Environmental performance and cost of equity:
Developed versus developing countries
S.R. (Saska) van Engen (s2017067)
Study program: MSc International Financial Management
Faculty of Economics and Business, University of Groningen, the Netherlands.
Supervisor: W. (William) Forbes
Co-assessor: M. (Mario) Hernandez Tinoco
Thesis information Abstract
History:
Draft: 17th May 2016
Received feedback: 28th of May 2016 Final: 13th of June 2016
Keywords:
Environmental performance CSR
Cost of equity capital
Word count: 16.934
2
1. Introduction
In recent decades, the topic of corporate social responsibility (CSR) has been heavily debated by academics. One of the focus areas has been the link between environmental performance1 and firm value or financial performance for firms based in the United States. A research review of Murphy (2002) focusing on research from the 1990s till 2002 indicates that the majority of the papers reviewed found a positive correlation between environmental performance and firm market value (Konar and Cohen, 2001; Dowell et al., 2000), stock market return (Gottsmand and Kessler, 1998; Cohen et al., 1995), profitability (Stanwick and Stanwick, 1998; Russo and Fouts, 1997) and a negative correlation with the cost of capital (Feldman et al., 1997; Garber and Hammitt, 1998).
Sharfman and Fernando (2008), El Ghoul et al. (2011) and Cheng et al. (2014) have continued to research the link between environmental performance of firms and the cost of (equity) capital. Sharfman and Fernando (2008) have found a negative correlation between the improved environmental risk management of US firms and their cost of equity capital, while the cost of debt capital does not yield significant results in their research2. Furthermore, firms will become more leveraged as they switch from equity to debt financing, enabling them to profit from the tax shield. El Ghoul et al. (2011) looked at the link between CSR and the cost of equity capital for US firms and they find that there is a negative correlation between these two. This especially holds for employee relations, product strategies and environmental policies. Cheng et al. (2014) used a broad spectrum of CSR performance metrics and data on firms from 49 countries. They find that stakeholder engagement and transparency of CSR performance reduce capital constraints; the social and environmental pillars of CSR especially drive this outcome. Unfortunately, Cheng et al. (2014) have not looked at how CSR is changing capital constrains on a country level but have simply looked at all firms in general.
1
Environmental performance is one of the pillars of Corporate Social Responsibility (CSR). According to ISO 14001 (requirements for environmental management), environmental performance results are achieved by managing the environmental aspects of the organization’s activities, products, processes, systems and services. Examples of criteria on which an organization is able to monitor its environmental performance are emission (CO2) reductions, reductions in toxic wastes, resource reduction and improving product processes.
2 The tradeoff seems unbeneficial for investors as higher investments in environmental performance will
3 CSR, and in this case environmental performance, can be perceived differently by investors, depending on the country or region in which the firm is located or listed on the stock exchange (Ammann et al., 2011). According to Cheng et al. (2014) and Hong et al. (2011) CSR can be seen as a luxury good. Only firms with sufficient funds and low capital constraints are able to invest in CSR. One would be able to conclude from the research of Cheng et al. (2014) and Hong et al. (2011) that CSR is most apparent in firms from developed countries, as these firms often have lower capital constrains and an abundance of funds (Pagano et al., 2002). However, the work of Cheng et al. (2014) shows that this argument does not hold. The most capital constrained firms show the strongest relation with CSR, hence these companies benefit more from investments in CSR as this will increase their ability to obtain capital and at a lower cost. Furthermore, CSR is expected to increase the amount of disclosure of a firm, as the firm wants its investors and stakeholders to be informed about its performance (Cheng et al., 2014). The cost of equity capital is influenced by disclosure levels (Botosan, 1997; Stulz 1999) as increased disclosure levels increase transparency and reduce information asymmetry and agency costs, leading to an overall reduction in the cost of equity capital. Therefore, the relationship between CSR and cost of equity capital would become more negative in countries with lower overall transparency and higher costs of obtaining information as CSR enhances the transparency of the firm. This is confirmed by Durnev and Kim (2005) as firms with high corporate governance and transparency levels have higher firm values when traded in “less investor-friendly countries3” versus firms located in countries like the United States.
The following research question will be investigated in this Thesis:
Do investors located in developing countries appreciate increases in environmental performance differently than investors located in developed countries?
The trade-off here is the implication that investors differ in their opinions on environmental performance based on the stock exchange they are trading on. Hence, an investor originating
3
In their research, Durnev and Kim (2005) have focused on the legal environment and the legal protection of investors. Less-investor friendly countries offer poor legal protection for equity investors. The strength of a country’s legal regime is accounted for by looking at the anti-director rights index and the rule-of-law index from the International Country Risk Guide. In these less-investor friendly environments, other types of ownership may flourish, as seen with the chaebol in South-Korea or Keiretsu in Japan. Note: while used extensively in previous
4 from the U.S. investing in the U.S. stock exchange (e.g. NYSE) will have a different opinion on environmental performance than an investor from Indonesia trading on the Indonesian stock exchange (e.g. IDX). When the U.S. investor values environmental performance more highly, an investment in environmental performance by a company will lead to a reduction of the firm’s cost of equity capital. The research question is inspired by Schröder (2014), who states that there is a high need for international research on the link between environmental performance and cost of equity capital.
With the usage of a panel data fixed effects model, environmental performance data of 3,729 firms from 35 countries is tested to see if the results from previous research hold. Furthermore, this international dataset allows me to test if there are differences between the effect from environmental performance on the cost of equity capital (measured by the capital asset pricing model) depending on the country in which the firm is active on the stock exchange. An interaction variable between the country in which the firm is traded and the environmental performance of the firm allows me to test for this difference. My results are similar to previous findings and find a significant negative relationship between the cost of equity capital and environmental performance. Furthermore, the impact of environmental performance is higher for countries with a low gross national income versus high gross national income countries.
The rest of the thesis will be structured as follows. Chapter 2 will provide an overview of the relevant literature in this field and construct the hypotheses that will be researched. Chapter 3 will develop an econometric model to test the hypotheses. Chapter 4 will discuss and describe the data used. Chapter 5 analyses the data in order to reach the conclusions in Chapter 6.
2. Literature review
5 2.1 Environmental performance
There are multiple views as to how environmental performance decreases the cost of equity capital. The resource-based view (Barney, 1991) explains that a more efficient usage of resources leads to lower input costs and decreases the environmental impact of production by lowering inputs. This raises the firm’s profits and has a lowering effect on its cost of equity capital. Other researchers have focused on the risk perception of investors in explaining a decreasing cost of equity capital due to improved environmental performance (Sharfman and Fernando, 2008; El Ghoul et al., 2011; Cheng et al., 2014). However, the environment is a given public good. Changes in the environment could have similar impacts on all firms, hence the risk is systemic and firms are not able to lower it by improving their environmental performance. Only when many or perhaps all firms are able to reduce their environmental impact on the planet, certain risks can be prevented or postponed (see section 2.2 for a broader insight).
