How is the MPI computed when there is incomplete data on both health indicators?

Incomplete data on the health dimension were collected in the Brazil’s Pesquisa Nacional por Amostra de Domicílios (PNAD). While PNAD collects rich data on social, economic and demographic characteristics of households and their members, some data that are needed for the MPI computation, such as basic nutrition characteristics (body weight and height of children and adults) are not collected. Although the survey collects household information on deaths of children (of any age), it does not collect the date of the death, so it is not possible to distinguish recent deaths from those that happened more than 5 years prior to the survey. Thus, the child mortality indicator has incomplete information on deaths of children. Given the lack on nutrition, this implies that the imperfect child mortality indicator would receive the full dimension weight of 1/3.

In such a case, the MPI methodology was modified to compensate for this limitation in the health dimension and two adjustments were performed: (i) only child mortality reported by mothers of ages between 15 and 49 was considered, and (ii) a household that is deprived only in child mortality cannot be declared multidimensionally poor. The first adjustment increases the likelihood that the reported deaths happened more recently and that they refer to a child under the age of 18 year, while the second removes the possibility that a household is declared multidimensionally poor based only on the death of a child that happened any time before the survey. This second rule means that an additional deprivation in at least one of the remaining indicators is needed for such a household to be declared multidimensionally poor.

Brazil is the only country that lacks information about nutrition and that lacks information about date of death for the child mortality indicator. There are 9 countries that lack data on deprivations in nutrition but with complete data on child mortality. Likewise, there are another 11 countries without date of death, but with data on deprivations in nutrition, therefore, the weight attached to child mortality is only 1/6 making the MPI estimates less dependent on the treatment of child mortality.