Borrowed Youth and the Geography of Production Investment: Article and Replication Materials

Borrowed Youth and the Geography of Production Investment explains how country demographic trajectories and multinational organisation determine the continuity of labour-based location advantage. Opening access, local cohort succession and recruitment provide different dated services over committed production capital. Home-linked routes, shared foreign arrangements and admission institutions change host rankings.

Completed public article and replication, version 3.24.0, 28 October 2026. This version clarifies when costly capital redeployment makes workforce renewal consequential, how recruitment expansion can induce admission-policy feedback, and how legal permission differs from operational staffing. It explains that negative conditional allocation coefficients are not causal effects of hiring. Total and average service weights are linked as I=A*K and J=A*Q; independent integration and optimisation verify the eighteen-year break-even coefficient 638.9660, reported as 638.97.

The held engineering-life manufacturing analysis uses 317 observations from 32 firms in nine hosts, with independent BEA production-machinery lives and an identical 299-observation duration comparison. The rotation contrast remains -110.25 (95 percent interval -520.28 to 299.78; positive-direction p=0.724). Joint positive service pricing and rotation are unestablished (p=0.763); familywise contradiction is also unestablished (p=0.712). These are observational allocations; intended plant life and delivered recruitment are not observed. No estimate, uncertainty interval or probability is changed by this final revision.

The documented twenty-year JLR host appraisal supplies a EUR106-226 million strategic hurdle after disclosed delay and aid adjustments. Its recruitment component is unobserved. The five-percent continuous-discount reference annuity is EUR8.38-17.88 million per year; JLR's internal rate is redacted. The article preserves other strategic factors and excludes double counting labour already priced in financial cash flows.

The manuscript contains 9,992 counted words, a 180-word abstract and 23 full substantive text pages before references. The main PDF has 41 pages and six figures; the appendix has 126 pages. Regression tables retain three-decimal probabilities, no significance stars and explicit economic magnitudes. All earlier estimates, proofs, figures and the complete appendix are retained.

Start with README.txt. Current manuscript, appendix and complete LaTeX are directly downloadable. Three verified current ZIP blocks, their restoration helper and nine frozen complementary base archives reconstruct the exact 1,858-file replication tree, verified by 1,857 checksum entries. Public aggregate inference reconstructs without WRDS; full reestimation uses saved licensed inputs obtained by the supplied subscriber queries. No new data were acquired. Licensed issuer rows are excluded. Third-party source rights remain; journal articles and the Lyons dissertation are linked rather than redistributed.

The exact declaration is: ChatGPT and Codex (OpenAI) were used in October 2026 for coding help, literature search, and red teaming and review. Reviewer files exclude author identifiers; public metadata supplies attribution. This public deposit is not a journal submission or acceptance.