The Impact of Climate Risk on Insurers’ Sustainable Operational Efficiency: Empirical Evidence from China

Abstract

The operational efficiency of insurance companies is crucial for their long-term stability and sustainable development. Climate risk has emerged as a significant factor affecting insurers’ operational performance in the context of global climate change and sustainable development goals. Although prior research provides a solid foundation, further exploration is needed to clarify how climate risk influences insurers’ efficiency and underlying mechanisms. This paper uses panel data from 248 Chinese insurance companies spanning 2011 to 2021 to construct a climate risk indicator and systematically examines the potential pathways through which this indicator influences operational efficiency. Precisely, absolute temperature deviation measures physical climate risk, and an entropy-weighted method captures climate transition risk; the DEA model evaluates operating efficiency. A fixed-effects model reveals that physical climate risk may adversely affect operational efficiency, while climate transition risk demonstrates a U-shaped relationship with efficiency. Mechanism analysis shows that physical climate risk increases exposure to natural disaster losses, whereas transition risk may encourage green insurance development. Heterogeneity emerges across insurer types and between coastal and non-coastal regions, with resilient infrastructure mitigating the adverse effects of physical risks and insurance technology driving gradual transformation to offset initial transition risks. Overall, this study expands the perspective on how climate risk shapes the insurance industry’s sustainable development, offering theoretical and practical insights for policymakers to optimize risk management and promote green finance.

1. Introduction

An insurance company’s operational efficiency is considered significant to its sound operation and sustainable development, directly affecting profitability, risk management, and market competitiveness. Efficient operations can optimize resource allocation, improve capital utilization efficiency, and enhance claims-paying capacity, ensuring sound business growth and long-term development. Conversely, inefficient operations may lead to rising costs, liquidity constraints, and underwriting risk imbalance, weakening market competitiveness and even affecting industry stability. At the industry level, the improved operating efficiency of insurance companies not only enhances industry resilience and optimizes capital allocation but also promotes product innovation, making the insurance industry more adaptable and providing more solid risk protection for the economy and society. In addition, the sound operation of the insurance industry is crucial to developing green finance and the transition of the economic system towards sustainability [1].

However, the uncertainties driven by climate change in recent years are reshaping the operating environment of the insurance industry, with the frequent occurrence of extreme weather, escalating disaster losses, and an accelerated low-carbon transition posing serious challenges to insurers’ business models and profitability. Global climate change intensifies its influence on sustainable socio-economic development, as extreme weather, rising sea levels, and ecological degradation emerge as significant global risks. Over the next decade, climate risk is consistently projected to be the most threatening risk category worldwide, with extreme weather events identified as the primary factor most likely to trigger a major crisis [2]. Data indicate that between 2000 and 2019, 11,000 extreme weather events occurred worldwide, resulting in 475,000 fatalities and economic losses totaling USD 2.56 trillion [3]. In recent years, the frequency and severity of natural disasters have continued to escalate, exemplified by Hurricane Daniel in the Mediterranean Sea and Tropical Cyclone Freddie in 2023, both of which resulted in substantial economic losses, and global temperatures are projected to reach a record high in 2024 (Figure 1). As a region susceptible to climate change, China also faces significant impacts, with 10.539 million hectares of crops affected by natural disasters in 2023 and direct economic losses exceeding RMB 330 billion [4,5]. The insurance industry bears a heightened level of climate risk exposure. As a result, it remains one of the financial sectors most directly affected by climate change. Scholars typically classify climate risk into two main components: physical and transition risks [6]. Physical climate risk originates from extreme weather events driven by climate change. Examples include typhoons, floods, and heat waves. These events worsen economic losses, increase insurers’ claims, and harm liquidity and profitability [7]. In contrast, transition risks primarily stem from low-carbon policy reforms, technological advancements, and shifting market expectations. These could trigger a decline in the value of high-carbon assets, curb investment returns, and prompt a restructuring of insurance product portfolios [8]. In this context, insurance companies must address both types of climate risks to maintain operational efficiency, highlighting the importance of investigating how climate risks affect insurers’ performance.

The research contribution of this paper is reflected in the following aspects: First, the existing studies mainly focus on the impact of climate risk on the insurance market, such as insurance demand, premium income, and other macro-level indicators, while there is a relative lack of micro-level exploration of how it plays a role in the internal operational efficiency of insurance companies. In addition, the literature has paid more attention to physical climate risk, while there is a lack of research on climate transition risk and its mechanism. This study conducts a micro-level analysis by systematically assessing the distinct effects of physical and transition climate risks on the operational efficiency of insurance companies, thereby offering a novel perspective on how climate risks influence the insurance industry. In addition, a comprehensive analytical framework is developed to elucidate the underlying mechanisms by which climate risk affects the operational efficiency of insurance companies. Specifically, physical climate risk can precipitate large-scale natural disaster losses, increasing insurers’ payout burdens and undermining their operational stability. In contrast, climate transition risk reshapes the insurance company’s business model and promotes the industry’s restructuring through the changes in the green insurance market. Third, this paper further introduces the factor of government intervention and examines the role of disaster-resistant infrastructure development and insurance technology development in mitigating the impact of climate risk. By emphasizing the government’s moderating role in responding to climate risk, this study enriches the policy-oriented perspective on how the insurance industry adapts to climate change. It provides empirical evidence for the government to optimize climate risk management policies and enhance the industry’s adaptive capacity. Finally, this study investigates the heterogeneous effects of climate risk on the operational efficiency of insurance companies by examining variations across company types and regional disparities, thereby offering a more comprehensive perspective on the adaptation strategies employed in diverse market environments and business models.

The rest of the paper is organized as follows: Section 2 provides a literature review of relevant studies; Section 3 presents the theoretical analysis and research hypotheses; Section 4, provides the variables, data sources, and measurements of the regression model; Section 5 presents the main findings; and Section 6 concludes and provides policy recommendations.

2. Literature Review

2.1. Factors Affecting the Operational Efficiency of Insurance Companies

Scholars have mainly explored the influencing factors of insurance companies’ operational efficiency in micro- and macro-level dimensions.

Micro-level research focuses on how internal governance, capital structure, and business models affect the efficiency of insurance companies. Firm size has always been an important factor affecting the operational efficiency of insurance companies. Wulandari and Cabanda found that insurance solvency and asset size negatively correlate with operational efficiency, suggesting that too large a size may increase management costs and reduce operational efficiency [10]. However, Fernández et al. examined the competitive advantages of firms of different sizes. They concluded that larger insurers benefit from economies of scale, while smaller firms improve market adaptability under flexibility [11]. In the Chinese market, Yao et al. point out that there are prominent economies of scale in the insurance market, and larger insurers have stronger profitability and monopoly advantage [12]. Regarding cost efficiency, Cummins et al. analyzed U.S. life insurers through the DEA–Malmquist index. They found that M&A can reduce operating costs and improve efficiency, especially for financially vulnerable life insurers [13]. Cummins and Weiss further compared the efficiency of integrated insurers with that of specialized property–casualty insurers and found that integrated operation can improve cost efficiency. It was found that while consolidation improves cost efficiency, profitability is relatively low, and life insurers are more efficient [14]. In terms of corporate governance, Wang et al. studied the effects of board structure, voting rights, and ownership on the efficiency of insurance companies. They found that these governance factors significantly affect the efficiency of property and casualty insurers but have a more negligible effect on life insurers [15]. Huang et al. concluded that the operating efficiency of U.S. insurance companies is affected by the proportion of independent directors, board size, and shareholding structure [16].

Macro-level studies focus on the systematic impact of economic cycles, market competition, and regulatory policies on insurers’ efficiency. Fluctuations in the economic cycle profoundly impact the underwriting business and capital management of insurance companies. Eling and Schaper and Li et al. show that the underwriting cycle of insurance companies is positively correlated with the economic cycle in the long run and that periods of recession may lead to a decline in premium income. In contrast, periods of economic growth promote the expansion of the insurance business [17,18]. Nourani et al. show that the efficiency of insurance companies in the two regions is higher when the financial markets are more liberalized and that the efficiency level of insurance companies in the two regions is higher when they are more liberalized. Nourani et al., on insurance markets in Asia and Europe, show that the higher the degree of liberalization of financial markets, the higher the efficiency of insurance companies, and the gap between the efficiency levels of insurance companies in the two regions is gradually narrowing [19]. Regarding globalization and market competition, Olasehinde-Williams and Balcilar found that economic globalization has brought opportunities for insurance companies in emerging markets and facilitated their internationalization. However, it has also intensified the pressure of market competition. Changes in regulatory policy are also an important factor affecting insurers’ efficiency [20]. Cummins and Rubio-Misas and Weiss and Choi find that higher regulatory standards may limit market competition. However, insurers are more likely to find profitability opportunities in a competitive or lax regulatory environment [21,22]. Overall, changes in the macroeconomic environment shape the competitive landscape and profitability patterns of the insurance industry and, to some extent, determine the operational efficiency of insurance companies. Overall, changes in the macroeconomic environment have shaped the competitive landscape and profitability model of the insurance industry and, to a certain extent, determined the operational efficiency of insurance companies. However, with the intensification of climate change, extreme weather events and low-carbon transition policies are increasingly becoming important factors affecting the insurance industry. Existing research on the factors affecting insurance companies still focuses on the traditional areas of macroeconomics, market competition, and regulatory policies, and research on how climate risk plays a role in the operational efficiency of insurance companies is still limited and lacks a systematic empirical analysis.

2.2. Impact of Climate Risk

Despite a continued rise in losses due to climate risk, its impact on the global economy and financial system is increasingly far-reaching and has become a significant source of systemic risk (NGFS) [23]. The discussion around the impacts of climate risk on the financial sector is increasingly broad, focusing on three main areas: securities, banking, and insurance.

Several scholars have investigated the multifaceted impacts of climate risk from various perspectives in the securities industry. On the stock market side, studies have shown that climate policy uncertainty significantly increases market volatility; e.g., Lasisi et al. find that climate policy uncertainty increases stock market volatility by changing investor expectations [24]. At the same time, Guo further points out that extreme weather events may trigger systemic risk and lead to a contraction in trading volume [25]. In terms of market pricing mechanisms, Bolton and Kacperczyk find that there is a significant “carbon premium” in the stocks of high-carbon-emitting firms, suggesting that investors are demanding compensation for climate risk, a finding that is confirmed in the Chinese market [26], and Wu et al. confirm that firms’ performance briefings with higher climate risk trigger a more vigorous market reaction [27]. Meanwhile, Xu et al. reveal a U-shaped, nonlinear relationship between climate transition risk and fossil energy stock returns in China [28]. On the investment opportunity dimension, the low-carbon transition has created structural opportunities. A case study by Sun et al. shows that mining firms with suitable governance structures and control of environmental costs remain financially successful in the face of climate risk [29]. In contrast, international comparisons show that renewable energy investments (e.g., solar, wind) have outperformed in risk-adjusted returns [30]. Notably, Krueger et al. find that more than 60% of institutional investors respond to climate risk through active risk management rather than simple divestment, with large ESG-oriented institutions preferring shareholder engagement strategies [31]. On the bond market side, climate risk pricing mechanisms show complexity. Allman confirms that climate risk premiums are embedded in corporate bond yields, raising the cost of debt financing for high-emitting firms by an average of 0.8–1.2 percentage points, while research in the area of infrastructure finance shows that climate adaptation costs can increase the probability of default on a project by a factor of three to five [32]. For risk hedging, Engle’s team innovatively constructs dynamic climate risk hedging portfolios. The simulated portfolios constructed by extracting climate news shocks through textual analysis show effective hedging capabilities in and out of the sample [33].

From the banking sector’s perspective, both physical and transformational risks have put pressure on bank operations and performance, including increasing the share of non-performing loans [34], reducing total deposits [35], and greatly amplifying systemic risk and the likelihood of loan defaults [36]. Meanwhile, research by Campiglio et al. and Scholtens indicates that banks should incorporate climate risk into their traditional credit risk management systems while undertaking proactive changes. These changes involve supporting renewable energy and environmental protection projects by adopting green financial instruments and low-carbon loans, ultimately balancing risk mitigation and business growth [37,38]. Li et al. use a cross-country comparative approach to study the impact of climate risk on credit supply. They explicitly examine the differences between the private and public sectors [39]. Recently, Cardenas conducted a systematic review of current bank management practices to address climate risk, pointing out that further refinement of risk models and expansion of data sources are important directions for future research [40]. Overall, the existing literature suggests that climate risk both poses a serious challenge to the banking sector and creates new profit opportunities for banks in the green transition process, requiring the banking sector and regulators to continuously optimize risk assessment methods and internal management frameworks in order to enhance the long-term robustness of the financial system.

