NEWSWORTHINESS OF INTIMATE PARTNER HOMICIDES: AN EXAMINATION OF RACE AND GENDER By Carissa M. Sorenson A THESIS Submitted to Michigan State University in partial fulfillment of the requirements for the degree of Criminal Justice – Master of Science 2020 NEWSWORTHINESS OF INTIMATE PARTNER HOMICIDES: AN EXAMINATION OF ABSTRACT RACE AND GENDER By Carissa M. Sorenson Newsmakers rely on common stereotypes and existing belief systems when determining which criminal instances receives news coverage (Gilliam & Iyengar, 2000; Gruenewald et al., 2013; Entman, 1992, Fishman, 1980; Freng, 2007). Previous research that examines the newsworthiness of crime have found that individual characteristics of the crime victim, individual characteristics of the crime perpetrator, and circumstantial characteristics of the crime are related to whether a criminal instance receives any news coverage and the amount of news coverage received. The present study builds upon previous research by examining the newsworthiness of a sample of intimate partners homicides, some of which involved Native American victims and perpetrators. Using the Montana Domestic Violence Fatality Review Commission 2017 report, this study analyzed news media coverage by four Montana news sources of intimate partner homicides occurring in the state between the years of 2000 to 2016. Findings indicate that circumstantial characteristics of the homicide incidents explained a greater amount of variance in the amount of news coverage the incidents received compared to crime victim and perpetrator characteristics. Dedicated to my guardian angels: Judy Rae Sorenson, Romona Jean Ryder, and Shirley Jean Quam. Thank you for teaching me the importance of compassion and grace. I love you. I miss you. iii ACKNOWLEDGEMENTS Thank you Dr. Merry Morash, Dr. Jennifer Cobbina, and Dr. Steven Chermak for sharing your wealth of knowledge and for your guidance. A huge thanks also goes to my family and friends for their steadfast encouragement. iv LIST OF TABLES.……………………………………………………………………………….vi TABLE OF CONTENTS INTRODUCTION.….……………………………………………………………………………..1 LITERATURE REVIEW………………..…………………....…………………………….……. ..3 A Description of News Production…………...……………………………………….…...3 The Influence of Crime News Coverage on Public Perceptions…………………………. 4 Newsworthiness of Criminal Instances……………………………………………..……..6 Biases in Crime News Coverage: Individual Characteristics……....……...…………...….8 Biases in Crime News Coverage: Circumstantial Characteristics………..…..….…….....13 METHODS………………………………………..……….……………………………...………14 Overview……………………………………………..…………………..……....……….14 Research Questions…………………………………..…………………………………..16 Methodology………………………………………..…………..……………………..…..17 RESULTS…………………………..…………………………………………...……..……….…27 Measurement of Variables………………..………………………………………20 Analytical Strategy……………………..………………………………………...24 DISCUSSION……..…………………………………………..………….……………..…….…41 Limitations………………………..……………………………………………………...43 Conclusion…………………………………....……………………..……………………44 APPENDIX…………………………………..…………..……………...…………………….…46 REFERENCES……………………………..………………..……………………………..….…49 v LIST OF TABLES Table 1 Independent Variables…………………………………..………………………………22 Table 2 Control Variables………………………………………………………..…………..…..23 Table 3 Descriptive Statistics of Articles (N=826)……………………….………………..….…27 . Table 4 Descriptive Statistics of Cases (N=112, N=97)…………….……………………..….…28 Table 5 Statistics for Chi-square Tests of Association of Variables with News Coverage (N=112) ………………………………………………………………………………………………..….30 Table 6 Statistics for Independent t-Tests for Comparison of Mean Number of Articles for Cases with at Least One News Story (N=97)…………………………………………….………..…….31 Table 7 Statistics for Independent t-Tests for Comparison of Mean Number of Words for Cases with at Least One News Story (N=97)………………………………………….………………...33 Table 8 Comparing Goodness of Fit for Models that Predict the Count of News Stories Occurrence of News Coverage …………………………………..……………………...……….35 Table 9 Model 1: Parameter Estimates (n=112)……………………………………….…..…....36 Table 10 Model II: Parameter Estimates (n=89)……………………………………….….…....37 Table 11 Robust Regression Analysis of Mean Word Count of Intimate Homicide News Coverage (n=96) …………………………………………………………………………………………...40 vi INTRODUCTION Crime news coverage plays an important role in shaping the general public’s perception of the prevalence and causes of crime in the community (Bjornstrom et al., 2010; Chermak, 1994b; 1995; Dowler, 2004a; Dowler, 2004b; Gilliam & Iyengar, 2000; Surette, 1990). A large body of previous research has examined discretion of newsmakers in deciding which criminal instances do and do not receive news media coverage (Bjornstrom et al., 2010; Buckler & Travis, 2005; Chermak, 1995, 1998; Dixon & Linz, 2000; Dowler, 2004a, 2004b; Gilliam & Iyengar, 2000; Gruenewald et al., 2009, 2013; Lundman,, 2003; Peelo et al., 2004; Pritchard & Hughes, 1997; Sorenson et al., 1998; Taylor & Sorenson, 2002; Weiss & Chermak, 1998). Evidence indicates that newsmakers rely on common cultural stereotypes and existing belief systems about race and gender when assessing newsworthiness of criminal instances (Gilliam & Iyengar, 2000; Gruenewald et al., 2013; Entman, 1992, Fishman, 1980; Freng, 2007). The reliance on cultural stereotypes has led to biases of newsmakers to favor crimes involving certain circumstantial factors, individual characteristics of the crime victims, individual characteristics of the crime perpetrator, and gender and race combinations of victims and perpetrators when assessing the newsworthiness of criminal incidents. The present study builds upon previous research by examining the newsworthiness of a sample of intimate partners homicides, some of which involved Native American victims and perpetrators. Specifically, the present study assesses the newsworthiness of intimate partner homicide incidents occurring in the state of Montana by conducting a news media distortion analysis. News media distortion analysis links specific crime incidents with the news articles written regarding the specific incident to examine the criteria newsmakers used when assessing newsworthiness of criminal instances (Gruenewald et al., 2009, 2013; Peelo et al., 2004; Pritchard & Hughes, 1997; Sorenson et al., 1998; Weiss & Chermak, 1998).The two data sources 1 used in the present study are the Montana Domestic Violence Fatality Review Commission 2017 report (Montana Department of Justice, 2017) and newspaper coverage of the intimate homicide incidents listed in the report. News coverage was in four Montana news sources. The present study analyzes the news sources’ assessment of newsworthiness by examining the relationship between characteristics of the victim and the perpetrator (who in some cases was found guilty by a court) and the occurrence and amount of news coverage. The literature review that follows next explains the context of news production. The literature review also summarizes the evidence showing that news coverage influences public opinion, which is relevant to the present study because it shows the importance of identifying biases in news coverage. The remainder of the literature review discusses findings from previous studies that assess newsmakers’ evaluation of the newsworthiness of crime. Specifically, the section “Newsworthiness of Criminal Instances” explains the cultural factors that influence newsworthiness and the variables included in previous analyses of newsworthiness. The following sections, “Biases in Crime News Coverage: Individual Characteristics” and “Biases in Crime News Coverage: Circumstantial Characteristics,” provide detail regarding the individual characteristics of the crime victim and perpetrator and circumstantial characteristics that are associated with biases in crime news coverage. These last two sections discuss which variables are associated with newsworthiness and the direction of the relationships between variables and the occurrence and amount of news coverage. 2 A Description of News Production LITERATURE REVIEW News stories are created and selected for publication by journalists and decision-makers (e.g., editors) in the field of news media. News sources have limited space to publish their articles, so journalists must use discretion to determine which community events are newsworthy and should receive news coverage (Bjornstrom et al., 2010; Buckler & Travis, 2005; Chermak, 1994a, 1994b, 1995; Dowler, 2004a; Gilliam & Iyengar, 2000; Lundman, 2003; Shudson, 1989). Newsworthy refers to “criteria by which news producers choose which of all known events are to be presented to the public as news events” (Surette, 1998, p. 60). Once an event is determined newsworthy, the journalist creates a story about the event. The news story is then either selected or rejected for publication (Chermak, 1994a, 1994b; 1995; Shudson, 1989). The ability of newsmakers to determine which events are newsworthy and the potential for news sources to shape public perceptions have led to the examination of the process for determining what factors lead newsmakers to deem some events as newsworthy and what factors lead them to reject stories (Bjornstrom et al., 2010; Buckler & Travis, 2005; Chermak, 1994a, 1994b, 1995; Dowler, 2004a; Gruenewald et al., 2009, 2013; Gilliam & Iyengar, 2000; Lundman, 2003; Shudson, 1989). Researchers who examine the production of news draw attention to the fact that news sources are businesses competing in a competitive market for viewership in order to earn a profit (Bjornstrom et al., 2010; Buckler & Travis, 2005; Chermak, 1994a, 1994b; 1995; Dowler, 2004a; Fishman, 1980; Shudson, 1989). In this market share perspective, the priority of news media sources is not to provide an accurate reflection of local and national events. Instead, the goal is to earn a profit. News media sources accomplish their profit-driven agenda by publishing news stories they believe will resonate with their viewership, 3 solicit their attention, and appeal to their entertainment needs (Bjornstrom et al., 2010; Buckler & Travis, 2005; Chermak 1994a, 1994b). Crime stories are one type of news story that meets news media sources’ goals of providing exciting and entertaining publications. News producers recognize the exciting nature of criminal activity and capitalize on its appeal to readers. Overall, news sources dedicate a significant portion of their space to stories about criminal activity and its aftermath, such as the police investigation and the trial and sentencing of the perpetrator (Dowler, 2004a; Freng, 2007; Gilliam & Iyengar, 2000; Gruenewald et al., 2009; Surette, 1998). Violent crime news stories are especially prevalent and are disproportionately published in the media compared to stories about non-violent crimes (Dowler, 2004a; Gruenewald et al., 2009; Freng, 2007; Surette, 1998). The popularity of violent crime news coverage has been attributed to the exciting, sensational, and dramatic nature of the news stories which draw in readers in the competitive market of mass media (Buckler & Travis, 2005; Chermak, 1994a, 1994b; 1995; Dowler, 2004a; Gilliam & Iyengar, 2000; Gruenewald et al., 2009). Typically, the more serious a crime, the more likely it will be deemed a newsworthy story that is selected for publication (Dowler, 2004a; Gruenewald et al., 2013). The Influence of Crime News Coverage on Public Perceptions Local newspapers are essential in informing the public of the social conditions in the community. News stories provide information that allows individuals to gauge events and people in their communities with information beyond what is provided by their everyday interactions with their environments (Bjornstrom, Kaufman, Peterson, Slater, 2010; Dowler, 2004a; Fishman, 1980; Gilliam & Iyengar, 2000; Schudson, 1989). Thus, local newspaper coverage