1
A meta-analytical review on the relationship between Corporate
Environmental Performance and Corporate Financial
Performance. Do emission trading schemes influence the
relationship?
Jan Taeke Galama 1
Supervisor: Prof. dr. L.J.R. Scholtens Co-assessor: Prof. dr. S. Homroy MSc. International Financial Management
Faculty of Economics and Business University of Groningen
Date: 10-01-2020 Abstract
The relationship between corporate environmental and corporate financial performance has been studied extensively in the research literature, and results are still contradictory. Existing meta-studies cover a wide range of environmental issues and consolidate the findings of multiple measures of environmental performance. This method hinders the investigation of how measurement approaches influence empirical findings. As such, this meta-analysis focuses on one specific environmental issue: climate change, and one operational performance outcome: corporate greenhouse gas performance. My meta-analysis includes 76 effect sizes from 34 studies, covering 172.117 observations from the period 1997-2019. I find a significant positive link between corporate greenhouse gas performance and financial performance. Additionally, I analyze whether differences in indicators for greenhouse gas performance and financial performance influence the empirical outcomes. With this meta-study, I also investigate whether industry pollution-intensity and country-level emission trading schemes stringency influence the relationship. The subgroup-analysis indicates that the relationship is significantly influenced by the corporate greenhouse gas and financial performance measures and stringent emission trading systems. However, not by industry pollution intensity.
Key words: Corporate Environmental Performance, Corporate Greenhouse Gas Performance, Corporate Financial Performance, Environmental Regulation, Emission Trading Schemes
2
1. Introduction
3 Several explanations for the inconsistent findings have been proposed, covering both methodological and theoretical issues (Endrikat, et al., 2014). First, existing literature measures CEP by a wide range of environmental issues (Albertini, 2013). Guenther and Hoppe (2014) argue that how the CEP construct is measured, is likely to affect the empirical results. As a result, it is hard to generalize theoretical arguments that support or reject a positive relationship between CEP and CFP. Second, next to the multidimensionality of the CEP construct, the multidimensional nature of the CFP construct is also argued to be a source of the inconclusive findings in the CEP-CFP literature (Busch & Hoffmann, 2011). CFP can be measured using both accounting and market-based performance indicators (Endrikat, et al., 2014). Hoffman and Bansal (2012) claim that the absence of consensus on selecting and defining the constructs of interest for CEP and CFP could be an explanation for the conflicting evidence on the relationship between CEP-CFP. Third, other moderating influences, such as the industry pollution intensity, and the effects of country-level factors are argued to be possible explanations for the conflicting findings on the CEP-CFP relationship (Endrikat, et al., 2014) (Dixon-Fowler et al., 2013) (Albertini, 2013).
The conflicting empirical findings in the existing literature are the reason for conducting this meta-study on the CEP-CFP relationship. Meta-analysis has proven to be a useful approach in areas where individual studies yielded inconclusive or even conflicting findings (as in the CEP-CFP literature) (Orlitzky, et al., 2003). A meta-analysis is a sophisticated approach to summarize, evaluate, and analyze empirical findings in a research field (Kirca & Yaprak, 2010).
4 the broader concept of CEP, they differ in several ways from other measures (Wang, et al., 2014). In contrast to other CEP measures, corporate emissions receive more attention by stakeholders (Busch & Hoffmann, 2011), impact the environment on a global level (Fujii, et al., 2012), and receive increasing attention of legislators (Wang, et al., 2014). By focusing on one specific measure of CEP: corporate GHG emissions, my study attempts to determine how corporate GHG emissions influence the CFP of companies. Apart from the multidimensionality of the CEP construct, the multidimensional nature of the CFP construct is also argued to be a source of the inconclusive findings in the CEP-CFP literature (Busch & Hoffmann, 2011). In this meta-analysis, I attempt to discover whether corporate GHG performance is more positively related to market or accounting-based measures of CFP. Furthermore, I investigate whether the industry pollution intensity moderates the relationship between corporate GHG Performance and CFP. Additionally, Endrikat et al. (2014) call for the inclusion of country-level factors. I adhere to this call by investigating whether one specific type of governmental policy, namely: emission trading schemes (hereafter: ETSs) influences the firm-level corporate GHG performance - CFP relationship
5 industrialized countries, which was signed and ratified by 187 countries. It is often seen as an essential starting point of the global battle against climate change (Böhringer, 2003). There are several types of carbon pricing policies, with an emissions trading scheme being the most prominent one (World Bank, 2019). By putting a price on emissions, the burden shifts back to the emitters who are responsible for the emissions (World Bank, 2019). Today, there are several examples of ETSs around the globe, varying in stringency and effectiveness (World Bank, 2019). Initial phases of ETSs are known for their limited effectiveness, due to numerous concessions regarding the stringency in order to create a political support base (Czerny & Letmathe, 2017). In later phases, ETSs are known for gradually becoming more stringent (Lian, et al., 2013). With this meta-study, I investigate whether the stringency of ETSs influences the CEP-CFP relationship.
6 This thesis is related to the aspects of international, financial and management as it uses meta-analytical techniques to consolidate the results of multiple studies that investigated the relationship between GHG performance and CFP. Studies that conducted their analyses based on international samples, which combined cover data on 35 countries. It further investigates whether a country-level factor (ETS stringency) influences the firm-level relationship between CEP and CFP. Furthermore, the investigated relationship is that of corporate GHG performance on corporate financial performance measured by market- and accounting-based indicators (e.g. Tobin’s Q, ROA, ROE). Additionally, management implications can be derived from this study. The CEP of a firm is part of the broader corporate social strategy. By investigating whether corporate GHG performance is related to improvements in CFP, this thesis can support managers to better understand how their sustainable strategies are related to financial performance.
The remainder of the paper is structured as followed: section 2 briefly discusses several theoretical approaches to explain the CEP-CFP relationship, discusses the existing meta-analysis, and discusses the development of hypotheses. Section 3 discusses the sample construction process, the coding procedures, and the meta-analytical methodology. Followed by section 4 that presents the results of the meta and subgroup analyses. Section 5 presents the conclusion, limitations, and recommendations for future research.
2. Theoretical background, existing meta-analyses, and hypotheses 2.1. Theoretical background
8 The trade-off theory argues that a trade-off between environmental and financial performance exists. The trade-off theory is built on the argument of Friedman's (1970) that increasing profits is the only social responsibility of firms. By investing in environmental performance, company resources are drawn away from profit-generating core business activities. Consequently, reducing a company's competitive position compared to companies that devote their investments solely to core business activities (Preston & O'Bannon, 1997). Friedman (1970) reasons that investments in the CEP of firms are an example of an agency problem, where the principal (the shareholder) and the agent (the managers) have conflicting interests. He argues that managers invest in CEP to pursue their own personal interests and career agendas at the costs of the shareholders. Hart and Zingales (2017) argue that Friedman (1970) is only partially right. They claim that shareholder welfare and market value are not equivalents and that companies should focus on the former. Companies should pursue policies consistent with the preferences of shareholders and not only pursue market value. In summary, one side argues that a win-win relationship between both constructs exists; the other argues that the relationship consists of a trade-off.
