Nighttime Lights in Unit-Level Small Area Estimation for Estimating per Capita Expenditure

Diaztri Hazam (School of Data Science, Mathematics, and Informatics, IPB University, Bogor, Indonesia)
Erfiani Erfiani (School of Data Science, Mathematics, and Informatics, IPB University, Bogor, Indonesia)
Anang Kurnia (School of Data Science, Mathematics, and Informatics, IPB University, Bogor, Indonesia)

Zero : Jurnal Sains, Matematika, dan Terapan
Vol 10, No 2 (2026): Zero: Jurnal Sains Matematika dan Terapan

The most widely used auxiliary variables for estimating per capita expenditure using small area estimation (SAE) are from National Socio-Economic Survey (SUSENAS) or Village Potential (PODES) data. Another alternative is remote sensing, which can quickly and cheaply identify area characteristics, such as nighttime lights (NTL). This study will compare the SAE model with auxiliary variables using PODES data, NTL data, and a combination of both. The method used is a unit-level SAE model with log-transformation to estimate per capita expenditure at the subdistrict level in Bandung Regency.  Model performance was assessed using Relative Root Mean Squared Error (RRMSE), Akaike Information Criterion (AIC), and Bayesian Information Criterion (BIC). Estimation from all models have a similar range. The model with auxiliary variables using PODES data and combined data has similar RRMSE, AIC, and BIC. The model with auxiliary variables using only NTL data has the smallest RRMSE, AIC, and BIC.