The Effect of Environmental Damage Costs on the Performance of Insurance Companies

Abstract

We examine worldwide Property and Casualty and Life and Health insurance companies from 2004 until 2023, implementing panel regression models and mediation analyses to show that insurers raise their reserves when they face increasing costs for their potential environmental damages, ultimately reducing their profitability and underwriting capacity. Our findings extend to the insurance sector the previous evidence on banks, demonstrating that environmental damages could affect profits and solvency of financial intermediaries. These insights are important especially for insurance managers and regulators.

1. Introduction

The growing trend of sustainability and environmental consciousness within the insurance industry incentivizes insurers to adopt responsible behaviors and green policies. The United Nations’ Principles for Sustainable Insurance (PSI) were launched in 2012 to guide the insurance industry in addressing environmental, social, and governance (ESG) risks and opportunities (https://www.unepfi.org/insurance/insurance/ (accessed on 27 August 2024)). “Sustainable insurance” (also known as “green” or “ethical insurance”) refers to integrating sustainable and eco-friendly practices in the insurance business, in order to improve performances, reduce risk, and contribute fostering the sustainability of the entire economic system and environment [1].

While the awareness of the corporate world toward sustainability is growing, how to assess corporate environmental “impacts” and environmental performance objectively and efficiently remains an open question [2]. Researchers and evaluators in the international development community discuss the definition of “impact” and its measurement. The most commonly used definition is the definition given by the Organisation for Economic Co-operation and Development (OECD): Positive and negative, primary and secondary long-term effects produced by an intervention, directly or indirectly, intended or unintended (https://cdn.odi.org/media/documents/10352.pdf (accessed on 27 August 2024)). While the concept of impact develops along multiple layers, the focus of this article is the cost incurred by corporations for their “environmental impact”. “Environmental impact” is any positive or negative change in environmental quality resulting from human interference, able to change the natural rhythm of the processes of a system [3].

Environmental impacts affect corporations on several aspects. The literature has treated this topic more frequently in relation to the management of non-financial firms like for example industrial, materials, or energy companies. However, financial intermediaries (primarily banks and insurers) play a key role for the transition to ESG and for the economy-wide risk management. Therefore, it is important to understand whether also financial corporations can create potential damages to the environment, ultimately affecting their performance. Although there is evidence that environmental damage costs can impinge on the profitability of banks [4], we could not find similar results for insurance companies.

To fill this gap of knowledge, we study environmental damage costs of worldwide insurers from 2004 until 2023. Using the S&P Capital Trucost database, we download environmental impact ratios, which estimate, in monetary terms, the negative externality on the society and the environment derived from the companies’ use of natural resources (water, minerals, metals, natural gas, oil, coal, forestry, and agriculture), the pollution of air, land or water, or the generation of greenhouse gas or waste. Environmental impact ratios indicate to what extent revenues would be liable for the company’s potential environmental damages. We reveal that the magnitude of environmental costs is not negligible in our sample. As we look more in detail to a few companies (see the Appendix A), we show that environmental costs absorb between 0.2% and 0.3% of revenues, with the larger share of expenses being due to greenhouse gas emissions along the supply chain.

Using panel data regressions, we show that greater environmental damage costs lead insurers to increase their reserves. In addition, mediation analyses reveal that this behavior reduces profits and underwriting capacity (i.e., high leverage). Finally, we also show that the price-to-book ratio of insurers is inversely related with their environmental impact ratios. Therefore, our interpretation is that insurers facing high environmental expenses are willing to set aside high levels of reserves to safeguard the stability and longevity of their business. However, this behavior may constrain growth opportunities, reflecting into lower profits, underwriting capacity, and equity value.

These findings contribute to extend the previous evidence on banks [4] to the broader set of financial intermediaries, demonstrating that environmental damages could sort a relevant effect on the profitability and leverage of insurers. Therefore, our intuition is that the inappropriate (intended or unintended) use of environmental resources is harmful for the environment, but also can impinge on corporate financial health. This holds not only for businesses that, more evidently, manage natural resources or interact more directly with the natural environment, but it applies also to the financial sector.

Our article differentiates from [4] showing that the mechanism driving the effect from environmental costs to business performance flows through insurance reserves. Reserves, indeed, represent an item that strongly characterizes the insurance industry. Overall, reserves are a fundamental component of a sound financial management of insurers. By maintaining adequate reserves, companies can mitigate risks, and safeguard the stability and longevity of their business. Therefore, the insights of this article are important for insurance managers, actuaries, and policy makers. We suggest that a more efficient reserve management would contribute to avoid that environmental costs could influence in a negative way the stability and financial integrity of insurance companies.

The article is organized as follows. Section 2 relates our topic to the most recent literature. Section 3 presents the data and the variables that we use in the analysis. Section 4 develops our main working hypothesis. Section 5 outlines the results. Section 6 concludes.

2. Literature

With the growing interest on corporate sustainability, a few studies treated the topic of sustainability in the insurance sector. Ref. [5] conduct a systematic literature review on the research about ESG in the insurance sector, reporting that, so far, scholars gave more attention on the areas of risk, underwriting and investment management, compared to other areas like, for example, claims management and sales. Moreover, within the existing studies, climate change and the environmental factor have received even stronger focus.

