Content-Type: text/html
This paper was presented at the Association for Education in Journalism and
Mass Communication in San Antonio, Texas August 2005.
If you have questions about this paper, please contact the author
directly. If you have questions about the archives, email
rakyat [ at ] eparker.org. For an explanation of the subject line,
send email to
[log in to unmask] with just the four words, "get help info aejmc," in the
body (drop the "").
(Jan 2006)
Thank you.
Elliott Parker
====================================================================
Partisan and Structural Balance in Newspaper Coverage of U.S. Senate
Races in 2004 With Female Nominees
By
Frederick Fico
Eric Freedman
Brad Love
A paper submitted to the Theory and Methodology Division of AEJMC for
consideration for presentation at the August 2005 Conference in San Antonio.
Frederick Fico is professor and Eric Freedman is assistant professor
in the School of Journalism at Michigan State University, where Brad
Love is a doctoral student in the Mass Media Ph.D. Program.
Abstract
Nine newspapers covering U.S. Senate races in 2004 were mostly
even-handed in the space and prominence given candidates. Reporter
gender, newsroom diversity and newspaper size were associated with
partisan imbalance giving more favorable treatment to Democrats.
The partisanship of a story's lead predicted the story's structural
imbalance, regardless of the party the imbalance favored. However,
story partisan and structural imbalances were negligibly related,
suggesting that news processing conventions rather than journalistic
partisanship produced the imbalance.
Key Words: fairness; balance; election reporting; bias
Media bias is seldom defined in a manner that permits it to be
reliably observed in a variety of contexts. More seldom still is
such bias related to journalist or news organization characteristics
that enable it to be predicted.
The first goal of this research, therefore, is to define two types
of news bias that are distinct in their origin. The first is a
structural bias in individual stories that favors one side in a
conflict. Structural bias may result from news reporting conventions
and limitations or the attention-commanding activities of the
conflict sides. The second is a partisan bias in which aggregate
news coverage systematically favors the liberal or conservative side
in a political conflict. Political bias results from journalists'
political orientations infiltrating coverage despite reporting
conventions mandating impartiality.
The second goal of this research, then, is to illuminate factors
causally related to these structural and political biases. Certainly
if bias exists in news reporting, journalists themselves need to know
what it is and even how they might minimize it. For their part,
citizens need to know the degree of caution and confidence they
should have in the political information they get from media. Both
the news media and the public will be better served by knowledge of
such biases than by assumptions of partisanship or purity.
THEORICAL FRAMEWORK
Shoemaker and Reese define five levels of influences on news media
content, with higher levels constraining the influence of those
below.1 Influences from four such levels include non-media
institutions affecting media organizations, specific media
organizational characteristics, work routines within media
organizations, and the characteristics of individual journalists.
This study explores election coverage bias potentially influenced
from these four levels. At the societal level, the operation of the
political process will both command media attention and shape the
coverage. At the organization level, a news organization's audience
reach determines the resources potentially available to do news
work. Organization goals and resources will in turn shape
goal-attainment routines, the kind of personnel hired, and the
rewards and punishments that reinforce desired behaviors. But the
backgrounds, beliefs and expertise of such individual workers will
also, to some extent, help shape the work they do.
Some research has employed this hierarchical model's conceptual
ordering to explore influences on political bias in news
content. Most research, however, has focused solely on ways of
defining and describing aspects of media content. For example, a
review of content analysis studies published in Journalism & Mass
Communication Quarterly since 1998 showed that barely a third of them
attempted to predict any quality of media content.2
This study brings into a hierarchical model variables that previous
research has empirically found to be directly and indirectly related
to conflict coverage bias. Two streams of research are
reviewed. One deals with studies of election coverage fairness and
balance. The second deals with influences on such election coverage balance.
Bias and Election Reporting
Science becomes biased when atypical observations of some phenomenon
produce misleading inferences about the phenomenon. In most
journalistic work, however, observations are indirectly relayed by
sources. Bias can therefore occur if source selection by reporters
produces atypical or incomplete perspectives or information about
some news topic.
