Comprehensive Nationally Representative Panel Data Generation
In contrast to the existing studies, the present study will draw upon the largest nationally representative panel dataset available so far for rural Bangladesh. It will build upon a panel dataset that already exists, with a sample of 6300 households drawn from 180 villages from all over Bangladesh. The sample was drawn following a sampling design very similar to that of the Household Income and Expenditure Survey (HIES) carried out by the Bangladesh Bureau of Statistics (BBS). However, the usefulness of the existing dataset is somewhat circumscribed by the fact that it consists of only two rounds of survey over an interval of just three years (in 2010 and 2013). Such a short interval may be good enough to answer some of the questions related to poverty dynamics (for example, the effects of prices, wages, some of the safety net measures, etc.)), but a longer interval is needed for a more effective analysis of poverty dynamics since many of the causal processes affecting poverty may take much longer to have their effects felt (form example, economic growth and growth-induced structural change, education, health, shocks, etc.). For this reason, the proposed research would carry out another round of survey – in 2020, giving an interval of a decade, which should be long enough to study many of the causal processes more effectively. The objective of generating a nationally representative panel over a period 2010-2020 is to assess rural poverty dynamics as well as the transformation that has taken place over time. The importance of studying poverty dynamics is self-evident in the context of Bangladesh, where nearly one-fourth of the rural population still live below the poverty line (as of 2015). Until recently, the scope of undertaking of such studies was limited by the lack of longitudinal data, i.e., repeated surveys of the same households over time. Very recently, a few pioneering studies have begun to emerge, but the usefulness of most of them is circumscribed either by the fact that they were designed narrowly to study the impact of particular interventions (e.g., Khandker and Samad, 2013; Quisumbing and Baulch, 2013), or by relatively small sample sizes (e.g., Hossain and Bayes, 2009).
Study Team: S.R. Osmani, Binayak Sen, Monzur Hossain, Minhaj Mahmud, Nahian Azad Shashi and Mahir A. Rahman