by the Committee of Concerned Journalists, February 18, 1998
From the earliest moments of the Clinton crisis, the press routinely intermingled reporting with opinion and speculation--even on the front page--according to a new systematic study of what and how the press reported.
The study raises basic questions about the standards of American journalism and whether the press is in the business of reporting facts or something else.
As the story was breaking, the two source rule for anonymous sources was not dead, but it was not the rule.
A large percentage of the reportage had no sourcing.
The study, designed by the Committee of Concerned Journalists and conducted by Princeton Survey Research Associates, involved a detailed examination of the 1,565 statements and allegations contained in the reporting by major television programs, newspapers and magazines over the first six days of the crisis. The goal was to find out what this cross section of the news media actually provided the American people and what the level of verification was.
Among the findings:
- Four in ten statements (41% of the reportage) were not factual reporting at all-here is what happened--but were instead journalists offering analysis, opinion, speculation or judgment.
- Forty percent of all reporting based on anonymous sourcing was from a single source.
- Only one statement in a hundred (1% of the reporting) was based on two or more named sources.
- News organizations that had better sources generally relied less on analysis and opinion in their reportage.
In a finding that may account for the widely reported public complaint that journalists rushed to judgment, the most common statement by journalists was a conclusion--that Clinton was in big trouble. That interpretation was reported even more often than the core allegations against the President, his denial and the ensuing investigation. The next two most common statements by journalists were also conclusions: that the President was dissembling and that impeachment was a possibility. From the first hours, journalists had, in effect, placed judgmental statements like quotation marks around the core fact on which the story was based.
As the story unfolded, the reliance on named sources and factual reporting tended to rise and the level of commentary and speculation dropped. But that also highlights the insistence to jump to conclusions, especially by news organizations that have the fewest facts.
The study raises such questions as: What are the standards for American journalism in this newly competitive atmosphere? Are we watching them change? Was the standard in the early days of this story, "do we think it's true?" or was the standard "how can we get it in?"
Other Overall Findings
Looked at another way, the picture that emerges is of a news culture that is increasingly involved with disseminating information rather than gathering it. For instance:
- If the amount of punditry and unverified reporting passed along from other news outlets is added together, it reveals that nearly one in three statements (30% of what was reported) was effectively based on no sourcing at all by the news outlet publishing it.
- Only one in four statements (26%) was based on named sources (overwhelmingly one named source).
- The rest, 23%, was what we called analysis--that is interpretative reporting attributed to some sourcing so that the audience could evaluate its credibility.
- The fact that almost half of all the reporting was punditry and analysis may be one reason the public is irritated with the press. Public opinion polls such as those by the Pew Research Center for the People and the Press showed that 80% of the public felt there was too much commentary in the coverage.
In all, 21% of the reporting was based on anonymous sources. Given the nature of a story involving a grand jury and an ongoing investigation, that may not be so surprising, and some of this reportage three weeks later holds up well.
In general, however, the track record of stories with multiple anonymous sources appeared far stronger than those with one.
For instance, weeks later one story that stands out for being unproven--that Monica Lewinsky kept a blue dress stained with DNA evidence of an affair--was initially based on a single anonymous source.
Nearly a week after the blue dress story was first aired on ABC and then repeated in several news outlets, including The New York Times, the FBI reported it had found no such evidence. It is possible today that such a dress exists and perhaps even was returned to Betty Currie, the White House secretary, according to yet another anonymously sourced story.
Yet this also may be a textbook example of consider the source. ABC described its source as "someone with specific knowledge of what it is Monica Lewinsky says really took place." In a subsequent interview with the New York Daily News, Linda Tripp's literary agent friend Lucianna Goldberg, a woman with a history of antipathy for Clinton and for engaging in dirty tricks for the Republican party, openly said that she was the source for the blue dress allegation. "The dress story? I think I leaked that." Goldberg told the Daily News laughing in a way that suggested she was mocking the press with this and other leaks. "I had to do something to get their (the media's attention). I've done it. I'm not unproud of it."
