Master Thesis Finance
Name: Sven van der Heide
Student number: s1913964 Study program: Master Finance
Supervisor: R.O.S. Zaal 16th of January 2014
The morality of the stock market
The possibility of adding CSR-scores to the beta, growth and size factors in a multifactor model for valuation purposes.
Abstract
This paper examines whether investors are willing to pay a premium to comply with their morality. Rational investors are expected to pay more for CSR firms ceteris paribus. In order to test this theory I identify the different factors that might mediate the relation between CSR and market performance. Subsequently I control for these factors by adding the following variables to the model: beta, market capitalization, price to book ratio, operating performance and changes in CSR. To further exclude any firm specific characteristics that might mediate the relation I employ a firm and time fixed effects model. As suggested by Ang and Schwarz (2010) I use an individual approach rather than a portfolio analysis to generate more precise estimates. This paper uncovers that the market is indeed willing to pay a premium for firms with a high CSR score and is unable to find significant relations between (a change in) CSR and operating performance and the beta. These results extend and are in line with the findings Dhaliwal (2001and 2014) and Ghoul (2011) that CSR is negatively related with cost of capital.
Keywords: Corporate social responsiblity (CSR), Socially Responsible Investing (SRI), Delta CSR, Asset Pricing, Morality, Fixed Effects Model.
Table of Contents
1. Introduction ... 1
2. Literature review ... 5
2.1 Theoretical importance of CSR for society ... 5
2.2 Relation between CSR and operational performance. ... 5
2.2.1 Theoretical relation between CSR and operational performance. ... 5
2.2.2 Empirical studies on the relation between CSR and operational performance ... 7
2.3 Relation between CSR and discount rate. ... 8
2.3.1 Theoretical relation between CSR and discount rate. ... 8
2.3.2 Emperical studies on the relation between CSR and discount rate ... 9
2.4 Hypotheses ... 10
3 Methodology ... 13
3.1 Choice for a firm and time fixed effects model ... 13
3.2 Model 1: relation between (delta) CSR and a firm’s market returns ... 13
3.3 Model 2: Relation between (delta) CSR and next year’s beta ... 16
3.4 Model 3: relation between (delta) CSR and operating performance ... 17
4. Sample ... 18
4.1 Data... 18
4.2 Descriptive data ... 19
5 Results ... 23
5.1 Relation between (delta) CSR and market returns (model 1) ... 23
5.2 Relation between (delta) CSR and next year’s beta (model 2) ... 25
5.3 Relation between (delta) CSR and operating performance (model 3) ... 25
5.4 Robustness analyses ... 30
6 Conclusion ... 32
7 Sources ... 34
1.
Introduction
The stock market is often associated with rational and/or greedy investors who look specifically at profits and risk, and is less often linked to socially conscious investors who are willing to sacrifice return for moral considerations. Consumers appear to be more morally conscious, as corporate social responsibility (CSR) is now one of the most influential factors when forming an opinion of a company (Dawkins & Lewis, 2003). In multiple countries, a majority of the people indicate that they are willing to pay a premium for environmentally and socially responsible products (Deloitte, CSR Europe & Euronext, 2003). This, and other factors, has led many companies to embrace and be transparent about their CSR policy. In 2013, more than 5000 companies worldwide have voluntarily disclosed CSR reports (Kell, 2014). Also on the side of investors CSR comes into play, illustrated by the surge in social responsible investments (SRI) in the last decade. According to data from the The Forum for Sustainable and Responsible Investment (US SIF) in 2011 3314 billion USD (from 166 billion USD in 1995) was invested following a SRI screening. However, these investments do not necessarily result from moral considerations of the investors. Research conducted by Nilsson, Jansson, Isberg, and Nordvall (2014) indicates that the most important predictors of the satisfaction of investors in SRI funds is financial performance. Although perceived environmental and social performance is positively related to satisfactions it is unlikely to generate satisfied investors. This is also demonstrated by institutional investors, such as APG, which state that investments in firms with a focus on CSR perform better in the long run.1 Numerous studies support this statement and find a positive relation between CSR and corporate financial performance (Orlitzky, Schmidt, & Rynes, 2003; Dixon-Fowler, Slater, Johnson, Ellstrand, & Romi, 2013). Thus although APG invests in firms that focus on CSR, the investment itself is not necessarily driven by moral considerations because the investor might benefit from better prospects of these firms. Nevertheless, it seems logical that any rational and responsible investor will choose a CSR firm (a firm with a higher than average CSR score) over a non-CSR firm (a firm with a lower than average CSR score), ceteris paribus, thus with equal risk and return. This creates additional demand for the shares of CSR firms driving up the share price, or driving down the return of investment for that firm. Eventually investors in CSR firms need to pay a premium in return for non-financial utility that they derive from this socially responsible investment. A survey of European fund managers and financial analyst initiated by Deloitte et al. (2003) supports this theory; respectively 51% and 37% of the Dutch managers and analysts indicate that they are willing to “grant a premium to companies which are environmentally and socially responsible”. This study aims to investigate whether this
1
effect can actually be observed on the stock market. Therefore this paper seeks an answer to the following question: are investors willing to pay a premium (sacrifice return) for a social responsible investment to comply with their morality (i.e. gain non-financial utility).
Research on this topic is complicated due to different, opaque ways in which CSR impacts market performance. Market performance comprises changes in stock price and distributed dividend in one year. According to the net present value model share prices are expected future cash flows discounted at an appropriate rate (Williams, 1938). It is therefore required to evaluate the different theories about the impact of CSR on the discount rate and on the cash flows (operational impact) in the short and long term. One of these theories posits that operational returns may vary as a result of increased sales to newly attracted customers, higher costs related to the investment necessary to implement CSR or lower costs due to more efficient utilisation of resources (Dhaliwal, Li, Tsang & Yang, 2011). The empirical studies on the effect of CSR on the operational results in the short and long term also show mixed, but mostly positive results (Orlitzky et al., 2003; Flammer, 2013). The discount rate, on the other hand, has a negative relation with CSR (Dhaliwal et al., 2011 and 2014; Ghoul, Guedhami, Kwok & Mishra, 2011). This negative relation is attributed to a lower perception of risk and/or a larger investor base.
Figure 1. The potential effects of CSR on a firm’s market performance (CFP). The effects of CSR on market performance as described in other papers are indicated in bold.
