The Impact of Residing in a Gang Territory on Adverse Birth Outcomes: Evidence from Los Angeles

Abstract

Gang violence remains an ongoing crisis in many communities in the United States. This paper assesses the potential association of gang-occupied neighborhoods with birth outcomes. Adverse birth outcomes serve as a “barometer” of population health, denoting both poor conditions for human development and portending future public health concerns. We draw upon (1) Los Angeles County Vital Statistics Birth Records (2008–2012), (2) GIS information on gang territory boundaries, (3) LA city geo-coded crime data, and (4) the 2010 U.S. Census and 2006–2010 American Community Survey. We find an association between gang-occupied neighborhoods and adverse birth outcomes; however, this association is largely explained by other neighborhood socio-demographic characteristics, crime notwithstanding. We also find that gangland neighborhoods tend to exacerbate the effects of crime for all birth outcomes, but only significantly so for small for gestational age births. Lastly, gang co-residence, crime, and other neighborhood demographics explain a substantial portion of socioeconomic and racial/ethnic disparities in adverse birth outcomes. Gangland neighborhoods appear to be a novel contributor to both population health and health disparities. Future studies should address these relationships in a broad range of metropolitan settings, paying careful attention to causal linkages and moderating effects of gangs and crime.

Introduction

Gangs and gang violence are an epidemic affecting many communities. Although reliable estimates are hard to come by, there are upwards of 30,000 gangs in the United States, with approximately 850,000 members [1]. Though female members are likely undercounted, approximately 93% of gang members are male and members are disproportionately Black and Hispanic youth [2]. Though the majority of youth are in a gang less than three years [3], the collective impact on their health and that of their surrounding communities can be quite severe. Gang members are at three times the risk for mortality than age-matched peers [4]; approximately 80% of deaths of known gang members are due to external causes, most frequently assault [5]. Homicides extend beyond members, as estimates suggest gang involvement in at least 16% of homicides generally [6]. Metropolitan neighborhoods where gang membership is concentrated also have higher rates of gun assault [7], leaving residents at risk for victimization. Even absent assault, community members may simply fear the risk associated with the presence of weapons, as other work finds a strong impact of gang membership on weapon possession [8].

Although gangs have spread into rural and suburban areas in recent decades, the majority of gang-related crime continues to occur within metropolitan areas [8]. Currently, Los Angeles County (LAC) is home to one of the largest gang populations in the world, numbering approximately 120,000 gang members aligned with nearly 1300 street gangs who are not confined to the stereotypical streets of South LA, but who also occupy suburban and beach areas of LAC (http://www.lapdonline.org/get_informed/content_basic_view/23466). Los Angeles has long been thought of as the epicenter of gang activity nationwide, with street gangs occupying neighborhoods in seven counties in Southern California and having migrated to hundreds of regions around the United States. The emergence of gang activity in war-torn countries in Latin America has also been tied to the large-scale deportation and emigration of gang members from Southern California. Gangs in LA account for more than 40% of all homicides (http://www.laalmanac.com/crime/cr03x.php) and can have a dramatic impact on communities, rendering residents fearful, stressed, and crowded out of social spaces making it difficult to walk, exercise, interact, and socialize with neighbors [9].

Living in gang-occupied (heretofore “gangland”) neighborhoods may affect health through a variety of potential mechanisms, including an increased likelihood of directly experiencing violence, increased risk of exposure to crime and violence, increased fear of crime and lower perceived safety, stress, and reduced social interaction and exercise/play [10]. Gangland co-residence may be associated with each of these factors, and its potential effects may not be reducible to neighborhood crime. Although the direct effects of violence, gang involvement, and crime have been thoroughly studied as a social problem [11], no current study has considered contextual impacts of living in gangland neighborhoods on birth outcomes. This paper seeks to assess the potentially unique relationship between gangland residence and a barometer of population health—adverse birth outcomes—using unique combinations of data.

Birthweight as a Contemporaneous Barometer of Population Health

We argue that birth outcomes are an appropriate measure of population health with which to study the effects of gangland residence because they are responsive to socioeconomic conditions and have far-reaching implications for health over the life course [12]. Birth outcomes are extraordinarily important for the health and wellbeing of infants and children [13, 14]. Poor birth outcomes are also associated with greater childhood learning problems and worse cognitive outcomes [15–17]. At the same time, the long arm of infant health reaches far into adulthood, affecting health, longevity, and economic outcomes [13, 14, 18]. Birth outcome disparities also mirror larger social inequalities [19], particularly with respect to socioeconomic status [20, 21]. Black-White disparities in birth outcomes are well documented [14], although Hispanic infants appear to be an exception as their better birth outcomes tend to belie their poorer socioeconomic status [22] due to healthy migrant selection and immigrant social networks [23]. Birth outcomes are also affected by the broader socioeconomic environment, including neighborhood poverty, income, segregation, residential instability, and violent crime [24]. Because of the responsiveness to dynamic social conditions and the mirroring of social inequalities, birth outcomes can be seen as a critical social barometer of population health.

