Borrowed Youth: Demographic Change and Multinational Location Advantage

Borrowed Youth develops a temporal account of demographic location advantage. Workers available at opening and routes to their successors deliver different services over committed capital's life. The model connects international resource bundling, global staffing, home-linked recruitment and shared foreign-affiliate organisation. Bayesian investment analysis separates a directional duration contrast from the joint restriction that both workforce sources attract capital and arrivals gain relative weight. Under the stated priors and approximate cluster-t summary likelihood, the primary directional posterior probability is 70.9%. On identical allocations it is 69.0% for machinery lives and 10.8% for accounting lives. The joint empirical restriction has below 0.1% posterior probability in the primary specification. These results do not establish the predicted investment pattern. No untested measurement explanation is presented as a finding. Jaguar Land Rover's Nitra-Puebla record supplies finalists, a twenty-year appraisal and funded and unfunded rankings that bound the funded winning margin by the EUR91 million award. Origin-specific recruitment agreements and delivery terms are unobserved. The JLR probabilities are conditional staffing scenarios, not identified probabilities of the firm's mechanism. Under a common flat-margin specification, both recruitment offers have identical likelihoods and their Bayes factor is one. The observed orderings constrain the margin without learning which offer the firm held. Assumed hiring shares, staffed-year value, tenure, discount rate, costs and prior route odds determine the scenario results. Version 3.60.0 corrects the interpretation of these probabilities and the empirical claims. It adds 5,924 conditional appraisal sensitivities, 27 panel prior and likelihood-scale comparisons, an explicit prior-to-posterior decomposition, independent integrations, and main-text data definitions and estimating equations. No observations are added and no regression is refitted. Earlier results and theory remain available. The manuscript distinguishes the full-rank asset panel from the underidentified capital-expenditure sample. The public release supplies anonymous manuscript and appendix PDFs, complete LaTeX, editable tables and reproducible code. Licensed WRDS rows require subscriber access; queries and public aggregate outputs are supplied. There are 2,117 replication files, 2,116 checksum entries, 9,994 words across all main PDF pages, and 20 full substantive pages before references. README.txt describes restoration. ChatGPT and Codex (OpenAI) were used in October 2026 for coding help, literature search, and red teaming and review. All figures were drawn with AI-assisted code from the model, demographic data and statistical estimates.