FINDINGS 2: FINANCIAL DIMENSIONS OF WELLBEING IN OLD AGE
The collaboration between the International Longevity Centre-UK (ILC-UK) and the Personal Finance Research Centre (funded by the ESRC) uses secondary data analysis to look at financial dimensions of wellbeing in older age. As well as analysing large-scale survey datasets, we are undertaking analysis of two qualitative datasets to explore the role that finances play in older people’s quality of life. This blog describes some of the methodological issues we have encountered in conducting secondary analysis of qualitative data collected by others.
1. Choice of datasets
Archiving large-scale survey datasets for secondary data analysis is relatively common, particularly in the case of publically funded research. The same is not presently true for qualitative data (at the time of writing the UK Data Service returns around 400 results for qualitative and mixed methods data, compared with 4,100 results for UK survey data). This may be due to factors such as lack of awareness among academics and researchers, conditions of funding or concerns about data protection and respondent anonymity.
As a result, we had a limited choice of datasets to explore the research questions we were interested in. When writing our research bid, we identified five potentially relevant datasets in the ESDS Qualidata Online database (now part of the UK Data Service). We rejected three of these datasets because they had very specific remits (e.g. focused on widowhood) or because they were too old (generally pre-2000).
The remaining two datasets (both from ESRC-funded projects) offered the most potential to address the research questions we wanted to answer:
• Boomers and Beyond: Intergenerational Consumption and the Mature Imagination, carried out between 2005 and 2007 by Keele University
• Adding Quality to Quantity: Quality of Life in Older Age, carried out between 2000 and 2002 by University College London.
We had to ask permission to use the Adding Quality to Quantity dataset (which we only did once we had secured funding) so we initially had limited information about its content. Access to the Boomers and Beyond dataset was unrestricted, so we were able to review some of the transcripts at the bid stage, which reassured us that it contained data relevant to our research questions. Even so, without any prior involvement in the research it was difficult to be sure of the extent and relevance of the data without analysing it in detail.
2. Answering new research questions with existing qualitative data
We found it challenging to impose new research questions on data that was collected for another purpose, even though the research topics covered similar ground.
Once we had secured funding and accessed both datasets, we found that the topics outlined in the interview topic guide, while relevant, were not always covered in sufficient detail for the purposes of our research questions. As qualitative topic guides are never intended to be prescriptive, this was not a surprise, but it did limit the depth of secondary analysis that we were able to undertake.
In addition, we were dependent on the socio-economic and demographic data that was made available. A particular issue for a study of financial well-being was that only one of the studies provided information about individual and household income. In the other study, income information was not provided with the dataset and respondents were not asked about income in the qualitative interviews.
In hindsight, we would have benefitted from having data from more than two studies, in order to increase our chances of identifying data relevant to the (new) research questions we wanted to answer. This would require even more coding and summarising of data than we undertook, which itself was more time-consuming than we anticipated.
3. The age of the data
The age of the data (whether qualitative or quantitative) is an important consideration when thinking about the relevance and validity of the findings. One of the qualitative studies we looked at was conducted between 2000 and 2002, the other between 2005 and 2007. So by the time we re-analysed the data, it was quite old. We were, however, unable to find any more recent data that was relevant to our research questions.
In our analysis, the age of the data was particularly pertinent in fast-changing areas of life such as communications. Mobile phones were far less common when the data was collected, as was access to the internet and activities such as online shopping. There have also been changes to pensions and benefits in the time since the data was collected.
4. Distance from data
Not surprisingly, we found the experience of re-analysing qualitative data that others had collected very different from analysing (or re-analysing) data we collect ourselves.
Because we were not involved in the study and we had relatively limited socio-economic and demographic data available to us, we were only able to ‘get to know’ the respondents to a certain extent. This meant that many of the nuances of the qualitative research process were lost to us. For example, we only had a limited sense of the respondents as individuals and we had little information about the home environment in which the interview was conducted. Both datasets were anonymous in terms of place as well as people, which meant we could not situate respondents in terms of geographical location either. In other words, we lost some of the immersive and ethnographic aspects that are unique to qualitative research.
Despite the challenges and limitations we have described, secondary data analysis is a useful and cost-effective research method. We have valued the opportunity to re-analyse these two datasets and look forward to sharing some of the findings as the project progresses.
Sharon Collard, Sara Davies and David Collings, Personal Finance Research Centre
This blog can also be found on the PFRC website.