Betsi Grabe, a professor in Telecom at Indiana University, gave a talk Friday assessing the degree of visual bias in network news coverage of U.S. presidential candidates from 1992-2004. Her study, part of a larger series of studies examining the so-called “liberal bias” in the media, focused exclusively on visual images rather than other forms of content. Grabe’s conclusion was that a conspiracy was unlikely and in fact suggests a persistent visual pattern which favors Republicans over Democrats.
(NOTE: I tried my hand at live blogging for this, so I’d be interested in knowing if anyone read the earlier version of this post and made any sense from my notes.)
Grabe used to be a broadcast news director for the South African Broadcasting Corporation during a period of severe restrictions and censorship. Few cared about the images and still saw news as being textual, even on film and television media that were defined by their visual properties. Grabe became part of the “first generation” of people with visual training. She resented having to submit for review both an initial script and the finished product to administrative review and editing, often leading to shouting matches (“I was very young, very idealistic.”) with the old, white men making the decisions.
Once in such a review session, she noticed one of the men with eyes closed listening to the words of the documentary and ignoring what was on the screen. That was Grabe’s first insight that they were responding more to words than visuals, giving her hope of being able to tell her story through images. Grabe would likely still be in South Africa fighting those battles if she didn’t feel compelled to take a leave of absence to come to the U.S. to access some academic papers being filtered from her country. After enjoying the stark contrast in freedom of press and availability of content, Grabe found herself planting roots in America and moving from journalism into academia.
“Liberal bias” is an old accusation, emerging in the mid-1960s and the Nixon campaigns that led to his Presidency. The accusations, most evident in popular press, are not supported by academic research. According to Grabe, however, the current scholarly investigation into media bias suffers from three glaring problems:
- Focus on content — Researchers typically ignore visuals when studying media bias. If visuals are acknowledged at all, the concentration is on its content, such as the presence of American flags behind and surrounding a speaker. Such content may be more a product of image handlers than journalists, who cannot always control the scenery or the behavior of candidates they cover. Yet, many researchers view it as evidence of media bias.
- Bias in selection of study — Newspapers are perceived as more important than television, despite network news remaining the most used source of information, especially for those who vote. “As long as the New York Times is objective,” quips Grabe, “we can all sleep at night.” Analysis of television is reduced by academics to something lesser than that of traditional newspapers and magazines.
- Sample too narrow — Research tends to look at volume of content first, often limiting the selection to a particular election or even a specific candidate. There is a definite lack of longitudinal data on media bias, with broad conclusions drawn from indicental coverage.
The goals of Grabe’s study aim to address these shortcomings. She investigated visual properties only and tried to come up with some quantitative measure of journalistic intent. The data also spanned the last four Presidential elections from the period between Labor Day and Election Day to give the research some longitudinal value.
Why are Visuals Important?
Grabe offers a couple compelling arguments about how humans use such information. There are both biological and linguistic differences from the written word. Referencing Andrew Parker’s In the Blink of an Eye, she points out that the development of life was incredibly slow for many millions of years. However, something happened after the first few distinct forms of life evolved to spark an explosion of diversity in animals. Parker speculates that the Cambrian period, known as the “Big Bang” for animal life on this planet, was tied to the development of sight. Eyes were useful for mating and finding food, and that led to more diversity and adaptation. Even the placement of the eyes of a head can distinguish between prey and predator, for example the rabbit (on each side of the head) and the owl (facing forward on the front side). Grabe admitted to having a difficult time finding a photo for her presentation of a rabbit taken from front since almost universally they are photographed from the side … where the eyes are located.
Humans have been able to make sense of the world through visuals much longer than through writing. Long before the first writing occurred about 5000 years ago, there was plenty of evidence of communication and information gathering through visuals. Even the early written forms were largely visual: pictographic representations on cave walls, ideographic characters representing things, and phonetic letters drawn from the starting sounds of physical objects. Half of our neural pathways are dedicated just to processing visual input. When received simultaneously with other sensory input, the visual is dominant over the other kinds of information cues.
Linguistically, there is a difference as well. Visual images have an analogical or indexical quality that is inherently authentic. The lack of constructed appearance is more easily accepted and trusted … although arguably the Age of Photoshop and CGI is changing that. Grabe pointed to coverage of O.J. Simpson’s murder trial by Time and Newsweek. In the same week, the two major magazines ran the same mugshot photo on their covers. However, Time‘s OJ photo was darker and appeared more ominous than that of Newsweek. It caused an uproar and accusations of racism and bias, but it may never have been noticed if the two covers didn’t appear at the same time and in the same context. Visuals also lack explicitly propositional syntax. Words make causal claims, comparisons and generalization. Visuals can make the same claims but do so without explicit cues.
