This text is based on a lectio praecursoria. Lectio praecursoria is an introductory lecture given by a doctoral candidate at a public examination. In the lecture, the candidate will introduce the background and the key concepts of his or her doctoral dissertation. The author defended her doctoral dissertation at the University of Helsinki in February 2020. Her dissertation can be found online here.
In my doctoral dissertation, Decarbonising energy regimes: Methodological explorations and empirical insights for policy, I advance research on decarbonisation policy and politics by answering methodological questions that help improve synergies between policy studies and energy transition studies. When I set out on my PhD endeavour in the Academy of Finland funded project Defend in 2015, it was widely argued that while decarbonisation presents a systemic challenge to our societies, it can be increasingly framed as a political challenge. In this context, I became zealous about how adopting a policy studies perspective can enrich our understandings of decarbonisation processes. In the energy transition scholarship, many have successfully initiated work on conceptual bridging to take advantage of policy-based approaches developed within the policy studies discipline. Importantly, I highlight in this dissertation that the conceptual advancement is inherently intertwined with and dependent on sound methodological practices. My dissertation therefore critically explores the best practices and added value of textual methods, both discursive approaches and the topic modelling machine learning method, while having decarbonisation as the empirical context. The empirical part of this dissertation yields novel insights into the development of the European Energy Union’s decarbonisation agenda and the discourse surrounding the decline of coal-fired power generation in the UK. This lectio praecursoria presents the main findings of my dissertation and offers advice for policy and practice.
I first intended to begin this speech by stating how urgent it is to tackle climate change, underlining the difficulty of the decarbonisation task ahead of us.
I intended to explain how literature has provided ample evidence of how fossil fuels have become deeply entrenched in the works of contemporary societies. How our lives are largely operated by carbon-intensive systems, incompatible with sustainability goals.
I even wanted to highlight that the challenge of delivering decarbonised energy systems is often attributed to complexity. Decarbonisation necessitates system change, yet, changing entire systems requires simultaneous changes across multiple dimensions, be they socio-technical, economic, institutional or cultural, cognitive and behavioural.
Instead, I wish to start by arguing that decarbonising the energy sector is, at its core, a relatively simple problem. It is simple.
It is simple in the sense that we have scientific consensus over what has to be done. Science tells us, and has been doing so for decades, that we need to act urgently to curb emissions by getting rid of fossil fuels. Moreover, decarbonisation can be viewed as a simple problem in the sense that we today, by and large, possess the technologies and knowhow necessary for fossil-free energy production.
Now, when decarbonisation becomes complex, I argue, is when politics kicks in.
We have seen governments around the world avoiding taking sufficiently ambitious steps to accelerate decarbonisation, and global climate talks being hamstrung by resistance even when scientific knowledge has been clearly communicated to decision-makers, most recently by brilliant young activists. And in places where there has been political will to initiate processes of decarbonisation, as is the case in many European countries including the UK or Denmark and also Finland, the pace of decarbonisation has been insufficiently fast due to political lock-ins.
This ultimately boils down to the characteristics of energy transitions: They are value-laden and highly contested processes, involving a wide range of actors with different views on sustainability problems, and preferences over most favourable policy targets and instruments. Also, given the urgency for rapid change, the literature widely argues that successful transitions will have to be deliberatively steered and accelerated by public policy.
Motivation and research questions
My statement about the decarbonisation challenge, while purposefully simplistic, illustrates the main motivation behind my research. Specialising in the fields of sustainability transitions and policy studies, I took an interest in a relatively recent line of research that has incorporated policy concepts into the existing, traditionally largely socio-technical and economic, frameworks of energy transitions.
In my dissertation, I identified a gap in the existing attempts to take advantage of policy-based approaches when conducting research on transitions. I argue that while conceptual bridging between policy and transition scholars has already been successfully initiated, there is a further need to advance the exchange of methodological insights to make the most of the conceptual transfer.
Identifying this gap led me down a path that was both methodological and empirical at heart. I took an interest in a compelling combination of textual approaches, discursive approaches and unsupervised machine learning methods, in particular that of topic modelling, while having decarbonisation as the empirical context.
Even if the interest in discursive approaches is nothing new in the policy studies circles, it is a relatively novel addition to the energy transitions research field. With discursive methods, energy scholars can refine their investigative tools to study complex and politically attuned phenomena like decarbonisation. Unsupervised machine learning methods, in turn, come with the benefit of scale and scope. We can cover large collections of policy relevant data, time- and resource effectively. It is also possible to both create novel data sets and run a model without imposing human bias to the analysis.
I wanted to explore this arguably tremendous and underutilised potential of textual methods — both discursive and computational — for the purposes of transitions research. I therefore set out to answer the following research question: What novel contributions can textual methodologies bring to the study of decarbonisation policy and politics, both in terms of methods and empirical insights?
