This research employs three primary research methods: a quantitative content analysis of eight national newspapers over the period January 2007 to March 2016; a frame analysis of a random sample of the corpus of articles produced by the content analysis, and a series of interviews with journalists, politicians and political advisors involved with the Irish media coverage of climate change during the period under review. In this section, I wish to set forth some general argument as to why content analysis is an appropriate method for this research, and also to describe how the content analysis was carried out.
Content analysis is “a research technique for making valid and replicable inferences from data to their context” (Krippendorf, 1989, p. 403). In order to allow for replication, however, the technique can only be applied to data that are durable in nature. Content analysis enables researchers to sift through large volumes of data with relative ease in a systematic fashion (GAO 1996). It can be a useful technique for allowing us to discover and describe the focus of individual, group, institutional, or social attention (Weber, 1990). It also allows inferences to be made which can then be corroborated using other methods of data collection. Krippendorf (1980, p. 51) notes that “much content analysis research is motivated by the search for techniques to infer from symbolic data what would be either too costly, no longer possible, or too obtrusive by the use of other techniques”. Content analysis is a something of an umbrella term for a variety of approaches to extracting and analysing data (Cavanagh, 1997) covering a family of analytic approaches ranging from impressionistic, intuitive, interpretive analyses to systematic, strict textual analyses (Rosengren, 1981). However, this flexibility can lead to difficulties. As Hsieh and Shannon (2005, p. 1277), citing Tesch (1990), point out, “although this flexibility has made content analysis useful for a variety of researchers, the lack of a firm definition and procedures has potentially limited the application of content analysis.” Furthermore, some scholars have critiqued the results produced by content analyses. For example, Nosty (2014, p. 8) gives a trenchant account of the inadequacy of content analysis media research:
“Content analyses, for example, verify the hypotheses that have already been discussed in many previous studies. Their journalistic discourse is poor, they are insufficiently specialised and the narrative tendency does not allow for the effects of an unsustainable industrial model to be perceived socially. They discover deficiencies in media content based on an ideal theoretical model based on the subjective perspective of the researcher, who takes an arbitrating position which is not always realistic. These studies, as a whole, are useful in asserting that the media do not fulfil the role that they are expected to play.
They even state that the media are unable to explain the weaknesses of a system in crisis, because they are part of its very core.”
However, this analysis seems more critical of the purposes to which media scholars put content analysis, rather than of the method itself.
In its strictly operational phase, content analysis provides for the coding of data according to criteria relevant to the researcher’s intent and the content of the research question. The process of sampling data, reducing the sample size and coding the data amounts to what Krippendorf (2004, p. 83) calls “data making”, and as such it is a precursor to the description and interpretation of the data. He recommends that a content analysis have the following elements and structure: unitising, sampling, recording/coding, reducing sample, drawing inferences, and narrating the answer to the research question (Krippendorff, 2004, p. 83).
In a content analysis of framing studies in leading communications journals (Matthes, 2009), the author searched a range of journals for “articles that identified, named and extracted media frames” (Ibid. p. 353). These were then coded for descriptive variables (media analysed, timeframe covered), conceptual variables (frame type, unit of analysis, frame definition etc.) and theory variables (whether hypotheses were tested, or whether a research question was considered). For this research, a search of the LexisNexis database was undertaken, using the search terms “climate change” OR “global warming” OR “greenhouse effect”. As LexisNexis does not permit searches that return over 3,000 results, 111 individual monthly searches were carried out, producing a corpus of 12,751 articles mentioning the search terms.
The newspaper titles searched are presented in Table 4.1 below. The eight titles represent a cross-section of daily (five) and Sunday (three) national newspapers, with five broadsheet and two tabloid titles represented, and one title (the Irish Independent) that published in broadsheet format for part of the timeframe under study, then appeared in both broadsheet and “compact” format (from February 2004) before finally appearing in the compact format only (from December 2012).
The discrepancies in the dates for which each title was available presented a challenge in recording meaningful data. Therefore, a method was employed whereby the total number of relevant stories for each title in each month was recorded, and this total was then divided by the total number of titles available for that month. This produced a total for the average number of climate stories per title per month, a figure which could be usefully compared to other datasets, such as the one maintained by the ICECAPs collective in the University of Boulder, Colorado (McAllister et al., 2017). To give an example: for the early months of 2007, only four of the eight titles were recorded in the LexisNexis database, and thus the total of climate change stories was divided by four; later in 2007 (from August), six titles were available, and the total was divided by six, and so on. It should be noted that the recorded content for the Evening Herald in LexisNexis covers less than a year, during which the title published just a single climate change story and therefore this title has been omitted from the analysis.
