Anonymous attribution during two periods of military conflict: Using logistic regression to study veiled sources in American newspapers

Journalism and Mass Communication Quaterly

Volumn 74, Issue 3, 1997, Pages 565-578

  Denham, Bryan E ^a  

^a MISSOURI STATE UNIVERSITY   (United States)

Author keywords

[No Author keywords available]

Indexed keywords

EID: 0040063372     PISSN: 10776990     EISSN: None     Source Type: Journal    
DOI: 10.1177/107769909707400310     Document Type: Article

References (37)

1. For additional discussion, see William Blankenburg, "The Utility of Anonymous Attribution," Newspaper Research Journal 12 (winter/spring 1992): 10-23; Bryan Denham, "Anonymous Attribution and Official News Sources in the Reporting of Bosnia and Somalia" (paper presented at the Southeast Colloquium of AEJMC, Roanoke, VA, 1996); Bryan Denham, "Anonymous Government Sources in the New York Times: Coverage of the Middle East Peace Accords, the Conflict in Bosnia and the Clinton Health Care Package" (paper presented at the annual meeting of AEJMC, Atlanta, GA, 1994); K. Tim Wulfemeyer and Lori McFadden, "Anonymous Attribution in Network News," Journalism Quarterly 63 (autumn 1986): 468-73; K. Tim Wulfemeyer, "How and Why Anonymous Attribution is Used by Time and Newsweek," Journalism Quarterly 62 (spring 1985): 81-86, 126; K. Tim Wulfemeyer, "Use of Anonymous Sources in Journalism," Newspaper Research Journal 4 (winter 1983): 43-50.
2. For additional discussion, see William Blankenburg, "The Utility of Anonymous Attribution," Newspaper Research Journal 12 (winter/spring 1992): 10-23; Bryan Denham, "Anonymous Attribution and Official News Sources in the Reporting of Bosnia and Somalia" (paper presented at the Southeast Colloquium of AEJMC, Roanoke, VA, 1996); Bryan Denham, "Anonymous Government Sources in the New York Times: Coverage of the Middle East Peace Accords, the Conflict in Bosnia and the Clinton Health Care Package" (paper presented at the annual meeting of AEJMC, Atlanta, GA, 1994); K. Tim Wulfemeyer and Lori McFadden, "Anonymous Attribution in Network News," Journalism Quarterly 63 (autumn 1986): 468-73; K. Tim Wulfemeyer, "How and Why Anonymous Attribution is Used by Time and Newsweek," Journalism Quarterly 62 (spring 1985): 81-86, 126; K. Tim Wulfemeyer, "Use of Anonymous Sources in Journalism," Newspaper Research Journal 4 (winter 1983): 43-50.
3. For additional discussion, see William Blankenburg, "The Utility of Anonymous Attribution," Newspaper Research Journal 12 (winter/spring 1992): 10-23; Bryan Denham, "Anonymous Attribution and Official News Sources in the Reporting of Bosnia and Somalia" (paper presented at the Southeast Colloquium of AEJMC, Roanoke, VA, 1996); Bryan Denham, "Anonymous Government Sources in the New York Times: Coverage of the Middle East Peace Accords, the Conflict in Bosnia and the Clinton Health Care Package" (paper presented at the annual meeting of AEJMC, Atlanta, GA, 1994); K. Tim Wulfemeyer and Lori McFadden, "Anonymous Attribution in Network News," Journalism Quarterly 63 (autumn 1986): 468-73; K. Tim Wulfemeyer, "How and Why Anonymous Attribution is Used by Time and Newsweek," Journalism Quarterly 62 (spring 1985): 81-86, 126; K. Tim Wulfemeyer, "Use of Anonymous Sources in Journalism," Newspaper Research Journal 4 (winter 1983): 43-50.
4. For additional discussion, see William Blankenburg, "The Utility of Anonymous Attribution," Newspaper Research Journal 12 (winter/spring 1992): 10-23; Bryan Denham, "Anonymous Attribution and Official News Sources in the Reporting of Bosnia and Somalia" (paper presented at the Southeast Colloquium of AEJMC, Roanoke, VA, 1996); Bryan Denham, "Anonymous Government Sources in the New York Times: Coverage of the Middle East Peace Accords, the Conflict in Bosnia and the Clinton Health Care Package" (paper presented at the annual meeting of AEJMC, Atlanta, GA, 1994); K. Tim Wulfemeyer and Lori McFadden, "Anonymous Attribution in Network News," Journalism Quarterly 63 (autumn 1986): 468-73; K. Tim Wulfemeyer, "How and Why Anonymous Attribution is Used by Time and Newsweek," Journalism Quarterly 62 (spring 1985): 81-86, 126; K. Tim Wulfemeyer, "Use of Anonymous Sources in Journalism," Newspaper Research Journal 4 (winter 1983): 43-50.
