Greenhouse Gas Emissions and the Financial Stability of Insurance Companies

Abstract

The recent losses and damages due to climate change have destabilized the insurance industry. As global warming is one of the most critical aspects of climate change, it is essential to investigate to what extent greenhouse gas emissions affect the financial stability of insurers. Insurers typically do not emit substantial greenhouse gases directly, while their underwriting and investment activities play a substantial role in enabling companies that do. This article uses panel data regressions to analyze companies in all insurance segments and in all geographic regions of the world from 2004 to 2023. The main finding is that insurers that increase their greenhouse gas emissions become financially unstable. This result is consistent in all three scopes (scope 1, scope 2, and scope 3) of emissions. Furthermore, the findings reveal that this impact is related to reserves and reinsurance. Specifically, reserves increase with greenhouse gas emissions, while premiums ceded to reinsurers decline. Thus, high-emissions insurers retain a significant share of carbon risk and eventually become financially weak. The results encourage several policy recommendations, highlighting the need for instruments that improve the assessment and disclosure of insurers’ carbon footprints. This is crucial to achieving environmental targets and improving the stability of both the insurance market and the economic system.

1. Introduction

Global warming and climate change are destabilizing the insurance industry. Regulators and environmental authorities recognize that, despite mitigation and adaptation efforts, the more recent (economic and non-economic) “loss and damage” due to climate change have been substantial (United Nations Environment Programme, 2023). Therefore, insurers face huge payouts that threaten the solvency of their businesses. Examining the responses activated by US insurers to environmental risks, Gupta et al. (2023) showed that many firms are not adequately prepared for the risks of climate change, exhibiting a relatively high level of financial weakness. The Bank for International Settlements (BIS) warns that there is a possibility that the global average temperature increase target under the Paris Agreement could be breached, pushing the world, the global economy, and financial systems into uncharted territory (Khoo & Yong, 2023). For underwriters, this could increase the difficulty in measuring, predicting, and distributing risks. Insurers recognize that, when their own emissions grow or when they underwrite carbon-intensive businesses, they are supporting unsustainable business practices (KPMG, 2023), which in the longer term would make their operations unfeasible.

Recent floods and extreme weather events in the United States have been blamed for fueling a crisis in the insurance market. As a result, insurance premiums have increased, and some insurance companies have started refusing to insure real estate (BBC, 2024). Evidently, the carbon footprint of insurers has an impact on their financial conditions, affecting the sustainability of the risk-sharing services that insurance provides. In turn, this has profound implications for the stability of the insurance market and the entire economic system. To our knowledge, this is the first study to empirically investigate how the financial stability of insurers is related to their scope 1, scope 2, and scope 3 emissions, outlining potential channels that determine this link. Addressing this topic is important for expanding knowledge about the impact of greenhouse gas emissions on insurance operations, which the current literature has investigated by looking at, for example, corporate valuation, performance, and capital costs. Even more importantly, the findings provide insight into policy making, supporting the need for rules and initiatives aimed at reducing carbon exposures within the insurance sector, and eventually enhance the health of the economic system beyond benefiting the environment. The article is organized as follows: Section 2 reviews the literature on the topic and develops working hypotheses. Section 3 presents the methodology. Section 4 outlines the results. Section 5 discusses the findings and concludes.

2. Literature and Development of Hypotheses

Emissions of CO₂ and non-CO₂ are the major causes of global warming (Montzka et al., 2011), which is one of the main aspects of climate change (NASA, 2024). The impact of greenhouse gas emissions on corporate financial decisions is multifaceted. Research has shown that carbon risk affects companies from different perspectives, ranging from corporate equity performance (Oestreich & Tsiakas, 2015; Wen et al., 2020; Bolton & Kacperczyk, 2021, 2024), valuation (Matsumura et al., 2014; Clarkson et al., 2015; Griffin et al., 2017; Hágen & Ahmed, 2024), investor decision-making (Krueger et al., 2020), to capital costs (Kim et al., 2015; Bui et al., 2020).

With the growing attention in the economy to environmental topics, a key issue for both corporate managers and policy makers has become understanding whether the carbon footprint of businesses can make them financially unstable. For example, Kabir et al. (2021) analyzed carbon emissions of non-financial firms worldwide, reporting that increasing emissions lead to a higher default risk. This is mainly due to corporate emissions having a negative effect on corporate cash flow volatility and profitability. Based on the S&P 500 index members, Perera et al. (2023) found that the intensity of carbon emissions of firms leads to a higher idiosyncratic volatility. Concerning the banking sector, Chabot and Bertrand (2023) developed a theoretical framework for the transmission of climate risks to financial institutions and the financial system. To test the prediction of the model, the authors use a panel of European financial institutions. They show that the transition climate risk, measured by greenhouse gas emissions (scope 1, scope 2, and scope 3), negatively affects financial stability at institutional and system-wide levels.

