2. Literature review

While there is extensive literature on the relationship between debt levels and economic development (Reinhart and Rogoff (Reference Reinhart and Rogoff2010), Eberhardt and Presbitero (Reference Eberhardt and Presbitero2015), Woo and Kumar (Reference Woo and Kumar2015)), relatively few studies have examined the role that fossil fuel abundance may play in shaping this relationship. Fossil resources may stimulate macroeconomic growth through the extraction sector and resource exports, while local combustion and use provide competitive energy. Fossil resource abundance is thus a source of economic wealth. Another option is to preserve underground resources for future use. Finally, abundance constitutes a form of implicit collateral that could lower borrowing costs. Underinvestment in the energy transition or overinvestment in the fossil fuel sectors may reinforce the role of abundance, as suggested by the carbon curse assumption (see Friedrichs and Inderwildi (Reference Friedrichs and Inderwildi2013), Chiroleu-Assouline et al. (Reference Chiroleu-Assouline, Fodha and Kirat2020)).

On the other hand, fossil fuel abundance raises concerns about transition risks, particularly losses associated with stranded assets. Climate policies are being implemented (Paris Agreement (2015) or COP28 agreement in Dubai (2023)) and are accelerating the phase-out of fossil fuels, making resource-rich countries more vulnerable. Investors may therefore anticipate a long-term economic slowdown in these countries, making their ability to repay debt more uncertain. The most vulnerable economies could be severely affected, especially in the event of external shocks (e.g., natural disasters, epidemics). These countries will no longer be able to finance their mitigation and adaptation strategies, further exacerbating their vulnerability. As a result, investors may demand a carbon risk premium, increasing borrowing costs.

These relationships between climate change, fossil fuels, and public finances have been addressed by three strands of literature. The first strand studies the impact of natural disasters on macroeconomic aggregates and public finances. Extreme events are likely to impact economic growth, productivity, and financial asset values. The literature shows that since the 1990s, a series of natural disasters have caused substantial economic losses. These losses may explain a decrease in public spending for environmental protection: a fall in GDP implies a decrease in public revenues and public spending. Climate change increases the frequency and the intensity of extreme weather events. There are direct and indirect economic impacts of natural disasters. Direct impacts refer to the damage to assets (e.g., property), with the losses occurring at the time of the disaster or shortly thereafter (like the destruction of houses, productive capital, infrastructure, crops, livestock...). The indirect impacts include interruptions of economic activities as well as any positive spillover effects due to the substitution of production (for example, spending on essential goods and health rather than R&D) and the demand for reconstruction. These indirect impacts capture the short- and long-term economic losses in economic production and consumption and any related economic recovery paths (see Fang et al. (Reference Fang, Lau, Lu, Wu and Zhu2018), Loayza et al. (Reference Loayza, Olaberria, Rigolini and Christiaensen2012), Botzen et al. (Reference Botzen, Deschenes and Sanders2019)).

The second branch examines the bilateral interactions between climate policies and public financing capacities. The poorest and most exposed countries cannot implement policies to adapt to climate change, especially after COVID-19 when public finances deteriorated further. On the other hand, environmental taxes raise revenue while discouraging greenhouse gas emissions. Costs associated with climate change could compromise the ability of some countries to repay their debt. As Dibley et al. (Reference Dibley, Wetzer and Hepburn2021) point out, public debt is legitimate as long as it finances an investment whose returns offset the debt burden. However, climate change threatens economic growth and therefore weakens the countries affected by extreme events. This risk is all the more important as the health context in 2020 and 2021 has forced most countries to increase their public debt. The vulnerability of these countries is growing and may lead the most fragile and indebted to default and enter into a sovereign debt crisis. Faced with this debt risk, which could be exacerbated by the consequences of climate change, investors will be tempted to require a higher risk premium, which would lead borrowers to offer higher interest payments on their debt. Beirne et al. (Reference Beirne, Renzhi and Volz2021) show how public borrowing costs can be affected by climate change. The main channels are the decline in capital, the fiscal consequences of natural disasters, and government spending related to adaptation and mitigation needs. They estimate the impact of climate risks on bond yields and find increased vulnerability and lower resilience to climate risks lead to higher bond yields. Cantelmo et al. (Reference Cantelmo, Melina and Papageorgiou2023) observe that, between 1998 and 2017, on average, during months when natural disasters occurred in Jamaica, the interest rate paid on government debt increased by 3.15 percent. Mallucci (Reference Mallucci2022) studies how natural disasters can exacerbate fiscal vulnerabilities and imply sovereign defaults, for seven Caribbean countries frequently hit by hurricanes. He shows that disaster risk reduces the government’s ability to issue debt and that climate change further restricts government access to financial markets. Furthermore, he predicts that in Caribbean countries, if the frequency and intensity of hurricanes increase as expected, credit spreads will increase by more than 30 percent. This result is also supported by Kling et al. (Reference Kling, Lo, Murinde and Volz2025) who estimate that countries vulnerable to natural disasters pay, on average, a 1.17 percent higher cost of debt compared to countries less exposed to climatic events.

