Measuring the Dynamics of Climate Change Communication in Mass Media and Social Networks with Computer-Assisted Content Analysis

To date, multiple authors have examined media representations of and public attitudes towards climate change, as well as how these representations and attitudes differ from scientific knowledge on the issue of climate change. Content analysis of newspaper publications, TV news, and, recently, Internet blogs has allowed for identification of major discussion themes within the climate change domain (e.g., newspaper trends, comparison of climate change discourse in different countries, contrasting liberal vs. conservative press). The majority of these studies, however, have processed texts manually, limiting textual population size, restricting the analysis to a relatively small number of themes, and using time-expensive coding procedures. The use of computer-assisted text analysis (CATA) software is important because the difficulties with manual processing become more severe with an increased volume of data. We developed a CATA approach that allows a large body of text materials to be surveyed in a quantifiable, objective, transparent, and time-efficient manner. While staying within the quantitative tradition of content analysis, the approach allows for an interpretation of the public discourse closer to one of more qualitatively oriented methods. The methodology used in this study contains several steps: (1) sample selection; (2) data preparation for computer processing and obtaining a matrix of keyword frequencies; (3) identification of themes in the texts using Exploratory Factor Analysis (EFA); (4) combining identified themes into higher order themes using Confirmatory Factor Analysis (CFA); (5) interpretation of obtained public discourse themes using factor scores; and (6) tracking the development of the main themes of the climate change discourse through time. In the report, we concentrate on two examples of CATA applied to study public perception of climate change. First example is an analysis of temporal change in public discourse on climate change. Applying CATA to a conservatively selected sample of 4043 articles published on climate change in The New York Times from 1995, we found a considerable change in major topics of discussion. One of the most significant tendencies is a gradual decline in the volume of material within the "Science" topic and an expansion of themes classified under the "Politics" topic. The second example is the analysis of public ability to detect climate change, in which we used a database of over 1 million Twitter messages on climate change that we have collected. We compared the intensity of tweeting on climate change with the "common-sense climate index" by Hansen et al (1999) and found that the weather extremes experienced at a certain location is immediately reflected in the number of tweets discussing climate change originating from that location. Although the CATA approach certainly has its limitations, we are convinced that it has a number of advantages over manual processing: it is able to analyze large textual bodies, is more time efficient, has a higher level of detail, enhances the richness of interpretation, and is able to reliably track discourse development through time.