Exposure to Local Violent Crime and Childhood Obesity and Fitness: Evidence from New York City Public School Students

Abstract

This paper estimates the relationship between neighborhood violent crime and child and adolescent weight and fitness. It uses detailed data from the Fitnessgram assessments of public school students in New York City matched to point specific crime data geocoded to students’ residential location. Our empirical approach compares the weight and fitness outcomes of students exposed to a violent crime on their residential H-block with those living in the same census tract but not exposed to violent crime in close proximity to their home. We find for adolescent girls, increases in BMI that range from 0.01 to 0.035 standard deviations and an increase in the probability of overweight of 0.5 to 1.7 percentage points. We find little evidence that BMI, obesity, and overweight change as a result of violent crime for adolescent boys, and younger children. Results are not explained by declines in physical fitness.

1. Introduction

Neighborhoods are an important social determinant of health. Neighborhood crime, in particular, has been associated with worse chronic health outcomes among adults and children including worse self-reported and cardiovascular health, decreased mental health, increased difficulty sleeping, and increased child asthma (Bencsik, 2018; Browning et al., 2012; Doyle et al., 2006; Dustmann and Fasani, 2016; Hessel et al., 2019; Hill et al., 2016; Merrill et al., 2021). Further, neighborhood crime also negatively affects birth outcomes (Currie et al., 2022; Goin et al., 2019; Grossman and Khalil, 2022) and the well-being of adults (Cornaglia et al., 2014). Exposure to community violence is a reality for many children in the nation’s urban centers. According to the FBI, in 2020, cities across the country have experienced an increase in violence¹. In 2020, New York City (NYC) – the focus of the present study – experienced its largest year-over-year increase in homicides since the 1970s (Mangual, 2021). Even before this recent rise in violent crime, some children had been living in areas with recurring violence. In 2016, over 77,000 public school students lived within a very short distance of more than ten violent crimes (homicides, aggravated assaults, and robberies).

Existing research provides strong evidence that violence negatively affects students’ academic outcomes (Bowen and Bowen, 1999; Schwartz 2022; Sharkey, 2010), socioemotional development (McCoy et al., 2016; Ramey and Harrington, 2019), and stress (Theall et al., 2017). Exposure to neighborhood crime may also affect child health, including body mass index (BMI) and the probability of obesity and overweight. Indeed, prior research shows that a child’s environment plays an important role in shaping their weight and fitness (Bader et al., 2013; Carroll-Scott et al., 2013; Morgan Hughey et al., 2017) and violent crime may moderate that relationship. Specifically, neighborhood crime may hinder the enjoyment of neighborhood amenities such as parks or playgrounds, limit the amount or frequency of time spent outside being active, and increase stress with detrimental consequences for child health and weight (Sandy et al., 2013).

Identifying whether neighborhood violent crime is linked with increased BMI or a higher risk of obesity and overweight among children and adolescents is important. According to the CDC, in 2017 and 2018, the prevalence of childhood obesity was 19.3 percent, and the prevalence was higher among adolescents aged 12 to 19 years old (21.2 percent) (Ogden et al., 2020). Childhood obesity is a significant public health problem as it is linked to negative physical and emotional health (An et al., 2017). Thus, understanding the role that neighborhoods play in worsening (or not) this problem is important. In this paper, we investigate this issue by estimating the effect of local neighborhood violent crime on child and adolescent weight and physical fitness.

Existing research that examines the relationship between crime and weight outcomes (BMI, obesity, overweight) yields mixed findings. While some studies show that neighborhood crime is associated with higher weight (Borrell et al., 2016; Daniels et al., 2021; Miranda et al., 2012; Theall et al., 2019), others find no relationship even when there is a negative association with physical activity (Burdette and Whitaker, 2005, 2004). In a recent review of the literature, Yu and Lippert (2016) echo these mixed findings. The literature that focuses on the relationship between neighborhood crime and physical activity is also mixed. Some studies find that neighborhood crime is negatively correlated with physical activity (Molnar et al., 2004; Yu and Lippert, 2016), while others find no association (Lovasi et al. 2011), and a third group of papers finds a positive correlation (Jago et al., 2006; Robinson et al., 2016). While valuable, existing studies often rely on small samples that may be underpowered to detect statistically significant changes and are largely cross-sectional. Further, prior research often uses self-reported measures of weight, physical activity, and exposure to neighborhood crime. The few papers that use objective crime measures often rely on large area-level crime rates that may imperfectly capture exposure to violence at the micro-neighborhood level. Finally, existing evidence is primarily descriptive – capturing associations between crime and weight outcomes (or physical activity) – leaving open myriad potential underlying causal mechanisms.

Our study overcomes many of these limitations. First, we use objective measures of weight and physical fitness obtained from the NYC Fitnessgram, an annual assessment of the weight and physical fitness of public school students in K-12 schools. Second, our data include the universe of NYC public school students, presenting a much larger sample than previous research, and allowing us to explore heterogeneity by gender and age. Third, we use point specific crime data that include the latitude and longitude of reported crimes from the New York City Police Department (NYPD). We geocode these data to a student’s residential location allowing us to construct crime exposure measures within a very small geographic area around a student’s home. This is a particularly important point as recent evidence shows aggregate measures of crime may lead to underestimating the relationship between crime and physical activity relative to more localized measures (Astell-Burt et al., 2015). Finally, the detailed nature of our crime data also allows us to explore the effect of violent crime as separate and distinct from the effect of property crimes.

The empirical approach we use exploits variation in the timing of violent crimes on a student’s residential H-block² relative to the dates of the Fitnessgram assessment each year, which vary by school, grade, and year. Specifically, we estimate regression models with a large complement of student demographic controls and census tract fixed effects, thereby comparing the weight outcomes of students exposed to violent crime within a very small area around their residence (H-block) with students not exposed to crime near their homes but living in the same neighborhood (census tract). This approach should yield credibly causal estimates of the relationship between violent crime and weight and fitness if, within neighborhoods, exposure to violent crime on a student’s H-block before the Fitnessgram is as if random after conditioning for student-level characteristics, grade, year, and census tract fixed effects. This approach allows us to isolate more robust estimates of the relationship between violent crime and child weight than prior studies in this area.

2. Data and Measures

This paper links three sources of administrative data. First, we use crime data from the NYPD including all reported crimes in NYC with their geographic location (latitude and longitude) for years 2006 to 2016. We geocode these crimes to a blockface and H-block.³ A blockface is a street segment between two intersections, and the H-block includes a central blockface and its adjacent blockfaces forming an H.

Second, we match these data to student-level administrative records from the NYC Department of Education (NYCDOE) for academic years 2009/10–2015/16 (henceforth 2016). These longitudinal data provide information on student demographic characteristics such as race and ethnicity, gender, eligibility for free or reduced-price lunch (a proxy measure to identify low-income students), and receipt of special education services, among others. Most importantly, these data include each student’s residential location, which we also geocode to the blockface and the H-block level. In doing so, we identify whether a crime occurred on a student’s residential blockface and its adjacent blockfaces (H-block) on a given day, month, or year.⁴ Capturing violent exposure at this micro-level is important because local crimes are more likely to affect behaviors (Astell-Burt et al. 2015). Finally, we use data from the NYC Fitnessgram - an annual fitness assessment for students in K-12 – to calculate BMI and derive indicators of obesity and overweight. Fitnessgram data also include measures of aerobic fitness (pacer test) as well as measures of strength, endurance, and flexibility (push-ups, curl-ups, trunk-lifts, and sit-reach left and right).⁵

The final sample is composed of students enrolled in traditional public schools aged 9 to 18 years old in grades 4 through 12 across years 2010 to 2016. To be included in our sample, students must have two consecutive years of Fitnessgram data. We exclude students in PreKindergarten to third grade because we do not have physical fitness measures for them. We also exclude students in full-time special education, those attending charter schools, and those in alternative schools. This accounts for a small percentage of NYC students. Roughly, 3 percent of students in grades 4 to 12 attend alternative schools or are enrolled in full time special education. The percent of charter school students has expanded over time from 2.8 percent in 2010 to 8.16 percent in 2016. Our final analysis sample has a total of 3,035,973 student-year observations (945,414 unique students).

