Muddling through meanings: doing discourse analysis in health policy research (part 2)

Written by Ida Okeyo, Sudha Ramani and Eleanor Whyle on 21 August 2018.

Discourse analysis offers a way of analysing ideas, ideologies, power relationships and contexts; of considering the role of language, rhetoric, narratives and framing to get a better understanding of the policy processes. But how does one conceptulise a discourse analysis study and choose the right methodological approach?

In this final part of their two-part blog, Ida Okeyo, Sudha Ramani and Eleanor Whyle share some lessons.

Introduction

As health policy analysis (HPA) fellows pursuing doctoral degrees, we ventured into discourse analysis (DA) with the aim of making better sense of policy documents, and examining the ideas, ideologies, power relations and contextual factors reflected in these documents. However, we quickly realised that making a methodological choice for DA is not easy.

Firstly, the term is broad and there are many methodologies to choose from. Secondly, while there is an expanse of literature on the theoretical concepts, relatively few practical, how-to guides are available – and what guidelines are available are not relevant to all HPA studies. In addition, the range of methods available indicates that it might be necessary to combine multiple methods in a single analysis.

After struggling with these issues ourselves, we held a number of informal reflective discussions to identify what had helped us make appropriate methodological choices. As with all research, we realised that the choice of DA method should be made based on the topic and theoretical underpinnings of the study. It was slightly easier for us to consider methodologies once we were clear on what we wanted to achieve at the end of the DA. Below, we share our key learnings.

Key learnings

Move beyond linguistics

The type of questions we were asking in our research clearly needed DA techniques that would enable us to move beyond counting word occurrences or analysing linguistics in the texts. It was helpful to think of the policy document as an interview transcript and therefore consider using a combination of deductive and inductive methods to analyse it.

For example, when analysing a minister’s public announcement of a new policy, a linguistic analysis that focuses on word choice (e.g. the use of evocative imagery to describe those who might reject the policy) or sentence structure (e.g. placing a powerful actor as the object of the sentence, rather than the subject) might aid in understanding the minister’s personal interests or sympathies. However, an analysis of the structure of the argument to defend the policy (i.e. whether it appeals to the authority of the WHO or the minister of finance; or whether the argument defends the policy in terms of how cheap it is rather than its capacity to protect the rights of vulnerable populations) can be used to inform an understanding of the ideologies underlying the decisions of policy-makers.

Explore combinations of methodologies/approaches

Initially, we focussed on finding one best-fit methodology for our studies. We soon realised that deriving or combining ideas from different methods might work better. While combining methods, we noted the importance of documenting the rationale behind our methodological choices.

Consider, for example, a study intended to explore how and why policy actors frame ideas during policy formulation and whether this changes over time. This may require borrowing concepts from process tracing, where one concentrates on specific events or moments in the policy and obtains information from interviews or documents to understand what factors led to the event. The same study may also include a DA method such as framing analysis, which accounts for the content of ideas, the process through which they are conveyed, and relates them to broader worldviews or social constructions. In addition, the study could be grounded in interpretive policy analysis that moves beyond a focus on factual analysis and considers that actors may have different interpretations of the policy content which can affect the policy development process. The study can therefore seek to understand the meanings and purposes that shape certain policy actions, especially those of the relevant policy actors. In this way one study can be enriched with concepts from various methodologies or approaches.

Allow for iterations

While doing DA, we realised that we have to move beyond the initial policy documents we had chosen to analyse in order to look for explanations of our findings. Hence, it was important to work with a flexible methodology that would allow for iterative selection and processing of documents.

For example, in the study of PHC units in India, the obvious documents to begin with were the national policies on health and reports of national committee meetings. Initial analysis of these documents revealed certain patterns in the description and evolution of PHCs. To understand the ‘why’ of such patterns, one needed to consider major events in global health and move beyond the initially selected policy texts. For instance, the period after the Alma Ata Declaration brought about a conspicuous change in the language and content of national-level policies on primary care in many countries. Also, local events such as war, drought or major political changes often led to healthcare issues taking a back-seat at policy level, despite global advocacy of primary care. An iterative consideration of documents that provided context to the PHC policy discourse explained why it unfolded in certain ways and allowed for depth in the discourse analysis.

‘Triangulate’ with others

To ensure the rigour of the DA findings, we found it useful to share some of the emerging ideas, themes and connections with others with expert knowledge of the subject, such as our supervisor and health policy analysts. There are likely to be multiple possible ways to make sense of the data, but sharing one’s interpretation with others can help to ensure that it is justified and resonates with those with different perspectives.

For example, if the discourse in a policy document is commonly assumed to be influenced by the country’s powerful for-profit pharmaceutical sector (and their attempts to ensure that new policies align with their interests), but your analysis of the discourse indicates that it is more aligned with neoliberalism in the development sector, you might find it helpful to discuss your data and analysis with an expert (i.e. someone who had direct involvement in the policy process). This can help to check whether your interpretation of the data accords with the experiences of those directly involved and may indicate that other possibilities need to be considered.

Conclusion

Our DA explorations have been fun, all the more so since we had a great group of policy fellows and mentors to discuss our thoughts with. It helped us to think of DA as an approach, rather than a methodology in itself. Reading methodological resources and other research based on DA also helped to clarify some of the confusion. However, we realised that it was unlikely that we would find any step-by-step set of instructions that would fit our DA needs exactly. So, in the end, we had to develop our own approach, by borrowing tools, techniques and tips from everything we read. Overall, it has been an exciting journey so far!