In this stream of research, a few article demonstrate that sustainability affects insurers’ financial performance. Nonetheless, the evidence is heterogeneous. For example, ref. [6] finds that sustainable insurers exhibit higher return-on-assets (i.e., are more profitable) and tend to purchase low reinsurance compared to less sustainable firms. In contrast, ref. [7] reports that high ESG insurers have low return-on-assets and pay high taxes. Other papers use market-based measures of equity performance. For example, ref. [8] reveal that upgrades in the ESG ratings of insurers lead to abnormal stock returns. Similarly, ref. [9] finds that stocks issued by Property and Casualty insurers with high ESG scores deliver positive excess returns. The author shows that also the price-to-book ratio, i.e., a widely employed measure of equity performance for insurers, is negatively related to the company’s ESG score. In line with these previous studies, our analysis uses insurers’ return-on-assets and price-to-book ratios in order to measure corporate performance and valuation. The innovative feature in our approach will be to test the impact on these dimensions from environmental damage costs.

In fact, in the broad framework of corporate sustainability, environmental impacts (and damages) cover a key aspect, with important consequences on financial management. Environmental impacts (and environmental damages) cover a key aspect of corporate sustainability, with important consequences on financial management. Most of the literature on this subject focuses on non-banking firms, for which the interplay with environmental resources is evidently straightforward. For example, ref. [10] uses environmental data from Japanese chemical companies showing that environmental costs decrease firms’ profitability. Analysing Japanese manufacturing firms, ref. [11] discover that greenhouse gas reductions increase their financial performance.

Other studies discussed environmental damage costs from an accounting perspectives. Among others, ref. [12,13] outline the main elements of environmental management accounting (EMA), which is performed by private or public corporations combining financial accounting, cost accounting and mass balances to increase material efficiency, reduce environmental impacts and risks, and reduce costs of environmental protection.

In the context of financial intermediaries (primarily banks and insurers), we have scarce evidence on environmental damages. Nonetheless, recent evidence in [4] shows that banks environmental damage costs affect in a negative way the profitability of banks. Evidently, the financial sector plays a key role in leading toward a more sustainable development [14,15], therefore it is important to understand more deeply if financial firms can create damages to the environment, and to what extent these damages may impinge on their financial health. As mentioned, the relevance of environmental aspects are recognized in the existing insurance literature [5], but we know very little about environmental impacts of insurance firms. To fill this gap of knowledge, our analysis uses environmental impact ratios, which quantify the damage (in economic terms) due to the usage of natural resources and the interaction with the environment.

Finally, dealing with the topic of environmental damages, we can also link our paper to the research on the effects of climate change on insurers. Ref. [16] highlight that insurers are adapting their business model to the realities of climate change and energy volatility. Mainstream science and customers have changed the way they construct buildings, transport people and goods, design products and produce energy. Consequently, insurers are expanding their efforts to improve their "green" behaviors and resilience to the climate change threat. Similarly, ref. [17] claims that insurers are ubiquitous players in the economy and society, therefore regulators and policy makers provide guidelines to the whole insurance sector to adopt responsible behaviors and shape climate policies in an effective way. Clearly, climate change poses serious concerns to insurance companies financial health. Ref. [18] argue that more frequent and intense catastrophic events may threaten the solvency of insurers. Ref. [19] explore the consequences on United States insurers’ profitability from massive losses due to extreme weather. The evidence suggests that the insurance industry can withstand a series of extreme shocks relatively well in the short-term. However, in the longer term, availability and affordability of insurance could be damaged, due to price and quantity adjustments, along with regulatory and public policy developments. In the framework of climate change adaption and mitigation, our results suggest that, a proactive behavior to climate change that reduces also insurers’ own environmental impacts could limit potential negative consequences on profits.

3. Data and Variables

Sample and Variables

S&P Global Trucost provides environmental data for worldwide companies classified by industries and geographies. We consult the “Environmental Register” of Trucost to obtain environmental data on insurance companies (Property and Casualty and Life and Health) from all geographies, i.e., Africa, Asia-Pacific, Europe, Latin America and the Caribbean, Middle East, United States and Canada. In particular, we download so-called environmental “impact ratios”. For each company, this number is the ratio between estimated environmental “damage costs” and total revenues. “Damage costs” quantify, in monetary terms, the negative externality associated with the use of a natural resource (water, minerals, metals, natural gas, oil, coal, forestry, and agriculture), the emission of a pollutant, or the generation of greenhouse gas or waste. Impact ratios include both direct damage costs, i.e., environmental costs due to a company’s direct operations, and also indirect damage costs, i.e., costs arising inside the company’s supply chain. For the assessment of environmental damages, Trucost uses mainly annually updated information disclosed by the company itself. Damage costs are computed by multiplying the company’s natural resources used or pollutants emitted (e.g., m3 of water or tCO2e) by environmental valuation coefficients. Valuation coefficients are factors that represent the average damage value, i.e., the external cost of damage to human, natural and built capital, resulting from an organization’s direct and indirect use of natural resources or the emission of pollutants. In lack of company’s disclosure, Trucost uses an econometrics environmental input–output model (EEIO) that approximates the damage originating from the company’s operations as well as its supply chain tiers. More information on Trucost can be found at https://www.spglobal.com/esg/trucost (accessed on 27 August 2024), while the outline about the methodology implemented for the collection of environmental data can be found at https://portal.s1.spglobal.com/survey/documents/SPG_S1_Trucost_Environmental_Data_Methodology.pdf (accessed on 27 August 2024). In the Appendix A we report an example of the environmental data classification provided by S&P Capital Trucost. We display items for a Property and Casualty insurer (State Farm Insurance, Bloomington, IL, USA) as well as a Life and Health insurer (MetLife, New York City, NY, USA).