The risk of default on public debt can be explained by the fact that financial markets currently do not fully take climate change risks into account in their measures of the risks associated with financial contracts. Loans to exposed countries for highly polluting projects may be made, while conversely, sustainable investment strategies are discarded (Monasterolo (Reference Monasterolo2020)). More generally, the mispricing of climate risks could lead to systemic risk and financial instability. For these reasons, it seems important that governments systematically assess their exposure to climate risks and disseminate these assessments and risks to economic actors. Only Ghana has undertaken this risk assessment to secure borrowing in response to the COVID-19 crisis (Dibley et al. (Reference Dibley, Wetzer and Hepburn2021)). This information requirement would inevitably result in higher borrowing costs for vulnerable countries, creating a clear disincentive to disclose such information.

The third branch analyses the impact of climate change and natural resources on public and private debt. Natural resources have a direct impact on public debt through the income that they generate. Ampofo et al. (Reference Ampofo, Jinhua, Bosah, Ayimadu and Senadzo2021) show that resource-rich countries have a bidirectional relationship between resource rents and public debt accumulation. In the short run, an increase in resource revenues reduces public debt, but in the long run, dependence on natural resources can increase debt due to price instability and poor fiscal policies. Similarly, high levels of public debt can influence the management of natural resources, in particular by encouraging governments to intensify extraction in order to finance debt servicing. This mechanism has been characterized as the debt-resource hypothesis (Neumayer (Reference Neumayer2005)). In the case of Ghana, Alhassan and Kwakwa (Reference Alhassan and Kwakwa2023) show that increased extraction of natural resources to meet budgetary needs has a negative impact on the environment, which could threaten long-term economic sustainability. Clootens and Magris (Reference Clootens and Magris2024) show that countries with abundant fossil resources benefit from favorable borrowing conditions when resource prices are high. However, when prices fall, investors perceive increased risk, leading to a rise in public debt costs.

These interactions become more complex as climate change affects the value of natural resources. The mechanisms at work are linked to transition risks and stranded assets. The rise of renewable energy and the decline in demand for fossil fuels affect the long-term value of fossil resources. This uncertainty can increase the risk premium associated with the sovereign debt of countries dependent on fossil fuels (Humphreys and Sandbu (Reference Humphreys and Sandbu2007)). This is clearly the case for carbon-intensive publicly owned assets, such as coal mines or energy utilities in several EU countries. These could become stranded assets, as they are no longer economically viable when carbon prices rise. An estimated $12 trillion of assets could become stranded by 2,050, equivalent to 3% of the capital stock (Banque de France (Reference Banque de France2019)).

A more recent literature has focused on the impact of these risks on the assets and debts of private firms. A consensus is emerging that companies bear an additional carbon cost when their efforts to adapt to climate change are weak. Bolton and Kacperczyk (Reference Bolton and Kacperczyk2023) examine how carbon emission levels and trends affect stock market returns on a global scale. The study examines how factors such as economic development, energy structure, and institutions influence the relationship between carbon emissions and stock market returns. They show that carbon emission levels and their growth are significantly correlated with stock market returns. Firms in high-emission countries and high-emission firms both tend to exhibit lower stock market returns.

