Crime,Perceived Danger,and Adiposity: The Role of Gender

Crime, Perceived Safety, and Adiposity

“Aging in place” is a critical concept for healthy aging as the immediate environments can provide certain kinds of satisfaction and support—and yet a distressing, disordered environment can also impose stress and high risks of illness among older adults. Several theories in social gerontological work, including the Person–Environment Congruence Model (Kahana, 1982) and the Disability Model (Lawton, 1983), emphasize the effects of the environment in determining the health and well-being of older adults. Building on these theories, numerous studies have indicated that older adults—who tend to remain in the same residential areas as they age—become increasingly vulnerable to environmental stress because they have accumulated a longer duration exposure than young adults to unhealthy features of their environment, and because the changes in biological and psychological functioning associated with normal aging may reduce older adults’ competency to combat stress (Glass & Balfour, 2003).

Among other sources of environmental stressors, objective measures of a dangerous neighborhood such as police-reported crime statistics have been considered important characteristics by which to understand how bad place matters for health in later life. For instance, evidence based on social disorganization theory suggests that neighborhoods with high levels of crime affect later-life health through their impact on worries and danger about the neighborhoods (Browning, Cagney, & Iveniuk, 2012; Kawachi & Berkman, 2003). Research on the built environment indicates that high-crime neighborhoods are more likely to be physically disordered, such that residents are presented with visible evidence of the breakdown of physical infrastructure in the local environment (e.g., run-down and abandoned buildings, vandalized physical structures, unkempt and dilapidated properties; LaGrange, Ferraro, & Supancic, 1992). Exposure to such disorderly built environments may be especially stressful to residents because weakened social order and a lack of social control exacerbate feelings of threat and danger (Burdette & Hill, 2008).

The concentration of neighborhood disorder and decay has also been associated with residents’ reduced access to healthy food options and with a lower likelihood of healthy behaviors such as exercise. For example, Larson and colleagues (2009) suggested that individuals living in low-income neighborhoods have limited access to healthy food options due to a high prevalence of “food deserts” clustered in their neighborhoods, potentially creating obesogenic environments. Lovasi, Hutson, Guerra, and Neckerman (2009) indicated that poorly maintained sidewalks and lacking public parks that are commonly featured in physically disordered neighborhoods may also limit residents’ opportunities for physical activity. Residents living in disadvantaged neighborhoods are more likely to engage in other unhealthy behaviors, such as smoking and drinking, possibly as a means of coping with stress (Diez Roux & Mair, 2010). Physical inactivity, smoking, and heavy drinking have been identified as behavioral risks that have adverse health consequences for older adults including adiposity (Gordon-Larsen, Nelson, Page, & Popkin, 2006; Kruger, Ham, & Prohaska, 2008).

Although overlooked in the literature of “Aging in Place,” a growing body of research has demonstrated that subjective feelings about a neighborhood can also contribute to health. This prior work posits that objective conditions of neighborhood disorder may affect health outcomes through their impact on subjective interpretations of neighborhood safety, which may increase fear and vulnerability among residents. For example, recent evidence has found that, relative to objective indicators of neighborhood danger, perceptions of neighborhood safety are more directly associated with depression (Latkin & Curry, 2003; Wilson-Genderson & Pruchno, 2013) and cognition (Lee & Waite, 2018). However, the current literature in the context of adiposity has used either subjective measures of neighborhood safety or objective measures of neighborhood danger interchangeably without distinguishing their independent pathways, direct and indirect, to disease risk (Eisenstein et al., 2011; Grafova et al., 2008), or has collapsed them into one measure which could potentially mask their differential impacts on health (Glass, Rasmussen, & Schwartz, 2006; Powell-Wiley et al., 2015). Because high-crime activity in the local area may be a mechanism that results in how people perceive the safety of their neighborhoods and surroundings, it is important to investigate the direct (objective) and potential indirect effects of crime on health through subjective interpretations.

