The impact of climate risk on credit supply to private and public sectors: an empirical analysis of 174 countries

Abstract

In recent years, risk has been increasingly a long-term environmental problem that cannot be underestimated due to its tremendous impacts on various sectors including banking sector. Accordingly, the credit supply to private and public sectors is affected by the increased climate risk. In order to examine the climate risk effect from an international comparison, this paper empirically investigates the impact of climate risk on credit supply by using a sample of 174 countries during 2000–2019 from the perspective of the difference between private and public sectors. The results show that climate risk has a significant negative effect on the credit supply to private sector and a positive effect on that to the public sector. Further, we provide new evidence that the climate risk effect has a more significant effect on the private and public sector credit supply in the high-income countries than that in the low-income countries, suggesting a quick risk contagion in the high-income countries.

Introduction

The failure of climate action, extreme weather, and biodiversity are currently the top three global risks with the highest probability of occurrence and the most severe consequences. Among them, the failure of climate action will be one of the most threatening risks in the next decade. It is evident that climate change will gradually become a tough constraint for global economic development in the future. The potential threat of climate-related financial risk to the stable and healthy development of the financial system is also rising. The climate change is transformed into financial risk by two main channels: first, climate change will increase climate risk, and frequent weather disasters will directly increase the economic loss of financial institutions; second, if climate changes are not effectively controlled, climate risk will indirectly increase the risk of financial institutions by reducing the productivity of the real sector of the economy. As a fundamental financial institution, banks have an important role in regulating macroeconomic leverage and allocating social capital. Climate risk may have impacts on banks’ credit business, leading to significant distorted credit markets and mismatch of financial resources (Garmaise & Moskowitz, 2009). Due to the differences in the objectives, structures, and organizational culture between the private and public sectors, which suffer from climate risks discrepantly, banks may adopt different credit supply strategies for the private and public sectors when facing climate risk. Based on this, we attempt to empirically investigate the impact of climate risk on the supply of bank credit to the private and public sectors from a country-level perspective.

The issue of climate risk has received increasing attention from scholars in recent years, and many studies have begun to analyze the relationship between climate risk and the financial system, including the credit supply (Clarvis et al., 2014; Oguntuase, 2020; Da Mata & Resende, 2020; Christophers et al., 2020; Nwani & Omoke, 2020; Birindelli et al., 2022), the governance of fund markets (Bowman & Minas, 2019; Cui & Huang, 2018; Lu et al., 2022), the insurance institutions’ responsibility (Surminski et al., 2016) and so on. Campiglio et al. (2018) recognize that climate risks cause losses to the balance sheets of financial market players, leading to the repricing of carbon-intensive assets and threatening the stability of financial markets. In the case of the banking sector, as climate change threatens the stable operation of financial markets, the banking sector as an important component of the financial industry is also affected by climate change. Lamperti et al. (2019) analyze the effects of climate change on the stability of the global banking system using an agent-based climate-macroeconomic model and find that climate change will make bank crises more frequent. Dafermos et al. (2018) argue that climate change will first exert the impact on the banking sector, with higher environmental standards, joint and several liability for emissions, reputational risk, and collateral damage increasing banks' climate change-related risks. Caby et al. (2022) demonstrate that climate risk and financial performance for banks in emerging and developed countries are positively correlated over the period 2011–2019.

As the core business of banks, credit has led to more and more studies to analyze the relationship between climate risk and bank credit. More specifically, regarding climate risk and bank credit, scholars focus on the following aspects: first, the impact of climate risk on bank credit pricing. Kleimeier and Viehs (2016) reveal the negative relationship between the cost of bank loans and the level of carbon dioxide emission. High-polluting firms must pay higher costs for their bank loans (Chava, 2014; Javadi & Masum, 2021). Jung et al. (2018) point out that both higher carbon emissions and carbon intensity of firms lead to higher credit costs, and the loan spreads of firms with voluntary carbon emission disclosure are lower than those of non-disclosed firms. Meanwhile, commercial banks' loan values are also depreciated by carbon tax shocks, and state-owned commercial banks are significantly more affected than joint-stock commercial banks (Chava, 2014). Second, the impact of climate risk on credit supply. Climate risks from climate change will inevitably cause imbalances in loan supply and demand, and such shocks may significantly affect macroeconomic activities (Golosov et al., 2014). For example, Berg and Schrader (2012) find that loan demand would increase following unforeseen shocks from climate change, but the proportion of applicants actually receiving loan supply would decrease due to the higher risks involved. Islam and Wheatley (2021) reveal that firms exposed to climate risk tend to use less credit due to the impact of liquidity shocks. Hosono et al. (2016) suggest that in the presence of shocks from extreme weather hazards, banks may shy away from lending under the impact of an extreme weather disaster, resulting in illiquidity in the financing market. In addition, David (2010) suggests that bank lending shyness is pronounced in developing countries after a climate disaster. Faiella and Natoli (2018) show that banks respond to climate risks such as abnormally high temperatures and floods by reducing credit amounts and credit approval rates. Third, the impact of climate risk on banks' credit transformation. Climate risk will prompt banks to launch more green credit projects, which promotes the green transformation of macroeconomy and makes credit resources flow more to low-carbon industries (Sun et al., 2019). Banks’ lending to green companies can reduce risk and create more innovative green projects, and refusing to lend to highly polluting industries can also reduce risk and improve banks' reputation (Fatica et al., 2021; Gangi et al., 2019). Highly profitable commercial banks have a strong awareness of social responsibility and actively develop green credit (Yin et al., 2021). Mésonnier (2022) investigates whether and how banks align green behavior in terms of credit allocation to carbon-intensive industries in France. Banking institutions with a high share of high pollution and emissions in their credit business are given more attention and scrutiny by relevant stakeholders and regulators during external audits and regulatory inspections on the transformation risks of the companies they support (Andersson et al., 2016). In order to comply with the policy requirements, banks have to use the extra credit balance to increase the proportion of green credit by reducing the interest rate of green credit and thus attracting more green loans (Dikau & Volz, 2021). As Bătae et al. (2021), banks often provide green financial products and services, such as green funds, to improve their competitiveness, reputation, customer loyalty, and profitability.

