Hotspots Lead and Crime

The spatial context of crime has been a topic of criminological inquiry for over

180 years (Quetelet, 1968). Few observations about crime and offending are supported as

consistently in the literature as the non-random distribution of index crime across space.

Findings repeatedly indicate over representation of index crime in areas of concentrated

urban disadvantage. In efforts to explain this phenomenon, criminologists have found

street crime occurs frequently in areas high in poverty, increased racial and ethnic

heterogeneity, and high residential turnover (Lowenkamp, Cullen, & Pratt, 2003; Park &

Burgess, 1925; Sampson & Groves, 1989; Shaw & McKay, 1942). Other explanations

include subculture of violence theses, intimating the derivation of countercultural norms

that cultivate in areas of disadvantage (Anderson, 1999; Wolfgang & Ferracuti, 1982).

Leading mainstream macro-level theories of crime charge that these socio-structural

and/or cultural variables reduce informal social control, increase disorganization, create

criminal opportunities and, in turn, concentrate crime.

Recently, crime and place experts have turned their attention to these theoretical

frameworks to inform work on crime hotspots (Weisburd, 2012). While the precise

definition of hotspot may vary subtly, the general understanding of a crime hotspot is, “an

area that has a greater than average number of criminal or disorder events, or an area

where people have a higher than average risk of victimization,” (Eck et al., 2005, pg. 2).

An NIJ special report on understanding crime hot spots regards neighborhood area hot

spots as informed by disorganization, ecological, and opportunity theories (Eck et al.,

2005). The same report attributes likely causes to neighborhood hot spots as: “low

collective efficacy, social fragmentation, concentrations of youth, economic

disinvestments, and concentrations of crime targets,” (Eck et al., 2005, pg.5).

However, a growing literature is elucidating other factors that may explain the

unequal distribution of index crime. Environmental justice studies have found evidence

of disproportionate siting of environmental harms proximal to low income minority

neighborhoods (Bullard, 1994, 1996; Checker 2007; Hipp & Lakon, 2010; Krieg, 1995;

Opp, 2012; Stretesky & Hogan, 1998). As such, it appears many of the same areas

carrying the burden of above average levels of index crime also carry the burden of

increased exposure to environmental contaminants that pose threats to human health and

behavior. In turn, as researchers observe crime hot spots, they can also observe pollution

hot spots (Oyana & Margai, 2010; Pastor, Morello-Frosch, & Sadd, 2006).

Pollution hot spots may be important causes of crime. Prior research has shown

that exposure to various environmental toxins possesses the ability to alter behavior.

Based on prior research, it is plausible that environmental hazard sites may emit toxins

into nearby communities threatening the neurological functioning and cognition of people

residing in those neighborhoods (Carpenter & Nevin, 2010). Some of these chemicals,

particularly heavy metals and dioxins, have the capacity to compromise the central

nervous system. The link between neurological deficiencies and exposure to numerous

types of environmental toxins is well documented (Colburn, Dumanoski, & Myers, 1998;

Kampa & Castanas, 2008; Miranda, Kim, Reiter, Galeano, & Maxson, 2009; Nevin,

2000), with perhaps the most prominent environmental neurotoxin being lead (Pb)

(Rodricks, 2007).

The research on the spatial distribution of crime and distribution of neurotoxins

converged in 2001 when Stretesky and Lynch assessed the relations between air lead (Pb)

levels2 and homicide rates across the United States. Lead (Pb) presented as a toxic agent

of interest because numerous extant studies demonstrated lead’s relationship with

variables established as correlates of crime and delinquency in individuals. For example,

individuals with elevated blood lead levels have been found to be more aggressive and

have lower IQ levels than individuals who do not have elevated blood lead levels.

Stretesky and Lynch found that counties with higher quantities of air lead also

experienced higher rates of homicide. Air lead level explained unique variance in

homicide rates even while controlling for the sociological predictors of crime widely

tested in the criminology literature, including: classification of area as urban, number of

persons, area in square miles, number of adults without a high school education, number

2 Empirical evidence has established there is a positive relationship between air lead level and blood lead

level. This will be discussed in further detail in subsequent chapters, but see also: Stretesky & Lynch,

2001; 2004

of young persons, number of persons living below the poverty line, number of black

residents, and classification of county as part of the South.

Following this study, criminology was posed with a new variable with important

implications related to place and crime. Subsequent studies went on to find relationships

between lead and violent crime and property crime at the county level across the United

States (Stretesky & Lynch, 2004). Other studies found relationships between lead and

juvenile delinquency assessing bone and teeth lead levels (Olympio et al., 2010;

Needleman et al., 2002). Another study assessed childhood blood lead level and crime

rates internationally across countries (Nevin, 2007). These studies support the notion that

lead is related to crime at the individual, county, and country levels.

Yet, as criminologists continue to study crime hotspots, the majority of studies

consistently employ theoretical frameworks oriented towards traditional predictors of

crime. Left unaddressed, it is reasonable to question whether hotspots of crime may be

indicative of hotspots of lead. Ambiguity in this area is problematic, as even the most

promising of policy initiatives may be of limited utility if other sociological and

individual level correlates of crime are addressed while ignoring environmental threats to

pro social human behavior. This gap in the literature is acknowledged by both

environmental crime scholars (Narag, Pizarro, & Gibbs, 2009) and crime hotspot experts

(Lersch & Hart, 2011). However, with the exception of one unpublished study assessing

lead and homicide across census tracts (Lynch, McGurrin, & Stretesky, 2001), the

relationship between crime and lead hotspots at the community level of analysis has yet

to be subjected to empirical assessment. This study proposes to fill this gap.

The Present Study

This study addresses a very straightforward research question: Is there a

relationship between hotspots of lead and hotspots of crime? Subsequently, what is the

nature and extent of lead’s impact on crime? This study explores these issues using the

city of Chicago’s Community Areas (n=77). In doing so, this study makes at least three

contributions to the literature. First, this study examines lead across community rates of

violent and nonviolent (property) index crimes to compare the predictive power of lead

exposure across types of offenses. Extant studies suggest lead exposure is a stronger

predictor of violent crime than non-violent crime (Pihl & Ervin, 1990). Second, this

study examines lead’s impact on adult and juvenile arrest rates to compare the predictive

power of lead relative to rates of juvenile and adult crime. Prior work has suggested that

youth are more vulnerable to the adverse impacts of lead and other pollutants than adults

(Landrigan, Rauh, & Galvez, 2010; Moore, 2003; Nevin, 2007). Finally, the study

conducts an exploratory assessment of the socio-demographics of community areas

relative to their lead exposure. Prior environmental justice studies have found that

minorities and populations at risk are disproportionately exposed to environmental

hazards (Bullard, 1994; Landrigan et al., 2010; Opp, 2012).

Overview of Chapters

This study is organized into seven chapters. Chapter two introduces lead, and

chronicles its various industrial uses from past to present. Literature exploring the

relationship between lead and emotional behavioral disorders, including ADHD,

impulsivity, learning disabilities, and IQ reduction is reviewed. Also reviewed are studies

that have documented a link between lead and delinquency and antisocial behavior in

youth, as well as lead and criminal offending in adults, at both micro and macro levels.

Lastly, chapter two discusses the spatial distribution of lead and introduces to the

discussion the concept of lead hotspots.

Chapter three serves to define, describe, and review literature on crime hotspots.

This chapter discusses the empirical standing of crime hotspot studies. Focus is given to

predicting variables used in crime hotspot studies, as well as the policy initiatives inspired

by such studies. In this chapter, the issues associated with the omission of lead from

community and crime studies will be demonstrated. Taken together with chapter two, an

argument will be made that the exclusion of lead in crime hotspot studies is an important

issue in need of resolve.

Data for this study was obtained for the city of Chicago in Cook County, Illinois.

Chapter four justifies Chicago as an ideal location for this research, due to the Chicago

area’s documented struggle with the presence of both lead and crime. This chapter

provides background on the Chicago area, and explains rates of lead exposure and crime

as they pertain to the region.

Following the review of relevant literatures and background information on this

study’s setting, chapter five presents the research hypotheses, methods, and measures.

Hypotheses are expressly stated, and information on the data collection procedure and

sample is provided. Measurement and the conceptualization of variables employed in

this study are also presented.

Chapter six presents the analytic procedures for the data and results from the

hypothesis tests. Descriptive information and bivariate relationships are provided.

Results from a series of negative binomial regressions, fixed effects negative binomial

regressions, and correlations are presented and interpreted. Finally, chapter seven

discusses results from the study relative to previous findings, as well as theory and policy

implications. Limitations of the study are discussed, and suggestions for future research

are provided.

Chapter Two:

Lead

Lead is classified as a heavy metal, akin to Mercury (Hg), Cadmium (Cd)

Chromium (Cd), and Manganese (Mn). It is connoted as Pb on the periodic table of

elements and belongs to the transition metals family. Lead is a soft, malleable substance,

and its ability to resist corrosion has made it appealing for numerous commercial and

industrial purposes. Lead can be present in the air, water, and soil, and once absorbed in

the human bloodstream, lead has a half-life of approximately thirty days. However, lead

(like other heavy metals) can bio-accumulate in bones and teeth (Kampa & Castanas,

2008) where it has a half-life of twenty-seven years (Needleman, 2009). Lead has such a

long half life in the environment that it is considered a stable element. This is problematic

since the extraction and processing of lead ores into lead products can create lead

pollution which is relatively stable once emitted into the environment. This chapter is

concerned with presenting an overview of lead as well as reviewing literature relevant to

lead’s industrial uses, lead’s impact on behavior, and the distribution of lead across place.

Industrial History & Uses of Lead

It has been estimated that lead was discovered in Turkey in 6500 B.C., and thus

has been mined and in use for thousands of years. Records indicate the Romans were

routinely mining and smelting lead from 500 BC to 300 BC (Gilbert & Weiss, 2006), and

that they had used it in the construction of pipes (Kinder, 2012), and even to flavor wine.

While steel and iron required painting to resist corrosion, lead naturally resisted

deterioration. This made lead desirable for a number of uses, and through the middle

ages, lead was used for roofing, coffins, tanks and gutters. Lead was also used at this

time in producing artworks in the form of ink, statues, ornaments, and for making strips

that joined together pieces of stained glass for church windows (Kinder, 2012). In later

years, lead’s malleability and resistance to erosion made it an appealing material to use to

protect electrical wires, and was continually used in the construction of pipes and

plumbing systems. It was not until the early 20th century that lead was assigned the two

industrial purposes that make lead most ubiquitous today: inclusion in paint and inclusion

in gasoline (Reyes, 2007).

Lead in paint. Because of lead’s resilience, when lead is added to paint, it

increases paints ability to remain intact and become resistant to chipping or peeling.

Lead has been used to give paint color or tint, and common forms of lead used for this

purpose are white lead (PbCO3) and vivid yellow lead chromate (PbCrO4) (Crow, 2007).

Leaded paint is also waterproof and opaque, making a small amount of lead paint able to

cover relatively large surfaces. Lead paint has been used to coat furniture and toys, and

used for interior and exterior purposes on houses. While alternative chemicals shared

these properties with lead, and were available as early as 1914 (e.g.- zinc based and

titanium based paints), lead remained highly prevalent in paint for several years. The

leaded paint industry was very profitable for manufacturers: in 1948 alone, Dutch Boy

leaded paint had grossed $320 million (Markowitz & Rosner, 2000).

The use of lead paint for indoor purposes was banned overseas as early as 1909

(Moore, 2003); however it was not until 1970 that the United States passed the Lead Paint

Poisoning Prevention Act marking the beginning of the phasing out of lead based paint.

In 1978 the use of lead in paint was drastically reduced in the United States by the federal

government. Following this legislation, lead was largely replaced with titanium dioxide

(Crow, 2007). Regardless, smaller levels of lead are still used in paint today. The federal

government currently permits up to 600 parts per million (ppm) of lead in household

paint (California Department of Education, 2012). The use of paint with higher

concentrations of lead is also lawful in the United States at certain industrial settings

(Crow, 2007).

Lead in gasoline. In 1921 researchers employed by the automobile industry

found that adding tetraethyl lead [(CH3CH2)4Pb] to gasoline curbed engine “knock”; a

loud noise resulting from an explosion of unburned fuel inside the engine(Davis, 2002).

Engine knock causes reduced fuel economy for the vehicle as well as wear on the engine.

Prior to tetraethyl lead’s discovery as an antiknock gas additive, General Motors had been

reliant on ethanol, which not only served as an antiknock additive to gasoline, but burned

“clean” (no carbon emissions), and increased horsepower. However, the automotive

industry could neither patent ethanol nor prevent consumers from producing it on their

own. Because of its patenting potential, tetraethyl lead presented the auto industry with a

more lucrative solution to engine knock. In 1923 DuPont, Standard Oil (later the Exxon

Corporation) and General Motors patented tetraethyl lead, and co-founded the Ethyl

Corporation for the purpose of producing tetraethyl lead (Davis, 2002). Within ten years

of its inception, leaded gasoline dominated 80% of the U.S. gasoline market, grossing an

annual profit of $300 million for General Motors and its shareholders (including DuPont,

who simultaneously profited from Dutch Boy leaded paint) (Kitman, 2000) (Markowitz

& Rosner, 2000).

Tetraethyl lead continued to be the most prominent antiknock agent used in

gasoline in the United States through the 1960s, and remained unregulated even after the

passage of the Clean Air Act in 1970 (Moore, 2003). In 1973 the Environmental

Protection Agency passed the first laws regulating lead in gasoline, marking the

beginning of a phase out of leaded gas. Amendments to the Clean Air Act in 1990

mandated the elimination of lead from all U.S. motor fuel by 1996. Ethanol, and other

compounds, would slowly be reintroduced as antiknock agents. In the U.S., tetraethyl

lead continues to lawfully be added to aviation fuel, and it is estimated that approximately

200,000 airplanes and helicopters in the United States are powered by leaded fuel

(Peeples, 2012). The U.S. government also currently allows .05 grams of lead or less per

gallon to lawfully be defined as “unleaded gasoline,” (Protection of Environment, 40

C.F.R. § 80.2). Thus, today, a car with a twelve-gallon tank may contain fuel with up to

.6 grams of lead.

Lead and Neurotoxicity

For nearly as long as lead has been used for commercial and industrial purposes,

it has been recognized that lead exposure poses serious threats to the healthy functioning

of the human brain. Greek physicists as early as the second century BC described lead as

“deadly,” with one physicist stating, “Lead makes the mind give way,” (Needleman,

2009, p.245). In 1786, Benjamin Franklin wrote a letter discussing his observation of lead

poisoning among printers in London, and emphasized that lead has been known to

threaten public health for at least sixty years. In 1892, Dr. J. Lockhart Gibson treated a

number of children with “severe neurological disease,” (Needleman, 2009, p.245) at

Brisbane Hospital in Australia. He determined the cause of the disease to be the lead in

the white paint that was being used to paint railings and porches in the region.

In the United States, as early as 1910, experts with clinical and research

experience with lead’s neurotoxicity publicized findings and urged representatives from

industry and government to ban lead (Markowitz & Rosner, 2000; Moore, 2003;

Needleman, 1997; Rosner & Markowitz, 1985). Doctors treating Ethyl factory

employees who exhibited violent insanity told the press their conditions were directly

linked to lead exposure. Employees at the Ethyl Corporation referred to tetraethyl lead

as, “looney gas,” and the plant as “the House of Butterflies,” because of hallucinations

brought on by lead exposure (Needleman, 1997). In 1924, Dr. Yendell Henderson, a

biochemist at Yale University (who formerly guided U.S. chemical warfare research

during World War I), wrote to the head of the U.S. Bureau of Mines5 that leaded gas

indicated, “General Motors is now deeply committed financially in reckless disregard of

the possible and even probable industrial and public health hazard,” (Henderson, 1924 in

Needleman, 1997, p.97). One year later, Dr. Alice Hamilton, an occupational health

expert at Harvard who had treated lead workers with brain damage, testified before the

U.S. Surgeon general and lead industry representatives that, “lead is a slow and

cumulative poison.” Also presented to the Surgeon General and industry representatives

was data from Columbia University’s Dr. Frederick Flinn, demonstrating death and

toxicity in animals exposed to lead. Markowitz and Rosner (2000) identified at least

thirteen scholarly publications documenting the relationship between lead exposure and

5 In the 1920s, the U.S. Bureau of Mines began conducting research on the safety of lead on grant funding

from General Motors. One stipulation of this funding was that the Ethyl Corporation would be granted

permission to screen all manuscripts prior to publication of findings, and no manuscript may be submitted

for publication without joint consent from the Ethyl Corporation and the Department of Interior. In the

same letter, Henderson addressed this, stating: “It seems to me extremely unfortunate that the experts of the

United States Government should be carrying out this investigation on a grant from the General Motors. I

feel very strongly that there is the most urgent need for an absolutely unbiased investigation,” (Needleman,

1997). On this point, Needleman (1997) himself comments this correspondence reveals, “the Bureau of

Mines, a government agency acting in a regulatory capacity, complicit with the industry. In straining to

protect the corporation and its product, the Bureau agreed to act as a shield against the press, to camouflage

the nature of the toxin, and to cede to the company control of the publication of any results,” (p.97). This

illustrates the early roots of lead’s relationship with political-economy.

adverse impacts on the nervous system published prior to 1928, inclusive of studies

documenting lead paint ingestion and effects on children.

While these early studies were very important, they generally understood

neurological damage from lead exposure as having one of two outcomes in humans: (1)

death, or (2) full recovery. The idea that someone could fully recover from lead

poisoning was first challenged in an analysis by Byers and Lord in 1943. Their study

examined twenty patients with elevated blood lead levels in infancy that were treated and

thought to have recovered. The patients were subsequently administered neurological

tests as children, where all but one demonstrated decreased IQ, learning problems, and

behavioral disorders. As children, all patients appeared healthy in terms of acute lead

exposure effects (e.g.) none reported/demonstrated nausea, dizziness, fever, headaches.

Byers and Lord’s study introduced the notion that neurological damage caused by lead

exposure may not only be physically asymptomatic, but also persistent and irredeemable8.

Following Byers and Lord, researchers published correlations between subclinical

blood lead levels and neuropsychological impairments (de la Burdé & Choate, 1975;

Kotok, 1972; Landrigan et al., 1975). In a landmark study published in the New England

Journal of Medicine, pediatrician Herbert Needleman and colleagues studied subclinical

lead levels and neurotoxicity using a sample of first and second grade children from

Boston. Needleman et al. (1979) tested subjects’ cognition as well as dentine lead levels,

and interviewed childrens’ parents and teachers (Needleman et al., 1979). Compared to

children with low dentine lead levels, teachers were found to rate children with high

dentine lead levels as significantly more: distractible (36% vs. 14%), hyperactive (16%

vs. 6%) and impulsive (25% vs. 9%). Children with high dentine lead levels had a mean

IQ score significantly lower than children with low dentine lead levels (102.1 vs. 106.6).

Importantly, these relationships held net of thirty nine control variables, including: social

class, age, sex, race, weight, history of pica, level of parental education, parental

occupation, and parental age at subject’s birth—overcoming a limitation of prior studies.

Since this study, Needleman and colleagues repeatedly found relationships between

subclinical bone and blood lead levels among these and other emotional behavioral

impairments (Needleman & Gatsonis, 1990; Needleman et al., 1990; Bellinger et al.,

1987; Bellinger, Stiles, & Needleman, 1992), and the study of lead’s neurotoxicity has

grown exponentially.

Lead and emotional behavioral disorders. A subset of lead and neurotoxicity

studies that continues to captivate scientists is lead exposure’s relationship with emotional

behavioral disorders. An emotional behavioral disorder can be described as a persistent

condition in which abnormal emotions or behaviors impede an individual’s ability to:

build relationships, learn, and/or self-regulate mood or actions. These impairments may

also manifest in the form of phobias and/or physical pain, and can include both

externalizing (anger, aggression) and internalizing (anxiety, depression) symptoms. Cecil

et al. (2008) found subjects who had increased blood lead levels as children demonstrated

significantly decreased development of gray brain matter

8 Even when individuals who have been exposed to lead have been properly screened and identified, the

effectiveness of interventions, medicines, and therapies remains topic of contemporary debate among

(specifically in the prefrontal cortex and anterior cingulate cortex) as adults; these regions

of the brain are responsible for regulating emotions, mood, executive functioning, and

decision making. Lead exposure has demonstrated relationships with depression, anxiety,

schizophrenia, conduct disorders, and inattention (Bouchard et al., 2009; Braun et al.,

2008; Opler et al., 2008; Roddicks, 2007; Roy et al., 2009; White et al., 1993).

Using the Bayley Scales of Infant Development, Mendelsohn et al. (1998) found

that children with blood lead levels between 10μg/dL and 24.9μg/dL reported

significantly greater levels of fear, withdrawal, and disinterest, differing from children

with decreased blood lead levels by a mean score of 14.2 points. Using the Achenbach

Child Behavior Checklist, Burns et al. (1999) found increases in cumulative blood lead

level among adolescent girls to be associated with corresponding increases in levels of

anxiety, depression, aggression, withdrawal, thought problems, and attention problems.

The same study found increases in cumulative blood lead level among adolescent boys to

be associated with increased aggression and attention problems (Burns et al., 1999).

Chiodo, Jacobson, and Jacobson (2004) also used the Achenbach Child Behavior

Checklist, and found children to exhibit significantly increased withdrawn and inattentive

behaviors at blood lead levels at as low as under 3 μg/dL.

Lead & ADHD. A growing literature has also documented lead’s relationship

with an especially prevalent emotional behavioral condition: attention deficit

hyperactivity disorder (ADHD) (Braun et al., 2006; Chiodo et al., 2007; Nigg et al., 2010;

Nigg et al., 2008; Roy et al., 2009; Wang et al., 2008). The DSM-IV describes ADHD as

scientists (Moore, 2003; Rogan et al., 2001; Taylor et al., 2011)

difficulty maintaining attention and focus, hyperactivity, and difficulty controlling

behavior. Nigg et al. (2010) found significant associations between childhood blood lead

level and parent reports of their children’s impulsivity and hyperactivity on the Connors

Rating Scale. Braun et al. (2006) assessed 4,704 respondents in the NHANES and found

parents of children with blood lead levels between 2 and 5 μg/dL were over four times

more likely to also report that their child has an ADHD diagnosis than parents of children

with blood lead levels less than .7 μg/dL. A case-control study of 1,260 subjects found

nearly 76% of respondents diagnosed with ADHD to have blood lead levels greater than

5 μg/dL- a significant difference from participants not diagnosed with ADHD (where

nearly 50% reported blood lead levels greater than 5 μg/dL) (Wang et al., 2008).

ADHD & crime. The link between ADHD and lead is relevant to criminology

because studies have linked ADHD to criminal and delinquent behavior (Dembo et al.,

2011; Pratt et al., 2002; Sibley et al., 2011; Teplin et al., 2006; Winters et al., 2008). In a

meta analysis, Pratt and colleagues (2002) found ADHD to carry a strong association to

crime and delinquency. Of particular interest to criminologists has been the impulsivity

component of ADHD, as it has been empirically linked to increased risk behavior

(Winters et al, 2008), delinquency (White et al., 1994), crime (Jones & Lynam, 2009;

Lynam et al., 2000; Yarbrough et al., 2011) and is theoretically argued to be an important

dimension of low self control (Gottfredson & Hirschi, 1990). A meta-analysis of twenty

one studies testing Gottfredson and Hirschi’s General Theory of Crime finds self control

a robust predictor of crime and analogous behavior (Pratt & Cullen, 2000). In a sample

of adult inmates, Cahill et al. (2009) find more than one in ten inmates with ADHD, while

Eme (2009) estimates one in four adults in the U.S. correction system has ADHD.

McCabe et al. (2002) find among 625 adjudicated youth, 15% of males and 21% of

females had ADHD. When these rates are compared to the rates of ADHD in the general

public- 3-7% of children and 4% of adults- ADHD appears overrepresented among

offending populations. This is consistent with the overrepresentation of all emotional

behavioral disorders among incarcerated populations, and associations between ADHD

and offending do not appear exclusive to violent offending.

Lead & intellectual functioning. ADHD maintains high co morbidity with one of

the earliest established neurological problems associated with lead exposure- learning

problems. Furthermore, lead exposure has been linked to decreases in IQ points and

intellectual functioning, even at very low exposure levels (Fulton et al, 1987; Canfield et

al., 2003; Chiodo et al., 2004; Landrigan, 2000; Lanphear et al., 2000; Lanphear et al,

2005; Miranda et al., 2007; Schwartz, 1994; Wasserman et al., 2000). In a meta-analysis

of lead exposure and childhood IQ, Schwartz (1994) finds across eight studies an overall

increase in blood lead level from 10 to 20 μg/dL resulted in a corresponding decrease of

2.6 IQ points. Gilbert and Weiss (2006) report that for every 1 μg/dL increase in blood

lead, a decrease of .87 IQ points can be anticipated. In an analysis of academic

achievement and neurotoxicity, Zahran et al. (2009) found children with elevated blood

lead levels underperformed significantly on standardized tests, and Miranda et al. (2007)

found fourth grade students with increased blood lead levels significantly

underperformed on end of grade math and reading tests. Studies have demonstrated that

lead exposure in early years (including infancy and exposure in utero) have long lasting

impairments on intelligence which were detectable in childhood (Bellinger et al., 1987;

Bellinger, et al., 1992; Wasserman et al., 2000) and again in young adulthood (Needleman

et al., 1990). Muldoon et al. (1996) state that while that the developing nervous systems

of children are more sensitive to lead exposure than adults, adult populations remain at

risk for damage associated with lead exposure. They observed decreased performance on

neuropsychological tests in a sample of elderly women at blood lead levels as low as 8

μg/dL (Muldoon et al., 1996).

Intellectual functioning & crime. The link between lead exposures and

diminished intellectual functioning is relevant to criminology because a relationship

between diminished intellectual functioning and criminal behavior has been observed

(Bellair & McNulty, 2010; Diamond, Morris, & Barnes, 2012; Hirschi & Hindelang,

1977; Lynam, Moffitt, & Stouthamer-Loeber, 1993; McGloin & Pratt, 2003). Lynam et al.

(1993) studied participants from the Pittsburg Youth Study and found on average juvenile

delinquents score eight IQ points lower than non-delinquents on standardized IQ tests.

Lynam et al. (1993) also found IQ to be a significant predictor of subsequent selfreported

delinquency. Mainstream theories of criminology also argue associations between

intellect and all crime. In their General Theory of Crime, Gottfredson and

Hirschi (1990) regard individuals with low self control to be “short-sighted” (p.90) and

describe crimes as “[requiring] little skill.” Gottfredson and Hirschi (1990) go on to

specify that they intend for their general theory of crime to, “explain all crime, at all

times,” (p. 117). Among a sample of prison inmates, Diamond et al. (2012) report a

mean IQ of 90, ten points below the mean score among the general population, and

associate this IQ reduction with inmate violence. A meta analysis finds that juveniles

with reduced academic performance offend more frequently, commit more serious

offenses, and are more likely to persist in offending (Maguin & Loeber, 1996).

In sum, lead has a lengthy history of widespread industrial use and established

neurotoxicity. Lead’s ability to cause long-term, asymptomatic effects on the human brain

were demonstrated empirically in the 1940s, and have since been replicated numerous

times over. Studies supporting these relationships find the association tenable using a

variety of subjects (human, non-human, children, adults), and employing a range of study

designs (cross sectional, longitudinal). Pertinent to a discussion of crime is lead’s

relationship with well-established crime correlates, especially emotional behavioral

disorders, impulsivity, ADHD, and intelligence. Taken together, it is reasonable to

suspect lead exposure may set into motion neurological conditions (mood, emotions,

behavior) conducive to criminal activity.

Lead and Crime

Given the well-established relationship between lead exposure and widely studied

causes and correlates of crime (impaired cognition, emotional behavioral disorders,

ADHD, impulsivity, learning problems, diminished IQ), a number of causal pathways

linking lead to crime may be explicated (Narag, Pizarro, & Gibbs, 2009). In the 1990s,

researchers became interested in assessing the dynamic between lead and criminal

behaviors directly, and empirical tests of this association began to emerge. Research has

supported associations between lead and property offenses (Nevin, 2007; Stretesky &

Lynch, 2004), and violent offenses (Lynch, McGurrin, & Stretesky, 2001; Nevin, 2000;

Reyes, 2007; Stretesky & Lynch, 2001). As of current, studies have tested and found

support for the relationship between lead and criminal behavior at both micro and macro

levels of analysis.

