Priced Out

This page describes the methodology and data used for the Priced Out report produced by ONE Data for the Development Finance Observatory.

This analysis examines how the cost of external public borrowing changed between 2020 and 2024 for low- and lower-middle-income countries, as well as other developing countries with exposure to multilateral and market finance. The report focuses on two related questions:

How have average interest rates changed across different creditor groups?
How much do repayment burdens differ once we account for maturity, grace periods, and repayment structure rather than interest rates alone?

To answer these questions, we combine creditor-level borrowing terms from the World Bank’s International Debt Statistics (IDS) with a present-value framework that compares the repayment cost of borrowing from different sources.

Data

This report primarily relies on the World Bank’s International Debt Statistics (IDS) database.

The core IDS indicators used in this analysis are:

Indicator	Unit	Code
Interest rate	%	`DT.INR.DPPG`
Maturity	years	`DT.MAT.DPPG`
Grace period	years	`DT.GPA.DPPG`
Interest payments, bilateral	current US$	`DT.INT.BLAT.CD`
Interest payments, multilateral	current US$	`DT.INT.MLAT.CD`
Interest payments, bondholders	current US$	`DT.INT.PBND.CD`
Interest payments, commercial banks	current US$	`DT.INT.PCBK.CD`
Interest payments, other private	current US$	`DT.INT.PROP.CD`
Commitments, bilateral	current US$	`DT.COM.BLAT.CD`
Commitments, multilateral	current US$	`DT.COM.MLAT.CD`
Commitments, private	current US$	`DT.COM.PRVT.CD`

For the present-value analysis, we also use market-equivalent sovereign borrowing rates from three sources:

IDS bondholder interest rates on new commitments
Secondary-market yields on outstanding sovereign bonds from CBONDS
Rating-implied sovereign spreads added to the US Treasury 7-year yield (calendar-year average of daily observations)

Country groups

To understand how the high-rate environment affects countries differently, we divide borrowers into three groups using World Bank IDA eligibility and borrowing status.

IDA-only – These are IDA-eligible countries that do not have access to IBRD borrowing. In practice, they rely primarily on concessional finance and have little or no direct access to international bond markets. IBRD-only – These are developing countries that are not IDA-eligible and are classified as creditworthy for IBRD borrowing. They tend to have broader access to non-concessional multilateral lending, bilateral finance, and market borrowing. Blend – This group includes IDA-eligible countries that are also creditworthy for some IBRD borrowing and countries borrowing on harder blend credit terms within IDA. These countries sit between full concessional protection and full market access. In the report, they are the group most exposed to the interaction between rising multilateral rates and tighter market conditions.

Creditor groups

We compare borrowing terms across five creditors:

World Bank-IDA and World Bank-IBRD as multilateral creditors. Paris Club (Australia, Austria, Belgium, Brazil, Canada, Denmark, Finland, France, Germany, Ireland, Israel, Italy, Japan, Netherlands, Norway, Russia, South Korea, Spain, Sweden, Switzerland, the United Kingdom, and the United States) as bilateral creditor. China, for which we consider both private and bilateral creditors, due to Chinese state-linked institutions being recorded under both categories. Bondholders as private creditors.

Data transformation

The IDS data is reshaped so that each row corresponds to a year–debtor–creditor combination, with each indicator stored in its own column.

To ensure each creditor group only accumulates the correct series, we use two mappings:

a mapping that defines which lending type or types each creditor group covers; and
a mapping that assigns each IDS indicator code to a lending type.

This matters because some counterpart groups span more than one lending type. For example, China is treated as a combined bilateral/private counterpart in the source processing, while multilateral institutions only use multilateral codes.

For each group, only indicators whose type matches that group’s declared lending type are summed. For example, China totals only include the relevant bilateral and private codes, whereas multilateral rows only include multilateral codes.

After this step, the data is filtered either by African countries or by the country groups used in the report.

Aggregation and weighting

To compare borrowing terms across groups, we compute commitment-weighted averages.

For each year–debtor–creditor combination, commitments are used as weights to calculate average interest rates. This ensures that larger loans contribute more to the average than smaller ones.

These country-level aggregates are then used to compute borrower-group averages for each creditor in 2020 and 2024.

Interest rate comparison

The first part of the analysis compares average interest rates faced by each borrower group across the five creditors listed above.

This allows us to trace how borrowing conditions changed between 2020 and 2024 across:

concessional multilateral lending;
non-concessional multilateral lending;
bilateral official finance;
and market-based borrowing.

These averages should be read as group-level summaries rather than the exact terms available to every country in a category. There is often substantial variation beneath the average, especially for private creditors. However, interest rates only tell part of the story.