5. For additional discussion, see William Blankenburg, "The Utility of Anonymous Attribution," Newspaper Research Journal 12 (winter/spring 1992): 10-23; Bryan Denham, "Anonymous Attribution and Official News Sources in the Reporting of Bosnia and Somalia" (paper presented at the Southeast Colloquium of AEJMC, Roanoke, VA, 1996); Bryan Denham, "Anonymous Government Sources in the New York Times: Coverage of the Middle East Peace Accords, the Conflict in Bosnia and the Clinton Health Care Package" (paper presented at the annual meeting of AEJMC, Atlanta, GA, 1994); K. Tim Wulfemeyer and Lori McFadden, "Anonymous Attribution in Network News," Journalism Quarterly 63 (autumn 1986): 468-73; K. Tim Wulfemeyer, "How and Why Anonymous Attribution is Used by Time and Newsweek," Journalism Quarterly 62 (spring 1985): 81-86, 126; K. Tim Wulfemeyer, "Use of Anonymous Sources in Journalism," Newspaper Research Journal 4 (winter 1983): 43-50.
6. For additional discussion, see William Blankenburg, "The Utility of Anonymous Attribution," Newspaper Research Journal 12 (winter/spring 1992): 10-23; Bryan Denham, "Anonymous Attribution and Official News Sources in the Reporting of Bosnia and Somalia" (paper presented at the Southeast Colloquium of AEJMC, Roanoke, VA, 1996); Bryan Denham, "Anonymous Government Sources in the New York Times: Coverage of the Middle East Peace Accords, the Conflict in Bosnia and the Clinton Health Care Package" (paper presented at the annual meeting of AEJMC, Atlanta, GA, 1994); K. Tim Wulfemeyer and Lori McFadden, "Anonymous Attribution in Network News," Journalism Quarterly 63 (autumn 1986): 468-73; K. Tim Wulfemeyer, "How and Why Anonymous Attribution is Used by Time and Newsweek," Journalism Quarterly 62 (spring 1985): 81-86, 126; K. Tim Wulfemeyer, "Use of Anonymous Sources in Journalism," Newspaper Research Journal 4 (winter 1983): 43-50.
7. For additional discussion, see Hugh M. Culbertson, "Leaks - A Dilemma for Editors as well as Officials," Journalism Quarterly 57 (autumn 1980): 402-408, 535; Hugh M. Culbertson, "Survey Shows Editors Divided on Handling Unnamed Sources," Presstime, August 1980, 25; Hugh M. Culbertson," Veiled Attribution - An Element of Style?" Journalism Quarterly 55 (autumn 1978): 456-65; James Kilpatrick, "'Trust Me' Journalism: An Identifiable Source is Fed Up," Washington Journalism Review (January-February 1988): 43-45.
8. For additional discussion, see Hugh M. Culbertson, "Leaks - A Dilemma for Editors as well as Officials," Journalism Quarterly 57 (autumn 1980): 402-408, 535; Hugh M. Culbertson, "Survey Shows Editors Divided on Handling Unnamed Sources," Presstime, August 1980, 25; Hugh M. Culbertson," Veiled Attribution - An Element of Style?" Journalism Quarterly 55 (autumn 1978): 456-65; James Kilpatrick, "'Trust Me' Journalism: An Identifiable Source is Fed Up," Washington Journalism Review (January-February 1988): 43-45.
9. For additional discussion, see Hugh M. Culbertson, "Leaks - A Dilemma for Editors as well as Officials," Journalism Quarterly 57 (autumn 1980): 402-408, 535; Hugh M. Culbertson, "Survey Shows Editors Divided on Handling Unnamed Sources," Presstime, August 1980, 25; Hugh M. Culbertson," Veiled Attribution - An Element of Style?" Journalism Quarterly 55 (autumn 1978): 456-65; James Kilpatrick, "'Trust Me' Journalism: An Identifiable Source is Fed Up," Washington Journalism Review (January-February 1988): 43-45.
