3. Research method
3.1. Content analysis
國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
24
3. Research method
3.1. Content analysis
A content analysis of news stories about two presidential candidates from two important daily newspapers in Slovakia during the four weeks of official presidential campaign time was conducted. Only two candidates who passed to the second round were included in the analysis.
There are various reasons to this decision: the remaining five candidates had substantially less support among voters than the two strongest candidates, so the remaining five candidates were probably reported as less viable independently of the gender.
There was also much less coverage for them independently of the gender. Because in the second round there were only two candidates, it is more consistent to include only these two candidates even in the examination of the first round. This way the competition between one strong male candidate and one strong female candidate can be observed.
Sme and Pravda are the largest daily non-tabloid newspapers in Slovakia. Their circulation is close to 70,000 (2007) each, they are non-partisan and they are general newspapers, not specialized in particular subject. The hard copies of the newspapers were used, so the prominence of coverage could be observed, also the database LexisNexis doesn`t contain any of the newspapers.
Due to the short time span a census was conducted. The study included the articles from both newspapers starting from March 6th, to April 4th, 20095. There were 26 issues for each newspaper, a total of 52 issues, including both election days and moratorium time, because
5 There are no newspapers on Sunday in Slovakia.
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
25
newspapers are allowed to report also during the moratorium period, although they cannot publish the results of polls.
The unit of analysis for this research are newspaper articles, all the articles whose main headline contains one or both candidates` names; the word candidate and all its possible versions due to Slovak declination and gender of nouns; the word election, but not the verb vote; and the word campaign were included.
The articles whose headline includes the word vote were not included, because these articles are more concerned about the voters, not about the candidates. The articles in form of interviews were not included because these kinds of articles don`t represent how the media report on candidates, but how the candidates talk about themselves. The articles placed in the section Advertising were also excluded, because it is a place paid by candidates and therefore it is expressing what the candidates want to say, this doesn`t indicate how the media are representing the candidates.
There were 173 articles which fulfilled the criteria to be included in the analysis, of which 94 were from daily newspaper Sme and 79 were from the daily newspaper Pravda. Nine articles included in the analysis based on the criteria mentioned above were not related to the two candidates, but to the elections in general sense, about other candidates, or about other politicians related to the elections. Because they fulfilled the criteria they remained in the analysis.
In order to analyze the content of articles, the coding scheme used by Miki Caul Kittilson and Kahn Fridkin (2008) in their study called ―Gender, Candidate Portrayals and Election Campaigns: A Comparative Perspective” was adapted. This increases the validity of the measurement and creates a possibility of comparison of results with this study.
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
26
The codebook was created and written in both English and Slovak language; coding sheet used only English. The researcher served as the first coder, and the second coder, a Slovak native female, coded approximately 10% of all the articles. There were two training sessions before coding the articles, the researcher offered explanations of all the concepts from the codebook, followed by a demonstration of how the researcher would have done coding of a selected article.
The 10% of the articles to be coded by the second coder were selected using a simple random sample, in order to do that, a list of all the articles was created and a table of random numbers was used to select the articles.
33 out of 40 variables showed 100% agreement between the coders, and the Krippendorff's Alpha of all these variables was 1. The remaining seven variables showed intercoder reliability with Krippendorff's Alpha from 0.633 to 0.927, these variables were all measured in interval-ratio levels, which makes it difficult to have 100% agreement.
For the variable length of article the Krippendorff's Alpha was 0.6336. The variable number of sentences about female candidate showed the Krippendorff's Alpha level of 0.7287.
6 This variable was coded as a total number of sentences in the article, the difference between the two coders was in all the cases one sentence. For example coder one coder the article as having 30 sentences while the coder 2 coded the same article as having 31 sentences. When coding all the articles, this disagreement was solved by counting the total number of sentences twice to prevent any mistakes
7 The differences between the coders were only of one sentence, and only in articles longer than 28 sentences. While coding all the articles, more attention was paid to who was the subject of each sentence in order to decide, whether the sentences were talking about female candidate or not.
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
27
Similarly, the variable number of sentences about male candidate noted the Krippendorff's Alpha level of 0.8018.
The variables number of horse race sentences for female candidate, number of horse race sentences male candidate, tone of article for female candidate, and tone of article for male candidate had Krippendorff's Alpha level between 0.857 and 0.927, which is according Neuendorf (2002) acceptable in most situations. To see the detailed results of intercoder reliability test see the Appendix II.