This problem of source selection goes directly to fairness and
balance in journalism. Fairness in journalism usually refers to
whether relevant sides in a conflict are included in news
coverage. Balance usually refers to the evenness with which such
conflict sides are treated relative to one another in that
coverage. Fairness is clearly a necessary but not sufficient
condition for balance in any single story. But in the aggregate of
stories on some conflict, individually unfair ones reflecting
opposing sides may "balance out" in the entire coverage.
Fairness and balance in reporting become problematic, of course,
when judgments of source availability, credibility, and truthfulness
must also be made. Fairness and balance become still more
problematic when news values such as impact or unusualness are part
of journalistic decision making. However, many political and social
issues involve conflict over values, meanings and priorities that
cannot be decided by an appeal to evidence. Support or opposition to
abortion, for example, may depend on the meanings and values one
brings to pregnancy rather than on a decisive biological fact. In
other words, most coverage of political conflict involves judgments
by journalists about the legitimacy of sources and their
contentions. In the political arena, journalistic judgments that
result in departures from fairness and balance may have troublesome
consequences for both public policy and for media credibility.
Indeed, the concern for fair and balanced election reporting flows
from both ethical and practical concerns that biased reporting may
actually affect public policy outcomes. During the decades following
World War II, such concerns were largely allayed by "minimal-effects"
studies, and by traditional agenda-setting research emphasizing news
media power to tell people "what to think about" rather than "what to
think."3 More recently, however, "attribute" agenda setting has
explored how coverage of positive or negative issue attributes can
influence the public's positive or negative evaluations of issues,
with real consequences on their electoral decision making.4 Also,
experimental framing research has demonstrated how news coverage
context can influence how news consumers respond to an issue.5
Empirical studies of news coverage political bias have focused
largely on U.S. presidential elections, tracking aggregate attention
to Republican and Democratic candidates in major newspapers and
television networks. Findings have been ambiguous, but one pattern
seems discernable (although barely so). Studies conducted during the
1960s and 1970s have (coinciding with the dominance of the
traditional agenda-setting assessment of media power) generally found
more even-handed coverage, while more recent work has illuminated
imbalances. For example, Evarts and Stempel found no clear bias in
their 1972 study of television networks, three major news magazines
and six major newspapers.6 Similarly, Hofstetter found mixed results
in his 1972 study of several media.7 Stovall, however, found
evenhanded newspaper attention to candidates in the 1980 election,
but not in the 1984 election, in which coverage favored the
Republican.8 Kenny and Simpson found imbalance favoring the
Republican by a Washington newspaper in the 1988 election, which they
attributed to its conservative owner.9 Lowry and Shidler found
television sound bites in 1992 and 1996 favoring the Democratic
candidate, which they attributed to network liberal bias.10 Domke et
al., however, found balanced positive and negative coverage in their
1996 study of two networks and 38 dailies.11
One problem with discerning patterns in these findings is that
differences may be specific to particular elections, particular
media, or particular organizations. But an additional complication
is the possibility that different findings reflect different ways of
conceptualizing and measuring political news bias. Studies assessing
tone of coverage, for example, can be subjective and hard to
replicate. Moreover, studies assessing aggregate partisan coverage
of an election may miss structural bias in typical stories.
Consequently, more objectively observable and replicable measures of
individual stories developed in a series of studies are used in this
research, thereby standardizing the measurement instruments for
bias. These studies, anchoring measurement in the prominence and
attention given partisan sources in individual election stories, have
found imbalance toward one or the other candidates to be
typical. Fico and Cote in newspaper studies of gubernatorial and
presidential elections, for instance, found readers would typically
encounter one-sided stories, or, if two sided, stories that gave much
more space and prominence to just one of the candidates.12 Carter et
al. found a similar pattern when adopting these measures to local
television election coverage of a governor's race.13 However, Fico
and Freedman, and Fico and Cote in newspaper studies and Fico et al.
in a broadcast study found imbalance related more to incumbency than
to the political party of the candidates.14
Influences on Election Reporting
This research also incorporates independent variables empirically
related in past studies to the partisan and structural bias dependent
variables, thereby facilitating replication or modification of
generalizations about factors related to bias. Moreover, this study
includes additional variables that may help to predict election
coverage bias.
