This study aims to diagnose where public climate signals remain visible across issuer grades, modal debt grades and bond-grade record distributions in China.
An archival universe of 1,468 interbank medium-term notes issued by 165 listed legal obligors yields 962 bond-grade records and modal debt grades for 149 issuers, of whom 105 also have issuer grades. Rating-availability diagnostics precede paired concentration, multiplicity-controlled sorting, measurement sensitivities, repeated imbalance-aware classification, ordered and multinomial logit and full-notch prediction.
These are descriptive associations within a static archival snapshot, not evidence of structural or dynamic transmission. Issuer Herfindahl–Hirschman index (HHI) is 0.689 vs 0.604 for paired modal debt grades; issuer-balanced bond HHI is 0.488, while raw bond-count HHI is 0.741 with an 85.4% AAA share. The broad national-emissions-trading-system sector-policy proxy is the strongest full-sample signal in the pre-declared three-proxy family, but significance does not survive a conservative definition. Weather sorting disappears after excluding Beijing, and disclosure separation is driven by transition mentions. Binary, ordinal, multinomial, notch and imbalance-aware diagnostics show no stable positive predictive increment.
The selected archival cross section has unmatched information dates, no verified rating-action history and no facility geocodes.
Credit systems should link obligors to facilities, emitters, guarantors and securities and preserve transparent overlays alongside stable grades.
The paper separates issuer compression, debt-grade resolution, issuance-frequency re-compression and proxy-measurement limits.