Micro level findings. Individual level studies have evinced a relationship

between lead exposure and adult criminal behavior. For example, Pihl and Ervin (1990)

measured lead levels in a sample of male adult inmates. They found that inmates

convicted of violent offenses had mean hair lead concentrations significantly higher than

inmates convicted of non-violent offenses (10.57ppm vs. 4.58ppm). Fergusson, Boden,

& Horwood (2008) studied dentine lead levels and crime among a birth cohort of 1265

participants. Results indicated that dentine lead level at age 6-9 was significantly related

to criminal convictions and self-reported criminal behavior at ages 14-21 (Fergusson et

al., 2008). Controlling for confounders, participants with childhood dentine lead levels

over 15 µg/g self-reported an average of about 13 criminal offenses as opposed to

participants with childhood dentine lead levels between 0-2 µg/g, who self-reported an

average of about 6 criminal offenses (Fergusson et al, 2008). Fergusson and colleagues

(2008) observed significant associations between lead and both property and violent

crime combined. Wright et al. (2008) and Denno (1990) have also found an association

between prenatal and childhood blood lead level and criminal arrests for violent crimes in

young adulthood.

Macro level findings. These associations have persisted in macro level studies

using aggregate measures of crime and lead at the international and national levels.

Nevin (2007) employed time series analyses and found associations between preschool

blood lead levels and index crime arrest rates across eight developed nations. Nevin

(2007) detects an 18 year lag impact of preschool lead exposure on burglary rates in eight

developed nations, five of which retain the significant association even when controlling

for unemployment. In the U.S., Nevin (2000) measured gasoline lead exposure with

annual U.S. gasoline consumption between 1941-1987 as reported by the U.S. Geological

Survey. Nevin (2000) found gasoline lead level to explain unique variance in the U.S.

violent crime rate from 1960-1998 while controlling for teenage pregnancy,

unemployment, maternal education level, and age. In 2007, Reyes also uses aggregate

crime and lead exposure rates in the U.S. to study the predictive power of childhood lead

exposure relative to the oft studied 1990s crime drop. Controlling for income,

employment, poverty, welfare dispersion, prison population, size of police force, beer

consumption, age, and teen pregnancy rates, Reyes (2007) found 56% of the drop in

violent crime in the 1990s is attributable to the reduction of lead in gasoline. Unlike

Nevin (2007), however, Reyes (2007) did not detect a significant relationship between

lead and property crime.

Macro level studies have also demonstrated an association between lead and

crime at the county level (Stretesky & Lynch, 2001; 2004). In their 2001 study, Stretesky

and Lynch test the relationship between air lead level and homicide across 3,111 U.S.

counties, controlling for nine other air pollutants and eight other relevant sociological

variables (race, population, area, poverty, education, age, urban area, and region of U.S.).

Findings revealed the relationship between lead and homicide to be significant, with

results of a negative binomial regression indicating an incident rate ratio of 4.12. This

suggests that homicide is over four times more likely in counties with the highest level of

air lead as opposed to counties without air lead, net of confounders. None of the other air

pollutants controlled for emerged as significant predictors of homicide. These findings

demonstrated not only that the relationship between lead crime was tenable at the county

level, but also, they suggest lead is a relevant characteristic of place for criminologists to

consider in ecological studies of crime.

This suggestion was bolstered in 2004, when Stretesky and Lynch demonstrated

that county air lead levels were significantly related to both county violent and property

crime rates. Inclusion of lead and other sociologically relevant variables in regression

models explained 44% the variance in violent crime and and 34% of the variance in

property crime. The same study also found air lead level and resource deprivation to

have an interactive effect on violent and property crime rates, suggesting the impact of

lead on crime is exacerbated in resource deprived areas.

A recent study uses aggregate statistics and finds a relationship between lead and

crime across six U.S. cities: Chicago, Minneapolis, Indianapolis, Atlanta, New Orleans,

and San Diego (Mielke & Zahran, 2012). Mielke and Zahran (2012) examine the

relationship between air lead (measured by calculating annual city estimates of air lead

from vehicle traffic from 1950-1985) and number of aggravated assaults reported to

police from 1972-2007, employing lags theoretically informed by the age-crime curve.

Mielke and Zahran (2012) control for income per capita as well as percent of population

between 15 and 24 years of age. Their findings suggest that for every 1% increase in lead

air a corresponding .46% increase in aggravated assault rate can be anticipated 22 years

following that increase. Mielke and Zahran (2012) stated that their full model explained

90% of the variance in reported aggravated assault rate across all of the cities in their

analysis.

Lead and Delinquency

In addition to associations found between lead and adult crime (Denno, 1990;

Wright et al., 2008), associations have also been found between lead and juvenile

delinquency (Needleman et al., 1996; Olympio et al., 2010).

The relationship between lead and juvenile offending at the individual level has

garnered particular interest among researchers (Denno, 1990; Dietrich et al, 2001;

Needleman et al., 2002; Olympio et al., 2010). This is reflective of trends in lead and

neurotoxicity in general, and is likely related to the fact that children are especially

vulnerable to environmental toxins (Landrigan & Carlson, 1995; Landrigan, Kimmel,

Correa, & Eskenazi, 2004; Moore, 2003; Reyes, 2007). In 1993, a report by the National

Research Council established that children are especially vulnerable to environmental

toxicity for four reasons. Landrigan et al (2004) describe these four reasons as (1) children

have a comparatively higher exposure to environmental agents, (2) children’s metabolisms

are under-developed, (3) developmental processes are susceptible to disruption during

phases of rapid growth, and (4) exposure early in the life course leads to longer time frames

for diseases to develop. What this implies is a differential age effect for the association

between lead exposure and human behavior. As such, one consequence of this may be

increased rates of juvenile delinquency.

Findings from the empirical literature document associations between lead

exposure and juvenile delinquency. For instance, Denno (1990) found childhood lead

poisoning to significantly predict juvenile delinquency among black men, following a

sample of black children through young adulthood. Needleman et al. (1996) found

increased bone lead burdens among participants in the Pittsburgh Youth Study to be

significantly related to self-reports of delinquency at age 11 (log transformed scores on

the SRD indicated 1.51 for a low lead group compared to 2.39 for a high lead group,

p=.04). The same study also found teachers and parents were more likely to indicate

higher rates of aggressive and delinquent behavior on the Child Behavior Checklist for

children with high bone lead burdens. These findings were replicated recently

implementing the same delinquency measures but using dentine lead level on a sample of

adolescents from Brazil (Olympio et al., 2010). A case control study of adjudicated

juvenile delinquents (compared to non-adjudicated youth) found after controlling for

confounders, adjudicated youth were four times more likely to have increased bone lead

burdens (greater than 25 ppm) than controls (Needleman et al., 2002). Dietrich et al.

(2001) studied this relationship prospectively, and found prenatal and postnatal blood lead

levels to significantly predict frequency of self-reported teenage delinquency. Dietrich

and colleagues (2001) used responses on the Self-Report of Delinquent Behavior to

measure delinquency, which includes items capturing both violent and non-violent

offenses.

In sum, lead appears to consistently relate to crime across several units of

analysis: individuals, cities, counties, and countries. Further, these studies have

employed a wide range of methodologies, and have documented a link between lead and

crime prospectively in longitudinal designs, utilizing birth cohorts, panel studies, and

time series techniques. By design, these studies illustrate the hypothesized temporal

ordering of lead as a causal mechanism in explaining crime. Lead and crime

relationships have also been observed in cross sectional studies using case control

designs. Taken together, these findings suggest that lead should not be discounted as a

factor in explaining crime. A study of lead and crime at the community level, however, is

further justified by reviewing literature discussing the factors that connect lead exposure

to place.

The Spatial Distribution of Lead

According to the EPA’s Toxic Release Inventory (TRI) database, in 2010 alone,

industries across the United States reported disposing, transporting, or releasing over 570

million pounds of lead and lead compounds. The EPA also states that over 300

Superfund sites on the National Priorities List indicate lead presence (suggesting the

presence of lead at over 25% of Superfund sites). In 2011, Mielke, Laidlaw, and

Gonzales analyzed soil lead levels across 90 U.S. urban areas and found anthropogenic

lead deposited from leaded gasoline sold between 1950 and 1982, demonstrating lead’s

ability to accumulate and persist in surface materials. According to the 2000 Census,

there are 17.4 million housing structures in the United States built before 1940 (a risk

factor for lead paint exposure), with other accounts estimating up to 38 million U.S.

housing units containing lead-based paint (Jacobs et al., 2002).

These sources of lead, however, are not randomly distributed. For example, using

TRI data, Masters, Hone, and Doshi (1998) find the distribution of lead to be highly

skewed, with over 80% of all U.S. counties reporting no release of lead or lead

compounds in 1991. For that same year (1991), TRI reports indicate industries reported

disposing of or releasing 40,311,787 pounds of lead or lead compounds. Thus, the

dispersion of over 40 million pounds of lead appears restricted to approximately 451 (out

of 3,111) U.S. counties. Similarly, Stretesky and Lynch (2001; 2004) found presence of

air lead across U.S. counties using the Cumulative Exposure Project (CEP) to range from

0 µg/m3 to .172 µg/m3, with a mean and median of .001, also indicative of skewness. In a

study assessing lead poisoning in North Carolina, Hanchette (2008) analyzes North

Carolina’s Childhood Lead Poisoning Prevention Program and finds positive spatial

autocorrelation among county lead poisoning prevalence rates each year from the years

1994 to 2005. In 2008, the EPA identified 21 areas in 22 counties where air lead levels

exceeded national ambient air quality standards. This observed disproportionality in the

distribution of lead, has resulted in what this study will refer to as lead hotspots- areas

with disproportionately high concentrations of lead.

Explaining the lead hotspot phenomenon. Explaining this phenomenon of lead

hotspots requires recognition that the spatial distribution of lead is inextricably wedded to

historical, political, economic, and social forces (Hanchette, 2008; Kitman, 2000; Lynch,

2004; Moore, 2003; Davis, 2002; Markowitz & Rosner, 2000; 2003; Needleman, 1997;

Rosner & Markowitz, 1985; Stretesky, 2003; Stretesky & Lynch, 2004). Consider that

lead’s neurotixicity was known by manufactuers and the U.S. government for over sixty

years before the first act of legislation was passed regulating lead presence in products.

Further, when policies were eventually implemented, they were not retroactive in nature,

in spite of lead’s bioaccumulative properties. As such, facilities built prior to lead paint

regulation carry higher concentrations of lead than newer structures (Jacobs et al., 2002;

Lanphear et al., 1998). Further, areas proximal to freeways, roadways, and heavy traffic

patterns carry higher lead concentrations, as new lead gasoline deposits commingle with

anthropogenic deposits (Needleman & Bellinger, 1991). Because the use of lead is still

permissible in some industries, and present at a number of Superfund sites, centers of

industry and smelteries also typify lead hotspots. Taken together: lead hotspots tend to be

urban areas (Mielke, 2005).

Urban ecology and community composition are also linked to historical, political,

economic and social factors (Lynch, 2004; Lynch & Michalowski, 2006). Housing

choices increase with income, and urban ecologists have illustrated a tendency of income

to increase as distance from industrial zones (Park & Burgess, 1924) and environmental

hazards also increase. Public policy has shaped community composition through history,

and housing practices such as redlining, regulation on interest rates and subprime

mortgage lending, as well as bussing and school segregation have left legacies in

American communities and shaped the composition of place (Wacquaint & Wilson,

1989).

The consequence of these relationships is observed differences in community

composition relative to race, class, and ethnicity (Park & Burgess, 1924). For example,

institutionalized racism and classism (zoning, covenants, redlining, subprime lending)

have tempered educational, occupational, and housing choices for individuals, resulting

cumulatively in the overrepresentation of minorities in poor, urban communities (Massey

& Denton, 1993). Comparisons of black and white communities reveal disparate income

levels, with black neighborhoods earning less overall than similarly situated white

neighborhoods (Pattillo, 1998). These segregation practices organize the sociological

composition of space in a way that makes race and class power lines visible; with

ecological goods (employment opportunities, cleaner environments) farther in distance,

and ecological bads (pollution, decay) increasingly proximal to minority groups most

marginalized by political-economic practices.

Lead’s distribution and urban ecology’s shared underlying relationship to these

factors force lead and communities to converge in time and space in such a way that is

reflective of overarching power dynamics in the United States (Lynch, 2004; Stretesky,

2003). In turn, the nonrandom distribution of lead parallels the nonrandom distribution of

income, race, and ethnic minority communities. Ecological studies observe sociological

differences between communities with and without increased lead presence (Hanchette,

2008; Stretesky, 2003). Because income precludes access to newer schools and housing

units, individuals of lower socioeconomic status become disproportionately restricted to

older buildings (increased lead paint and lead plumbing presence). Historically,

communities lacking the social capital and resources to organize and resist interstate and

expressway construction were disproportionately targeted by builders (increased leaded

gasoline presence). This same logic can be extended to environmental hazards and waste

sites, which since the 1970s, have tended to follow a path of least resistance (Burns et al.,

2008). These historical, political, economic, and social factors act in concert, and

generate observable, consistent characteristics of lead hotspots.

Correlates of lead hotspots. Racial disparities have been observed across lead

hot spots. In an assessment of lead exposure across U.S. counties, Stretesky (2003) found

counties with the largest proportions of black youth contained 7.9% more air lead than

counties with no black youth. Further, the same study found that counties with the largest

presence of white youth have nearly 10% less air lead than counties with the smallest

proportion of white youth. Stretesky (2003) found these compositional changes were

observable even while controlling for urbanization, income, location of manufacturing

sites, age of house, and housing values. Mielke et al. (1999) studied soil lead levels

across census tracts in New Orleans and found 60% of New Orleans black population

resided in tracts with high levels of lead in soil. By comparison, the same study found

37% of New Orleans’ white population to be residing in tracts with high levels of lead in

soil. Lanphear, Weitzman, and Eberly (1996) found black children significantly more

likely than white children to be exposed to lead dust and to be residing in houses with

poor interior paint conditions.

Income disparities are also observable across lead hot spots. Lanphear and

Roghmann (1997) found income to have a significant, inverse relationship with level of

lead dust found in home. Diawara et al. (2006) studied levels of lead in topsoil in Pueblo,

Colorado and find higher proportions of low-income communities situated closer to

leadcontaminated areas. Jacobs et al. (2002) collected a nationally representative sample

of 831 housing units between 1998 and 2000 and found 35% of households with incomes

less than $30K a year resided in housing units with lead hazards, and 38% of households

living in poverty resided in housing units with lead hazards. By comparison, the same

study found 19% of households earning equal to or greater than $30K a year to live in

units with lead hazards, and 22% of households not in poverty to live in units with lead

hazards.

Differences in ethnicity are visible across lead hotspots. In addition to observed

income disparities, Jacobs et al. (2002) find differences in rates of Hispanic residents

occupying housing units with lead hazards: 32% of Hispanic or Latino residents reside in

units with lead hazards present as opposed to 24% of non-Hispanic or Latino residents.

Diawara et al. (2006) observe that populations in census tracts with lowest lead soil levels

are nearly 80% non-Hispanic white. They also observe that as soil lead level increases

within census tract, there is a corresponding significant, positive increase in the

percentage of Hispanic residents (Diawara et al., 2006). In Sargent et al.’s (1997)

multivariate analysis of predictors of elevated childhood blood lead levels (>10 µg/dL),

as percentage of recent immigrants increased among census tracts, a corresponding

increase of .56 was observed in percentage of children with elevated blood lead levels.

This relationship held even while controlling for other significant socio-demographic

variables.

In the same study, researchers found a relationship between residential mobility

and lead hot spots. Sargent et al. (1997) analyze census tracts in Providence, Rhode

Island, examining socio-demographic predictors of childhood blood lead levels. Sargent

et al. (1997) found the percentage of owner occupied houses and the percentage of

standalone vacant homes to be associated with more than 60% of the variation in elevated

childhood blood lead. In addition to the above referenced association between percent of

recent immigrants (b=.56, p = .003) and elevated childhood blood lead level, this study

found a significant relationship between percentage of vacant houses (b= 5.18, p =

<.0001) and elevated childhood blood lead level. Sargent et al. (1997) found these

predictors significant while simultaneously controlling for eleven other sociodemographic

variables. Their final parsimonious model, which included percent screened, households

on public assistance, percent homes built prior to 1950, houses vacant, and percent of

recent immigrants was associated with 83% of the variance in elevated childhood blood

lead.

Connecting lead hotspots to exposure patterns. Disparate distributions of lead

have been linked to disparate exposure patterns. Studies evincing this link demonstrate

an important relationship between environmental lead and human body lead burden

(Brunkenreff, 1984; Hayes et al., 1994; Lanphear et al., 1996; Lanphear et al., 1998;

Pirkle et al, 1994; Reyes, 2007; Schwartz & Pitcher, 1989; Stretesky & Lynch, 2001).

For example, Hayes et al. (1994) finds a strong correlation (r=.8, p<.001) between air

lead level and blood lead level, with a regression model finding each decrease in .1 µg/m3

in mean air lead level (when air lead was around 1 µg/m3) to be associated with a

corresponding .56 µg/dL decrease in median blood lead level. When federal policy

initiatives called for a reduction of lead in gasoline, an analysis of the National Health

and Nutrition Surveys (NHANES) revealed the average blood lead level for Americans

aged 1 to 74 decreased from 12.8 µg/dL at NHANES II (1976-1980) to 2.8 µg/dL at

NHANES III (1988-1991) (Pirkle et al., 1994). This marked an overall decrease in

national blood lead level by 78%. Associations between air lead, soil lead, and blood

lead levels have also been observed (Laidlaw et al., 2012; Zahran et al., 2011).

Aschengrau et al. (1994) report every 1000-ppm decrease in soil lead level corresponded

with a 1.12-1.35 µg/dL decrease in blood lead level. Thus, prior studies demonstrate that

the presence of environmental lead implies a presence of lead in the human body, and

vice versa.

As anticipated, it logically follows that groups disproportionately exposed to

environmental lead sources are more likely to carry higher body lead burdens (Bernard &

McGeehin, 2003; Brody et al., 1994; Scinicariello, Abadin, & Murray, 2011; Lanphear et

al., 1998; Pirkle et al., 1994). Bernard and McGeehin (2003) assess the NHANES III

from 1988-1994 and found percentage of non-Hispanic blacks with blood lead burdens

between 5-10 µg/dL to be twice the rate of non-Hispanic whites with the same blood lead

levels (32% vs 15%). When blood lead level increased from 10-20 µg/dL, non-Hispanic

blacks reported these levels at three times the rate of non-Hispanic whites (12% vs 4%).

Further, individuals living at or below the poverty level were more than twice as likely to

report a blood lead level of 5-10 µg/dL as those living above poverty (30% vs 14%).

According to Brody et al.’s (1994) findings, when income and location of residence were

considered, race effects became exacerbated: blacks who lived in cities or reported low

income were more than twice as likely as whites to report elevated blood lead levels.

Lanphear et al. (1996) reach similar conclusions on a smaller sample of residents of

Rochester, New York, finding race and income to be significantly related to blood lead

level and lead dust exposure.

Socio-demographic relationships with lead distribution and exposure patterns are

consistent with broader findings in environmental justice research: race, class, and

ethnicity are significantly related to environmental hazards. Specifically, environmental

hazards, like lead, appear to be disproportionately close in proximity to communities with

high percentages of racial and ethnic minorities, and high rates of poverty (Lerner, 2005;

Brulle & Pellow, 2006; Bullard 1994; 1996; Checker, 2007; Hipp & Lakon, 2010; Opp,

2012; Pastor, Morello-Frosch, & Sadd, 2006; Stretesky & Lynch, 2002). A metaanalysis

of 49 environmental equity studies finds racial inequity consistently emerged as a

significant variable relating to environmental hazards, regardless of level of analysis or

type of hazard examined (Ringquist, 2005). In a study assessing school segregation and

proxy to EPA identified environmental hazard sites (including Superfund sites), Stretesky

& Lynch (2002) find schools with high rates of racial and ethnic minority student

enrollment and high rates of students participating in free or reduced lunch programs are

closer in distance to hazard sites. Environmental justice studies have grown

exponentially over the past twenty years; however, environmental studies remain

underrepresented (constitute less than 4%) of the criminology literature (Burns et al.,

2008; Lynch, McGurrin, & Fenwick, 2004).

In sum, the spatial distribution of lead appears as nonrandom. Data indicates that

lead highly concentrates in some areas over others, resulting in lead hotspots. Studies

suggest lead hotspots tend to be urban areas, have high rates of racial and ethnic minority

populations, have high rates of poverty, and evince trends of residential mobility. This, in

turn, has resulted in disproportionately high exposure patterns for minority groups,

particularly African Americans, Hispanics, city residents, and low income individuals.

Combined with the adverse behavioral outcomes associated with lead, it is

suggested lead exposure could contribute to disproportionate rates of crime and

delinquency in: urban areas, areas with high rates of poverty, areas with high percentage

of racial and ethnic minorities, and residential mobility. A consequence of this would be

disproportionate rates of racial and ethnic minorities, urban residents, and low income

individuals involved in the criminal justice system. In spite of criminology

acknowledging race, ethnicity, and urban disadvantage as correlates of community crime

rates, lead exposure’s relationship with crime has yet to be tested at the community level.

Further, lead exposure has yet to be integrated into community level crime theories.

Chapter Three:

Mainstream Interpretations of Crime and Place

In the previous chapter, it was suggested that the distribution of lead and crime are

related, and that as a result, the distribution of lead may play a role in explaining the

geographical distribution of crime. However, lead represents a variable that has yet to be

included in mainstream criminology theory and community-level empirical assessments

of crime. What remains to be discussed in this study are the explanatory variables that

have been included in mainstream criminology theory and community-level empirical

assessments of crime. The aim of chapter three is to review literature that documents

what mainstream criminologists have established about the spatial distribution of crime.

Also presented are the community level policy initiatives that have been informed by

mainstream criminology theory and research. Chapter three concludes by arguing why

these theories and policies are limited with regard to lead hotspots.

The Spatial Distribution of Crime

According to FBI’s Uniform Crime Report, in the year 2010, there were

approximately 10,329,135 index crimes reported in the United States. Of these crimes,

1,246,248 (12%) were violent crimes and 9,082,887 (88%) were property crimes.

Estimates for violent crime were obtained by adding the totals of reported murders

(14,748), rapes (84,767), robberies (367,832), and aggravated assaults (778,901).

Estimates for property crime were obtained by adding the totals of reported burglaries

(2,159,878), larceny-thefts (6,185,867), and motor vehicle thefts (737,142). Alternatively

phrased, the sum of index crimes can be broken down as: 59.8% larceny-theft, 20.9%

burglary, 7.5% aggravated assault, 7.1% motor vehicle theft, 3.6% robbery, .8% rape, and

.1% murder.

The distribution of index crime, however, is not random. For example, when

index offenses are broken down by region, differences between the Western, Southern,

Northeastern, and Midwestern United States can be observed. The UCR indicates the

property crime rate (per 1,000 inhabitants) in the South (34.39) is 1.19 times higher than

the property crime rate in the West (28.87), 1.21 times higher than the Midwest (28.34),

and 1.63 times higher than the Northeast (21.16). A similar pattern emerges across

regions for violent offenses.

Crime also is distributed differentially across states. Table 1 uses UCR data to

display the top and bottom five states with respect to rate of index offending. Each state’s

overall crime rate was calculated using the number of index offenses committed in each

state and each state’s population. The national index crime rate in 2010 was

approximately 33.46 offenses per 1,000 U.S. residents. The state with the highest index

crime rate was South Carolina, with 44.98 index offenses per 1,000 inhabitants.

According to this data, offenses in South Carolina occurred twice as frequently as

offenses in North Dakota, South Dakota, or Idaho (the bottom three states for index

offending rates, respectively). Offenses in South Carolina, Tennessee, Texas, Louisiana,

and Florida occurred at a rate that is more than twice as frequent as offenses in North

Dakota. The disproportionality of crime has also been evinced at the county level (Baller

Table 1. Top and Bottom Five States for Rate of Index Crimes, per Uniform Crime

Report, 2010

Top 5

Bottom 5

Name

Rate*

# Offenses

Name

Rate*

# Offenses

South

Carolina

44.98

208,055

North Dakota

19.93

13,408

Tennessee

42.71

271,053

South Dakota

21.21

17,268

Texas†

42.33

1,064,477

Idaho

22.17

34,751

Louisiana

41.97

190,243

New York

23.33

452,138

Florida

41.01

771,004

New

Hampshire

23.53

30,980

* Per 1,000 inhabitants

†Texas alone accounted for approximately 10.31% of all index crime in the U.S. in 2010

Table 2. Annual Rates and Counts of Index Crimes by Community Type, per Uniform

Crime Report, 2010-2008

2010

2009

2008

Community

Type

Crime

Type

Rate* Count

Metropolitan Violent 4.28 1,107,026 4.59 1,177,758 4.89 1,242,047

Statistical Non- 30.47 7,874,193 31.60 8,113,233 33.52 8,514,199

Area Violent

Total 34.75

Cities

Outside

Metropolitan

Statistical

Area

Violent

Non-

Violent

Total

4.00

36.00

40.00

724,070

804,410

36.58

40.55

733,161

812,607

37.46

41.38

747,713

825,945

Non-

Metropolitan

Counties

Violent

Non-

Violent

1.95

16.06

58,882

484,624

2.02

15.70

61,194

474,577

2.05

16.81

61,733

506,003

Total

18.00

543,506

17.72

535,771

18.86

567,736

* Per 1,000 inhabitants

et al., 2001; Beaver & Wright, 2011; Kposowa, Breault, & Harrison, 1995; Phillips,

2006).

Crime rates also vary by community type or with respect to the designation of

communities on a continuum from metropolitan to nonmetropolitan areas. Table 2

presents UCR data from the years 2008, 2009, and 2010, classified by community type.

The UCR rates and counts are categorized by area of occurrence, specifically,

metropolitan statistical area, cities outside metropolitan areas, or nonmetropolitan

counties. According to these figures, approximately 87% of all index offending in 2010

occurred in metropolitan statistical areas.

In 2010, rates of violent index offending were highest in metropolitan statistical

areas, with violent offenses being more than twice as frequent in these areas than

nonmetropolitan counties. Metropolitan statistical areas accounted for 89% of all violent

index offenses in 2010, and 87% of non-violent (property) crimes. Further, these

observations appear stable across each year. These data are consistent with research that

finds index crime, like lead, concentrates in cities and urban areas (Bursik, 1984; Cullen

& Levitt, 1999; Venkatesh, 2000; Massey, 1990; Massey & Denton, 1993). This appears

particularly true of violent crime (see also Krivo & Peterson, 1996).

Crime is also found to be distributed non-normally within community areas,

including across census tracts (Kubrin, Squires, Graves, & Ousey, 2011; Krivo &

Peterson, 1996; Garcia, Taylor, & Lawton, 2007; Griffiths & Chavez, 2004; Mears &

Bhati, 2006; Morenoff, Sampson, & Raudenbush, 2001; Sampson & Wilson, 1995;

Logan & Stults, 1999). For example, Krivo and Peterson (1996) examine index crime

using 1990 census data from 177 tracts in Columbus, Ohio. They find a strong, positive

correlation between violent crime and extreme disadvantage (r=.717), and find rates of

both property and violent crimes to be significantly higher in extremely economically

disadvantaged tracts (Krivo & Peterson, 1996). Mears and Bhati (2006) study

neighborhoods in Chicago and find that a one standard deviation increase in

neighborhood resource deprivation is associated with a corresponding 60% increase in

homicides.

More recently, criminologists have taken up the study of crime across

microgeographical areas (addresses, streets, blocks), where they have continued to find

evidence of non-random distributions of crime (Braga, Papachristos, & Hureau, 2010;

Braga & Weisburd, 2010; Eck, Clarke, & Guerette, 2007; Sherman, Gartin, Buerger,

1989; Weisburd, Morris, & Goff, 2009). In a widely-cited study, Sherman et al. (1989)

analyze 323,979 police calls for service over the course of one year, across 115,000

addresses and intersections in Minneapolis. Their findings revealed that 50% of all calls

originated at 3% of all places (Sherman et al. 1989). Alternatively stated, approximately

161,990 calls came from 3,450 areas, meaning that each of these areas where calls for

service are concentrated generate approximately 47 annual calls for service.

In their study, Weisburd et al. (2009) examined juvenile crime on street segments

in Seattle across a fourteen year time frame (1989-2002). Their findings indicate that

across all years studied, 100% of juvenile arrests occurred within 2-5% (less than 1,400)

of all street segments (n=29,849) in Seattle. The same study found over the same time

frame that 50% of all juvenile arrests occurred within less than 1% (less than 299) of all

of Seattle’s street segments (Weisburd et al., 2009). Further, Weisburd and colleagues

(2009) state that the locations of these delinquency concentration areas appeared

relatively stable over the course of fourteen years.

