Strengthening internal horizontal equity: Detection of inconsistencies and dynamic updating of urban cadastral values using machine learning and bootstrap

Cadastral systems are central to property taxation, urban planning, and territorial governance, yet in many cities—particularly in the Global South—they suffer from outdated information, poor data quality, and inconsistent valuation. These limitations undermine both horizontal and vertical equity, weakening fiscal capacity and public trust. This paper proposes an integrated framework that combines machine learning and bootstrap-based simulation to improve the internal consistency and dynamic updating of cadastral values. First, we use Extreme Gradient Boosting (XGBoost) to estimate Expected Cadastral Values (ECV) based on observable property and location characteristics. Deviations between observed and predicted values are then analyzed to systematically detect anomalies associated with data errors, outdated records, or inconsistent application of valuation criteria. Second, we implement a bootstrap-based simulation to estimate Simulated Cadastral Values (SCV) under hypothetical changes in urban infrastructure, thereby isolating and quantifying the impact of improved accessibility on land values. The methodology is applied to Aracataca, Colombia, a data-constrained context characterized by limited institutional capacity for frequent updates. Results show that anomalies follow clear spatial and structural patterns, particularly in transition areas between valuation zones and in locations with low-quality data. The simulation further demonstrates that infrastructure improvements can generate significant increases in land values, highlighting the need for timely cadastral updates. By shifting the focus from prediction alone to consistency and adaptability, this study contributes to the literature on automated valuation models and provides a scalable, evidence-based tool to strengthen equity, transparency, and fiscal performance in cadastral systems.