Environmental impacts are part of the range of topics and challenges that corporate sustainability poses to insurance boards and executive level managers (Golnaraghi, 2023), especially as insurance companies expand globally and must manage complex corporate structures (Pranugrahaning et al., 2023). Gatzert et al. (2020) provided an overview of the most relevant sustainability risks and opportunities that involve insurers at different levels. For example, investment strategies can benefit from the growing size of sustainable asset classes. Insurance liabilities, on the other hand, are significantly affected by the increasing severity of climate change and weather risks, complicating the assessment and quantification of insurance costs.1 However, insurers can also adapt and benefit from environmental risks by incentivizing low-carbon-related purchases/assets or by developing methods that incorporate carbon risk in premium settings and claim management. Using data from the Taiwanese market, Ho et al. (2018) illustrated that the sustainable development of the insurance business is considerably dependent on environmental practices, with climate change being one of the most important criteria for evaluating how effectively corporate social responsibility is integrated with the insurance business strategy and management processes. Chiaramonte et al. (2016) approximated corporate sustainability with environmental, social, and governance (ESG) ratings, showing evidence that the environmental dimension plays a stronger role in enhancing the financial health of insurers.

Carbon risk is one of the main environmental challenges for insurers, as it is a source of significant instability for firms. Dlugolecki (2008) argued that global warming has several implications for underwriters: “Catastrophe models are wrongly calibrated; premiums are too low; exposures are too high; claim handling capacity is inadequate; and credit ratings are too generous.” Recent evidence illustrates substantial losses and damages to property and life as a result of natural catastrophes, resulting in huge payouts that, in turn, threaten the solvency of many businesses. Insurers recognize that, by underwriting carbon-intensive businesses, they support unsustainable business practices (KPMG, 2023). Therefore, given their pivotal role in the net zero transition, insurers are currently struggling to manage their carbon footprints to avoid carrying the negative consequences related to climate change risks. For example, Dawson et al. (2022) used the company Allianz as a case study, analyzing its carbon emissions compared to other firms in the sector, and describing how it has achieved emissions reductions in response to pledges of net zero emissions.

The stability of insurance operations is affected by risks from climate change both in the short term and long term (KPMG, 2022). The increasing frequency and severity of natural disasters and extreme weather events make it more difficult for insurers to predict losses and appropriately price insurance claims. Unstable businesses also implies problems in the affordability and availability of insurance (European Insurance and Occupational Pensions Authority, 2023). The Bank of England points out that the stability of insurers, as financial intermediaries, is threatened by climate risks in the form of physical, transition, as well as “liability risk”, which is directly related to investment provision and insurance services (Bank of England, 2019). Finally, evidence based on United States data from Bressan and Du (2025) indicates that the increase in greenhouse gas emissions throughout the country dampens the value of the insurer’s equity. The authors interpret this finding by arguing that investors discount the value of insurance companies with higher expected returns when faced with greater climate change risks.

While carbon emissions seem to impact the insurer’s stability, the literature has not examined the drivers of this effect. Both reserves and reinsurance can be relevant to this mechanism. In fact, uncertainty related to climate change affects the ability to establish adequate reserve levels, making it more difficult for insurers to quantify the likelihood and size of the influence that climate change has on both their own operations and the operations of their customers. Therefore, carbon exposure would lead to higher reserve levels that are arguably necessary to cover possible volatility in future payments. However, if insurers react to the frequent and increasingly serious nature of environmentally destructive events by continually increasing premiums and reserves, they may affect their underwriting capacity and profitability, a solution that, in the longer term, would be unfeasible and bring about instability.

Reinsurance plays a key role in helping insurance companies pay for large, unexpected losses caused by natural disasters. However, the high frequency of extreme weather events has led reinsurers to receive many claims. A recent report from the Fitch Ratings Rating Agency points out that the incidence of climate losses has prompted reinsurers to step back, as the reinsurance industry has become more averse to supporting “secondary risks”, which are smaller but more frequent extreme weather events (Fitch Ratings, 2024). Another rating agency, S&P, said that “more than half of the top 20 global reinsurers maintained or reduced their natural catastrophe exposures during the January 2023 renewals, despite the improved pricing terms and conditions and growing demand” (S&P, 2023). Thus, increasing the exposure to carbon risk would decrease the availability of reinsurance. In other words, high-polluting insurers would end up retaining a larger share of the risk.

Despite the concerns raised among policy makers and practitioners in the insurance industry about carbon footprints, existing research presents a gap in providing evidence related to the stability of insurers. This article aims to explore this issue. The main argument is that greenhouse gas emissions expose insurers to environmental risks that ultimately make companies financially unstable. This effect concerns both direct and indirect emissions, measured by scope 1, scope 2, and scope 3 emissions. Arguably, both reserves and reinsurance are two important drivers of this impact. To summarize, the following working hypotheses (HPs) will be tested in the next section by means of panel data analyses:

HP1. 

Insurers’ carbon emissions make insurance firms financially unstable.

HP2. 

Scope 1, scope 2, and scope 3 emissions reduce the stability of insurance firms in a similar way.

HP3. 

Increasing greenhouse gas emissions make insurers financially unstable through the growth of reserves.

HP4. 

