18
Once the sample was drawn, a team of 20 advanced undergraduate researchers
who had been trained to code the articles and news transcripts’ frames were each
assigned a set of approximately 100 transcripts and articles.
18
News stories were coded
from FPR Hegemonic, FPR Dominant, Balanced, IPC Dominant, to IPC Hegemonic. A
final category, Not Applicable, was included to capture any stories that mentioned El
Salvador, Nicaragua, or Central America but offered no framing of the conflict.
19
As a measure of public opinion, I then collected 55 national public opinion polls
conducted between 1981 and 1988, which asked respondents their views on U.S.-Central
American relations. These surveys were drawn primarily from two sources, the Gallup
organization (21) and ABC-Washington Post (22).
20
These were supplemented by seven
New York Times-CBS and five NBC-Wall Street Journal surveys. My dependent variable
is the percentage of public opinion support for U.S. policy towards Central America as
measured by these polls. The mean of public support is 27.96 and the standard deviation
is 6.215. To control for possible wording effects I included dummy variables for each of
the different wordings used in the various surveys.
My independent variables were then constructed from the content and frame
analysis of New York Times articles and ABC Nightly News broadcasts found in the
month immediately prior to the dates on which the surveys were conducted. Specifically,
18. The coding sheet is included in Appendix A. The important variable for this study is the final question,
what frame did the article / transcript present in describing U.S. policy towards Central America: FPR or
IPC.
19 . Tests of inter-coder reliability were conducted throughout the process. The final tests conducted
achieved approximately 82% agreement among coders and was statistically significant at p<.001 using
correlation analysis (Pearson’s R). To control for chance agreement I tested the results using
Krippendorff’s alpha, which yielded an inter-coder reliability of .65.
20. The ABC-WP surveys consist of several different question wordings. The Gallup surveys ask the same
question changing only the name of the country or region: El Salvador, Nicaragua, or Central America.
Convention