Data and Code for "Auditing the Robustness of Corporate Carbon-Performance Claims: A Source-Traceable Accountability Study of Semiconductor Foundries"
收藏资源简介:
This dataset and reproducibility package supports a source-traceable audit of corporate carbon-performance claims using longitudinal evidence from Taiwan Semiconductor Manufacturing Company (TSMC) and United Microelectronics Corporation (UMC) over 2015–2024. The package contains a harmonized company-year panel, page-level source-evidence registers, historical-restatement and input-correction records, physical- and revenue-based carbon-intensity indicators, Tapio decoupling outputs, claim-stability assessments, disclosure-consistent-window analyses, location-based and market-based Scope 2 sensitivity results, additive logarithmic mean Divisia index (LMDI) decomposition outputs, figure-source data, and executable Python analysis code. The strict dual-method Scope 2 comparison is limited to TSMC during 2019–2024. The analysis evaluates whether sixteen analyst-specified carbon-performance claims remain Stable, Conditionally stable, Sensitive, or Not assessable under source-supported measurement and reporting alternatives. Particular attention is given to activity-denominator choice, historical restatements and input corrections, disclosure-consistent analytical windows, and Scope 2 accounting basis. The results are descriptive rather than causal, and the two firms are not treated as representative of the semiconductor industry. Version 2.2 adds reproducible activity guard-band checks at 0%, 0.5%, 1%, and 2%, including 144 annual detail records, eight summary records, and a matching supplementary worksheet. The selected observations, baseline analytical results, and sixteen claim-stability classifications remain unchanged from Version 2.1. The underlying observations were derived from publicly available annual reports, sustainability reports, Form 20-F filings, and climate-related disclosures issued by TSMC and UMC. Original corporate PDF reports are not redistributed; source references and page-level locators are provided for traceability. Missing accounting variants are not imputed. The ZIP contains the analysis code, data, documentation, and supporting outputs. A separate XLSX file provides the spreadsheet companion.