We have in the sample a total of 1866 insurer-year observations.

Table 1. Number of insurance companies in the sample.

Geography	Property and Casualty	Life and Health	Total
Africa	13	69	82
Asia-Pacific	250	333	583
Europe	145	151	296
Latin America and Caribbean	7	0	7
Middle East	29	9	38
United States and Canada	590	270	860
Total	1034	832	1866

displays the sample composition across geographies and insurance segments. We have a larger number of Property and Casualty insurers, mainly located in the United States and Canada, Asia-Pacific, and Europe. The Life and Health insurers in the sample, instead, are concentrated in Asia-Pacific, followed by United States and Canada, and Europe.

For each company, the environmental impact ratio is denoted with $IR$, and divides the total environmental damage costs by the total revenues of the company. Thus, $IR$ represents potential costs if the insurer would be held responsible for its environmental damages. Our goal is to test whether environmental impacts measured by $IR$ are associated with the level of insurance reserves. To measure reserves, we use the following variables. $LN\_RESERVES$ is the natural logarithm of total insurance reserves and liabilities for insurance and investment contracts (in dollar terms). $RESERVES\_EQ$ are total policy reserves as a multiple of GAAP equity, while $RESERVES\_AS$ are reserves for insurance and investment contracts as a percent of total assets.

Moreover, we verify the effect from $IR$ on leverage, profitability, and equity value. Our measure for leverage ($LEV$) is the ratio of gross premiums to policholder surplus [20,21,22,23]. We also tested models using the ratio of net premiums written to surplus, but results changed very marginally. For this reason, we let them available upon request. This ratio measures the efficiency with which the insurer uses its capital resources to generate business. An insurers with a relatively low $LEV$ is not fully utilizing its capital, and it has more room for growth, i.e., has higher capacity to underwrite new policies. In contrast, a high $LEV$ indicates a more aggressive underwriting and greater risk. For insurers with high leverage, the exposure to pricing errors is also larger. Potential losses due to underpricing of policies are related to the amount of premiums written, while policyholder surplus measures the cushion available to absorb such losses [24].

To assess profitability, we employ alternatively the net margin ratio or the return-on-assets. The margin ratio ($NETMARGIN$) is net income divided by total net premiums earned [20,21]. It calculates the degree of profit of the insurer produced from its total revenue. It measures the amount of net profit that a company obtains per dollar of revenue gained. Ref. [25] shows that operating margins are positively correlated with the rate of solvency. Thus, a high (low) margin ratio is a signal for high (low) financial solidity. The return-on-assets ($ROA$) is the ratio of net income to total assets, and high $ROA$ denotes higher profitability. For example, ref. [7] uses the return-on-assets to assess the profitability of insurers in relation to corporate sustainability.

To assess the value of the company’s equity, we compute the price-to-book ratio ($PB$) [26]. Ref. [27] recommend to use book value multiples in valuing insurance companies compared to earnings multiples. The authors argues that, unlike nonfinancial firms, the book value of equity seems to be a reasonable predictor of future earnings of insurers, as proved for example by evidence from [24]. Finally, in our regressions the variable $SIZE$ accounts for the company’s size, and is computed as the natural logarithm of total assets [20].

The definitions of all our variables are summarized in

Table 2. Definition of variables.

Variables	Definition
$IR$ (%)	Total environmental damage costs (direct and indirect) in percentage to total revenues. Environmental damage costs are computed multiplying water used, waste generated, air pollution generated, greenhouse gas emissions, land and water pollution generated, natural resources used (minerals, metals, natural gas, oil, coal, forestry, and agriculture) by the respective environmental impact valuation coefficients elaborated by S&P Global Trucost. Valuation coefficients quantify the average damage value, i.e., the external cost of damage to human, natural and built capital, resulting from an organization’s direct and indirect use of natural resources or the emission of pollutants. In lack of company’s disclosure, Trucost uses an econometrics environmental input–output model (EEIO) that approximates the damage originating from the company’s operations as well as its supply chain tiers.
$LN\_RESERVES$	Natural logarithm of total insurance reserves and liabilities for insurance and investment contracts.
$RESERVES\_EQ$ (%)	Total policy reserves as a percent of GAAP equity.
$RESERVES\_ASS$ (%)	Total reserves for insurance and investment contracts as a percent of total assets.
$LEV$	Gross premiums written divided by policyholder surplus
$NETMARGIN$ (%)	Net income as a percent of total premiums earned.
$ROA$ (%)	Net income as a percent of total assets.
$SIZE$	Natural logarithm of total assets
$PB$	Market value of equity divided by total book value equity. Market value of equity is calculated as the product between average share price and total shares outstanding.
$PRED\_PB$	Predicted values from a regression of $PB$ on return-on-equity, and fixed effects for segment, geography, and time.
E (0–100)	S&P Capital IQ environmental score from ESG scores.