Gender Differences in the Association Between Environmental Stressors and Adiposity

Men and women may respond differently to the stress of living in a dangerous and distressing community. First, gender differences in biobehavioral responses to stress may shed light on gender differences in health related to neighborhood crime. According to Taylor (2011; Taylor et al., 2000), when women are in stressful conditions, they may be more likely than men to adopt a “tend and befriend” strategy in which they manage stressful conditions by creating, maintaining, and relying on social groups for comfort and safety. Threatening circumstances trigger an increase in oxytocin, a hormone that can motivate affiliative behaviors which, with supportive others, can reduce potentially damaging biological stress responses including those that arise from elevated glucocorticoid levels (Taylor, 2011). However, if affiliation occurs with antagonistic and unsupportive others, oxytocin may worsen physiological stress responses (Stafford, Chandola, & Marmot, 2007). Men, however, typically follow the “fight-or-flight” pattern (Taylor et al., 2000), a behavioral strategy which is supported by the release of adrenaline into the blood stream to increase blood pressure and heart rate. The fight-or-flight response also results in high levels of glucocorticoids, which have beneficial effects over the short term (e.g., mobilization of energy reserves) but have cumulative adverse effects on cognitive and physical health over the long term. Second, men and women may use different coping strategies when exposed to stressful circumstances. Evidence suggests that men engage in unhealthy behaviors such as cigarette smoking (Todd, 2004) and binge drinking (Cooper, Russell, Skinner, Frone, & Mudar, 1992) in response to stress, the result of which may be weight loss. Women, on the contrary, are more likely than men to cope with stress by eating (Jackson & Knight, 2006). An empirical study by Reagan and Hersch (2005) showed that, for women, the frequency of binge eating increases with depressive symptoms, and with spending a large fraction of daily life in disadvantaged neighborhoods.

These findings on gender differences in response to stress have led researchers to speculate as to distressing neighborhood conditions and gender differences in adiposity risk. In a community-based sample of older adults, for example, Glass et al. (2006) found that a composite variable of violent crime and other neighborhood disadvantage measures was significantly associated with a higher BMI. With an adult sample from the Dallas Heart Study, Powell-Wiley et al. (2015) found an association between moving to more deprived neighborhoods (e.g., indexed by neighborhood-level violence, social cohesion and physical environment) and weight gain. Despite their attempts to expand our understanding about how dangerous neighborhoods shape BMI in later life, neither study examined how these associations differ by gender. Only a couple of studies have focused on gender differences in older age and have found that neighborhood characteristics are more strongly associated with BMI for women than men (Christman et al., 2016; Powell-Wiley et al., 2017). Most of the foregoing studies are based on localized samples (e.g., from single cities), and, as pointed out earlier, fail to differentiate between objective measures of crime activity and subjective interpretations of neighborhood safety. Violent crime has been the predominant measure of crime in past research, although evidence suggests that violent crime tends to be localized and occurs only rarely compared with other types of crime such as burglary. In fact, burglary is more prevalent than most other types of crime (e.g., second only to larceny; see Lauritsen & Rezey, 2013). In contrast to violent crime, burglary is unpredictable in location and timing. The unpredictable nature of burglary and possibilities of home violation are especially disruptive to feelings of safety and security, which may then magnify a sense of fear and vulnerability (Coakley & Woodford-Williams, 1979). This is particularly true for older adults for whom the home becomes increasingly important as mobility and functioning decreases. However, how burglary is associated with later-life health has been largely understudied.

Furthermore, none of these studies considered waist circumference and its association with crime (except for Powell-Wiley et al., 2017). According to Epel et al. (2000), excessive abdominal fat is a greater health risk than is whole-body fat. Abdominal fat serves as an indicator of prolonged exposure to high levels of glucocorticoids that may result from chronic stress exposure (McEwen, 1998, 2003). Although the findings above consider neighborhood context as a risk factor for adiposity, additional research employing both BMI and waist circumference is needed; the significance of neighborhood context on adiposity-related outcomes for older adults will be better understood with a more nuanced approach to adiposity status.

Sample

The National Social Life, Health and Aging Project (NSHAP W2) is a population-based study of older community-dwelling Americans. The NSHAP sample is selected using a multistage area probability design, and data collection consists of an in-person interview that includes biomeasure collection, and a leave-behind questionnaire that is returned by postage-paid mail (O’Muircheartaigh, English, Pedlow, & Kwok, 2014). The full sample of respondents available in NSHAP W2 was 2,261. Given that crime is concentrated in densely populated urban areas, we limited our analytic sample to urban-dwelling respondents (N = 1,708; 75.81%). We identified urban areas as metropolitan areas of more than 50,000 residents based on the 2010 Rural-Urban Commuting Area (RUCA) Codes 1 through 3 (United States Department of Agriculture, 2016). The analytic sample consisted of 823 males (48.2% of sample) and 885 females.