2.2 Existing meta-analysis:
In order to describe the contribution of my meta-study to the existing literature, the following section will discuss several prominent narratives, systematic, and meta-analytical reviews that have been conducted so far.
9 can be used to integrate conflicting research findings. It is the most objective approach to summarize, integrate, correct, and evaluate research findings in a research field. It can explicitly measure the effects of observed model specification variation and can estimate biases precisely (Stanley & Doucouliagos, 2012). Inconclusive findings in a research field (as in the CEP-CFP literature) contribute to the number of scholars that try to consolidate empirical findings to arrive at a consolidated conclusion (Endrikat, et al., 2014). Research on the question whether and when it pays to be green has been the subject of several narrative, systematic, and meta-analytic reviews (Endrikat, et al., 2014). However, the question whether the stringency of ETSs influences the relationship between CEP and CFP has after an extensive internet search, to the best of my knowledge, been neglected.
10 review, in contrast to a narrative review, is a method that attempts to include all studies, meeting the predefined search criteria. When clear search criteria are used and documented, a systematic review is replicable by reviewers (Stanley & Doucouliagos, 2012). The method was employed several times in the CEP-CFP literature (e.g., Molina-Azorin and Claver-Cortes, 2009; Lewandowski, 2015). The technique, however, similar to vote counts, ignores the role of sample size. This can make conclusions biased (Orlitzky, et al., 2003). By statistically aggregating the results of independent empirical studies, the meta-analytical review has much greater precision than other forms of reviews (Hunter, et al., 1982). In comparison with empirical studies, a meta-analysis uses the empirical outcomes of multiple studies and attempts to present a generalized outcome. Misspecification biases have been commonly found in all areas of empirical research and can be large enough to influence how a phenomenon in question is seen (Stanley & Doucouliagos, 2012). Meta-analytical techniques allow the modeling of the effect of observed model specification variation and are thereby able to estimate the misspecification biases (Stanley & Doucouliagos, 2012).
11 relationship. However, none of the early meta-analytical reviews on the CSP-CFP literature accounted for the multidimensional nature of the CEP and the CFP constructs. Additionally, these early meta-analyses obtained their effect sizes from empirical studies performed before 1997. Global concerns about climate change led to a breakthrough in international climate policy in the year 1997. The Kyoto Protocol, which was signed and ratified by 187 countries, contained legally binding emissions targets for industrialized countries. It is often seen as an essential starting point of the global battle against climate change (Böhringer, 2003). As such, the relationship between CSP and CFP may have changed and more recent studies may yield different results (Endrikat, et al., 2014).
12 sample), which may provide explanations for the inconsistent findings in the empirical literature. The study examines the differences between process and outcome-based CEP but does not separate between different types of outcome-based CEP in their subgroup analysis. As such, the review is not able to arrive at consolidated conclusions about corporate GHG performance. The more recent study of Busch and Lewandoski (2018) uses a sample of 65 empirical results from 32 empirical studies to study the causal relationship between CEP and CFP. They investigate whether the used constructs for both CEP and CFP predetermine the empirical outcomes. The study focuses only on one environmental issue, namely: climate change. The authors only include the operational performance dimension of climate change: corporate carbon performance, which is measured by a firm's total carbon dioxide emissions. The authors find carbon emissions vary inversely with CFP, which they measure by a wide variety of both accounting and market-based indicators. Furthermore, the study also finds variances in results between the different indicators for financial performance. Table 1 gives an overview of the existing meta-analytical literature that investigates the CEP-CFP relationship.
Table 1: existing meta-studies and the main findings
AUTHORS RELATIONS HIP PERIOD REGIO N N SELECTION FINDINGS ORLITSKY ET AL, (2003)
CSP-CFP 1972-1997 Global 17 Period CSP-CFP relationship is positive in nature and several
factors moderating the relationship ALLOCHE AND
LAROCHE, (2005)
CSP-CFP 1972-1996 Global 82 Period and
CSP construct
CSP is strongly related to CFP and both the measurement and method of the empirical study affect the outcomes. DIXON FOWLER ET
AL., (2013)
CEP-CFP 1970-2009 Global 39 Period CFP is significantly influenced by CEP. Several
important contingencies moderate the relationship.
ALBERTINI, (2013) Environmental
Management – CFP
1975-2011
Global 52 Period A positive relationship between environmental
management and CFP. The relationship is influenced by performance measures and other contingencies. ENDIRKAT ET A,
(2014)
CEP-CFP / CFP-CEP / bidirectional
All Global 149 CEP-CFP
measures
The results indicate that there is a positive and partially bidirectional relationship between CEP and CFP. They find several moderating effects.
BUSCH AND LEWANDOSKI, (2018)
Corporate Carbon Performance
All Global 32 CEP measure Corporate carbon performance positively related to CFP
Outcomes vary with CFP measures
14 select the included studies and empirical estimates only based on objective selection criteria. Table 2 gives an overview and description of the most commonly emitted GHGs.
Table 2: Most emitted greenhouse gases (IPCC, 2014)
Greenhouse Gases Emission sources Carbon Dioxide (co₂) approx. 82%
of all GHG emissions
Carbon dioxide enters the atmosphere through burning fossil fuels (oil, natural gas, and coal), solid waste, trees, and wood products. Other chemical reactions in manufacturing processes can also release it. Conversely, plants can remove carbon dioxide from the atmosphere.
Methane (CH₄) approx. 10% of all GHG emissions
Methane is emitted by the agricultural sector and during the production and transportation of coal, natural gas, and oil. It can also be emitted as a result of the decay of organic waste in landfills. Nitrous Oxide (N₂O) approx. 6% of
all GHG emissions
Nitrous oxide is both emitted during agricultural and industrial activities and by the combustion of fossil fuels and waste.
Fluorated gases approx. 3% of all GHG emissions
Several fluorinated gasses have a ‘high global warming potential’ and are emitted from a variety of industrial processes.
Note: table 2 gives an overview of the most emitted types of GHG, their contribution to total GHG emission, and the primary source of emission. Carbon dioxide, Methane, Nitrous Oxide and Fluorated gases are reported. The less emitted Hydrofluorocarbons, Perfluorocarbons, Sulphur Hexafluoride, and Nitrogen Trifluoride emissions are not reporting in the table.
15 others find that CEP is more strongly related to accounting-based indicators of CFP (Dixon-Fowler, et al., 2013). Previous studies have researched both types of proxies for financial performance, but provide conflicting results, which could be explained in differences in the CEP constructs (Busch & Hoffmann, 2011). With this meta-analysis, I investigate whether accounting or market-based indicators are more positively related to corporate GHG performance.
