The Debt Sustainability Framework for Low-Income Countries (LIC-DSF), developed by the IMF and World Bank, serves as a vital instrument for assessing debt-related risks in vulnerable economies. With many low-income countries experiencing escalating debt pressures — particularly in the wake of the COVID-19 pandemic and higher global interest rates — the importance of the LIC DSF has grown significantly.
This note by Clemens Graf von Luckner aims to achieve two goals: 1. “lifting the hood” of the LIC-DSF to analyze its mechanics for predicting the risk of debt distress; and 2. Identifying the shortcomings of the current model. The first section of the paper explores the complex steps involved in transforming projections (especially of debt indicators such as external public debt in present value to GDP; or debt services to revenues) into a risk measure: how likely is this country to default? The second section reveals that a number of choices made to accomplish this task lack transparency and efficiency. Graf von Luckner proposes a more straightforward procedure that reduces the risks of manipulation and the sharp discontinuities that have plagued recent debt restructurings.
The LIC-DSF uses a statistical model to predict the risk of debt distress, influencing policy decisions and access to financing for low-income countries. While civil society’s attention has largely focused on macro-financial projections, a key component of the DSF is its risk model: assessing when countries are at low, moderate or high risk of debt distress. This risk model relies on drawing lessons from past episodes of distress in order to be able to detect new ones. As a result, what is considered “risky” crucially depends on the way past information is incorporated (Figure 1).
To achieve this, the current LIC-DSF groups together countries with vastly differing economic conditions into the same risk categories. The central index for this classification is called “debt-carrying capacity” (DCC): figure 2 shows how various DCC groups lead to different final assessments of the risk of debt distress. While the framework is relatively transparent overall, and the need for “use of judgment” in assessments is made clear, the DCC often relies on projections that can be manipulated.
Evidence of some manipulation stems from the existence of sharp discontinuities around the thresholds. This finding is not new, and researchers (including from international financial institutions themselves) have documented its existence for a broad range of reasons (see Lang and Presbitero, 2018, and the works cited therein). However, what is new is that despite efforts following the previous review to reduce space for discretionary practices, they seem to persist. The data presented here suggests that meddling with the DSA is still commonplace, even after the last review, whose goal had been to make staff judgment more explicit (Figures 3 and 4).
However, simply predicting the probability of distress is not sufficient: the framework must also provide a rule for countries facing an excessively “high” probability. In the last review, a technical choice was made to emphasize the importance of reducing missed crises relative to “false alarms”. As a result, developing countries may have tighter limits to their ability to invest than what would be warranted under a more balanced framework. International institutions should be more transparent about how those parameters are chosen, as it involves both technical aspects and but policy decisions. Putting more emphasis on avoiding “false alarms” would lead countries – especially those where other variables (such as CPIA or reserves level) are relatively “better positioned”, to borrow and invest more to grow.
The paper explores three proposals to improve the framework: 1. allow automatic model selection instead of relying on human judgment to predict crises; 2. remove the grouping of countries; and 3. achieve better balance in risk weights. In this improved framework, the crisis detection model would be based on a smaller set of variables and be more resilient to structural changes in global financing conditions. Additionally, instead of grouping countries by “debt-carrying capacity” and losing information on each country's probability of default, each country would be assigned a risk assessment for distress based on a small set of structural and debt-related variables. Finally, policy decisions would be based on an assessment of what level of risk is acceptable: specifically, a probability threshold above which the risk of debt distress would be deemed “high”.
These changes aim to create a more refined, transparent, and accurate framework that better serves the needs of low-income countries, all while maintaining fiscal responsibility. With the 2026 review of the LIC-DSF on the horizon, it is crucial for policymakers and civil society to critically assess the framework, ensuring it effectively strikes a balance between development goals and responsible debt management.