Currently, the models for valuating firms and evaluating market performance, such as Fama and French’s three factor model (1993) and the capital asset pricing model (Sharpe, 1964), look mainly at risk and return and lack a morality component. Both these models use the covariance of the stock’s movements with the market, the beta, as a proxy for measuring risk. Fama and French’s three-factor model adds two additional factors: the growth (price to book ratio) and the size (market capitalization) factor. Although literature is not unambiguous about the purpose of these factors: the factors could be explained as proof that CAPM is oversimplified, missing additional risk-factors or as proof of an inefficient stock-market due to irrational investors (Fama & French 2004). Neither of these theories captures a possible morality of the stock market. Thus, the models do not incorporate the idea that an investor derives utility from an investment in a CSR firm and is therefore willing to pay a premium.
sacrificing firm value because CSR score drives down the discount rate, which offsets potentially lower operating returns related to the CSR project. Third, governments can reduce legislation in favour of better information-based instruments in this area as the market can regulate itself in a more efficient manner.
The study contributes in different ways to existing literature. Most importantly, and as suggested by Aguines and Glavas (2012), it distinguishes the different links of CSR with market performance. Unique is that the study emphasizes the importance of morality as a utility adding factor which might impact market performance. To my knowledge, this study is the first to take- changes in CSR-scores into account separately, which provides a clearer view of the mechanism behind the relation. To further extract potential mediating effects it adds several control variables(changes in operating income, beta, size and growth effects) and employs a firm and year fixed effects model. In addition, it assesses stocks individually rather than aggregated in portfolios. According to Ang and Schwarz (2010), this approach provides a more efficient test of whether factors are priced. It extends the work of Dhaliwal et al. (2011) and 2014) and Ghoul et al. (2011) who found a negative relation between CSR and the implied cost of capital, by examining whether this link can also be observed on the market. As the effect of CSR on consumer attitude towards the firm depends heavily on the sector, location and other factors (Ambec & Lanoie, 2008), the effect of CSR on operational performance also depends on these factors. In contrast, the stock market is far more homogeneous, therefore the results of this study regarding the effect of CSR on the discount rate is likely to be more broadly applicable. In addition, the study uses a new, recent, transparent and extensive database consisting of 4000 firms located in different countries. The study looks at risk-adjusted returns and in contrast to other studies, this study does not assess returns of fund managers as the skills of the managers might impact the return of the CSR portfolio.
2.
Literature review
In literature different theories exist that explain and forecast the effect of CSR on a firm’s market performance. Market performance depends on the distributed dividends of a firm during the year and the difference in the value of that firm at the end of that year relative to the value at the beginning of that year. Mostly, firms are valued using the net present value model. This model discounts cash flows at an appropriate discount rate. Most of the theories regarding the effect of CSR are related to the operational performance of the firm, so the way in which CSR affects cash flows. Although the topic of this research focuses on the effect of CSR on the discount rate, it is nevertheless important to touch upon the theories related the operational side. These theories are not only relevant to get a better feel for the subject, but they are also essential for effectively extracting the morality component from other implications of CSR.
2.1
Theoretical importance of CSR for society
Although it might seem clear that focusing on CSR positively affects society, Friedman (1970) argues otherwise and states that “there is one and only one social responsibility of business to use its resources and engage in activities designed to increase its profits so long as it stays within the rules of the game, which is to say, engages in open and free competition without deception or fraud.” This implicates that companies should focus primarily on their own interests while governments should deal with provisions of public goods and externalities. If companies focus on their own interests, competition between companies increases, which lowers prices and enables the public to decide for themselves to donate to a charity organisation of choice. However, this theory is flawed, as it assumes that property rights are defined so that no external effects exist. In reality, externalities do exist, and it would require additional efforts of charity organisations to deal with the externalities related to the activities exerted by companies who strive to maximize shareholder value. This induces several inefficiencies: it is generally more costly to mitigate externalities than to prevent them, it creates more agency costs and it might induce free rider problems (Statman, 2008).
2.2 Relation between CSR and operational performance.
Even in a world where no externalities exist, companies could still implement CSR as long-term market value is related to the treatment of many stakeholders. This view is supported by numerous theories that cover the operational consequences of CSR
2.2.1 Theoretical relation between CSR and operational performance.
links, three related to the revenue side and four related to the cost side. They argue that improving CSR scores can increase customer loyalty and thereby can support sales. Especially public administrations might be more willing to buy from CSR companies as guidelines prescribe to take environmental performance into account when evaluating potential suppliers. CSR might also help to exploit niches in environmentally conscious markets by creating new product categories, which is illustrated by the surge of revenues generated by products with an eco-label. This market increased from 51 million Euro in 2000 to 644 million Euro in 2004. A third possibility to increase revenues, although of less importance, concerns the development of technology required to improve CSR. These newly developed technologies can be sold, and can thus generate profits. Implementation of CSR might also result in costs reduction. By continuously complying with applicable legislation, liability costs could be reduced and fines and litigation can be avoided. Following Porter’s Hypothesis (1995), Ambec and Lanoie(2008) argue that pollution is an incomplete utilisation of resources. Hence, resources can be utilised more efficiently by reducing pollution, which generates a reduction of expenditures on energy and materials. Also, by improving CSR, the image of the company improves. This motivates not only employees to perform better, but also stimulates employees to act more like ambassadors. So it might reduce costs related to illnesses, absenteeism, recruitment and turnover. Last, the element most closely related to the actual subject of this thesis, is the implications CSR has on costs of capital. Ambec and Lanoie (2008) argue that CSR facilitates a company’s access to finance as it opens the doors to corporate social responsible investors. Also, banks are no longer required to evaluate environmental performance to assess potential litigation risk, which might reduce interest costs. Third, the additional disclosures related to CSR efforts increase awareness of the firm and positively influences an investor’s perception of the firm. These links are summarized in table 1.
Possibilities to Increase Revenues Possibilities to Reduce Cost
i) Better Access to Certain Markets iv) Regulatory Cost
ii) Possibility to Differentiate Products v) Cost of Material, Energy and Services iii) Selling Pollution-Control Technologies vi) Cost of Capital
vii) Cost of Labour
Table 1. The links between CSR and economic performance.
stakeholders into account. This enables the firm to adapt more easily to external demands and reduces the likelihood that a firm relies too heavily on a small group of stakeholders (Freeman & Evan, 1990). In contrast to the above-mentioned theories, slack resources theory posits that the relation between CSR and market performance is reversed: a firm with good financial performance, or good foresights, might be more willing to implement corporate responsible activities (Waddock & Graves, 1997).
