Tehran Stock Exchange Corporate Financial Distress Prediction Dataset (1392–1402): Analysis-Ready Company-Year Panel and Reproducibility Materials
Year: 2026
Type: dataset
Abstract: An analysis-ready company-year panel of Iranian listed non-financial companies for one-year-ahead corporate financial distress prediction, covering Jalali fiscal years 1392–1402, together with the contracts, dictionaries, quality-control records and audit surfaces needed to reproduce and check its construction. Primary modeling surface: data/analysis_ready_main_rule_a_stage125.csv — 1,012 company-year pairs, 119 companies, 115 columns, 80 positive and 932 negative outcomes. The other three analysis-ready files are prespecified robustness surfaces, not alternative headline datasets. Audit surfaces: the four audit/audited_pairs_*.csv files retain company-year pairs that the leakage-safe timing rule excludes. They are audit surfaces and are not all model-ready; fitting a model on them reintroduces look-ahead. Availability is an assumption. Predictor admission uses a prespecified fixed four-Jalali-month regulatory lag. Row-level publication timestamps were never collected and no individual filing date was verified, so the availability date is a proxy, not an observed publication timestamp. Provenance is file-level and incomplete: a source file is recorded for 1,303 of 1,331 upstream rows and absent for 28. Quality-control coverage differs across accounting checks (1,273 to 1,312 evaluable rows depending on the check). What the archive holds: release_manifest.json describes 25 payload files. The archive itself contains 27 members — those 25 payload files plus release_manifest.json and SHA256SUMS.txt, which are integrity records about the payload and are not themselves manifest payload files. Every one of the 115 released columns is documented in RELEASE_COLUMN_DICTIONARY.csv, one row each, with definition, type, unit, role, model eligibility, provenance, temporal reference, missing-value semantics, formula, limitations and the exact committed source each fact came from. The released company-year panel contains researcher-compiled company financial-statement fields transcribed from publicly accessible CODAL disclosures, together with author-derived variables and annotations. No TSETMC-derived and no World Bank-derived field is included in this release; those sources relate only to the wider study. No purchased, confidential, personal or human-participant data were used. Original source PDFs and raw provider responses are not redistributed. The CC BY 4.0 licence covers only the authors' own compilation, annotations and release metadata to the extent the authors hold rights in them; it does not relicense third-party source materials. The authors' rights position rests on a determination by the human author that the source data were publicly and freely accessible and need no separate provider permission — a determination by the author, not a provider licence and not a verification of any provider's published terms. See SOURCE_AND_LICENSE_NOTES.md and LIMITATIONS.md.
Authors: Asghari, Abtin, Mehraein, MohammadMehdi
Language: English
Subfield: Accounting
Domain: Social Sciences
Open Access status: green
OpenAlex ID: W7204033281