|
Some of the most
promising approaches for obtaining high-resolution maps use existing
data and generally require little or no additional data-collection
efforts. One approach combines survey and census data using
econometric techniques to overcome typical limitations in the
geographic coverage of household welfare surveys and the lack of
direct welfare indicators in censuses (Alderman et al., 1999; Elbers
and Lanjouw, 2001; Statistics South Africa, 2000).
A multidisciplinary, multi-institution
effort is now in progress in Kenya, Tanzania and Uganda aimed at
producing high-resolution poverty maps. Research teams associated
with the statistical bureaus of each country are involved in a
careful analysis and testing in order to develop a robust
statistical model. The first stage of the analysis starts with a
household survey such as a welfare monitoring survey (WMS) to obtain
a reliable estimate of household expenditure (y). Household
expenditure is then used to calculate more specific poverty measures
linked to a poverty line.
Most household surveys are designed to
produce statistically reliable results at a relatively coarse level,
such as the six units of analysis (urban and rural estimates for
three regions) used in Ecuador by Elbers and Lanjouw (2001).
Population and housing censuses typically do not collect information
on household expenditures, but provide complete coverage of a
country and can be aggregated to small statistical or administrative
areas such as villages and communities.
Combining survey and census data can
overcome the typical limitations of surveys (too small a sample size
to create high-resolution maps) and censuses (lack of direct welfare
measures), and it allows the production of statistically reliable
poverty indicators at much higher resolution, such as at the parish
level—1,000 units of analysis—in Ecuador) (Elbers and
Lanjouw, 2001).
This approach requires a common set of
explanatory variables (x) at the household level in both the survey
and the census. Typically, these variables are such household
characteristics as household size, educational level and quality of
housing, and are obtained from the same or similar questions in the
respective questionnaires.
The next step involves statistically
estimating the relationship between household expenditure (y) and
household characteristics (x) in the survey. Once robust
relationships between y and x have been established, the researcher
can apply these estimated relationships to the same variables, i.e.
household characteristics (x) in the census, to predict per capita
household expenditures.
Other small-area estimation (SAE) approaches
have extended this statistical estimation of poverty and include
other explanatory factors going beyond the variables in the census
or survey, by using geographic information system (GIS) data layers
to calculate information on, for example, distance to markets and
crop production.
Recent poverty maps
produced for South Africa highlight their potential usefulness
(Alderman et al., 1999; Statistics South Africa, 2000). Government
institutions in South Africa were the first major users of these new
poverty maps, but other researchers are also beginning to use the
new data. The poverty maps have been instrumental, for example, in
guiding educational outreach campaigns to slow the further spread of
cholera outbreaks that started in Durban in September 1999. By
combining epidemiological information with poverty data and
demographic profiles from the census, the outbreak of the disease
could be clearly linked to some of the poorest areas in Durban.
Researchers also established a close correlation between access to
sanitation and clean drinking water (from more detailed census data)
and the prevalence of the disease. Other users working on crime
issues combined information on the location of crime hot spots with
the poverty maps and census data to characterise communities
neighbouring these hot-spot areas. This helped to formulate first
hypotheses on potential linkages between community characteristics
and crime.
References
Alderman
H., Babita M., Lanjouw J.O., Lanjouw P., Makhatha N., Mohamed A.,
Ozler B. and Qaba O. 1999.
Is
census income an adequate measure of household welfare? Combining
census and survey data to construct a poverty map of South Africa.
Draft report, available from Statistics South Africa (StatsSA)
www.statssa.gov.za.
Elbers
C. and Lanjouw P. 2001.
Intersectoral transfer, growth, and inequality in rural Ecuador. World
Development 29(3): 481–496.
Statistics
South Africa. 2000. Measuring poverty in South Africa. Statistics
South Africa (StatsSA). Pretoria www.statssa.gov.za/Publications/Publications.html
|