At the individual reporter level, studies have found that typical
election story leads give play to only one side in a conflict. Fico
and Cote in studies of newspaper coverage of gubernatorial and
presidential races also found that such partisan leads predicted that
candidate's domination of the rest of the story.15 Given that leads
set the "agenda" for a story, this should not be
surprising. Interestingly, however, imbalance in many stories was
not necessarily related to imbalance in the sourcing, but rather to
the ordering of sources within stories.
Several studies have explored a second individual-level factor,
reporter gender's relation to election stories. Fico and Freedman in
a governor's race study in which a female Democrat ran against a male
Republican found that male reporters gave more attention to the male
Republican while female reporters gave more attention to the female
Democrat.16 In a subsequent study, Freedman and Fico found that
female reporters were more likely than male reporters to use
non-partisan female sources in such election stories.17 Zeldes and
Fico also found this gender difference in a study relating the gender
of network news reporters covering the 2000 presidential race to
their use of sources.18
Studies have also looked at how newsroom-level influences may shape
story balance in election reporting. Fico and Cote and Fico and
Freedman found that the more prominently stories were displayed, the
more balanced they tended to be in their treatment of electoral
opponents.19 The researchers attributed this to heightened editorial
concern for balance of stories more likely to get public (and
partisan) scrutiny, and to a socialization effect on reporters who
model their own work on the qualities of stories given more prominent display.
Few studies have directly tried to measure and relate how journalist
political orientation may affect the balance of their
stories. Richardson and Lancendorfer, however, found a newsroom
"diversity index" related to its editorial support for affirmative
action as a means to enhance diversity in higher education.20 They
suggest that a large minority population in a news organization's
circulation area pressures news organization management to hire more
minorities, and that more minority representation in a newsroom
communicates to news staff about management's own beliefs on
diversity. Craft and Wanta found that the relative proportion of
male and female editors at newspapers influenced the beats assigned
male and female reporters, including the likelihood that women
reporters would be assigned political stories.21 The implication of
these studies is that both management beliefs and management
demographic characteristics have direct and indirect influence on how
reporters do their work. Even more, such a management commitment to
newsroom diversity is plausibly related to a broader political
liberalism that may influence election coverage.
Fico and Freedman found that circulation, considered in that study a
proxy variable for news organization resources, was related to the
existence of specialty bureaus at the newspapers, and therefore
indirectly related to content qualities.22 Certainly it is
plausible that the more resources a news organization has, the more
likely it will be able to hire more staff, including women to help
achieve diversity goals.
Finally, political parties and the candidates themselves will set
the context in which news organizations cover elections. One
well-researched context is the incumbency or challenger status of
candidates.23 Lowry and Shidler, for example, explicitly tested
and refuted the hypothesis that challengers would get more positive
coverage than incumbents in broadcast news election
coverage.24 More recently, Fico and Freedman and Fico et al.
assessed the partisan and structural balance of races that were open
compared to those in which a challenger sought reelection.25 Open
races were characterized by more structurally balanced, even-handed
individual stories.
HYPOTHESES
The Shoemaker-Reese hierarchical model and the studies discussed
above provide guidance for hypotheses linking these variables to
partisan and structural balance. Past research is replicated and
extended, and measures are standardized and replicated to facilitate
generalization across studies. Finally, this research approach is
applied to a geographically broader set of elections and to races
that are less commonly studied for their coverage balance.
Structural Balance Hypotheses
These hypotheses assume that newsroom conventions and resources are
more likely to account for structural than political bias.
H1: Stories with partisan leads are more structurally imbalanced than others.
H2: More prominent stories are less structurally imbalanced than
others.
H3: Stories covering open races are less structurally imbalanced than others.
Partisan Balance Hypotheses
The following hypotheses specifically assume (for the sake of
argument) that a liberal bias of journalists results in story
partisan bias favoring Democratic candidates.
H4: Stories by women reporters will have more partisan imbalance than others.
H5: Stories by newsrooms with higher newsroom diversity scores will
have more
partisan imbalance than others.
