Forecasting Municipal Financial Distress in South Africa: A Machine-Learning Approach

The study uses a compiled municipality–year panel for South African municipalities covering the 2018/19 to 2022/23 financial years (July–June). The analytic dataset is provided as an Excel panel (Main_Panel_model_ready.xlsx) containing the final predictors and the outcome variable used for one-year-ahead financial distress prediction, with municipality and year identifiers to support time-structured evaluation. An accompanying panel (Main_Panel_with_NT13_Composite.xlsx) contains the reconstructed composite-score distress labels based on the NT-13 framework. The replication package includes Jupyter notebooks for data preparation, distress label construction, main model estimation, and robustness analyses, together with supplementary tables and a Top-30 prioritisation output for the out-of-time test year.