10. For additional discussion, see Hugh M. Culbertson, "Leaks - A Dilemma for Editors as well as Officials," Journalism Quarterly 57 (autumn 1980): 402-408, 535; Hugh M. Culbertson, "Survey Shows Editors Divided on Handling Unnamed Sources," Presstime, August 1980, 25; Hugh M. Culbertson," Veiled Attribution - An Element of Style?" Journalism Quarterly 55 (autumn 1978): 456-65; James Kilpatrick, "'Trust Me' Journalism: An Identifiable Source is Fed Up," Washington Journalism Review (January-February 1988): 43-45.
11. Neuharth was a guest on the 4 June 1994 edition of This Week with David Brinkley on ABC. His segment, on anonymous sources, came in the wake of anonymous tips being reported without verification in the O.J. Simpson case.
12. Douglas Anderson, "How Newspaper Editors Reacted to the Post's Pulitzer Prize Hoax," Journalism Quarterly 59 (autumn 1982): 363-66.
13. These issues have been explored in previous studies of anonymous attribution and media sociology. For example, see Daniel Hallin, Robert Manoff, and Judy Weddle, "Sourcing Patterns of National Security Reporters," Journalism Quarterly 70 (winter 1993): 753-66; Daniel Riffe and Gail Johnson, "Unnamed Sources in White House Coverage," (paper presented at the annual meeting of AEJMC, Washington, DC, 1995).
14. These issues have been explored in previous studies of anonymous attribution and media sociology. For example, see Daniel Hallin, Robert Manoff, and Judy Weddle, "Sourcing Patterns of National Security Reporters," Journalism Quarterly 70 (winter 1993): 753-66; Daniel Riffe and Gail Johnson, "Unnamed Sources in White House Coverage," (paper presented at the annual meeting of AEJMC, Washington, DC, 1995).
15. In addition to the studies cited in notes 1 and 2, see F. Dennis Hale, "Unnamed News Sources: Their Impact on Perceptions of Stories," Newspaper Research Journal 5 (winter 1983): 49-56; Daniel Riffe, "Relative Credibility Revisited: How 18 Unnamed Sources are Rated," Journalism Quarterly 57 (winter 1980): 618-23.
16. In addition to the studies cited in notes 1 and 2, see F. Dennis Hale, "Unnamed News Sources: Their Impact on Perceptions of Stories," Newspaper Research Journal 5 (winter 1983): 49-56; Daniel Riffe, "Relative Credibility Revisited: How 18 Unnamed Sources are Rated," Journalism Quarterly 57 (winter 1980): 618-23.
17. It is important to note that the AP stories analyzed were written by AP reporters, not reporters from member newspapers.
18. Preliminary searches revealed that "Bosnia" and "Somalia" were the most effective search terms, as nearly all coverage focused on some aspect of each conflict. As indicated, coverage of both conflicts peaked in 1993, with extensive coverage in the year prior and the year after. Inclusion of news reports about the 1995 commitment of U.S. forces to Bosnia would have introduced a problem for the comparative study, as there has been virtually no coverage of Somalia since 1994; even then, many of the articles that mentioned the conflict in Somalia focused primarily on the problems occurring in Haiti, Liberia, and Rwanda. For randomization, intervals were created in which a population of articles from a given year was divided by the number of articles needed for the sample; a table of random numbers was then used to select one article from each interval. As an example, in 1993 the search term "Bosnia" yielded 6,317 articles in the Associated Press; this represented 42% of all AP articles containing that search term between 1992 and 1994. To arrive at a sufficient number of paragraphs for inferential statistics (3,000 for each conflict in each newspaper, before attrition), a total of 164 AP articles about Bosnia had to be downloaded. Forty-two percent of 164 is 69, and 6,317 divided by 69 is 92. Thus, the 1993 intervals were 0, 92, 184, 276, etc., and the random number added to each figure yielded the article to be downloaded. This stratification ensured that (1) articles published throughout a given year would be included in the analysis, and (2) the sample would include an appropriate number of articles from each year.