Finally, the two dependent variables in this research -- partisan
and structural imbalance -- should be systematically
related. Logically, stories that are imbalanced on the partisan
index because of a political bias (liberal or conservative) on the
part of the journalists should also be systematically imbalanced on
the structural balance index.
The reverse, however, would not logically hold, since random
structural imbalance should equally benefit both opponents over time. Hence:
RQ1: How is the partisan imbalance of stories related to their
structural
imbalance?
METHOD
This study content analyzed the election coverage of the largest
daily newspapers in nine states that held U.S. Senate elections in
2004.26 These races were selected because a woman candidate was
running against a male opponent. In eight of the races, the women
were Democrats. In five races, the women were incumbents. In one
race a women challenged a male incumbent, and in three races women
contested open seats.
Races in which a woman ran against a man were selected for study
because they should provide the strongest possible evidence for
partisan imbalance. In other words, if a liberal bias is evident,
all eight races in which a Democratic woman ran against a Republican
man should have coverage giving more prominence and space to the
Democrat. If structural bias is a better explanation, than attention
to the races should not be patterned to favor the Democrat.
The elections were studied from Labor Day to Election Day. The
individual hard news story was the unit of analysis. Editorials,
letters, op-ed pieces and analyses were excluded, as well as "Q & A"
type stories. A story had to have at least three paragraphs on the
U.S. Senate election to be included.
Partisan and Structural Balance Measures
These partisan and structural dependent variable measures were taken
from the Fico and Freedman study of the 2002 governor's race in
Michigan. Each election story was assessed on the basis of four
components combined into an index for each story. These included the
total number of paragraphs containing source assertions supporting
each candidate, whether partisans for one or both candidates made
assertions in the lead, whether partisan opponents made assertions in
the second through fifth story paragraphs, and whether partisan
opponents made assertions in the sixth through tenth paragraphs.27
To create the partisan balance index, each of the four components was
first judged to favor the Republican, the Democrat, to be balanced or
to be irrelevant (no partisan assertions were made). For example,
the total paragraph measure was considered to favor the Republican if
more Republican-support paragraphs appeared than Democrat-support
paragraphs. The measure was considered to be balanced if an equal
number of paragraphs contained assertions supporting each
candidate. In a similar fashion, the lead, paragraphs
two-through-five, and paragraphs six-through-ten story components
were judged to favor one or the other candidate or be balanced. The
partisan imbalance of a story as a whole was then determined by
subtracting the number of components dominated by the Republican from
the number dominated by the Democrat. The resulting story Partisan
Balance score could range from +4 (all four components dominated by
the Democrat) to -4 (all four components dominated by the Republican).
To create the Structural Balance index, each of the four components
was assessed identically as in the partisan index, but in this index
the number of components dominated by the Republican in a story was
subtracted from the number that favored the Democrat, and the
absolute value then taken. The resulting story index score could
range from 0, indicating a story structurally balanced between the
two candidates (and also balanced on the partisan index) to 4,
indicating that the same candidate (whether Republican or Democrat)
was favored on each of the four components.
Explanatory Variables
Individual-level variables were story lead and reporter gender. The
story's lead was assessed as a nominal-level variable based on
whether it contained a partisan assertion supporting just one side
(regardless of the side any such assertion supported). Reporter
gender, a nominal-level variable, was determined from bylines, but
newspapers were contacted to determine gender when names were ambiguous.
The newsroom-level variable included was story prominence. Story
prominence was an ordinal-level variable on which an inside page
story scored 0, a section front page story scored 1, and a Page One
story scored 2.
At the organizational level, newsroom diversity and newspaper
circulation were considered. A newsroom's diversity index was a
ratio-level variable determined by dividing its proportionate
newsroom diversity by the proportionate diversity of the news
organization's surrounding community. A 0 indicated no correspondence
between the two, and a score of 1 indicated that a newsroom's
diversity matched the diversity of the surrounding
community.28 Circulation rank was an ordinal-level variable used as
a measure of news organization resources.29 Finally, type of race
covered, a societal-level factor, was a nominal-level variable, with
open races scored as 1 and others as 0.
