The costs of variable interest rate debt
Recently, developed economies have raised their policy rates to unprecedented levels. This marks a significant event in the history of the European Central Bank, while the Federal Reserve has not seen such levels since the early 2000s. What are the potential risks for emerging markets as a result of this?
In advanced economies, there are concerns about a positive interest rate-growth differential (r-g) which makes it difficult to reduce debt. Among others, Olivier Blanchard has expressed concern about this issue. The situation also poses a significant threat to emerging markets and developing countries, as refinancing has become increasingly harder. In fact, a large number of countries are now excluded from bond markets. Moreover, many developing countries have variable interest debt, which means that the cost of higher interest rates is immediately transferred to the debtor. This was noted last year by Marcello Estevao, who highlighted that around 30% of the external debt of "poor" countries that are eligible for IDA financing is variable. This indicates a growing use of such mechanisms.
This blog aims to deepen this analysis and to highlight the exposure of governments to variable debt. The World Bank's annual International Debt Statistics (IDS) provides useful information on external variable debt on a counterparty basis, identifying creditors based in China, France, and various Multilateral Development Banks (MDBs) like the World Bank. The issue of variable debt and its impact has been a topic of concern for quite some time. Recently, the “Triple Agenda report” on MDBs highlighted that net transfers by MDBs may become negative in 2023 due to higher interest payments from their clients. Another well-known source of variable interest loans is the IMF GRA (non-concessional) loan, which has seen a rise in cost and is well documented by Sandor Tordoir and Tobias Krahnke.
This note documents the broad rise of variable public debt, which has become a small but significant source of development finance, representing about $622 billion in 2021, or approximately 3.9% of GDP of developing countries. The study uses countries eligible to World Bank financing (“IDS countries”) with GDP data in the WEO. In some cases, it is not possible to identify variables debt owned by the government or the private sector. After excluding China, Russia and Ukraine from the sample, the results cover 118 countries. This represents 22 % of public external debt for developing countries in 2021.
The study looks across income groups and creditors, finding that MDBs are the largest providers of variable-interest loans, especially in middle-income economies. In 2023, the additional interest cost will represent more than 7.5% of the total of debt service for low-income countries (LICs) and upper-middle income countries (UMICs), and approximatively 10% for lower-middle income countries (LMICs). MDBs offer hedges such as rate-fixing options or interest rate swaps, but reporting on them is limited and seems inconsistent across institutions.
How large is variable debt for developing countries?
The International Debt Statistics (IDS) is a World Bank database that provides information on external debt. The data is collected from the World Bank’s clients, on a loan-by-loan basis, including details such as grace period, maturity, interest rate, creditor, etc.. The database publishes aggregated data on the volumes and terms of countries’ external debt. The latest IDS 2023 report mentions the risks of variable interest loans but does not quantify their impact.
However, the database has information on the proportion of debt that is variable. The reference rate for this type of loan is unknown, but it is likely to be connected to the unsecured market, such as the London Interbank Offered Rate (LIBOR), and its successor, the Secured Overnight Financing Rate (SOFR) (plus a fixed spread). The information available on variable rate debt is not very useful when presented at a general level because it combines external liabilities from both private and public sectors. However, when looking at information on counterparty location, we can determine the amount of variable debt owed to each creditor by geographic location [1].
This data gives us a relatively comprehensive picture of borrowers’ exposure to interest rate risks. Additionally, we can supplement this information with extracts from a sovereign bonds database (Cbonds), where we can access the exact reference rate, and extract bonds with LIBOR/SOFR as the reference rate.
Variable debt is larger for middle-income borrowers
In developing countries, the amount of PPG variable debt has more than doubled since 2008, from $241.7 to $622.6 billion in 2021, corresponding to a rise from 2.4% to 3.9% of GDP respectively (figure 1). However, this significant increase is smaller than the remarkable rise of external debt in recent years, which increased from 10 to 17% of GDP between 2008 and 2021. As a result, while variable debt is on the rise, it has actually declined as a share of external public debt. This is because bonds, which are mostly at a fixed rate, have replaced loans that are often at variable rates.
The borrowing practices vary across regions and income groups. Countries that are eligible for concessional loans usually borrow at a fixed rate. In 2021, LICs had the lowest share of variable debt, which was about 16 percent and has been decreasing since 2019 (figure 2). Those levels are higher for lower-middle-income countries, at about 24,8%. Meanwhile, UMICs have seen a decline in the share of variable PPG debt from 26 percent in 2008 to 20.6% in 2021 [2]. Regional trends also differ. Latin America and the Caribbean (LAC) countries have experienced a persistent decrease in the share of variable PPG since 2009, while Sub-Saharan Africa (SSA) countries have seen a similar decline since 2016. On the other hand, the Middle East and North Africa (MENA) and South Asia regions have experienced sustained growth dynamics over the recent years, reaching respectively 27% and 28%.
What are the possible explanations?