Measures

We define violent crimes following the Uniform Crime Reports (UCR) Part I. This definition includes homicides, aggravated assaults and robberies.⁶ We focus on violent crimes because these are serious crimes against persons and are most likely to affect behavior, and prior work has shown violent crime affects other child outcomes such as school performance and socioemotional outcomes (Sharkey 2010; Sharkey et al. 2014; Ramey and Harrington 2019). We also create a measure of property crime following the UCR Part II definition that includes burglary, grand larceny, grand larceny of motor vehicle theft, and arson.⁷ This allows us to test the robustness of our results to controlling for overall neighborhood disorder, and to test whether property crimes also affect child weight and fitness.

We consider a student exposed to violent crime if they live on an H-block in which a homicide, robbery, or aggravated assault occurred in the year between Fitnessgram assessments. The Fitnessgram assessment is not administered at the same time every year and administration varies by grade and school. This means that two students of the same age in the same year would be tested at different times depending on their grade and school. To determine exposure to violent crime – and the number of exposures – we identify the month of the Fitnessgram for each student in year t and count the number of crimes on a student’s H-block since the Fitnessgram month in year t-1. To give an example, if a student completes the Fitnessgram assessment in October of their fourth grade in 2011, we count all the crimes that happened on their residential H-block between this month and the Fitnessgram assessment month in the third grade of 2010. For some students this Fitnesssgram to Fitnessgram period is roughly a full year (12 months), but for others it could be longer or shorter.⁸ It is important to note that our definition of exposure to violent crime does not assume that the student witnessed the crime or that they were the victim of the crime. Because H-blocks are very small geographic units around a child’s home, and violent crimes are such serious offenses that reverberate across communities it is likely that students or their family had direct or indirect knowledge of the crime, for example through sirens or increased police presence in affected blocks (Schwartz, 2022; Sharkey, 2018). We construct exposure to property crime measures similarly.

The main outcomes of interests are students’ BMI measured in standard deviations that we create using CDC growth charts and standardized by age and sex. We also construct indicators of child obesity (BMI at the 95^th percentile or greater for their age and sex) and overweight (BMI between the 85^th and below the 95^th percentile for their age and sex).

Fitness measures include indicators for whether students are within the healthy fitness zone (HFZ) for their age and sex on six indicators of physical fitness: pacer, curl-up, push-up, trunk-lift, seat-reach right, and seat-reach left. These measures equal to one if the student is in the HFZ for their age and sex and they are zero if students are not in the HFZ. For example, a girl 10 years of age would be in the HFZ for the pacer if she completes between 9 and 54 laps (running back and forth between two points). She would be in the HFZ for curl-ups if she completes between 12 and 26 curl-ups and so on.⁹ Using these variables, we construct measures of aerobic fitness (HFZ for the pacer), strength and endurance (HFZ on curl-ups, push-ups, and trunk-lifts), and flexibility (HFZ on seat-reach right and seat-reach left). We also combine all of these in a “fitness index”. The index ranges from 0 for students that are never in the HFZ across indicators to 6 for students who are always in the HFZ across indicators.

3. Empirical strategy

As we briefly mentioned earlier, the centerpiece of our empirical strategy is a regression model linking exposure to violent crime on a student’s H-block to objective measures of weight (BMI, obesity and overweight) and fitness within census tracts. That is, we compare the outcomes of children exposed to violent crime on their H-block to those of children living in the same neighborhood (census tract) but not exposed to violent crime in very close proximity to their residential location (a typical census tract includes roughly 16 H-blocks.) To be precise, we estimate:

$$y_{ict}={\alpha }_t+{\mathit{\text{Crime}}}_{ict}\beta +{X^{\prime }}_{ict}\theta +{\mu }_m+{\gamma }_g+{\delta }_c+{\epsilon }_{ict}$$

Where y is a weight or fitness outcome for student i in census tract c and year t. Crime consists of three indicator variables that equal 1 if students were exposed to one to 10 crimes only, 11 to 20 crimes only, and 21 or more crimes. We estimate this model with violent crimes only and also with violent and property crimes in the same regression. We also estimate alternative linear and quadratic specifications using the number of violent crimes in students’ H-block instead of the three dummy variables. The model also includes student demographic controls (X’) such as race and ethnicity, low-income status (constructed based on participation in free or reduced-price lunch), participation in special education, limited English proficiency, home language English, foreign born, and age in years. Additionally, the model includes grade (γ_g), year (α_t), census tract fixed effects (δ_c), and Fitnessgram month fixed effects (μ_m). Year fixed effects help us control for unobserved factors that may affect all students in a given year, while census tract fixed effects allow us to control for time invariant unobserved neighborhood characteristics. Fitnessgram month fixed effects allow us to account for any seasonality that could affect both weight outcomes and crime. The addition of census tract fixed effects allows to compare students exposed to violent crime in their H-block to those not exposed but living in the same census tract (neighborhood).

We also estimate all models separately by gender and age to address heterogeneity between boys and girls, and children and adolescents. For example, young children’s weight may be more affected by crime in close proximity to their home if their parents keep them inside to protect them as they are less independent to travel to other neighborhoods by themselves. Adolescents, in contrast, may travel farther away from home and the presence of violent crime near their home may have a smaller effect on their weight and fitness. To the extent that parents treat boys and girls differently or that adolescent boys and girls cope with stressful events such as neighborhood violence differently (Rasmussen et al., 2004) we may also see differences by gender. We split our sample in two age groups: 9 to 13 years old and 14 to 18 years old. For ease of interpretation, we estimate all regressions with binary outcomes using linear probability models. Coefficients in these regressions should be interpreted directly as percentage point changes in the outcome. Standard errors are clustered at the census tract level.¹⁰

4. Results

Descriptive statistics

We begin by comparing the demographic characteristics of children exposed to violent crime and not stratifying the sample by age (Table 1). Most students in our sample (76.6 percent) live on an H-block with at least one violent crime. Of these, 44.8 percent are between 9 and 13 years of age and 38.8 percent are between 14 and 18 years of age. There are some differences between students exposed to violent crime and those not exposed. Among students 9 to 13 years old (columns 1 and 2) and 14 to 18 years old (columns 3 and 4), students who are Black or Hispanic are more likely to be exposed to violent crime than those who are White or Asian. Specifically, 27.51 of 9 to 13 year-olds who experience at least one violent crime exposure on their residential H-block are Black, while 43.62 percent of students in this age group are Hispanic. In contrast, among 9 to 13 year-olds exposed to at least one crime on their H-block, only 11.8 and 17.07 percent are White or Asian. Among adolescents (14 to 18 year-olds) exposed to at least one incident of violence on their H-block, 31.42 percent are Black and 41.13 percent are Hispanic. Among adolescents not exposed to violence, 19.07 percent are Black and 27.25 percent are Hispanic.

Table 1:

Mean demographic characteristics by exposure to neighborhood violent crime and age, NYC public school students, 2010–2016

	Age 9 to 13 years old		Age 14–18 years old
	Not exposed (1)	Exposed (2)	Not exposed (3)	Exposed (4)
Black	15.73	27.51	19.07	31.42
Hispanic	27.92	43.62	27.25	41.13
Asian	24.07	17.07	25.95	17.22
White	32.27	11.79	27.55	10.07
Girls	49.96	50.84	49.56	50.27
Age	11.50	11.53	15.87	15.87
Ever low income	80.98	93.09	84.59	94.18
Foreign born	11.73	14.56	20.40	23.32
LEP	6.01	11.16	5.39	9.34
SPED	12.49	13.15	9.86	11.77
Home language English	60.42	54.94	56.44	53.42
Observations = 3,035,955	431,107	1,361,229	279,595	964,024
%	14.20	44.84	9.21	31.75

Notes: Sample restricted to public school students 9 to 18 years old in grades 4 to 12 from AY 2009/2010 to 2015/2016 with at least two consecutive Fitnessgram measures. Violent crime includes: robberies, homicides, and aggravated assaults.

