IMF Policy Paper: Review of the Debt Sustainability Framework in Low-Income Countries: Proposed Reforms

The Debt Sustainability Framework for Low-income Countries (LIC DSF) has been the cornerstone of assessments of risks to debt sustainability in LICs.


The Debt Sustainability Framework for Low-income Countries (LIC DSF) has been the cornerstone of assessments of risks to debt sustainability in LICs.


1. The framework for debt sustainability analysis for LICs (LIC DSF) was introduced in 2005. The LIC DSF uses the Country Policy and Institutional Assessment (CPIA, an index produced annually by Bank staff) to classify countries based on their debt-carrying capacity, and a set of threshold levels for selected debt burden indicators linked to elevated risk of debt distress that are statistically estimated for varying levels of debt-carrying capacity. Baseline macroeconomic projections and stress test scenarios are evaluated relative to these thresholds, and used in conjunction with staff judgment to assign risk ratings of external debt distress. The value of these risk ratings is, of course, dependent on the realism of medium- to long-term macroeconomic projections, which are critical inputs to the DSF produced by country teams (see Appendix I for an overview of the framework).

2. The DSF has remained the cornerstone of assessments of risks to debt sustainability in LICs, with important operational implications for multilateral institutions and other creditors. The DSF risk assessment of a country plays an important role in the Fund’s debt limits policy (DLP) and the Bank’s non-concessional borrowing policy (NCBP), and also influences the grant-loan funding mix of IDA support for the country. Many MDBs (including the AfDB and IaDB) have also linked their lending policies to the DSF risk assessment. Debtors often use it to inform borrowing decisions, albeit with many expressing concerns that the framework is limiting their opportunities to borrow for development. CSOs have shown increased interest in the LIC DSF and have advocated its use as a tool to help determine how to finance the large development needs of LICs in the context of the efforts to achieve the Sustainable Development Goals (SDGs).

3. The DSF has been reviewed on three occasions, most recently in 2012 (see Box 1 for a summary of the Board discussion). New features were added to the DSF in the 2012 review, to incorporate further country-specific information into the determination of risk ratings and to pay greater attention to domestic debt vulnerabilities in the risk assessment. Some other potential additions—such as embedding macro-linkages in stress tests and enhancing the modeling of the investment-growth nexus—were explored but ultimately left for future work.

4. Since 2005, the economic environment in which many LICs operate has changed significantly, resulting in potentially important gaps in the DSF. The evolving financing landscape and the challenges emerging from a weaker environment have reshaped the nature of risks facing LICs. Financing sources that have increased in importance include borrowing from non-Paris Club creditors, from domestic markets, and from international bond markets—most notably for “frontier” LICs that have attracted foreign portfolio investors. As a result, LICs are increasingly exposed to a wider set of vulnerabilities, including from market volatility.1 The DSF, in its current form, lacks tools to assess these market-related risks.

Board Discussion on the 2012 Review of the Debt Sustainability Framework for Low-Income Countries

Executive Directors of IMF and WB Boards agreed to introduce new features into the LIC DSF to improve its assessment of the risk of debt distress. These features were added to incorporate further country-specific information in the determination of risk ratings. The probability approach was introduced to complement the assessment of the risk of external debt distress in “borderline” cases, where risk ratings arguably fell in between categories. The introduction of remittances-augmented debt thresholds was meant to capture better the enhanced repayment capacity of countries receiving large remittances. The review also proposed strengthening the analysis of total public debt by introducing public debt benchmarks to support the assessment of the overall risk of debt distress. It also called for an in-depth analysis to determine the extent of vulnerabilities emanating from private external debt.

Directors saw merit in developing stress tests that included macro-linkages and looking more closely into the links between investment and growth:

  • Stress tests. There was strong support for exploring stress tests that include macro-linkages. Many Directors wanted the inclusion of such stress tests to be experimental and only optional for country teams. One Director thought such stress tests should be the norm where data allows.

  • Investment-growth nexus. There was broad support for exploring further the links between investment and growth. Several Directors raised the concern that since too little is known about the effect of investment on growth, it is best to be prudent and keep the framework conservative.

Directors unanimously supported simplifying the DSF template. Some Directors cautioned though that it should not come at the expense of adequately capturing risks.

5. In the external consultation process that has supported the review, stakeholders have called for gaps and complexities in the framework to be addressed.2 In particular, external stakeholders have emphasized the need to: (i) ensure the DSF remains balanced in its treatment of risks versus borrowing opportunities; (ii) strengthen the ability of the DSF to provide adequate early warning of potential stress, including by better incorporating relevant country-specific information into the framework; (iii) ensure that the framework is appropriately aligned with the evolving nature of risks facing LICs, including by enhancing the assessment of domestic debt, better accounting for liquidity risks, and improving the assessment of contingent liabilities; (iv) introduce tools to illuminate the realism of macro projections, particularly the investment-growth nexus; (v) expand the stress testing framework to assess key risk scenarios, such as the impact of natural disasters; and (vi) provide more differentiation in characterizing the extent of vulnerabilities for countries assessed to be at moderate risk of debt distress. While calling for these issues to be tackled, stakeholders also emphasized the need to simplify where possible what is already a complex framework.

6. The review has identified a package of reforms that would strengthen the DSF while providing for some simplification of the toolkit. Key reforms include: (i) strengthening the assessment of countries’ debt-carrying capacity by drawing on an expanded set of country-specific and global factors, instead of relying exclusively on the CPIA score; (ii) introducing tools to facilitate closer scrutiny of baseline macroeconomic projections; (iii) recalibrating standardized stress tests to better reflect the actual scale of shocks, while adding tailored scenario stress tests to better evaluate risks stemming from natural disasters, volatile export prices, market-financing shocks, and contingent liability exposures; (iv) providing a richer characterization of debt vulnerabilities (including an enhanced assessment of domestic debt vulnerabilities and new tools for assessing vulnerabilities to shifts in market financing conditions and for better discriminating across countries within the moderate risk category); and (v) enhancing the guidance for a more even-handed application of judgment. At the same time, the framework would be simplified by substantially reducing the number of debt indicators, thresholds, and standardized stress tests.

7. The remainder of this paper is organized as follows. The next section identifies key areas for reform, informed by an assessment of the LIC DSF performance (subsection A) and its methodology (subsection B). The third section proposes reforms to the framework to align it to the evolving nature of risks facing LICs and to address the identified weaknesses, with the aim to make it more comprehensive, more transparent, and simpler. The fourth section assesses the effects of the proposed changes by presenting back-testing results. The fifth section summarizes how the process of engagement around producing a DSA would change and how the Bank and the Fund would support the implementation process. The sixth section examines the appropriate discount rate to use in the DSF. The final part of the paper concludes, summarizing the main takeaways and recommendations from the review, and raises issues for Board discussion.

Identifying Key Areas for Reform

A review of the performance of the DSF and its methodology points to areas where there is scope for improving the framework. On performance, the quality of medium-term debt projections, the calibration of stress test shocks and thresholds, and the integration of country-specific information into the framework could all be enhanced. On methodology, the identification of external debt distress episodes, the specification of the statistical model for predicting debt distress, and the procedure to derive debt thresholds are all areas for improvements. Further, the design of stress tests could be strengthened (by incorporating macroeconomic linkages among key variables) while tools to assess market-related risks could usefully be added.

A. Assessing the LIC DSF Performance

8. To help map out potentially useful reforms to the DSF, it is important to understand first where it is performing well and not so well. The evaluation follows the overall structure of the framework; that is how it utilizes inputs and filters them to arrive at a risk rating. Since the framework starts by comparing baseline projections and stress test scenarios to thresholds, two immediate questions concern how good a job baseline projections and stress tests are doing in characterizing the possible evolution of debt; and how well the thresholds have been set. The evaluation then covers the performance of rules being used to assign risk ratings, including the use of judgment. It concludes by stepping back to bring in case studies and the additional perspective that they can bring.

Performance in Characterizing the Likely Evolution of Debt: some weaknesses

9. While debt projections have been accurate in the near-term, they have tended to underestimate outcomes in the medium-term. Over one and two-year projection horizons, the median (absolute) unexpected change was about 5 percent of GDP for both external and total public debt, with the median deviation more than doubling for both external and total public debt over a five-year horizon (Figure 1. a–b). About 40 percent of DSAs produced during 2007–10 contained unexpected changes in debt over a five-year horizon in excess of 15 percentage points of GDP (Figure 1. c);3 for small states, this occurred in more than half of the DSAs. Deviations on this scale are particularly common in DSAs for countries that are currently at high risk of external debt distress.

Figure 1.
Figure 1.

Performance of the DSF in Anticipating Debt Developments

Citation: Policy Papers 2017, 006; 10.5089/9781498346351.007.A001

Source: Fund staff calculations.

10. Sizable medium-term underestimation of debt outcomes are more common, and seem to be related to particular shocks:

  • The majority of DSAs with sizable medium-term deviations involve an underestimation of debt outcomes. Among external and public DSAs with sizable unexpected changes in debt (larger than 15 percentage points of GDP), about 70 and 80 percent, respectively, underestimate debt outcomes (Figure 1. d).4 This point is clearly captured in the projected dynamics of the median public debt to GDP ratio across LICs for each of the last five DSA vintages (Figure 2); it is noteworthy that projections quickly revert to a downward debt path even after significant short-term upward shifts.

  • Unexpected changes in debt ratios have been primarily driven by fiscal deviations and balance of payments (BOP) shocks. Figure 1. e–f show the decomposition of unexpected changes over a 5-year horizon for the total public debt and external debt-to-GDP ratios, for all DSAs produced during 2007–10 and for those DSAs with sizable positive deviations.

  • ➢ For total public debt, the analysis signals that, in addition to the primary deficit, unanticipated positive residuals have contributed to sizable deviations (suggesting that the materialization of contingent liabilities has played a role). Growth and exchange rate shocks have been less important.

  • ➢ For external debt, the DSA methodology does not allow to disentangle between the contribution of private and public sector elements in the balance of payments. However, the evidence points to a much larger role of financial account flows driving unexpected changes in total external debt. This suggests larger than expected access to sources of finance (a manifestation of the changes in the financing landscape).

Figure 2.
Figure 2.

Long-term Optimism Bias

(Public debt/GDP, median)

Citation: Policy Papers 2017, 006; 10.5089/9781498346351.007.A001

Source: Staff calc Source: Fund staff calculations.

11. The standardized stress tests used in the DSF to characterize risk scenarios have generally not been well aligned with actual drivers of debt changes:

  • A close look at the impact of stress tests on total public and external debt projections shows a mismatch with the impact observed in the data (Figure 3. a). The simulated impact on debt from shocks to the primary balance is modest compared to the actual impact of primary balance deviations. One problem has been the way the shock is calibrated. Post-shock values in stress tests are set at one standard deviation below historical averages, with the key assumption being that historical averages are good predictors for the next 2–3 years. However, this assumption did not hold for the primary balance: actual primary surpluses (deficits) turned out to be consistently lower (larger) than historical averages during 2008–15 (see Annex I).5 In contrast, the simulated impact on the debt-to-GDP ratio from shocks to the exchange rate is large relative to the actual impact of realized deviations (and with the magnitude of the exchange rate shock correctly calibrated -see Annex I-, the larger simulated impact on the debt ratio suggests that key macro interaction factors are not captured in the stress testing framework).

