The box plot is drawn on a coordinate plane. The horizontal axis is labeled “Country” and contains two categories from left to right: “Small” and “Large”. The vertical axis is labeled “Prediction error” and ranges from 0.0 to 1.2 in increments of 0.2 units. For “Small”, the box extends from 0.10 to 0.34, with a median at 0.20. The lower whisker reaches near 0.00, and the upper whisker extends to 0.70. Multiple outliers appear above the upper whisker, ranging from 0.70 to 0.95. For “Large”, the box extends from 0.10 to 0.35, with a median at 0.20. The lower whisker reaches near 0.00, and the upper whisker extends to 0.72. Numerous outliers appear above the upper whisker, ranging from 0.72 up to about 1.22, including several extreme values above 1.0. Note: All numerical data values are approximated.Boxplot of prediction error by country. Note(s): This boxplot compares prediction error distributions between the small and large economies. The small country exhibits a narrower spread and a lower median error, indicating a more effective learning environment due to greater graph cohesion. All simulations use a unified GNN architecture trained on partially observable data enriched through synthetic generation methods. The Excel workbook attached to this submission includes all raw data and results. The R scripts used for the analysis and the Python scripts used for GNN modeling are available upon request or in the supplementary materials repository. Source(s): Author's simulations based on synthetic data calibrated to household survey and administrative records