Anonymous Sources at The New York Times

Wednesday, April 1, 2009
By Sharad Goel

Anonymous Source

Last week the Public Editor of The New York Times, Clark Hoyt, criticized the paper’s common practice of using anonymous sources, often in flagrant violation of its own policy. Drawing on several recent examples, Hoyt demonstrates that casually granting anonymity facilitates personal attacks without repercussion, provides cover for those who make false or misleading statements, and generally damages the credibility of the paper. (This is a problem Glenn Greenwald has written about extensively.)

Despite his censure, citing a 2008 study by a group of journalism students at Columbia University, Hoyt states that the use of anonymous sources at The Times has decreased by about half since a strengthened policy was instituted in 2004. However, somewhat puzzlingly, a search of two news archives shows that the number of NYT articles in which sources spoke on “condition of anonymity” has steadily grown since 2000, both before and after the policy change.

The Columbia study was based on an examination of six issues of the paper before and six issues after Executive Editor Bill Keller wrote a February 2004 memo outlining the paper’s new, stricter policy on the use of anonymous sources. The authors defined:

“an anonymous source as one where information was attributed to someone whose identity is indiscernible to the lay reader. This includes attributions such as “sources said” or “experts said,” without further description of the speaker or the reasons for granting anonymity.”

By this measure, as Hoyt states, the study finds an approximately 50 percent reduction in the use of anonymous sources after the new policy was adopted.

As opposed to analyzing specific issues of the paper, I estimated the use of anonymous sources by: (1) searching the Google News Archives for NYT articles containing any of the phrases “condition of anonymity,” “requested anonymity,” or “anonymous source;” and (2) searching The Times’ Archive for articles containing the phrase “condition of anonymity.” (Note: The Times’ archive search does not appear to support the broad match used on Google News.) Estimates from both archives indicate a steady rise in the use of anonymous sources by The Times since 2000.

Use of Anonymous Sources in the New York Times

There are several possible explanations for the rather stark discrepancy between the Columbia study and my own analysis. I suspect, though, the difference is largely a result of the archive search identifying only instances of requested anonymity—in which sources explicitly ask to remain anonymous—whereas the Columbia study also includes instances of designated anonymity—in which reporters fail to name a source despite their willingness to be cited (e.g., abstract references to “expert” opinion). While requested anonymity may have risen (as indicated by the archive search), that increase may have been offset by a drop in designated anonymity. Importantly, designated anonymity is often not nearly as consequential and prone to abuse as is requested anonymity, and the two should not be counted together.

Given these seemingly contradictory findings, further investigation is warranted on the use of anonymous sources, and in particular, on the asserted effectiveness of the The Times’ policy to curb their use. At the very least, the conclusion that the 2004 policy halved the use of anonymous sources at The Times likely requires important caveats.

Illustration by Kelly Savage

Tags: ,

  • doug rivers

    why the sharp drop between 1997 and 2000? It’s hard to believe the use of anonymous sources tripled over this period, which suggests that the method of identifying anonymous sources might change with different editors. Do all three searches terms increase similarly? Do you get the same patterns if you search for “asked not to be identified” or “senior officials” or similar terms?

    p.s. great blog!

  • http://www.cam.cornell.edu/~sharad/ Sharad Goel

    Doug, I agree that the precipitous drop from 1998 to 2000 is likely due to the very real limitations of this quick and dirty method of analysis. Interestingly, though, the four phrases “condition of anonymity,” “requested anonymity,” “anonymous source,” and your suggested phrase, “asked not to be identified” all increase two to threefold between 2000 and 2008, indicating at least some level of robustness. Given the myriad ways in which reporters refer to anonymous sources (e.g., “a senior State Department official”, “a Democratic strategist”, etc.), it seems quite difficult to mechanically measure use of anonymous sources.

    On a related note, a 2004 study commissioned by the then Public Editor of The Times, Daniel Okrent, found a slight INCREASE in the use of anonymous sources after the stricter policy was instituted. This again suggests that quantifying the use of anonymous sources is quite sensitive to how exactly one defines and identifies anonymity.

  • http://www.mahdian.info Mohammad

    Does the conclusion change if you normalize by the total number of articles? I imagine the total number of article hasn’t remained constant over time.

  • http://www.cam.cornell.edu/~sharad/ Sharad Goel

    Mohammad, the number of articles per year in The Times does appear to be relatively stable over the last 20 years. The “null search” on the NYT archive returns between 80K and 100K articles in any given year since 1990 (about 250/day), with 2008 actually a low point.

  • Achtung Cyclist

    Failing to bring my own data, I will say that as reader, I noticed an increase in perfunctory explanations for anonymity under the new policy.

    Could it be that the number of anonymous sources did not change so much as the number of words devoted to introducing or rationalizing an anonymous source did?

  • http://www.cam.cornell.edu/~sharad/ Sharad Goel

    Achtung, some version of your conjecture is almost certainly true. The observed increase in sources who speak on “condition of anonymity” may result from attempts to provide context, justification, or standardization. Given the upward trend since 2000, however, I think it is less likely that such reporting changes were in response to the 2004 policy.

    To be clear, given the limits of my analysis, I doubt there was actually a threefold increase in the use of anonymous sources between 2000 and 2008. But I also think it is unlikely that the stricter policy halved the use of sources who requested anonymity.