Credit Consensus Data: The Credit Benchmark Dataset
An independent consensus view of credit risk, built from the 1-year probability of default estimates contributed by 40+ of the world’s leading banks — the foundational dataset behind every Credit Benchmark analytic and delivery channel.
The foundation of
the platform
The Credit Risk Dataset is Credit Benchmark’s primary asset. Over 40 contributing banks submit the 1-year Probability of Default (PD) estimates their own credit teams produce — obligor-level estimates on senior unsecured exposure in the wholesale and commercial book, linked to each bank’s internal rating system. Credit Benchmark aggregates those contributions into a single consensus view of credit risk for each entity. Every analytic and every delivery channel elsewhere in the platform is derived from this dataset.
This differs structurally from a traditional credit rating. The consensus is an average of the views of banks with real lending relationships and genuine exposure to the names they assess — “skin in the game” — rather than a single issuer-paid opinion. Individual bank views remain completely confidential, and because the network collectively covers far more of the market than any agency, the dataset reaches deep into entities that carry no public rating. The consensus is published weekly.
Why the consensus model matters
Five characteristics distinguish the dataset. It reflects the credit risk views of institutions with skin in the game, not a single issuer-paid opinion. Bank lending reaches all corners of the market, including obligors that have traditionally been unrated. It is independent and anonymized, with bank risk teams acting as a private-side function that is walled off from other public markets activities at the banks. It is based on banks’ one-year probability of default estimates, providing a more stable view of credit risk and helping to reduce procyclicality. And bank’s refresh their internal views at different cadencies reflecting either regulatory enforced refreshed or ongoing deal flow.
How the
consensus
is built
- Contribution.
40+ banks submit obligor-level 1-year PD estimates linked to their internal ratings, covering Corporates, Financials, Funds and Governments. Scope is senior unsecured exposure in the wholesale/commercial book; retail and facility-level PDs are out of scope. Banks deliver at least monthly via browser-based CB Secure or automated SFTP. - Entity resolution.
Submissions are matched to a single Credit Benchmark entity (a unique CBId) before any calculation, by a dedicated team of specialists who prioritize identifiers such as LEI and ticker and cross-check commercial and public reference data. - Validation.
Contributed PDs pass automated and manual checks, including outlier detection, cross-bank checks and trend analysis, before entering the consensus. - Aggregation.
The Consensus PD is the simple unweighted average of the contributed PDs. Credit Benchmark publishes a consensus only where at least two banks contribute. - Mapping.
The Consensus PD is mapped to the Credit Benchmark 21-category rating scale (aaa, highest quality, through to d, default) to produce the Consensus Credit Rating, with a finer 101-bucket CCR100 scale available for granular work.
Core dataset components
Credit Consensus Ratings
What it is: An independent letter rating (the Consensus Credit Rating, or CCR) for a legal entity, derived from the averaged PD estimates of multiple contributing banks — a single, market-based reference point for the entity’s creditworthiness.
How it is derived: Contributed 1-year PDs are averaged into a Consensus PD, which is mapped to the 21-category scale (aaa to d). A rating is published only where at least two banks contribute.
What is included: The CCR on the 21-category scale; the consensus PD average; the finer CCR100 value; an investment-grade / high-yield flag; the published contributor count (shown as “MIN” below five contributors, “5+” above); and the CCR Source — “Consensus” where three or more banks contribute, or “Implied” where exactly two do (a non-distorting obfuscation point protects the pool in that case).
How it is used: Independent validation and calibration of internal models; customer onboarding and credit decisioning; portfolio monitoring; pricing and valuation; regulatory and third-party risk validation; and entity reference mapping.
Delivery & documentation: API, Web App, Excel Add-In and SFTP, plus Snowflake, Databricks, AWS, Bloomberg and Preqin.
Probability of Default Data
What it is: The consensus PD that underpins every CCR — credit risk expressed as a 1-year default probability rather than a letter grade.
How it is derived: A simple unweighted average of the contributed 1-year PDs. The same average is also mapped to the CCR100 scale: indexed 1–100, with a default category, that preserves more of the underlying signal than the 21 letter grades.
What is included: The Consensus PD average; the CCR100 midpoint PD; the 21-category midpoint PD; and the published PD boundaries for each rating category (expressed in basis points — for example, bbb spans 20–30 bps).
How it is used: Ranking entities that sit within the same letter grade; detecting small movements that do not cross a rating threshold; and feeding quantitative models — PD calibration and validation, CVA and pricing, and RWA work — with a continuous input.
Delivery & documentation: API, Web App, Excel Add-In, SFTP and the cloud/marketplace channels.
