METHODS

Sample selection and data sources

This investigation focused on stories about homicides published in the Los Angeles Times. The Los Angeles Times serves a very diverse population and is second only to the New York Times in newspaper circulation in the US. From 1990 to 1994, the Los Angeles Times published 2782 stories about a total of 1241 homicide victims. The paper prepared a large dataset on the articles as well as on all criminal homicides that occurred in the county during those years. The data provided the basis for a five part series published in December 1996 that was nominated for a Pulitzer Prize and comprised part of the data for the present study.

The Los Angeles Times database provided story information (for example, section of paper, page number, number of words, publication date) as well as homicide data from state and county offices including the California Department of Justice, Office of Vital Records and Statistics, Los Angeles County Department of the Coroner, and Los Angeles County Municipal and Superior Courts. The state and county offices provided information on victims and suspects including age, gender, ethnicity, and relationship.

The sample of articles used for this study was chosen by stratifying the 2782 homicide articles using two variables: ethnicity of the victim (Asian, black, Hispanic, white) and relationship of the victim to the suspect (acquaintance, child, gang, intimate, stranger). Stories about Hispanic victims, who constitute the largest ethnic group of homicide victims reported in the Los Angeles Times, were used as index cases for matching purposes. Forty five Hispanic victim articles were randomly selected, 10 from each adult relationship category—acquaintance, gang, intimate, and stranger—and five from the child category. (The number of stories about the homicide of a child is less simply because fewer such articles were published.) Articles for the Asian, black, and white strata were obtained by matching on two variables, victim-suspect relationship and publication date, using stories published as closely as possible in time to those about Hispanic victims. Because fewer stories were published about Asian and white homicide victims, a match could not be found for each index article. Table 1 reports the strata and numbers of the articles selected.

Data coding process

Copies of the 156 selected articles were made from microfiche and news coverage variables were abstracted. All articles went through two full reviews by independent coders (graduate students). After the first review, variable definitions were refined and a second review was conducted. The average inter-rater reliability for all nine abstracted variables was 88%, with agreement ranging from 84% to 98%. Some variation between reviews was expected due to coder subjectivity as well as slight alteration of definitions due to refinement. As a further consistency check, a 10% random sample of the reviewed articles was drawn and a third review was conducted by the second reviewer. Classifications remained consistent for eight of the variables; one variable, “genre”, was again refined and the variable was recoded again for all 156 articles.

Newspaper coverage variables examined

This study focused on two aspects of covering a homicide story—prominence in the newspaper and story frame. Three of the variables used to assess an article's prominence were taken directly from the Los Angeles Times database—section of the paper, page in section, and number of words—while two were abstracted from article copies. These included the placement of an article on a page (above or below the fold) and whether photographs were included in a story.

Story framing goes beyond the facts of a homicide to add depth to the persons involved, provide a social context, and provide an overall angle on how to think and feel about an incident. The following story frame variables were abstracted from the sample of articles and coded dichotomously according to the process described above: (1) victim background, (2) suspect background, (3) issue background, (4) use of a hook, (5) focus of the story, (6) genre, and (7) tone.

Victim and suspect background were coded as positive when any descriptive information was provided about the victim or suspect beyond basic facts such as age, gender, and race or ethnicity. In other words, any attempts to provide more depth to the victim or suspect were coded as “yes” for victim and suspect background. Similarly, if there was any mention of the broader social context and issues of concern related to the homicide, then issue background was coded positively.

A hook was defined as a first or second sentence in an article that focused on a catchy human interest element to the story rather than on facts. Hooks are used to draw a reader into a story. For example, a story with a hook might start out with, “A community is in shock today as a woman much loved and respected was murdered . . .”, whereas a story without a hook might begin with, “A 30 year old woman was killed . . .”. Each story was coded to indicate whether or not it had a hook.

The focus of a story was categorized as episodic or thematic. A story was coded as episodic if it focused on a particular incident or incidents rather than on the broader issue. To be classified as thematic, a story had to focus on a broader issue of which the fatal violence was a part (for example, child abuse, gang violence). For example, a thematic article might embed a gang homicide incident within a larger story about the history of gang violence in a particular neighborhood or a child's death within a story about the larger issue of child abuse.

Genre was coded as primarily factual or opinion based. Stories were coded as factual when journalists primarily transmitted unattributed statements and relatively objective facts. In contrast, stories were coded as opinion based when they were dominated by quotes from any sources or subjective personal descriptions (that is, beyond demographics), or when they were mainly issue narratives (that is, more like story telling than fact telling). Quotes generally add a personal angle to a story; for example: “Her brother said, `She was a devoted mother'” or “The officer said, `This is the worst case I've seen'”. An example of a subjective personal description is: “Neighbors described the man as friendly and always willing to lend a helping hand”. These are subjective because other persons might not have perceived the person or situation in the same way, thus, it is information that gives the story a selected angle.