  • Some stress tests have rarely played a role in the determination of the external risk rating. Simulated shocks to exports, nominal depreciation, other flows (transfers and FDI), and a combination shock have played the main role in signaling risks of external debt distress. The external financing alternative scenario and the real GDP growth and GDP deflator stress tests are rarely identified as sources of risk (Figure 3. b–f).

Figure 3.
Figure 3.

Adequacy of Stress Tests

Citation: Policy Papers 2017, 006; 10.5089/9781498346351.007.A001

Sources: LIC DSAs during 2008–15, and Fund staff calculations.

12. Some important risks facing sub-groups of LICs have been only partially covered by the standardized stress tests used in the LIC DSF, or by other means:

  • Risks from natural disasters have been narrowly covered in LIC DSAs. Less than 50 percent of countries severely exposed to natural disasters (as identified in IMF (2016)) used a customized scenario to assess their associated impact on debt vulnerabilities.

  • Few country teams have sought to directly assess the risks to public debt positions stemming from the realization of contingent liabilities. Recent examples of such analysis have included Nicaragua (2015), which examined the transfer of private external debt related to the oil collaboration with Venezuela onto the government balance sheet, and Solomon Islands (2016), which looked at potential liabilities from power purchase agreements.

  • Commodity price shocks have also not been directly assessed in relevant cases. Over the last two years, 8 of 11 LICs that experienced a downgrade in their risk rating were commodity exporters; the generic export shock scenario in the DSF did not capture the full impact on export and fiscal revenues from price shocks in these cases.

Evaluation of debt indicators and thresholds: some not playing a role

13. Some debt burden indicators have played a minor role in signaling the risk of debt distress:

  • Debt service indicators have rarely determined risk signals of external debt distress. Only in 8 percent of all DSAs, high risk signals of external debt distress have been exclusively informed by breaches of debt service thresholds (Figure 4. a). Moreover, the distance of debt service indicators to their thresholds in the run up to recent debt distress events remained significant (Figure 4. b). In some cases, even after sizable fiscal deviations for several years and rising market risks, debt service thresholds were not breached. These facts suggest that either debt service measures contain little information in terms of signaling debt distress (which is counter-intuitive) or that the relevant thresholds have been poorly estimated.

  • Among debt stock indicators, the PV of external debt to fiscal revenues appears to be redundant in determining external risk signals. Across all LIC DSAs since the inception of the framework, there was only one case in which this threshold was breached under the baseline without breaches in other debt stock indicators (Figure 4. c). In contrast, the PV of external debt to exports has signaled high risk of debt distress in about 80 percent of all cases. This could reflect the information content of these debt stock indicators and/or the appropriateness of the existing debt stock thresholds.

Figure 4.
Figure 4.

Performance of the DSF in Setting Risk Ratings

Citation: Policy Papers 2017, 006; 10.5089/9781498346351.007.A001

Source: Fund staff calculations.

14. As regards the new features added in the 2012 review (Box 1), the framework has a high bar for applying these, and the added value of making use of them has been limited:

  • Remittances-augmented thresholds. To inform the determination of risk ratings, 13 out of 69 countries have incorporated remittances as an integral part of the framework. Of those, five saw upgrades in their risk ratings. Extending the application of the remittances-augmented debt thresholds to all countries for which remittances data are available since 2014 (not only those that qualify under DSF rules) would not have improved any risk rating, and would have produced 14 downgrades. This outcome is the combined result of the way in which remittances-augmented thresholds were derived in 2012 and the highly-skewed distribution of remittances.6

  • Probability approach (which involves moving beyond discrete thresholds set by the CPIA-based country classification to consider all country-specific information in the model). Since the new features became operational in 2014, only nine out of 69 countries have used the probability approach to inform their risk ratings (leading to one final risk rating improvement). Had it been more broadly applied to all DSAs produced since 2012 it would have upgraded the risk ratings in 17 DSAs in the sample, while there would have been a downgrade of risk ratings in 28 of them.

Performance of rules to set risk ratings: too many false alarms and scope for deeper risk analysis

15. Overall, the mechanical framework underlying the DSF produces a high rate of false alarms.7 While debt burden indicators perform poorly in terms of their individual capacity to predict debt distress, the joint use of all debt burden indicators—with a breach by any one sufficing to signal high risk—significantly improves the DSF’s overall predictive performance. Still, the results imply that the existing framework yields a large rate of false alarms (about 50 percent of the cases in which debt distress is not observed) and a more modest rate of missed crises (about 20 percent of the cases in which debt distress is observed).8 Updating the sample period and re-estimating debt thresholds based on the existing methodology would imply a significant tightening of debt thresholds, leading to a much larger rate of false alarms (about 60 percent) and a very low rate of missed crises (about 8 percent).

16. The use of staff judgment for the final determination of external risk ratings has generally been beneficial, but has not been evenly applied. Since the inception of the DSF, judgment has been applied to override the mechanical risk signal in about 25 percent of all cases, leading to an upgrade about 80 percent of the time. Staff judgment has reduced the rate of false alarms vis-à-vis the mechanical application of the framework by about 10 percentage points. Country teams have typically used judgment to override risk signals driven by marginal and/or temporary breaches, although this does not happen in all cases. For instance, the mechanical risk signal has been overridden in 48 percent of all marginal breaches (measured as a 5 percent deviation from the threshold).

17. Deeper attempts to disentangle risks have not been a strong feature:

  • The moderate risk category conceals diverse debt vulnerabilities in a large number of countries. Figure 4. d shows that the number of years in which debt indicators are projected to breach thresholds under stress tests varies widely, with some countries having only temporary breaches while others showing protracted ones. A similar result is found for the size of breaches, where some moderate risk countries have large breaches while others have minor ones.

  • The new feature added during the 2012 review—to provide for a deeper assessment of total public debt and an overall risk rating—has been only sporadically used. Since 2014, DSAs documented 26 countries with significant vulnerabilities emanating from total public debt.9 However, only 10 countries reported an overall risk rating, and only six of those provided an in-depth discussion of domestic public debt vulnerabilities. Most DSAs did not provide a discussion of the extent of vulnerabilities stemming from rollover risks, the increasing participation of non-residents in domestic local-currency bond markets, or from the structure of domestic public debt.

18. Finally, the DSF timeframe for considering risks appears excessive. Breaches of debt burden indicators are concentrated during the early years of the projection horizon (Figure 4. e–f). The fact that debt forecasts become subject to increasing underestimation errors as the projection horizon is extended is likely playing a role in this result, but even with better forecasts these would still be subject to significant uncertainty.

Case Studies

19. Select case studies show that the DSF has been successful in signaling impending debt difficulties in several country cases, but its effectiveness has been limited in others due to a variety of factors. In several cases (e.g., Chad, Mongolia) the framework signaled impending debt difficulties one year or more in advance, capturing the impact of expansionary fiscal policy and the external debt service outlook. However, other case studies highlight areas in which the framework would benefit from improvements, such as securing stronger baseline debt projections (e.g., Maldives, Djibouti); enhancing the analysis of key risks in stress tests (e.g., Sri Lanka, Samoa); and expanding the assessment of broader risks (e.g., Ghana, Central African Republic). See Annex II for more details.

B. Assessing the Model Specification and Methodology Underpinning the LIC DSF

20. A closer look at the technical approach underlying the DSF framework also reveals potential areas for reform. Underpinning the DSF is a core model of debt distress, estimated by looking at actual debt distress episodes. Thus, two immediate questions arise concerning how well debt distress episodes have been defined—an incorrect definition would reduce performance—and how well specified and accurate the model is. Since the framework then works off thresholds derived from the model (i.e., comparing baseline forecasts and stress test scenarios to them), another key question concerns the methodology used to derive these thresholds. Finally, it is also worthwhile to explore limitations in the existing methodology for stress tests.

21. The accuracy of the estimated core statistical model depends on correctly identifying external debt distress episodes, but the existing approach has limitations (see also Annex III):

  • The existing approach focuses on identifying severe debt distress episodes by requiring distress signals to be observed for at least three consecutive years.10 Distress events such as pre-emptive restructurings that take less than three years are ignored.

  • The identification of distress signals based on IMF disbursements could be improved. The emergence of debt difficulties is captured at the time cumulative IMF disbursements exceed 50 percent of quota, providing either a false alarm—sooner or later many GRA arrangements breach such a cutoff without necessarily involving debt difficulties—or a missed or late call.

  • The identification of distress signals based on Paris Club restructurings warrants revision. A debt restructuring deal with creditors, currently used to signal the start of the distress episode, typically comes well after the onset of debt problems. In addition, HIPC treatments are over-represented given that multi-round debt restructurings between the decision and completion points are considered as separate signals.

22. The specification of the core statistical model, which relates the probability of external debt distress to debt burden indicators and country characteristics, could be enhanced. While the existing model has done a reasonable job of accounting for historical patterns of external debt distress, it makes use of few variables and, when the sample period is updated (to include seven years of extra data) and the revisited identification procedure of debt distress episodes used, its performance deteriorates sharply.

23. There are limitations in the methodology used to derive debt thresholds:

  • Thresholds are derived individually for each of the five debt indicators without regard to the information contained in other debt indicators. This is at odds with the DSF’s aggregation rule, which looks at all debt indicators and produces a risk signal if there is a threshold breach for any indicator. This may have introduced a downward bias in the estimation of debt thresholds (see Berg and others, 2014).

  • The framework would benefit from a transparent policy choice for the tolerance of missed crises and false alarms. Instead of direct specification of weights on the rate of missed crises (type I error) and false alarms (type II error) for deriving debt thresholds, the present framework averages the results over a range of weights (from 50 to 75 percent) for type I error. This in the view of some observers is too conservative and may have introduced a downward bias in the estimation of debt thresholds (see Berg and others, 2014).

  • Country-specific information could be better integrated into the methodology for classifying countries. Countries are classified into weak, medium or strong performers based only on the strength of their policies and institutions, measured by the CPIA score, without regard to other factors that can be shown to be of value in predicting external debt distress.

24. The stress testing framework can be improved along two dimensions:

  • The shocks that are modeled could seek to capture macro-linkages among key variables. The lack of macro-linkages seems to be a key reason for the miscalibration of the impact of the exchange rate shock (paragraph 11). In any event, lack of macro-linkages contradicts recent experiences with LICs and evidence in the literature (see Aisen and Hauner, 2008, for instance, on the primary balance and borrowing cost relationship).

  • A common set of shocks should be applied to external and total public debt. For example, the primary balance shock has been considered only in public DSAs but not in external DSAs, although deteriorations in fiscal positions can affect external public debt and debt service through the channels of both higher public external borrowings and higher borrowing costs. Similarly, some simulated shocks to external public debt did not inform the assessment of total public debt vulnerabilities.

25. Finally, the DSF does not provide explicit tools for assessing risks associated with market financing. In recent years, several frontier LICs have received significant amounts of market and other forms of non-concessional financing, thereby transforming their public debt profiles and the nature of risks to which they are exposed (e.g., risks from shorter maturities and a more diverse creditor base). As a result, countries with market access are increasingly exposed to more frequent spikes in financing needs, particularly due to issuances of bonds with bullet amortization.11 In addition, external debt service indicators alone may no longer be sufficient to capture the extent of liquidity risks (Figure 5). Indeed, the experience in some countries (e.g., Ghana and Sri Lanka) showcases the challenges that country teams had in accounting for rollover risks in the face of growing fiscal deficits and the resulting high gross financing needs (GFNs), and often had to rely on their own analysis outside of the LIC DSF.

