RACE AND ETHNIC REPRESENTATIONS OF LAWBREAKERS AND VICTIMS IN CRIME NEWS: A NATIONAL STUDY OF TELEVISION COVERAGE

Abstract

Research on racial-ethnic portrayals in television crime news is limited and questions remain about the sources of representations and how these vary for perpetrators versus victims. We draw from power structure, market share, normal crimes, racial threat, and racial privileging perspectives to further this research. The reported race or ethnicity of violent crime perpetrators and victims are modeled as functions of: (1) situational characteristics of crime stories; and (2) contextual characteristics of television market areas. The primary data are from a stratified random sample of television newscasts in 2002–2003 (Long et al. 2005). An important innovation of our work is the use of a national, more generalizeable, sample of local news stories than prior researchers who tended to focus on single market areas. Results indicate that both the context of the story itself and the social structural context within which news stories are reported are relevant to ethnic and racial portrayals in crime news. We find limited support for power structure, market share, normal crimes and racial threat explanations of patterns of reporting. Racial privileging arguments receive more extensive support.

Media representations of crime shape public opinion in important ways, including through the frequency with which, and how they present criminal participants and victims. For example, views of the nature of the crime problem and who, or what, is responsible for said problem in a locale may be shaped by the extent to which specific groups are over- or under-represented as perpetrators or victims in crime news relative to other groups or their share of criminal involvement or victimization. On the one hand, if media sources overrepresent certain groups (e.g., males, people of color, etc.) as perpetrators, this may promote racial or gender stereotypes or reinforce public hostility toward such groups (e.g., Barlow, Barlow, and Chiricos 1990; Dixon, Azocar and Casas 2003; Dixon and Linz 2000a, 2000b; Russell 1998). On the other hand, overrepresentation of the victimization of certain groups (e.g., females, Whites, etc.) may promote misleading views of what populations are the most vulnerable to crime, or who should fear crime. If media sources shape public opinion in these ways and such opinion has its counterpart in the development of criminal justice policies, differential treatment may be the result (e.g., harsher penalties for crimes that are more typically committed by certain perpetrators; greater attention to reducing the victimization of members of certain groups who may be erroneously perceived as being at greater risk for victimization) (Bobo and Johnson 2004; Russell 1998).

In light of their potential influence, researchers have sought to assess empirically the nature and outcomes of media representations of crime. Further, because of the assumed differential group impact of such representations, a number of scholars have directed their attention to how racial and ethnic groups are portrayed in crime news stories. Several research issues have been explored: (1) the extent to which news coverage represents actual patterns of participation in crime by different ethnic and racial groups (e.g., Gilliam et al. 1996; Lundman 2003, 2004); (2) the manifest content of racial portrayals and/or the extent to which these involve racial typifications (e.g., Chiricos and Eschholz 2002; Gilliam et al. 1996; Lundman 2003, 2004); and (3) the impact of race and ethnic portrayals on audiences' concerns about crime and views of who, or what, is responsible for the crime problem (e.g., Altheide 1997; Dixon and Linz 2000b; Entman 1992; Entman and Rojecki 2000).

Despite the proliferation of studies, findings about the nature of media representations of crime vis-á-vis race-ethnic groups are not straightforward. Research provides some evidence that racial minorities (especially, African Americans) are overrepresented in news stories focused on perpetrators of violent crime (e.g., Dixon and Linz 2000a). However, this is not a uniform finding across investigations. For example, some studies indicate that the representation of members of subordinate populations as crime perpetrators in news media is consistent with their “real” share of arrests or offending (e.g., Chiricos and Eschholz 2002). Others indicate that Whites rather than Blacks or Latinos are overrepresented as perpetrators (e.g., Dixon et al. 2003). To date, research has not provided a clear explanation of these disparate patterns.

Such inconsistent findings may result from several limitations of extant work. First, with some notable exceptions (Altheide 1997; Chiricos and Eschholz 2002; Dixon and Linz 2000a, 2000b; Dixon et al. 2003; Poindexter, Smith and Heider 2003; Sacco 1982), existing work examines news coverage in the print media. This is true despite the fact that in the contemporary United States people get most of their news from television, and often regard television as the most important, credible, and trustworthy news source (GlobeScan 2006). Given this, efforts to assess ethnic and racial portrayals in news media should consider television news in detail. Second, investigations of crime news coverage have often examined news in one local area (Chiricos and Eschholz 2002; Dixon and Linz 2000a, 2000b; Entman 1992; Sacco 1982; Sheley and Ashkins 1981). This emphasis comports with the recognition that local news tends to be more influential in determining individuals' perspectives than national news (Dorfman, Thorson and Stevens 2001; Liska and Baccaglini 1990; Poindexter et al. 2003). Yet, considering a single news market makes it difficult to fully examine the role of social structural context in outcomes, and to determine the extent to which patterns found in individual places hold when the full range of variation in important conditions is considered.

Third, research on the representation of race and ethnic groups in news coverage has tended to focus on the consequences of stories (racial stereotyping, fear of crime, hostility toward groups, equation of news coverage with world reality). This focus is understandable in view of contemporary interests in: (1) reducing sensationalism and increasing accuracy and fairness in news coverage (Gilliam et al. 1996; Heider 2000; Klein and Naccarato 2003; Rosenstiel, Gottlieb, and Brady 2000); and (2) setting news stories in their broader context to foster improved understanding of social problems like crime (Coleman and Thorson 2002; Dixon et al. 2003; Iyengar 1991; Rogers and Thorson 2001). Yet, properly addressing these and other concerns may depend upon having a more fundamental understanding of the situational and contextual reasons for over-and/or under-representation of different groups in news about crime. To address this possibility, greater attention must be given to the factors that predict crime news portrayals by ethnicity and race. Answers are needed to a variety of questions. In what types of contexts are particular race and ethnic groups inaccurately versus accurately portrayed in news? Do media distinguish among representations of males versus females? Are race and ethnic representations equally distorted or accurate for homicides and other types of violent crime?

Given our limited knowledge regarding such issues, the research reported below examines the predictors of ethnic and racial representations of perpetrators and victims of violent crime in a stratified random sample of news reports drawn from local television stations across the United States as well as from broadcasts of news on four national networks: ABC, CBS, CNN, and NBC. Including both local and national network news stories in the analysis helps to ensure that our sample represents the breadth of market sizes in which television stations operate, as well as the full range of violent crime news stories to which viewers are exposed. Relying on these news reports, we ask: (1) how do situational components of crime stories and aspects of the social structural context, such as the race and ethnic composition, and level of violent crime in the area, influence variation in outcomes; and (2) to what degree do such predictors have similar or different effects for representations of perpetrators versus victims.

<A>PAST RESEARCH AND REMAINING ISSUES

What answers has research provided regarding the nature and predictors of the representation of race and ethnic groups in television news media? One approach to this question assesses the extent to which news stories featuring particular groups have crime as the central theme. Robert M. Entman took this approach in his studies of local news programs in Chicago (Entman 1992) and of network television news (Entman 1994). Both studies revealed that crime is one of the most common subjects of stories featuring Blacks (one of the top two subjects in local news in Chicago, and one of the top three in national news stories). The studies also revealed other interesting patterns. For example, the evaluation of local television news in Chicago demonstrated that when suspects are mentioned by name, 72 percent are White and only 28 percent are Black. And the study of network news found that among crime stories in which Blacks are portrayed as perpetrators, 77 percent are violence- or drug-related, while this was true for only 42 percent of crime stories featuring White perpetrators (Entman 1994).

Paula M. Poindexter, Laura Smith and Don Heider (2003) assessed how a variety of race and ethnic groups were featured in local television news stories for 26 stations in 12 cities for periods during the late-1980s through 1998. Among other issues, they sought to address the question: “What is the presence of people of color and how are they framed in local television news” (2003: 328).¹ They found that Asians, Latinos, and Native Americans were seldom the subjects of television news reports. Further, although African Americans were not an uncommon subject of television news stories, they “were more likely to be newsworthy because they had committed a crime” (Poindexter et al. 2003:533). Notably, crime was the most common topic of news stories (22 percent), with 60 percent of these indicating the perpetrator's race or ethnicity. Approximately half the stories report offenders' race as White (51 percent) and 41 percent as Black. However, Black- focused stories were more than two times as likely as White-focused stories to be about crime (i.e., 69 percent and 28 percent, respectively).