A firm that improves its environmental performance reduces the risk of future claims from governmental organizations concerning e.g. emission reductions (Sharfman and Fernando, 2008). Sharfman and Fernando (2008) state that there are current and future hazards that carry uncertain financial consequences for a firm. By improving environmental performance, the firm reduces its current and future hazards, thus reducing the uncertainty concerning future financial claims (e.g. settlements, fines or litigation costs). An example of this is the Exxon Valdez oil spill case of 1988, which cost the company 8.7 billion U.S. dollars in clean-up costs and fines (Husted, 2005). Another risk a firm faces is the risk of reputational loss (Robinson et al., 2008), stemming from large disasters that could have been prevented by investing in environmental risk management. Reputational loss could lead to reductions in revenue and hence decreasing firm market values. Furthermore, investment in environmental performance, especially when investments are done in the R&D department, could become an additional source of income. The firm would be able to license new technologies, benefiting the firm in multiple ways.
6 for non-green stocks (only ‘non-green investors’ remain for these non-green stocks). These ‘green investors’ are provided with utility from green firms which adds value to them above the normal return on an investment. This large pool of green investors has a decreasing effect on cost of equity capital for green firms, by raising the supply of capital for any given level of demand, while the smaller group of non-green investors reduces funds available for firms that do not invest in environmental performance, by reducing supply for a fixed level of demand, hence the cost of equity capital increases for non-green firms. When this gap between green and non-green firms’ cost of equity becomes large enough to offset investments, investor behavior is able to influence corporate behavior. Mackey et al. (2007) build on this theory and state that firms will maximize their firm market value by arbitraging the market (depending on the amount of green investors and the amount of green firms) to resolve imbalances between the supply of capital to green and un-green ventures. They are able to do so by investing in environmental performance, disinvestment or by keeping their current policy. One should keep in mind that increasing or decreasing involvement in environmental performance does come at a certain cost, thus the benefits might not outweigh the costs. According to the US Social Investment Forum (2014), 6,57 trillion U.S. dollars are invested according to CSR criteria, nearly 17 percent of all investments in the U.S. in 2013. In Europe, investments conforming to the CSR criteria also accounted for 17 percent of the market, with a volume of 2,5 trillion euro in 2013 (Schröder, 2014).
El Ghoul et al. (2011) follow the above two mentioned paths through which CSR influences the cost of equity capital (the reduction in the risks of future claims and the increase of green investors) and add a few other dimensions to the debate. According to them, CSR reduces information asymmetry. Investors want information on a firm before they are willing to invest. Disclosure4 of information is higher for high CSR firms, as firms want to signal their greener production efforts to investors and other stakeholders like consumers (Dhaliwal et al., 2009). This is the so-called ‘signaling by the firm’ and allows bad practice (poor CSR policies) to be driven
4
7 out by good practice. According to El Ghoul et al. (2011), the other two stages that are important to prevent information asymmetry are ‘coverage by the media and analysts’ and ‘reception by investors’. According to Fernando et al. (2010), non-green firms (toxic and gray stocks according to their paper5) have a higher likelihood of penalties, litigation and other costs concerning the environment. This would lead to investors relying heavily on analyst reports during their stock selection, hence, non-green firms reach a higher analyst coverage than green firms for the abovementioned reasons (as found by Fernando et al., 2010). Lastly, green investors will be focused solely on green firms, and thus will only act upon information received from/about these firms. A higher information flow from green firms also leads to increased transparency and higher disclosure levels (Cheng et al., 2014). According to Botosan (1997) and Stulz (1999), increased transparency and disclosure levels have a negative effect on cost of equity capital. Furthermore, stakeholder engagement is higher in green firms, leading to a reduction of agency costs (Cheng et al., 2014)
Signaling has been found to be highly effective by Lys et al. (2015). Their research has shown that firms anticipate a stronger financial performance in the future, and hence decide to invest in CSR. With their research they have ruled out the so-called “charity motive” that states that firms solely invest in CSR to benefit society as CSR investments positively effect a firm’s future performance in favor of a form of enlightened self-interest. Furthermore, a firm’s financial performance does not benefit from CSR investments. Rather, they state that CSR investments are caused by management’s perspective of the firm’s financial performance. Hence, CSR investments made by a firm signal to the market that management expects a solid financial future with increased profits.
2.2 Cost of equity capital
According to Sharfman and Fernando (2008), the cost of equity and debt capital is linked to the return investors demand on their investment. The higher the rate of return, the riskier the firm and the higher the financing costs are to the firm. Moreover, they state that the cost of equity and debt
8 capital combined, in the firm’s weighted-average cost of capital (WACC), is used to discount a firm’s future cash flows and determine the viability of future projects. Hence, it determines which projects are worth pursuing and which are not profitable under the current cost of capital. Indirectly, the cost of capital is thus also linked to a firm’s market value as this is calculated by the firm’s future cash flows discounted by its weighted-average cost of capital (Konar and Cohen, 2001; Dowell et al., 2000). El Ghoul et al. (2011) agree with Sharfman and Fernando (2008) that the cost of equity capital is important in determining the firm’s overall required rate of return. The cost of equity capital is also linked to the amount of systematic risk the firm is exposed to (Stulz, 1999; Sharfman and Fernando, 2008). Systematic risk is often explained by looking at the beta of the firm, obtained by the Capital Asset Pricing Model (CAPM). The firm’s beta explains the tendency of its stock to move along with the market. A high beta indicates that the stock is heavily influenced by changes in the overall economy (Stulz, 1999). However, since the CAPM was revealed to be empirically flawed (Fama and French, 1992), other models have been developed that are better capable to predict the stock returns of a firm, and thus a firm’s cost of equity capital. Examples of these are the Fama and French (1993) three-factor model, the Carhart (1997) four-factor model and the Fama and French (2015) five-factor model. These models include the excess returns of small cap stocks over big cap stocks and the excess between value stocks and growth stocks. The explanative power of these models has increased vis-à-vis the CAPM model, with the Fama and French (1993) three-factor model explaining 90 percent of a portfolios return. Part 3.1 will explain in why I still use the CAPM model in my thesis.
Global warming creates systematic risk (e.g. the unknown hazards in the paper of Sharfman and Fernando, 2008) by having a depressing effect on the overall economy and making production conditions harder, hence increasing the riskiness of the market and thus for all firms. Sustainable production and investing in environmental performance lowers adaptation costs for climate change in the future. Lowering costs while revenues remain equal leads to higher profits and a lower cost of equity capital.
9 (1999) argues that the firm’s stock trade solely within the home market, leading to low diversification options for investors and local investors carrying all the risk. However, in a highly globalized world where capital markets are not restricted, foreign (global) investors are also able to purchase the local firm’s stock. This improves the investment climate for both local and foreign investors. Foreign investors are able to diversify between different markets, resulting in the returns of the global investor’s portfolio to become more stable as diversification leads to smoothening out individual business cycles. Under open capital markets, local investors no longer have to bear all the risks within the economy but these risks are now also shared with the foreign investor. Cetorelli and Goldberg (2011) state that banks have globalized and are thus able to service their customers globally. The financial crisis of 2007/2008 has highlighted the globalized banking sector as it acted as one of the main transmission channels. Regarding equity financing, international correlation between equity markets has increased, as has already been indicated by Longin and Solnik (1995) for the time period between 1960 and 1990. Later research by Bekaert et al. (2009) indicates that there is no evidence of increased correlations in stock exchange returns (except for the EU market). Hence, there is no general agreement on stock markets being globalized. Globalization on a firm level has been debated by Rugman and Verbeke (2004) who state that firms are still very nationally or regionally oriented and solely 9 out of the 500 firms in their research are truly global6. 320 of the 500 firms are oriented towards their home region (either North America, Europe or Asia-Pacific).