Figure 5.
Figure 5.

Gross Financing Needs and External Debt Service

(Percent of GDP)

Citation: Policy Papers 2017, 006; 10.5089/9781498346351.007.A001

Source: Fund staff calculations.

Proposed Reforms

26. The assessment of the DSF performance and methodology has pointed to areas where the framework should be strengthened. The approach taken in refining the framework has been guided by two overarching principles. First, there is a need to preserve the core architecture of the DSF—model-based results complemented by judgment (Figure 6). A model helps ensure consistent and transparent application of the framework grounded in empirical observations. At the same time, a model cannot capture the intricacies of every country, so a role for judgment is imperative. Second, there is a need to ensure continued balance in the application of the DSF. The framework should continue providing countries early warnings of debt distress—as these episodes are very costly when they occur—but without generating multiple false alarms (which would unnecessarily advise against additional borrowing).

Figure 6.
Figure 6.

Structure of the Reformed LIC DSF

Citation: Policy Papers 2017, 006; 10.5089/9781498346351.007.A001

27. The specific reforms proposed would correct weaknesses and gaps in the framework, adapt it to the evolving circumstances facing LICs, make it more transparent, and yet simpler to use. Specifically, the limitations noted with the DSF methodology would be addressed by updating the approach to identify debt distress episodes and the specification of the core model. These changes would then allow the integration of key country-specific information into the classification and risk assessment of countries. This would also serve to make the framework more transparent and complete. New realism tools—to understand the key assumptions underpinning macroeconomic projections—would both support the implementation of the new approach to classify countries and help address the problems identified with debt projections. Greater simplicity would be achieved by dropping the various features that have proved, in practice, to add little value to the assessment of risks. Finally, added tools would help with a deeper assessment of risks and firmer application of judgment. Figure 6 and Table 1 summarize the proposed reforms.

Table 1.

Proposed Reforms to the LIC DSF

article image

28. The rest of this section is organized as follows: Subsection A covers adaptations to the core debt distress model; subsection B discusses revisions to the approach to classify countries; subsection C discusses new macro realism tools; subsection D covers modifications to the framework to determine mechanical risk signals; subsection E covers the analysis of other risk factors; subsection F covers enhanced guidance for the application of judgment; and subsection G summarizes the elements needed to draw conclusions on the risk of debt distress.

A. Strengthening the Statistical Model for Predicting Debt Distress

29. The core engine of the DSF framework is the statistical model to predict debt distress. The model identifies the key factors affecting the probability of debt distress, and the DSF uses the estimated model to derive debt thresholds consistent with different carrying capacities for debt. To be effective at predicting debt distress, the model must build off of the right identification of debt distress episodes, and incorporate the right explanatory factors.

30. The estimation of the core model in this review has used a revised procedure for identifying external debt distress episodes to improve model performance. The aim has been to eliminate episodes that do not represent debt distress while bringing in those that clearly do. Debt distress episodes would now also include distress periods that last only one or two years (except when the shorter episode is driven solely by the occurrence of external arrears).12 The existing criteria used to identify distress episodes would also be redefined: (i) the identification of distress signals based on IMF disbursements would be refocused on large upfront financing disbursements; (ii) information from new debt restructuring databases would be incorporated; and (iii) a better treatment of the timing of debt restructurings would be added (see Annex III for details).

31. To strengthen predictive capacity, the model specification has been expanded to incorporate additional explanatory variables (see Table AIII.2 in Annex III). In particular, the new model includes two important proxies of capacity to repay (international reserves scaled by imports, and remittances scaled by nominal GDP), as well as a proxy for global shocks (world growth).13 These additional controls have been used extensively in the literature on sovereign debt crises and sovereign borrowing (e.g., Manasse and others (2003), Manasse and Roubini (2005), Gelos and others (2011), Catão and Milesi-Ferretti (2014)). To further improve its accuracy and applicability, the new model has been estimated based only on data from all LICs (the previous model also included data from MICs).

32. Alternative variables and approaches were considered, but not incorporated as they did not improve predictive performance:

  • Other macro controls considered include real GDP per capita, the current account balance/GDP and foreign direct investment/GDP, a measure of the country risk premium, global interest rates, commodity prices, the terms of trade, and measures of natural disasters and conflicts.14 None improved the predictive performance of the baseline specification.

  • The role of trade openness—which has been shown to be relevant in predicting debt distress in emerging market economies—was explored empirically. However, trade openness seems to be less relevant for LICs when it is included alongside remittances in the statistical model, which partly reflects the high correlation between the two variables. The model including only remittances performs better.

  • Consideration was given to estimating different models for specific country groupings (e.g., small states)—but this was not technically feasible given the limited number of debt distress episodes within smaller country groups. In any event, a specialized model is not needed for such cases, given the scope for exercise of judgment to complement the statistical model. Moreover, new tools (i.e., tailored scenario stress tests; see more below) would allow key vulnerabilities facing small states to be incorporated into the risk assessment.

B. Classifying Countries by their Debt-carrying Capacity

33. A second core element of the framework concerns how a country’s debt-carrying capacity is assessed. This helps determine the thresholds against which debt projections are compared to derive mechanical risk signals (as discussed below, an aggregation rule and risk-weighting scheme are also technically necessary to set the thresholds). The framework at present utilizes just one explanatory factor in the core statistical model—the country policy and institutional assessment (the CPIA, produced every year by Bank staff)—to classify a country’s debt-carrying capacity.

34. The new framework would use the additional information in the updated core statistical model to move away from relying exclusively on the CPIA to classify countries (Figure 7). To this end:

  • A composite indicator would be constructed, covering the CPIA, country growth, reserve coverage, remittances, and world growth (with weights given by the estimated coefficients of these variables in the model). The cross-country distribution of the average composite indicator over a 10-year period (2005–14) can then be used to establish classification cutoffs. In this framework, a country’s debt-carrying capacity would be assessed as weak if its composite indicator fell below the 25th percentile of this distribution, medium if it fell between the 25th and 75th percentiles, and strong if its composite indicator fell beyond the 75th percentile (see Annex III for details).

  • Countries would then be fit into this classification scheme based on the country-specific composite indicator.15 In this approach, forward-looking elements can be integrated into the classification of countries; it is proposed that the country-specific composite indicator be calculated based on the latest five years of historical data and the first five years of projections.16 Mixing historical data and projections allows the framework to capture ongoing changes in the outlook for countries’ fundamentals. Changes in a country’s classification would continue to require two consecutive signals in which the country’s composite indicator exceeds its classification cutoff, to mitigate concerns about undue volatility in applying the framework. Note that since the country-specific composite indicator would be anchored on a long-term average (10 years of data), this would attenuate concerns that countries’ classifications could be influenced by cyclical considerations.

Figure 7.
Figure 7.

New Methodology for Classifying Countries Debt-carrying Capacity

Citation: Policy Papers 2017, 006; 10.5089/9781498346351.007.A001

35. The proposed new classification methodology has significant advantages over relying exclusively on the CPIA score:

  • The new approach would give the overall framework greater predictive power. Ignoring the rich information brought in by the additional country-specific variables would misclassify countries and raise both missed crises and false alarms.

  • Threshold effects—the fact that a small change in the CPIA can produce large changes in debt thresholds under the current framework—would be mitigated. The several components of the new composite indicator—whose co-movements are not perfectly synchronized—and the long-term averages used for its computation would make it less likely that a movement to another classification category would be due to a change in only one variable.

  • The use of forward-looking information would allow country authorities to understand the relationship between their policy framework and their debt-carrying capacity, enhancing engagement. More generally, classification would be more transparent, as the additional country-specific information is objective and readily available to country teams, country authorities, and other stakeholders.

36. Staff considered an alternative to the debt threshold approach to classification, but does not see it as viable. Conceptually, a debt threshold approach has some shortcomings compared to a probability threshold approach. The advantage of the latter is that it uses country-specific information more efficiently to predict debt distress (see Berg and others 2014), while avoiding the so-called threshold effects implied by grouping country-specific information into three classification categories.17 On the other hand, conditional on a soundly-based classification of countries’ debt-carrying capacity, a debt threshold approach has a number of operational advantages over a probability approach: (i) it rules out situations resulting from the strict application of the statistical model where debt thresholds would take extreme values in case of countries with very weak or very strong fundamentals (which would imply zero borrowing space for some LICs and implausibly-high debt limits for others); (ii) debt targets in the context of Fund-supported programs, the Fund’s DLP and the Bank’s NCBP can be more directly mapped to debt thresholds than to probability cutoffs; and (iii) debt thresholds are more intuitive and easier to interpret than some maximum level for the probability of debt distress.

C. Improving the Realism of Baseline Projections

37. To support stronger baseline debt forecasts, and implementation of the new classification methodology, the new DSF would include realism tools. Such tools, which are already included in the DSF for market access countries, provide a point of comparison for forecasts, whether drawing on the country’s own history, cross-country experience, or on relationships drawn from economic theory. They help to inform users of situations where important drivers of the macroeconomic baseline debt projections deviate markedly from experience (either with an optimistic or a pessimistic bias). They are not meant to be prescriptive (in the sense of requiring that the baseline scenario comply with comparator experience), but rather to highlight key assumptions underpinning the projections, thereby focusing discussion on these features of the macroeconomic framework and their realism.

38. A first realism tool would present a decomposition of past and projected drivers of debt dynamics. Specifically, DSF users would be shown the evolution of projections of external and public debt to GDP ratios over DSA vintages (one-year and five-years ago). It would provide several summary charts to help DSF users identify and scrutinize marked changes in historical and projected drivers of debt dynamics (Figure 8 shows an illustrative country). For example, a high past contribution of unexpected primary deficits would caution against projecting an excessive reliance on fiscal adjustment. This tool would also shed light on differences between the historical and projected contributions of the current account and FDI flows to external debt, which may signal potential optimism or pessimism in projected reserve accumulation

Figure 8.
Figure 8.

Drivers of Debt Dynamics—Baseline Scenario

Citation: Policy Papers 2017, 006; 10.5089/9781498346351.007.A001

Source: Fund staff calculations.

39. A second tool would contrast projected fiscal adjustment with the historical experience of countries. The comparison group focuses on a set of LICs requesting a Fund-supported program, as these countries generally have faced a need to adjust their fiscal positions.18 The tool would present the distribution of observed headline primary fiscal adjustment over a three-year horizon, against which a country’s projected primary fiscal adjustment would be compared. A fuller discussion of the credibility of the fiscal path would be called for if the projected primary fiscal adjustment over any three years during the projection horizon exceeds, say, 2½ percent of GDP, which is approximately equal to the top quartile of the distribution of observed primary fiscal adjustment (Figure 9 shows an illustrative case).19

Figure 9.
Figure 9.

Realism of Planned Fiscal Adjustment

Citation: Policy Papers 2017, 006; 10.5089/9781498346351.007.A001

Source: Fund staff calculations.1/ Data cover Fund-supported programs in LICs (excluding emergency financing) from 1990 to 2013. Only the 3-year adjustment since the program request per country is included (horizontal axis). Percent of sample on vertical axis.