can shape public perception of social conditions and community safety (Chermak, 1994a, 1994b; 1995; 4 Dowler, 2004a; Dowler, 2004b; Gilliam & Iyengar, 2000; Gruenewald, Chermak, Pizarro, 2013; Gruenewald, Pizarro, Chermak, 2009; Lundman, 2003; Shudson, 1989; Surette, 1990). Crime is an easily identifiable threat to community safety, which causes the public to have a vested interest in understanding the nature and causes of crime in the area. Because most people do not have personal experience with criminal offending or victimization, their perception of crime in the community is largely developed from crime news coverage (Bjornstrom et al., 2010; Gilliam & Iyengar, 2000; Surette, 1990). In fact, research indicates that at least 95% of the population relies on news media as their primary source of local crime information (Surette, 1990). Given that the general public often relies on local news coverage to learn about crime in the area, individuals’ perceptions of the prevalence and causes of crime can be shaped by news coverage (Bjornstrom et al., 2010; Chermak, 1994b; 1995; Dowler, 2004a; Dowler, 2004b; Gilliam & Iyengar, 2000; Surette, 1990). The combination of the general public’s reliance on news sources for information about the prevalence of crime and news sources’ tendency to dedicate a significant portion of space to crime news creates a worrisome phenomenon. News coverage of crime typically focuses on the particular incident, e.g., the circumstances of the crime, the police investigation, and the perpetrator(s) and victim(s) of the crime (Chermak, 1994a, 1994a, 1994b, 1995; Dowler, 2004a; Dowler, 2004b; Gruenewald et al., 2009, 2013). The focus on individual crimes in news coverage makes it difficult for the general public to understand the underlying social forces that drive the crime rate (Dowler, 2004a). Without an understanding of any broader social conditions responsible for crime, the general public may be quick to assume that characteristics of individuals discussed in crime stories are the causes of crime (Bjornstrom et al., 2010; Dowler, 2004a; Dowler, 2004b; Gruenewald et al., 2009; Freng, 2007; Surette, 1990). In fact, previous research shows that the overrepresentation of 5 individuals of particular racial categories in crime stories leads to the false perception that people belonging to the racial category are inherently criminal and are responsible for most or all of the crime in the community (Bjornstrom et al., 2010; Campbell, 1995; Dixon & Linz, 2000; Dowler, 2004a; Dowler, 2004b; Gilliam & Iyengar, 2000; Gruenewald et al., 2009; Freng, 2007). Newsworthiness of Criminal Instances The ability of newsmakers to determine which criminal instances are newsworthy and news sources’ capacity to influence public perception about the prevalence and causes of crime has led researches to analyze which characteristics of criminal instances make some crimes newsworthy while other crimes do not receive any media attention (Bjornstrom et al., 2010; Buckler & Travis, 2005; Chermak, 1995, 1998; Dixon & Linz, 2000; Dowler, 2004a, 2004b; Gilliam & Iyengar, 2000; Gruenewald et al., 2009, 2013; Lundman,, 2003; Peelo et al., 2004; Pritchard & Hughes, 1997; Sorenson et al., 1998; Taylor & Sorenson, 2002; Weiss & Chermak, 1998). In general, research has found that cultural stereotypes influence newsmakers’ appraisal of newsworthiness. News sources rely on common narratives to make assumptions about who is involved in criminal incidents (Gruenewald et al., 2013) and tend to report criminal instances that reflect existing belief systems about race and gender. For example, violent crimes that involve a minority male perpetrator and white female victim receive more extensive coverage compared to crimes involving other race and gender perpetrator and victim combinations (Bjornstrom et al., 2010; Buckler & Travis, 2005; Gruenewald et al., 2013). The trend has been attributed to the cultural goal to protect white femininity and the history of fear of white female victimization at the hands of black males (Bjornstrom et al., 2010) The reliance on existing belief systems in crime news reporting in turn fosters the perpetuation of stereotypes based on race and 6 gender belief systems among the general public (Gilliam & Iyengar, 2000, Chermak, 1995; Entman, 1992, Fishman, 1980; Freng, 2007). Analyses that indicate a bias in assessing newsworthiness of crime incidents include two types of variables: individual and circumstantial (Bjornstrom et al., 2010; Buckler & Travis, 2005; Chermak, 1994a, 1994b, 1995, 1998; Dowler, 2004a; Dowler, 2004b; Gruenewald et al., 2009. 2013). Individual characteristics include information about the perpetrator and victim, such as race, age, and gender (Bjornstrom et al., 2010; Buckler & Travis, 2005; Chermak, 1998; Gruenewald et al., 2009, 2013). Circumstantial characteristics include variables that indicate the circumstances of the crime itself, such as location of the incident, number of victims, and the type of weapon used. Examining both individual and circumstantial variables allows researchers to control for the influence of circumstantial factors when analyzing the effect of individual characteristics of the perpetrator and offender on the newsworthiness of the incident. Although research findings have been inconsistent regarding the direction and strength of the relationship between occurrence and amount of news coverage and certain individual and circumstantial characteristics, a large body of research indicates that individual and circumstantial characteristics both play an influential role among newsmakers when assessing newsworthiness of criminal incidents (Bjornstrom et al., 2010; Buckler & Travis, 2005; Chermak, 1995, 1998; Dixon & Linz, 2000; Dowler, 2004a, 2004b; Gilliam & Iyengar, 2000; Gruenewald et al., 2009, 2013; Lundman,, 2003; Peelo et al., 2004; Pritchard & Hughes, 1997; Sorenson et al., 1998; Taylor & Sorenson, 2002; Weiss & Chermak, 1998). Individual and circumstantial characteristics that have been shown to have the most significant influence on newsworthiness of criminal incidents are the race, gender, and age of the crime victim and perpetrator (Bjornstrom et al., 2010; Buckler & Travis, 2005; Chermak, 1998; Dixon & Linz, 7 2000; Dowler, 2004b; Gruenewald et al., 2009, 2013; Peelo et al., 2004; Prichard & Hughes, 1997; Sorenson et al., 1998), use of firearms (Buckler & Travis, 2005; Dowler, 2004a; Gruenewald et al., 2009, 2013), and multiple victims (Buckler & Travis, 2005; Chermak, 1995, 1998; Dixon & Linz, 2000; Gruenewald et al., 2009, 2013; Peelo et al., 2004; Sorenson et al., 1998; Weiss & Chermak, 1998). Biases in Crime News Coverage: Individual Characteristics Studies that examine the influence of crime perpetrator race compare the occurrence and extent of news coverage of crimes involving minority and white perpetrators while controlling for circumstantial and other individual variable characteristics (Buckler & Travis, 2005; Prichard & Hughes, 1997). Although violent offenders are predominantly white (Bjornstrom et al., 2010), several studies have indicated that crime incidents involving minority crime perpetrators are more likely to be reported in news stories compared to crime incidents involving white perpetrators (Bjornstrom et al., 2010; Dixon & Linz, 2000). Crimes involving African American perpetrators have been shown to be significantly more likely to appear in news media compared to white perpetrators in other analyses (Bjornstrom et al., 2010; Dixon & Linz, 2000; Sorenson et al., 1998). Results of a study by Buckler and Travis (2005) showed that crimes involving an African American or Asian crime perpetrator were 4.8 times more likely to be reported by news sources compared to crimes involving white perpetrators. However, it is important to note that some research has found the opposite trend, with crimes involving white perpetrators receiving more coverage compared to crimes involving minority perpetrators than was warranted by official law enforcement data (Dowler, 2004b; Prichard & Hughes, 1997). Research comparing the occurrence of news coverage of crimes involving Hispanic and African American perpetrators has found that crimes involving Hispanic perpetrators are more likely to receive 8 news coverage (Gruenewald et al., 2009, 2013). So far, studies that have examined the influence of crime perpetrator race on newsworthiness have not included Native Americans as a racial category in the analysis. The age and gender of the crime perpetrator also influence the occurrence and amount of news coverage a criminal incident receives (Chermak, 1998; Gruenewald et al., 2009; Peelo et al., 2004; Prichard & Hughes, 1997; Sorenson et al., 1998). Crimes involving perpetrators under the age of 18 (Peelo et al., 2004) and perpetrators over the age of 60 (Chermak, 1998; Sorenson et al., 1998) have been shown to receive more news coverage compared to crimes involving perpetrators in other age groups. Studies examining the effect of perpetrator gender on newsworthiness have had mixed results. Research has shown that crimes involving male perpetrators receive more coverage compared to crimes involving female perpetrators (Prichard & Hughes, 1997; Gruenwald et al., 2009), crimes involving female perpetrators receive more coverage compared to those involving male perpetrators (Chermak, 1998), and that there were no significant differences in the extent of news coverage of crimes involving male and female perpetrators (Sorenson et al., 1998). Individual characteristics of crime victims also influence whether or not a criminal incident will be reported by news sources and the amount of coverage it receives (Bjornstrom et al., 2010; Buckler & Travis, 2005; Chermak, 1995; Dixon & Linz, 2000; Dowler, 2004b; Gruenewald et al., 2009, 2013; Prichard & Hughes, 1997; Sorenson et al., 1998; Weiss & Chermak, 1998). Research by Bjornstrom et al. (2010) suggests that the race of the victim plays a more significant role in predicting the occurrence and length of news coverage than the race of the perpetrator (Bjornstrom et al., 2010). Studies comparing the occurrence of news coverage of crimes involving white and minority victims have had consistent findings; crimes involving 9 white victims are more likely to be reported by news sources and to receive more extensive news coverage compared to criminal instances involving minority victims (Bjornstrom et al., 2010; Buckler & Travis, 2005; Chermak, 1995; Dixon & Linz, 2000; Gilchrist, 2010; Gruenewald, Pizarro, Chermak, 2009; Juwani & Young, 2006; Prichard & Hughes, 1997; Sorenson et al., 1998; Weiss & Chermak, 1998). The length of crime news stories has also been found to be related to the race of the victim; longer stories are more likely to involve white victims compared to non-white victims (Buckler & Travis, 2005; Dowler, 2004b). Previous studies examining the influence of the gender of the victim have also had consistent findings. Crimes involving female victims are more likely to be reported in news media and to receive more extensive coverage than those involving male victims (Bjornstrom et al. 2010; Buckler & Travis, 2005; Chermak, 1995; Dixon & Linz, 2000; Prichard & Hughes, 1997). Combinations of gender and race also play a significant role in assessing the newsworthiness of crimes, as crimes involving white female victims and minority male perpetrators are overrepresented in news media compared to other race and gender combinations (Bjornstrom et al, 2010). Victim age has also been shown to predict whether or not a crime was reported in the news, with victims under the age of 18 and over the age of 60 receiving disproportionate news coverage compared to victims in other age categories (Chermak, 1995; Dixon & Linz, 2000, Sorenson et al., 1998). Previous deviant behavior of the victim