Increasing greenhouse gas emissions make insurers financially unstable by decreasing reinsurance.

3. Methods

The data source for this analysis is S&P Capital IQ, which provides data on carbon emissions, as well as accounting data for worldwide insurers. The companies are all publicly listed and operate in one of the following segments of insurance: Financial guarantee, life and health, managed care, mortgage guarantee, multi-line, property and casualty, and title insurance.2 Based on these screening criteria, the final sample consists of 2043 firm-year observations from 2004 to 2023. The composition of the sample by insurance segments is shown in

Table 1. Number of observations by insurance segments.

Segment	N
Financial Guaranty	30
Life and Health	605
Managed Care	106
Mortgage Guarantee	48
Multi-line	349
Property and Casualty	856
Title Insurance	49
Total	2043

.

According to the GHG Protocol Corporate Standard, a company should separately account for and report its greenhouse gas emissions into three “scopes”. Scope 1 emissions are direct greenhouse gas emissions that occur from sources that are owned or controlled by the company. Scope 2 covers indirect emissions from the purchase and use of electricity, steam, heating, and cooling. Scope 3 includes all other indirect emissions that occur in the upstream and downstream activities of an organization (Partnership for Carbon Accounting Financials, 2022). The emissions associated with investment and insurance activity are defined as scope 3 emissions, and are known as financed or insurance-associated emissions (KPMG, 2023)

For each scope, the total annual values (in tons of CO2e) of greenhouse gas emissions are obtained. The data source provides the figures that combine company-reported data and modeled data. The model data are calculated using process-based approaches and multisector Environmentally Extended Input–Output (EEIO) modeling.3 In the analysis, the natural log of the emission values of scope 1, 2, and 3 are GHG1, GHG2, and GHG3, respectively.4

Figure 1 shows the averages of insurer greenhouse gas emissions over the years. Despite a decreasing trend over time, the figure highlights how scope 3 greenhouse gas emissions widely overcome scope 1 and scope 2 emissions. A similar pattern in the magnitude of scope 1, scope 2, and scope 3 emissions, also concerning banks, has been outlined by Bressan (2025). This pattern depends on the business model of insurers, which, in fact, are “unique in the need to consider both investment and insurance emissions” (KPMG, 2023). In 2020, the reporting finance portfolio emissions of European insurers were “over 700 times larger than direct emissions” (Deloitte, 2023). More recently, it has been calculated that, in the insurance industry worldwide, more than 95 percent of emissions fall under scope 3 (KPMG, 2023).

The financial stability of insurers is measured using two alternative measures. The first measure is the combined ratio (CR), that is, the sum of incurred losses, loss adjustment expenses, plus other underwriting expenses, divided by earned premiums (Rejda, 2005). A combined ratio below 100% indicates that the company is making an underwriting profit, whereas a ratio above 100% suggests an underwriting loss. Doherty and Garven (1995) provide theoretical evidence for the negative relationship between the combined ratio and the solvency rate of a corporation. In empirical research, the combined ratio is used to examine the financial conditions of insurers. For example, Browne and Hoyt (1995) report that a ratio could indicate unfavorable underwriting results and lower profitability. Chen and Wong (2004) used the combined ratio to study the main determinants of financial health of insurers. Bressan and Du (2024a) employed the combined ratio to show that, by purchasing reinsurance, insurers become financially more solid. Regulators closely monitor insurers’ combined ratios to ensure that they maintain a sufficient level of capital to cover claims while also operating sustainably. From a regulatory standpoint, a persistently high combined ratio can trigger increased scrutiny and potential intervention to safeguard the insurer’s solvency and the interests of the insured. For example, in the European framework, the European Insurance and Occupational Pensions Authority (EIOPA) uses the combined ratio to report on insurers’ solvency and profitability (European Insurance and Occupational Pensions Authority, 2025).

The second measure assessing the financial stability of insurance companies is the underwriting leverage, or premium-to-surplus ratio, i.e., the ratio of net premiums written to policyholder surplus (PS). The policyholder surplus is the total assets net of any liabilities and represents the amount of money that the firm has left over after paying its claims and other expenses. The premium-to-surplus ratio, which is sometimes referred to as the “insurance exposure”, has been advocated on occasion as a rule-of-thumb indicator of insolvency (Lai, 2006). Leng and Meier (2006) examined underwriting cycles, using the premium-to-surplus ratio to measure insurer risk-taking behavior. In fact, a high ratio indicates that an insurer has a high level of risk exposure compared to its surplus, which could result in financial difficulties if it experiences a significant loss. On the other hand, a lower ratio indicates greater financial strength for the company. As a rule, regulators set a ratio of less than 3-to-1 premium-to-surplus that insurance companies must adhere to in order to remain relatively healthy. In the United States, the premium-to-surplus ratio is used within the Insurance Regulatory Information System (IRIS) to predict the financial strength of insurers and assess the risk of insolvency (National Association of Insurance Commissioners, 2023). Rating agencies also use the premium-to-surplus ratio to evaluate an insurer’s financial profile and assign credit ratings. Working hypotheses 3 and 4 refer to the effects on reserves and reinsurance. The measure for reserves (RES) is the ratio of reserves to policyholder surplus (National Association of Insurance Commissioners, 2023). This quantity denotes the amount that the firm has set aside for potential claims compared to the total assets it possesses (minus its liabilities) and is inversely related to the insurer’s ability to effectively serve its clients. In fact, a high RES signals that the firm might need to pay a higher amount in losses compared to what it could potentially sell to raise cash. The measure of reinsurance (REINS) is the ratio of ceded premiums to gross premiums (Bressan & Du, 2024a). Ceded premiums are the premiums that an insurer pays to another company (the reinsurer) to cover a portion of its liabilities, while gross premiums are the total premium an insurance company receives from its policyholders. A high REINS indicates that the insurer transfers a large portion of its premium to a reinsurer to mitigate its risk exposure, thus ensuring financial stability.