In summary, research on climate risk on the pricing, supply, and transformation of bank credit has been well-established. Based on the previous studies, the objectives of this study can be explained in two folds. Firstly, in the face of climate change, many countries worldwide, from developed to developing countries struggle to reduce the negative impact of climate risks (Abid et al., 2022; Nga, 2022). The technically advanced countries have attained sustainable development goals through green innovation and green finance, such as China, Japan, and South Korea (Tolliver et al., 2021). The developed countries also respond to climate change to promote high productivity low-carbon projects (Hosono et al., 2016; Mésonnier, 2022). At the same time, developing countries become aware of the climate risk and take some actions (Aftab et al., 2022; Dougherty et al., 2020; Khan et al., 2022; Sun et al., 2019). Global climate change has threatened economic growth and financial development in developed and developing countries (Tol, 2009). However, to the best of our knowledge, little academic research adds insight for both countries to study the climate risk and credit supply. Consequently, it is crucial to consider the impact of climate risk on credit supply from an international perspective.

Second, private sector credit as a percentage of GDP is usually employed as the indicator to measure the credit supply (Batabyal & Chowdhury, 2015; Chen & Chang, 2021; Denizer et al., 2000; Levine et al., 2000; Samargandi et al., 2015). As well as private sector credit, we believe that public sector credit is also an important part of the credit supply. Although a lot of literature studies the differences between the private and public sectors (Chan et al., 2011; Goodwin, 2004; Lyons et al., 2006), there is currently less attention on the impact of climate risk on public sector credit, especially the different impacts to private and public sectors. To address this, this paper empirically investigates the impact of climate risk on the private sector and public sector credit supply based on panel data for 174 countries from 2000 to 2019, particularly focusing on the heterogeneous effect of climate risk on the different sectors.

This paper contributes to the literature in three important ways at least. First, it contributes to the growing literature investigating the impact of climate risk on credit supply (Campbell & Slack, 2011; Levine et al., 2018). A common argument is that serious natural disasters will lead to the unbalance of the financial system and impact the banks’ credit service (Brei et al., 2019; Cortés & Strahan, 2017; Ivanov et al., 2020). Expanding this argument, we contribute to this growing literature by presenting that the regular climate risks instead of the extreme climate events have been recognized by the banks and induced the changes in credit supply.

Second, we divide bank credit supply to the private sector and public sector, and study the influences of the climate risks on the credit supply of the private sector and public sector, respectively. Joo and Chung (2014) compare the characteristics of private and public sectors, especially in objective and organizational culture. Considering the impact of climate risks, the private sector and public sector take different responses to climate change (Victor, 2012; Koh & McCarthy, 2018; Barry & Adeyemi, 2022). Based on the previous studies, there may be different mechanisms of the impact of climate risk on credit supply to the private sector and public sector. Therefore, we analyze the impacts of climate risk on the credit supply of private and public sectors and compare the differences.

Third, by providing an international comparison among 174 countries, we consider the heterogeneity of income level and analyze the influences of climate risks on credit supply in high-income countries and low-income countries. It manifests that climate risk in high-income countries has a greater negative impact on private sector credit supply and a greater positive impact on the supply of public sector credit supply.

The rest of this paper is arranged as follows: Sect. 2 is the theoretical analysis and hypothesis; Sect. 3 illustrates the data and methodology; Sect. 4 presents the empirical results; Sect. 5 argues the treatment of endogeneity and conducts the robustness checks. Section 6 discusses the results. Section 7 concludes.

Theoretical analysis and hypotheses

Climate risk affects the economic and financial system through two main channels: physical risk and transition risk (Monnin, 2018; Margherita et al., 2019; Allen et al., 2020; Nehrebecka, 2021). Physical risk is mainly caused by the impact of extreme weather events and long-term changes in climate patterns, while transition risk refers mainly to the uncertainties that may arise from low-carbon socio-economic transitions, such as policy changes and shifts in market preferences (Batten et al., 2020). On the one hand, physical risk and transition risk can directly affect banks' credit business. Sudden climatic events are likely to cause drastic fluctuations in commodity prices, increase uncertainty faced by the macroeconomy, increase market interest rate volatility, increase market risk exposure faced by banks, and reduce banks' willingness to provide credit services (Mulwa et al., 2017), while banks will hoard more liquid assets to cope with potential liquidity risks, thus reducing the supply of credit. Driven by the policy of low-carbon economic transformation, banks' credit strategies will again be subject to intervention from the government. The government will support banks' credit resources to state-owned enterprises (SOEs) by intervening to support SOEs to actively cope with climate risks and promote the low-carbon transformation of the state-owned economy.

On the other hand, physical and transformation risks indirectly impact banks' credit business by affecting the development of the real economy. Extreme climate events of physical risk lead to the shrinkage of a large number of collateral assets of enterprises, causing extensive damage to physical assets such as land and equipment, generating huge construction and maintenance costs, reducing the productivity of the real economy sector through bank financing channels, which in turn affects banks' credit business (Monnin, 2018). The transition to the low-carbon economy will limit the future use of fossil fuel energy, leading to a reduction in revenue capacity for the high-carbon energy sector. At the same time, the implementation of the carbon emissions trading market and carbon tax policies will increase the operating costs faced by the high-carbon sector (Grippa et al., 2019).

Impact of climate risk on private sector credit supply

The private sector refers to business enterprise organizations that are market-regulated and aim to maximize organizational interests. When climate risk increases, the problem of information asymmetry will appear in the private sector. On the one hand, the private sector lacks of good climate risk information disclosure, income certificates, and asset certificates, which makes banks face a higher risk of adverse selection (Cameron et al., 2018; Adhikari & Chalkasra, 2021). On the other hand, when exposed to climate risk, the private sector especially the small private sector is motivated to take more risks (Xu et al., 2022), which makes banks face a greater risk of moral hazards. Banks are usually averse to the uncertainty of information asymmetry; therefore, they have more incentive to overestimate the losses and reduce the credit supply. Hence, increased information asymmetry might lead banks to decrease credit supply to the private sector.