2.3 Developed versus developing countries
Regarding the differences between developed and developing countries, no research has been published to my knowledge concerning the link between environmental performance and the cost of equity capital. This is confirmed by Schröder (2014) who calls for more international research on this topic.
The research done by Cheng et al. (2014) has by far the most international perspective. Their sample includes firms from 49 different countries, which range from the United States and
6 Rugman and Verbeke (2004) used data from 2001, the amount of global firms might have fluctuated since then.
10 multiple European countries to Brazil and Thailand. According to their findings, CSR is not the luxury good which it was perceived to be. On the contrary, companies that are capital constrained will benefit more from involvement in CSR as CSR often increases the amount of disclosure. Their research unfortunately has not focused on differences between countries, rather they created two groups: capital constrained and non-capital constrained firms.
Ammann et al. (2011) have done research on the relationship between corporate governance and firm value, and helpfully have also made a distinction between developed countries as part of their analysis. They show that the relationship between corporate governance and firm value is not significant for Japan, while it is significant for the UK, Canada, Australia and France. This difference would be due to the difference in how CSR is valued by Japanese investors according to them. Furthermore, the investor climate is different in Japan with the main bank-based Keiretsu families dominating the capital market for Japanese firms, so it may be easier to invest in order to reap a collective/public good. Overall, Ammann et al. (2011) find a strong significant relationship between corporate governance and firm value for their sample, hence, firms located in developed countries (excluding the US) benefit from investment in corporate governance.
11 It is of importance to mention here that the firm’s activities can be placed outside of the country it’s traded in. For example, Philips is traded on the Dutch stock market, but has many production plants based in Asia. The environmental performance of the firm is for the firm as a whole, not solely for its activities in the country in which its traded on the stock market. Investors in the Netherlands thus bear litigation risks when something happens in one of Philips’s plants in Asia. However, regarding emissions, the whole planet is affected, not solely the stakeholders in South Korea or shareholders in the Netherlands. The main aim of this thesis is to see if shareholders in the Netherlands react positively to the increased environmental performance of all the activities of Philips and their peers. On the other hand, regarding possible waste or spills from e.g. BP, the stakeholders originating in the country in which production takes place (e.g. Nigeria) are, needless to say, more effected by these events than the shareholders based in the country in which the company is incorporated (e.g. in the UK). Area segmented information from the firm would solve measurement issues surrounding this problem. Unfortunately, I have not been able to obtain such segmental information, thus this remains an area on which future literature hopefully will improve.
2.4 Hypotheses
12 al. (2015) explain the relationship through a different causative mechanism. According to them, the link between CSR and future financial performance is due to signaling to the market. Their research shows that firms investing in CSR anticipate on increased financial performance in the future, hence they signal to the market that their insider knowledge predicts increased profits. CSR in itself thus does not increase the financial performance of the firm, it simply is a side benefit.
In order to confirm the previous findings of the literature, the first hypothesis will test the relationship between the environmental performance of the firm and its cost of equity capital. According to the findings of Sharfman and Fernando (2008), El Ghoul et al. (2011) and Cheng et al. (2014), a negative relationship is expected. This hypothesis will confirm that the data used results in the same conclusions as those of the before mentioned academics. Hence, hypothesis one results in:
H1: An increase in environmental performance leads to a reduction in cost of equity capital.
As I am highly interested in the perceived benefits from environmental performance by investors, I will also measure if the country in which the stock is traded influences this relationship. This has been encouraged by Schröder (2014) who calls for more international research on this topic. Previous literature (see part 2.3) has found linkages between disclosure levels, corporate governance and firm performance, in combination with low shareholder protection rights. Following from this literature, one could come up with the following hypothesis to test whether investors perceive environmental performance differently across countries, as due to its perceived, or real, side benefits like increased transparency and reduced information and agency costs.
H2: Firms traded7 in developing countries benefit more from increased environmental performance versus firms traded in developed countries, measured by the associated reduction of the cost of equity capital.
13 The following section will discuss the econometric model used to research these relationships. Chapter 4 discusses the dataset used and chapter 5 shows the regression results. Finally, chapter 6 concludes my thesis.
3. Methodology
As stated during the climate conference in Paris (December 2015), global warming is an existential issue which needs to be dealt with sooner rather than later, almost regardless of cost8. Among the influencers in this issue are investors and firms. On the one hand, investors are able to pressure firms to reduce their impact on the environment (e.g. see theory developed by Heinkel et al., 2001, explained in part 2.1). On the other hand, firms themselves can also pioneer in this role without being pressured by stakeholders9. Do investors really appreciate the efforts of firms to reduce their impact on the environment?
In order to answer the first hypothesis, the following econometric model is used:
𝑟𝐸,𝑖𝑡 =
𝛽1+ 𝛽2 𝑒𝑛𝑣𝑖𝑟𝑜𝑛𝑚𝑒𝑛𝑡𝑎𝑙 𝑝𝑒𝑟𝑓𝑜𝑟𝑚𝑎𝑛𝑐𝑒𝑖,𝑡−1+ 𝛽3ln(𝑠𝑖𝑧𝑒)𝑖𝑡+ 𝛽4 𝑙𝑒𝑣𝑒𝑟𝑎𝑔𝑒𝑖𝑡+ 𝜃𝑐 + 𝜆𝑡+ 𝜇𝑖𝑛𝑑𝑢𝑠𝑡𝑟𝑦+ 𝜀𝑖𝑡
For the second hypothesis, an interaction term is added between the country in which the stock is traded and environmental performance. Hence the econometric model will transform into:
𝑟𝐸,𝑖𝑡 = 𝛽1+ 𝛽2 𝑒𝑛𝑣𝑖𝑟𝑜𝑛𝑚𝑒𝑛𝑡𝑎𝑙 𝑝𝑒𝑟𝑓𝑜𝑟𝑚𝑎𝑛𝑐𝑒𝑖,𝑡−1+
𝛽3𝐶𝑜𝑢𝑛𝑡𝑟𝑦 𝑑𝑢𝑚𝑚𝑦 ∗ 𝑒𝑛𝑣𝑖𝑟𝑜𝑛𝑚𝑒𝑛𝑡𝑎𝑙 𝑝𝑒𝑟𝑓𝑜𝑟𝑚𝑎𝑛𝑐𝑒 𝑖,𝑡−1+ 𝛽4ln(𝑠𝑖𝑧𝑒)𝑖𝑡+ 𝛽5 𝑙𝑒𝑣𝑒𝑟𝑎𝑔𝑒𝑖𝑡 + 𝜃𝑐+ 𝜆𝑡+ 𝜇𝑖𝑛𝑑𝑢𝑠𝑡𝑟𝑦+ 𝜀𝑖𝑡
8 According to NASA, March 2016 has been the 7th month in a row breaking global temperature records. There
is even a high chance of the 1.5 degrees Celsius increase agreed upon during the Paris climate conference being reached this year. (see: http://www.theguardian.com/environment/2016/may/16/april-third-month-in-row-to-break-global-temperature-records)
