Diverging economic futures under demographic uncertainty: a global hierarchical Bayesian model of the association between population and GDP

This dataset contains the posterior distributions (NetCDF format) and projection baseline files from a global hierarchical Bayesian analysis investigating the nonlinear association between population and aggregate economic output (GDP). The repository includes the complete posterior samples required to reproduce all figures, scenario projections, out-of-sample temporal backtests, and sensitivity analyses presented in the manuscript "Diverging economic futures under demographic uncertainty." The included files are: hierarchical_model_rcs_v2.nc: The primary production posterior (flat xarray format) used for the demographic forward pass to 2050 and the baseline pipeline. hierarchical_wa_posterior.nc: The posterior distribution from the linear working-age companion model, used to assess demographic expansion-contraction asymmetry and project long-run macroeconomic realignment to 2100. trace_fixed005.nc: ArviZ InferenceData for the strongly regularized (fixed first-difference penalty, RW SD = 0.05) spline specification, yielding a smooth, structurally sub-unitary elasticity. Used for model comparison (LOO-CV) and supplementary diagnostics. trace_adaptive.nc: ArviZ InferenceData for the data-driven (adaptive ridge) spline specification, yielding a volatile, super-unitary decline elasticity. Used for model comparison (LOO-CV) and main-text sensitivity analyses. These files provide the probabilistically coherent basis for projecting global economic trajectories under United Nations World Population Prospects (WPP) demographic scenarios. They are designed to be executed alongside the Python/PyMC analysis scripts and projection engine hosted on the project's GitHub repository (https://github.com/denovo2021/EPOCH) to ensure strict computational reproducibility.