40. A third tool would provide benchmarks for assessing growth assumptions in light of public investment dynamics or fiscal adjustment:

  • Assessing the investment-growth nexus. A formal assessment of the impact of public investment on growth is beyond the scope of the DSF.20 The proposed tool would use a simple growth accounting framework to decompose projected growth rates into a: (i) contribution from the changes in the government capital stock due to public investment dynamics, and (ii) contribution from other sources (see Annex IV for details on the methodology). These two sources of growth could then be compared with historical data and past projections. Figure 10.a shows an illustrative case in which public investment was projected to be scaled-up in the 2013 DSA vintage. The tool’s output would trigger a deeper discussion of the underlying assumptions underpinning growth projections in the context of changes in public investment, which is generally now absent in DSA write-ups.

  • Assessing the impact of fiscal adjustment on growth. Negative growth surprises have been identified as the main factor derailing fiscal consolidations (see Mauro and Villafuerte, 2013). While many factors may contribute to this, a key issue to avoid is overly optimistic assumptions about fiscal multipliers. This tool would shed light on the impact of the planned fiscal adjustment on growth projections under a range of plausible fiscal multipliers and persistence parameters, allowing a comparison with the baseline projected growth path (Figure 10. b) (see Annex IV for details on the methodology).21

Figure 10.
Figure 10.

Realism of Baseline Growth Projections

Citation: Policy Papers 2017, 006; 10.5089/9781498346351.007.A001

Source: Fund staff calculations.1/ Left-hand chart shows differences in projected public and private investment rate over two DSA vintages; the right-hand chart compares the simulated contribution of government capital and other factors to real GDP growth over two DSA vintages and based on historical data.2/ Bars refer to annual projected fiscal adjustment and lines show possible real GDP growth paths under different fiscal multipliers.

D. Determining Mechanical Risk Signals

41. A key output of the framework is a mechanical risk signal. This is derived by comparing debt projections in the baseline forecast and under stress tests with debt thresholds, drawing on an aggregation rule and risk-weighting scheme (i.e., how much weight is placed on the rate of missed crises versus that of false alarms). The mechanical risk signal is derived from a simple rule: a country is signaled as low risk if all debt burden indicators are below their corresponding thresholds under the baseline and stress test scenarios; moderate risk if there is any breach under the stress tests but not under the baseline; and high risk if there is at least one breach under the baseline.

42. Two methodological improvements would be made to the technical approach to setting thresholds, to address problems noted in the previous section. First, a transparent policy choice for the tolerance of missed crises and false alarms would be introduced. It is proposed that this be set at 2:1, recognizing the importance of keeping a strong early warning framework of the risk of debt distress (the complex averaging procedure in the existing framework produces an effective weight around this level). Second, debt thresholds for the three classification categories would be derived in a manner that is consistent with the DSF’s aggregation rule, which would efficiently rebalance the role of debt burden indicators per their capacity to signal the risk of debt distress (see Annex III).

43. The procedure for generating mechanical risk signals would be considerably simplified in the new framework, even while a more comprehensive stress testing framework is introduced (Figure 11). The broad approach to mechanical risk signals—the aggregation rule— would remain unchanged as it has been beneficial for predicting debt distress. However, the number of thresholds, debt burden indicators, and standardized stress tests would be considerably reduced. The reduced standardized stress tests would in turn make room for tailored scenario stress tests, better capturing key risks. The rest of this subsection considers each of these in turn.

Figure 11.
Figure 11.

Key Simplifications to the Framework for Determining External Risk Signals

Citation: Policy Papers 2017, 006; 10.5089/9781498346351.007.A001

Debt Burden Indicators and Thresholds

44. Streamlining would address the redundancies noted in the previous section:

  • The indicator on the present value of external debt-to-fiscal revenues would be dropped. After incorporating all proposed changes into the framework, results confirm that this indicator is redundant (including or excluding this indicator does not affect the rate of missed crises and false alarms). This confirms the findings of the previous section (subsection A), which showed that this indicator individually has not in the past played a role in signaling high risk of debt distress.

  • The use of remittances-augmented thresholds would be discontinued. The inclusion of remittances directly in the statistical model informs the country classification of debt-carrying capacity, removing the need for this more complex and indirect approach.

  • The projection horizon underpinning the determination of mechanical risk signals would be shortened from 20 to 10 years. The evidence shows that threshold breaches are concentrated in the first five years of projections and that there is high uncertainty in long-term projections (see previous section). The new DSF template would continue to report 20 years of projections as is the case in the present template, while staff judgment on projected developments during the 10–20-year period could be used to override the mechanical risk signal. This approach would focus greater attention on the determinants of long-term breaches of thresholds, enhancing the transparency of DSAs.

45. With these changes, the predictive performance of the reformed framework significantly improves on existing benchmarks (Table 2). Under the proposed reforms, the probability of false alarms is significantly reduced (by 10 percentage points), while the model’s capacity to anticipate debt distress is improved. For comparison, staff also assessed the implications of keeping the existing model specification and methodology for deriving thresholds, and re-estimated the thresholds based on the updated data (1970–2014). It was found that the existing re-estimated model explains the set of new episodes poorly, with tighter thresholds driving up the mechanical rate of false alarms to about 70 percent, without any gain in overall predictive performance (see Annex III for more details).

Table 2.

Predictive Power of Existing and Reformed Framework

article image
Source: Fund staff calculations.

Weighted sum of errors (weight on type I error equal to 67 percent).

Performance of the existing framework in predicting debt distress episodes as identified by the new approach. Type I and II errors are estimated at 18 and 48 percent, respectively, based on the existing approach to identify debt distress episodes.

Performance of the existing framework in predicting debt distress episodes as identified by the new approach, after re-estimating debt thresholds using the existing methodology based on the new sample (1970–2014).

Performance of the reformed framework.

46. The re-estimated external debt thresholds imply some additional room for countries to borrow, provided they manage their debt service well. The new framework would maintain or increase debt stock thresholds for all countries, but countries deemed to have weak or medium debt-carrying capacity would face lower debt service thresholds (Table 3). Since the present framework had set debt service thresholds at what were typically non-binding levels (see previous section), it is not surprising that the improved methodology produces lower debt service thresholds. The impact also needs to be understood considering the changes to the classification scheme. Since the new scheme upgrades classifications after bringing in key fundamentals (see next section), effectively not all countries face lower debt service thresholds.

Table 3.

Existing and Re-estimated External Debt Thresholds

article image
Source: Fund staff calculations.

Standardized Stress Tests

47. The new framework would apply a smaller set of standardized shocks in the external and public DSAs, to address the redundancies noted in the previous section. The GDP deflator stress test and all permanent shocks would be dropped (although the “historical” scenario would remain as a realism check). The data show that, historically, most shocks have been of a temporary nature. The smaller common set of standardized stress tests to be applied would include shocks to: (i) real GDP growth, (ii) export growth, (iii) primary balance, (iv) other flows (transfers and FDI), (v) exchange rate depreciation, (vi) contingent liabilities, and (vii) a combination shock.22 These shocks would be applied to both external and public DSAs to eliminate inconsistencies in the risk assessment.

48. The set of standardized shocks would also be re-calibrated/re-designed, to address the misalignments noted in the previous section:

  • Interactions among the main macro variables in response to shocks would be introduced. These include: (i) the export growth shock would negatively impact GDP growth; (ii) the exchange rate shock would positively affect inflation and net exports, which would mitigate the impact of a sharp FX depreciation on debt ratios; (iii) the real GDP growth shock would negatively affect inflation and the primary balance; and (iv) the primary balance shock would increase commercial borrowing costs. The modeling of elasticities for these interactions draws on an event study and on empirical evidence from the existing literature (see Annex I).

  • The point of application of shocks would be adjusted. For all stress tests, post shock values would be set to the historical average minus one standard deviation—the current practice in the stress testing framework—or the baseline projection minus one standard deviation, whichever is lower for the relevant periods. This proposed change would reduce the likelihood of underestimation of primary balance shocks, but would also help improve the accuracy of all stress test shocks.

  • The contingent liability shock would be redesigned (the existing “other debt-creating flows” stress test). The existing contingent liability shock would be re-calibrated to 5 percent of GDP (from 10 percent) aimed at capturing potential liabilities stemming from financial sector vulnerabilities.23 A tailored shock would expand its scope where relevant, to capture potentially larger financial sector risks or those resulting from exposure to liabilities not covered under the debt concept in the DSA (see below).

Tailored Scenario Stress Tests

49. The reformed framework would also introduce new tailored scenario stress tests where appropriate (see Annex I for details on triggers and calibration). This would fill a key gap in the framework without overburdening the DSF, and align it with the recommendation of the IMF’s 2014 Surveillance Review to have better tailored scenario analysis. These stress tests would only be triggered for countries assessed to be exposed to particular risks (although they could be used on a discretionary basis in other cases). The scenarios’ default settings would be based on the median shock for LICs, but in each case country teams would be encouraged to customize the scenario parameters to reflect country-specific considerations. As in the case of the standardized stress tests, tailored scenario stress tests could move a mechanical risk signal from “low” to “moderate”.

50. The tailored stress tests would reflect risks that are common to groups of LICs. For some countries, more than one test might apply. Annex I provides details on each test.

  • Natural disasters. The scenario would be required for LICs identified as particularly vulnerable to natural disasters (see IMF, 2016). The scenario’s default magnitudes would involve a one-off shock to public debt (capturing fiscal impacts), and real GDP and export growth decline in the year of the shock, calibrated based on data covering natural disasters during 1950–2015. Country teams would be encouraged to adjust the parameters of the scenario to country-specific circumstances, explaining the basis for the adjustments, including any assumptions about the impact of natural disasters already embedded in the baseline scenario.

  • Contingent liabilities supplement. The re-calibrated standardized contingent liability shock would be supplemented with a tailored shock triggered when significant elements of the public sector are not covered by the public debt concept used in the DSA (e.g., other parts of general government, public private partnership exposures, or unaccounted state enterprise exposures) or when financial sector risks to the sovereign balance sheet are deemed larger than the size modeled under the standardized contingent liability shock.24 The scenario would involve a onetime increase in the debt ratio, scaled to the size of potential exposures in those sectors not covered by the country-specific public debt concept.

  • Commodity export price shock. The scenario would be triggered for LICs where commodity exports represent at least 80 percent of merchandise exports.25 The proposed scenario would capture the impact of a sudden one standard deviation decline in commodity export prices (using the distribution underlying WEO forecasts), with macro interactions incorporated, based on staff event analysis and recent studies (IMF, 2015c and Aslam and others, 2016).

  • Market-financing stress test. The shock would be triggered for LICs with access to market-based financing. A tailored combination shock would capture the impact of stressed external commercial borrowing conditions resulting from a deterioration in global risk sentiment, covering increases in the cost of new external commercial borrowing, temporary nominal depreciation, and shortening of maturities of new external commercial borrowing.

51. The use of customized alternative scenarios would continue to provide flexibility to assess other country-specific risks not covered by standardized and tailored stress tests. Although standardized and tailored stress tests ensure comparability among countries, certain idiosyncratic risks may be overlooked (e.g., delays in the implementation of megaprojects, active conflict scenarios), and magnitudes or duration of certain shocks may be underestimated. The DSF template would be enhanced to better allow users to fully design customized scenarios to address such country-specific circumstances.26 The revised Staff Guidance Note would also address good practice in handling such risks, including those related to conflicts.

