Reproducibility Package for "Cross-Country Pooling in Unemployment Forecasting: Heterogeneous Gains and Negative Transfer across OECD Countries under Simulated Publication Lags"

This dataset and reproducibility package accompanies the study “Cross-Country Pooling in Unemployment Forecasting: Heterogeneous Gains and Negative Transfer across OECD Countries under Simulated Publication Lags.” The package contains frozen OECD-derived analysis extracts, derived modelling datasets, executable Python code, machine-readable point and probabilistic forecast results, robustness and leave-one-country-out outputs, model-tuning records, validation and leakage-control checks, independent-rerun checks, dependency records, and SHA-256 manifests required to inspect and reproduce the numerical analysis. The study evaluates local and cross-country pooled models for monthly unemployment forecasting across OECD countries at 3- and 6-month horizons under simulated publication lags. The package supports the primary point-forecast comparisons, probabilistic forecasting analysis, robustness exercises, and transportability checks reported in the associated manuscript. The underlying source data were obtained from the OECD Data Explorer. The included public analysis extracts are deterministic subsets of the frozen source snapshot and are restricted to the observations required for the registered analyses. Source provenance, data-vintage information, extraction boundaries, and checksums are documented within the package. Users should consult README.md and RUN_GUIDE.md before reproducing the analysis.