has also been shown to predict the amount of news coverage a crime receives. Generally, victims with a history of deviant behavior receive less news coverage compared to victims without deviant backgrounds (Gruenewald et al., 2009; 2013). Interestingly, previous deviant behavior has a different effect across gender subgroups. A study by Gruenewald et al. (2013) found that male victims with deviant histories received 10 significantly less coverage compared to males without deviant histories. The study also found a positive relationship between female victims with deviant backgrounds and news coverage of the crime, although the results were not significant (Gruenewald et al., 2013). In general, newsmakers overrepresent crimes involving victims who are white compared to non-white, female compared to male, exceptionally young and old compared to other age groups, and without a history of deviant behavior (Bjornstrom et al., 2010; Buckler & Travis, 2005; Chermak, 1995; Dixon & Linz, 2000; Dowler, 2004b; Gruenewald et al., 2009, 2013; Prichard & Hughes, 1997; Sorenson et al., 1998; Weiss & Chermak, 1998). The biases found in news coverage regarding victim characteristics indicate that newsmakers prioritize reporting on crimes that involve victims perceived to be vulnerable and thus worthy of public sympathy (Gruenewald et al., 2009; 2013). The biases in newspaper coverage also indicates that crimes involving victims with limited social ties receive less news media attention (Gruenewald et al., 2009, 2013; Lundman, 2003; Pritchard & Hughes, 1997). Certain combinations of the race of the victim and perpetrator involved in crime have also been found to result in biases in the assessment of the newsworthiness of criminal incidents (Buckler & Travis, 2005; Gruenewald et al., 2013). Buckler and Travis (2005) found that homicides that involved a white victim and non-white perpetrator were more likely to be reported by news sources and to receive more extensive coverage compared to other race combinations. An examination of minority homicides by Gruenewald et al. (2013) found that intra-racial homicides were significantly more likely to be reported by news source when the homicide involved a Hispanic victim and perpetrator compared to an African American victim and perpetrator. 11 Overall, previous research examining the biases of newsmakers in assessing the newsworthiness of criminal incidents supports the argument that newsmakers rely on common narratives regarding race and gender and on cultural stereotypes when making their assessments of newsworthiness (Campbell, 1995; Chermak, 1995; Entman, 1992; Fishman, 1980; Gilchrist, 2010; Gilliam & Iyengar, 2000; Jiwani & Young, 2006; Razack, 2000; Wilcox, 2006). Newsmakers tend to underrepresent criminal incidents based on characteristics of the victim and perpetrator, overrepresent crimes involving certain race combinations of the victim and perpetrator (Buckler & Travis, 2005; Gruenewald et al., 2013), provide less news coverage of crimes involving victims with fewer social ties (Gruenewald et al., 2009, 2013; Lundman, 2003; Pritchard & Hughes, 1997), and express less outrage regarding crimes involving minority victims compared to white victims (Dowler, 2004b). The overrepresentation of minority crime perpetrators has been found to influence the perception that minorities are inherently more criminal and are responsible for most criminal activity (Bjornstrom et al., 2010; Dowler, 2004a; Dowler, 2004b; Entman, 1994; Gilliam & Iyengar, 2000;). News media have also been shown to present a racialized image of criminality that perpetuates stereotypical beliefs about minorities in the general public (Bjornstrom et al., 2010; Campbell, 1995; Dowler, 2004a; Dowler, 2004b; Entman, 1994; Fishman, 1980; Gilliam & Iyengar, 2000). The decision of newsmakers to highlight crimes that follow expected social scripts (e.g. vulnerable elderly victims, white victimization by minority perpetrators) further perpetuates the continued stereotyping of certain populations (Gilliam & Iyengar, 2000, Chermak, 1995; Entman, 1992, Fishman, 1980), and the underreporting of crimes based on characteristics of the victim further marginalizes the experiences of populations with fewer social bonds (Campbell, 1995; Gilchrist 2010; Jiwani & Young, 2006; Wilcox, 2005). 12 Biases in Crime News Coverage: Circumstantial Characteristics Several circumstantial variables that characterize criminal instances have been shown to predict news coverage (Buckler & Travis, 2005; Chermak, 1995, 1998; Dixon & Linz, 2000; Dowler, 2004a, 2004b; Gilliam & Iyengar, 2000; Gruenewald et al., 2009, 2013, Lundman, 2003; Peelo et al., 2004; Sorenson et al., 1998; Taylor & Sorenson, 2002; Weiss & Chermak, 1998). Circumstantial characteristics that have been repeatedly shown to influence the assessment of newsworthiness of a homicide incident are oftentimes related to the seriousness of the offense. Crimes that are more serious in nature, such as violent crimes and homicides, are much more likely to be reported in the news and to receive more extensive coverage by news sources (Chermak, 1998; Gruenewald et al., 2013). Crimes involving the use of firearms are more likely to be reported by news sources compared to crimes involving other types of weapons or no weapons (Buckler & Travis, 2005; Dowler, 2004a; Gruenewald et al., 2009, 2013). A bias in news coverage also occurs concerning the number of victims impacted by a crime. Criminal instances that involve multiple victims compared to a single victim are more likely to receive news coverage (Buckler & Travis, 2005; Chermak, 1995, 1998; Dixon & Linz, 2000; Gruenewald et al., 2009, 2013; Peelo et al., 2004; Sorenson et al., 1998; Weiss & Chermak, 1998). The number of crimes committed during criminal incidents is also positively associated with news coverage (Chermak, 1998). Homicides that involve a more sensational motive and/or method of murder receive significantly more news media coverage (Gruenewald et al., 2009; Peelo et al., 2004). 13 Overview METHODS The present study builds upon previous research that assess the newsworthiness of crime by determining which characteristics of criminal incidents are more likely to predict the incident’s being covered at all and the amount of coverage by news media. It fills a gap in previous studies by analyzing data on a sample of intimate partner violence crimes that includes some Native American crime victims and perpetrators. Although media coverage of this group’s involvement in intimate partner violence has not been studied before, it is important to study crimes involving this group as the minority experience of the Native American population is further compounded compared to other racial minorities due to their history of colonization and forced segregation on tribal reservations (Freng, 2007; Gilchrist, 2010; Greenfield & Smith, 1999; Jiwani & Young, 2006; Mann & Zatz, 2002; Mihesuah, 1996). Examining news coverage of crimes involving Native American is also important considering the higher rates of criminal victimization and offending among Native Americans men and women compared to other racial categories (Bachman, Zaykowski, Kallmyer, Poteyeva, & Lanier, 2008; Freng, 2007; Gilchrist, 2010; Golberg, Champagne, &Singleton, 2007; Greenfield & Smith, 1999; Jiwani & Young, 2006). For example, the National Crime Victimization Survey indicates that the rates of physical and sexual assault against Native American women are much higher compared to the rates of victimization of white and African American women (Bachman et al., 2008). Violence against Native American women is compounded on tribal reservations compared to areas off a tribal reservation. Counties consisting of primarily tribal lands have extremely high assault and homicide rates among Native American women compared to Native American women in the general population (Bachman et al., 2008; 14 Golberg et al., 2007). Some counties with primarily tribal land have homicide rates of women that are over 10 times the national average rate (Bachman, 2008). Previous research on the representation of Native Americans in news media has noted the common stereotypes of Native Americans as savage, heathen, thieving, drunk, violent, and bloodthirsty, which solidifies the view of Native Americans as inherently criminal (Mann & Zatz, 2002; Mihesuah, 1996). The pervasive stereotypes of the Native American population as criminal and their rates of involvement in crime as victims and perpetrators highlight the need to analyze the characteristics that make intimate partner violence incidents involving Native Americans newsworthy. Previous research on the circumstantial and individual characteristics that make some crimes newsworthy often has compared official crime data to news media coverage of the crimes identified in the official crime data (Gruenewald et al., 2009, 2013; Peelo et al., 2004; Pritchard & Hughes, 1997; Sorenson et al., 1998; Weiss & Chermak, 1998). Such research is referred to as news media distortion analysis. News media distortion analyses use official crime data to guide a search for news coverage written about each crime incident included in the data. The criminal incidents are linked to all news articles written about the particular incident, and the news coverage is then analyzed to determine if characteristics of the crime victim, crime perpetrator and circumstances of the criminal incident predict the occurrence of any news coverage and the amount of news coverage (Buckler & Travis, 2005; Gruenewald et al., 2009, 2013). The occurrence of any news coverage, number of news stories published, word count of the stories, and placement in the newspaper or news broadcast have been used in previous studies as indicators of newsworthiness. (Buckler & Travis, 2005; Gruenewald et al., 2009, 2013). Homicide incidents are often used as the unit of analysis in media distortion analysis rather than 15 other types of crimes because homicides are the most likely to come to law enforcement’s attention and thus are included in official crime data (Buckler & Travis, 2005; Lundman, 2003). The increased likelihood that homicide incidents appear in official crime data ensures that the entire universe of homicide incidents are included in the analysis. The present study is a news media distortion analysis of all intimate partner homicides occurring in the state of Montana from February 15, 2000 to December 28, 2016. News coverage from four Montana news sources was analyzed. Montana has a higher proportion of Native American residents compared to most other states, with Native American comprising 6.4% of the population in Montana and 0.9% of the population in the United States (U.S. Census Bureau, 2020). Native Americans are disproportionately victimized by intimate partner homicide in Montana. Although Native American make up 6.4% of the population in Montana, they are 11% of intimate partner homicide victims (Montana Department of Justice, 2017). Research Questions The present study addresses the following research questions: Q1: Are victim race and gender related to whether a domestic homicide is covered in the news? Q2: Are victim race and gender related to the amount of coverage in the news? Q3: Are perpetrator race and gender related to whether a domestic homicide is covered in the news? Q4: Are perpetrator race and gender related to the amount of coverage in the news? Q5: Are white female victims with Native American male perpetrators more likely to be covered in the news than other race and gender combinations? Q6: What independent variables either most increase the odds of being covered at all in the news or explain the greatest amount of variance in amount of crime news coverage? 