Exposure to abnormal floods is believed to have negative short- and long-term consequences for welfare and health in poor countries, and such impacts are likely to grow worse with continued anthropogenic climate change. However, two common proxies for flood exposure, self-reported exposure and rainfall, are problematic. This paper describes a method for constructing objective measures of flood exposure using satellite data. Using the case of Bangladesh in the period 2002-2011, we show that (a) self-reported exposure has an important, non-random bias in that it responds much more strongly to actual exposure in areas where floods are relatively rare, and (b) rainfall is at best weakly correlated with floods.
The data sits under “Additional Materials”. It’s in Stata format. The folder contains the stata .do file ‘guiteras_jina_mobarak_replic.do’. In the ‘data’ subfolder are two data sets. ‘cmns_2005_union_flood.dta’ contains data on floods in each of the CMNS unions. ‘rain_flood_corr_hist.dta’ contains a series of correlation coefficients necessary to plot figure 1 of the paper.
Child and Maternal Nutrition Survey 2005: The CMNS data are property of the Bangladesh Bureau of Statistics (BBS). Data were obtained in person by the authors in 2012. Contact details for the BBS can be found on their website at www.bbs.gov.bd