To answer my research question, I structured my doctoral dissertation as a two-step process: in the first step, I studied the two groups of textual methods and, in the second step, I applied the methods to empirical cases of decarbonisation. As one case, I analysed public discourse surrounding the decline of coal-fired electricity in the UK, one of the first countries to have almost completely abandoned coal use, in the period of 2000-2017. As another case, I conducted a big data analysis of the European Energy Union, a European energy policy reform project, in its formative years from 2015 to 2018.
The added value of discursive approaches
The results stemming from my jointly-written article  reviewing the existing use of discursive approaches indicate that taking a discursive lens has brought about novel analytical standpoints for energy transitions research.
These include, for example, shedding more light on political ideology and state orientation – in other words, how policy salience and legitimacy are developed in different political contexts. Discursive approaches also allow gaining insights into the ‘publics’; helping us understand how actors adopt, react to or influence energy policy alternatives. Another aspect that the results suggest is that discursive methods enrich the analysis of institutional and policy change.
Echoing these findings, our article  on coal phase-out in the UK, where my co-author and I conducted an analysis on newspaper data, exemplified how a discursive approach allows us to identify the different ways various actors delegitimize or legitimize coal use in their discourse. We gained an understanding of the different storylines actors used in depicting the possibility of phasing out coal from power generation.
The analysis suggests that coal phase-out occurred much faster and with significantly less resistance than what transitions research would generally give cause to expect. There were some areas of contestation and relevant discursive shifts, like that surrounding the Carbon Capture and Storage technology. Incumbents and policy makers became very active in the discourse and tried to re-establish the legitimacy of coal with this idea of a ‘technology fix’. Interestingly, while the discourse by coal incumbents was for a long time similar to that of the government, their discourse about coal actually shifted before the government’s official phase-out pledge of 2015. Incumbent actors had already been preparing for the decline of coal by developing renewable energy technologies, which resulted in little resistance.
On the whole, the investigation of discursive methods revealed how energy transition scholars can gain added value by applying these methods to analyse policy and politics. Critically, however, the results also clearly show that thus far scholars have mainly relied on a few, well-established methods and that, in order to exploit the full potential of discursive approaches, there would be room to engage in further methodological exploration and development.
This is exactly what sparked my interest in topic modelling and caused me and my co-authors to consider, how, if at all, the unsupervised algorithm could be used in a discourse analytical endeavour.
Topic modelling for qualitative textual analysis
Topic modelling refers to an automated algorithm that organises a corpus into word clusters based on word co-occurrences and probability distributions. The appeal is evident: TM does not require any predefined classification of training data by humans. With the method, we can automatically and quickly analyse large data sets. Interpretation is shifted to a later stage, after the method has produced its output. And this output, clusters of words, represents the latent thematic structure of the documents analysed.
It is no surprise, then, that in recent years, policy scholars have tried to use topic modelling to examine various policy concepts, including aspects of discursive environments, semantic or thematic categories, issue definitions, narratives or frames. However, the results from my dissertation show that there has actually been a mismatch between what the method produces as outcome and the ways this outcome has been interpreted. I argue that seeking to gain straightforward qualitative or policy relevant value from the topic modelling output risks being misleading.
To illustrate this, I want to consider the concept of discourse. Distilling discourse analytical information from text is an extremely complex cognitive task including an understanding of literal, metaphorical, contextual and inter-textual meanings. Conversely, due to its functions, topic modelling is unable to take into account semantics, syntax or place of words in the documents. As topic modelling was developed to only take into account the words contained in the document corpus, so it is also unable to ask the critical question often posed in discourse analysis: is something purposefully left unsaid in the text? As a result, it becomes clear that the topic modelling output cannot and should not be equated with the concept of discourse.
While adopting this critical stance on topic modelling, I do not mean to deny the vast potential of computational methods for policy studies research or even for the use of discourse analysis. On the contrary, I argue with my co-authors that topic modelling brings added value to textual analysis when used in mixed-method designs . To exemplify this, we propose embedded and sequential designs for incorporating topic modelling into the analytical processes of different qualitative textual methods.
On the whole, the analysis of topic modelling shows that when carefully designed to match the real-world phenomena they are applied to, topic modelling can bring about novel angles to a given research topic, contributing to challenge and potentially transcend, current knowledge and practices.
We made use of this possibility to obtain such new takes in our application of topic modelling to European Commission’s policy data on the Energy Union . The case of the Energy Union is, in itself, very intriguing. Prior studies have found, by examining smaller datasets, that the project is used for advancing conflicting policy aims, and therefore risks being unable to ambitiously drive decarbonisation objectives. Applying an unsupervised approach to this phenomenon thus gave an interesting vantage point: It became possible to ask, what a big data angle on policy texts reveals in this case? What are the main policy priorities? And, are there signals of energy and climate policy convergence?
