Borrowing costs after debt relief
The Covid-19 pandemic is putting government finances of many developing countries under severe strain (Djankov and Panizza 2020). In response, a range of proposals and calls for action have been floated by experts and policy makers (Bolton et al. 2020a, 2020b, Bulow et al. 2020; Horn et al. 2020; Landers et al. 2020). In a short time, the international community – under the leadership of the G20 – agreed to help poor countries by offering a suspension of debt servicing due in the second half of 2020. Under the Debt Service Suspension Initiative (DSSI) participating countries can ask their bilateral lenders to defer debt service repayments by three years without affecting the net present value (NPV) of public debt. The size of the liquidity provision under the DSSI is non-trivial. For all eligible countries it amounts to $10.2 billion and accounts for about one fifth of the fiscal shortfall due to the Covid-19 shock. However, many eligible countries so far have been reluctant or refused to participate in the DSSI. This may seem a puzzling response to what at first sight is free money at time of great need. Yet, these countries fear that DSSI participation may signal debt sustainability problems that could trigger a downgrades of sovereign ratings and increases in sovereign borrowing costs.1
In a recent paper (Lang et al. 2020), we provide a first assessment of the short-term impact of the DSSI on sovereign bond spreads. In particular, we test whether the potential benefits from short-term liquidity provision outweigh any stigma effects that may be associated with participation in the debt relief initiative. Estimating the effect of debt relief on sovereign bond spreads is usually challenging, as debt relief initiatives are typically not randomly allocated. Comparing beneficiaries of debt relief to other countries is thus not informative. The case of the DSSI, however, allows us to construct plausible counterfactuals. In contrast to most debt restructurings, the DSSI was announced simultaneously for all 73 eligible countries and, thus, was not tailored to the needs of individual countries. Also, the eligibility criteria were based on preexisting income thresholds rather than on financing needs or on the severity of the shock, which crucially influence borrowing costs.
Sovereign borrowing costs declined by about 300 basis points
We exploit this event to analyse its impact on sovereign bond spreads of the 16 DSSI-eligible countries with international market access and available daily data. We used the synthetic control method (SCM) developed by Abadie and Gardeazabal (2003) and now increasingly used in similar contexts (see Marchesi and Masi 2020). For each DSSI-eligible country we construct a synthetic control (or “doppelganger”) combining countries from a pool of middle-income non DSSI-eligible countries.2
Figure 1 shows our main result. Comparing sovereign bond spreads of DSSI-eligible countries with their synthetic controls shows that sovereign spreads significantly declined after debt relief. Several days after the DSSI was announced, the spreads of eligible countries were down by about 300 basis points (bps) more than in comparable, non-treated doppelganger countries. This average effect differs across countries, but it is negative for all borrowers that can benefit from the debt relief. This result is robust to different model specifications, including the generalized synthetic control method (Xu 2017). Moreover, a set of placebo tests in space and time show that the effect on spreads is due to the DSSI and cannot be explained by the (contemporaneous) request of an IMF programme.
Figure 1 Sovereign bond spreads in DSSI-eligible countries versus their synthetic controls
Notes: The figure plots the difference between the actual sovereign bond spreads and those of the synthetic control (spread gap) for the DSSI-eligible countries. The red solid line is the average of the country-specific spread gaps. The grey solid lines refer to countries which have joined the DSSI as of September 17, 2020, while the grey dashed lines refer to countries which have not formally requested to join the initiative (Ghana, Honduras, Kenya, Mongolia, Nigeria, and Uzbekistan). The vertical lines indicate the announcement of the DSSI on April 15, 2020 (solid line) and the first participation in the DSSI on May 1, 2020 (dashed line). The dots signal the country-specific participation in the DSSI. See description in the main text. Source: Bloomberg, Our World in Data, and IMF World Economic Outlook.
The drop in spreads seems to be due to liquidity provision
To discriminate between two mechanisms that could drive the results, we test for heterogeneous effects of debt relief. We focus on two sources of heterogeneity—the size of the DSSI relief and the share of private creditors in debt service—and estimate their effects in a difference-in-difference setting using the local projection method. This analysis shows that the decline in bond spreads for DSSI-eligible countries is larger for countries that happen to have a larger share of debt service due in the eligibility period (between May and December 2020, Figure 2, Panel A). By contrast, the decline in spreads does not depend on the importance of private creditors (Figure 2, Panel B). As there is no increase in spreads—not even for countries owing a large share of repayments to private creditors—these results do not support the presence of a stigma effect. Instead, the results are consistent with a positive liquidity effect due to the postponement of the debt service due in 2020.
