In lieu of an abstract, here is a brief excerpt of the content:

Appendix IV Content Analysis We perform a content analysis of 26 media outlets from 7 national media systems, totaling 3,883 stories (Table 1.1). For television, we only code the main regular newscasts , but not special programs. We decide not to do a full-scale content analysis of all the data collected due to (a) resource constraints; (b) sampling efficiency; and (c) methodological limitations. From the sampling perspective, a complete sample may not yield additional insights that justify the tremendous additional effort. Moreover, we feel that the current status of the arts in content analysis is inadequate for revealing the thematic , rhetoric, and syntagmatic features of media discourses (Pan and Kosicki, 1993). We use the content analysis as a supplement to the discourse analysis, in setting the parameter and in providing some, albeit limited, basis for generalizing our inferences. This content analysis does not include Japan. None of us reads or speaks Japanese . We hire a Japan-educated media scholar to read the sampled newspaper issues and watch taped television newscasts; he translates each story title and writes down a thematic summary of the story. We hold frequent communication with him to get an overall perspective. Having defined the coding variables, we proceed to develop a coding scheme. The research assistant assembles an assortment of newspaper articles and television newscasts for us to do a trial round of coding. We study the intercoder reliability coefficient with a view to revising the coding scheme and refining the operational definition of each variable. After three trial runs we finalize the coding scheme (Appendix V). The level of agreement among the four coders on the “factual” variables (such as the size and time of a story) reaches above 90 percent. They agree around 80 percent on the more “subjective ” variables (such as the “tone”). Table IV-1 summarizes the amount of media coverage. In terms of the size of newspaper coverage (in squared centimeters), Hong Kong and the PRC rank highest, followed by Taiwan, Britain, and Australia. The United States ranks low because the Wall Street Journal and two “local” papers (the Chicago Tribune and the Des Moines Register ) devote a smaller space to the handover, but the three elite papers (the New York Times, the Washington Post, and the Los Angeles Times) all produce large amounts of stories. The amount of television coverage (in seconds) differs only slightly. CCTV (China) ranks first, followed by TVB (Hong Kong) and TTV (Taiwan), and Australian, British, and US networks. 215 The handover, as a prescheduled calendar event, has a clear life cycle. Media coverage peaks on June 30 and July 1. Table IV-2 shows the standardized size of daily media coverage by country in three phrases of the study period. The score is generated by dividing the size of an item (in newspaper squared centimeters or in TV seconds) by the range of story size (maximum minus minimum) of a particular media outlet. During the prehandover preparatory period (June 16–29), daily media coverage is only 11 percent to 33 percent of the amount produced in the peak period. The posthandover coverage (July 2–5) tails off to range from 7 percent to 39 percent of the amount produced in the peak period. We construct a summary measure of source diversity based on the content data. We coded the top ten sources quoted (Table 3.5). Where there were multiple sources in a story, we coded only the first ten sources. Based on the data, we classified the sources in terms of (a) their national affiliation; (b) social position (officials, political opinion leaders , business elite, media and journalists, academics and think thanks, other kinds of opinion leaders, and ordinary people.) For each of the two dimensions, we calculated an H-statistic, or the entropy measure in the information theory (Shannon and Weaver, 1964). The statistic is calculated in the following formula: H ⫽ ⫺ ⌺ pi ln pi where, pi ⫽ the percentage of cases falling into category i; i ⫽ 1 to I, with I equals to the total number of categories involved; ln pi ⫽ the natural logarithm of pi. The resulting H statistic has a minimum of 0 (least diverse), if all the cases fall into a single category. The maximum value (most diverse) is the natural logarithm of the 216 Appendix IV Table IV-1 Amount of Media Coverage1 Number Total of Number newspaper Total TV space per Time per outlets of Space Airtime paper channel coded Items (cm2 ) (seconds) (cm2 ) (seconds) PRC 3 717...

Share