. After winsorizing the variables at the 1st and 99th percentiles to mitigate the potential influence of outliers, for each segment we display descriptive statistics in

Table 3. Descriptive statistics for insurers.

	Mean	Median	Min	Max	Standard dev
Property and Casualty
$IR$	0.3698	0.2600	0.2212	81.9201	2.5429
$LN\_RESERVES$	15.9111	16.2134	9.1615	20.5232	1.7715
$RESERVES\_EQ$ (%)	2.9801	2.3756	0.0000	42.2954	2.6771
$RESERVES\_ASS$ (%)	54.7823	57.7870	0.0000	88.6401	17.1367
$LEV$	1.5351	1.2944	−18.991	41.9522	2.0101
$NETMARGIN$ (%)	11.4809	9.5645	−12.9733	43.6601	12.4267
$ROA$ (%)	2.62758	2.3633	−17.6890	16.2553	3.7643
$PB$	162.1000	131.3238	17.6282	585.6034	107.3765
$SIZE$	16.4211	16.7340	9.7496	21.1779	1.7571
E	37.7800	30.0000	3.0000	97.0000	24.0901
Life and Health
$IR$	0.2762	0.2601	0.2204	1.7302	0.08316
$LN\_RESERVES$	17.8625	18.1673	10.6423	21.9211	1.8073
$RESERVES\_EQ$ (%)	10.4209	7.0615	0.0000	42.2948	9.4681
$RESERVES\_ASS$ (%)	64.7522	71.8589	0.0000	93.2867	20.5233
$LEV$	1.4187	1.1290	0.2993	13.9511	1.1686
$NETMARGIN$ (%)	12.2190	10.8956	−12.9799	43.6622	11.8418
$ROA$ (%)	1.0615	0.6627	−17.6817	14.7387	2.4170
$PB$	142.7000	113.6056	16.9637	585.6865	102.3837
$SIZE$	18.7301	19.0242	11.1326	23.2101	2.0070
E	44.9600	44.0000	2.0000	98.0000	23.9805

, while correlation coefficients in

Table 4. Correlation coefficients—Property and Casualty insurers.

	$IR$	$LN\_RESERVES$	$RESERVES\_EQ$	$RESERVES\_ASS$	$LEV$	$NETMARGIN$	$ROA$	$SIZE$	E
$IR$	1.000
$LN\_RESERVES$	0.040	1.000
	(0.247)
$RESERVES\_EQ$	0.130 ***	0.275 ***	1.000
	(0.000)	(0.000)
$RESERVES\_ASS$	0.062 *	0.393 ***	0.581 ***	1.000
	(0.069)	(0.000)	(0.000)
$LEV$	0.031	−0.019	0.717 ***	0.321 ***	1.000
	(0.416)	(0.632)	(0.000)	(0.000)
$NETMARGIN$	−0.006	−0.024	−0.248 ***	−0.224 ***	−0.183 ***	1.000
	(0.858)	(0.502)	(0.000)	(0.000)	(0.000)
$ROA$	−0.010	−0.018	−0.362 ***	−0.220 ***	−0.176 ***	0.774 ***	1.000
	(0.765)	(0.613)	(0.000)	(0.000)	(0.000)	(0.000)
$SIZE$	0.045	0.971 ***	0.207 ***	0.244 ***	−0.004	0.028	−0.015	1.000
	(0.153)	(0.000)	(0.000)	(0.000)	(0.921)	(0.433)	(0.655)
E	−0.000	0.370 ***	0.290 ***	0.329 ***	0.047	−0.096 *	−0.155 ***	0.351 ***	1.000
	(0.994)	(0.000)	(0.000)	(0.000)	(0.409)	(0.059)	(0.001)	(0.000)

See definitions of variables in Table 2. * p < 0.10, *** p < 0.01.

and

Table 5. Correlation coefficients—Life and Health insurers.

	$IR$	$LN\_RESERVES$	$RESERVES\_EQ$	$RESERVE{S}_{A}SS$	$LEV$	$NETMARGIN$	$ROA$	$SIZE$	E
$IR$	1.000
$LN\_RESERVES$	0.103 **	1.000
	(0.014)
$RESERVES\_EQ$	0.087 **	0.182 ***	1.000
	(0.035)	(0.000)
$RESERVES\_ASS$	0.024	0.041	0.543 ***	1.000
	(0.565)	(0.331)	(0.000)
$LEV$	0.113 *	0.151 **	0.277 ***	0.155 **	1.000
	(0.077)	(0.024)	(0.000)	(0.021)
$NETMARGIN$	0.022	−0.180 ***	−0.080 *	−0.213 ***	−0.149 **	1.000
	(0.618)	(0.000)	(0.069)	(0.000)	(0.040)
$ROA$	−0.045	−0.294 ***	−0.302 ***	−0.266 ***	0.217 ***	0.293 ***	1.000
	(0.243)	(0.000)	(0.000)	(0.000)	(0.001)	(0.000)
$SIZE$	0.072 **	0.956 ***	0.083 **	−0.105 **	0.095	−0.080 *	−0.243 ***	1.000
	(0.042)	(0.000)	(0.045)	(0.012)	(0.136)	(0.068)	(0.000)
E	0.120 **	0.360 ***	0.150 ***	0.070	−0.341 ***	−0.001	−0.140 ***	0.402 ***	1.000
	(0.011)	(0.000)	(0.006)	(0.205)	(0.000)	(0.989)	(0.005)	(0.000)

See definitions of variables in Table 2. * p < 0.10, ** p < 0.05, *** p < 0.01.

