Understanding the impact of pollutants: the effect of air, water and waste pollutants on international firm
performance
Aileen Booijink
MSc Business and Economics Financial Management Track Department of Business Studies
Uppsala University 31 January 2017 Abstract
This study investigates how different type of pollutants influence international firm performance. The dataset covers 1804 firms from 43 countries and 20 industries. Five
different types of pollutants are used as well as five different financial performance measures.
The paper uses industry-specific fixed effects as estimation method and finds that the type of pollutant influences the relationship between environmental performance and firm
performance. In general, the relationship between pollutants and firm performance is
negative, however water pollutants are an exception. Moreover, pollutants appear to be more negatively correlated with accounting performance measures than with market performance measures. Additionally, two country level factors show that the country of origin matters.
TABLE OF CONTENT
1 Introduction 3
2 Literature and hypotheses development 5
3 Data and methods 13
4 Results 12
4.1 Descriptive statistics 17
4.2 Full sample 21
4.3 Country specific 26
4.4 Industry specific 31
5 Conclusion 24
Appendices 36
References 47
1.INTRODUCTION
Corporate environmental performance is expected to directly or indirectly influence firm performance. Over the years, environmental performance has become more important to firms’ stakeholders (Iwata and Okada, 2011), as they take this performance measure into account in their evaluation of the firm. Therefore, firms have an incentive to reduce their environmental impact. However, different stakeholders may have differentiated preferences when assessing environmental issues, and thus may different environmental performance measures lead to mixed effects on firm performance measures.
In contrast to financial performance measures, environmental performance measures are not standardized. There is a risk of mismeasurement of environmental performance if only some pollutants are taken into account and others are ignored (Horvathova, 2012). All environmental issues and type of pollutants have different characteristics, such as the severity of damage, scope of pollution, the time before damage appears, and existence of (international) regulations and protocols. Therefore, different stakeholders may assign different degrees of importance to different environmental issues and pollutants (Iwata and Okoda, 2011). It is found that the type of environmental performance measure affects the relationship between environmental and financial performance (Horvathova, 2010).
Consequently, it is important to include different types of pollutants when examining their impact on firm performance.
Porter (1991) argues that pollution signals economic inefficiency, and therefore firms have incentives to improve their environmental performance, as it may be beneficial for them. Economic literature, however, has treated environmental issues as inconsistencies between private and social benefits that have to be solved through government intervention.
Yet, it is possible that the market solves environmental issues without the intervention of governments, as firms may have incentives to reduce their emissions (Iwata and Okada, 2011). Investigating the relationship between environmental performance and financial performance may thus not only have important firm implications, but also policy implications.
Several studies investigate the relation between pollution and firm performance, examples are Hart and Ahuja (1996), Konar and Cohen, 2001; Wagner, 2001; Iwata and Okada, 2011; Nishitani and Kokubu (2011); Horvathova (2012); Wang et al. (2013);
Scholtens and van der Groot (2014); Lee et al. (2015). However, in my opinion these studies do not sufficiently address the difference between the various types of pollutants.
The key question addressed in this paper is how different types of pollutants affect individual firm performance. This is done by taking into account five different pollutants:
carbon dioxide, nitrogen oxide, sulphur oxide, water pollutants and waste, both hazardous and non-hazardous waste. The effect of each of these pollutants on different firm performance measures is investigated in order to avoid biases. These firm performance measures are return on assets, return on equity, stock returns, beta and Altman’s Z-score. This study focuses on the period 2007-2015. Data availability of both pollutants and financial performance measures is the main motivation for this period. The main hypothesis to be tested is that pollution negatively affects firm performance. Moreover, I will analyse the differences on a country and industry level. To my knowledge, there is no study that includes these five different pollutants as environmental performance measures to investigate their individual effects on financial performance.
This paper consists of three different aspects: international, financial and managerial.
The international aspect comes to light in different ways. First, the used sample is highly international as it consists of 43 countries. Moreover, the effect of pollutants on firm performance is tested for two country-level characteristics: income level and carbon dioxide emission level per capita. These characteristics allow a comparison on the country-level between pollutants and firm performance. The financial aspect is captured in firm performance; five different financial measures are used to measure firm performance.
Implications drawn from the financial and international aspects are the basis for the managerial aspect.
The main finding of this paper is that the type of pollutant affects the relationship between the level of emissions and firm performance. Moreover, the strength of the impact is different for distinctive performance measures. Additionally, I found that the income level and carbon dioxide level of a country and the “dirtiness” of an industry have an influence on this relationship for some pollutants.
The remainder of the paper is organized as follows. The second section reviews the main literature and develops the hypotheses. The third section details the data and
methodology. The fourth section provides the results, while the fifth section concludes and discusses limitations of this study.
2.LITERATURE AND HYPOTHESES DEVELOPMENT
Many studies have attempted to study the relationship between corporate social performance (CSP) and corporate financial performance (CFP). A positive relationship between CSP and CFP was found by three meta-studies of Orlitzky et al. (2003), Wu (2006) and Margolis et al.
(2009). One of the components of the broad scope of CSP is environmental performance, and pollutant emissions are only one type of measure of environmental performance. Previous studies that attempt to relate environmental and financial performance over time have often led to conflicting results (Konar and Cohen, 2001; Wagner, 2001; Iwata and Okada, 2011).
According to a meta-analysis of Horvathova (2010), 55% of studies find a positive relation between environmental performance and financial performance, 30% of studies find no effect, and 15% of studies find a negative effect.
