Methodology
Introduction
This document details the methodology of CountryRisk.io’s framework for determining the Shadow Rating. We use four different models to determine the Shadow Rating , which indicates the possible range of sovereign credit ratings for foreign currency debt obligations based on a set of economic indicators. The Shadow Rating is a purely quantitative assessment that excludes qualitative indicators or adjustments.
Applications
The Shadow Rating has two main applications:
- Identify credit rating pressure: The Shadow Rating helps determine whether a country’s macro fundamentals suggest a forthcoming upgrade or downgrade from the Big Three rating agencies.
- Proxy ratings for non-rated sovereigns: Where a country does not receive a rating from one of the Big Three agencies, the Shadow Rating offers an indication of how that country would be rated based on its fundamentals.
Methodology
Every model has its own strengths and weaknesses, which means that no single model can fully capture sovereign risk. So, we base the Shadow Rating on the results of four statistical models. By aggregating multiple models, we get a better understanding of the underlying sovereign risk and reduce the dangers that come with over-reliance and over-confidence in one model.
Currently, we base the Shadow Rating on the following four models:
- CountryRisk.io’s Sovereign Risk Score applies a scoring-based approach to traditional sovereign risk indicators. You can learn more about the Sovereign Risk methodology and its indicators here .
- CountryRisk.io’s ESG Risk Score applies a scoring-based approach to traditional sovereign risk relevant indicators, along with social and environmental aspects that are relevant to sovereign risk. You can learn more about the ESG methodology and its indicators here.
- The Regression Tree Model is part of the model family of decision-tree learning and is widely used in the context of classification. The purpose of a regression tree is to classify a variable (i.e. sovereign rating) based on several explanatory variables (e.g. GDP per capita).
- The Multi-nominal Model predicts the probability that a variable (i.e. sovereign ratings) belongs to a certain class (e.g. AAA) based on a set of explanatory variables (e.g. GDP per capita).
We present the results of each model using the standard letter ratings from AAA to C.
Data
All four models take the average sovereign rating of DBRS, Fitch, Moody’s and S&P as the target variable. Explanatory variables for the four models are:
| Risk Section | Sovereign Risk Score | ESG Sovereign Risk Score | Regression Tree Model | Multi-Nominal Model |
|---|---|---|---|---|
| Economic growth prospects | GDP per capita Real GDP growth (5 year average) Real GDP volatility (5 year window) Gross national savings (% GDP) Trade openness Research and development expenditures (% GDP) Researchers in R&D (per million people) Unemployment rate Youth unemployment rate Labour force participation rate | GDP per capita Real GDP growth (5 year average) Real GDP volatility (5 year window) Gross national savings (% GDP) Trade openness Research and development expenditures (% GDP) Researchers in R&D (per million people) Unemployment rate Youth unemployment rate Labour force participation rate | GDP per capita Real GDP growth (5 year average) Real GDP volatility (5 year window) | GDP per capita Real GDP growth (5 year average) Real GDP volatility (5 year window) |
| Institutions and governance | Rule of law Control of corruption Government effectiveness Regulatory quality Voice and accountability Political stability Level of statistical quality | Rule of law Control of corruption Government effectiveness Regulatory quality Voice and accountability Political stability Level of statistical quality | Rule of law Control of corruption | Rule of law Control of corruption |
| Monetary stability | Inflation rate (5 year average) Inflation volatility (5 year window) Change of domestic credit to GDP ratio (5 year window) Real interest rate | Inflation rate (5 year average) Inflation volatility (5 year window) Change of domestic credit to GDP ratio (5 year window) Real interest rate | ||
| Fiscal solvency and public debt | General government debt to GDP Public external debt to GDP Public external debt to total external debt Revenue efficiency | General government debt to GDP Public external debt to GDP Public external debt to total external debt Revenue efficiency | General government debt to GDP Revenue efficiency | General government debt to GDP Revenue efficiency |
| Sovereign liquidity | Fiscal balance (% of GDP) Current account balance (% of GDP) Export growth (5 year average) Interest payments to tax revenues Debt service (% of exports) | Fiscal balance (% of GDP) Current account balance (% of GDP) Export growth (5 year average) Interest payments to tax revenues Debt service (% of exports) | Fiscal balance (% of GDP) Current account balance (% of GDP) | Fiscal balance (% of GDP) Current account balance (% of GDP) |