Theodore Chiricos and Sarah Eschholz (2002) examined the criminal typification of race and ethnicity in television crime news in Orlando in 1998. They found two patterns that are indicative of this type of typification. First, their analysis revealed that Blacks and Hispanics are disproportionately portrayed as criminal suspects in television news stories. Here, the disproportionality is greater for Hispanics than for Blacks. More specifically, Blacks were nearly two and a half times and Hispanics over five and a half times more likely to be presented in roles as criminal suspects than Whites. Second, Chiricos and Eschholz's assessment of the relative frequency of the portrayal of race and ethnic groups as criminal suspects versus as victims or in conventional roles (e.g., as police officers, reporters, or other positive role model) revealed that the two groups of color are four times more likely to be portrayed as suspects than as victims of crime, while Whites are about evenly portrayed in these two roles. Further, Whites were more likely to appear as role models than as criminal suspects, Blacks appeared about evenly in these two types of roles, and Hispanics were much less likely to appear as role models than as criminal suspects.

In brief, studies by Entman (1992) and Poindexter et al. (2003) underscore that crime is frequently covered in the news, and show that compared to Whites, racial and ethnic minorities are disproportionately likely to be covered in crime stories. Chiricos and Eschholz (2002), along with Poindexter et al., show that the frequency with which minorities (Blacks and Hispanics) are portrayed in sympathetic (as crime victims) or conventional roles relative to being portrayed as criminal suspects is greater than that for Whites. However, none of these studies assesses the predictors of such differential representation.

Prior research has also given attention to how extensively subordinate race and ethnic groups are portrayed in television crime news relative to their “actual” participation in crime. In addition to examining the criminal typifications of race and ethnicity, Chiricos and Eschholz (2002) examined racial and ethnic typifications of crime by quantitatively assessing the differential representation of Whites, Blacks and Hispanics as crime suspects in local television news relative to the groups' representation in Orlando's population and/or as criminal arrestees Their analysis revealed that, in general, neither Blacks nor Whites are overrepresented in television crime news relative to one another or to their proportions in the population; Blacks are underrepresented, and Whites slightly overrepresented compared to their arrest proportions. Hispanics were found to be overrepresented in crime news relative to their proportion in the population. However, data were not available to determine Hispanics' representation in news compared to their levels of arrest. Based on a qualitative assessment, Chiricos and Eschholz discovered that within crime news stories, Blacks and Hispanics are represented as more threatening than their White counterparts.

Travis L. Dixon and associates (Dixon and Linz 2000a, 2000b; Dixon et al. 2003) focused on the “accuracy” of coverage of race and ethnic groups in local crime stories. Some of their findings are consistent with the hypothesis that compared to Whites, Blacks and other minorities are likely to be overrepresented in crime stories, and in more negative ways. However, the results depend upon the standard of comparison (see below), whether the stories portray perpetrators or victims, the specific group (e.g., Blacks or Hispanics) under consideration, and the types of crime being covered. Dixon and Linz's (2000a) analysis of local television news in Los Angeles indicate that Blacks, Whites, and Hispanics may be overrepresented, underrepresented, or accurately portrayed in television news depending upon the particular aspect of overrepresentation in question: intergroup, one group compared to another group (e.g., Blacks to Whites); interrole, whether a group is portrayed more frequently as perpetrators versus victims; or interreality, representation of a group in the media compared to its representation in crime statistics. Applying the intergroup standard, Dixon and Linz found that Whites were more likely than African Americans and Hispanics to be portrayed as victims, and this is true whether homicide or another type of crime is examined. Their interreality comparison revealed a complex picture. Blacks were overrepresented as perpetrators of homicide, while Latinos along with Whites were underrepresented in this role. By contrast, Whites were overrepresented and Latinos were underrepresented as victims of homicide, but Blacks were accurately represented, not overrepresented, in this role. Finally regarding interrole comparisons, Dixon and Linz found that that Blacks and Latinos were more likely to be portrayed as perpetrators than as victims, while Whites were more likely to be portrayed as victims than perpetrators.

Dixon and associates (2003) analysis of national network television news programs also revealed complex patterns regarding the depiction of ethnic and racial groups. They found that Whites are more likely than Blacks or others (Hispanics, Asians, and Native Americans) to be shown as either perpetrators or victims. Their interrole analysis showed that all racial groups are more likely to be portrayed as perpetrators than as victims, and their interreality comparisons showed that the portrayal of racial groups as perpetrators is on a par with their rates of arrest. However, Whites are overrepresented as victims compared to statistics on their levels of victimization, while Blacks are underrepresented in this role.

Overall, two conclusions seem warranted from existing studies. First, research on television news portrayals of race-ethnic groups is relatively sparse, and existing studies are not necessarily designed to explain the observed patterns of ethnic and racial representation. Rather, descriptions of how groups are portrayed in television news are often preludes to a focus on the social-psychological or other consequences of such depictions. Second, studies make it clear that racial and ethnic portrayals in television news cannot be described in simple terms. Instead, patterns appear to be quite contingent (on the nature of the comparison, the group under consideration, and perhaps regional location) in ways that call for explanation.

Below, we take up the task of investigating the determinants of ethnic and racial representations in television crime news by conducting a multivariate logistic regression analysis of the race or ethnicity of crime victims or perpetrators as reported in news stories. To the best of our knowledge, prior research has not examined how specific substantive factors influence race or ethnic reporting of crime subjects net of one another. This study is innovative in several additional ways. First, the analysis is based on a national sample of local news stories. Prior work has emphasized the importance of local news, and focused on coverage of crime news in a single market area (e.g., Chiricos and Eschholz 2002; Dixon and Linz 2000a, 2000b; Entman 1992; Gilliam et al. 1996; and Pritchard and Hughes 1997). However, because single market area studies vary in what they reveal about the degree to which different race and ethnic groups are represented in local news coverage, it is important to consider the contribution of social context to media depictions. Examining local news stories in a variety of areas also allows for consideration of the full range of variation in predictors. Poindexter and associates (2003) study is an exception to the focus on a single market. Recall that they examine race and ethnic representations in television news coverage for 26 different stations in 12 different cities. However, Poindexter et al.'s sample is not representative. As such, they call for additional work that addresses the external validity question raised by prior work, including their own, by “increasing the number of markets and news stories and randomly selecting local television newscasts from all television stations across the United States” (2003:234). Our work attempts to heed this call.

Second, our analysis treats perpetrators' and victims' race or ethnicity as outcomes. Prior studies make it clear that the portrayal of groups in crime news varies depending upon these roles. Yet, there is no systematic research examining the comparative structural factors that explain why minority perpetrators are characterized differently from minority victims in television news. The data used in this study allow us to begin to fill this gap in the literature. Third, we follow Chiricos and Eschholz (2002), Dixon and Linz (2000a), and Poindexter et al. (2003) in considering television news representation of Whites, Blacks, and Hispanics. In part because the sample of stories examined are often from a single local area, some prior studies have tended to compare the representation of Blacks versus Whites only (e.g., Entman 1992, 1994). However, despite their common minority status, there are real differences in the positions of, perceptions about, and treatment of Blacks and Hispanics in U.S. society (e.g., Bobo and Hutchings 1996; Bobo and Suh 2000; Charles 2003, 2006; Schuck, Lersch, and Verrill 2004; Steffensmeier and Demuth 2000), which may also be reflected in crime news stories. Thus, it is important to undertake multivariate studies that simultaneously consider these three major race and ethnic groups.