9 Stakeholders also include the government and their clients, besides shareholders.
14 In which 𝑟𝐸 is the cost of equity capital, which is the dependent variable. The model uses the CAPM to obtain the cost of equity capital. The independent variable is environmental performance. The control variables are firm size and leverage. Furthermore, the model controls for country (𝜃𝑐), year (𝜆𝑡) and industry (𝜇𝑖𝑛𝑑𝑢𝑠𝑡𝑟𝑦) fixed effects10. For hypothesis 2, interaction variables between the country dummies and environmental performance are added to the econometric model. These interaction variables will indicate the relationship between the environmental performance of a firm traded in a specific country and the effect on its cost of equity capital. The interaction term between U.S. firms and environmental performance has been left out in order to avoid the dummy variable trap. The U.S. firms will then become the reference group, which will be beneficial as previous research has found a lot of proof on the negative relationship between environmental performance and cost of equity capital for firms traded in the U.S.
The expected signs of the variables are as follows. According to previous literature, there is a negative relationship between environmental performance and cost of equity capital. For one unit increase in environmental performance, cost of equity capital will go down (Sharfman and Fernando, 2008; El Ghoul et al., 2011). An increase in firm size will have a decreasing effect on the cost of equity capital and leverage will have a positive effect on the cost of equity capital according to El Ghoul et al. (2011).
In researching (variables of) CSR, it is important to keep in mind that cost of equity capital is able to influence CSR. Therefore, time lags are needed to deal with the problems of reverse causality and endogeneity (Sharfman and Fernando, 2008; Cheng et al., 2014; El Ghoul et al., 2011). Reverse causality problems occur as firms take into consideration their cost of equity capital when determining whether to engage in environmental performance (El Ghoul et al., 2011). Waddock and Graves (1997) present the reverse causality hypothesis, namely: increased CSR performance enhances the relationship with stakeholders, resulting in better financial performance. Furthermore, the results of Waddock and Graves (1997) also indicate that better financial performance leads to increased CSR performance of the firm. Both hypotheses of
10 Whether running the regression analysis with fixed effects is applicable is tested by the Hausman test. The
15 Waddock and Graves (1997) are taken into account by el Ghoul et al. (2011), Cheng et al. (2014), Hong et al. (2011), and Sharfman and Fernando (2008), hence the problems of reverse causality and endogeneity should be taken seriously in research on this topic. Sharfman and Fernando (2008) found that one year of lagged environmental performance is sufficient to deal with endogeneity11. Cheng et al. (2014) have ruled out reverse causality in their research12. El Ghoul et al (2011) have gone to great lengths to ensure that reverse causality, omitted variable bias and endogeneity were not driving their results13. Overall, the main econometric equation of this thesis has been altered by lagging the environmental performance variable with one year, following Sharfman and Fernando (2008).
3.1 Cost of equity capital
As explained in section 2.2, numerous literature has uncovered a link between the cost of equity capital and environmental performance of a firm14. Cost of equity capital will be obtained from the capital asset pricing model (CAPM), following the work of Stulz (1999) and Sharfman and Fernando (2008). The model that is used in this research is the local CAPM model of Stulz (1999) :
𝑟𝐸𝑖, = 𝑟𝑓𝑐+ 𝛽𝑖(𝑟𝑚𝑐− 𝑟𝑓𝑐)
In which the 𝑟𝐸𝑖is the return on investment that investors demand as the reward, and hence this is the cost of equity capital for the firm. Next, 𝑟𝑓𝑐 is the risk-free interest rate which often is the
11 Sharfman and Fernando (2008) have run their analysis by lagging environmental performance between three to
zero years and by taking the average of three years and have obtained significant results solely for a one year lag.
12
Cheng et al. (2014) have divided firms in three groups ranging from low to highly capital constrained, compared to other firms in the same country and the same year, and found that the relation between CSR and capital constrains is the weakest for the least capital constrained firms, while it is the strongest for the most capital constrained firms. Hence, the luxury good argument does not seem to hold.
13
El Ghoul et al. (2011) have added the following variables related to CSR and financial constrains; the log of institutional investors, increases in CEO compensation when stock prices increased by 1 per cent, an antitakeover provisions index, log of number of analysts tracking the firm and an index of financial constraints. The results do not alter the baseline regression signs and significance levels. Furthermore, specifically for endogeneity, they have substituted the cost of equity capital for changes in cost of equity capital and have included the initial CSR score when entering the same into the equation, both do not change the significance and sign of their CSR coefficient, thus ruling out endogeneity.
14 On the other hand, the work of Lys et al. (2015) have indicated that first a firm’s management needs to
16 interest rate on government bonds or so-called treasury bills, specific to the local country. The average return in the local market is noted by 𝑟𝑚𝑐. Lastly, the variable that fluctuates per firm is the firm’s beta (𝛽𝑖). The beta of a firm indicates whether the stock price of a firm follows the movement of the market, it indicates the systematic risk vulnerability of the firm. When the firm’s stock price perfectly follows the market average, beta will have the value of 1. When the market returns increase, but the firm’s stock price decreases or increases less rapidly, beta has a value below 1, it is a ‘defensive’ stock. Conversely, when the firm’s stock price increases at a higher rate than the market, the beta will have a value above 1, it is an ‘aggressive’ stock. To sum up, the firm’s beta shows the correlation between the firm’s returns and the overall market return15.
The reason for choosing the local CAPM model (Stulz, 1999) is because credit ratings of a firm are often linked to sovereign credit ratings (Borensztein, Cowan and Valenzuela, 2013) (except for large global firms of which there are few, see Rugman and Verbeke, 2004). As the creditworthiness of a firm depends on the creditworthiness of its country, the country of origin16 would also affect the cost of equity capital. Furthermore, cost of equity capital is also linked to the credit rating of a firm (Lee et al., 1996), hence the link between country and cost of equity capital. Moreover, Fama and French (2012) state that global models are not applicable when one tries to explain regional differences in cost of equity capital. Distinguishing across country differences is the aim of the second hypothesis, hence the use of the local CAPM model is preferred.
During recent decades, many models have been developed that have improved the CAPM model or are able to predict cost of equity capital in more detail. These models include the Fama and French (1993) three-factor model, the Carhart (1997) four-factor model and the Fama and French (2015) five-factor model. Unfortunately, these models make usage of the SMB17 and HML18
15 The CAPM was revealed to be empirically flawed (Fama and French, 1992), but remains the best possible option
for researching my hypotheses, due to a lack of data.
