A Study to Estimate Head Count Index in Small Areas with M-quantile Regression Model Case Study : Poverty in Bogor District Year 2015

Abstract

Poverty should be overcome with data. Problem arises when poverty is identified in sub-distric level, yet the data are in district level. Alternatively, M-quantile regression modeling in small area estimation as an indirect estimation approach can be done to measure poverty level in sub-district region with the use of district-scaled or even province-scaled data. In this paper, a Monte Carlo simulation will be conducted to empirically evaluate M-quantile modeling which contaminated area random effect and individual random effect to estimate head count index. M-quantile modeling is chosen because it is quantile-based semiparametric method which guarantees statistical estimation to be robust. Both direct and indirect estimations are performed and the the results of both estimations will be compared in each scenarios. The goodness of fit will be measured with bias and root mean squared error (RMSE). The result shows that M-quantile modeling is effective when there are outliers in individual random effect. Finally, results of application of M-quantile regression modeling to National Socio-economic Survey in Indonesia are presented.