Gang Activity/Presence and Health

The impacts of crime on health are widely documented [25]. The direct effects of crime victimization (e.g., homicide and injuries) can impact mental health and physical and cognitive functioning and may increase maladaptive stress behaviors [26]. Crime or violence-related fear and stress can likewise impact the psychological wellbeing of mothers during pregnancy [27] and trigger stress responses which can be harmful for fetal growth or reduce gestational age [28, 29]. Evidence for the underlying mechanisms is still being accumulated, but proposed stress pathways include impacting growth-related hormones, blood flow, elevating blood pressure, and risk of infection [30]. Evidence from discrete and repeated terror attacks implicates the stress pathway and specifically the restriction of intrauterine growth in early pregnancy [31]. A gang-related stress response during pregnancy seems likewise conceivable, as mothers report feeling very unsafe and utilize coping behaviors such as isolation or vigilance in gang areas [32]. Additionally, victimization or fear of crime may impact adverse health behaviors (e.g., smoking and drinking) that may be especially harmful for fetal development or may preclude a mother’s use of prenatal care, although the evidence is mixed on this point [33]. With respect to poor birth outcomes, there are established associations with perceived crime [24] and exposure to violence [28–30]. However, most of this work does not disentangle generalized crime or violence from that which is gang related, although higher rates of crime and violence may be suggestive of gang violence, and in some studies, measures of crime may be primarily driven by gangs or drug trafficking [29]. Certainly, the impact of violence and crime may be compounded in gangland communities.

Gang violence directly affects gang members as well as their romantic partners. Recent research shows that gang membership increases the risk of being victimized and this is only partially due to selection into gang affiliation [34]. Gang-related homicides disproportionately impact Non-Hispanic Blacks, Hispanics, males, and those in late adolescence and early adulthood (15–29) [35] and closely track known membership patterns [36]. Gang members who survive the high rate of homicide that confronts them at increasingly younger ages are still at greater risk for a host of poor outcomes as adults, including drug abuse, drug dependence, and poor general health [37]. Membership also increases the prevalence of mental disorders (e.g., anxiety disorders and psychosis) and other adverse outcomes such as substance misuse [38]. Studies have also shown that gang involvement, either self-involvement or romantic involvement with a member, is associated with greater sexual promiscuity [39], lower condom use among young women [40], and higher risk of pregnancy among Latina adolescents [40, 41]. If resulting pregnancies are, in turn, at greater risk for poor birth outcomes, due to close proximity to violence, drugs, and other threats, this may operate as yet another mechanism through which gang territories impact births. While much is known about the effects of crime and violence on health and while some research documents the health implications of gang membership, less is known about the larger community effects of having to live among gang members, particularly for women.

Gang activity can have far-reaching effects on communities that extend beyond crime, due to the nuances of gang activity and the nature of crimes that make gang-related crime very different from typical crime. First, gang-related crimes are unique in that they are more likely to occur outside, during daylight hours, are more likely to involve a firearm, and include a uniquely terror-inducing type of crime—drive-by shootings [42]. Second, outdoor disputes over gang territory and the openness of the drug trade—often occupying scenic communal spaces—are unique in gang-occupied neighborhoods and can result in increased exposure and fear of crime, stress among residents, restriction of mobility, and limitation of neighborhood social interaction [43]. Third, gang-related violence acts as a catalyst for additional and more frequently lethal crimes, relative to incidents of non-gang violence [44]. The mere presence of gang members, violence and crime notwithstanding, is a public nuisance with unknown health implications. Fourth, crime remains highly underreported [45, 46], particularly in areas with high levels of crime; in fact, gangland residence may be one of the factors related to the underreporting of crime [47]. Witness intimidation is rife in gangland communities such that many crimes remain unsolved due to norms that limit talking to police (i.e., “snitching”). Violence and threats can be perpetrated against community members that choose to talk to police and/or testify in court—in violation of norms established initially to protect the community from racist policing, but which now tend to protect violent and homicidal criminals [48].

Neighborhoods with gang activity, high levels of crime, and/or other forms of social disorder are also often those with greater levels of socioeconomic disadvantage, which is likewise tied to poor birth outcomes [49, 50]. Given the scant evidence that exists regarding the relationship between gang activity exposure and birth outcomes, it is unclear whether gang activity in neighborhoods is detrimental to birth outcomes, independent of its association with crime and socioeconomic disadvantage, or whether it exacerbates the effects of these more established contextual risk factors.

Current Study

Gangland residence has health implications for a number of reasons that are directly related to increased crime. First, violent crime affects community health through increased mortality and an increased rate of injuries. Second, fear of crime has known public health effects that are not reducible to objective crime rates. Third, indirect exposure to violence also has known public health implications. Witnessing crimes, hearing second-hand about local crimes, and experiencing crime have all been shown to be deleterious. However, independent of relationships with crime rates, the health implications of gangland residence have never been directly studied. We believe this is a major oversight due to (a) the uniqueness of gang-related crimes, (b) the fact that gang activities can limit social interaction, and (c) nuances limiting collection of local crime data. In short, while the direct effects of gang violence are obvious (e.g., threats, injuries, and homicide) and the risks associated with gang membership are well documented—the indirect effects on population health are less well known. Even for those not participating in gang activity—the vast majority of gangland residents—living in neighborhoods in which gang members are a daily presence and gang conflict is rife may have unknown, but potentially far-reaching, impacts on public health. Gangland residence may also be an unexplored factor for poorer birth outcomes among less educated, non-White mothers who already experience other forms of economic disadvantage and discrimination [51]. Additionally, given the known race and education-specific patterns of gang membership and housing segregation [51], such mothers may also be more physically and/or socially proximate to gang members within neighborhoods. Thus, we also query whether gangland residence explains racial/ethnic and educational inequalities in birth outcomes.

