Endogenous weights and multidimensional poverty: A cautionary tale

Multidimensional poverty measures have become a standard feature in assessments. A large and growing body of work uses endogenous (data driven) weights to compute multidimensional poverty. We demonstrate that broad classes violate key properties indices such as monotonicity subgroup consistency, without which evaluation policy targeting are seriously compromised. Using data from Ecuador Uganda we show these violations widespread. Our results can be extended other composite welfare the widely used asset indices.