OPTIMISING ANTI-POVERTY TRANSFERS WITH QUANTILE ... - Ivie
You also want an ePaper? Increase the reach of your titles
YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.
13<br />
to the poorest of the poor. The ‘p-type’ rule is applied to the case of the head-count<br />
index (FGT index with α =0). In that case, ‘p-type’ transfers just express that it<br />
is least costly to start transferring from the poorest of the poor and sweeping up<br />
the income distribution, if the aim is merely to reduce the number of the poor.<br />
For the economically most interesting cases (k convex), the above algorithm<br />
depends on Euler conditions, including the budget condition. No negative transfer<br />
is necessary and the condition of positive transfers is never binding under perfect<br />
targeting. We shall show how to implement a similar procedure under imperfect<br />
targeting and even with X multivariate.<br />
2.1.1. The imperfect information case<br />
Unfortunately, perfect targeting is not feasible because incomes cannot be perfectly<br />
observed.<br />
Nevertheless, since the household living standard is correlated<br />
with some observable characteristics, it is possible, as in Glewwe (1992), to think<br />
about minimizing an expected poverty measure subject to the available budget<br />
for transfers and conditioning on these characteristics. In practice, the approach<br />
followed in the literature falls short from such a lofty ambition. Practitioners, including<br />
Glewwe, design the APTS by merely replacing unobserved living standards<br />
with OLS predictions based on observed variables and working with the observed<br />
sample as if it was the global population. We shall refine this approach.<br />
Under perfect information, the social planner needs to use observed characteristics<br />
X rather than unobserved incomes y to implement the transfers.<br />
We<br />
therefore consider the following optimisation program.<br />
min<br />
{t(X)}<br />
Z +∞ Z +∞<br />
−∞ −∞<br />
P (z, y + t(X)) dF (y, X)