Our hypotheses are as follows:

• Hypothesis 1 (H1)

  Gangland residence will have an independent effect on adverse birth outcomes, net of neighborhood crime, and socio-demographics.

• Hypothesis 1 (H2)

  Gangland residence will exacerbate the effect of neighborhood-level crime and/or neighborhood-level socioeconomic disadvantage.

• xHypothesis 1 (H3)

  Gangland residence and neighborhood-level crime will partially account for individual-level educational and racial/ethnic disparities in adverse birth outcomes.

Data and Methods

Data Sources

Our data for this project come from four primary sources. First, our core data are restricted Los Angeles Vital Statistics Birth Records (2008–2012). These data contain all live births in Los Angeles City for every calendar year from 2008 to 2012 and include all information available on a standard birth certificate form. Although vital statistics data are not error-free and medical records remain the gold standard for reporting of birthweight, several validation studies show that birth certificate data are highly accurate in reflecting birthweight and concordance between medical records and birth certificates ranges from 87 to 100% [52, 53]. While the gold standard for estimating gestation length remains the last menstrual period and date of birth, these data are less reliable [54]. That said, our goal is not to estimate prevalence of these outcomes, and as long as poor recall is not correlated with gangland residence, then our results should remain unbiased. Finally, each home address record was geo-coded to create a linkable census-based code (e.g., census tract).

Second, we use GIS gang territory files compiled by a gang researcher from nearly a decade of collaborative research with the LA Police Department (LAPD) and the LA County Sherriff’s Department (LACSD) [55]. The mapping of gang territories in Los Angeles began in 1972 with the LA Times’ publication of a map of the most active Black street gangs according to the 77th Division of the LAPD. This map was reproduced in Mike Davis’ seminal historical ethnography of Los Angeles, City of Quartz. Steve Jablonsky, a Parole Officer, updated these maps to include all gangs, relying on information from intelligence fieldwork, and Mark Poirier updated them using gang graffiti to redefine boundaries. More recently, Alonso [56] updated these maps in 1999 and then again in 2010, using data from the original maps, combined with interviews of active gang members (who tended to exaggerate their boundaries) and young community residents (who tended to have more accurate knowledge of boundaries) as well as a thorough scouring of the streets for aggressive graffiti. The results of these efforts crystallized into the map shown in Fig. 1, where gang territories current to 2010 are overlaid onto the census tracts that constitute municipal Los Angeles. Residence in a gangland neighborhood was determined by overlaying the gang maps on top of the coordinates for each birth in order to determine whether home addresses for each birth lay inside or outside of gang boundaries.

Fig. 1.

Gang boundaries overlaid on LA metro area census tracts

Third, the crime data come from the LA Times’ Mapping LA and are updated on a weekly basis as data are collected from official sources. Crimes are assembled from the LAPD and LACSD official crime data that is ultimately reported to the FBI’s Uniform Crime Reports and from the LA City Planning Department’s census-based estimates and include date-stamped incidents of aggravated assault, burglary, grand theft auto, homicide, rape, robbery, theft, and theft from vehicle. Aggravated assault, homicide, rape, and robbery are defined as violent crimes; burglary, grand theft auto, theft, and theft from vehicle are defined as property crimes. We aggregated crimes to the level of the census tract, as crime is best captured as a neighborhood-level attribute, rather than a county-level trait, at least for the study of birth outcomes. Given that local crime can vary unpredictably over short time spans [57], we pooled all violent crimes for the entire study period to improve the robustness of our results.

Fourth, we used 2010 U.S. Census data and 2006–2010 American Community Survey data at the census-tract level, including median household income, % of population in poverty, % female-headed households, and % of population with less than a high school education, to develop an index of neighborhood socioeconomic disadvantage. These characteristics were also appended to each birth record.

Sample

Our sample comprises 143,223 births, restricted to all live, singleton births to mothers residing in the City of Los Angeles between the years 2009 and 2012. While we had access to 2008 birth certificate data, births occurring in 2008 were excluded because we did not have access to crime data for 2008. Less than 1% of births occurring between 2009 and 2012 had missing crime values and were dropped from the sample as well.

Additionally, we applied listwise deletion to all births with missing values for mother’s education, mother’s age, mother’s nativity, sex of child, parity, or adequacy of prenatal care. For all other control variables, we created flags for missing values and included the flags in our models so that the records with those values would not be dropped. Note that we do not find that our results change if we apply listwise deletion to all control variables, however. When listwise deletion is applied, an additional 20,450 births are dropped from the sample, but our results are robust to the exclusion of births with any missing demographic or pregnancy behavior values (results available upon request).

Measures

Our key dependent variables include the following adverse birth outcomes: (1) a dichotomous marker of low birthweight (<2500 g), (2) a dichotomous marker of preterm birth (<37 weeks), and (3) a dichotomous marker of small for gestational age (<10th percentile of weight for gestational age and gender). All three are robust markers of future morbidity and mortality, although they may have slightly different correlates and potentially discrete etiologies. Births with extreme values for birthweight (<500 g or >5500 g) or gestation length (<22 weeks or ≥45 weeks) were excluded.