Banks are concerned with the repayment ability of lenders when making credit decisions, while the borrowers’ repayment ability decline because of the decrease in profitability, solvency, and operation capacity (Atta-Mensah, 2016; Dafermos et al., 2018; Huang et al., 2018; Olovsson, 2018). Thus, the profitability, liquidity, and operating costs of lenders become their critical indicators in credit. The climate risk mainly affects the private sector through the following three channels: first, climate risk reduces the revenues and profit of firms (Huang et al., 2018). When climate risks increase, equipment damage and weather restrictions can affect the normal production of the private sector, causing a decline in revenue. Climate change also has an impact on the cost of raw materials. For example, highly-polluted suppliers face strict environmental regulatory policies, which will lead to higher and more volatile raw material costs under supply uncertainty. The decline of revenue and the increase of costs impair the profitability of the private sector (Sakhel, 2017; Natalia, 2021); second, when climate risk increases, a large number of collateral assets of private sector enterprises shrink, resulting in the depreciation of collateral such as land and equipment, and the reduction of collateral asset value. The reduction in the value of collateral assets means that the borrower's solvency decreases (Graff Zivin & Neidell, 2014); third, when climate risk increases, higher carbon emission standards, and environmental requirements will seriously increase the operating costs. The private sector spends more costs on adjusting their production activities to meet environmental standards, such as replacing equipment, using clean energy, and applying carbon taxes. Climate change leads to higher operation and maintenance costs due to the shortened life span of equipment or infrastructure. Meanwhile, the promotion of technological alternation and industrial upgrading results in the stranding or impairment of outdated technologies and the increase of operating costs in replacing equipment (Pankratz et al., 2019).

In summary, climate risk impairs the profitability, solvency, and operation capacity of the private sector, which increases banks' credit risk and reduces their credit supply to the private sector; at the same time, since banks have serious "ownership preferences" and "large enterprise preferences" when allocating financial resources, these preferences are more unfavorable when climate risk increases. Based on the above analysis, this paper proposes hypothesis 1.

Hypothesis 1

Climate risk has a negative impact on private sector credit supply.

Impact of climate risk on public sector credit supply

The public sector refers to organizations such as governments and state-owned enterprises that aim to achieve the public interest of society. Many public sector enterprises are in industries that are natural monopolies, such as the water, electricity, oil, and gas industries are well-known examples. For the public sector, the impact of climate risk on profitability, solvency, and operation capacity also appears in the public sector. However, compared to the private sector, the public sector has a more transparent corporate governance structure and more open information, and the influence of information asymmetry is less (Javadi & Masum, 2021). Furthermore, the public sector generally has a large scale and strong organizational structure. So public sector is more resilient to risks than the private sector. We mainly focus on the impact of climate risk on public sector credit supply from a government perspective for two reasons: first, government ownership of banks and regulation of credit markets are common throughout the world (La Porta et al., 2002), As Calomiris and Haber (2015) states, “banking system is an outcome of political deal making.” It is universal for the regulation of credit allocation; second, the public sector plays an important role in facilitating climate finance, for example by leading the development of a strong climate information architecture to further improve decision-making and risk pricing (Georgieva & Adrian, 2022). Thus, the public sector is more likely to receive government attention and has more affected by government intervention.

Driven by the policy of low-carbon economy transition, banks' credit strategy to the public sector will be subject to intervention from the government (Switzer & Wang, 2013). On the one hand, the government will directly interfere with banks' financial resource allocation behavior by means of administrative-style orders. When climate risk increases, the government will directly intervene to push banks' credit resources toward public sector enterprises, support state-owned enterprises to actively cope with climate risks, and promote the low-carbon transformation development of the state-owned economy. D'Orazio (2022) discovers that the government and regulators take green credit allocation measures to support sustainable development, such as concessional loans and green lending quotas. On the other hand, the government may indirectly influence banks' credit resource allocation behavior by distorting the risk-return characteristics of some of the enterprises it prefers or supports through direct or "invisible" subsidies or guarantees. When climate risk increases, the government's strong support for state-owned enterprises, through subsidies for public sector credit and financial products or the establishment of lending incentives, promotes banks to increase their credit allocation to public sector enterprises. Lamperti et al. (2019) find that green credit guarantees reduce carbon dioxide emissions and contribute to sustainable economic growth.

In summary, with the development of green finance and low-carbon transition economy, the increase of climate risk makes the government prompt banks to increase the supply of credit to public sector enterprises, directing banks' resources to state-owned enterprises and promoting the low-carbon transition of public sector enterprises. Based on the above analysis, this paper proposes hypothesis 2.

Hypothesis 2

Climate risk has a positive impact on the supply of credit to the public sector.

Data and methodology

Data

This study uses the annual data over the period of 2000–2019 period. The main reasons of period selection are presented as follows: first, since the beginning of the twenty-first century, countries around the world have been paying more and more attention to the impact of climate change on economic development. Since 2000, human carbon dioxide emissions have been increasing at a rate of more than 2% per year, while the amount of carbon dioxide that can be absorbed by natural ecosystems has been decreasing (see Fig. 1). Wang et al. (2020) state that carbon emissions can impair financial growth. Thus, we select 2000 as the beginning of the period. Second, according to Feyen et al. (2021), the macro-financial shock caused by the COVID-19 pandemic precipitated a global economic recession and put severe pressure on bank balance performance. In order to avoid the effect of COVID-19, we exclude data from 2020 and beyond. Therefore, we choose 2000–2019 as the sample period.

Fig. 1.

The carbon dioxide emissions of the world.

Source: Climate Watch. 2020. GHG Emissions. Washington, DC: World Resources Institute. Available at: climatewatchdata.org/ghg-emissions

In the selection of the sample countries, we collect all countries with climate risk and bank credit data and delete some countries based on two considerations: first, the countries deleted are small economy sized and are not representative in our study; second, the current sample includes sufficient countries and covers all representative countries. Therefore, we believe that the data process can support us to draw common conclusions.