There are several reasons for these inconclusive findings. First, early studies are problematic because of the use of small samples and the absence of objective environmental performance measures (Konar and Cohen, 2001; Wagner et al., 2001). Second, moderating factors such as the size of the firms, growth and country were not taken into account in early studies (Wagner et al., 2001). Whereas more recent studies have dealt with the problems of sample size and moderating factors, there are some other remaining problems. Which are, third, the inconclusive findings may be affected by the fact that higher firm performance is reached not only through environmental efficiency, but also through efficiency in other production processes. (Filbeck and Gorman; 2004). Fourth, papers studying the relationship between environmental and financial performance use different environmental and financial indicators. Environmental indicators may be capturing a single specific dimension, like carbon dioxide emissions, or a broad category, such as an environmental rating. Financial indicators can be distinguished between e.g. accounting measures, that are backward looking and market measures that are more forward looking. Fifth, there are methodological problems as the research designs and their quality widely differ among the studies. (Ullman, 1985;
Olsthoorn et al., 2001; Derwall et al., 2005; Horvathova, 2012; Gonenc and Scholtens, 2017).
There are three general approaches that are regularly used to study the association between environmental performance and financial performance, namely portfolio analyses, event studies and regression analyses (Horvathova, 2012). Portfolio analysis compares returns on portfolios including firms with high environmental performance with portfolios including firms without a scale of environmental performance. Event studies analyses the impact of an event related to environmental performance on the firm’s stock returns.
Regression analysis studies the relationship between a firm’s environmental performance, its characteristics and its financial performance (Horvathova, 2012). Literature directly considering the association between pollutants as carbon dioxide, nitrogen and sulphur oxide, water pollutants and waste, and firm performance through regression analysis is the most relevant for the purpose of this study.
Konar and Cohen (2001), Nishitani and Kokubu (2011) and Wang et al., (2013) investigated the relationship between environmental performance and firm performance by using only one firm performance measure; Tobin’s q. Whereas Konar and Cohen used the aggregate pounds of emitted toxic chemicals, Wang et al., measured the greenhouse gas emissions, and Nishitani and Kokubu used the firm’s carbon dioxide productivity. Both Konar and Cohen and Nishitani and Kokubu found that a reduction in emissions could enhance firm value. However, Wang et al., found that higher greenhouse gas emissions correlate with a stronger Tobin’s q.
Konar and Cohen studied the influence of the aggregate pounds of toxic chemicals emitted per dollar revenue and the number of environmental lawsuits against the firm on Tobin’s q for 321 manufacturing firms in the United States. Using regression analysis, they found that poor environmental performance is negatively related to Tobin’s q. More specifically, both the emission of toxic chemicals and the number of lawsuits have a statistically negative effect, however only the emission of toxic chemicals has also an economically significant effect on Tobin’s q. The authors suggest that firms with better environmental reputations and performance have higher intangible assets. Firms go beyond complying with environmental regulations to obtain a good environmental reputation, since they are rewarded for these actions in the marketplace (Konar and Cohen, 2001). Nishitani and Kokubu (2011) used data from 641 manufacturing firms in Japan over the years 2006- 2008, to come to a similar conclusion as Konar and Cohen (2001). They found that a reduction of greenhouse gas emissions leads to a higher Tobin’s q. The authors argue that stockholders and investors regard a reduction of greenhouse gas emissions as an intangible value, and that this might be because the reduction is expected to lead to higher profitability and lower environmental liabilities (Nishitani & Kokubu, 2011). Wang et al. (2013) focused on greenhouse gas emission data of 69 Australian firms in 2010. They found that higher greenhouse gas emissions correlate with a stronger Tobin’s q, using linear regression models.
The authors suggest that this different finding may be explained by the unique country structure of Australia. Emission intensive industries (e.g. mining and manufacturing) lie at the heart of Australia’s economy (Wang et al., 2013). Care should be taken when trying to
generalize these studies’ results, as they all focus on only one country, the manufacturing industry is overrepresented and the study period is rather short.
Studies which included more firm performance measures include for example Hart and Ahuja (1996), Wagner et al., (2002); Alvarez (2012), Horvathova, (2012), and Lee et al., (2015). Hart and Ahuja (1996) found that emission reductions have a positive effect on firm performance. They studied 127 manufacturing, mining or production firms from the S&P 500 in 1988 and 1989. Emission reductions were measured by computing the percentage change of emissions for each firm. Return on sales and return on assets were used to measure operating performance, and return on equity was used to measure financial performance.
Conducting multiple regression analysis, Hart and Ahuja found that emission reductions appear to have a positive impact on firm performance. They report a difference between operating and financial performance, as operating performance is positively affected in the following year while financial performance is only benefited after two years. Similarly, Horvathova (2012) investigated the effect of an index of different pollutants on the return on assets and return on equity of 136 firms from the Czech Republic, with the use of multiple regression analysis. The index was created from emission data from the European Pollutant Release and Transfer Register. This database contains emission data on 93 pollutants related to air, water and waste. The study does not explicitly mention which exact pollutants are included in the study. Horvathova (2012) found that higher emissions increase both return on assets and return on equity in the following year, but decrease performance after two years.
However, as an index is be used, nothing can be implied about the impacts of individual pollutants on firm performance.
Whereas the preceding studies all focused on one country, Alvarez (2012) used emission data of firms from 21 countries. In this study, the effect of the amount of carbon dioxide emissions on return on assets and return on equity was investigated by using multiple regression analysis. Only a significant negative relationship between carbon dioxide emissions and return on assets was found. Alvarez (2012) seeks to explain this result by suggesting that it takes time between the first efforts a firm makes to reduce emissions and making an actual profit because of the reductions. However, the study lacks a clear suggestions or explanation of why this difference between return on assets and return on equity was found.