16 Methodology Data for the present study come from two sources. The first source of data is the Montana Domestic Violence Fatality Review Commissions (MDVFRC) 2017 report (Montana Department of Justice, 2017), which was published August 2017. The MDVFRC was created by the state legislature in 2003 to study domestic violence homicides to provide recommendations to prevent such fatalities. To acknowledge the disproportionate number of Native American domestic homicide victims and perpetrators and the unique cultural aspects of such incidents, the commission formed the Native American Domestic Violence Fatality Review Team in 2014 to specifically review domestic homicide incidents that occur in Indian Country. The MDVFRC 2017 report includes data on all intimate partner homicides that have occurred in the state of Montana from February 15, 2000 to December 28, 2016. The data indicates that 112 intimate partner homicide incidents occurred in the state during the timeframe. Of the 112 intimate partner homicide incidents, 18 incidents involved Native American victims. The 112 intimate partner homicide incidents resulted in 175 fatalities. Some of the incidents ended in the death of the perpetrator and/or additional victims besides the intimate partner of the perpetrator (Montana Department of Justice, 2017). The data in the MDVFRC 2017 report is listed in a table format and includes the intimate partner homicide victims’ first and last name, age, and race. The MDVFRC 2017 report also includes the location of the homicide by city, date of the homicide, and type of weapon used. The second source of data are newspaper articles published on each intimate partner homicide incident detailed in the MDVFRC 2017 report. The sample of newspaper articles come from four Montana newspaper sources. The articles are accessed through the online newspaper database NewsBank, which contains a comprehensive collection of 5,817 world news sources (Newsbank, 2019). The NewsBank database stores articles from twenty-two Montana news 17 sources. The selection of news sources for the present study was based on the following criteria; the sources are newspapers published in Montana, the database includes articles from the sources beginning January 1, 2000 or earlier, and the source covers statewide crime and court news. The four news sources that met the criteria and that are included in the present study are as follows: Billings Gazette: Published daily in Billings, Montana and is the top circulating newspaper in the state with a weekday circulation of 22,933 and a Sunday circulation of 25,186 (SRDS Media Planning Platform, 2019). Billings Gazette local coverage area includes southwest Montana and northeastern Wyoming. Billings Gazette also covers statewide, national, and world news (Billings Gazette, 2019). Bozeman Daily Chronicle: Published daily in Bozeman, Montana and is the fifth top circulating newspaper in the state with a weekday circulation of 10,406 and a Sunday circulation of 10,437 (SRDS Media Planning Platform, 2019). Bozeman Daily Chronicle has a total audience, including online readers, of more than 51,000. Bozeman Daily Chronicle local coverage area includes southwest Montana. Bozeman Daily Chronicle also covers statewide, national, and world news (Bozeman Daily Chronicle, 2019). Helena Independent Record: Published daily in Helena, Montana and is the sixth top circulating newspaper in the state with a weekday circulation of 8,830 and a Sunday circulation of 10,143 (SRDS Media Planning Platform, 2019). Helena Independent Record covers Helena area, state, regional, and national news (Helena Independent Record, 2019). Montana Standard: Published daily in Butte, Montana and is the seventh top circulating newspaper in the state with a weekday publication of 7,641 and Sunday circulation of 8,325 (SRDS Media Planning Platform, 2019). Montana Standard local coverage area is southwest 18 Montana. Montana Standard also covers statewide, regional, and national news (Montana Standard, 2019). Sample The unit of analysis for the present study is the intimate partner homicide incident. Each of the 112 intimate partner homicide incidents listed in the MDVFRC 2017 report was coupled to all newspaper coverage of the incident published by Billings Gazette, Bozeman Daily Chronicle, Montana Standard, and Helena Independent Record within three years following the date of the homicide incident. A three-year timeframe ensures that all intimate homicide incidents are allowed the same amount of time for article publication regarding each incident given that the most recent of the incidents listed in the report occurred on December 28, 2016. The three-year timeframe should capture the newspaper coverage of the homicide incident, the criminal investigation, and court proceedings including appeals (Buckler & Travis, 2005). The sample frame of newspapers articles for the present study was developed by using the newspaper online database NewsBank. Every intimate partner homicide incident listed in the MDVFRC 2017 was searched for newspaper coverage regarding each incident from the four newspaper sources analyzed in the present study: Billings Gazette, Bozeman Daily Chronicle, Montana Standard, and Helena Independent Record. The search for newspaper coverage for each homicide incident began by choosing one of the four newspaper sources to search. Next, a ‘date’ filter was set to capture a three-year timeframe of news coverage following the homicide incident. The information included in the date filter is the month the homicide occurred, the year the homicide occurred, and the month and year three years following the homicide. For example, the date filter for a homicide incident that occurred on January 01, 2000 would be set as January 2000 to January 2003. 19 Once the search filters are set, the MDVFRC 2017 report provided keyword search terms for each intimate partner homicide incident. The keywords from the report used to search for newspaper coverage on each homicide incident include the victim name and the city where the homicide occurred. The word and phrases used to search for newspaper coverage are as follows: 1. “Victim first and last name” as given in the MDVFRC 2017 report 2. Various spellings of victim first and last name to account for misspelling of the names in news coverage 3. “Homicide location” AND “death” 4. “Homicide location” AND “homicide” 5. “Homicide location” AND “murder” The article search procedure discussed above was repeated for the four newspaper sources and conducted for every intimate homicide incident listed in the MDVFRC 2017 report. The article search was conducted separately for each newspaper source because searching for articles from the four sources at once does not produce reliable search results. Articles that discuss the homicide victim and/or the homicide itself were included in the sample frame of articles. Once the sample frame of newspaper articles was collected, each article was analyzed to collect additional variable data that is not provided by the MDVFRC 2017 report. Measurement of Variables Dependent Variables: The dependent variables in the present study are whether the incident is covered at all in the news and for incidents that have at least one news story, the number of articles, and the average number of words per story. Whether or not there is any news coverage is coded as a dichotomous variable with 0= no news articles published and 1= news articles published. The amount of news media coverage for each homicide incident is measured 20 in two ways for incidents that received at least some news coverage. The first is the number of articles published regarding the homicide incident. The second is the average number of words published regarding each incident (see Appendix). The average number of words per incident was determined by dividing the total number of words published by the total number of articles published on each incident. Independent Variables: The independent variables for the present study are the race and gender of the intimate partner homicide victim and the gender of the perpetrator (see Table 1). Data regarding the race of the victim is from the MDVFRC 2017 report (0= non-Native American, 1 = Native American). The gender of the intimate partner homicide victim and perpetrator are not explicitly stated in the MDVFRC 2017 report. Victim gender was determined by a combination of news coverage and inferences of gender based on the victim’s first name as listed in the MDVFRC 2017. Data regarding perpetrator gender is gathered from the newspaper coverage of the homicide incident. Newspaper coverage discloses the gender of the crime victim or perpetrator by referring to them with gender specific labels (e.g. women, man, wife, husband, girlfriend, boyfriend). Based on frequencies of existing combinations among victim and perpetrator race and gender, the plan was to create dummy variables to reflect combinations that have been highlighted by previous research to be related to biases in newspaper coverage. For example, the plan was to create a dummy variable that would code an incident involving a female non-Native American homicide victim and male Native American perpetrator as 1, and all other combinations as 0. Due to small numbers and missing data, these dummy variables could not be computed. 21 Table 1 Independent Variables Variable Measurement Source Victim Race Victim Gender Perpetrator Gender 0 = non-Native American 1= Native American MDVFRC 2017 report 0 = female 1 = male 0 = female 1 = male Newspaper Articles Newspaper Articles Control Variables: Additional variables that have been shown to influence the news coverage a criminal incident receives were included in the present analysis as control variables (see Table 2). The control variables include individual characteristics of crime victims and perpetrators and circumstantial characteristics of crime. Control variables regarding the characteristics of individuals involved in the intimate partner homicides are from the MDVFRC 2017 report and newspaper coverage. The MDVFRC 2017 report includes the age of the intimate partner homicide victim. Perpetrator age was gathered from newspaper coverage of the homicide incidents. Circumstantial characteristics from the MDVFRC 2017 report are also related to the seriousness of the intimate partner homicide. The MDVFRC 2017 report lists the number of additional victims of the intimate partner homicide incident and notes if the additional victims were under 18 years old. Additional victims are victims that are not the intimate partner of the perpetrator. The additional victim variable was coded as 0 = no additional victim and 1 = additional victim. A dummy variable to indicate whether or not a perpetrator fatality occurred was also included in the analysis as an indicator of the seriousness of the incident. The MDVFRC also reports the type of weapon used in each intimate partner homicide incident. Since 22 crimes involving firearms receive more news coverage (Buckler & Travis, 2005; Dowler, 2004a; Gruenewald et al., 2009, 2013), the weapon variable in the present analysis was coded as 0 = no firearm use, 1 = firearm use. Several circumstantial characteristics are related to the location of the crime. The variable tribal reservation indicates whether the homicide occurred on or off a tribal reservation (0=no; 1=yes). Location population distinguishes the location of the homicides as either rural or urban (0= less than 5000; 1 = greater than 5000). Tribal reservations are in remote areas, and the variables tribal reservation and location population are intended to compare the amount of newspaper coverage for incidents in rural and remote areas with coverage of incidents in more populated areas. This comparison adds to the existing literature, as previous analysis of the newsworthiness of crimes draw media samples from urban areas. The variable location