Responding to the Needs for Decentralized Planning: Small Area Poverty Estimates

Type Conference Paper - 10th National Convention on Statistics (NCS)
Title Responding to the Needs for Decentralized Planning: Small Area Poverty Estimates
Publication (Day/Month/Year) 2007
URL papers/ips-14/ips14-01.pdf
Micro-level estimation of poverty has been increasingly becoming prominent among statistical and economic tools geared towards assessment of welfare indicators. The strength of this methodology comes from its capability to address the
limitation imposed by detailed household surveys that include reasonable measures of income or expenditures but are rarely representative at low levels of disaggregation to yield statistically reliable poverty estimates. In particular,
information from census (or other large sample) data, which are of sufficient size to allow disaggregation at local level but do not have data on welfare indicators, is combined with detailed welfare information available from household surveys. This is implemented by constructing a model of (log per capita) income / expenditure using the survey data set, restricting reg ressors to those that can be linked to households in both data sets. The expected level of poverty indicator is then estimated given the census -based observable characteristics of the population of interest using the
"synthetic" parameter estimates derived from the model. In the Philippines, official estimates of poverty incidence are available at the national, regional and provincial levels. In response to the need of stakeholders, NSCB implemented a project, Estimation of Local Poverty in the Philippines, through funding assistance of World Bank ASEM Trust Fund. The project’s outputs include poverty statistics for the 1,623 municipalities, with the year 2000 (being both a census and survey year) as the reference period. At present, the estimates are being updated for the year 2003, a survey but a non-censal year in the country. However, the small area estimation methodology gets a little complicated when the reference years of the two data sources are not the same. (Hence, the linking of explanatory variables from survey to census is not straightforward). When census is conducted less frequently than household surveys, small area poverty estimation for intercensal years requires the availability of either panel data or time-invariant regressors that can be extracted from the data sets. This paper discusses the methodology of constructing welfare model(s) for censal and intercensal years. Further, the paper enumerates the empirical usefulness of small area poverty estimates to policy-makers, stakeholders and antipoverty organizations in the country, based on the experience gained in the dissemination and utilization of the outputs of the project on the Estimation of Local Poverty in the Philippines project.

Related studies