E. Analyzing Other Potential Risk Factors

52. The existing DSF calls on staffs to examine potentially significant risks not captured in the core statistical model. In particular, staffs are called upon to conduct a deeper analysis of domestic debt vulnerabilities when public debt-GDP ratios are approaching or exceed estimated benchmark levels; to assess the risks to the balance of payments position and to government exposure to contingent liabilities of elevated levels of private external debt; and to assess the risks posed by shifts in market sentiment to debt positions in LICs that have a high share of public debt contracted on market terms with private external creditors. Where assessed to be significant, these risks can affect the overall risk of debt distress—a broader concept than the risk of experiencing external debt distress (IMF-WB, 2013b).

53. The new framework would continue this approach. It would include new tools to help users assess public debt-related risks and market financing-related risks, and a tool to help users better understand, in summary form, the diverse risk characteristics of countries at moderate risk of debt distress. The DSF template would automatically report on these broader risk assessments through a heat map.

Total Public Debt

54. The new DSF would continue to require an assessment of risks stemming from domestic debt levels, through an analysis of total public debt. Although external PPG debt remains the largest component of total public debt in most LICs, a formal analysis of risks from total public debt is warranted because a) domestic debt markets are an increasingly important source of financing for many LICs, and b) in some frontier LICs featuring deeper financial markets and less restrictions on capital flows, non-residents have increased their participation in local and regional debt markets, blurring the distinction between domestic and external debt.27

55. The methodology to assess public debt-related risks has been strengthened. Lack of adequate data on domestic arrears is a key impediment to identifying domestic debt distress episodes. An alternative identification procedure has been used to capture de facto domestic default, based on two proxies (see Annex III). A noise-to-signal modelling approach is then used to derive benchmarks for the PV of total public debt—contrasting with the probit models deployed for this purpose in the 2012 LIC DSF review.28 29

56. The newly-estimated benchmarks for the present value of total public debt are not substantially different from the existing benchmarks (Table 4).30 As in the analysis of external debt distress, the new benchmarks imply a lower rate of false alarms compared to the one achieved by the existing framework at the time of the 2012 review. Estimated benchmarks are robust to variations in the definitions of the two proxies of de facto domestic default. The framework would signal high risk of public debt distress if any of the four external debt burden indicators or the public debt indicator breach their corresponding threshold/benchmark; moderate risk if the thresholds/benchmark were breached in stress tests; and low risk if the thresholds/benchmark were not breached in either the baseline or stress tests scenarios.

Table 4.

Existing and Re-Estimated Benchmarks for Total Public Debt

article image

Based on NTS approach, using 0.67 weight on typ e I error. Benchmarks are rounded up to the nearest 5 percent.

Based on the probit model and weight on type I error ranging from 0.5 to 0.75.

Market-financing Pressures

57. The new DSF would include a tool to signal rollover risks from external commercial borrowing exposures, where such exposure is significant. The debt service indicators in the existing framework provide some indication of risks under the baseline scenario and the proposed market stress test (discussed in subsection D above) examines the impact of shifts in market sentiment on the four debt indicators tracked in the framework. However, while the behavior of these indicators is informative, the historical experience suggests that they are not sufficient, by themselves, to fully capture the extent of liquidity risks faced by countries with market access under the baseline scenario.31

58. The new tool would assess whether estimated gross public financing needs (as a share of GDP) and prevailing EMBI spreads point to risks of experiencing debt distress. Applying the noise-to-signal approach to external debt distress episodes in a sample of frontier LICs and MICs between 1995 and 2015 produces estimated benchmark levels for public gross financing needs to GDP (a comprehensive measure of liquidity pressures) and for bond spreads (a measure of market risk perception) (see Annex V for details). The two indicators tend to move together in the run-up to debt distress (Figure 12).

Figure 12.
Figure 12.

GFNs and EMBI Spreads Around External Debt Distress Episodes

Citation: Policy Papers 2017, 006; 10.5089/9781498346351.007.A001

Source: Fund staff calculations.Note: Data is on annual basis since the mid-1990s for which the two indicators are available.

59. Market-financing vulnerabilities would be deemed significant if there were consistent early warning signals across the two indicators. A breach of both benchmarks (GFNs above 14 percent of GDP and EMBI spreads higher than 570 bps) would signal heightened liquidity needs at times of adverse external market conditions, thereby increasing rollover risks and the likely need to seek exceptional financing. In such circumstances, country teams would be expected to provide an in-depth assessment of liquidity needs and the composition of the creditor base, as input into the final conclusions of the risk assessment.

Moderate Risk Category

60. The new framework would provide a tool to help users better understand the nature of debt vulnerabilities in countries rated as facing a moderate risk of debt distress. One way to illuminate the diversity of debt vulnerabilities is by measuring how far a country’s debt burden indicators are from crossing the debt thresholds under the baseline scenario. This distance to breaching thresholds or “space to absorb shocks” can be assessed against the distribution of shocks that have been observed in the past to lead to a downgrade of countries to high risk of debt distress. This uses existing information in DSAs and does not require a view about the likelihood of shocks (as an “outlook” would).

61. Using the tool, countries at moderate risk of debt distress would be characterized per their assessed “space to absorb shocks”. Taking this approach, a country rated as facing moderate risk would be characterized as having:

  • Limited space to absorb shocks” if a median-size observed shock would lead to threshold breaches under the baseline, excluding cases of marginal and/or temporary breaches (i.e., if there is at least one debt burden indicator with insufficient distance to thresholds under the baseline scenario to withstand a shock equivalent to the median observed shock, estimated at 20 percent of the threshold for debt stock indicators, and 12 percent for debt service indicators).

  • Substantial space to absorb shocks” if threshold breaches would not occur under the baseline in all but shocks in the top quartile of the observed distribution (i.e., if all debt burden indicators have sufficient distance to thresholds under the baseline scenario to withstand all but top quartile observed shocks, estimated at 40 percent of the threshold for debt stock indicators, and 35 percent for debt service indicators).

This assessment of “space to absorb shocks” would shed light on the robustness of the debt position of countries in the moderate risk category and the potential for slipping into high risk of debt distress; it would not have operational implications for Fund/Bank debt policies (see Annex VI for details on the methodology).

F. The Application of Judgment

62. Judgment would continue to play an important role in the framework. The mechanical risk signals from the framework would provide a first cut at determining the external risk rating, but this could be modified on the basis of staff judgment. Any application of judgment would need to be justified in the DSA write-up of the final risk assessment. Factors that could justify an adjustment in the external risk rating due to judgment would include:

  • Breaches of thresholds that are assessed to be marginal and/or transitory in nature. Given the incomplete explanatory power of the probit model (as is the case for any econometric tool), the strict mechanical application of the DSF’s ratings approach—where the breach of any one threshold signals high risk, could unduly penalize countries with marginal and/or transitory breaches. Staff would be expected to avoid mechanical application of the risk assessment rules in such circumstances, while retaining the ability to do so if a convincing case can be made. A technical specification of the circumstances under which this presumption would apply will be developed for the Staff Guidance Note.

  • Concerns as to the severity of domestic debt vulnerabilities could justify adjustment of the risk of external debt distress in cases where non-residents hold a sizable share of domestic government debt; such concerns would need to be flagged in the summary risk assessment even if they did not rise to a level that warranted adjustment of the external risk rating.

  • Concerns about the exposure of the sovereign to external market-financing pressures could also justify adjustment of the risk of external debt distress if sufficiently severe, and would warrant mention in the summary risk assessment even if not sufficient to justify adjustment of the external risk rating.

  • Other country-specific considerations (risks and mitigating factors) not captured by the statistical framework—such as the availability of sizable public financial assets (i.e., partially offsetting public debt levels), conflict risks, long-term considerations (e.g., vulnerability to climate change), availability of insurance-type arrangements, and collateralized financing arrangements. The Staff Guidance Note, to be prepared by IMF and WB teams, would discuss the considerations under which these factors should be considered for the final determination of external risk ratings.

G. Drawing Conclusions on the Risk of Debt Distress

63. Having conducted the various tests and assessments discussed above, and brought in judgment on other factors, staffs would be expected to provide:

  • A rating of the risk of external debt distress, using the established categories: low, moderate, high, or in debt distress.32

  • A rating on the overall risk of debt distress, where vulnerabilities linked to public domestic debt levels are a serious concern: low, moderate, high, or in debt distress.

  • A full discussion of the main risks to this assessment, including from factors such as data coverage, macroeconomic uncertainty, policy implementation risks, and global factors, among others.

64. The summary risk assessment would be expected to explain how the staffs reached their conclusion and to provide perspective on the evolution of debt-related vulnerabilities over time. This would cover the results derived from the mechanical features of the framework as well as explanations of the decisions made to deviate from the mechanical risk signals. The summary would also be expected to comment on the evolution of debt risks and vulnerabilities over time, and identify key risks/mitigating factors that could shift the risk assessment going forward.

Back-Testing Results

65. To help better illuminate the properties of the reformed DSF, staff has back-tested its mechanical features. It is important to emphasize that mechanical back-testing results cannot be taken as a literal interpretation of how country classifications, external risk ratings, and the analysis of other risk factors would evolve. By necessity, back-testing results are based on a vintage of macro data and projections, and most DSAs (about 95 percent) used in the simulations were produced during 2015–16. Likewise, country classifications were informed by macro data and projections from the 2016 Fall WEO. Moreover, mechanical risk signals do not factor in full customization of the tailored stress tests, nor staff judgment, which as noted in the second section of the paper, has been applied in about 25 percent of DSAs since the DSF inception.

66. On country classifications, back-testing suggests more upgrades than downgrades relative to the existing CPIA-based classification. Out of 67 countries for which LIC DSAs are currently produced, the CI-based classification keeps the existing CPIA-based classification unchanged for the majority (40 countries, 60 percent of the sample), upgrades 22 countries (33 percent), and downgrades 5 countries (7 percent) (Table 5.a). This relative stability is in part explained by the fact that the CPIA still accounts for an important share of the CI (45 percent). Most upgrades are driven by large reserve positions and/or sizable remittances flows (Table 5.b). These results confirm that each variable in the composite indicator plays a meaningful role in classifying countries’ debt carrying capacity (see Annex III).

Table 5.

New Country Classification of Debt-Carrying Capacity

article image

67. On mechanical risk signals, back-testing points to a balanced impact of the framework. Out of 67 countries currently producing LIC DSAs, the reformed framework would keep the existing mechanical external risk signals unchanged for the majority of countries (47 countries, 70 percent of the sample), upgrade 12 countries (18 percent), and downgrade 8 countries (12 percent) (Table 6). Overall, back-testing results are not very different from those under the current framework at the aggregate level, with the distribution across risk signal categories remaining broadly unchanged.

Table 6.

External Risk Ratings and Transition Matrix of Countries’ Mechanical Risk Signals

article image

68. On the assessment of other risk factors, back-testing results give a sense of how they could be deployed:

  • For public debt, assessments would closely parallel those given in the current DSF. As expected, the broadly unchanged re-estimated total public debt benchmarks lead to a very few changes in mechanical signals of heightened risks from total public debt vulnerabilities (i.e., breaches of the total public debt benchmark). Most assessments under the existing framework (about 90 percent of the cases) are reaffirmed under the reformed DSF (Table 7).