proximity distinguishes how far away a location is from one of the four cities of newspaper publication that are included in the present study (0 = city of publication or within 100 miles; 1 = greater than 100 miles from city of publication). Location proximity examines whether the location of the homicides and distance from a city of news publication will influence the amount of coverage it will receive. This comparison also adds to the literature, as previous studies examine news content that is produced and reporting on the city in which the content was published. Table 2 Control Variables Variable Name Measurement Source Victim Age Continuous Variable MDVFRC 2017 Report Perpetrator Age Continuous Variable Newspaper Articles Perpetrator Fatality 0 = no 1 = yes MDVFRC 2017 Report 23 Table 2 (cont’d) Additional Victims Firearm Use 0= did not involve additional victims 1 = did involve additional victims 0= no 1 = yes Tribal Reservation 0 = no 1 = yes Location Proximity 0 = city of publication or within 100 miles 1 = greater than 100 miles from city of publication MDVFRC 2017 Report MDVFRC 2017 Report MDVFRC 2017 Report MDVFRC 2017 Report Location Population 0 = less than 5000 1 = greater than 5000 Montana- demographics.com Analytical Strategy The data were analyzed using IBM SPSS Statistics 25 and STATA 14. Univariate analysis was conducted for all independent, control, and dependent variables to examine the frequencies, means, and measures of dispersion. The univariate results were examined to determine which variables to include in the bivariate and multivariate analyses. For example, categorical variables must include cases in all categories of the variable. Bivariate Analysis Bivariate analysis was conducted to examine the strength and direction of the relationship between each independent and control variable to each dependent variable. A Chi-square analysis was conducted to examine the categorical independent variables’ relationship with the occurrence of any news coverage of the homicide incidents (Q1, Q3). Chi-square analysis examines the relationship between categorical independent and categorical dependent variables by determining if the differences in the frequencies of categories are statistically significant. The Chi-square statistic provides a test of the significance of a comparison of observed with expected frequencies. Expected frequencies are the frequencies that are expected in each cell of the Chi- 24 square table if the null hypothesis is assumed, and observed frequencies are the actual frequencies in each cell. The larger the Chi-square statistic the more the observed frequencies differ from the expected frequencies. Rejecting the null hypothesis of a Chi-square analysis indicates that differences in frequencies of the categories of the independent variables are statistically significant, which indicates that the independent and dependent variables are related. It is important to note the limited use of Chi-square analysis with small or highly skewed samples. No more than 20% of the cells in chi-square tables should have expected frequencies of five or less. If more than 20% of the cells have expected frequencies of five or less, categories should be combined if the variable includes more than two categories (Weisburd & Britt, 2014). However, for 2 X 2 tables, Yates correction was used when cell frequencies were below five (Weisburd & Britt, 2014). The independent variables’ relationship with amount of news coverage (Q2, Q4) was analyzed using independent sample t-tests. Independent sample t-tests compare the mean values of two separate groups to assess if the difference in means of the dependent variable is statistically significant (Hahn Fox & Jennings, 2014; Ho 2014; Weisburd & Britt, 2014). The data for the present study meet the assumptions of an independent sample t-tests that the dependent variable is interval/ratio level and the independent variable is nominal level consisting of two groups that are independent of one another (Ho 2014; Weisburd & Britt, 2014). Multivariate Analysis Multivariate analysis was conducted to determine which variables result in the greatest increase of odds that the incident was covered at all or that explain the greatest amount of variation or variance of news coverage net the effects of control variables. 25 The occurrence of any news coverage (Q1, Q3, Q6) was examined with a logistic regression analysis. Regression analyses assess the relationship between variables by determining how well the independent variable predicts the dependent variable (Hahn Fox & Jennings, 2014; Ho 2014; Weisburd & Britt, 2014). Logistic regression is unique compared to other regression analyses because it can predict dichotomous dependent variables by limiting the dependent variable to values of zero and one (Weisburd & Britt, 2014). For incidents with at least one news story, the amount of news coverage (Q2, Q4, Q6) was examined using negative binomial regression to predict the number of news stories and robust regression to predict the average number of words in a news story. Negative binomial regression models count variables and is an appropriate test when analyzing over-dispersed dependent data (UCLA, 2020). Robust regression assesses the relationship between variables by determining how well the independent variables predicts the dependent variable to determine which variable explains the greatest amount of variance of the dependent variable. Robust regression is better test for the present analysis compared to OLS regression because robust regression can include outlier cases in the analysis (Western, 1995). 26 RESULTS The sample frame of news stories for the present study consists of 826 newspaper articles published by four Montana news sources. The majority of articles were published by the Billings Gazette (46.1%), followed by Montana Standard (23.4%), Helena Independent Record (16.5%), and Bozeman Daily Chronicle (14%) (see Table 3). Of the 112 intimate partner homicide incidents listed in the MDVFRC 2017 report, 15 incidents received no newspaper coverage and 97 incidents had at least one newspaper article published that discussed the homicide victim(s) and/or the homicide incident (see Table 4). The number of articles published regarding each homicide incident ranged from 0 articles to 79 articles, with a mean of 7.38 articles published per incident (SD=11.13). For the incidents with at least one story, the mean was 8.53 (SD=11.55). The average word count of news coverage published for all incidents ranged from 0 words to 2644.5 words (mean = 355.1530, SD= 356.57). For incidents with at least one story, the average word count ranged from 55 words to 2644.5 words (mean=410.07, SD=352.5). Table 3 Descriptive Statistics of Articles (N=826) n % Mean Std. Deviation Range News Source Billings Gazette Bozeman Daily Chronicle Helena Independent Record Montana Standard 381 116 136 46.1 14.0 16.5 193 23.4 Article Count Word Count 7.38 3799.10 11.13 0 - 79 7084.90 0 - 57317 Of the 112 homicide incidents, 18 involved a Native American intimate partner victim (16.1%) and 94 (83.9%) involved a non-Native American intimate partner victim. The race of the homicide perpetrator was not disclosed in the MDVFRC 2017 or in the newspaper coverage 27 regarding the homicide incidents and was not included in the current analysis. Female intimate partner victims were involved in 78 (69.6%) of the incidents and male intimate partner victims were involved in 34 (30.4%) of the incidents. The gender of the homicide perpetrator was available for 96 of the 97 incidents that received any news coverage. The newspaper coverage disclosed that 67 (59.8%) of the homicide perpetrators are male and 29 (25.9%) are female. None of the homicides that received news coverage involved same-sex couples. The age of the intimate partner victim ranged from 18 to 90 years old, and the age of the perpetrators ranged from 19 to 85 years old. The homicide perpetrators had a slightly higher mean age (42.97) compared to the intimate partner victims (42.54). Regarding the circumstantial characteristics of the intimate partner homicide incidents, 48 (42.9%) involved the fatality of the perpetrator (suicide or shot by officer) and 64 (57.1%) did not involve a perpetrator fatality. Ten of the homicide incidents included additional victim(s) that were not the intimate partner. Firearms were used in 71 (63.4%) of the homicide incidents. The remaining 41 incidents involved either stabbing, strangulation, blunt-force trauma, run-over with vehicle, hired killer, or pushed from cliff. Table 4 Descriptive Statistics of Cases (N=112, N=97) Race of Victim Native American Non-Native American Gender of Victim Female Male Gender of Perpetrator* Female Male Age of Victim Age of Perpetrator** n 18 94 78 34 29 67 % Mean Std. Deviation Range 14.70 14.08 18-90 19-85 42.54 42.97 16.1 83.9 69.6 30.4 25.9 59.8 28 Table 4 (cont’d) Firearm Use Yes No Perpetrator Fatality Yes No Additional Victims Yes No Tribal Reservation Yes No Location Population Less than 5000 Greater than 5000 Location Proximity Less than 100 miles Greater than 100 miles Articles No Articles Published Article(s) Published Article Count All cases Cases with news coverage Average Word Count All cases Cases with news coverage *Missing (N=23) **Missing (N=16) 71 41 48 64 10 102 13 84 46 51 54 43 15 97 63.4 36.6 42.9 57.1 8.9 91.1 13.4 86.6 47.4 52.6 55.7 44.3 13.4 86.6 7.38 8.53 355.15 410.07 11.125 11.544 0-79 1-79 356.57 0 – 2644.5 352.50 55 – 2644.5 Chi-Square analyses were conducted to examine the relationship between the dichotomous variable reflecting whether there was any news coverage and several nominal independent variables (see Table 5). Perpetrator gender was not included in the analysis since only cases that received some news coverage had data regarding perpetrator gender. The results of the analyses indicate that there is not a significant relationship between news coverage and the race of the victim, gender of the victim, use of a firearm, fatality of perpetrator, whether there was an additional victim involved whether the homicide occurred on a tribal reservation, the 29 proximity of the location of the incident to the city of publication, or the population of the location. Table 5 Statistics for Chi-square Tests of Association of Variables with News Coverage (N=112) No News Coverage News Coverage Non-NA NA Female Male No Yes No Yes No Yes No Yes % 80.0 20.0 73.3 26.7 40.0 60.0 46.7 53.3 93.3 6.7 86.7 13.3 Less than 100 53.3 miles Greater than 46.7 100 miles Less than 40.0 5000 Victim Race Victim Gender Firearm Use Perpetrator Fatality Additional Victims Tribal Reservation Proximity to City of Publication Population n 12 3 11 4 6 9 7 8 14 1 13 2 8 7 6 9 % 84.5 15.5 69.1 30.9 36.1 63.9 58.8 41.2 90.7 9.3 86.6 13.4 55.7 44.3 47.4 n 82 15 67 30 35 62 57 40 88 9 84 13 54 43 46 X2(1) p .005a .946 .001a .974 .086 .769 .776 .378 .000a 1.00 .000a 1.00 .000a 1.00 52.6 51 .288 .592 Greater than 60.0 5000 a Yates correction was used due to number of cells with expected value less than 5. Independent samples t-tests showed that three variables are significantly related to the total number of news articles about an incident (see Table 6). Stories about Native American 30 victims were covered in many fewer articles than stories about non-Native American victims. Homicides involving non-Native American victims were covered in more than twice the number of articles (M=9.38) than homicides involving Native American victims (M=3.27). The same pattern appeared comparing homicides occurring on a tribal reservation compared to homicides occurring off a tribal reservation. Homicides occurring off a tribal reservation (M=9.36) were covered in more than twice the number of articles of homicides occurring on a tribal reservation (M=3.15). The