Table 4.1: Irish national newspaper records available in the LexisNexis database. Title Dates available Irish Times June 1992-present Irish Independent July 2006 – present Irish Examiner Aug 2007 – present Irish Daily Mail Feb 2012-present Evening Herald Feb 2008 – Jan 2009 Sunday Independent Oct 2006 – present Sunday Business Post Aug 2007-present Sunday Tribune Sept 2001-Feb 2011
This process comprised the unitising phase of Krippendorf’s (2004) content analysis process. Subsequently, the sampling stage was undertaken. Krippendorf acknowledges that sampling is necessary in order to deal with large amounts of data, and suggests that random sampling is the best way to deal with “texts that stem from regularly appearing publications”. In this type of systematic sampling, every kth unit is extracted, with k is a constant. Thus the “rhythm” of the publication is reflected in the sampled data (Ibid. p. 115). LexisNexis allows for sample reduction by permitting the selection of articles containing >500 words. This strategy reduced the sample of climate change articles from the eight newspaper titles to 6,959 articles. Employing a further systematic sampling strategy of selecting every 10th article further reduced the sample to 706. These three datasets, all climate change stories, climate change stories over 500 words, and every 10th climate change story over 500 words show the same trends (Fig 4.1).
Fig 4.1: Comparison of data samples
This sample of 706 newspaper articles about climate change were then coded for a number of descriptive variables. These were:
(i) Publication (ii) Headline
(iii) Publication date (iv) Word count (v) Story type (vi) Author
(vii) Section/Page number (viii) Science
(ix) Climate focus
(x) Dominant frame
Several of these variables are self-explanatory. Others, such as “story type”, “science”, “climate focus” and “dominant frame” require further explanation. Each of the 706 stories was coded as to story type, a classification which recorded whether the story was (a) a straightforward news story, written in a news style, with relatively few sources, (b)
0 500 1000 1500 2000 2500 3000 3500 4000 2007 2008 2009 2010 2011 2012 2013 2014 2015
Area chart comparing data samples
a complex news story, with several sources and providing more detailed context, (c) a feature article, (d) an editorial, expressing the stance of the newspaper on a given topic, (e) an opinion article, (f) a regular column. The science classification recorded whether the science of climate change was accepted or portrayed as a given, in which case the story was coded as “settled science”; in cases where the science of climate change was not accepted, these stores were coded as “contested science”. The climate focus classification recorded whether climate change was the primary focus of the article, or whether it was a secondary focus. Stories were also coded according to their dominant frame. The approach to coding frames is dealt with in the following section. Articles were coded in their entirety for the presence of several frames, and articles typically contained more than one frame. In this descriptive phase, the frame present in the first paragraph (commonly called the lead or intro [or lede in US journalism]) or opening section of the article.
Schäfer et al, in their “best practice” guide for media coverage of climate change researchers, state that “To analyse issue attention as a process of selecting one issue over others given limited editorial space, however, it is necessary to normalise these measures, e.g. by relating the number of CC-related articles to all articles published in a given medium over a given time span (see Schmidt et al. 2013; Schäfer et al. 2014). Such measures control for the size of the news hole and thus yield data comparable across outlets, time periods, or countries.” (2016, p. 7). Therefore, the media coverage of climate change is compared to all news coverage in the period under study. This was achieved by setting the parameters of the search within the LexisNexis database (date range and newspaper titles) but leaving the search term field blank. This method returned all news items for the titles with the specified dates. As LexisNexis does not allow searches to
return more than 3,000 articles, this exercise entailed 670 individual searches (five searches per month, over 134 months). The findings of this analysis are set out in Chapter 5, section 5.3.
As this thesis is also concerned with the influence of the Green Party on levels of media attention for climate change, a second content analysis was undertaken to discover the extent to which media coverage of climate change also mentioned the Greens. Again, the LexisNexis database was searched for the time period of January 2006 to December 2011 for all articles mentioning “climate change”, “global warming” and “greenhouse effect”; subsequently, the same search was undertaken again, this time with the added search terms of “Green Party” or “Greens”. The results of this content analysis are discussed in Chapter 6, section 6.7.1.