Low-income students – those that have ever received free or reduced-price lunch – are also overrepresented among students exposed to violent crime on their H-block, regardless of age. As for other demographic characteristics, students exposed to neighborhood violence are more likely to be classified as having limited English proficiency, but differences in their placement in special education, language spoken at home, and age relative to students not exposed to neighborhood violence are not large.

Table 2 shows differences in weight and fitness outcomes between students exposed to different levels of violence. Among 9 to 13 year-olds in the sample, 59.5 percent were exposed to 1 to 10 violent crimes on their H-block, 12.25 percent were exposed to 11–20 crimes and 4.2 percent lived on the most violent H-blocks with more than 20 incidents of violence. Trends look similar among adolescents with the majority exposed to 1 to 10 crimes on their H-block, roughly 13 percent exposed to 11 to 20 crimes, and 4.61 percent exposed to more than 20 crimes. In general, students exposed to violent crime have higher BMI and probability of obesity and overweight than students not exposed, and weight outcomes worsen with increased local violence for children (9 to 13 years old) and adolescents (14 to 18 years old) (Panels A and B). To be exact, children exposed to 1 to 10 crimes on their H-block have an average BMI of 0.66 standard deviations while those exposed to more than 11 violent crimes have a mean BMI of 0.76 to 0.77 standard deviations. Approximately 22 percent of children exposed to 10 violent crimes or fewer have obesity, while this percentage increases to 26 for 9 to 13 years old exposed to more than 20 violent crimes on their H-block. Adolescents experience similar patterns, on average, the BMI of those exposed to 1 to 10 violent crimes is 0.48 standard deviations, it increases to 0.58 for adolescents exposed to 11 to 20 crimes and to 0.61 for adolescents exposed to more than 20 violent crimes on their H-block. Obesity and overweight are also higher for adolescents with 11 or more exposures to violence. Overall, fitness outcomes are also slightly worse for students exposed to violent crime compared to those not exposed, and for those with 11 or more violent crime exposures relative to those with fewer exposures.

Table 2:

Mean outcomes by number of exposures to neighborhood violent crime and age, NYC public school students, 2010–2016.

	Not exposed (1)	Exposed to 1–10 crimes (2)	Exposed 11–20 crimes (3)	Exposed to more than 20 crimes (4)
A.	Age 9 to 13 years old
Weight outcomes
zBMI	0.56	0.66	0.76	0.77
Obesity	0.19	0.22	0.25	0.26
Overweight	0.38	0.42	0.45	0.46
Observations = 1,792,336	431,107	1,066,423	219,498	75,308
%	24.05	59.50	12.25	4.20
Fitness outcomes
Fitness index	3.77	3.49	3.32	3.37
Aerobic fitness	0.43	0.39	0.37	0.37
Strength	0.53	0.47	0.43	0.45
Flexibility	0.57	0.51	0.48	0.48
Observations = 1,748,535	421,809	1,040,050	213,706	72,970
%	24.12	59.48	12.22	4.17
B.	Age 14 to 18 years old
Weight outcomes
zBMI	0.41	0.48	0.58	0.61
Obesity	0.14	0.15	0.18	0.19
Overweight	0.30	0.32	0.36	0.36
Observations = 1,243,637	279,613	746,280	160,348	57,396
%	22.48	60.01	12.89	4.62
Fitness outcomes
Fitness index	3.73	3.48	3.26	3.24
Aerobic fitness (pacer)	0.24	0.23	0.23	0.23
Strength (push up, curl up, trunk lift)	0.56	0.50	0.43	0.43
Flexibility (seat reach right and left)	0.59	0.55	0.51	0.50
Observations = 1,207,337	271,603	724,314	155,806	55,614
%	22.50	59.99	12.90	4.61

Notes: Sample restricted to public school students 9 to 18 years old in grades 4 to 12 from AY 2009/2010 to 2015/2016 with at least two consecutive Fitnessgram measures. Violent crime includes: robberies, homicides, and aggravated assaults.

Taken together these descriptive analyses show that students exposed to violent crime on their residential H-block have higher weight than those not exposed. Further, students with more violent crime exposures in the prior year have higher BMI and higher probability of obesity and overweight than those exposed to fewer crimes. That said, these unadjusted estimates could also be reflecting sociodemographic differences between students exposed to neighborhood violence and not such as those presented in Table 1, and that also correlate with weight and fitness. We turn to regression analyses in the next section to derive more robust estimates.

Exposure to violent crime and student weight and fitness outcomes

Table 3 shows results for weight outcomes stratified by gender and age group. Panel A presents estimates for students 9 to 13 years of age and Panel B shows results for adolescents 14 to 18 years old. Exposure to violent crime increases BMI and the probability of obesity for boys 9 to 13 years old living in neighborhoods with 11 or more violent crimes. Specifically, young boys exposed to 11 to 20 violent crimes have a BMI that is 0.018 standard deviations higher than those not exposed, and they are 0.06 percentage points more likely to have obesity than boys not exposed to violent crime (Panel A, columns 4 and 5). Point estimates associated with exposure to 21 or more crimes are not statistically significant but are similar in magnitude. Further, the probability of overweight is roughly 1 percentage point higher for boys and girls exposed to more than 20 violent crimes compared to those not exposed (columns 13 and 6).

Table 3:

Regression results, exposure to violent crime and weight outcomes by gender and age, NYC public school students, 2010–2016

	Girls			Boys
	zBMI (1)	Obesity (2)	Overweight (3)	zBMI (4)	Obesity (5)	Overweight (6)
A. Age 9–13 years old
1–10 crimes	0.003 (0.004)	−0.001 (0.001)	0.002 (0.002)	0.008 (0.005)	0.003 (0.002)	0.003 (0.002)
11–20 crimes	0.010 (0.007)	−0.001 (0.002)	0.003 (0.003)	0.018* (0.007)	0.006* (0.003)	0.006 (0.003)
21 or more crimes	0.020 (0.010)	0.005 (0.004)	0.012** (0.004)	0.017 (0.011)	0.007 (0.004)	0.009* (0.005)
Observations	907,469	907,469	907,469	884,863	884,863	884,863
R-squared	0.061	0.038	0.045	0.043	0.034	0.035
B. Age 14 to 18 years old
1–10 crimes	0.010* (0.005)	0.003* (0.001)	0.005** (0.002)	0.005 (0.005)	0.000 (0.002)	0.001 (0.002)
11–20 crimes	0.021** (0.007)	0.003 (0.002)	0.009** (0.003)	0.014 (0.008)	0.004 (0.003)	0.005 (0.003)
21 or more crimes	0.034** (0.010)	0.006 (0.004)	0.015** (0.005)	−0.001 (0.012)	0.002 (0.004)	−0.001 (0.005)
Observations	623,208	623,208	623,208	620,423	620,423	620,423
R-squared	0.079	0.039	0.051	0.047	0.027	0.030

Standard errors in parentheses clustered by census tract (**p<.01 *p<.05)

Notes: Student controls include: Black, Asian, White, ever low income, home language English, foreign born, limited English proficiency, special education and age in years. All models include year, grade fixed effects, Fitnessgram month fixed effects, and census tract fixed effects. Sample restricted to public school students with at least two consecutive Fitnessgram measures. BMI measured in standard deviations, standardized for each gender and age. AY 2010–2016.

Panel B shows results for adolescents. Adolescent girls are the most affected by neighborhood violence. They have a BMI that is 0.01 standard deviations higher when exposed to 1 to 10 violent crimes and it increases with the number of exposures. For those exposed to 11 to 20 crimes BMI is 0.021 standard deviation higher compared to adolescent girls not exposed. Among those within the most violent H-blocks, BMI is 0.034 standard deviations higher (column 1). These increases in BMI as a result of neighborhood violence seem to result in higher probability of overweight. Exposure to one to 10 violent crimes results in a 0.5 percentage point increase in the probability of overweight, while exposure to 11 to 20 crimes increases the probability of overweight by 0.9 percentage points. As before, adolescent girls in the most violent H-blocks that see more than 20 crimes have a probability of overweight that is 1.5 percentage points higher (column 3). Adolescent boys do not seem to be affected by neighborhood violent crime (columns 4 to 6). Results from linear and quadratic specifications largely confirm these conclusions (Table A2 and A3 in appendix).