Our key independent variables of interest are gangland residence, neighborhood socioeconomic disadvantage, and neighborhood crime. All neighborhood measures are proxied by census tracts. We measured gangland residence by overlaying the GIS gang territory map onto mothers’ geo-coded addresses. We coded mothers as gangland = 0 if their address was not within the boundaries of a gang territory and gangland = 1 if their address was within the boundaries of a gang territory. To measure neighborhood socioeconomic disadvantage, we developed a normalized, census-tract level index of median household income, % of population in poverty, % female-headed households, and % of population with less than a high school education, measured during the 2010 U.S. Census (for all births that occurred from 2008 to 2012). For our analyses, we divide the index into tertiles and use the cut points to code census tracts as low poverty, mid poverty, or high poverty. Our measure of crime is a census-tract level count of violent crimes for the years 2009–2012, provided by the LA Times’ Mapping LA data project.

We also control for a series of individual-level demographic characteristics, derived from the Vital Statistics birth records. These controls include sex of child [male (omitted), female], type of payer for birth [private insurance (omitted), government, self-pay, other, unattended, none, no charge, missing], mother’s race/ethnicity [Non-Hispanic White (omitted), Non-Hispanic Black, Non-Hispanic Other, Hispanic, Missing], father’s race/ethnicity [Non-Hispanic White (omitted), Non-Hispanic Black, Non-Hispanic Other, Hispanic, Missing], mother’s age [<20 (omitted), 20–34, 35–39, 40+], education level [<high school degree (omitted), high school degree, BA or more], nativity [foreign born (omitted), U.S. born], birth parity based on the Kleinman and Kessel (1987) index [first birth (omitted), low, high], and pre-pregnancy BMI [underweight (omitted), normal, overweight, obese, missing].

Lastly, we control for a series of pregnancy behavior indicators derived from the Vital Statistics birth records, including WIC participation [no (omitted), yes, missing], mother’s smoking behavior during pregnancy [did not smoke (omitted), smoked, missing], weight gain during pregnancy [<16 lbs. (omitted), 16–40 lbs., >40 lbs., missing], and adequacy of prenatal care [inadequate (omitted), adequate, intermediate, adequate plus].

Statistical Analysis

We examine relationships between birth outcomes and census-tract- and individual-level characteristics using logistic regression models, adjusted for the non-independence of residents of the same census tract through clustered standard error estimation. Variables at the individual level include demographic characteristics (Demog) and pregnancy behavior indicators (Preg) from the birth certificate data as well as a dichotomous indicator of gangland residence (Gang). Variables at the census-tract level (CT) include neighborhood socioeconomic disadvantage (NSES) and the violent crime count for 2009–2012 (Crime). We first describe the broad associations of gang territories to local NSES and crime distribution using several figures. Next, we estimate models of the effect of gangland residence on the relative odds of an adverse birth outcome, controlling for demographic characteristics and neighborhood socioeconomic disadvantage, with standard errors clustered at the census-tract level. Specifically, we fit an equation where AB_ij = 1 indicates an infant has an adverse birth outcome and AB_ij = 0 indicates an individual does not. Gang_ij is the gangland residence indicator for individual i in census tract j where i = 1,2,…,n_j (number of individuals within CT = j), and j = 1,…,J (number of CTs). Below we demonstrate our model sequence for specifications without (Eqs. 1.1–1.4) and with (Eq. 2.1–2.4) interaction terms. These equation (Eq.) numbers will be used in our results and tables for easy identification.

$$\mathit{\text{logit}}\left({\mathit{AB}}_{\mathit{ij}}=1\right)={\beta }_{0j}+{\beta }_1{\mathit{\text{Gang}}}_{\mathit{ij}}+{\beta }_2{\mathit{\text{Demog}}}_{\mathit{ij}}+e_{\mathit{ij}}$$ 1.1

$$\mathit{\text{logit}}\left({\mathit{AB}}_{\mathit{ij}}=1\right)={\beta }_{0j}+{\beta }_1{\mathit{\text{Gang}}}_{\mathit{ij}}+{\beta }_2{\mathit{\text{Demog}}}_{\mathit{ij}}+{\beta }_3{\mathit{\text{NSES}}}_{\mathit{ij}}+e_{\mathit{ij}}$$ 1.2

$$\mathit{\text{logit}}\left({\mathit{AB}}_{\mathit{ij}}=1\right)={\beta }_{0j}+{\beta }_1{\mathit{\text{Gang}}}_{\mathit{ij}}+{\beta }_2{\mathit{\text{Demog}}}_{\mathit{ij}}+{\beta }_3{\mathit{\text{NSES}}}_{\mathit{ij}}+{\beta }_4{\mathit{\text{Crime}}}_{\mathit{ij}}+e_{\mathit{ij}}$$ 1.3

$$\mathit{\text{logit}}\left({\mathit{AB}}_{\mathit{ij}}=1\right)={\beta }_{0j}+{\beta }_1{\mathit{\text{Gang}}}_{\mathit{ij}}+{\beta }_2{\mathit{\text{Demog}}}_{\mathit{ij}}+{\beta }_3{\mathit{\text{NSES}}}_{\mathit{ij}}+{\beta }_4{\mathit{\text{Crime}}}_{\mathit{ij}}+{\beta }_5{\mathit{\text{Preg}}}_{\mathit{ij}}+e_{\mathit{ij}}$$ 1.4

$$\mathit{\text{logit}}\left({\mathit{AB}}_{\mathit{ij}}=1\right)={\beta }_{0j}+{\beta }_1{\mathit{\text{Gang}}}_{\mathit{ij}}+{\beta }_2{\mathit{\text{NSES}}}_{\mathit{ij}}+{\beta }_3{\mathit{\text{Demog}}}_{\mathit{ij}}+{\beta }_4{\mathit{\text{Crime}}}_{\mathit{ij}}+{\beta }_6{\mathit{\text{Gang}}}_{\mathit{ij}}{\mathit{\text{NSES}}}_{\mathit{ij}}+e_{\mathit{ij}}$$ 2.1