. We notice that $IR$ does not differ considerably between the two groups, as the median $IR$ is 0.26 in both segments. This value is close in magnitude to impact ratios estimated for banks by [4]. Reserves are higher for Life and Health insurers, for which $RESERVES\_EQ$ and $RESERVES\_ASS$ are respectively 7.0% and 72%, compared to lower values equal to 2.4% and 57.8% for Property and Casualty insurers. Instead, Property and Casualty insurers seem to be more profitable, with median $ROA$ equal to 2.4% with respect to 0.7% $ROA$ of Life and Health insurers. However, $NETMARGIN$ is close to 10% inside both groups. The correlation tables reveal a positive correlation between $IR$ and reserves, with maxim correlation equal to 0.13. For Property and Casualty insurers the correlation is significant with $RESERVES\_EQ$ and $RESERVES\_ASS$, while for Life and Health insurers the correlation is significant with $RESERVES\_EQ$ and $LN\_RESERVES$.

4. Development of the Working Hypothesis

In the first part of our analysis, the goal is to test whether there is a significant association between environmental costs (measured by $IR$) and insurance reserves. By maintaining adequate reserves, insurers can mitigate risks and remain financially stable. Companies need to consider a variety of factors to ensure the adequacy of their reserves, not only related to claims severity and pricing, but also factors associated to economic conditions and cost management. The German supervisory authority (Bundesanstalt für Finanzdienstleistungsaufsicht –BaFin) points out that sustainability risks constitute a relatively new type of risk for insurers, emphasizing that insurance’ reserves should take into account potentially adverse changes in the environmental conditions (see https://www.bafin.de/SharedDocs/Downloads/EN/Merkblatt/dl_mb_Nachhaltigkeitsrisiken_en.html (accessed on 23 September 2024)). Therefore, our conjecture is that high environmental impacts would require insurers to hold larger reserves that would contribute to offset ongoing costs.

In a second step of the analysis, we want to verify if the effect from environmental costs on reserves reflects also on the business performance of insurers, as measured by profits ($NETMARGIN$ and $ROA$) and leverage ($LEV$). In fact, the previous literature shows that high levels of reserves may impinge on insurers’ profits and solvency. For example, ref. [28] study the practice of over-reserving, arguing that one of the primary causes lies in the desire of insurers to develop a safety margin to help offset possible future adverse loss experience, thus providing increased security for policyowners. The authors develop a theoretical model proving that over-reserving reduces net income, also causing some deterioration of the financial strength of the company as measured by our variable $LEV$. Ref. [29] examines reserve errors inside property and liability insurers, arguing that high reserve levels impair surplus, diminishing also policyholders’ confidence. Finally, ref. [30] proposes a model to explain the underwriting cycle, in which insurer profits and capacity are positive correlated.

Therefore, based on this previous evidence, we advance the following working hypothesis which we will test empirically in the next session:

Working hypothesis: Raising costs for environmental damages as measured by high environmental impact ratios lead insurers to increase their reserves, ultimately decreasing their profitability and capacity.

5. Results

5.1. Effect from Environmental Damage Costs on Reserves

To test our working hypothesis more formally, we first estimate the following equation of bank j’s reserves in year t on environmental impact ratios:

$$\begin{array}{c}\hfill {Reserves}_{r,j,t}={\alpha }_{r,j,t}+{\beta }_{r,j,t}I{R}_{j,t}+{\delta }_s+{\lambda }_g+{ϵ}_t.\end{array}$$

(1)

The subscript r denotes the measure for reserves among $LN\_RESERVES$, $RESERVES\_EQ$, and $RESERVES\_ASS$. $IR$ is the impact ratio. $\delta$ are fixed effects for the insurance segment (s) between Property and Casualty and Life and Health. $\lambda$ are fixed effects for the geographic area (g) among Africa, Asia-Pacific, Europe, Latin America and Caribbean, Middle East, United States and Canada. Finally, $ϵ$ are year (t) fixed effects. Standard errors are clustered at the firm level.

In line with our conjecture,

Table 6. Regressions of reserves on environmental impact ratios.

	(1)	(2)	(3)
Regressors	$\mathit{LN}\mathbf{\_}\mathit{RESERVES}$	$\mathit{RESERVES}\mathbf{\_}\mathit{EQ}$	$\mathit{RESERVES}\mathbf{\_}\mathit{ASS}$
$IR$	0.0290 ***	0.0758 ***	0.2874 ***
	(0.0110)	(0.0272)	(0.0421)
Observations	1416	1435	1440
R-squared	0.2910	0.3660	0.1950
Segment fixed effects	yes	yes	yes
Geography fixed effects	yes	yes	yes
Time fixed effects	yes	yes	yes

See definitions of variables in Table 2. Robust standard errors in parentheses. *** p < 0.01.

shows that reserves increase significantly in environmental impact ratios. The effect on $RESERVES\_ASS$ is higher in magnitude, with coefficient close to 0.29. Therefore, it seems that insurers facing increasing environmental damage costs set aside also more money for unexpected losses.

