ABSTRACT

Poverty mapping is crucial in order to find which are the critical regions to
which policies aimed at reducing poverty should be targeted and then allocate the
corresponding funds in a rational way. Unfortunately, when detailed maps are re-
quired for smaller regions or population subgroups, often official surveys do not
have enough sample data within all the target regions to provide reliable regional
estimates. Those regions or population subgroups that are not well covered by the
sample are called "small areas". For those areas, direct estimators, which use solely
the data from the corresponding area, do not have enough precision. Indirect es-
timators assume models that help to "borrow strength" from related areas. Small
area estimation of poverty indicators is a challenge because most of poverty indi-
cators are non linear with complex shapes. The basic procedures for small area
estimation of general non linear parameters will be reviewed. More recent contri-
butions that try to extend the basic methods to a wider range of situations will
be also described. The goodness of these methods will be illustrated by the results
of simulation studies. Poverty maps obtained in an application with Spanish data
from the Survey on Income and Living Conditions will be also shown.

Keywords: empirical Bayes; hierarchical Bayes; linear mixed models; poverty in-
dicators; small area estimation.