Paper short abstract:

We compare energy policy document framings in the EU and the US through a topic modelling approach. The machine learning method enables the analysis of large text corpora. In this way, we provide novel insights into how different technology frames may matter in energy transition governance.

Paper long abstract:

Transforming the current energy governance systems towards sustainability faces numerous interconnected challenges. Reframing energy-policy towards one that facilitates energy transition is one of them. This paper starts from the non-linear approach to policy-making process, arguing that policy language is not a neutral medium. Instead, the use of language, and the values, interests and institutional aspects embedded in the development of policy discourse are believed to play a strategic role when articulating policy and promoting new technologies. This article seeks to provide insights into transition governance policy framing by presenting a case study from the European Union and the United States. We use a natural language processing method, dynamic topic modelling, to analyse a big data corpus of circa 5700 energy policy documents. The strength of this method is that it allows the analysis of large amounts of text, covering significantly more ground than humans could qualitatively read. Topic models are machine learning methods that map word co-occurrence in documents to find word probability distributions that are interpretable as coherent topics or frames. By analysing and comparing the evolution of the topic structure in these cases, we evaluate what vocabulary is used in framings and whether elements of strategic technology frames are present in the documents. Moreover, we seek to unravel whether existing frames are supporting the status quo or encouraging the emergence of novel frames on technologies. In doing so, our analysis seeks to open a novel vantage point into how policy frames matter in policy change and technology development.