Putting homicide followed by suicide in context: Do macro‐environmental characteristics impact the odds of committing suicide after homicide?*

• Published date: 01 February 2019
• Author: Emma E. Fridel,Gregory M. Zimmerman
• DOI: http://doi.org/10.1111/1745-9125.12195
• Date: 01 February 2019

Received: 7 February 2018 Revised: 20 July 2018 Accepted: 23 July2018

DOI: 10.1111/1745-9125.12195

ARTICLE

Putting homicide followed by suicide in context:

Do macro-environmental characteristics impact

the odds of committing suicide after homicide?*

Emma E. Fridel Gregory M. Zimmerman

School of Criminology and Criminal Justice, Northeastern University

Correspondence

Gregory M. Zimmerman, PhD, Graduate

ProgramDirector and Associate Professor,

Schoolof Criminology and Cr iminal Justice,

Northeastern University,431 Churchill Hall,

360Huntington Avenue, Boston, MA 02115.

Email:g.zimmer man@northeastern.edu

∗Additionalsupporting information can be

foundin the listing for this article in the Wiley

OnlineLibrar y at https://onlinelibrary.wiley.

com/doi/10.1111/crim.2019.57.issue-1/

issuetoc.

Abstract

Homicide followed by suicide remains an understudied

phenomenon in the criminological literature. This is due, in

part, to methodological and statistical limitations—much of

the extant research includes small samples and has not kept

pace with quantitative advances. Moreover, scholarship on

homicide–suicide has been focused almost exclusively on

individual risk factors, discounting contextual influences.

In this study, we examine whether macro-environmental

characteristics affect the odds of suicide after a homicide.

We use data on 24,373 homicide and homicide–suicide

cases distributed across 3,019 cities and 48 U.S. states from

the National Violent Death Reporting System to examine

the direct effects of structural factors on the odds of suicide

after a homicide; and whether structural characteristics

condition the impact of the victim–offender relationship

on the odds of homicide–suicide. Hierarchical logistic

regression models indicate that macro-level concentrated

disadvantage decreases the odds of homicide–suicide.

Furthermore, concentrated disadvantage attenuates the

odds of suicide after the homicide of an intimate partner,

child, family member, or friend, relative to the killing

of a stranger. The findings reveal that researchers should

account for the context in which homicide–suicide occurs;

failure to do so may unintentionallydiscount a key correlate

34 © 2018 American Society of Criminology wileyonlinelibrary.com/journal/crim Criminology. 2019;57:34–73.

FRIDEL AND ZIMMERMAN 35

of homicide–suicide and artificially inflate the effects of

the micro-environment.

KEYWORDS

concentrated disadvantage, homicide, homicide–suicide, neighborhood

effects

1INTRODUCTION

Compared with its component acts, homicide followed by suicide is an understudied phenomenon in

the criminological literature. This is due, in part, to its rarity—homicide–suicide has an estimated rate

of .22 per 100,000 U.S. persons, or 4.0 percent of all annual homicides and 1.5 percent of all annual

suicides (Large, Smith, & Nielssen, 2009; Logan et al., 2008), which are rare events themselves. The

infrequency with which homicide–suicide occurs poses several challenges.First, until now, researchers

have mainly discounted the utility of studying homicide–suicide as a distinct behavior, instead regard-

ing it as a variant of either homicide or suicide and studying it accordingly (Liem & Nieuwbeerta,

2010), despite the knowledge that homicide–suicide has a devastating impact on familiesand commu-

nities (Liem, 2010), garners national attention that can be used to shape the sociopolitical landscape

(McPhedran et al., 2015), and inspires copycat crimes (Towers, Gomez-Lievano, Khan, Mubayi, &

Castillo-Chavez, 2015). Second, and relatedly, many studies on homicide–suicide have methodological

and statistical shortcomings, including small sample sizes, low statistical power, inflated effect sizes,

overestimation of both type I and type II errors, inefficiency of model estimates, and low reproducibil-

ity of results (Button et al., 2013). Third, homicide–suicide has no formal legal definition (because

charges are not typically filed) and weak theoretical grounding on which few empirical tests are based

(Liem, 2010; Riedel, 2010). As a result, much of the extant research on homicide followedby suicide is

epidemiological in nature; has been focused on descriptions of isolated events rather than on scientific

assessment of larger sampling frames; has examined whether homicide–suicide is primarily suicide

driven, primarily homicide driven, or a different form of violence entirely; and has produced mixed

results (Eliason, 2009; Flynn et al., 2009; Liem, Postulart, & Nieuwbeerta, 2009; McPhedran et al.,

2015; Panczak et al., 2013; Reckdenwald & Simone, 2017).

