Complements, substitutes and multidimensional deprivation measurement

Since the inception of multidimensional poverty, it has been acknowledged that individuals’ poverty levels should depend on whether the different variables included in a poverty measure are substitutes (i.e., ‘very similar’ in nature, or ‘close to each other’) or complements (i.e., not ‘easily interchangeable’). This chapter reviews how current approaches in the literature on multidimensional poverty indices have addressed the issue of complementarity and substitutability across variables - differentiating whether the trade-offs or compensations across variables occur in different parts of the domain. Unfortunately, most commonly used approaches implicitly assume that all variable pairs are either complements or substitutes, an unreasonable and extremely rigid assumption that severely limits the possibility of modeling non-trivial interactions between variables in realistic ways. This chapter also reviews and proposes new directions of research attempting to introduce more realism and flexibility when modeling the relationship between pairs of variables.