Tellimer's proprietary Sovereign Probability of Default model
5 million data points spanning 20+ years
Tellimer’s probability of default model draws on more than 5 million data points spanning over 20 years to provide a single intuitive score, updated monthly, that reflects the likelihood of sovereign default in the next 12 months. Designed specifically for emerging and frontier markets, the model blends macroeconomic, political, and market signals to capture both the slow-moving structural risks and the fast-moving shocks that precede debt crises
It includes core liquidity and solvency metrics like debt-to-GDP and reserves-to-imports ratios.
It accounts for bond maturity profiles, oil price impacts on trade balances, and domestic political risks—to name just a few.
It also integrates economic surprise indicators, such as missed targets, sharp forecast revisions, and unexpected data outturns.
Unlike traditional ratings or databases
Unlike traditional ratings or databases, the model produces a monthly updated probability score, capturing developments in near real-time.
Built using our point-in-time (PIT) historical database, our model shows what the PD would have been at each moment in time. This makes it ideal for backtesting and understanding how risk has evolved.
The best of both worlds
Our PD model is built in two tiers, combining the best of traditional economic modelling and modern machine learning:
Tier 1 – (Logistic Regression):Anchored in established economic theory, Tier 1 captures the build-up of sovereign vulnerabilities using a transparent and interpretable logistic regression framework. Every variable has a clear, traceable coefficient - so you always know what’s driving risk higher or lower.
Tier 2 – (LightGBM):Tier 2 captures faster-moving accelerants, such as inflation surprises, political unrest, or unexpected data outturns. Built using LightGBM, this layer boosts the model’s responsiveness to shocks while preserving the stability provided by the Tier 1 structural signal.
Unlike traditional ratings or databases
Together, the two tiers create a model that is:
Interpretable when you need to explain the risk,
Reactive when risk evolves fast,
And robust across regions, timeframes, and use cases.
With coverage of 84 emerging and frontier market sovereigns, our monthly PD estimates make it easy to compare risk across countries, regions, and time.