A drawback of the studies discussed so far is that these studies only included accounting measures for financial performance. Lee et al. (2015), used a market evaluation measure (Tobin’s q) and an accounting measure (return on assets), to evaluate the effect of
carbon emissions on firm performance. This study investigated the impact of carbon emissions on 362 Japanese manufacturing firms, using fixed effects. Carbon dioxide emissions were scaled by a firm’s value of assets. They found that a firm is penalized for negative environmental performance by the market and that carbon dioxide emissions decrease firm value. Moreover, they found that the market penalizes poor environmental performance more consistently than that the market rewards good environmental performance.
In contrast, Wagner et al. (2002) aggregated sulphur dioxide, nitrogen oxide and chemical oxygen demand in an index and found a negative relationship between financial performance and environmental performance. The authors tested the ‘traditionalist’ view against the ‘revisionist’ view using three-stage least squares. The traditionalist view suggests a uniformly negative relationship between environmental and financial performance.
According to this view, decreasing marginal benefits are the outcome of pollution abatement.
On the other hand, the revisionist view argues that pollution abatement may lead to a competitive advantage in the long run, as it may lead to innovations that offset the costs.
Financial performance was measured in terms of return on equity, return on sales and return on capital employed. The authors found evidence to support the traditionalist view for the impact of the pollutant index on return on capital employed. However, they used a relatively small sample size (n=70 and n=80) and only studied the paper manufacturing industry.
Wagner et al. (2002) suggest that these findings may be very specific to the paper industry, and that an analysis of more industries is needed. Although Wagner et al. (2002) used an adjustment factor to adjust the index for the individual contribution of the different pollutants, the individual impact of each pollutant on firm performance is not studied when using an index.
Another study reporting a negative correlation between environmental performance and firm performance is Sarkis and Cordeiro (2001). They studied 482 firms from the United States in 1992. The United States Environmental Protection Agency’s toxic releases inventory was used to calculate environmental performance, which, in this study, consisted of total emission releases and waste generated. Return on sales was used to measure short-term firm performance. They found that return on sales is negatively affected by improved environmental performance in the short-run (1 year). Sarkis and Cordeiro (2001) suggest that this result is found because of the higher organizational costs that arise because of pollution prevention approaches in the short-term.
However, the aforementioned studies only included air pollutants as proxy for environmental performance, such as carbon dioxide, sulphur oxide, nitrogen dioxide, etc. It is important to also include other types of pollutants as all environmental issues and type of pollutants have different characteristics. The dangerousness of pollutant emissions varies among each pollutant. Iwata and Okada (2011) and Horvathova (2012) suggest that the relationship between pollutants and firm performance is influenced by the dangerousness of the pollutant; the more dangerous pollutants are for the environment, the stronger the negative impact on firm performance. Few of the previous studies have included other environmental performance measures such as water pollutants and waste. One of these studies is Cormier and Magnan (1997), who investigated the correlation between water pollutants and firm performance of 28 Canadian firms in 1986-1991. They found by using pooled fixed effects OLS regression, a negative relationship between water pollutants and firm performance, implying that a higher amount of pollution leads to lower stock market valuation of the individual firm. Cormier and Magnan (1997) argue that pollution creates implicit environmental liabilities. Investors will subtract these environmental liabilities from a firm’s stock market valuation, therefore are higher pollution levels resulting in lower stock market valuations. Al-Tuwaijri et al. (2004) use waste as environmental performance measure. More specifically, they use the ratio of toxic waste recycled to total toxic waste generated. Industry-adjusted annual returns are used to measure a firm’s financial performance. Three-stage least squares is used to examine the association between the toxic waste ratio and financial performance of 198 firms in 1994. They found that good environmental performance is associated with higher firm performance. Al-Tuwaijri et al.
(2004) argue that environmental and financial performance are related to management quality. According to them, a good manager acts in the long-term interest of the firm, accepts the social responsibility of the firm and therefore adopts a strategy to control the firm’s pollution levels.
Iwata and Okada (2011) also recognize the importance of capturing the different characteristics of environmental issues together in one study, as they use greenhouse gas emissions and waste as environmental performance measures. An industry-specific fixed effects model is employed to estimate the relationship between greenhouse gas emissions and waste of 268 Japanese manufacturing firms in the period 2004-2008. They found that financial performance measures responded differently depending on the environmental issue.
Financial performance was not affected by waste, while greenhouse gas reductions improved
several financial performance measures including, return on assets, return on investment, return on invested capital and Tobin’s q-1.
Although the discussed literature assumes that direction of the causality is from environmental performance to financial performance, I recognize that the causality can also run from financial to environmental performance. For example, Scholtens (2008) found some preliminary evidence that direction of the causality predominantly runs from financial to environmental performance. Two different techniques are used, OLS with distributed lags and Granger causation. Nevertheless precedence is not identical to causality, much more often financial performance precedes social performance than the other way around (Scholtens, 2008). However, various non-financial stakeholders relation to corporate social responsibility were left out in this evaluation, and some critical variables affecting financial and social performance are missing in this analysis. Stakeholder theory and the trade-off view assume that the relationship runs from environmental performance to financial performance. Whereas stakeholder theory assumes a positive relationship, the trade-off theory assumes a negative relationship between environmental and financial performance.
Yet, the causality between environmental performance and financial performance should be further investigated.
Because of the findings relating to the stakeholder theory of relative recent studies discussed above, I expect a negative relationship between the level of pollutant emissions and firm performance. The results of studies from Nishitani and Kokubu (2011); Iwata and Okada (2011); Horvathova (2012); and Lee et al., (2015) provide reasons to believe that the relationship is negative. Therefore, I come to the following hypothesis.