The Study
Grabe’s team of researchers measured volume and visual weight of network news coverage of recent Presidential races. She also examined the structure of the content, which more than anything else is in the journalist’s control.
- Volume — The number and duration of single-party stories were examined, as well as who dominated multi-party stories (predominately coded as “nobody dominated”). Volum also includes a count of sound bytes by party affiliation.
- Visual Weight — The architecture of the story can be described as a range of techniquess used in the industry that embed more and more suspicion about authenticity and objectivity. At the most trusted end of the specturm, journalists present their stories as a reader (someone sitting and reading the copy) or voice over (an off-camera reading). Then comes VO/SOT (voice over, sound on tape), where architecture starts to be more suspect, followe by package, interview, an in-depth package (like a series of reports offered over several broadcasts). Also of interest was the placement of a story within the newcast. A story that leads or is at least included in the first block of stories before commercial are considered more important and noticeable than those appearing later in the broadcast.
- Composition — This deals with packaging, not the content of a story. There are some widely-accepted conventions within the industry that are used for specific effect, and this is where journalists have the most say in how their work is conveyed to an audience. In editing, the “Goldilocks effect” occurs when a speaker has the last say (earlier can be a positive or negative, but never as strong as the final word). “Lip Flap” is the use of interview material as video coverage for narration. When the shot is too close, not using the speakers voice is considered a major violation of video journalism production standards. It is both confusing and symbolically allows the journalist to put words into the speaker’s mouth.
The camera controls are also tools for the journalist to shape a story. If the angle is looking up to the subject, it harkens the viewer back to days as a child looking up at an adult authority figure and carries a positive connotation. It is a negative to shoot a speaker from above. The length of the shot is also important. Extreme closeups (violates personal space) and long shots (no intimacy) are negatives, where closeups and medium shots balance the personal connection with environmental context. Likewise, zooming out is a negative because it pushes the viewer away from the speaker.
The samples included 178 newscasts and 437 campaign stories aired by three major networks (ABC, CBS, NBC). The stories were restricted to a period between Labor Day to Election Day in four presidential campaign cycles (1992, 1996, 2000 and 2004).
The results showed some differences in each measure. Volume, which is where the existing literature focuses, is not a good measure since it neglects the natural drama that journalists are drawn to. Ross Perot, for example, helped lift the “Other” category through his efforts as a candidate from the Reform Party. When he left the scene, so did the media coverage. When examinging Weight and Composition, the Republicans were noticeably favored. In fact, coded scores for Democrats were almost flatlining in terms of editing and camera shot selection.
In trying to answer the original charge of liberal media bias, Grabe draws the conclusion that such a conspiracy is unlikely. Volume tells only one dimension, but there is actually a persistent pattern of visually favoring Republicans over Democrats. Rather than attribute this to some designed master plan, Grabe suggests this is the effect of self-censorship by newsmakers. It is documented that part of the Republican strategy since Nixon has been to accuse the media of liberal bias (As former Republican Party Chair Rich Bond once told the Washington Post: “If you watch any great coach, what they try to do is ‘wrok the refs.’ Maybe the ref will cut you a little slack next time.”). As a result, journalists — afraid of losing their credibility as an objective source of news — overcompensate with decisions favorable to Republicans. The same may be happening in academia. Grabe admitted that she couldn’t bring herself to declaring a “conservative media bias” for fear of accusations of being part of a “liberal academic” agenda. Still, the results here show Republicans at the very least benefitted from the network coverage.
Discussion
Some questions naturally arise from this work. There isn’t any real distinguishing being done between the potential conspirators — the network suits — and the journalists. Grabe indicated that, at least at the level of a Presidential race, the journalists exert great control over how their stories are put on the air, including shot selection and editing decisions. That would make a Republican bias their responsibility, however, it would be helpful to try to distinguish between the input of network policies, journalistic intent, and political handlers.
I also wonder if bias is embedded in the context of the story, apart from such journalistic choices. The volume data clearly matches with the victors and the interest in a major 3rd party candidate (Ross Perot). If journalism is to blame, it may be more of a criticism of lazy investigation than an injection of political bias. There are many different forces at work during coverage of a Presidential campaign: The slow-moving transition of political power … A boom in technology … Diffusion of viewer interests due to cable television and the Internet … Contemporary events that involve specific individuals and spark additional coverage. Why, for instance, did this study not extend all the way back to Nixon campaigns, which was identified as the origin of this accusation? These kinds of factors need to be separated out of the data to see whether the trends hold.
Most importantly for my interests, can this data be examined as a complex system? Is there a “Wisdom of the Crowds” effect happening, where the individual editorial decisions of journalists are identifying winners? Particularly with the volume data, the bias may have been predicting which party would win, rather than causing it to happen. I also think that analysis of a network of political events, individuals and campaign coverage might yield some new information about how they relate.