ESG disclosure quality predicts emissions action — or its absence

Five independent claude-sonnet-4-6 calls, one per persona — Skeptic, Fact-Checker, Researcher, Compliance-Guard, Falsification-Auditor. Each writes its hardest objection from its own seat, paired with the methodological resolution it would accept. Run on the static catalogue spec Jun 3, 2026; a live council for any topic is at /try.

1. Skeptic #01
   Raised

   The observed correlation between ESG disclosure quality and emissions reduction is almost certainly confounded by Science Based Targets initiative (SBTi) commitment status acting as a common cause upstream of both variables. Companies that formally enroll in SBTi are contractually required to produce CDP-aligned, scope-3-disaggregated reporting—mechanically inflating their LLM-assessed specificity scores—while simultaneously facing reputational and contractual pressure to hit validated reduction milestones, driving the downstream emissions delta. This means any Spearman ρ > 0.25 between disclosure quality and emissions trajectory could reflect SBTi selection entirely, with disclosure quality carrying zero independent causal weight on actual decarbonization behavior.

   Resolved

   Cross-reference every issuer in the sample against the SBTi public company database (sciencebasedtargets.org/companies, matched by LEI or ISIN) to generate a binary `SBTi_committed` indicator, then rerun the primary analysis as an OLS regression of 24-month emissions delta on disclosure-quality decile while including `SBTi_committed`, log(market cap), and GICS 2-digit sector fixed effects as covariates. If the coefficient on disclosure-quality decile shrinks to |t| < 1.96 or collapses by more than 60% of its unadjusted magnitude after `SBTi_committed` enters the model, the hypothesis is falsified—the mechanism is target commitment, not filing quality. A Sobel-Baron-Kenny mediation decomposition using `SBTi_committed` as mediator would further quantify what share of the raw correlation is attribution-safe for the disclosure-quality channel.

2. Fact-Checker #02
   Raised

   Scope-3 emissions—the primary outcome variable—carry a GHG Protocol estimation uncertainty of ±30% to ±200% depending on category (Category 1 purchased goods vs. Category 11 use-of-sold-products), meaning the 5% reduction threshold for top-decile filers is indistinguishable from methodological noise; a company switching from spend-based to activity-based scope-3 accounting can generate apparent 20–50% "reductions" with zero real-world change. LLM-assessed specificity scores lack calibrated inter-rater reliability: cross-provider ESG rating divergence studies (Berg et al. 2022) report pairwise correlations of only ~0.54 between major raters, implying a scoring reliability coefficient r_xx ≈ 0.50; combined with a scope-3 data reliability of r_yy ≈ 0.60 for partially-verified CDP submissions, classical attenuation bias

   Resolved

3. Researcher #03
   Raised

   The dominant uncontrolled confounder is prior adoption of Science-Based Targets (SBTi commitments), which operates through a dual channel that mechanically inflates both sides of the correlation: SBTi membership obligates firms to produce scope-3-inclusive, quantified, audit-ready disclosures — directly raising their LLM-assessed specificity score — while simultaneously requiring verified emissions reductions under threat of public delisting, generating the very emissions trajectory the hypothesis attributes to disclosure quality. Because both the treatment variable (high filing specificity) and the outcome variable (24-month emissions decline) are jointly downstream of a pre-existing target-framework commitment, the estimated Spearman ρ is upward-biased by an omitted common cause rather than any causal link running from disclosure quality to operational action.

   Resolved

   The issuer panel should be merged with the SBTi public company database (downloadable at sciencebasedtargets.org/companies-taking-action, which records firm name, ISIN, commitment date, target type, validation status, and sector) and SBTi membership vintage should be included as a binary covariate — or, more rigorously, the primary Spearman correlation should be re-estimated within the subsample of issuers holding no active SBTi or CDP-verified equivalent commitment at the start of the 24-month measurement window. If the disclosure quality–emissions delta correlation retains ρ > 0.25 and the top-decile magnitude threshold (≥5% reduction) within this non-committed subsample, the hypothesis survives the confounder; if the correlation collapses toward the falsification threshold of |ρ| < 0.05 in that restricted panel, the original result is attributable to target-framework selection rather than disclosure quality per se.