16 The country of origin is the country in which the firm has its headquarters and is traded on the stock
exchange. Firms can be listed on multiple stock exchanges, however, solely data from the stock exchange of the country in which the firm is headquartered is used in this Thesis.
17 SMB stands for ‘small minus big’, which refers to the market capitalization of a firm. Fama and French
17 variables. The data on these two variables are available for the U.S. and the UK market, but not for other countries. As my thesis makes usage of data from over 35 countries, these models are not applicable to the research. This is one of the limitations that hopefully can be improved upon in future research.
3.2 Environmental performance
The main aim of this thesis is to examine the link between the increased environmental performance of a firm and the cost of equity capital. Environmental performance is often seen as one of the three pillars of CSR (see graph 3.1), the other two being social performance and corporate governance. The firm’s impact on natural systems (air, land, water and complete ecosystems) is tested. The environmental score measures the ability of the firm to avoid environmental risks and exploit environmental opportunities. The full environmental performance dataset contains 327 carefully collected variables and is composed of the following pillars: resource reduction, emission reduction and product innovation.
The resource reduction pillar focusses on the efficient usage of natural resources and reflects the ability of the firm to reduce material, energy and water usage, and improvements in the supply chain to become more eco-efficient. Emission reduction is focused on reduction of air emissions (e.g. F-gases, NOx, SOx, greenhouse gases and ozone-depleting substances). Furthermore, hazardous waste and spillage are measured, as is the firm’s overall impact on biodiversity. R&D efforts on inventing and producing environmentally friendly products or services is part of the product innovation pillar.
The overall environmental score is used to make sure that all of these variables influencing the environment are taken into account. The score reaches from 0 to 100, with 0 indicating very poor environmental performance and 100 excellent environmental performance. The scores for the three individual pillars are tested together with the first regression in order to make sure that the outcome of the regression does not depend on only one of these pillars. A grasp of which
18 HML stands for ‘high minus low’. According to Fama and French (1993), firms with a low price-to-book
18 variables have been taken into account in constructing this environmental performance index score19 can be found in the appendix (table A2).
Graph 3.1 The connection between CSR and environmental performance
3.3 Developed versus developing countries
In order to research hypothesis 2, a distinction must be made between developed versus
developing countries. Therefore, country dummies multiplied by environmental performance are included in the second regression. These interaction terms will show the effect of increasing environmental performance for a firm traded in a specific country on the cost of equity capital. These effects are on a country scale, hence these countries could then be indicated on their level of development based on their GNI per capita (obtained from the World Bank database20). Later on in the robustness checks section, dummy variables will be created based on the Human Development Index and the Doing Business rankings and are then interacted with environmental performance in order to see if these have a different impact on countries with a low index/ranking versus a high index/ranking.
19 The index has been composed by Datastream Asset 4 ESG professionals on the basis of all available information
within the Asset 4 ESG database.
20 In order to obtain data, please visit: http://data.worldbank.org/indicator/NY.GNP.PCAP.PP.KD
Corporate Social Responsibility Environmental Performance Resource
Reduction Reduction Emission innovation Product Social
19 3.4 Control variables
According to Gebhardt et al. (2001), it is important to control for firm size21 and financial leverage, as these two factors highly influence the cost of equity capital. Larger firms give out more information to the market (Gebhardt et al., 2001) leading to a lower information asymmetry and cost of equity capital. The natural logarithm of the market value of the firm is a proxy for firm size (Chen et al., 2009). For financial leverage of the firm, market leverage is used which is the ratio of long-term debt at t-1 and market value. Modigliani and Miller (1958) state that cost of equity capital is an increasing function of debt in the firm’s capital structure, which has been proven by Fama and French (1992), hence the relationship should be positive.
Moreover, the model should control for country, year and industry fixed effects. Country fixed effects are needed as business environments differ per country; e.g. the cost of capital in general may be higher overall for the country. Industry fixed effects control for differences across industries as some industries might be more prone to environmental management or simply have an overall higher/lower cost of equity capital, for example the nuclear fuel or forestry industries. Year fixed effects will take into account differences across years, which is especially helpful in the turbulent years after the financial crisis of 2007/2008.
3.5 Research design
The data used to answer my hypotheses is quantitative data obtained mainly from Thomson Reuters Datastream. As heteroscedasticity influences my data22, White’s robust standard errors are used to be on the conservative side regarding hypotheses testing. In order to assure that a fixed effects model suits my data, I’ve run two Hausman tests in order to check for the preferred model type: the first on fixed effects versus pooled OLS and the second for fixed versus random effects. The benefit of using a panel data model is that it allows for individual heterogeneity (coefficients of the variables are allowed to differ over time and/or per firm). The downside of using a fixed effects model is that variables that do not variate over time will cancel out (Hill et
21
Size is also one of the components used in the newer generation of models predicting stock returns (the Fama and French three-factor model etc.), however, the variable SMB is not available for all countries, hence solely the natural logarithm of market capitalization is used.
22 When running a modified Wald test for groupwise heteroscedasticity on a fixed effects regression model,
20 al., 2012). This would result in not being able to use country and industry fixed effects in my model. These time-invariant variables can be included in the random effects model as this model will subtract the time-mean (Hill et al., 2012). The first Hausman test allows me to reject the null hypothesis and conclude that fixed effects model is preferable over pooled OLS. Secondly, the Hausman test on fixed versus random effects shows that the random effects model is not appropriate as the null hypothesis is rejected, hence the fixed effects model is used. The downside of this is that country and industry fixed effects will cancel out in my econometric model as they do not change over time.
Hence, the econometric models to investigate my hypotheses will become:
𝑟𝐸,𝑖𝑡 =
𝛽1+ 𝛽2 𝑒𝑛𝑣𝑖𝑟𝑜𝑛𝑚𝑒𝑛𝑡𝑎𝑙 𝑝𝑒𝑟𝑓𝑜𝑟𝑚𝑎𝑛𝑐𝑒𝑖,𝑡−1+ 𝛽3ln(𝑠𝑖𝑧𝑒)𝑖𝑡+ 𝛽4 𝑙𝑒𝑣𝑒𝑟𝑎𝑔𝑒𝑖𝑡+ 𝜆𝑡+ 𝜀𝑖𝑡
And:
𝑟𝐸,𝑖𝑡 = 𝛽1+ 𝛽2 𝑒𝑛𝑣𝑖𝑟𝑜𝑛𝑚𝑒𝑛𝑡𝑎𝑙 𝑝𝑒𝑟𝑓𝑜𝑟𝑚𝑎𝑛𝑐𝑒𝑖,𝑡−1+
𝛽3Country dummy ∗ environmental performance 𝑖,𝑡−1+ 𝛽4ln(𝑠𝑖𝑧𝑒)𝑖𝑡+ 𝛽5 𝑙𝑒𝑣𝑒𝑟𝑎𝑔𝑒𝑖𝑡 + 𝜆𝑡+ 𝜀𝑖𝑡
Unfortunately, country and industry fixed effects have cancelled out due to the usage of a panel fixed effects regression as these variables do not change over time. However, the VIF test of multicollinearity of most of the dummy variables for the industry and country23 have high VIF values indicating possible multicollinearity. Therefore, omitting these variables due to using the fixed effects model is not such a large problem for my dataset. Furthermore, the previous literature that have used the country and industry fixed effects have used a pooled least squares model, which would increase the need of these fixed effects, while a panel fixed effects model allows to control for individual heterogeneity. When performing a Wald test on year dummies, I
23 The countries with the highest VIF values are the countries that have the most observations (U.S., Canada,
Japan, United Kingdom and Australia), see also table 4.1c for the sample distribution across countries. (H1)
21 am able to reject the null hypothesis that the coefficients of the year dummies are together equal to zero. Therefore, year fixed effects are needed in the regressions.