Low and middle-income countries borrow from creditors with very different terms. Interest rates do not provide a meaningful comparison across creditor types. A low interest rate with a short maturity may impose a heavier burden than a higher rate spread over decades. Grace periods, maturities, interest rates and repayment profiles all shape the true cost to the borrower.

The ‘present value’ and ‘grant element’ of a loan are useful concepts to condense this complexity into single numbers. The grant element is simply the difference between the face value and the present value of future debt service (with some assumptions), expressed as a percentage of face value.¹

The IDS grant equivalent series computes the grant element using a conventional 10% discount rate,² and the IMF/World Bank operational framework uses a fixed 5% rate.³ These fixed rates make sense for classification and comparability (i.e. determining if a loan is concessional, enabling cross-country aggregations). But they miss the significant differences in market rates available to different countries.

The Present Value Repayment Ratio

We propose a measure which captures the present value of the contractual debt service (at commitment), discounted at the borrower’s market equivalent rate. This measure is a Present-Value Repayment ratio (PVR), which we can express as a dollar amount (PV per $100 committed). Compared to the grant equivalent, this places the emphasis on the cost to the borrower, rather than on the generosity of the lender.

This measure uses the same mathematical operation as the grant element (discounting future debt service) but applies a country-specific market-equivalent yield as the discount rate, instead of a uniform policy benchmark.

Using the borrower’s market yield as the discount rate allows us to assess how much the borrower would save, in present value, relative to borrowing the same amount in market terms.⁴

For a loan with a face value $FV$, a contracted interest rate $i$, maturity $M$ and grace period $GP$, discounted at the country’s market-equivalent borrowing rate $d$:

\[ PVR = \frac{100}{FV} \sum_{t=1}^{M} \frac{DS_t}{(1+d)^t} \]

Where $DS_t$ is total debt service (principal plus interest) in year $t$, determined by the interest rate $i$, grace period $GP$, and the assumed repayment profile.

A PVR of approximately 100 means that the creditor is lending at terms approximately equivalent to the borrower’s market rate.

A PVR under 100 means that the creditor is lending below the borrower’s market rate. The gap (100 - PVR) is the market-equivalent implicit subsidy per $100 committed.

The relationship between the PVR and the standard grant element (GE) is

\[ PVR = 100 (1-GE) \]

Where the GE is computed using the borrower’s market yield as the discount rate. The two measures contain the same information. The proposed PVR just frames it as a potential cost difference.⁵

It’s important to note that, with publicly available data, we cannot directly calculate the PVR based on all contractual details (principal, interest, service charges, commitment fees, other charges), inflows as disbursements net of upfront fees, and borrower’s market-equivalent discount factors at each tenor. For this we would need loan-by-loan microdata on disbursement schedules, fee structures, repayment modalities and currencies, which is not publicly available.

But we can estimate the PVR with public IDS data as a stylised commitment-basis approximation, which uses weighted average terms (interest rate, maturity, grace period), commitment amounts and a flat country-specific discount rate.⁶

The denominator of the PVR is the commitment amount, not the present value of borrower inflows. This is based on what’s available via IDS, though we note it’s not necessarily ideal since official loans can disburse slowly over several years. For instruments that disburse quickly (bonds, bilateral loans), the difference is small. For project loans with multi-year disbursements (common for MDB lending), the commitment basis underestimates the true PVR because it ignores the time value of delayed disbursements.⁷

The debt service schedule

Given the average terms⁸ of new commitments reported in the IDS, we construct a stylised repayment schedule. The repayment profile assumption varies per creditor type:

For official creditors (bilateral and multilateral) we use equal principal payments as the baseline amortisation profile. This is the dominant repayment structure for most official lending. During the grace period only interest is due (i.e debt service is face value times interest rate). During the repayment period (years after the grace period and up to maturity), equal annual principal payments plus interest on the declining outstanding balance.
For bonds (private creditors) we use bullet repayment. That means periodic coupon payments with full principal repaid at maturity, which is the standard structure for sovereign Eurobond issuances. Annual debt service is the face value times the interest rates for all years before maturity, with the face value due at maturity.

The IDS reports interest rate on the commitments as of the commitment date. For fixed-rate instruments this rate applies through the loan’s life. For floating-rate instruments (e.g IBRD loans priced at SOFR + a margin), the commitment-date rate is a snapshot of the prevailing interest rate but it isn’t a forecast of rates over the loan’s life.⁹

This means that for the bond-vs-IBRD comparisons, our estimates using peak-period IBRD rates are conservative. However, using commitment-date rates as a proxy is standard practice (even if it’s an approximation). The DAC Converged Statistical Reporting Directives and the IMF/World Bank grant element calculators also use single fixed interest rate inputs.