The interest rates for loans are largely determined by the policies of the creditors. Non-concessional lending windows of Multilateral Development Banks (MDBs) tend to shift interest risk to borrowers and they are not the only ones. Bilateral creditors, China in particular, frequently do the same. Since we cannot differentiate between official and private lenders, we only analyse based on geographic origins.
Loans from bilateral and private sources are by far the main driving force behind variable loans in low-income countries (figure 3). The institutions based in China and the Paris Club have an equal volume of variable loans, which amounts to $7.9 billion and $7.4 billion respectively, or about 1.6% of GDP each. Unlike in middle-income countries where multilateral loans represent by far the largest source of variable finance, MDB loans are almost always concessional in LICs, and have fixed rates. On the other hand, in middle-income countries, large Regional Development Banks (RDBs) such as IDB, ADB and AfDB, along with IBRD offer $285 billion worth of variable PPG debt, or 1.8% of middle-income GDP.
Concerns about Increasing Interest Charges on Variable Rates
To estimate the impact of interest rate hikes on developing countries, we need to consider the volumes of loans offered by most MDBs, which are tied to interbank markets. For instance, LIBOR was used in the past, while SOFR is used currently. It is easy to calculate coupons for 2022 and 2023 based on these rates. However, for Chinese loans, there is more variability because, according to a recent AidData report, most loans were priced relative to the LIBOR until 2017, after which there was a shift to the Shanghai Interbank Offered Rate (SHIBOR). Given the high volumes of loans borrowed between 2013 and 2017, it is reasonable to assume that most existing loans are still tied to LIBOR.
Furthermore, to consolidate our estimates, we used the financial statements of the IBRD and the three main RDBs (ADB, IDB, and AfDB) to obtain the exact share of floating-rate loans. As explained in more detail in the last section, loans with a floating rate can be fixed at the request of the debtor country through loan conversions or interest rate swaps. In other words, the initial amount of floating-rate loans varies over time. However, in the case of IBRD, IDB, and ADB, IDS data points only take into account the initial proportion of floating-rate loans, and do not consider the above-mentioned fixing options [3]. Therefore, we lowered the share of floating-rate loans for 2021 reported in IDS based on the FY 2022 financial statements of MDBs.
The expected costs were likely substantial. SOFR averaged about 4.9% in 2023, and was worth 5.3% in December. As a result, floating-rate loans could increase debt service by a total of $29 billion (figure 4), of which $17.1 billion are accounted for by MDBs and about $9.5 billion by bilateral and private institutions.
Indeed, MDB loans are also becoming more expensive: for instance, interests on IBRD loans have risen to a range between 6 and 8%. The G20 Independent Export Group report feared that net transfers would be close to zero in 2023, with interest costs jumping from $9 billion in 2021 to $35 billion in 2023. Without considering hedging practices, our estimates would be similar. However, when taking them into account, net inflows from MDBs still fall significantly, but remain positive, by $10 billion. The hedging practices of MDBs are examined in the next section.
A useful comparison point is the projected interest cost for 2023. How much do variable interest loans exacerbate the situation? Our estimates suggest that the interest charge will increase by 20% in LICs, by approximately 30% for LMICs, and by slightly more than 25% for UMICs (figure 5). Those numbers are large: the amount of additional interest was substantial last year. Those have likely exacerbated the liquidity difficulties in countries most affected.
What countries are the most affected ?
As expected, large countries bear the highest absolute cost. India will face a loss of $3.4 billion, followed by Brazil at $1.9 billion and Mexico at $1.7 billion.
When considered as a percentage of their GDP, the situation becomes more complicated, with the most impacted countries incurring approximately 0.6-0.8% of their GDP as a result of the crisis. This countries with opaque debt structures such as Congo-Brazza and Angola, and also countries where debt risks have risen significantly, such as Tunisia and Jamaica.
For at least 10 countries, additional interest costs have added to expected debt service payments by more than 20%. However, it is important to note that this figure does not indicate an imminent risk. It representes a drag on liquidity, but mot countries with large variable interest also have a relatively prudent debt service structure, where MDB loans predominate relative to other less concessional borrowings, such as bonds.
In summary, large emerging markets (EMs) are likely to be significantly affected in absolute terms. For instance, Brazil and Mexico may face negative flows. However, only a few countries will be significantly impacted in terms of debt sustainability. In most cases, the significant impact will be relative to existing obligations only if countries were relatively prudent. Additionally, this reflects a natural hedge where countries with both bond and IBRD exposure blend fixed and flexible rates, in line with debt managers' risk preferences. Overall, the point made by the IEG that new net positive flows will fall is true. For instance, IBRD in fiscal year 2023 (covering June 2022 – June 2023) showed a clear decline in net flows from $12 billion to $4.5 billion, largely due to $5 billion in additional interest paid by countries.
Hedging practices across MDBs
The large share of MDBs in these results raises a crucial question: what are their policies to alleviate the interest rate risks for countries? This question is more difficult to investigate than it seems, and data recording and publishing practices do not seem concordant across MDBs.