The impact of climate risk on the insurance industry has been an area of growing academic interest since 1997, when Tucker suggested that the insurance industry could be a potential advocate for mitigating climate change due to uncertainty about the timing and extent of climate change [41]. Since that seminal work, considerable debate has emerged regarding the role of the insurance sector in managing climate change risks. It is acknowledged for its capacity to enhance preparedness and proactive measures before climate events and for its effectiveness in mitigating the impacts and losses in the aftermath of such events [42,43]. The pivotal role of the insurance sector in disaster management is well documented [44]. The demand for insurance products evolves as the frequency of climate risks and extreme events increases and risk exposures widen. Existing research sheds light on how climate risk influences insurance demand [45,46], identifying key factors, including consumers’ willingness to pay, their awareness of risk [47,48], and regional education levels as critical determinants of public demand [49]. Moreover, several studies are closely related to this paper’s focus, including firm-level investigations within individual countries and cross-country comparative analyses. Several scholars have adopted a single-country perspective to examine how climate risks influence insurance companies at the firm level. Using data from 159 Chinese insurers, Gong finds that precipitation changes increase underwriting and investment risks, while temperature changes do not significantly affect insurers’ risk-taking behavior [50]. Further, Chen and Lin show that both physical and transition climate risks negatively impact the profitability of Chinese insurance companies [51]. Drawing on data from Korean insurers, Sewan demonstrates that physical climate risk significantly affects return on assets and solvency ratios, underscoring the financial vulnerability of insurance firms to climate-related shocks [52]. Other studies take a cross-country approach to examine how climate risks affect insurance companies across different institutional and regional contexts. Gatzert and Reichel empirically examine the determinants and value of climate-related risk and opportunity awareness among insurance companies in the United States and Europe. Their findings show that larger European and P&C insurers are more likely to manage climate-related risks and opportunities and that such awareness positively impacts firm value [53]. Li and Li, using insurance company data from 73 countries, find that the impact of climate change on the insurance industry is pervasive, systemic, and characterized by delayed transmission and predominantly short-term effects [54].

3. Theoretical Analysis and Research Hypotheses

3.1. Theoretical Analysis of Climate Risk on Insurers’ Efficiency

Climate physical risks mainly include acute extreme weather events (e.g., typhoons, floods, droughts) and chronic climate change (e.g., rising temperatures, sea level rise), which pose systemic challenges to insurers’ operational efficiency. First, the frequency of extreme weather events directly pushes up insurers’ claims expenses and increases pressure on liquidity, thus weakening the organization’s financial stability and daily operational capacity [7]. Second, physical climate risk also indirectly affects operational efficiency by affecting market supply and demand. Insurers often adopt strategies to increase premiums or tighten coverage in the face of rising potential losses in high-risk areas (IPCC, 2022). However, premium increases significantly reduce the willingness of low-income, low-risk groups to insure in low-risk areas, affecting the penetration of the insurance market [55,56,57]. Some studies have shown that frequent climate disasters may increase insurance awareness among the population, prompting businesses and individuals to increase insurance coverage [58]. However, sharp fluctuations in premium income may adversely affect insurance companies’ capital adequacy and operational efficiency. Moreover, some insurers in high-risk areas have withdrawn from the market due to persistent losses, leading to further concentration of regional risks and posing more significant challenges to the operational management of the industry as a whole [59]. On the other hand, climate physical risks can also alter the volatility of market asset prices and portfolio stability. Extreme weather events may impact markets such as stocks, bonds, and real estate, increasing market volatility and exposing investment assets held by insurers to valuation and liquidity risks [60]. For example, extreme weather events may cause damage to infrastructure and disruptions to business operations, which in turn may affect stock prices and bond credit ratings of the relevant industries, resulting in losses to an insurer’s investment portfolio [61]. At the same time, high claims demand may also force insurers to liquidate some of their long-term investment assets in the short term to meet funding needs; in unfavorable market environments, such asset sales at a discount can further reduce investment returns [62]. Although these effects are first reflected in profitability and investment returns, since profitability levels are closely related to fund flows, capital allocation, and operational management, their fluctuations will eventually be transmitted into a decline in overall operating efficiency. Based on this, we propose the hypothesis:

Hypothesis 1 (H1a).

Climate physical risk reduces insurers’ operational efficiency.

Climate transition risk refers to the uncertainties arising from policy adjustments, technological change, and changes in market demand during the global transition to a low-carbon economy. In the early stages of the transition, high-carbon assets may depreciate rapidly with the transition to a low-carbon economy, leading to a decline in the value of stranded assets and reducing the return on high-carbon assets held by insurers [8]. At the same time, insurers, as key underwriters of high-carbon sectors, have had to adjust their underwriting strategies quickly in response to increasingly stringent carbon regulations. Such adjustments often take the form of raising premium levels or reducing lines of business in order to manage risk in the new regulatory environment [63]. However, such rapid adjustments are often accompanied by higher policy restructuring costs, and insurers holding more brownfield assets are more sensitive to policy changes, making adjustments even more complex [64]. In addition, stringent carbon regulation could also push up underwriting risk in high-carbon sectors, forcing insurers to reduce the size of their underwriting in traditional sectors, leading to lower or sharply more volatile premium income [65]. In the early stages of the transition, insurers’ operating efficiency declined significantly due to a series of short-term shocks, including declining asset values, rising costs, unstable premium income, and the pressure to reorganize internal resources. With the gradual deepening of the low-carbon economic transition and the gradual stabilization of the market environment, insurance companies have adjusted their business structure and investment portfolios while adapting to the new policies and market demands. At this time, insurers are gradually opening up new profit growth points by expanding green insurance business (e.g., climate liability insurance, carbon credit insurance) [66,67] and optimizing their investment strategies to increase the allocation of sustainable investment (ESG) assets, to reduce the climate risk in their overall investment portfolios [68]. Although this transformation brings higher uncertainty and adjustment costs at the initial stage, insurance companies gradually establish a stable underwriting model and risk management system after adaptation and adjustment, diversify their business structure, and improve their operational efficiency. As the transition risk promotes changes in the enterprise’s internal structure and market positioning, it will eventually be transformed into a positive force that enhances overall operational efficiency. Based on this, we propose the hypothesis:

Hypothesis 1b (H1b).

Climate transition risk exhibits a U-shaped, nonlinear relationship with the operational efficiency of insurance companies.

3.2. Mechanisms of Climate Risk Transmission to Insurers’ Efficiency

Natural catastrophe loss exposures reflect an insurer’s exposure to potential and actual losses due to extreme weather events. Natural loss exposure focuses on the threat of potential risks to the business’s assets, operations, and financial soundness and the size and scope of direct and indirect losses that could result from these risks. First, climate risk will increase the frequency and severity of natural disasters, and there is a cyclical relationship between climate risk and natural disasters such as droughts and floods [69,70]. The frequency of natural disasters will increase insurers’ payout expenses, leading to rising liquidity pressures [71,72]. Research shows that the amount of insurance claims due to natural disasters has been increasing globally in recent years, and some insurance companies in high-risk areas have even experienced solvency crises [73]. At the same time, reinsurance companies have increased their reinsurance premium rates due to frequent catastrophes, which has led to a further increase in the underwriting costs of primary insurers [74] and aggravated the financial pressure on the industry. In addition, traditional actuarial models based on historical data cannot accurately respond to the uncertainty of future climate risks, leading to pricing bias and imbalance between premium income and claims expenses [75], further weakening insurers’ operational efficiency. Extreme weather also causes a surge in demand for post-disaster claims, forcing insurers to invest more resources in risk assessment, data collection, and claims management [76]. These additional costs expose companies to large direct payouts in the event of a catastrophic event and to the managerial and operational pressures associated with exposure to natural disaster losses. In addition, physical climate risk also increases the difficulty of insurers’ operational management, requiring them to formulate more sophisticated underwriting strategies to cope with the significant differences in climate risk in different regions [77,78]. While such strategic adjustments are necessary for risk management, they also lead to higher operational costs and resource deployment difficulty, further reducing overall operational efficiency. Based on the above analysis, it can be seen that climate physical risk not only directly exacerbates claims and cost pressures by enhancing insurers’ exposure to catastrophe losses in high-risk environments but also indirectly transmits into a decline in overall operational efficiency by increasing operational management complexity. Therefore, we propose the following hypothesis:

Hypothesis 2a (H2a).

Climate physical risk can reduce the operating efficiency of insurance companies by increasing their exposure to natural disaster losses and thus reducing their operating efficiency.

The development of green insurance products is an advanced strategy by insurance companies to consider the threat of climate transition; green insurance aims to address environmental hazards and promote sustainability [79]. At the early stage of climate transition risks, the development of green insurance products faces a number of constraints that have a significant negative impact on the operational efficiency of insurance companies. First, green insurance relies on emerging low-carbon technologies and engineering innovations that are not supported by adequate historical loss data, leading to greater uncertainty in risk assessment and product pricing, making continuous claims data difficult to obtain, and thus making underwriting more difficult [80,81]. Secondly, the lack of uniform green product standards in the market is prone to the phenomenon of greenwash, which makes it difficult to accurately quantify the actual risks and benefits and further raises the cost of policy reconstruction and management [82]. In addition, in order to meet the requirements of the low-carbon transition, insurance companies need to invest a lot of resources in information collection, personalized risk assessment, and customer service, and these additional costs exacerbate the operational burden in the short term. Combined with the above factors, green insurance products at the initial stage often become an amplifier of operational risks, which ultimately leads to a decline in the overall operational efficiency of insurance companies. With the deepening of the low-carbon economic transition and the continuous improvement of policies and technologies, green insurance products are gradually getting rid of the initial uncertainty, which helps to improve the operating efficiency of insurance companies. First, with government incentives and green financial support, low-carbon industries such as renewable energy are developing rapidly, providing a good opportunity for insurers to develop new, niche businesses. By offering customized green insurance products, insurers that have entered this market earlier can not only reduce underwriting risks but also achieve long-term profitable growth. Second, environmentally and sustainability-conscious customers tend to exhibit lower risk profiles, and this “positive” adverse selection effect can help reduce underwriting risk, improve the risk structure of insurance products, and, in turn, increase operational efficiency [83]. Finally, by actively promoting green insurance, insurance companies can establish a green brand image, enhance market competitiveness, and further expand their market share. With the continuous improvement of the risk assessment and pricing system, the positive effect of green insurance will gradually appear in the later stage of the transition, thus promoting the improvement of the overall operational efficiency. Therefore, we propose the following hypothesis:

Hypothesis 2b (H2b).

Green insurance significantly mediates the nonlinear effect of climate transition risk on insurers’ efficiency.

3.3. The Role of Stock of Disaster-Resistant Infrastructure and Insurance Technology

Disaster-resistant infrastructure (DRI) mainly consists of disaster prevention and mitigation construction in various regions, such as levees, flood control systems, emergency communication equipment, and public safety facilities. These infrastructures can effectively reduce losses in natural disasters and mitigate the impact of disasters on businesses and society. Improved disaster resilience infrastructure for insurance companies can help reduce large-scale losses due to extreme weather events, thereby mitigating the negative impact of physical climate risks on operational efficiency. When a region has a high level of disaster resilience infrastructure, the extent of damage after a disaster will be relatively low and claims expenditure and the frequency of payments will be reduced. Easing the financial pressure on businesses to recover from a disaster helps stabilize the insurer’s liquidity and capital adequacy ratio. It also reduces the operation and management costs caused by disaster risk. Therefore, we propose the following:

Hypothesis 3a (H3a).

Resilient infrastructure mitigates the negative impacts of climate physics.

