Investigating variations in the sustainability of poverty escapes is a big challenge, yet one we are currently working on. Success in this endeavour will contribute to a better understanding of how escapes from extreme poverty and deprivation can be sustained over time.
The research attempts to undertake an in depth analysis of how poverty reduction can be measured and compared across Rwanda, Ethiopia and Tanzania, in the project focused on understanding and supporting sustained pathways out of extreme poverty and deprivation. Given the cross-country comparisons required in this study, we decided to use an innovative approach based on a mixed-methods training using blended learning.
In this blog I first explain what blended learning is. I then go on to provide a simple explanation of qualitative comparative analysis (QCA) and how statistics and QCA together can be used to explore causality, and I thereafter provide some resources to help researchers and practitioners engage in QCA in their respective studies.
Exploring Blended Learning
Blended learning is where a teacher or trainer has some face-to-face time but also has online-mediated interaction with the learners. The online-mediated work is usually moderated by someone who checks that uploads or the comments made in emails are polite, kind, and appropriate. The idea of a ‘safe space’ applies to all learning situations. It means, for example, that no racist comments would be tolerated; nationalism would be toned down before a message is approved for release to a wide partnership group and any sexism would be actively discouraged. Within a safe learning space, the partners can then reveal weaknesses and revel in having their questions answered, without any sense of fear or worry. Sometimes it’s hard for firms to create such safe spaces because of the wage hierarchy, but at CPAN as part of its ESRC project, we are trying it out. We are doing so as we believe that without a safe space, you simply cannot learn new techniques.
Moreover, blended learning can be a good example for other current and future CPAN projects to adopt, especially for those that are remotely managed. Within that, the workshop has been structured around three distance-meetings. The initial one is for them to do without me being present, having sent some background reading suggestions. Then we hold two face-to-face (computer based) virtual workshops.
Blended learning also involves creating relevant online video or audio. It is usually expensive to make new video because of the editing stage, which can take up to six hours minimum for one hour of good quality raw video time. There is also a time-consuming rendering stage (creating the online version) and an uploading stage, e.g. to Youtube. In view of these costs we are simply making a link to a video that the Economic and Social Research Council (ESRC) has sponsored.
One such video that I have created is located in the website EPrints of the Research Methods Festival, a huge conference held at Bath University in Summer 2016. This was funded by the National Centre for Research Methods. Similar government investments in key research infrastructure can subsidise international partnership costs. This will be even more common once the Global Challenge Research Fund gets going in 2017.
Exploring QCA, A Method for Studying Causes and Effects
We may use statistical methods to test for causality of some factors for a particular outcome using both qualitative and quantitative data. Typically, the outcome variable might be whether a household has experienced ‘Poverty Reduction’ or has experienced a ‘Sustained Poverty Escape’, wherein their income have risen above the poverty line and their position in subsequent years has remained above the poverty line. An alternative is to look at asset trajectories. Clarity on the ‘outcome’ is crucial for tests and debates to work well on poverty reduction. It is common convention in research and policy settings to use adjusted per capita household income or expenditures and compare this to a national or district-level poverty threshold.
To thereafter test for causality, we also need to develop an understanding of causes of our outcome of interest. These causes will typically include variables such as household demographics including information around age-group, gender, whether widowed, overall health, and disability. Then we add to that the interventions, e.g. membership in a noon meals scheme or a microfinance scheme; pension available in a 5-year period; and housebuilding subsidy. These tend to be yes/no variates, called ‘binaries’. We set up a big matrix as a table with one line per household, summarising change over time and initial structural conditions. We have given an example here.
We next look for patterns in this set of data. A statistical F test can be used to check that the results are robust. Weights can also be used as appropriate. This new method of analysis is based on crisp and fuzzy set membership in the set of households that experienced poverty reduction. For more details on the process and accompany examples from a host of publications, click here.
At the end of the day, our aim is to develop a better understanding of the drivers of poverty trajectories and reduction according to quantitative and qualitative data sources, all the while enabling cross-country comparisons to be made. Stay tuned for more as results emerge in the coming months.
We will meet twice face-to-face to review each country team’s progress with the research. I hope the blended learning will work.
Written by Wendy Olsen, University of Manchester
Visit the Integrated Mixed Methods Network on Facebook.