Moneter Poverty Estimate With Molina dan Rao Method Application in County and City Malang

Abstract

The strategy of poverty solution needs the availability of accurate data regarding the poverty itself. One of the most important things on the poverty data is the poverty measurement. The poverty measurement is assumed to become a powerful instrument for the policy makers to focus their attention toward the living condition of poor people. The concept of poverty which is used by Badan Pusat Statistik (The Statistic Center) is the monetary approach by measuring the poverty based on the expenses per capita of the house holders. The indicators used by the BPS in measuring the poverty are the monetary approach developed by Foster, et. al. they are (1) the percentage of poor people [Head Count Index (HCI), P0], (2) Poverty Gap Index [(PGI), P1], and (3) the Distributional Sensitive Index [(DSI), P2]. The estimation calculation of the monetary poverty measurement is done directly by BPS which is based on the data of Survei Sosial Ekonomi Nasional (Susenas)/The National Survey of Socio Economy. This direct estimation method was not be able to give the good accurateness if it is conducted on the small scale sample, therefore the statistical result gained through this method will show the big various score as well as the lack of accuracy. Nevertheless, this situation can be solved by a method called the small area estimation method. This study aims to analyze the sample size on the monetary poverty estimation which is used by BPS and to find the alternative solution on the monetary poverty measurement estimation in the case small sample area. The approach of the study is divided into two groups – the simulation and the application. The simulation was conducted through nine scenarios based on the pattern of which has the combined parameter to make the same expectation score E(). It was also conducted through resampling data by using various size of sample, for 500 repetition. The application in this study is from two data sources, they are the data from the Survei Sosial Ekonomi Nasional (Susenas) in 2008 and the data from the Potensi Desa (PODES)/the village potency in 2008 in East Java. From Susenas data, there are two variables – the data of householders’ expenses per capita as the treatment variable (), and the data of village amount which has kelurahan status as the control variable for the householder member sample (The PODES is used as the source for supporting i.e the proportion of kelurahan status in each kecamatan. This data is used as the control variable for non-sample householder. The result on the simulation approach shows that if the sample size is small, so the estimation value resulted on the direct estimation method is unbiased with big variance. This situation is proved by the simulation comparison of all samples’ size and also by looking into the index bias behavior of bias Relative Bias (RB), Absolute Relative Bias (ARB), and Relative Mean Square Error (RMSE). Whereas, the result from the Bayes empiric found non-zero result. Therefore, the direct estimation can be corrected by the Bayes empiric estimation in term of small scale sample.