Hierarchical Bayes Small Area Estimation for Gender Parity Index in Education: Case Study East Java Province

Rifdatun Ni'mah, Nur Iriawan

Abstract: Gender gaps can be identified through indicators i. e. Gender Parity Index (GPI). It is calculated based on the ratio of the School Participation Rate (SPR) of women to men at every level of education. The direct estimate of GPI is obtained from the result of the survey which has a small sample size for each level of education. An estimation variable with a few available samples can be done by Small Area Estimation (SAE). Our applied Hierarchical Bayesian (HB) approach for SAE using model linking log linier to estimate the IPG on the three levels of education in East Java province. Markov Chain Monte Carlo (MCMC) with Gibbs sampling methods are consider to solve the computational aspect. HB SAE under the Fay - Herriot model is obtained using some auxiliary variable such as a number average of household members, expenditure average per capita per month, the number of building schools and region’s education budgets. There are two significant results. First, HB has a better strength in explaining the variability in areas with small sample sizes and could reduce CV’s direct survey estimates about ± 29.21%. Second, cross validation’s result shows that the HB SAE model for all levels of education is fit to the actual observations and reliable.

Keywords: Small Area, Hierarchical Bayes, MCMC, Gender Parity