Table 4 shows results for fitness outcomes. Overall, we find little evidence that exposure to neighborhood violence affects students’ fitness for children (9 to 13 years old) and adolescents (14 to 18 years old). That said, we see small declines in the fitness index of 0.017 and 0.030 points for young girls exposed to 1 to 10 crimes and 11 to 20 crimes. These drops are quite small and represent a decline of less than one percent relative to the mean for girls in this age group. Interestingly, we also see worse aerobic fitness for young boys. Specifically, young boys exposed to one to 10 crimes are 0.4 percentage points less likely to be in the HFZ for aerobic fitness, this probability is −0.7 percentage points and −1.1 percentage points for boys exposed to 10–20 crimes and more than 20 crimes, respectively. These represents declines of roughly 1 to 2 percent over the mean for boys this age.¹¹

Table 4:

Regression results, exposure to violent crime and fitness outcomes by gender and age, NYC public school students, 2010–2016

	Girls				Boys
	Fitness Index (1)	Aerobic Fitness (2)	Strength (3)	Flexibility (4)	Fitness Index (5)	Aerobic Fitness (6)	Strength (7)	Flexibility (8)
A. Age 9–13 years old
1–10 crimes	−0.017* (0.007)	−0.001 (0.002)	−0.003 (0.002)	−0.005* (0.002)	−0.011 (0.006)	−0.004* (0.002)	−0.006** (0.002)	0.000 (0.002)
11–20 crimes	−0.030** (0.011)	−0.003 (0.003)	−0.005 (0.003)	−0.007* (0.003)	−0.014 (0.011)	−0.007* (0.003)	−0.004 (0.003)	−0.001 (0.003)
21 or more crimes	−0.009 (0.017)	−0.006 (0.004)	−0.007 (0.005)	0.003 (0.005)	−0.005 (0.017)	−0.011* (0.005)	−0.004 (0.004)	0.003 (0.005)
Observations	885,957	885,957	885,957	885,957	862,575	862,575	862,575	862,575
R-squared	0.127	0.078	0.077	0.077	0.084	0.070	0.056	0.054
B. Age 14 to 18 years old
1–10 crimes	−0.010 (0.007)	−0.000 (0.002)	−0.003 (0.002)	−0.003 (0.002)	−0.014* (0.007)	−0.004* (0.002)	−0.003 (0.002)	−0.002 (0.002)
11–20 crimes	−0.018 (0.010)	−0.000 (0.002)	−0.004 (0.003)	−0.005 (0.003)	−0.026* (0.010)	−0.003 (0.003)	−0.006 (0.003)	−0.005 (0.003)
21 or more crimes	−0.007 (0.015)	−0.002 (0.004)	−0.000 (0.005)	−0.003 (0.004)	0.004 (0.015)	0.005 (0.005)	−0.004 (0.005)	0.001 (0.004)
Observations	608,012	608,012	608,012	608,012	599,319	599,319	599,319	599,319
R-squared	0.111	0.035	0.073	0.069	0.067	0.038	0.045	0.037

Standard errors in parentheses clustered by census tract (** p<0.01, * p<0.05)

Notes: All models include student controls: Black, Asian, White, ever low income, home language English, foreign born, limited English proficiency, special education, age in years. All models include year, grade fixed effects, Fitnessgram month fixed effects, and census tract fixed effects. Sample restricted to students with two consecutive Fitnessgram measures and fitness information on all indicators Fitness index ranges from 0 for children and adolescents that are in the healthy fitness zone for none the fitness measures to 6 for children and adolescents who are in the healthy fitness zone for all of the fitness measures. AY 2010–2016.

Taken together, results in these analyses show that exposure to violent crime worsens BMI, obesity, and overweight among adolescent girls. There is no corresponding decline in fitness suggesting reduced physical activity may not be the main mechanism explaining changes in BMI, obesity, and overweight for this group. We find little evidence that crime is associated with changes in weight and fitness for adolescent boys with some suggestive evidence of higher BMI and probability of obesity and overweight among younger boys and declines in aerobic fitness for this group.

Accounting for exposure to property crime

Results so far focus on violent crimes because these are more serious crimes committed against persons and are likely to generate more attention (Schwartz et al. 2022). If it is violent crime rather than general neighborhood disorder – or other characteristics correlated with crime – driving the results, our estimates should remain unchanged if we add additional variables capturing exposure to property crime. Further, while there is limited evidence on the effects of different types of crimes on weight outcomes and fitness, in general, exposure to property crime has been linked with null findings and one negative association with physical fitness (Yu and Lippert 2016). Thus, these analyses not only test the robustness of our results but also contribute to the literature by providing robust evidence of the type of crimes that are more likely to affect child and adolescent weight and fitness.

Results for violent crime remain largely unchanged when we account for exposure to property crime (Tables 5 and 6). That said, we no longer see evidence that exposure to violent crime negatively affects BMI, obesity, and overweight among young boys. As before, adolescent girls are the most affected by neighborhood violence. Their BMI worsens with the number of exposures increasing from 0.01 standard deviations for girls exposed to 1 to 10 crimes, to 0.22 standard deviations for those exposed to 11 to 20 crimes, and to 0.035 standard deviations for those exposed to 21 or more local crimes. These changes in BMI seem to translate into higher probability of overweight among this group. Point estimates for overweight are similar in magnitude to those reported in the prior section. Interestingly, when we account for exposure to property crimes, we see an increase in BMI for adolescent boys (Table 5, column 4) of 0.018 standard deviations after exposure to 11–20 crimes and an increase of 0.07 percentage point in the probability of overweight for this group (Table 5, columns 4 and 6).

Table 5:

Regression results, exposure to violent and property crime and weight outcomes by gender, 9 to 13 year-olds, NYC public school students, 2010–2016

	Girls			Boys
	zBMI (1)	Obesity (2)	Overweight (3)	zBMI (4)	Obesity (5)	Overweight (6)
Violent crime
1–10 crimes	0.002 (0.004)	−0.000 (0.001)	0.001 (0.002)	0.006 (0.004)	0.002 (0.002)	0.003 (0.002)
11–20 crimes	0.006 (0.007)	−0.001 (0.002)	0.002 (0.003)	0.014 (0.007)	0.005 (0.003)	0.004 (0.003)
21 or more crimes	0.012 (0.011)	0.002 (0.004)	0.009* (0.005)	0.012 (0.011)	0.005 (0.004)	0.007 (0.005)
Property crime
1–10 crimes	−0.001 (0.005)	−0.005** (0.002)	0.001 (0.002)	0.010 (0.006)	0.004 (0.002)	0.002 (0.002)
11–20 crimes	0.003 (0.007)	−0.003 (0.002)	0.002 (0.003)	0.013 (0.007)	0.004 (0.003)	0.003 (0.003)
21 or more crimes	0.019 (0.010)	0.003 (0.003)	0.007 (0.004)	0.014 (0.011)	0.006 (0.004)	0.006 (0.004)
Observations	907,469	907,469	907,469	884,863	884,863	884,863
R-squared	0.061	0.038	0.045	0.043	0.034	0.035

Standard errors in parentheses clustered by census tract (**p<.01 *p<.05)

Notes: Student controls include: Black, Asian, White, ever low income, home language English, foreign born, limited English proficiency, special education and age in years. All models include year, grade fixed effects, Fitnessgram month fixed effects, and census tract fixed effects. Sample restricted to students with at least two consecutive Fitnessgram measures. BMI measured in standard deviations, standardized for each gender and age. AY 2010–2016.