The regression models control for additional firm-specific aspects. These include financial leverage, measured with the ratio of debt to total assets (DEBT), and profitability, assessed with the return on assets (ROA), i.e., the ratio of net income to total assets. Furthermore, the ratio of gross premiums to assets (GP) accounts for the underwriting business, while the ratio of investment income to assets (INV) captures to what extent the firm generates income by investing policyholder funds in various assets, so to earn returns through interest, dividends, and capital appreciation.

The definition of the variables is summarized in

Table 2. Definition of variables.

Variables	Definition
GHG1	Log of total scope 1 greenhouse gas emissions.
GHG2	Log of total scope 2 greenhouse gas emissions.
GHG3	Log of total scope 3 greenhouse gas emissions.
CR	Combined ratio, i.e., the sum of incurred losses, loss adjustment expenses, plus other underwriting expenses, divided by earned premiums.
PS	Premium-to-surplus ratio, i.e., the ratio of net premiums written to policyholder surplus. Policyholder surplus is total assets minus total liabilities.
RES	Ratio of reserves to policyholder surplus.
REINS	Ratio of ceded premiums to gross premiums.
DEBT	Ratio of total debt to total assets.
ROA	Ratio of net income to total assets.
GP	Ratio of gross premiums to total assets.
INV	Ratio of investment income to total assets.
ESG	ESG score of the company.

. After winsorizing the variables at the 1st and 99th percentiles, descriptive statistics and pairwise correlation coefficients are reported, respectively, in

Table 3. Descriptive statistics.

	Mean	Min	Max	Std. Dev.
Scope 1 emissions (tons CO2e)	21,645	0	3,157,004	129,130
Scope 2 emissions (tons CO2e)	43,765	0.74	2,071,311	111,439
Scope 3 emissions (tons CO2e)	514,961	5.48	13,800,000	1,003,674
CR	92.8301	31.0756	184.2003	15.9115
PS	1.0817	0.0014	4.3780	0.7228
RES	5.6780	0.0000	37.0878	6.5525
REINS	14.1301	0.0000	74.3503	14.6476
DEBT	0.0830	0.0000	0.5569	0.1051
ROA	0.0198	0.0120	0.1591	0.0284
GP	0.2647	0.0000	1.0993	0.2086
INV	0.0274	0.0015	0.1532	0.0244
ESG	42.8500	2.0000	91.0000	20.2901

The estimation period is 2004–2023, except for ESG, which is available in the period 2013–2023. Table 2 defines the variables.

and

Table 4. Pairwise correlation coefficients.

	GHG1	GHG2	GHG3	CR	PS	RES	REINS	DEBT	ROA	GP	INV	ESG
GHG1	1.000
GHG2	0.8840 ***	1.0000
GHG3	0.8901 ***	0.9074 ***	1.0000
CR	0.2191 ***	0.2264 ***	0.2124 ***	1.0000
PS	0.1724 ***	0.1690 ***	0.1741 ***	0.2493 ***	1.0000
RES	0.1461 ***	0.1492 ***	0.2008 ***	0.1800 ***	0.30501 ***	1.0000
REINS	−0.2540 ***	−0.2856 ***	−0.3347 ***	0.0400	−0.2334 ***	−0.0781 **	1.0000
DEBT	0.1000 ***	0.1080 ***	0.0760 ***	−0.0651 *	−0.1581 ***	−0.2223 ***	−0.0863 ***	1.0000
ROA	−0.0521 **	−0.0360	−0.0431 *	−0.4554 ***	−0.0880 ***	−0.3552 ***	−0.0890 ***	0.0561 **	1.0000
GP	−0.2567 ***	−0.2586 ***	−0.3041 ***	0.0574	0.4165 ***	−0.4293 ***	0.1941 ***	−0.1410 ***	0.2060 ***	1.000
INV	0.0370	0.0770 ***	0.0720 ***	−0.0150	−0.0160	0.1980 ***	−0.1841 ***	−0.0864 ***	0.1051 ***	−0.1781 ***	1.0000
ESG	0.3680 ***	0.4240 ***	0.5590 ***	0.0401	−0.0241	0.2010 ***	−0.1944 ***	−0.0258	−0.1119 ***	−0.2188 ***	−0.0315	1.0000

The estimation period is 2004–2023, except for ESG, which is available in the period 2013–2023. Table 2 defines the variables. *, **, and *** indicate significance at the 10%, 5%, and 1% level, respectively.

