Wing S. Supermarkets, other food stores, and obesity - The atherosclerosis risk in communities study. American journal of preventive medicine. 2006;30(4):333–339. doi: 10.1016/j.amepre.2005.11.003. [DOI] [PubMed] [Google Scholar]
  • 7.Morland KB, Evenson KR. >Obesity prevalence and the local food environment. Health & Place. 2009;15(2):491–495. doi: 10.1016/j.healthplace.2008.09.004. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 8.Papas MA, et al. The Built Environment and Obesity. Epidemiologic reviews. 2007;29(1):129–143. doi: 10.1093/epirev/mxm009. [DOI] [PubMed] [Google Scholar]
  • 9.Boardman JD, et al. Race Differentials in Obesity: The Impact of Place. Journal of Health and Social Behavior. 2005;46(3):229–243. doi: 10.1177/002214650504600302. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 10.Lovasi GS, et al. Built environments and obesity in disadvantaged populations. Epidemiologic reviews. 2009;31(1):7–20. doi: 10.1093/epirev/mxp005. [DOI] [PubMed] [Google Scholar]
  • 11.Larson NI, Story MT, Nelson MC. Neighborhood environments: disparities in access to healthy foods in the US. American journal of preventive medicine. 2009;36(1):74–81. doi: 10.1016/j.amepre.2008.09.025. e10. [DOI] [PubMed] [Google Scholar]
  • 12.Timberlake JM. Racial and Ethnic Inequality in the Duration of Children's Exposure to Neighborhood Poverty and Affluence. Social Problems. 2007;54(3):319–342. [Google Scholar]
  • 13.Logan JR. Separate and unequal: The neighborhood gap for blacks, Hispanics and Asians in Metropolitan America. Project US2010 Report. 2011:1–22. [Google Scholar]
  • 14.Quillian L. How Long are Exposures to Poor Neighborhoods? The Long-Term Dynamics of Entry and Exit from Poor Neighborhoods. Population Research and Policy Review. 2003;22(3):221–249. [Google Scholar]
  • 15.Sharkey P. Stuck in place: Urban neighborhoods and the end of progress toward racial equality. University of Chicago Press; Chicago: 2013. [Google Scholar]
  • 16.Ratitle Peterson, Krivo LJ. American Sociological Association's Rose Series in Sociology. Russell Sage Foundation; New York: 2010. Divergent Social Worlds: Neighborhood Crime and the Racial-Spatial Divide. [Google Scholar]
  • 17.Boone-Heinonen J, Gordon-Larsen P. Obesogenic Environments in Youth: Concepts and Methods from a Longitudinal National Sample. American journal of preventive medicine. 2012;42(5):e37–e46. doi: 10.1016/j.amepre.2012.02.005. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 18.Dietz WH. Critical periods in childhood for the development of obesity. The American Journal of Clinical Nutrition. 1994;59(5):955–9. doi: 10.1093/ajcn/59.5.955. [DOI] [PubMed] [Google Scholar]
  • 19.Adair LS. Child and adolescent obesity: Epidemiology and developmental perspectives. Physiology & Behavior. 2008;94(1):8–16. doi: 10.1016/j.physbeh.2007.11.016. [DOI] [PubMed] [Google Scholar]
  • 20.Lee JM, et al. Getting heavier, younger: trajectories of obesity over the life course. Int J Obes. 2010;34(4):614–623. doi: 10.1038/ijo.2009.235. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 21.Burdette AM, Needham BL. Neighborhood Environment and Body Mass Index Trajectories From Adolescence to Adulthood. Journal of Adolescent Health. 2012;50(1):30–37. doi: 10.1016/j.jadohealth.2011.03.009. [DOI] [PubMed] [Google Scholar]
  • 22.Aber JL, et al. Development in context: implications for studying neighborhood effects. In: Brooks-Gunn J, Duncan GJ, Aber JL, editors. Neighborhood Poverty: Context and Consequences for Children. Russell Sage Foundation; New York: 1997. [Google Scholar]