Table 6:

Regression results, exposure to violent and property crime and weight outcomes by gender, 14–18 year-olds, NYC public school students, 2010–2016

	Girls			Boys
	zBMI (1)	Obesity (2)	Overweight (3)	zBMI (4)	Obesity (5)	Overweight (6)
Violent crime
1–10 crimes	0.011* (0.005)	0.004* (0.001)	0.006** (0.002)	0.006 (0.005)	0.001 (0.002)	0.002 (0.002)
11–20 crimes	0.022** (0.007)	0.004 (0.002)	0.011** (0.003)	0.018* (0.008)	0.005 (0.003)	0.007* (0.003)
21 or more crimes	0.035** (0.011)	0.007 (0.004)	0.017** (0.005)	0.007 (0.012)	0.004 (0.004)	0.002 (0.005)
Property crime
1–10 crimes	−0.002 (0.005)	−0.003 (0.002)	−0.001 (0.002)	−0.005 (0.007)	−0.002 (0.002)	−0.002 (0.003)
11–20 crimes	−0.004 (0.007)	−0.006* (0.002)	−0.003 (0.003)	−0.008 (0.008)	−0.002 (0.003)	−0.003 (0.003)
21 or more crimes	−0.002 (0.010)	−0.002 (0.003)	−0.003 (0.004)	−0.021 (0.012)	−0.004 (0.004)	−0.007 (0.005)
Observations	623,208	623,208	623,208	620,423	620,423	620,423
R-squared	0.079	0.039	0.051	0.047	0.027	0.030

Standard errors in parentheses clustered by census tract (**p<.01 *p<.05)

Notes: Student controls include: Black, Asian, White, ever low income, home language English, foreign born, limited English proficiency, special education and age in years. All models include year fixed effects, grade fixed effects, Fitnessgram month fixed effects, and census tract fixed effects. Sample restricted to students with at least two consecutive Fitnessgram measures. BMI is measured in standard deviations standardized for each gender and age. AY 2010–2016.

Overall, there is little evidence that exposure to property crimes worsens weight outcomes suggesting it is the most serious crimes that correlate with child and adolescent weight changes. In particular for adolescents, property crime point estimates are generally not statistically significant and they are also smaller in magnitude.

Findings for physical fitness are also robust to the addition of property crimes (Tables A6 and A7 in appendix). We still see lower scores on the fitness index for young girls, and worse aerobic fitness for young boys exposed to local violence. To be specific, boys exposed to 1 to 10 violent crimes are 0.5 percentage points less likely to be in the healthy fitness zone for aerobic fitness, 0.8 percentage points less likely after exposure to 11 to 20 crimes and 1.2 percentage point less likely after exposure to more than 20 violent crimes. We do not see any statistically significant results associated with property crimes. In contrast to our weight estimates, we find little evidence that exposure to neighborhood crime negatively affects the fitness outcomes of adolescents.

Results in this section support the hypothesis that not all neighborhood crimes matter equally. Indeed, we show evidence that exposure to local violent crime is more detrimental for child, and especially, adolescent weight than exposure to property crime. While teasing apart the individual effect of violent and property crime is complicated by the correlation between them, our results suggest our findings are not driven by general neighborhood disorder but rather by the more disruptive effects of neighborhood violence.

Robustness tests: alternative specifications

Students spend much of their time in school, and schools vary widely in their characteristics. Some schools include large athletic facilities while others have smaller areas or limited space for intense or moderate physical activity. Thus, fixed elements of the school built environment may be omitted variables in these analyses. To address this concern, and the fact that students do not always attend their neighborhood school, we estimate all regressions with both census tract and school fixed effects (Tables A8 and A9 in appendix). School fixed effects control for time invariant unobserved characteristics related to the school built environment that may correlate with weight and fitness. Overall, coefficients associated with the weight outcomes remain unchanged with the addition of school fixed effects. Boys exposed to 11 or more crimes have a BMI that is 0.016 standard deviations higher. As before, adolescent girls experience the largest increases in BMI and the probability of overweight. This last set of results is also robust to controlling for exposure to property crimes.

With the addition of school fixed effects, we still see a reduction in aerobic fitness for boys exposed to 11 to 20 crimes of 0.5 percentage points but point estimates for the other crime exposure levels are no longer statistically significant. As before, we find little evidence that violent crime affects the fitness outcomes of adolescents.

Finally, we re-estimate our main specifications with census tracts by year fixed effects instead of census tract and year fixed effects to account for potential unobserved time varying census-tract specific confounders that are correlated both with crime and obesity. Results are reported in Tables A10 and A11 in appendix. Overall, our results remain largely unchanged. Adolescent girls’ weight outcomes are the most affected by local violence, and we see little change for adolescent boys, and young girls and boys, with a few exceptions similar to those reported in Table 3. Fitness results also remain consistent with the main specification reported in Table 4. In sum, we can rule out that findings in this paper are driven by differential time trends across census tracts or by unobserved time invariant school characteristics.

5. Conclusion

This paper estimates the relationship between neighborhood violent crime and child and adolescent weight and physical fitness. Results show meaningful gains in BMI and the probability of overweight among adolescent girls (14 to 18 years of age) that increase with the level of exposure. For adolescent girls living within the most violent H-blocks, BMI increases by 0.035 standard deviations and the probability of obesity by 1.7 percentage points compared to girls in the same neighborhood but not exposed to local violence. To put these findings in context, these represent an increase of 7.4 and 5.4 percent over the mean, respectively. These results are robust to controlling for exposure to property crimes as well as to the addition of school fixed effects and census tract by year fixed effects. We also find suggestive evidence that boys 9 to 13 years old exposed to 11 or more violent crimes on their H-block have approximately 0.018 standard deviations higher BMI (roughly a 2 percent increase over the group sample mean), compared to others in their neighborhood that live farther away. Young girls and boys have a 1 percentage point increase likelihood of overweight. Results are robust to the addition of school fixed effects and census tract by year fixed effects. Controlling for property crimes makes estimates for boys statistically insignificant though point estimates remain similar. In general, we find little evidence of a relationship between physical fitness and exposure to local violence. Point estimates are usually insignificant or represent very small changes. In other words, we find little support for the hypothesis that declines in physical fitness due to neighborhood crime explain increases in weight outcomes. Further, we see little evidence of deleterious weight effects due to property crime, suggesting it is violent crime - the most serious and salient crimes against persons - that affect weight rather than general neighborhood disorder. This is especially salient in the case of adolescent girls.

As for the magnitude of these effects, they are comparable or even larger than those of school-based interventions in NYC. For example, increases in the BMI of adolescent girls as a result of neighborhood violence far exceed the estimated effects of the installation of waterjets in NYC public schools which yielded BMI decreases of 0.025 (.022) standard deviations for boys (girls) (Schwartz et al. 2016). Among boys 9 to 13 years old, violent crime increases BMI by more than half of the waterjet reductions. Further, declines in the probability of overweight as a result of waterjets in schools are smaller than the increases we found in the present paper for adolescent girls.

There are some limitations in this study worth noting. First, our study sample includes public school students in specific ages, grades, and school settings (i.e. excludes charter, private, and alternative schools, students in full-time special education, and students younger than 9 years of age). Thus, caution may be warranted in generalizing to students of other ages or in different school settings. Second, we acknowledge our crime data may reflect some underreporting, or misclassification of crimes, thereby introducing measurement error in our crime counts. Third, while we provide estimates of the negative effect of neighborhood violent crime on weight, this paper sheds little light on mechanisms beyond the examination of fitness outcomes. Future work should explore whether and to what extent changes in weight outcomes are driven by increased sedentary behaviors, stress-related behaviors, or a combination of both. Further, the differences in effects by age and gender suggest future work exploring differential mechanisms is well warranted. For example, existing work suggests that girls and boys use different coping strategies when exposed to neighborhood violence (Rasmussen et al. 2004) and experience different emotional responses (Javdani et al., 2014). Future work should also try to uncover the long-term consequences of exposure to violent crime, and whether and to what extent living in violent neighborhoods during early childhood is associated with higher weight in late childhood or adolescence.