Measurement Validity and Reliability
The balance indices weight the prominence given the assertions of
partisans. This emphasis on prominence is reader-driven. Stories
present information sequentially and many readers who start a story
may not finish it.30 Consequently, stories that present opposing
candidate assertions sequentially may not be perceived as balanced by
readers who leave the story before the opponent's positions or
supporters are presented.
A coder reliability test was conducted on all dependent and
independent variables using about 10 percent of the sample. Both
percentage of agreement and Scott's Pi tests correcting for chance
agreement were used. Percentage of agreement scores ranged from 95
percent to 100 percent. Scott's Pi scores ranged from .93 to 1.0.
Data Analysis
These data include the universe of relevant stories. Nonetheless,
inferences for broader application are made, and statistical
significance tests are therefore used to identify especially
important findings. Study hypotheses were tested using multiple
regression, in a causal model identifying influences on the two
dependent variables. Independent Variable Beta weights were used to
assess the relative strength and direction of influence of these
variables. The Explained Variance statistic is used to assess the
total influence of these independent variables on partisan and
structural balance.
Finally, an overarching question is whether partisan or structural
bias is a better description of any bias found in the reporting. The
partisan and structural balance indices used in this study will have
a modest-to-strong relationship only if stories systematically favor
one of the parties. That regression Beta will be positive if the
Democratic Party is favored, and negative if the Republican Party is
favored. A Beta at or near 0 indicates that while individual stories
are imbalanced, that imbalance does not systematically favor one of
the parties.
RESULTS
Some 175 stories were relevant for the analysis from the nine
newspapers. Individual newspapers ranged from 9 to 39 stories on the
studied Senate races, with a mean of 19.3 stories per
newspaper. About 31 percent of stories ran on Page One, with another
43 percent running on a section front page. About 86 percent of
stories were done by staff reporters, with the rest accounted for by
bureau specialists or other news services. Just over 30 percent of
the stories were written by women reporters. The Diversity Index for
the newspapers ran from .13 to .88, with a mean of .46. On average,
therefore, newspaper staffs were about half as diverse as the
surrounding communities. About 47 percent of the stories covered the
three open races in three states.
The Partisan and Structural Balance indices indicated substantially
even-handed coverage of the elections. (See Table 1) The Partisan
Balance index ranged from -4 to +4, with a mean of -.29, indicating a
slight bias toward Republican candidates. Nearly 46 percent of the
stories ranged between -1 and +1 on the partisan index. The
Structural Balance index ranged from 0 to 4, with a mean of
1.6. Only 23 percent of stories achieved imbalance scores as high as 3 or 4.
Hypotheses Tests
Two of the five study hypotheses were strongly supported and two
were weakly supported. One was contradicted.
Hypothesis One, predicting that stories with a partisan lead would
be more structurally imbalanced than others, was supported. This is
consistent with past research findings and with news writing
conventions mandating that stories emphasize the angle established in
the lead. (See Table 2) Interestingly, however, partisan leads
were not systematically more likely to include Democrat assertions
than Republican assertions, as indicated by the negligible .02 Beta
predicting Partisan Imbalance favoring Democrats.
Hypothesis Two, predicting that more prominent stories would be less
structurally imbalanced than others was supported, consistent with
past research. (See Table 2). But while the direction of the
relationship was as predicted, its Beta was small. Moreover, stories
that were more prominent were more likely to give more space and
attention to Democratic assertions, although again only weakly so.
Hypothesis Three, predicting that open races would be covered with
less structural imbalance than other races, was contradicted. (See
Table 2) In fact, stories covering open races were slightly more
likely to be structurally imbalanced. Moreover, such coverage tended
to give more space and attention to Republican candidates than to
Democratic ones, as indicated by the negative Beta for the Partisan index.
Hypothesis Four, predicting that stories by women reporters would be
more imbalanced in favor of Democrats, was strongly supported. (See
Table 2). Inconsistently, however, stories by women were likely to
be less structurally imbalanced than others, although only weakly so.
Finally, Hypothesis Five, predicting that news organizations with
higher diversity scores would be more likely to give more space and
attention to Democrats, was also supported, but only weakly so. (See
Table 2) Such diversity was negligibly related to Structural Imbalance.
