Examining the Relationship Between Weapon Type and Relationship Type in American Homicides: A Bayesian Approach

• Author: Joshua J. Reynolds,Paul C. Bürkner
• DOI: 10.1177/1088767920976191
• Date: 01 November 2021
• Published date: 01 November 2021
• Subject Matter: Articles

https://doi.org/10.1177/1088767920976191

Homicide Studies

2021, Vol. 25(4) 295 –312

© 2020 SAGE Publications

Article reuse guidelines:

sagepub.com/journals-permissions

DOI: 10.1177/1088767920976191

journals.sagepub.com/home/hsx

Article

Examining the Relationship

Between Weapon Type

and Relationship Type in

American Homicides:

A Bayesian Approach

Joshua J. Reynolds1 and Paul C. Bürkner2

Abstract

The weapon type used in a homicide predicts the victim-perpetrator relationship.

However, there are some limitations in this past research including the common data

analytic strategies. Our purpose was to build a model of weapon type, predicting

relationship type, and to address previous limitations. We examined 363,927

homicides and used Bayesian multilevel categorical regression. In addition to analyzing

weapon type (final model consisted of 16 weapon categories), we examined the

victims’ sex, age, and race as covariates and modeled the data across states and

counties. Results indicate that weapon type is highly informative, however, the age of

the victim and sex of the victim interact in important ways.

Keywords

homicide, weapons, Bayesian statistics, multilevel models, categorical regression

Introduction

Predicting what kind of relationship a homicide victim and perpetrator have, using

details from the crime scene, is a practical question that concerns all homicide inves-

tigators. For example, if a detective knows the suspect is likely to be an acquaintance,

then that could help to steer the time sensitive investigation, by prioritizing investigat-

ing a victim’s neighbors, friends, employer, etc. One variable that has received some

consideration in understanding victim- perpetrator relationship, is the type of weapon

1University of Scranton, Scranton, PA, USA

2Independent Junior Research Group Leader, Cluster of Excellence SimTech, University of Stuttgart,

Germany

Corresponding Author:

Joshua J. Reynolds, Stockton University/University of Scranton, 101 Vera King Farris Drive, Galloway,

NJ 08205-9441, USA.

Email: josh.j.ren@gmail.com

976191HSXXXX10.1177/1088767920976191Homicide StudiesReynolds and Bürkner

research-article2020

296 Homicide Studies 25(4)

used in the h omicide. If a homicide detective could increase the chances of correctly

identifying the suspect by knowing the strength of the relationship between weapon

type and relationship type, then this information is potentially valuable. Further, it

seems that homicide detectives consider a variety of crime scene evidence when trying

to predict who is the perpetrator. If this is helpful or harmful, partly depends on the

veracity of the relationship. Past research on this topic exists, but some of the inferences

that can be drawn are limited due size of the homicide database, the necessity of col-

lapsing across categories (e.g., combining different weapons types into one group), and

choice of statistical procedure. This research analyzes over three decades of homicides

from the Federal Bureau of Investigation’s Supplementary Homicide Reports (SHR)

and homicides not reported to the FBI, but obtained through the Freedom of Information

Act, by the Murder Accountability Project (Hargrove, 2019). In this research, Bayesian

statistics are applied to model the relationship between weapon type and relationship

type and how it may change due to a victims age, sex, and race. By analyzing a larger

set of cases, including more weapon types, and applying a more advanced statistical

approach, this research makes a substantial contribution by allowing us to build a pre -

dictive model and produce parameter estimates that are more intuitive.

Victim–Perpetrator Relationship

There is no single scheme for categorizing the victim-perpetrator relationship, but

regardless of scheme, there is typically a distinct pattern within homicides. According

to the 2017 data from the National Incident based Reporting System (NIBRS), the

most common relationship type (38%) is, known to victim and other, which means that

the victim and perpetrator knew each other, but were not related. Additional categories

include family members (15%), family members and other (1%), stranger (10%), and

all other (37%). In the Supplementary Homicide Report there are 28 relationship cat-

egories, with three main groups, family member, acquaintance outside the family, and

victim not known to the offender. However, there are a variety of ways researchers

collapse or combine categories. For example, in describing homicide trends in the US,

Cooper and Smith (2011) used multiple categorization schemes including stranger,

spouse, other family, boyfriend/girlfriend, and other acquaintance. Although the

majority of cases did not have a known victim-perpetrator relationship, among those

that did, approximately half were cases in which the perpetrator was an acquaintance.

Similarly, Morgan and Kratcoski (1986) found that the majority of homicide victims

were killed by someone who was not a stranger, although the proportion of stranger to

non-stranger killings does fluctuate over time. Just as there are a variety of victim-

offender relationship types and categorization schemes, so too are there a variety of

weapons used in homicides.

Weapon Type

Data from NIBRS (United States Federal Bureau of Investigation, 2017a 2017b) has

five groups of weapons (for homicide offenses), firearms, knife/cutting instrument,