Insurance technology (IT) mainly uses advanced technologies such as digital tools, big data analytics, artificial intelligence, and blockchain to optimize risk assessment, pricing, claims management, and customer service. When the risk of climate transition is low (i.e., on the left half of the U-curve), insurance technology effectively helps insurers capture policy changes and market dynamics promptly by improving the efficiency of information collection, enhancing risk early warning capabilities, and optimizing data integration, thus adjusting underwriting strategies and risk pricing, and reducing the negative impacts of depreciation of high-carbon assets and policy adjustments [84]. At this point, advanced technology reduces internal management costs and policy restructuring costs. It enables companies to respond more flexibly to the uncertainties at the early stage of the transition, thus curbing the negative impact of climate transition risks on operational efficiency. However, when climate transition risk is at a high level (i.e., on the right half of the U-curve), although a high level of insurance can standardize the risk management process and improve pricing accuracy, excessive standardization and automation of technological systems tends to entrench established decision-making patterns, limiting firms’ flexibility in responding to market opportunities, which in turn dampens positive incentives insurers can gain through business innovation and investment restructuring [85]. In addition, over-reliance on technology may exacerbate information mismatch between the public and insurers, weakening the public’s sense of fairness in transactional behavior and thus reducing trust in insurers, further inhibiting positive effects [86]. In sum, in moderating the relationship between climate transition risk and operational efficiency, insurance technology effectively mitigates adverse shocks in the low-risk phase and dampens positive incentives in the high-risk phase, thus reducing the volatility of the impact of climate transition risk on insurers’ operational efficiency.

Hypothesis 3b (H3b).

Insurance technology negatively moderates the impact of climate transition risk on insurers’ efficiency.

4. Research Design

4.1. Sample Selection and Data Sources

This study empirically analyzes the impact of climate risk on the operational efficiency of insurance companies based on panel data of China’s insurance industry from 2011 to 2021. The choice of this time interval is based on the following two considerations: On the one hand, since 2011, China’s policy system for addressing climate change has been gradually improved, and the concept of green development and the strategy of sustainable development has gradually gained popularity. The Twelfth Five-Year Plan (2011–2015) has incorporated climate change and energy saving into the national development strategy for the first time. Meanwhile, the frequent occurrence of extreme weather events at home and abroad has gradually made climate risk an important issue that the insurance industry needs to deal with in the long term. On the other hand, 2021 is the opening year of the 14th Five-Year Plan, marking a new stage in China’s green economy development, which not only continues the deepening of climate-related policies but also provides a complete database for assessing the long-term impact of climate risk on the insurance industry. Therefore, the study period covers the key stages from policy practice to market response and can effectively reveal the dynamic impact of climate risk on the efficiency of insurance companies.

The sample selected for this study includes insurance companies that continuously disclosed their operating data in the China Insurance Yearbook between 2011 and 2021, covering 197 insurance companies and their annual panel data. The sample of insurance companies includes life insurance companies, property and casualty insurance companies, and reinsurance companies, which can more comprehensively reflect the general impact and heterogeneous characteristics of climate risk on the insurance industry. During the sample construction process, companies with special treatment and data with significant outliers are excluded to ensure the robustness of the research findings.

The data in this study are mainly sourced from authoritative public databases to ensure the comprehensiveness and accuracy of the data. The climate physical risk data come from the National Center for Environmental Information (NCEI) under the National Oceanic and Atmospheric Administration (NOAA), which systematically records climate change and temperature fluctuations in different regions, providing a scientific basis for measuring the impact of extreme weather events. The climate transition risk data, on the other hand, integrate multidimensional information, including data on climate policy dynamics provided by the Global Climate Risk Integration Database (GCRID), data on public concern reflected in the Baidu Index, and changes in energy structure recorded in the China Energy Statistics Yearbook, ensuring the breadth and timeliness of data sources. Macroeconomic control variables are mainly from the China Statistical Yearbook published by the National Bureau of Statistics, and operational and financial data at the insurance company level are taken from the China Insurance Yearbook. Regarding mediating variables, data on disaster loss exposure are taken from the China Meteorological Disaster Yearbook, reflecting the pressure on insurers to pay out claims and the economic losses caused by natural disasters. At the same time, the Eps database provides data on green insurance to measure the insurance industry’s role in promoting sustainable development and low-carbon transition. In addition, the data related to the moderating variables are also sourced from authoritative statistical organizations. The data on disaster-resistant infrastructure come from the National Bureau of Statistics, which characterizes the construction of disaster prevention and mitigation capacity in each region. In contrast, the insurance science and technology data come from the Digital Finance Research Center of Peking University, which measures the application level of digital transformation and technological innovation in the insurance industry. By combining the above multidimensional data sources, the researchers ensure the data’s breadth, authority, and timeliness, laying a solid empirical foundation for an in-depth analysis of how climate risk affects the operational efficiency of insurance companies.

4.2. Definition of Variables

4.2.1. Explanatory Variables

In examining the impact of climate risk on the efficiency of insurance companies, this paper subdivided climate risk into two dimensions, climate physical risk and climate transformation risk, to fully portray the potential impact of climate change on the insurance industry. Climate physical risk mainly describes the direct environmental impacts due to extreme weather events or long-term climate change, which may exacerbate insurers’ payout pressure and affect their operational stability. On the other hand, climate transition risk reflects systemic changes due to policy adjustments, changes in consumer (investor) preferences, and technological innovations, which could profoundly impact insurers’ efficiency.

1.

Climate physical risk

In existing studies, scholars have used various methods to measure climate physical risk, and these methods have focused on measures of temperature, precipitation, and extreme weather events. Early studies typically used metrics such as temperature levels, magnitude of temperature change, precipitation, and frequency of storms as proxy variables for climate physical risk. For example, Deschênes and Greenstone explored how climate variables reflect the risks faced in economic activities by analyzing the effects of temperature and precipitation changes on agricultural output [87]; Dell et al. examined the negative effects of temperature fluctuations on economic growth using historical temperature anomaly data, thus indirectly revealing the potential of temperature risk on the overall economic system The potential threat of temperature risk to the whole economic system is indirectly revealed by Dell et al. [88]. On the other hand, Kousky illustrates the utility of using the frequency of extreme events as a metric by systematically reviewing the economic impact of extreme weather events (e.g., heat waves and heavy rainfall) on natural hazard risk [89]. Despite the explanatory power of these indicators in specific contexts, there are limitations to the direct use of temperature data or the frequency of extreme weather events, which may be influenced by geographic location, for example, leading to a decrease in comparability between different regions [90,91].

Against this background, scholars have begun to seek a more stable and comparable measure of climate risk. Wang et al. and Yang et al. used the absolute value of temperature deviation as a measure of the physical risk of climate in their study [92,93]. This method calculates the absolute value of annual temperature deviation based on the historical average temperature, thus eliminating, to a certain extent, the interference of geographic location on climate variables and making the climate risk between different regions more comparable.

This paper employs absolute temperature deviation values to measure climate physical risk, building on previous research. Specifically, it calculates the deviation between the average temperature of each province in a given year and the province’s annual average temperature over the sample period (2011–2021). Then, it takes the absolute value of this deviation. The calculation formula is presented below:

$$CP{R}_{ct}=\left|T_{c,t}-\overline{T_c}\right|$$

(1)

where $CP{R}_{ct}$ represents the climate physical risk of province c in year t, $T_{c,t}$ represents the average annual temperature of the province in year t, and $\overline{T_c}$ represents the average temperature of the province over the sample period (2011–2021). This method can effectively capture the degree of abnormal temperature fluctuations in each region and thus reflect the potential impact of climate physical risk on the insurance industry.

The temperature data in this paper come from the National Center for Environmental Information (NCEI), a division of the National Oceanic and Atmospheric Administration (NOAA). This dataset provides a long-term monitoring record of global temperatures and can ensure the authority and reliability of the data. In addition, NCEI’s temperature data cover all provinces in China, providing robust data support for the climate risk measurements in this study.

2.

Climate transition risks

In existing studies, scholars usually use a variety of methods to measure climate transition risk, mainly including measures of policy uncertainty, changes in market preferences, and technological innovation risk. Earlier studies focus more on policy factors. For example, Gavriilidis et al. and Gambhir measure policy-driven climate transition risk using carbon pricing, economic abatement costs, and textual news data analysis [94,95]. Another group of studies focuses on market preference changes, arguing that shifts in investor and consumer preferences for green products and a low-carbon economy may affect firms’ financing costs and market competitiveness. For example, Guo constructed the Global Climate Concern Index based on the Google search index to measure market sensitivity to climate issues [96]. In addition, some studies consider technological change to be an important factor influencing the climate transition, suggesting that the diffusion of new energy technologies and advances in carbon emission control technologies may affect traditional high-carbon industries’ profitability and business model [97]. Although these studies provide a rich theoretical basis for measuring the risk of climate transition, it may be difficult for a single indicator to reflect the complexity of climate transition fully. Therefore, in recent years, scholars have begun to adopt a multidimensional and comprehensive measurement approach to capture the systemic risks associated with the climate transition more comprehensively.

Based on the above research, this paper constructs the Climate Transition Risk Index (CTRI) from policy uncertainty, market preference change, and technological innovation. First, regarding climate policy uncertainty, policy adjustments and regulation changes may exacerbate enterprises’ operational uncertainty and affect their investment decisions and long-term strategic layout [98]. For this reason, this paper selects the Climate Policy Uncertainty Index (CPU), which is based on the analysis of official policy documents and news texts to measure the degree of uncertainty of the Chinese government in climate governance [99]. Secondly, regarding market preference changes, consumers and investors are paying increasing attention to green products and the low-carbon economy [100,101], which may affect insurers’ risk-pricing and product-innovation strategies. This paper adopts the Baidu Index (BI) to measure the public’s attention to climate-related issues, which can better reflect the market’s sensitivity to climate issues by analyzing users’ search behavior through big data. Finally, regarding technological innovation, the decarbonization transition of energy structure is an important indicator of climate transition risk [102]. In this paper, the clean energy share (CES) is used to measure the use of clean energy in each province, which reflects the progress of regional energy structure adjustment, and a higher CES means that enterprises face more substantial pressure for low-carbon transition.

In order to comprehensively portray the climate transition risk, this paper utilizes the entropy weighting method to construct the climate transition risk index (CTR) at the regional level based on three indicators, namely, policy uncertainty (CCPU), market concern (BI) and clean energy structure (CES). (See

Table 1. Description of sub-indicators of climate transition risk.

Primary Indicator	Secondary Indicator	Indicator Description	Data Sources
CTR	Climate policy change	Climate policy uncertainty index	the Global Climate Risk Integration Database
	Changes in market preferences	Baidu index for climate keywords	Baidu Index
	Changes in technological innovation	Share of clean energy	China Energy Statistics Yearbook

). The entropy weight method is an objective assignment method that can automatically determine each index’s weights according to the size of information entropy, reduce subjective bias, and ensure the scientificity and comparability of the index. The specific steps are as follows:

First, this paper adopts the normalization process to standardize the raw data, eliminating the influence of differences in the scale and magnitude of different indicators on the construction of the composite index. Considering that the three selected secondary indicators are all positive, they are processed using the positive extreme deviation standardization method to ensure that the standardized results are comparable and consistent.

$${p_{ct,j}}^{\prime }={\displaystyle \frac{p_{ct,j}-\min (p_{ct,j})}{\max (p_{ct,j})-\min (p_{ct,j})}}$$

(2)

where $p_{ct,j}$ is the original value of the jth indicator for province c in year t, and ${p_{ct,j}}^{\prime }$ is the normalized value.

The share of the jth indicator in the cth province in year t is then calculated:

$$P_{ct,j}={\displaystyle \frac{{p_{ct,j}}^{\prime }}{{\displaystyle \sum _{i=1}^n{p_{ct,j}}^{\prime }}}}$$

(3)

Calculate the information entropy $e_j$ accordingly:

$$e_j=-{\displaystyle \frac{1}{\ln (n)}}{\displaystyle \sum _{i=1}^n{P}_{ct,j}\ln P_{ct,j}}$$

(4)

The final indicator weights are obtained at $W_j$:

$$W_j={\displaystyle \frac{1-e_j}{{\displaystyle \sum _{j=1}^3(1-e_j)}}}$$

(5)

Based on the above calculations, this paper integrates the three indicators through the weighted average method to construct the Climate Transition Risk Index (CTR) as follows:

$$CT{R}_{ct}=W_1\cdot CCP{U}_{ct}+W_2\cdot B{I}_{ct}+W_3\cdot CE{S}_{ct}$$

(6)

where $CT{R}_{ct}$ denotes the climate transition risk index of province c in year t, and $W_1$, $W_2$, and $W_3$ are the weights of the three dimensions of policy, market preference change, and technological innovation, respectively. This index integrates the three dimensions of climate policy, market preference, and technological innovation, which can comprehensively reflect the level of climate transition pressure at the regional level and is suitable as the core explanatory variable in the subsequent regression model.

