Page 328 - Contributed Paper Session (CPS) - Volume 6

CPS1949 Khalid S.
unsettled. They relate mainly on four points: (i) the identification of the
relevant dimensions of deprivation; (ii) the aggregation of the dimensional
indications; (iii) the choice of the weighting scheme; and (iv) the determination
of the poverty line by dimension. Each of these methodological items widely
affects the targeting of the poor and the choice of relevant economic and
social policies especially those aiming to reduce poverty and inequalities.
The main purpose of this work is to assess the robustness of the
multidimensional indices poverty according to different weighting schemes.
So, it is necessary to make the tour of the methodological frames, with a
particular focus on the approach of the capabilities of A. Sen, which establishes
an adequate frame to measure the multidimensional poverty. To implement
this approach, three methodologies of measurement will be applied while
adopting a range of weighting schemes: (i) the fuzzy logic method; (ii) the
Bourguinon and Chakravarty method; and (iii) the AF method.

2. Weighting schemes: presentation of the main statistical methods
Every measure of multidimensional poverty sets somehow a weight to each
well-being dimension. However, the weighting scheme can vary in its
specification and the way it affects the estimation of weighted indices. From
then on, the robustness of the multidimensional indices to a range of
weighting schemes continues to be a serious challenge.
In this regard, it is important to analyze the sensibility of the poverty indices
according to various weighting schemes. Since 1988, different statistical
weighting schemes have been designed to facilitate the summing of
dimensional indices in a composite index. Generally, three statistical functions
of weighting can be distinguished, such as: The specifications of Desai & Shah
(1988), Cerioli & Zani (1990) and Betti & Verma (1998) and Betti & al (2007).
Although there are several possible formulations of these types of functions,
we present below some that are usually used to determine the weighting
coefficients:
i) The function of normalized weighting proposed by Cerioli and Zani (1990):

with f ) is the weight attached to the observation
of the sample ,xij ϵ [0, 1] denotes the value of a particular deprivation item j.
This formulation shows that the weight attributed to the factor j is an
inverse function of its degree of deprivation.
Ceriolis and Zani also developed another not logarithmic format:

ii) The function of normalized weighting proposed by Desai and Shah (on
1988)

317 | I S I W S C 2 0 1 9