Second, the meta-analysis of Endrikat et al. (2014) calls for the inclusion of country-level factors, such as differences in regulatory environmental systems. From several country regulatory environmental systems, economists argue that cap-and-trade systems like ETSs are the most cost-effective way to reduce the environmental impact of countries (Bowen, 2018). The Dynamic Integrated model of Climate and the Economy (the DICE model) developed by Nordhaus (1993) attempts to determine the most effective strategy for coping with the threat of global warming. Using his model, Nordhaus (2007) compared both quantity-oriented approaches and price-type approaches. He found that price-type approaches effectively integrate the economic costs and benefits of emission reductions and are more effective. World Bank (2019) offers a cohesive overview of the current state and developments regarding carbon-pricing mechanisms. Today, there are already 28 ETS initiatives active in regional, national, and subnational jurisdictions. These carbon pricing mechanisms are covering 11 gigatons of CO2 emissions, representing 20.1% of the global emissions (World Bank, 2019). Multiple existing empirical studies investigated the effect of ETS on the CFP (e.g., Clarkson et al., 2015; Czerny and Letmathe, 2017). Nonetheless, an extensive internet search did not result in the finding of any previous meta-analysis that includes the effect of ETS on the CEP-CFP relationship.
16 for the most polluting industries are stricter and that polluting firms employ in general different strategies for reducing their emissions. My study explicitly investigates whether the relationship between corporate GHG performance and CFP is different for studies conducted in polluting-intense industries in comparison tostudies which are conducted in diverse industries.
In conclusion, the contribution of my meta-study is to systematically aggregate the conflicting empirical findings in the corporate GHG performance-CFP literature, complementing the cumulative findings in the CEP-CFP literature. By conducting subgroup analyses on several methodological topics, the moderating effect of differences in methodological artifacts, industry pollution intensity, and country-level ETS stringency can be examined.
2.3 Hypotheses development
17 the debate on the relationship between corporate GHG performance and CFP is polarized around two lines: the win-win and the win-lose approaches. Scholars that find a positive relationship between both constructs attribute their findings commonly to the concept of eco-efficiency. The eco-efficiency concept follows the reasoning that productivity gains through reduction of the use of materials, and improvements in the manufacturing processes and utilization of waste can improve the operation efficiency of firms (Kuo, et al., 2010). Eco-efficiency is expressed as the ratio of product value divided by its environmental burden (Verfaillie & Bidwell, 2000). Improved efficiency via the reduction of emissions and the utilization of by-products and waste can lead to both lower costs and more innovation, improving the competitive advantages of firms (Kuo, et al., 2010). Orsato (2006) argues that eco-efficiency emerging from the environmental performance can be seen as a business strategy, which can lead to both cost and differentiation advantages. Moreover, stakeholders are demanding more efficient companies. Market agents can change consuming and investing behavior according to the GHG performance of companies. Consumers may avoid buying products from companies that have low corporate GHG performance. As such, companies can improve their financial performance by reaping the reputational benefits associated with cleaner production, which can result in social legitimacy (Hart & Ahuja, 1996).
18 that the costs of emission reduction can differ widely between specific types of technology and over time. The relationship between corporate emissions performance and CFP is not straightforward, and different factors determine whether the relationship is positive or negative. I, therefore, investigate whether improvements in corporate GHG performance positively or negatively affects the CFP of firms. The following hypotheses are tested:
H1A: The overall relationship between corporate GHG performance and corporate financial
performance is positive
H1B: The overall relationship between corporate GHG performance and corporate financial
performance is negative
19 whether data from voluntary or mandatory reporting schemes is more related to CFP. Hence, the following hypothesis is tested:
H2: The type reporting scheme influences the empirical outcomes in the corporate GHG and corporate financial performance literature
20
H3A: Corporate GHG performance influences corporate financial performance more
positively when corporate environmental performance is measured by relative corporate emission variables than with absolute corporate emission variables
H3B: Corporate GHG performance influences corporate financial performance more
positively when corporate environmental performance is measured by absolute corporate emission variables than with relative corporate emission variables
Several measures of CFP are used in the research literature (Dixon-Fowler, et al., 2013). Most studies use either accounting-based measures or market-based measures as indicators for CFP (Albertini, 2013). Accounting-based indicators often encompass return on Assets (ROA) and return on equity (ROE) and return on sales (ROS) (Danso, et al., 2019). Accounting-based indicators reflect the internal capabilities of the firm and the efficiency to generate value by using its assets, rather than external perceptions of the performance (Orlitzky, et al., 2003). Market-based measures such as a price-earnings ratio, price per share, or earnings per share capture long-term and intangible effects (Dowell, et al., 2000). They include estimations of the firm’s prospects and reflect the notion of external stakeholders (Orlitzky, et al., 2003). Table 3 presents the description of the commonly used market and accounting-based measures for CFP in the literature.
Table 3: Measures of corporate financial performance used in the CEP-CFP literature
Note: table 3 gives an overview of some widely used indicators for corporate financial performance in the literature. It provides information about the dimension of the indicators, a short description of what the indicator measures, and an a study using the type of CFP indicator. The following indicators are included: Return on Assets, Return on Equity, Return on Investment, Return on Sales, Tobin’s Q, Total Stock Return, and Price Earnings Ratio.
MEASURE DIMENSION DESCRIPTION EXAMPLES
RETURN ON ASSETS Accounting The ratio of income to total assets. It indicates the profitability of the firm relative to its assets
Aggarwal and Dow (2012) RETURN ON EQUITY Accounting The ratio of income to total equity. It indicates the profitability of the firm relative to its
equity
Brouwers et al. (2018) RETURN ON
INVESTMENT
Accounting The ratio of the benefits of an investment divided by the costs of the investment operating. It measures the efficiency of the investment.
Czerny and Letmathe (2017) RETURN ON SALES Accounting The ratio of the operating income to net sales. It measures the profitability of a firm. Fujii et al. (2012) TOBINS Q Market The value of the firm’s assets in relation to the total market value of the firm. It is able to
incorporate the intangible effect of environmental influences.
Busch and Hoffmann (2012) TOTAL STOCK
RETURN
Market The appreciation of the stock price plus dividends paid. It is an indicator for the stock market performance.
Trumpp an Guenther (2015) PRICE EARNINGS
RATIO
Market The ratio of the share price relative to the earnings of the company. It indicates the amount of dollars an investor is willing to pay for a dollar of earning. It gives information about the subjective value of the firm.