<A>EXPLAINING PATTERNS OF RACE-ETHNIC GROUP REPRESENTATIONS AS PERPETRATORS OR VICTIMS IN TELEVISION CRIME NEWS

We draw on power structure, market share, normal crimes, racial threat, and racial privilege arguments in developing models of race-ethnic representations in television crime news. The power structure argument focuses on how the race or ethnicity of those with power in the media affects story selection, while the market share and normal crimes approaches are grounded in ideas about how aspects of crime make such incidents more or less newsworthy. As such, these perspectives underlie some prior studies of the nature and outcomes of media representations of crime. Racial threat and racial privilege themes are rooted in sociological and criminological traditions regarding the provision for, and imposition of, higher or lower levels of criminal punishment for different populations. Together this set of orientations draw attention to: (1) the situational components of criminal incidents that may help to account for which race-ethnic groups are reported in crime news; and (2) aspects of the social structural context that may shape the portrayals of groups in these reports. We discuss these approaches in turn to set the stage for hypotheses about the predictors of race-ethnic representations in television crime news. Because many of our arguments contrast the representation of race and ethnic groups as perpetrators versus victims, it is important to keep in mind throughout that crime is predominantly intra-racial in nature; Whites are most likely to victimize Whites, Blacks tend to victimize other Blacks, and so on.

Some scholars emphasize news organizations' power structure in assessing whether and how features of a crime make a story newsworthy. The vast majority of print and television media is owned by Whites and employ Whites as reporters and anchors, positions which typically encompass the power to affect story selection (Klein and Naccarato 2003; Poindexter et al. 2003). Thus, the biases and prejudices of Whites may be reflected in media output for marketability or simply because these representations are encompassed by the world-views of those who make the determination of what to cover and how to cover it (Heider 2000; Klein and Naccarato 2003; Rosenstiel et al. 2000). If so, it would be reasonable to expect that race and ethnic minorities (as members of subordinate populations) would be overrepresented as offenders in crime stories but underrepresented as victims.

The market share perspective makes a different argument about what drives story selection and suggests partially contrasting expectations. Due to intense competition among news stations for market share, media staff may face strong pressure to focus on stories that draw viewers, advance reporters' careers, or draw on important sources such as law enforcement (Chermak 1994; Sheley and Ashkins 1981). Such stories may be newsworthy according to one or more of the following criteria identified by Pamala Shoemaker and Stephen Reese (1996): timeliness, proximity, novelty, prominence, conflict, magnitude, and human interest. Thus, if treatment of groups in the news depends upon their “newsworthiness” as determined by novelty and human interest, then it seems possible that crime stories involving more powerful groups as offenders (or victims) would likely be judged as more newsworthy than those involving populations that are more often associated with crime (i.e., minorities).

The concept of “normal crimes” was introduced by Sudnow (1965) in recognition of the fact that many types of criminal incidents can be characterized as involving a “normal” set of circumstances, with typical accompanying social characteristics (e.g., race, gender, and social class) of offenders and victims. He argued that such crimes likely garner little special attention, and instead are processed as a matter of routine in an efficient, businesslike fashion using a plea bargain and the “going rate” for penalties. By contrast, justice officials direct more attention to crimes with more “unusual” offenders, victims and circumstances. David Pritchard and Karen D. Hughes (1997) offer a normal crimes type perspective in exploring news coverage of homicide. They argue that deviance is a prime component of newsworthiness. In their analysis of newspaper reports of homicides in Milwaukee County, Wisconsin, they identify and examine the extent to which three forms of deviance (statistical, status, and cultural), or departures from normal, are relevant to how characteristics of a crime influence its newsworthiness. First, unusual or statistically deviant events are more likely to receive coverage than routine criminal incidents. Second, for Pritchard and Hughes, “news is about the powerful.” Therefore, to the extent that the status of participants in crime (suspects or victims) departs from high status, white males (status deviance), the less news attention the crime will receive. Third, crimes are likely to receive more attention to the extent that they depart from dominant cultural norms about what is normal, healthy or clean (cultural deviance), as when participants are involved with drugs or gangs, or victims are regarded as especially vulnerable (e.g., children, women, the elderly). From Pritchard and Hughes' perspective, statistical deviance and status deviance both portend greater reporting of White perpetrators or victims compared to their African American or Hispanic counterparts (for whom street crime is more “normal”). Because cultural deviance is a function of acts, the race (and presumably ethnicity) of participants is not theoretically relevant to cultural deviance newsworthiness.

Racial threat arguments have long been articulated, and have gained prominence as an explanation of race relations (see, Blalock 1967). This perspective emphasizes that dominant groups view large minority populations as threatening their advantaged positions with respect to political ascendancy, finite economic resources, and freedom from minority-on-White crime. In light of this racial threat, it is hypothesized that the dominant group (i.e., Whites) takes steps to reduce encroachments on its prerogatives, which results in more pronounced prejudice and discrimination in settings where minority populations are larger. Criminologists have drawn on racial threat arguments in investigating a variety of outcomes that reflect the demand for control and punishment of subordinate groups e.g., police size and expenditures, sentence severity, use of the death penalty (e.g., Baumer, Messner and Rosenfeld 2003; Behrens, Uggen and Manza 2003; Earl, Soule, and McCarthy 2003; Eitle, D'Alessio and Stolzenberg 2002; Harring 1983; Jacobs and Carmichael 2001; Jacobs and Helms 1999; King and Wheelock 2007; Stolzenberg, D'Alessio and Eitle, 2004). In these investigations, crime control initiatives and harsh sanctions are seen as mechanisms by which dominant groups respond to threats imposed by large and growing subordinate populations. To our knowledge, the racial threat hypothesis has not been applied in examining racial and ethnic representations in crime news. Yet, as a vehicle that often reflects the world views of audiences, news media might well be a means by which dominant groups express claims for criminalization of members of groups perceived as encroaching upon its privileges. If so, this should be reflected in a greater representation of Blacks and Hispanics as offenders (and Whites as victims) in places with larger proportions of these minority groups.

Related to racial threat, we also anticipate that TV news stories in places with more extensive violent crime problems will more likely portray Blacks and Hispanics as perpetrators and less likely portray them as victims than their White counterparts. To elaborate, criminologists and race scholars make it clear that in the contemporary period, concerns about certain types of crime have become particularly salient to racial threat as a factor influencing crime and justice outcomes (Bontrager, Bales and Chiricos 2005; Quillian and Pager 2001; Russell 1998; Wilson 1992). Violent crime, along with drug and gang crime has become closely associated with racial and ethnic minorities (especially African Americans and Hispanics). Indeed, Russell's assessment of media portrayals and public images of Blacks suggest that, with respect to African American males, racial status and criminal status have become fused as reflected in her reference to the “criminal blackman.” Lincoln Quillian and Devah Pager (2001) show that the size of the young Black male population in a neighborhood is positively associated with perceptions of neighborhood crime levels. And, James Q. Wilson (1992) notes that fear of crime is a major reason that Whites prefer not to live in communities with large Black populations. Thus, in the context of this study, levels of violence in television market areas may be particularly germane to the representation of African Americans and Latinos compared to Whites in television crime news.

As used in criminological literature, the notion of racial privilege draws attention to the salience of characteristics of crime victims in how seriously victimization is treated and in the outcomes of justice processing (Bachman 1996; Baumer, Messner, and Rosenfeld 2000; Bontrager et al. 2005; Hawkins 1986; Kleck 1981; Maxwell, Robinson, and Post 2003; Williams, Demuth and Holcomb 2007). Such arguments assert that the justice system operates in a manner that takes more seriously crimes involving White victims than those affecting other groups, especially if Whites' victimization is at the hands of non-Whites. Thus, crimes against White victims are investigated with more intensity and punished more severely than crimes against victims of color. Put differently, victimization of persons of color is devalued (seen as representing less loss or harm to the community) while White victims are privileged (recognized as having incurred serious loss with harm to the community). In the media literature, a somewhat parallel devaluation interpretation has been given to the finding that newspapers underreport the victimization of minorities compared to actual crime reports (see, for example, Chermak and Weiss 1998; Johnstone, Hawkins, and Michener 1994; Peelo et al. 2004; Sorenson, Manz and Berk 1998). To the extent that this type of devaluing and privileging of the victimization of groups occurs, it should be reflected in the differential portrayal of racial and ethnic groups in crime news. Specifically, there should be greater representation of White victims than of Black and Hispanic victims.

The historical legacy of slavery in the United States includes the privileging of White female victims believed to be violently attacked (most notably in the form of rape or sexual assault) by Black males. In contrast, violence against Black women by either White or Black men did not provoke high levels of outrage or punishment (Karmen 2004). According to some, this legacy of racial privilege for White females continues to affect criminal justice decisions and processes (Price and Sokoloff 2004; Wriggins 2004). It may also affect which groups are represented in crime news.