5.2. Effect from Environmental Damage Costs on Profits and Leverage

In

Table 7. Regressions of profits (columns 1 and 2) and leverage (column 3) on environmental impact ratios.

	(1)	(2)	(3)
Regressors	$\mathit{NETMARGIN}$	$\mathit{ROA}$	$\mathit{LEV}$
$IR$	−0.0694 *	−0.0174 **	0.0260 ***
	(0.0382)	(0.0086)	(0.0044)
$SIZE$	−0.1297	−0.2251	−0.0442
	(0.4927)	(0.1472)	(0.0481)
Observations	1320	1541	940
R-squared	0.0864	0.1091	0.0489
Segment fixed effects	yes	yes	yes
Geography fixed effects	yes	yes	yes
Time fixed effects	yes	yes	yes

See definitions of variables in Table 2. Robust standard errors in parentheses. *** p < 0.01, ** p< 0.05, * p< 0.1.

we test if $IR$ plays a significant role on profitability and leverage. We use the model in (1) using as dependent variable alternatively $NETMARGIN$, $ROA$, and $LEV$. The equations include also a control for $SIZE$, to assess if we estimate a significant effect from environmental costs at the net of the corporate size. We find that the coefficient on $IR$ is significantly negative on both $NETMARGIN$ and $ROA$, in line with evidence of [4] based on environmental damage costs of banks, showing that banks’ profitability is negatively correlated with environmental impact ratios. Leverage, instead, increases in $IR$, i.e., companies become financially unstable with increasing environmental impacts.

According to our working hypothesis, the effect from $IR$ on reserves that we displayed in Table 6 should mediate the findings in Table 7. To test this argument analytically, we conduct a mediation analysis. Statistical mediation analysis is about quantifying the indirect effect of an independent variable (X) on the dependent variable (Y) through a third variable called the mediator (M). That is, X has a direct effect on Y, while has also an indirect effect on Y mediated by M. We follow the procedure of [31] that modifies the method originally introduced by [32]. The procedure works as follows. In a first step we estimate the following system of equations:

$$\begin{array}{c}\hfill Y={\beta }_0+cX+ϵ\\ \hfill M={\beta }_0+aX+ϵ\\ \hfill Y={\beta }_0+bM+c^{\prime }X+ϵ.\end{array}$$

(2)

The first equation estimates the direct effect from X on Y. The second equation tests the relationship between X and the mediator variable, which is assumed to be significant, and important also to explain the ultimate effect observed on Y. In the third equation we regress Y on M partialling out, or statistically controlling for X. If there is full mediation, both the coefficients c and b must be statistically significant, as this means that Y and M may be correlated because X causes both. If c^′ is not significant, it indicates that the magnitude of the effect from X to Y is reduced to zero after controlling for the mediator.

In a second step, we can compute the Sobel’s z-test [33] to test explicitly that the direct effect reduces significantly through the indirect (mediated) path:

$$\begin{array}{c}\hfill z={\displaystyle \frac{a\ast b}{\sqrt{b^2{s}_a^2+a^2{s}_b^2}}},\end{array}$$

(3)

where a and b are the coefficients estimated by the system of simultaneous equations, while $s_a$ and $s_b$ are the standard errors of the two coefficients. For example, if z≥±1.645, then the mediation (c−c^′ or a ∗ b) is statistically significant at 0.10. Complete mediation implies that the z is significant and the direct path from X to Y captured by the coefficient c is not significant. See more details about this procedure at https://davidakenny.net/cm/mediate.htm (accessed on 27 August 2024).

We use $NETMARGIN$, $ROA$, and $LEV$ as the outcome variable (Y) for three separate mediation analyses. The independent variable (X) is $IR$, while the mediator (M) is $RESERVES\_ASS$. This procedure allows to assess if the effect from environmental costs on profits and leverage is channeled by reserves.

Table 8. Effect from $IR$ on $LEV$ mediated by $RESERVES\_ASS$.

	Mediator (M)	Outcome (Y)
	$\mathit{RESERVES}\mathbf{\_}\mathit{ASS}$	$\mathit{LEV}$
Independent variable (X): $IR$	0.3801 **	0.00788
	(0.1880)	(0.0223)
$RESERVES\_ASS$		0.0341 ***
		(0.0041)
Observations	850	850
Sobel’s z-test	1.9690 **
Indirect effect/Direct effect	62%

See definitions of variables in Table 2. Standard errors in parentheses. *** p < 0.01, ** p< 0.05.

,

Table 9. Effect from $IR$ on $NETMARGIN$ mediated by $RESERVES\_ASS$.

	Mediator (M)	Outcome (Y)
	$\mathit{RESERVES}\mathbf{\_}\mathit{ASS}$	$\mathit{NETMARGIN}$
Independent variable (X): $IR$	0.3673 *	0.0292
	(0.2097)	(0.1465)
$RESERVES\_ASS$		−0.154 ***
	(0.0191)
Observations	1318	1318
Sobel’s z-test	−1.7172 *
Indirect effect/Direct effect	7.3%

See definitions of variables in Table 2. Standard errors in parentheses. *** p < 0.01, * p< 0.1.

and

Table 10. Effect from $IR$ on $ROA$ mediated by $RESERVES\_ASS$.