$$\mathit{\text{logit}}\left({\mathit{AB}}_{\mathit{ij}}=1\right)={\beta }_{0j}+{\beta }_1{\mathit{\text{Gang}}}_{\mathit{ij}}+{\beta }_2{\mathit{\text{NSES}}}_{\mathit{ij}}+{\beta }_3{\mathit{\text{Demog}}}_{\mathit{ij}}+{\beta }_4{\mathit{\text{Crime}}}_{\mathit{ij}}+{\beta }_5{\mathit{\text{Preg}}}_{\mathit{ij}}+{\beta }_6{\mathit{\text{Gang}}}_{\mathit{ij}}{\mathit{\text{NSES}}}_{\mathit{ij}}+e_{\mathit{ij}}$$ 2.2

$$\mathit{\text{logit}}\left({\mathit{AB}}_{\mathit{ij}}=1\right)={\beta }_{0j}+{\beta }_1{\mathit{\text{Gang}}}_{\mathit{ij}}+{\beta }_2{\mathit{\text{NSES}}}_{\mathit{ij}}+{\beta }_3{\mathit{\text{Demog}}}_{\mathit{ij}}+{\beta }_4{\mathit{\text{Crime}}}_{\mathit{ij}}+{\beta }_6{\mathit{\text{Gang}}}_{\mathit{ij}}{\mathit{\text{Crime}}}_{\mathit{ij}}+e_{\mathit{ij}}$$ 2.3

$$\mathit{\text{logit}}\left({\mathit{AB}}_{\mathit{ij}}=1\right)={\beta }_{0j}+{\beta }_1{\mathit{\text{Gang}}}_{\mathit{ij}}+{\beta }_2{\mathit{\text{NSES}}}_{\mathit{ij}}+{\beta }_3{\mathit{\text{Demog}}}_{\mathit{ij}}+{\beta }_4{\mathit{\text{Crime}}}_{\mathit{ij}}+{\beta }_5{\mathit{\text{Preg}}}_{\mathit{ij}}+{\beta }_6{\mathit{\text{Gang}}}_{\mathit{ij}}{\mathit{\text{Crime}}}_{\mathit{ij}}+e_{\mathit{ij}}$$ 2.4

The individual error e_ij is the respondent error that is assumed to be normally distributed with mean 0 and variance σ². In Eq. 1.2, we include a measure of the crime count (Crime_ij) at the census-tract level; in Eq. 1.3, we include a term for the interaction of gangland residence and neighborhood socioeconomic disadvantage. Lastly, in Eq. 1.4, we include our individual-level indicators of pregnancy behaviors, which we conceptualize as potential mechanisms through which gangland residence is likely to influence birth outcomes. Eqs. 1.1–1.4 will be used to test Hypotheses 1 and 3 (H1 and H3). The specification of interactions terms in Eqs. 2.1–2.4 will be used to test our second research hypothesis (H2).

Results

We describe the characteristics of births in Table 1. First, 62.48% of mothers live in gangland territory. We see that overall 5.58% of births are low birthweight (LBW), 9.06% are preterm birth (PTM), and 11.21% are small for gestational age (SfGA). Births taking place outside of gangland territories tend to have a smaller chance of adverse birth outcomes than those which take place in ganglands. For example, while only 4.87% of births outside of ganglands are LBW, 6.0% within ganglands are LBW. Most births in Los Angeles are happening for Hispanic parents. Overall, 62.90% of mothers and 56.10% of fathers are Hispanic. However, the share of Hispanic parents is much smaller outside of gangland territory. Outside of gangland areas, the largest share of births is happening for White parents (38.78% mothers and 40.78% fathers). Whether or not the parents live in gangland territories reveals other important differences. For example, mothers tend to be more educated outside of ganglands; 47.58% of mothers outside of ganglands are college educated while 13.68% of mothers in these areas have less than a high school degree. When it comes to pregnancy behaviors, there are less notable differences between parents in gangland verses non-gangland areas. For example, while a higher share of parents in gangland areas were rated with inadequate care, 9.80% compared to 5.65% in non-gangland areas, slightly more mothers smoked in non-gangland areas, 0.67% compared to 0.48% in gangland areas. Lastly, gangland areas also appear associated with other neighborhood characteristics. Mothers who live in gangland areas experience greater shares of high poverty (50.93%, compared to 6.24% for mothers who live outside of gangland areas) and higher rates of violent crime (301.66, compared to 143.74 in non-gangland areas).

Table 1.