By examining the Commission’s high-level policy priorities before and after the launch of the Energy Union project, my co-authors and I found that the Energy Union’s agenda has in fact been increasingly geared towards decarbonisation objectives. In addition, the evidence suggests that the reform project has also exerted a streamlining effect on the climate vs. security and energy efficiency vs. affordability debates. This is a change from the pre-Energy Union era, which was marked by the difficulty to combine climate and energy policy agendas. Finally, the results also imply that there is little thematic focus on phasing-out incumbent fossil fuels from European energy regimes.
Implications for research and policy
In policy studies, it has, on many occasions, been demonstrated that a comprehensive discourse analysis of environmental policy dilemmas not only requires expertise across different disciplines but also an understanding of diverse epistemological takes on the research topic. While my findings resonate with these notions, they also extend them in showing that there is also a need for diverse methodological expertise in these processes. I argue that as the often very general sounding terms, such as discourse, frame or narrative, each have an entire conceptual literature behind them, solid methodological understanding of these underpinnings is a prerequisite for any attempt to thoroughly analyse policy and politics.
Furthermore, and critically, the results highlight that integrating computational approaches into social scientific research endeavours requires strides to be taken in stepping outside of the ‘algorithmic black boxes’. As the example of the topic modelling method shows, applying powerful computational methods without thorough methodological considerations risks ending up producing unintended false practices, even if the intensions for these analyses were methodologically ambitious.
I call for more in-depth methodological dialogue, as this would make it possible to avoid and overcome pitfalls that combining the traditionally very different approaches of computational and societal research often entails. This kind of dialogue would also be critical and beneficial in the case of topic modelling, as the algorithm is being constantly improved, with aims to make it better account for contextual sensitivity.
Finally, I wish to turn to some conceptual implications and contributions that the empirical cases make for the energy transitions literature.
The lessons learned from the UK case indicate that, unlike what is sometimes suggested, incumbent technology decline does not necessarily have to be ‘a call against all incumbency’. Instead, incumbent power suppliers can be incorporating old and new technologies into their operations, and thus be promoting a reconfiguration pathway for technology decline as opposed to immediately overthrowing coal. While these reconfiguration pathways may be a practical way forward, and can work as an encouraging example for the numerous other European countries that have recently pledged to phase-out coal; I argue that policymakers should pay careful attention to how best to govern the processes of technology decline.
One challenge is the risk of phase-out becoming an oxymoron if coal is substituted by other fossils such as natural gas, which has, to some extent, been the case in the UK. This highlights the importance of how alternatives to coal are presented and defined in policy discourse.
Moreover, recent instances surfacing from the EU seem to consolidate the findings of my dissertation that give encouraging signals of decarbonisation policy integration at the EU-level. Since the time period covered in my data set, the European Commission has been further promoting decarbonisation. Today, through the European Green Deal, it is proposing to address some of the critical areas and limitations of the Energy Union framework that my research pointed to. For example the lack of policy interest in deliberatively phasing out coal and guaranteeing the participation of fossil-fuel reliant member states to advance ambitious climate policy.
Recent instances in the EU have also beautifully demonstrated the power of discourse in policy, and the importance of the ways in which we give meaning to decarbonisation. In May last year, the UK parliament used its voice to declare Climate Emergency, and, a few months later, the European Parliament followed this example. And only recently, the chief executive of BP pledged to cut the fossil fuel company’s emissions to net-zero by 2050.
I argue that these kind of discursive shifts are vital in our attempts to tackle climate change. I therefore urge policymakers to embrace this alarmist discourse to make swift policy decisions, while simultaneously promoting hopeful, fact-based visions of decarbonised and sustainable futures. With discursive action, we could begin reframing our climate crisis predicament, turning devastation into an opportunity filled with hope. Discursive action can prompt us to finally face the reality and start implementing the solutions needed. The solutions that are, at their core, very simple.
 Isoaho, K., & Karhunmaa, K. (2019) A critical review of discursive approaches in energy transitions. Energy Policy, 128, 930–942. https://doi.org/10.1016/j.enpol.2019.01.043
 Isoaho, K. & Markard, J. (2020) The Politics of Technology Decline: Discursive Struggles over Coal Phase-Out in the UK. Review of Policy Research. https://doi.org/10.1111/ropr.12370
 Isoaho, K., Gritsenko, D. & Mäkelä, E. (2019) Topic modelling and qualitative text analysis for policy studies. Policy Studies Journal. https://doi.org/10.1111/psj.12343
 Isoaho, K., Moilanen, F., and Toikka, A. (2019) A Big Data View of the European Energy Union: Shifting from ‘A Floating Signifier’ to an Active Driver of Decarbonisation? Politics and Governance, 7 (1): 28. http://dx.doi.org/10.17645/pag.v7i1.1731
Karoliina Isoaho, PhD, is an environmental policy scholar specialised in energy and climate policy. Isoaho defended her doctoral dissertation at the University of Helsinki in February 2020. Currently, she works as an environmental inspector in transportation and sustainable mobility at the City of Helsinki. Isoaho holds an MSc in Environment and Development from the London School of Economics and Political Science and a BA in Modern Languages from the University of Southampton.