In sum, our study relies on large sample of public school students in New York City and objective measures of weight, physical fitness, and violent crime to show that local neighborhood violence increases BMI and the probability of obesity and overweight for adolescent girls in particular.

Highlights.

• Exposure to local violent crime increases the BMI and probability of overweight of adolescents girls (14 to 18 years old).

• Adolescent girls exposed to more than 20 crimes on their H-block have a BMI that is 7.5 percent higher than the average BMI for their group.

• They also experience a 5.5 percent increase in the probability of overweight relative to their group mean.

• In general, we find no change in physical fitness as a result of neighborhood crime, with a few notable exceptions. Boys 9 to 13 years of age experience small declines in aerobic fitness of roughly 1 to 2 percent relative to the average for boys in this age group.

• We find no evidence that property crime affects weight or fitness outcomes.

Appendix

Table A1:

Mean student characteristics by H-block geocoding, NYC public school students, 2010–2016

	Has H-block		H-block missing
Black	0.27	0.44	0.31	0.46
Hispanic	0.39	0.49	0.38	0.49
Asian	0.19	0.39	0.17	0.37
White	0.15	0.36	0.14	0.34
Girl	0.50	0.50	0.50	0.50
Age in years	13.41	2.51	13.29	2.49
Ever low income	0.91	0.29	0.90	0.30
LEP	0.11	0.31	0.15	0.35
SPED	0.12	0.33	0.12	0.32
English at home	0.55	0.50	0.57	0.50
Foreign born	0.19	0.39	0.23	0.42
Observations	3,484,902		168,656
	95.38		4.62

Note: Students missing H-block are those in our full dataset for whom we were unable to match H-block information before other sample restrictions.

Table A2:

Regression results, exposure to violent crime and weight outcomes by gender and age, linear specification, NYC public school students, 2010–2016

	Girls			Boys
	zBMI (1)	Obesity (2)	Overweight (3)	zBMI (4)	Obesity (5)	Overweight (6)
A. Age 9–13 years old
HBlock Crimes (10s)	0.009* (0.003)	0.002 (0.001)	0.004** (0.001)	0.009* (0.004)	0.003* (0.002)	0.004* (0.002)
Observations	907,469	907,469	907,469	884,863	884,863	884,863
R-squared	0.061	0.038	0.045	0.043	0.034	0.035
B. Age 14 to 18 years old
HBlock Crimes (10s)	0.009* (0.004)	0.002 (0.001)	0.005** (0.002)	−0.001 (0.004)	0.002 (0.001)	0.000 (0.002)
Observations	623,208	623,208	623,208	620,423	620,423	620,423
R-squared	0.079	0.039	0.051	0.047	0.027	0.030

Standard errors in parentheses clustered by census tract (**p<.01 *p<.05)

Notes: Student controls include: Black, Asian, White, ever low income, home language English, foreign born, limited English proficiency, special education and age in years. All models include year fixed effects, grade fixed effects, Fitnessgram month fixed effects, and census tract fixed effects. Sample restricted to students with at least two consecutive Fitnessgram measures. BMI measured in standard deviations, standardized for each gender and age standardized for each gender and age. AY 2010–2016.

Table A3:

Regression results, exposure to violent crime and weight outcomes by gender and age, quadratic specification, NYC public school students, 2010–2016

	Girls			Boys
	zBMI (1)	Obesity (2)	Overweight (3)	zBMI (4)	Obesity (5)	Overweight (6)
A. Age 9–13 years old
HBlock Crimes (10s)	0.007 (0.004)	0.002 (0.001)	0.003* (0.001)	0.009* (0.004)	0.003* (0.002)	0.004* (0.002)
HBlock Crimes (10s) Squared	0.000** (0.000)	0.000 (0.000)	0.000** (0.000)	0.000 (0.000)	0.000 (0.000)	0.000 (0.000)
R-squared	907,469	907,469	907,469	884,863	884,863	884,863
Observations	0.061	0.038	0.045	0.043	0.034	0.035
B. Age 14 to 18 years old
HBlock Crimes (10s)	0.009* (0.004)	0.002 (0.001)	0.005** (0.002)	−0.001 (0.004)	0.002 (0.001)	0.000 (0.002)
HBlock Crimes (10s) Squared	−0.000 (0.000)	−0.000 (0.000)	−0.000 (0.000)	−0.000 (0.000)	−0.000 (0.000)	−0.000* (0.000)
R-squared	623,208	623,208	623,208	620,423	620,423	620,423
Observations	0.079	0.039	0.051	0.047	0.027	0.030

Standard errors in parentheses clustered by census tract (**p<.01 *p<.05)

Notes: Student controls include: Black, Asian, White, ever low income, home language English, foreign born, Limited English proficiency, special education and age in years. All models include year fixed effects, grade fixed effects, Fitnessgram month fixed effects, and census tract fixed effects. Sample restricted to students with at least two consecutive Fitnessgram measures. BMI measured in standard deviations standardized for each gender and age. AY 2010–2016.

Table A4:

Regression results, exposure to violent crime and fitness outcomes by gender and age, linear model, NYC public school students, 2010–2016

	Girls				Boys
	Fitness Index (1)	Aerobic Fitness (2)	Strength (3)	Flexibility (4)	Fitness Index (5)	Aerobic Fitness (6)	Strength (7)	Flexibility (8)
A. Age 9–13 years old
HBlock Crimes (10s)	−0.008 (0.006)	−0.003 (0.001)	−0.004* (0.002)	0.000 (0.002)	−0.006 (0.006)	−0.004* (0.002)	−0.001 (0.002)	−0.000 (0.002)
Observations	885,957	885,957	885,957	885,957	862,575	862,575	862,575	862,575
R-squared	0.127	0.078	0.077	0.077	0.084	0.070	0.056	0.054
B. Age 14 to 18 years old
HBlock Crimes (10s)	−0.003 (0.005)	−0.000 (0.001)	−0.001 (0.002)	−0.001 (0.001)	−0.004 (0.005)	0.001 (0.001)	−0.002 (0.002)	−0.000 (0.001)
Observations	608,012	608,012	608,012	608,012	599,319	599,319	599,319	599,319
R-squared	0.111	0.035	0.073	0.069	0.067	0.038	0.045	0.037

Standard errors in parentheses clustered by census tract (**p<.01 *p<.05)

Notes: Student controls include: Black, Asian, White, ever low income, home language English, foreign born, limited English proficiency, special education and age in years. All models include year fixed effects, grade fixed effects, Fitnessgram month fixed effects, and census tract fixed effects. Sample restricted to students with at least two consecutive Fitnessgram measures and fitness information on all indicators Fitness index ranges from 0 for children and adolescents that are in the healthy fitness zone for none the fitness measures to 6 for children and adolescents who are in the healthy fitness zone for all of the fitness measures. AY 2010–2016.