Overall, the press often did little to offer audiences a hint of the possible bias of anonymous sources that might have colored the reliability or completeness of what they were leaking. This was particularly true in some of the stories that remain unverified. One such story, for instance, is that a White House steward told the Grand Jury that he had witnessed an intimate encounter between Lewinsky and the President. The Wall Street Journal attributed the story simply to "two individuals familiar with (the steward's) testimony." Similarly, another story that remains unproven was an ABC report that more than one White House staffer, perhaps secret service agents, witnessed an intimate encounter between Lewinsky and the President. ABC attributed this story simply to "several sources."
Many of the anonymous sources in this crisis--even those close to events--might have an axe to grind and needed to be treated with greater discretion than many of the stories demonstrated. In general, indeed, the press tended to make information look better sourced than it was.
When one news organization broke an especially controversial story that others couldn't confirm, there was widespread tendency by other media to pick it up without verifying it. The day after ABC reported the blue dress story, for instance, the percentage of reporting attributed only to other news organization spiked to 18%, the highest single day in the study.
Sometimes journalists seemed fascinated with the most salacious details, even if unverified, such as the meaning of oral sex or the background of Monica Lewinsky. On the Today Show January 22, for instance, Matt Lauer repeatedly tried to get Newsweek's Michael Isikoff to admit whether he had "heard anything" about a semen stained dress. Even after Isikoff said an answer would be irresponsible, Lauer pressed him, for the third time. "You're not telling me whether you've ever heard of it?"
What We Looked At
The study measured a snapshot of the news media culture in the first week of the story. From Wednesday January 21 through Saturday, January 24, we studied the nightly newscasts, prime time magazines and specials, and relevant segments of Larry King and Charlie Rose, Nightline, the morning news shows, the front page coverage of the New York Times, Los Angeles Times, St. Louis Post Dispatch, the Washington Post, and the Washington Times. Added to that universe, we studied the Sunday network talk programs and the Monday news magazines, Time and Newsweek.
Based on ratings, influence, and the degree to which their work found their way into other reports, these outlets represented a fair picture of how Americans learned about this story. Indeed, because we wanted to study those outlets that presumably were doing original reporting or interviewing, we deliberately did not include local television, the most popular news source, in the study.
In order to most thoroughly and accurately record press performance, the study did not just measure stories, since some contained more than one key point. It measured instead the key assertions inside stories. Thus in a piece stating that Monica Lewinsky alleged having sexual relations with the President and that Clinton denied the allegation, these two statements were measured separately.
The goal was to find out what the news media was actually providing audiences. How much of the coverage of this story was factual reporting-here is what happened? What was the level of sourcing for that reporting? How much was analytical-that is analysis attributed to some reporting or evidence in a way that the audience can evaluate how it was arrived at?
How much fell into a different category-one you might call punditry? We included here three categories of assertions. 1) Opinion, which is analysis not attributed to anything. 2) Speculation, which is opinion based on facts that do not yet exist. 3) Judgment-an unequivocal assertion that leaves no room for dissent-Clinton is liar, Clinton cannot survive.
When it comes to analysis or punditry, the study measured what journalists themselves asserted, not what their sources or TV interviewees had to say.
What the Press Reported
The most common statement by journalists in the first days of the story was interpretative: that Clinton was in big trouble. Most often--more than a third of the time--reporters based this conclusion on their own opinion or speculation. Roughly a quarter of the time, journalists offered this as an analysis but cited some reporting to support it. Only 17% of the time did journalists cite named sources for this conclusion. Eleven percent of the time it was cited to another media source.
The second most common assertion--that Clinton denied the allegations--was usually attributed to Clinton himself in interviews he had granted.
Given the limited number of reporters who actually had listened to the tapes or interviewed Linda Tripp, most news organizations did not have any confirmation of the major allegation that drove this story--that Lewinsky had talked about having an affair with Clinton and the possibility of lying about it. In only 4% of cases was that allegation attributed to a named source. In more than six in ten cases it was attributed to other sources or offered as part of an analysis. In a third of the cases the news organization offered anonymous sources for that statement.
