Elites’ Polarization in Climate Change Dialogue: The Data, the Insights, and the Tool to Disintegrate the Building Components of Polarization

The debate on climate change and the sociopolitical polarization within this debate, has been the focus of a range of studies. While much of this research has focused on the public, there is a growing body of work examining the role and behavior of polarization created by policy elites. These studies highlight that policy elites play a considerable role in shaping public opinion and policy agendas; nonetheless, the characteristics of polarization within elite segments, for example, the various organizations influential in policy making, is missing from such studies. In this work, we used survey data on organizations influential in climate policy to perform a data collection on Twitter (X) platform. In particular, we collected the content generated by the closely affiliated members of four organizational classes: government, scientific, business, and civil societies in 9 countries across the globe over the course of 6 years. We collected all the tweets and retweets relevant to the account of those actors. Using lemmatization and keyword filtering for the language(s) relevant to each country, we filtered the data to separate the content on the climate change discussions from the other data. We employed this data to analyze the characteristics of the activity of elites on Twitter and its polarization patterns, comparing the organization types, and organizational levels (by grouping the users according to their ranks in their organization), within each country as well as across different countries, also inspecting the effect of global patterns (i.e., relevant international events). We firstly show the activity pattern in climate change debate is completely different from the gross activity patterns. Moreover, we illustrate that while countries generally differ in their climate change activity pattern, at time points corresponding to the global events, specific countries show similar increase in their activity. Interestingly, the activity in all four classes of organizations is strongly correlated in all the countries. To study the polarization behavior of the climate change debate, we use community detection and find that, expectedly, the debate consists of two sides as two communities of unequal sizes. We show that the countries differ according to whether the majority of business accounts side with the majority of other organization types. We also investigate the polarization patterns employing different polarization metrics. Moreover, we introduce a new polarization metric that has the advantages of the well-known polarization measures, but can also be broken into the sum of the contributions of the subnetworks (each consisted of actors pertaining to an organization type) within the full network. We show that each contribution equals the polarizations of the subnetworks times the structural influence of that subnetwork. Using this metric, we show when an organization type is adding high polarization to the system because it is highly polarized and when it is because of its high influence while its polarization is similar or even less than the other types. These findings contribute to better understanding of the drivers of polarization in the climate change dialogue, as well as in other domains, and their role and behavior.