Tone was classified as non-emotional or emotional. Stories were coded as non-emotional if feelings were not transmitted in the coverage. Stories were considered emotional if one or more sources, generally personal, raised feelings such as sadness, loss, or shock in the story. The following are some “emotional” tone examples: “The principal said that the whole school mourned the loss of their hero”, or “His mother said, `I don't know how I will ever get over this'”, or “The neighbor said, `I just can't believe this happened'”.

Data analysis

Descriptive statistics were generated to provide general information about the sampled homicide articles. Simple analyses were conducted using cross tabulations and χ² tests to examine associations between the predictor variables and each of the outcome variables. In addition to findings that meet the usual criterion of p<0.05, we included those at p<0.10 due to their potential substantive interest to some readers.

Multivariate analyses were conducted using binary, ordinal, and categorical logistic regression equations modeled to test the relationship between the two predictors (victim ethnicity and victim-suspect relationship) and the identified newspaper coverage variables. Adjusted odds ratios and 95% confidence intervals were calculated to measure differences in newspaper coverage by the two predictors. Stories with white victims and victims whose assailants were strangers were chosen as the reference groups since these victims tend to be more represented in the news than others.

RESULTS

Description of newspaper coverage

The majority of sampled homicide articles were printed in the second section of the newspaper on page one or two of that section (see table 2). At least half were printed “above the fold”, and most consisted of 500 or fewer words. Most stories (71.2%) did not include photographs.

Most articles did not report on the victim's or the suspect's background, nor were background issues discussed. Most articles did not contain a hook (69.9%) and were predominantly episodic (85.9%), factual (74.4%), and non-emotional (66.0%). Additionally, half of the homicides were murder-robberies (17.3%), murder-suicides (16.0%), or gang motivated (16.7%).

In the following results sections, findings of substantive importance will be described as well as those that reached statistical significance.

Newspaper coverage by predictors: cross tabulations

The prominence or visibility of a homicide story was generally not associated with victim ethnicity (see table 3). Although there were significant differences in story placement on a page by victim ethnicity (χ² = 13.6₍₆₎, p = 0.04), it is important to note that page placement could not be determined for 24% of the stories. Stories about black victims were least likely to be classified as “above the fold” (χ² = 9.65₍₁₎, p = 0.002), while stories about Asian victims were most likely to be classified “above the fold” (χ² = 5.90₍₁₎, p = 0.015). Homicides of child and gang victims were given the most visibility overall in the newspaper, although none of these associations were statistically significant. Gang homicide stories were more likely than others to appear in the first section of the newspaper (25.0%), to have over 500 words (44.4%), and to have photographs (37.0%). Child victim stories were more likely than others to be printed on page one (47.1%) and to be above the fold (70.6%).

As shown in table 4, white and Asian victims were more likely to have a story with a positive or more descriptive frame. (Although of substantive interest, none of these associations were statistically significant.) Articles about white homicide victims were more likely to discuss the victim's background (36.8%), have a hook (34.2%), and be opinion dominant (34.2%); stories about Asian victims were more likely to discuss the suspect's background (25.0%), have a thematic focus (21.4%), and have an emotional tone (42.9%). White (18.4%) and Asian (17.9%) victim stories were both more likely to include a discussion of background issues.

Overall, stories regarding child and gang homicide victims appear to have been framed with more of a human interest angle than those concerning other homicide victims. Child victim articles were most likely to contain background information about both the victim and the suspect, however only differences for suspect background were statistically significant (overall χ² = 29.5₍₈₎, p = 0.000; child victim χ² = 8.97₍₁₎, p = 0.003). Gang homicide articles were more likely to have a thematic focus and to discuss background issues, but only issue background differences were statistically significant (overall χ² = 17.1₍₄₎, p = 0.002; gang victim χ² = 12.91₍₁₎, p = 0.000).

In contrast, stories about intimate partner homicides were less likely to have humanistic story frames. These articles rarely discussed background issues (3.0%) or made use of a hook (9.1%). They were also rarely thematic (3.0%), opinion based (9.1%), or emotional in tone (15.2%). Statistically significant differences were found regarding issue background (overall χ² = 17.1₍₄₎, p = 0.002; intimate homicides χ² = 4.57₍₁₎, p = 0.03) and use of a hook (overall χ² = 11.8₍₄₎, p = 0.02; intimate victims χ² = 8.80₍₁₎, p = 0.003). Although overall comparisons for genre and tone did not reach statistical significance (both were p<0.10), specific comparisons of intimate partner homicide compared with other articles were statistically significant for genre (χ² = 6.01₍₁₎, p = 0.01) and tone (χ² = 6.61₍₁₎, p = 0.01).