  • 23.Auchincloss AH, et al. Neighborhood health-promoting resources and obesity risk (the multi-ethnic study of atherosclerosis) Obesity. 2013;21(3):621–628. doi: 10.1038/oby.2012.91. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 24.Wodtke GT, Harding DJ, Elwert F. Neighborhood Effects in Temporal Perspective: The Impact of Long-Term Exposure to Concentrated Disadvantage on High School Graduation. American Sociological Review. 2011;76(5):713–736. doi: 10.1177/0003122411420816. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 25.Sampson RJ, Morenoff JD, Gannon-Rowley T. Assessing "Neighborhood Effects": Social Processes and New Directions in Research. Annual Review of Sociology. 2002;28:443–478. [Google Scholar]
  • 26.Elder GH Jr, Giele JZ, editors. The Craft of Life Course Research. The Guilford Press; New York and London: 2009. [Google Scholar]
  • 27.Nau C, et al. Community socio economic deprivation and obesity trajectories in children using electronic health records. Obesity. 2014 doi: 10.1002/oby.20903. 00(00) [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 28.Lee H, Harris KM, Gordon-Larsen P. Life Course Perspectives on the Links Between Poverty and Obesity During the Transition to Young Adulthood. Population Research and Policy Review. 2009;28(4):505–532. doi: 10.1007/s11113-008-9115-4. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 29.McTigue KM, et al. Urban Neighborhood Features and Longitudinal Weight Development in Girls. American Journal of Preventive Medicine. 2015;49(6):902–911. doi: 10.1016/j.amepre.2015.05.021. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 30.Hernandez DC, Pressler E. Accumulation of childhood poverty on young adult overweight or obese status: race/ethnicity and gender disparities. Journal of Epidemiology and Community Health. 2014;68(5):478–484. doi: 10.1136/jech-2013-203062. [DOI] [PubMed] [Google Scholar]
  • 31.Brener ND, et al. Reliability and validity of self-reported height and weight among high school students. Journal of Adolescent Health. 2003;32(4):281–287. doi: 10.1016/s1054-139x(02)00708-5. [DOI] [PubMed] [Google Scholar]
  • 32.Strauss RS. Comparison of measured and self-reported weight and height in a cross-sectional sample of young adolescents. International Journal of Obesity. 1999;23(8):904–908. doi: 10.1038/sj.ijo.0800971. [DOI] [PubMed] [Google Scholar]
  • 33.Matthews SA. Spatial polygamy and the heterogeneity of place: studying people and place via egocentric methods. In: Burton LM, et al., editors. Communities, Neighborhoods, and Health: Expanding the Boundaries of Place. Springer; New York: 2011. [Google Scholar]
  • 34.Diez Roux AV, Mair C. Neighborhoods and health. Annals of the New York Academy of Sciences. 2010;1186(1):125–145. doi: 10.1111/j.1749-6632.2009.05333.x. [DOI] [PubMed] [Google Scholar]
  • 35.Crowder K, Downey LC. Inter-neighborhood migration, race, and environmental hazards: modeling micro-level processes of environmental inequality. American Journal of Sociology. 2010;115(4):1110–1149. doi: 10.1086/649576. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 36.Geolytics Neighborhood Change Database, 1970-2010. 2012 Available from: http://www.geolytics.com/USCensus,Neighborhood-Change-Database-1970-2000,Products.asp.
  • 37.Robins JM, Hernan MA, Brumback B. Marginal Structural Models and Causal Inference in Epidemiology. Epidemiology. 2000;11(5) doi: 10.1097/00001648-200009000-00011. [DOI] [PubMed] [Google Scholar]
  • 38.Welch C, Bartlett J, Petersen I. Application of multiple imputation using the two-fold fully conditional specification algorithm in longitudinal clinical data. The Stata Journal. 2013 [PMC free article] [PubMed] [Google Scholar]
  • 39.Leventhal T, Dupere V, Brooks-Gunn J. Neighborhood influences on adolescent development. In: Lerner RM, Steinberg L, editors. Handbook of adolescent psychology. John Wiley & Sons, Inc.; 2009. [Google Scholar]
  • 40.McLeod JD, Almazan EP. Connections between Childhood and Adulthood. In: Mortimer, Shanahan, editors. Handbook of the Life Course. Springer; 2004. [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

1
NIHMS756647-supplement-1.docx (109.9KB, docx)

RESOURCES