Twelve Internet web sites were monitored for six days during the heat of the primary election season: February 23, 24, 27, 28 and March 5 and 6. Each day, we downloaded the sites four times, based roughly around the average workday. Download sessions began at 9:00 a.m., 12:30 p.m., 4:30 p.m. and 9:00 p.m. Each download took between 20 and 45 minutes [6]. Web sites were selected to develop a sample of the election news offered on the Internet today. We wanted to include the most popular portals and news sites as well as some magazines and news sites of the national newspapers. The portals and news sites were based on Media Metrix's February, 2000 ratings. We chose the top four portals that contained election news (Netscape, Yahoo!, Go and MSN ) and the top four web sites that contained election news (AOL News, MSNBC, Pathfinder, CNN). We then added two magazine sites (Salon and the National Review) and two sites from national newspapers (the Washington Post and the New York Times.)
We also selected two national newspapers (the Washington Post and the New York Times) to study as a control group.
Inclusion and Screening
Once the web sites were selected, the political front page for each web site was found. The front pages selected were defined as the most direct way to political news from the site's home page. The AOL political front page is the news site from aol.com, not their subscriber-only portal. The Yahoo! political page studied was the political page accessed from their own "News" link across the top banner, not the link from their "Sites"category located in the body of the page.
The political home pages are as follows:
The political front pages and the lead stories on those pages were captured through one of three Internet service providers (ISPs): Earthlink, Erols or AOL. Testing was done to be sure the pages did not differ among the servers. Each page and lead story was captured, saved to a disk and printed out landscape style. Downloads took between 20 and 45 minutes. Exact times of each download are noted, and the order was rotated.
On Feb. 23, due to technical difficulties, the 4:30 PM AOL News and the 9:00 PM Washington Post downloads were incomplete. The 4:30 AOL download was irretrievable and that download was inputted at missing data. Therefore, AOL News had one fewer download than the other sites. For the Washington Post, the front page file was corrupted and irretrievable. Therefore, the front page data was inputted as missing data, but the lead story was fully coded. The Post had one fewer front page than did other sites.
Any missed downloads on the Pathfinder pages were fully retrievable as the site did not change in a week's time. And Pathfinder staff was able to tell us exactly what had appeared and where it now was on the site.
The print newspaper sample consisted of election-related front page stories. They were captured via Lexis-Nexis and compared to the actual newspaper. There were few enough stories that we could monitor each front page individually to capture all stories that were at least 50% about the 2000 election.
The resulting project sample consisted of 72 political front pages and 286 lead stories as well as 25 newspaper stories, all of which were fully coded and are included in the final data analysis.
Coding Process
Researchers worked with a detailed, standardized coding scheme. All pages were first coded for basic inventory variables: source, date, download time, etc. Then, front pages for the 9 AM downloads only were coded for content variables: number of stories, amount of original reporting, number of interactive elements, links to external sites, number of feature stories, etc. Next, the lead story of each front page was coded. First it was coded for content: number of sources, general topic, amount of unfiltered material, changes in lead story, etc. Then the story was coded for intent variables: story trigger and frame.
In all cases, coders worked with a defined set of rules per variable. Of particular note is the 50% rule in effect for story frame. When calculating story frame, coders identified all text that implies which frame the story is told around. If 50% or more of the text is told around that frame, the story is coded as having that frame. If multiple frames are used, but one did not dominate, stories are coded as straight news or no dominant frame.
Intercoder Reliability
Intercoder reliability measures the extent to which individual coders, operating independently of one another, reach the same coding decisions. Tests were performed throughout the project. No systematic errors were found. In addition, the coding supervisor reviewed decisions on intent variables when necessary and made any changes needed to bring coders to agreement.
6 There were two exceptions to this timetable. The initial download, 9 a.m. on Feb. 23, began at 9 a.m. and was completed at 10:32 a.m. The 12:30 p.m. download on February 23, began at 12:30 p.m. and was completed by 1:30 p.m.