.

To develop an initial understanding of the relationship between greenhouse gas emissions and financial characteristics of insurers, the averages of the variables are calculated for subsamples of firms with low/medium/high emissions and indicated in

Table 5. Averages of variables for insurers with low/medium/high greenhouse gas emissions.

GHG3	CR	PS	RES	REINS
Low	89.8252	0.9282	3.7338	19.1046
Medium	93.8303	1.110	6.1656	12.4263
High	97.5959	1.4910	8.1446	8.8308
Dunn test (Low vs. High)	35.6430 ***	36.9988 ***	67.6731 ***	55.5707 ***

The estimation period is 2004–2023. Table 2 defines the variables. The Dunn’s test (Dunn, 1964) performs pairwise comparisons between firms with low and high emissions. The statistics have a chi-squared distribution. *** indicate significance at the 1% level.

. The subsamples are defined every year using the sample distribution. That is, high-emissions insurers have GHG3 above or equal to the 90th percentile, medium-emissions insurers have GHG3 between the 30th and the 90th percentile, while low-emissions insurers have GHG3 below or equal to the 30th percentile. The numbers reveal that high-emissions insurers are financially weaker (HP1), since their combined ratio and the premium-to-surplus ratio are, respectively, 9% and 62% higher than those of low-emissions insurers. In addition, reserves are considerably higher among high-emissions firms (HP3), while the insurance purchased is lower (HP4). Based on Dunn’s test (Dunn, 1964), all these differences are statistically significant at the 1% level of significance. Similar comparisons are also found in all three subsamples when the firms are classified according to the distribution of GHG1 and GHG2. However, for brevity, these outcomes are omitted while remaining available upon request.

To test the working hypotheses HP1 and HP2 formally, the following equation estimates the effect on the financial stability of bank j in year t from its greenhouse gas emissions:

Financial stability_f,j,t = α + β GHG_s,j,t + γControls_j,t + Fixed effects + ϵ_j,t.

(1)

The subscript f stays, alternatively, for the combined ratio (CR) and the premium-to-surplus ratio (PS). The subscript s denotes the scope of the emissions, as the model is tested separately for GHG1, GHG2, and GHG3. The set of control variables includes DEBT, ROA, GP, and INV. Fixed effects capture characteristics that are invariant across time and geographic region (Africa, Asia–Pacific, Europe, Latin America and the Caribbean, the Middle East, the United States, and Canada), while the standard errors are clustered by the firm. Finally, the term α and the term ϵ represent, respectively, a constant and an error term. The working hypothesis HP1 predicts that the coefficient β is positive in both models for CR and PS, as the two quantities are inversely related to corporate financial stability, meaning that firms become more unstable when their greenhouse gas emissions increase. Moreover, according to the working hypothesis HP2, this pattern is homogeneous across emission scope 1, scope 2, and scope 3.5

To test the working hypothesis HP3, the following equation estimates the effect of the greenhouse gas emissions of the bank j in year t on its reserves (RES):

RES_j,t = ϕ + δGHG_s,j,t + ψControls_j,t + Fixed effects + ζj,t.

(2)

Again, the model includes controls, fixed effects, a constant (ϕ) and an error term (ζ). HP3 predicts that the coefficient on δ is positive, indicating that with growing emissions, the insurer increases its reserves with respect to the available surplus.

Finally, the validity of the working hypotheses HP4 is verified with the equation of reinsurance purchased (REINS) on greenhouse gas emissions:

REINS_j,t = υ + σGHG_s,j,t + ρControls_j,t + Fixed effects + κ_j,t.

(3)

The explanatory variables have the same composition as in the previous models. Based on HP4, the coefficient on σ is negative, revealing that a high-polluting firm gives low premiums to reinsurers, thus showing a higher level of risk retention.

4. Results

Table 6. Regressions in insurer financial stability (combined ratio) on greenhouse gas emissions.

	(1) CR	(2) CR	(3) CR	(4) CR	(5) CR	(6) CR
GHG1	1.7106 ***	0.9259 *
	(0.4591)	(0.4732)
GHG2			1.9137 ***	1.1961 **
			(0.4982)	(0.4862)
GHG3					1.8373 ***	1.0286 *
					(0.552)	(0.552)
DEBT		−0.0971		−0.1132		−0.0923
		(0.0883)		(0.0890)		(0.0871)
ROA		−0.0489 ***		−0.0485 ***		−0.0487 ***
		(0.0073)		(0.0073)		(0.0073)
GP		0.3415 ***		0.3458 ***		0.3506 ***
		(0.0687)		(0.0687)		(0.0690)
INV		0.9824 *		0.9391		0.9647 *
		(0.5651)		(0.5725)		(0.5741)
Constant	79.2764 ***	80.2902 ***	75.4122 ***	75.9698 ***	71.2112 ***	74.6378 ***
	(4.3083)	(7.2181)	(5.1771)	(7.8864)	(7.1811)	(9.2711)
Fixed effects	Yes	Yes	Yes	Yes	Yes	Yes
Observations	1154	939	1154	939	1154	939
R-squared	0.0483	0.5871	0.0511	0.5916	0.0453	0.5874