Adiposity Measures

Two measures of adiposity were investigated as dependent variables: waist circumference and BMI. Waist circumference was measured in inches and converted to centimeters. Height and weight were measured by field interviewers, and BMI was derived by dividing weight in kilograms by height in meters squared. Cases with extreme (unrealistic) BMI and waist values that suggested errors in measurement and/or transcription were dropped from analyses (n = 46, or 2.7% of cases). Of the remaining values, statistical outliers (i.e., greater than 3 SDs from the mean) were treated as missing (n = 27, or 1.6% of 1,662 retained cases). Sensitivity analyses truncating outlier values to 3 SDs from the mean did not substantively change our findings, so we elected to leave these values coded as missing.

In addition, waist circumference and BMI were dichotomized to generate measures of abdominal and general obesity, respectively. Abdominal adiposity was defined as a waist circumference greater than 102 cm in men and greater than 88 cm in women (National Heart, Lung, and Blood Institute, 1998). BMI was used to classify respondents as obese (BMI > 29.9) based on guidelines from Centers for Disease Control and Prevention (2016).

Police-Reported Crime and Perceived Neighborhood Context

County-level geocoded crime data were derived from the Federal Bureau of Investigation (FBI) Uniform Crime Report (UCR) and purchased from Applied Geographic Solutions (AGS). These data included rates of murder, rape, robbery, assault, burglary, larceny, and motor vehicle theft, standardized to represent county-level rates as a proportion of the national rate (e.g., burglary index = [number of burglary incidents at counties × 100] / number of national burglary incidents). This burglary index was right-skewed, so we divided burglary rates into tertiles (e.g., low, medium, and high) for analytic purposes. We chose to model burglary at the county-level because we believe that (a) informed by recent research on “activity space,” older adults may be exposed to a much broader range of space than only their residential location (Cagney, Browning, Jackson, & Soller, 2013) and (b) disorder has a spillover effect across adjacent neighborhoods (Sampson, 2012; Skogan, 1992). In assessing perceived neighborhood context, we constructed scales for perceived neighborhood danger (α = .80) and distrust (α = .67) from the NSHAP respondents’ report on their local areas. In NSHAP, local areas were defined as everywhere within a 20-min walk or within about a mile of respondent’s home. For a scale of perceived neighborhood danger, respondents were asked to indicate, using a 5-point Likert-type scale (from strongly disagree coded as 1 to strongly agree coded as 5), how afraid they are to go out at night, how much trouble is expected, and whether they are taking a big chance walking alone after dark in local areas. We considered the items used in the scale of perceived neighborhood danger to capture the degree of feelings of fear of individuals’ reactions to community crime. Perceived neighborhood distrust was defined by items that assess whether neighbors can be trusted, get along, share values, and help each other, and whether the neighborhood is close-knit, using the same 5-point scale as was used for perceived neighborhood danger. We then reverse-coded the scale so that higher values reflect greater neighborhood distrust and lack of security. The correlation matrix indicates that objective and subjective neighborhood measures were positively correlated with each other, with perceived danger exhibiting a larger correlation with burglary than neighborhood distrust (Supplemental Table 1). Validation of the measurement of neighborhood perceptions has been published elsewhere (Cornwell & Cagney, 2014).

Health Behaviors

Previous research identified physical inactivity, cigarette smoking, and heavy drinking as risk behaviors that have adverse impacts on older adults’ weight status (Kruger et al., 2008; Watson et al., 2016). Smoking status was assessed as (a) never (reference group), (b) ever, and (c) currently smoking. To measure daily alcohol consumption (number of drinks per day), we grouped respondents into 3 categories: (a) none (reference group), (b) moderate (1 to 3 drinks/day), and (c) heavy (+3 drinks/day). Evidence indicates that in older adults, physical inactivity (sedentariness) is associated with excess weight (Gordon-Larsen et al., 2006), and that replacing sedentary time with even light intensity physical activity has been associated with lower BMI (Bann et al., 2015). We, therefore, compared respondents who reported no activity (coded = 1) with those who reported doing any activity (reference group).