Given this background, this paper proposes to select data on climate risk and bank credit from 174 countries over the period 2000–2019. All data in this paper are obtained from the University of Notre Dame public database, WDI, and GFDD databases. The climate risk index is from the University of Notre Dame public database, the macro variables at the country level are from the WDI database, and the microvariables at the bank level are from the GFDD database.

Variable construction

Dependent variables

The University of Notre Dame-Global Adaptation Index (ND-Gains), an emerging, country-level, leading indicator for predicting climate change issues and the adaptive capacity of countries, was chosen to measure climate risk. Since 1995, the University of Notre Dame has ranked countries annually on their resilience to natural hazards and their ability to adapt to changes in natural hazards. The ND-Gains are calculated based on the following formula.

$$\text{ND}-\text{Gains score}=(\text{Readiness score}-\text{Vulnerability score}+1)\times 50$$ 1

The vulnerability index is calculated from 36 indicators, including the proportion of imported cereals to domestic cereals, the rate of freshwater extraction, the number of deaths due to disease caused by climate change, changes in biodiversity (MC1), the proportion of vulnerable people (under 14 or over 65 years old), and the number of flood disasters. The readiness index is calculated from 9 indicators, including infrastructure resilience, quality of government regulation, higher education enrollment rate, and the number of patent applications per capita.

As can be seen from the formula, the ND-Gains index is a negative indicator of climate risk, so Gains are taken as a proxy for climate risk. The greater Gains, the higher the climate risk.

Core independent variables

The supply of credit in this paper includes both private sector and public sector credit supply and the corresponding explanatory variables are: private sector credit as a percentage of GDP (PriCredit) and public sector credit as a percentage of GDP (PulCredit). Batabyal and Chowdhury (2015) state that private sector credit as a percentage of domestic GDP as a percentage of GDP is usually used as an indicator of the banking system; Samargandi et al. (2015) also argue that the private sector share of GDP can measure the development of the banking system; Levine et al. (2000) argue that private sector credit has a positive relationship with the development of the banking system. Thus, private sector credit as a percentage of GDP can be used as a proxy variable to measure the supply of bank credit to the private sector. Similarly, in order to study the difference in the supply of bank credit to the private and public sectors, the ratio of public sector credit to GDP is chosen as a proxy variable for the supply of bank credit to the public sector (Table 1).

Table 1.

Variable names, symbols and definitions

Variables	Symbols	Definitions
Independent variable
Climate risk index	Gains	The Gains measure the country's vulnerability and preparedness for climate risks
Dependent variables
Private sector credit	PriCredit	Private sector credit as a percentage of GDP
Public sector credit	PulCredit	Public sector credit as a percentage of GDP
Control variables
Economic growth	Gdpg	GDP growth rate
Inflation	lnCPI	Logarithm of the Consumer Price Index
Import and export trade	Trade	Total import/export trade as a percentage of GDP
Bank earnings	Return	Net income after taxes of the banking sector as a percentage of average annual equity
Non-performing loan rate	Npl	Non-performing loans in the banking sector as a percentage of total loans
Bank deposits	Deposit	Deposits in the banking sector as a percentage of GDP

Control variables

Based on existing studies (Bordo et al., 2016; Ward & Shively, 2017; Laséen et al., 2017; Niepmann & Schmidt-Eisenlohr, 2017; Korri & Baskara, 2019; Yin, 2019; Nguyen et al., 2020; Phan et al., 2021; Chen & Chang, 2021; Rusdiyanto et al., 2020; Demir, 2021), this paper selects control variables at the country level and the bank level that prior studies suggest being associated with credit supply.

1. GDP growth rate (Gdpg), which measures economic growth. As richer countries have better infrastructures and more mature financial markets, which may be a determinant of credit supply in the effect of climate risk (Chen & Chang, 2021).

2. Logarithm of the consumer price index (lnCPI), which measures the level of inflation, is selected. Ikpepsu (2021) points out that bank credit can be hindered by inflation. Meanwhile, in order to mitigate the adverse effects caused by possible heteroskedasticity, the CPI data are smoothed by taking the natural logarithm of the CPI and the lnCPI is used to measure the level of inflation in this paper;

3. Total import and export trade as a percentage of GDP (Trade), which measures the level of trade and the degree of openness of the country. Niepmann and Schmidt-Eisenlohr (2017) studies the relationship between international trade level and band credit. Referring to Chen and Chang (2021), we choose the total import and export trade as a percentage of GDP as the proxy of trade level.

4. The ratio of net income after tax to average annual equity (Return), which measures the profitability of the banking industry; Ruziqa (2013) and Beck et al. (2013) find the adverse relationship between bank performance and credit. Ruziqa (2013) concludes that higher profitability contributes to lower bank credit risk. Return on equity (ROE) and return on total assets (ROA) are usually used as proxies for the performance of banks (Ekinci & Poyraz, 2019). Compared to ROA, ROE illustrates the profitability of the banks’ net assets, which considers the leverage and capital structure of banks. As most banks maintain highly leveraged operations, ROE is more appropriate to measure bank financial performances.

5. The ratio of non-performing loans to total loans (Npl), which measures the stability of the banking industry; It estimates the quality and safety of the bank credit and allows us to assess the sustainability of bank credit risk (Caby et al., 2022; Mohaddes et al., 2017). Thus we choose Npl as an indicator of the stability of banks.

6. The ratio of deposits to GDP (Deposit), which measures the size of the banking industry, is selected. Yin (2019) reveals a strong relationship between the ratio of deposits to GDP and bank credit supply. If a country has a higher ratio of bank deposits to GDP, the banking system's role in providing credit to the economy is more important. Phan et al. (2021) illustrate that the ratio of deposits to GDP is usually used as proxies for the deposit level of a country. Therefore, deposits should be considered in our analysis.