The table reports estimates from panel regression models for CR according to Equation (1). The estimation period is 2004–2023. Table 2 defines the variables. Standard errors in parentheses are clustered at the firm level. *, **, and *** indicate significance at the 10%, 5%, and 1% level, respectively.

and

Table 7. Regressions in insurer financial stability (premium-to-surplus ratio) on greenhouse gas emissions.

	(1) PS	(2) PS	(3) PS	(4) PS	(5) PS	(6) PS
GHG1	0.0692 *	0.1439 ***
	(0.0351)	(0.0271)
GHG2			0.0735 **	0.1475 ***
			(0.0352)	(0.0300)
GHG3					0.0762 **	0.1600 ***
					(0.0341)	(0.0309)
DEBT		−0.0064 **		−0.0083 **		−0.0077 **
		(0.0031)		(0.0031)		(0.0032)
ROA		−0.0006 ***		−0.0006 ***		−0.0006 ***
		(0.000)		(0.000)		(0.000)
GP		0.0234 ***		0.0233 ***		0.0238 ***
		(0.0022)		(0.0022)		(0.0030)
INV		0.0098		0.0102		0.0074
		(0.0115)		(0.0120)		(0.0120)
Constant	0.5517 **	−0.0617	0.4246	−0.3139	0.1987	−0.8351 *
	(0.2504)	(0.3097)	(0.2937)	(0.3683)	(0.3763)	(0.4289)
Fixed effects	Yes	Yes	Yes	Yes	Yes	Yes
Observations	1397	1074	1397	1074	1397	1074
R-squared	0.0303	0.4261	0.0288	0.4204	0.0300	0.4311

The table reports estimates from panel regression models for PS according to Equation (1). The estimation period is 2004–2023. Table 2 defines the variables. Standard errors in parentheses are clustered at the firm level. *, **, and *** indicate significance at the 10%, 5%, and 1% level, respectively.

report the estimates of the model (1), testing the effect of greenhouse gas emissions on the combined ratio and the premium-to-surplus ratio of the insurers. The estimated sign on the three types of emissions is always positive and significant, supporting the validity of HP1 and HP2. That is, insurers are financially unstable as their emission scopes (scope 1, scope 2, and scope 3) increase. The signs on the control variables indicate that insurers are more stable as they are also more profitable, less levered, earn high income from their investments, and underwrite high premiums.

Table 8. Regressions in insurer reserves on greenhouse gas emissions.

	(1) RES	(2) RES	(3) RES	(4) RES	(5) RES	(6) RES
GHG1	0.4844 *	0.5716 *
	(0.2462)	(0.3041)
GHG2			0.5123 **	0.6999 **
			(0.2581)	(0.3252)
GHG3					0.7417 **	0.7217 *
					(0.3061)	(0.3796)
DEBT		−0.1538 ***		−0.1605 ***		−0.1545 ***
		(0.0352)		(0.0341)		(0.0338)
ROA		−0.0064 ***		−0.0064 ***		−0.0063 ***
		(0.0011)		(0.0011)		(0.0011)
GP		−0.1006 ***		−0.1004 ***		−0.0988 ***
		(0.0354)		(0.0338)		(0.0338)
INV		0.5331 **		0.5165 **		0.5098 **
		(0.2581)		(0.2551)		(0.2474)
Constant	1.8239	4.1109	0.9367	1.9880	−3.1554	0.3071
	(1.9435)	(3.7257)	(2.4062)	(4.2315)	(3.5746)	(5.4766)
Fixed effects	Yes	Yes	Yes	Yes	Yes	Yes
Observations	2043	1292	2043	1292	2043	1292
R-squared	0.021	0.422	0.022	0.427	0.043	0.429

The table reports estimates from panel regression models for RES according to Equation (2). The estimation period is 2004–2023. Table 2 defines the variables. Standard errors in parentheses are clustered at the firm level. *, **, and *** indicate significance at the 10%, 5%, and 1% level, respectively.

presents the estimates of the model (2). The results support the validity of HP3, suggesting that insurance reserves increase when the firm is more exposed to carbon risk.

Finally,

Table 9. Regressions in insurer-purchased reinsurance on greenhouse gas emissions.