Our paper lies at the intersection of these three strands of literature and complements previous findings by focusing on public debt at a large scale. We therefore use original data on fossil resources and the cost of public debt, incorporating a broader geographical and temporal dimension. Unlike most of the literature, which primarily explores the impact of public debt costs and tax revenue needs on fossil resources (Ampofo et al. (Reference Ampofo, Jinhua, Bosah, Ayimadu and Senadzo2021), Alhassan and Kwakwa (Reference Alhassan and Kwakwa2023), Neumayer (Reference Neumayer2005)), our focus is on the reverse causality. Nevertheless, the theoretical work of Clootens and Magris (Reference Clootens and Magris2024) suggests that a decline in the value of fossil resources, potentially rendering them stranded assets, should lead to an increase in the cost of public debt. To our knowledge, our study is the first to empirically test this relationship, and our findings reveal a nuanced effect. We find that financial markets play an ambiguous role in the face of fossil resources. Indeed, while the abundance of fossil resources reduces the cost of government borrowing, the level of CO2 emissions per $ of GDP increases this cost, which seems contradictory. Yet, our results suggest a clear conclusion: fossil fuels should be preserved. We argue that fossil resources are important for debt sustainability. Therefore, countries should focus on exploring and conserving fossil reserves rather than depleting them, as this strategy can help resource-rich nations avoid an increase in public debt costs.

3. Data

3.2 Descriptive analysis

The main descriptive statistics are presented in Table 1 and Figures 2 to 5. They highlight broad disparities between countries in economic and financial variables, natural resource endowments, and exposure to low-carbon transition risks. The level of subsoil wealth is illustrative of these differences; some countries lack natural resources while others are extremely well-endowed.

Table 1.

Summary statistics

Figure 2 provides an overview of the average government bond yields across a subsample of 72 countries,Footnote ⁷ revealing considerable heterogeneity among national economies. The average government bond yield over the period in Venezuela (17.97%) or Ghana (18.68%) is nearly seventeen times higher than that of Japan (1.09%). This variability reflects underlying differences in economic fundamentals and sovereign risk profiles. Countries distinguished by prudent fiscal policies, low debt-to-GDP ratios, and robust institutional frameworks tend to exhibit markedly lower yields. Exemplifying this trend are nations such as Japan (even with high levels of debt), Switzerland, Denmark, and Luxembourg, whose government bond yields remain exceptionally low. These yields highlight their status as safe-haven assets, underpinned by strong financial credibility and minimal perceived risk. In contrast, countries such as Venezuela, Pakistan, South Africa, Ghana, and Kyrgyzstan exhibit substantially higher yields, often exceeding 10%, indicative of heightened sovereign risk and macroeconomic instability.

Figure 2.

Average yields on government bonds by country (1995–2019)

Figure 3 provides a comparative analysis of the average treasury bill rates across a subsample of 97 countries, including a notably larger proportion of developing nations than Figure 2. This disparity arises from the more significant constraints faced by developing countries in accessing long-term debt markets, compelling them to rely more heavily on short-term borrowing. The observed rates exhibit considerable variation, ranging from near 0% to over 25%, reflecting pronounced economic and financial disparities among the countries in the sample. Four distinct groups can be identified: (i) countries such as Japan and Saudi Arabia, which exhibit exceptionally low rates, below 1%; (ii) European countries like Greece, Italy, and Spain, which display relatively low rates between 2% and 5%, indicative of moderate but persistent sovereign risks linked to high debt levels; (iii) a diverse group of fifty countries with moderate rates ranging from 5% to 15%, predominantly comprised of developing economies; and (iv) countries with high rates, largely represented by the least developed nations, such as Zambia, Malawi, and Laos, where rates exceed 20%. These elevated rates reflect substantial risk premiums, often tied to macroeconomic instability, high inflation, or political uncertainty.

Figure 3.

Average rates on treasury bills by country (1995–2019).

Figure 4.

Average debt to GDP ratios by country (1995–2019).