Sample characteristics, Los Angeles live births, 2008–2012, by gang territory status

Demographic characteristics	No gang (53,739)	Gang (n = 89,484)	Total (n = 143,223)
% of population:	37.52	62.48	100
Adverse birth outcomes
Low birthweight	4.87	6.01	5.58
Preterm birth	7.80	9.82	9.06
Small for gestational age	10.43	11.68	11.21
Neighborhood socioeconomic disadvantage
Low poverty	70.41	8.64	31.81
Mid poverty	23.35	40.44	34.03
High poverty	6.24	50.93	34.16
Average violent crime count	54.86	125.52	98.40
Demographic controls
Mother’s education
No HS degree	13.68	42.81	31.88
HS degree	38.75	47.03	43.92
BA or more	47.58	10.16	24.20
Mother’s age
<20 years	4.01	11.01	8.38
20–34 years	67.17	73.10	70.88
35–39 years	22.23	12.48	16.14
40+ years	6.58	3.41	4.60
Mother’s race/ethnicity
NH White	38.78	6.41	18.55
NH Black	5.43	9.40	7.91
NH Other	17.38	5.92	10.22
Hispanic	37.93	77.90	62.90
Missing	0.48	0.37	0.41
Father’s race/ethnicity
NH White	40.78	6.51	19.37
NH Black	6.17	7.91	7.25
NH Other	14.16	5.00	8.44
Hispanic	33.64	69.59	56.10
Missing	5.25	11.00	8.84
U.S. born	53.17	43.66	55.80
Sex of child: female	48.22	48.84	48.61
Parity
First birth	48.25	36.52	40.92
Low	48.94	55.29	52.91
High	2.80	8.20	6.17
Payer for birth
Private insurance	60.59	21.37	36.08
Government	34.79	76.27	60.70
Self-pay	2.18	0.90	1.38
Other	2.21	1.28	1.62
Unatt, none, no charge	0.08	0.08	0.08
Missing	0.16	0.11	0.13
Body mass index
Underweight	5.43	3.26	4.07
Normal	57.71	41.97	47.88
Overweight	20.54	27.36	24.80
Obese	13.70	24.41	20.39
Missing	2.62	2.99	2.86
Pregnancy behaviors
WIC participation
Yes	39.58	84.11	67.40
Missing	0.38	0.29	0.33
Smoked during pregnancy
Yes	0.67	0.48	0.55
Missing	0.84	3.60	2.57
Weight gain during pregnancy
<16 lbs	10.23	18.46	15.37
16–40 lbs	68.53	63.81	65.58
>40 lbs	17.90	14.01	15.47
Missing	3.34	3.72	3.58
Adequacy of prenatal care
Inadequate	5.65	9.80	8.24
Adequate	2.53	3.48	3.13
Intermediate	33.81	37.34	36.01
Adequate Plus	58.01	49.38	52.62

We present the distribution of gang territories and violent crime as well as gang territories and socioeconomic status in Fig. 2. This figure shows that gang territories are spread across the city, though they tend to cluster in select parts including in the central/eastern portions, which includes areas in around the downtown as well as the “South Central” section of the city; the southern reach of the city, near Long Beach; and select northern parts of the city, in the San Fernando area. Both violent crime and socioeconomic status are disproportionately clustered in the central section of the city. There is a fair amount of overlap between violent crime and the presence of gangs, something suggested by Table 1, although there are many instances where violent crime can be found outside of gang territories and vice versa. This spatial pattern suggests a moderate association between ganglands and crime. However, we also find that gangland territories closely overlap with areas with mid poverty and high poverty.

Fig. 2.

Gang boundaries overlaid on census-tract level violent crime

We report the findings of our logistic regression analysis in Table 2. For ease of interpretation, we only directly report on our focal predictors, but the full models are available upon request. We find in our base models that residence in ganglands is significant and positive for all birth outcomes; that is, gangland residence adversely affects birth outcomes (Eq. 1.1). The strength of these associations, however, varies by outcome; the chances of LBW increase by 11.5% for mothers in ganglands (1–1.115 × 100), 7.3% for PTB, and 4.3% for SfGA. However, the full significance (p<0.05 or less) of these associations is nearly lost with the introduction of the other neighborhood-level controls (Eq. 1.2). The exact association of the other neighborhood-level controls to birth outcomes depends on the outcome. For example, SfGA is the only birth outcome to have a significant, albeit weak, association with low birthweight. Next, all outcomes have an association with SES, but only LBW and PTB have associations that exist independent of neighborhood violent crime (Eq. 1.3). While PTB has a more robust association with high poverty than middle poverty, LBW has more robust associations with middle poverty than high poverty.

Table 2.

Gangland residence and adverse outcomes

Variables	Low birthweight				Preterm birth				Small for gestation
	1.1	1.2	1.3	1.4	1.1	1.2	1.3	1.4	1.1	1.2	1.3	1.4
Gangland	1.115***	1.065+	1.065+	1.066+	1.073**	1.032	1.032	1.025	1.043*	1.009	1.008	1.010
	(0.034)	(0.037)	(0.037)	(0.037)	(0.025)	(0.027)	(0.027)	(0.027)	(0.022)	(0.024)	(0.024)	(0.024)
Neighborhood disadvantage
Low poverty (reference)
Mid poverty		1.05*	1.097*	1.126**		1.068*	1.058+	1.116***		1.041	1.026	1.019
		(0.045)	(0.044)	(0.046)		(0.032)	(0.031)	(0.035)		(0.029)	(0.029)	(0.029)
High poverty		1.125**	1.099+	1.078		1.113**	1.081*	1.071+		1.101**	1.054	1.033
		(0.051)	(0.054)	(0.053)		(0.037)	(0.040)	(0.040)		(0.035)	(0.036)	(0.036)
Violent crime count			1.0002	1.0002			1.0002+	1.0002+			1.0004**	1.0003**
			(0.000)	(0.000)			(0.000)	(0.000)			(0.000)	(0.000)
Demographic controls	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Pregnancy behavior indicators	No	No	No	No	No	No	No	No	No	No	No	Yes
Constant	0.052***	0.051***	0.050***	0.100***	0.092***	0.090***	0.089***	0.133***	0.052***	0.051***	0.194***	0.334***
	(0.004)	(0.004)	(0.004)	(0.009)	(0.006)	(0.006)	(0.006)	(0.011)	(0.004)	(0.004)	(0.011)	(0.023)
Observations	143,223	143,215	143,215	143,215	143,223	143,215	143,215	143,215	143,223	143,215	143,215	143,215

