Identifying Poverty Vulnerability Patterns in Indonesia using Cheng and Chruch’s Algorithm

Abstract

Poverty remains a significant issue in developing countries, including Indonesia, where in 2022, the number of people living in poverty reached 26.36 million, with a poverty rate of 9.57%. The Central Statistics Agency (BPS) measures poverty using a basic needs approach, defined as the inability to meet essential food and non-food needs through expenditure. Individuals are considered poor if their average monthly per capita expenditure is below the poverty line. Research on poverty has evolved into a more multidimensional understanding, The Multidimensional Poverty Index (MPI), which identifies deprivation across three key dimensions: health, education, and living standards. This study aims to identify patterns of poverty vulnerability by applying the Cheng and Church (CC) algorithm through a biclustering approach using data from BPS. This quantitative method utilizes 13 multidimensional poverty indicators across 34 provinces. The CC algorithm begins by setting a threshold, followed by removing rows and columns with the largest residuals, adding qualifying rows and columns, and substituting elements to prevent overlap. The quality of the bicluster is then evaluated based on the Mean Squared Residue (MSR) value until optimal groups are formed. The results indicate that a threshold of 𝛿 = 0.01 generates seven biclusters with the lowest mean squared residual (0.0065), signifying optimal bicluster quality. Further validation using the Liu and Wang index reveals less than 50% similarity with other thresholds, reinforcing the uniqueness of these findings. MSR serves as a measure of homogeneity within the bicluster, similar to how uniform the level of poverty is within a region. If families have similar expenditures and are below the poverty line, they face similar challenges, resulting in a low MSR value. In contrast, the Liu and Wang index compares regions with different poverty alleviation strategies. These findings provide valuable insights for policymakers. For example, in bicluster 7, where specific interventions are needed in Papua and West Kalimantan, which face local challenges such as reliance on agriculture, low education levels, and limited access to sanitation and clean water.