21 Albertini (2013) and Orlizky et al. (2003) find accounting-based indicators to be more positively related to CEP. Nonetheless, Dixon-Fowler et al (2013) argue that market-based indicators are more positively related to CEP. Arguments for a more positive association between CEP and accounting-based measures support the view that corporate GHG performance commonly results from relatively small projects with short amortization (Yang, et al., 2014). The eco-efficiency concept reasons that investments in corporate GHG performance could be converted into future accounting-based improved performance (Ambec & Lanoie, 2008). A more positive result between corporate GHG performance and market-based CFP implies that investors value carbon emissions and use off-balance sheet valuation discount for GHG emissions (Griffin, et al., 2017). Investors could value improvements in GHG performance since such improvements can reduce regulatory business risks and can become more valuable due to changes in future consumption patterns (Albertini, 2013). Orlitzky et al. (2003) suggest that the multidimensional nature of the CEP construct could be an explanation for the inconsistent findings. In summary, arguments can be made for both accounting-based and for market-based measures to be more positively related to corporate GHG performance. Therefore, this meta-analysis tests whether corporate GHG performance is more positively related to accounting-based or market-based measures for CFP. The following hypotheses are tested:
H4A: Corporate environmental performance is more positively related to prior market-based
corporate financial performance than to prior accounting-based corporate financial performance
H4B: Corporate environmental performance is more positively related to prior
22 The focus in empirical CEP-CFP studies has primarily been on industrial companies, as they are concerned with the toxic emission resulting from their manufacturing processes (King & Lenox, 2001). Some studies on the CEP-CFP relationship explored specific industrial sectors; others have collected data from multiple industries to present a more generalizable outcome (Albertini, 2013). Several studies highlight the differences between industries (e.g., Russo and Fouts, 1997; Busch and Hofmann, 2011). The World Commission on Environment and Development (1987) classified the chemical, oil & gas, utility, and pulp and paper industry as the most polluting or pollution-intense industries. Jo et al. (2014) find the weakest relationship between CEP and CFP in the banking industry, which they argue is the most eco-friendly sector. They argue that the CEP-CFP relationship could be stronger in more polluting-intense industries. Konar and Cohen (2001) find that the magnitude of the relationship between CEP and CFP varies across industries. They also claim that the relationship is more profound in the traditionally polluting industries. Hart and Ahuja (1996) argue that the most polluting industries did not show substantial reductions in their carbon emissions over the years, and ‘low hanging fruit’ still exists in these industries. The largest impacts on CFP accrue to ‘high polluters’ since they can still make plenty of low-cost improvements. In less-polluting industries, investments in CEP tend to become more expensive, as further improvements in CEP means raising capital and investments in technology. In summary, it appears that the pollution intensity of the industry in which the study is performed influences the results of the study. Therefore, the following hypothesis is tested:
H5: The industry pollution intensity moderates the relationship between CEP and CFP, so that
24
Figure 1: Systematic representation of the relationship between environmental regulation and improved profitability and competitiveness according to Porter’s hypothesis (Ambec & Barla, 2006)
25 to more innovation, strengthening the relationship with CFP. As such, innovation and corresponding improved productivity induced by ETSs may offer financial benefits to firms and can enhance their competitiveness compared to non-regulated competitors (Joltreau & Sommerfeld, 2018). An advantage of ETSs over other emission policy instruments is that it allows participating companies to tailor their compliance strategies in a way that reduces their costs (Kruger, 2005). It leaves companies with three alternative strategies: reducing GHG emissions to meet the requirements, buy emission rights, or reducing emission to a level below the legal requirements and sell the excess emission rights (Sandoff & Schaad, 2009). Since all strategies increase or decrease the associated costs of emissions, ETS can make the relationship between corporate GHG performance and CFP more positive (Czerny & Letmathe, 2017). In summary, ETSs may, similar to other environmental instruments, results in both product and process offsets and associated increases in CFP. However, ETSs also integrate the costs which are otherwise born by the public in the corporate GHG– CFP relationship. These costs can be reduced by investing in corporate GHG performance. As such, the relationship with CFP is expected to be more positive in regions with more stringent ETSs policies. Therefore, the following hypothesis is tested:
H6: The relationship between Corporate GHG performance and CFP is more positive for
firms that operate in countries with more stringent ETSs than for firms that operate in countries with less ETSs stringency.
3. Methodology
26 estimation techniques, and are therefore expected to contain some degree of error (Hunter, et al., 1982). Results from a meta-analysis may include a more precise estimate of the effect of a construct, than any individual study contributing to the pooled analysis (Tavakol, 2018). Typically, the data of a meta-analysis consists of estimates of some economic association, namely: effect sizes or empirical results. In meta-analysis, the weighted average of this economic association is calculated (Stanley & Doucouliagos, 2012). This enables the generation of objective and comprehensive summary effects and is a powerful tool for improving empirical findings (Endrikat, et al., 2014). Following other recent meta-analytical studies (e.g., Albertini, 2013; Endrikat et al., 2014; Busch and Hoffmann, 2018), my study follows a three-step methodology. First, the description of the systematic search for effect sizes between both indicators is presented. Secondly, a description of the included effect sizes is given, and the coding procedures are discussed. Thirdly, the calculation of mean correlations and the meta-analytical procedure are discussed. Finally, the nonparametric Mann-Whitney-Wilcoxon test, which is employed for testing whether subgroups are significantly different and the robustness tests are described.
3.1 Sample
27 GHG performance, climate change, GHG emissions, CO2 emissions, environmental management, environmental regulation, and carbon-pricing. The electronic search was conducted using EBSCO, ScienceDirect, JSTOR, Emerald, and Google Scholar. Existing meta-analyses (e.g., Dixon-Fowler et al., 2013; Endrikat et al., 2014; Busch and Lewandoski, 2018) also include papers based on a search in references of related empirical papers and working papers which were presented during conferences. These methods can, according to Hunter (1982), lead to systematic biases. I, therefore, do not employ these methods..