Note: coefficients are reported as odds ratios. Standard errors (clustered by census tract) are reported in parentheses. Only singleton births are included in these analyses. Results for control variables are not reported in the table and can be requested from the lead author. Models include the following demographic controls: child’s sex, type of insurance for child’s birth, mother and father’s race/ethnicity, mother’s age, education level, nativity, parity, and BMI. Pregnancy behavior indicators include WIC participation, mother’s smoking behavior during pregnancy, weight gain during pregnancy, and adequacy of prenatal care

^***p < 0.001, ^**p < 0.01, ^*p < 0.05, ⁺p < 0.10

We also interact ganglands with neighborhood SES to determine if they have any shared association with adverse birth outcomes (Eq. 2.1–2.2). We find that none of these interactions are significant (results not shown, available upon request). In short, the association of gangland residence to birth outcomes can largely be explained by other neighborhood characteristics, and any significant effects are largely additive. Finally, we test whether gangland residence compounds the observed crime effects (Eq. 2.3–2.4) and find that this is the case for only SfGA (results not shown, available upon request). That is, while crime and gangland residence are both associated with adverse birth outcomes (although only crime is statistically significant), we find that gangland residence exacerbates the crime effects for SfGA, although to a marginally small extent.

Finally, we examine whether adverse birth outcome disparities are explained by our key variables (Figs. 3 and 4). Model sequence largely follows the progression of variables added in Table 2, except that we specify a baseline model with “disparity” variables only in order to assess its reduction over subsequent models We explore both race/ethnicity and educational disparities in sequential models. We present only the disparity odds ratios in Figs. 3 and 4. Exploring racial/ethnic disparities in low birthweight (see Fig. 3), we note a baseline disparity of 2.66 for NH Blacks (166% more likely to experience low birthweight) and 1.39 for Hispanics (39% more likely), relative to NH Whites, our reference group. Controlling for gangland co-residence in model 2 results in an OR reduction to 2.47 for NH Blacks and 1.29 for Hispanics. Further controls for crime in model 3 reduce ORs to 2.37 and 1.27 for NH Blacks and Hispanics, respectively. Controlling for neighborhood socioeconomic status in model 4 further reduces the ORs to 2.29 and 1.22 respectively. Adding individual-level controls for maternal socio-demographic variables (age, parity, insurance status, education, nativity, parity, BMI, and child sex) further reduces the NH Black OR to 1.80 and reduces the Hispanic OR to statistical non-significance. Controlling for pregnancy behaviors (WIC participation, maternal smoking, weight gain, and prenatal care adequacy) had little effect on either race/ethnicity estimate. These general patterns also held for both preterm birth and small for gestational age, although the baseline and residual racial/ethnic disparities were slightly smaller in both instances.

Fig. 3.

Odds ratios for racial/ethnic disparities in birth outcomes (relative to Non-Hispanic White). Note: lighter-colored bars denote odds ratios with a p value greater than 0.05. Models: 1: race_ethnicity, 2: race_ethnicity + gang, 3: race_ethnicity + gang + crime, 4: race_ethnicity + gang + crime + ses_index, 5: race_ethnicity + gang + crime + ses_index + demographic characteristics, 6: race_ethnicity + gang + crime + ses_index + demographic characteristics + pregnancy behaviors

Fig. 4.

Odds ratios for educational disparities in birth outcomes (relative to less than high school). Note: lighter-colored bars denote odds ratios with a p value greater than 0.05. Models: 1: race_ethnicity, 2: race_ethnicity + gang, 3: race_ethnicity + gang + crime, 4: race_ethnicity + gang + crime + ses_index, 5: race_ethnicity + gang +crime + ses_index + demographic characteristics, 6: race_ethnicity + gang + crime + ses_index + demographic characteristics + pregnancy behaviors

Turning next to the educational disparities presented in Fig. 4, we first estimate baseline advantages for having a college degree relative to not having a high school diploma. College graduates are 24% less likely to give birth to a low birthweight infant, and this effect is statistically significant. Controlling for gangland co-residence reduces this effect to 17% less likely and crime further reduces this advantage to 13%. Neighborhood SES reduces the advantage to 9%, but controlling for demographic characteristics suppresses this effect, expanding it back to 24% and pregnancy behaviors further expand the advantage to 27% lower likelihood. Although the general pattern of explanation (net of gangland, crime, and neighborhood SES) and then suppression (demographics and pregnancy behaviors) holds for both preterm birth and small for gestational age, the educational disparities are more pronounced for preterm birth and less pronounced for small for gestational age, relative to low birthweight.