	(1) REINS	(2) REINS	(3) REINS	(4) REINS	(5) REINS	(6) REINS
GHG1	−1.8910 ***	−2.0774 ***
	(0.6423)	(0.6263)
GHG2			−2.2211 ***	−2.5145 ***
			(0.5431)	(0.6286)
GHG3					−2.7073 ***	−2.7977 ***
					(0.5464)	(0.6537)
DEBT		−0.0092		0.0173		−0.0016
		(0.0831)		(0.0816)		(0.0784)
ROA		−0.0057		−0.0061		−0.0067
		(0.0063)		(0.0063)		(0.0063)
GP		0.0800		0.0845		0.0765
		(0.0756)		(0.0691)		(0.0641)
INV		−1.0237 **		−0.9368 *		−0.8796 *
		(0.5086)		(0.5154)		(0.4955)
Constant	29.0578 ***	23.0737 ***	34.4906 ***	30.3081 ***	46.1347 ***	38.7178 ***
	(5.1553)	(7.4666)	(5.2071)	(7.4489)	(6.7525)	(8.3765)
Fixed effects	Yes	Yes	Yes	Yes	Yes	Yes
Observations	1744	1182	1744	1182	1744	1182
R-squared	0.0654	0.1847	0.0819	0.1970	0.1124	0.2145

The table reports estimates from panel regression models for REINS according to Equation (3). The estimation period is 2004–2023. Table 2 defines the variables. Standard errors in parentheses are clustered at the firm level. *, **, and *** indicate significance at the 10%, 5%, and 1% level, respectively.

shows the estimates of the model (3) testing HP4. The negative and significant sign on greenhouse gas emissions reveals that high-polluting insurers cede a small share of their premiums to reinsurers, i.e., they retain high risk.

To stress the robustness of HP1 and HP2, alternative measures of greenhouse gas emissions are tested. In

Table 10. Regressions in insurer financial stability on lagged greenhouse gas emissions.

	(1) CR	(2) CR	(3) CR	(4) PS	(5) PS	(6) PS
GHG1t−1	1.5988 ***			0.0834 **
	(0.4445)			(0.0384)
GHG2t−1		1.8221 ***			0.0871 **
		(0.4711)			(0.0407)
GHG3t−1			2.0444 ***			0.0945 **
			(0.5352)			(0.0391)
Constant	78.0731 ***	68.1910 ***	64.1601 ***	0.3889	−0.0660	−0.3139
	(5.0691)	(6.7496)	(7.7151)	(0.2597)	(0.4607)	(0.5108)
Observations	1083	1083	1083	1314	1314	1314
R-squared	0.0823	0.0842	0.0900	0.1390	0.1343	0.1408

The table reports estimates from panel regression models for CR (columns 1 to 3) and PS (columns 4 to 6) according to Equation (1) using one-period lagged values of greenhouse gas emissions. The estimation period is 2004–2023. Table 2 defines the variables. Standard errors in parentheses are clustered at the firm level. **, and *** indicate significance at the 5%, and 1% level, respectively.

, the model (1) is estimated using lag values of one period of GHG1, GHG2, and GHG3, providing results consistent with the findings in Table 6 and Table 7. In

Table 11. Regressions in insurer financial stability on an alternative measure for greenhouse gas emissions.

	(1) CR	(2) CR	(3) CR	(4) PS	(5) PS	(6) PS
Scope1/Assets	0.0966 ***			−0.0016
	(0.0283)			(0.0063)
Scope2/Assets		0.1559			0.0217 **
		(0.1111)			(0.0108)
Scope3/Assets			0.0323 **			0.0054 ***
			(0.0164)			(0.0024)
Constant	90.7938 ***	86.1059 ***	88.5732 ***	0.8802 ***	0.8741 ***	0.7593 ***
	(3.1223)	(4.5971)	(3.4061)	(0.0996)	(0.0975)	(0.0932)
Fixed effects	Yes	Yes	Yes	Yes	Yes	Yes
Observations	1154	1154	1154	1397	1397	1397
R-squared	0.0430	0.0461	0.0473	0.1067	0.1617	0.2316

The table reports estimates from panel regression models for CR (columns 1 to 3) and PS (columns 4 to 6) according to Equation (1) using the ratio of greenhouse gas emissions on total assets. The estimation period is 2004–2023. Table 2 defines the variables. Standard errors in parentheses are clustered at the firm level. **, and *** indicate significance at the 5%, and 1% level, respectively.

, we perform regressions on the ratio of greenhouse gas emissions to total assets (Ali et al., 2023). The sign of emissions is still positive, with scope 3 emissions revealing a more strongly significant impact on insurer stability. Therefore, these additional outcomes support the robustness of the baseline results and the plausibility of the working hypotheses.

Finally, to verify the robustness of the findings more carefully, ESG ratings are employed. The data source used in the analysis provides environmental, social, and governance (ESG) ratings starting from 2013. Therefore, we can only focus on a subsample of the initial dataset. The variable ESG ranges from 0 to 100, and a higher number means that the firm has a stronger ESG performance. As the subsample is relatively small, the estimates are conducted using the pooled observations.

Table 12. Regressions in insurer financial stability on greenhouse gas emissions controlling for ESG scores.