4.2.2. Explained Variables

In the research of measuring the operating efficiency of insurance companies, scholars at home and abroad generally adopt the data envelopment analysis (DEA) method or stochastic frontier analysis (SFA) method. The SFA method is based on the form of the set production function, which can separate the stochastic error from the loss of efficiency to a certain extent [103]. However, the method relies on a specific function setting, possibly leading to efficiency measures sensitive to assumptions. In the insurance industry, which is a highly heterogeneous field, the traditional SFA method may be complicated in accurately characterizing the efficiency of insurance companies of different sizes. Therefore, more and more studies have adopted the DEA method to measure the efficiency of insurance companies [104,105]. The DEA method, which does not require a predefined production function, enables efficiency comparisons across heterogeneous samples and is widely applied in the finance and insurance sectors. Therefore, this paper adopts the DEA method to measure insurance companies’ operating efficiency better to reflect their resource allocation and business operation efficiency. The DEA method is a nonparametric method based on linear programming, which is suitable for multi-input–multi-output efficiency evaluation problems. The classical DEA models mainly include the CCR model and the BCC model. Among them, the CCR model assumes that the production unit is under the condition of constant return to scale (CRS), which applies to the efficiency analysis of large-scale enterprises. The BCC model, on the other hand, allows variable returns to scale (VRS) and is more suitable for the efficiency measurement of enterprises of different sizes [106]. Given the significant differences between insurance companies in operation mode and business scale, this paper adopts the DEA-BCC model for measurement to consider the different scale effects of insurance companies. The DEA-BCC model assumes that the decision-making unit (DMU) operates under variable returns to scale (VRS), which can effectively identify the technical and scale efficiencies and, thus, more accurately evaluate the operating efficiency of insurance companies. The DEA model assumes that the sample contains n decision-making units, each of which inputs m production factors and outputs s return indicators. Let $x_{i,j}$ denote the ith input used by the jth insurance company, $y_{r,j}$ denote the r output produced by the jth insurance company, and $v_i$ and $v_r$ denote the weights of the inputs and outputs, respectively. Then, the objective function of the DEA-BCC model is as follows:

$$\begin{array}{cc}\hfill \underset{\theta ,\lambda }{{\max }^{\theta }}& \\ \hfill s.t.& {\displaystyle \sum _{j=1}^n{\lambda }_j{x}_{i,j}}\le \theta x_{i,j},i=1,2,\dots ,m\hfill \\ & {\displaystyle \sum _{j=1}^n{\lambda }_j{y}_{i,j}}\le y_{r,j},r=1,2,\dots ,s\hfill \\ & {\displaystyle \sum _{j=1}^n{\lambda }_j}=1,{\lambda }_j\ge 0,\forall j\hfill \end{array}$$

(7)

where $\theta$ is the efficiency value and ${\lambda }_j$ is the weighting variable. When $\theta =1$, and all constraints are satisfied, the DMU is considered technically efficient; when $\theta <1$, the DMU is technically inefficient.

In the existing studies on the measurement of insurance company efficiency, different scholars selected different input and output indicators according to the study’s purpose and the data availability. Zhao et al. selected operating and administrative expenses, taxes and surcharges, service charges, commission expenses, and claims expenses as inputs and earned premiums and investment income as outputs [107]. In their study of insurance industry efficiency, Tone et al. chose overhead, fixed assets, and invested assets as inputs and earned premiums and investment income as outputs [108]. Biener et al. used headcount, materials, business services, equity capital, and surplus as inputs and total invested assets, smoothed losses, and net premium income as outputs [109]. Kwehet al. used operating expenses, debt capital, and equity capital as inputs, and the outputs are earned premiums after reserve increases and investment income [110]. Cummins et al. selected administrative labor, agent labor, materials and business services, and financial equity capital as inputs, and the outputs are short-term and long-term insurance claims and the value of actual losses on invested assets [14]. Yang chose labor expenses, general operating expenses, capital equity, and incurred claims as inputs and net premium income and net income as outputs [111].

From the existing studies, when scholars measure the efficiency of insurance companies using the DEA method, they usually take human capital and financial capital as the primary input variables. In contrast, premium income and investment income are taken as the primary output variables. These studies provide a solid practical foundation for measuring the efficiency of insurance companies and, at the same time, flexibly adjust the input and output indicators according to different research needs to more accurately reflect the operational efficiency and market performance of insurance companies. In addition, some scholars have proposed that claims expenses should also be included when considering output factors. Insurance companies are not only profit-making enterprises but also bear the responsibility of social security. By diversifying and transferring risks, the insurance business promotes mutual assistance among the insured while easing the government’s financial pressure, reducing social friction, and maintaining social stability. Therefore, considering claims expenditure as an effectual output helps to measure the operational efficiency of insurance companies more comprehensively, focusing on their profitability and social functions. Based on the results of previous research and the research needs of this paper, this paper selects the number of employees, paid-in capital, and operating expenses as input indicators and premium income, investment income, and payout expenses as output indicators when measuring the efficiency of insurance companies. These input and output variables are specified in

Table 2. Selection of input–output variables for DEA modeling.

Variable Type	Variable Name	Variable Description	Data Sources
Input variables	Number of employees	Number of employees of insurance companies (in persons)	China Insurance Yearbook
	Paid-up capital	Paid-up capital of the company (in millions)
	Business expense	Fee and commission expenses + operating and administrative expenses (in millions)
Output variables	Premium income	Income from insurance operations (in millions)
	Investment income	Investment income (in millions)
	Claims expenditure	Claims expenditure (in millions)

.

In this paper, all representative domestic insurance companies recorded in the 2011–2021 China Insurance Yearbook were selected as research samples, and their operational efficiencies were measured using the DEA-BCC model.

Table 3. Descriptive statistics of insurance company efficiency.

Year	Mean Value	Upper Quartile	Standard Deviation	Minimum Value	Maximum Values
2011	0.448	0.372	0.302	0.026	1.000
2012	0.448	0.370	0.282	0.017	1.000
2013	0.396	0.301	0.299	0.019	1.000
2014	0.333	0.160	0.360	0.012	1.000
2015	0.250	0.080	0.322	0.004	1.000
2016	0.372	0.281	0.294	0.009	1.000
2017	0.414	0.300	0.308	0.029	1.000
2018	0.390	0.279	0.310	0.001	1.000
2019	0.398	0.278	0.319	0.016	1.000
2020	0.376	0.258	0.316	0.007	1.000
2021	0.329	0.232	0.294	0.020	1.000
Property insurance	0.287	0.211	0.267	0.001	1.000
Life insurance	0.384	0.279	0.307	0.006	1.000
Total	0.376	0.279	0.314	0.001	1.000

shows the descriptive statistics of the measured efficiency of insurance companies.

According to the statistical results in Table 3, the operating efficiency of Chinese insurance companies during the period 2011–2021 shows significant fluctuation characteristics, reflecting the changes in the internal and external environments faced by the industry at different stages. Overall, the annual average value of insurance company efficiency is 0.376, with a standard deviation of 0.279, showing significant differences in resource allocation and operational efficiency among insurance companies within the sample. With efficiency values ranging from a low of 0.001 to a high of 1.000, the differences in operational and management capabilities between insurance companies are evident.

Observed on an annualized basis, the efficiency of the insurance industry experienced significant fluctuations over the sample period. From 2011 to 2013, the average efficiency declined steadily from 0.448 to 0.396, dropping to the lowest point of the sample period, 0.250, in 2015. this trend may reflect the fact that China’s insurance market was subjected to the dual impacts of policy adjustments and intensified competition in the industry during the early stages of its development. With the industry’s enhanced ability to adapt to market demand and the gradual increase in policy support, the efficiency value has rebounded since 2016 and reached a second-highest point of 0.414 in 2017. However, insurers’ efficiency fell back again after 2020 due to the impact of the new crown epidemic, showing the negative impact of external shocks on the industry’s short-term operating efficiency. Regarding business type, the annual average of total life insurers’ efficiency was 0.384, significantly higher than that of total P&C insurers at 0.287. This difference may stem from life insurers’ long-term capitalization and risk management strengths. In contrast, P&C insurers’ efficiency performance was relatively low due to their shorter business cycles and sensitivity to market volatility. In addition, life insurers typically have more substantial investment and innovation capabilities, further enhancing their efficiency performance.

4.2.3. Mediating Variables

The mediating variables include disaster loss exposure (DLE) and green insurance (GI), which measure the transmission mechanisms through which climate physical and climate transformation risks affect the operational efficiency of insurance companies. Disaster loss exposure (DLE) is measured by the ratio of a company’s claims expenditure to that of the province in which it operates multiplied by the number of people affected by the disaster in that province. DLE captures the insurer’s underwriting risk exposure due to extreme weather events and natural disasters. Higher exposure to catastrophe losses means insurers must absorb more excellent economic shocks, which may exacerbate liquidity pressures, reduce capital allocation efficiency, and affect long-term operational stability. As physical climate risk rises, insurers may be exposed to more frequent catastrophe losses, which could undermine their operational efficiency. Green insurance (GI) is measured using the ratio of environmental pollution liability insurance revenues to total insurance revenues, reflecting the insurance industry’s role in promoting sustainable development and supporting the low-carbon transition. Its development not only inhibits the expansion of high-carbon assets by raising the underwriting threshold of high-pollution and high-carbon industries but also guides the flow of capital to low-carbon and sustainable industries and promotes economic restructuring by increasing the supply of green insurance products [112]. While green insurance strengthens environmental risk management, it also enhances insurance companies’ risk identification and pricing ability, making them more adaptable and stable in the process of low-carbon transition.

4.2.4. Moderating Variables

Moderating variables include the stock of disaster resilient infrastructure (DRI) and insurance technology (IT), which measure the buffering role that climate physical risk and climate transformation risk may play in affecting the operational efficiency of insurance companies, respectively. The stock of disaster resilient infrastructure (DRI) is measured by the per capita stock of materials such as vehicles and farm machinery, pipelines and lights, and buildings. It reflects the level of construction of regional disaster prevention and mitigation capacity. Adequate disaster resilience infrastructure helps reduce direct economic losses from extreme weather events. It reduces the pressure on insurers to settle large-scale claims due to natural disasters, thereby mitigating the impact of physical climate risks on their operational efficiency [113]. Insurtech (IT) is measured by multiplying the ratio of insurance revenue to claims expenditure of the province in which it is located by the insurance service sub-index in the NU Digital Inclusive Finance Index of the previous year of the province [114,115], representing the level of insurers’ use of technology to optimize product innovation, enhance risk pricing ability and improve service efficiency. With the development of insurance technology, insurance companies can more accurately assess and manage the risks brought about by the low-carbon transition process, improve market adaptability, and thus help mitigate the negative impact of initial climate transition risks on operational efficiency [116].

4.2.5. Control Variables

Existing studies show that the operational efficiency of insurance companies is affected by a combination of the macroeconomic environment and the firm’s characteristics [117]. GDP growth rate and inflation level are usually used to measure the macroeconomic conditions that affect the insurance market’s development potential and cost structure [17]. At the micro level, investment returns and firm size are considered to be important determinants of insurers’ profitability and resource allocation efficiency [118,119], while payout ratios, human capital structure, and risk-taking capacity directly affect their operational stability and underwriting capacity [120,121]. This paper incorporates the following control variables in the regression analysis to ensure the robustness of the model estimation:

(1)

GDP growth rate (GDPG): Economic growth drives the expansion of the insurance market and raises the demand for insurers’ business. This paper measures the macroeconomic environment using GDP growth rate data from the China Statistical Yearbook.

(2)

Consumer price index (lnCPI): Inflation affects insurers’ underwriting costs and investment returns. This paper uses the CPI’s logarithmic form to minimize the impact of volatility.

(3)

Return on investment (ROI): Insurance companies rely on investment income to maintain profitability, and higher investment returns enhance corporate risk resistance. This paper calculates the ratio of ROI to total assets as a measurement indicator.

(4)

Firm size (ES): Larger insurers will likely enjoy cost advantages and be more competitive. This paper uses the logarithm of total assets to measure firm size.

(5)

Claims ratio (CR): The claims ratio reflects an insurer’s level of underwriting risk, and higher claims ratios may erode profitability. This paper uses the ratio of claims expense to premium income as a measurement indicator.

(6)

Human structure (HUM): The level of human capital affects insurance companies’ risk management and product innovation abilities. This paper measures the human capital structure by selecting the ratio of employees with bachelor’s degrees or above to the total number of employees.

(7)

Risk-taking capacity (RTC): Underwriting capacity affects insurance companies’ stability in the face of market volatility, and more substantial risk-taking capacity helps improve operational efficiency. This paper adopts the ratio of insurance revenue to total assets as a measure.