Covariates

Baseline demographic characteristics included age, race/ethnicity, living arrangement, and education attainment. The range of respondents’ age was 62 to 90 years. We grouped respondents into three age brackets: (a) 62-70, (b) 71-80, and (c) 81-90, to capture any differences across subgroups at older ages (i.e., young old, middle old, and oldest old). Race/ethnicity was self-reported as (a) Non-Hispanic White, (b) Non-Hispanic African American, (c) Hispanic, and (d) Other. Because we believe that living alone may decrease feelings of security and safety, and increase fear of crime and victimization compared with living with someone, we controlled for living arrangement (living alone = 1; otherwise = 0). Educational attainment was classified into four levels: (a) less than high school, (b) high school or equivalent, (c) some college, and (d) college or more. Previous victimization experiences are highly correlated with fear of crime (Lauritsen & Rezey, 2013), and we therefore included prior victimization (i.e., having been a victim of a violent crime in the past 2 years) as a covariate. Chronic disease burden was also included as a covariate because it may be related to adiposity. The NSHAP comorbidity index includes 15 prevalent health conditions in older adults. Detailed information on the use of chronic condition data to calculate a comorbidity index has been published elsewhere (Vasilopoulos et al., 2014). Neighborhood disadvantage and physical disorder measures were included as covariates because they have been documented to be related to both crime and adiposity. We constructed the disadvantage measure by standardizing and summing the following items from 2006-2009 American Community Survey: (a) percent households living in poverty, (b) percent households receiving public assistance, (c) percent adults (⩾25 years olds) with a high school diploma or less, and (d) percent unemployed (⩾16 years olds; α = .63). We measured physical disorder using field interviewers’ responses to two questions about the respondent’s immediate built environment: “How well kept is the building in which the respondent lives?” and “How well kept are most of the buildings on the street (1 block, both sides) where the respondent lives?” Responses ranged from very well kept (= 1) to very poorly kept (= 4; needs major repairs). Values were averaged across the two items (α = .83), with higher scores indicating greater physical disorder.

Analytic Strategy

Because we expected that multiple neighborhood social contexts produce the ramifications of stress, including older adults’ adiposity, we fit sequential models, beginning by examining adiposity as a function of objective crime, namely, burglary (Model 1 in Table 2). In Model 2, we added neighborhood perceptions—neighborhood danger and distrust—to assess the extent to which neighborhood perceptions explain the association between burglary and adiposity measurements. In Model 3, we added health behavior variables to test their explanatory capacity. Waist circumference and BMI are continuous variables, so we fit ordinary least squares (OLS) linear regression models. In the case of dichotomous variables for abdominal and general adiposity, multivariate logit regression models were employed. Because we expected that responses to environmental stressors may differ by gender, all analyses were completed separately for men and women. Because NSHAP respondents are spread out across the United States rather than clustered within counties, multilevel modeling with individuals nested within counties is not feasible. In the current study, descriptive statistics and all analytical models were adjusted for survey design (STATA 14). This procedure corrected for multiple stages of cluster sample design and unequal probability of selection to ensure nationally representative results with unbiased standard errors.

Descriptive Statistics

As shown in Table 1, men, relative to women, had a larger waist (105.5 vs. 94.3) and a higher BMI (29.1 vs. 28.7), and were more likely to be obese (39.3% vs. 36.4%). Women were more likely than men to have abdominal adiposity (64.1% vs. 52.5%). In addition, men were more likely to live in deleterious, distressing neighborhood contexts characterized by higher levels of burglary and perceived neighborhood distrust. Despite a lower exposure to burglary, women were more likely to perceive neighborhoods as dangerous. Notably, men were more likely to engage in unhealthy behaviors. For example, 14.5% and 7.8% of men reported current use of cigarettes and heavy drinking, compared with women (10.8% and 1.4%, respectively). However, approximately 10% more women reported that they do not exercise.

OPEN IN VIEWER

Table 1.

Descriptive Statistics for Key Variables, Weighted and Adjusted for Survey Design.

Variable name	Men (n = 823)		Women (n = 885)
	Percentage (n a )/M	SD	Percentage (n)/M	SD
Outcome variables
Waist circumference (cm)	105.50	14.16	94.34	15.13
BMI	29.09	5.20	28.74	6.04
Abdominal adiposity	52.50% (432)		64.06% (567)
General adiposity	39.29% (323)		36.42% (332)
Crime index
Burglary index	108.62	50.78	107.93	50.83
Neighborhood characteristics
Disadvantage index	0.16	0.08	0.16	0.07
Physical disorder index	0.52	0.68	0.57	0.67
Perceived danger index	2.34	0.91	2.46	0.83
Perceived distrust index	2.57	0.52	2.50	0.54
Individual characteristics
Age
62-70	52.54% (432)		44.40% (393)
71-80	31.47% (259)		34.51% (305)
80 and up	15.99% (132)		21.09% (187)
Race/ethnicity
White, non-Hispanic	76.95% (633)		77.57% (686)
Black, non-Hispanic	10.91% (90)		12.44% (110)
Hispanic	8.99% (74)		8.04% (71)
Other	3.15% (26)		1.95% (18)
Living alone	27.39% (225)		38.69% (342)
Education
<High school	14.02% (115)		17.79% (157)
High school	20.76% (171)		25.74% (228)
Some college	28.06% (231)		34.37% (304)
BA or more	37.16% (306)		22.10% (196)
Prior victimization	4.19% (35)		3.27% (29)
NSHAP comorbidity index	2.21	1.42	2.35	1.26
Health characteristics
Smoking
Never	24.96% (205)		49.96% (442)
Ever	60.50% (498)		39.26% (347)
Currently smoking	14.54% (120)		10.78% (96)
Drinking
None	33.41% (275)		46.77% (414)
Moderate	58.81% (484)		51.83% (459)
Heavy	7.78% (64)		1.40% (12)
No exercise	17.80% (146)		28.73% (254)

