About PrezPix

The Methodology of this Study 

Do news outlets show bias in their selection of photos of politicians?

Photographs are not objective. In fact they can never be. But how partisan are news outlets in their choices of how to visually cover candidates running for office?  Are the ups and downs of the polls paralleled in ups and downs in the photographs of the candidates?

This study analyzed the photos that 21 major news outlets in the United States used to cover the 2012 presidential election — from the GOP primaries among Mitt Romney, Rick Santorum, Newt Gingrich and Ron Paul, in February and March, to the general election campaign between President Barack Obama and Governor Romney in September and October.

This study makes an effort to describe in a systematic way the components of a very large body of photos taken during several months of great political activity.  The study then attempts to methodically analyze those photographs.  Finally, in its conclusions, the study summarizes that analysis while yet remaining cognizant that no two viewers of a photograph will understand that image exactly the same:  each will bring his or her biases to the image, and both will look at and react to that image differently, especially if that image has some political content, as all of those studied do.

“Photographs taken from eye level suggest an equal relationship between the subject and viewer.”

— from a study of presidential photographs presented by Keith Greenwood at the 2005 annual conference of the International Communication Association

Challenges in conducting a research study of media use of photos

Historically very few research studies — either in the public sphere or in academe — have taken on the challenge of studying the photographic coverage of events, especially political events.

Watchdog groups — some with partisan agendas, others with an interest in fact-checking — regularly monitor candidates’ political advertisements and speeches.  Similarly activists and academics review media coverage of candidates, typically focusing on the extent of coverage or the content of that coverage.   There are very few instances, however, where there has been a systematic effort made to evaluate the photographic coverage of political candidates.

The research team assembled for this PrezPix study conducted an extensive literature review and found only a handful of other studies that it could use to inform the current project.  There was no recent study discovered that was of sufficiently similarity that it could provide a definitive template for selecting and categorizing images, or for coding those found.  After several months of review of existing/previous studies and of evaluating those studies’ strengths and limitations, the ICMPA research team generated its own questions and categories, then conducted a series of mock studies to further refine those selections, categorizing and coding rubrics.

It is perhaps important to note here, that despite the dearth of photographic study precedents, the researchers discovered that that people of all kinds — from everyday citizens to hardened social scientists — love to view and respond to photos.  In fact a motivation for this study was the phenomenal growth of Instagram and Pinterest as photographically driven social media platforms.  This PrezPix study draws on that interest.

Background to research study

Researchers at the International Center for Media and the Public Agenda at the University of Maryland, College Park, set out to consider the photographic coverage of the 2012 primary and general election seasons, and determine whether there had been any significant differences in how major U.S. news outlets — as accessed via their online news platforms — visually depicted the four remaining major Republican candidates for president in late February through late March 2012 and depicted the Republican and Democratic nominees for president in September and October 2012.

The researchers first compiled a list of top online news outlets, and considered outlets that are traditionally print, broadcast (TV and radio), and online.  Researchers considered a lengthly list of possible outlets, in part to ensure that there would be geographic diversity (including states and regions that would have upcoming primaries) and editorial diversity.   After several “dry runs” of Pinterest pinning to evaluate the provisional list of outlets, the GOP primary list of news outlets was whittled down to 18, as it became evident that quite a few outlets on the original list were not daily covering the GOP primary battles.  For the fall general election three of those original 18 news outlets (the Atlanta Journal Constitution, the Dallas Morning News and Philly.com) were dropped, so that three news outlets from swing states and states hosting the presidential debates (the Cleveland Post Dispatch, the Denver Post and the Miami Herald) could be added to the list.

Using the web platform Pinterest, researchers daily checked the news sites between February 25 to March 25, pinning every available photo of the four major GOP candidates. That time frame was selected to coincide with major primaries, including  Super Tuesday, March 6.  To organize the photographs, the researchers created boards on Pinterest for each news outlet and candidate.  Researchers followed the same process in fall 2012, pinning all available photographs of President Obama and Governor Romney for three weeks:  September 17 – 23, October 1 – 7 and 15 – 21.  The fall dates were selected to surround the scheduled presidential debates:  the first week fell before the first debate, the second week was the week of the first presidential debate on October 3, and the third week was the week of the second presidential debate on October 16.

The chart below lists the 21 news outlets surveyed in this PrezPix study, and further lists the numbers of photographs found and pinned for each of the candidates.  Note that these figures, while representing the number of photographs pinned, do not necessarily tally with the total number of photographs of the various candidates on those sites.  As noted below, some photos on some sites could not be physically be pinned and were therefore as a consequence not reviewed for this study.

Researchers made no distinctions in their pinning between photos taken by staff members of the news outlets and photographs taken by the wire services — although that distinction was (when known) mentioned on the Pinterest board captions.  With the advent of digital photography, there are so many images taken of every single event, that whether a news outlet relies on its own photographer(s) or on the photo agencies there is almost always dozens, if not hundreds of photos to select from.  Researchers also pinned photographs that were repeatedly used.  Most outlets recycled a substantial number of photos, using a single image to illustrate different articles (both on the same day and on different days). Every time a photo of a candidate appeared on the site, researchers counted and pinned it again — even when it was a duplicate.

Researchers for this study concluded that editors were exercising their choice of what to publish when they posted one (or even a handful) of photos to illustrate an article or to be a stand-alone image.  Their choices might be based on deadlines needs, budget, perceived news values, aesthetics, or some other determination.  It is this study’s effort to analyze that active choice that makes this study of especial value.

Simultaneously with pinning the photos to the Pinterest boards, the researchers designed a multi-question survey, using the online survey tool SurveyMonkey, aimed at recording what elements appeared in each of the pinned photos.  That survey also attempted to assess whether those elements created a overall negative or positive impression of the candidate pictured.