To summarize, we draw on the above themes in proposing that the representation of race and ethnic groups in television news coverage is a function of situational characteristics of the crimes, and contextual characteristics of the local Designated Market Areas in which crime stories are presented. Designated Market Areas, or DMAs, refer to the geographic areas defined by the Nielsen Media Research Company as making up particular television markets. They consist of groups of counties that comprise the major viewing audience for the television stations located in particular metropolitan areas. DMAs do not overlap one another, and every county in the United States belongs to only one DMA (Long et al. 2005). Yet, positing the directions of effects of situational and contextual variables is not straightforward because the approaches sometimes lead to contradictory expectations. With this caution in mind, we pose the following hypotheses.

Baseline models

In baseline models of ethnic and racial representations in crime news, power structure, racial threat and racial privilege arguments would predict that Blacks and Hispanics will be overrepresented as offenders and underrepresented as victims. Conversely, Whites will be underrepresented as offenders but overrepresented as victims. Market share and normal crimes themes suggest the contrasting expectation that Whites will be overrepresented as both perpetrators and victims in news coverage.

Situational influences

Consistent with power structure and racial privilege arguments, crime incidents involving violent index offenses, multiple victims, and female victims will be associated with a greater likelihood of reporting minority perpetrators (particularly Blacks who are disproportionately involved in violent index crimes) as well as White victims. These situational features make crime stories more newsworthy and provide a vehicle for White owners of news media to express privileging of more valued victims while protecting White offenders, and presenting minority perpetrators as particularly dangerous in committing violent crimes, victimizing vulnerable subjects, or committing especially heinous offenses such as those with multiple victims. As a practical matter, privileging in our analyses thus may be seen in the absence of reporting White perpetrators. In contrast, a market share approach would suggest that situational factors which create newsworthiness would be used by decision-makers to cater to white audiences (as the largest market share) and increase the representation of whites as both offenders and victims. Neither the “normal crimes” nor racial threat approaches speak to the influence of these situational factors.

Social structural context

In line with racial threat arguments, crime news in DMA contexts with higher rates of violent index crime and higher percentages of Blacks and Hispanics will be associated with a greater likelihood that members of groups of color will be represented as perpetrators. Market share arguments suggest that higher percentages of minority populations will increase the reporting of both minority perpetrators and victims. Conversely, in these contexts, racial privileging will be evident in a greater representation of Whites as victims. The other approaches do not argue a role for such contextual characteristics.

Population size and region are treated as control variables in the analyses.

<A>DATA AND METHODS

<B>Data, Sample, and Units

Data for the analyses were collected by Long and associates (Long, et al. 2005). They consist of a stratified random sample of television news stories broadcast in 2002 and 2003. Therefore, we can confidently make inferences only about coverage during those years. However, the literature provides no empirical or theoretical reason to believe that patterns of coverage of race in reports of crime is likely to have substantially changed over the few intervening years in a national sample of television news coverage. The Long et al. data are used to measure the dependent and independent variables characterizing the news story. Long, Slater, and their associates provided the data to us as a secondary data file with variables already coded.² Notably, the news stories were content coded by two trained graduate students under the auspices of the lead researcher, who met regularly with them to refine the coding scheme and discuss coding discrepancies. Our analyses are based on stories that make explicit mention of current, specific violent crimes. Coders excluded stories about: violent crime as a general phenomena without referencing a specific crime; policy or program stories; stories about civil cases (unless they involved a violent crime, which is rare); terrorism; sports violence occurring among participants during an event; government-sponsored violence; historical violent crimes (e.g., well-known historic assassinations); or uncontrolled animal attacks. As reported in Slater, Long, and Ford (2006), identification of violent crime stories was double-coded on randomly-selected subsamples at the outset and when 25 percent, 50 percent, and 75 percent of the programs had been coded. Cohen's kappas for identification of qualifying stories ranged from .76 to over .9.

The sampling frame for selecting newscasts was stratified by both date and geographic location. Regarding date, a constructed month (four weeks) was created for each year (Riffe, Lacy and Fico 2005), stratified by quarter and day of week to ensure that the sample of stories is unaffected by the seasonality of news events and coverage decisions.³ For each selected date, a stratified random sample of Designated Market Areas was taken.⁴ These DMAs were divided into six strata by market size (number of households) so that each stratum contained approximately the same number of households. As such, the first stratum includes the three largest markets (leading to a greater likelihood of these large markets being included in the analysis), while the last contains 119 markets. In the end, the percent of the population served was roughly equivalent across strata, 15–18 percent each (Long et al. 2005:312). On each selected date, one DMA per stratum was picked at random. The evening news broadcast from one local TV station (chosen to balance network affiliations across the six DMAs) was selected from within each sampled DMA. Stories from the evening news broadcasts for all four national networks (ABC, CBS, CNN, NBC) were randomly sampled separately from the Vanderbilt Archives. Thus, on each day, broadcasts from each of the four national networks were sampled in addition to one broadcast from each of the six DMA strata. We recognize that CNN is solely a national news network while ABC, CBS and NBC broadcast both national and local news. However, because we control for whether or not stories are from national versus local broadcasts, we do not believe this to be problematic (See Appendix A for a list of the DMAs included in the analysis and their frequencies).

To define the geographic correspondence between DMAs and the geographic units of the data sources for contextual measures of population composition and crime rates, we use information from the Broadcasting and Cable Yearbook 2002–2003 (R.R. Bowker 2002) to determine the county composition of the DMAs. Each DMA is composed of a county or group of counties in which the largest share of viewing is to stations located in that same market area. They vary in population size from just over 12,000 to 16.8 million (R.R. Bowker 2002). Non-overlapping DMAs cover the entire continental United States, Hawaii and parts of Alaska. Whole counties are used in defining DMAs and each county is assigned to only one DMA (Neilson Media Research 2009). Population data are from the 2000 Census of Population (U.S. Bureau of the Census 2000), with county-level counts aggregated to form estimates for DMAs. Data for constructing rates of criminal offending are from the Federal Bureau of Investigation's (FBI) “Uniform Crime Reports” (UCR) for 2002 and 2003 (U.S. Department of Justice 2004, 2005).

<B>Measurement and Operationalizations

Table 1 presents the operationalizations and descriptive statistics for the variables in our analysis. We discuss the measurement of the outcomes and predictors in turn.

Table 1.

Descriptive Statistics and Operationalizations for all Variables (N=660).

Variable	Operationalization	Descriptive Statistics
Outcomes
Ethnicity of Perpetrator in Broadcast	The ethnicity of the perpetrator in the broadcast coded White, Black, Hispanic, Other, or Not indicated.	Percent
	White	26.82
	Black	21.06
	Hispanic	6.97
	Other	2.88
	Not indicated	42.27
Ethnicity of Victim in Broadcast	The ethnicity of the victim in the broadcast coded White, Black, Hispanic, Other, or Not indicated.	Percent
	White	27.88
	Black	9.24
	Hispanic	3.79
	Other	2.12
	Not indicated	56.97
Predictors
Story Content		Mean	Standard Deviation
Violent Index Crimea?	The crime was a violent index crime.	.902	.298
Sexual Assault?	The crime was a sexual assault.	.161	.367
Multiple Victims?	The crime had multiple victims.	.368	.483
Female Victim?	The victim was female.	.362	.481
White Female Victim?	The victim was a white female.	.168	.374
DMA Context
Percent White in the DMAb	Percent of the population that identifies as non-Hispanic white in the Designated Market Area.	68.275	15.655
Percent Black in the DMAb	Percent of the population that identifies as non-Hispanic black in the Designated Market Area.	11.988	8.528
Percent Hispanic in the DMAb	Percent of the population that identifies as of Hispanic ethnicity of any race in the Designated Market Area.	13.437	13.104
Percent Other Racec in the DMAb	Percent of the population that identifies as non-Hispanic other race in the Designated Market Area.	6.300	4.435
Violent Index Crime Rated	Violent index crime rate in the Designated Market Area.	495.184	166.740
Population	DMA Population	55,600,000	108,000,000
Region
Northeast	Northeast Census Region	.155
Midwest	Midwest Census Region	.173
South	South Census Region	.296
West	West Census Region	.196
National	United States	.182

^aViolent index crime includes murder, sexual assault, assault and robbery.