Table A5:

Regression results, exposure to violent crime and fitness outcomes by gender and age, quadratic specification, NYC public school students, 2010–2016

	Girls				Boys
	Fitness Index (1)	Aerobic Fitness (2)	Strength (3)	Flexibility (4)	Fitness Index (5)	Aerobic Fitness (6)	Strength (7)	Flexibility (8)
A. Age 9–13 years old
HBlock Crimes (10s)	−0.016 (0.015)	−0.001 (0.003)	−0.005 (0.003)	−0.003 (0.004)	−0.002 (0.014)	0.000 (0.004)	−0.002 (0.003)	−0.001 (0.003)
HBlock Crimes (10s) Squared	0.003 (0.005)	−0.001 (0.001)	0.000 (0.001)	0.001 (0.001)	−0.001 (0.004)	−0.001 (0.001)	0.000 (0.001)	0.000 (0.001)
Observations	885,957	885,957	885,957	885,957	862,575	862,575	862,575	862,575
R-squared	0.127	0.078	0.077	0.077	0.084	0.070	0.056	0.054
B. Age 14 to 18 years old
HBlock Crimes (10s)	−0.007 (0.009)	0.002 (0.002)	−0.001 (0.003)	−0.005 (0.003)	−0.002 (0.014)	0.000 (0.004)	−0.002 (0.003)	−0.001 (0.003)
HBlock Crimes (10s) Squared	0.001 (0.002)	−0.001 (0.000)	0.000 (0.001)	0.001 (0.001)	−0.001 (0.004)	−0.001 (0.001)	0.000 (0.001)	0.000 (0.001)
Observations	608,012	608,012	608,012	608,012	862,575	862,575	862,575	862,575
R-squared	0.111	0.035	0.073	0.069	0.084	0.070	0.056	0.054

Standard errors in parentheses clustered by census tract (**p<.01 *p<.05)

Notes: Student controls include: Black, Asian, White, ever low income, home language English, foreign born, limited English proficiency, special education and age in years. All models include year fixed effects, grade fixed effects, Fitnessgram month fixed effects, and census tract fixed effects. Sample restricted to students with at least two consecutive Fitnessgram measures and fitness information on all indicators Fitness index ranges from 0 for children and adolescents that are in the healthy fitness zone for none the fitness measures to 6 for children and adolescents who are in the healthy fitness zone for all of the fitness measures. Ay 2010–2019.

Table A6:

Regression results, exposure to violent and property crime and fitness outcomes, 9–13 year-olds, NYC public school students, 2010–2016.

	Girls				Boys
	Fitness index (1)	Pacer (2)	Strength (3)	Flexibility (4)	Fitness index (5)	Pacer (6)	Strength (7)	Flexibility (8)
Violent crime
1–10 crimes	−0.017** (0.007)	−0.001 (0.002)	−0.002 (0.002)	−0.005** (0.002)	−0.012 (0.006)	−0.005* (0.002)	−0.007** (0.002)	0.000 (0.002)
11–20 crimes	−0.033** (0.011)	−0.003 (0.003)	−0.004 (0.003)	−0.008* (0.003)	−0.014 (0.011)	−0.008** (0.003)	−0.004 (0.003)	−0.001 (0.003)
21 or more crimes	−0.011 (0.017)	−0.006 (0.004)	−0.004 (0.005)	0.002 (0.005)	−0.004 (0.018)	−0.012* (0.005)	−0.002 (0.005)	0.003 (0.005)
Property crime
1–10 crimes	0.000 (0.008)	−0.001 (0.002)	−0.001 (0.002)	0.002 (0.002)	0.003 (0.008)	0.002 (0.002)	0.001 (0.002)	−0.000 (0.002)
11–20 crimes	0.008 (0.011)	−0.000 (0.003)	−0.002 (0.003)	0.004 (0.003)	0.001 (0.011)	0.003 (0.003)	0.002 (0.003)	−0.001 (0.003)
21 or more crimes	−0.001 (0.016)	−0.001 (0.004)	−0.009* (0.004)	0.003 (0.005)	0.001 (0.016)	0.004 (0.004)	−0.003 (0.005)	0.001 (0.005)
Observations	885,957	885,957	885,957	885,957	862,575	862,575	862,575	862,575
R-squared	0.127	0.078	0.077	0.077	0.084	0.070	0.056	0.054

Standard errors in parentheses clustered by census tract (**p<.01 *p<.05)

Notes: Student controls include: Black, Asian, White, ever low income, home language English, foreign born, limited English proficiency, special education and age in years. All models include year fixed effects, grade fixed effects, Fitnessgram month fixed effects, and census tract fixed effects. Sample restricted to students with at least two consecutive Fitnessgram measures and fitness information on all indicators Fitness index ranges from 0 for children and adolescents that are in the healthy fitness zone for none the fitness measures to 6 for children and adolescents who are in the healthy fitness zone for all of the fitness measures. AY 2010–2016.

Table A7:

Regression results, exposure to violent and property crime and fitness outcomes, 14–18 year-olds, NYC public school students, 2010–2016

	Girls				Boys
	Fitness index (1)	Pacer (2)	Strength (3)	Flexibility (4)	Fitness index (5)	Pacer (6)	Strength (7)	Flexibility (8)
Violent crime
1–10 crimes	−0.014* (0.007)	−0.001 (0.002)	−0.003 (0.002)	−0.004* (0.002)	−0.014* (0.007)	−0.005* (0.002)	−0.003 (0.002)	−0.002 (0.002)
11–20 crimes	−0.026* (0.011)	−0.003 (0.002)	−0.005 (0.003)	−0.007* (0.003)	−0.026* (0.011)	−0.005 (0.003)	−0.006 (0.003)	−0.005 (0.003)
21 or more crimes	−0.011 (0.016)	−0.006 (0.004)	0.001 (0.005)	−0.004 (0.005)	0.001 (0.016)	0.001 (0.005)	−0.004 (0.005)	0.002 (0.004)
Property crime
1–10 crimes	0.013 (0.008)	0.003 (0.002)	0.001 (0.003)	0.005* (0.003)	0.003 (0.008)	0.003 (0.002)	0.002 (0.003)	−0.000 (0.002)
11–20 crimes	0.031** (0.011)	0.007** (0.002)	0.005 (0.003)	0.007* (0.003)	0.000 (0.010)	0.003 (0.003)	0.002 (0.003)	−0.000 (0.003)
21 or more crimes	0.005 (0.016)	0.007 (0.004)	−0.004 (0.005)	0.002 (0.005)	0.009 (0.016)	0.009* (0.005)	0.002 (0.005)	−0.002 (0.004)
Observations	608,012	608,012	608,012	608,012	599,319	599,319	599,319	599,319
R-squared	0.111	0.035	0.073	0.069	0.067	0.038	0.045	0.037

Standard errors in parentheses clustered by census tract (**p<.01 *p<.05)

Notes: Student controls include: Black, Asian, White, ever low income, home language English, foreign born, limited English proficiency, special education and age in years. All models include year fixed effects, grade fixed effects, Fitnessgram year fixed effects, and census tract fixed effects. Sample restricted to students with at least two consecutive Fitnessgram measures and fitness information on all indicators Fitness index ranges from 0 for children and adolescents that are in the healthy fitness zone for none the fitness measures to 6 for children and adolescents who are in the healthy fitness zone for all of the fitness measures. AY 2010–2016.

Table A8:

Regression results, exposure to violent crime and child weight outcomes by gender, census tract and school fixed effects, NYC public school students, 2010–2016

	Girls			Boys
	zBMI (1)	Obesity (2)	Overweight (3)	zBMI (4)	Obesity (5)	Overweight (6)
A. Age 9–13 years old
1–10 crimes	0.001 (0.004)	−0.001 (0.001)	0.001 (0.002)	0.008 (0.005)	0.003 (0.002)	0.003 (0.002)
11–20 crimes	0.007 (0.007)	−0.001 (0.002)	0.002 (0.003)	0.016* (0.007)	0.006* (0.003)	0.005 (0.003)
21 or more crimes	0.016 (0.010)	0.003 (0.004)	0.011* (0.004)	0.015 (0.011)	0.006 (0.004)	0.009* (0.004)
Observations	907,436	907,436	907,436	884,837	884,837	884,837
R-squared	0.074	0.045	0.053	0.055	0.042	0.043
A. Age 14–18 years old
1–10 crimes	0.011* (0.005)	0.003* (0.001)	0.005** (0.002)	0.006 (0.005)	0.000 (0.002)	0.001 (0.002)
11–20 crimes	0.022** (0.007)	0.003 (0.002)	0.010** (0.003)	0.015 (0.008)	0.004 (0.003)	0.006 (0.003)
21 or more crimes	0.034** (0.010)	0.006 (0.004)	0.015** (0.005)	−0.001 (0.012)	0.002 (0.004)	−0.001 (0.005)
Observations	623,169	623,169	623,169	620,396	620,396	620,396
R-squared	0.095	0.052	0.066	0.057	0.037	0.039

Standard errors in parentheses clustered by census tract (**p<.01 *p<.05)

Notes: Student controls include: Black, Asian, White, ever low income, home language English, foreign born, limited English proficiency, special education and age in years. All models include year fixed effects, grade fixed effects, Fitnessgram month fixed effects, census tract fixed effects, and school fixed effects. Sample restricted to students with at least two consecutive Fitnessgram measures. BMI in standard devisations standardized for each gender and age. AY 2010–2016.