Historical Time Series
What it is: The published history of consensus ratings and PD for each entity and aggregate, with full month time stamps; weekly refreshes of the ‘latest’ month provide currency.
How it is derived: Each weekly publication is retained and linked into a continuous monthly time series at both the entity and aggregate level.
What is included: Consensus PD, CCR, CCR100 and the distribution fields over time, plus rating-change fields measured over 1-, 3-, 6-, 9- and 12-month windows. Credit Benchmark has published consensus data since May 2015 — over 10 years of monthly history.
How it is used: Back-testing and model validation; trend and migration analysis; constructing transition matrices; and scenario and stress calibration.
Delivery & documentation: Bulk and historical pulls via API, SFTP and the cloud channels; recent history via the Web App and Excel Add-In.
Consensus quality
and movement
Trend Analysis
Movement in the consensus over time. Published as the Opinion Change Indicator — Improving, Deteriorating or Stable, based on the month-on-month movement of the underlying contributions — alongside rating-change fields over 1-, 3-, 6-, 9- and 12-month windows. The Opinion Change Indicator is designed to distinguish a genuine shift in bank opinion from a move caused only by a change in which banks contributed.
Dispersion Analysis
The spread and shape of contributed views for an entity — how much the contributing banks agree. Published as the relative standard deviation of contributed PDs (CCRRSD), a skew measure (CCRSkew), an Agreement Indicator (High, Medium or Low) and an Outlier Indicator (Optimistic, Balanced or Pessimistic), together with the highest and lowest ratings in the distribution (CCR Max and CCR Min). Tight dispersion signals strong conviction in the consensus; wide dispersion flags names where bank views are divided.
Point-in-Time Consensus
Credit Benchmark’s core consensus is based on banks’ one-year probability of default estimates. The Point-in-Time (PIT) Consensus applies the same aggregation model to a different, forward-looking risk perspective: contributing banks submit full point-in-time PD term structures — credit risk as assessed today and projected forward — through a separate submission stream that feeds its own consensus. It is the dataset behind Credit Benchmark’s Impairment Benchmarking Service for IFRS 9 and CECL.
What it is
A separate consensus dataset of forward-looking PD term structures, built from contributed point-in-time estimates and available to the banks that contribute PIT data. It runs out to 10 years and is published for both a baseline and a scenario-weighted view.
How it is derived
A distinct contribution stream run through the same aggregation logic as the core consensus. Each bank’s PIT submission is linked to its existing submission via the same internal-ID mapping, so consensus measures resolve to a single Credit Benchmark entity. A consensus is published where at least three banks contribute to an entity.
What is included
Consensus PD term structures from year 1 to year 10, on both cumulative and marginal curves and for both baseline and scenario-weighted views — at single-name (entity) level and aggregated by entity type, industry, super-sector, sector and sub-sector. Each aggregate is published for all contributors, for peers excluding your own submission, and for your own data, alongside dispersion and conservative/aggressive positioning so a bank can see exactly where it sits.
How it is used
IFRS 9 and CECL impairment benchmarking — comparing your PD term structures, staging assumptions and ECL drivers against peers; model calibration, monitoring and validation; assessing the impact of economic scenarios and their weights; stress testing; and audit support. It also enables direct TTC-versus-PIT and 1-year-versus-curve comparison.
Delivery
Published quarterly as a set of data files plus an accompanying analysis report, delivered through the Web App Inbox or a secure file channel (CB Secure / SFTP).
Point-in-Time Consensus demonstrates the flexibility of the contribution model: the same aggregation engine applied to a different, forward-looking measure of risk. It also enables direct comparison with the core consensus offering.
Metadata & Reference Information
What it is
The entity universe covered by the dataset, plus the descriptive metadata that lets clients map, filter and integrate the data.
How it is derived
Contributed entities are resolved to a single CBId and enriched with Credit Benchmark’s classification and identifiers.
What is included
Entity identifiers — CBId, legal name, LEI, primary-equity ISIN / CUSIP / ticker, and ultimate-parent CBId, name and country; a six-level industry classification (Entity Type → Sub-Type → Super Sector → Industry → Sector → Sub-Sector); a four-level geographic classification (Region Group → Region → Country of risk → Subdivision, with ISO codes); and entity characteristics including a CRA-rated flag and a public/private flag. Entity types span Corporates, Financials, Funds and Governments across global markets.
How it is used
Entity mapping and portfolio matching; coverage assessment; and clean integration into internal systems — the data works alongside a client’s own schema.
Delivery & documentation
Coverage and metadata accompany every delivery channel; large-scale reference mapping is supported through Matching as a Service (see Platform & Delivery).
Build on the dataset
Request a coverage check on your own entity universe, or explore the analytics and delivery options built on the consensus dataset.