4. Data
4.1 Time period and countries
The sample size depends on the amount of data available from Datastream. See table 4.1a-c for the sample distribution across industry, year and country. The time period chosen for this study mainly depends on availability of the data. A period from 2009 till 2014 would result in the most complete dataset. Furthermore, the financial markets have been able to adapt to the economic and financial situation after the financial crisis of 2007/2008. The total dataset comprises data from 3,729 firms.
In the dataset, there is a high dependency on the manufacturing of plastics, leather, concrete, metal products, machinery, and equipment industry (SIC code 3), the finance, insurance, and real estate industries (SIC code 6) and the manufacturing of foods, textile, lumber, publishing, chemicals, and petroleum products industry (SIC code 2), as these three industries comprise more than half of the dataset (see table 4.1b). Furthermore, a few of the main developed countries dominate the dataset. This is due to the limited availability of data from less developed countries. The U.S., Japan, United Kingdom, Australia and Canada have the largest amount of observations, totaling to more than half of the dataset (62.47 percent of total observations), of which the U.S. has more than 25% of total observations.
Table 4.1a Sample distribution across years
22
Table 4.1b Sample distribution across industries
SIC Industry Categories N %
0 Agriculture, forestry, and fishing 114 0.51
1 Mining and construction 3,004 13.57
2 Manufacturing of food, textile, lumber, publishing, chemicals, and petroleum products
3,201 14.46
3 Manufacturing of plastics, leather, concrete, metal products, machinery, and equipment
4,291 19.38
4 Transportation, communications, electric, gas, and sanitary services 2,904 13.12
5 Trade 1,925 8.70
6 Finance, insurance, and real estate 4,201 18.98
7 Personal, business, and entertainment services 1,790 8.09
8 Professional services 709 3.20
Total 22,139 100
Table 4.1c Sample distribution across countries24
Country N % Australia 1,956 8.84 Austria 96 0.43 Belgium 162 0.73 Brazil 340 1.54 Canada 1,770 7.99 Denmark 156 0.70 Finland 150 0.68 France 588 2.66 Germany 522 2.36 Greece 30 0.14 Hong Kong 708 3.20 Hungary 24 0.11 Indonesia 210 0.95 Ireland 84 0.38 Italy 270 1.22 Japan 2,472 11.17 Korea 636 2.87 Malaysia 245 1.11 Mexico 174 0.79 Netherlands 198 0.89 New Zealand 102 0.46
24 Environmental performance data (Asset4 ESG data) is available for Chile, China, Czech Republic, Egypt,
23 Norway 108 0.49 Philippines 150 0.68 Poland 168 0.76 Portugal 50 0.23 Singapore 270 1.22 South Africa 750 3.39 Spain 264 1.19 Sweden 300 1.36 Switzerland 402 1.82 Taiwan 792 3.58 Thailand 198 0.89 Turkey 162 0.73 United Kingdom 1,818 8.21 United States 5,814 26.26 Total 22,139 100 4.2 Data collection
Many previous studies that have researched the linkage between the cost of equity capital (or firm performance in general) and environmental performance often focus solely on the United States. The KLD dataset has often been used in research, as has the TRI dataset. However, these datasets are unfortunately not applicable to answer the research question and investigate the hypotheses of this thesis as it only offers data on U.S. firms. Therefore, the method of Cheng et al. (2014) is followed and data is obtained from Thomson Reuters Datastream, specifically the ASSET 4 ESG dataset that focusses on CSR based data.
The ASSET 4 ESG dataset is panel based and offers environmental, social and governance (ESG) performance scores of firms. According to Cheng et al. (2014), investors managing €2.5 trillion in assets use the dataset for their investment decisions. Overall, the Thomsom Reuters Datastream ASSET4 ESG dataset offers information on more than 4,000 firms and composes 327 collected variables on environmental performance.
Regarding the information needed to calculate the cost of equity capital (the CAPM method), all variables are obtained from Datastream. For the risk-free interest rate25, the 3 month treasury bill
25
24 data of the home country is obtained. The beta26 of a firm is composed by Datastream who look at the volatility of the stock in relation to the volatility of the local market. The beta used is constant over the duration of the period examined (from 2009 – 2014) but differs by firm. Local market return is based on the local currency and on a high proportion of stocks traded in the market. See the Appendix, table A1 for the amount of stocks used per country to obtain the local market return.
Furthermore, the ISO country codes and SIC codes for industry classification are obtained from Datastream. The ISO country code is for the country in which the firm is headquartered and trades on the stock exchange. In case the firm trades on multiple stock exchanges, the headquarter country is used for the country code. For the SIC code, the business segment that provides the most revenue will be used to determine the industry to which the firm belongs. The market value of equity is computed by multiplying the share price of a firm (local currency) by the number of ordinary shares issued and obtained from Datastream. Long-term debt is composed of all interest bearing financial obligations, excluding items which are due within one year, also obtained from Datastream. All data is on a yearly basis.
4.3 Data characteristics
In table 4.2 the descriptive statistics of the dataset are presented. The CAPM has a mean of 5.80, indicating that the average return over 2009-2014 has been 5.80% on a yearly basis for the whole dataset. The returns have deviated tremendously, with losses of 158.95 percent to profits of 158.35 percent. Therefore, the CAPM data has been winsorized to insure that outliers do not drive the regression results. The winsorized CAPM has a slightly lower mean of 5.76 and lower standard deviation. The range of the winsorized CAPM has improved. The minimum CAPM is
http://www.dsta.nl/Actueel/Resultaten_emissies). For France, the data could not be obtained through the website of the responsible authority, hence, the data was obtained through http://uk.investing.com/rates-bonds/france-3-month-bond-yield-historical-data. For spain, the same method as used for the Netherlands applies, the website is: http://www.tesoro.es/en/deuda-publica/historico-de-estadisticas/subastas-2001-2014. Finally, Italy hardly gives out any 3-month t-bills, therefore 6-month t-bills are used and obtained through:
http://www.dt.tesoro.it/en/debito_pubblico/dati_statistici/rendimenti_composti_lordi_all_esmissione.html.
26 Datastream does not produce the beta coefficient for all countries. This is the case for many of the firms
25 now minus 47.94 and the maximum 56.01. the CAPM has been winsorized two-sided with 1% on both sides27.