H1a: Pollutants have a negative effect on firm performance
This hypothesis is rather general, as I expect a negative relationship between each pollutant and firm performance. Orlitzky et al. (2003) and Wu (2006) suggest that corporate social performance is more highly correlated with accounting-based firm performance measures, in comparison to market-based measures. Therefore, I expect that the impact of the pollutants is more severe for return on assets, return on equity and Altman’s Z-score than for the annual market returns and beta. Moreover, the order of magnitude may vary as the distinctive pollutants have different impacts on the environment. As assessing the exact impact of each pollutant on the environment is not within the scope of this study, I give only a short description of the impact of the different pollutants. Carbon dioxide is one of the main
greenhouse gases, and emissions of carbon dioxide contribute to the global warming potential. Heat is trapped in the lower atmosphere by greenhouse gases like carbon dioxide.
The ability of a greenhouse gas to absorb infrared radiation and its concentrations in the atmosphere, determine the radiative forcing, the degree of warming the gas transmits (Tucker, 1995). However the radiative forcing of, for example, methane is stronger than of carbon dioxide, because carbon dioxide is released in these high quantities and persists for about 100 years in the atmosphere it captures the majority of responsibility for the global warming potential (Tucker, 1995). Nitrogen oxide and sulphur oxide are acidic substances that cause air pollution by forming acids. These acidic substances can damage the environment directly and indirectly. Directly through damaging plants, materials and buildings, and indirectly through acidifying the soil. Water pollutants also have a direct impact on the environment as they can pollute freshwater, lowering the freshwater quality (Chapman, 1996). Different types of waste produced can have different impacts on the environment. Organic waste may rot, but it can also generate for example methane gases.
Synthetic waste is problematic as it can produce toxic substances, for example through burning the synthetic waste.
Firm emissions have attracted increasing attention from firm’s stakeholders for several reasons. First, stakeholders become more aware of climate change and environmental issues, and their negative consequences. Stakeholders are becoming more concerned about a firm’s emission levels, as industrial processes are largely been held accountable for climate change. Moreover, emissions have the possibility to have an impact on every company, in every country, and in every sector (Lee et al., 2015). Because of the severe impacts of nitrogen and sulphur oxides on the environment, I expect that stakeholders assess higher pollution levels of these pollutants more negatively than other pollutants. Therefore, I expect these pollutants have a stronger influence on firm performance than carbon dioxide, water pollutants and waste. Again, I expect that this impact is stronger for the accounting-based firm performance measures than for the market-based measures.
H1b: Nitrogen and sulphur oxides have a stronger negative influence on firm performance measures than carbon dioxide, water pollutants and waste.
Next to the studies that examined the relationship between pollutant emissions and firm performance, some studies have investigated the relationship between pollutant emission and the country income level. For the international aspect of this paper, I take the country
income level into account when assessing the relationship between pollution and firm performance. Grossman and Krueger (1992) proposed the Environmental Kuznets Curve (EKC) hypothesis for the relationship between environmental pollution levels and country income. They suggested that relationship between pollution levels and income is an inverted- U shaped curve, implying that pollution levels will increase when a country develops, but will decrease as income passes beyond a turning point. Several other studies have also investigated this relationship, for example Selden and Song, (1994), Tucker (1995), and List and Gallet (1999). Usually, the EKC hypothesis is tested using sulphur dioxide as the dependent variable. Grossman and Krueger (1992) and Shafik and Bandyopadhyas (1992) found that the turning point of sulphur dioxide was around an income of $5000 per capita.
List and Gallet (1999) found evidence for an inverted-U shaped relationship between nitrogen oxide and sulphur dioxide and country income, by studying United States emission data over the period of 1929-1994. However, for example, Tucker (1995) found that there is also a positive relationship between carbon dioxide emissions and country GDP. This study used regression analysis to investigate the relationship between emitted carbon dioxide and the economies of 137 countries in the period of 1971 to 1991. GDP and carbon dioxide emissions are both scaled on a per capita basis. Tucker (1995) found that the level of carbon emissions is decelerated when higher income levels are reached.
Unlike EKC hypothesis studies, I do not take countries’ income growth, but income level into account. Countries with a high-income level are beyond the turning point, implying a decrease in pollution emissions. The decline in pollution emissions may partially be explained by the possibility that more knowledge and advanced technologies are available in these countries, enabling cleaner production methodologies. Moreover, stakeholders become more aware and assign more importance to the amount of pollutions emitted. Stakeholders from countries with higher income levels may increasingly demand environmental protection (Tucker, 1995). Improved and increased laws and regulations can require a firm-level decrease of pollutions emitted. Therefore, I expect that the negative relationship between pollutants emitted and firm performance is stronger for firms in countries with higher incomes. Stakeholders may punish firms from these countries more as they assign more importance to these environmental performance issues relative to stakeholders from the lower income countries. This leads me to the following hypothesis.
H2: The negative relationship between pollutants emitted and firm performance is stronger for firms with higher incomes.
3.DATA AND METHODS
The main purpose of this study is to examine the effects of different pollutants on firm performance. This study investigates the relationship between each pollutant and each of the firm performance measures for a large international sample of 1804 firms. I will explain the sample construction later in this section.
Therefore, the basic model specification is expressed as follows:
Firm performanceijt = β0 + β1Debtit + β2Growthit + β3Sizeit + β4 R&Dit (6) + β5Pollutantiet + µcj + εit
Where i denotes the firm; j indicates the firm performance measure; c expresses the country; t shows the period; and e denotes the type of pollutant. µ captures the industry-specific fixed effects and ε is the standard error term. ROA, ROE, Market returns, Beta and Z-score,
measure firm performance. Wagner et al. (2002) found that a non-linear relationship between environmental performance and firm performance was not robust. Furthermore, studies as Hart and Ahuja (1996); Iwata and Okada (2011); Horvathova (2012) also used this basic linear model specification to test their hypotheses.
