Ratings Radar
Tellimer Ratings Radar is a forward-looking model that identifies the direction of sovereign credit ratings over a 12-month horizon. It combines signals from Moody’s, S&P and Fitch — including rating levels, outlooks, recent actions and cross-agency disagreement — with Tellimer’s proprietary sovereign probability of default (PD) model, capturing changes in underlying credit risk that are not yet reflected in agency ratings. These are synthesised into a single cross-country ranking designed to highlight where credit quality is improving or deteriorating ahead of rating agency actions.
Model output and architecture
Each sovereign is assigned an Outlook percentile (0–100), ranking countries from strongest positive to strongest negative credit momentum, and grouped into five categories: Upgrade watch, Positive, Stable, Negative and Downgrade watch.
The model does not produce explicit probabilities. Instead, each percentile is calibrated to historical outcomes, allowing users to observe the historical frequency of upgrades and downgrades across the distribution.
The model consists of separate upgrade and downgrade random forest machine learning models, combined into a single Outlook percentile. It uses 46 features across four groups: rating levels and dispersion, rating momentum, agency outlooks, and Tellimer’s proprietary sovereign PD, its macroeconomic data and its dynamics.
Driver attribution is provided for each signal, highlighting the role of agency signals, cross-agency disagreement and changes in underlying credit risk.
Model performance
The model is trained on data from 2010 to 2021 and evaluated out-of-sample from 2022.
Across 124 historical crossings of the investment grade / high yield (BBB-/BB+) and distressed (B-/CCC+) boundaries, the model signalled the correct direction ahead of the first agency move in 90% of cases, with an average lead time of 17 months.
In out-of-sample analysis:
Sovereigns placed on Upgrade Watch saw their next rating change be an upgrade in 90% of cases
Sovereigns placed on Downgrade Watch saw their next rating change be a downgrade in 80% of cases
Out-of-sample case studies:
Ghana (2022 downgrade cycle): Downgrade Watch signal ~22 months before the first agency downgrade
Bolivia (2023–24 downgrade cycle): Sustained negative signal ~24 months ahead of a staggered multi-agency downgrade sequence
Senegal (2025 distressed downgrade): Negative signal ~24 months ahead; Downgrade Watch ~12 months before crossing into distress
Oman (2024–25 upgrade to investment grade): Upgrade Watch signal ~23 months before the first upgrade
Key capabilities
Early identification of rating direction ahead of agency actions
Cross-country ranking of upgrade and downgrade risk across EM and frontier markets
Monitoring of rating-sensitive thresholds, including IG/HY boundaries
Driver-based interpretation, combining agency signals with proprietary PD dynamics
Systematic complement to discretionary analysis, providing a consistent forward-looking signal