28
Table 4: Summary of study data and characteristics of the 34 included studies
CFP Measure CEP Measure
Paper Authors Period Region/country Industries Accounting-based Market-Based Indicator specification
Reporting scheme
1 Aggarwal and Dow (2012) 2008-2009 US Multiple ROA Tobin’s Q Relative Voluntary
2 Brouwers et al. (2018) 2005-2012 Europe Multiple ROA, ROE Tobin Q Relative Mandatory
3 Brzobahaty and Jansky (2010) 2004-2006 Czech Republic Multiple polluting Inpa, Inra, Inca Relative Mandatory
4 Busch and Hoffmann (2011) 2005-2007 Global Multiple ROA, ROE Tobin’s Q Relative Voluntary
5 Busch, Lehmann, Hoffmann
(2012)
2003-2009 Global 464 industries D/a, CF To Assets Total market risk,
Unsystematic risk, Systematic risk
Absolute Voluntary
6 Chakrabarty and Wang (2013) 2001-2009 US Manufacturing,
Utilities
SALES effectiveness, ROE, ROA Absolute Mandatory
7 Chapple, Clarkson and Gold
(2013)
2007-2009 Australia Multiple V, Earnings Relative Mandatory
8 Clarkson, Li, Pinnuck and
Richardson (2015)
2006-2009 Europe ROA Absolute Mandatory
9 Clarkson, Chapple and Gold
(2011)
2005-2006 Australia Multiple ROA Tobin’s q Absolute Mandatory
10 Czerny and Letmathe (2017) 2005-2012 Europe Multiple Profit margin, ROA, ROE, ROIC,
P/L, Profit per employee, EBIT margin
Absolute Mandatory
11 Dangelico and Pontrandolfo
(2015)
2011 Italy Multiple Market Performance Relative Mandatory
12 Delmas, Nairn‐Birch and Lim (2015)
2004-2008 US Multiple ROA, ROE Tobin’s Q Absolute Voluntary
13 Fujii, Iwata, Kaneko and
Managi (2012)
2006-2008 Japan Manufacturing ROA, CT, ROS Relative Voluntary
14 Gallego‐Alvarez, Garcia‐
Sanchez, and de Silva Vieira (2014)
2006-2009 Global ROA Relative Voluntary
15 Gallego-Álvarez, Segura, and Martínez-Ferrero (2015)
2006-2009 Global Multiple ROA, ROE Absolute Voluntary
16 Griffin, Lont and Sun (2017) 2006-2012 Global Multiple PRCC Relative Voluntary
17 Hatakeda, Kokubu, Kajiwara, Nishitani (2012)
2006-2008 Japan Manufacturing Profitability Tobin’s Q Relative Mandatory
18 Huo, Huang, Wu (2014) 2001-2006 Japan Chemical
Automobile, Electronic
Net income Absolute Voluntary
29
20 Huo, Huang, Wu (2014) 2001-2006 Japan Chemical
Automobile, Electronic
Net income Absolute Voluntary
21 Jung, Herbohn and Clarkson (2016)
2009-2013 Australia Multiple Cost of debt Relative Mandatory
20 Kim, An and Kim (2015) 2007-2011 South Korea Multiple COE, ROA Relative Mandatory
22 Lannelongue et al (2015) 2011 Spain Multiple ROA, ROE, Profits Absolute Voluntary
23 Lee and Min (2015) 2001-2010 Japan Manufacturing Tobin’s q Relative Voluntary
24 Luo and Tang (2014) 2011 Australia Multiple Direct exposure to tax Market return Relative Mandatory
25 Makridou et al (2019) 2006-2014 Europe Multiple EBITTA, Current ratio, Solvency
ratio
Relative Voluntary
26 Matsumura et al (2014) 2006-2008 US Multiple MKTE Relative Voluntary
27 Misani and Pogutz (2014) 2007-2013 Global Industrial ROS, ROA Tobin’s q Relative Voluntary
28 Nishitani and Kokubu (2012) 2006-2008 Japan Manufacturing Tobin’s q Relative Mandatory
29 Rokhmawati et al (2015) 2015 Indonesia Manufacturing ROA Relative Mandatory
30 Saka and Oshika (2014) 2012 Japan multiple MVE Absolute Voluntary
31 Trumpp and Guenther (2015) 2008-2012 Global Multiple ROA, TSR Relative Voluntary
32 Wang et al. (2014) 2010 Australia Multiple Sales Tobin’s q Absolute Voluntary
33 Yagi and Managi (2018) 2011 Japan Manufacturing Sales Absolute Voluntary
34 Yang, Shi, Meng, Zeng (2014) 1999-2010 China Multiple ROA Relative Voluntary
30
3.2 Coding
A coding procedure was developed to collect the data required for the analysis. The empirical findings or effect sizes of the included individual studies are the unit of analysis in meta-analytical methodology (Hedges & Olkin, 1985). Effect sizes are gathered from two types of statistics: Pearson moment correlations and partial-correlations. Pearson product-movement can usually be derived from the correlation table in the empirical studies. For studies that did not report correlation tables, the effect sizes are calculated from the reported t-statistics and degrees of freedom using the following formula:
𝑟 = √𝑡2/(𝑡2− 𝑑𝑓) (1)
For studies that do not report the t statistics, the t statistic is calculated backward from the standard errors, significance level, or p values. The t statistic, which can be used to calculate the effect sizes, is constructed backward by the following equation:
𝑡 = 𝑏1
𝑆𝐸 (2)
31 low compared to Dixon-Fowler et al. (2013) (292 effect sizes from 39 studies), Albertini (2013) (205 effect sizes from 52 studies), and Endrikat et al. (2014). (245 effect sizes from 149 studies). However, the number of included observations is similar to the meta-study of Busch and Lewandoski (2017) (68 effect sizes from 32 studies), which also only include studies that investigate the relationship between GHG emissions and CFP. They argue that research that focuses on GHG emission is still limited. Appendix A gives an overview of all included individual effect sizes and corresponding sample sizes.
Figure 2: Results and effect sizes measures
Note: figure 2 gives information on the included effect sizes in this study. A total of 49 positive; 23 negative effect sizes are included in the meta-study; in 3 cases, no effect relationship was observed. A total of 47 effect sizes are gathered using Pearson product-correlations, and 28 are based on partial correlation coefficients. The sample consists of 34 studies, 75 effect sizes, and a total of 172,117 observations.
32
Figure 3: CFP characteristics and sample study results
Note: Figure 3 gives information on CFP measures in the sample studies. A total of 47 effect sizes are measured using accounting-based indicators for CFP. From which 27 indicate a positive, 19 a negative, and one measure no relationship. A total of 28 observations are measured using market-based indicators for CFP from which 22 measure a positive, 4 a negative relationship, and 2 do not observe any relationship. The sample consists of 34 studies, 75 effect sizes, and 172,117 observations.
Figure 4 displays the characteristics of how corporate GHG performance construct differs between studies. Most studies use relative emissions in contrast to absolute emission as an indicator of CEP. Furthermore, the majority of observations are collected based on voluntary instead of mandatory reporting schemes.
Figure 4: GHG performance measurement characteristics of sample studies
Note: figure 4 gives information on how GHG performance is measured in the sample studies. A total of 13 papers use absolute indicators, and 21 use relative indicators to measure corporate GHG performance. Data is collected 14 times by mandatory and 20 times by voluntary reporting schemes. The sample consists of 34 studies, 75 effect sizes, and 172,117 observations.
33 studies that only included heavy polluting industries (chemical, oil & gas, utility, and pulp & paper) and effect sizes that come from studies that investigated multiple industries, I constructed two subgroups. This method allows me to determine whether the relationship between corporate GHG performance and CFP is stronger for firms operating in polluting-intense industries. Figure 5 shows the total number of individual firms in each sector in the sample.
Figure 5: Number of firms in different industries
Note: figure 5 reports the number of individual firms in different industries in the sample studies based on standard industrial classification. A total of 9820 firms are active in the manufacturing industry; 2210 individual firms are active in the service industry; 1858 firms are active in the transportation, communication, electric and gas and sanitary services industries; 1233 firms are active in the finance, insurance, and real estate industries; 1167 firms are active in the retail trade industry; 302 firms are active in the wholesale trade industry; 159 firms are active in the mining industry; 114 firms are active in the construction sector; 369 firms are active in the agriculture, forestry and fishing industries; from a total of 7357 firms, the industry is unknown. A total of 24,589 individual firm observations are included in the sample studies.