Furthermore, the environmental performance differs between a score of 8.3 and 95.08, with an average of 51.48. The average natural logarithm of firm size is 23.05, resulting in an average market value of 10,244 million28 local currency units. The minimum market value is 0.5 million and the maximum is 304,594 billion (all converted back from natural logarithms to millions of local currency units). These numbers are hard to interpret as they are all in local currency units. One U.S. dollar is for example approximately 107 Japanese Yen29 or 1,161 South Korean Won. It is possible that the firm value outlier of 304,594 billion local currency units is in South Korean Won and thus would the firm would have a value of 262 billion U.S. dollars. The firm value in local currency is used as the CAPM is also based on the local market. These large numbers increase the need to use a natural logarithm of firm value to ease comparison between firms. The mean leverage is very low, at 0.001, indicating that for every local currency unit (e.g. dollar) of assets, there are 0.001 local currency units of debt and thus 0.999 local currency units of equity. A possible explanation for why this number is so low is that long-term debt is divided by market capitalization instead of total assets, following the research of Chen et al. (2009). In the robustness section (Part 5.2), two other measures of leverage are tested, namely total debt divided by total assets and long-term debt divided by total assets to see if these measurements make a difference in the sign and significance of leverage.
The CAPM has fluctuated a lot over the time period, see table 4.3. The CAPM had a mean of -21.35 during the financial crisis year in 2009, after that, stock markets have recovered but remained volatile. Environmental performance has also improved after the financial crisis, leading to (slightly) increasing overall scores.
27 Winsorizing a variable by 1% on both sides means that the 1% of the highest / lowest values lying on the
outer ends of both tails will be changed to the next value counting inwards from the extremes. Winsorizing the CAPM has been done to limit extreme outliers.
28 As firm size is computed by ln(market size), e23,05=10,244 million.
26
Table 4.2 Descriptive statistics
N Mean Std. dev. Min. Max.
Independent variables CAPM1 22,139 (22,139) 5.80 (5.76) 18.82 (17.75) -158.95 (-47.94) 158.35 (56.01) Dependent variables Environmental performance2 18,573 51.48 32.25 8.3 95.08 Emissions 18,573 51.42 32.35 7.33 95.01 Resources 18,573 51.75 32.26 6.81 95.28 Product 18,573 49.53 31.82 10.33 97.7 Control variables Firm size 21,316 23.05 2.59 13.12 33.35 Leverage 21,282 0.001 0.013 0 1.65
Notes: 1. CAPM numbers in brackets are for winsorized values of the CAPM.
2. All numbers for environmental performance and the variables it is composed off (emissions, resources and product) are lagged by 1 year.
Firm size is the natural logarithm of market capitalization. Leverage is the long term debt at t-1 divided by market capitalization. Std. Dev., Min., and Max. are abbreviations for Standard Deviation, Minimum, and Maximum respectively.
Table 4.3 Evolution of the CAPM and environmental performance
2009 2010 2011 2012 2013 2014 Mean CAPM1 -21.67 (-21.09) 19.18 (18.79) 4.07 (4.11) -0.22 (-0.18) 18.57 (18.23) 11.22 (11.16) 5.80 (5.76) Environmental performance 51.05 51.56 51.49 51.59 51.73 51.35 51.48
Notes: 1. CAPM numbers in brackets are for winsorized values of the CAPM.
27 against expectations30. There is also a negative correlation between firm size and leverage. Overall, there is nearly no relationship between the CAPM, environmental performance, firm size and leverage. There is somewhat of a relationship between firm size and the environmental performance of a firm.
Table 4.4 Correlation matrix31
(1) (2) (3) (4) (5) (6) (7) CAPM 1.00 Environmental performance -0.02 1.00 Product -0.01 0.84 1.00 Emission -0.02 0.93 0.65 1.00 Resource -0.02 0.93 0.65 0.85 1.00 Firm size 0.02 0.28 0.28 0.27 0.23 1.00 Leverage 0.00 0.01 0.02 0.00 0.00 -0.04 1.00
5. Analysis
This section will evaluate the results of the regression analysis. Firstly, the results of hypothesis 1 will be discussed, namely whether there is an relationship between environmental performance and cost of equity capital. These findings will be compared with the findings of Sharfman and Fernando (2008), El Ghoul et al. (2011) and Cheng et al. (2014), among others. Moreover, differences between countries are evaluated on the basis of the interaction between country dummies and environmental performance. Furthermore, robustness checks will ensure that the results provided by the first two regressions hold (part 5.2).
5.1 Results regression analysis: environmental performance and the cost of equity capital In table 5.1, the first hypothesis is tested. Firstly, model (1) tests the effect of environmental performance on the cost of equity capital without the usage of year fixed effects and control variables. The environmental performance variable is highly significant at the 1% level. However, the sign is positive, indicating that an increase in environmental performance increases cost of capital. The sign changes when year fixed effects are included in model (2) and the
30 Sharfman and Fernando (2008) have found that when environmental performance increases, leverage of the firm
also tents to increase. Hence, one would expect a high(er) relationship between environmental performance and leverage.
28 variable remains highly significant. It is important to note that much of the explanatory power of the model comes from adding in year fixed effects; R2 increases from 0.0127 to 0.6113. This indicates that much of the change in the cost of equity capital is due to differences by year. The CAPM is composed of yearly data that is country specific (the local market return and the risk-free interest rate both differ per year), hence it is not unexpected to see such a large increase in R2 between running a regression with versus without year fixed effects. Furthermore, the years after the financial crisis of 2008 have been very volatile, leading to large differences in the CAPM.
Models (3) and (4) of table 5.1 have included the control variables firm size and leverage. When looking at model (3) which does not include year fixed effects but does include control variables, we see that all variables are highly significant. However, the signs of the environmental performance and firm size variables are not as expected. From previous research, one would expect to see a negative sign for both of these. On the other hand, leverage has the expected sign and is significant; an increase in leverage will have a positive effect on the cost of equity capital according to El Ghoul et al. (2011). Adding the two control variables to the equation does increase the explanatory power of the model from 0.0127 to 0.0800. Furthermore, the intercept is highly significant and thus I am able to conclude that it is different than zero. Adding in year fixed effects in model (4) does give the expected signs, however, the two control variables have become insignificant. Hence, I am not able to say with certainty that these signs do hold. The environmental performance variable remains highly significant and has turned negative, following the findings of previous literature. Furthermore, the intercept remains significantly different from zero. Adding year fixed effects has led to a large increase in explanative power of the model, however, the R2 has decreased when comparing model (4) with model (2). This might be due to adding in the two control variables that are insignificant, reducing the overall explanative power of the model.
29 overall variable environmental performance is composed. Furthermore, the inclusion of the pillars do not alter the findings. From now onwards, solely the variable environmental performance will be used.
