Thomas Sohnesen, The World Bank

"Drought indicators and measured drought impact. The case of the 2015-drought in Ethiopia"


In 2015, Ethiopia was supposedly hit by a large-scale drought which prompted emergency responses from both government and international organizations. This study assesses the impact of the drought on rural households’ well-being approximated by consumption using a two-period panel data set. Using satellite- based drought indicators, we find that though there are signs of a widespread meteorological drought measured by rain and temperature, there is no sign of a widespread agricultural drought measured through vegetation (NDVI). Crop loss and production estimates also do not show sign of an agricultural drought. None of the meteorological or agricultural drought indicators show any impact from drought on consumption for rural areas in general. Looking at subpopulations, different drought indicators show limited impact for different subgroups, which is consistent with the drought indicators not being spatially correlated.

The observed no-impact of drought is unlikely to be explained by the distributed assistance in response to the drought, even though those with worse than normal vegetation are more likely to receive the assistance. Using self-reported drought exposure, we find a large negative impact on consumption. However, these results are likely driven by biased self-reporting reflecting drought impact as much as drought exposure, also known as endogeneity. The latter is problematic as self-reported data has underpinned most previous studies of impact of drought in Ethiopia. Hence, we find that the general existence of a drought heavily depends on the choice of drought indicator, and its impact on population is equally dependent on choice of drought indicator. Across time, we find good correlation between meteorological and agricultural drought in the smaller Belg season, but no correlation in the agriculturally more important Meher season. The results call for a much better understanding of drought indicators themselves, and their relationship to impact on households.

Contact person: Neda Trifkovic