For each of the 8,000-plus photos collected, the coders recorded:

  • Whether the candidate was pictured in close up, from a medium range or from a far away distance.
  • Whether the candidate’s gaze was looking up, level or down
  • How the candidate was photographed: the photo was a ‘head shot,’ he was addressing an audience,  he was mingling with an audience, he was posing for a photo, he was speaking with media, etc.
  • Who (if anyone) also appeared in the photo with the candidate, including family members, staff, supporters, journalists, etc.
  • The expression on the candidate’s face, from positive (smiling, laughing, etc.) to neutral (listening, can’t see face, etc.), to negative (frowning, glaring, etc.)
Following the answers to those questions, coders asked two subjective questions to try to evaluate aesthetic considerations not otherwise “codeable”:
  • What adjectives best described the candidate in that specific photo?  (Options included:  charismatic/confident, friendly/approachable, honest/genuine, depressed/frustrated, vulnerable/lonely, cocky/arrogant, can’t see face, etc.)
  • What characterization best described the candidate in that specific photo? (Options included:  leader, family man, man of the people, isolated individual, stern taskmaster, stiff figure, etc.)
All these questions then factored into the final evaluation of each photograph:
  • Was the candidate in that specific photo portrayed very positively, slightly positively, neutrally or slightly negatively or very negatively.  (In cases where the factors didn’t lead to a clear response, the prompt was the following:  “If you were the candidate’s staffer would you be happy, unhappy or non-committal about the portrayal of your candidate in the photograph?”)

Following the recording of that data in SurveyMonkey, researchers then downloaded the SurveyMonkey data into Excel spreadsheets that could be uploaded into IBM’s Collaborative User Experience research group platform ManyEyes.  Using Many Eyes and Google Fusion researchers generated several types of data visualizations, such as bubble and pie charts, depending on their unique data sets, in order to best analyze the data.  The resulting analysis was broken down in two ways:  by individual news outlets and by candidate.  Those data visualizations informed the analysis of the coverage of individual candidates and the of specific news outlets.

Working with Pinterest

According to its website, Pinterest is an online pinboard that “lets you organize and share all the beautiful things you find on the web.” Researchers used Pinterest’s “Pin it” button to attach the found photos to “boards” on Pinterest, boards organized by by candidate and by news outlet. Throughout the months of this study, researchers pinned thousands of photos to a hundred and ten different boards.

When possible the researchers “selected” with a mouse the photo caption to the photo that they wanted to pin.  That “selection” then automatically appeared as a “caption” on the Pinterest pin.  When captions were absent or not available to grab (because of formatting issues), researchers selected headlines and other meta-type data, including bylines and datelines.  Researchers’ ability to grab information was idiosyncratic to the individual sites and to the types of pages on those sites; therefore there is no absolute uniformity to what textual information appears on the photo pins.

A benefit to researchers in using Pinterest was the ease in grabbing and organizing the photos of the candidates found in the 21 news outlets. Also, due to the nature of the “Pin it” button, each pinned photo was also automatically linked back to the original source, providing a high level of accuracy and transparency.  (NB:  when URLs on the originating sites are updated, the links from the Pinterest boards are broken, and as a result it is no longer possible to view the pinned photograph in its original location.)

Researchers did find, however, some significant limitations to using Pinterest — although it should be noted that those limitations did not outweigh the very strong advantages.  In fact this study would not have been possible without the use of Pinterest, or an analogue-type platform.

One limitation of Pinterest’s platform is that while it allows the organization of photos grabbed from various sites to be categorized into (“pinned on”) specific boards, once the photos were pinned to a board they could not be rearranged. This created some problems when organizing photos for coding and analysis.

From the other side, the side of the news outlets, there were also limitations.  Primary among those were that there were discrepancies in what photographs could be “pinned” on the various sites.  Some sites allowed the pinning of thumbnail photos — say in a frontpage teaser to an internal article — while other outlets did not.  A very few sites allowed photos in photo galleries to be pinned, while most only allowed photos that accompanied articles to be pinned.  (Note that specific problems in syncing Pinterest with individual news outlets are recorded on the pages relating to those outlets.  See also the page Pinterest as a Research Tool.)

The Research Team

Dr. Susan Moeller supervised the research for ICMPA at the Philip Merrill College of Journalism at the University of Maryland, College Park.  Seventeen Merrill College journalism students took roles in the pinning of photos to Pinterest and in the research, analysis and writing/editing of the findings of photographs of the GOP primaries, and 18 Merrill students took similar roles in pinning of photos to Pinterest and in the identification, analysis, and writing up of the findings of photographs of the general election.  Josh Cooper designed the PrezPix logo.  Minsoo Chung, a post-doctoral researcher assisted in the editing and oversight of this study.


  • Deema Alfadl
  • Julie Baughman
  • Michelle Chan
  • Joshua Cooper
  • Felicia Garay-Stanton
  • Catherine Irwin
  • Sarah Katz-Hyman
  • Alexa Kravitz
  • D’Ambour Lewis
    • Jason Lewis
    • Travis Mewhirter
    • Rebecca Pang
    • Julie Peak
    • Claire Saravia
    • Kaila Stein
    • Conor Walsh
    • Jeffrey Williamson

  • Gaby Arancibia
  • John Brill
  • Kaitlin Bulavinetz
  • Haley Bull
  • Matthieu Drotar
  • Erin Egan
  • Lauren Holstein
  • Emma Kantrowitz
  • Greg Kohn
  • Taylor Lewis
  • Gabby Siskind
  • Naomi Tesfai
  • Chris Trevino
  • Leah Villanueva
  • Jared Wasserman
  • Tyler Weyant
  • May Wildman
  • Kate Yoon