| External debt sustainability | Net external debt (% of GDP) Net external debt (% of exports) Short-term external debt to FX reserves Import coverage (in months) External financing requirements IMF reserves adequacy ratio Short-term external debt to total external debt Foreign currency denominated external debt to total external debt | Net external debt (% of GDP) Net external debt (% of exports) Short-term external debt to FX reserves Import coverage (in months) External financing requirements IMF reserves adequacy ratio Short-term external debt to total external debt Foreign currency denominated external debt to total external debt | ||
| Private sector strength | Non-performing loans to total loans Regulatory capital to risk-weighted assets Return on equity Liquid assets to short-term liabilities Household debt to GDP | Non-performing loans to total loans Regulatory capital to risk-weighted assets Return on equity Liquid assets to short-term liabilities Household debt to GDP | ||
| Climate change and renewable energy | CO2 emissions Renewable energy consumption Renewable electricity output Access to clean fuels and technologies for cooking Emissions of carbon dioxide per unit of GDP ND-GAIN Index ND-GAIN Vulnerability Index | |||
| Biodiversity | Protected areas Air pollution Deforestation | |||
| Education | Completion rates Attainment rates Enrolment ratios Pupil-to-teacher ratios Literacy rates | |||
| Health, food insecurity and poverty | INFORM Risk Index Lack of coping capacity index Life expectancy Net migration Quality of health care system Intentional homicides Mortality rates Immunisation rates Public health expenditures Poverty ratio Access to sanitation services | |||
| Labour market, social safety nets and equality | Income equality Unemployment rates Proportion of seats held by women in parliament Individuals using the internet Account ownership at a financial institution Age dependency rate Proportion of unemployed receiving beneftis Coverage of social safety programs Unsentenced detainees as a proportion of overall prision population |
As the table above shows, the Regression Tree and Multi-nominal Models incorporate a much smaller number of indicators than the Risk Score Models. This is because the scoring-based Risk Score Models can handle missing data better than the other statistical models, which require a balanced sample to estimate the model. All models use annual data from 1990 through present (or latest available data).
Country coverage
The underlying geographic universe covers 190 countries and territories. However, country coverage is different across indicators. Besides country coverage, we selected the indicators on the basis of other criteria, such as:
- Available history: Is there a long history of regular updates? This allows us to assess whether an indicator is too volatile.
- Reporting lag/latest datapoint: When was the index last updated? Will it be updated again in the future?
- Methodology changes: Is the methodology used for calculating the index revised regularly? Frequent and significant changes lead to a lack of comparability over time, while modest changes suggest that the indicator continues to be developed to reflect a changing environment.
- Basis of indicator: Is the indicator based on original (survey) data, or is it a composite of other indicators?
| Share of Available Indicators | < 20% | 20% < 40% | 40% < 60% | 60% < 80% | 80% < 100% |
|---|---|---|---|---|---|
| Data Quality Classification | Very Poor | Poor | Medium | Good | Very Good |
As part of the Shadow Rating calculation, CountryRisk.io also provides a quantitative measure of data quality for each country. We base the data quality measure on the number of available indicators for each country divided by the total number of indicators included in the model. The mapping table between the share of available indicators and data quality is shown above. This only applies to the scoring-based models.
Governance process
- Update frequency: We update our Sovereign Risk Index and publish it on the CountryRisk.io Insights Platform towards the end of each month. In addition, we update the data on an ad hoc basis whenever substantial new information becomes available.
- Model review and adjustments: CountryRisk.io strives to continuously improve its methodology, such as by incorporating new high-quality indicators as and when they become available. CountryRisk.io also consults external experts to review the model and any adjustments we make to it. We will reflect any changes in future versions of this methodology document.