Narrowing the development gap has long been and continues to be a key element of government aspiration worldwide. Since 2015, the Government of Indonesia has implemented the village fund (VF) transfer to enhance its rural economy, especially in remote areas. The impact of the VF on village development may vary greatly depending on the village’s location. This study examines the causal effects of VF transfer on the rural economic growth of underdeveloped villages in Indonesia. Using a nighttime light dataset at the village level as a proxy for rural economic growth and a regression discontinuity design in time, we found a significant improvement in rural economic growth in underdeveloped villages after the implementation of VF transfer. Our study confirms that the underdeveloped villages in East Indonesia are growing faster than those in West and Central Indonesia. The average growth of nightlight after the implementation of VF is approximately 156% in East Indonesia, 141% in Central Indonesia, and 98% in West Indonesia compared to the growth of pre-VF. Therefore, there is a strong argument to review the current formula of the VF to narrow the rural development gap in Indonesia.

This study examines the impact of village fund (VF) transfers on rural welfare in Indonesia. We use the monthly night-time light (NTL) dataset as a proxy for rural welfare for around 75,000 villages from 2014 to 2019. Using Regression Discontinuity in Time (RDiT), we find that VF positively impacts rural welfare, as the value of Average Light Intensity (ALI) increases by almost 97% after the implementation of VF. The highest influence of VF transfers is revealed to be in underdeveloped villages. This indicates preliminary evidence of regional convergence, where poor villages grow faster than developed villages. In addition to the RDiT, the Difference in Difference (DiD) estimations show that the rural welfare of villages that received VF early in 2015 was 19% higher than those without VF in 2015.

This study aims to develop a data-driven framework for classifying rural and urban areas at the village level in Toba Regency by integrating official statistical data, geospatial big data, and machine learning techniques. The current regional classification still relies on the 2020 baseline and may not adequately reflect recent socio-spatial transformations occurring at finer administrative levels. To address this limitation, this study integrates Village Potential Statistics (PODES) data with spatial indicators derived from big data sources, including population density from WorldPop and the Built-Up Index (BUI) extracted from satellite imagery. The integration of these datasets enables a more comprehensive representation of settlement patterns, spatial development intensity, and demographic distribution across villages. Three supervised machine learning algorithms were implemented to this study: Support Vector Machine (SVM), Naïve Bayes, and Random Forest, with model evaluation using accuracy, precision, recall, and F1-score. The analysis results show that the Random Forest algorithm provides the best performance. Based on the best model, of the 244 villages analyzed, 156 areas were classified as rural and 88 areas as urban. These results indicate a change in status in 47 villages compared to the previous classification. These findings indicate that integrating official statistical data with big data and machine learning methods can capture the dynamics of regional development more adaptively, potentially serving as a complementary approach for compiling regional classifications and formulating more targeted development policies.