It is clear that risk management offices within MDBs are aware of the risks borne by borrowers. They have established tools to manage these but what is less clear, is how much these tools are effectively used in practice.
-
The World Bank provides loans to countries through two main lending tools - the International Development Association (IDA) and the International Bank for Reconstruction and Development (IBRD). While IDA loans have fixed terms, IBRD offers Flexible Loans (IFLs) as its primary lending tool. These loans come with several options to manage interest rate risk, including conversions and swaps. Conversions allow changes in loan currency or interest rate basis. Swaps involve exchanging cash flows, such as shifts between fixed and floating reference rates, interest rate caps, collars, and automatic rate fixings. The IBRD 2023 financial report reveals that around $203 billion of its $243 billion portfolio is referenced to variable rates. This means that $40 billion worth of floating loans have been fixed over time, which accounts for 17% of the total. Additionally, the World Bank acts as an intermediary for swap operations for loans that are not on its books, an activity covering about $8 billion in 2023 (p.49). It is worth noting that these fixings do not seem to be taken into account in IDS, which reports 100% variable rates from IBRD. Therefore, we uniformly reduce all variable debt outstanding obtained from IBRD by 17% (as we have not found a more detailed country-by-country source).
-
Regional development banks differ from the World Bank in that they group concessional and non-concessional loans together, and there are no common reporting standards for hedging practices.
- The Asian Development Bank’s main product is the Flexible Loan Product (FLP), designed to cater to borrower demands for customised loan products that effectively address project needs and manage external debt. Under this framework, ADB offers sovereign borrowers various options to mitigate interest rate and exchange rate risks. Borrowers can request principal amount conversions, including changes to standard or nonstandard currencies, alterations to loan currency, and interest rate conversions (floating to fixed or vice versa). ADB reported one interest rate conversion totaling $107 million in the first half of 2023 (p.11), indicating a relatively low use of these conversions. In addition, concessional fixed loans are offered by the same entity, and thus grouped in reporting to IDS. Based on table 8 of the financial report (p.13), we estimate that $8.6 billion are converted from variable to fixed, or about 9% of the total.
- The Interamerican Development Bank also offers a flexible loan arrangement, the Flexible Financing Facility, with options similar to those provided by IBRD and ADB. Table 11 (p.29) provides a detailed breakdown of outstanding loans, from which hedging can be deduced (with some assumptions on concessional and non-sovereign components). Taking these numbers into account, we obtain an effective floating share equal to 51.7 %.
- As the others, the African Development Bank ‘s standard non-concessional loan is floating with conversion options, under the Fully Flexible Loans (FFLs) framework. It also has a concessional arm, the African Development Fund, whose financial reports is separate, but whose loans are grouped with AfDB in IDS reporting. ADF's outstanding loans, which are all fixed-rate, amounted to $19.6 billion (p.139) at the end of 2021 , while the total amount of outstanding loans of the AfDB for the same period reached $28 billion (p.78). In addition, about $4 billion are hedged (p.39 of the financial report), which is approximately 15%. These numbers do not quite match what we find in IDS (where 38% is considered as based on variable rates), so we leave them unchanged.
The issue of inconsistent reporting aside, the main question remains: Why do multilateral development banks (MDBs) transfer the interest rate risk to their clients? The goal is to manage the risks of assets and liabilities using derivatives, ultimately ensuring that both are only dependent on the short-term interest rate. This decision is driven by the maturity mismatches between the assets and liabilities of MDBs. It is often difficult for MDBs to match the maturity of their loans in their liabilities, which exposes them to interest rate risk. However, it is worth considering if there are better ways to protect sovereigns from interest rate fluctuations, especially in a world where interest rates have risen significantly above zero.
Conclusion:
Developing countries are vulnerable to unpredictable changes in interest rates due to a combination of bilateral and multilateral loans. Additionally, loans from the International Monetary Fund (IMF) under General Resources Account (GRA) and Special Drawing Rights (SDR) liabilities, as well as the Resilience and Sustainability Trust for countries above a threshold (not considered in this context) add to the overall shock of debt service. On average, this amounts to 9% of aggregate debt service.
Is it necessary to have better hedging tools in place?
The answer is most likely yes, but it depends on whether debt management offices can effectively use them. In the past, countries could have reduced their exposure by fixing interest rates when they were low. However, fixing rates now would be expensive. Despite this, there is still value in hedging against possible future shocks that could reverse current trends. Additionally, it would be helpful if MDBs, which are transparent about their loan terms, could provide more consistent information on the supply of hedging instruments and their use by sovereigns.
Footnotes
- [1] For instance, we can know how much Benin owes German institutions, and what share of those loans is variable, but not whether this loan originates from a private German bank or from an official lender.
- [2] The income categories are built using the 2021 income categories and then are used for the entire period 2008-2021.
- [3] In the case of the AfDB, we have not been able the amount of variable PPG debt reported in IDS numbers.