Modeling Spatial Ordinal Logistic Regression and The Principal Component to Predict Poverty Status of Districts in Java Island

This study is intended to determine the factors that affect poverty-stricken districts in Java Island by using spatial ordinal logistic regression (SOLR). This study is urgent for local governments since they have a mission for the welfare of their citizens, especially alleviating poverty that occurred in each region. The evaluation is based on the best model and to evaluate SOLR model with and without handling multicollinearity. The principal component (PC) is for handling multicollinearity. This research used secondary data (Kusumaningrum, 2010)[6]. Results showed the ordinal logistic regression (OLR) model with all explanatory variables spatial has Correct Classification Rate (CCR) value of 50%. Then the OLR with PC model has a CCR of 24%. So it can be concluded that SOLR model is better than OLR with PC model. Based on spatial model without PC, there are five variables that influence the level of district poverty on Java island, namely: the number of farmers, small industries, families without electricity, hospitals and spatial variable on districts poverty. The variables that have negative correlation are the number of farmers, families without electricity and spatial variables. Increase in the number of farmers and families without electricity will reduce the chances of a district becoming rich. The spatial correlation is negative, which means that if a district is surrounded by districts with high poverty levels, the chances of a district to become more wealthy will decline.