2.2.2 Empirical studies on the relation between CSR and operational performance
Many studies have tried to empirically test if CSR affects a firm’s corporate financial performance (CFP) using different measures. Orlitzky et al. (2003) conduct a meta analysis consisting of 52 studies yielding a total sample size of 33,878 observations. Overall, results show a positive and significant correlation between CSR and CFP. In addition, they divide the studies in subsets based on the measure used to assess a firm’s Corporate Social Performance (CSP disclosures, CSP reputation rating, social audits, managerial CSP principles and values) and a firm’s CFP (market-based, accounting based (ROA, ROE, EPS)) to investigate whether the relation is similar across subgroups. The study indicates that the correlation depends on the operational construct of CFP; accounting based indicators show higher correlation than market-based equivalents. However, different measures of CSP are highly correlated, which refutes speculation about the differences in outcomes of previous studies. Furthermore, the findings suggest that corporate environmental performance has a limited relationship with CFP compared to other measures of CSR. They argue that CSP is positively correlated with managerial competencies, contributes to organizational knowledge about the firm’s market, social, political, technological, and other environments, and thus enhances organizational efficiency. CSP also improves the reputation and goodwill with its external stakeholders. These aspects are therefore considered as potential mediating effects between CSR and CFP.
In line with Orlitzky et al. (2003), they find that deviations in the methodology applied in the various studies do not result in significant differences. In addition they find that it does not matter whether CFP is measured simultaneously or lagged, which indicates that the effects of CSR are incorporated gradually. In contrast to Orlitzky et al. (2003) they find that market-based measures show a larger effect on CEP than other performance measures.
Findings of a recent study conducted by Flammer (2013) also support a positive relation between CSR and operational performance. She indicates that, as suggested by Lanoie and Ambec (2008), labour productivity increases and sales grow after the board adopts CSR policies. These studies support the existence of a positive relation between CSP and CFP, measured with both accounting- as market based returns. The effect CSR has on accounting based returns indicates that operational returns are indeed affected by implementing CSR. It is difficult and probably not feasible nor useful to analyse through which links CSR is exactly affecting the operational side of a company as it depends on the industry, the consumers the properties of the firm, and other political and social demographic characteristics (Lanoie and Ambec, 2008).
2.3
Relation between CSR and discount rate.
The mechanisms through which CSR affects market performance are unclear as well. As stated, market performance depends on future cash flows and the appropriate discount rate. So an increase in market return can be attributed to a change in expected future cash flows or a change in the applicable discount rate. As this study is interested in the morality of the market, to which extent the market is willing to pay a premium for shares of CSR firms, changes in share price associated with non-moral considerations have to be excluded. Therefore examining the links through which CSR might impact the discount rate is essential.
2.3.1 Theoretical relation between CSR and discount rate.
of irresponsible behaviour on potential, future litigation risk, which in turn might lead to lower volatility, hence a lower beta. Starks (2009) suggests that CSR alters a firm’s risk profile by decreasing risk associated with regulatory, supply chain, litigation, product and technology risk.
2.3.2 Emperical studies on the relation between CSR and discount rate
Multiple studies evaluate these theoretical arguments empirically, including a meta-analysis conducted by Renneboog, Ter Horst and Zhang (2008). This study suggests that the explosive growth in investments in CSR companies (social responsible investment (SRI)) should impact stock return of CSR firms as risk is diversified between more investors. They look at previous studies that analysed the risk-adjusted returns of SRI. The results are mixed: in the US and UK there is little evidence that risk-adjusted returns of SRI differ from conventional funds. However, in continental Europe and Asia SRI funds underperform compared to regular funds supporting the view that investors are willing to accept a lower return. This view is supported by the fact that investments in SRI funds are less sensitive to lagged negative returns compared to conventional funds, indicating SRI investors include other factors in their investment decision (Bollen, 2007). Renneboog et al. concludes that the puzzle of the effect of CSR on firm performance remains unsolved. Mollet and Ziegler (2014) confirm the findings of Renneboog, using a more recent dataset that covers both European and US stock market from 1998 to 2009. They correct for risk, size, value, and momentum factors and conclude that the prices of CSR firms are therefore correctly priced, that is, the stocks are neither overpriced nor underpriced
Additionally they find that non-financial disclosing has similar, moderating effects on firm performance.
Ghoul et al. (2011) also examine the effect of CSR on the ex-ante cost of equity for large US-firms. They use ex-ante cost of equity implied in analyst earnings forecast and stock prices because these forecast are not subject to noisy realized returns and it allows them to control for differences in growth rates. They also control for fixed year and industry effects as well as various firm-specific determents. In line with Dhaliwal et al. 2011, they find that improving employee relations, environmental policies and product strategies lower cost of equity. The results suggest that CSR in employee relations, environmental policies and product strategies are most effective in lowering firms’ cost of equity, while there is no effect in community relations, diversity and human rights. Overall they conclude that higher CSR scores significantly lower cost of equity.
With regard to the relation between CSR and the perceived risk of a firm, the findings of Feldman, Soyka and Ameer (1997) are relevant. They show that investors indeed perceive firms focusing on CSR as less risky. Luo and Bhattacharya (2009) find that CSR is associated with a decrease in a firm’s idiosyncratic risk. This does not necessary mean that CSR decreases the beta as well, because investors might be able to diversify the risk away. Hughes, Liu and Liu (2007) do find evidence that disclosure of CSR practices reduces the covariance of a firm’s cash flow and thereby a firm’s beta.
2.4
Hypotheses
Classification Altered revenues Altered costs Net operating effects: Better Access to Certain
Markets
Regulatory Cost Possibility to Differentiate
Products
Cost of Material, Energy and Services Selling Pollution-Control
Technologies
Cost of Labour Organizational efficiency Compliance costs Non-moral considerations of the investors Estimated riskiness Searching costs Better prospects
Morality premium Larger investor base,
Preference for CSR companies, Lower required rate of return Table 2. Summary and classification of effects of CSR on market performance.
By controlling for the potential operational effects of CSR this paper aims to uncover the effect of CSR on the discount rate. The discount rate of firms might be affected in two ways: through a premium investors are willing to pay for CSR firms and through a lower perception of risk. The former, in line with Merton’s Framework, is expected to show a negative relation with the discount rate: a higher CSR score decreases cost of capital and thus market performance. Hence, I postulate the following hypothesis :
H1: investors are willing to pay a premium to invest in a firm with a higher CSR score compared to a firm with a lower CSR score.
As stated, a change in CSR might be associated with a lower perception of risk. This should result in a lower volatility of share prices, and therefore might impact the beta of a firm(Hughes et al., 2007). This is tested with the second hypothesis:
H2: A change in CSR scores reduces the beta of a firm.