Taken together, areas with high concentrations of crime in space have been

defined by criminologists as crime hotspots (Eck et al., 2005; Paulsen & Robinson,

2009). On the issue of defining what is meant by crime hot spot, Eck and colleagues

(2005) state, “Though no common definition of the term hot spot of crime exists, the

common understanding is that a hot spot is an area that has a greater than average number

of criminal or disorder events, or an area where people have a higher than average risk of

victimization” (p.2). This study draws on Eck and colleagues (2005) understanding of

crime hotspots to facilitate an analysis of the relationship between lead and crime

hotspots at the community level. For purposes of this research the term “area” is

conceptualized as community area as defined by the U.S. Census Bureau and the

University of Chicago (Chicago Police Department, 2008).

Explaining Crime Hotspots

While the study of crime hotspots has yielded a great deal of attention from

criminologists in recent years (Lersch & Hart, 2011), researchers have been observing the

existence of crime hotspots for over a century. In doing so, scholars have documented a

number of community-level attributes that correlate with crime hotspots. Further,

researchers have incorporated these covariates into theories that attempt to explain why it

is that crime is concentrated in some areas over others. The observation of crime

hotspots, correlates of crime hotspots, and explanations of crime hotspots are all rooted in

the work of nineteenth century statisticians and sociologists (Levin & Lindesmith,1937).

The cartographic school of criminology. Levin and Lindesmith (1937) attribute the

earliest systematic studies on the ecology of crime to the work of attorney and

cartographer Andre-Michel Guerry and astronomer and statistician Adolph Quetelet.

Guerry used cartography to map occurrences of crime in France. As part of his method,

Guerry divided France into five sections (north, west, south, east, and centre), and

mapped data on rates of crime by type from the years 1825 to 1830. Guerry also

employed maps to study co-occurance of crime and gender, age, climate, education,

illegitimacy, and philanthropy (Paulsen & Robinson, 2009). Among his findings were the

observations of higher rates of property crime than violent crime in Northern France,

while the opposite appeared true in Southern France. From his maps, it became Guerry’s

speculation that poverty, lack of education, and population density acted as causes of

crime.

Quetelet earned his doctorate in mathematics from the University of Ghent. He

subsequently became interested in whether laws of regularity found in the physical

sciences applied to human behavior. Quetelet, like Guerry, was among the first to employ

probability and statistics to assess demographic data (Beirne, 1987). In doing so, he was

able to quantitatively

assess patterns and consistencies in human behavior, including crime and place. In, “A

Treatise on Man: and the Development of his Faculties,” Quetelet (1842), analyzes data

on the distribution of crime across France and across countries. Congruent with Guerry,

Quetelet found that property crime in France occurred with a greater frequency than

violent crime (approximately 3:1). He also observed fluctuation in crime rates across

departments of France, by dividing departments into three classes: (1) departments where

rates of violent and property crime were above average (e.g. Corse, Landes), (2)

departments where rates of violent and property crime were below average (e.g. Creuse,

Indre), and (3) departments where either rates of violent or rates of nonviolent crime were

below average (e.g. Var, Hautes-Alpes). Quetelet observed that departments that had

more industrial establishments, greater inequity between “riches and wants,” greater

influxes of people, and greater heterogeneous racial and ethnic populations also displayed

evidence of higher crime rates. On this matter, Quetelet concludes:

“The countries where frequent mixture of the people takes place; those in which

industry and trade collect many persons and things together, and possess the

greatest activity; finally, those where the inequality of fortune is most felt, all

things being equal, are those which give rise to the greatest number of crimes,”

(1842, p.95).

The work of both Guerry and Quetelet inspired English scholars, where after

studies on the ecology of crime continued to emerge in the nineteenth century. For

example, Levin and Lindesmith (1937) describe the work of Henry Mayhew, who used

official statistics to map juvenile delinquency, female criminality, recidivism, and other

types of crime across London, as well as across counties in England and Wales from 1841

to 1850. Much like contemporary UCR data reveals, Mayhew found counties that had

large cities in them had higher rates of juvenile delinquency. Much like contemporary hot

spot research, Mayhew broke down the county containing London and found 41% of all

juvenile offenders were concentrated into one of seven police districts, with another 24%

concentrated to another. These studies by Guerry, Quetelet, Mayhew, and others

concerning the distribution of crime from 1830 to 1880, comprise what has been referred

to as the cartographic school of criminology (Paulsen & Robinson, 2009).

The contributions of the cartographic school of criminology, especially the work

of Quetelet, would go on to influence the research of French sociologist Emile Durkheim.

In the late 1800s, Durkheim studied the sociological changes associated with the

emerging shift from agrarian-based economies to industrialized economies, or what he

described as the transition from mechanical to organic societies. Durkheim (1892)

theorized that these shifts resulted in a phenomenon that he termed anomie—which

translates literally into “normlessness.” Anomie is described by Durkheim as being

present when fast-paced societal changes take place, producing a state of normlessness

for individuals in which individuals become unclear concerning the content of the

collective conscience. Durkheim argued that these rapid structural changes disrupt the

collective conscience, or a society’s uniform sense of behaviors that are right or wrong

(Lersch & Hart, 2011). Durkheim went on to explain that when individuals cannot rely

criminology, see Beirne (1987).

on the collective conscience to regulate their behavior, they rely only on their own

hedonistic tendencies, and their wants and desires become unchecked. This observation

has been interpreted as a society losing its ability to control the behavior of its people

(Kornhauser, 1978). Durkheim (1897) also pointed out that as wants and desires grew,

and there was also a failure in the structure of society increase the means to meet these

wants and desires, feelings of despair and frustration emerge, a position linked to social

strain (Agnew, 1992; 2001; 2005).

Durkheim examined the effect of anomie in a study of suicide rates across Europe.

Durkheim, like Guerry, Quetelet, and Mayhew, used quantitative data and cartography to

study human behavior. When he mapped suicide rates across gender, region, ethnicity,

and religion, he observed different rates of these behaviors across the subgroups under

examination. Among his findings were that in times of economic transition (anomic

conditions), suicide rates increased (Durkheim, 1897). Durkheim

(1897) also describes the “world of trade and industry” as inherently anomic, and finds

higher suicide rates among professionals employed by trade and industrial sectors.

Durkheim (1897) coined the phrase anomic suicide, which he described as depending,

“not on the way in which individuals are attached to the society but on the way in which

it controls them,” (p. 283). While Durkheim never wrote about crime or delinquency

explicitly in Suicide, his work is often regarded as the cornerstone for many

contemporary criminological theories (Vold & Bernard, 1979).

In sum, the existence of crime hotspots has been demonstrated by researchers

since the nineteenth century. The earliest studies of crime and place appeared from

18301880, and are referred to as the cartographic school of crime. The earliest observed

correlates of concentrated crime were reported by Guerry, Quetelet, and Mayhew as

poverty, education, industry, population density, race, ethnicity, and mobility.

Explanations as to why human behavior differs across place are elaborated upon by

Durkheim in the 1890s, when he introduces the concept of anomie. It is noteworthy that

the correlates of crime hotspots established by the Cartographic School of Crime would

later come to be established as correlates of lead hotspots (Diawara et al., 2006; Jacobs et

al., 2002; Lanphear et al., 1996; Lynch, 2004; Mielke et al., 1999; Sargent et al., 1997;

Stretesky, 2003;).

The Chicago school. In the 1920s and 1930s, sociologists affiliated with the

University of Chicago and the Institute for Juvenile Research in Chicago, like Durkheim,

became interested in the impact of industrialization on deviant behavior, crime and social

disorganization. Specifically, these scholars were interested in the relationship between

industrialization and urban ecology. At the turn of the century, the growth of industry in

the United States spurred changes in the composition of American cities, including the

expansion of industry as well as growth in immigrant populations. The city of Chicago

exemplified much of the changes typical of American cities at the time, and therefore

acted as “the perfect natural laboratory” (Kubrin, Stuckey, & Krohn, 2009, p.83) for the

study of industrialization and urban sociology. These studies included observations of

community level crime and delinquency. Given their university affiliation, and setting for

their work, these scholars and their studies are commonly referred to as the Chicago

school.

Two of the most prominent scholars of the Chicago school were Robert Park and

Ernest Burgess. Park and Burgess (1925) drew parallels between patterns of survival and

behavior in natural ecosystems and patterns of survival and behavior in cities. Park and

Burgess argued that cities expanded outward in concentric zones, and developed a model

referred to as the Concentric Zone Theory. Park and Burgess observed five concentric

zones, with the first centermost zone being the Central Business District (CBD). At the

turn of the century the CBD was quickly growing and expanding, and continually

encroaching on communities in zone 2 -- the transitional zone. The second zone was

termed the transitional zone because of the high rates of change the CBD imposed on this

zone. The transitional zone was observed to carry higher concentrations of immigrant

residents, deteriorated housing, and factories than other zones (Park & Burgess, 1925).

Often, residents who could afford to leave the transitional zone did so, while residents of

lower socioeconomic status were restricted to this area. Park and Burgess termed zone 3

the working-class zone, and this area was predominately occupied by skilled blue collar

workers, housed in tenements. Finally, the outmost zones (zone 4 and 5) were termed the

residential zone and commuter zone, respectively. These zones were comprised of

suburban communities commonly occupied by middle and upper class families.

Park and Burgess were human ecologists, and were not analyzing crime and

delinquency expressly. However, the concentric zone theory is relevant to the discussion

of crime because the concentric zone model presented the composition of urban

landscape in a way that would allow social scientists to explore the relationship between

community-level characteristics and human behavior (including crime) in the new

American city.

Social disorganization theory. The first social scientists to incorporate the

concentric zone theory in an explanation of community level differences in crime were

Clifford Shaw and Henry McKay (1942). Shaw and McKay were interested in the

relationship between community processes and delinquency. Shaw and McKay justified

research in this area because a theory that describes why crime concentrates in some

areas over others had yet to be proposed. To develop such a theory, Shaw and McKay

mapped Park and Burgess’ concentric zones in the city of Chicago, and simultaneously

mapped the addresses of juvenile delinquents. In their study, Shaw and McKay found

disproportionately high levels of delinquency in areas closest to the CBD (zones I and II).

Their explanation as to why crime concentrates in these communities is called social

disorganization theory (Shaw & McKay, 1942).

In its earliest form, social disorganization theory explained crime using concepts

found in cultural, strain, and control theories. Most criminologists, however, understand

social disorganization theory as it was interpreted in 1978 by Ruth Kornhauser.

Kornhauser understands social disorganization theory as a community-level control

theory. Social disorganization has been described as: “an inability of inhabitants to

control the behavior of residents and users of neighborhood space because of deleterious

social conditions,” (Paulsen & Robinson, 2009, p.53). Social disorganization parallels

Durkheim’s discussion of anomie, and failure to recognize the collective conscience,

because residents of socially disorganized communities are theorized to lack a unified set

of community values. Criminologists argue that residents of socially disorganized

communities do not foster relationships with one another, and demonstrate few or no

social ties with each other (Sampson, Raudenbush, & Earls, 1997). That is, residents may

not spend time with one another, may be unable to identify their neighbors, or may have

little in common with other residents (Kubrin et al., 2009). Conversely, socially

organized communities are understood as areas where residents interact, know each other,

and share common interests and values (Kubrin et al., 2009).

Social disorganization theory suggests the degree to which residents are bonded to

one another, and committed to a unified set of community values, translates into the

degree of informal social control the community is able to exercise (Sampson et al.,

1997). Informal social control is theorized to be an important mechanism for reducing

community crime and delinquency. For example, in socially organized communities,

residents may frequently occupy public space (walking dogs, jogging, etc.) and

inadvertently reduce crime by maintaining physical presence in the community (e.g.,

reduce criminal opportunities). Residents who know and recognize each other’s children

may be more likely to intervene personally to stop delinquency and status offending.

Thus these informal actions are theorized to decrease offending.

In contrast, in socially disorganized communities, the absence of informal social

control is theorized to promote crime and delinquency (Kornhauser, 1978). Just as

Durkheim related suicide to anomie, Shaw and McKay connected delinquency to lack of

informal social control. For example, in socially disorganized communities, residents

may be reluctant to occupy public spaces, as they may have little or no desire to interact

with neighbors. In these areas it is also possible that residents may not detect the

presence of a community to integrate themselves into. Without uniform community

values, the wants of residents become unrestrained, and residents may perceive desirable

offending opportunities in their neighborhoods. Absent of relationships with neighbors,

residents may be more likely to victimize their neighbors. Residents may also be

dissuaded from intervening to stop delinquent acts or status offenses if the community’s

children are strangers to them.

Key variables. Shaw and McKay argue that three variables in particular impact

community social (dis)organization. Shaw and McKay list these variables as (1)

socioeconomic status (SES), (2) racial and ethnic heterogeneity, and (3) residential

mobility. It is noteworthy that Shaw and McKay’s three variables mirror the crime

correlates established by the Cartographic School of Crime, demonstrating a time-tested

stability between these factors and index crime. Social disorganization theory posits that

these three factors act exogenously to impact informal social control and, in turn, crime.

That is, social disorganization theory holds that informal social control mediates the

relationship between SES, racial and ethnic heterogeneity, residential mobility and crime

(Sampson & Groves, 1989). It is hypothesized that SES co-varies positively with

informal social control. At the same time, it is hypothesized that racial and ethnic

heterogeneity and residential mobility have an inverse relationship with informal social

control (Kornhauser 1978; Morenoff et al., 2001; Sampson & Groves, 1989). Social

disorganization theory holds that informal social control has an inverse relationship with

crime and delinquency. In this section, the exogenous sources of social disorganization—

SES, racial and ethnic heterogeneity, and residential mobility-- are described in turn.

Socioeconomic status (SES). Social disorganization theory suggests that poverty

weakens the strength of community ties, and in turn, threatens social control. Poverty has

been defined by social disorganization theorists as, “the presence of lower class people in

a community,” (Paulsen & Robinson, 2009). Criminologists have theorized that poverty

relates to crime because inequity drives individuals to engage in criminal behaviors

(Michalowski, 1985; Sampson, 1995), and impoverished areas lack capital and resources

that promote pro-social human behavior. In tests of social disorganization theory, poverty

is often measured by calculating the proportion of lower class individuals in a

community, often using census data.

The relationship between poverty and crime has been demonstrated empirically in

numerous studies (Bursik, 1984; Krivo & Peterson, 1996; Pratt & Cullen, 2005; Sampson

& Groves, 1989; Shaw & McKay, 1942; Venkatesh, 2000; Wilson 2009; for exceptions

see: MacDonald, Hipp, & Gill, 2012; Ousey & Kubrin, 2009; Sampson, 2008; Wikström

& Loeber, 2000). Shaw and McKay (1942) found that areas of Chicago with the highest

rates of committed youth also had the highest percentages of families on welfare and

lowest median monthly rent. In their test of social disorganization theory, Sampson and

Groves (1989) found that as socioeconomic status increased, there was a corresponding

decrease in community rates of burglary, auto-theft, vandalism, and presence of street

corner peer groups. A meta-analysis of mainstream criminology’s macro-level predictors

and theories of crime finds the effect of poverty on crime to be strong and stable across

over 200 studies (Pratt & Cullen, 2005). Bursik (1984) finds a strong, positive

correlation between community delinquency rates and community poverty levels

(r=.676). Krivo and Peterson (1996) regressed tract-level property crime rates and violent

crime rates on a number of theoretically relevant community level contextual variables.

Their findings indicated that extreme poverty exerted a significant independent effect on

property crime (b=.2486) as well as a significant independent effect on violent crime

(b=17.3926).

Racial & ethnic heterogeneity. A second key variable theorized to impact

community social control is racial and ethnic heterogeneity. Racial and ethnic

heterogeneity has been described by criminologists as the degree to which a community

is racially and/or ethnically diverse, or the extent to which differing racial or ethnic

groups are present in a community (Paulsen & Robinson, 2009). Social disorganization

theory proposes that individuals of differing racial and ethnic backgrounds are unlikely to

share similar values, attitudes, and beliefs. Social disorganization theory further argues

that language (and other communication) barriers prevent social ties from developing

between individuals of differing racial and ethnic backgrounds. For these reasons, social

disorganization theorists argue that communities with high levels of racial and ethnic

heterogeneity are less capable of exercising informal social control, and in turn, carry

higher crime rates. Racial and ethnic heterogeneity has often been measured by using

Lieberson’s Index of Heterogeneity (Lieberson, 1969), the index of diversity, (see, for

example, Osgood & Chambers, 2000) and Herfindahl index (Gibbs & Martin, 1962) to

analyze census data. Recently, researchers have captured ethnic heterogeneity

dynamically by calculating an ethnic churning variable (Hipp, Tita, & Boggess, 2009).

Empirical support exists for the link between racial and ethnic heterogeneity and

crime (Hipp, 2007; Kubrin, 2000; Osgood & Chambers, 2000; Sampson & Groves, 1989;

Shaw & McKay, 1942; for exceptions see: Desmond & Kubrin, 2009; Stowell &

Martinez, 2007; Nielsen, Lee, & Martinez, 2005). In a study of 119 Seattle communities,

Kubrin (2000) examined the impact of racial heterogeneity on violent and property

crimes. Kubrin’s (2000) findings revealed that racial heterogeneity and violent (r=.52)

and non-violent (r=.23) crimes significantly and positively co-vary. Net of confounders,

racial heterogeneity was found to be associated with increased rates of rape, robbery,

assault, burglary, and auto-theft (Kubrin, 2000). Hipp (2007) found that across census

tracts in 19 U.S. cities, ethnic heterogeneity significantly predicted increases in rates of

aggravated assault, robbery, burglary, and motor vehicle theft. In their study, Osgood and

Chambers (2000) reported that a 43% increase in ethnic heterogeneity corresponded with

a doubling of violent index crime arrests among juveniles.

Residential mobility. The third key variable in social disorganization theory is

residential mobility. Residential mobility can be described as the rate at which residents

move in or out of a community (Kubrin et al., 2009). Researchers have observed that

some communities are relatively stable in terms of residency -- that is, residents commit

to living in their communities for decades. In contrast, some communities display a high

degree of transiency -- that is, residents engage in shorter term commitments, and

frequently move in or out of the area. Social disorganization theorists argue that

difference between stability and instability in residential composition impacts social

control because residents are less able to know, trust, and interact with one another if

residents are continually moving in or out of the community (Kubrin et al., 2009). In this

way, the ties that do form between residents in unstable communities are constantly

threatened by the high rates of population change. Moreover, social disorganization

theorists argue that residents may be more committed to a community if they have a

longterm investment in the property (e.g., mortgage) rather than shorter-term

engagements (e.g., lease to rent; e.g.Alba, Logan, & Bellair, 1994; DiPasquale & Glaeser,

1999; Squires & Kubrin, 2006 ). That is, residents with long term commitments may be

more compelled to stop crime in their communities, because they are less likely/able to

leave the area in the event that crime becomes a widespread community problem.

Furthermore, crime threatens property values, and mortgage holders may be more

compelled to control crime if they stand to lose money. Residential mobility has been

measured using census data, including percentages of renters, or proportion of individuals

who had moved from another dwelling within the past five years (Kubrin, 2000; Osgood

& Chambers, 2000;

Sampson et al., 1997; Warner & Pierce, 1993). .

Numerous studies have documented a link between increases in residential

mobility and increases in crime (Osgood & Chambers, 2000; Sampson, 1985; Sampson &

Groves, 1989; Shaw & McKay, 1942; Warner & Rountree, 1997; For exceptions see:

Kingston, Huizinga, & Elliott, 2009; Warner & Pierce, 1993). Shaw and McKay (1942)

found that areas with the lowest percentages of home owners had the highest rates of

juvenile delinquency. Osgood and Chambers (2000) find that rates of arrest for juvenile

violence double with every 24% increase in residential turnover over the course of a five

year period. Warner and Rountree (1997) studied communities in Seattle and found that

increases in residential stability are associated with significant decreases in assault and

burglary rates, net of confounders. Sampson (1985) found that increases in residential

mobility were associated with increased rates of personal victimization as reported on the

National Crime Victims Survey (robbery or larceny with contact, rape, aggravated

assault, and simple assault).

Following Shaw and McKay’s 1942 analyses, studies that examined the

relationship between SES, racial and ethnic heterogeneity, residential mobility and crime

represented an important start for social disorganization theory. However, one criticism

of Shaw and McKay’s test, and a number of the subsequent studies, was that informal

social control (or disorganization itself) was often not measured in the analysis. For

many years, researchers assumed that SES, racial and ethnic heterogeneity, and

residential mobility’s impact on crime was mediated by social control (disorganization),

without actually including measures for social control (disorganization) in their model(s).

For these reasons, among others, social disorganization theory largely fell out of favor in

crime studies through the 1970s.

This changed in the 1980s, particularly in 1989 when Sampson and Groves

empirically tested social disorganization theory by including measures for social control

(disorganization). To assess the endogenous relationship of social control

(disorganization), Sampson and Groves (1989) measured strength of local friendship

networks, presence of unsupervised teenage peer groups, and degree of organizational

participation. Results from a weighted least squares regression revealed that the

relationship low SES, high ethnic heterogeneity, high residential mobility, and

community crime and delinquency rates were mediated by social control measures

(supportive of social disorganization theory’s hypotheses) (Sampson & Groves, 1989).

Sampson and Groves’ study has been referred to as a “criminological classic”

(Lowenkamp, Cullen, & Pratt, 2003, p.351), and a 2003 replication of the study

demonstrated Sampson and Groves’ findings tenable more than a decade after initial

publication (Lowenkamp et al., 2003). Furthermore, re-analysis of Sampson and Groves’

data using LISREL replication techniques revealed support for social disorganization

theory (Veysey & Messner, 1999; Lowenkamp et al., 2003).

In more recent years, a meta-analysis of macro-level crime studies (n=214) finds

social disorganization theory to garner strong support in the criminology literature (Pratt

& Cullen, 2005). Eck et al. (2005) proclaim social disorganization and ecological

theories of crime as pertinent theoretical explanations of crime in community areas.

Furthermore, Eck et al. (2005) ascribe social fragmentation as a likely cause of

community-level hotspots of crime.

Contemporary updates and expansions. As tests of social disorganization theory

continued to appear in the literature, new variables were incorporated into

communitylevel crime hotspot studies. The inclusion of these variables uncovered new

correlates of crime that became relevant to social disorganization and community crime

researchers.

For example, Sampson and Groves (1989) included family disruption and urbanization as

exogenous variables in their test of social disorganization theory, and found a positive,

significant relationship between these variables, measures of social disorganization, and

crime. In addition, contemporary tests of social disorganization theory uncover

relationships between population density, unemployment, percent of young males and

crime (Kubrin et al., 2009). Because a number of these variables are highly interrelated,

researchers have needed to address multicolinearity in their models. To overcome this issue,

social disorganization researchers have constructed a variable called “concentrated

disadvantage” which contains measures for poverty, unemployment, racial composition,

divorce, female-headed households, and households on public assistance (Kubrin et al.,

2009; MacDonald & Gover, 2005; Sampson, Raudenbush, & Earls, 1997). MacDonald and

Gover (2005) found concentrated disadvantage a significant predictor of youth-onyouth

homicide across 159 U.S. cities.

As models of social disorganization expanded to include new variables,

reformulations of social disorganization theory began to emerge in the criminology

literature. Clear (2007) identifies three of the most prominent efforts to update social

disorganization theory as: (1) Sampson’s Collective Efficacy, (2) Bursik and Grasmick’s

Systemic Theory, and (3) Weatherburn and Lind’s Economic Stress-Induced Offender

Motivation (ESIOM) Paradigm. These theories focus on explaining community level

crime differences, and add to social disorganization frameworks the idea of social

cohesion and trust among community members (Sampson et al., 1997), differing types of

social control (Bursik & Grasmick, 1993), and persistent economic hardship

(Weatherburn & Lind, 2001). Clear (2007) also presents Coercive Mobility as a new and

relevant theoretical framework that includes the impact of concentrated incarceration on

community level crime hotspots (see also Clear, Rose, Waring, and Scully, 2003).

In sum, with respect to mainstream criminology theories designed to explain

community level crime hotspots, social disorganization theory emerges as a leading

theoretical explanation (Eck, 2005). Moreover, contemporary reformulations, updates,

and extensions of social disorganization theory imply that criminologists continue to

perceive community level crime hotspots as a relevant social problem. It also implies

that criminologists continue to search for more refined theoretical explanations of

community level crime hotspots. In spite of these latter points, and in spite of lead’s

empirical relationship with race, ethnicity, residential mobility, urbanization, and poverty,

lead has yet to be factored into empirical assessments of community level crime.

Crime Hotspot Policy Initiatives

Interest in crime hotspots is not restricted to criminologists and social science

researchers. Policy makers, practitioners, government representatives, and law

enforcement officials have also demonstrated a vested interest in addressing community

level crime hotspots. That said, social disorganization theory has informed numerous

policy initiatives, including some programs that have been in effect for over seventy

years (Lersch & Hart, 2011). A number of these programs have grown in size and scope,

and consume considerable amounts of tax dollars and resources. Lersch and Hart (2011)

highlight three examples of policy initiatives influenced by the tenets of social

disorganization theory, including: The Chicago Area Project, community policing, and

various federal government directives.

The Chicago area project (CAP). The Chicago Area Project (CAP) was

founded in the 1934 by Clifford Shaw himself. CAP began in three Chicago

communities which demonstrated high rates of juvenile delinquency. Shaw emphasized a

“self-help” approach in reducing community crime rates, and believed that juvenile

delinquency in these areas could be reduced by organizing parents and residents of the

community as well as approaching youth and attempting to dissuade them from

delinquency. Shaw’s initiatives spread throughout Chicago. By the early 1970s, CAP

was working with twenty-two community committees, and by the 1980s, CAP was

inspiring similar programs across Illinois. CAP remains active today, and collaborates

with over 40 grassroots organizations and projects that promote community-level

advocacy, organizing, and service provisions. CAP proclaims its mission statement as

one that has not changed from the programs inception: “To work toward the prevention

and eradication of juvenile delinquency through the development and support of affiliated

local community self-help efforts in communities where the need is greatest,” (Chicago

Area Project, 2012).

CAP’s mission relates to the tenets of social disorganization theory in that CAP

holds that the solution to community rates of juvenile delinquency rely in unifying and

organizing community members. The parallels between social disorganization theory and

CAP is even more clearly conveyed in CAP’s goal, which is stated as: “to develop special

projects and established locally controlled organizations that implement the directives put

forth in CAP’s mission and philosophy. Projects and affiliates are mandated to positively

impact areas in the Chicago vicinity with high rates of juvenile delinquency or other

symptoms of social disorganization,” (Chicago Area Project, 2012a). It would seems,

then, that CAP is expressly interested in establishing community social control (e.g.,

“locally controlled organizations) and ameliorating social disorganization, because these

directives are thought to reduce juvenile delinquency

(CAP’s mission).

While CAP’s mission is amicable, some evaluations of CAP describe CAP’s

impact on juvenile crime as, “negligible at best,” (Lersch & Hart, 2011, p.132). It has

been pointed out that even Shaw relied increasingly less on official juvenile crime data to

demonstrate CAP’s effectiveness (Lersch & Hart, 2011; Schlossman, Zellman, and

Shavelson, 1984). Regardless, there are currently 55 organizations in high-risk

communities in Chicago that utilize the CAP model, and over 200 across the state of

Illinois (Chicago Area Project, 2012b) . Funding for CAP draws heavily from Illinois’

Community Youth Services grant line, but CAP also obtains funding from donors. While

funding for the Community Youth Services grant line continues to decline, and has been

threatened with elimination, in 2010 $5.4 million was appropriated to the line (Illinois

Human Services Commission, 2011). CAP representatives suggest this grant line serves

70,000 children, whom CAP argues may otherwise be at risk of entering the justice

system (Sardin, 2011).

Community policing. Social disorganization theory informs a widespread policy

movement in law enforcement termed community policing. Community policing has

been described as a philosophy of policing that emphasizes partnership between law

enforcement and the community in identifying and resolving local crime problems

(including delinquency, fear of crime, disorder, and decay). These sentiments are

conveyed by Wilson and Kelling (1982), who state: “The essence of the police role in

maintaining order is to reinforce the informal control mechanisms of the community

itself.” The key components of community policing have been listed as community

partnerships, organizational transformation, and problem solving (Department of Justice,

2009). Examples of community policing initiatives include the Department of Justice’s

Weed and Seed Program, and Chicago’s Chicago Alternative Policing Strategy (CAPS).