^bDemographic variables are taken from the 2000 census.

^cOther race included those that identified as non- Hispanic Asian, American Indian or Alaska natives, Native Hawaiians, other races or multiple races.

^dCalculated per 100,000 residents in the DMA.

Outcome (Dependent) Variables

Two outcome variables are considered in the analyses: (1) the race or ethnicity of the perpetrator; and (2) the race or ethnicity of the victim, as reported in broadcast news stories. We followed Dixon and Linz's (2000a; 2000b) procedure for coding race and ethnicity. Race-ethnicity of the perpetrator is a five category nominal-level dependent variable, distinguishing among White (the reference category), Black, Hispanic, Other, or Not-indicated race and ethnicity. Perpetrators' race or ethnicity was determined from the content of broadcast stories based on one or more of the following indicators: the perpetrator was shown on the tape; a mug shot was shown; an artist's sketch was shown; a photo was shown; the race or ethnicity was stated (or inferred by a stated surname); a family member was shown, or a different news report on the same incident indicated the person's race-ethnicity.⁵ If race or ethnicity could not be determined by these methods, it was coded as not indicated. If there were multiple perpetrators, the race-ethnicity of the most prevalent race or ethnic group was assigned; if there were equal numbers from different racial or ethnic groups, the minority group was coded. For example, if a group of four perpetrators was shown in which two were White and two were Black, Black was coded.⁶ The second outcome variable distinguishes among the race-ethnicity of victims using the same categorization and criteria. Note that approximately 20 percent of the stories were randomly sampled for double-coding to assess reliability; this was done for the 2002 and 2003 samples separately to assess possible coder drift. Cohen's kappas for perpetrator ethnicity remained at .86 and ranged from .83–.95 for victim ethnicity, .73–.80 for number of victims, .89–.94 for victim gender, and .72–.84 for type of violent crime. As kappa is a conservative statistic, these are very good levels of reliability (Landis and Koch 1977).

Predictor (Independent) Variables

We examine the roles of explanatory variables representing three broad contexts which may affect the portrayal of race and ethnic groups: (1) the situational context of a particular report, including characteristics or content of the story itself; (2) the social structural context of the DMA in which the television station broadcasts and competes (i.e., race-ethnic composition and crime level); and (3) controls for region and population size.

<C>Situational Context (Story Content)⁷

To assess the importance of the nature of the crimes that are the subject of the news stories, we distinguish between crimes that are similar to the FBI's violent index offenses⁸ and non-index crimes (e.g. kidnapping, car jacking, breaking and entering with persons in home). Incidents involving homicide (e.g., murder, attempted murder, manslaughter, slayings), robbery, murder-suicide, assault, or sexual assault (e.g., rape, sexual abuse, sexual molestation) are coded as violent index crimes. Multiple Victims is a dummy indicator distinguishing between crimes involving more than one victim and single victim crimes. A dummy-coded variable for female victim differentiates between crimes in which the victim was female versus male. If there were multiple victims, the more common gender is assigned; if there were equal numbers, the story is coded as having female victims. Analyses of perpetrators' race-ethnicity also include a dummy variable for White female victim to identify any unique effect of this gender-race status compared to all others. Of note, we conducted some analyses that included story length as an additional situational context measure. However, this factor was not significant in any model; thus, we do not report the findings of these supplemental analyses.

<C>Designated Market Area Context

To capture the influence of the larger social environment on the media's portrayal of the race-ethnicity of perpetrators and victims, we include a number of DMA-level contextual variables, constructed by aggregating count data from the county to the DMA level before creating the measures.⁹ Racial and ethnic composition of the DMA is measured as the percentage of the total DMA population reported in the 2000 Census as a member of each race or ethnic group. Four variables are included: percent non-Hispanic White, percent non-Hispanic Black, percent Hispanic, and percent Other including non-Hispanics Asians, American Indians or Alaska natives, Native Hawaiians, other races, or multiple-race individuals. We capture the seriousness of the crime problem in the area by including the violent index crime rate operationalized as the incidence of reported homicides, robberies, rapes, and aggravated assaults per 100,000 persons in the DMA in the same year as the date of the broadcast (either 2002 or 2003).

Some readers might wonder about the applicability of Designated Market Area characteristics (socio-demographic contexts) to the national news networks. “DMA” characteristics are in fact highly relevant for the national news given our interest in how the social contexts of the markets within which news stations are situated affect their portrayal of race and ethnicity in the news stories which they broadcast to their markets. Conceptually, for the national news stations the appropriate market is the entire nation, so we use the social characteristics of the nation as a whole to measure the “DMA” characteristics for the national news stations. Moreover, the inclusion of national and local broadcasts in the same analysis means that we capture the diversity of media markets in which stations operate, from the smallest geographic locales to the entire country. This also ensures that our sample includes the complete spectrum of crime news stories to which viewers are exposed, from those of purely local interest to those which captivate the nation. We do consider the possibility that analyzing stories broadcast by both local and national stations could affect our results. Consequently, we include a control for this in our models (discussed next), and in our supplementary results section we report the outcome of a sensitivity analysis (which shows no systematic differences).

<C>Controls

We include two controls for the broader social context within which television stations operate. First, because population size is related to crime, sense of community, and other social processes, we control for its possible influence on television crime news portrayals of race and ethnicity of participants. Population is measured as the number of persons living within the DMA according to 2000 census data. Region is a set of four dummy variables, indicating whether the local broadcasting station (DMA) is located in one of four broad Census regions (Northeast, Midwest, South, and West); the reference category for this set is the United States as a whole which is the context for national network news broadcasts. This coding scheme permits us to assess whether or not portrayals vary across regions and between locally and nationally broadcast stories.

<B>Statistical Analyses

Multinomial logistic regression was used to analyze the two outcome variables (race-ethnicity of perpetrator and race-ethnicity of victim) given that each consists of five unordered categories. However, analyses of these data potentially must take into account how the sampled news stories are clustered by date or geography as such clustering may induce non-independence (unmeasured, idiosyncratic common causes of the outcomes) in the sample data. For example, a sensational crime is very likely to be covered by newscasts in many disparate DMAs on the same date. Analogously, a local crime spree (or evolving story) may result in continued local coverage within a DMA across a number of dates. In either case, it is not possible to assume that the cases are independent, and thus the calculated standard errors will be biased and inconsistent, typically resulting in significance test values being upwardly biased (Pendergast et al. 1996).

In the event that diagnostic tests indicate the presence of clustering effects in the data, it is possible to calculate robust standard errors for a multinomial logistic regression to estimate the standard errors properly for a single, specified form of clustering. Diagnostic testing is complicated by the fact that no standard test exists for examining the presence of clustering effects in multinomial logistic regression, although one is available in STATA for binomial logistic regression (i.e., for a dichotomous outcome). Consequently, we diagnosed the presence of clustering effects by constructing five dichotomous contrasts from each outcome variable (each category versus all the others) and then estimating a random effects models (REM) for binomial logistic regression. We estimated two REMs for each dichotomous contrast, one specifying that the sample is clustered by date and the other that the sample is clustered by DMA. The likelihood ratio (LR) tests for the statistical significance of the clustering effects suggested that the analyses should take into account clustering by date but not by DMA. For the race-ethnicity of the perpetrator, four of the five LR tests were significant for clustering by date (with the other marginally significant, p= .06), while none was significant for clustering by DMA. Similarly for the race-ethnicity of the victim, four of the five LR tests were significant for clustering by date, while only one was significant for clustering by DMA. The model results presented below utilize robust standard errors correcting for clustering by date of broadcast.

<A>RESULTS

We begin our presentation of the multivariate results by exploring how the portrayal of the race or ethnicity of perpetrators in television newscasts varies across aspects of story situational factors (e.g., severity of crime) and social context (e.g., race and ethnic composition of DMAs). We then turn to the results for the representation of victims in crime news stories.