  • For market-based risks, two countries out of 17 LICs with market access would be flagged (Table 8). Five countries do not breach any benchmark, while in the other cases either only one benchmark is breached or the assessment is inconclusive due to unavailability of EMBI spreads.

  • Finally, back-testing of the methodology to identify the extent of “space to absorb shocks” in the moderate risk category shows that it can usefully discriminate across countries. Of the 32 countries currently rated as facing moderate risk of debt distress, 16 would be characterized as having “limited space to absorb shocks” while 6 would be characterized as having “substantial space to absorb shocks”. Of note, 7 out of the 16 countries flagged as “with limited space to absorb shocks” already have a mechanical high risk signal (Table 9).

Table 7.

Back-testing of Total Public Debt Benchmarks

article image
Table 8.

Back-testing of Market-Financing Risk Indicators

article image

Max GFN over a 5-year projection horizon based on data from the latest available DSA.

EMBI spreads correspond to the latest available 3-month average.

Table 9.

“Space to Absorb Shocks” in the Moderate Risk Category

article image

Discount Rate for the LIC DSF

69. Since the DSF uses indicators of debt that focus on present value, an important consideration is the discount rate used to determine this present value. Too high a rate can underestimate the debt burden (and risks); too low a rate can over-estimate it (and over-state risks). Given drawn out and large movements in global interest rates since the global financial crisis, the issue is of particular relevance.

70. The methodology used to set the LIC DSF discount rate has been updated twice in the last 12 years:

  • When the LIC DSF was introduced in 2005 the adopted discount rate was the then-prevailing level of the USD Commercial Interest Reference Rate (CIRR) of 5 percent. This was to be adjusted by 100 bps whenever the 6-month average USD CIRR deviated from that level by at least 100 bps for a 6-month period. Based on this rule, the discount rate was adjusted to 4 percent in 2009 and 3 percent in 2012. However, these developments were primarily driven by low interest rates in developed countries and were thus seen to have unjustifiably narrowed LICs’ assessed borrowing space against the backdrop of relatively stable financing terms in LICs.

  • In 2013, reflecting concerns that had arisen, staff proposed a change to the methodology by using as a reference point a long-term average (10-year average of the USD CIRR) plus a margin to reflect the term premium for loans with maturities longer than the bond yields used to derive the CIRR. The long-term average aimed to smooth out the effect of low global interest rates that was assessed to be cyclical at the time. As a result, the discount rate was set at 5 percent.33

71. As requested by Executive Directors (see IMF-WB, 2013a), staff has assessed whether there is justification for an update of the discount rate. Staff considered two methodologies:

  • The existing CIRR methodology, which links the discount rate to market rates in advanced economies. This has several benefits: (i) the CIRR is transparent and readily available, and (ii) it yields a PV of debt that can be broadly interpreted as the amount of risk-free investment a country needs to make to be able to service its external debt.34 Applying the existing methodology today, the discount rate would be 4.75 percent.

  • A methodology linking the discount rate with LICs’ average per capita income growth. The right discount rate for the purpose of assessing debt sustainability is arguably the one that reflects LICs’ economic fundamentals and capacity to repay debt. A concept based on the nominal GDP growth rate addresses these concerns. Intuitively, higher per capita income growth implies a greater repayment capacity as future debt service obligations represent a smaller burden on future taxpayers. With the median nominal dollar GDP growth rate for LICs at 7 percent and the median population growth rate at 2.2 percent, the discount rate based on this methodology would be 4.8 percent.35

72. The review proposes to keep the current 5 percent discount rate, revisiting this decision in future DSF reviews, and to maintain a unified discount rate for the LIC DSF, the debt limits policy (DLP), the non-concessional borrowing policy (NCBP), and the grant element calculator. Both approaches above point to keeping the discount rate at 5 percent at present. Conceptually, this could differ from the rate used for the DLP, NCBP, and the grant element calculator. However, operational complications can arise from different discount rates producing different PV of debt for the same underlying loans assessed in the LIC DSF versus the DLP and NCBP (see IMF-WB, 2013a).

Issues in Implementation

73. The proposed features of the reformed framework would facilitate a deeper discussion of debt risks and vulnerabilities between Fund and Bank staffs and country authorities. This expanded discussion would respond to feedback received during consultations with country authorities, particularly about the need for stronger consideration of country-specific circumstances. Thus:

  • Under the new approach to classifying countries’ debt-carrying capacity, there would be greater engagement with country authorities to discuss key elements of macro projections (and thus the country’s policy framework) that play a role in determining a country’s classification. In principle, the new classification could continue to be done annually using the country teams’ projections reflected in the Fall WEO. However, there could be circumstances in which an update may be warranted based on projections submitted to WEO during the Spring or at the time of producing DSAs. The operational aspects of updating countries’ classifications, including transitional rules to avoid undue volatility in classifications, would be discussed in the Staff Guidance Note.

  • The customization of tailored stress tests would require discussion with country authorities about the appropriate scenario parameters given a country’s experience with these shocks.

  • The technical and/or country-specific considerations underlying the application of judgment and the analysis of other risk factors would be fully disclosed and justified in DSA write-ups after discussion with country authorities. The DSA write-up would incorporate the authorities’ views in this regard.

  • Finally, in a framework with some additional room to borrow, but where greater attention to debt service risks is required, more policy dialogue would be needed on debt management capacity and strategy.

74. The implementation of the reformed DSF would be supported in two ways:

  • Updated materials to support implementation (expected to be available by end-2017):

    • ➢ The review would be followed by a thorough revision and simplification of the Staff Guidance Note, aimed at providing a user manual of the DSF template. The Guidance Note would go beyond the simple application of the framework to cover issues such as: factors that guide the application of judgment, including the appropriate use of the probability approach; use of the framework to better understand fiscal space (building on the work done to date for EMs and advanced economies); use of the framework to make judgments about debt sustainability; and issues in modeling specialized types of risk (such as conflict-related risks). The Guidance Note would also address data requirements (with a view to encourage incremental improvements in debt coverage data reporting); and the applicability of the LIC versus MAC DSA frameworks. It is expected that a Staff Guidance Note will be issued, to support implementation, by end-December 2017.

    • ➢ The DSF template would be re-programmed, simplified, and streamlined, with automated features for ease of implementation and transparency. Composite indicator calculations would be automated and transparently reported, as would be the assumptions underlying other tailored scenario shocks.

  • Interactions with country authorities, through three different channels:

    • ➢ Joint IMF-WB high-level seminar during the 2017 Annual Meetings, focused on clarifying the conceptual underpinnings of the new framework.

    • ➢ Technical discussions with country counterparts during Fund/Bank missions.

    • ➢ Expanded training program, including through onsite and online courses and seminars. With support from donors, external training would be increased from nine missions during FY 2017 to sixteen during FY 2018, including four workshops in regional training centers during the Fall and nine training missions during the first half of 2018.36 This expanded training would help countries to build capacity to prepare their own DSAs and, ultimately, fully own the revamped DSF template. Particular attention would be given to assisting countries with weak capacity, with training tailored to their own specific circumstances.

75. The staffs propose that the new framework becomes effective on July 1, 2018, predicated on the assumption that the updated staff Guidance Note and DSF template are finalized by the end of December 2017. The effectiveness date would ensure teams had adequate time to complete DSAs for all countries under the new DSF prior to July 1, 2019 – the cut-off date for determination of changes to country “traffic lights” under the IDA grant allocation framework applicable to IDA’s fiscal 2020 allocations and IDA19 projections. To maintain this timetable, it will be important to disseminate supporting materials to staffs and country officials in a timely manner, complemented by internal and external training in use of the methodology. Once the new framework becomes effective, country teams would make use of the new DSF in their next cycle of engagement with Fund and/or Bank staff.37 Such interaction, coupled with hands-on experience in tailoring the DSF template to country data and circumstances, are the key elements in building up technical capacity in assessing risks to debt sustainability.

76. Bank/Fund staff would continue providing support after the rollout of the new framework. Continuing training and other technical assistance would support capacity building in the use of the revamped LIC DSF as well as integrating debt management strategies when producing DSAs. In this connection, staff would explore options to link the DSF with debt management tools, in particular the Medium-term Debt Management Strategy (MTDS) analytical tool, for both analysis and policy formulation.

77. The flexibility embedded in the DLP and NCBP should help smooth any implications arising from changes in final risk ratings in the wake of this review.

  • In the case of the DLP, if a country’s risk rating is upgraded, the country would face looser debt conditionality (or no debt conditionality) under a Fund-supported program. If, instead, its risk rating is downgraded, it could still benefit from the flexibility embedded in the current DLP. For instance, a country whose risk rating is downgraded from moderate to high could still access non-concessional resources to finance critical development projects or to implement debt management operations that improve the overall debt profile. A country whose rating is downgraded from low to moderate would generally continue benefiting from the flexibility of the DLP as it provides room for non-concessional borrowing (provided the borrowing does not lead to a deterioration in the moderate risk rating). Where the risk rating of a member under an existing arrangement changes as a result of the new DSF, staff would seek to reach new understandings when changes in debt conditionality are called for by the DLP. Such revisions to debt conditionality would be considered by the Executive Board in the subsequent staff report.

  • Similar considerations will apply under the NCBP for all IDA-only non-gap countries that are either grant recipients in the current fiscal year or MDRI recipients, irrespective of their risk rating of debt distress. All countries will continue to be provided with a choice between loan by loan exceptions to the NCBP, with countries at low or moderate risk of debt distress also provided the option to seek nominal ceilings on new non-concessional external PPG borrowing. Countries at low or moderate risk of debt distress with adequate capacity may also opt for a PV ceiling on total new external PPG borrowing. The options available to countries will remain consistent with the conclusion of the joint IMF-WB debt management capacity assessment. IDA’s grant/loan allocation mix will not be affected by the changes in country risk ratings resulting from the application of the new DSF until the fiscal year 2020 (starting July 1, 2019).38


78. The reforms proposed in this paper imply a significant overhaul of the DSF, bringing significant advantages to all stakeholders. The proposed reforms would address several technical criticisms of the existing framework, and reduce an excessive rate of false alarms (with improved predictive power to signal debt distress events). By incorporating additional key country-specific information into country classification and derivation of thresholds, risk assessments in the new framework would be better customized to country conditions and would provide the basis for a richer dialogue on how policy actions affect debt-related risks. Finally, new realism checks and stress test reforms, and the enhanced guidance on the application of judgment, coupled with greater attention to domestic debt vulnerabilities and market-financing risks, will allow a more informed dialogue on the risks and trade-offs that national policy-makers face.

Issues for Discussion

79. Directors may wish to offer views on the following:

  • Do Directors agree that an expanded set of country-specific information (including reserve coverage and remittances flows) should be used to underpin the assessment of countries’ debt-carrying capacity in the framework?

  • Do Directors see merit in introducing realism tools to promote a deeper understanding of key assumptions underlying baseline macroeconomic projections, including the assumed relationship between public investment and growth, and to support stronger debt projections?

  • Do Directors agree with the need to streamline the mechanical framework, by reducing the number of debt indicators, debt thresholds and standardized stress tests, and to rebalance the debt thresholds?