The definitions of all variables used in the analysis are summarized in

Table 4. Definition of variables.

Variable Type	Variable Name	Variable Symbol	Variable Interpretation	Source of Variables
Independent Variable	Climate physical risk	CPR	Absolute temperature deviation values are used to measure the intensity of risk due to natural hazards and extreme weather events.	NOAA
	Climate transition risks	CTR	Calculated based on entropy power method, synthesizing policy, social preference, and technological innovation changes.	GCRID, China Energy Statistics Yearbook
Dependent Variable	Operational efficiency of insurance companies	Efficiency	efficiency values measured by the DEA-BCC model.	Measured by DEAP2.1 software
Control Variable	GDP growth rate	GDPG	Representing the level of growth of the national economy	China Statistical Yearbook
	Consumer price index CPI	lnCPI	Consumer price index in logarithmic form, reflecting inflation
	Investment income	ROI	Ratio of investment income to total assets	China Insurance Yearbook
	Enterprise size	ES	Logarithm of total assets
	Compensation rate	CR	Ratio of claims expense to premium income
	Manpower structure	HUM	Ratio of employees with bachelor’s degree or above to total employees
	Risk-bearing capacity	RTC	Ratio of insurance income to total assets
Intermediary Variable	Exposure to natural disaster losses	DLE	Percentage of company claims expenditure multiplied by the number of people affected in the province	China Climate Hazards Yearbook, etc.
	Green insurance	GI	Environmental pollution liability insurance income/total premium income	Eps database, etc.
Moderator Variable	Stock of disaster-resistant infrastructure	DRI	Per capita stock of materials such as vehicles and agricultural machinery, plumbing and lighting, buildings, etc.	(China) National Bureau of Statistics (NBS)
	Insurance technology	IT	The company’s share of premium income multiplied by the insurance services sub-index of the BYU Digital Inclusion Index for that year	Digital Finance Research Center, Peking University

.

4.3. Model

4.3.1. Fixed Effects Model

This paper focuses on climate risk’s impact on insurance companies’ efficiency. It selects panel data from 2010 to 2021 and develops the following econometric model: model (8) analyzes the direct impact of climate physical risk on insurance company efficiency, while model (9) tests the nonlinear effect of its climate transition risk by including a quadratic term for climate transition risk.

$$Efficienc{y}_{ict}={\mathsf{\beta }}_0+{\mathsf{\beta }}_1\cdot CP{R}_{ct}+{\displaystyle \sum _{\mathrm{n}=2}^8{\mathsf{\beta }}_{\mathrm{n}}\cdot X_{ict}}+u_i+{\epsilon }_{ict}$$

(8)

$$Efficienc{y}_{ict}={\mathsf{\beta }}_0+{\mathsf{\beta }}_1\cdot CT{R}_{ct}+{\mathsf{\beta }}_2\cdot CT{R_{ct}}^2+{\displaystyle \sum _{\mathrm{n}=3}^9{\mathsf{\beta }}_{\mathrm{n}}\cdot X_{ict}}+u_i+{\epsilon }_{ict}$$

(9)

In particular, $Efficienc{y}_{ict}$ is the explanatory variable representing the operational efficiency of insurance company i located in province c in year t; $CP{R}_{ct}$ and $CT{R}_{ct}$ are the core explanatory variables representing the climate physical risk and climate transition risk in province c in year t, respectively; $CT{R_{ct}}^2$ represents the quadratic term of the climate transition risk, which is used to capture its potential nonlinear effects. ${\mathsf{\beta }}_0$, ${\mathsf{\beta }}_1$, ${\mathsf{\beta }}_2$, and ${\mathsf{\beta }}_{\mathrm{n}}$ are the coefficients of the explanatory variables and control variables, respectively; $X_{ict}$ represents the control variables, including multiple potential influences at the macro and micro levels; $u_i$ is the individual fixed effect of the insurance company; and ${\epsilon }_{ict}$ is the random error term. It should be noted that the climate effect is a time series variable, so in order to avoid multicollinearity caused by the introduction of time fixed effects, the time fixed effect [50,122] is not controlled in the model.

4.3.2. GMM Models

In this study, a dynamic panel system GMM model was used to replace the fixed effects model for robustness testing, and dynamic models (10) and (11) were constructed by introducing lagged one-period dependent variables into the baseline regression models (8) and (9).

$$Efficienc{y}_{ict}={\mathsf{\beta }}_0+{\mathsf{\beta }}_1\cdot Efficienc{y}_{ic,t-1}+{\mathsf{\beta }}_2\cdot CP{R}_{ct}+{\displaystyle \sum _{\mathrm{n}=3}^9{\mathsf{\beta }}_{\mathrm{n}}\cdot X_{ict}}+u_i+{\epsilon }_{ict}$$

(10)

$$Efficienc{y}_{ict}={\mathsf{\beta }}_0+{\mathsf{\beta }}_1\cdot Efficienc{y}_{ic,t-1}+{\mathsf{\beta }}_2\cdot CT{R}_{ct}+{\mathsf{\beta }}_3\cdot CT{R_{ct}}^2+{\displaystyle \sum _{\mathrm{n}=4}^{10}{\mathsf{\beta }}_{\mathrm{n}}\cdot X_{ict}}+u_i+{\epsilon }_{ict}$$

(11)

where $Efficienc{y}_{ic,t-1}$ denotes a one-period lag in the value of the dependent variable, which is used to capture the dynamically changing characteristics of the insurance company’s efficiency, and the other variables are consistent with the baseline regression model.

4.3.3. Models of Mediating Effects

Due to the limitations of Equations (8) and (9) in exploring only the direct impact of climate risk on the operational efficiency of insurance companies, based on Baron and Kenny’s approach, this paper further constructs mediation effect models (12)–(13) and (14)–(15) to explore how climate physical risk and climate transition risk, respectively, affect the efficiency of insurance companies [123].

$$DL{E}_{ict}={\mathsf{\beta }}_0+{\mathsf{\beta }}_1\cdot CP{R}_{ct}+{\displaystyle \sum _{\mathrm{n}=2}^8{\mathsf{\beta }}_{\mathrm{n}}\cdot X_{ict}}+u_i+{\mathsf{\epsilon }}_{ict}$$

(12)

$$Efficienc{y}_{ict}={\mathsf{\alpha }}_0+{\mathsf{\alpha }}_1\cdot CP{R}_{ct}+{\mathsf{\alpha }}_2\cdot DL{E}_{ct}+{\displaystyle \sum _{\mathrm{n}=3}^9{\mathsf{\alpha }}_{\mathrm{n}}\cdot X_{ict}}+u_i+{\mathsf{\epsilon }}_{ict}$$

(13)

$$G{I}_{ct}={\mathsf{\beta }}_0+{\mathsf{\beta }}_1\cdot CT{R}_{ct}+{\displaystyle \sum _{n=3}^9{\mathsf{\beta }}_n\cdot X_{ict}+u_i+{\mathsf{\epsilon }}_{ict}}$$

(14)

$$Efficienc{y}_{ict}={\mathsf{\alpha }}_0+{\mathsf{\alpha }}_1\cdot CT{R}_{ct}+{\mathsf{\alpha }}_2\cdot G{I}_{ct}+{\mathsf{\alpha }}_3\cdot G{I_{it}}^2+{\displaystyle \sum _{\mathrm{n}=4}^{10}{\mathsf{\alpha }}_{\mathrm{n}}\cdot X_{ict}}+u_i+{\mathsf{\epsilon }}_{ict}$$

(15)

where the mediator variable $DL{E}_{ict}$ is the natural disaster loss exposure of insurance company i located in province c in year t, and $G{I}_{ct}$ is the degree of green insurance in province c in year t.

4.3.4. Moderating Effects Model

This paper constructs the moderating effects models (16) and (17), drawing on the methodology of Lin and Feng [124], to test the moderating effects of the stock of disaster-resistant infrastructure and insurance technology on climate risk, which affects insurance companies’ operational efficiency.

$$Efficienc{y}_{ict}={\mathsf{\beta }}_0+{\mathsf{\beta }}_1\cdot CP{R}_{ct}+{\mathsf{\beta }}_2\cdot DR{I}_{ct}+{\mathsf{\beta }}_3\cdot CP{R}_{ct}\cdot DR{I}_{ct}+{\displaystyle \sum _{\mathrm{n}=4}^{10}{\mathsf{\beta }}_{\mathrm{n}}\cdot X_{ict}}+u_i+{\epsilon }_{ict}$$

(16)

$$\begin{array}{l}Efficienc{y}_{ict}={\mathsf{\beta }}_0+{\mathsf{\beta }}_1\cdot CT{R}_{ct}+{\mathsf{\beta }}_2\cdot CT{R_{ct}}^2+{\mathsf{\beta }}_3\cdot I{T}_{ct}+{\mathsf{\beta }}_4\cdot CT{R}_{ct}\cdot I{T}_{ct}\\ +{\mathsf{\beta }}_5\cdot CT{R_{ct}}^2\cdot I{T}_{ct}+{\displaystyle \sum _{\mathrm{n}=6}^{12}{\mathsf{\beta }}_{\mathrm{n}}\cdot X_{ict}}+u_i+{\epsilon }_{ict}\end{array}$$

(17)

where $DR{I}_{ct}$ denotes the stock of disaster-resistant infrastructure in province c in year t, and $I{T}_{ct}$ denotes the level of insurance technology in province c in year t.

5. Empirical Analysis

5.1. Descriptive Statistics

Table 5. Results of descriptive statistics.

Variable	Sample Size	Mean Value	Standard Deviation	Minimum Value	Maximum Values
Efficiency	1801	0.376	0.314	0.001	1
CPR	1801	0.067	0.110	0	1
CTR	1801	0.462	0.111	0.110	0.907
GDPG	1801	0.091	0.038	−0.053	0.282
lnCPI	1801	4.628	0.010	4.606	4.664
ROI	1800	0.042	0.174	0	5.433
ES	1801	9.445	2.098	5.422	16.130
CR	1796	0.342	0.314	0.002	1.988
HUM	1775	0.548	0.298	0.018	2.192
RTC	1801	0.361	0.250	0.006	1.196

presents the descriptive statistics for the main variables. The mean operating efficiency among insurers is 0.376, suggesting a generally low efficiency level overall. However, a few firms reach full efficiency. The standard deviation of 0.314 points to notable differences across companies.

Regarding the central explanatory variables, the average climate physical risk exposure is 0.067, indicating that most insurers face moderate climate-related threats. At the same time, the maximum value of 1 implies that some companies confront considerable risks from extreme weather events. Meanwhile, the mean climate transition risk is 0.462, with a standard deviation of 0.111, underscoring variations in insurers’ capabilities and performance when dealing with climate transition challenges.

Among the macroeconomic control variables, the mean economic growth rate is 0.091, and the mean consumer price index is 4.628. Both display slight deviations, suggesting relatively stable economic and inflationary conditions in the sampled regions. Regarding firm-specific controls, the average ROI is 0.042, with a standard deviation of 0.174, indicating substantial discrepancies in investment outcomes among insurers. Firm size averages 9.445, with a standard deviation of 2.098, reflecting significant unevenness in size distribution. The mean values of capital structure, human resource structure, and composite claim rate are 0.342, 0.548, and 0.361, respectively, each with standard deviations of 0.314, 0.298, and 0.250. These variations collectively reveal pronounced differences in insurers’ capital operations, workforce allocation, and claims management efficiency.

5.2. Baseline Regression and Analysis

Columns (1) and (3) of

Table 6. Baseline regression results.

	(1)	(2)	(3)	(4)
	Efficiency	Efficiency	Efficiency	Efficiency
CPR	−0.4210 ***	−0.3419 ***
	(−11.3240)	(−9.7571)
CTR			−0.5660 **	−0.8802 ***
			(−2.0957)	(−3.3791)
CTR_sq			0.8307 ***	0.7153 ***
			(2.8470)	(2.6270)
GDPG		0.1630		0.3715 ***
		(1.5134)		(3.3811)
lnCPI		3.5862 ***		3.0297 ***
		(8.1414)		(6.2623)
ROI		0.1604 ***		0.1624 ***
		(7.1897)		(7.0910)
ES		0.0733 ***		0.0978 ***
		(11.5439)		(13.0922)
CR		0.0656 ***		0.0813 ***
		(3.8523)		(4.5693)
HUM		0.0365		0.0512 *
		(1.3858)		(1.8940)
RTC		0.2828 ***		0.3143 ***
		(10.3728)		(11.1323)
_cons	0.4037 ***	−17.0603 ***	0.4495 ***	−14.5366 ***
	(86.1601)	(−8.2578)	(7.2508)	(−6.4081)
N	1801	1769	1801	1769
R2	0.0745	0.2238	0.0119	0.1849

t statistics in parentheses. * p < 0.1, ** p < 0.05, *** p < 0.01.

show the regression results of climate physical risk and climate transition risk on the operating efficiency of insurance companies without the inclusion of control variables, while columns (2) and (4) show the regression results of climate physical risk and climate transition risk on the operating efficiency of insurance companies with the inclusion of control variables, respectively.