Third, this paper aims to evaluate the effect of CSR on operating performance. Previous studies find ambiguous results but the meta analysis of Orlitzky et al. (2003) and Dixon-Fowler et al. (2013) reveals that the relation between CSR and operating performance is predominately positive. Also, according to the slack resource theory good financial prospects enables a company to invest in CSR, indicating that operating performance is (positively) related to CSR. This leads to the following hypothesis:
Fourth, the different effects that CSR has on market performance are expected to be incorporated in the share price at the moment of the change in CSR. There are several factors that might affect market performance positively: increased revenues and decreased costs on the operational side, and reduced estimated risk and a premium for CSR firms on the financial side. However, operating performance can also be affected negatively, as improving CSR might require investments in durable assets. The net effect of CSR on operational performance is translated in a change in the expected future cash flows, which translates in a higher share price according to the discounted cash flow model. The impact CSR might have on the estimated risk and morality premium are included in the discount rate. Therefore, if an investor acknowledges the impact of CSR on these factors directly after the change occurs, all these effects can immediately be incorporated in the price. According to the theories regarding these effects, a change in CSR results in net increased performance and a decreased discount rate. Hence the following hypothesis is formulated:
H4: An increase in a firm’s CSR score results in a one-time increase of market performance.
Figure 2 summarizes and illustrates all the potential effects of CSR on market performance.
Figure 2. This graph illustrates the potential effects of 50% increase in CSR score in year 2 on several variables.
With on the the first verticale axis CSR score, net income, market capitalization (divided through 10) and the CSR score, on the secondary axis the instric discount rate (which represent s the riskfree rate, the size and growth effect) , the beta, the morality premium (which is negative) and the discount rate are which is the cumulation of these factors are depicted. Assuming rational, responsible investors, no growth other than the growth as a result of the implementation of CSR, this increase in CSR of 50% increases the growth rate, lowers the beta, increases the morality premium(hence lowers the discount rate). This results in a time surge in market returns after which it stabilizes at a level that is lower than the initial discount rate.
0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16 0.18 0.2 0 20 40 60 80 100 120 140 160 1 2 3 4 5 6 7 8 9 10 11 Morality Premium Beta
Intristic Discount rate CSR
3
Methodology
3.1
Choice for a firm and time fixed effects model
In contrast to the majority of the literature regarding the effect of CSR on market performance that use a Fama Macbeth regression (Petersen, 2009), I employ a firm and time fixed effect model to assess the returns of individual firms. This model has several advantages over other models (such as Fama Macbeth or OLS (with and without Newey West standard erros); it results in efficient estimates for beta and is unlikely to have biased estimates for the standard error if the firm fixed effect are present (Petersen, 2009 and Ang & Schwarz, 2010). Although the estimates for the beta of individual firms are subject to more noise and capture idiosyncratic risk (problems that Fama Macbeth regression mitigate), it does not shrink the cross sectional dispersion of beta, which results in a less efficient estimate for the beta due to a higher standard error. According to Ang and Schwarz (2010), the potential efficiency losses related to the Fama Macbeth regression are severe and outweigh the gains of the approach related to decreasing idiosyncratic risk. In addition the Fama Macbeth regression results in downwards biased estimates for the standard error if firm fixed effects are present (e.g. leverage, dividend payout and investments). Although theory rejects firm fixed effect, the Hausman test indicates that the model is appropriate.2 Thus by employing a time and firm fixed effect model for all my models I ensure that the results are unbiased and efficient (Petersen, 2009).
3.2
Model 1: relation between (delta) CSR and a firm’s market returns
My first model estimates the effect of CSR and changes in CSR on the firm’s market return. Following related literature (e.g. Dhaliwal et al. 2011 and 2014; Mollet & Ziegler 2014) I include the factors developed by Fama and French: the beta, size and growth. The beta corrects for non-diversifiable risk of each corporation, and the size- and growth (firms with a high market to book ratio) component for anomalies found by Fama and French of underperforming large cap and growth firms. The proxy I use for environmental performance (further explained in section 4.1) reflects how well a company uses best management practices to avoid environmental risks and capitalize on environmental opportunities. Thus, in line with Ghoul’s Theory, this proxy might result in a lower beta for the firm and therefore an increase in share prices. By including the beta I control for the possible associated errors. Cortez and Silva (2011) find evidence that socially responsible funds are strongly exposed to small cap and growth stocks, therefore I add size and growth as control variables. Normally, studies include HML (return of the 50% firms with the highest P/B ratio and market capitalization subtracted from the 50%
2
firms with the lowest P/B ratio) and SMB (market capitalization subtracted from the 50% firms with the lowest P/B ratio and market capitalization) to assess whether a portfolio is exposed to size or growth firms. Values for HML and SMB are usually obtained from Kenneth French’ website3 (see Mollet & Ziegler, 2014). Therefore these studies do not test whether portfolios containing firms with small market capitalization outperform portfolios containing firms with large market capitalization, but merely estimate the coefficients for the HML/SMB factor and based on that coefficient they determine whether a size or growth factor is present within the portfolio. For individual firms this approach is even more questionable, as the coefficient functions basically as a proxy for the size and growth factor. Because I use individual stocks it makes sense to determine the size and growth factors based on the actual market capitalization, and P/B ratio (further explained in section 4). Model 1.1 only contains the factors of CAPM, Model 1.2 does include Fama and French 3 factors:
Model 1.1: 𝑅𝑖,𝑡𝑒 = α𝑖+ (𝐶𝑆𝑅𝑖,𝑡)𝛽1+ 𝐵𝑒𝑡𝑎𝑖,𝑡𝑅𝑚,𝑡𝑒 𝛽2 + 𝜀𝑖,𝑡
Model 1.2 : 𝑅𝑖,𝑡𝑒 = α𝑖+ (𝐶𝑆𝑅𝑖,𝑡)𝛽1+ 𝐵𝑒𝑡𝑎𝑖,𝑡𝑅𝑚,𝑡𝑒 𝛽2 + 𝑆𝑖𝑧𝑒𝑖,𝑡𝛽3+ 𝐺𝑟𝑜𝑤𝑡ℎ𝑖,𝑡𝛽4+ 𝜀𝑖,𝑡 With(𝐶𝑆𝑅𝑖,𝑡) as a score for CSR for firm i in year t,𝑅𝑖,𝑡𝑒 and 𝑅
𝑚𝑡𝑒 as the excess return (𝑅𝑖,𝑡− 𝑅𝑓𝑡)
and (𝑅𝑚,𝑡− 𝑅𝑓𝑡). Where (𝑅𝑖,𝑡) is the stock returns for firm i in year t, (𝑅𝑚,𝑡) is the market returns for market m in year t and 𝑅𝑓𝑡 is the risk free rate in year t. 𝐵𝑒𝑡𝑎𝑖,𝑡 is the covariance of the of the firm’s shares with the market divided through the variance of the market. 𝑠𝑖𝑧𝑒𝑖,𝑡 and 𝑔𝑟𝑜𝑤𝑡ℎ𝑖,𝑡 represent respectively the natural logarithm of the market capitalization and the
square root of the absolute value of the market to book ratio. 𝜀𝑖,𝑡 is an error term.