Figure 4 illustrates the average debt-to-GDP ratios for the countries in our sample, calculated over the period from 1995 to 2019. The figure highlights the substantial heterogeneity in public debt levels relative to the size of national economies. The debt-to-GDP ratios vary considerably, ranging from approximately 10% to over 140%, reflecting pronounced disparities in public debt management practices across nations. Of particular interest is the contrast between resource-rich nations and those with limited access to such resources. For example, economies like Russia and Saudi Arabia report some of the lowest debt-to-GDP ratios, while countries such as Lebanon, Japan, and Greece exhibit particularly high average ratios. Meanwhile, countries like France, Spain, and Austria occupy an intermediate position within this spectrum.

Figure 5 illustrates the average carbon intensity of GDP across 121 countries in our sample, measured in kilograms of CO2 emissions per U.S. dollar of GDP. The data reveal significant variability in carbon intensity, ranging from less than 0.1 to over 1.5 kilograms per dollar of GDP. The countries with the highest carbon intensities include Mongolia, Moldova, South Africa, and Russia.

A closer analysis of Figure 5 suggests that carbon intensity is at least partially associated with the abundance of fossil resources, reflecting the carbon curse phenomenon. This relationship appears particularly pronounced in developing and emerging economies. Notably, eleven of the twenty countries with the highest carbon intensities are also among the twenty most resource-rich nations in fossil fuels, including Russia, Saudi Arabia, South Africa, Algeria, Libya, India, Iraq, Azerbaijan, Malaysia, Vietnam, and Egypt.

Figure 5.

Average carbon intensity by country (1995–2019).

Before moving on to the empirical estimation, we present several correlations between the key variables to better understand the relationship between sovereign debt costs and transition risks. Figure 6 shows that, when country fixed effects (unobserved heterogeneity) are considered, the relationship between government bond yields and oil abundance (represented by the different colored dots) is negative, pointing out that oil-abundant countries pay lower interest rates on their sovereign debts (even if the overall relationship is positive, as represented by the red line).

In the same line of thought, Figure 7 shows a negative and statistically significant relationship between government bond yields and the share of renewable energy in total energy consumption. Hence, the greater the share of renewable energy in a country’s energy mix, the lower its debt cost.

Figure 8 shows that higher CO2 intensity of GDP is associated with more expensive debt or higher interest rates. These preliminary correlations suggest that financial markets reward fossil resource abundance (as a measure of solvency) while penalizing their use (to meet the net-zero emissions target).

Figure 6.

Government debt costs and oil abundance.

The next section empirically tests these relationships.

Figure 7.

Government debt costs and renewable energy consumption.

Figure 8.

Government debt costs and carbon intensity of GDP.