From the regression coefficients in columns (2) and (4), it can be seen that the coefficient of climate physical risk on the efficiency of insurance companies is −0.3419, which is statistically significant at the 1% level, which indicates that climate physical risk significantly reduces the operational efficiency of insurance companies. The coefficient of climate transition risk on the efficiency of insurance companies is −0.8802, and the coefficient of squared term is 0.7153, which are both significant at a 1% level, indicating that the impact of climate transition risk on the efficiency of insurance companies presents a U-shape characteristic; i.e., the efficiency is negatively affected when the transition risk is lower, but when the transition risk is further increased, this negative impact is gradually weakened and turned into a positive impact. This U-shape relationship passes the utest test.

The possible reasons for these results are as follows: On the one hand, the exacerbation of physical weather risks directly increases the pressure on insurers to pay claims and increases their financial instability. Frequent extreme weather events lead to higher property and personal claims for insurers, weakening liquidity and affecting capital adequacy. In addition, the uncertainty of weather disasters makes it difficult for insurers to price their products accurately and increases the cost of risk management, further reducing operational efficiency. In extreme cases, persistently high payouts may force some insurers to adjust their scope of coverage or even withdraw from the high-risk market, thus affecting overall operational stability. On the other hand, climate transition risk affects insurers’ efficiency mainly through asset structure adjustment and business adaptability. When the transition risk is low, insurers still hold many high-carbon assets and fail to optimize their investment portfolios promptly, leading to asset depreciation and lower investment returns after tightening the policy, thus weakening operational efficiency. In addition, insufficient transition pressure may cause insurers to underinvest in green insurance products and business innovation, affecting their ability to adapt to future markets. However, as transition risks intensify and market demand for low-carbon financial products increases, insurers are accelerating the layout of their green investments, reducing their reliance on high-carbon assets, and promoting business model innovation, thereby enhancing their competitiveness and operational efficiency. Compared with Sewan and Chen’s study, the benchmark regression in this paper not only verifies the negative effect of climate physical risk but also reveals that climate transition risk’s impact on insurance companies’ operating efficiency shows a U-shaped relationship [51]. In conclusion, these benchmark regression results provide strong empirical support for Hypotheses 1a,b.

5.3. Endogeneity Test

In this study, the dependent variable climate risk may be endogenous. One possible explanation is that measurement errors in climate risk, which are incorporated into the residual terms, could result in its endogeneity. Another factor is the omission of relevant variables during the modeling process, which may contribute to this endogeneity. To address these potential issues, we employ an instrumental variables approach. For climate physical risk, this paper refers to Nunn and Qian’s approach of selecting a time-varying variable versus a historically invariant variable to characterize the time-varying nature of the instrumental variable. It draws on Zhang et al.’s study of constructing a panel of data as an instrumental variable for climate physical risk by using a cross-section of the topographic relief data multiplied by the absolute temperature deviation across the country in the current year (Terrain) [113,125]. The panel data were constructed as an instrumental variable for climate physical risk (Terrain) by multiplying the topographic relief cross-section data with the national absolute temperature deviation of the year. As the topographic relief reflects the topographic differences within the province, its magnitude indirectly affects the degree of infrastructure development, emergency response capacity, and the concentration of economic activities. The greater the degree of undulation, the more decentralized the resource allocation and disaster prevention and control within the region; thus, the region is more vulnerable to the impacts of climate physical risks in the event of an extreme weather event. The absolute temperature deviation of the country as a whole is correlated with the absolute temperature deviation of the province but not with the operational efficiency of the insurance company situated in that province. Therefore, the selection of Terrain is consistent with the instrumental variable construction rule. The larger the value of the cross-multiplier between the two, the greater the local climate physical exposure and the more likely it is to be affected by climate physical risk. For climate transition risk, this paper draws on Yang and Li’s approach to select the carbon dioxide emissions of Chinese provinces calculated in CEADs China Carbon Accounting Database as the instrumental variable [46]. CO₂ emissions reflect regional industrial structure, energy use characteristics, and transition pressure, which are closely related to climate transition risk but do not directly affect the efficiency of insurance companies and thus meet the requirements of instrumental variables. In order to avoid the problem of “forbidden regression,” this study adopts a hierarchical instrumental variable method with primary and secondary terms in modeling; i.e., the primary term of CEADs is used as the instrumental variable for the primary term of climate transition risk. The secondary term of CEADs is used as the instrumental variable for the secondary term of climate transition risk.

Table 7. Regression results of instrumental variable approach.

	Phase I			Phase II
	(1)	(2)	(3)	(4)	(5)
	CPR	CTR	CTR_sq	Efficiency	Efficiency
Terrain	0.0771 ***
	(6.3517)
CEADs		0.2027 ***	0.3281 ***
		(3.6764)	(6.2615)
CEADs_sq		0.0496 ***	0.0390 ***
		(4.8778)	(4.0414)
CPR				−0.7499 ***
				(−3.2722)
CTR					−3.6401 **
					(−2.3519)
CTR_sq					2.8565 **
					(2.1324)
GDPG	−0.4655 ***	0.1685 ***	0.1603 ***	−0.0482	0.4908 ***
	(−6.1009)	(4.4102)	(4.4148)	(−0.2972)	(3.3095)
lnCPI	0.4760	−2.2549 ***	−1.8826 ***	3.4727 ***	0.7916
	(1.4153)	(−14.3534)	(−12.6088)	(7.5101)	(0.5580)
ROI	0.0066	0.0150 *	0.0120	0.1638 ***	0.1785 ***
	(0.4143)	(1.8927)	(1.5900)	(7.0350)	(6.9137)
ES	−0.0395 ***	0.0522 ***	0.0469 ***	0.0590 ***	0.1417 ***
	(−8.7736)	(23.3869)	(22.1181)	(5.7067)	(4.8808)
CR	−0.0269 **	0.0442 ***	0.0432 ***	0.0563 ***	0.1113 ***
	(−2.2111)	(7.2640)	(7.4576)	(3.0543)	(3.7916)
HUM	−0.0256	0.0121	0.0070	0.0254	0.0689 **
	(1.3612)	(1.2849)	(0.7860)	(0.9044)	(2.2875)
RTC	−0.0572 ***	0.0573 ***	0.0509 ***	0.2625 ***	0.3652 ***
	(−2.9358)	(5.9030)	(5.5189)	(8.5998)	(8.3513)
F-statistic	40.34	22.29	32.93
Anderson canon. corr. LM stat.				39.526 [0.0000]	29.056 [0.0000]
Cragg-Donald Wald F-statistic				40.344 [16.38]	14.718 [7.03]
N				1767	1767

t statistics in parentheses. * p < 0.1, ** p < 0.05, *** p < 0.01. Note: values in [] correspond to the p-value of the LM test, and values in [] correspond to the Stock–Yogo 10% critical value.

reports the instrumental variable estimation results. In the instrumental variable test for climate physics risk, column (1) indicates the first-stage estimation results, and the F-value of the one-stage regression is 40.34, indicating that the instrumental variable Terrain is exceptionally strongly correlated with climate physics. Column (4) represents the second-stage estimation results. In the above results, the Anderson canon.corr. LM statistic is significant at the 1% level, indicating that the selected instrumental variable is identifiable; the Cragg–Donald Wald F statistic is greater than the Stock–Yogo threshold of 10%, indicating that the weak instrumental variable test is passed. This result shows that climate physical risk still hurts insurers’ operational efficiency, consistent with the previous benchmark regression results. For the instrumental variable test for climate transition risk, columns (2) and (3) indicate the first-stage estimation results and the first-stage regression F-values of the correlation test are 22.29 (primary term) and 32.93 (secondary term), respectively, with Prob > F of 0.0000 for both, indicating that the primary and secondary terms of the instrumental variable CEADs are significantly correlated with climate transition risk. Column (5) represents the second-stage estimation results. In the above results, the Anderson canon. corr. LM statistic is significant at the 1% level, indicating that the selected instrumental variables are identifiable; the Cragg–Donald Wald F statistic is greater than the 10% Stock–Yogo threshold, indicating that it passes the weak instrumental variables test. Thus, after mitigating endogeneity, the results reported for the instrumental variables are generally consistent with those of the benchmark regression.

5.4. Robustness Tests

5.4.1. Adjustment of Sample Period

To assess whether the regression findings are influenced by anomalous years, this study excludes data from 2020. That year, the southern region of China was severely affected by COVID-19 and endured its worst flooding since 1998, potentially disrupting insurers’ operational efficiency. After the sample period was adjusted, the revised regression outcomes, shown in columns (1) and (2) of

Table 8. Robustness test results.

	(1)	(2)	(3)	(4)
	Efficiency	Efficiency	Efficiency1	Efficiency1
CPR	−0.3546 ***		−0.0902 **
	(−9.6080)		(−2.5105)
CTR		−0.8384 ***		−0.7702 ***
		(−3.0567)		(−2.9590)
CTR_sq		0.6399 **		0.6342 **
		(2.1999)		(2.3311)
GDPG	0.0797	0.5045 ***	0.3491 ***	0.4218 ***
	(0.5530)	(3.5425)	(3.1601)	(3.8419)
lnCPI	3.7025 ***	2.8966 ***	3.0593 ***	2.5303 ***
	(7.8790)	(5.6024)	(6.7704)	(5.2345)
ROI	0.1530 ***	0.1554 ***	0.1996 ***	0.2029 ***
	(6.7241)	(6.6407)	(8.7195)	(8.8683)
ES	0.0727 ***	0.0966 ***	0.1001 ***	0.1137 ***
	(10.7757)	(11.8825)	(15.3610)	(15.2420)
CR	0.0768 ***	0.0938 ***	0.0366 **	0.0452 **
	(4.1938)	(4.8794)	(2.0961)	(2.5438)
HUM	0.0369	0.0527 *	0.0472 *	0.0543 **
	(1.3071)	(1.8162)	(1.7456)	(2.0094)
RTC	0.2953 ***	0.3320 ***	0.2950 ***	0.3117 ***
	(10.3820)	(11.2298)	(10.5472)	(11.0480)
_cons	−17.5889 ***	−13.9359 ***	−14.9592 ***	−12.4525 ***
	(−7.9796)	(−5.7516)	(−7.0585)	(−5.4939)
N	1592	1592	1769	1769
R2	0.2343	0.1924	0.2155	0.2185

t statistics in parentheses. * p < 0.1, ** p < 0.05, *** p < 0.01.

, indicate no substantial changes in the direction or significance of climate physical risk and climate transition risk. Specifically, the coefficient for climate physical risk is −0.3546 and remains significant at 1%. Meanwhile, the primary term for climate transition risk has a coefficient of −0.8384, and its secondary term is 0.6399, which is at least significant at the 5% level. These results suggest that the influence of climate physical and transition risks on insurers’ efficiency persists even after omitting 2020 data.

5.4.2. Replacement of Explanatory Variables

The SBM-DEA method is also employed to recalculate insurers’ operating efficiency for the robustness test and to reinforce the reliability of this study’s findings. Unlike the traditional DEA approach, the SBM-DEA method incorporates both desired and undesired outputs and effectively addresses potential slack variables in the efficiency assessment, enhancing accuracy and comparability.

Specifically, the number of employees, paid-in capital, and operating expenses serve as input indicators. At the same time, premium and investment income are designated as desired outputs, and claims expenses are undesired outputs. After replacing the explanatory variables, the results in columns (3) and (4) of Table 8 indicate that the coefficient for climate physical risk is −0.0902, which is significant at the 5% level. Meanwhile, the primary and quadratic terms for climate transition risk are −0.7702 and 0.6342, respectively, both also significant at the 5% level. These findings demonstrate that the effect of climate risk on insurers’ efficiency remains consistent whether using the conventional DEA or the SBM-DEA approach, thus further substantiating the robustness of the study’s conclusions. By incorporating the SBM-DEA method, the robustness test strengthens the persuasiveness of the empirical results.