Both of these programs are designed to encourage law enforcement to work in

conjunction with the community residents to identify, prevent, and respond to local crime

problems. These programs work towards these objectives by holding meetings with

residents, or having community-assigned beat officers.

Community policing is related to social disorganization theory because it

emphasizes the use of law enforcement to promote informal social control (Wilson &

Kelling, 1982). In a review of research that studies police effectiveness, Weisburd and

Eck (2004) find that community policing efforts appear to reduce community fear of

crime, and show promise in reducing crime when paired with problem oriented policing

strategies. However, the same study fails to find a consistent relationship between

families moving from welfare to workforce. It has been estimated that WIT has successfully placed

approximately one third of its clients into employment positions (Lersch & Hart, 2011).

community policing strategies alone and occurrences of crime or disorder (Weisburd &

Eck, 2004). Regardless, recent reports suggest community policing strategies are widely

implemented by law enforcement. A 1997 study of 1,637 law enforcement agencies

revealed 58% of agencies had implemented a community policing strategies, and another

27% were in the process of implementing community policing (Fridell, 2004). The

Bureau of Justice Statistics reported that in 2003, 58% of police departments (that

employed 82% of all officers) used full-time community policing officers (Hickman &

Reeves, 2006). Furthermore, almost half of all departments (employing 73% of officers),

advertised a mission statement that included aspects of community policing (Hickman &

Reeves, 2006). Since the 1990s, community oriented policing initiatives have received

over $11 billion in federal grant support.

Federal government initiatives. In addition to CAP and community policing,

social disorganization theory influences a number of policy initiatives at the federal level.

In 1995, the Office of Juvenile Justice and Delinquency Prevention (OJJDP) compiled a

matrix of community based initiatives designed to prevent violence and/or create

economic opportunities that may, in turn, prevent delinquency. This matrix identified

government and private entities. The OJJDP identified 36 programs in total, including: 7

through the Department of Housing and Urban Development, 6 through the Department

of Justice (including community policing), 5 through the Department of Justice, 5 though

the Department of Health and Human Services, and 3 though the Department of Labor.

These projects vary in goals and implementation, but share the commonality that they all

include an, “interdisciplinary local planning board that has included in its focus

improving the lives of at-risk children and families,” (OJJDP, 1995, p.1). Because these

programs work towards restoring communities and orienting neighborhoods towards

selfhelp and informal social control, they incorporate social disorganization theory

ideology. One example of a federal government initiative with these objectives is the

Department of Housing and Urban Development’s Empowerment Zone (EZ), Enterprise

Community (EC), and Renewal Communities (RC) endeavors. These initiatives

designate communities as an EZ, EC, and/or RC, which in turn entitles communities to

federal resources (e.g., programs which work with schools in designated areas to reduce

truancy, increase academic performance, etc.). These designations also offer incentives

to entrepreneurs to create market opportunities in such communities (e.g., recruit

employees locally). Lersch and Hart (2011) estimate that collectively entrepreneurs who

establish businesses in EZ/EC/RC zones are eligible for up to $10 billion in incentives

from the federal government. Lersch and Hart (2011) identify EZ/EC/RC efforts to

rebuild economically depressed areas as a “necessary ingredient” to reduce community

crime and delinquency. Sherman et al. (1997) describe these programs as promising

when a particular goal is explicated.

The Problem

Analysis of mainstream community and crime literature reveals that while many

variables have been tested with respect to explaining the distribution of crime, to date,

only one unpublished study examining community levels of lead and homicide (Lynch,

McGurrin and Stretesky, 2001) has tested the role lead may play in explaining

community level differences in crime and delinquency rates. This void creates

limitations in the study and control of crime across communities with respect to (1)

theory and (2) policy.

Theory. The absence of lead from the study of community crime is problematic

for mainstream macro level crime theory because it is possible that observed associations

between mainstream predictors and crime are biased. There are a few reasons to suspect

this.

One of which is lead’s demonstrated ability to cause changes in human behavior

(Canfield et al., 2003; Landrigan, 2000; Lanphear et al., 2005; Lanphear et al., 2000;

Needleman et al., 1990; Needleman et al., 1979; Schwartz, 1994; Wasserman et al., 2000;

White et al., 1993). Impacts that lead has on human behavior include diminished IQ,

impulsivity, aggression, mood disorders, and decision making. Because many of these

factors have been found to be correlates of crime, it is reasonable to suspect that the

presence of lead in an environment may set into motion criminal behaviors and

delinquency. For example, it is reasonable to infer that the presence of lead may

adversely impact the ability of a community to exercise informal social control, and/or

hamper collective efficacy. Because lead impacts IQ, intelligence, academic

performance, and cognition (Canfield et al., 2003; Landrigan, 2000; Lanphear et al.,

2005; Lanphear et al., 2000; Needleman et al., 1990; Needleman et al., 1979; Schwartz,

1994; Wasserman et al., 2000; White et al., 1993; Zahran et al., 2009), high lead body

burden may adversely impact community residents’ ability to communicate with each

other. Diminished IQ and intelligence also may impact a community’s ability to

establish, recognize, or articulate a unified set of positive values and norms. Because

exposures to lead is associated with mood and affect (Burns et al., 1999; Cecil et al.,

2008; Bouchard et al., 2009), lead exposure is highly likely to impact the social cohesion

in a community. Relatedly, community context of collective efficacy is described as “the

linkage of mutual trust and the willingness to intervene for the common good,”

(Sampson, Raudenbush, & Earls, 1997, pg. 919). Given lead’s association with decision

making (Cecil et al., 2008), it is reasonable to suspect that community lead levels may

impact the “willingness to intervene” component of collective efficacy. Taken together, it

is reasonable to infer that the presence of lead in a community may play out collectively

as diminished social control and/or collective efficacy.

Moreover, because social disorganization theory is presented as a general theory

of crime, it proposes to explain variation across all types of community crime. That is,

informal social control is proposed by social disorganization theorists to suppress

property crime as well as violent crime. Thus, if lead impacts the ability of a community

to exercise informal social control, variations in both property and violent crime can be

anticipated. This is consistent with findings that observe associations between lead,

crime, and delinquency inclusive of violent and property offenses (Masters et al., 1994;

Stretesky & Lynch, 2004). However, tests of mainstream macro-level crime theory do

not include lead as a variable.

Another reason the relationship between mainstream macro level predictors of

crime and community crime rates may be biased, relate to the patterns in the spatial

distribution of lead and crime. Evidence suggests lead, crime, and mainstream theoretical

predictors of crime follow similar spatial distribution patterns. For example, social

disorganization (emerging as a leading mainstream explanation for community-level

differences in crime—see Eck, 2005; Pratt & Cullen, 2005) explicates three variables

specifically that correlate with crime: SES, racial/ethnic heterogeneity, and residential

mobility. However, as demonstrated in chapter two, literature that has studied the

distribution of lead has found lead to be correlated with: impoverished communities

(Diawara et al., 2006; Jacobs et al., 2002), neighborhood racial composition (Stretesky,

2003; Mielke et al., 1999; Lanphear et al., 1996), neighborhood ethnic composition

(Jacobs et al., 2002), and areas with demonstrated residential mobility (Sargent et al.,

1997). Thus, lead, crime, and delinquency appear pulled together in time and space

(Lynch, 2004). This has yet to be empirically examined at the community level (but for

exception, see Lynch, McGurrin, and Stretesky, 2001).

Results of macro-level assessments of lead and crime provide additional evidence

that suggest observations between traditional crime predictors and community crime rates

is bias (Lynch, 2004; Masters et al., 1994;Nevin, 2000; Nevin, 2007; Reyes, 2007;

Stetesky & Lynch, 2001; Stretesky & Lynch, 2004). Because of lead’s demonstrated

association with criminal behavior in cities, counties, and countries, it is reasonable to

suspect that high concentrations of lead in communities may lead to collective

neighborhood dysfunction, including criminal behavior. This argument is strengthened

by the fact that prior macro-level studies have found that lead exerts an effect on crime

rates net of other factors including: unemployment (Masters et al., 1994; Nevin, 2007;

Reyes, 2007; Stetesky & Lynch, 2004), poverty rate (Reyes, 2007), resource deprivation

(Stretesky & Lynch, 2004), abortion (Reyes, 2007), race (Stretesky & Lynch, 2001),

population structure and density (Masters et al., 1994; Stretesky & Lynch, 2004), level of

education (Masters et al., 1994; Strestesky & Lynch, 2001), percent of the population that

is young (Reyes, 2007; Stretesky & Lynch, 2004), and percent divorced (Stretesky &

Lynch, 2004). Moreover, these associations have been observed between lead and

property crime and as well as lead and violent crime (Masters et al., 1994; Stretesky &

Lynch, 2004). Consistent with these observations, Narag and colleagues (2009)

summarize, “high levels of lead exposure add an additional structural ill and obstacle

(emphasis mine) for residents of disadvantaged communities, perhaps adding to higher

levels of crime in these areas,” (pg. 965).

Empirical assessment is needed to assess whether tests of social disorganization

theory may be omitting an important explanatory variable, implying results from

traditional models may be biased. Given that social disorganization theory is grounded in

control theory (Kornhauser, 1979), it is possible that the factors that suppress crime may

be more effective in communities without lead than communities that contain lead. Since

increased presence of lead in individuals, countries, counties, and cities has been

associated with increased criminal behavior and crime rates, it is reasonable to infer that

increased presence of lead in communities may be associated with increased rates of

crime. That is, neurotoxicity, rather than or in addition to disorganization, may emerge is

a relevant factor in explaining and theorizing about the distribution of crime across

communities.

Policy. Not only is mainstream criminology theory limited with regard to lead’s

impact on community crime rates, but community crime control policies may also be

limited by this void (Lersch & Hart, 2011). There are at least two ways in which policy

may by adversely impacted by this omission in the literature. First, crime control policies

that are informed by the tenets of social disorganization theory are costly and expansive

initiatives. These efforts to promote community organizing represent multi-billion dollar

investments from taxpayers. Should lead demonstrate to contribute to community level

crime rates, it would appear that these expensive policy initiatives may be undermined by

the presence of lead in those communities, if lead and crime are also related. This position

is underscored by the fact that empirical assessments of a few of the programs designed

to promote community organization (i.e., CAP) reveal these programs are not a panacea

for crime. It would appear that these initiatives may be limited in some way, and an

empirical assessment is needed to discern if this limitation is failure to address

community lead exposure.

Second, in the event that an empirical assessment of the lead crime relationship

reveals an association between lead and crime, new policy initiatives may be supported.

For example, inform initiatives that reduce the presence of lead may present as promising

new directions for reducing community crime rates. This suggestion would be consistent

with extant research that suggests reducing lead levels may be a cost-effective initiative in

reducing other social problems (Needleman 2004; Reyes, 2007). For instance, in 1991

the Public Health Service conducted a cost-benefit analysis on abatement of housing

structures built using lead-based paint (approximately 80% of structures built before

1950), across a thirty year period. These estimates suggested that the project would cost

approximately $33.7 billion. However, the return from raised IQ points and avoided

health care costs alone was estimated to be $61.7 billion. Needleman (2004) called the

Public Health Service’s estimate conservative because it had not factored in avoided

delinquency (or cardiovascular diseases), which Needleman provides as effects of lead

exposure. A recent study by Gould (2009) suggests that a reduction in the average

preschool blood lead level of 1 μg/dL would save $1.8 billion associated with the direct

costs of crime alone. Considering crime among other costs, Gould (2009) went on to

estimate that each dollar invested in lead paint hazard control would correspond with a

return ranging from $17 to $221, or a total savings of at least $180 billion. The outcome

of a community-level assessment of lead and crime rates may help to inform such cost

benefit analyses.

To ensure that theoretical explanations of community crime rates are not

missspecified, and to ensure that costly policy ventures to promote community organizing

to deter crime are not undermined by the presence of lead, a study is needed to

empirically assess the relationship between lead and crime at the community level.

Chapter Four:

Bridging the Gap: Lead, Crime, and Chicago

The previous chapters have provided a review of the literature on distributions of

lead and crime, justifying the need for an empirical assessment of lead and crime hotspots

at the community level of analysis. To conduct such an assessment, this study analyzes

lead and crime across communities in Chicago, Illinois. The purpose of chapter four is to

justify the selection of the city of Chicago. Chapter four opens with a short background

on the city of Chicago. Next, data and findings from extant publications are provided to

demonstrate disproportionately high levels of both lead and index crime in Chicago. This

chapter concludes with a discussion of community-level research in Chicago, noting prior

studies and data that suggest a lead and crime hotspot overlap in Chicago communities.

About Chicago

The city of Chicago was incorporated in 1837 (City of Chicago, 2012). Chicago

is located in Northeastern Illinois, within Cook County, along the shore of Lake

Michigan. Because of the city’s location, Chicago was regarded as a hub of trade and

commerce through America’s westward expansion in the 1800s (City of Chicago, 2012).

Since 1790, Chicago has ranked within the top 10 most populous cities in the United

States (U.S. Census, 2010). According to Shaw and McKay (1942) in 1840, Chicago had a

population of 4,470. By 1880, Chicago’s population boomed to 500,000 persons, and by

1910, that figure quadrupled. In 1940, Chicago had a population of over 3 million, 91.7%

of which identified as white, and 8.2% of which identified as black (Wirth & Furez, 1938).

Along with a growing population, Chicago also expanded geographically over time.

According to Shaw and McKay (1942), in 1830, the area designated as Chicago subsumed

approximately one-half of a square mile of land. By 1889, Chicago would come to occupy

over 40 square miles, and by 1940, Chicago consisted of approximately 134 square miles

of territory (Wirth & Furez, 1938).

Today, according to the U.S. Census, the city of Chicago is approximately 227.63

square miles. In 2010, Chicago has a population of 2,695,598, concentrating

approximately 21% of the entire population of Illinois (U.S. Census, 2010). Today,

Chicago’s population is 45% white, 32.9% black, and 28.9% Hispanic or Latino/a.

Chicago currently houses eleven Fortune 500 companies, and it has been estimated that

approximately 86 million people visit Chicago each year (City of Chicago, 2012).

Chicago maintains its reputation as a trade and commerce epicenter, with the O’Hare and

Midway International airports constituting the busiest aviation centers in the U.S, while

50% of all U.S. rail freight continues to pass through the city (City of Chicago, 2012). In

2010, Chicago was deemed by the U.S. census the third most populous city in the United

States (U.S. Census, 2010).

Lead in Chicago

The presence of lead in Chicago may be unsurprising when one considers the

level of pollution in Chicago. The Environmental Protection Agency’s Toxic Release

Inventory (TRI) maintains data on the release, transfer, and handling of over 600 toxic

chemicals, including, but not limited to, lead and lead compounds. Based on TRI data, in

2010, the Right-to-Know network (2012) ranked the state of Illinois in the top ten states

for releases of toxic waste. Also according to 2010 TRI data, Chicago housed over one

hundred facilities that reported the transfer or release of toxic chemicals into the air,

water, or land, with each facility generating an average of nearly thirty thousand pounds

of toxic waste (Right-to-Know Network, 2012). The American Lung Association’s State

of the Air report (2012) ranks the Chicago area as one of the nations’ most polluted

regions with regard to short-term particle pollution.

There is a documented presence of lead in Illinois, Cook County, and Chicago.

According to the Illinois Department of Public Health (2011), the state of Illinois ranks

especially high in the United States in terms of rates of lead-poisoned children (with lead

poisoned being defined as testing with a blood lead level at or above 10 μg/dL). In 2009,

the Illinois Department of Public Health (2011) provided case management services to

3,171 lead-poisoned children, and initiated 2,624 environmental investigations.

Furthermore, the department estimates that currently approximately 2 million houses in

Illinois contain leaded paint (Illinois Department of Public Health, 2011). In 1999, the

EPA ranked Cook County in the 95th percentile nationally for highest emissions of lead

compounds. TRI data reveals that in 2010, over one hundred thousand pounds of lead

and lead compounds were released in Cook County (Right-to-Know Network, 2012).

The city of Chicago demonstrates a unique struggle with lead. Since 2005, over two

hundred thousand pounds of lead and lead compounds have been generated in Chicago

(Right-to-Know Network, 2012). The Illinois Department of Public Health (2011)

reports that in 2009 nearly 28 thousand children tested positive for a blood lead level at or

above 5 μg/dL, strictly in Chicago. Most of the children tested (92%) were younger than

six years of age, and these data suggest that in 2009 nearly 11% of all

Chicago children age six and under have been identified as carrying a blood lead level of

5 μg/dL or higher. Figures 1 and 2 present data reported by the Illinois Department of

Public Health from 2008 (2011). Figure 1 compares the rates at which tested children

have been found to carry blood lead levels at or above 5 μg/dL. What these data reveal is

that approximately one in four children tested for lead in Chicago carry a blood lead level

at or above 5 μg/dL. Furthermore, a comparison of Chicago strictly to the rest of Cook

County and Illinois reveals that children tested for lead in Chicago carry elevated blood

lead levels at over twice the rate of the rest of the county and state, respectively. Figure 1

is congruent with Figure 2 which demonstrates that of 49,571 children with elevated

blood lead levels in Illinois, 61% were concentrated in Chicago.

Because of the strong, positive correlation that has been detected between blood

lead level and environmental lead presence (Hayes et al., 1994; Pirkle et al., 1994),

figures 1 and 2 also imply that disproportionate amounts of lead in Illinois concentrate in

Chicago’s environment. Hayes’ and colleagues (1994) analyze the relationship between

air and blood lead levels among children in Chicago explicitly. Their study tests the

relationship between air lead levels in Chicago and children’s blood lead levels as

Figure 1. Rate at Which Tested Children Found to Have Blood Lead Levels (BLL) at or

Above 5 μg/dL Within Jurisdiction, 2008*

SOURCE: Illinois Lead Program Surveillance Data, 2006-2009 and Illinois Center for Health Statistics, as

reported by the Illinois Department of Public Health (2011) in “Illinois Lead Program Surveillance Report –

2009”.

* The number of total children tested includes children who have been tested more than once. However, in

instances where children tested positive and were retested, only the highest venous or capillary result was

recorded (Illinois Department of Public Health, 2011). For this reason, these figures are likely to provide a

more conservative estimate.

Figure 2. Percentage of All Illinois Children Tested and Found to Have Blood Lead

Levels at or Above 5 μg/dL by Jurisdiction, 2008, (n= 49,571)

SOURCE: Illinois Lead Program Surveillance Data, 2006-2009 and Illinois Center for Health Statistics, as

reported by the Illinois Department of Public Health (2011) in “Illinois Lead Program Surveillance Report –

2009”.

captured by the City of Chicago Department of Health Division of Laboratories from

1968 through 1988. Hayes et al. (1998) observe that the median blood lead level declined

from 30 μg/dL in 1968 to 12 μg/dL in 1988, corresponding significantly to air lead levels

(r=.8, p<.0005). It may be reasonably inferred, then, that lead is present in both

Chicago’s physical environment and in the bodies of Chicago residents.

Crime in Chicago

Index crime also has a documented presence in Chicago’s environment and

surrounding areas. In the state of Illinois, 2010 UCR data indicate that index crimes

occurred across the state at a rate of approximately 31.16 crimes per 1,000 residents

(totaling 399,824 crimes). This can be broken down further into a violent crime rate of

4.35 crimes per 1,000 residents, and a property crime rate of 26.81 crimes per 1,000

residents. Index crime in Illinois, like index crime nation-wide, has steadily been on the

decline since the early 1990s. According to the State of Oregon Criminal Justice

Commission (2012), the state of Illinois ranked # 17 in the nation for violent index crime

rate in 2010. The same report (2012) found Illinois violent crime rate to be higher than

the three lowest ranked states for violent crime (New Hampshire, Vermont, and Maine)

combined. The Circuit Court of Cook County of the State of Illinois represents the

largest judicial circuit in Illinois, and is also one of the largest court systems in the world

(State of Illinois Circuit Court of Cook County, 2012). By some accounts, the Cook

County Criminal Courthouse is the largest and busiest felony courthouse in the United

States (Bogira, 2005). The Circuit Court of Cook County employs more than 400 judges,

and handles over 1.2 million cases each year (State of Illinois Circuit Court of Cook

County, 2012).

According to the Chicago Police Department’s annual report (2010), there were

152,031 index crimes in the city of Chicago in 2010. Of these crimes, approximately

20% were violent crimes and 80% property crimes. Considering Chicago’s population,

the 2010 Index Crime rate in Chicago was approximately 56.3 crimes per 1,000 persons.

By comparison, the national index crime rate in 2010 was approximately 33.5 crimes per

1,000 persons. Figure 3 compares rates of index offending in Chicago, Cook County,

Illinois, and the United States from 2005 to 2009. What these data suggest is that index

offending occurs at disproportionately higher rates in Chicago when compared to county,

state, and national rates. The Illinois State Police (2009) also report that offenses

reported by the Chicago Police Department account for 36.4% of the total index crime

offenses in the state of Illinois.

Similar to the data that was reported for lead, Figure 4 charts index offense rates

across Illinois. Calculations were based on data provided by the Illinois State Police in

the Crime in Illinois’ 2009 Annual Uniform Crime Report. These data reveal that the

index crime rate in Chicago alone is substantially higher than index crime rates across the

rest of Cook County and the state of Illinois. Moreover, this is a trend that maintains both

when Chicago is included in the crime rate configuration for Cook County and

Illinois, and when Chicago’s figures are omitted. When compared to data in Figure 2, it

would appear that trends in lead across Illinois are very similar to trends in index crime

across Illinois.

Figure 3. Trends in Annual Index Crime Rate (per 1000 persons) By Jurisdiction,

20052009

SOURCE: Illinois State Police (2009). The Executive Summary of the Crime in Illinois 2009, Annual

Uniform Crime Report.

Figure 4: Index Crime Rate across Illinois, Cook County, and Chicago, 2009

SOURCE: Illinois State Police (2009). The Executive Summary of the Crime in Illinois 2009, Annual

Uniform Crime Report.

In sum, it appears that just as rates of elevated blood lead are disproportionately

high in Chicago, so are the crime rates. For these reasons the city of Chicago is

especially well suited for an assessment of the lead and crime connection. However,

because this study seeks to test this relationship at the community level of analysis,

justification for the selection of Chicago would be further enhanced if it could be

demonstrated that hotspots of lead and crime exist at the community level. To

demonstrate that extant data and literature studying Chicago’s communities illustrate the

presence of such hotspots, a discussion of lead and crime across Chicago’s communities

is provided.

Lead, Crime, and Chicago’s Communities

Chicago has interested researchers not only in its entirety as a city, but also as an

amalgam of communities. Chicago’s communities have posed such long-standing

intrigue among researchers that Robert Sampson has referred to the history of community

research in Chicago as “unparalleled” (2012, p.76). Indeed, the communities of the city

of Chicago have captivated sociologists and criminologists alike since the work of the

Chicago School scholars (Kotlowitz, 1992; Park & Burgess, 1925; Shaw & McKay,

1942; Sampson, 2012; Sampson, 2002; Sampson, Morenoff, & Raudenbush, 2005;

Sampson et al., 1997; Venkatesh, 2000; 2008; Visher & Farrell, 2005; Zorbaugh, 1929).

One of the most renowned research endeavors to take up the study of Chicago’s

neighborhoods is undoubtedly the Project on Human Development in Chicago

Neighborhoods (PHDCN). The PHDCN was established in 1994 by Robert Sampson and

colleagues to examine development and context across neighborhoods in the city of

Chicago (Sampson, 2012; Sampson et al., 1997; Earls & Visher, 1997). The PHDCN

broke up the Chicago area into 343 neighborhood clusters (NCs). NCs were

neighborhoods larger than a single census tract, but smaller than the Chicago Community

Areas. NCs were designed to maintain relative internal socio-demographic homogeneity.

Studies from the PHDCN and otherwise have detected differentials in the

distribution of crime across communities in Chicago, suggesting community-level crime

hotspots in Chicago (Block and Block, 1993; Lyons, 2006; Sampson, 2012; Sampson &

Raudenbush, 1999; Sampson et al., 1997; Visher & Farrell, 2005). For example, Block

and Block’s (1993) study of street-gang motivated crimes in Chicago found that the two

communities with the highest mean rates of street crime (East Garfield Park and

Humboldt Park), had mean rates of street crime that were seventy-six times higher than

the rate of street gang crime in the two safest communities (Mount Greenwood and

Edison Park). The same study explores rates of street-gang motivated homicide, and

finds concentrations of street-gang motivated homicide in two corridors of Chicago

community areas along the Northwest and Southwest sides. Analyses demonstrated that

across Chicago, 22% of communities had no street-gang related homicides, while one

community (Lower West Side) averaged over six per year (Block & Block, 1993).

Sampson and Raudenbush (1999) found that Chicago neighborhoods with higher levels

of concentrated disadvantage also experienced significantly higher rates of homicide,

robbery, and burglary. Visher and Farrell (2005) found that over half of former male

prisoners returning to Chicago were released to one of seven community areas (Austin,

Humboldt Park, North Lawndale, West Englewood, East Garfield Park, Roseland, and

Auburn Gresham). Of all released prisoners, 52.4% returned to prison (Visher & Farrell,

2005).

Studies also illustrate unequal distributions of lead across Chicago communities.

Oyana and Margai (2010) studied spatial and temporal patterns in pediatric lead exposure

across Chicago census tracts in 1997, 2000, and 2003. The distribution of blood lead

level was skewed, with some census tracts emerging with no cases of elevated blood lead

levels among children, and one tract reporting as many as 106 cases. These researchers

found that Chicago’s Westside emerges as the region with the highest risk of elevated

child blood lead levels, followed by Chicago’s Southside and Far Southside regions

(Oyana & Margai, 2010). Oyana and Margai observe that, “the high-risk areas are fairly

consistent over the seven-year period, yielding persistent and visually distinct hot spots

for pediatric lead poisoning,” (2010, p. 56). Oyana and Margai’s (2010) findings are

consistent with data released from the Chicago Department of Public Health. The

Chicago Department of Public Health (2008), identifies seven Chicago communities

where 4.5% or more tested children had blood lead levels >10 μg/dL: Austin, West

Garfield Park, Fuller Park, Avalon Park, West Englewood, Englewood, and Greater Grand

Crossing. By comparison, in thirty Chicago communities only 1.5% or fewer tested

children were found to have blood lead levels > 10 μg/dL.

When taken together, findings from extant data and community-level crime and

lead studies in Chicago allude to hotspot overlay. For example, Block and Block’s (1993)

findings indicate Humboldt Park and East Garfield Park as among the most dangerous in

terms of gang-crimes. Data from the Chicago Department of Public Health (CDPH)

(1998) report rates of children tested and found to carry blood lead levels >10 μg/dL.

CPDH data indicate that children tested in the Humboldt Park and East Garfield Park

Communities had elevated blood lead at rates of 23% and 28%, respectively. By

comparison, Block and Block (1993) name Edison Park and Mount Greenwood as two of

the safest communities in terms of gang-crime. The CDPH (1998) indicates that children

tested in both of these communities report elevated blood lead levels at rates of 1%.

Likewise, the 1998 Chicago Police Department’s Annual Report reveals that 60

homicides occurred in Austin in 1998, indicating that more homicides occurred in 1998 in

Austin than in any other Chicago community. The CDPH (1998) indicated that in 1998,

32% of children tested in Austin carried blood lead levels > 10 μg/dL.

Alternatively stated, in 1998, over 1,000 children in Austin were found to have elevated

blood lead levels.

In sum, both lead and index crime have a documented presence in the city of

Chicago. This chapter presented data from the UCR, Illinois State Police, and Illinois

Department of Public Health to demonstrate that both lead and crime are

disproportionately high in Chicago. Moreover, graphical depictions of the data illustrate

the substantial impact that the Chicago area has on state and county rates of lead exposure

and index crime. Because this study is concerned with community level hotspots of lead

and crime, a brief discussion of Chicago’s numerous community-level crime studies is

presented. It is noted that Chicago’s communities are among the most widely studied in

criminology and sociology, giving Chicago a uniquely extensive history of community-

level studies. Prior work demonstrates that the skewed distribution of crime produces

hotspots in Chicago communities. Likewise, extant studies and data illustrate a non-

normal distribution of lead across the city of Chicago. Finally, merging these findings

intimates overlap between lead and crime in Chicago communities, making Chicago a

truly ideal setting for assessing the relationship between lead and crime

hotspots.

Chapter Five:

Data, Methods, & Measures

A review of the literature finds that prior studies suggest a causal relationship

between lead exposure and crime. While this association appears tenable across varying

levels of analysis (individual, county, city, nation, time), and linkages between lead and

homicide hotspots have been demonstrated (Lynch et al, 2001), additional assessment at

the community level of analysis is needed (Lersch & Hart, 2011; Narag et al., 2009).