population of the location of the incident was also significantly related. Homicides that occurred in a city with a population of over 5000 people received over twice the amount of news coverage (M=11.22) compared to homicides occurring in cities of less than 5000 people (M=5.54). Table 6 Statistics for Independent t-Tests for Comparison of Mean Number of Articles for Cases with at Least One News Story (N=97) Test for Test for Equality of Variances Differences Victim Race Non-Native American Native American Victim Gender Female Male Perpetrator Gender* Female Male Firearm Use No Yes N 82 15 67 30 29 67 35 62 between Means µ 9.38 3.88 8.01 9.67 SD 12.30 3.27 9.62 15.13 8.59 8.60 14.43 10.24 10.14 15.58 F p 5.527 .021 ta(df) 3.45(84)a .656 .420 .003 .959 -.649(95) -.004(94) p .001 .585 .997 7.61 8.50 3.837 .053 1.037(95) .302 31 Table 6 (cont’d) Perpetrator Fatality No Yes Additional Victims No Yes Tribal Reservation No Yes Proximity to City of Publication Less than 100 miles 57 40 88 9 84 13 9.72 6.83 9.36 3.15 12.80 9.36 1.124 .292 1.22(95) .226 7.82 15.44 11.46 10.55 .163 .687 -1.91(95) .059 12.18 1.82 6.059 .016 4.363(95)a .000 54 10.56 10.641 Greater than 100 miles 43 5.98 12.239 1.479 .227 1.969(95) .052 Population Less than 5000 Greater than 5000 46 51 5.54 7.49 11.22 13.78 5.979 .016 -2.552(79)a .013 a When p < .05 for Levene’s test for equality of variances, t was calculated without assuming equality of variances. *One case has missing data so (N =96) Victim race, whether the homicide occurred on a tribal reservation, and the population of the location were also related to the average word count published on each incident (see Table 7). Stories involving non-Native American victims had more than twice as many average words per incident (M=4994.18) compared to stories involving Native American victims (M=1065.07). Homicides that occurred off a tribal reservation had nearly seven times the mean number of words (M=4953.24) compared to homicides occurring on a tribal reservation (M=725.15). Stories occurring in larger cities had more than three times the mean number of words published (M=2365.75) compared to stories occurring in smaller cities (M=6209.31). Overall, the results of 32 the bivariate analysis indicate that homicides involving Native American victims and that occur on tribal reservation and in small communities receive significantly less news coverage. Table 7 Statistics for Independent t-Tests for Comparison of Mean Number of Words for Cases with at Least One News Story (N=97) Test for Test for Equality of Variances p F Differences ta(df) between Means p Victim Race Non-Native American Native American Victim Gender Female Male Perpetrator Gender* Female Male Firearm Use No Yes Perpetrator Fatality No Yes Additional Victims No Yes Tribal Reservation No Yes City of Publication Less than 100 miles Greater than 100 miles N 82 15 67 30 29 67 35 62 57 40 88 9 84 13 54 43 µ SD 4994.18 1065.07 7930.12 1515.66 3928.30 5410.10 5316.91 10844.46 4646.72 4232.84 10829.29 5538.73 5234.75 3907.79 10322.54 5215.84 4787.79 3814.88 8659.056 5312.46 3878.58 9353.78 7432.91 5813.82 4953.24 725.15 7850.56 652.47 483.592 317.748 412.761 230.963 6.993 1.976 .775 2.453 .703 .059 7.175 1.888 .163 -.905(95) .010 4.096(95)a .381 .193(94) .404 .632(95) .121 .842(95) .808 -2.14(95) .009 4.829(89)a .173 2.356(95) .000 .368 .847 .402 .529 .035 .000 .021 33 Table 7 (cont’d) Population Less than 5000 Greater than 5000 a When p < .05 for Levene’s test for equality of variances, t was calculated without assuming equality of variances. *One case has missing data so (N=96) Before the multivariate analyses were conducted, a collinearity diagnostic test was run to 46 2365.74 51 6209.31 4020.75 9205.40 8.119 .005 2.709(70)a .008 examine potential collinearity between the independent and control variables. The results indicate that there are no concerns with multicollinearity between variables. The largest variance inflation factor value was 3.744 and the smallest value for tolerance was .267. A logistic regression analysis was conducted to determine which variables result in the greatest increase of odds that the homicide incident received any news coverage. The full model included victim race, victim gender, victim age, firearm use, whether there was a perpetrator fatality, whether there were additional victims, whether the incident occurred on a tribal reservation, the proximity of the location to the city of publication, and the population of the location. Perpetrator age was excluded from the model due to missing cases. The omnibus test for the model was not significant (X2 = 4.139, df=9, p = .902). Also, the hypothesized relationships between each independent variable and the dependent variable were not statistically significant. Two models were tested to examine the relationship between predictor variables and the number of articles published regarding each homicide incident. The first model included all cases and the predictor variables victim race, victim gender, victim age, firearm use, perpetrator fatality, additional fatality, whether the homicide occurred on a tribal reservation, location proximity to the city of publication, and location population. Perpetrator gender and perpetrator age were excluded from the first model due to missing data. The second model was tested for 34 only cases that received some news coverage. The second model includes perpetrator gender and perpetrator age in addition to the predictor variables used in the first model. For the two models, negative binomial regression and Poisson regression models were compared to determine the model that best fit the data (see Table 8). The log likelihood, AIC, and BIC values were used to compare the goodness of fit of the techniques. The results indicate that negative binomial regression best fit the data based on the smaller log likelihood, AIC, and BIC values. Negative binomial regression was then selected as the statistical technique to identify the predictors of the number of news articles. Table 8 Comparing Goodness of Fit for Models that Predict the Count of News Stories Occurrence of News Coverage LL df AIC BIC Model All Cases -615.285 102 1250.571 1277.756 Poisson with robust standard -322.641 101 667.281 697.185 Negative binomial with robust standard errors errors Cases with News Coverage -467.238 77 958.477 988.34 Poisson with robust standard errors -267.70 76 561.40 593.752 Negative binomial with robust standard errors The omnibus test for the first model predicting the number of news articles published was statistically significant (x2 = 36.30; df = 9; p = .000). The Wald chi-square tests showed that the variables, perpetrator fatality, additional victim, and location proximity were significantly related to the number of news articles published (see Table 9). Holding all other independent variables constant, incidents in which there was a perpetrator fatality had significantly fewer news stories 35 than those without a perpetrator fatality. The number of stories for incidents with a fatality is 39% less than for incidents without a fatality. Incidents in which there were additional victims received significantly more stories than those without additional victims. The number of stories for incidents with additional victims is 105% greater than for incidents without additional victims. Incidents that occurred less than 100 miles from a city of publication were covered in significantly fewer stories than those occurring greater than 100 miles from a city of publication. The number of stories for incidents less than 100 miles from a city of publication is 49% less than for incidents not occurring greater than 100 miles from a city of publication. The remaining variables were not statistically significantly related to the number of news articles published. Table 9 Model 1: Parameter Estimates (n=112) Victim Race Victim Gender Firearm Use Perpetrator Fatality Additional Victim Tribal Reservation Table 9 (cont’d) B -.907 .035 -.405 -.501 .719 -.270 Location Population Location Proximity Victim Age .168 -.665 -.014 Hypothesis Test Std. Error Wald Chi-Square df Sig. Exp(B) .5225 .2218 .2252 .2015 .3099 .5648 .2196 .1886 .0073 3.010 .026 3.239 6.186 5.382 .228 .587 12.449 3.754 1 1 1 1 1 1 1 1 1 .083 .873 .072 .013 .020 .633 .444 .000 .053 .404 1.036 .667 .606 2.052 .763 1.183 .514 .986 The omnibus test for the second model (tested with cases that had at least some news coverage) was also statistically significant (x2 = 41.00; df = 11; p = .000). The variables, firearm use, perpetrator fatality, additional victim, and location proximity to a city of publication, were significantly related to the number of news articles published (see Table 10). Holding all other 36 independent variables constant, incidents in which there was firearm use had significantly fewer news stories than those without firearm use. The number of stories for incidents with firearm use is 40% less than for incidents without firearm use. Incidents in which there was a perpetrator fatality had significantly fewer news stories than those without a perpetrator fatality. The number of stories for incidents with a fatality is 35% less than for incidents without a fatality. Incidents in which there were additional victims received significantly more stories than those without additional victims. The number of stories for incidents with additional victims is 104% greater than for incidents without additional victims. Incident that occurred less than 100 miles from a city of publication received significantly fewer stories than those occurring greater than 100 miles from a city of publication. The number of stories for incidents 100 miles or less from a city of publication is 55% less than for incidents occurring greater than 100 miles from a city of publication. The remaining variables were not statistically significantly related to the number of news article published. Table 10 Model II: Parameter Estimates (n=89) Hypothesis Test Victim Race Victim Gender Perpetrator Gender Firearm Perpetrator Fatality Additional Victim Tribal Reservation Location Population Location Proximity Perpetrator Age Victim Age B Std. Error Wald Chi-Square .789 1.291 .964 4.614 4.352 5.143 .668 .000 15.915 .295 .993 -.522 .477 .430 -.505 -.435 .710 -.525 -.005 -.791 -.006 -.009 .5878 .4201 .4381 .2352 .2084 .3132 .6424 .2315 .1982 .0102 .0095 37 df 1 1 1 1 1 1 1 1 1 1 1 Sig. Exp(B) .374 .593 .256 1.612 .326 .032 .037 .023 .414 .982 .000 .587 .319 1.537 .603 .647 2.035 .591 .995 .454 .994 .991 Comparing the first model (all cases and excluding perpetrator gender and age) and the second model (only cases with some news coverage and including perpetrator gender and age), the variables perpetrator fatality, additional victim, and location proximity to the city of publication were significantly related to the number of news articles published holding all other independent variables constant. In the second model, whether a firearm was used was also significantly related to the number of news articles. Whether there were additional victims involved had the greatest effect on the number of news articles predicted, with cases that involved additional victims having 105% and 104% increase in the number of articles predicted in the first and second models, respectively. The result is consistent with previous findings that crimes that involve multiple victims tend to receive more news coverage compared to crimes that involve a single victim (Buckler & Travis, 2005; Chermak, 1995, 1998; Dixon & Linz, 2000; Gruenewald et al., 2009, 2013; Peelo et al., 2004; Sorenson et al., 1998; Weiss & Chermak, 1998). The increase in coverage for homicides that involved more than one victim is consistent with the theme found in previous research that crimes considered more serious in nature tend to receive more news