  • Do Directors support the proposed enhancements to the stress testing framework, including the recalibration and new features of standardized stress tests (i.e., macro-interactions) as well as the inclusion of tailored scenario stress tests to assess key risks facing LICs?

  • Do Directors see a need for a better assessment of broader risks stemming from high domestic debt levels and market-financing pressures, and a tool for characterizing countries in the moderate risk category in terms of its “space to absorb shocks”?

  • Do Directors see merit in enhancing the Staff Guidance Note to ensure a more uniform and transparent application of judgment?

  • Do Directors support the proposed timeline for the implementation of the framework?

Appendix I. Overview of the LIC DSF

1. A model explaining the probability of external debt distress is at the core of the methodology underpinning the DSF. Given an approach to identify external debt distress episodes, the probability of experiencing such events is explained through probit models controlling for debt burden indicators, the strength of institutions and policies—measured by the CPIA—, and country growth.

2. Given an estimated model, countries are classified per their borrowing capacity in three categories: weak, medium, and strong performers. This categorization is exclusively based on the CPIA. Once countries are classified, debt thresholds derived for each category from the probit models are assigned.

3. As a first step for determining a risk rating, risk signals are obtained from the mechanical application of the framework. Based on key inputs—baseline macro projections and a battery of stress tests—forecast for five debt burden indicators (PV of PPG external debt to GDP, exports, and fiscal revenues, and PPG external debt service to exports and fiscal revenues) are produced. These in turn are compared against their respective debt thresholds, after which risk signals are determined based on the following aggregation rule:

  • (i) if none of the forecasted debt burden indicators exceed their corresponding thresholds under the baseline and stress test scenarios, the DSF would signal a low risk of debt distress;

  • (ii) if all baseline forecasts are below their thresholds but at least one forecast exceeds its threshold under the stress test scenarios, the DSF would signal a moderate risk;

  • (iii) if at least one baseline debt forecast exceeds its threshold, the DSF would signal a high risk;

  • (iv) significant or sustained breach of thresholds, actual or impending debt restructuring negotiations, or the existence of arrears would generally suggest that a country is in debt distress.

4. Next, the mechanical risk signals are combined with staff judgment. This helps bringing in country-specific considerations and/or technical elements (e.g., whether breaches are marginal and/or one-off) that cannot be captured by the core model to make a final determination of the risk rating of external debt distress (Low/Moderate/High/In Debt Distress).

5. In the 2012 review, new features were incorporated to the framework aimed to capture country heterogeneity where relevant. These included: i) the use of remittances-augmented thresholds, only applied to countries receiving sizable remittances;1 ii) the introduction of the “probability approach”, an alternative technical methodology, that was to be applied at “borderline cases”;2 and iii) qualification of risks stemming from total public debt or private external debt.


Current Structure of the LIC DSF

Citation: Policy Papers 2017, 006; 10.5089/9781498346351.007.A001


  • Abbas, S. M. A., and J. E. Christensen, 2007, “The Role of Domestic Debt Markets in Economic Growth: An Empirical Investigation for Low-Income Countries and Emerging Markets,” IMF Working Paper No. 07/127, (Washington: International Monetary Fund).

    • Search Google Scholar
    • Export Citation
  • Abbas, S. M. A., N. Belhocine, A. El Ganainy, and M. Horton, 2010, “A Historical Public Debt Database,” IMF Working Paper No. 10/245, (Washington: International Monetary Fund).

    • Search Google Scholar
    • Export Citation
  • Aisen, A., and D. Hauner, 2008, “Budget Deficits and Interest Rates: A Fresh Perspective,” IMF Working Paper No. 08/42. (Washington: International Monetary Fund).

    • Search Google Scholar
    • Export Citation
  • Aron, J., R. MacDonald, and J. Muellbauer, 2014, “Exchange Rate Pass-through in Developing and Emerging Markets: A Survey of Conceptual and Policy Issues, and Empirical Findings,” Journal of Development Studies, 50(1): pp. 10143.

    • Search Google Scholar
    • Export Citation
  • Aslam, A., S. Beidas-Strom, R. Bems, O. Celasun, S. Kılıç Çelik, and Z. Kóczán, 2016, “Trading on Their Terms? Commodity Exporters in the Aftermath of the Commodity Boom,” IMF Working Paper No. 16/27, (Washington: International Monetary Fund).

    • Search Google Scholar
    • Export Citation
  • Bahmani-Oskooee, M., and A. Ratha, 2004, “The J-Curve: A Literature Review,” Applied Economics, 36(13): pp. 137798.

  • Baldacci, E., I. Petrova, N. Belhocine, G. Dobrescu, and S. Mazraani, 2011, “Assessing Fiscal Stress,” IMF Working Paper No. 11/100. (Washington: International Monetary Fund).

    • Search Google Scholar
    • Export Citation
  • Beers, D. T., and J.-S. Nadeau, 2015, “Database of Sovereign Defaults,” Bank of Canada Technical Report No. 101.

  • Berg, A., E. Berkes, C. Pattillo, A. Presbitero, and Y. Yakhshilikov, 2014, “Assessing Bias and Accuracy in the World Bank—IMF’s Debt Sustainability Framework for Low-Income Countries,” IMF Working Paper No. 14/48. (Washington: International Monetary Fund).

    • Search Google Scholar
    • Export Citation
  • Berg, A., E. Buffie, C. Pattillo, R. Portillo, A. Presbitero, and L. Zanna, 2015, “Some Misconceptions about Public Investment Efficiency and Growth,” IMF Working Paper No. 15/272. (Washington: International Monetary Fund).

    • Search Google Scholar
    • Export Citation
  • Bom, P., and J. Ligthart, 2014, “What Have We Learned from Three Decades of Research on the Productivity of Public Capital?Journal of Economic Surveys, 28(5): pp. 889916.

    • Search Google Scholar
    • Export Citation
  • Buffie, E., A. Berg, C. Pattillo, R. Portillo, and L. Zanna, 2012, “Public Investment, Growth, and Debt Sustainability: Putting Together the Pieces,” IMF Working Paper No. 12/144. (Washington: International Monetary Fund).

    • Search Google Scholar
    • Export Citation
  • Calderon, C., E. Moral-Benito, and L. Serven, 2015, “Is Infrastructure Capital Productive? A Dynamic Heterogeneous Approach,” Journal of Applied Econometrics, 30(2): pp. 17798.

    • Search Google Scholar
    • Export Citation
  • Catão, L. A. V., and G. M. Milesi-Ferretti, 2014, “External Liabilities and Crises,” Journal of International Economics, 94: pp. 1832.

    • Search Google Scholar
    • Export Citation
  • Cespedes, L. F., and A. Velasco, 2013, “Was This Time Different? Fiscal Policy in Commodity Republics,” NBER Working Paper No. 19748.

    • Search Google Scholar
    • Export Citation
  • Cruces, J. J., and C. Trebesch, 2013, “Sovereign Defaults: The Price of Haircuts,” American Economic Journal: Macroeconomics, 5(3): pp. 85117.

    • Search Google Scholar
    • Export Citation
  • Das, U. S., M. G. Papaioannou, and C. Trebesch, 2012, “Sovereign Debt Restructurings 1950–2010: Literature Survey, Data, and Stylized Facts,” IMF Working Paper No. 12/203. (Washington: International Monetary Fund).

    • Search Google Scholar
    • Export Citation
  • Gelos, R. G., R. Sahay, and G. Sandleris, 2011, “Sovereign Borrowing by Developing Countries: What Determines Market Access?Journal of International Economics, 83: pp. 24354.

    • Search Google Scholar
    • Export Citation
  • Giovanni, A., and M. de Melo, 1993, “Government Revenue from Financial Repression,” The American Economic Review, Vol. 83, No. 4, pp. 95363.

    • Search Google Scholar
    • Export Citation
  • International Monetary Fund-The World Bank, 2012, “Revisiting the Debt Sustainability Framework for Low-Income Countries,” (January) (Washington).

    • Search Google Scholar
    • Export Citation
  • International Monetary Fund-The World Bank, 2013a, “Unification of Discount Rates used in External Debt Analysis for Low Income Countries: Staff Proposals,” (October) (Washington).

    • Search Google Scholar
    • Export Citation
  • International Monetary Fund-The World Bank, 2013b, “Staff Guidance Note on the Application of the Joint Bank-Fund Debt Sustainability Framework for Low Income Countries,” (November) (Washington).

    • Search Google Scholar
    • Export Citation
  • International Monetary Fund-The World Bank, 2015, “Public Debt Vulnerabilities in Low-Income Countries: The Evolving Landscape,” (November) (Washington).

    • Search Google Scholar
    • Export Citation
  • International Monetary Fund, 2007, “The Changing Dynamics of the Global Business Cycle,” Chapter 5, World Economic Outlook (October). (Washington).

    • Search Google Scholar
    • Export Citation
  • International Monetary Fund, 2012, “Commodity Price Swings and Commodity Exporters,” Chapter 4, World Economic Outlook (April). (Washington).

    • Search Google Scholar
    • Export Citation
  • International Monetary Fund, 2014, “Assessing Reserve Adequacy—Specific Proposals,” (December). (Washington).

  • International Monetary Fund, 2015a, “Making Public Investment More Efficient,” (June). (Washington).

  • International Monetary Fund, 2015b, “Exchange Rates and Trade Flows: Disconnected?Chapter 3, World Economic Outlook (October). (Washington).

    • Search Google Scholar
    • Export Citation
  • International Monetary Fund, 2015c, “Where Are Commodity Exporters Headed? Output Growth in the Aftermath of the Commodity Boom”, Chapter 2, World Economic Outlook (October). (Washington).

    • Search Google Scholar
    • Export Citation
  • International Monetary Fund, 2016, “Small States’ Resilience to Natural Disasters and Climate Change: Role for the IMF,” (November) (Washington).

    • Search Google Scholar
    • Export Citation
  • International Monetary Fund, 2017, “Estimating the Stock of Public Capital in 170 Countries: January 2017 Update”.

    • Search Google Scholar
    • Export Citation
  • Kraay, A., and V. Nehru, 2006, “When is Debt Sustainable?The World Bank Economic Review, 20(3): pp. 341365.

  • Laeven, L., and F. Valencia, 2013, “Systemic Banking Crises Database,” IMF Economic Review, 61: pp. 22570. (Washington: International Monetary Fund).

    • Search Google Scholar
    • Export Citation
  • Ligthart, J., and R. Martin-Suarez, 2011, “The Productivity of Public Capital: A Meta-Analysis,” in Manshanden, W. and W. Jonkhoff, eds. Infrastructure Productivity Evaluation. (New York: Springer).

    • Search Google Scholar
    • Export Citation
  • Mallik, G., and A. Chowdhury, 2001, “Inflation and Economic Growth: Evidence from Four South Asian Countries,” Asia-Pacific Development Journal, Vol. 8, Issue 1, pp. 12335.

    • Search Google Scholar
    • Export Citation
  • Manasse, P., and N. Roubini, 2005, “‘Rules of Thumb’ for Sovereign Debt Crises,” IMF Working Paper No. 05/42. (Washington: International Monetary Fund).

    • Search Google Scholar
    • Export Citation
  • Manasse, P., N. Roubini, and A. Schimmelpfennig, 2003, “Predicting Sovereign Debt Crises,” IMF Working Paper No. 03/221. (Washington: International Monetary Fund).