34 distinct policies per country per year that affect the renewable energy sector. Correspondently, other types of measures have also been proposed (e.g., single policy event measures, multiple types of surveys, and measures based on shadow prices). However, these measures are rather indirect and do not measure carbon constraints, which makes cross-country analysis of differences in ETS stringency limited (Botta & Kozluk, 2014). Multiple indices are developed to measure the performance climate policies of governments and enable country comparison (Bernauer & Bohmelt, 2013). Two indices that can be used to measure ETS stringency are the CCPI of Burck et al. (2005) and the C3-I developed by Bernauer and Böhmelt (2013). The CCPI was first released in 2005 and tracks efforts of countries to battle climate change. The database covers the effort of 58 countries between 2005 and 2019. The broader C3-I offers a dataset including 172 countries for the period 1996-2008. Both indices capture overall performance scores as well as performance in terms of political behavior and emissions (Bernauer & Bohmelt, 2013). Table 5 compares the stringency indices used in this study.
Table 5: the used ETS stringency indices
Note: table 5 presents an overview of the three used ETS stringency indices. It includes information on the number of included countries, the covered time period, how the policy component is assessed, the weights of the emission relative to policy components, and a definition of the stringency measure.
According to Bernauer and Bohmelt (2013), the methodology to arrive at the C3-I, and the CCPI indices are closely related. The CCPI evaluates the emission component based on trends and emission levels, and the C3-I also evaluates the trends and levels relative to income. The policy component is assessed by expert assessment in the CCPI and by observed behavior in the C3-I.
Climate Change Performance Index
Climate Change Cooperation Index Climate Action Tracker
Countries included 58 172 61
Time-period covered 2005-2019 1996-2008 2011-2019
Policy component Expert assessment Observed behavior Proposed future emission cuts Weighing of emissions
relative to policy
80%/20% 50%/50% 100%
Definition The CCPI tracks countries' efforts to combat climate change. The score is based on total emissions, renewable energy use, and climate policies (Burck, et al., 2005).
The C3-I captures the overall
performance as well as the performance in terms of political behavior and emissions; it allows for a global comparison of the climate policies of countries (Bernauer & Bohmelt, 2013).
36 ETS stringency study ranks). When studies did not provide information on the included number of observations from individual countries, they were excluded from the ranking. The method allows dividing the studies into four groups (from stringent to less stringent ETSs), with the use of two indexes, even though the scales and methodologies of both indices are not exactly the same. The method follows the following reasoning: when a country is ranked higher compared to other countries in a given year, a study performed in the country in that year is expected to observe higher carbon constraints than a study performed in a lower scoring country in that year (Botta & Kozluk, 2014). However, both indices measure historical output and emission trends in a wider range of environmental policies and do not necessarily measure the future carbon constraints faced by companies (Bernauer & Bohmelt, 2013).
37
Table 6: studies ranked on their environmental policies
Study number Subgroups based on C3I & CPI CAT
Rank Group Group
4 Busch and Hoffmann (2011) - - -
31 Trumpp and Guenther (2017) - - -
5 Busch, Lehmann, Hoffmann (2012) - - -
14 Gallego‐Alvarez, Garcia-Sanchez, de Silva Viera. (2014) - - -
12 Delmas, Nairn-Birch and Lim (2015) - - -
27 Misani and Pogutz (2014) - - -
19 Iwata and Okada, (2011) 28 L Sufficient
1 Aggarwal and Dow (2012) 27 L Insufficient
6 Chakrabarty and Wang (2013) 26 L Insufficient
16 Griffin, Lont and Sun (2017) 25 L Insufficient
15 Gallego-Álvarez, Segura and Martinez-Ferrero. (2015) 24 L Insufficient
20 Jung, Herbohn and Clarkson (2016) 23 L Insufficient
34 Yang, Zeng, Shi, Meng, (2014) 22 L Insufficient
30 Saka and Oshika (2014) 21 M-L Sufficient
33 Yagi and Managi (2018) 20 M-L Sufficient
29 Rokmawati, Sathye and Sath (2015) 19 M-L Sufficient
17 Hatakeda, Kobubu, Kaijwara, Nishitani (2012) 18 M-L Moderate
21 Kim, An and Kim (2015) 17 M-L Sufficient
13 Fujii, Iwata, Kaneko and Managi (2012) 16 M-L Sufficient
3 Brzobahaty and Jansky (2010) 15 M-L Moderate
22 Lannelongue et al (2015) 14 M Moderate
24 Luo and Tang (2014) 13 M Insufficient
23 Lee, Min and Yook (2015) 12 M Sufficient
28 Nishitani and Kokubu (2012) 11 M Sufficient
26 Matsumura et al (2014 10 M Sufficient
2 Brouwers et al. (2018) 9 M Insufficient
10 Czerny and Letmathe (2017) 8 M Moderate
18 Huo, Huang, Wu (2010) 7 H Sufficient
8 Clarkson, Li, Pinnuck and Richardson (2015) 6 H Insufficient
7 Chapple, Clarkson and Gold (2013) 5 H Sufficient
9 Clarkson, Chapple and Gold (2011) 4 H Insufficient
11 Dangelico and Pontrandolfo (2015) 3 H Sufficient
25 Makridou et al (2019) 2 H Medium
32 Wang et al. (2014) 1 H Insufficient
38
3.3 Meta-analytical procedures
Previous meta-analytical reviews on the CEP-CFP relationship employed both the aggregation techniques of Hunter and Smith (1980) (hereafter: HS) (e.g., Albertini., 2013; Orlitzky et al., 2003) and the Hedges-Olkin-type meta-analysis (hereafter: HOMA) (e.g., Endrikat et al., 2014; Busch and Lewandoski., 2014). The choice for a specific method is expected to influence the results (Field, 2003). Johnson et al., (1995) give a comparison of meta-analytical techniques, and they argue that the HS method does not attempt to correct biases in the effect sizes before deriving mean effect sizes. My meta-study, therefore, uses the HOMA method to aggregate the individual findings and to correct for individual study artifacts (e.g., overestimation of the population effect size in small sample studies). Additionally, I used the HS method as a test for the robustness of the results. Field (2003) argues that when employing a relatively small sample, using the fixed-effect model over the random-effects models can bias results. Since the sample of this study is relatively small, the random-effect model is used to calculate the summary effects.