Table A9:

Regression results, exposure to violent crime and fitness outcomes by gender, census tract and school fixed effects, NYC public school students, 2010–2016

	Girls				Boys
	Fitness index (1)	Aerobic fitness (2)	Strength (3)	Flexibility (4)	Fitness index (5)	Aerobic fitness (6)	Strength (7)	Flexibility (8)
A. Age 9–13 years old
1–10 crimes	−0.009 (0.006)	−0.001 (0.002)	0.000 (0.002)	−0.003 (0.002)	−0.006 (0.005)	−0.003 (0.002)	−0.003* (0.002)	0.001 (0.002)
11–20 crimes	−0.016 (0.009)	−0.003 (0.002)	−0.002 (0.003)	−0.003 (0.003)	−0.003 (0.009)	−0.005* (0.003)	−0.001 (0.003)	0.002 (0.003)
21 or more crimes	0.006 (0.013)	−0.001 (0.004)	−0.005 (0.004)	0.006 (0.004)	0.007 (0.014)	−0.006 (0.004)	−0.002 (0.004)	0.006 (0.004)
Observations	885,932	885,932	885,932	885,932	862,556	862,556	862,556	862,556
R-squared	0.279	0.205	0.191	0.188	0.229	0.175	0.146	0.175
B. Age 14–18 years old
1–10 crimes	−0.016** (0.006)	−0.002 (0.002)	−0.003 (0.002)	−0.004 (0.002)	−0.011 (0.006)	−0.004* (0.002)	−0.003 (0.002)	−0.001 (0.002)
11–20 crimes	−0.021* (0.009)	−0.002 (0.002)	−0.006* (0.003)	−0.004 (0.003)	−0.014 (0.009)	−0.003 (0.003)	−0.005 (0.003)	−0.001 (0.003)
21 or more crimes	−0.006 (0.013)	−0.003 (0.003)	−0.002 (0.004)	−0.000 (0.004)	0.020 (0.014)	0.006 (0.004)	−0.003 (0.004)	0.005 (0.004)
Observations	607,984	607,984	607,984	607,984	599,301	599,301	599,301	599,301
R-squared	0.249	0.143	0.162	0.157	0.179	0.125	0.103	0.125

Standard errors in parentheses clustered by census tract (**p<.01 *p<.05)

Notes: Student controls include: Black, Asian, White, ever low income, home language English, foreign born, limited English proficiency, special education and age in years. All models include year fixed effects, grade fixed effects, Fitnessgram month fixed effects, school fixed effects, and census tract fixed effects. Sample restricted to students with at least two consecutive Fitnessgram measures and fitness information on all indicators Fitness index ranges from 0 for children and adolescents that are in the healthy fitness zone for none the fitness measures to 6 for children and adolescents who are in the healthy fitness zone for all of the fitness measures. AY 2010–2016.

Table A10:

Regression results, exposure to violent crime and weight outcomes by gender, census tract by year fixed effects, NYC public school students, 2010–2016

	Girls			Boys
	zBMI (1)	Obesity (2)	Overweight (3)	zBMI (4)	Obesity (5)	Overweight (6)
A. Age 9–13 years old
1–10 crimes	0.003 (0.005)	−0.001 (0.002)	0.002 (0.002)	0.009 (0.005)	0.003 (0.002)	0.004 (0.002)
11–20 crimes	0.012 (0.007)	−0.000 (0.003)	0.004 (0.003)	0.017* (0.008)	0.007* (0.003)	0.006 (0.003)
21 or more crimes	0.027* (0.011)	0.006 (0.004)	0.014** (0.005)	0.018 (0.012)	0.009 (0.004)	0.011* (0.005)
Observations	907,412	907,412	907,412	884,788	884,788	884,788
R-squared	0.072	0.048	0.055	0.055	0.044	0.047
A. Age 14–18 years old
1–10 crimes	0.013** (0.005)	0.002 (0.002)	0.007** (0.002)	0.006 (0.006)	0.000 (0.002)	0.002 (0.002)
11–20 crimes	0.023** (0.008)	0.002 (0.003)	0.010** (0.003)	0.015 (0.008)	0.005 (0.003)	0.006 (0.003)
21 or more crimes	0.034** (0.011)	0.004 (0.004)	0.015** (0.005)	0.002 (0.013)	0.003 (0.004)	0.001 (0.005)
Observations	623,092	623,092	623,092	620,315	620,315	620,315
R-squared	0.093	0.053	0.066	0.062	0.042	0.046

Standard errors in parentheses clustered by census tract (**p<.01 *p<.05)

Notes: Student controls include: Black, Asian, White, ever low income, home language English, foreign born, limited English proficiency, special education and age in years. All models include grade fixed effects, Fitnessgram month fixed effects, and census by year fixed effects. Sample restricted to students with at least two consecutive Fitnessgram measures. AY 2010–2016.

Table A11:

Regression results, exposure to violent crime and fitness outcomes by gender, census tract by year fixed effects, NYC public school students, 2010–2016

	Girls				Boys
	Fitness index (1)	Aerobic fitness (2)	Strength (3)	Flexibility (4)	Fitness index (5)	Aerobic fitness (6)	Strength (7)	Flexibility (8)
A. Age 9–13 years old
1–10 crimes	−0.021** (0.007)	−0.002 (0.002)	−0.004 (0.002)	−0.005** (0.002)	−0.014* (0.007)	−0.004* (0.002)	−0.007** (0.002)	0.000 (0.002)
11–20 crimes	−0.032** (0.011)	−0.004 (0.003)	−0.004 (0.003)	−0.008* (0.003)	−0.013 (0.011)	−0.006* (0.003)	−0.002 (0.003)	−0.001 (0.003)
21 or more crimes	−0.021 (0.018)	−0.010* (0.004)	−0.008 (0.005)	0.001 (0.005)	−0.015 (0.018)	−0.012* (0.005)	−0.002 (0.005)	−0.001 (0.005)
Observations	885,898	885,898	885,898	885,898	862,497	862,497	862,497	862,497
R-squared	0.150	0.102	0.099	0.099	0.109	0.093	0.076	0.079
B. Age 14–18 years old
1–10 crimes	−0.012 (0.007)	−0.000 (0.002)	−0.003 (0.002)	−0.004 (0.002)	−0.017* (0.007)	−0.006** (0.002)	−0.003 (0.002)	−0.003 (0.002)
11–20 crimes	−0.020 (0.011)	−0.000 (0.003)	−0.004 (0.003)	−0.006* (0.003)	−0.030** (0.011)	−0.005 (0.003)	−0.006 (0.003)	−0.007* (0.003)
21 or more crimes	−0.016 (0.016)	−0.003 (0.004)	0.001 (0.005)	−0.006 (0.004)	−0.007 (0.016)	0.001 (0.005)	−0.004 (0.005)	−0.001 (0.004)
Observations	607,894	607,894	607,894	607,894	599,207	599,207	599,207	599,207
R-squared	0.129	0.057	0.091	0.088	0.087	0.058	0.064	0.058

Standard errors in parentheses clustered by census tract (**p<.01 *p<.05)

Notes: Student controls include: Black, Asian, White, ever low income, home language English, foreign born, limited English proficiency, special education and age in years. All models include grade fixed effects, Fitnessgram month fixed effects, and census tract by year fixed effects. Sample restricted to students with at least two consecutive Fitnessgram measures and fitness information on all indicators Fitness index ranges from 0 for children and adolescents that are in the healthy fitness zone for none the fitness measures to 6 for children and adolescents who are in the healthy fitness zone for all of the fitness measures. AY 2010–2016.