Multivariate analyses

Consistent with the simple cross tabulation findings, newspaper coverage (that is, prominence and story frame) was not statistically associated with victim ethnicity when victim-suspect relationship was taken into consideration (see table 5). However, some of the findings may be of substantive interest. For example, stories about Asian victims were less likely to discuss the victim's background and less likely to be opinion based, adjusted odds ratio = 0.33 and 0.34, respectively, both with p<0.10.

In contrast, victim-suspect relationship predicted aspects of coverage more frequently, particularly the way in which a story was framed. Coverage differed most consistently for one type of relationship—intimate partner homicides. In comparison with stranger homicide articles, those concerning intimate victims had less overall visibility and more perfunctory story frames than the others examined. Nine out of 12 adjusted odds ratios indicated lower likelihoods of prominence and descriptive, humanistic story frames for intimate victim articles; three of these findings were statistically significant. Findings for other types of relationships will be presented first.

Stories about gang homicides tended to garner greater visibility and more descriptive story frames, with almost 75% of the variables examined having adjusted odds ratios greater than 1. Two of these findings were statistically significant. Gang homicide articles were almost 10 times more likely to discuss the relevant background issue (in this case, gang violence), and they were almost six times more likely to be printed in the first section of the newspaper (both p<0.05). In contrast, none of the examined variables were statistically significant for acquaintance homicides, although there was a trend for these stories to appear more frequently in the front section of the paper (p<0.10).

Child victim homicides also tended to be given greater prominence and more descriptive story frames, with two thirds of the variables examined having adjusted odds ratios greater than 1. Stories about child homicide victims were 6.5 times more likely to discuss the suspect's background (p<0.05); the suspect was generally a parent. Of substantive interest, these articles were more likely to address the relevant background issue of child abuse and to be printed on the top half of the page (both p<0.10).

The overall picture for intimate partner homicide stories stood in contrast to all other victim-suspect categories. Seventy five per cent of the examined variables had adjusted odds ratios less than 1, suggesting that overall, these articles were less visible and less likely to have a descriptive, humanistic story frame. While no prominence variables were statistically significant, intimate homicide articles tended not to have photos (p<0.10). Three of the story frame variables were statistically significant at p<0.05. Stories about intimate partner homicide victims were less likely to be emotional, less likely to be opinion dominated, and less likely to make use of a hook to draw in the reader (adjusted odds ratios = 0.23, 0.23, and 0.13, respectively).

Although “focus” was not statistically significant, it is important to note that just one out of the 33 articles about intimate partner homicide had a “thematic” rather than an “episodic” story frame; the headline for this article was “When women abuse their spouses”. Thus, despite the fact that most of the intimate homicide victims in these articles were women, the one article with a thematic story frame focused on male victims of abuse. The article stated that “95% of domestic abuse victims are women”, then focused on the fact that the woman (in the article) accused of killing her husband “contradicted the statistics”.

For about one third of the intimate homicides, the article reported additional details about the relationship between the victim and suspect: they were estranged (21%), the victim was leaving the suspect (6%), and the victim had written but not filed a restraining order against the suspect (3%). Relationship problems or acute arguments were reported in just over one third (36%) of the intimate homicide stories.

DISCUSSION

Newspaper coverage of homicide

One factor commonly believed to be related to news coverage is ethnicity. Previous research has documented the association between victim ethnicity and whether a homicide is covered at all.^12,14 Other work has found that story frames often differ when comparing stories about risk for black and white victims.²⁰ We found no such association for the nature of homicide coverage. Ethnicity of the homicide victim, when victim-suspect relationship was taken into consideration, was not related to the nature of coverage in a large, metropolitan newspaper.

Homicide reporting differed by victim-suspect relationship, even when victim ethnicity was taken into account. The most consistent differences were observed in how articles about homicides by intimate partners compared to those about homicides by strangers. Gang related and child homicides were also covered differently, while those committed by acquaintances were covered very similarly to homicides by strangers. Stories about intimate homicide victims were less likely to be given a human interest angle through the use of opinions and emotional tones, or by opening with sentences that would hook the reader. Articles about gang homicides were more likely to appear in the first section of the newspaper and also to address broader issues related to the homicide. While 37% of gang homicide articles addressed the issue of gang violence, only one of the 33 intimate partner articles addressed domestic violence as a social issue. Moreover, the sole story that discussed domestic violence as an issue focused on women's violence against men, a type of violence that is less likely to result in injury than men's violence against women.