Before calculating the summary effects, the effect sizes are first transformed to a standard normal metric by Fisher’s z transformation to address the potential distribution skewness in meta-analyses (Hedges & Olkin, 1985). The Fisher’s r to z transformation was performed with the following formula:
𝑍𝑖 =1
2∗ 𝐿𝑜𝑔𝑒 1+𝑟𝑖 1−𝑟𝑖 (3)
Where Z is the transformed partial correlation. In line with Hedges and Olkin (1985), the weight assigned to the individual effect sizes is a variance component that consists of both the between-study and the within-between-study variance. The within-between-study variance is calculated as followed:
39 The between-study variance is calculated by
τ2=𝑄−(𝑘−1) 𝐶 (5) Where 𝑄 = ∑ 𝑤𝑖 𝑍𝑖− (∑𝐾𝑖=1𝑤𝑖 𝑍𝑖)2 ∑𝐾𝑖=1𝑤𝑖 𝐾 𝑖=1 (6) 𝑤𝑖= 1 𝑉𝑤𝑖𝑡ℎ𝑖𝑛 (7) 𝐶 = ∑𝐾𝑖=1𝑊𝑖 − ∑𝐾𝑖=1𝑤𝑖 2 ∑𝐾𝑖=1𝑤𝑖 (8)
The random-effect aggregated effect size is calculated using the sum of the between-study and the within-study variance (Hedges & Olkin, 1985), which is calculated as followed:
𝑉𝑖= 𝑉𝑤𝑖𝑡ℎ𝑖𝑛+ τ2 (9)
Relatively larger samples produce more accurate estimates than relatively smaller samples (Stanley & Doucouliagos, 2012). In line with Hedges and Olkin (1985) weights are assigned to each effect size based on the inverse value of the sum of the between and within-study variance by the following equation:
𝑊𝑖 = 1 𝑉𝑖 (10)
The mean effect size and the standard error of the mean effect size are calculated in line with Hedges and Oskin (1985) using the following equations
𝑧̅ = 𝑟
40
𝑆𝐸(𝑧̅𝑟) = √ 1
∑𝐾𝑖=1𝑊𝑖 (12)
The confidence interval for the aggregated effect size is calculated by
𝐶𝐼𝑈𝑝𝑝𝑒𝑟= 𝑧̅ + 1.96 ∗ 𝑆𝐸(𝑧̅𝑟)(13)
𝐶𝐼𝐿𝑜𝑤𝑒𝑟= 𝑧̅ − 1.96 ∗ 𝑆𝐸(𝑧̅𝑟) (14)
In line with Hedges and Olkin (1995) all values are transformed back to correlation units by
𝑟𝑖= 𝑒2𝑧𝑖−1 𝑒2𝑧𝑖+1 (15)
41
Figure 6: Forest plot of the effect sizes and confidence intervals
Note: figure 6 is a forest plot, which is a graphical display of all included effect sizes. The numbers 1 until 75 correspond to the individual effect sizes and confidence intervals that are gathered from the sample studies. Number 76 corresponds to the aggregated effect size and confidence intervals. A total number of 75 effect sizes from 34 studies are included in the sample.
42 In line with both Hunter and Smith (1982) and Hedges & Olkin (1985), I test for the publication bias. Following existing meta-analytical studies, the failsafe-N of Rosenthal (1979) is calculated. The failsafe-N test calculates the number of insignificant studies that have to be included in the sample to make the aggregated effect size statistically insignificant (Stanley & Doucouliagos, 2012). The publication bias is based on the argument that empirical studies with significant results have a higher probability than studies with insignificant results to be published. In line with Hunter and Smith (1982) and Hedges and Oskin (1985) the number of additional scores that have to be included to make the aggregated effect size insignificant at the 5% level is calculated as followed:
𝑘 ∗ [𝑍𝑠
𝑍𝑎]
2
− 𝑘 (16)
Where 𝑍𝑎 is the critical upper-tail value of the normal distribution, and 𝑍𝑠 is calculated as
followed:
𝑍𝑠 =(∑𝐾𝑖=1 𝑍𝑠𝑐𝑜𝑟𝑒𝑠)
√𝑘 (17)
43 Nevertheless, since the sample size of this study is relatively small, the Q statistic may provide a misleading measure of heterogeneity, it should be carefully interpreted (Tavakol, 2018) (Hedges & Olkin, 1985). Finally, in order to test whether subgroups differ significantly from one another since the effect-sizes of the subgroups are unpaired, the non-parametric Mann-Whitney-Wilcoxon test is performed. The test does not assume normally distributed or paired data, which is the case in my sample (Fay & Proschan, 2010). The effect-sizes in the subgroups are not weighed as differences in sample size would make the differences significant by definition. The results of the Mann-Whitney-Wilcoxon test can be found in appendix B.
4. Results
44
Table 7: results meta-analysis
Note: Table 7 summarizes the results of the meta-analysis based on the Hedges and Olkin (1985) method. It first describes the overall aggregated relationship between corporate GHG performance and CFP. Next, the results of the different subgroup analyses are presented. The aggregated effect sizes for the subgroups for different reporting types and the indicator specification of the corporate GHG performance construct are given; the effect sizes for the market and accounting based CFP indicator specification and the industry carbon intensity are reported; the ETS stringency hypothesis using two different methods is reported. For ETS stringency based on the C3-I and CCPI, the ‘high’ group consists seven studies conducted in the most stringent environments, the following seven studies from subsequently lower ETS stringency environments form the group ‘medium,’ the seven following studies form the group ‘medium-low and the studies conducted in the lowest ETS stringent regions studies form the group low. The CAT ETS stringency measure has resulted in 3 groups, studies which for the group ‘sufficient’ are from regions with sufficient policies for reaching the UN climate goals. Moderate forms the group of studies which are performed in countries with moderate policies, and inadequate forms the groups of studies which are performed in inadequate performing countries. For the group ‘global / no data available,’ the study was conducted globally, or no information about the studied country was. k = number of effect sizes; N= total sample size; r= aggregated effect size 95% CI = 95% confidence intervals for the aggregated effect sizes; p= probability Q= Q statistic Pq= probability of Q statistic pQ-bet= p-value of between-group heterogeneity
The results of the subgroup analysis are presented in table 7. Subsequently, the results of the conducted Mann-Whitney-Wilcoxon test can be found in Appendix B, table A1 and A2. The results show that the empirical findings of the CEP-CFP relationship are influenced by the type of measurement of corporate GHG performance. When emissions are measured by voluntary reporting types, it is positive and significant related to CFP at the 1% level (r=0.08, p=0.00). When emissions are measured using mandatory reporting types, the relationship was
k N r 95% CI Z p Q pQ PQbet Overall 75 172,117 0.0717 0.04-0.10 4.52 0.000 5120.69 0.000 Moderators Reporting type Mandatory 29 124,188 0.05 0.001-0.10 2.17 0.030 269.23 Voluntary 46 47,929 0.08 0.040-0.117 4.00 0.000 4743.28 0.396 CEP Indicator specification Absolute 31 65,146 0.10 0.06-0.15 1.39 0.00 1547.59 Relative 44 106,971 0.05 0.01-0.09 2.32 0.01 3914.88 0.367 CFP measures Market-based 28 74,635 0.12 0.07-0.18 4.53 0.000 1114.96 Accounting based 47 97,482 0.04 0.00-0.08 2.25 0.012 4066.45 0.412 Pollution intensity of industry Only pollution-intensive industries 25 27,877 0.06 0.008-0.102 2.28 0.0228 762.06 Mixed industries 23 81,616 0.14 0.079-0.197 4.61 0.000 3315.84 0.026 No data on industries 27 62,624 0.04 -0.02-0.08 1.85 0.065 685.73
ETS stringency C3I CPI High 12 65,288 0.18 0.08-0.29 4.33 0.000 3136.47 Medium 16 16,464 0.06 -0.01-0.111 2.32 0.020 99.65 Medium-low 15 23,463 0.05 -0.04-0.15 1.28 0.199 678.26 Low 15 8,769 0.06 0.02-0.11 2.99 0.003 39.05 0.061 Global / no data on country 17 58,135 0.04 -0.015-0.097 1.42 0.156 590.47
ETS stringency CAT
46 positively related to CFP could indicate that ETSs related costs are an important cost factor for businesses, therefore making the relationship more positive.