5.4.3. Alternative Empirical Model

Table 9. System GMM regression results.

	(1)	(2)
	Efficiency	Efficiency
L. Efficiency	0.2945 ***	0.5251 ***
	(3.9202)	(2.6615)
CPR	−0.3890 ***
	(−6.3711)
CTR		−1.4488 **
		(−2.1259)
CTR_sq		1.7052 **
		(2.4663)
GDPG	0.2175 *	−0.5018
	(1.8181)	(−1.6255)
lnCPI	4.5825 ***	5.4997 **
	(6.4942)	(2.4602)
ROI	0.1574 ***	0.1639 ***
	(4.3808)	(4.9131)
ES	0.0603 ***	0.0475 **
	(7.0419)	(2.4942)
CR	0.0998 **	0.0639
	(2.2306)	(1.2186)
HUM	0.1281 ***	0.1040 **
	(3.0867)	(2.5327)
RTC	0.0832 *	0.0830 **
	(1.7916)	(2.2574)
_cons	−11.0926 ***	−25.5065 **
	(−5.2270)	(−2.4494)
Observations	1555	1555
AR (1)	0.000	0.020
AR (2)	0.374	0.677
Hansen test	0.221	0.548

t statistics in parentheses. * p < 0.1, ** p < 0.05, *** p < 0.01.

presents the system GMM regression results. The coefficient for climate physical risk is −0.3890, significant at the 1% level. The primary and secondary terms for climate transition risk are −1.4488 and 1.7052, respectively, each significant at the 5% level, indicating a pronounced U-shaped effect of transition risk on insurers’ efficiency. These findings are consistent with the baseline regression. Meanwhile, the lagged dependent variable exhibits coefficients of 0.2945 and 0.5251, both positive and significant at the 1% level, suggesting that past efficiency positively influences current efficiency.

AR(1) and AR(2) tests were performed to confirm model validity to detect autocorrelation. The AR(1) test yields p-values below 0.05, confirming first-order autocorrelation, whereas AR(2) shows p-values of 0.374 and 0.677, both exceeding 0.1 and thus indicating no second-order autocorrelation which is in line with the setup of the system GMM model. In addition, the Hansen test produces p-values of 0.221 and 0.548, both above 0.1, confirming that the chosen instruments are valid and ruling out over-identification concerns. Overall, these results affirm the reliability of the study’s findings.

5.5. Heterogeneity Test

5.5.1. Heterogeneity of Company Type

This study further divides the sample into property, casualty, and life insurance companies according to their type to explore the heterogeneous impact of climate risk on the operational efficiency of insurance companies under different business models.

As shown in

Table 10. Heterogeneity test of company type.

	Property Insurance			Life Insurance
	(1)	(2)	(3)	(4)	(5)
	Efficiency	Efficiency	Efficiency	Efficiency	Efficiency
CPR	−0.2910 ***			−0.4031 ***
	(−5.4201)			(−7.5139)
CTR		−0.6027	−0.2337 **		−0.8044 *
		(−1.5440)	(−2.0565)		(−1.8728)
CTR_sq		0.3841			0.8248 *
		(0.9880)			(1.8079)
GDPG	0.1350	0.3311 **	0.3369 **	0.3098 *	0.4624 ***
	(0.8651)	(2.0747)	(2.1127)	(1.9027)	(2.7380)
lnCPI	2.8613 ***	2.5625 ***	2.6559 ***	3.2337 ***	3.1091 ***
	(4.0260)	(3.3449)	(3.4935)	(5.0530)	(4.3672)
ROI	0.1393 ***	0.1362 ***	0.1342 ***	0.2700 ***	0.2700 ***
	(4.7401)	(4.5259)	(4.4715)	(7.1457)	(6.8636)
ES	0.0221	0.0603 ***	0.0596 ***	0.0774 ***	0.0936 ***
	(1.4938)	(3.4302)	(3.3935)	(9.4951)	(9.6451)
CR	0.0739 **	0.0744 **	0.0772 **	0.0777 ***	0.0880 ***
	(2.2840)	(2.2213)	(2.3135)	(3.0860)	(3.3164)
HUM	0.1183 **	0.1296 ***	0.1245 **	−0.0027	−0.0001
	(2.4349)	(2.5959)	(2.5072)	(−0.0842)	(−0.0032)
RTC	0.3516 ***	0.3714 ***	0.3693 ***	0.2514 ***	0.2736 ***
	(9.3542)	(9.5638)	(9.5241)	(4.8033)	(4.9899)
_cons	−13.4207 ***	−12.2229 ***	−12.7299 ***	−15.4631 ***	−14.9137 ***
	(−4.0119)	(−3.3977)	(−3.5752)	(−5.1509)	(−4.4628)
N	639	639	639	741	741
R2	0.2327	0.1997	0.1983	0.2777	0.2192

t statistics in parentheses. * p < 0.1, ** p < 0.05, *** p < 0.01.

, the regression results show that CPR significantly negatively impacts property and casualty insurers and life insurers. However, the shock is more significant for life insurers. This difference is closely related to the business characteristics of the two types of insurance companies. Property and casualty insurers mainly underwrite short-term property losses. They can cushion the impact of extreme weather events by adjusting coverage, increasing premium pricing, and reinsurance mechanisms. In contrast, life insurers mainly underwrite long-term policies and must maintain asset-liability matching. Their operating efficiency is more affected by increased mortality rates, fluctuations in asset returns, and increased pressure on long-term claims due to extreme weather events. In addition, life insurers hold many long-term investment assets, subject to more significant asset value volatility in the event of heightened physical weather risks, further exacerbating operational instability.

For climate transformation risk (CTR), the response differs significantly between property–casualty and life insurers. Among P&C insurers, the primary term of CTR is significantly negative, while the secondary term is insignificant, indicating that rising transition risk directly undermines operational efficiency without a U-shaped pattern. This may be because P&C insurers mainly underwrite corporate and industrial risks, which are more exposed to carbon policy tightening and industrial restructuring, leading to heightened claims uncertainty and reduced efficiency. In contrast, life insurers exhibit a pronounced U-shaped relationship: in the early stages, the depreciation of high-carbon assets reduces investment returns, but as the transition deepens, they adjust portfolios, increase green asset allocations, and develop low-carbon-related products, thereby improving efficiency. Benefiting from long-term fund management, life insurers demonstrate more substantial adaptive capacity and resilience under high transition risk. This pattern aligns with Liu et al. [45], who found a U-shaped relationship between climate risk and life insurance demand and an inverse S-shape for non-life insurance in OECD countries. Although their focus is on demand and ours on operational efficiency, the underlying mechanism is similar: life insurers are better equipped to adapt due to their long-term investment horizons. This international evidence supports our findings’ plausibility and potential generalizability across varying institutional and regulatory contexts.

5.5.2. Heterogeneity of Location

This study divides 31 provinces into coastal and inland regions, based on the geographic location of the provinces where the insurance companies are located, to investigate whether there is regional heterogeneity in the impact of climate physical risk and climate transition risk on the operational efficiency of insurance companies.

As shown in

Table 11. Heterogeneity test of location.

	Coastland		Interior
	(1)	(2)	(3)	(4)	(5)
	Efficiency	Efficiency	Efficiency	Efficiency	Efficiency
CPR	−0.3261 ***		−0.3664 ***
	(−7.3011)		(−6.5743)
CTR		−1.1846 ***		−0.5176	−0.2584 **
		(−3.3606)		(−1.2470)	(−2.2771)
CTR_sq		1.1197 ***		0.2728
		(2.9606)		(0.6492)
GDPG	0.1517	0.3271 **	0.1877	0.3946 **	0.3969 **
	(0.9903)	(2.0748)	(1.2187)	(2.5160)	(2.5329)
lnCPI	2.5036 ***	1.8089 ***	5.2679 ***	4.8603 ***	4.9305 ***
	(4.3655)	(2.8722)	(7.5695)	(6.3113)	(6.4693)
ROI	0.1472 ***	0.1516 ***	1.0543 ***	0.9747 ***	0.9665 ***
	(6.7849)	(6.8057)	(3.2701)	(2.9380)	(2.9168)
ES	0.0584 ***	0.0821***	0.0987 ***	0.1242 ***	0.1252 ***
	(7.6940)	(8.7429)	(8.8472)	(9.9538)	(10.1220)
CR	0.0850 ***	0.0900 ***	0.0491 *	0.0779 ***	0.0792 ***
	(3.6762)	(3.7444)	(1.9369)	(2.9225)	(2.9782)
HUM	0.0223	0.0379	0.0559	0.0624	0.0602
	(0.6412)	(1.0594)	(1.3879)	(1.5010)	(1.4533)
RTC	0.2587 ***	0.2858 ***	0.3084 ***	0.3441 ***	0.3438 ***
	(7.9644)	(8.4885)	(6.3766)	(6.8728)	(6.8711)
_cons	−11.9172 ***	−8.6813 ***	−25.1241 ***	−23.3775 ***	−23.7703 ***
	(−4.4391)	(−2.9451)	(−7.6691)	(−6.4395)	(−6.6436)
N	1027	1027	742	742	742
R2	0.2199	0.1848	0.2544	0.2115	0.2110

t statistics in parentheses. * p < 0.1, ** p < 0.05, *** p < 0.01.

, the regression results indicate that CPR significantly negatively impacts insurance company operational efficiency in both regions. However, the extent of its impact is more pronounced in the inland region. Common sense suggests that coastal areas may face more severe natural disaster losses than inland due to frequent exposure to extreme weather events such as typhoons and heavy rains. However, this result may be due to the relatively weak infrastructure and lower disaster resilience in inland areas. Compared to coastal areas, inland areas have a weaker disaster response and repair capacity after extreme weather events, making the pressure on insurers to pay out more prominent. In addition, the relatively immature insurance market in inland areas, with smaller assets and weaker risk diversification capacity of insurance companies, makes them more prone to a decline in operational efficiency when the physical risks of climate rise.

For CTR, the coastal region shows a significant U-shaped relationship. In contrast, the inland region is significant only in the primary term, and the direction is negative, indicating differences in the mechanism of climate transition risk on the efficiency of insurance companies in different regions. Coastal regions have higher levels of economic development and more mature green finance and green insurance markets, and insurance companies can adapt to the transition environment by adjusting their asset structure and expanding their green insurance business under higher climate transition risks, thus enhancing their operational efficiency. However, in the inland region, limited by the development level of industrial structure and financial market, insurance companies are less capable of adapting to the transition risk, so the rise of transition risk is mainly reflected in the negative impact on their operational efficiency. In addition, inland regions, where the proportion of traditional high-carbon industries is higher, the transition process may face a more serious risk of asset depreciation, further dragging down insurance companies’ operating efficiency.

5.6. Mechanism Analysis

In the benchmark regression, it was demonstrated that climate physical risk negatively affects the operational efficiency of insurance companies. Through what intrinsic mechanisms do climate physical risk affect insurers’ operational efficiency? In this paper, we argue that increased physical climate risk exposes insurers to higher natural disaster loss exposures, forcing insurers to increase their costs to meet potential catastrophe payout liabilities and affecting their operational efficiency. According to the results in columns (1) and (2) of

Table 12. Mechanism analysis results.

	(1)	(2)	(3)	(4)
	DLE	Efficiency	GI	Efficiency
CPR	1.2271 ***	−0.3349 ***
	(3.1572)	(−9.5368)
DLE		−0.0062 ***
		(−2.7062)
CTR			0.0509 ***	−0.1059
			(21.8885)	(−1.2262)
GI				−36.0807 ***
				(−3.4378)
GI_sq				215.0427 ***
				(3.3961)
GDPG	13.1362 ***	0.2500 **	−0.0193 ***	0.2729 **
	(11.0081)	(2.2406)	(−5.4530)	(2.3821)
lnCPI	−4.6970	3.5682 ***	−0.1061 ***	2.7385 ***
	(−0.9628)	(8.1186)	(−6.8389)	(5.4806)
ROI	−2.0238 ***	0.1478 ***	0.0002	0.1623 ***
	(−8.1899)	(6.5023)	(0.2828)	(7.0982)
ES	−1.3937 ***	0.0649 ***	0.0029 ***	0.0980 ***
	(−19.8080)	(9.1529)	(11.9837)	(12.5713)
CR	−3.1727 ***	0.0483 ***	0.0016 ***	0.0841 ***
	(−16.7854)	(2.6100)	(2.8167)	(4.7230)
HUM	0.4112	0.0391	0.0012	0.0524 *
	(1.4081)	(1.4870)	(1.4073)	(1.9385)
RTC	−1.8369 ***	0.2724 ***	0.0008	0.3128 ***
	(−6.0796)	(9.8931)	(0.9248)	(11.0915)
_cons	29.3160	−16.9339 ***	0.5209 ***	−11.8834 ***
	(1.2812)	(−8.2128)	(7.1781)	(−4.8215)
N	1767	1767	1769	1769
R2	0.4403	0.2283	0.6119	0.1876

t statistics in parentheses. * p < 0.1, ** p < 0.05, *** p < 0.01.

