The Identification Ceiling: ESG Rules and Bank Soundness at Country Level

¶ Archived manuscript for The Identification Ceiling: ESG Rules and Bank Soundness at Country Level. ABSTRACT ¶ [manuscript ka asli 247-word abstract] WHAT THE PAPER FINDS ¶ A recorded policy year does not recover the legal effect, the covered banks, or the first filing. Official-source review verifies regime-level first-report timing for 17 of 38 records, and 330 of 356 extracted candidate texts remain unreviewed. The treatment variable that the literature uses is therefore a publication date, not an identified legal exposure. ¶ No static association survives six-outcome Holm correction. The accounting funding-cost proxy changes sign between raw and winsorised specifications, and the observed-standard-error minimum detectable effect is 2.8 times the point estimate. The panel has limited power for effects near the estimated magnitude. ¶ Pseudo-adoption intervals exclude zero for four of six outcomes, so the design fails its own falsification test. The ceiling then follows formally: under observed standard errors a country-level design needs roughly 1,455 treated systems to detect a 0.05-standard-deviation effect, almost ten times the 149-system universe. Minimising over every treatment–control split of that universe leaves a hard floor of 0.287 standard deviations. ¶ The bound tightens under clustering and multiplicity conventions and does not depend on the paper's own point estimates. It is a statement about the design, not about the sign or size of any effect. WHAT THE PAPER PROPOSES ¶ The policy-facing output is a sequenced data agenda rather than a list of data sources. Priorities are ordered by what a bank-level design needs first, not by ease of collection: rule chronology and covered-entity scope; bank obligation, filing, assurance, and enforcement; direct funding prices with matched funding bases; harmonised banking outcomes with metadata and vintages; and verified comparison-policy histories. ¶ The same 1,455-unit requirement that is unreachable across 149 countries is routine at bank level, where a single large jurisdiction contains thousands of institutions inside the scope of a reporting mandate. The binding constraint is linked records, not estimator choice. DATA ¶ 149 banking systems, 2,270 country-years, 2000–2024. Sources are IMF Financial Soundness Indicators via the DBnomics API, World Bank World Development Indicators via the World Bank API, and the KSTZ (2023) mandatory-disclosure policy workbook from its Open Science Framework deposit. ¶ No private or licenced data are used. Six banking outcomes are analysed separately rather than combined into an index. REPRODUCIBILITY ¶ The replication package accompanying the manuscript holds the raw API responses, a SHA-256 manifest with retrieval timestamps, and the acquisition and estimation scripts. Every input is public and every transformation is archived rather than reported from a private extract. ¶ Two components do not reproduce at the reported magnitude from public sources. The manuscript states both explicitly in Section 7, with the direction of the discrepancy, rather than leaving them for a reader to discover. The paper's conclusions do not depend on either. KEYWORDS ¶ sustainability reporting; banks; financial soundness; funding costs; disclosure regulation; difference-in-differences; staggered adoption ¶ JEL codes: G21, G28, M48 LICENCE ¶ CC BY 4.0. CITE AS ¶ Finallen (2026). The Identification Ceiling: ESG Rules and Bank Soundness at Country Level. Zenodo. https://doi.org/10.5281/zenodo.23910721