Model construction

This paper selects panel data for 174 countries from 2000 to 2019 and develops the following econometric model.

$$\text{Credit}={\beta }_0+{\beta }_1{\text{Gains}}_{\mathit{it}}+{\beta }_2{\text{Gdpg}}_{\mathit{it}}+{\beta }_3{\text{lnCPI}}_{\mathit{it}}+{\beta }_4{\text{Trade}}_{\mathit{it}}+{\beta }_5{\text{Return}}_{\mathit{it}}+{\beta }_6{\text{Npl}}_{\mathit{it}}+{\beta }_7{\text{Deposit}}_{\mathit{it}}+u_{\mathit{it}}$$ 2

In Eq. (2), credit is divided into two categories, PriCredit for private sector credit supply and PulCredit for public sector credit supply. β₀ is a constant term; β₁ is the coefficient of climate risk, by estimating this coefficient, the impact of climate risk on credit supply can be known; β₂, β₃ and β₄ are the coefficients of country-level control variables Gdpg, lnCPI, and Trade, respectively. Gdpg denotes the country's level of economic growth; lnCPI denotes the country's level of inflation; Trade denotes the country's level of import and export trade; and are the coefficients of the bank-level control variables Return, Npl, and Deposit, respectively; Return denotes the profitability of the banking sector; Npl denotes the credit quality of the banking sector and Deposit denotes the size. i and t denote country and year, respectively; u_it are error terms.

Empirical findings

Descriptive statistics

Table 2 shows the descriptive statistical results for the main variables. It can be seen that the maximum value of private sector credit (PriCredit) is 986.1, the minimum value is 0.391, and the standard deviation is 51.06, indicating that the private sector credit of each country fluctuates greatly during the sample period in this paper. The maximum value, minimum value and standard deviation of public sector credit (PulCredit) are 75, 0.003 and 10.46, indicating that there are also great differences among public sector credit of different countries. Gains also showed significant differences in the maximum value of − 26.99, the minimum value of − 77.83, the standard deviation of 11.07, and these differences prompted further studies of the heterogeneity of climate risk effects.

Table 2.

Descriptive statistical results

Variables	Observations	Mean	SD	Minimum	Maximum
PriCredit	3212	47.95	51.06	0.391	986.1
PulCredit	3130	10.42	10.46	0.003	75
Gains	3480	− 48.53	11.07	− 77.83	− 26.99
Gdpg	3432	3.849	5.229	− 62.08	123.1
lnCPI	3236	4.551	0.379	1.068	7.916
Trade	3220	86.43	49.99	0.175	437.3
Return	2703	12.86	14.19	− 132.6	259
Deposit	3169	50.77	48.15	0.61	770.3
Npl	2108	6.922	7.443	0	74.1

Baseline regression

In this paper, we select gains as a proxy variable for climate risk and empirically test the effect of climate risk on private sector credit supply and public sector credit supply. Based on Eq. (2), F-test and Hausman test are conducted to determine the specific form of the panel regression model by combining the sample data. The results show that all p-values of the F-test are less than 0.1, so the mixed-effects model is not selected. All p-values of the Hausman test are less than 0.05, so the random-effects model is not selected. In summary, this paper chooses a two-way fixed effects model, i.e., controlling for country and time effects, and conducts hypothesis tests for private sector credit supply and public sector credit supply, respectively.

Columns (1) and (2) of Table 3 show the regression results of climate risk on the credit supply of the private sector and public sector, respectively, and it can be seen that the coefficients of climate risk on private sector credit supply and public sector credit supply are − 0.7253 and 0.1848, respectively, and both are statistically significant at 1% level. The results indicate that climate risk has a significant negative relationship with private sector credit and a significant positive relationship with public sector credit. For the private sector, when climate risk increases, the profitability and solvency decrease, operation costs increases, which makes banks face more credit risk and thus reduce the supply of credit to the private sector; for the public sector, when climate risk increases, the government influences the credit resource allocation behavior of banks through direct or indirect interventions to promote the low-carbon transformation of public sector enterprises development, resulting in a credit bias of bank credit to public sector enterprises.

Table 3.

Baseline regression results

	(1)	(2)
	PriCredit	PulCredit
Gains	− 0.7253***	0.1848***
	(− 4.7393)	(2.9329)
Gdpg	− 0.9453***	− 0.1279***
	(− 9.1291)	(− 2.9815)
lnCPI	3.4077**	0.6821
	(2.0573)	(1.0024)
Trade	− 0.0858***	− 0.0102
	(− 4.2534)	(− 1.2154)
Return	− 0.0606***	− 0.0201**
	(− 2.7589)	(− 2.2326)
Deposit	0.7044***	0.0229**
	(25.2242)	(1.9788)
Npl	− 0.0297	0.0945***
	(− 0.5056)	(3.9152)
_cons	− 25.0591**	15.9286***
	(− 2.1800)	(3.3751)
N	1704	1683
R2	0.4745	0.1406

Robust t-statistics in parentheses; ***p < 0.01, **p < 0.05, *p < 0.1

Considering the influence of control variables, in the regression of climate risk on private sector credit supply, GDP growth rate, international trade, banking sector profitability Return and banking sector deposit level are significantly negative at 1% level, the coefficient of Deposit is significantly positive at 1% level, and the price level lnCPI is significantly positive at 5% level; in the regression of climate risk on public sector credit supply regressions, GDP growth rate and banking sector profitability Return are significantly negative at 1% and 10% levels, respectively, and banking sector deposit level Deposit and non-performing loan ratio Npl are significantly positive at the 5% and 1% levels, respectively, i.e., rising price level has a positive impact on both credit supply, and an increase in bank deposits as an indicator of GDP Deposit has a positive impact on private sector credit and the public sector have a positive impact, and the findings are in line with theoretical expectations.

Endogeneity

As described by Giusy et al. (2020), there may be a bidirectional effect between climate risk and credit supply, and the p-values of both regressions are less than 0.05 by Hausman's test, indicating that the benchmark regression does have the problem of estimation bias due to endogeneity. Therefore, this paper adopts an instrumental variable approach to mitigate the endogeneity problem, drawing on Demir (2021) by using bank-level control variables as endogenous variables and using the lagged terms of the endogenous variables as instrumental variables, choosing the first-order lags of bank earnings Return, bank deposits Deposit, and non-performing loan ratio Npl as instrumental variables for the regressions.