According to the studies mentioned in the literature review (summarized in table 2 and graph 1), a change in CSR potentially impacts (future) financial returns in numerous ways. If (future) financial returns are indeed affected, rational, forward looking investors can react to this development immediately by incorporating this forecast in the discounted cash flow model. This consequently results in a one-time surge in market returns at the time of the change in CSR. As mentioned, Ghoul et al. (2011) argue that perceived riskiness is affected by CSR. In case of rational, forward looking investors the change in perceived riskiness is directly incorporated in the price, and causes the share prices to increase (future returns are discounted at a lower rate). However, the beta is calculated based on a two year period, therefore, a change in perceived riskiness is not directly included in the beta. This might positively affect the coefficient of CSR and thus induce a biased estimate. The same logic is
applicable for a possible effect of CSR on the required return due to the premium investors are willing to pay. The required return goes down, resulting in a rise in share price.
I account for both these factors by adding the change in CSR score: delta CSR (Δ𝐶𝑆𝑅𝑖𝑡,𝑖𝑡−1) as a variable to the model. The effect of a change in financial forecast, perceived risk and morality on market performance occurs at the moment of the change in CSR, thus delta CSR prevents possible biases to enter the morality component (CSR). Delta CSR can therefore be interpreted as the net effect of a change in CSR. The following model is constructed:
Model 1.3: 𝑅𝑖,𝑡𝑒 = α𝑖+ (𝐶𝑆𝑅𝑖,𝑡)𝛽1+ (Δ𝐶𝑆𝑅𝑖,𝑡;𝑖,𝑡−1)𝛽2+ 𝐵𝑒𝑡𝑎𝑖,𝑡𝑅𝑚,𝑡𝑒 𝛽3 + 𝑆𝑖𝑧𝑒𝑖,𝑡𝛽4+ 𝐺𝑟𝑜𝑤𝑡ℎ𝑖,𝑡𝛽5+ 𝜀𝑖,𝑡
With (Δ𝐶𝑆𝑅𝑖𝑡,𝑖𝑡−1) is the CSR score at the end of year t-1 subtracted from the CSR score at the end of year t.
Model 1.4a: 𝑅𝑖,𝑡𝑒 = α𝑖+ (𝐶𝑆𝑅𝑖,𝑡)𝛽1+ 𝐵𝑒𝑡𝑎𝑖,𝑡𝑅𝑚,𝑡𝑒 𝛽2 + 𝑆𝑖𝑧𝑒𝑖,𝑡𝛽3+ 𝐺𝑟𝑜𝑤𝑡ℎ𝑖,𝑡𝛽4+ Δ𝑅𝑂𝐴𝑖,𝑡;𝑖,𝑡−1𝛽7+ 𝜀𝑖,𝑡 Model 1.4b: 𝑅𝑖,𝑡𝑒 = α𝑖+ (𝐶𝑆𝑅𝑖,𝑡)𝛽1+ 𝐵𝑒𝑡𝑎𝑖,𝑡𝑅𝑚,𝑡𝑒 𝛽 2 + 𝑆𝑖𝑧𝑒𝑖,𝑡𝛽3+ 𝐺𝑟𝑜𝑤𝑡ℎ𝑖,𝑡𝛽4+ Δ𝑂𝐼𝑖,𝑡;𝑖,𝑡−1 𝑀𝑎𝑟𝑘𝑒𝑡 𝐶𝑎𝑝𝑖,𝑡𝛽8+ 𝜀𝑖,𝑡
With Δ𝑅𝑂𝐴𝑖,𝑡;𝑖,𝑡−1 as the change in ROA (net income/total assets) measured at the end of year t compared to the ROA at the end of year t-1 and Δ𝑂𝐼𝑖,𝑡;𝑖,𝑡−1
𝑀𝑎𝑟𝑘𝑒𝑡 𝐶𝑎𝑝𝑖,𝑡 as the change in operating income in year to compared to year t-1 divided through market capitalization.
3.3
Model 2: Relation between (delta) CSR and next year’s beta
If indeed delta CSR is positively associated with market returns this might attributed to a change in expected cash flows or/and a change in the discount rate. Model 3 examines the former effect and model 2 looks at the former. More specific, model 2 evaluates the relation between a change in CSR score and the beta in the year after the change.
A positive coefficient of delta CSR, suggests that investors incorporate changes in risk directly after the change in CSR occurs. In addition I will check whether CSR affects the beta in the long run, by adding the CSR score as a variable to the model. Because a change in beta might also be associated with other factors, such as a change in the market capitalization of the firm, a change in the P/B ratio and changed financial performance, those variables are added to control for the possible disturbances. Similar to the previous regression, two alternative versions are constructed to control for changes in financial performance: delta NI/assets and delta ROA.. The following two regressions are estimated:
Model 2a: 𝐵𝑒𝑡𝑎𝑡+1 = α𝑖+ (𝐶𝑆𝑅𝑖,𝑡)𝛽1+ Δ𝐶𝑆𝑅𝑖,𝑡;𝑖,𝑡−1𝛽2+ Δ𝑆𝑖𝑧𝑒𝑖,𝑡;𝑖,𝑡−1𝛽3+ Δ𝐺𝑟𝑜𝑤𝑡ℎ𝑖,𝑡;𝑖,𝑡−1𝛽4+ Δ𝑅𝑂𝐴𝑖,𝑡;𝑖,𝑡−1𝛽5+ 𝜀𝑖,𝑡
Model 2b: 𝐵𝑒𝑡𝑎𝑡+1= α𝑖+ (𝐶𝑆𝑅𝑖,𝑡)𝛽1+ Δ𝐶𝑆𝑅𝑖,𝑡;𝑖,𝑡−1𝛽1+ Δ𝑆𝑖𝑧𝑒𝑖,𝑡;𝑖,𝑡−1𝛽3+ Δ𝐺𝑟𝑜𝑤𝑡ℎ𝑖,𝑡;𝑖,𝑡−1𝛽4+
Δ𝑂𝐼𝑖,𝑡;𝑖,𝑡−1
𝑀𝑎𝑟𝑘𝑒𝑡 𝐶𝑎𝑝𝑖,𝑡𝛽5+ 𝜀𝑖,𝑡
3.4
Model 3: relation between (delta) CSR and operating performance
Finally, the study evaluates the effect of a change in CSR score on the financial performance. In this model I add both size, growth, beta and changes in those variables as control variables. This results in the following models:
4.