Research Question
A key question in this study focused on whether any news coverage
imbalance was more likely due to the partisan ideology of the
journalists or to factors more attributable to the news reporting
process. The relationship between the Partisan and Structural
Balance indices was -.05, one of the weakest relationships found in
the study. In other words, although stories were frequently
imbalanced structurally, that imbalance nearly as often favored
Republicans as Democrats, and does not seem to reflect a systematic
political bias operating though the coverage of the races studied.
IMPLICATIONS
Several findings emerge with implications for news coverage of the
political process. Perhaps the most straightforward is that
journalists already control one of the tools that influences both the
balance of stories and the potential for news audiences to perceive
bias. News writing conventions mandating placement of the most
important or dramatic developments in story leads often result in
imbalanced stories that give more attention and space to one side in
an electoral contest. This emphasis does not reflect political bias
on the part of journalists as much as it may the news savvy of
sources who know how to capture a headline, and, of relevance to this
study, the lead. Certainly journalists may take more trouble to
balance leads, or even to write more neutral leads that do not yield
as much control to political sources. Failure to do so may
needlessly arouse perceptions of news bias attributed to political
motives of journalists.
A second finding, however, does not point to something so easily
controlled. The gender of reporters makes a difference in the
partisan balance of stories. In this study, stories written by women
were more likely to give space and attention to Democrats, while
stories written by men were more likely to do the same to
Republicans. These are not large differences, as the much smaller
effect of gender on structural balance shows. Nonetheless, it ought
not to be there, and its reasons ought to be explored in newsrooms.
Some other relationships may be notable, but they are smaller and
less confidence must be placed in them. Newsroom diversity, taken as
a measure of management political liberalism, had a small effect
producing partisan imbalance favoring Democrats. However, story
prominence and news organization size had somewhat stronger effects
in that direction as well. Consistent with past studies, story
prominence did tend to produce less structural imbalance, however.
Finally, the type of race covered had mixed effects on partisan and
structural balance. Open races tended to be associated with partisan
imbalance giving more space and prominence to Republicans. Open
races also tended to have a positive effect on structural imbalance,
contrary to predictions of a negative effect. This is not consistent
with previous research, and more work is needed to determine if this
finding is an anomaly or whether generalizations from earlier
research need to be modified.
Certainly other factors must be considered. Indeed, only about 10
percent of the variation in partisan or structural imbalance could be
accounted for by the factors included in the analysis. One factor,
the specialty status of the reporters, shown in past research to be
an influence on balance, was not a factor in this study because the
overwhelming majority of stories were staff-produced.
Much more needs to be learned, of course, about factors affecting
news coverage
of electoral conflict. An obvious next step is to determine how the
partisan sources cited in these stories articulate the bias
quantified in this research. For example, do partisan sources mostly
extol themselves and advance their positions? Do they mostly
criticize or defame their opponents? Also, how are sources other
than partisans incorporated into the coverage, and how do they affect
the partisan and structural balance of stories?
This research also took a rigorously quantitative approach to bias
measurement that limits interpretation. Standardized definitions and
measures will help advance our knowledge of these influences on news
bias. But so too will focused interviews with journalists, closer
observation of news gathering processes, and more attention to the
community and institutional contexts in which news organizations
work. All these approaches should be brought to bear. The end
result may be understandings that enable us to deliver news and
information with less bias and better precision and context. Both
the news media and the public would be better served by that result.
REFERENCES
1. Pamela Shoemaker and Stephen Reese, Mediating The Message: Theories of
Influence on Mass Media Content 2nd ed. (White Plains, N.Y.: Longman, 1996).
2. Only a fifth of these studies of media content explicitly dealt
with how news
media reported political or social conflict.
3 Jim Dearing and Everett Rogers, Agenda Setting (Thousand Oaks, CA:
Sage, 1996).
4. Sei-Hill Kim, Dietram A. Scheufele, and James Shanahan, "Think About
It This Way: Attribute Agenda Setting Function of the Press and the
Public's Evaluation of a Local Issue," Journalism & Mass
Communication Quarterly 79 (spring 2002): 7-25.