4. The empirical model

We aim to investigate the impact of low-carbon transition risks on sovereign debt costs. To this end, we extend a standard model from the macroeconomic and financial literature that examines the determinants of sovereign bond yields (Edwards (Reference Edwards1984), Edwards (Reference Edwards1986), Beirne and Fratzscher (Reference Beirne and Fratzscher2013)) by incorporating proxy variables that capture exposure to low-carbon transition risks as additional explanatory variables. In this framework, the cost of borrowing for countries serves as the dependent variable, while the explanatory variables include the most commonly used macroeconomic determinants in the literature, such as GDP growth, fiscal balance, the debt-to-GDP ratio, inflation rate, resource rents as a share of GDP, exchange rate, and institutional quality.Footnote ⁸ The choice of control variables for the pricing of sovereign risk is crucial. The theoretical literature (Eaton and Gersovitz (Reference Eaton and Gersovitz1980), Eaton and Gersovitz (Reference Eaton and Gersovitz1981), Arellano (Reference Arellano2008), Eichengreen and Hausmann (Reference Eichengreen and Hausmann1999)) identifies and supports a wide range of determinants of the cost of public borrowing and sovereign risk, while empirical studies (Edwards (Reference Edwards1984), Edwards (Reference Edwards1986), Eaton and Gersovitz (Reference Eaton and Gersovitz1980), Eichengreen and Mody (Reference Eichengreen and Mody2000), Min (Reference Min1998), Beck (Reference Beck2001), and Ferrucci (Reference Ferrucci2003)) test their relevance. Most empirical studies converge on the relevance of variables such as the level of indebtedness, economic growth, the inflation rate, the exchange rate, and the fiscal balance as key determinants of sovereign risk across contexts. Indeed, the debt-to-GDP ratio reflects a country’s level of indebtedness and is widely used as a key indicator of its solvency. An increase in this ratio signals a deterioration in solvency, which heightens the perceived risk of default and, consequently, leads to higher borrowing costs. The inflation rate is expected to exert a positive effect on the cost of sovereign borrowing. Rising inflation prompts lenders to require higher nominal interest rates to offset the anticipated loss of purchasing power, thereby increasing borrowing costs. In addition, high or volatile inflation heightens uncertainty about future price levels, which raises the inflation risk premium embedded in interest rates. Persistent inflation may also signal weak monetary or fiscal discipline, leading investors to perceive greater sovereign risk and demand higher yields. Furthermore, elevated inflation can reflect broader macroeconomic imbalances, such as an increased likelihood of a balance of payments crisis, reinforcing perceptions of economic instability and a higher probability of default, which in turn raises the risk premium on sovereign debt. High growth rates are expected to reduce the cost of borrowing as sustained economic growth strengthens fiscal capacity by increasing tax revenues and improving budget balances. This reduces sovereign risk premiums and borrowing costs. Fiscal balance is an indicator of a government’s fiscal discipline. A surplus reduces the risk of default, while a deficit must be financed through additional borrowing, thereby weakening the country’s solvency. Exchange rate movements affect sovereign borrowing costs primarily through valuation effects on foreign-currency-denominated debt and through risk premiums, as exchange rate depreciation raises debt servicing costs and increases perceived default risk, thereby increasing borrowing costs (Eichengreen and Hausmann (Reference Eichengreen and Hausmann1999)). Our baseline panel regression model is specified as follows:

(1) \begin{equation} Y_{it}=\alpha +\beta R_{it}+\gamma Z_{it}+\delta _{i}+\lambda _{t}+\varepsilon _{it} \end{equation}

where $i$ refers to the country and $t$ to the time period. $Y$ refers to the borrowing cost. This variable is proxied in the macroeconomic literature as the sovereign bond yield.Footnote ⁹ In this article, we proxy the cost of borrowing by two alternative measures: the 10-year sovereign bond yield and the treasury bill rate, which respectively capture long-term and short-term costs of sovereign borrowing. $R_{it}$ represents the exposure to low-carbon transition risk faced by country $i$ in year $t$ . We proxy this latter variable using various measures among subsoil resource wealth, CO2 intensity of GDP, and the share of renewables in total energy consumption. We first introduce these variables individually with the controls, then we combine them to see how they simultaneously affect debt costs. Subsoil resource wealth encompasses several variables that allow us to distinguish between fossil resources, oil resources, natural gas resources, and coal resources. The inclusion of these subsoil resource wealth variables allows us to test whether different fossil resources can be considered stranded assets or, conversely, whether they continue to be valued by financial markets. All subsoil resource wealth variables are expressed in logarithms in the estimation framework. The other proxy variables for exposure to low-carbon transition risks will help explore the existence of a risk premium linked to the carbon intensity of GDP and/or a reward associated with the deployment of renewable energies. $Z$ is a matrix of control variables that includes the debt-to-GDP ratio, the inflation rate, GDP growth to control for the macroeconomic environment, the budget balance, exchange rates, the resource rent-to-GDP ratio, and institutional quality. $\gamma$ represents a vector of parameters associated with control variables. In panel data models, we control for unobserved heterogeneity using fixed or random effects, with the $\delta _{i}$ terms capturing unobserved country-specific characteristics. We favor a fixed-effects specification for two main reasons. First, unobserved country-specific factors are likely to be correlated with the explanatory variables included in equation (1), making the random-effects GLS estimator inconsistent, whereas the fixed-effects within estimator remains consistent. Second, a Hausman test rejects the null hypothesis of no systematic difference between the two estimators, in favor of the fixed-effects specification.Footnote ¹⁰ $\lambda _{t}$ denotes a full set of time-varying fixed effects serving to capture exogenous shocks that could affect all countries of our sample in the same way, and $\varepsilon _{it}$ is the error term.