Year: 2026
Type: article
Abstract: Scope 3 emissions have become central to climate disclosure, but remain the least reliable component of firms' carbon footprints. This paper argues that the core measurement problem lies not only in the aggregate Scope 3 total, but in the heterogeneous quality of its underlying categories. Because some categories are far more difficult to observe, verify and compare than others, the aggregate figure embeds substantial noise—and, as we show, systematic omission. Using firm‐level disclosures from the CDP (Carbon Disclosure Project) database over 2013–2023, we implement a category‐level framework that first diagnoses, for each of the 15 GHG Protocol categories, whether a firm fully omits, partially under‐reports or soundly discloses it and then estimates the missing magnitude by machine learning. We find that hidden Scope 3 amounts to about 22% of reported emissions on a value‐weighted basis, while under‐reporting is near‐universal across firms—the median firm conceals 64% of its corrected footprint—but concentrated in tonnage among the largest emitters. Omission is not diffuse: it concentrates on use of sold products, the single category that most directly captures a firm's exposure to transition risk and that alone accounts for roughly two‐thirds of hidden emissions. Correcting these disclosures therefore reshapes firms' relative positions in the carbon distribution—44% of declarants change carbon‐intensity quintile—and selectively reveals the most financially material dimension of their carbon exposure. Reported Scope 3 carries a reliability ratio of at most 0.65 with respect to corrected exposure (0.62 within sectors), so carbon‐premium estimates built on it are attenuated by a third or more—distortions that affect ESG scoring, portfolio construction, transition‐risk assessment and the design of disclosure standards.
DOI: 10.1002/ijfe.70330
Language: English
Subfield: Strategy and Management
Domain: Social Sciences
SDG: Climate action
Open Access status: hybrid
Estimated APC paid: $3,400
Award: G7682321522
OpenAlex ID: W7234630517