<B>Explaining Variation in the Portrayal of the Race or Ethnicity of Perpetrators

We first estimated a baseline model with no predictors and compared the relative incidence of the reported ethnicity or race of non-White perpetrators to Whites to determine if, as expected, there was an initial disproportionality in the reporting of these characteristics of perpetrators relative to groups' representation in the population (results not shown). In this intercept only model, the odds ratios represent the relative chances that members of a given race or ethnicity are more or less likely to be portrayed as a perpetrators than are Whites. The odds ratios are .785 (NS) for Blacks, .260 (p<.001) for Hispanics, .107 (p<.001) for Others, and 1.576 (p<.001) for Not indicated. The odds ratios suggest that Blacks are not significantly more or less likely to be reported as perpetrators than Whites, while both Hispanics and Other races are less likely to be reported than Whites. Race or ethnicity is more likely to be Not indicated in television news reports than to be reported as White. Thus, although official national crime statistics indicate that violent perpetrators are predominantly White (with the partial exception of homicides), Whites are not significantly more likely than Blacks to be portrayed as perpetrators in television news stories.¹⁰ The media's depiction of Hispanics and Other groups as perpetrators, in contrast, is consistent with national statistics showing that they are much less likely than Whites to be perpetrators of violent crimes. Such a relative overreporting of Black offenders accords with the predictions of power structure, racial threat and racial privileging explanations, but the media representation of Hispanics and Others is more consistent with market share and normal crimes expectations.

Table 2 presents the odds ratios and standard errors (in parentheses) of the analysis of television news reporting of perpetrators' race or ethnicity for the indicators of the situational characteristics of the reported crime (story content) and DMA context. The Likelihood Ratio Index indicates that the addition of these predictors increases the log likelihood of the model by 11.9 percent of the maximum possible change. Supplementary analyses (not shown) show that adding each set successively has significant effects. That is, we examined a model where only the story content measures are added to the baseline model (χ² = 145.42). Similarly, the addition of the contextual measures yields a significantly better explanatory model (change in χ² = 64.22). We discus each set of factors in turn.

Table 2.

Odds Ratios from Multinomial Logistic Regression^a of Perpetrator's Race and Ethnicity in Media Coverage of Violent Crime on Story Content and DMA Context (N=660).

Predictors	Black Perpetrator	Hispanic Perpetrator	Other Perpetrator	Race-Ethnicity Not Indicated
Story Content
Violent Index Crimeb	2.306* (.508)	1.807 (.661)	.956 (1.003)	1.561 (.372)
Female Victim	1.682 (.407)	1.181 (.530)	1.072 (.660)	1.032 (.379)
White Female Victim	.039*** (.561)	.507 (.736)	.065** (1.075)	.100*** (.457)
Multiple Victims	.656 (.333)	.732 (.405)	.714 (.839)	339*** (.251)
DMA Context
Violent Index Crime Rate in DMAc	1.000 (.001)	1.000 (.002)	.999 (.002)	1.000 (.001)
Percent Black in DMA	1. 025 (.023)	1.036 (.040)	1.009 (.031)	1.004 (.020)
Percent Hispanic in DMA	1.005 (.019)	1.081*** (.020)	.998 (.034)	1.003 (.015)
Percent Other Race in DMA	.961 (.077)	1.035 (.052)	1.095* (.047)	1.002 (.033)
Control Variables
DMA Population (millions)	1.078 (.062)	.931 (.053)	1.107 (.099)	1.013 (.037)
Northeast	1.40×1009 (17.079)	7.77×10−09 (15.066)	7.83×1011 (26.643)	48.122 (10.452)
Midwest	9.95×1008 (17.274)	3.42×10−09 (15.022)	7.55×1011 (27.312)	35.742 (10.442)
South	1.47×1009 (17.470)	1.20×10−09 (15.459)	2.70×1012 (27.634)	43.530 (10.710)
West	8.51×1008 (17.600)	1.18×10−09 (15.120)	1.52×1011 (28.153)	39.210 (10.570)
Intercept	3.96×10−10 (17.094)	1.57×1007 (15.099)	8.19×10−13 (27.182)	.057 (10.345)
Percent in Outcome Category	21.1%	7.0%	2.9%	42.3%
Log Likelihood	−774.816
Likelihood Ratio Index	11.9%

^aThe base outcome is White (category percent = 26.8%). Robust standard errors calculated for clustering by date of broadcast.

^bViolent index crime includes murder, sexual assault, assault and robbery.

^cCalculated per 100,000 residents in the DMA.

^*Significant at .05 level, one tailed.

^**Significant at .01 level, one tailed.

^***Significant at .001 level, one tailed.

Although only a few predictors have significant effects, the pattern of results reveals an interesting story. If a particular news story is about an index crime, this increases the odds of mentioning a Black perpetrator over a White, consistent with power structure and privileging arguments. If the news story concerns a White female victim, this lowers the likelihood of mentioning Black or Other perpetrators relative to Whites or of even identifying the race or ethnicity of the perpetrator at all. Put differently, if the victim is a White female this generally increases the likelihood of identifying the perpetrator as White, reflecting the reality that Whites (rather than other racial group members) harm Whites. Note that the effect of female victim is not significant. Given the coding of these variables, this means that stories about minority female victims do not affect the portrayal of perpetrator's race or ethnicity and indicate a privileging of white (female) victims. Stories about multiple victims reduces the likelihood of the perpetrator's ethnicity not being identified.

Turning to the DMA context, we find that population composition plays only a small role in shaping media coverage of suspected criminal perpetrators. These limited significant predictors are in line with expectations from racial threat and market share perspectives. The larger the percent Hispanic and the percent Other in the DMA, the more likely the corresponding group is portrayed as perpetrators. Although stories about violent index crimes affect the reporting of perpetrators' ethnicity or race, the local rate of actual violent index crime has no effect. Finally, neither region nor population is influential.

<B>Analyses of Media Depictions of Victims' Race or Ethnicity

As with perpetrators, we estimated baseline model for victims to establish if there is an initial disproportionality in television news reporting of the race or ethnicity of non-White victims compared to White victims (results not shown). In this case, the odds ratios reflect the chance that members of the group in question are more or less likely than Whites to be portrayed as victims. The odds ratios are .332 (p<.001) for Blacks, .136 (p<.001) for Hispanics, .076 (p<.001) for Others, and 2.044 (p<.001) for Not indicated. Above we noted that TV news stories are not significantly less likely to report Blacks than Whites as perpetrators. But notably, stories are less likely to depict Blacks than Whites as victims, providing a clear contrast between media's portrayal of Blacks as perpetrators versus victims. Indeed, Blacks, Hispanics, and Others are all less likely than Whites to be portrayed as victims.

As was the case for the perpetrator models, we also examined a model (not shown) in which the indicators of the situational context alone were added to the baseline model. This model and the model displayed in Table 3, which includes the story content and broader social structural context measures, are each a significant improvement in the fit of the model (χ² = 134.15 and 92.36, respectively). According to the Likelihood Ratio Index, the full model explains nearly 15.6% of the possible change in the log-likelihood.

Table 3.

Odds Ratios from Multinomial Logistic Regression^a of Victim's Race and Ethnicity in Media Coverage of Violent Crime on Story Content and DMA Context (N=660).

Predictors	Black Victim	Hispanic Victim	Other Victim	Race-Ethnicity Not Indicated
Story Content
Violent Index Crimeb	2.625 (.673)	1.109 (.503)	.939 (.865)	1.537 (.479)
Female Victim	.338** (.379)	.228** (.605)	.491 (.655)	.183*** (.290)
Multiple Victims	.562 (.410)	2.969* (.521)	.940 (.563)	1.292 (.229)
Sexual Assault	-	-	-	7.116*** (.403)
DMA Context
Violent Index Crime Rate in DMAc	1.002* (.001)	1.003 (.003)	.996* (.003)	.999 (.001)
Percent Black in DMA	1.018 (.030)	1.060 (.074)	1.113*** (.032)	1.018 (.015)
Percent Hispanic in DMA	1.015 (.014)	1.097*** (.018)	1.041 (.029)	1.003 (.012)
Percent Other Race in DMA	.986 (.081)	1.011 (.075)	.977 (.064)	.986 (.029)
Control Variables
DMA Population (millions)	1.010 (.060)	.823** (.088)	1.042 (.077)	1.012 (.036)
Northeast	51.068 (16.619)	4.91×10−24† (24.451)	2.73×1005 (21.636)	37.862 (10.040)
Midwest	24.637 (16.683)	7.03×10−24† (24.671)	5.84×1004 (21.881)	32.160 (10.087)
South	11.150 (17.097)	1.02×10−24† (25.245)	2.26×1004 (22.130)	50.982 (10.288)
West	9.789 (17.001)	1.75×10−24† (25.126)	3.98×1004 (22.251)	31.552 (10.142)
Intercept	.003 (16.365)	2.64×1021† (24.700)	1.82×10−06 (21.567)	.068 (9.958)
Percent in Outcome Category	9.2%	3.8%	2.1%	57.0%
Log Likelihood	−614.367
Likelihood Ratio Index	15.6%

^aThe base outcome is White (category percent = 27.9%). Robust standard errors calculated for clustering by date of broadcast.