4. Compliance-Guard #04
   Raised

   Before this hypothesis crosses its declared SUPPORTS threshold, citing the correlation as established would expose downstream users to liability under SEC Rule 10b-5 and Section 17(a) of the Securities Act, because the binary "greenwashing" label applied to bottom-quartile filers constitutes an implied assertion of material misrepresentation — a characterization that, if acted upon by investors or plaintiff counsel, creates a defamatory or market-manipulative instrument grounded in an unvalidated LLM-scored proxy metric. Simultaneously, premature use of the predicted CSRD Article 8 audit-failure correlation in any audit, rating, or enforcement referral context would misapply the CSRD's third-party assurance standard (ESRS General Requirements, paragraphs 114–120) by substituting an unvalidated computational score for the limited-assurance engagement required of accredited statutory auditors. The SEC's climate-disclosure rulemaking (Release No. 33-11275, currently under judicial stay) and IFRS S2 paragraph 64 both condition scope-3 materiality findings on issuer-specific verification, not sectoral correlation inference; citing this hypothesis as if it satisfies either standard before the ρ > 0.25 / N ≥ 500 gate is reached would constitute a material misstatement of the evidentiary basis in any regulatory submission, litigation filing, or ESG rating product.

   Resolved

   No issuer-level characterization — including but not limited to "greenwashing," elevated litigation-risk flag, or CSRD audit-failure prediction — may be published or transmitted to any third party until the Spearman rank correlation has been independently replicated on a hold-out sample (minimum N = 200 issuers outside the training set) using CDP-verified Scope 3 actuals as the emissions ground truth, and the LLM-based specificity scores have been cross-validated against at least one human-expert blind-coded subsample of ≥ 50 filings with inter-rater agreement κ ≥ 0.70. All interim outputs must carry the mandatory disclaimer: "This analysis reflects a falsifiable research hypothesis that has not yet met its pre-registered evidentiary threshold; it does not constitute a finding of material misstatement, regulatory non-compliance, or greenwashing under SEC Rule 10b-5, CSRD Article 8, or IFRS S2, and must not be relied upon as such in litigation, audit, or investment decisions." Formal standard adoption — including any mapping to SBTi Net-Zero Standard sectoral pathways or SEC climate-rule scope-3 materiality thresholds — requires separate validation by a qualified third-party assurance provider before the correlation output may be cited in a regulatory or fiduciary context.

5. Falsification-Auditor #05
   Raised

   The principal threat to falsifiability is systematic sector-composition confounding inflating the observed Spearman ρ above the FALSIFIES threshold of |ρ| < 0.05 even under a true null of no causal link between disclosure quality and emissions

   Resolved

Unlike the static stress tests above (synthesised against the frozen catalogue spec), this is what a 3-voice council found in the most recent biweekly review. Refreshed on the 1st and 15th of each month at 09:00 UTC. Each voice runs one bounded web search via Anthropic's web_search_20260209 tool, cites what it finds, and recommends a verdict. The verdict aligns with the curated catalogue status (forming).

1. forming

   May 30, 2026 · added to catalogue at status "forming"

• ClientEarth and U.S. climate-plaintiff firms (Sher Edling, Hagens Berman) concentrate ≥65% of new securities-fraud climate filings on bottom-quartile ESG-disclosure issuers, as classified by the SEC's structured-data climate scorecard, within 24 months of final rule enforcement.
• MSCI ESG Ratings and Sustainalytics face at least 5 formal methodology-review proceedings under ESMA's ESG Ratings Regulation (EU 2023/2631) for systematic scope-3 over-scoring of low-quality filers, initiated by 2028.
• S&P Global Ratings embeds ESG-disclosure-quality scores into its ESG credit indicators, widening investment-grade corporate spreads for bottom-decile filers by 20–40 bps relative to top-quartile counterparts within two annual rating-review cycles by end-2027.
• Engine No. 1 and Calvert Research and Management adopt bottom-quartile disclosure-quality flags as primary proxy-targeting criteria, directing ≥$15 billion in shareholder-resolution pressure at low-scoring Scope 3 reporters during the 2027 proxy season.
• FTSE Russell removes bottom-decile ESG-disclosure-quality issuers from the FTSE4Good Global index at its December 2027 annual rebalance, triggering passive-fund divestiture of $10–18 billion in affected equities within 90 days of the rebalance date.