Note. BMI = body mass index; NSHAP = National Social Life, Health and Aging Project.

a

Values presented in parentheses are raw numbers while we obtain percentages adjusted for survey design.

Multivariate Models

Is the association between burglary and adiposity greater in women than men?

Results in Table 2 showed that burglary is positively associated with waist circumference for women, not for men. As shown in Model 1s, living in the highest burglary areas increased waist circumference by 4.55 cm (p < .05) among women controlling for individual sociodemographic factors. Similarly, we found gender differences in the association between burglary and BMI. Women residing in the highest burglary communities were more likely to have a higher BMI, but again, not so for men (β = 1.42, p < .05). Results were strikingly different for men. Among men, Model 1s of Table 2 indicates a negative relationship between living in a high-burglary neighborhood and waist circumference and BMI. However, this relationship was not statistically significant throughout the models.

OPEN IN VIEWER

Table 2.

Multivariate Regression Predicting Gender Differences in the Association Between Crime and Waist Circumference/BMI.

	Waist circumference						BMI
	Men			Women			Men			Women
	Model 1	Model 2	Model 3	Model 1	Model 2	Model 3	Model 1	Model 2	Model 3	Model 1	Model 2	Model 3
Objective neighborhood disorder
Crime (burglary)
Low (= ref.)
Medium	−0.15	−0.66	−0.31	1.27	1.02	0.59	−0.17	−0.20	−0.03	0.62	0.75	0.44
High	−0.15	−0.66	−0.31	4.55*	4.44*	4.05*	−0.31	−0.35	−0.36	1.42*	1.29*	1.07
Subjective neighborhood disorder
Perceived danger		0.94	1.10		1.49*	1.56*		−0.04	−0.01		1.18***	1.14***
Perceived distrust		0.69	0.88		−0.81	−1.53		0.37	0.35		−0.25	−0.57
Health behaviors
Smoking
Never (= ref.)
Ever			0.61			0.10			−0.47			−0.34
Currently smoking			−3.17			−8.01***			−2.05*			−3.12**
Drinking
None (= ref.)
Moderate			−0.15			−0.59			−0.08			−0.71
Heavy			−6.27*			−3.29			−0.20			−3.16*
No exercise			−0.21			4.90**			0.12			2.28***
Covariates a	X	X	X	X	X	X	X	X	X	X	X	X

Note. BMI = body mass index; NSHAP = National Social Life, Health and Aging Project.

a

All models are adjusted for covariates including age, race/ethnicity, living arrangement, educational attainment, prior victimization, neighborhood socioeconomic status, physical disorder, and NSHAP comorbidity index.

*

p < .05. **p < .01. ***p < .001.

To examine whether burglary increases risk for exceeding adiposity thresholds, we fitted the logistic regressions with abdominal and general adiposity as the outcomes (Supplemental Table 2). Consistent with the results from Table 2, we found that living in neighborhoods with high levels of burglary were associated with adiposity only for women. As seen in Figure 2, for women, living in neighborhoods with high-burglary rates increased their odds of abdominal adiposity (odds ratio [OR] = 1.87, p < .01) and general adiposity, although marginally significant (OR = 1.52, p < .10; results from final Model 3s). We did not see the same pattern in men. To formally evaluate whether the association between burglary and adiposity differed between men and women, we used “seemingly unrelated estimation” and the Paternoster test (Paternoster, Brame, Mazerolle, & Piquero, 1998). Results confirmed that the regression coefficients for burglary coefficients (both Table 2 and Supplemental Table 2) differed significantly between men and women.

OPEN IN VIEWER

Figure 2.

Odds Ratios Predicting Gender Differences in the Association between Burglary and Abdominal/General Adiposity.