There are infinitely many alternative benchmark weights that decision makers could choose to measure multidimensional poverty. To overcome the resulting uncertainty, we derive a feasible range of multidimensional poverty that considers all admissible weights within the chosen lower and upper bounds of weights. We use Kenyan and Canadian data to illustrate the use of our methodology, which is an adaptation of existing methods for portfolio analysis based on stochastic dominance. These two-empirical analyses suggest that different weights allocated to poverty dimensions can produce very different multidimensional poverty outcomes for a given population even in cases with small weight perturbations.
- Multidimensional poverty measurement
- Stochastic dominance
- OR in societal problem analysis
- Mixed integer programming