Although for both accounting-based and market-based indicators the relationship between CEP and CFP is positive, CEP is more positively related to CFP when market-based measures are used (r=0.12, p=0.00), instead of accounting-based measures (r=0.04, p=0.012) (pQb= 0.412) The Mann-Whitney-Wilcoxon test presented in Appendix B table A1, indicates that the differences between both subgroups are highly significant at the 1% level (p=0.000). Hence, hypothesis 4A (Corporate GHG performance is more positively related to prior market-based than to prior accounting-market-based CFP) is accepted, which is not the case for hypothesis 4B (Corporate GHG performance is more positively related to prior accounting-based than to prior market-based CFP). The positive association between accounting-based CFP and corporate GHG performance indicates that investments in corporate GHG performance can be converted into future accounting-based CFP, in line with the eco-efficiency concept (Kuo, et al., 2010). However, the significantly stronger relationship between market-based indicators of CFP and corporate GHG performance indicates that investors include the firms’ GHG emissions as a negative component of equity value and incorporate the value of investments of a firm in its corporate GHG performance in their market valuation (Griffin, et al., 2017). The market could value improvements in corporate GHG performance because they can reduce regulatory business risks associated with emissions and can attract the growing group of environmentally aware consumers in the future (Griffin, et al., 2017).
47 table A1, differences between the subgroups are significant at the 5% level (p=0.031). H5 (industry carbon intensity moderates the relationship between corporate GHG performance and CFP, the corporate GHG performance-CFP relationship is stronger in more polluting industries) is therefore not accepted. The Corporate GHG performance - CFP relationship seems to be weaker for studies conducted in pollution-intense industries than for studies conducted in multiple industries. This finding indicates that firms’ in more polluting industries do not have more financial benefits from emission reductions than other industries. A possible explanation for this finding could be that over the years forced by tighter regulation pollution-intense industries have already picked the ‘low hanging fruits’ (Delmas, et al., 2015). As such, the financial benefits of investments in pollution reduction of firms in pollution intense industries could have become smaller compared to they used to be.
48 Moreover, when subgroups are developed based on Climate Action Tracker, the studies conducted in countries from which the policy is evaluated as sufficient show the most positive relationship between corporate GHG performance and CFP (r=0.012 p=0.000) compared to medium scoring (r=0.06 p=0.063) and inadequate scoring (r=0.05 p=0.030) (pQb=0.164) subgroups. Nonetheless, the Mann-Whitney-Wilcoxon test in Appendix B table A2, shows that the subgroups are not significantly different. Based on the above results of the two subgroup analyses, H6 (the corporate GHG performance – CFP relationship is stronger in more stringent ETS regions) cannot be accepted. Results indicate that the relationship between corporate GHG performance and CFP is significant and positive for all subgroups, but is only significantly more positive in the most stringent ETS environments. A possible explanation could be the fact that initial phases of ETSs are characterized by low stringency, bureaucracy, and little influence on innovation (Czerny & Letmathe, 2017). These early phases are known for the free allocation of emission rights, low emission prices, and many industries excluded (relatively medium ETS stringency) (Abrell, et al., 2011). The subgroups, which consist of studies conducted in medium and medium-low stringency regions, could consist of studies performed in relatively early phases of ETS. This is, however, not tested in my study.
4.2 robustness tests
49 partial-correlations (r=0.06, p=0.004) (pQb=0.247). However, both measures yield results in line with the total sample. Lastly, I tested for the presence of the publication bias, in line with Rosenthal (1991). The failsafe-N is calculated, which indicated moderate existence of the publication bias: 245 additional null-effect studies are required to make the summary effect size insignificant (Hunter, et al., 1982). This finding can be explained by the fact that this study only includes studies that investigate the relationship between GHG emissions and CFP, and the number of studies on the topic is growing but still limited (Chapple, et al., 2011).
5. Conclusion
This study examined one specific environmental performance outcome (GHG emissions) and its relationship with CFP over the period 1997-2019. The goal of this paper was to demonstrate the existence of the relationship between GHG performance and CFP and to provide insights into the methodological and business environmental factors that influence the empirical outcomes. Including 75 effect sizes from 34 studies, the main conclusion of this paper is that a significant positive relationship between corporate GHG performance and CFP exists. This conclusion is in line with previous meta-studies on the CEP-CFP relationship (e.g., Albertini, 2013; Dixon Fowler et al., 2013; Endrikat, 2014), and the study of Busch and Lewandoski (2018) on the relationship between GHG emissions and CFP.
50 studies, which could influence the results. Furthermore, this study finds that the choice of CFP construct leads to differences in the Corporate GHG performance-CFP relationship. In line with the findings of the meta-studies of Dixon-Fowler et al. (2013) and Busch and Lewandoski (2018), corporate GHG performance is more positively related to CFP when CFP is measured by market-based indicators, instead of accounting-based indicators. The study of Albertini (2013) and Endrikat (2014) found contradictory evidence but did not differentiate between different measures of CEP. This could explain the differences in outcomes.
Furthermore, I reject the notion that the corporate GHG performance - CFP relationship is stronger in pollution-intensive industries. As such, I cannot conclude that the industry carbon intensity is moderating the CEP-CFP relationship. This is in line with the findings of Albertini (2013). Finally, based on two subgroup analyses, the findings of this study do not show that the corporate GHG performance and CFP relationship is more positive in countries with more stringent ETS. Although the corporate GHG performance – CFP relationship is significantly more positive in the countries with the most stringent ETSs, there are little differences between countries with low and medium ETS stringency.
51 employing a different method. Indifferent of the business environment, my study shows that it pays to be green. Finally, the findings of this paper can be used to develop future empirical and meta-studies on the CEP– CFP relationship.