Table 5.1 Environmental performance and the cost of equity capital (Hypothesis 1)
ENV (1) ENV (2) ENV (3) ENV (4) ENV_PRO (5) ENV_RES (6) ENV_EMI (7) ENV 0.193*** (0.015) -0.049*** (0.011) 0.122*** (0.015) -0.050*** (0.011) -0.020*** (0.008) -0.038*** (0.009) -0.043*** (0.011) Size 11.148*** (0.464) -0.276 (0.315) -0.319 (0.315) -0.295 (0.315) -0.289 (0.315) Lev 57.587*** (6.009) 2.010 (6.758) 2.838 (6.939) 1.678 (6.749) 1.800 (6.798) Intercept -3.675*** (0.753) -19.467*** (0.595) -258.861*** (10.622) -13.096* (7.231) -13.515* (7.230) -13.250* (7.211) -13.129* (7.207) Year fixed effects
No Yes No Yes Yes Yes Yes
Entity fixed effects
Yes Yes Yes Yes Yes Yes Yes
N 18,573 18,573 18,510 18,510 18,510 18,510 18,510
R2 0.0127 0.6113 0.0800 0.6109 0.6103 0.6107 0.6108
Notes: ENV stands for environmental performance at time t-1, Size for firm size measured as the natural logarithm of market value, Lev for leverage measured as long-term debt at t-1 divided by market value. Models (1) and (3) tests hypothesis one but have no year fixed effects. Models (2) and (4) present the regression results corresponding to hypothesis one. Models (5), (6) and (7) report the effects of product innovation, resource reduction and emission reduction respectively instead of environmental performance in order to see if one of these factors mainly drive the overall results.
Robust standard errors reported in parentheses. * Statistical significance at the 10% level. ** Statistical significance at the 5% level. *** Statistical significance at the 1% level.
30 The pillar in which the company invests (product innovation, resource reduction or emission reduction) would all yield a similar negative reaction on the cost of equity capital, but the threshold of equity capital cost reduction fluctuates per pillar. If the firm is solely investing in environmental performance for the sake of reducing its cost of equity capital, one could argue that it is best to invest in emission reduction as the coefficient of this variable is the highest of the three. However, the overall environmental performance remains more important than solely focussing on one factor. Hypothesis one can be accepted on the basis of the test results provided in table 5.1.
The introduction of year fixed effects has a large impact on the sign of the variable environmental performance. Following the model of Heinkel et al. (2001) and Mackey et al. (2007) (see part 2.1 for an explanation of the model), firms are able to arbitrage the market for green investors, but there are costs involved in doing so. As no-one is able to say with certainty how the market will react during a certain year (the year dummy value cannot be predicted beforehand), and the general long-term tendency is that environmental performance will result in decreasing the cost of equity capital, the advice would be to keep investing in environmental performance. This is the case unless a firm would want to take on large costs that occur when changing its strategy, together with running the risk of losing the benefit of a reduction in the cost of equity capital.
31
Table 5.2 Environmental performance and the cost of equity capital including country interaction terms (Hypothesis 2)
32 New Zealand -0.257 (0.164) -0.135 (0.086) Philippines -0.783*** (0.183) -0.689*** (0.161) Poland -0.125 (0.111) -0.168*** (0.044) Portugal -0.746** (0.335) -0.392* (0.233) Sweden -0.194*** (0.074) -0.298*** (0.057) Singapore -0.331*** (0.059) -0.326*** (0.057) Thailand -0.737*** (0.113) -0.723*** (0.114) Turkey -0.482*** (0.090) -0.481*** (0.060) Taiwan -0.141*** (0.047) -0.117*** (0.034) South Africa -0.194*** (0.056) -0.266*** (0.063) Intercept -260.040*** (10.601) -13.721* (7.144)
Year fixed effects No Yes
Entity fixed effects Yes Yes
N 18,510 18,510
R2 0.0915 0.6190
Notes: ENV stands for environmental performance at time t-1, Size for firm size measured as the natural logarithm of market value, Lev for leverage measured as long-term debt at t-1 divided by market value. Model (1) reports the regression results for hypothesis two without year fixed effects. Model (2) adds in year fixed effects to the equation to test for hypothesis two. Models (3), (4) and (5) report the effects of product innovation, resource reduction and emission reduction respectively in order to see if one of these factors mainly drives the overall results. Solely the results for country*environmental performance interaction terms with a t-score higher than 1 are shown in the table as they are highly insignificant, leading to not including 4 interaction variables within the table (results for Austria, Belgium, Ireland and Norway have been omitted for this reason).
Robust standard errors reported in parentheses. * Statistical significance at the 10% level. ** Statistical significance at the 5% level. *** Statistical significance at the 1% level.
33 However, the findings are not in line with the findings from the regression results of hypothesis 1 that yield a negative relationship between the cost of equity capital and environmental performance when year fixed effects are included. Most of the highly significant interaction variables yield negative relationships between environmental performance and the cost of equity capital for the firms traded in that specific country.
Overall, by looking at the regression results excluding year fixed effects, I am able to conclude that firms traded in Brazil, the United Kingdom, Hong Kong, Indonesia, Korea, Mexico, Malaysia, Philippines, South Africa, Sweden, Singapore, Thailand, Turkey, Taiwan (all 1% significance), Australia, Portugal (both 5% significance), Canada, Hungary and the Netherlands (all 10% significance) all perceive benefits from investing in environmental performance as their cost of equity reduces by increased environmental performance. For firms traded in Italy (1% significance) and Denmark (10% significance), the cost of equity capital will increase when environmental performance increases. The interaction variables for firms traded in Austria, Belgium, Greece, Ireland, New Zealand, Norway and Spain are all insignificant.
In model (2) from table 5.2, the year fixed effects have been added. Here, the control variables firm size and leverage have become insignificant, as was the case in model (4) from table 5.1. The interaction terms for Australia, Canada, Switzerland, Germany, Finland, France, Japan and Poland have increased in significance. The interaction term of Portugal has decreased in significance from 5 to 10 percent. The interaction terms for Denmark, Italy and the Netherlands are no longer significant. Furthermore, all interaction terms that are significant (at either 1, 5 or 10 percent) provide proof for a negative relationship between environmental performance and the cost of equity capital for firms trading on the
34 country’s stock exchange, regardless of the year. To summarize, hypothesis two can be accepted on the basis of these results, as low GNI per capita (developing) countries have a higher benefit of investment in environmental performance than high GNI per capita (developed) countries, although firms traded in almost all countries benefit.
In graph 5.1 on the previous page, a visual representation of scale of mean GNI per capita is presented, including the countries in my dataset. The interaction variables for the Philippines, Indonesia and Thailand (low GNI per capita countries) show large negative relationships of -0.689, -0.626 and -0.723 versus smaller benefits for firms traded in the high GNI per capita countries Singapore, Switzerland, Sweden and Germany of -0.326, -0.125, -0.298 and -0.093 (Germany is significant at the 5 % level, the rest all at the 1% level). Hence, one would be able to conclude that for the low GNI per capita countries discussed above, investment in environmental performance would lead to a larger decrease in the cost of equity capital than is the case for the high GNI per capita countries. Firms traded on stock exchanges in low GNI per capita countries should be encouraged to increase their investments in environmental performance when the cost reduction on equity capital outweighs the cost of environmental investments.
5.2 Robustness checks