	(1) CR	(2) CR	(3) CR	(4) PS	(5) PS	(6) PS
ESG	−0.0316	−0.0388	−0.0824 **	−0.0043	−0.0044	−0.0067
	(0.0368)	(0.0370)	(0.0410)	(0.0040)	(0.0040)	(0.0051)
GHG1	1.4936 **			0.0797 *
	(0.7091)			(0.0496)
GHG2		1.8049 **			0.0830 *
		(0.7486)			(0.0486)
GHG3			2.1570 **			0.1028 *
			(0.9526)			(0.0537)
Constant	81.6760 ***	77.4440 ***	69.9026 ***	0.5907 *	0.4661	0.0867
	(6.9206)	(7.9767)	(11.7216)	(0.3206)	(0.3756)	(0.5339)
Observations	543	543	543	660	660	660
R-squared	0.0321	0.0426	0.0437	0.0387	0.0344	0.0400

The table reports estimates from panel regression models for CR (columns 1 to 3) and PS (columns 4 to 6) according to Equation (1) controlling for ESG scores (ESG). The estimation period is 2013–2023. Table 2 defines the variables. Standard errors in parentheses are clustered at the firm level. *, **, and *** indicate significance at the 10%, 5%, and 1% level, respectively.

shows a pattern similar to the baseline results found in Table 6 and Table 7. That is, despite controlling for the ESG score, greenhouse gas emissions remain negatively related to insurer financial stability, with a significant impact on all regressions in Table 12. The negative sign on ESG indicates that ESG scores influence in a positive way the financial stability of insurers. However, in statistical terms, this effect is quite small. The results do not change in quality either, as we tested the environmental pillar separately or when we interacted greenhouse gas emissions with ESG ratings. As all these interaction terms were not statistically significant, we did not report the results in the table, but we made them available on request.

5. Discussion and Conclusions

This article examines global insurance companies from 2004 to 2023. The results of the panel regression show that insurers that increase their greenhouse gas emissions become financially unstable in all emission scopes (scope 1, 2, and 3). Higher emissions are related to larger reserves and lower premiums ceded to reinsurers, indicating that high-polluting insurers retain significant carbon risk, which harms their financial stability.

Few important implications can be drawn from this research. First, the findings suggest that actions aimed at reducing the carbon footprint of insurers are necessary to maintain financial stability, preventing the burden of carbon risks from falling on governments and individuals. This insight recommends that insurance managers redirect the major capital flows associated with investments and underwriting companies to carbon-neutral activities to make the company financially healthy (Braun et al., 2019). In fact, stable financial conditions could eventually translate into more favorable capital costs (Chava & Purnanandam, 2010), efficiency (Nguyen & Nghiem, 2015), and optimal decision-making (Brown et al., 2016). Moreover, evidence that reinsurance is decreasing in relation to insurer carbon footprints raises concerns about the sustainability of insurance against larger hazards, which is a pressing issue that has already been highlighted in public debates (Financial Times, 2024; S&P, 2024). Therefore, the insight for financial authorities is that climate change policies aimed at reducing the carbon footprints of insurers would not only positively impact the environment but also improve the solvency of insurers, with benefits in terms of insurance availability and affordability. At the same time, however, insurers could end up withdrawing from carbon-intensive sectors, thereby increasing the protection gaps. This is an important trade-off that policy makers should carefully consider.

There are a few limitations in the results of the current study. First, like most studies in this field, this research has to deal with the issue of the quantity and quality of data on corporate greenhouse gas emissions. In fact, due to the lack of homogeneity in carbon accounting worldwide, the results suffer from different legislation and practices. Evidence shows that greenhouse gas emissions are often complex to measure, inaccurate, and also facilitate greenwashing practices, making it more difficult to obtain a reliable picture of the carbon risk entailed by corporations (Gatzert et al., 2020; Pitrakkos & Maroun, 2020; Bajic et al., 2023; Callery, 2023; Gheyathaldin Salih, 2024).

Insurers, in particular, following the principles of the GHG Protocol (The Greenhouse Gas Protocol Initiative, 2024) and the PCAF Global Standard for Managing and Reporting GHG (Partnership for Carbon Accounting Financials, 2020), will need to adopt a systematic approach to calculate scope 1, 2, and 3 emissions. However, an insurer’s clients (that is, supply chain partners) may be located in different jurisdictions and can range from large corporations that have well-established disclosure and reporting standards to small businesses and individuals with no formal data collection processes. From such a diverse base, it seems very difficult to obtain data and calculate insurance-associated emissions (KPMG, 2023).

Future research could use broader financial risk indicators and measures for default-like z-scores (Fiordelisi & Marques-Ibanez, 2013; Chiaramonte et al., 2016), or the distance to default (Bharath & Shumway, 2008; Jessen & Lando, 2015) to provide broader and more robust evidence for these results. Our analysis involved several insurance segments to obtain a sufficiently large sample. However, the availability of larger datasets for single segments of insurance would allow us to address the research hypotheses in a more specific way for different businesses. Finally, extending the baseline models by including variables capturing physical risk exposure (e.g., catastrophic loss ratios, weather-related claims) would also help understand more deeply the relevant role of greenhouse gas emissions in financial dimensions of insurers. All these extensions are left to scholars for their future research agendas.