19. In terms of operationalizing "hard news," articles had to report specific issues and events and not contain editorial commentary. They had to originate from a news desk, be published independent of the editorial page and contain a dateline. No articles identified as "news analysis" were included, nor were any that were clearly features (e.g., a 3,500-word, one-source article of how the conflict in Bosnia had affected a local merchant). Ideally, only "hard news" articles would have been downloaded from Dialogue; however, search parameters did not allow these articles to be gathered independently of others. Because certain types of articles in the sample were not analyzed, then, considerable attrition did occur. Attrition did not pose methodological problems, since the analysis focused on articles with specific characteristics and generalized only to "hard news" reporting in the conflict coverage of three news organizations.
20. As mentioned, this generated a value between 0 and 1 for every article in the study. To facilitate data analysis, the values were then truncated, multiplied by 10, rounded to the lesser of two whole numbers (i.e. if 1<=ratio<2, then Level of Attribution=1), and the resulting figures formed the response variable. So, if the Washington Post really did use more anonymous attribution than the Los Angeles Times or the Associated Press -as observed in a cross-tabulation with the news paragraph as the unit of measurement - then most of its articles would appear to the right on a given number line. For example, consider a Post article with a value of .56, a Times article with a value of .38, and an AP article with a value of .29. Truncate each value, multiply by 10, round to a whole number, and they appear at the following points: 1 <2 < 3 < 4 < 5 < 6 AP LAT WP After all values were rounded, the LOGISTIC procedure in SAS was used to fit the best model to the data. Level of Attribution was the only variable in the study that contained subjective components, such that intercoder reliability needed to be examined. The author and a professional journalist coded 100 paragraphs for Level of Attribution and reached agreement 95 times. In the study, each paragraph was coded for the presence of one source. If more than one source appeared in a paragraph, the first source mentioned was coded. Given the brevity of newspaper paragraphs and the observations made in the pilot study, the presence of more than one source per paragraph was infrequent and thus not problematic. Additional coding rules can be reviewed by obtaining the author's doctoral dissertation, "Anonymous Attribution and Official News Sources in the Reporting of Bosnia and Somalia: A Study of Conflict Coverage in American Newspapers," University of Tennessee, Knoxville, 1996.
21. Michael X. Della Carpini, "Critical Symbiosis: Three Themes on President-Press Relations," Media Studies Journal 8 (spring 1994): 185-97; Daniel Riffe and Gail Johnson, "Unnamed Sources in White House Coverage" (paper presented at the annual meeting of AEJMC, Washington, DC, 1995).
22. Michael X. Della Carpini, "Critical Symbiosis: Three Themes on President-Press Relations," Media Studies Journal 8 (spring 1994): 185-97; Daniel Riffe and Gail Johnson, "Unnamed Sources in White House Coverage" (paper presented at the annual meeting of AEJMC, Washington, DC, 1995).
23. Joseph Spear, Presidents and the Press: The Nixon Legacy (Cambridge, MA: The MIT Press, 1984); Mark Hertsgaard, On Bended Knee: The Press and the Reagan Presidency (NY: Farrar Straus Giroux, 1988); H.R. Haldeman, The Haldeman Diaries (NY: Putnam, 1994).
24. Joseph Spear, Presidents and the Press: The Nixon Legacy (Cambridge, MA: The MIT Press, 1984); Mark Hertsgaard, On Bended Knee: The Press and the Reagan Presidency (NY: Farrar Straus Giroux, 1988); H.R. Haldeman, The Haldeman Diaries (NY: Putnam, 1994).