The fourth most common statement in the first week was that Lewinsky was negotiating for immunity with Kenneth Starr's office. Due in large part to the visibility of Lewinsky's attorney, this was most often attributed to a named or anonymous source. A third of the time it was analysis.
The next two most common statements were particularly judgmental: that Clinton was engaged in double talk and that impeachment was a possibility.
When it came to impeachment, four out of ten times that statement was attributed to named source, making it one of the hardest sourced allegations in the study. An equal amount of the time it was came from reporters offering their own opinion, speculation or judgment.
As for Clinton dissembling, the most common basis for that assertion was reporter's own opinion, speculation or judgement, about a third of the time. A quarter of the time reporters offered some attribution for that analysis. In one out of five cases it was attributed to a named source. Another one out of five times it was attributed to another news outlet.
Breakdown by News Genre
Overall, 59% of the reportage was factual reporting--it described what had happened. This reporting had several levels of verification: from multiple named sources, to a single anonymous source to another news outlet.
Looking just at this universe of factual reporting, substantially less than half, only 42%, was based on named sources.
More than a third, 35%, was based on anonymous sources. Another 21% was unverified by the news outlet reporting it and instead was taken from some other news outlet.
Thus, in all, more than half of the universe of factual reporting, 58%, was based on anonymous sourcing or another news outlet.
Clearly, at least in the first week of this story, it was not always the rule to leaven interpretation with evidence that would allow the consumer to assess how much the reporter knew. Roughly half the time, there was no evidence offered. The lines between opinion and analysis were not closely observed in the news pages or the news programs.
Overall 4 in 10 statements by journalists were interpretation (everything from attributed analysis to speculation). Of this interpretative universe, slightly more than half (55%) was analysis attributed or supported by some reporting, thus allowing the consumer to assess its credibility.
The rest, 45%, might be called punditry--that is the interpretation was not supported by any sourcing) Broken down, 18% of all interpretative reporting was opinion; 21% was speculation (opinions about events that had not yet happened); and 5% was judgment, (unequivocal conclusions by the reporter that left no room for disagreement--the president is a liar, the president cannot survive.)
The Washington Post was the most aggressive of the newspapers studied in using anonymous sources-including a single anonymous source. Only 16% of its reporting in the first few days of the story was based on named sourcing, significantly lower than the average. On the other hand, 38% of its reporting was based on two anonymous sources, and 26% of its reporting was based on a single anonymous source, in both cases more than triple the average.
The New York Times was more conservative: In it's pages, 53% of the reportage was based on named sources. Less than 8% of its reporting was based on a single anonymous source. At the Los Angeles Times, 43% of its reportage was based on named sources, and 9% on a single anonymous source. The Washington Times based 36% of its reportage on named sources and 3.4% on a single unnamed source. On the other hand, the Washington Times was also more subjective in its reportage. It published more than double the amount of analysis of newspapers (23%) and more than double the amount of speculation (6%).
There were notable differences between networks, as well. CNN had the lowest level of reporting based on named sources, 18.5%, versus 22% at NBC, 24% at ABC and 26% at CBS.
CNN also stood out for allowing its reporters to engage in opinion unattributed to any reporting whatsoever. Nearly 30% of all their reportage was opinion. That is higher than any other network, or any other genre of new outlet.
On its Sunday program, CNN Late Edition, 26% of all statements journalists made on the program were unattributed opinions, more than double any other Sunday talk program other than the McLaughlin Group (which was 25% opinion). Late Edition, however, did not engage in any speculation or judgment. Thus, when opinion, speculation and judgment are factored together as total punditry, Late Edition had the lowest percent of statements (remaining at 26%). The McLauglin Group had the most statements that were total punditry (68%) followed by Meet the Press with 42% punditry.
On its nightly newscast, CNN The World Today, similarly, the level of unattributed and opinion and speculation were double that of any other evening newscast, 15% opinion and 10% speculation.
