3. Exploring exception-al circumstances: a mixed methods approach
3.3. A framework for integration
The discussion of mixed methods has been limited until now to discrete phases of the research project. Bryman set out a list of research purposes124; Creswell and Plano Clark, among others, advanced a range of research designs summarising the order of research and the integration of data125. Plowright has built on their work to create a framework for mixed methods research which, he claims, can be used to structure and support data integration regardless of the type (or types) of data used126. Dubbed the Framework for Integrated Methodology, or FraIM, his model eschews the distinction between qualitative and quantitative data. At each stage of the project, research choices are shaped by the level of involvement or ‘mediation’ on the part of the researcher, which dictates both the methods used and the ‘degree of structure’ the researcher imposes on the data collected via that method, giving rise to numerical or narrative data.
124
Bryman, Social Research Methods, pp. 608-9
125
Creswell and Plano Clark, ‘Choosing a Mixed Methods Design’, pp. 53-106
126 David Plowright, Using Mixed Methods: Frameworks for an Integrated Methodology, (Thousand Oaks,
He suggests that there are only three methods of data collection: observation; asking questions; and artefact analysis, or the study of documents, pictures, films etc. The ‘level of mediation’ supports the choice of method – the lower the level the more naturalistic the research setting. For example, a researcher may go unnoticed when they are observing a particular event, ensuring that it unfolds much as it would have done had the researcher not been present. Asking questions is highly mediated; outside of the research setting, it is unusual for people to indicate their answers to questions by ticking a series of boxes.
Once the broad research strategy has been chosen, the next stage is to consider the level of structure required, which dictates the data collection tools to be used. Researchers may ask questions in a number of ways – the most common in a research setting being a questionnaire or an interview. In the former, the researcher has imposed a significant degree of structure on the format of the questions in advance – particularly by dictating the possible answers to each question. The latter is less structured, as interview subjects are able to answer questions in their own words. Bryman has suggested that this structuring means that quantitative research tends to be driven by the researcher’s concerns, while a qualitative approach ‘takes the subject’s perspective as the point of departure’; combining the two provides balance and is further justification for employing both approaches127.
The degree of pre-structuring or pre-coding determines whether the data returned is numerical as with a questionnaire containing closed questions, or narrative as with interviews. The question may not be different, but how it is asked does vary, so too the results generated. It depends on the research question as to how structured, or not, the data collection should be. In this way, numerical and narrative data are seen as different points on a scale rather than as mutually exclusive elements not to be mixed. In this, Plowright is supported by Tashakkori who sets out a continuum for qualitative and quantitative data integration. All research
127 Alan Bryman, ‘Quantitative and Qualitative Research’ in Clive Seale (ed.), Social Research Methods: A
includes qualities that can be counted and quantified by frequency, even if only in vague terms such as ‘some’, ‘few’ or ‘all’, while quantitative data is interpreted narratively128.
This leads Plowright to suggest a three-dimensional model for integrated research, mapping the data collection approach against the strategies for ‘data source management’: experiment; case study; and survey. It indicates the types of data that will be generated depending on the case selection, sampling and structuring decisions and suggests how that data could be pulled together into a coherent whole. The model (Figure 3.2) comprises 18 options based on the choices made during the project. Options 8 and 17 both adopt a survey strategy and ask questions, but the former takes a structured approach to generate numerical data, while the latter uses semi-structured interview questions to generate narrative data.
Figure 3.2: Three-dimensional model of an integrated approach to designing research
Source: Plowright, 2011
Visualising the decisions made during research in this way presents a useful check for researchers on their research choices and the compatibility of the various types of data generated. Using the same data source management strategy and method of collection to
128
See Figure 19.1 in Charles Teddlie, Abbas Tashakkori and Burke Johnson, ‘Emergent Techniques in the Gathering and Analysis of Mixed Methods Data’, in Nagy Hesse-Biber and Leavy (eds.) Handbook of Emergent Methods, p. 390 and also Denscombe, The Good Research Guide: For small-scale social research projects, 3rd ed., (Maidenhead: Open University Press, 2007), p. 119
create both numerical and narrative data is likely to result in data that can be synthesized more easily than taking a narrative approach to artefact analysis in one part of the project and an experimental approach to generate numerical data elsewhere.
It is also more likely to generate what Plowright terms ‘warrantable research’129 – that is research, conducted in a credible and auditable way, such that any claims made or conclusions drawn stand up to scrutiny. The reason being that working through the project in a logical fashion from the original research question through case selection to data collection and analysis provides the support, or ‘backing conditions’, which can be used to justify the final conclusion. For example, explicit analysis of the theoretical and policy context means that the researcher is able to quote evidence from the literature that supports the arguments advanced. Similarly, the researcher can demonstrate that appropriate decisions have been made at each stage.
So how can the FraIM approach be applied to the current project on the cultural exception? As shown in Figure 3.3, the data being collected are all forms of survey data. In terms of the recorded interviews, the researcher may not have asked the questions personally, but questions were asked about French cultural policy and thus it will be possible to combine their review with analysis of the primary interviews. Similarly, the data collected for both the quantitative analysis of diversity and to understand the development of the video-on-demand market in France are derived from document analysis; it is simply that for the former numerical data was pulled from the CNC publications, and in the latter, narrative information was gathered. As long as the researcher understands the basis upon which the data have been generated, then these elements should be compatible. As mentioned, Creswell and Plano Clark provide us with the model for integrating the elements in a parallel research design, resulting in valid or warrantable conclusions.
129
Figure 3.3: Three-dimensional model applied to data collected on the cultural exception
Source: Plowright, 2011