In this study, firm performance is measured through five financial performance measures of which three are value and return related and two measures relate to risk. As to the former, I use return on assets to capture operational performance, return on equity for financial performance and annual stock market returns to measure market performance. Each of the firm performance measures captures the behaviour and assessment of different
stakeholders with various interests (e.g. government, local societies, investors, employees, consumers, financial agencies, trading partners and stockholders) (Iwata and Okada, 2011).
To illustrate, for example, local societies who suffer directly from environmental pollution caused by a firm, might behave and assess the firm differently than other stakeholders who have only a monetary relationship with the firm, and therefore do not directly suffer from the
environmental pollution. Iwata and Okada (2011) found that stakeholders such as
stockholders, investors and financial agencies take also the long-run firm (both environmental and financial) performance into account, while stakeholders such as trading partners do not really care about a firm’s environmental performance in the short-run. Therefore, in order to be able to capture the different stakeholder assessments, five firm performance measures are included.
Operational performance is measured by using return on assets, in order to capture a firm’s internal performance on the balance sheet. Several other studies have also used return on assets as a measure for firm performance (e.g. Jaggi and Freedman, 1992; Hart and Ahuja, 1996; Alvarez, 2012 and Iwata and Okada, 2011). Financial performance is captured by return on equity. A number of studies, for example Jaggi and Freedman, 1992; Hart and Ahuja, 1996; Wagner et al., 2002; Alvarez, 2012; Iwata and Okada, 2011, also use this accounting measure for financial performance. However, a drawback of these performance measures is that return on assets and return on equity are both accounting measures that are based on historical information and thus more vulnerable to manipulation (Wang et al., 2013). Therefore, I also include a market measure for financial performance. According to Feldman et al., (1996) stock market returns are commonly used as a market measure to measure financial performance, this measure is also used by for example Klassen and McLauglin, 1996; Gilley et al., 2000; Scholtens and van der Groot; 2014 and Oestreich and Tsiakas, 2015.
Additionally, two specific measures of risk are included, as it seems that a firm’s degree of pollution may have an effect on risk. Gonenc and Scholtens (2017) found that including risk measures are useful in understanding the relationship between environmental and firm performance. More specifically, especially for oil & gas firms, good social
performance reduces systematic risk (beta) and business risk. McGuire et al. (1988) suggest that a firm’s financial risk may be increased through low levels of environmental
responsibility. Moreover, investors may consider investments in firms with low levels of environmental responsibility as more risky, because the stakeholders view the low
environmental responsibility as low management skills (Alexander and Buchholtz, 1978).
Pollution is an aspect of a firm’s environmental responsibility. Firms with high social responsibility, as compared to other firms, are less sensitive to some external events, and therefore may have lower market-based and accounting-based risk (McGuire et al., 1988). I include a firm’s systematic risk by including beta. Additionally, business risk, in terms of a firm’s Altman’s Z-score, is included to also capture an accounting measure of risk.
This study includes different pollutants to capture the individual effects of the pollutants on firm performance. Previous studies have often included an aggregated index when different pollutants were used. However, this may be misleading as the individual dangerousness of a pollutant varies among each type of pollutant (Horvathova, 2012).
Moreover, when an index is used, the impact of one individual pollutant on firm performance is not investigated. Therefore, the five different pollutants are included individually in this study. A firm’s level of carbon dioxide (CO2), sulphur oxide (SOx), nitrogen oxide (NOx), water pollutants and waste is measured in tonnes divided by its net operating revenues to allow firm comparisons. To measure the actual impact on the environment absolute emission levels should be used. However, as it is not the purpose of this study to investigate a firm’s actual impact on the environment, I use the scaling to be able to compare the firms.
To control for the influence of firm-level characteristics on the relationship, I included four control variables. Indebtedness is measured as debt to total assets (Horvathova, 2012;
Wang et al., 2013; Lee et al., 2015), growth is captured by the change in total assets (Iwata and Okada, 2011), the log of total assets is taken to include firm size (Hart and Ahuja, 1996;
Konar and Cohen, 2001; Horvathova, 2012; Alvarez, 2012;) and research and development expenses are divided by total assets to measure research and development intensity (Hart and Ahuja, 1996; Konar and Cohen, 2001; Wang et al., 2013; Lee et al., 2015). A full description of the variables used in this study can be found in table B1 in the appendix.
The sample used is drawn from two data sources. This study includes all firms from Thomson Reuters Datastream ASSET4 ESG data, for which data was available for at least one of the pollutants used in this study. According to Gonenc and Scholtens (2017) is the ASSET4 ESG database preferred over MSCI because of the reporting consistency of ASSET4, and because the database provider, Thomson Reuters, also provides financial performance information for the same companies. Only research & development expenses were obtained from Orbis, as the financial performance measures as well as the other control variables were collected from Thomson Reuters Datastream. Based on all available data, the sample consists of unbalanced panel data, which includes 1804 firms from 2007 to 2015 from 20 industries and 43 countries.
I use World Bank country classification data to determine a country’s income level.
