2.1. Setting context: risk perception and climate change There are multiple approaches we could have applied to develop our conceptual framework for exploring barriers to and modes of adaptation in agriculture in Bangladesh, including work by Smit and Skinner (2002), Howden et al. (2007) and asset based approaches, that map links between assets and livelihoods (Heltberg et al., 2009). These approaches focus on limits (Adger et al., 2007, 2009) and barriers to adaptation (Moser and Ekstrom, 2010). However, in trying to understand both scientific and cultural factors, and trying to assess the role played by formal institutional communities of practice, we adopted a risk perception framework as our conceptual frame, and applied the model used by Raymond and Robinson (2013) that assessed the perspective of formal institutions and rural communities of practice about factors impeding or helping adaptive practice. We take their definition of institutions (after North, 1990) as our own: “groups which follow rules and procedures that are created, communicated and enforced through channels widely accepted as official, such as courts, legislatures and bureaucracies”. Understanding people's knowledge of climate change has been a major theme in research on risk perception, and can help garner many insights (Maslin and Austin, 2012; Leiserowitz, 2006) enabling institutions and policy makers to understand local perceptions of policy preferences and to develop adaptation policies accordingly (Patt and Schroter, 2008; € Leiserowitz, 2006). Perception matters for climate change adaptation.
2.2. Site selection To investigate this issue, we applied a case study approach to data collection and chose two Upazilas (sub districts) situated in the north-west region of Bangladesh: Nachole and Chapainawabganj Sadar. Both are found within the Chapainawabganj district in Bangladesh. These sites were selected given their strong reliance on agricultural production, documented likelihood to experience drought and diversity of climatic zones. Currently, drought, more than flooding or submergence, is causing major yield reduction in agricultural production as compared to problems of flooding or submergence in this region (World Bank, 2013). Although results were geographically focussed, overall insights from this work will enable deeper understanding of how institutions can interact with farmers in this and similar areas (Alam and Rahman, 2014). Islam and Walkerden (2015, p.1710) advance a similar argument to justify their own study which is similarly based on two case studies: “though the current study is based on only two study villages, the findings … are expected to apply generally”.
2.3. Data collection We used mixed methods to undertake the research. The use of combined methods is well established in the literature. In this case, we undertook a quantitative survey of farmers in the case study sites, and undertook a series of qualitative semi-structured interviews. Firstly, we conducted a survey to assess farmer perceptions about the role of institutions in enabling climate change adaptation in agriculture. The survey provided an avenue to explore the farmers' perception of climate change, external institutional interventions including assistance, training, and provision of information by formal institutions. Farmers who were part of an institutionally sponsored climate adaptation project or activity in Bangladesh were termed as ‘affected farmers’ and farmers not involved in the programs sponsored by formal institutions were termed ‘unaffected farmers’. Farmers were then asked to identify and then assess the interventions by formal institutions. A total of 30 farmers including 15 affected and 15 non-affected farmers agreed to participate in the survey, which provided us with the opportunity to see the difference in experience and perception between those who had, and those had not directly used or applied institutionally resourced interventions/adaptations.
2.4. Data analysis We analysed the survey data by using a statistical software package, PASW Statistics 18.0 (formerly SPSS Statistics) (Norusis, 2010) and thematic analysis with the interviews. The thematic analysis allowed us to identify and analyse patterns within the data, which we then coded. Coding processes included; (a) becoming familiar with the content via close reading, (ii) generating initial codes, (iii) searching for themes in the text and then assigning content to those codes/categories and then (iv) reviewing the whole (Raymond and Robinson, 2013).