coverage (Chermak, 1998; Gruenewald et al., 2013). However, the variables perpetrator fatality and firearm use did not support the hypothesis that the occurrence of a perpetrator fatality and the involvement of a firearm would result in more news articles. In both models, the occurrence of perpetrator fatality and the use of a firearm both resulted in a decrease in the number of news articles predicted compared to homicides not involving a perpetrator fatality and homicides that involved the use of a firearm. The predictor variables’ relationship with the prominence of news coverage following a homicide incident was examined by testing a model that predicted the average word count in 38 each story published regarding each incident. STATA 14 was used to run a robust regression analysis. Mean word count was calculated by dividing the total number of words by the number of articles published regarding each homicide incident. The analysis was used to examine the predictor variable’s relationship with the average number of words per story after controlling for the other independent variables. The robust regression model included the predictor variables victim race, victim gender, victim age, perpetrator gender, perpetrator age, firearm use, perpetrator fatality, additional victim, whether the homicide occurred on a tribal reservation, location population, and location proximity to a city of news source publication. Only cases that received some news coverage were included in the robust regression model. Table 11 presents the results of the robust regression analysis. The results of the robust regression test indicate that the model fits that data and is significantly related to the predicted mean number of words published on homicide incidents (F = 2.32, df = 11, 84, p=.016). None of the individual predictor variables are significantly related to the mean number of words published when controlling for other independent variables. Although the predictor variables were not significantly related to the mean number of words published, it is worth the direction of their relationship to the dependent variable. Concerning individual characteristics of the homicide victim and suspect, the variables perpetrator gender (p =.100) and victim age (p =.110) explain the greatest among of variance of the mean number of words published. Being a male perpetrator is positively related to the mean number of words published. Thus, being female is negatively related to the mean number of words published. Victim age is negatively related to mean word count, which indicates that homicides involving younger victims are predicted to receive more average words compared to homicides involving 39 older victims. Regarding circumstantial characteristics of the homicides, the variable location proximity (p = .147) explained the greatest amount of variance of the mean number of articles published. Homicide incidents occurring greater than 100 miles from the city of publication were predicted to receive fewer mean number of words compared to homicides occurring less than 100 miles from a city of publication. Table 11 Robust Regression Analysis of Mean Word Count of Intimate Homicide News Coverage (n=96) Victim Race Native American Non-Native Americana Victim Gender Male Femalea Victim Age Perpetrator Gender Male Femalea Perpetrator Age Firearm Use Yes Noa Perpetrator Fatality Yes Noa Additional Victims Yes Noa Tribal Reservation Yes Noa Location Population Greater than 5000 Less than 5000a Location Proximity Greater than 100 miles Less than 100 miles aReference category Coef. Std. Err. T -141.205 134.624 -1.05 - .980 - -1.61 1.66 - -1.10 -.63 - -.17 - .73 - -.46 - -.01 - -1.46 - - 99.50 - -3.326 - 101.468 - 2.062 172.295 103.507 - -1.682 -39.670 - - 1.528 62.827 - -9.612 55.20 - - 63.538 86.841 - - -61.973 134.06 - - -.490 58.922 - -76.610 - - 52.354 - 40 P>|t| .297 - .330 - .110 .100 - .274 .529 - .862 - .466 - .645 - .993 - .147 - DISCUSSION The present study sought to determine which individual and circumstantial characteristics of intimate partner homicides are more likely to predict whether a homicide is covered by news sources and the amount of news coverage the incident receives. Of special interest to the present study is analyzing the relationship of victim race, victim gender, and perpetrator gender on whether a homicide receives news coverage and the amount of news coverage. The present study fills a gap in previous studies of the effect of race by analyzing data on a sample of homicides that includes Native American victims. Typically, Native American crime victims and perpetrators are not included in studies that examine the newsworthiness of crime. Several of the present study’s research questions (Q1, Q3, Q6) examine predictor variables’ relationship to whether an intimate partner homicide is covered by the news sources included in the sample. The predictor variables included in the logistic regression analysis were all expected to be significantly related to receiving any news coverage. Specifically, homicides involving female victims and non-Native American victims were hypothesized to have greater odds of being covered compared to homicides involving male victims and Native American victims. Surprisingly, the logistic regression model and the individual variables were not significantly related to whether a homicide received news coverage. Despite the lack of model significance when examining whether a homicide received news coverage, the independent variables did significantly predict the amount of news coverage. The circumstantial characteristics – whether an additional victim was involved, whether a perpetrator fatality occurred, and the homicide locations’ proximity to the city of news publication – were significantly related to the number of articles about the homicide incident. Whether a firearm was used was statistically significant in the model that included homicide 41 cases that received at least some news coverage. Victim race, victim gender, and perpetrator gender were not statistically significant in either of the models. Robust regression was used to identify the predictors of the mean number of words in news stories. Although the model significantly fit the data, none of the individual variables are significantly related to the mean number of articles when controlling for other independent variables. Being a male perpetrator is positively related to the mean number of words published compared to being female perpetrator. Among the circumstantial characteristics of the homicides, the variable location proximity explained the greatest amount of variance of the mean number of articles published. Overall, the results of the present study indicate that circumstantial characteristics of homicides explained the greater amount of variation in amount of news coverage compared to victim and offender characteristics. Specifically, the involvement of additional victims in homicide events was consistently significant across models. The location of the homicide was important as well; location proximity to a city of news publication was statistically significant to the number of articles published in the negative binomial regression model. Although the relationship between victim race and amount of news coverage was not always statistically significant, the relationship was in the expected direction across models with homicides involving Native American victims receiving less coverage. The relationship between gender and amount of news coverage was not in the hypothesized direction, with homicides involving male victims and perpetrators receiving more coverage compared to incidents involving female victims and perpetrators. The results of the present study, especially the relationship between individual characteristics and the amount of news coverage that contradicted the proposed hypotheses, 42 highlight to need for future research to further examine the relationship of individual characteristics of crime victims and suspects. The analyses indicate that homicides involving Native American victims and that occur on tribal reservations receive less news coverage than homicides involving non-Native American victims and that occur off tribal reservations. This highlights the marginalization of Native Americans involved crime and sheds light on the importance of future research in examining newsworthiness of crimes that involve Native American victims and suspects. Limitations The present study had several limitations that could be overcome in future research. One limitation of the present study is the small sample size of homicides involving Native American victims. Instead of a statewide approach, future research that examines the newsworthiness of crimes involving Native Americans could analyze news coverage of homicides occurring in counties with high populations of Native Americans across several states. A larger sample of homicides involving Native Americans would allow researchers to examine the interaction effect of victim and perpetrator race and gender. A second limitation is the reliance on newspaper coverage for collecting perpetrator gender and age data, as there may be biases in reporting. Another drawback is that the perpetrator data was unable to be collected for instances that did not receive any newspaper coverage. However, in the future the data of the present study could be supplemented with the FBI’s Supplementary Homicide Report. The specifics of the homicide incident listed in the Montana Domestic Fatality Review Commission Report could be used to guide a search for the incidents, which will all be listed in the SHR. The SHR discloses demographics of the victim and 43 perpetrator such as race, gender, and age. Thus, the SHR can be used to collect perpetrator gender, age, and race information for all the homicide incidents examined in the current study. Other limitations involve the methodology of newspaper article collection. The criteria for articles included in the sample of newspaper coverage was that the article mentioned the homicide incident. The news articles did not have to be solely about the homicide incident. For example, some article discussed domestic violence policies and mentioned specific intimate partner homicide incidents as warnings of the potential escalation of domestic violence. Future analysis and data collection should consider only articles that focus on one particular incident. Otherwise, the number of words and number of articles is not unique for each incident. In the future, it is recommended that only articles that solely discuss the homicide incident and investigation be included in the sample of articles. The second limitation concerning the methodology is the time frame. A three-year timeframe of news coverage following the homicide incident was chosen to collect news coverage of the homicide incident, investigation, and potential court proceeding. However, the three-year time frame could result in biases concerning the number of articles and mean number of words published regarding the homicide incidents. Specifically, homicide incidents that result in an extended trial period and appeals would have more coverage than incidents for which the perpetrator plead guilty. It is recommended that future research investigating the newsworthiness of homicide incidents include a shorter time frame of newspaper articles (e.g. one year) or only include news coverage of the homicide incident/investigation and exclude court coverage. Conclusion Despite the limitations of the current study, the study did contribute to the literature by examining the influence of being a Native American victim on the occurrence and amount of 44 news coverage the crime receives compared to being a non-Native American victim. The present study also filled gaps in previous studies by comparing rural and urban areas and the amount of news coverage crimes receive. The research also has implications that might be considered in the training and practice of news journalists. News journalists would benefit from training that addresses common cultural stereotypes of the Native American population to address biases that journalists may foster when covering news stories that involve Native Americans. The study also has implications regarding the practice of homicide investigations. Of the 97 intimate partner homicide incidents that received some news coverage, none involved same-sex couples. The trend may imply biases among crime scene investigators in determining which homicides are considered intimate partner homicides and which homicides are not. Law enforcement officers may benefit from training regarding gender bias and its impact on decision making when investigating homicides and other interpersonal crimes. Given the current study’s findings, future research should further examine the representation of Native American crime victims and suspects in crime news coverage. Future research that examines newsworthiness of crimes should also consider demographics of the location of the crime such as location population, location remoteness. Specifically, future studies should compare the newsworthiness of crime that occur on tribal reservations to explore if the remoteness of tribal reservations increases the marginalization of Native American crime victims and suspects in crime news coverage. 