The discount rate

The discount rate is the borrowing country’s market-equivalent sovereign borrowing cost in US dollars. This is the yield it would face (or does face) when borrowing on international capital markets.¹⁰ We take this to be the borrower’s opportunity cost of capital (i.e the rate at which each future dollar of debt is valued).

For this we use a 3-tier cascade of publicly available data, used in order of preference:

IDS bond terms on new commitments. Where IDS reports the average interest rate on new commitments from bondholders for a given country-year, we take this rate as the most direct proxy for the country’s observed market borrowing cost.¹¹
Secondary market yields on outstanding bonds (CBONDS data). For countries that did not issue new bonds in a given year but have outstanding bonds trading on secondary markets, we use end-of-year indicative yield-to-maturity (YTM) data from CBONDS. For each country, we compute an issue-amount weighted average YTM across all outstanding hard currency sovereign bonds. This provides a direct, market-observed measure of what investors demand to hold that country’s debt, even in years without new issuances.¹²
Implied yield from sovereign credit ratings (Damodaran data). For countries with no bond market data in either IDS or CBONDS (smaller borrowers, prolonged default with no trading activity), we construct a synthetic yield using Aswath Damodaran’s annual updated mapping of Moody’s sovereign credit ratings to default spreads.¹³ Each rating is associated with a typical default spread in basis points, derived from CDS markets and corporate bond data. We add this spread to the US Treasury 7-year yield (which we take from FRED) to produce an implied borrowing rate.¹⁴

This includes two important simplifications, both driven by data constraints. First, we’re using a flat discount rate across maturities instead of a term structure that would discount each future payment at its own tenor-specific rate. Second, we discount all instruments in USD terms instead of doing it in each loan’s denomination currency. Both simplifications align with standard grant element frameworks from the IMF/WB and OECD DAC.

Quantifying the cost of Bond financing vs IBRD

When a country issues a sovereign bond on international capital markets rather than borrowing from a multilateral lender such as the IBRD, how much more does it commit to paying in present value terms? The PVR can provide a direct answer by reducing the comparison to a single calculation.

With the PVR, a newly issued bond priced at par has (by definition) a PVR of exactly 100 when discounted at its own yield. To quantify the excess cost of bond financing we take the commitment-weighted average IBRD lending terms for each year (interest rate, maturity, grace period) and compute the PVR of those terms for each country in the bond-issuing universe, using that country’s own market-equivalent rate as the discount rate. Using stylised average terms rather than country-specific IBRD terms allows the analysis to cover all bond-issuing countries, not just those that borrowed from IBRD in the same year.¹⁵

The excess cost per $100 is then $100 - PVR_{IBRD}$ and the aggregate excess cost for a given bond issuance is the face value of the bond multiplied by $\frac{100 - PVR_{IBRD}}{100}$.

This excess cost is determined by the interaction of 3 gaps:

The interest rate gap as higher bond coupons increase the cash flows of the bonds relative to those of the IBRD loan
The maturity gap as longer IBRD maturities push principal repayments further into the future where they are more heavily discounted by the high market yield
The grace period which defers principal payments into the future where they are also more heavily discounted by the high market yield.

The relative contribution of each factor depends on the discount rate. At low yields the interest rate gap dominates and the maturity/grace period effects are modest. But at high yields, the maturity and grace period effects become larger.

Country-years in active debt restructuring or in severe debt distress (Ghana 2022-24, Grenada 2021-23, Sri Lanka 2020-24, Zambia 2020-24) are excluded from headline aggregates, as are non-standard bond issuances that do not reflect market pricing (e.g. debt-for-nature swaps).

Ultimately this analysis quantifies the unit cost premium of bond financing relative to multilateral terms.¹⁶ It does not claim that the country could have substituted bond issuance with an equivalent volume of MDB lending. IBRD and IDA lending is supply-constrained and cannot always meet developing countries' financing needs. For context, total IBRD commitments to all blend countries averaged approximately $1.7 billion per year over this period, compared to an average of $8.1 billion per year in bond issuance. Countries may also prioritise market access, even at higher rates than equivalent IBRD-lending.