	Mediator (M)	Outcome (Y)
	$\mathit{RESERVES}\mathbf{\_}\mathit{ASS}$	$\mathit{ROA}$
Independent variable (X): $IR$	0.3820 *	0.0003
	(0.2256)	(0.0391)
$RESERVES\_ASS$		−0.0397 ***
		(0.0046)
Observations	1409	1409
Sobel’s z-test	−1.6680 *
Indirect effect/Direct effect	2.2%

See definitions of variables in Table 2. Standard errors in parentheses. *** p < 0.01, * p< 0.1.

show estimates from the three systems of equations, together with the Sobel’s z-statistics. The results show that full mediation for $LEV$ is accepted with a 5% level of significance, while for $NETMARGIN$ and $ROA$ mediation is accepted with a 10% level of significance. Clearly, the mediation analysis reveals that reserves explain in a considerable way the impact that environmental costs have on leverage. In fact, the 62% of the effect of $IR$ on $LEV$ is mediated by $RESERVES\_ASS$. We also verified that the results are similar applying the procedure in [34], based on a bootstrap test of the indirect effect (a ∗ b). Furthermore, we tested that the procedure gave similar results also using the other two proxies for reserves ($LN\_RESERVES$ and $EQUITY\_ASS$). All these latter results are omitted for brevity but available from the author upon request. Therefore, the mediation analyses provide results supporting the plausibility of our working hypothesis. That is, through an effect on reserves, high costs for environmental damages lower profits and underwriting capacity.

5.3. Effects from Environmental Damage Costs on Price-to-Book Ratios

We now verify if environmental costs have an impact on the company equity valuation. To approximate the insurer’s equity value we use the price-to-book ratio ($PB$). In addition, to improve the reliability of the valuation metric, we follow [27], who proposes to use a method that simultaneously extracts information from both book value and earnings instead than using univariate price multiples. The approach is to condition the price-to-book ratio on the return-on-equity, so that the resulting valuation reflects earnings in addition to book value. The author shows that this method improves the valuation accuracy of book value multiples. Therefore, to implement this method, we run a preliminary regression of $PB$ on return-on-equity plus fixed effects, calling the predicted values $PRED\_PB$. In a second stage, we estimate a panel regression of $PRED\_PB$ on $IR$ controlling for the company’s $SIZE$.

In

Table 11. Regressions of equity valuation on environmental impact ratios.

	(1)	(2)	(3)	(4)
Regressors	$\mathit{PB}$	$\mathit{PB}\mathbf{\_}\mathit{RES}$	$\mathit{PB}$	$\mathit{PRED}\mathbf{\_}\mathit{PB}$
$IR$	3.3083 ***	0.4488 ***	−21.3734 ***	−3.8147 *
	(0.3490)	(0.0863)	(6.9911)	(2.1407)
$IR$ × $IR$			0.3028 ***	0.0523 **
			(0.0846)	(0.0265)
$SIZE$	−16.2018 ***	−0.3227	−15.9722 ***	−0.2848
	(4.5791)	(1.1334)	(4.5959)	(1.1367)
Observations	1507	1537	1507	1537
R-squared	0.1470	0.3839	0.1492	0.3834
Segment fixed effects	yes	yes	yes	yes
Geography fixed effects	yes	yes	yes	yes
Time fixed effects	yes	yes	yes	yes

See definitions of variables in Table 2. Robust standard errors in parentheses. *** p < 0.01, ** p< 0.05, * p< 0.1.

columns (1) and (2) we find that the effect from $IR$ on $PB$ and $PRED\_PB$ is positive and significant. To gain a deeper insight, in columns (2) and (3) we test models that include also a quadratic term ($IR$×$IR$). These models reveal an interesting behavior: While the coefficient on $IR$ is highly negative, the coefficient on the squared term is positive although smaller in magnitude. Thus, we observe that relationship between environmental costs and equity value of insurers is non linear (i.e., U-shaped). When environmental costs are relatively low, it seems that market values decrease more largely than book values. Arguably, this may be due to the negative effects on growth opportunities and reputation that lead investors to discount stock values at higher returns. However, when environmental costs are more considerably high, the value of insurers would increase. This sign could depend from the increasing reserves that we found previously. Therefore, our results are only partially consistent with the findings of [4], who proves that price-to-book ratios of banks decrease in their environmental impact ratios. In fact, the author does not test non-linear effects as in our models. In contrast, the pattern that we outlined raises some concerns about the usage of the price-to-book ratio as a standard benchmark for the valuation of insurers. Our intuition is that, when environmental costs are substantial, increased reserves would reduce equityholders’ surplus in a greater proportion than the market value of equity, leading to a higher $PB$.

5.4. Environmental Impact Ratios and Environmental Ratings

Finally, we verify whether the relationship between $IR$ and the business performance of insurers is robust also while we control for the companies’ environmental performance as measured by the environmental score of their ESG ratings. In our database we could find ESG ratings available only from 2014, and we have a sample of 900 firm-year observations with not-missing $IR$ and environmental score (E). However, for these firms few accounting data are not available, therefore the following regressions could be estimated on a subsample which is approximately the 40% of our initial sample.