45 APPENDIX 46 Codebook Variables used to Search for Newspaper Articles 1. First name victim_firstname (from MDVFRC 2017 report) o String variable 2. Last name victim_lastname (from MDVFRC 2017 report) o String variable 3. Date of homicide incident date (from MDVFRC 2017 report) ▪ String variable ▪ Format: MM/DD/YYYY 4. Location of homicide incident by city location_city (from MDVFRC 2017 report) ▪ String variable: name of city Dependent Variables Main Dependent Variables 1. Number of Articles article_totalcount (from newspaper articles) ▪ Total number of newspaper articles published on each IPH incident o Numerical variable 2. Total Number of Words article_totalwordcount (from newspaper articles) ▪ Total number of words published on each IPH incident o Numerical variable (word count of all articles published on the incident) 3. Mean Number of Words meanwordcount ▪ Total number words divided by total number of articles Descriptive Measure of Newspaper Articles 4. Number of words per article article1_wordcount (from newspaper articles) Independent Variables 1. Race victim_race (from MDVFRC 2017 report) ▪ 0= non-Native American ▪ 1 = Native American 2. Gender victim_gender (from newspaper articles) ▪ 0 = female ▪ 1 = male ▪ 9 = not applicable (no news coverage of incident or not mentioned in news coverage) 3. Gender perpetrator_gender (from newspaper articles) ▪ 0 = female ▪ 1 = male ▪ 99 = not applicable (no news coverage of incident or not mentioned in news coverage) 47 Control Variables 1. Age victim_age (from MDVFRC 2017 report) ▪ Continuous variable 2. Age perpetrator_age (from newspaper articles) ▪ Continuous variable 3. Firearm use during intimate partner homicide incident firearm (from MDVFRC 2017 report) ▪ 0 = no ▪ 1 = yes 4. Additional Victims Present additionalvictim_present (from MDVFRC 2017 report) ▪ 0= did not involve additional victims ▪ 1 = did involve additional victims *Additional victims are any victims of intimate partner homicide incident that are not the intimate partner victim or the perpetrator of the homicide 5. Tribal Reservation ▪ 0 = no ▪ 1 = yes 6. Location Population ▪ 0 = 5000 or less ▪ 1 = greater than 5000 7. Location Proximity ▪ City of publication or within 100 miles ▪ Greater than 100 miles from city of publication 48 REFERENCES 49 REFERENCES Billings Gazette. 2019. Retrieved from https://billingsgazette.com/. Bachman, R., Zaykowski, H., Kallmyer, R., Poteyeva, M., & Lanier, C. (2008). Violence against American Indian and Alaska Native women and the criminal justice response: What is known. Washington, DC: National Institute of Justice. Bjornstrom, E. E. S., Kaufman, R. L., Peterson, R. D., & Slater, M. D. (2010). Race and ethnic representations of lawbreakers and victims in crime news: A national study of television coverage. Social Problems, 57(2), 269–293. doi: 10.1525/sp.2010.57.2.269. Bozeman Daily Chronicle. 2019. Retrieved from https://www.bozemandailychronicle.com/. Buckler, K., & Travis, L. (2005). Assessing the newsworthiness of homicide events: An analysis of coverage in the Houston Chronicle. Journal of Criminal Justice and Popular Culture, 12(1), 1-25. Campbell, C. P. (1995). Race, myth and the news. Thousand Oaks, CA: Sage Publications. Chermak, S. M. (1994a). Body count news: How crime is presented in the news media. Justice Quarterly, 11(4), 561–582. doi: 10.1080/07418829400092431. Chermak, S. M. (1994b). Crime in the news media: A refined understanding of how crime becomes news. In Gregg Barak (ed,), Media, process and the social construction of crime: Studies in newsmaking criminology. New York, NY: Garland Publishing. Chermak, S. M. (1995). Victims in the news: Crime and the American news media. San Fransisco, CA: Westview Press. Chermak, S. M. (1998). Predicting crime story salience: The effects of crime, victim, and defendant characteristics. Journal of Criminal Justice, 26(1), 61-70. doi:10.1016/S0047- 2352(97)00055-X. Dowler, K. (2004a). Comparing American and Canadian local television crime stories: A content analysis. Canadian Journal of Criminology and Criminal Justice, 46(5), 573– 596. doi: 10.3138/cjccj.46.5.573. Dowler, K. (2004b). Dual realities? Criminality, victimization, and the presentation of race on local television news. Journal of Crime and Justice, 27(2), 79–99. doi: 10.1080/0735648x.2004.9721196. Dixon, T. & Linz, D. (2000). Overrepresentation and underrepresentation of African Americans and Latinos as lawbreakers on television news. Journal of Communication, 50(2), 131- 154. doi: 10.1111/j.1460-2466.2000.tb02845.x. 50 Entman, R. (1994). Representation and reality of the portrayal of Blacks on network television news. Journalism Quarterly, 71(3), 509-520. doi: 10.1177/107769909407100303. Fishman, M. (1980). Manufacturing the news. Austin, TX: University of Texas Press. Freng, A. (2007). American Indians in the news: A media portrayal in crime articles. American Indian Culture and Research Journal, 31(1), 21–37. doi: 10.17953/aicr. 31. 1. y3r2n486 11025346. Gilchrist, K. (2010). “Newsworthy” victims? Exploring differences in Canadian local press coverage of missing/murdered Aboriginal and white women. Feminist Media Studies, 10(4), 373–390. doi:10.1080/14680777.2010.514110. Gilliam Jr, F. D., & Iyengar, S. (2000). Prime perpetrators: The influence of local television news on the viewing public. American Journal of Political Science, (44)3, 560-573. doi: 10.2307/2669264. Goldberg, C., Champagne, D., & Singleton, H. V. (2007). Final report: Law enforcement and criminal justice under Public Law 280. US Department of Justice, Washington, DC. Greenfeld, L. A., & Smith, S. K. (1999). American Indians and crime. US Department of Justice, Office of Justice Programs, Bureau of Justice Statistics. Gruenewald, J., Chermak, S. M., & Pizarro, J. M. (2013). Covering victims in the news: What makes minority homicides newsworthy? Justice Quarterly, 30(5), 755–783. doi: 10.1080/07418825.2011.628945. Gruenewald, J., Pizarro, J., & Chermak, S. M. (2009). Race, gender, and the newsworthiness of homicide incidents. Journal of Criminal Justice, 37(3), 262–272. doi: 10.1016/j.jcrimjus.2009.04.006. Hahn Fox, B., & Jennings, W. G. (2014). How to write a methodology and results section for empirical research. Journal of Criminal Justice Education, 25(2), 137-156. doi:10.1080/ 10511253.2014.888089. Ho, R. (2014). Handbook of Univariate and Multivariate Data Analysis with IBM SPSS (2nd ed.). Boca Raton, FL: Taylor & Francis Group. Helena Independent Record. 2019. Retrieved from https://helenair.com/. Jiwani, Y., & Young, M. L. (2006). Missing and murdered women: Reproducing marginality in news discourse. Canadian Journal of Communication, 31(4). doi:10.22230/cjc.2006v 31n4a1825. 51 Lundman, R. J. (2003). The newsworthiness and selection bias in news about murder: Comparative and relative effects of novelty and race and gender typifications on newspaper coverage of homicide. Sociological Forum (18)3, 357-386. doi:10.1023/ A:1025713518156. Mann, C. R., & Zatz, M. S. (Eds.). (2002). Images of color, images of crime: Readings. Los Angeles, CA: Roxbury. Mihesuah, D. A. (2010). American Indians: stereotypes & realities. Atlanta, GA: Clarity Press, Inc. Montana Cities by Population. (2018). Retrieved June 15, 2020, from https://www.montana- demographics.com/cities_by_population. Montana Department of Justice, Office of Consumer Protection and Victim Services. (2017). Report to the legislature: Montana domestic violence fatality review commissions. Retrieved from https://media.dojmt.gov/wp-content/uploads/MDVFR-2017-Legislative- Report-Final.pdf. Montana Standard. 2019. Retrieved from https://mtstandard.com/. Negative Binomial Regression. (n.d.). Retrieved July 01, 2020, from https://stats.idre.ucla.edu/ NewsBank. 2019. Retrieved from https://infoweb-newsbank-com.proxy2.cl.msu.edu/ apps/ stata/dae/negative-binomial-regression/. news/advanced-search?p=AWNB&f=advanced. Peelo, M., Francis, B., Soothill, K., Pearson, J., & Ackerley, E. (2004). Newspaper reporting and the public construction of homicide. British Journal of Criminology, 44(2), 256-275 doi: 10.1093/bjc/44.2.256. Pritchard, D., & Hughes, K. D. (1997). Patterns of deviance in crime news. Journal of Communication, 47(3), 49-67. doi: 10.1111/j.1460-2466.1997.tb02716.x. Razack, S. H. (2000). Gendered racial violence and spatialized justice: The murder of Pamela George. Canadian Journal of Law and Society, 15(2), 91–130. doi:10.1017/s08293201 00006384. Reiner, R., Livingstone, S., & Allen, J. (2003). From law and order to lynch-mobs: Crime news since the second World War. In Mason, P. (Eds.), Criminal visions: Media representations of crime and justice (pp. 13-32). Cullompton, UK: Willan Publishing. Robust Regression. (n.d.). Retrieved June 22, 2020, from https://stats.idre.ucla.edu/stata/dae/ robust-regression/. 52 Schudson, M. (1989). The sociology of news production. Media, Culture & Society, 11(3), 263–282. doi: 10.1177/016344389011003002. Sorenson, S. B., Manz, J. G., & Berk, R. A. (1998). News media coverage and the epidemiology of homicide. American Journal of Public Health, 88(10), 1510-1514. doi: 10.2105/ajph. 88.10.1510. SRDS Media Planning Platform. 2019. Retrieved from https://next-srds-com.proxy1.cl.msu. edu/ portal?newsess=y. Taylor, C. A., & Sorenson, S. B. (2002). The nature of newspaper coverage of homicide. Injury Prevention, 8(2), 121-127. doi: 10.1136/ip.8.2.121. Surette, R. (1990). The Media and criminal justice policy: recent research and social effects. Springfield, IL, U.S.A.: C.C. Thomas. Surette, R. (1998). Media, crime, and criminal justice: Images and realities, 2d ed. New York, NY: Wadsworth Publishing. United States Census Bureau. (2020). Retrieved from www.census.gov. Weisburd, D., & Britt, C. (2014). Statistics in Criminal Justice (4th ed.). New York, NY: Springer. Western, B. (1995). Concepts and suggestions for robust regression analysis. American Journal of Political Science, 39, 786–817. Weiss, A. & Chermak, S., M. (1998). The news value of African-American victims: An examination of the media’s presentation of homicide. Journal of Crime and Justice 11(2). 71-88. doi: 10.1080/0735648X.1998.9721601. Wilcox, P. (2005). Beauty and the beast: Gendered and raced discourse in the news. Social & Legal Studies, 14(4), 515–532. doi: 10.1177/0964663905057593. 53