Moreover, empirical studies on homicide–suicide have been conducted almost exclusively at the

individual level—theory and literature as to how the macro-sociological context impacts homicide

followed by suicide are sparse. Yet, scholars have long recognized that one avenue to advance crim-

inology as a discipline is macro–micro integration. For example, Parsons (1937), Weber (1922), and

later Coleman (1986) argued for methodological individualism (in contrast to methodological holism),

under which macro-level changesare g roundedin individual action. Relatedly, micro-level and macro-

level theories alone have historically been unable to explainmore t han 20 percent of variation in crim-

inal offending (Elliott, 1985; Muftic, 2009). Accordingly, recent calls to bridge the epistemological

divide between micro- and macro-criminology (Matsueda, 2017; Messner, 2012; Rosenfeld, 2011), as

well as laudable efforts at multilevel theorizing (Wikström, 2010), have been made. Concurrently, the

neighborhood effects literature (Sampson, 2012) and research on the person-context nexus (Baumer &

Arnio, 2012) have included multilevel statistical analysisas a way to examine the inextricable linkages

between individuals and their social contexts.

Guided by the neighborhood effects literature and research on the socio-structural correlates of

homicide (Land, McCall, & Cohen, 1990; Sampson, Morenoff, & Gannon-Rowley, 2002; Sharkey and

36 FRIDEL AND ZIMMERMAN

Faber, 2014), we examine the mechanisms through whichthe broader social context impacts homicide

followed by suicide. Our analysis is guided by two primary research questions. First, we investigate

whether macro-level concentrated disadvantage has a direct effect on the likelihood of suicide after

homicide. Grounded primarily in attribution theory (Heider, 1958), we hypothesize that pervasive vio-

lence in disadvantaged areas facilitates the external attribution of blame and desensitizes residents to

violence, thereby decreasing the guilt associated with homicide and reducing the odds of subsequent

suicide.

Second, we focus on the relevanceof the social context for victim–offender dynamics—arguably the

most prominent differentiating factor between homicide and homicide–suicide (Allen, 1983; Palmer &

Humphrey,1980; Selkin, 1976; Stack, 1997). Specifically, we examine whether the macro-environment

attenuates the salience of the victim–offender relationship for committing suicide after homicide. We

posit that the internal attribution of guilt associated with killing an intimate partner, family member,

or associate (relative to a stranger) will be reduced in disadvantaged areas with frequent violence—

where residents may be desensitized to violence, justify violence, and even tolerateviolence (Sampson

& Bartusch, 1998)—thereby reducing the odds of committing suicide after the homicide of a close

associate.

We thus contribute to the literature by formulating and testing hypotheses that link the social con-

text to the individual act of homicide–suicide, grounded in theories situated at both the micro and

macro levels. Additionally, we examine our hypotheses using data from a large sample of homicide

(N=22,960) and homicide–suicide (N=1,413) cases from the National Violent Death Reporting Sys-

tem (NVDRS), a state-based active Websurveillance system of death records established by the Centers

for Disease Control and Prevention (CDC). Our sample size accommodates a test of several types of

intimate partners (spouses, ex-spouses, and partners), family members (children, parents, and other

family members), associates (friends, acquaintances, and other known persons), and strangers. Finally,

we examine contextual effects and cross-level interaction effects using three-level logistic regression

models to nest the 24,373 homicide and homicide–suicide sample cases within 3,019 cities distributed

across 48 contiguous U.S. states. Our narrative begins with a brief review of the individual-level cor-

relates of homicide followed by suicide beforetur ning to the relevance of the macro-social context for

homicide and homicide–suicide.

2CONCEPTUAL BACKGROUND

2.1 Individual-level correlates of homicide–suicide

Scholarship on the etiology of homicide–suicide has been focused almost exclusively on individual

differences and micro-environmental characteristics. In the extant literature, scholars have indicated

that homicide–suicide perpetrators are more likely than homicide perpetrators and suicide victims to

be male (Panczak et al., 2013), White (Logan et al., 2008; Wolfgang, 1958), adult (as compared with

adolescent for “homicide only” and adolescent or elderly for “suicide only”; Banks, Crandall, Sklar,

& Bauer, 2008; Liem et al., 2009), and married or separated (relative to single; Panczak et al., 2013;

Rosenbaum, 1990). In addition, homicide–suicide disproportionately involves firearms: The odds of

firearm usage are approximately 5 times greater and 11 times greater among homicide–suicides, respec-

tively,than among homicides and suicides (Carcach & Grabosky, 1998; Panczak et al., 2013). Scholars

also generally haveagreed that homicide–suicide shares a common set of risk factors with intimate part-

ner homicide and family homicide (Haines, Williams, & Lester, 2010; Heron, 2017; Manning, 2014;

Salari & Sillito, 2016). Most notably, these acts typically involve a history of interpersonal conflict