    • Search Google Scholar
    • Export Citation
  • Mauro, P. and M. Villafuerte, 2013, “Past Fiscal Adjustments: Lessons from Failures and Successes,” IMF Economic Review, 61(2): pp. 379404. (Washington: International Monetary Fund).

    • Search Google Scholar
    • Export Citation
  • Panizza, U., 2008, “Domestic and External Public Debt in Developing Countries,” Unctad Discussion Paper No. 188.

  • Panizza, U., F. Sturzenegger, and J. Zettelmeyer, 2009, “The Economics and Law of Sovereign Debt and Default,” Journal of Economic Literature, 47:3, pp. 65198.

    • Search Google Scholar
    • Export Citation
  • Pennings, S., 2017, “Long-term Growth Model v4.0: Model Description,” available at

    • Search Google Scholar
    • Export Citation
  • Rand, J., and F. Tarp, 2002, “Business Cycles in Developing Countries: Are They Different?World Development, Vol. 30, No. 12, pp. 207188.

    • Search Google Scholar
    • Export Citation
  • Reinhart, C. M., and K. S. Rogoff, 2009, “This Time is Different: Eight Centuries of Financial Folly,” Princeton University Press.

  • Reinhart, C. M., and K. S. Rogoff, 2010, “From Financial Crash to Debt Crisis,” American Economic Review, Vol. 101, No. 5, pp. 1676706.

    • Search Google Scholar
    • Export Citation
  • Reinhart, C. M., and K. S. Rogoff, 2011, “The Forgotten History of Domestic Debt,” The Economic Journal, Vol. 121, Issue 552, pp. 31950.

    • Search Google Scholar
    • Export Citation
  • Reinhart, C. M., and M. B. Sbrancia, 2015, “The Liquidation of Government Debt,” IMF Working Paper No. 15/7. (Washington: International Monetary Fund).

    • Search Google Scholar
    • Export Citation
  • Spatafora, N., and I. Samake, 2012, “Commodity Price Shocks and Fiscal Outcomes,” IMF Working Paper No. 12/112. (Washington: International Monetary Fund).

    • Search Google Scholar
    • Export Citation
  • UNCTAD, 2015, “State of Commodity Dependence, 2014,” (New York and Geneva: United Nations Conference on Trade and Development).

  • Yehoue, E., 2005, “International Risk Sharing and Currency Unions: The CFA Zones,” IMF Working Paper No. 05/95. (Washington: International Monetary Fund).

    • Search Google Scholar
    • Export Citation

See Public Debt Vulnerabilities in LICs: The Evolving Landscape (IMF-WB, 2015).


The technical work of the review has been informed by a broad external consultation process. This has included dialogue with authorities from developing countries and staffs of multilateral development banks (including at the 2016 (Lusaka) and 2017 (Vienna) DMF Stakeholders’ Forums, the 2016 African Caucus in Cotonou, at the 2016 and 2017 Spring Meetings -including bilateral discussions with Pacific Islands Governors and Ministers—, and the 2017 Multilateral Development Bank Meeting on Debt Issues in Washington, DC), and with members of the Paris Club. Staff have also sought feedback from civil society organizations (including through an open web-based consultation and different events during the 2016 Annual Meetings).


Debt service indicators have relatively much smaller unexpected changes. The share of DSAs with unexpected changes above 15 percentage points (in absolute value) ranges between 0 and 8 percent.


Among DSAs with sizable unexpected changes in debt, the share of those underestimating debt outcomes rises from about 55 percent to more than 80 percent when moving from the 1- to the 7-year projection horizon.


This relationship between historical and actual primary balances continues to hold when the sample is expanded to cover the longer period of 2000–15.


The highly-skewed distribution of remittances implies that only countries receiving sizable remittances would benefit from the estimation of remittances-augmented thresholds based on the full sample of countries. Not surprisingly, countries that have benefited from the use of remittances-augmented thresholds received remittances flows as percentage of GDP and exports way in excess of the eligibility cutoffs (averages of 30 percent of GDP and 150 percent of exports).


In case of individual debt indicators, missed crises are defined as cases in which debt distress is observed but the associated debt threshold was not breached, while false alarms are defined as cases in which debt distress is not observed but the associated debt threshold was breached. Type I and II errors are missed crises and false alarms as a share of all cases in which debt distress is observed and non-observed, respectively.


The rates of false alarms and missed crises are measured using historical data during the sample period (i.e., 1970– 2014). It is not straightforward to assess how policy responses to high risk ratings may have affected the number of false alarms, but any such impact may be only present in the few last years of the sample period in which such ratings were produced (2008–2014).


Such vulnerabilities are signaled in relation to the relevant benchmark of the PV of total public debt to GDP ratio.


The DSF relies on three distress signals to identify conditions under which a country is experiencing external debt difficulties: i) cumulative IMF disbursements from the General Resource Account (GRA)—under Stand-By Arrangements (SBA) and Extended Fund Facilities (EFF)—exceeding 50 percent of the member’s quota; ii) restructuring of claims held by Paris Club creditors; and iii) accumulation of arrears on external PPG debt in excess of 5 percent of the outstanding stock of external PPG debt.


Since 2005, 14 LICs have issued 29 Eurobonds worth US$20 billion. Of the 29 issuances, 25 or US$17 billion had bullet payments. Principal repayments of sovereign external debt (in percent of exports of goods and services) by frontier LICs are projected to exceed those of the 17 largest EMs over the next five years (see IMF-WB, 2015).


This helps rule out episodes associated with one-off and temporary occurrence of external arrears (including for technical reasons) that do not necessarily signal debt distress.


See Box AIII.2 and AIII.3 (Annex III) on data sources and issues, including a discussion of reserve measurement in currency unions.


Sub-indices of the CPIA could not be considered due to insufficient data.


The determinants of the country-specific composite indicator would be shown in the DSA output to facilitate understanding of what is driving the classifications and where any changes have come from.


Forecasts of the additional variables in the composite indicator are routinely produced in the WEO database and individual DSAs. Recognizing its slow-moving nature, the forecast for the CPIA rating would consist of its most recent value.


As highlighted in the 2012 review, threshold effects occur when small changes in the predictors of debt distress lead to discrete jumps in debt thresholds. For instance, in the current framework, a weak-performer (say with a CPIA score of 3.24) would face a debt-to-GDP threshold of 30 percent, whereas a medium performer (say with a CPIA score of 3.26) would face a threshold of 40 percent (see IMF (2012), Appendix 1).


While the tool would ideally use cyclically-adjusted primary balances, these are difficult to compute for LICs given the high degree of uncertainty in estimating output gaps. At the same time, measuring fiscal adjustment as the change in headline primary balances across all LICs may include cases in which improved fiscal performance was driven by exogenous factors (e.g., coming on stream of natural resource projects).


This issue will be addressed in the Staff Guidance Note.


A more detailed analysis may be done outside the confines of the DSF, as a means of informing Fund-supported programs, World Bank growth diagnostics, and the policy dialogue more generally. Available tools supported by Fund and Bank staff for this purpose include the IMF’s Debt-Investment-Growth model (see Buffie and others, 2012), and the World Bank’s Long-Term Growth model (see Pennings, 2017).


While a useful check for the consistency of growth and primary balance projections is to uncover the growth path consistent with a neutral fiscal stance and assess its realism, such an approach is challenging in the case of LICs. This is because LICs’ borrowing capacity is very weak in the absence of fiscal adjustment, casting doubts about the likelihood of such a counterfactual.


This stress test reflects a scenario where multiple shocks hit the economy at the same time. It applies half the magnitudes of all stand-alone shocks, and incorporates individual macro interactions as assumed under each individual shock.


This is based on the average increase in debt-to-GDP ratios observed for 44 banking crisis episodes for LICs since the 1980s, as in Laeven and Valencia (2013).


This stress test would help eliminate a disincentive to achieve full and transparent disclosure of public sector contingent liabilities.


This is based on UNCTAD (2015), which is broadly in line with the definition used in IMF-WB (2015). This cutoff captures 36 countries (see Annex I).


If a user yet wished to consider one of the dropped stress tests, this would also offer a channel to do that.


This is particularly important in countries where external debt is defined on a currency denomination basis, which seems to be the norm in LICs, given data limitations on debt by residency (a Fund’s staff survey in 2016 revealed that more than 50 percent of LICs report their external debt on currency basis).


Data on the PV of total public debt is the sum of the PV of external PPG debt and nominal domestic public debt, which is assumed to be contracted on market terms.


The use of probit models for analysis of public debt vulnerabilities would produce a composite indicator different from the one derived in the external debt risk assessment—potentially leading to a classification of countries’ debt-carrying capacity different from that developed in analyzing external debt distress. This would create considerable conceptual confusion, while also increasing the complexity of the general framework.


Benchmarks were rounded up to align the strong classification category with the MAC DSA high-risk benchmark.


This is even more compelling in those cases where non-residents have increased their participation in local and regional debt markets, blurring the distinction between domestic and external debt.


Actual or impending debt restructuring negotiations, the existence of arrears, or a significant or sustained breach of thresholds would generally suggest that a country is in debt distress (see IMF-WB, 2013b). The Staff Guidance Note will elaborate further.


See “Unification of Discount Rates Used in External Debt Analysis for Low-Income Countries” (IMF-WB, 2013a).


The CIRR overstates the risk-free rate, thus underestimating the required investment, as the different export credit agencies (ECAs) agreed to set the CIRR as the average of market rates for long-term government debt in their own currency, plus one percent. This was designed to ensure that ECAs do not unfairly subsidize trade by setting their lending rates too low.


The median dollar nominal GDP growth rate is 7 percent when based on (i) 5 years of history and 5 years of projections, (ii) 10 years of history and 10 years of projections, or (iii) 5 years of history and 10 years of projections, using data submitted in the latest LIC DSAs. The median population growth rate is 2.2 when based on (i) 5 years of history and 5 years of projections or (ii) 10 years of history and 5 years of projection, using WEO data. A longer-term average has the benefit of smoothing out cyclical components and avoiding abrupt changes in the discount rate.


During 2016–17, 175 representatives from 40 countries participated in one-week training sessions held in several regional centers and individual countries around the world, the bulk of which were financed through the Debt Management Facility. Most of them benefited from accessing the IMF-WB online module on the LIC DSF ahead of their face-to-face training. Overall, over the same period, nearly 850 government officials familiarized themselves with the online LIC DSF materials, through the IMF Massive Open Online Course on debt sustainability and debt management.


Missions conducted before the DSF enters into effect will produce DSAs based on the old framework, as will the ensuing staff documentation. In cases where multiple missions are needed to complete the relevant documentation for Board meetings, it will be the timing of the final mission that determines whether the new or old frameworks should be applied.


Each year, IDA determines its grant/allocation mix for the next 12 months based on the external risk ratings available by end-June.


Countries eligible to incorporate remittances into the DSF mechanical analysis have remittances flows greater than 10 percent of GDP and greater than 20 percent of exports of goods and services. Both ratios are measured on a backward-looking, three-year average basis.


A borderline case is defined as one where the largest breach, or near breach, of a threshold under any scenario falls within a 10-percent band around the threshold.

Review of the Debt Sustainability Framework for Low Income Countries: Proposed Reforms
Author: International Monetary Fund