Because findings from prior studies imply overlap between hotspots of lead and hotspots

of crime across communities in Chicago, Illinois (Block & Block, 1993; The Chicago

Department of Public Health, 1998; Lynch et al., 2001; Oyana & Margai, 2010), the city

of Chicago is selected for this assessment. Chapter five details the methodology

employed for this study. To begin, the relationships to be tested and hypotheses

statements are listed. Details on the unit of analysis, data, and sample follow. Finally, the

conceptualization and operationalization of independent, dependent, and control variables

is provided.

Research Questions & Hypotheses

This study has very specific research aims. First, this study sets out to understand

the distribution of lead and crime across communities in the city of Chicago. Next, this

study seeks to examine the relationship between lead and crime across communities in

the city of Chicago. Finally, this study explores the distribution of lead as an

environmental justice concern for Chicago residents. As such, this study’s research

questions are broken down into thirteen testable hypotheses.

The Distribution of lead and crime across communities. This study seeks to

address a very straightforward research question: Is there a relationship between hotspots

of lead and hotspots of crime? With respect to this research question, three hypotheses

are explicated:

H1: Lead is distributed non-randomly across Chicago communities, creating lead

hotspots.

H2: Crime is distributed non-randomly across Chicago communities, creating

crime hotspots.

H3: Communities which are lead hotspots will also emerge as crime hotspots.

Prior literature suggests that crime and lead have heavily skewed distributions

across place (Hanchette, 2008; Lynch et al., 2001; Lynch, 2004; Masters et al., 1994;

Mielke et al., 1999; Stretesky & Lynch, 2001; 2004). The criminological literature offers

no clear definition of a hotspot. On this point, Chainey, Reid and Stuart (2003) note:

“Several techniques and algorithms are used in practice for the generation of continuous

surface hotspot maps, all of which have different merits. These mainly relate to their ease

of use, applications to different types of events, visual results, and interpretation. Few of

these methods help to distinguish a consistent defining threshold that helps the analyst

decide when a cluster of crimes can be defined as a hotspot,” (pg. 23).

For example, Eck (2005) described a crime hotspot as, “an area that has a greater than

average number of criminal or disorder events,” (p.2). While “greater than average” may

not be the only way to describe a hotspot of crime or lead, this is one approach for

defining hotspots. As such, this study conceptualizes a hotspot as a community area with

a significantly greater than average presence of lead or crime, and assumes a background

level of lead and crime throughout all of Chicago.

Tests of hypotheses 1 and 2 set out to examine if lead hotspots and crime hotspots

emerge across the city of Chicago. To assess if lead and crime concentrate in the same

areas of Chicago, tests of hypothesis 3 specifically sets out to examine overlay between

hotspots of crime and hotspots of lead, should they emerge following tests of hypothesis

one and two. Hypotheses 1, 2, and 3 are based on prior findings, and are theoretically

grounded in the concept that political economic arrangements force lead and crime

together in time and space (Hanchette, 2008; Lynch, 2004; Lynch & Stretesky, 2003;

Stretesky & Lynch, 2004).

The Relationship between lead and crime in communities. This study also

addresses related question if a relationship between lead and crime emerges, such as what

is the nature and extent of the relationship between lead and crime at the community

level? Specifically, this study explores the conditions under which a relationship between

lead and crime remains tenable. Based on prior research, it is predicted that: H4:

Community lead presence is consistently associated with increases in community crime,

controlling for traditional criminological predictors of crime.

H5: Community lead presence will be a more consistent predictor of violent crime

than property crime.

H6: Community lead presence will be a more consistent predictor of juvenile

crime than adult crime.

H7: Community lead presence will be a more consistent predictor of juvenile

property crime than adult property crime.

H8: Community lead presence will be a more consistent predictor of juvenile

violent crime than adult violent crime.

Tests of hypothesis 4 concern the relationship between community levels of lead

and crime with respect to the predictors of crime documented in social disorganization

research. Specifically, hypothesis 4 examines whether in the presence of social

disorganization variables, lead accounts for increases in crime. Contingent on support for

Hypothesis 4, tests of Hypotheses 5 through 8 examine lead’s impact on crime across

varying traits of offenders and types of offenses. Prior literature suggests that lead

promotes violence and aggression particularly (Burns et al., 1999; Needleman et al.,

1996), and previous studies have found lead to be a stronger predictor of violent crime

than property crime (Pihl & Ervin, 1990). Hypothesis 5 is informed by these findings.

Likewise, extant studies have argued that children are particularly vulnerable to

environmental contaminants (Landrigan & Carlson, 1995; Moore, 2003), including lead.

For this reason, Hypothesis 6 predicts that lead will be a more consistent predictor of

juvenile offending than adult offending. Hypotheses 7 and 8 build on Hypotheses 5 and

6, exploring the conditions under which lead may have the most persistent explanatory

power. It is hypothesized that lead will be an especially consistent predictor of variation

in violent juvenile offending.

Environmental justice issues. Finally, this study takes into consideration

environmental justice literature which documents a disproportionately high presence of

environmental hazards in disenfranchised communities (Lerner, 2005; Brulle & Pellow,

2006; Bullard 1994; 1996; Checker, 2007; Hipp & Lakon, 2010; Krieg, 2005; Opp, 2012;

Pastor, Morello-Frosch, & Sadd, 2006; Stretesky & Lynch, 2002). As such, an

exploratory assessment of lead and environmental justice in Chicago hypothesizes: H9:

Percentage of black residents in a community will correspond positively and significantly

with level of lead in a community.

H10: Percentage of Hispanic residents in a community will correspond positively

and significantly with level of lead in a community.

H11: Disadvantage will correspond positively and significantly with level of lead

in a community.

H12: Disadvantage and percentage of black residents will interact and correspond

positively and significantly with level of lead in a community.

H13: Disadvantage and percentage of Hispanic residents will interact and

correspond positively and significantly with level of lead in a community.

These last five hypotheses are informed by environmental justice literature and

race and political economy theory (Stretesky & Lynch, 2002; 2004). Chicago’s diverse

population and ethnic enclaves provide an opportunity to assess lead and environmental

justice concerns in the city (Sampson, 2012). With respect to race and ethnicity, because

prior work finds levels of lead higher in minority communities than non-minority

communities (Brulle & Pellow, 2006; Bullard 1994; 1996; Checker, 2007; Hipp &

Lakon, 2010; Opp, 2012; Pastor, Morello-Frosch, & Sadd, 2006; Ringquist, 2005),

Hypotheses 6 and 7 anticipate an overrepresentation of lead in predominately black and

Hispanic communities. Hypothesis 8 anticipates an over-representation of lead in

economically disadvantaged areas (see also Krieg, 2005). Hypotheses 9 and 10 examine

the presence of lead when both community minority status and economic status interact

and are simultaneously considered. Hypotheses 11, 12 and 13 present variations of

hypotheses 9 and 10 which may also be evident.

Unit of Analysis

Because this study contributes to the literature by studying lead and crime at the

community level, the necessary unit of analysis is the community. In this study,

community is conceptualized as Chicago Community Area (CCA). Currently, Chicago is

divided into 77 distinct CCAs. Each community is assigned a name and its own unique

number ranging from 1 through 77. The CCAs are recognized by politicians, community

leaders, and government entities, and as such, are meaningful to residents and to the city

of Chicago (Sampson, 2012). Moreover, the CCAs have historical relevance, and have

been recognized in Chicago for decades.

In the 1920’s, to facilitate the study of communities across Chicago, the

University of Chicago’s Social Science Research Committee established 75 community

areas across the city (Hudson, 2006). These 75 community areas were non-overlapping

and contiguous. The committee used five criteria to distinguish the communities from

one another: (1) settlement, growth, and history of the area, (2) local identification with

the area, (3) local trade area, (4) distribution of membership of local institutions, and (5)

natural and artificial barriers (Hudson, 2006, pg. 23). In subsequent years, when the U.S.

Census Bureau began defining the boundaries for census tracts in Chicago, boundaries

were drawn so that they overlapped with the predefined boundaries of the 75 community

areas. That is, census tracts aggregate up to corresponding community areas (Hudson,

2006; Visher & Farrell, 2005), and Chicago’s community areas exist as aggregates of

census tracts (Block & Block, 1993; Zhang, 2012).

In 1960, O’Hare was annexed as the 76th community area, and in 1980,

Edgewater became recognized as the 77th community area (Zhang, 2012). The

boundaries of the tracts have undergone only minor changes over time, and construe what

are contemporarily recognized as the 77 Chicago Community Areas (CCA). Today, the

city of Chicago officially recognizes the CCAs for city planning and service delivery

purposes (Hudson, 2006; Visher & Farrell, 2005), and the CCAs are also readily known

by residents, the media, and administrative agencies (Sampson, 2012). Researchers have

used the CCAs as the unit of analysis in their studies of street-gang crime (Block &

Block, 1993), homicides (Sampson, 2012), prisoner re-entry (Visher & Farrell, 2005), and

hate crime (Lyons, 2006).

The use of the CCAs to measure community has many benefits. For example,

CCA’s are recognized by the local government and policy makers in terms of service

delivery and regional planning, making results from this study more amenable to policy

implications. Residents also identify and recognize the CCAs. Additionally, because the

boundaries of the CCAs have remained relatively stable over several years, the CCAs

bear historical relevance, and contemporary studies of the CCAs can be meaningfully

compared to prior works given the long term stability of these community areas.

Relatedly, because other researchers have utilized the CCAs, the use of the CCAs as the

unit of analysis in this study allows for findings to be readily interpreted in the context of

extant work. A map of the Chicago Community Areas is provided in Appendix A.

The Ecological fallacy. In 1950, W.S. Robinson published “Ecological

Correlations and the Behavior of Individuals,” in the American Sociological Review. In

this seminal piece, Robinson articulates the distinction between individual correlations

(descriptive properties of individuals) and ecological correlations (descriptive properties

of groups, often rates, means, or percentages). Robinson explains that the purpose of the

essay was to express the mathematical difference between individual correlations and

ecological correlations, and in the process, illustrate the consequences of using ecological

correlations in place of individual correlations. To make his point, Robinson uses race,

ethnicity, and illiteracy data to argue how the strength and/or direction of a relationship

between two variables is dependent upon the level at which (individual, ecological) the

data was collected. The methodological pitfall Robinson calls attention to -- using

aggregate data to draw conclusions about individuals -- has come to be known as “the

ecological fallacy.” Selvin (1958) coined the phrase and described the ecological fallacy

as occurring when, “relationships between characteristics of individuals are wrongly

inferred from data about groups,” (p. 613).

Robinson’s essay was critical of aggregate data, and the ecological fallacy had far

reaching consequences for social science research. One of those consequences is the

development of the tradition of methodological individualism in the social sciences

(Mills, 1959), including criminology (Michalowski, 2009; Sampson, 2012; Subramanian

et al., 2009). Robinson’s critique of aggregate data is limited, however, by his inherent

assumption that social science researchers are only concerned with individual level

processes (Sampson, 2012; Subramanian et al., 2009). When social science researchers

seek to understand community level processes, it is most appropriate to analyze

ecological data. On this point, Sampson (2012) explains:

“Robinson’s mistake, and that of many readers, was to assume that ecological

researchers only cared about individual-level inferences. Rather than arguing

against ecological or neighborhood-level research, the right message was to make

clear distinctions among units of analysis and to appropriately frame analytical

questions… It follows that if the main goal is to explain rates of variation across

neighborhoods rather than individual differences, Robinson’s critique does not

hold…” (p. 39).

Following Sampson’s argument, the current study is not making an ecological

fallacy since its emphasis is on the community level of analysis, and not the individual.

The research questions addressed in this study concern variation in community crime and

lead rates: not variation across the health and behavior of individuals. All data used in

this study are qualities or traits of communities represented as rates and percentages.

Data & Sample

Data for this study were obtained from multiple secondary sources, each source

providing relevant data on the total sample of 77 CCAs (n = 77). These sources include

the City of Chicago (2012), the U.S. Census Bureau (2000; 2010), the American

Community Survey (2010), the Chicago Department of Public Health (2008), and The

Chicago Police Department (2008). The City of Chicago19 provides shapefiles for the

city of Chicago, including files identifying: the boundaries of the CCAs, Chicago census

tracts, and each Chicago census tract’s corresponding CCA. The city of Chicago

provides this data for 2000 census tracts and 2010 census tracts. Given the overlap

between census tracts and CCAs in the city of Chicago, it is possible to aggregate census

data to CCA (Block & Block, 1993; Lyons, 2006; Zhang, 2012). In 2010, the city of

Chicago identified 873 census tracts within Chicago that comprised CCAs, and in 2010

identified 801 census tracts within Chicago that comprised CCAs. Socio-demographic

information for this sample was obtained for each of these tracts from the 2000 U.S.

Census, 2010 U.S. Census, and 2010 American Community Survey. It follows then, that

data from each tract was aggregated up to the appropriate CCA20.

are sufficient to explain collective outcomes,” (p. 40). Both the individual and ecological fallacy are

explored empirically in a contemporary re-analysis of Robinson’s data (see Subramanian et al., 2009).

19 See: https://data.cityofchicago.org/

20 While some census tract boundaries changed (merged, split) between 2000 and 2010, the boundaries for

the Chicago Communtiy Area remained unchanged. The city of Chicago provides 2010 census tract data,

Data for this study was also obtained from the Chicago Police Department. The

Chicago Police Department produces an annual report on index crime, case clearances,

domestic violence, traffic safety, calls for service, and related law enforcement

information. These annual reports are available for download on the Chicago Police

Department website for years 1967 through 2010. Beginning in 1998, the Chicago Police

Department began providing data on index crime as it occurred in the CCAs. Crime data

for this study was obtained using figures publicly available from the 2000 through 2010

Annual report. To obtain information about age of offenders (juvenile vs. adult index

offenses), a Statistical & Crime Data Request was filed with the Research and

Development Division of the Chicago Police Department. Data on arrests for index

crimes by CCA and age of offender (adult or juvenile) was provided for the year 2008,

and translated into age-based rates of offending for each community.

Finally, data from the Chicago Department of Public Health was collected for this

study. Per the Illinois Lead Poisoning Prevention Act, Illinois law mandates every

practicing physician in the state to screen children age 6 months through 6 years old for

lead, if the child resides in a zip code that the Illinois Department of Public Health deems

“high risk,” (Illinois Lead Poisoning Prevention Act, 410 ILCS 45). Currently, the

department deems all zip codes in Chicago to be high risk (Lead Safe Illinois, 2006).

According to the Illinois Department of Public Health, a blood lead level (BLL) less than

10 μg/dL, is considered to be permissible for children age 16 and younger (regardless of

and designates tracts to CCA. Census data was obtained by requesting from American Fact Finder data for

all census tracts (or parts) fully within/partially within the city of Chicago, Illinois. As a measure of quality

assurance, after tracts were aggregated, population totals for each CCA were matched to CCA population

the fact that findings from the medical literature recommend much lower thresholds).

However, if a child is screened and found to carry a BLL greater than or equal to 10

μg/dL, that child is considered to have an elevated BLL. Physicians are required to report

results of lead screenings (elevated and otherwise) to the Illinois Department of Public

Health (which oversees the Illinois Lead Program) (Illinois Lead Poisoning Prevention

Act, 410 ILCS 45). In turn, the Chicago Department of Public Health uses laboratory

results reported to the IPHD, to construct rates of children tested and found to have

elevated BLLs by Chicago Community Area. This information is publicly available and

was accessed online via the City of Chicago Data Portal.

Independent Variable

The independent variable of interest in this study is lead. To measure lead, this

study relies upon the Chicago Department of Public Health’s data on children age six and

under screened for lead (Chicago Department of Public Health, 2012), as an indicator for

the presence of lead in a community. Specifically, this study uses the variable “Percent

Elevated,” for the years 1999 through 2010. According the Chicago Department of

Public Health, in this dataset, percent elevated is the proportion of children tested that

were found to have elevated BLLs. This figure is determined by dividing the number of

children found to have elevated BLLs by the total number of children tested and

multiplying by 100, and produces the community measure of lead in this study. Per the

Department’s standards, “elevated” is defined as a venous blood sample found to be at a

BLL at or above 10 μg/dL. The Chicago Department of Public Health reports this

totals reported by the city of Chicago and the Chicago Police Department for 2000 and 2010, respectively.

Results indicate tracts were aggregated accurately.

proportion for each of the Chicago Community Areas. Increases in proportion of children

under age six tested with elevated BLL will be interpreted as increases in community

presence of lead.

The use of proportion of children with elevated BLL is an appropriate indicator

for community presence of lead because the link between body lead burden and presence

of lead in the environment is robust and well documented (Aschengrau et al., 1994;

Brunkenreff, 1984; Hayes et al., 1994; Laidlaw et al., 2012; Lanphear et al., 1996;

Lanphear et al., 1998; Pirkle et al, 1994; Reyes, 2007; Schwartz & Pitcher, 1989;

Stretesky & Lynch, 2001; Zahran et al., 2011). Because the presence of blood lead has

been found to be positively, significantly correlated with presence of lead in the

environment (e.g., air, soil), it acts as a strong indicator of community lead presence.

These associations in prior literature establish that it is reasonable to infer that if the

residents of a community are carrying high levels of lead in their bloodstreams it is

related to a corresponding high level of lead present in their community. Conversely, it is

reasonable to infer that if a community has relatively few residents with lead in their

blood, it is reasonable to infer that the community has less of a presence of lead. For this

reason, presence of lead is measured using this indicator across all hypotheses.

Dependent Variable

The dependent variable in this study is crime. In this study crime is

conceptualized as index crime. Index crimes refer to the eight offenses labeled Part I-

Index crimes by the FBI, and include: murder, rape, aggravated assault, robbery,

burglary, larceny-theft, motor vehicle theft, and arson. The Chicago Police Department

documents each of these acts as index crimes, with the exception of rape, as the Chicago

Police Department documents instead acts of sexual assault. Sexual assault is described

in Chicago Police Department Annual Reports as “broader than the traditional definition

of ‘rape,’” and includes, “any sexual assault- completed or attempted, aggravated, or

nonaggravated- committed against any victim, female or male,” (Chicago Police

Department, 2008, pg. 29). The number of reported incidents of index crimes occurring

in each CCA is provided in the Chicago Police Department’s Annual Report, with counts

for each offense listed separately (e.g, counts are provided for each type of index

offense). This study uses crime data from 2000 - 2010 to test the relation between lead

and crime. To test H4, crime is measured using each CCA’s total number of index crime,

as reported in Chicago Police Department Annual Reports. Next, the count of index

crime is transformed into the rate of index crime. Rate is determined by dividing the

number of CCA index crimes by the corresponding CCA total population.

To test H5, it is necessary to distinguish types of crime for the dependent variable.

Testing of H5 will measure violent crime rate by using the total count of murders, sexual

assaults, aggravated assaults, and robberies as reported in the Chicago Police

Department’s Annual Reports from 2000 to 2010. Testing of H5 will measure property

crime rate by using the total count of burglaries, thefts, motor vehicle thefts, and arsons

as reported in the Chicago Police Department’s Annual Reports from 2000 to 2010.

noted in Chapter 6.

To test H6, H7, and H8, it is necessary to distinguish the age of the offender

(juvenile vs. adult) as well as type of offense (violent vs. property). This data was

provided by the Chicago Police Department in response to a Statistical & Crime data

request for a breakdown of index crimes committed by adults and index crimes

committed by juveniles by CCA. In tests of H6, juvenile crime is defined as an arrest of

a person age 17 or under for an index crime. In tests of H6, adult crime is defined as an

arrest of a person age 18 or over for an index crime. In tests of H7, juvenile property

crime is captured using the total count of arrests of a person age 17 or under for

burglaries, thefts, motor vehicle thefts, or arsons. In tests of H7, adult property crime is

captured using the total count of arrests of a person age 18 or over for burglaries, thefts,

motor vehicle thefts, or arsons. In tests of H8, juvenile violent crime is measured by

taking the total count of arrests of a person age 17 or under for murders, sexual assaults,

aggravated assaults, or robberies. In tests of H8, adult violent crime is measured by

taking the total count of arrests of a person age 18 or over for murders, sexual assaults,

aggravated assaults, or robberies . Data for tests of H6, H7, and H8 rely on index crime

arrests in 2008. Rate is determined by dividing juvenile arrests by the total population

age 17 and under, and dividing adult arrests by the total population 18 and over. It was

necessary to use incident arrest rate as opposed to incident rate, because demographics of

offender (e.g., age), are not always known or accurately reported for incident rates.

Naturally, this limits the ability of results from tests H6, H7, and H8 to be compared to

findings from H1 – H5. This will be reiterated throughout the analysis chapter, and in the

discussion chapter.

Control Variables

Social disorganization theory and communities and crime research informed the

decision to include a number of control variables in the analysis. As such, this study uses

data from the 2000 U.S. Census, 2010 U.S. Census, and 2010 American Community

Survey to include measures for socio-economic status, racial/ethnic heterogeneity, and

residential mobility as informed by the social disorganization literature (Kornhauser,

1978; Shaw & McKay, 1942). Additional control variables were added based on findings

from the communities and crime literature. These variables include: percent black,

percent Hispanic, percent divorced (Liska & Bellair, 1995), population size and density

(Morenoff, Sampson, & Rauenbush, 2001; Sampson & Raudenbush, 1999), and percent

population between 15-24 (Pridemore & Grubesic, 2011).

Disadvantage. Socioeconomic disadvantage is measured using a principle

component index of the following variables: median income, percentage of families

living below the poverty level, percentage of households receiving public assistance,

percent unemployed, and percentage of female headed households. Higher values on this

index will be interpreted as increased levels of economic disadvantage, and conversely,

lower values on this index will be interpreted as decreased levels of economic

disadvantage. The inclusion of an index for economic status is informed by social

disorganization theory, which predicts that increases in disadvantage weakens community

informal social control, and in turn, increases crime (Kornhauser, 1978; Sampson &

Groves, 1989; Shaw & McKay, 1942). The selection of the indicators for disadvantage

is informed by prior research (Sampson, Raudenbush, & Earls, 1997; Stretesky & Lynch,

2004).

Two disadvantage indices were created for 2000 and 2010, respectively. For the

2000 SES index, the first component extracted had an Eigenvalue of 4.22, which is well

above the recommended cutoff of 1 (Kaiser, 1961). Additional factors obtained

Eigenvalues less than 1 (ranging from .021 to .529), and thus, were not retained. Four of

five indicators loaded positively onto the retained factor at .4120, .4502, .4638, and

.4762. Median income loaded inversely at -.4309. The 2000 SES index obtained a

Cronbach’s α = .8029. For the 2010 SES index, the first component extracted had an

Eigenvalue of 3.95, which is well above the recommended cutoff of 1 (Kaiser, 1961).

Additional factors obtained Eigenvalues less than 1 (ranging from .0546 to .6034), and

thus, were not retained. Four of five indicators loaded positively onto the retained factor

at .4092, .4736, .4692, and .4199. Median income loaded inversely at -.4602. The 2010

SES index obtained a Cronbach’s α = .7539. Tests of hypotheses four through eight, as

well as hypotheses nine, eleven, and thirteen will employ this index to measure

disadvantage.

Racial & ethnic heterogeneity. Population heterogeneity was measured using

the Herfindahl index (Gibbs & Martin, 1962; Hipp, Tita, & Boggess, 2011; Vélez &

Richardson, 2012). The Herfindahl Index will be used to capture the degree to which

communities are racially or ethnically diverse with respect to four racial/ethnic groups:

white, black, Hispanic, and other populations. Following Hipp et al., (2011), this study

estimates the Herfindahl Index applying the following formula:

WHERE:

G = the proportion of a given racial/ethnic group present in a community

j = a given racial/ethnic group (white, black, Hispanic, other)

J = the total number of racial/ethnic groups

Higher values on the Herfindahl index are to be interpreted as higher levels of

heterogeneity, and likewise, lower values on this index will be interpreted as lower levels

of heterogeneity (with a value of zero indicating perfect homogeneity). The inclusion of

an index for racial/ethnic heterogeneity is informed by social disorganization theory,

which predicts that increases in racial/ethnic heterogeneity weaken community informal

social control, thereby increasing crime.

For this study, ethnicity was captured by classifying Chicago’s population into

those that identified on the Census as either Hispanic or Non-Hispanic. To capture race,

those who identified as Non-Hispanic were classified into one of four categories: White,

Black, Asian, and Other. Here, the Other category includes individuals who identify as

Native American Indian, Hawaiian, as a race outside of the options provided on the

Census, or individuals who identify as multi (2 or more) racial. Tests of Hypotheses 4

through 8 will use this measure to control for racial/ethnic heterogeneity.

Residential mobility. Residential mobility will be measured using a principle

component index of the following variables: percentage of families who have moved

from another residence in the past five years, percentage of vacant houses, and percentage

of houses occupied by a renter. Higher values on this index will be interpreted as higher

rates of residential mobility, and lower values on this index will be interpreted as lower

rates of residential mobility. The inclusion of an index for residential mobility is

informed by social disorganization theory, which predicts that increases in residential

mobility will decrease informal social control, and in turn, increase crime (Kornhauser,

1978; Sampson & Groves, 1989; Shaw & McKay, 1942).

Tests of Hypotheses 4 through 8 will use this measure to control for residential mobility.

Two residential mobility indices were created for 2000 and 2010, respectively. For the

2000 residential mobility index, the first component extracted had an Eigenvalue of 2.00,

which is well above the recommended cutoff of 1 (Kaiser, 1960). Additional factors

obtained Eigenvalues less than 1 (.7731 and .2179), and thus, were not retained.

Factor loadings ranged from .4957 to .6599, and a Cronbach’s α = .7449 was obtained.

For the 2010 residential mobility index, the first component extracted had an Eigenvalue

of 2.244, factor loadings ranging from .5373 to .6368, and a Cronbach’s α = .8281 was

obtained.

Other controls. Since the inception of social disorganization theory, a number of

studies examining communities and crime have uncovered additional variables that bear

significant associations with community crime rates (Hipp, 2007; Liska & Bellair, 1995;

Logan & Stults, 1999; Lyons, 2007; Morenoff, Sampson, & Rauenbush, 2001; Sampson

& Raudenbush, 1999; Velez & Richardson, 2012). These variables include percent black,

percent Hispanic, percent divorced, percent population between 15-24, and population

density. These variables were measured with responses taken from U.S. census data,

with the exception of population size and density, which will be calculated by dividing

the total population of a CCA (determined from the U.S. Census) by the land area in

square miles (provided by the city of Chicago shapefile for the CCAs). Because these

variables have proved to be significantly associated with crime in prior studies, they are

also included in the present analysis. In addition, the analytic approach of this study calls

for tests of spatial autocorrelation. Thus, spatial lag variables will be employed as

necessary. These variables are used as controls in tests of hypotheses four through eight.

Percent black and percent Hispanic are also used in the exploratory analysis of

environmental justice concerns.

Statistical & GIS Software Used

In this study, data were analyzed using Stata 11 (StataCorp., 2009), GeoDa

(Anselin, Syabri, & Kho, 2006), and ArcMap 10 (ESRI, 2011). Creation of principal

component indices, bivariate, and multivariate techniques was carried out with the use of

Stata 11. The generation of spatial weights, the spatial lag variable, and the Morans I

statistic was carried out using GeoDa. Finally, maps of the Chicago Community Areas

provided in Figures 5, 6, and 7 were generated using ArcMap10.

In sum, this study aims to explore the relationship between hotspots of lead and

hotspots of crime. In doing so, three hypothesis statements are tested to explore the

existence of lead and crime hotspots, and their potential overlap. To understand the

nature and extent of the relationship between lead and crime at the community level,

additional hypotheses statements are tested addressing the conditions under which the

relationship does (or does not) remain tenable. Five final hypotheses are tested to explore

potential environmental justice concerns across Chicago’s communities. Data from a

variety of secondary sources are used in this analysis, including the Chicago Department

of Public Health (2007), the Chicago Police Department, and the U.S. Census (2000).

Chapter five has specified the measurement of each variable in this analysis, thus setting

the foundation for the analysis and results detailed next in chapter six.

Chapter Six:

Analysis and Results

Chapter six presents the analysis and results for this study. Before the results

related to tests of hypotheses statements are presented, a series of preliminary analyses

are performed, and descriptive statistics on the sample are provided. Next, the results

from tests of the first three hypotheses statements are presented to explore whether there

are hotspots of lead and hotspots of crime among Chicago communities. Then, five

hypotheses relating to the nature and extent of the lead-crime relationship are tested.