The regression results for private sector credit supply indicate that the explanatory variables remain significantly negative at the 1% level. In the identification test, the Anderson LM statistic is 62.492, with a p-value less than 0.01 indicating that the original hypothesis of “underidentified instrumental variables” is significantly rejected at the 1% level; in the weak instrumental variables test, the Cragg–Donald Wald F-statistic is 21.405, which is greater than the In the weak instrumental variables test, the Cragg-Donald Wald F-statistic is 21.405, which is greater than the 5% threshold, indicating that there are no weak instrumental variables. The results of the instrumental variables regression for public sector credit supply indicate that the explanatory variables remain significantly positive at the 1% level. In the identification test, the Anderson LM statistic is 61.623 with a p-value less than 0.01 indicating that the original hypothesis of “underidentified instrumental variables” is significantly rejected at 1% level; in the weak instrumental variables test, the Cragg-Donald Wald F-statistic is 21.101, which is greater than the In the weak instrumental variable test, the Cragg-Donald Wald F-statistic is 21.101, which is greater than the 5% threshold, indicating that there is no weak instrumental variable problem.

The results in Table 4 show that after mitigating the endogeneity problem, climate risk has a significant negative impact on private sector credit supply and a significant positive impact on public sector credit supply, and the results of the instrumental variables regression are generally consistent with the results of the benchmark regression.

Table 4.

Regression results of instrument variable method

	(1)	(2)
	PriCredit	PulCredit
Gains	− 0.6746***	0.1863***
	(− 4.2625)	(2.7775)
Gdpg	− 1.0608***	− 0.0034
	(− 7.3173)	(− 0.0553)
lnCPI	3.0559	1.0022
	(1.6374)	(1.2679)
Trade	− 0.0940***	− 0.0197**
	(− 4.3901)	(− 2.1415)
Return	− 0.0486	− 0.1281***
	(− 0.4581)	(− 2.8511)
Deposit	0.7842***	0.0241
	(19.3373)	(1.3888)
Npl	− 0.3002***	0.1143***
	(− 3.7292)	(3.3617)
N	1538	1521
R2	0.4515	0.0518

Robust t-statistics in parentheses; ***p < 0.01, **p < 0.05, *p < 0.1

Additional robustness checks

Using an alternative proxy variable

Following Demir (2021), the lagged term of the explanatory variable Gains is used as a proxy variable, and the results from columns (1) and (2) of Table 5 show that replacing the explanatory variables, the climate risk index has a significant negative effect on private sector credit supply and a significant positive effect on public sector credit supply, in line with the results of the benchmark regression.

Table 5.

Robustness checks

	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)
	PriCredit	PulCredit	PriCredit	PulCredit	PriCredit	PulCredit	PriCredit	PulCredit
L. Gains	− 0.6727***	0.2296***
	(− 4.3961)	(3.6670)
Gains			− 0.8096***	0.2076***	− 0.7253***	0.1848**	− 0.2730**	0.1706***
			(− 5.0702)	(3.0111)	(− 3.0305)	(2.5002)	(− 2.4305)	(2.9883)
Gdpg	− 0.9479***	− 0.1381***	− 0.9010***	− 0.1318**	− 0.9453***	− 0.1279*	− 0.3622***	− 0.0939**
	(− 9.1454)	(− 3.2342)	(− 7.3026)	(− 2.4749)	(− 8.3055)	(− 1.9534)	(− 4.6095)	(− 2.3598)
lnCPI	2.0771	0.9432	4.4567***	0.3343	3.4077*	0.6821	− 4.3778***	2.1747***
	(1.1921)	(1.3250)	(2.6106)	(0.4542)	(1.9675)	(1.0955)	(− 3.4900)	(3.4249)
Trade	− 0.0983***	− 0.0103	− 0.0788***	− 0.0257***	− 0.0858***	− 0.0102	− 0.0344*	0.0211**
	(− 4.8386)	(− 1.2294)	(− 3.7567)	(− 2.7739)	(− 3.1225)	(− 0.7282)	(− 1.9079)	(2.3090)
Return	− 0.0602***	− 0.0207**	− 0.0800***	− 0.0386***	− 0.0606**	− 0.0201*	− 0.0777***	0.0120
	(− 2.7370)	(− 2.3007)	(− 2.7555)	(− 3.0798)	(− 2.4411)	(− 1.9179)	(− 4.2132)	(1.2865)
Deposit	0.6811***	0.0219*	0.7147***	0.0237*	0.7044***	0.0229*	0.9146***	0.0652***
	(24.3163)	(1.9031)	(23.6582)	(1.8021)	(9.4718)	(2.0495)	(34.9359)	(4.9081)
Npl	− 0.0191	0.0917***	− 0.0517	0.1058***	− 0.0297	0.0945***	− 0.0458	0.0810***
	(− 0.3177)	(3.7260)	(− 0.8559)	(4.0648)	(− 0.2445)	(4.8666)	(− 1.0483)	(3.6684)
Reserve							− 0.3046***	− 0.0877***
							(− 11.4177)	(− 6.4814)
FDI							0.0820*	− 0.0341
							(1.9232)	(− 1.5838)
Urban							0.0517	− 0.0613
							(0.4400)	(− 1.0261)
_cons	− 16.0681	21.0344***	− 34.5979***	19.9219***	− 25.0591	15.9286**	18.2673	10.9813*
	(− 1.2099)	(3.8827)	(− 2.8726)	(3.8406)	(− 1.6337)	(2.5612)	(1.4950)	(1.7697)
N	1651	1632	1428	1413	1704	1683	1192	1185
R2	0.4669	0.1460	0.4922	0.1410			0.6957	0.2645

Robust t-statistics in parentheses; ***p < 0.01, **p < 0.05, *p < 0.1

Adjusting sample period

Considering the impact of the global financial crisis on credit supply in 2008, the sample of the year of the event is removed from the paper. The data from 2008 to 2010 are excluded and thus regressed again. Columns (3) and (4) in Table 5 show the regression results. The results show that a 1-unit increase in the climate risk index would reduce private credit supply by 0.8096% and increase public sector credit by 0.2076%, supporting hypotheses 1 and 2, indicating that the results are robust.