Countries are divided into four different income groups: low, lower-middle, upper-middle, and high. Country income is measured in US dollars, using gross national income (GNI) per capita (World Bank, 2016). Countries are classified in the low-income group if the GNI per capita is or is less than $1,025. The lower-middle group consists of countries with a GNI per
capita between $1,026 and $4,035 and countries with a GNI per capita between $4,036 and
$12,475 are classified in the upper-middle group. The high-income group consists of
countries with a GNI per capita of $12,476 or more. To investigate if there is a difference in the relationship between pollutant emissions and firm performance based on firm income, I group the low and lower-middle income countries and the upper-middle and high-income countries together. A country’s level of carbon dioxide emissions is also obtained from World Bank. Carbon dioxide emissions are measured in metric tonnes per capita. The world average is 4.996 (World Bank, 2013). If a country is emitting more than the world average, it is considered as a ‘high emitting’ country, and if it is emitting less than the world average it is considered as ‘low emitting’ country.
As estimation method, I use the fixed effects model because the sample consists of 9 year unbalanced panel data. I use industry-specific fixed effects to control for unobserved industry-specific fixed effects that may have an influence on financial performance.
Endogeneity issues arising from the unobserved industry-, country- and year-specific effects are dealt with using this estimation method (Iwata and Okada, 2011; Lee et al., 2015).
Endogeneity issues are particularly associated with omitted variables. A fixed effects model minimizes this problem as it deals with the omitted variable by using within-group variations over time (Lee et al., 2015). The omitted variable causes an endogeneity bias as it contains heterogeneity that affects the dependent variable, but is not observed by the included
regressors. The fixed effects model can control for the unobserved heterogeneity. The robust standard errors are clustered at the frim level. In panel data estimated standard errors are not independently distributed, but the residuals will be correlated across years within each firm.
To take this error structure into account, standard errors are clustered at the firm level.
To avoid outliers affecting the estimation results, the data is winsorized at 0.01 and 0.99. A new variable identical to the initial variable is generated when a variable is
winsorized. Winsorizing replaces the highest and lowest values, by the next inward value taken from the extremes in the new variable. In comparison to previous papers (e.g. Hart and Ahuja, 1996; Konar and Cohen, 2001; Wagner et al., 2002; Nishitani and Kokubu, 2011;
Iwata and Okada, 2011; Wang et al., (2013); Lee et al., 2015), this sample is highly
international, focuses on more industries, includes a more recent period and uses a broader scope of pollutants and financial performance measures. Besides taking operational, financial and stock market performance into account, I also include risk characteristics and use five
different pollutant emissions to be able to investigate the relationship between pollution and firm performance at a detailed level.
4.RESULTS
In this section, he analysis proceeds as follows. First, I report the descriptive statistics. Next, I report the results for the effect of the different pollutants on firm performance in tables 1-5 for each of the five firm performance measures. Third, I will separately run the regressions again for the countries United Kingdom, Japan and the United States, as these three countries are overrepresented in this sample. Moreover, I test the influence of country level income on the relationship between pollutant emissions and firm performance for the other 40 countries.
Then, I divide the sample into ‘clean; and ‘dirty’ industries following Mani and Wheeler (1998).
4.1 Descriptive statistics
Table 1 presents the descriptive statistics for all the variables. Return on assets, return on equity and annual market returns are presented in percentages. Waste and carbon dioxide emissions have the highest averages of the five different pollutants. Regarding the air pollutants, carbon dioxide, nitrogen and sulphur oxide, this is not a surprise. Carbon dioxide is generally emitted in higher quantities than nitrogen and sulphur oxide. However, the latter are more harmful to the environment. Waste is captured in this study as total waste produced by the firm in tonnes, including hazardous and non-hazardous waste. Therefore, it is not surprising that the average of this pollutant is significantly higher than for the other pollutants.
Table 1
Descriptive statistics of all variables.
Mean Median Min. Max. SD Kurtosis Skewness
ROA .06 .06 -.19 .33 .07 6.19 .27
ROE .14 .12 -.64 1.30 .23 11.50 1.30
MR .11 .08 -.75 1.82 .44 5.26 .97
BETA 1.02 .97 -.40 3.61 .63 6.10 1.14
ZSCORE 1.67 1.60 -.66 4.64 .96 3.58 .48
CO2 .30 .03 .00 6.15 .88 29.12 4.89
NOX .00 .00 .00 .02 .00 25.11 4.48
SOX .00 .00 .00 .04 .01 34.39 5.34
WATER .00 .00 .00 .18 .02 77.82 8.57
WASTE 1.51 .00 .00 69.89 9.00 47.10 6.63
DEBT .56 .57 .09 1.00 .20 2.73 -.10
GROWTH .70 .04 -1.00 43.08 4.79 68.28 8.00
SIZE 7.25 7.15 5.21 10.36 1.12 2.89 .63
R&D .01 .00 .00 .18 0.03 25.94 4.59
Table 2 shows the distribution of the countries that are represented in the sample. Firms from Great Britain, Japan, and the United States represent 47% of all firms in the sample (13%, 15%, and 19% respectively). Moreover, 89% of the firms are from a high-income country and 89% of the firms are from a country that has a high level of carbon dioxide emissions per capita. The high percentage of developed countries may be explained by two reasons. First, the data coverage of Thomson Reuters ASSET4 data was originally limited to European and US firms, but has expanded over the years. Data collection of developed countries data is generally easier and more reliable than of developing countries data. It is possible that the database contains more developed countries than developing countries. Second, improved and increased laws and regulations might require better reporting of firm pollution, especially in developed countries.
Table 2
The distribution of countries in the sample. Additionally, a country’s income level and general carbon dioxide level is presented here.
