Hierarchical bayesian estimation of poverty incidence for the provincial level in the Philippines

Poverty incidence is defined as the inability of a household to meet the poverty threshold. In order to provide reliable statistics at the provincial level, the government relied on direct small area estimation. The poverty incidence estimates provided by the direct estimation methods have noticeably large standard errors.A possible solution to this problem is through hierarchical Bayesian estimation. Bayesian statistics uses Gibb\u27s sampling and simulation techniques to provide estimates with small Markov Chain Monte Carlo (MCMC) errors. In this paper, the hierarchical Bayesian beta-binomial model was used in order to provide estimates for the poverty incidence at the provincial level using the 2006 Family and Income Expenditure Survey (FIES). Results show that the ranking of the ten poorest provinces using the hierarchical Bayesian estimation is different from the ranking using small area estimation method. Furthermore, the estimates using hierarchical Bayesian estimation procedures have lower standard errors