Sample
4.1 Data
Data is collected from the asset 4g ESG database and the Datascope database, both present in Thomson Reuters Datastream, for the period 1/2005 to 12/2013. The 4g ESG database is used because it conforms with the suggestions made by Waddock and Graves (1997), to ‘use a measure that offers a broad scope so it can be applied across a wide range of industries and large samples of companies’. The 4g ESG database consists of several indicators: economic performance, environmental performance, social performance and corporate governance performance. Every indicator is based on a multi-step verification conducted by a team of financial professionals to control and ensure high quality of data. 900 data points are used to establish 250 key performance indicators, which are listed in one of the four performance pillars. By calculating z-scores for these indicators the performance can be measured relative to other firms. Berry and Junkus (2012) find that investors consider environmental and sustainability issues as most important when evaluating the CSR performance of a firm. Hence I include the environmental and social pillar to assess the CSR score. According to the description provided by Thomson Reuters the environmental and social pillars measure, respectively the impact of a firm on natural systems, such air, land and water, and the capacity of a firm to generate trust and loyalty with its workforce, customer and society. Similar to Loannou and Serafeim (2012) I calculate the CSR score as the equally weighted average of the social and environmental score. The key performance indicators linked to environmental performance are resource reduction, emission reduction and product innovation, and to assess social performance data point score for employment quality, health and safety, training and development, diversity, human rights, community and product responsibility are used (Appendix 1).
The P/B ratio is requested from the DataScope database. If the book value of a firm approaches zero or is even smaller than zero, the P/B ratio becomes rather extreme or even negative. This problem is circumvented by setting all values larger than 15 or smaller than 0 equal to 15. Changes in relative performance is accounted for by subtracting the return on assets (ROA) before the year of interest (Y-1) from the ROA in the year of interests (Y). In contrast to measures like percentage change of net income or absolute change in income, it controls for the size of the change in performance relative to size company. Risk free rates are obtained from the Kenneth French website.4
Next, firms are excluded for which not all variables are available, this yields 18790 observations. The remaining companies are listed in 42 different countries, mostly located in the USA or Canada.
4.2
Descriptive data
Table 3 presents the data. The highest CSR score yields 97.67 achieved by BMW in 2006. Although this score is high, it does not present a problem in the dataset as the data is measured relatively and thus high scores can be expected. The maximum change in CSR score of 81.61 can be problematic however. According to the dataset Lotte Chemical’s CSR score rose from 12.28 in 2010 to 93.89 in 2011. On the website of Lotte Chemical no indication can be found that the social awareness of the firm increased dramatically in 2011. Therefore it is more likely that this, and other sudden extreme changes in the CSR score, result from measurement errors and should thus be excluded. The same goes for the high maximum returns and low minimum returns that are connected to the financial crisis and should be excluded from the data. P/B ratios, become very sensitive ones the book value drops significantly which cause the extreme high and even negative market to book values. Last, the most extreme values for ROA are unlikely to be related to CSR and might affect the results. Given the outcomes of previous studies CSR is unlikely to be related to these extreme outcomes therefore all these variables the .5% highest (except for P/B as a high P/B is still better than a negative P/B) and lowest .5% are excluded from the database.
After winsorizing, the average firm in the sample has a CSR score of 53.95 out of 100 points that changes with 1.76 points each year. The return of this firm is rather high with 12.17%, which is quite remarkable as it exceeds the risk corrected return of the market by more than 6 percent. The average P/B ratio is 2.94 and the average market capitalization is 13.47 billion dollars.
4
Table 3. This table presents the descriptive data of the unadjusted sample. Market capitalization in billions.
Mean Median Maximum Minimum Std. Dev. Skewness Kurtosis
Mean Median Maximum Minimum Std. Dev. RI-rf 12.17% 9.55% 242.20% -0.83 0.419065 MI-rf 5.52% 11.45% 68.29% -0.66 0.218878 CSR score 53.95 56.15 97.67 5.92 29.64652 CSR Score Delta 1.76 0.03 48.21 -27.39 9.843296 Beta (y) 1.09 1.06 5.88 -0.71 0.458277 Beta *(MI-rf) 5.32% 6.99% 1.79 -1.37 0.255998
Beta delta (y, y-1) 0.01 0 2.37 -1.61 0.199111
Market capitalization $13.47 $4.91 $504.00 $0.00 $28.56 LN market capitalization 15.51 15.41 20.04 7.88 1.287798 LN market capitalization delta 0.04 0.08 4.18 -2.73 0.436298 P/B ratio 2.94 1.82 812.07 -15.38 11.95004 P/B sqrt 1.5 1.36 3.87 0.22 0.616476 P/B sqrt Delta -0.02 -0.02 3.87 -3.07 0.285791 ROA 0.0422 0.0203 1.77 -1.03 0.083646 ROA Delta -0.0017 0 0.42 -0.43 0.060182 Delta NI/assets 0.0019 0.0029 1.71 -2.57 0.070015 Net income $862.84 $261.21 $45,220.00 $-38,732.00 2571.062 Operating income $728.42 $82.98 $36,800.00 $-1,450.00 783
5
Results
5.1
Relation between (delta) CSR and market returns (model 1)
Appendix 2 presents the correlation coefficients for the variables of model 1.
Table 5 shows the results for Model 1 with yearly market returns as dependent variable. First, I discuss a simplified version of Model 1 based on the CAPM (Model 1.1 version 2). This individual and time fixed model includes beta (to adjust for risk) but lacks the 2 additional factors of the Fama and French 3 factor model (size and growth). The coefficients of CSR and delta CSR are in line with my expectations respectively negatively and positively correlated with market returns and are significant at the .01 level. This indicates that market returns are indeed lower for CSR firms and that the net effect of an increase in CSR does have net positive effects on market returns (which might be attributed to lower beta, a CSR premium or better operating performance). One thing to note is that, while I do not use portfolios, which might have negative effects on the effectiveness of the beta because some idiosyncratic risk is included in the estimate, the coefficient for beta is almost 1 and is significant at a 1% confidence level, indicating that it efficiently estimates the risk associated with the firm.