The nature of news media coverage generally reflects a combination of economic pressures and the professional values of the journalist.¹⁷ Once they decide to cover a homicide, the Los Angeles Times appears to do so in a comparable manner across ethnic groups and victim-suspect relationship, with intimate homicides being the main exception. Newspapers may be reluctant to “aggressively” cover stories about domestic violence due to the belief that domestic violence is a private or family matter.^21,22 Unfortunately, the subjective choices made in media coverage can lead to unrealistic perceptions and biases regarding the issue of violence against women, some of which tend to blame the victim or otherwise shift ideas about responsibility.^22–28

Because this study reports on news articles that were published before the O J Simpson case, one might wonder whether news reporting on domestic violence has shifted since the news media rediscovered domestic violence.²² Maxwell and colleagues found that while the amount of coverage regarding domestic violence did increase after the murders of Nicole Brown-Simpson and Ronald Goldman, the nature of the coverage did not differ.²⁹ Before their murders, 23% of the articles primarily about domestic violence were socially rather than episodically focused; a percentage that remained consistent afterwards. Accordingly, Maxwell et al also found that most coverage continued to focus blame and responsibility on individual rather than social solutions.

Whereas the amount of media coverage may be proportional to the level of attention an issue receives, the nature of the coverage is important in determining the type of attention it gets. News story frames influence how people think about issues and how they assign responsibility for causes and solutions.^18,30 Responsibility for solving problems is assigned more often to government or society when an incident is discussed in its broader social context than when it is described as an isolated event.¹⁸ Thus, from a public health perspective, a thematic frame is desirable. Nonetheless, episodic coverage is five times more common than thematic coverage of violence in television news.¹⁹ Stories that draw in the reader, evoke sympathy, and present opinions are also more likely to garner attention and interest. Not only can framing have an impact on public opinion, but it can also have a direct or indirect influence on policy decisions (for example, Gandy et al,²⁰ Feigenson and Bailis,³¹ and Jones³²).

Study limitations and future research

Many factors may contribute to how a story is reported in the news media. The focus of the present investigation focused on two key victim variables. Whereas victim-suspect relationship was a consistent predictor of the nature of homicide coverage, it is important to keep in mind that a number of potential confounders were not examined. Two obvious ones are gender and number of persons killed in an incident. Whereas only about a third of the homicide victims overall were female, the vast majority of intimate victims were female, which may explain part of the findings. Although most of the homicides had just one victim, intimate homicides had the highest percentage of multiple victims. Associations between potential covariates (for example, female gender and intimate partner homicides) precluded simultaneous examination of other factors.

Future research using a different design and a larger sample size may benefit from examining multiple potential covariates simultaneously. For example, previous research found that while intimate partner homicides received less newspaper coverage than expected, homicides with female victims were more likely to receive coverage.¹⁴ In addition, the present study examined stories in one large, metropolitan newspaper during a five year period. Future research involving newspapers from other locales and other time periods will help elucidate whether and how news media coverage relates to characteristics of a homicide.

IMPLICATIONS FOR PREVENTION

In summary, the nature of homicide reporting is critical because the news media can influence public beliefs and perceptions, help establish specific agendas, and influence public policy. Differences in the amount and nature of news coverage are likely to have a corresponding influence on the perceived importance of and potential solutions for an issue. Given the findings of this and other studies, for example Dorfman et al¹⁹ and Maxwell et al,²⁹ one could expect members of the general public not to take a population based or public health view in considering the basis of and potential solutions for violence. Given that the primary goal of commercially-run media, even news media, is not necessarily education, it is especially critical that public health professionals work to frame stories in a way that will garner media attention.^9,10,33–36 Violence prevention program and policy strategies are likely to gain more attention, support, and resources when efforts are also made to help journalists frame their stories in a way that invokes public health notions of causality and supports public health solutions to violence.

Before discussing an issue with the media, it is critical for public health professionals to consider how they want their issue to be framed and ultimately portrayed to the public and policy makers. While journalists are the decision makers regarding how to write a story, guidance from public health professionals may help to invoke a public health framework. Some critical elements to consider when deciding how to frame a story include the attributed cause(s) of the problem, the effects of the problem and/or proposed solution, who supports and opposes these viewpoints, who is responsible for the solution(s), and the credibility of the information sources.³³ For specific examples and case studies on framing public health issues including violence, see Wallack et al,⁹ Chapman and Lupton,³³ and the Berkeley Media Studies Group.³⁴