5. D. M. McLeod and B. H. Detenber, "Framing Effects of Television News
Coverage of Social Protest," Journal of Communication 49 (1999):
3-23; V. Price, D. Tewksbury, and E. Powers, "Switching Trains of
Thought: The Impact of News Frames on Readers' Cognitive Responses,"
Communication Research 34 (1997): 481-506; and C. H. de Vreese, "The
Effects of Frames in Political Television News on Issue
Interpretation and Frame Salience," Journalism & Mass Communication
Quarterly 81 (spring 2004): 36-52.
6. Dru Evarts and Guido Stempel, "Coverage of the 1972 Campaign by TV,
News Magazines and Major Newspapers," Journalism Quarterly 51 (winter
1974): 645-76.
7. C. Richard Hofstetter, Bias in the News: A Study of Network News Coverage
of the 1972 Election Campaign (Columbus: Ohio State University Press, 1976).
8. James Glen Stovall, "The Third Party Challenge of 1980: News Coverage of
Presidential Candidates," Journalism Quarterly 62 (summer 1985):
266-71; Games Glen Stovall, "Coverage of 1984 Presidential
Campaign," Journalism Quarterly 65 (summer 1988): 443-49, 484.
9. Keith Kenny and Chris Simpson, "Was Coverage of the 1988 Presidential
Race by Washington's Two Major Dailies Biased?" Journalism Quarterly
70 (summer 1993): 345-55.
10. Dennis Lowry and Jon Shidler, "The Sound Bites, the Biters and
the Bitten:
A Two Campaign Test of the Anti-Incumbent Bias Hypothesis in Network
TV News," Journalism & Mass Communication Quarterly 75 (winter 1998): 719-29.
11. David Domke, David P. Fan, Michael Fibison, Dhavahn V. Shah, Steven S.
Smith, and Mark D. Watts, "News Media, Candidates and Issues, and
Public Opinion in the 1996 Presidential Campaign," Journalism & Mass
Communication Quarterly 74 (winter 1997): 719-37.
12. Frederick Fico and William Cote, "Partisan and Structural Balance of News
Coverage of the 1998 Governor's Race in Michigan," Mass Communication
& Society 5 (spring 2002): 165-82; Frederick Fico and William Cote,
"Fairness and Balance of Stories in Newspaper Coverage of the 1996
Presidential Election, Journalism & Mass Communication Quarterly 71
(spring 1999): 124-37.
13. Susan Carter, Frederick Fico, and Jocelyn McCabe, "Partisan and Structural
Balance in Local Television Election Coverage," Journalism & Mass
Communication Quarterly 79 (spring 2002): 41-53.
14. Frederick Fico and Eric Freedman, "Bureau, Wire Reporters Write More
Balanced Stories," Newspaper Research Journal 25 (spring 2004):
44-57; Fico and Cote, "Partisan and Structural Balance of News
Coverage of the 1998 Governor's Race in Michigan;" Frederick Fico,
Geri Alumit Zeldes and Arvind Diddi, "Partisan and Structural Balance
of Local Television Election Coverage of Incumbent and Open
Gubernatorial Elections," Journalism & Mass Communication Quarterly
81 (winter 2004): 897-910.
15. Fico and Cote, "Partisan and Structural Balance in News Coverage of
the 1998 Governor's Race in Michigan;" Fico and Cote, "Fairness and
Balance of Stories in Newspaper Coverage of the 1996 Presidential Election."
16. Fico and Freedman, "Bureau, Wire Reporters Write More Balanced Stories."
17. Eric Freedman and Frederick Fico, "Male and Female Sources in Newspaper
Coverage of Male and Female Candidates in Open Races for Governor in
2002." Forthcoming in Mass Communication & Society.
18. Geri Alumit Zeldes and Frederick Fico, "Race and Gender: An Analysis of
the Sources and Reporters in the Networks' Coverage of the Year 2000
Presidential Campaign." Forthcoming in Mass Communication & Society.