25. Joseph Spear, Presidents and the Press: The Nixon Legacy (Cambridge, MA: The MIT Press, 1984); Mark Hertsgaard, On Bended Knee: The Press and the Reagan Presidency (NY: Farrar Straus Giroux, 1988); H.R. Haldeman, The Haldeman Diaries (NY: Putnam, 1994).
26. Blankenburg, "The Utility of Anonymous Attribution"; Denham, "Anonymous Government Sources in the New York Times."
27. Blankenburg, "The Utility of Anonymous Attribution"; Denham, "Anonymous Government Sources in the New York Times."
28. Hallin, Manoff, and Weddle, "Sourcing Patterns of National Security Reporters."
29. In terms of operational definitions, Executive sources consisted of all members of the Executive branch and all Cabinet members. Definitions for the remaining source types follow. Legislative: All members of Senate, all members of House of Representatives; U.S. Military: All references to Pentagon and military, except president; Generic Official(s): No description other than "U.S. official(s)"; Private Sector: American sources independent of legitimated institutions; United Nations: All U.N. representatives in New York and abroad; Bosnian: All Bosnian sources, official and private sector; Serbian: All Serbian sources, official and private sector; Others in Region: All others (e.g., Croatian), official and private; Somalian: All Somalian sources, official and private; Other International: All sources independent of previous countries; No Source: No source in paragraph; Institution: Institutional source apart from institutions above (e.g., "The French Foreign Ministry").
30. For additional information about logistic regression, see Alan Agresti, Categorical Data Analysis (NY: John Wiley & Sons, 1990).
31. To demonstrate this goodness of fit mathematically, the -2 Log Likelihood Score with one predictor variable in the model was 40.729 with one degree of freedom and a p-value of .0001. This indicates a significantly better fit than a model including just the intercept. At step 2, with a second predictor included, the -2 Log Likelihood Score was 51.152 with two degrees of freedom. The difference between step 1 and step 2, then, was 10.423 with one degree of freedom, thus indicating the model with two predictors to be a significantly better fit. At step 3, the -2 Log Likelihood Score was 57.977 with three degrees of freedom, a difference of 6.825 with one degree of freedom from step 2. This indicates a significant improvement when the third predictor was entered.
32. At step one, the computer entered Associated Press, yielding a -2 Log Likelihood Score of 38.349 with one degree of freedom. At step two, the computer entered United Nations, and -2 Log Likelihood Score was 48.297 with two degrees of freedom. The difference between the two scores, 9.95 with one degree of freedom, was significant. At step three, the Los Angeles Times was entered and it increased the -2 Log Likelihood Score to 55.124 with three degrees of freedom. The difference in scores between steps two and three, 6.827 with one degree of freedom, was again significant, indicating the three-variable model as the best fit to the data. Its Residual Chi Square statistics (Chi Square=3.7708, d.f.=5, p=.5829) are consistent with the goodness of fit standards described earlier. The values in this regression closely resemble those from the earlier stepwise procedure. An additional regression tested for interactive effects, but revealed nothing significant.
33. Hallin, Manoff, and Weddle, "Sourcing Patterns of National Security Reporters."
34. Denham, "Anonymous Government Sources in the New York Times."
35. Riffe and Johnson, "Unnamed Sources in White House Coverage"; Carpini, "Critical Symbiosis."
36. Riffe and Johnson, "Unnamed Sources in White House Coverage"; Carpini, "Critical Symbiosis."
37. Hertsgaard, On Bended Knee.