Next I check whether this relation changes when delta CSR is dropped from the model (model 1.1 version 1). I expect a positive relation because several studies find that CSR (without controlling for changes in CSR) is positively and significantly associated with market performance (Arx & Ziegler, 2014; Orlitzky et al., 2003). In contrast to the findings of Arx and Ziegler, the coefficient of CSR is again negative, although less significant. This might be explained by the differences in methodology used in this study compared to the methodology of Arx and Ziegler’s study. For instance, their measure for CSR is divided in two parts, one part measures the CSR score compared to the industry average, and the second part measures the environmental and social performance of the industry to which the firm belongs. The measured coefficients for these factors are contradictory: the coefficient for the corporation is positive while the coefficient for the industry is negative. As this study combines these two factors in one measure for CSR, the significance might be mitigated. In addition their study uses a different dataset to determine the CSR score, evaluates portfolios and uses a dummy to qualify a firm as either a CSR firms (firms with a high CSR score are qualified as a CSR firm) or a non CSR firms (firms with a low CSR score are qualified as no CSR firm). When I estimate a OLS model without fixed year effects to verify my results with their results, the coefficient does become positive but not significant.
significance level. The significance level of the coefficients for CSR and CSR delta for this are larger than for model 1.1. This indicates that the control variables mediate the relation between CSR and a firm’s market performance in a positive way offsetting the premium related to CSR firms.The coefficient of CSR is not affected when delta CSR is dropped from the model. The coefficient of the growth factor contradicts the expected sign and is positively related with market returns and significant at the 1% level. These results are in accord with the findings of Ang and Schwraz (2010) when they look at individual stocks.
As stated, studies normally include HML and SMB to assess whether a portfolio is exposed to size or growth firms. Therefore these studies do not test whether portfolios containing on average firms with a small market capitalization outperform portfolios containing firm with large market capitalization, but merely test the correlation between the weekly/monthly returns of the portfolio and the HML/SMB factor and make determine whether a size or growth factor is present within the portfolio. I verify the positve association between growth and market return by constructing a portfolio consisting of the 50% firms with the highest P/B ratio. This portfolio outperforms a portfolio based on firms with a low P/B ratio.
Third, I estimate model 1.3. This model adds a control variable for operational performance to model 1.2. Two different versions of this model are constructed: model 1.3a contains the change in return on assets (delta ROA) and model 1.3b includes the change in operating income divided through market capitalization (Delta OI/market capitalization). The results for both models are similar to the findings for model 1.2: CSR and delta CSR impact market returns respectively negative and positive, both statically significant at the .01 level. Because this model controls for the potentially mediating effects of operating performance and beta (as described in theoretical background) on the relation between CSR and market performance, the results indicate that investors are indeed willing to pay a premium for a CSR firm. Therefore hypothesis 1 ‘investors are willing to pay a premium to invest in a firm with a higher CSR score compared to a firm with a lower CSR score’ can be accepted. Although these results are statistically significant, the economic impact is rather small: a firm with a CSR score that is 10 points higher than a similar other firm, has a market return that is .026% lower than that other firm.
increase in market returns that is associated with a higher CSR score as it might be explained by higher expected cash flows (which is related to operating income) or/and a change in beta.
5.2
Relation between (delta) CSR and next year’s beta (model 2)
Appendix 3 provides the correlation coefficient for Model 2.
Table 6 presents the results for Model 2 with next year’s beta as dependent variable. Similar to the previous model, I will start to discuss a simplified version of the model (model 2.1). Only the explanatory variables (CSR and delta CSR) are included in this firm and year fixed effect model. CSR and delta CSR are negatively related with next year’s beta. CSR is significantly related with the beta at the .10 level and the coefficient for delta CSR is not significant. This might indicate that investors do not forecast that a change in CSR is related with reduced risk but in the long run such an effect does exist, because, for example, the returns of the firms become less volatile. The relation becomes marginally more significant when delta CSR is omitted. Adding delta size and delta growth factor does improve the results further. CSR becomes even significant at the 5% level when delta CSR is dropped. In neither of the models CSR delta is significant. Model 2.3a and 2.3b perform similarly. Therefore, hypothesis 2: ‘a change in CSR scores reduces the beta of a firm’ can partly be confirmed. On the one hand, firms with a higher CSR score do have lower beta’s, but this effect is not directly recognized by investors at the moment of the change in CSR.
5.3
Relation between (delta) CSR and operating performance (model 3)
Appendix 4 presents the correlation coefficient for Model 3.
Table 7 provides the results for Model 3. The dependent variable for model 3a and 3b are respectively delta ROA and delta OI/market cap. Beta, the size factor, growth factor, delta size and delta beta, delta growth, and delta price to book ratio are added as control variables.
Model 1.1 model 1.2 model 1.3a model 1.3b Expected sign 1 2 1 2 1 2 1 2 CSR Score - -0.00195*** -0.00240*** -0.00217*** -0.00266*** -0.00212*** -0.00260*** -0.00216*** -0.00265*** -7.78 -8.91 -8.77 -9.96 -8.61 -9.79 -8.73 -9.92 CSR Delta + 0.00129*** 0.00136*** 0.00134*** 0.00136*** -4.52 -4.85 -4.8 -4.85 Beta *MI-rf + 1.057*** 1.061*** 0.981*** 0.985*** 0.976*** 0.979*** 0.982*** 0.986*** -104.28 -104.36 -93.9 -94.08 -93.56 -93.74 -93.95 -94.14 Size - -0.0744*** -0.0776*** -0.0757*** -0.0788*** -0.0735*** -0.0767*** -7.64 -7.96 -7.81 -8.12 -7.55 -7.86 Growth - 0.184*** 0.185*** 0.175*** 0.176*** 0.183*** 0.183*** -26.67 -26.75 -25.28 -25.36 -26.48 -26.56 ROA Delta + 0.440*** 0.439*** -10.51 -10.49 Delta OI/ Market cap + 0.000170** 0.000170** -2.66 -2.66 Constant 0.170*** 0.192*** 0.0212 0.0488* 0.0351 0.0623** 0.0208 0.0484* -12.45 -13.25 -1.01 -2.24 -1.67 -2.87 -0.99 -2.23 Observations 18160 18160 18160 18160 18160 18160 18160 18160 R-squared 0.4211 0.4219 0.4475 0.4484 0.4524 0.4516 0.4478 0.4486 Table 5. This table shows the results for model 1 with as dependent variable the firm’s market returns minus the risk free rate. A firm and time fixed effects model is
used with control variables for the three Fama and French (1994) factors (Beta, Size and Growth) in addition two alternative measures for operating performance are added (ROA Delta and Delta OI/Market Cap). The dataset covers the years from 1/2005 till 12/2013. The top and bottom number indicate respectively the corresponding coefficient and t-statistic.
