Estimating multidimensional development resilience

Existing measures of resilience are typically based on a single well-being indicator. This is problematic in contexts where households face deprivations across multiple dimensions. We develop a multidimensional resilience measure, integrating probabilistic moment-based resilience measurement approaches with multidimensional poverty measurement methods. Applying these to household panel data from Ethiopia, we show that univariate and multidimensional resilience measures based on expenditure-based poverty, dietary diversity, and livestock asset holdings can yield varied inferences on the ranking of households as well as potential impact of development interventions. Univariate resilience measures constructed using consumption expenditure, dietary diversity and livestock asset holdings show distinct temporal and spatial distributional patterns. But while univariate measures are weakly correlated with one another and with different well-being metrics, multivariate measures exhibit much stronger rank correlations. When we contrast univariate measures of resilience to multidimensional measures of resilience, we find that the latter vary less over the study period; multidimensional resilience measures seem to capture more “persistent or structural” vulnerability and associated capacity of households. We also demonstrate the differences in these univariate and multivariate measures, including the potential of the composite multidimensional resilience measures for supporting targeting processes.