Finally, the results from the empirical tests of the final five hypotheses are presented.

Those hypotheses explored the distribution of lead in Chicago in relation to

environmental justice theory and explanations.

Preliminary Analyses

Before addressing this study’s research hypotheses, a short series of preliminary

analyses were performed. The purpose of the preliminary analysis was to describe the

data and examine the bivariate relationships between the lead, crime, and control

variables. The preliminary analysis begins by presenting means and standard deviations

for variables that convey key demographic information about the sample. Next, the

bivariate relationships among the variables are assessed empirically.

Descriptive data. Descriptive data for all the control variables (years 2000 and

2010, respectively) are presented in Table 3. These data reveal that in 2000, the average

Chicago Community Areas concentrated an average of 37,610 individuals. In 2000, the

Chicago Community Areas were approximately 31.1% white, 40.89% black, and 21.76%

Hispanic. These data also indicate that the average

Table 3. Sample Demographics for Chicago Community Area Control Variables, 2000

and 2010 Census (n = 77)

Variable

Minimum

Maximum

Area (in square miles) 2.999 1.970 .6067

13.3377

2000

Total Population 37610.6 24443.8 3294 117527

Density 13741.91 7282.774 896.402

35789.09

% Hispanic .2176 .2514 .0059 .8890

% Black .4089 .4109 .0016 .9809

% Divorced .0902 .0258 .0422 .1560

% 15 – 24 yrs old. .1465 .0292 .0840 .2238

Heterogeneity* .3713 .2114 .0437 .7119

% White .3118 .2984 .0032 .9328

% Asian .0435 .0862 .0003 .6071

% Other .0615 .0915 .0052 .6213

SES Index 0 2.05 -3.834 6.225

% Female HH .2186 .1259 .0366 .5979

% Under Poverty .1714 .1238 .0150 .5358

% Public Assist. .0812 .0682 .0023 .2903

% Unemployed .0660 .0358 .0175 .1825

Median Income 46298.32 29408.28 13979.00

138336.5

Residential Mobility 0 1.417 -2.457 3.371

% Vacant .0824 .0540 .0198 .2736

% Moved .3805 .0939 .2143 .6540

% Rent .5174 .2208 .0886 .9106

2010

Total Population 35007.77 22361.95 2876 98514

Density 12886.35 6736.292 956.3817

32522.51

% Hispanic .2553 .2809 .0068 .8920

% Black .3918 .4027 .0025 .9778

% Divorced .0943 .0291 .0393 .1738

% 15 – 24 yrs old. .1490 .0323 .0771 .2550

Heterogeneity .3713 .2111 .0437 .7119

% White .2826 .2817 .0027 .8836

% Asian .0541 .1022 0 .7259

% Other .0159 .0084 .0032 .0413

SES Index 0 1.97 -3.920 4.578

% Female HH .2146 .1265 .0302 .6074

% Under Poverty .1800 .1100 .0190 .5521

% Public Assist. .0397 .0271 .0046 .1354

% Unemployed .0811 .0353 .0290 .1983

Median Income 58491.64 29408.4 17687.

167163.

Residential Mobility 0 1.498 -2.899 3.004

% Vacant .1243 .0581 .0442 .2941

% Moved .3704 .1150 .1733 .6276

% Rent .5027 .1952 .1088 .8860

* Indented items indicate those which were included in scales, but not expressly in models

population of the Chicago Community Area declined from 2000 to 2010. With respect to

race and ethnicity, Chicago Community Areas in 2010 have slightly fewer white or black

residents and slightly more Hispanic residents. The median income across all Chicago

Community Areas was $46,298.32 in 2000 and rose to $58,491.64 in 2010.

The means and standard deviations for lead and crime (from 1999 through 2010) are

presented separately in Table 4. With respect to lead, the average percentage of tested

children who had elevated blood lead level (BLL> 10 μg/dL) per year is presented along

with standard deviations. The data reveal that in 1999, on average, approximately

12.64% of a Chicago Community Area’s tested children carried an elevated blood lead

burden. The data also reveal that this rate declined annually, with the most recent

estimates indicating approximately 1% of a Chicago Community Area’s tested children

carry an elevated blood lead burden. A similar trend emerged across crime rates. Table 4

includes means and standard deviations for annual rates of violent, property, and total

index crimes in Chicago Community Areas (per 1,000 residents). Across all years, the

highest average rates of violent, property, and total crime per community occurred in

2000, at approximately 18, 61, and 79 per 1,000 residents, respectively. In 2010, these

rates dropped to approximately 12, 45, and 57 per 1,000 residents, respectively.

Finally, descriptive information for arrests for juvenile and adult index crime in

2008 are presented in Table 5. Because of missing data, the Riverdale community (#54)

is excluded from this analysis. Overall, in 2008, these data suggest that the rate of

juvenile arrests for index crime surpassed the rate of adult arrests for total, property, and

violent index crimes, respectively. With respect to frequency, arrests for property crime

are more frequent than arrests for violent crimes. Similar to the total distribution of index

crimes, arrests for juvenile and adult index offenses appears skewed. For example, the

rate of arrests per 1,000 for violent juvenile offenses ranged from communities

Table 4. Sample Means, Standard Deviations, Minimums, and Maximums for Lead, Violent Crime, Property Crime, and Total

Crime, 1999 through 2010

Year

Variable

1999

2000

2001

2002

2003

2004

2005

2006

2007

2008

2009

2010

% Children. Elevated Blood

Lead

12.64

10.87

8.61

6.82

4.78

4.23

2.79

1.96

1.73

1.15

.96

.93

10.04

9.11

7.17

6.02

4.27

3.78

2.15

1.57

1.39

.85

.59

.71

Min

Max

39.5

35.6

30.2

23.2

17.3

16.7

7.2

3.5

2.4

Violent Crime Rate† M

17.80

17.03

16.62

14.35

13.53

13.31

13.06

12.70

12.71

12.00

12.11

---

14.47

13.47

13.00

10.94

10.28

10.29

9.95

10.11

10.40

9.49

10.19

Min

---

1.07

1.27

1.33

.80

.99

1.03

.71

1.03

.71

1.04

.536

Max

---

65.50

57.47

51.39

45.45

41.56

45.10

40.64

37.25

47.66

34.50

36.86

Property Crime Rate†

60.87

56.07

53.40

53.47

52.36

48.71

48.88

46.44

46.46

43.63

45.30

---

50.11

44.56

38.99

37.32

38.09

34.51

35.20

33.64

33.68

30.52

24.04

Min

---

16.34

16.24

10.92

9.32

9.76

12.17

11.37

9.33

11.81

11.90

7.78

Max

---

417.44

376.92

327.31

307.30

313.83

287.1

279.53

274.35

278.62

261.66

144.45

Total Crime Rate† M

78.67

73.11

70.02

67.82

65.89

62.03

61.95

59.15

59.18

55.63

57.41

---

57.50

51.15

45.75

43.15

43.53

40.60

41.49

39.84

39.85

35.94

32.30

Min

---

17.41

17.67

12.26

10.13

11.19

12.23

12.08

10.39

12.52

13.06

8.31

Max

---

442.76

396.93

348.73

323.84

329.63

306.51

299.79

296.00

298.08

281.73

160.29

* Lead is computed for all 77 communities each year with the exception of the Riverdale (#54) community for the years 2003 through 2010. In these

years, for lead only the means and standard deviations for n=76 are provided. † Rate per 1,000 residents

108

with 0 (Edison Park- #9 and Forest Glen- #12) to communities with 34 (Fuller Park- #37).

Adult violent offenses ranged from .07 per 1,000 (Forest Glen- #12) to 39 (Fuller Park-

#37).

Table 5. Sample Means, Standard Deviations, Minimums, and Maximums for Juvenile

Index Crime Arrest Rate and Adult Index Crime Arrest Rate, 2008 (n = 76)

Variable

Minimum

Maximum

Juvenile Crime

16.57

40.96

.43

356.49

Adult Crime

11.64

15.06

.57

107.40

Juvenile Property

11.38

37.45

.43

327.89

Adult Property

8.75

11.27

.50

68.53

Juvenile Violent

5.19

5.67

34.08

Adult Violent

2.89

4.71

.07

39.28

Bivariate relationships. The next analyses involved testing the bivariate

relationships between the variables of theoretical interest. The first analysis explored the

bivariate relationship between the control variables, lead, and crime. Results are

presented below in Table 6. Here, associations between the Census 2000 control

variables, 1999 lead variable and 2004 crime average are tested. The data are drawn from

different years to reflect causal processing order and assumptions concerning effect lags

made in prior studies. Because prior literature suggests lead may have a lagged effect on

crime, a five year lag for lead was examined (the percent of tested children who had

elevated blood lead levels in 1999). To control for annual variations in crime rates, the

average rate of index crime for the years 2003, 2004, and 2005 were used. Results from

theses analyses suggest significant, positive relationships with crime and: (1) percent

black (r=.4582, p <.01), percent divorced (r=.4889, p <.01), (2) disadvantage (r=.3054, p

111

<.01), (3) population density (r=.4989, p <.01), and (4) lead (r=.4920, p <.01). A

significant inverse relationship emerged between percent Hispanic and crime (r=-.3238, p

<.01). The inverse relationship observed between crime and heterogeneity (r=-.2024, p =

.0775) was insignificant.

An additional analysis examined the bivariate relationship between the control

variables, lead, and crime 2010. Results from Pearson product-moment correlations are

presented below in Table 7. Here, associations between the Census 2010 control

variables, 2005 lead variable and 2010 crime average are tested. To control for annual

variations in crime rates, the average rate of index crime for the years 2009 and 2010

were used (2011 Annual Data has not yet been made available). Also, because lead data

was not available from Riverdale (Community # 54), Riverdale was not included in this

bivariate assessment (n = 76). Results of this analysis indicate significant, positive

associations between crime and: (1) percent black (r=.5934, p<.01), (2) percent divorced

(r=.4214, p<.01), (3) percent of population between 15 and 24 (r = .4703, p<.01), (4)

disadvantage (r = .4323, p <.01), (5) residential mobility (r = .5911, p <.01), and (6) lead

(r = .6234, p <.01). As illustrated, the largest relationship was between lead and crime.

Significant inverse relationships were found between crime and: percent Hispanic (r = -

.4066, p <.01), and heterogeneity (r = -.3772, p <.01). The bivariate associations between

110

2005 lead, 2010 census variables, and 2010 crime appear relatively similar to associations

found using 1999 lead, 2000 census, and 2004 crime.

In sum, results from the preliminary analysis appear congruent with this study’s

research hypotheses with respect to the association between lead and crime. In both

series of tests, there was a significant, positive association between mean BLL

concentrations in communities and community levels of crime. In addition, the

descriptive statistics suggest a skewed distribution for both lead and crime. For example,

in 1999, the distribution of lead ranged from 0% to 39.5%, and had a mean of 12.64%.

Likewise, crime in 2004 ranged from rates of 11 crimes per 1,000 residents to 329 crimes

per 1,000 residents, with a mean of 65.89 crimes per 1,000 residents. These findings are

consistent with the hypotheses that lead and crime are not normally distributed in

Chicago. A lagged association between lead and crime was also observed in the bivariate

analysis. Thus, additional analysis assessing the relationship between lead and crime

appear justified.

Hypothesis Tests

The results section of this study proceeds by re-stating the research questions and

corresponding hypotheses from chapter 5. Following each hypothesis is a description of

the analytic approach taken to test the hypothesis. Results are reported for each

hypothesis in turn.

Table 6. Bivariate Associations between Lead (1999), Crime Incidents (2004), and Census Control Variables (2000), (n= 77)

Correlation

Matrix

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

(9)

(10)

(11)

1. Area

---

2. Density

-.2545*

---

3. Hispanic

.0244

.2784*

---

4. Black

-.0748

-.2250*

-.5912**

---

5. Divorced

.0240

-.1118

-.6150**

.7071**

---

6. 15 – 24

-.0280

.3101**

.4736**

.0621

-.2676*

---

7. Heterogeneity

.1145

.3055**

.4543**

-.6630**

-.3784**

.1647

---

8. Disadvantage

-.0834

-.1254

-.2183

.7935**

.3369**

.3736**

-.5246**

---

9. Mobility

-.1622

.3528**

-.1358

.3214**

.3051**

.4144**

-.0764

.5155**

---

10. Lead

-.0623

.0122

-.2171

.7438**

.4104**

.3027**

-.4962**

.7220**

.4019**

---

11. Crime -.0759 -.1741 -.3238** .4582** .4889** .1187 -.2024† .3054** .4989** .4920** ---

**p<.01; *p<.05

Table 7. Bivariate Associations between Lead (2005), Crime Incidents (2010), and Census Control Variables (2010), (n= 76)

Correlation

Matrix

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

(9)

(10)

(11)

1. Area

---

2. Density

-.2772*

---

3. Hispanic

.0335

.1774

---

4. Black

-.0718

-.3294**

-.5935**

---

5. Divorced

-.1366

-.2733*

-.5828**

.7578**

---

6. 15 – 24

-.1194

.1387

.1146

.2172†

.0159

---

7. Heterogeneity

.1329

.3121**

.2712*

-.6716**

-.4320**

-.1089

---

8. Disadvantage

-.0831

-.2670*

-.0685

.7405**

.3952**

.3402**

-.5942**

---

9. Mobility

-.1404

.3768**

-.2162†

.3425**

.2145†

.4584**

-.1157

.4243**

---

10. Lead

-.0215

-.1130

-.2349*

.7020**

.3468**

.3832**

-.5327**

.7822**

.5167**

---

11. Crime -.0239 -.1193 -.4066** .5934** .4214** .4703** -.3772** .4323** .5911** .6234** ---

**p<.01; *p<.05; †p <.10

112

115

The Distribution of lead and crime across communities. The first research

question posed by this study was stated previously as “is there a relationship between

hotspots of lead and hotspots of crime?” To address this research question, the

distribution of lead and crime across the Chicago Community Areas must be examined.

As such, three hypotheses are explicated. The first of these hypotheses is:

H1: Lead is distributed non-randomly across Chicago communities, creating

lead hotspots.

To test this first hypothesis, the distribution of lead across Chicago’s communities

was examined. To begin, Shapiro-Francia and Shapiro Wilk tests were performed on the

data to examine the distribution of the lead variable in the year 1999 (Shapiro & Francia,

1972; Shapiro & Wilk, 1965). In this case, both the Shapiro-Francia and Shapiro-Wilk

test the null hypothesis that lead is normally distributed. Results for the Shapiro-Wilk

= .8927, p <.01) and Shapiro-Francia (W’ = .90556, p <.01) reject the null that lead is

distributed normally across Chicago communities. The 1999 lead variable was also

tested for skewness and kurtosis, using Stata’s skewness and kurtosis normality test

(Acock, 2008). Results indicate that based on skewness, there is a statistically

significant lack of normality (p <.01). Lead did not appear significantly non-normal

based on kurtosis (p=.315), and a joint test of normality based on both skewness and

kurtosis indicates significant non-normality in the lead variable (p <.01).

116

Because the research question in this study concerns spatial dependence,

additional analyses were performed to examine the spatial distribution of lead across

Chicago. As such, the data were examined for spatial autocorrelation To test for spatial

autocorrelation, a global Moran’s I statistic was calculated. The Moran’s I statistic tests

the null hypothesis that there is no spatial dependence associated with a given feature

across a geographic area (Anselin, 1988). Values of the Moran’s I range from -1 to +1,

with higher positive values indicating high, positive spatial autocorrelation (similar

values clustered together), and lower negative values indicating high negative spatial

autocorrelation (values surrounded by dissimilar values). Results produced a Moran’s I

of .536, associated with a z-score of 7.68 (p <.01), signifying spatial autocorrelation in

the distribution of lead across Chicago communities. That is, communities that have

higher levels of lead tend to be significantly closer in proximity to one another.

To test for lead hotspots, the lead variable was standardized into z-scores. In the

present study, the average presence of lead in a Chicago Community Area in 1999 was

12.64% (i.e, the percent of a community’s tested children carrying elevated blood lead

burdens). Eighteen communities ranked above average with respect to this indicator for

the presence of lead in the communities of Chicago. An analysis of the standardization

distribution of the lead variable revealed six communities with a presence of lead that

was significantly above the standardized average. In ascending order, these communities

are: Austin (34.2% of tested children, p <.05), West Englewood (34.4% of tested

children, p <.05), West Garfield Park (35% of tested children, p <.05), East Garfield Park

117

(37.8% of tested children, p <.05), Englewood (38.9% of tested children, p <.01), and

Fuller Park (39.5% of tested children, p <.01).

Importantly, when employing the data to identify and isolate lead hotspot, it

becomes clear that lead is not randomly distributed across Chicago communities.

Results indicate approximately six lead hotspots or communities carrying lead burdens

which are significantly higher than the rest of Chicago community areas. As such, these

results are supportive of the first hypothesis.

H2: Crime is distributed non-randomly across Chicago communities, creating

crime hotspots.

To test the second hypothesis, the distribution of crime across Chicago’s

communities was examined. First the Shapiro-Francia and Shapiro Wilk tests were

performed on the data to examine the distribution of the crime rate variable in the year

2004 (Shapiro & Francia, 1972; Shapiro & Wilk, 1965). Again, to control for annual

fluctuation in crime rates, the average crime rate across 2003, 2004, and 2005 was used.

In this case, both the Shapiro-Francia and Shapiro-Wilk test the null hypothesis that lead

is normally distributed. Results for the Shapiro-Wilk (W = .74894, p <.01) and

ShapiroFrancia (W’ = .73429, p <.01) reject the null hypothesis that crime is distributed

normally across Chicago communities. Next, the 2004 crime variable was tested for

skewness and kurtosis, using Stata’s skewness and kurtosis normality test (Acock, 2008).

Results indicate that based on skewness, there is a statistically significant lack of

118

normality (p <.01). Crime in 2004 also appeared significantly non-normal based on

kurtosis (p <.01). A joint test of normality based on both skewness and kurtosis indicates

significant nonnormality in the 2004 crime variable (p <.01).

Similar to tests of H1, additional analyses were performed to examine the spatial

distribution of crime across Chicago. As such, the data were examined for spatial

autocorrelation again using a global Moran’s I. Results produced a Moran’s I of .536,

associated with a z-score of 8.50 (p <.01), signifying spatial autocorrelation in the

distribution of crime across Chicago communities. That is, communities that have higher

levels of crime tend to be significantly closer in proximity to one another.

To test for crime hotspots, the crime variable was standardized into z-scores. The

variable was examined for areas with crime rates above average and significantly above

average. This analysis found the average crime rate in a Chicago Community Area to be

approximately 65 crimes per 1,000 persons. This analysis found thirty two communities

where crime rates exceeded the average crime rate in Chicago. The standardization of

the crime variable revealed three communities with a crime rate that was significantly

above average. In ascending order, these communities are: Near South Side

(approximately 154 crimes per 1,000 persons, p <.05), Fuller Park (approximately 154

crimes per 1,000 persons, p <.05), and Loop (approximately 319 crimes per 1,000

persons, p <.01).

These results suggest that crime is not randomly distributed across Chicago

communities. Results identify approximately three crime hotspots or communities in

119

which crime rates are significantly higher than the rest of Chicago community areas. As

such, these results are supportive of the second hypothesis.

H3: Communities which are lead hotspots will also emerge as crime hotspots.

The results from tests of hypotheses 1 and 2 are mapped simultaneously in Figure

5. These results indicate nine communities in total which have significantly high rates of

lead and/or crime. However, only one community, Fuller Park (# 37), emerged as having

significantly high levels of lead and significantly high rates of crime. Of the remaining

communities, six had significantly high levels of lead, but not significantly high rates of

crime. Two communities had significantly high rates of crime, but not significantly high

levels of lead.

While these results suggest that the communities with the highest levels of lead

and highest levels of crime may not overlap, a relationship between community levels of

lead and community crime rates may still exist. To examine this possibility, the bivariate

relationship between community lead level and community crime rate was tested using

Pearson product-moment correlations. The results of this test are presented in Table 8.

Significant relationships were detected between lead and crime for all years measured,

with the sole exception of 2009 lead and 2009 crime rate. Significant coefficients range

from a minimum value of .2929 to a maximum value of .7941. Bivariate results suggest

the strongest significant association was

120

Figure 5. Distribution of Lead (1999) and Average Total Crime Rate (2004) Across

Chicago Community Areas, (n=77)

found between lead in 1999 and crime rate in 2010, while the weakest significant

association was found between lead in 2008 and crime rate in 2009. Sizes of correlation

121

coefficients suggest that, in general, the association between lead and crime appeared

strongest overall in 2010. The magnitudes of the coefficients also suggest that the

relationship between lead and crime becomes stronger as lead lingers in the environment.

Taken together, it would seem that with the exception of Fuller Park, communities with

significantly high levels of lead and significantly high rates of crime generally do not

coincide precisely as dual lead hotspot and crime hotspot locations. Communities may

not overlap specifically with respect to hotspots for a number of reasons, including the

fact that lead is not the sole cause of crime, and other features of communities contribute

to the aggregate level of crime in a community. While crime and lead hotspots do not

overlap exactly, bivariate associations between community lead level and community

crime rates provide empirical support of a relationship between the distribution of lead

and crime. Specifically, when lead and crime are examined continuously, positive,

significant relationships emerge between lead and crime, revealing that community

levels of lead and crime do positively and significantly covary. With the exception of

2009 levels of lead and 2009 rates of crime, the bivariate association between

community levels of lead and rates of crime emerged as significant in Chicago from

1999 through 2010.

Thus, given the prevailing definition of hotspots, and the overlap of lead and

crime hotspots, it is appropriate to conclude that no support is found for the third

hypothesis in the strictest sense. However, the correlation between community lead level

and community crime rate warrants further investigation. This relationship is examined

in greater depth in the section that follows.

Table 8. Correlation Coefficients for Lead and Total Crime Incident Rate 1999 through 2010

Lead

Total Crime Rate †

1999

2000

2001

2002

2003

2004

2005

2006

2007

2008

2009

2010

2000

.4074**

2001

.4437**

.4010**

.3938**

---

2002

.4671**

.4177**

.4162**

.3894**

---

2003

.4885**

.4398**

.4343**

.4063**

.3952**

---

2004

.4755**

.4272**

.4191**

.3938**

.3818**

.4092**

---

2005

.5097**

.4658**

.4612**

.4375**

.4164**

.4500**

.4535**

---

2006

.5220**

.4803**

.4710**

.4462**

.4301**

.4586**

.4700**

.4013**

---

2007

.5311**

.4880**

.4762**

.4513**

.4475**

.4621**

.4718**

.3972**

.3478**

---

2008

.5372**

.4886**

.4824**

.4552**

.4515**

.4772**

.4845**

.4133**

.3647**

.2841*

---

2009

.5185**

.4722**

.4725**

.4448**

.4362**

.4572**

.4670**

.3885**

.3393**

.2929*

.1578

---

2010

.7941**

.7766**

.7822**

.7520**

.7261**

.7615**

.7491**

.7081**

.6714**

.6015*

.3641**

.4139**

* p<.05; **p<.05

† Per 1,000 Residents

NOTE: Crime data were available for all communities for the years 1999 through 2010. However, lead data were only

available for all communities from 1999 through 2002. From 2003 through 2010, one community- Riverdale (Community

#54)- was missing data for lead. Thus, lead’s relationship with crime from the year 2003 forward is assessed using only lead

and crime rates for 76 communities.

120

124

The Relationship between lead and crime in communities. The second research

question that this study asks is, “what is the nature and extent of the relationship between lead

and crime at the community level?” Results from tests of hypotheses 1 and 2 revealed that lead

and crime appear to concentrate in some communities more so than others. A test of hypothesis

3 finds that while hotspots may not overlap completely, community crime level and community

lead level have a significant, positive relationship to one another. The next four hypothesis tests

in this study explore the conditions under which (if any) the positive, significant relationship

between a community’s lead level and a community’s crime rate persists.

H4: Community lead presence is consistently associated with increases in community

crime, controlling for traditional criminological predictors of crime.

Hypothesis four examines whether the relationship between community lead level and

community crime rate consistently endures while controlling for the theoretical predictors put

forward in the orthodox criminology literature on the distribution of crime at the community

level. The test of this hypothesis proceeded in four steps. First, a cross sectional assessment

was conducted examining the association between lead and crime while controlling for relevant

variables from the 2000 census. Next, a cross sectional assessment was conducted examining

the associations between lead and crime while controlling for theoretically relevant variables

from the 2010 census. Finally, a fixed-effects negative binomial regression is used to examine

125

lead and crime across the Chicago community areas over time (Allison, 2009). Due to missing

data, the Riverdale community (#54) was excluded from these analyses.

The statistical technique selected for assessing the cross-sectional relationships between

lead (lagged) and total crime rate was a negative binomial regression. This technique is justified

because key assumptions of the OLS regression are violated when a count-based dependent

variable (such as crime rate) is analyzed as if it were continuous (Osgood, 2000). Particularly,

crime rates are sensitive to population size, impacting the variance of the error term

differentially across ecological units of different sizes, a condition known as heteroskedasticity.

Also, since the crime rate can only take on positive values, the data distribution will always be

skewed, potentially producing error terms with highly skewed distributions. Heteroskedasticity

is problematic because it impacts the size of the standard error, and in turn, may bias results of

hypothesis tests (Schroeder, Sjoquist, & Stephan, 1986). Criminologists have addressed this

issue by relying on Poisson regression models and negative binomial regression models when

working with count-based outcomes. Here, the negative binomial regression model is preferred

to the Poisson regression model, because overdispersion was detected in the dependent variable.

That is, the variance (s2=1792) greatly exceeded the mean (M=65), making the negative

binomial regression preferable (Long, 1997).

Because the independent variable in this study has demonstrated to be spatially

dependent, the measurement of lead may potentially introduce redundancy into a multivariate

model (Mitchell, 2005). To account for spatial autocorrelation in regression models, a spatial lag

variable was generated for lead. A spatial lag can be understood as a “weighted average of

126

neighboring values,” (Anselin, 2004). In this case, “neighboring” was modeled using the queen

criterion. That is, communities that share contiguous boundaries as well as vertices were

considered to be “neighboring.” “Values,” were the lead levels of a given neighboring

community. To generate the spatial lag variable, a first-order contiguity-based spatial weights

matrix, utilizing the queen criterion, was generated in GeoDa using the Chicago Community

Area polygon shape file. Next, the spatial weights file was used to generate the weighted

average of lead in neighboring communities per each Chicago Community Area. These values

were retained to include in subsequent multiple regression models. Thus, controlling for spatial

autocorrelation.

Next, data for this study were also examined for multicollinearity. While

multicollinearity does not impact the underlying assumptions required of regression,

multicollinearity can impact the size of standard errors and slopes for highly correlated

independent variables (Allison, 1999). Ecological data, such as the data used in this study are

particularly susceptible to multicollinearity. To test for multicollinearity, the variance inflation

factor and condition index number test (Belsley, Kuh, & Welsch, 1980) were examined across

lead and the control variables. VIF values suggest multicollinearity among these variables

(mean VIF of 6.05), particularly among disadvantage (VIF of 8.03) and percent black (VIF of

20.80). To address this, the percent black variable was log transformed, and collinearity

diagnostics were re-assessed only on variables with significant bivariate associations with lead

and crime in 2000 and 2010. These adjustments reduced the VIF for percent black to 4.55 and

the VIF for disadvantage to 4.36. The mean VIF following these adjustments was 3.19, and the

127

conditional index number was 20.0655. All VIFs fell under recommended cutoffs of 5 (Walker

& Madden,

2012), and the conditional index number falls under the suggested cutoff of 30 (Walker &

Madden, 2012). Collinearity diagnostics for all models are reported in Appendix B.

Following these diagnostics, three negative binomial regressions were performed to

examine the relationship between lead and crime, controlling for other sociologically relevant

predictors. Results of these regressions are presented in Table 8.

In the first model, the average total crime rate in 2004 was regressed on variables that

emerged as having significant bivariate associations with crime. Specifically, average crime rate

in 2004 was regressed on disadvantage, mobility, heterogeneity, logged percent black, percent

Hispanic, percent divorced, lead, and the spatial lag of lead. Both lead and the spatial lag of lead

emerged as having a positive and significant association with average total crime rate. Mobility

and percent divorced also emerged has having positive, significant relationships with average

total crime rate. Counter to what was hypothesized, disadvantage emerged as having a

significant inverse association with crime rate in 2004. Percent black wasnot significantly

associated with crime rate (p=.088).