Conclusions

Gang activity has far-reaching effects on communities that extend beyond crime, due to the nuances of gang activity and the nature of crimes that make gang-related crime very different from typical crime. First, they are more likely to occur outside during daylight hours and more likely to involve a firearm than other types of similar crimes (e.g., robberies and homicides). Second, disputes over territory and drug markets often occur in communal spaces, increasing stress in these neighborhoods. Although research has been done on the direct effects of gang activities on physical and mental health and premature mortality, no extant research has studied the larger contextual effect of living in gangland territories. Our study is the first to explore the possible independent effect of gangland co-residence on a malleable measure of population health: adverse birth outcomes. We also explore whether violent crime might have compounding stress effects in gang territories. Finally, we explore the extent to which gangland co-residence and crime exposure contribute to educational and racial/ethnic disparities in health.

Our results confirm that the probability of being born low birthweight is heightened in gangland neighborhoods, and this is partially, but not fully, accounted for by neighborhood crime and neighborhood disadvantage. The deleterious effects of gangland co-residence are also fully accounted for by neighborhood disadvantage (and to a small extent, crime) in the case of preterm births and small for gestational age, however. Although it is fairly clear that gangs do not modify the neighborhood SES effects, it does appear that gang residence exacerbates the crime effects for small for gestational age birth outcomes. It is not entirely obvious why the results differ for SfGA, but some scholars prefer using this measure because it isolates those infants who are underweight, irrespective of gestation and have not met their growth potential, and if the strongest pathway is a stress-specific restriction of intrauterine growth, this outcome may be the most consistent with the stress pathway.

Furthermore, gangland co-residence, crime exposure, and neighborhood socioeconomic disadvantage account for a very large percentage of educational and racial/ethnic differences in birth outcomes, underscoring how neighborhood segregation contributes to population health disparities, particularly for adverse birth outcomes which have far-reaching consequences for population health and health disparities.

However, given that our analysis is cross-sectional, it is impossible to identify the causal factors involved and difficult to assess whether compounded neighborhood disadvantage led to gang activity or whether gang activity drove out socioeconomically advantaged residents. As such, it is still plausible that gangland neighborhoods may be deleterious to population health, but the mechanism operates through a departure of more advantaged neighbors.

Our results show mixed but largely negative effects of gangland co-residence on birth outcomes, but further research is warranted. Specifically, future studies can attempt to parse out gang activity longitudinally to uncover whether specific events or protracted crime exposure may create a stressful environment that has effects on the larger community, beyond the gang members themselves. The potential impact on infant mortality and selective fertility should also be examined in more detail in the future.

In addition, given the widespread application of gang injunctions—and their assorted impacts on curtailing civil liberties simultaneous with short-term crime reductions—special attention should be paid to uncovering the long-term effects of injunctions on crime and population health. Los Angeles specifically engaged in a protracted effort to rid the streets of gang members by implementing broad gang injunctions in several of the gangland areas studied here. These injunctions, including the designation of precise physical boundaries required by law, could be used to explore the effects of changing gang dynamics in neighborhoods over time.

Similarly, it has been well documented that gangs from Los Angeles have spread to other communities around the United States. It would then be possible to more clearly draw lines between poorer communities that were affected by this gangland “diaspora” and those that were not in order to more directly establish causal relationships or at least to explore longitudinal change and the broader public health impacts.

Certainly, our results point to the significance of a deleterious contextual effect of gangland neighborhoods, but this is just a single step in the direction of uncovering more complex patterns and establishing causality. Future efforts will require mining public resources and merging data from multiple sources in order to untangle some of the complex relationships that we note here. At the same time, public health interventions may need to incorporate gathering information about the public health that includes more detailed knowledge of contextual effects such as gangland co-residence and its associated stressors.

What is already known on this subject?

➢ Previous studies have explored the impact of several contextual factors on adverse birth outcomes, including crime and neighborhood socioeconomic disadvantage. At the same time, several studies have explored the effects of gang participation on the health and wellbeing of gang members.

➢ On the other hand, no studies have integrated these two approaches to explore the potentially deleterious effects of residing in neighborhoods occupied by gang members to examine whether there is a unique effect of gang co-residence, whether this effect is additive or exacerbates crime effects, and whether gang co-residence is responsible for health disparities.

What this study adds

➢ Utilizing restricted birth certificate data in California, merged with gang territory maps, socioeconomic neighborhood data from the Census/ACS, and local crime data from the LAPD, we find that significant gang co-residence effects are largely explained by neighborhood disadvantage, although we are unable to untangle the exact direction of these effects. Gangland residence remains deleterious for low birthweight, however.

➢ At the same time, gang co-residence exacerbates the effects of crime for small-for-gestational-age infants. Most significantly, gang co-residence, neighborhood disadvantage, and crime all explain a much larger proportion of health disparities (both socioeconomic and racial/ethnic) than individual-level health behaviors—arguing again for an increased focus on the environmental differences that lead to health inequity in disparity populations.

Author Contributions

BKF conceived and designed the study and obtained funding to carry out the study. BKF also requested the restricted data and received IRB approval from both the CDPH and the USC IRB. JG used GIS methodologies to overlay crime counts and gang boundaries on the birth certificate data, and ANB created the neighborhood socioeconomic measures for appending to the birth certificate data. KT was responsible for data cleaning, merging, and statistical analysis. BKF wrote the initial draft of the paper with substantial assistance from ANB, and KT and JG wrote various portions of the data/methods and results. All authors edited the manuscript and agree to be accountable for all aspects of the work.

Data Availability

Data are restricted and cannot be shared, but are available via restricted data request from the CA Department of Public Health Vital Statistics Division.

Compliance with Ethical Standards

Consent for Publication

Not required.

Ethics Approval

None.

Provenance and Peer Review

Not commissioned; externally peer reviewed.