Country N Income
level
CO2
level
Country N Income
level
CO2
level
Australia 87 High High Austria 11 High High
Belgium 13 High High Bermuda 12 High High
Brazil 46 Upper Low Canada 97 High High
Cayman Islands 8 High High Cyprus 1 High High
Denmark 18 High High Finland 23 High High
France 77 High High Germany 54 High High
Great Britain 227 High High Greece 1 High High
Hong Kong 22 High High Hungary 2 High Low
Indonesia 9 Lower Low Ireland 15 High High
Italy 19 High High Japan 268 High High
Jersey 9 High High Korea 58 High High
Liberia 1 Low Low Malaysia 14 Upper High
Mauritius 1 Upper Low Mexico 16 Upper Low
Netherlands 28 High High New Zealand 7 High High
Norway 15 High High Panama 1 Upper Low
Papua New Guinea 1 Lower Low Philippines 9 Lower Low
Poland 10 High High Singapore 11 High High
South Africa 74 Upper High Spain 34 High High
Sweden 35 High Low Switzerland 38 High High
Taiwan 55 High Low Thailand 12 Upper Low
Turkey 9 Upper Low United States 347 High High
Virgin Islands (GB) 1 High High
Income level CO2 level
High 1612 High 1607
Upper middle 172 Low 197
Lower middle 19
Low 1
Total 1804 Total 1804
Table 3 presents the correlation matrix between the different variables. As expected, carbon dioxide, nitrogen and sulphur oxide and waste are all negatively correlated with return on assets, return on equity, annual market return and Altman’s Z-score, and positively correlated with beta. However, water pollutants are positively correlated with return on assets, return on equity and annual market return. These correlations suggest that there is an impact of
Table 3 Correlation matrix of all variables. 1 2 3 4 5 6 7 8 9 1011121314 ROA (1) 1.00 ROE (2) .79* 1.00 MR (3) .15* .13* 1.00 BETA (4)-.13*-.13*-.05* 1.00 ZSCORE (5) .55* .43* .10*-.10 1.00 CO2 (6) -.07*-.05*-.05*.01 -.22*1.00 NOX (7)-.01 -.01 -.03*.05*-.21*.65*1.00 SOX (8)-.03*-.03*-.02 .03*-.20*.61*.66*1.00 WATER (9) .07* .02 .01 -.08*-.04*.12*.10*.18*1.00 WASTE (10)-.10*-.10*-.07*.22*-.17*.18*.18*.32*.27*1.00 DEBT (11)-.17* .11*-.04*-.09*-.23*.07*.00 -.01 -.07*-.19*1.00 GROWTH (12) .02* .02* .02*.01 0.00-.00 -.03 -.03 -.02 -.02*-.01 1.00 SIZE (13) -.14*-.12*-.02*0.02*-0.08*-.11*-.27 .22*-.15*-.13*.03*.15*1.00 R&D (14) .08* .04* .03*0.02*0.05*-.07*-.10*.08*-.02 -.06*-.09*-.03*-0.20*1.00 * show significant values at the 10% significance level
pollutants on firm performance and that it differs per pollutant. Moreover, the three air pollutants highly correlate with each other (>0.6), therefore they are used separately in the regressions.
4.2 Full sample results
Table 4 presents the estimation results for the effect of the different pollutants on ROA, for all industries and all countries, using country-, industry-, and year-fixed effects. It shows preliminary evidence that how pollutant emissions affect financial performance is different for each type of pollutant. The effect of nitrogen oxide and water pollutants is not statistically significant on return on assets, whereas carbon dioxide, sulphur oxide and waste have significant negative impacts on return on assets. Noticeably, the coefficient of water pollutants is positive, while carbon dioxide, nitrogen oxide, sulphur oxide and waste have the expected coefficient signs. Carbon dioxide and sulphur oxide are significant at the 5% level, while waste is significant at the 1% level. These results provide thus partial support for hypothesis 1a. Carbon dioxide and waste have a rather weak impact compared to the strong negative effect of sulphur oxide. However the coefficient of nitrogen oxide is also strongly negative, it is not found to be significant. Hypothesis 1b is thus only supported for sulphur oxide and not for nitrogen oxide.
These results are consistent with the findings of Iwata and Okada (2011); Alvarez (2012); and Lee et al., (2015). Iwata and Okada (2011) found a negative relationship between greenhouse gas emissions and return on assets, while Alvarez (2012) and Lee et al. (2015) both found a significant negative relationship between a firm’s carbon dioxide emissions and its return on assets. Other studies, for example, Hart and Ahuja (1996) and Horvathova (2012) also found a negative relationship between pollutant emission levels and firm performance, however Hart and Ahuja (1996) only found a significant increase in operating performance after one year, while Horvathova (2012) found this after two years. Both of these studies used a relatively small sample size and a short time period. The numbers of firms studied were 127 and 136, respectively and they used timeframes of 2 and 5 years. In contrast to Iwata and Okada (2011), I do find a significant negative influence of waste on return on assets, although it is rather weak. They did not find significant effects of waste emissions on financial performance. After discussing the results of pollutant emissions on return on assets, I will now turn to the other accounting measure, return on equity.
Table 4
Results for effect of different pollution emissions on financial performance for the full sample. Where return on assets is the dependent variable.
ROA ROA ROA ROA ROA
DEBT -0.078*** -0.100*** -0.092*** -0.082*** -0.090***
[0.008] [0.010] [0.011] [0.012] [0.010]
GROWTH -0.000** -0.000*** -0.000*** -0.000* -0.000***
[0.000] [0.000] [0.000] [0.000] [0.000]
SIZE -0.004 0.002 0.004 0.006 -0.001
[0.003] [0.004] [0.004] [0.004] [0.003]
R&D 0.088 0.303*** 0.239** 0.075 0.093
[0.060] [0.098] [0.120] [0.102] [0.073]
CO2 -0.003**
[0.001]
NOX -0.790
[0.750]
SOX -0.758**
[0.326]
WATER 0.076
[0.108]
WASTE -0.001***
[0.000]
Constant 0.110*** 0.102*** 0.135*** 0.066** 0.104***
[0.037] [0.028] [0.035] [0.029] [0.024]
R2 0.203 0.279 0.278 0.345 0.240
Observations 10866 3969 3795 2187 7246
F-statistic 37.98*** 23.29*** 22.18*** 21.53*** 31.54***
*,** and *** show significant values at, 10%, 5% and 1% significance levels, respectively The numbers in the parentheses are the robust standard errors clustered at the firm level
Table 5 provides the results for the effect of each pollutant on return on equity.