The next negative binomial regression regressed crime rates in 2010 on lead, the spatial

lag for lead, residential mobility, racial/ethnic heterogeneity, the log of percent black, percent

Hispanic, and percent divorced. To capture the five year lagged effect of lead, the 2005

community lead and the spatial lag of lead were used. Annual fluctuation in crime rate was

adjusted for by averaging the total crime between the years 2009 and 2010 (data from 2011 are

128

not yet available). Control variables are drawn from the 2010 Census. Results are presented in

Table 9. Controlling for confounders, the relationship between 2005 lead and 2010 crime rate

was outside the range of the test for statistical significance (p = .075). The spatial lag of lead did

not emerge as significant in 2010, nor did percent Hispanic or percent divorced. Increases in the

log of percent black, and residential mobility were associated with significant increases in crime

rate. The relationship between disadvantage and crime rate again emerged as significant and

inverse.

Table 9. Negative Binomial Regressions and Fixed Effects Negative Binomial Regression

Predicting Total Average Crime Rate with Lead and Other Sociological Predictors, (n = 76)

Crime 2004 Crime 2010 Crime 2004-2010

Fixed Effects

Variables b SE b SE b SE

Lead

.017* .007

.054† .031

.009** .001

Spatial Lag: Lead

.024* .011

.076 .054

--- ---

Disadvantage

-.122**.034

-.118**.036

--- ---

Mobility

.163** .031

.158** .029

--- ---

Heterogeneity

.150 .254

-.427† .241

--- ---

(ℓn)% Black

.070† .041

.147** .043

--- ---

% Hispanic

.001 .002

.003 .002

--- ---

% Divorced

.058** .022

.019 .019

--- ---

Intercept

Model Diagnostics

2.77** .253

3.10** .272

4.78** .075

χ2

89.99

93.06

115.42

-2 log likelihood

-326.177

-312.857

-2839.983

Pseudo-R2

.1212

.1295

---

† p <.10, *p < .05, **p < .01

129

Finally, a third model examined the associations between lead and crime over time. The

association between lead and crime is assessed longitudinally using a fixed effects negative

binomial regression model (Allison, 2005; 2009). The fixed effects regression technique was

selected for many reasons. First, the fixed-effects approach is well suited to examine within-unit

change, and this study seeks to examine within-community associations between lead and crime

(Allison, 2009). This is relevant theoretically, as social disorganization theory aims to explain

why some communities are more crime-prone than others (Shaw & McKay, 1942), and the

fixed-effects negative binomial regression has been employed for this purpose in prior literature

(for example, see Martinez, Stowell, & Lee, 2010). Second, the fixed effects model provides

researchers an analytical technique for modeling change over time, and its utilization here allows

for a longitudinal assessment of the relationship between lead and crime. Third, the fixed effects

regression model controls for time-invariant factors that have not been measured and expressly

included in the model. That is, the fixed-effects model is especially well suited to handle

omitted variable bias (Allison, 2009). The negative binomial regression was selected because of

the count-based nature of the dependent variable.

Results for the fixed effects negative binomial regression model are presented in Table 9.

In this analysis, the total count of crime in a community area was regressed on lead. A five year

lag was used for the lead variable. Because data for control variables were only available for

census years (2000 and 2010, respectively), these variables were not included in the fixed effects

regression model. Overall, the fixed effects model fit the data well (X2 = 115.42, p<.01).

Results indicate that lead and crime are positively and significantly associated, even while

130

controlling for time-invariant community characteristics (b=.009, p<.01). This observation is

consistent with the observations from the two cross-sectional models.

In sum, a significant, positive relationship emerges between the presence of lead and the

presence of crime in a community in two out of three assessments for total crime. This

association is detected in a cross sectional assessment of lead’s impact on crime in 2004. These

data suggest that the relationship remains tenable even when controlling for other theoretically

relevant predictors: disadvantage, racial/ethnic heterogeneity, mobility, percent black, percent

Hispanic, and percent divorced in 2004. However, in an examination of the same variables in

2010, the significant association between lead and crime did not persist. Finally, a third analysis

evinced a positive, significant relationship between lead and crime longitudinally, controlling

only for time invariant factors. What these results imply is a somewhat inconsistent association

between lead and total rates of crime. As such, support for hypothesis four is not found.

However, because the inconsistency may be attributable to offense type, further analyses are

justified. This issue is explored in the test of hypothesis five.

H5: Community lead presence will be a more consistent predictor of violent crime than

property crime.

Hypothesis five concerns the strength of the relationship between lead and crime with

respect to type of crime. Based on prior literature, it is hypothesized that lead will be a stronger

predictor of violent crime than property crime (Stretesky and Lynch, 2004). Total crime rates

131

are divided separately into rates of violent crime and rates of property crime. The test of this

hypothesis proceeded in three steps. First, a cross sectional assessment was conducted

examining the association between lead and 2004 violent crime and property crime while

controlling for relevant variables from the 2000 census. Next, a cross sectional assessment was

conducted examining the associations between 2010 lead and violent crime and property crime

while controlling for theoretically relevant variables from the 2010 census. Negative binomial

regressions are used for the cross sectional assessments. Finally, a fixed-effects negative

binomial regression is used to examine lead and violent crime and property crime across the

Chicago community areas over time (Allison, 2009). Due to missing data, the Riverdale

community (#54) was excluded from these analyses.

The distribution of lead and violent crime and lead and property crime are mapped in

Figures 6 and 7, respectively. While both maps contain the same lead data (1999), Figure 6

displays the distribution of average violent crime rates in 2004, while Figure 7 displays the

distribution of average property crime rates in 2004. A side by side comparison of these maps

suggests differences in the distribution of violent crime as opposed to property crime. For

example, Loop (#32) emerges as a community that is 2.5 standard deviations above the Chicago

community mean in terms of property crime, but only .35 standard deviations above the mean

for violent crime. Likewise, Englewood (#68) is nearly 2 standard deviations above the mean

with respect to violent crime (1.9), but less than one half of a standard deviation above the mean

for property crime (.41). Together, these maps elucidate a possible explanation for the somewhat

132

contradictory findings in the test of hypothesis four. However, multivariate statistics are

necessary to employ more stringent criteria to examine this possibility.

As such, the average violent crime rate in 2004 was regressed on lead, the spatial lag of

lead, disadvantage, mobility, heterogeneity, the log of percent black, percent Hispanic, and

percent divorced. Annual fluctuation in crime rate was adjusted for by averaging the violent

crime rate across the years 2003, 2004, and 2005. Results are presented in Table 10. In this

analysis, both lead and the spatial lag of lead emerge as significantly and positively associated

LEFT: Figure 6. Distribution of Lead (1999) and Average Violent Crime Rate (2004) Across Chicago (n=76)

RIGHT: Figure 7. Distribution of Lead (1999) and Average Property Crime Rate (2004) Across Chicago (n=76)

129

135

with the violent crime rate in 2004. Likewise, mobility, the log of percent black, percent

Hispanic, and percent divorced also emerge as positively and significantly associated with

the violent crime rate in 2004. Of the predictors, only disadvantage and heterogeneity do

not emerge as significantly associated with violent crime in 2004.

Results differed in the next analysis, which regressed the average property crime

rate in 2004 on predictors. Annual fluctuation in property crime rate was adjusted by

averaging the property crime rate across the years 2003, 2004, and 2005. Findings from

this analysis are presented in Table 10. Lead’s effect in this analysis was outside the range

of the test for statistical significance (p = .053), while the spatial lag of lead emerged as

positive and significantly related to property crime. Mobility and percent divorced also

emerged as positively and significantly associated with property crime rates.

Disadvantage emerged as significantly related to property crime, although in the opposite

direction as hypothesized by social disorganization theory.

Property crime and violent crime were also regressed on predictors for the year

2010. Results from these analyses are presented in Table 11. With respect to violent crime

rate in 2010, results closely paralleled the 2004 analysis. Lead emerged as positively and

significantly related to violent crime in 2010. Mobility, the log of percent black, and

percent Hispanic again emerged as positively and significantly related to violent crime in

2010. Percent divorced was no longer significant, and heterogeneity and disadvantage

continued to remain below significance.

136

Table 10. Negative Binomial Regressions Predicting 2004 Violent Crime and Property

Crime with Lead and Other Sociological Variables, (n = 76)

Violent Crime

Property Crime

Variables

b SE

Lead

.019**

.005

.014† .007

Spatial Lag: Lead

.021*

.009

.023* .012

Disadvantage

.038

.035

-.155**.036

Mobility

.076*

.032

.171**

.033

Heterogeneity

-.149

.237

.190

.268

(ℓn)% Black

.166**

.048

.050

.043

% Hispanic

.005*

.002

.000

.002

% Divorced

.047*

.022

.060**

.023

Intercept

.919*

.245

2.59**

.266

Model Diagnostics

χ2

175.66

76.93

-2 log likelihood

-182.405

-313.035

Pseudo-R2

.3250

.1094

† p <.10, *p < .05, **p < .01

With respect to property crime in 2010, a few notable changes emerged compared

to the 2004 analysis. Neither lead nor the spatial lags for lead were significantly associated

with property crime rates in 2010. Mobility and the log of percent black emerged as

significantly associated with property crime rates in 2010. Specifically, as mobility and

log of percent black increased, associated increases in rate of property crime were

observed. Disadvantage emerged as being significantly related to property crime rates.

The relationship, however, was again in the opposite direction than hypothesized Table 11.

137

Negative Binomial Regressions Predicting 2010 Violent Crime and Property Crime with

Lead and Other Sociological Variables, (n = 76)

138

† p <.10, *p < .05, **p < .01

Table 12. Fixed Effects Negative Binomial Regressions Predicting Violent Crime and

Property Crime with Lead, 2004 through 2010

† p <.10, *p < .05, **p < .01

Violent Crime

Property Crime

Variables

b SE

Lead

.063*

.027

.049 .032

Spatial Lag: Lead

.098†

.052

.061 .056

Disadvantage

.022

.040

-.140**.038

Mobility

.097*

.031

.161** .032

Heterogeneity

-.424

.269

-.382 .254

(ℓn)% Black

.264** .056

.129** .046

% Hispanic

.006** .002

-.002 .002

% Divorced

.026 .020

.018 .020

Intercept

.782* .304

3.00** .286

Model Diagnostics

χ2

170.25

73.87

-2 log likelihood

-177.595

-300.199

Pseudo-R2

.3240

.1096

Violent Crime

Property Crime

Variables

b SE

Lead

.010** .001

.009** .001

Intercept

4.25** .091

4.72** .076

Model Diagnostics

χ2

77.46

96.78

-2 log likelihood

-2222.1209

-2760.0981

139

by social disorganization theory. Racial/ethnic heterogeneity, percent Hispanic, and percent

divorced did not appear to be significantly associated with property crime rate in 2010.

Next, the longitudinal relationship between lead and violent crime as well as lead

and property crime was assessed using negative binomial fixed effects regression. Based

on results from prior studies, a five year lag of lead was used to predict crime rates from

2004 through 2010. Results from this assessment are presented in Table 12. Lead

emerged as positively and significantly associated with both violent crime (b=.010, p<.01)

and property crime (b=.009, p<.01), while controlling for all time-stable factors. These

results suggest that increases in percent of lead are positively and significantly associated

with increases in community violent crime rates, net of time-stable factors. Results also

suggest that increases in percent of lead are positively and significantly associated with

increases in community property crime rates, net of time-stable factors.

In sum, the relationship between lead and crime does appear to differ slightly with

respect to violent and property crime. For example, in the 2004 cross sectional model, lead

was positively and significantly associated with violent crime, but the lead’s impact on

property crime in 2010 was outside the range of the test for statistical significance. In

2010, lead was positively and significantly associated with violent crime, but not with

property crime. Results from the fixed effects regression reveal a positive and significant

association between lead and both property and violent crime. The coefficient for violent

crime is only slightly larger than the coefficient for property crime. The results suggest

that the association between lead and property crime appears to weaken when controlling

140

for the sociologically relevant variables in the cross sectional models. Overall, data appear

to support hypothesis 5.

H6: Community lead presence will be a more consistent predictor of juvenile crime

than adult crime.

Hypothesis six concerns the persistence of the relationship between lead and crime

with respect to age. Prior literature finds that young people are especially vulnerable to

environmental toxins (Chatham-Stephens, Mann, Schwartz, & Landrigan, 2011-2012;

Landrigan & Carlson, 1995; Moore, 2003). As such, it is hypothesized that lead will be a

stronger predictor of juvenile arrest rates than adult arrest rates. To test this hypothesis,

first the juvenile arrest rates for index offenses in 2008 were regressed on relevant

sociological predictors, including lead. Next, the adult arrest rates for index offenses in

2008 were regressed on relevant sociological predictors, including lead. A five year lag of

lead was used (2003 lead levels) in both models, as were controls from the 2000 census.

Negative binomial regressions are used to examine associations between juvenile arrest

rates and adult arrest rates. Due to missing data, the Riverdale community (#54) was

excluded from these analyses.

Results from the negative binomial regression testing the association between

juvenile arrests and adult arrests are presented in Table 13. In this analysis, lead emerges as

positively associated with the juvenile arrest rate, but the association is slightly beyond the

test for statistical significance (p=.056). The spatial lag for lead is not significantly

141

associated with juvenile arrest rate in 2008, and neither is heterogeneity or percent

Hispanic. Mobility and the log of percent black emerge as positively and significantly

associated with juvenile arrest rates. Disadvantage emerges as significantly related to

juvenile arrests, but the two variables are inversely related.

Table 13. Negative Binomial Regressions Predicting 2008 Arrests for Juvenile Index

Crime and Arrests for Adult Index Crime with Lead and Other Sociological Variables, (n

= 76)

Juvenile Arrests

Adult Arrests

Variables

b SE

Lead

.076† .040

.119** .032

Spatial Lag: Lead

.057 .067

.030 .053

Disadvantage

-.414**.073

-.179**.064

Mobility

.284** .065

.248** .057

Heterogeneity

-.202 .602

.236

.472

(ℓn)% Black

.273** .094

.108

.078

% Hispanic

.007 .005

.007†

.004

% Divorced

.112* .052

.103*

.042

Intercept

.010 .584

.024

.496

Model Diagnostics

χ2

76.74

77.58

-2 log likelihood

-253.235

-224.433

Pseudo-R2 .1316 .1474

† p <.10, *p < .05, **p < .01

Next, adult arrest rates in 2008 were regressed on lead and other relevant

predictors. Findings are presented in Table 13. Results differ slightly from juvenile arrest

142

rates. Unlike the juvenile arrest rate model, lead emerges as positively and significantly

associated with the adult arrest rate in 2008. Also unlike the juvenile arrest rate model,

the log of percent black is no longer significantly related to arrest rate; also, percent

Hispanic emerges as positively associated but insignificant predictor (p=.100). Similar to

the juvenile arrest model, mobility is positively and significantly associated with adult

arrests, as is percent divorced. The spatial lag for lead and heterogeneity are not

significantly associated with either juvenile or adult arrests.

Taken together, the relationship between lead and arrest rate does appear to differ

slightly with respect to adult and juvenile arrests. However, the differences are the inverse

of the association predicted by this hypothesis. Specifically, the association between lead

and juvenile arrests emerges as outside the range of significance (p=.056), while the

relationship between lead and adult arrests emerges as significant (p=.000). As such,

support is not found for hypothesis 6.

H7: Community lead presence will be a more consistent predictor of juvenile

property crime than adult property crime.

Hypothesis seven concerns the persistence of the relationship between lead and

crime with respect to age of offender and property crime. While support for hypothesis 6

was not found (and results suggested the opposite- that effects of lead may be more robust

for adult arrests), tests of hypothesis five indicate differences among predictors of violent

143

crime and property crime. Hypothesis seven anticipates that lead will be a more

persistent predictor of juvenile property crime arrests than adult property crime arrests.

To test this hypothesis, first the juvenile property crime arrest rates in 2008 were regressed

on relevant sociological predictors, including lead. Next, the adult arrest rates for index

offenses in 2008 were regressed on relevant sociological predictors, including lead. A five

year lag of lead was used (2003 lead levels) in both models, and predictors from the 2000

census were included. Negative binomial regressions are used to examine associations

between juvenile property crime arrest rates and adult property crime arrest rates. Due to

missing data, the Riverdale community (#54) was excluded from these analyses.

Table 14. Negative Binomial Regressions Predicting 2008 Arrests for Juvenile Property

Crime and Arrests for Adult Property Crime with Lead and Other Sociological Predictors,

(n = 76)

Juvenile Property

Adult Property

Arrests

Variables

Lead

.091†

.048

.114**

.035

Spatial Lag: Lead

.057

.080

.041

.057

Disadvantage

-.516**

.086

-.221**

.068

Mobility

.337*

.077

.279**

.061

Heterogeneity

-.505

.736

.272

.510

(ℓn)% Black

.267*

.110

.109

.084

% Hispanic

.007

.006

.005

% Divorced

.097

.061

.095*

.045

Intercept

-.272

.693

-.169

.534

Model Diagnostics

χ2

74.01

68.57

-2 log likelihood

-226.016

-208.374

144

Pseudo-R2

.1407

.1413

† p <.10, *p < .05, **p < .01

Table 14 displays results from the negative binomial regression examining

predictors of 2008 juvenile property crime arrests. Results of this model were very

similar to the regression of total juvenile arrests on the same predictors. Again, increases

in lead are slightly outside the range of a significant effect (p = .059). The spatial lag for

lead did not emerge as significantly associated with juvenile property crime arrests.

Increases in mobility and the log of percent black emerged as significantly associated with

increases in juvenile property crime arrests. Disadvantage demonstrated to be inversely

related to juvenile property crime arrests. Unlike the association between divorce and

total juvenile arrests, the association between divorce and juvenile property crime arrests

did not emerge as significant.

Results from the negative binomial regression model examining predictors of 2008

adult property crime arrests are also presented in Table 14. The results of this model are

also very similar to the regression examining total adult arrests on the same predictors.

Lead was related positively and significantly to arrests for adult property crimes.

Residential mobility and percent divorced were also positively and significantly

associated with adult property crime arrest rates. An inverse association was detected

between disadvantage and adult property crime arrests. Unlike the model examining total

adult arrest rates, percent Hispanic was not found to be significantly related to arrests for

adult property crimes.

145

Taken together, the relationship between lead and adult property crime arrests and

juvenile property crime arrests is very similar to trends in lead and adult and juvenile arrests

overall. Lead’s impact on juvenile property crime arrests falls outside the range of

significance (p=.059), while lead impact on adult property crime arrests remains significant

(p=.001). What this suggests is, similar to total arrest rates for juveniles and adults, lead

appears to be a stronger predictor of adult arrest rates than juvenile arrest rates. These

results are similar to results of the test for hypothesis six. Thus, these data do not appear to

support hypothesis seven.

H8: Community lead presence will be a more consistent predictor of juvenile

violent crime than adult violent crime.

Hypothesis eight concerns the strength of the relationship between lead and crime

with respect to age of offender and violent crime. While support for neither hypothesis 6

nor hypothesis 7 was found (and results suggested the opposite effect - that the effect of

lead may be stronger for adult arrests), tests of hypothesis five indicate differences among

predictors of violent crime and property crime. Hypothesis eight anticipates that lead will

be a stronger predictor of juvenile violent crime arrest rates than adult violent crime arrest

rates. To test this hypothesis, first the juvenile violent crime arrest rates in 2008 were

regressed on relevant sociological predictors, including lead. Next, the adult violent crime

arrest rates for index offenses in 2008 were regressed on relevant sociological predictors,

including lead. A five year lag of lead was used (2003 lead levels) in both models, and

146

predictors from the 2000 census were included. Negative binomial regressions are used to

examine associations between juvenile violent crime arrest rates and adult violent crime

arrest rates. Due to missing data, the Riverdale community (#54) was excluded from

these analyses.

Results from the negative binomial regression testing the association between

predictors and juvenile violent crime arrests are presented in Table 15. In this analysis,

lead emerges as positively associated with the juvenile arrest rate, but the association is

once again slightly outside the measure of statistical significance (p=.062). The spatial lag

for lead is not significantly associated with juvenile violent crime arrest rate in 2008, and

neither is heterogeneity or percent Hispanic. Mobility, the log of percent black, and

percent divorced emerge as positively and significantly associated with juvenile violent

crime arrest rates. Disadvantage emerges as significantly related to juvenile arrests, but

the two variables are inversely associated.

Results from the negative binomial regression testing the association between adult

violent crime arrests are also presented in Table 15. Results illustrate differences between

predictors of arrests for juvenile violent crime and arrests for adult violent crime. Unlike

the juvenile arrest rate model, lead emerges as positively and significantly associated with

the adult violent crime arrest rate in 2008. Also unlike the juvenile arrest rate model, the

percentage of Hispanic residents emerges as significant, but the log of percent black,

mobility, and disadvantage do not emerge as significantly related to violent crime arrest

rate. The only predictor that appears to be significantly associated with both arrests for

juvenile violent crime and adult violent crime is percent divorced.

147

Taken together, results from both models are similar to findings from tests of

hypotheses six and seven. Specifically, lead’s relationship with arrests for violent juvenile

crime falls outside the range of significance. However, lead’s relationship with arrests for

adult violent crime remains significant. As such, it does not appear as though lead is a more

consistent predictor of juvenile violent crime than adult violent crime. Instead, these results

imply that lead may actually be a more consistent predictor of adult violent crime than

juvenile violent crime. Thus, the data does not appear to support hypothesis eight.

Table 15. Negative Binomial Regressions Predicting 2008 Arrests for Juvenile Violent

Crime and Arrests for Adult Violent Crime with Lead and Other Sociological Predictors,

(n = 76)

148

† p <.10, *p < .05, **p < .01

In sum, results from tests of hypotheses four through eight suggest that the

relationship between lead and crime is sensitive to type of crime and age of offender. The

findings discussed here appear to support hypotheses four and five. That is, the

relationship between lead and crime appears to persist even when controlling for other

sociologically relevant variables. Also, lead appears to be a stronger, more consistent

predictor of rates of violent crime over rates of property crime. Findings do not, however,

appear to support hypotheses six, seven, or eight. These data suggest that lead may be a

stronger predictor of adult arrest rates rather than juvenile arrest rates. This observation

appeared to persist regardless of type of crime (property or arrest).

Environmental justice issues. The last five hypotheses take into consideration

environmental justice literature which documents a disproportionately high presence of

environmental hazards in disenfranchised communities (Lerner, 2005; Brulle & Pellow,

Juvenile Violent

Adult Violent

Arrests

Variables

Lead

.065†

.035

.154**

.039

Spatial Lag: Lead

.020

.059

-.062

.067

Disadvantage

-.206**

.070

.009

.098

Mobility

.162**

.061

.116

.085

Heterogeneity

-.045

.506

.014

.622

(ℓn)% Black

.281**

.093

.109

.115

% Hispanic

.005

.012*

.006

% Divorced

.103*

.045

.122*

.055

Intercept

-.754

.541

-1.38*

.691

Model Diagnostics

χ2

65.04

73.16

-2 log likelihood

-171.880

-129.903

Pseudo-R2

.1591

.2197

149

2006; Bullard 1994; 1996; Checker, 2007; Hipp & Lakon, 2010; Krieg, 2005; Opp, 2012;

Pastor, Morello-Frosch, & Sadd, 2006; Stretesky & Lynch, 2002). This final series of

hypothesis tests represent a brief exploratory analysis that examines associations between

the distribution of lead and race, class, and ethnicity across Chicago Community Areas.

It is expected:

H9: Percentage of black residents in a community will correspond positively and

significantly with level of lead in a community.

H10: Percentage of Hispanic residents in a community will correspond positively

and significantly with level of lead in a community.

H11: Disadvantage will correspond positively and significantly with level of lead

in a community.

H12: Disadvantage and percentage of black residents will interact and correspond

positively and significantly with level of lead in a community.

H13: Disadvantage and percentage of Hispanic residents will interact and

correspond positively and significantly with level of lead in a community.

Tests of hypotheses 9 through 10 proceeded by examining the distribution of lead

in 2010 and race, ethnicity, and class variables from the 2010 census (as this was the most

150

recent year of data available). Disadvantage was captured using the principal component

index from 2010 census variables. An interaction term was developed between

disadvantage and percent black to test hypothesis 12. An interaction term was also

developed between disadvantage and percent Hispanic to test hypothesis 13. Next,

Pearson product moment correlations were used to examine the strength and direction

between race, ethnicity, and class variables and lead.

Results of the Pearson product moment correlation are presented in Table 16. With

respect to hypothesis nine, a positive and significant association between percent black and

lead is found. The size of the relationship (r=.6340, p <.01) implies a moderately strong

association between percentage of black residents and a communities lead level.

Regarding hypothesis ten, percent Hispanic and level of lead do not appear to be

significantly related. With respect to hypothesis eleven, a positive and significant

association between community disadvantage and lead emerges. The size of the

relationship (r=.6312, p <.01) suggests a moderately strong association between degree of

community disadvantage and level of lead. Regarding hypothesis twelve, a strong,

significant, positive relationship was detected between percent black and disadvantage

(r=.7424, p<.01), illustrating interaction. An interaction term between percent black and

disadvantage was also found to be significantly related to percent of lead. The value of the

coefficient (r=.5860, p<.01) suggests a positive, moderately strong relationship between

lead and communities that are both economically disadvantaged and have a high

percentage of black residents. Lastly, a test of hypothesis thirteen does not detect a

significant relationship between percentage of Hispanic residents and level of

151

disadvantage. An interaction term between percent Hispanic and disadvantage does

produce a relatively weak significant relationship with lead (r=.2558, p<.01). However,

given that percent Hispanic by itself is not significantly associated with lead or

disadvantage, the significance of the interaction term with lead is likely attributable to the

larger influence of disadvantage.

Table 16. Bivariate Associations between Percent Black, Percent Hispanic,

Disadvantage, Interaction Terms, and Lead, 2010 (n= 77)

Variable

(1)

(2)

(3)

(4)

(5)

(6)

1. Lead

---

2. % Black

.6340**

---

3. % Hispanic

-.1525

-.5935**

---

4. Disadvantage

.6312**

.7424**

-.0709

---

5. %Black*Disadvantage

.5860**

.7689**

-.3778**

.8565**

---

6. %Hispanic*Disadvantage

.2558*

.2825*

.1858

.5639**

.2821*

---

*=p<.05; **p<.01 With respect to environmental

justice questions, support is detected for some, but not all, hypotheses. Pearson product

moment correlations detect significant associations between lead and percent black,

disadvantage, and an interaction between disadvantage and percent black. Thus, the data

appear to support hypotheses nine, eleven, and twelve. However, Pearson product moment

correlations fail to detect significant associations between lead and percent Hispanic. Thus,

the data do not appear to support hypothesis ten. Findings also suggest that percent Hispanic

and disadvantage do not co-vary, and as such, support is not found for hypothesis thirteen.

Conclusion

In sum, chapter six presents the analytic procedure and results for tests of all

thirteen study hypotheses. A summary of the results for the tests of each hypothesis can be

found in Table 17. Both lead and crime appear to have skewed distributions across

152

Chicago Community Areas. However, an examination of communities that are crime

hotspots (communities with significantly high rates of crime) and communities of lead

hotspots (communities with significantly high levels of lead) does not detect consistent

overlap. The relationship between lead and crime does appear to be sensitive to type of

offense, with more consistent associations found for rates of violent offenses. The

relationship between lead and crime also appears sensitive to age, but with lead appearing

to lend more predictive power to adult arrest rates than juvenile arrest rates. Finally, lead

appears positively and significantly related to percent black, disadvantage, and the

interaction of percent black and disadvantage. However, lead does not emerge as

significantly associated with percent Hispanic. The implications of these findings are

discussed in detail in chapter 7.

153

Table 17. Summary of Hypotheses Tests

Supported?

Hypothesis

Yes

H1: Lead is distributed non-randomly across Chicago

communities, creating lead hotspots.

H2: Crime is distributed non-randomly across Chicago

communities, creating crime hotspots.

H3: Communities which are lead hotspots will also emerge as

crime hotspots.

H4: Community lead presence is consistently associated with

increases in community crime, controlling for traditional

criminological predictors of crime.

H5: Community lead presence will be a more consistent predictor

of violent crime than property crime.

154

H6: Community lead presence will be a more consistent predictor

of juvenile crime than adult crime.

H7: Community lead presence will be a more consistent predictor

of juvenile property crime than adult property crime.

H8: Community lead presence will be a more consistent predictor

of juvenile violent crime than adult violent crime.

H9: Percentage of black residents in a community will correspond

positively and significantly with level of lead in a community.

H10: Percentage of Hispanic residents in a community will

correspond positively and significantly with level of lead in a

community.