19: Fico and Cote, "Partisan and Structural Balance in News Coverage of the
1998 Governor's Race in Michigan; Fico and Cote, "Fairness and
Balance of Stories in Newspaper Coverage of the 1996 Presidential
Election; and Fico and Freedman, "Bureau, Wire Reporters Write More
Balanced Stories."
20. John D. Richardson and Karen M. Lancendorfer, "Framing Affirmative
Action: The Influence of Race on Newspaper Editorial Responses to the
University of Michigan Cases," Press/Politics 9 (fall 2004): 74-94.
21. Stephanie Craft and Wayne Wanta, "Women in the Newsroom: Influences of
Female Editors and Reporters on the News Agenda," Journalism & Mass
Communication Quarterly 81 (spring 2004): 124-138.
22. Fico and Freedman, "Bureau, Wire Reporters Write More Balanced Stories."
23. Peter Clarke and Susan Evans, Covering Campaigns (Stanford, Calif:
Stanford University Press, 1983).
24. Lowry and Shidler, "The Sound Bites, the Biters and the Bitten."
25. Fico and Freedman, "Bureau, Wire Reporters Write More Balanced Stories;"
26. Fico et al., "Partisan and Structural Balance of Local Television Election
27. Coverage of Incumbent and Open Gubernatorial Elections."
26. Newspapers were: Anchorage Times, Democrat-Gazette, Los Angeles Times,
Miami Herald, Atlanta Journal-Constitution, Baltimore Sun, Saint
Louis Post-Dispatch, The State, and the Seattle Times.
27. Assertions were defined as sentences linked to identified sources
by verbs of
attribution. Such verbs included those of speaking, such as "said,"
"charged," "claimed," etc. Verbs denoting states of mind or feeling
such as "believes," "wants," etc. were also considered to establish
attribution when it was clear they were being used as synonyms for
verbs of speaking.
28. Newsroom diversity was based on a survey by the ASNE. The community's
diversity was assessed using U.S. Census Data. The index was
developed by Dedman and Doig: B. Dedman and S.K. Doig, "Does Your
Newspaper Reflect Its Community?" Paper presented at the annual
meeting of the American Society of Newspaper Editors, New Orleans,
LA, April, 2003. It can be accessed at http://www.asu.edu/cronkite/asne>.
The diversity index of one sample newspaper was unavailable. This newspaper
was given an index score equal to the mean of the others so that its
other data could be used in the multivariate analysis. Tabachnick
and Fidel recommend this procedure because it is conservative,
lowering the correlations using the missing data variable (and
therefore making it harder to support hypotheses). See, Barbara G.
Tabachnick and Linda S. Fidell, Using Multivariate Statistics (New
York, Harper & Row, 1983).
29. Circulation rank was used instead of raw circulation because of the great
disparities among the circulations of newspapers used in the study.
30. Schramm found in a post-World War II study that between 25 and 50 percent
of readers would leave a story by the sixth paragraph. Wilbur
Schramm, "Measuring Another Dimension of Newspaper Readership,"
Journalism Quarterly 24 (1947): 293-306. Newspapers must now compete
with television and the Internet, and other media.
Table 1: Partisan and Structural Imbalance Index Distributions
(Percent of stories with
index scores.*)
Partisan Imbalance Structural Imbalance
Index Index
Scale Scores
Most Imbalanced 4 2 4
3 7 19
2 12 31
1 11 26
Most Balanced 0 19 19
-1 15 NA
-2 19 NA
-3 11 NA
Most Imbalanced -4 2 NA
N 175 175
*Total percents may not equal 100 due to rounding.
Table 2: Influences on Partisan and Structural Imbalance in Stories
(Coefficients are
Betas. Betas marked with * are significant beyond
the .05 level).
Influence Level
And Type Partisan
Imbalance Structural (Favoring
Democrats) Imbalance
Societal
Race Type -.12 .08
Organizational
Circulation .14 -.03
Diversity Index .07 -.02
Newsroom
Story Prominence .08 -.05
Individual
Reporter Gender .20* -.07
Partisan Lead .02 .33*
Story
Partisan Imbalance NA -.05
Equation Significance .14 .01
R-Squared .06 .12