Sampling Method and Unit of Analysis
A content analysis was conducted of four national newspapers during the first three weeks of the invasion phase of the war in Iraq and three weeks during what has turned into an occupation phase of the war. The analysis focused on a sample of 452 articles provided by the Los Angeles Times, Chicago Tribune, Washington Post, and New York Times about military operations. The Los Angeles Times was selected because it is a prominent West Coast newspaper, the Washington Post was selected because it is a prominent East Coast paper, The Chicago Tribune because of it influence in the Midwest, and New York Times because of its national influence.
The articles were chosen using the terms “war in Iraq” as search terms in the lead paragraph or headline. Full text articles were retrieved from Lexus Nexus or Newsbank Full Text Newspapers. The dates chosen for the invasion phase were March 20 to April 9, 2003, and November 1 to 21, 2004, for the occupation phase. The 2003 dates selected marked the beginning of the invasion into Iraq concluding with April 9, when journalists coined “The Fall of Baghdad,” which was symbolized by the fall of the statue of Saddam Hussein and considered as the end of the invasion phase of the war. Since the invasion phase ran the course of 21 days, the same number of days was selected for the analysis of the occupation phase to provide an equivalent comparison between the invasion and occupation phases. The dates of the occupation phase were chosen based upon embedded numbers attained from the Department of Defense (DoD). The numbers had increased to 96, which was the highest number of embeds since the invasion. After performing a preliminary search, it was decided to stratify the sample over the days. The sample was taken by selecting every third day from the time frame. This provided a total of six days with a sample size of more than 450.
The unit of analysis was each single news story about a person, unit, or event with a clear beginning, middle, and end, excluding about 100 editorials, opinion pages, or commentaries as well as non-newsworthy items.
Code Training
Six Department of Defense public affairs personnel, who were all enrolled in the Joint Communication Course at the University of Oklahoma, conducted the content analysis. A written coding instrument was developed to code the sample. Coding norms were established during a supervised six-hour training session. The effective intercoder reliability ratings (Rosenthal, 1984, 1987) are reported. Six DoD students coded 8% of the sample and established a high degree of standardization to the coding process resulting in effective intercoder reliability of .93 for the invasion phase. Effective intercoder reliability for the three coders for the occupation phase ranged from. 87 to .91. Following the training sessions, coders, who are veteran media analysts working in the public relations and broadcast media field, worked separately.
Variables Measured
The independent variable featured in the investigation was news correspondent status, which was operationalized based on information provided about units as embedded, non-embedded, or unknown. An embedded reporter was defined in the literature as a media representative remaining with a unit on an extended basis (SECDEF, 2003). Coders distinguished between embedded and non-embedded correspondents based on dateline information that was provided in the newspaper articles. In some instances, it was unclear whether the correspondent was embedded or non-embedded, in which case they were coded as unknown.
The investigation featured nine dependent variables. Overall tone of coverage toward the military was assessed with a global attitude measure adapted from Burgoon, Cohen, Miller, and Montgomery (1978). It consisted of six 7-interval scales, including: good/bad, positive/negative, wise/foolish, valuable/worthless, favorable/unfavorable, and acceptable/unacceptable. The alpha reliability rating of this measure was a=.98. Depiction of the trustworthiness of the military personnel covered in news reports was assessed using the Individualized Trust Scale (ITS), developed by Van Lear and Trujillo (1986) based on four, 7-interval items including trusting/untrusting, candid/deceptive, and sincere/insincere, and honest/dishonest. The reliability rating of the trustworthiness scale was a=.99.
The extent to which each unit employed framing was measured with a single 7-interval scale: episodic/thematic. The scale was used previously by Pfau (2004).
Affect conveyed in each story unit was measured using eight 6-items interval scale by Dillard, Plotnick, Godbold, Freimuth, and Edgaret (1996). The 0-6-interval scales measured five emotions with multiple item indicators. The emotion scale was based on the previous work of Dillard and colleagues (Dillard, et. al., 1996). Emotions featured included: anger (angry, irritated, and annoyed), a=.94; surprise (surprise, astonished, and amazed), a=.89; puzzled (puzzled, bewildered, and confused), a=.58; sad (sad, dreary, and dismal), a=.70; and fear (fearful, afraid, and scared), a=.98.
Authoritativeness is the “extent to which the article was perceived as possessing expertise relevant to the communication topic and can be trusted to give an objective opinion on the subject” (Goldsmith, Lafferty & Newell, 2000, p. 43). Authoritativeness of the story was measured using the authoritativeness dimension of the Source Credibility Scale (SCS) introduced by McCroskey in 1966, and examined by McCroskey et al., (1974). This scale was adapted to media as opposed to source. The reliability was a=.98. The 1-7 interval scales adjective pairings for the authoritativeness dimension included unreliable/reliable, uninformed/informed, unqualified/qualified, unintelligent/intelligent, worthless/valuable, inexpert/expert, dishonest/honest, unfriendly/friendly, unpleasant/pleasant, awful/nice, sinful/virtuous, and selfish/unselfish.