Replication data and code: Does Graduate Unemployment Reveal a Skills Problem? Evidence from Bangladesh

Replication data and code for Does Graduate Unemployment Reveal a Skills Problem? Evidence from Bangladesh. Version 2026-11-01 corresponds to the current article: 26-page manuscript, 14-page online appendix, eight publication figures and five tables. The archive includes four original quarterly Bangladesh Labour Force Survey 2024 person files, frozen public survey tables and ILOSTAT inputs, public 2022 Enterprise Surveys aggregates, a public 2024 American Community Survey Bangladesh-born extract, analysis scripts and documentation. It reproduces completed-qualification and age/sex comparisons, available-non-searcher accounting, observation-quality sensitivity, demographic standardization, occupational overeducation and the article's employer/migration context. Additional historical and wider contextual analyses are labelled ancillary rather than pooled into one population. The full original-input pipeline passed on 29 October 2026. Source-bound checks were repeated after the BBS publication-year and repository citation updates: 229 prose checks, 431 classified numeral occurrences and 181 result-table cells. Source changes in this release do not change empirical estimates. Restricted Enterprise Surveys/FAT establishment files obtained for optional independent checks are excluded. The separate historical hiring/AI study is not part of this article release. This record archives replication materials and does not claim journal acceptance or peer review. Author-created code and documentation are available under CC BY 4.0; third-party datasets, reports and extracts retain their original terms and attribution requirements. The Other (Attribution) licence identifies this mixed-source package. See RIGHTS_AND_ATTRIBUTION.md. The complete ZIP is provided in five numbered parts. The supplied assemble_replication.py downloads and verifies the parts and reconstructs the exact full editor ZIP. Current editable manuscript sources and figures are also supplied separately. The abstract is 132 words; the estimates and analytical scripts are unchanged by this editorial revision. The final editorial revision responds to coauthor feedback by connecting the objective, findings, existing measurement literature, contribution and policy implications in the abstract and conclusion. The 132-word abstract includes two existing author-year references. All empirical estimates remain unchanged.