Similar to the effect on return on assets, the pollutants carbon dioxide, sulphur oxide, and waste have a significant negative impact on return on equity, with a weak impact of carbon dioxide and waste and a strong impact of sulphur oxide. Again, this provides partial evidence for hypotheses 1a and 1b. The relationship is negative, however it is not statistically significant for all the pollution types. Again, the negative coefficients are stronger for carbon oxide and sulphur oxide than for the other pollutants, with only the coefficient of sulphur oxide being significant. These results also imply that carbon dioxide has a slightly higher negative influence on return on equity than on return on assets. Yet, carbon dioxide is now significant at the 10% level only. Additionally, sulphur oxide seems to have a bigger negative impact on return on equity than it has on return on assets, with a significance level of 1%.
There is no difference of the impact of waste on these firm performance measures. These
results are similar to the results of Iwata and Okada (2011), Alvarez (2012), and Horvathova (2012). Wagner (2002) found that there is a negative relationship between pollutants and return on equity. However, Wagner used an index of high correlating pollutants and could not make any inferences about their individual impact on return on equity.
Table 5
Results for effect of different pollution emissions on financial performance for the full sample. Where return on equity is the dependent variable.
ROE ROE ROE ROE ROE
DEBT 0.144*** -0.075* -0.053 -0.058 0.097**
[0.038] [0.044] [0.046] [0.059] [0.043]
GROWTH -0.000 -0.001** -0.001*** -0.001** -0.001***
[0.000] [0.000] [0.000] [0.001] [0.000]
SIZE -0.015* 0.002 0.005 0.018 -0.013
[0.009] [0.012] [0.013] [0.016] [0.011]
R&D 0.202 0.666* 0.514 0.359 0.203
[0.183] [0.365] [0.385] [0.401] [0.225]
CO2 -0.010*
[0.006]
NOX -2.533
[1.805]
SOX -2.301***
[0.853]
WATER 0.010
[0.243]
WASTE -0.001***
[0.001]
Constant -0.049 0.054 0.205* 0.053 0.108
[0.124] [0.104] [0.113] [0.114] [0.085]
R2 0.135 0.167 0.177 0.229 0.154
Observations 10736 3943 3778 2182 7179
F-statistic 23.31*** 23.29*** 22.18*** 21.53*** 31.54***
*,** and *** show significant values at, 10%, 5% and 1% significance levels, respectively The numbers in the parentheses are the robust standard errors clustered at the firm level
Table 6 shows the estimation results of each pollutant on the stock market return of a firm. Again, except for water pollutants, the coefficients are negative. However, only carbon dioxide and waste have a significant negative influence on a firm’s stock market return.
Carbon dioxide is significant at the 5% level, while waste is significant at the 10% level only.
These results provide partial support for hypothesis 1a. Hypothesis 1b is not supported, while only the coefficient of carbon oxide is more negative than for the other pollutants, the coefficients of carbon and sulphur oxide are not significant. These estimation results imply
that the market only takes the well-known pollutants, carbon dioxide and waste, into account when assessing the firm performance. These results are comparable to Orlitzky et al. (2003) and Wu (2006). They stated that the correlation between environmental performance and market based measures is less than for accounting performance measures. It may be that investors are not concerned about environmental performance of firms, and only take its financial performance into account, when assessing the firm’s performance. Moreover, these results are similar to results of Klassen and McLaughlin (1996) and Oestreich and Tsiakas (2015), who found that stock market returns were increased through good environmental performance.
Table 6
Results for effect of different pollution emissions on financial performance for the full sample. Where market stock return is the dependent variable.
MR MR MR MR MR
DEBT -0.080*** -0.099*** -0.087** -0.028 -0.054**
[0.021] [0.035] [0.036] [0.046] [0.025]
GROWTH 0.024 0.195*** 0.184*** 0.177*** 0.165***
[0.023] [0.042] [0.025] [0.030] [0.028]
SIZE -0.015** -0.034*** -0.029** -0.055*** -0.011
[0.007] [0.012] [0.013] [0.014] [0.008]
R&D 0.049 0.459** 0.275 0.093 0.053
[0.143] [0.230] [0.232] [0.262] [0.170]
CO2 -0.011**
[0.004]
NOX -2.272
[2.241]
SOX -0.007
[1.123]
WATER 0.293
[0.330]
WASTE -0.002**
[0.001]
Constant -0.592*** -0.561*** -0.434*** -0.097 -0.602***
[0.053] [0.125] [0.099] [0.097] [0.101]
R2 0.316 0.346 0.348 0.392 0.330
Observations 10168 3654 3487 1993 6761
F-statistic 64.55*** 29.42*** 28.31*** 24.38*** 45.90***
*,** and *** show significant values at, 10%, 5% and 1% significance levels, respectively The numbers in the parentheses are the robust standard errors clustered at the firm level
Table 7 provides the results of the effects of each pollutant on beta. Except for water pollutants, is the effect of the other pollutants in line with hypothesis 1a and 1b. The coefficients are