The IDS defines the grant equivalent as a measure of “the overall cost of borrowing”. ↩
The IDS metadata for the grant element series states: "the grant equivalent of a loan is its commitment (present) value, less the discounted present value of its contractual debt service; conventionally, future service payments are discounted at 10 percent." ↩
IMF & World Bank (2013). [Unification of Discount Rates Used in External Debt Analysis for Low-Income Countries](https://www.imf.org/external/np/pp/eng/2013/100413.pdf. The unified discount rate is set at 5% and applies across all IMF/World Bank debt-related calculations for low-income countries, including the Debt Sustainability Framework and concessionality assessments. See also the IMF Grant Element Calculator. ↩
This is not the same as the borrower’s expected fiscal burden under full repayment, nor a measure of lender generosity or effort. Sovereign yields include default risk premia. That means that a country in severe financial distress may have a very high market yield, which mechanically pushes the PVR down (making all lending appear very cheap relative to the market alternative). This doesn’t mean that the debt is inexpensive for the borrower, it just means that the market is comparatively very expensive. ↩
We use a different label (instead of ‘Grant Equivalent’) for two reasons. First, the discount rate is different: standard grant element series in IDS use a fixed 10%, and the IMF/World Bank operational framework uses a fixed 5%, while our measure uses a country-specific market yield. A grant element "of 40%" means something quite different depending on whether it was computed at 5% or at Kenya's 9.75%, and using the same term for both invites confusion. Second, the framing is different: the grant element is conventionally presented as the percentage of the loan that is "gift". Instead of placing the focus on the creditors’ concession, the PVR frames it as “dollars of repayment per $100 of debt committed”, placing the focus on the borrowers’ cost. ↩
Using commitment-date terms is standard institutional practice. The current OECD DAC Converged Statistical Reporting Directives (DCD/DAC(2024)40/FINAL, paragraph 26) define a variable interest rate loan as one where "the margin is fixed at the time of the commitment while the market reference rate is subject to change after the commitment of the loan." The CRS Reporting Checklist (DCD/DAC/STAT(2024)11)) instructs reporters to record the prevailing reference rate at commitment. The grant element is then computed from these commitment-date terms. The IMF/World Bank grant element calculator similarly takes a single fixed interest rate as input. In all cases, variable-rate loans are evaluated at the rate prevailing when committed, not at a forecast of future rates. ↩
The standard IMF/WB grant element calculator also assumes full upfront disbursement, so this simplification is shared institutional practice. ↩
IDS reports commitment-weighted average terms for each creditor-debtor-year. Reconstructing a payment schedule from averages basically produces a representative loan that may not correspond to any actual loan. This reconstruction could diverge in more or less significant terms depending on the heterogeneity of the terms provided by a creditor to a debtor in a given year. ↩
This can create a cycle-dependent bias. When base rates are at peaks, the IDS reported rate overstates the expected cost of floating rate-loans, producing a PVR that is too high, making the IBRD lending look more expensive than it is likely to be in practice. The opposite is true when base rates are low. ↩
We discount at the full yield (instead of spread) because it represents the borrower’s market-equivalent cost of capital. A bond issued at par, when its cash flows are discounted at its own yield, produces a PVR of exactly 100. That means that in the PVR scale any cheaper-than-market instrument produces a PVR below 100. ↩
We use this data when bonds disbursements in IDS exceed US$ 50 million. We also exclude non-standard instruments (debt swaps, sanctions settlements). ↩
A couple of caveats. The secondary market YTM reflects the current market price of existing bonds at their residual maturities, not the yield on hypothetical new issuance. The issue-amount-weighted average blends multiple residual maturities into a measure that is closer to an average of the curve yield than a specific-tenor benchmark. For the purposes of a flat discount rate (which is already an approximation) this is a reasonable proxy for the country’s market borrowing costs. Additionally, the YTM indicator from CBONDS may reflect dealer quotes and not just actual transaction prices (especially for thinly traded bonds). For countries with active secondary markets the difference is small but for less liquid countries the quotes may not be as precise. ↩
There are limitations to this dataset. Damodaran's country risk premiums are derived from Moody's sovereign credit ratings, which can lag actual credit deterioration by 1-2 years. The spread mapping is calibrated primarily from CDS markets and corporate bond data for larger, more liquid sovereigns, and may not extrapolate well to small borrowers with no market presence. Countries in prolonged default or without any sovereign rating fall outside the mapping entirely. For our purposes these limitations do not have a significant impact. We use the Damodaran data only for the 7-12 countries per year where no bond market data exists at all, and only to construct a spread (not to make claims about individual country pricing). The dataset is freely available at Damodaran's NYU page and is widely used in sovereign and corporate valuation. ↩
We use 7 years because it's closer to the average duration of emerging-market sovereign bonds than a 10-year benchmark. ↩
We use this approach for aggregate or cross-country comparisons. For specific case-studies, where data is available, we use actual IBRD terms. ↩
MDB loans carry de facto preferred creditor status. They are excluded from restructuring, and borrowers prioritise their repayment. The country's market yield includes a default risk premium that does not apply in the same way to MDB obligations. Discounting MDB cash flows at the full market yield therefore somewhat overstates their PVR advantage relative to a risk-adjusted comparison. This bias runs opposite to the floating-rate bias, which can understate the MDB advantage. ↩