Figure 2 Liquidity provision versus stigma
A) Size of the DSSI relief
B) Share of private creditors
Notes: The figures plot the impulse response functions of the differential effect of the DSSI announcement (t = 0) between eligible and ineligible countries on sovereign bond spreads. Panels A and B split the sample between eligible countries which received a DSSI relief above or below 0.5 percent of GDP and those with debt service due to private creditors above or below 60 percent of total debt service due under the DSSI (both thresholds are median values). See description in the main text. Data source: Bloomberg and IMF World Economic Outlook.
The international community is currently discussing the possibility to extend the current initiative to suspend debt service in developing countries to 2021. Our results suggest that this simple NPV-neutral debt moratorium—involving no haircut for creditors—can indeed help countries weather the crisis.
Our findings also add to the broader literature on debt restructuring. They show that rapid and unconditional provision of debt rescheduling to countries that face short-term liquidity shocks may provide an effective instrument of financial support that can help avoid hard defaults (Trebesch and Zabel 2017). In addition, our results support the design and adoption of simple state-contingent debt instruments with floating grace periods to help poor countries mitigate their exposure to adverse shocks (Cohen et al. 2008).
Two final qualifications are important. First, our results could be generalised to other situations in which countries face a short-term crisis. In the presence of severe negative shocks, only postponing debt service could contribute to a reduction in borrowing costs. However, this does not imply that the suspension of debt service will be the optimal response to the Covid-19 crisis in the months to come. If the shock persists, the liquidity crisis could evolve into a solvency crisis, as a change in the long-term growth rate of the economy would affect debt sustainability. In such a scenario, a debt stock reduction could be required to reduce debt overhang and restore debt sustainability. Second, our analysis looks at NPV-neutral debt relief provided by the official sector. How markets would react if private creditors also joined the initiative (as asked by the G20 and the main international financial institutions) remains an open question.
Abadie A and J Gardeazabal (2003), “The Economic Costs of Conflict: A Case Study of the Basque Country”, American Economic Review 93 (1): 113-132.
Bolton P, L Buchheit, P-O Gourinchas, M Gulati, C-T Hsieh, U Panizza and B Weder di Mauro (2020a), “Born Out of Necessity: A Debt Standstill for COVID-19”, CEPR Policy Insight no 103.
Bolton P, M Gulati and U Panizza (2020b), “Legal air cover”, VoxEU.org, 13 October.
Bulow J, C Reinhart, K Rogoff and C Trebesch (2020), “The Debt Pandemic”, IMF Finance & Development, Fall.
Cohen, D, H Djoufelkit-Cottenet, P Jacquet and C Valadier (2008), “Lending to the Poorest Countries: A New Counter-Cyclical Debt Instrument”, Working Paper 269, OECD Development Centre.
Djankov S and U Panizza (2020), “COVID-19 in developing economies: A new eBook”, VoxEU.org, 22 June.
Horn S, C Reinhart and C Trebesch (2020), “China’s overseas lending and the looming developing country debt crisis”, VoxEU.org, 4 May.
Landers C, N Lee and S Morris (2020), “More Than $1 Trillion in MDB Firepower Exists as We Approach a COVID-19 “Break the Glass” Moment”, Center for Global Development.
Lang V, D Mihalyi and AF Presbitero (2020), “Debt Relief, Liquidity Provision, and Sovereign Bond Spreads”.
Marchesi S and T Masi (2020), “Debt restructuring in the time of COVID-19: Private and official agreements”, VoxEU.org, 4 May.
Trebesch C and M Zabel (2017), “The output costs of hard and soft sovereign default”, European Economic Review 92: 416-432.
Xu Y (2017), “Generalized Synthetic Control Method for Causal Inference with Time-Series Cross-Sectional Data”, Political Analysis 25: 57–76.
1 See reports by international institutions (IMF 2020, World Bank 2020), Think Tanks (ODI 2020) and press coverage in The Economist and Reuters, among others. More details on the DSSI can be found here and on the World Bank website.
2 Given that the dynamics of sovereign spreads depend on fiscal and economic performance, we take real GDP growth, the current account, the fiscal balance, and public debt (all as shares of GDP) as macroeconomic variables to construct the synthetic control. Also, to compare countries with similar bond spreads dynamics before the DSSI, we match on spread levels at specific dates. Finally, to take into account differences in the intensity of the Covid-19 crisis, we use the number of cases per capita. See Lang et al. (2020) for details.