^bViolent index crime includes murder, sexual assault, assault and robbery.

^cCalculated per 100,000 residents in the DMA.

^*Significant at .05 level, one tailed.

^**Significant at .01 level, one tailed.

^***Significant at .001 level, one tailed.

^†Significant at .05 level, two tailed.

Story context plays a part in explaining victim portrayals by race and ethnicity. If a story describes a female victim, this decreases the likelihood of not identifying the victim's ethnicity or race or of portraying the victim as Black or Hispanic. That is, such news stories are more likely to identify the victim as White, consistent with racial privileging, normal crimes, market share and power structure arguments. And if a story describes multiple victims, this increases the odds of depicting Hispanic victims,¹¹ contrary to our expectation. Stories covering sexual assaults increase the likelihood of not describing the victim's ethnicity or race, likely reflecting the practice of trying to keep a victim's identity confidential. Indeed, sexual assault is only used as a predictor for the category in which race and ethnicity are not indicated because there are too few cases (none in some instances) for reliable estimation.

Our assessment of the influence of DMA context reveals a number of effects on ethnic and racial portrayals of crime victims, suggesting context may be more influential in coverage of victims than it is for perpetrators. A higher level of violent index crime in the DMA increases the likelihood of depicting a Black victim but decreases the odds of portraying the victim as from an Other group. The likelihood of reporting a Hispanic victim goes up as the percent Hispanic in the DMA increases, in line with an audience market share approach. There is no parallel effect of Other race-ethnic composition; instead the percent Black increases the portrayal of Other victims. Unlike the perpetrator analysis, the control variables have significant effects for this outcome. A larger DMA population decreases the likelihood of reporting on Hispanic victims, and each census region negatively affects the likelihood of featuring Hispanic victims. This means that local newscasts in each region (relative to the national news) are systematically less likely to report on Hispanic victimization.

<B>Supplementary Analyses

We undertook several sensitivity analyses in order to assess the robustness of our results to alternative analytic strategies, such as separately analyzing homicides versus other violent crimes or limiting the sample to local news broadcasts. (Tables available upon request.) We find essentially comparable results in these supplementary analyses although some scattered differences are seen. The multivariate homicide analyses of victims' race or ethnicity shows the only systematic difference; the reporting of multiple victims significantly affects all the victim race and ethnic categories. As before, multiple victims increase the likelihood of reporting Hispanic compared to White victims, but in the homicide only analysis this factor explicitly decreases the likelihood of reporting Black and Other victims relative to Whites. The homicide only analyses also show some small differences in the effects of race and ethnic DMA composition; in particular, the reporting of Hispanic victimization is increased by the percent Black but decreased by the percent Other in the DMA. The analysis of local news broadcasts (compared to the analysis that combines local and national news broadcast data) shows no systematic differences in the effects of the predictors for either perpetrator or victim representations. Further, the differences found are mainly limited to changes in effects on the reporting of Others as either perpetrators or victims.

<A>DISCUSSION AND CONCLUSIONS

The research reported above is an important extension of current literature on media representations of racial and ethnic groups in lawbreaking and as victims of crime (e.g., Chiricos and Eschholz 2002; Dixon and Linz 2000a, 2000b; Klein and Naccarato 2003; Entman and Rojecki 2000). We evaluated hypotheses drawn from five perspectives: power structure (e.g., Poindexter et al. 2003); market share (e.g., Chermak 1994); “normal crimes” (Pritchard and Hughes 1997), racial threat (e.g., Blalock 1967; Eitle et al. 2002), and racial privileging (e.g., Baumer et al. 2000; Kleck 1981; Williams et al. 2007). Drawing on television news reports for a broad cross-section of stations and Designated Market Areas nationwide, we were able to address heretofore unanswered questions regarding the determinants of ethnic and racial representations. Further, because of the innovative features (focus on a representative sample of TV crime news stories for a national set of local markets) of the data set that we used (Long et al. 2005), the observed results are more generalizable than those of earlier studies. Importantly, our work yields two broad conclusions. First, and most generally, the analyses indicate partial, though not overwhelming, support for each of the noted perspectives, with perhaps the most consistent support for the privileging explanation. Second, the findings demonstrate that aspects of the situational context of crime stories (especially the race-ethnicity and sex of the victim), and the social structural context of the market area of the broadcast (population composition, and for victimsʽ race or ethnicity, the level of violence in the area) are contributors to television news reporting of perpetrators' and victims' race or ethnicity. Below we elaborate on what specific findings support the various perspectives and offer some observations about the general importance of the work and directions for future research.

Starting with the results for portrayals of offenders, we would expect Whites to have a higher likelihood of being reported on if reporting reflects offending incidents, because they are the most populous group. We did not find evidence of a significant difference in the number of portrayals of White perpetrators relative to Blacks in our base models. To us, this suggests a relative overreporting of Blacks compared to Whites. We also found underreporting of Hispanic perpetrators relative to Whites. We interpret the results for Blacks as consistent with power structure, racial threat and racial privileging arguments. Recall that the power structure approach focuses on news coverage as reflecting the world views of powerful (and White) owners and employers of media. These powerful actors may not be inclined to represent Whites to the degree that members of this group perpetrate crimes. This could be at the expense of overreporting subordinate groups' members as offenders. That is, owners of TV news might use their resources to protect and privilege White over Black perpetrators. The non-significant difference in the reporting of Black versus White perpetrators is also consistent with racial threat arguments. Here, the idea is that racial minorities may be perceived as threats (politically, economically, and criminally) to the white majority. Thus, the frequent reporting of Black perpetrators may be a case of media reflecting dominant groups' views through delegitimizing Blacks by portraying them as criminal. Support for power structure and privileging arguments is also evident in the significant positive effect of index crime news stories on the reporting of Black perpetrators.

The finding that Hispanic perpetrators are underreported in television crime news relative to Whites is consistent with the market share perspective and with normal crimes arguments. These approaches focus on what makes crime (or other) stories newsworthy. From a market share standpoint, reporting on Hispanics might be regarded as of little interest to a predominantly White audience. Pritchard and Hughesʽ normal crime type perspective emphasize that crime stories are less newsworthy when they pertain to groups whose status departs from that of high status White males (as would be the case with Hispanics), and when they involve “usual” suspects (relative to their population size, Hispanic violent crime is more usual than White violent crime). Thus, from either a market share or normal crimes perspective the underreporting of Hispanic perpetrators is understandable. Our findings also demonstrate greater reporting of Hispanic perpetrators when the size of this population in the DMA is larger. This pattern is also consistent with a market share perspective, and suggests that as the Hispanic population grows so will TV news interest in covering the criminal involvement of this group to gain in market share of this audience.

Several findings do not comport with our themes about the representation of racial and ethnic groups as perpetrators in crime stories. For example, contrary to our hypotheses based on power structure and racial privilege arguments, White female victimization is associated with decreases rather than increases in the likelihood of reporting Black perpetrators. Our assessment is that this finding likely reflects the fact that victimization of White females is disproportionately at the hands of White males than males of other colors, rendering news reports of Black suspects and offenders irrelevant in the context of White female victimization. Another departure from our expectations is the fact that the size of the Black population in DMAs does not increase the reporting of minority perpetrators (either Black or Hispanic), suggesting a lack of merit for our racial threat or market share claims. Yet there are several possible explanations that do not challenge the general merits of these theories. For example, it may be that the settings or time-frame for this investigation means that mechanisms are at play (e.g., degree of residential segregation) that reduce the threat to dominant groups of these populations, while other settings and time frames would support the threat hypothesis. Similarly, threat and market share explanations of crime news patterns may be more salient for rapidly increasing populations, such as Hispanics or Blacks gaining population in areas where they have traditionally been a small component of the population. Additional research is required to shed light on these possibilities.