, every one-unit increase in climate physical risk reduces the operating efficiency of insurance companies by 0.3349 units. Meanwhile, a 1.2271 unit increase in the insurer’s exposure to natural catastrophe losses due to climate physical risk can indirectly lead to a 0.0076 unit decrease in operating efficiency. This result validates Hypothesis 2a.

Concerning climate transition risk, green insurance acts as a mediating variable that significantly transmits the effect of climate transition risk on insurer efficiency. The results in columns (3) and (4) of Table 12 show that climate transition risk (CTR) significantly and positively affects green insurance (GI), suggesting that as the risk of climate transition increases, insurers introduce more green insurance products in order to adapt to the policy and market needs to cope with the risks arising from the transition. However, the significance of the direct effect of climate transition risk on insurers’ efficiency decreases after considering green insurance as a mediating variable, suggesting that the effect of climate transition risk on efficiency is mainly transmitted through green insurance. Further analysis shows that green insurance mediates the impact of climate transition risk on insurer efficiency in a U-shaped manner. Specifically, at the beginning of the transition, the introduction of green insurance instead leads to a decrease in efficiency because insurers need to invest more resources in adapting and innovating green insurance products, which increases initial costs and decreases efficiency. However, as the transition risk intensified, green insurance helped insurers bring in more premium and investment income, improving insurers’ operational efficiency. This U-shaped effect reflects the initial negative effect of green insurance and the gradual recovery and positive effect with the intensification of transition risk. As a result, Hypothesis 2b is verified.

5.7. Moderating Effects

Table 13. Moderating effects.

	(1)	(2)
	Efficiency	Efficiency
CPR	−1.7507 **
	(−2.5520)
DRI	−0.0825 **
	(−2.4299)
CPR× DRI	0.1039 **
	(2.0313)
CTR		−1.4141 ***
		(−4.7764)
CTR_sq		1.1779 ***
		(3.7917)
IT		−0.1932 **
		(−2.0968)
CTR× IT		1.0496 ***
		(2.6559)
CTR_sq× IT		−0.9370 **
		(−2.0840)
control variable	YES	YES
_cons	−15.6429 ***	−15.1317 ***
	(−7.1719)	(−6.6357)
N	1769	1769
R2	0.2279	0.2056

t statistics in parentheses. ** p < 0.05, *** p < 0.01.

demonstrates the results of the moderating effects, further revealing the moderating role of the stock of disaster-resistant infrastructure and insurance technology in the impact of climate risk on the operational efficiency of insurance companies.

First, the stock of disaster-resistant infrastructure significantly mitigates the negative impact of physical climate risk on the operational efficiency of insurance companies. With the improvement of disaster-resistant infrastructure, insurance companies can cope with physical climate risks more effectively, reduce the pressure to pay claims caused by natural disasters, and maintain liquidity and capital adequacy, thus improving operational efficiency. In addition, an adequate stock of disaster-resistant infrastructure improves post-disaster recovery capacity, enabling insurers to respond quickly to emergencies and ensure timely and effective claims settlement. The government plays an active role in supporting the construction of disaster-resistant infrastructure, helping to strengthen the disaster response system through policy guidance and financial investment, which provides a more solid operating environment for insurance companies. Based on the above analysis, Hypothesis 3a is empirically supported.

From the results in column (2) of Table 10, it can be found that the coefficient of the interaction term between the CTR quadratic term and insurance technology (IT) (CTR_sq× IT) is significantly negative and heteroscedastic with CTR_sq at the 1% level of significance, i.e., insurance technology negatively moderates the U-shaped impact of climate transition risk on insurers’ efficiency so that the U-shape curve becomes flatter, and at the same time, ${\mathsf{\beta }}_1{\mathsf{\beta }}_5-{\mathsf{\beta }}_2{\mathsf{\beta }}_4>0$, suggests that a rightward shift of the inflection point delays the emergence of the inflection point (as shown in Figure 2). This implies that insurance technology helps insurers take advantage of scientific and technological innovation to adapt more effectively in the face of low climate transition risk at the early stage of the transition, optimize risk assessment and resource allocation, and thus mitigate the negative impact of climate transition risk on operational efficiency. With high insurance technology support, insurers are more inclined to go for an incremental transition process, i.e., to complete it as far as possible without significantly affecting the efficiency of insurers, thus resulting in a delayed emergence of the inflection point. Therefore, the above results support Hypothesis 3b.

6. Research Findings and Policy Recommendations

6.1. Conclusions of the Study

Operational efficiency in the insurance sector is widely acknowledged as a key determinant of long-term stability and the industry’s sustainable development. This efficiency is closely tied to profitability, risk management capabilities, and overall market competitiveness. Against the backdrop of escalating global climate change and the pursuit of sustainable development goals, climate risk has become a critical factor shaping insurers’ operational performance. Although numerous studies have provided valuable insights into how climate risk influences the insurance industry, there remains a need for deeper exploration of how climate risk affects operational efficiency and the underlying mechanisms involved. In response, this paper devises a climate risk indicator using panel data from 248 Chinese insurance firms spanning 2011 to 2021, systematically examining the various channels through which climate risk may influence operational outcomes. In particular, the study uses absolute temperature deviation values to measure physical climate risk and applies the entropy weighting method to develop a comprehensive indicator reflecting climate transition risk. It then employs the DEA model to assess insurers’ operating efficiency. The study builds on these metrics by employing a fixed effects model to evaluate the direct influence of climate risk on operational efficiency. Furthermore, it investigates the processes through which climate risk exerts its impact and the moderating roles of disaster-resilient infrastructure and innovations in insurance science and technology. The primary conclusions drawn from this investigation are as follows:

First, physical climate risks have a significant negative impact on the operational efficiency of insurance companies. The frequency of extreme weather events increases the claims pressure faced by insurers and leads to decreased liquidity and profitability, thus weakening their operational efficiency. Meanwhile, the impact of climate transition risk shows a U-shaped relationship. At the beginning of the transition, the increase in climate transition risk led to the depreciation of high-carbon assets held by insurers, resulting in a decline in operational efficiency; however, with the rise in market demand for low-carbon products, insurers have gradually adjusted their business models to actively adapt to the transition environment, which has, in turn, improved their operational efficiency.

Further analysis reveals that the impact of climate physical risk on the operational efficiency of insurance companies is mainly transmitted through increased exposure to natural disaster losses. With the intensification of climate physical risk, the pressure on insurers to face catastrophe losses gradually increases, leading to a decline in operational efficiency. On the other hand, the impact of climate transition risk is mainly realized through the intermediary role of green insurance. As climate transition risk increases, insurers have increased their investment in green insurance products to adapt to policy and market changes. While such investment may initially lead to higher costs and affect operational efficiency, as the green insurance market expands, insurers receive more benefits from it, ultimately contributing to improvements in operational efficiency.

In addition, the construction of disaster-resistant infrastructure and the application of insurance science and technology help to mitigate the adverse impact of climate risk to a certain extent. By strengthening the construction of disaster resilience infrastructure, losses caused by extreme weather events can be effectively reduced, thus mitigating the impact of physical climate risks on operational efficiency. Insurance technology, on the other hand, especially in the early stages of the low-carbon transition, can help insurance companies cope with the risks of the climate transition more efficiently through innovative technologies and push them to realize a gradual transition, thus enhancing operational efficiency in the later stages of the transition.

Finally, this study reveals considerable heterogeneity in how climate risk affects various types of insurers and different regions. In terms of types of insurers, property-casualty insurers seem to be more vulnerable to the impact of physical climate risks, while life insurers show more substantial adaptive capacity in dealing with climate transition risks; at the regional level, insurers in coastal regions show stronger adaptability and resilience in coping with climate risks and especially have certain advantages in the field of green insurance. On the other hand, inland regions face more severe climate risk challenges due to relatively weak infrastructure, and insurers need to focus more on improving their resilience and ability to cope with transition risks.

6.2. Policy Recommendations

Based on the findings of this paper, climate physical risk has a more significant negative impact on the operational efficiency of insurance companies by increasing their exposure to natural disaster losses. In contrast, climate transition risk may indirectly affect insurance companies’ operational efficiency by promoting green insurance growth. In addition, disaster resilience infrastructure can help mitigate the adverse effects of climate physical risk to a certain extent. At the same time, the application of insurance technology may also serve as a buffer in the early stages of climate transition risk, thus helping insurers achieve progressive transition. The study also shows significant adaptive differences between different types of insurers and insurers in different regions in responding to climate risks. Accordingly, policymakers can implement targeted measures that leverage these differentiated characteristics to bolster insurers’ capacity to manage climate risks, reinforce industry resilience, and enable a more effective response to transitioning to a low-carbon economy. Based on these insights, this paper presents the following three policy recommendations:

First, to mitigate the impact of physical climate risk on the operational efficiency of insurance companies, the government should implement comprehensive measures. These measures must focus on strengthening disaster-resilient infrastructure and upgrading risk management capabilities. The government can appropriately increase investment in disaster-resistant infrastructure construction in areas with high climate physical risk. In order to reduce natural disaster losses caused by extreme weather events at the source, the pressure on insurance companies to pay claims must be reduced. In addition, the government should also prompt insurance companies to establish a more complete disaster risk assessment system and enhance their ability to recognize and respond to natural disaster loss exposure. Insurance companies can respond in advance by strengthening risk monitoring and early-warning mechanisms, thereby effectively reducing the adverse impact of physical climate risks on operational efficiency.

Second, because climate transition risks may put more significant pressure on insurance companies at the initial stage, and in order to help them better adapt to the transition to a low-carbon economy, it is recommended that the government encourage and support insurance companies to increase their investment in insurance science and technology and to make use of advanced technologies such as big data and artificial intelligence to enhance their risk assessment, pricing, and management capabilities. With the help of technology, insurance companies can identify and control climate transition risks more accurately, thus alleviating the initial operational pressure to a certain extent and gradually improving operational efficiency. In addition, in order to reduce the adaptation costs that may be incurred during the initial period of green insurance promotion, the government may consider encouraging insurance companies to invest more in the research development and innovation of green insurance products through measures such as tax incentives and financing support, at the same time strengthening market publicity to enhance consumers’ awareness of and demand for green insurance products.

Third, in response to the differences exhibited by different regions and types of insurance companies in responding to climate risks, the government should adopt differentiated policy support tailored to local conditions and insurance policies. For property insurers facing more significant pressure from physical climate risks, the government can help mitigate risks by promoting reinsurance mechanisms and disaster response capacity building. In contrast, life insurers can better cope with climate transition risks, and the government should support them by investing more in green investment and green insurance product innovation. At the regional level, insurance companies in coastal areas have a relative advantage in green insurance and disaster resilience. Therefore, the government should strengthen its support for inland areas to enhance their disaster resilience infrastructure and green financial development to strengthen their ability to cope with climate risks.

Although this study was conducted within China’s institutional and regulatory context, where government involvement plays a central role in shaping responses to climate risk, the insights derived may still have broader relevance. The nonlinear effects of transition risk identified in our analysis offer meaningful implications for countries undergoing low-carbon transitions, highlighting the dynamic and stage-specific challenges that insurers are likely to face. Moreover, the moderating roles of disaster-resilient infrastructure in alleviating the operational impact of physical risks and insurance technology in supporting insurers’ gradual adaptation to transition risks provide valuable references for emerging markets. While our analysis is grounded in a policy-driven environment, we acknowledge that insurers may rely more heavily on competitive forces, pricing signals, and investor expectations to shape their climate strategies in more market-oriented economies. These differences indicate the need for context-sensitive policy approaches to strengthen insurers’ climate resilience. Future research could benefit from cross-country comparisons and a closer examination of how market-based versus government-led systems influence insurers’ responses to climate risk. This would help improve the generalizability and policy relevance of empirical findings.