Altering models

As the panel data in this paper has the characteristic of “long N short T,” it may cause the heteroscedasticity problem. In this paper, the method of Driscoll and Kraay (1998) was adopted to solve the heteroscedasticity problem of panel data. As Driscoll and Kraay (1998), the error structure is set as the autoregression of heteroscedasticity and specific order. When the time dimension increases gradually, the standard error is robust to the cross-section correlation and time correlation of the general form. Because this Driscoll and Kraay (1998) estimation approach uses the nonparametric technique to estimate the standard error, the number of sections is not limited, so when the number of sections N is much larger than the period number T, the estimation will not be greatly affected. The results of column (5) and column (6) are basically consistent with the baseline regression results, which proves the robustness of the results.

Adding controls

To mitigate the endogeneity of missing variables, we also add a series of controls by following recent studies (Carmignani et al., 2021; Emefiele et al., 2022; Sugimoto & Enya, 2022). The ratio of bank liquid reserves to bank assets (Reserve), which reflects the bank reserves level of a country. Emefiele et al. (2022) suggest that there is a significant effect of bank reserves on bank credits. Lower bank reserve ratio will enhance the credit supply. The net inflows of foreign direct investment as a percentage of GDP (FDI), which measures the ability of attracting foreign investment of a country. Foreign capital flows into banks in the form of direct investment increases the bank credit supply (Sugimoto & Enya, 2022). The ratio of the urban population to the total population (Urban), which reflects the level of urbanization. Carmignani et al. (2021) show that the level of urbanization has a positive impact on access to credit. Consequently, the above three variables (Reserve, FDI, and Urban) should be included as controls for this study. The results from columns (7) and (8) of Table 5 show that adding Reserve, FDI, and Urban as controls. Comparing the results with the results in Table 3, the estimations provide similar results.

Heterogeneity

Considering that there may be heterogeneity in national income level in the impact of climate risk on credit supply of private and public sectors, the samples are grouped and returned according to the World Bank standard by income level. The World Bank divides national income level into: High-income countries, upper-middle-income countries, lower-middle-income countries, and low-income countries. In this paper, considering the sample size when grouping, both high-income and upper-middle-income are classified into the high-income group, and both lower-middle-income and low-income are classified into the low-income group, and excluding 2 countries without ratings, there are 100 countries in the high-income group and 72 countries in the low-income group, and then grouped for regression, and the results are shown in Table 6.

Table 6.

Heterogeneous effect

	(1) High-income	(2) Low-income	(3) High-income	(4) Low-income
	PriCredit	PriCredit	PulCredit	PulCredit
Gains	− 0.7911***	− 0.2309**	0.2512***	0.1419**
	(− 3.4956)	(− 1.9826)	(2.6101)	(2.1423)
Gdpg	− 1.2392***	− 0.2336**	− 0.1311**	− 0.1219**
	(− 9.3434)	(− 2.4078)	(− 2.3058)	(− 2.2098)
lnCPI	8.2538***	− 2.3343	− 1.2140	3.9262***
	(3.4041)	(− 1.3672)	(− 1.1830)	(4.0434)
Trade	− 0.1194***	0.0447**	− 0.0323***	0.0558***
	(− 4.8676)	(2.0499)	(− 3.0501)	(4.4691)
Return	− 0.0430*	− 0.1755***	− 0.0221**	0.0185
	(− 1.6979)	(− 5.4047)	(− 2.0597)	(1.0017)
Deposit	0.4833***	1.0943***	0.0255	0.0211
	(13.2143)	(45.1597)	(1.6287)	(1.5318)
Npl	0.0949	− 0.2453***	0.1168***	0.0752***
	(1.0331)	(− 5.6617)	(2.9942)	(3.0496)
_cons	− 33.6446**	1.8133	31.4742***	− 8.9910*
	(− 2.0065)	(0.1987)	(4.4346)	(− 1.7319)
N	1209	495	1190	493
R2	0.4283	0.8664	0.1332	0.3054

Robust t-statistics in parentheses; ***p < 0.01, **p < 0.05, *p < 0.1

Table 6 shows that for private sector credit supply, private sector credit supply is more strongly affected by climate risk in high-income countries than in low-income countries, probably because of the faster development of inclusive finance and micro and small credit in high-income countries and thus a larger share of private sector credit, with a ratio of private sector credit to public sector credit of 35.02:1 in high-income countries compared to 5.41:1 in low-income countries. The higher proportion of private sector credit implies that private sector credit is more vulnerable to climate risk in high-income countries. Therefore, when climate risk increases, climate risk has a greater impact on the supply of private sector credit in high-income countries.

For public sector credit supply, the impact of climate risk on public sector credit supply is more dramatic in high-income countries than in low-income countries, probably because the government plays a more important role in the allocation of bank credit resources due to the improved infrastructure development and financial system in high-income countries, and the government promotes the transition to a low-carbon economy in the public sector through bank credit, thus promoting banks to increase credit allocation to the public sector. Therefore, when climate risk increases, the impact of climate risk on public sector credit supply is greater in high-income countries.

Discussion

This study empirically investigates the impact of climate risk on credit supply by using a sample of 174 countries during 2000–2019 from the perspective of the difference between private and public sectors. We present our main findings as follows: first, climate risk has a significant negative impact on credit supply to the private sector and a significant positive impact on the public sector. When climate risk increases, the private sector's profitability, and solvency decrease, while operation costs increase, making banks more vulnerable to credit risk and decreasing the supply of credit to the private sector. This finding is consistent with Hosono et al. (2016) and Islam and Wheatley (2021). When climate risk increases, the government directs or indirectly intervenes in banks' credit resource allocation behavior to encourage the public sector's low-carbon transformation, resulting in a credit supply heterogeneity of banks to the public sector. It is consistent with Lamperti et al. (2019) and D'Orazio (2022).