ABC's Nightline was the most factual news outlet of all those studied. More than 76% of all statements on Nightline were factual reporting. It had the highest level of reporting based on named sources of any TV show, 35%, and also one of the higher levels of reporting based on a single anonymous source, 15%. While it engaged in less analysis and punditry, Nightline also tended more often than other news outlets to air reporting from other news organizations it had been unable to verify itself, 17%.
The single most aggressive news organization when it came to relying on a single anonymous source was ABC News. Across all its programs, 14% of ABC News reporting was based on a single blind source. That compares with 8% for all the news media, and is double any other TV network. Of all of all news outlets studied, ABC News Good Morning America relied on a single blind source 22% of time, nearly triple the average. Prime time news magazines were the most analytical genre of program and had the least reportage based on named sources; 42% of what journalists said on such programs was analysis, and 21% of what they reported was based on named sources.
There were also distinct differences between evening network newscasts. CBS Evening News was the most judgmental (5.6% versus 2.6% at ABC's World News and 0% at NBC Evening News and 0% at CNN's The World Tonight).
The Sunday Shows
The Sunday shows relied far more on speculation, judgment and opinion than the rest of the press. In all, four in ten statements on the Sunday shows were opinion, speculation or judgment by reporters offered without any attribution as to what the reporter based that opinion on, more than double the press generally.
The McLaughlin Group defies categorization. Seventy percent of what appeared on that program was punditry (25% opinion, 36% speculation, 7% judgment). That is nearly double the level of punditry on either night-time talk shows like Larry King or Sunday talk programs like Meet the Press.
The length of time Newsweek had spent working this story, and its access to Linda Tripp and other sources driving it, showed in the study. Newsweek had roughly double the amount of reporting based on named sources (30%) versus Time (13%) in the first week. Newsweek also had less reporting based on other news outlets (2% versus 14% in Time).
Perhaps because it had more original reporting, Newsweek also had less analysis (33%) than did Time (49%).
Looking at a different category, the news magazines were the most aggressive when it came to inferring lessons about Bill Clinton's interior life or psychological motivations in this story. In all of the reporting, for instance, there were 10 instances in which journalists suggested that Bill Clinton had a sex addiction. Six of these occurred in one Monday's editions of Time and Newsweek.
In general, traditional news outlets tended to invest more in reporting and verification. Less than three percent of the reportage in newspapers, and only 12% on the nightly newscasts was punditry, compared with roughly 25% in all of news media.
Days of the Week
The press tended to leap to conclusions early on this story and then pull back. More than four in ten statements on the first day were either analysis or punditry (43%), declining each day thereafter until the Sunday talk programs, when it spiked upward again. Similarly, the reliance on named sources grew over time, rising from 17% on Wednesday to 36% on Saturday.
Selection/Inclusion of Broadcasts and Publications
Newspapers, magazines, and news broadcasts were selected on an ad hoc basis to provide a snapshot of nationally influential media., keeping in mind the importance of audience, ownership and editorial diversity.
Newspaper and magazine stories were either downloaded in their entirety from the NEXIS database, publication websites or were clipped from the publication in which they originally appeared. (Note: when the NEXIS database produced similar stories from the same newspaper, but different editions, the longer of the two stories was coded.)
Broadcast stories transcripts were acquired via network websites or professional transcript services.
Coders analyzed each news story(1) in its entirety, identifying the initial appearance of any individual statement/allegation within that story: these each became acase. (A news story was likely to contain multiple statements/allegations, and therefore, multiple cases.) Coders next analyzed all references to the particular case within said story, and the sourcing attributed in each instance.
When a statement or allegation had multiple appearances within one story it was considered only one case. Coders then identified all attribution cited by the journalist, and coded the case on the basis of the highest level of sourcing that appeared.
Intercoder reliability measures the rate at which two coders, operating independently of one another, code the same material in the same way. This monitoring occurred throughout the coding process, and no significant systematic errors were found.
Some of the columns in the tables may not add to 100% due to rounding.
1. For talk programs and Sunday shows, news stories were generally defined by commercial breaks. The exception: when the anchor/host went to a correspondent outside the studio or presented a prerecorded piece, that was considered a separate story.