Regarding the race of victim outcome, the findings lend particular support to the privileging and power structure perspectives, with partial support for the market share approach. Our basic analyses reveal that racial and ethnic minorities are less likely to be portrayed as victims than Whites. Notably this finding is consistent with: (1) racial privileging arguments that greater attention is given to the victimization of Whites than to non-Whites; (2) power structure claims that White owners of media are more interested in presenting news about White victims; and (3) the normal crimes perspective that White victims are more likely to be reported because they are more unusual (particularly White female victims) or because news is about more powerful groups (Pritchard and Hughes 1997). Although the market share approach is also consistent with the representation of White victims, it is partly contradicted by the portrayal of black and Hispanic perpetrators. The analyses further reveal that stories about female victims are significantly more likely to report victims' race as White than as Black or Hispanic. This is true even though White females are among the least likely groups to be victimized by violent crime, and as such is consistent with racial privileging arguments holding that television and other news media reflect the broader societal value placed on White female victims over their Black and Hispanic counterparts. The finding is also in line with market share arguments and the statistical deviance component of the normal crimes explanation, which both suggest that audience interest is piqued when Whites (who are less likely to experience criminal violence) are victimized.

Based on market share themes, we also hypothesized that higher percentages of Blacks and Hispanics would increase the reporting of minority victims. This hypothesis was only partially supported. In line with expectations, the percentage Hispanic is positively related to reporting Hispanic victims. However, there is no evidence that the percentage Black in the DMA influences the reporting of Black victims. The differential role of Hispanic and Black composition may reflect that Black populations are more stable in areas while Hispanic populations are growing (and rapidly so), with television news attempting to capture this increasingly large group by reporting crime victims from this population.

What are the implications of these findings? Although we have not focused on how the media affects individuals' schemas or world views on crime (as social psychological perspectives do), the privileging of White victims found in our results is highly pertinent in this respect. The natural converse of this result is that victimization in minority communities is routinely minimized. To the extent that television media are responsible for generating public support for and/or altering views of social groups, this differential attention to race-ethnic victims provides a misleading picture of which groups are most vulnerable to violent victimization. If viewers repeatedly see White victims, their world view will be altered to reflect this such that Whites come to believe they have more reason to be fearful of crime than other groups or than is realistic. Moreover, the consequences may reach beyond an unrealistic assessment of the likelihood of victimization for Whites. The minimization of attention to groups that, in fact, bear the brunt of victimization may lead to a lack of support for policies oriented to addressing their vulnerability to violence.

Overall, our results suggest that no single perspective is sufficient for understanding the portrayal of race and ethnicity in television news stories. On the one hand, this suggests the need for theoretical work to develop an integrated approach built on these eclectic perspectives. On the other hand, there remains a need for further empirical research in this area. In modeling the determinants of media representations of race and ethnic groups in crime news, attention should be given to additional situational components of stories. Of note, alcohol and drug use are not among the situational characteristics of crime stories included in our models. Examining such factors could shed light on whether racial and ethnic overtones in reporting attend news story when these factors, which are frequent elements of violent criminal activity, are involved in crime incidents. Other situational factors that should be explored include the ages of perpetrators and victims (especially younger ones), gang involvement, victim-offender relationship, motive, and stage of criminal justice processing.

Beyond the study's implications for various perspectives on media portrayals, a key contribution of our work is to demonstrate that the social structural characteristics of market areas matter, particularly for media depictions of crime victims. The fact that demographic variables are significant in explaining race and ethnic portrayals may indicate that violent crime stories are being marketed to particular viewer demographics. This seems particularly likely for market areas with a substantial Hispanic population. We found that the percentage Hispanic increases reporting on both Hispanic perpetrators and victims. Yet, our interpretation of these population effects as market share effects is tempered by the fact that patterns do not hold for other race and ethnic outcomes. There were a few other scattered effects of DMA characteristics. Clearly, additional work is needed to explore the mechanisms and processes whereby ethnic and racial population composition and other local contextual factors influence television crime reporting of victims' and perpetrators' race or ethnicity. However, if certain groups are represented fairly while others are not, and if this is routinely done to gain market share despite detrimental consequences of over- or under-representation, a claim of institutional racism could be made. Our results raise this possibility.

We also believe that future research will benefit from attention to additional aspects of the social-structural context of the market area in which a story is aired. For example, researchers should explore the influence of race-, ethnic- and gender-specific levels of offending and victimization in the DMA. Data limitations restricted us to considering only general levels of reported violence. Economic factors might be considered as well, such as the percent unemployed in the DMA. Because social structural context is relatively unexplored as a predictor of the outcomes of concern in this study, there are numerous directions that future studies could take. Our research offers a point of departure for this endeavor.

Finally, we note that our analysis sets the stage for future work on the sources of representation of ethnic groups in a wider set of TV newscasts, not just for stories focused on crimes but more broadly for all types of news stories. Scholars should keep in mind that findings from other studies show that stories featuring Blacks are predominantly crime stories, while only a minority of stories featuring Whites are about crime. Thus, Whites are privileged in how they are portrayed not just within crime stories but across all stories as well.

Appendix

Table A. 1.

Frequency of Stories for Designated Market Areas (N=660).

Designated Market Area	n
Abilene/Sweetwater	1
Albany	2
Albany/Schenectady/Troy	3
Albuquerque/Santa Fe	5
Atlanta	4
Augusta	4
Austin	2
Bakersfield	4
Baltimore	2
Baton Rouge	2
Birmingham	7
Boston	9
Buffalo	5
Champaign/Springfield/Decatur	2
Charlotte	5
Charlottesville	3
Chicago	35
Cincinnati	5
Cleveland	5
Colorado Springs/pueblo	2
Columbia, SC	3
Columbus, OH	1
Dallas/Ft.Worth	9
Davenport/Rock Island/Moline	2
Dayton	5
Denver	9
Detroit	13
Duluth/Superior	2
Fairbanks	1
Flint/Saginaw/Bay City	1
Fresno/Visalia	5
Ft. Myers Naples	2
Ft. Smith/Fayetteville/Springdale	4
Grand Rapids/Kalamazoo/Beaver Creek	4
Green Bay/Appleton	3
Greensboro/High Point/Winston Salem	1
Greenville/New Bern/Washington	1
Greenville/Spartenburg/Asheville	4
Harrisburg/Lancaster/Lebanon/York	6
Hartford/New Haven	12
Houston	18
Huntsville Decatur	2
Indianapolis	3
Jackson, ms	5
Jacksonville	2
Juneau	3
Kansas City	3
Knoxville	5
Laredo	5
Little Rock/Pine Bluff	1
Los Angeles	38
Louisville	2
Madison	4
Marquette	2
Memphis	3
Miami/Ft. Lauderdale	9
Milwaukee	2
Minneapolis/St. Paul	4
Nashville	3
New Orleans	7
New York	19
Norfolk/Portsmouth/Newport News	3
North Platte	2
Oklahoma City	5
Omaha	2
Orlando/Daytona Beach /Melbourne	4
Peducah/Cape Girardeau/Harrisburg/Mt. Vernon	5
Philadelphia	15
Phoenix	4
Pittsburgh	15
Portland, OR	5
Portland/Auburn	5
Providence/New Bedford	1
Raliegh/Durham	6
Richmond/Petersburg	7
Roanoke/Lynchburg	1
Rockford	2
Sacramento/Stockton/Modesto	11
Salt Lake City	5
San Antonio	1
San Diego	10
San Francisco/Oakland/San Jose	7
Seattle/Tacoma	16
Shreveport	1
South Bend	3
St. Louis	4
Syracuse	6
Tampa/St. Petersburg	8
Toledo	2
Tri-cities, TN, VA	6
Tucson	4
Tulsa	2
Tyler/Longview	5
Utica	5
Waco/Temple/Bryan	3
Washington D.C.	11
Watertown	1
West Palm Beach/Ft. Pierce	9
Wichita/Hutchinson	2
Youngstown	1
National Media Market	120