Second, both private and public sector credit supply are more strongly affected by climate risk in high-income countries than in low-income countries. For the private sector, climate risk has a more significant impact on the supply of private sector credit in high-income countries, mainly due to a larger share of private sector credit in high-income countries. Thus, private sector credit is more vulnerable to climate risk in high-income countries. For the public sector, government plays a more important role in the allocation of bank credit resources due to the well-developed infrastructure and better quality of governance in high-income countries, the impact of climate risk on public sector credit supply is greater in high-income countries. Our study provides some interesting evidence that may not be consistent with other scholars (Chen & Chang, 2021; Lee et al., 2022): The complete infrastructure construction and financial system have two folds in high-income countries. On the one hand, complete infrastructure construction and better quality of governance facilitate the role of government in credit allocation in high-income countries and promote the green transformation of the public sector. On the other hand, mature financial systems creates a higher proportion of private sector credit and the more convenient risk contagion in high-income countries, which results in the vulnerability to climate risks of the high-income countries.

Limitations of this study and directions for further research: this study reveals the impact of climate risk on bank credit supply to the private sector and public sector from an international perspective. Compared to firm-level or industry-level research in this field, this study focuses on the comparisons from a macro perspective, which inevitably leads to the neglect of some micro influences. The limitations in this study need further research and broader research scope.

Our investigation suggests the following future directions in this field as well: first, the features of credit should be taken into account when studying the impact of climate risk on the credit supply, such as the maturity of credit, the interest rate of credit, and so on. Firms exposed to climate risk are more likely to use long-term credit than short-term credit (Huang et al., 2018; Islam & Wheatley, 2021). Nguyen et al. (2020) find that banks charge higher interest rates on credit for firms exposed to climate risk. Therefore, investigating the maturity and cost of credit may generate discrepant findings which are also significant. Second, the industry of borrowers is also an important factor of credit supply. For example, it is obvious that firms operating in agriculture are more sensitive to climate change than other industries. Banks are more cautious about the credit supply to firms in agriculture. Third, the features of banks may be determinants of credit supply under the impact of climate risk. Yin et al. (2021) argue that compared to non-state-owned banks, state-owned banks have leading positions in green credit supply. Hence, the ownership of banks affects the credit supply to private and public sectors in the face of climate risks. In addition, the geographical location of the bank should be regarded as a determinant of the effect of the climate risks.

Conclusions and implications

This paper has investigated the impact of climate risk on credit supply by using panel data of 174 countries during 2000–2019. We have conducted a heterogeneity analysis of climate risk effects on private and public sectors’ bank credit for these countries. The results show that climate risk has a significant negative impact on credit supply to the private sector and a significant positive impact on credit supply to the public sector. Further, both private and public sector credit supply are more strongly affected by climate risk in high-income countries than in low-income countries.

Academic implications: these findings together provide some important academic implications. Firstly, the existent and common argument is that serious natural disasters will lead to the unbalance of the financial system and impact the banks’ credit service (Brei et al., 2019; Cortés & Strahan, 2017; Ivanov et al., 2020). Climate risk is an important component of natural disasters. Our findings confirm the previous studies of natural disasters to a certain extent and further deepen the research on the mechanism of the impact of climate risk on bank credit supply. Secondly, this study focuses on the heterogeneous effect of climate risk to private and public sectors and investigates the different mechanisms of impact of climate risk on credit supply, which fill the gap in existent research. The heterogeneous effects mainly come from the differences in objectives, organizational structure, and organizational culture between the private and public sectors. Thirdly, we study the impact of climate risk on bank credit supply from the international perspective and obtain common conclusions from a macro view. We also provide new evidence about the amplifying effect of climate risk in high-income countries compared to other scholars. While other scholars have noted that well-established infrastructure in high-income countries improves resilience to risk and national governance in credit allocation to the private sector, this study also indicates that the mature financial systems and the more convenient risk contagion in high-income countries lead to a higher vulnerability of the private sector's credit to the climate risk.

Managerial implications: according to the findings above, there are some managerial implications as follows. For the private sector, climate risk brings new changes to the profitability, solvency, and operational capacity of the private sector. In the face of climate risk, the private sector should raise awareness of climate risk, make corporate strategies, optimize organizational structure, and improve climate risk information disclosure to enhance climate risk management capabilities. For the public sector, it should make full use of credit provision and accelerate its own low-carbon transition. At the same time, the public sector can cooperate closely with policymakers, regulators, local governments, and enterprises to enhance the synergies of low-carbon transition. For banks, climate risk has impacts on risk assessment and capital management. In terms of risk assessment, banks should attach great importance to climate risk, fully consider the links and transmission pathways between climate risk and other risks, and incorporate climate risk into bank comprehensive risk management systems. In terms of capital management, climate risk should be regarded as an important consideration in credit decisions, and promote innovation in green financial products.

Policy recommendations: based on our findings, the following policy recommendations are proposed to cope with the current complex context of climate change: first, governors should enhance the quality of information disclosure, since the current pricing of climate risk is too low, mainly due to insufficient information disclosure requirements, resulting in the failure of capital market to identify the climate risk. The climate risk aggravates the information asymmetry in the market and impairs the efficiency of capital allocation. Therefore, the government should improve climate risk management and construct a complete information disclosure system of climate risk. Second, the climate risk should be incorporated into banks' credit policies. Especially for the private sector in high-income countries, with the development of inclusive finance, the private sector credit is more vulnerable to climate risk. For banks, specific credit supply strategies should be designed for high-climate-risk industries to manage climate risk. By developing quantitative and stereotypical tools to assess portfolio exposures in geographic regions and industries with high-climate risk. Particularly, the impact of climate risk on enterprises should be considered to avoid the physical and transition risks of climate change from exacerbating adverse impacts on bank customers when supplying credit to the private sector. Third, the complete infrastructure development and economic system are beneficial to prevent climate risk. For public sector credit, in high-income countries, a complete infrastructure construction and financial system is conducive to the government's promotion of low-carbon economic transformation in the public sector through bank credit and increased support for the low-carbon economic transformation of the public sector. Therefore, governors should devote to improving production technology and raising income levels, while stimulating R&D development to strengthen risk control capabilities.

Data availability

The data would be available from the authors.