H11: Disadvantage will correspond positively and significantly with

level of lead in a community

H12: Disadvantage and percentage of black residents will interact

and correspond positively and significantly with level of lead in a

community.

H13: Disadvantage and percentage of Hispanic residents will

interact and correspond positively and significantly with level of

lead in a community.

Chapter Seven:

Discussion and Conclusion

155

The purpose of this final chapter is to briefly summarize the results of this study, as

presented in chapter six. Following this summary, the limitations of this study are

acknowledged and discussed. Next, the implications these results have on criminological

theory are presented, as well as the implications these results have for community crime

reduction endeavors. Lastly, directions for future research are proposed, followed by this

study’s conclusion.

Summary of Findings

The objective of this study was to address three interrelated research questions: Is

there a relationship between hotspots of lead and hotspots of crime? Stemming from this

question, this study also asked, what is the nature and extent of the relationship between

lead and crime? Lastly, does the distribution of lead pose environmental justice concerns

with respect to race, class, and ethnicity in Chicago communities? To resolve these

questions, thirteen research hypotheses were explicated, and empirically evaluated in the

previous chapter.

The Distribution of lead and crime across communities. The first research

question asked in this study asks if there is a relationship between hotspots of lead and

hotspots of crime. As such, the first three hypotheses addressed the issue of identification

and overlay of crime and lead hotspots. In doing so, the distribution of lead and crime

across Chicago’s community areas was examined. It became apparent that lead and crime

hotspots do, in fact, exist among Chicago’s communities.

Counter to what was hypothesized, however, a geographic overlay did not emerge

between lead and crime hotspots; of the nine lead or crime hotspots, only one community

156

(Fuller Park) appeared to concentrate significantly high levels of lead and crime. One

explanation for this may be that in this sample, the distribution of index crime appeared

sensitive to type of offense. That is, the communities with the highest rates of property

crime (top five, in ascending order: Chatham, Fuller Park, Near West Side, Near South

Side, and Loop) were not synonymous with the communities with the highest rates of

violent crime (top five, in ascending order: Englewood, West Garfield Park, East Garfield

Park, Fuller Park, and Washington Park). The observation that hotspots differ with respect

to offense type has been observed previously in the literature (see, for example: Sherman,

Gartin, & Buerger, 1989; Weisburd et al., 1993). Moreover, prior literature suggests lead

may be a stronger predictor of violent crime than property crime (Pihl & Ervin, 1990), and

differential distributions of hotspots of violent crime versus hotspots of property crime are

consistent with this. A Pearson product moment correlation reveals a consistent,

significant, positive bivariate association between lead and total crime every year from

1999 through 2010 (with the sole exception of the 2009 lead rate and 2009 crime rate).

Multivariate analyses testing this association appeared justified. Finally, it is also plausible

that the measure of hotspots itself impacts the analysis. As indicated in chapter six, the

literature suggests that there may be more than one way to measure hotspots, and that the

results provided here may be a result of the methodology that has been employed. Thus,

future research needs to address this issue in greater depth before any definitive conclusion

about the relationship between crime and lead hotspots can be offered.

The Relationship between lead and crime in communities. The next research

question this study addressed the nature and extent of the relationship between lead and

157

crime. Specifically, hypotheses four through eight examined the condition(s) under which

the observed bivariate association between lead and crime persists. While a significant

relationship is observed between lead and total crime rate in 2004, net of confounders, the

association falls beyond the range of significance in 2010. Controlling only for time

invariant factors, a significant association is observed between changes in lead and

changes in crime rate. Taken together, lead does not appear to consistently have a

significant impact on incidents of total crime.

To examine if these somewhat conflicting results are attributable to type of crime,

additional analyses were performed. A chart summarizing outcomes of all multivariate

analyses (tests of hypotheses four through eight) is presented in Table 18, with X denoting

a significant association at a minimum of at least the .05 level. Overall, results of the

multivariate analyses suggest that the relationship between lead and crime is sensitive to:

(1) the measurement of crime (arrest versus incident), and subsequently, (2) type of

offense, and (3) age of offender.

Results of this study illustrate important differences with respect to lead’s impact

on crime when the measurement of the dependent variable (“crime”) changes from crime

incident rate to crime arrest rate. Arrests for crime (capturing traits of an offender) are

distinguishable from incidents of crime (demographics like age may be unknown). That

is, results from tests examining lead’s impact on arrests for crime may not be directly

comparable to tests examining lead’s impact on incidents of crime. For example, lead

does not appear to consistently predict total crime incident rates- that is, lead is

significantly associated with total crime incident rates in 2004 net of confounders, but not

158

2010. Lead is also significantly associated with change in total crime incidents between

2004 and 2010, net of time-stable variables. Separating property crime incident rates and

violent crime incident rates reveals important differences with lead’s influence on crime

incident rates. Specifically, lead appears to be significantly and consistently associated

with violent crime incident rates. That is, lead appears to significantly predict violent

crime, net of mainstream sociological predictors as well as unmeasured time-stable

variables. However, the association between lead and incidents of property crime appears

less consistent. Specifically, a significant relationship is found between lead and the

change on incidents of property crime net of time-stable variables. However, when other

sociological variables are included in the model, the association between lead and

property crime incidents is no longer found to be significant. Because prior findings

substantiate a relationship between lead, violence, and aggression, it was hypothesized

that lead would emerge as a more consistent predictor of violent crime than property

Table 18. Summary of Multivariate Findings

Total Crime

Property Crime

Violent Crime

Variable

2004

2010

Change Adult*

Juvenile*

2004

2010 Change Adult*

Juvenile*

2004

2010

Change Adult*

Juvenile*

Lead

Lag

Disadv.

Mobility

Heterog.

ln%Black

%Hisp

%Divrcd.

X = p <.05 * = 2008 Arrest Rate -- = Not Tested

152

crime. Overall, observations with respect to crime incidents were consistent with this

hypothesis.

These associations between lead and type of offense did not, however, emerge in

the examination of crime arrests. Overall, the relationship between lead and crime

arrests appears similar across arrests for total crime, property crime, and violent crime.

That is, while significant changes were observed between lead and type of offense when

the dependent variable was crime incidents, the same relationships were not observed

when the dependent variable was crime arrests. Thus, these data suggest that lead

impacts arrests for property crimes and violent crimes similarly. One reason the data may

show this may be that counts for reported crime arrests were lower overall than counts of

reported crime incidents. As such, the heterogeneity which existed in the crime incident

data may not have been as pronounced in the crime arrest data. Consider, for example,

these data suggest that in 2008 there was an average of approximately 59 crime incidents

(per 1000 persons) reported, as opposed to approximately 12 arrests (per 1000) persons

reported.

Relatedly, the spatial lag of lead was not found to exert a significant impact on

arrest rates. Across total, violent, and property arrests, the association between the spatial

lag of lead and arrests was found to be null. This also may be attributable to the

comparatively low frequency of arrests. On the other hand, the spatial lag of lead does

appear to exert an impact on crime incident rates. The spatial lag of lead exerts a

significant, positive impact on incidents of total crime, violent crime, and property crime

but only in 2004. The spatial lag of lead does not appear to exert significant influence on

162

153

crime incident rates at all in 2010. To better understand this discrepancy, the distribution

of lead in 1999 and 2005 were mapped and compared. These maps are presented in

Figures 8 and 9, respectively. What these maps suggest are changes with respect to the

amount and distribution of lead from 1999 to 2005. Specifically, changes emerge which

may impact the amount of lead present in the calculation of the weighted average of lead

in communities surrounding areas with the highest levels of lead. For example, the

Austin community (#25) carries a lead presence greater than 1.5 (but less than 2.5)

standard deviations above the mean in both 1999 and 2005. However, lead levels in both

communities bordering Austin on the north (#18 Montclair and #19 Belmont Cragin)

“increased” from 1999 to 2005 with respect to distance from the mean. That is, in 1999

these communities were more than .50 standard deviations below the mean, while in

2005 they were within a half a standard deviation of the mean. Similarly, lead levels in

Austin-bordering West Garfield Park (#26), appeared to move farther above the mean,

while lead levels in the contiguous East Garfield Park (#27) appeared to decrease towards

the mean. Likewise, in 1999 both West (#26) and East (#27) Garfield Park carried lead

levels greater than 1.5 standard deviations from the mean. However, lead levels in North

Lawndale (#29) decreased from 1.76 standard deviations above the mean in 1999 to 1.39

standard deviations above the mean in 2005. Together, these changes suggest that

LEFT: Figure 8. Distribution of Lead (1999) Across Chicago Community Areas (n=76)

RIGHT: Figure 9. Distribution of Lead (2005) Across Chicago Community Areas (n=76)

155

165

the impact of the spatial lag of lead on communities may be attributable to changes in

lead levels that impact the calculation of the spatial lag variable.

Analyses examining the association between lead, age, and crime arrest rate

revealed unexpected nuances between lead and arrest rates. Specifically, it was

anticipated that lead would exert more of an impact on juvenile arrest rates than adult

arrest rates. This was not observed. Overall, the relationship between arrests for total

juvenile index crimes, juvenile property crimes, and juvenile violent crimes consistently

emerged as outside the range of significance. Conversely, the association between

arrests for total adult index crimes, property crimes, and violent crimes was consistently

found to be positive and significant (p<.01). What these results seem to imply is that the

impact of lead on adult crime is positive and consistent, while lead bears no significant

impact on juvenile arrests. One possible explanation for this may be that the threshold

for lead’s impact on crime arrest rates may be longer than anticipated. This would be

consistent with prior literature which finds early childhood lead exposure impacts adult

criminality

(Nevin, 2007; Wright et al., 2008). Alternatively, it may be plausible that lead’s

relationship with juvenile behavior simply does not manifest in the form of criminal

behaviors (but, for contradiction, see Needleman et al., 1996 & Olympio, 2010). It

would appear that additional studies are necessary to further specify the age-graded

effects of lead on crime. In any event, with respect to a five year lag, age of offender

does seem to influence in the relationship between lead and crime arrest, albeit not the

type of influence that was anticipated.

166

It is also possible that the lead-crime association measured here as a function of

arrest also plays a role in impacting the outcomes. That is to say, since arrests represent

some dimension of police discretion, when used as a measure of crime, arrests will tend

to under-represent the total universe of crime. Moreover, it is possible that some aspect

of police discretion is distributed in ways that reduce the relationship between lead and

crime when measured by arrest. Thus, additional research on this issue is desirable, and

is needed to address other aspects of the lead-crime relationship. Additional studies, for

example, could include alternative indicators of crime, if available, such as reported

crimes, or as some studies suggest, calls for police services.

In sum, what multivariate results imply (with respect to lead and crime), are that

the relationship between lead and crime appears sensitive to measurement of crime, and

subsequently, type of crime, and age of offender. Specifically, these data suggest that

lead’s impact on crime incident rates is not directly comparable to lead’s impact on crime

arrest rates. Data suggest that this is true with respect to community level of lead as well

as the spatial lag of lead. The association between lead and crime incident rate appears

sensitive to type of offense. Specifically, lead was observed to be a more consistent

predictor of violent crime than property crime. On the other hand, the association

between lead and crime arrests appears robust to type of offense. That is, lead appears to

impact crime arrests similarly; regardless of if the arrest is for a property offense or

violent offense. The association between lead and crime arrest does, however, appear

sensitive to age of offender. Unexpectedly, lead emerges as a more consistent predictor

of adult arrest rates than juvenile arrest rates. All significant cross sectional associations

167

emerged net of confounding sociological variables. All significant longitudinal

associations emerged net of unobserved time-invariant factors.

Environmental justice issues. The last research question posed in this study

asks, does the distribution of lead pose environmental justice concerns with respect to

race, class, and ethnicity in Chicago communities? To answer this question, an

exploratory analysis examined the bivariate associations between community lead level

and community: percentage of black residents, percentage of Hispanic residents, score on

the disadvantage index, interactions between race and class, and interactions between

ethnicity and class. Overall, results suggest positive, significant relationships between

lead and: percent black, disadvantage, and both interaction terms. What this suggests is

that there does appear to be covariance between the distribution of lead and the

distribution of race and class across Chicago communities. That is, as the percentage of

black residents, percentage of economically disadvantaged residents, or the percentage of

residents who are both black and economically disadvantaged increases, corresponding

increases in lead can also be anticipated.

This observation is consistent with extant findings in the environmental justice

literature (Brulle & Pellow, 2006; Bullard, 1994; 1996; Checker, 2007; Krieg, 2005;

Opp,

2012). Moreover, these results suggest a relationship between political

disenfranchisement and the distribution of lead. Under-resourced and disenfranchised

communities may not only bear a higher presence of lead from industrial sources and

expressways, but also may lack the capital to promote lead abatement or other lead clean

168

up initiatives. Residents of these communities may also face additional hurdles with

respect to medical treatment and interventions. Additionally, these communities may

lack the resources to compel politicians and other policy makers to address

environmental justice concerns in their communities. These results suggest lead

exposure adds to a host of public health issues confronted by disenfranchised

communities.

Contrary to what was hypothesized, observations with respect to percent

Hispanic and the distribution of lead were not consistent with findings from prior studies

(for example, see Diawara et al., 2006; Jacobs et al., 2002). It was anticipated that lead

and percent Hispanic would positively co-vary. However, the association between lead

and percent Hispanic did not emerge as significant. Moreover, the association between

percent Hispanic and the disadvantage index also failed to emerge as significant. Thus,

while the interaction term between percent Hispanic and disadvantage did emerge as

having a relatively weak, significant relationship, it is highly likely that this was solely

the influence of disadvantage. What these data suggest is that Chicago Community

Areas that concentrate the highest values of disadvantage are not necessarily the same

communities that have the highest concentrations of Hispanic residents. This

observation may set this study apart from environmental justice research that examines

locations where ethnicity, disadvantage, and lead positively co-vary.

Taken together, these latter findings from exploratory environmental justice

analyses suggest a need for future study with respect to the distribution of lead, class, and

race in Chicago. Additional research appears justified especially in regards to the

169

distribution of lead and percent black and economic disadvantage. Previous literature

also has called attention to this matter. For example, Moore (2003) observes: “African

American children in low-income families in the United States have the highest exposure

to lead of any segment of the American population,” (pg. 13). In 2001, the National

Association for the Advancement of Colored People (NAACP) also called attention to

the disparate exposure of minority youth to lead, especially leaded paint in older homes

(Moore, 2003). Findings from this study suggest evidence of environmental racism and

classism across Chicago’s community areas. Given that these findings come from data

seven years after Moore’s observation, and over ten years following the NAACP’s call

for action, it would appear that lead exposure still may be of particular concern for

communities with high concentrations of minority residents.

Study Limitations

This study is not without limitation. One way in which this study may be limited,

is the reliance on official data (incident and arrest counts from the Chicago Police

Department) to measure crime. Use of official data to measure crime has been critiqued

by criminologists because many forms of crime go unreported and/or undetected by

police. Some studies suggest significant differences in reports of crime when self-report

data are compared to official data, with respect to the same sample (Farrington et al.,

2003; Kirk, 2006). To be sure, data in this study suggest differences in predictors when

measurement of the dependent variable changes from counts of crime incidents to counts

of crime arrests. Relatedly, given that juveniles may be less likely than adults to be

arrested, official arrest data may underestimate the “true” amount of juvenile crime.

170

However, with respect to this study, there are several benefits to using official data. First,

availability of data over multiple years allowed for both cross sectional and longitudinal

analyses. While offender demographics (e.g., age) escape measurement in crime incident

data (in some cases, this information is unknown, or potentially inaccurate), the use of

arrest data allowed this study to consider juvenile and adult arrest rates. Moreover, that

lead demonstrates associations with incidents of violent crime, and arrests of adult crime,

adds confidence to the conclusion that lead matters with respect to community crime

rates. Interestingly, a prior study’s comparison of self reported crime and official arrest

data in Chicago, suggests that while the counts of self report and arrest data differ

significantly, the relative impact of predictors of crime across the two data types did not

(Kirk, 2006). This implies that even if crime data on the Chicago Community Areas had

been obtained via self report (instead of via official records), at least one prior study

using data from the same city suggests that the outcome of the analyses may have stayed

the same.

Another limitation of this study is in the measurement of community via

boundaries for the Chicago Community Areas. The average area for a Chicago

Community Area is approximately 3 square miles, and each Chicago Community Area

concentrates an average of approximately 38,000 people. Relative to other studies of

community crime rates, this may be perceived as quite large, and potentially “too big” to

capture neighborhood effects. For example, Sampson and colleagues’ Project on Human

Development in Chicago Neighborhoods (PHDCN) aggregates Chicago census tracts

into approximately 300 neighborhood clusters (compared to 77 community areas).

171

However, a number of reasons make the Chicago Community Areas a meaningful unit of

analysis for this study. One of which is that the Chicago Community Areas represent a

smaller unit of analysis than has been used in prior ecological studies of lead and crime;

the Chicago community areas are smaller units than, for example, counties or cities.

Moreover, the Chicago Community Areas are identifiable to residents, politicians,

community leaders, and government entities (Sampson, 2012). The Chicago Police

Department states in annual reports that, “the Community Areas remain the most widely

used geographic units by Chicago planning agencies, advocacy groups, and service

providers,” (2009, p. 34). Finally, the Chicago Community Areas have been relied upon

in prior criminological analyses (Block & Block, 1993; Lyons, 2006; Sampson, 2012;

Visher & Farrell, 2005).

A closely related limitation of this study is that data is drawn from one

geographical region (the city of Chicago). This may limit the generalizability of findings

to communities in other locations. However, Chicago’s similarities to other northeastern

cities make it particularly well suited for studies of urban ecology. That said, a

replication of this study using communities within other cities represents an important

direction for future research. Likewise, data that captures annual neighborhood change

(such data was not available for this sample) as opposed to strictly on census years, may

provide for the ability to capture lead’s influence while controlling for dynamic

neighborhood change.

Finally, the measurement of lead introduces additional limitations. Specifically,

the percentage of children in a community with elevated blood lead levels only

172

represents children who have blood lead levels above 10 µg/dL. As such, children who

tested with blood lead levels greater than 0 µg/dL but less than or equal to 10 µg/dL are

not included.

This is problematic with respect to the research question, because the amount of lead in

each community may actually be underestimated. That is, 10 µg/dL creates a “floor” that

may have a censoring effect on the actual amount of lead present in a community.

However, there are reasons this indicator for community lead presence remains

appealing. One of which is that the amount of lead in community resident’s blood allows

for the fact that there are multiple sources of lead in a given community (e.g., soil, air,

water). The use of one structural characteristic (e.g., houses built prior to 1950) fails to

capture other community lead sources (e.g., lead in soil, lead in air from gasoline, leaded

paint on playground structures, lead in new pipes). Moreover, because of the use of a

threshold which is deemed by many researchers to be too high, this makes the indicator a

conservative estimate. Alternatively stated, the relationship between lead and crime may

actually be more consistent than results suggest.

Implications for Theory

These results pose implications for future theorizing about communities and

crime. Specifically, it was argued that the omission of lead from mainstream crime

theory may be problematic, given leads demonstrated association with crime rates at

other macro levels of assessment (Masters et al., 1994; Nevin, 2007; Reyes, 2007;

Stretesky & Lynch, 2004). Given social disorganization theory’s enduring popularity in

orthodox criminology (Kubrin & Weitzer, 2003; Pratt, 2005) it becomes especially

173

prudent to examine any hypothesized macro-level predictor of crime concurrently with

social disorganization variables. As such, the community impact of lead on crime was

examined while holding constant disadvantage, racial/ethnic heterogeneity, and

residential mobility in twelve different negative binomial regressions. These regressions

captured three different time frames, and multiple dependent variables. Multivariate

results in particular bear relevance for future tests of social disorganization theory.

Overall, predictors of social disorganization theory performed relatively

consistently across crime incidents as well as crime arrests. Additionally, social

disorganization predictors performed relatively consistently across juvenile arrests and

adult arrests, as well as incidents of property crime and violent crime. It was expected

that social disorganization theory would perform consistently across all models, as social

disorganization theory is designed to be a general theory of crime. That is, social

disorganization theory proposes to explain community crime, regardless of type of

offense or age of offender.

However, there were nuances with respect to disadvantage, heterogeneity, and

mobility’s respective associations with community crime. For instance, while

disadvantage emerged as significant in nine out of twelve models, in all nine instances,

the association with crime was inverse. That is, an increase in disadvantage was found to

be significantly associated with decreases in crime. This appeared true of all property

crime measures (incidents, arrests, adult, and juvenile). The relationship was also

inverse across all instances of total crime. Some differences were observed with respect

to violent crime, however. For example, disadvantage was not found to be significantly

174

associated with violent crime incident rates in 2004 or 2010. Nor was disadvantage

found to be significantly associated with adult arrests for violent crime. Although,

disadvantage did emerge as significantly associated with arrests for juvenile violent

crime.

While the direction of these relationships was unexpected, there are a few

possible explanations for the inverse relationships. One of which may be the different

distributions of crime with respect to violent crime versus property crime. While this has

been observed in other hotspot studies, this observation is conflicting with social

disorganization theory, which presents as a general theory. Given that the highest

proportion of property crime incidents were reported in the Loop community (central

downtown area), perhaps this suggests that Loop concentrates more “suitable targets,”

(Cohen & Felson, 1977). Another reason for the disparity may be that perhaps incidents

are more likely to be reported in areas where social organization is high. That is, these

observations may be a function of reliance on official data.

With respect to racial/ethnic heterogeneity and residential mobility, these

associations also appeared to have a more nuanced relationship with crime. Interestingly,

racial/ethnic heterogeneity did not emerge as significantly related to crime in any model.

That is, the relationship between community racial/ethnic heterogeneity and crime

emerges as null each time. While these results were contrary to what was hypothesized,

these findings are somewhat consistent with the mixed impact of racial and ethnic

heterogeneity on crime that recent tests of social disorganization are finding (Martinez,

Stowell, & Lee, 2010).

175

In contrast to both heterogeneity and disadvantage, residential mobility

performed very consistently, and in the hypothesized direction. Residential mobility

emerged as significantly and positively associated with crime in every model, with the

exception the examination of violent adult arrest rates. Given mobility’s robustness

across most dependent variables used in this study, it appears residential mobility is an

important component to consider in community crime rates.

Moving forward, given that lead was found to bear significant associations with

crime, net of pertinent sociological variables, these results suggest that lead be

considered in future theoretical models. Additionally, researchers may be interested in

testing integrated models of lead and crime, centering around testing multiple pathways

between lead exposure, relevant sociological predictors, and crime. Such models are

presented by Narag et al. in 2009. Moreover, this study illuminates the role that green

macro level predictors of crime in estimation procedures. That is, criminology may

continue to look towards relevant environmental information with respect to the

distribution of crime. The integration of mainstream criminology theories with green

perspectives is emergent, and can be observed with respect to other theories and other

criminal outcomes (for example, Agnew, 2012).

Implications for Policy

These results also bear importance for community crime policy initiatives.

Chicago in particular seems to host a number of policy initiatives geared towards crime

prevention and reduction (i.e., Chicago Area Projects). One such initiative that has

recently garnered attention is the work of the Violence Interrupters. The Violence

176

Interrupters initiative is an operation of the Chicago Project for Violence Prevention

(formerly CeaseFire Chicago), and the program identifies as a community crime

prevention and violence prevention program. The Violence Interrupters are a team of

interventionists committed to responding personally to instances of violence in Chicago

communities. The Violence Interrupters aim to change community norms and attitudes

about violence through public outreach and education. The Violence Interrupters operate

across Chicago and target their efforts in communities experiencing high levels of

violence. These communities include communities identified by this study as high

violence communities- for example, Englewood and West Garfield Park. The Violence

Interrupters Initiative is identified as a promising program initiative by the Office of

Justice Programs, has been associated with reductions in shots fired and gang

involvement across micro-geographical units (Skogan, Hartnett, Bump, & DuBois,

2008), and recently received a $1 million grant from the Mayor of Chicago to continue

its efforts

(Benzine, 2012).

While the efforts of the Violence Interrupters, for example, are associated

empirically with reductions in shots fired, gang participation, and are undoubtedly

amicable, results from this study imply that the endeavors of the Violence Interrupters

(and similar programs) may be enhanced by policy initiatives that also address

community lead levels. Given that findings in this study reveal lead an especially

consistent predictor of violence crime incidents (net of confounders), violence prevention

initiatives that address the clean up of lead concurrently with initiatives like the Violence

177

Interrupters, may witness more profound reductions in violence, or other enhanced

benefits. Likewise, policy endeavors that criminalize the inclusion of lead in products

may also serve to add additional public health benefits. Findings from this study suggest

that reductions in community rates of violent crime may be an example of such a benefit.

Directions for Future Study

The findings from this study present several important directions for future study.

For example, the mixed findings with respect to the distribution of incidents of violent

crime versus incidents of property crime invite additional study. For example,

criminologists may wish to examine these distributions further, perhaps using additional

mainstream criminological frameworks. Frameworks may include examining the impact

of relative deprivation, routine activities, or strain perspectives and their association with

the distribution of crime, net of lead. Future examinations of community crime may

wish to incorporate lead into the theoretical framework. Moreover, data that can capture

the organization of the community may allow for a more complete test of social

disorganization theory.

Future studies may also wish to examine a mediating or moderating association

between lead, social disorganization, and crime. This may be accomplished, for

example, through the use of structural equation models to more specifically explicate

pathways from community lead presence, social disorganization, and crime. Narag et al.

(2009) introduce a number of possibilities with respect to theoretical integration. Results

from this study suggest that these integrated pathways may be well worth exploring.

178

Additionally, future studies may wish to replicate this study in different areas, or

different measures of crime. For example, if the association between lead and violent

crime incidents and/or adult arrest rates were found to be persistent when measuring

crime with self report data, or through the use of qualitative data, this would strengthen

the argument that lead should be integrated into theory and policy initiatives.

Conclusion

The objective of this study was to address three interrelated research questions: Is

there a relationship between hotspots of lead and hotspots of crime? Stemming from this

question, this study also asked, what is the nature and extent of the relationship between

lead and crime? Lastly, does the distribution of lead pose environmental justice concerns

with respect to race, class, and ethnicity in Chicago communities?

Chapter one introduced the aims of the study, and provided an organizational

overview of the proceeding chapters. Chapters two and three reviewed relevant literature

on distributions of lead and crime, and argue why a study that examines the geographic

overlay of lead and crime is needed. Chapter four provided background on the study

setting, Chicago, Illinois- an area with a documented presence of both lead and crime.

Chapter five provided details on the research methodology, and presented the translation

of the three major research questions into thirteen testable hypotheses. These hypotheses

were tested subsequently in chapter six, and results were discussed in previous sections

of this chapter.

Is there a relationship between hotspots of lead and hotspots of crime? The

answer to this question appears to be dependent on type of offense. When measuring

179

crime incidents, there appears to be an observable overlap between the distribution of

lead and violent crime across communities. The overlay appears less consistent when

measuring property crime incidents, or total crime incidents (the latter of which is largely

influenced by property crime).

What is the nature and extent of the relationship between lead and crime? These

data suggest that a positive, significant relationship between lead and incidents of violent

crime that endures even while controlling for mainstream predictors of community crime

rates. These data also suggest that when the weighted average of lead in neighboring

communities is high enough, the lead in neighboring communities may impact incidents

of crime (property and violent) within a community. Results also suggest that changes of

lead level within a community are associated with changes in crime rates (total, violent,

and property), while holding constant unobserved, time-stable attributes. However, these

associations are most consistent with respect to incidents of violent crime. Finally,

increases in lead are associated with significant increases in adult arrest rates (regardless

of offense type). These data do not find evidence that lead is significantly associated

with juvenile arrest rates.

Does the distribution of lead pose environmental justice concerns with respect to

race, class, and ethnicity in Chicago communities? These data suggest that the

distribution of lead does seem to positively co-vary with race and class. Specifically,

communities with high concentrations of black residents, economically disadvantaged

residents, and concentrations of residents that are both black and economically

disadvantaged, appear to carry higher levels of lead. Associations between percent

180

Hispanic and lead, however, did not emerge as significant. Thus, these data imply

environmental concerns with respect to race and class, but not with respect to ethnicity as

measured by percent Hispanic.

Given that under certain conditions, the relationship between lead and crime

persists net of confounders, this study offers criminologists examining communities and

crime a new variable to contend with when theorizing about community crime rates.

Additionally, the empirical associations between lead and violent crime especially

provide policy makers with a relevant issue to address when confronting community

crime and violence. Future studies will be necessary in elucidating where lead falls in a

theoretical pathway to community crime. Future studies may also help inform the role

lead reduction/removal may play in policy initiatives that address community crime,

especially violent crime.

181

Students also viewed