Binary Logistic Regression Modeling on Household Poverty Status in Bengkulu Province

Abstract

Binary logistic regression is a statistical method used to analyze the relationship between one or more predictor variables and a binary or dichotomous response variable. Poverty is an issue in every province in Indonesia. One of the provinces with a relatively high poverty rate is Bengkulu Province, ranking seventh in Indonesia with a poverty rate of 14.62%. The Central Bureau of Statistics of Bengkulu Province (2023) explains that efforts to reduce poverty must involve all levels of society. Various government programs and policies in various fields such as health, social, and other areas are continuously being implemented to reduce the number of households classified as poor. Identifying the characteristics of households in Bengkulu Province by poverty status is important to study, as it serves as a

reference to ensure that government programs are implemented according to the target. One method that can be used to identify household characteristics is binary logistic regression. This study aims to model the poverty status of households in Bengkulu Province using binary logistic regression and to identify the factors that influence it. The data used are social and economic household data from March 2022. The response variable used is household poverty status (poor and not poor), while the predictor variables include the ownership of toilet facilities, the source of lighting, floor area, family size, and per capita calorie consumption. Modeling is done using binary logistic regression with simultaneous and partial parameter significance tests, as well as model fit tests. The analysis results show that the factors

significantly influencing household poverty status in Bengkulu Province are the ownership of toilet facilities, the source of household lighting, floor area, family size, and per capita calorie consumption. The formed binary logistic regression model has a classification accuracy of 89.98% with a sensitivity of 18.34% and a specificity of 98.61%.