In Table 4 and Table 5 the correlation between $IR$ and E is negative, but small in magnitude and not significant. Despite the low correlation, we verify if including E in our regressions would change the effect from $IR$.

Table 12. Regressions of reserves (columns (1)–(3)), leverage (column (4), and profits (columns (5) and (6)) on environmental impact ratios controlling for environmental scores.

	(1)	(2)	(3)	(4)	(5)	(6)
Regressors	$\mathit{LN}\mathbf{\_}\mathit{RESERVES}$	$\mathit{RESERVES}\mathbf{\_}\mathit{EQ}$	$\mathit{RESERVES}\mathbf{\_}\mathit{ASS}$	$\mathit{LEV}$	$\mathit{NETMARGIN}$	$\mathit{ROA}$
$IR$	0.0197 ***	0.0870 ***	0.3149 ***	0.0233 ***	−0.0954 **	−0.0245 ***
	(0.0060)	(0.0313)	(0.0400)	(0.0021)	(0.0423)	(0.0089)
E	0.0259 ***	0.0160	0.0616	−0.0062 **	−0.0308	−0.0184 **
	(0.0065)	(0.0201)	(0.0654)	(0.0030)	(0.0277)	(0.0091)
Observations	748	752	753	451	671	844
R-squared	0.3939	0.3265	0.1802	0.1984	0.0906	0.1241
Segment fixed effects	yes	yes	yes	yes	yes	yes
Geography fixed effects	yes	yes	yes	yes	yes	yes
Time fixed effects	yes	yes	yes	yes	yes	yes

See definitions of variables in Table 2. Robust standard errors in parentheses. *** p < 0.01, ** p< 0.05.

shows that the coefficient on $IR$ has same sign as in Table 6 and Table 7, i.e., positive and highly significant on reserves and leverage, while negative on profitability. Therefore, environmental damage costs play a relevant role on reserves, profits, and leverage also when we control for environmental ratings. The influence of E on our dependent variables is less evident. There is a positive impact on reserves measured by $LN\_RESERVES$, while the other two measures of reserves do not change considerably. Similarly, $ROA$ diminishes in E, but the change of $NETMARGIN$ is not relevant. Finally, we notice that $LEV$ reacts oppositely to $IR$ and E: The underwriting capacity seems to decrease with environmental costs, while improves in environmental ratings. Overall, these outcomes suggest that environmental damage costs and environmental ratings measure different aspects of insurance companies’ sustainability. However, our baseline outcomes reveal to be robust also controlling for environmental ratings.

6. Conclusions

Using data from worldwide Property and Casualty and Life and Health insurers from 2004 to 2023, we find that costs for potential environmental damages increase insurance reserves significantly, ultimately harming profits and underwriting capacity. At the same time, we also observe that price-to-book ratios, i.e., a standard benchmark for the valuation of insurers, decrease non-linearly in environmental impact ratios.

The key insight from our findings is that costs for environmental damages, which typically are a serious concern inside non-financial firms (e.g., industrial, energy, or materials sectors), seem to be an important matter also for insurance companies. In fact, we show that insurers’ profits, leverage, and price-to-book ratios are significantly associated with environmental impact ratios. The implication for insurance managers and policy makers is that improvements in the environmental management of insurers could reflect into greater profitability and value. In addition, we also recommend to insurance managers and investors to take into consideration that potential environmental damage costs could affect valuation measures based on equity book values.

Nonetheless, we acknowledge that the analysis presents few limitations that follow-up research could contribute overcoming. First, one could exploit the granularity of the database to study more deeply the composition of environmental damage costs. As we show in our Appendix A, the largest share of the total environmental damage costs seems to be generated in the supply chain. Moreover, among the different categories of environmental costs, it appears that costs for air pollution, green house gas emissions and water use are much larger than costs for the use of land and water pollutants, the use of natural resources and the waste produced. Research focused on supply chain management could examine how these costs are generated along the supply chain, so that we could better disentangle the mechanisms that drive insurers to increase reserves when they incur greater environmental damage costs.

Furthermore, accounting studies could dig deeper into the composition of reserves and their relation with environmental damages. We used in our analysis figures that are standardized for global insurers, nonetheless focusing on specific countries could allow to separate more specifically certain figures. To give an example, in our database the Canadian insurance statements separate the general and contingency reserve from earthquake reserve, mortgage reserve and nuclear reserve.

Regarding the sample composition, it would also be possible also to include in the sample multi-line insurance firms, for which our database provides environmental data too. This could be one way to verify whether business diversification is not a relevant issue for our findings. Concerning effects on company valuation, as already mentioned, our results could be extended by testing alternative measures of value, like for example price-to-earnings ratios or Tobin’s Q. Finally, we recognize that, based on our focus on reserves, we could only test effects from environmental impact ratios on insurance liabilities. However, it will be interesting to verify if environmental impacts are related also to insurers’ assets, ultimately reflecting on their risk-taking.

Overall, all these suggestions would help to improve our understanding about insurers’ costs for their environmental damages, and their consequences on financial management. This article is a first piece of research on this topic, which is important for improving sustainability standards in the financial sector.