Model 2.1 Model 2.2 Model 2.3a Model 2.3b Expected Sign 1 2 1 2 3 1 2 1 2 CSR score - -0.000497* -0.000486* -0.000570** -0.000556* -0.000569** -0.000555* -0.000262 -0.000554* -2.39 -2.14 -2.73 -2.44 -2.73 -2.44 -1.24 -2.43 CSR Delta - -0.0000267 -0.000262 -0.0000353 -0.0000354 -0.0000352 Growth -0.12 -1.24 -0.15 -0.15 -0.15 Delta -0.0234** -0.0236** -0.0234** -0.0234** -0.0234** -0.0236** -0.0233** -2.98 -3 -2.98 -2.98 -2.98 -3 -2.97 Size Delta - -0.0306*** -0.0301*** -0.0307*** -0.0308*** -0.0308*** -0.0303*** -0.0308*** -6.26 -6.16 -6.26 -6.16 -6.16 -6.06 -6.29 Growth delta - 0.00455 0.00457 -0.13 -0.13 Delta OI/ Market cap - 0.00508 0.0000376 -0.14 -0.71 Constant 1.127*** 1.127*** 1.131*** 1.101*** 1.130*** 1.131*** 1.130*** 1.101*** 1.130*** -100.04 -93.02 -100.13 -527.88 -93.15 -100.13 -93.15 -527.42 -93.14 Observations 15833 15833 15813 15813 15813 15813 15813 15813 15813 R Squared 0.0004 0.0004 0.0049 0.0045 0.0049 0.0049 0.0049 0.0045 0.005 Table 6. This table shows the results for model 2 with as dependent variable the firm’s beta at the end the next year. A firm and time fixed effects model is used with control
variables for changes in three Fama and French (1994) factors (Beta, Size and Growth) in addition two alternative measures for operating performance are added (ROA Delta and Delta OI/Market Cap). The dataset covers the years from 1/2005 till 12/2012. The top and bottom number indicate respectively the corresponding coefficient and t-statistic.
Model 3a Model 3b Expected sign 1 2 1 2 CSR score +/- 0.0000168 0.0000201 -0.0487 -0.0462 -0.35 -0.39 -1.52 -1.34 CSR Delta + -0.0000091 -0.00672 -0.17 -0.19 Beta *MI-rf +/- -0.00653** -0.00656** -9.331*** -9.352*** -2.84 -2.85 -6.11 -6.1 Size +/- 0.00534*** 0.00533*** 3.365** 3.360** -3.42 -3.41 -3.24 Growth +/- 0.0124*** 0.0124*** -4.049** -4.028** -5.87 -5.87 -2.9 -2.87 Beta Delta + 0.00142 0.00143 -2.425 -2.418 -0.56 -0.57 -1.45 -1.44 Size Delta + 0.0294*** 0.0295*** 5.422*** 5.430*** -18.79 -18.78 -5.21 -5.21 Growth Delta - -0.0110*** -0.0110*** -0.774 -0.784 -5.48 -5.48 -0.58 -0.59 Constant -0.105*** -0.105*** -43.65** -43.71** -4.51 -4.51 -2.82 -2.82 Observations 18137 18137 18137 18137 R-sq: 0.0483 0.0483 0.0059 0.0059
Table 7. This table shows the results for model 3a and 3b with as dependent variable the operating performance
measured respectivelyas a change in the ROA (ROA Delta) and a change in the operating income divideded through the market capitalization (Delta OI/Market Cap). . A firm and time fixed effects model is used with control variables for the three Fama and French (1994) factors(Beta, Size and Growth) and changes in those variables (Beta Delta, Size Delta and Growth delta).The dataset covers the years from 1/2005 till 12/2013. The top and bottom number indicate respectively the corresponding coefficient and t-statistic.
5.4
Robustness analyses
I conduct several additional tests to check the robustness of my results. First, I estimate the models using an ordinary least square regression. I add year and country dummies to all versions of the model to control for various potential mediating effects. This results in the following models:
Model 1: 𝑅𝑖,𝑡𝑒 = (𝐶𝑆𝑅𝑖𝑡)𝛽1+ (Δ𝐶𝑆𝑅𝑖𝑡,𝑖𝑡−1)𝛽2+ 𝐵𝑒𝑡𝑎𝑖𝑡𝑅𝑚𝑡𝑒 𝛽3 + 𝑆𝑖𝑧𝑒𝑖𝛽4+ 𝐺𝑟𝑜𝑤𝑡ℎ𝑖𝛽5+ Δ𝑅𝑂𝐴𝑖𝑡,𝑖𝑡−1𝛽7+ 𝐷𝐶,𝑦𝛽8,9+ 𝜀𝑖,𝑡 Model 2 𝐵𝑒𝑡𝑎𝑡+1= (𝐶𝑆𝑅𝑖𝑡)𝛽1+ Δ𝐶𝑆𝑅𝑖𝑡,𝑖𝑡−1𝛽1+ Δ𝐵𝑒𝑡𝑎𝑖𝑡,𝑖𝑡−1𝛽3 + Δ𝑆𝑖𝑧𝑒𝑖𝑡,𝑖𝑡−1𝛽4+ Δ𝐺𝑟𝑜𝑤𝑡ℎ𝑖𝑡,𝑖𝑡−1𝛽5+ Δ𝑅𝑂𝐴𝑖𝑡,𝑖𝑡−1𝛽5+ 𝐷𝐶,𝑦𝛽6,7+ 𝜀𝑖,𝑡 Model 3a: Δ𝑅𝑂𝐴𝑖𝑡,𝑖𝑡−1= (𝐶𝑆𝑅𝑖𝑡)𝛽1+ Δ𝐶𝑆𝑅𝑖𝑡,𝑖𝑡−1𝛽1+ Δ𝐵𝑒𝑡𝑎𝑖𝑡,𝑖𝑡−1𝛽3 + Δ𝑆𝑖𝑧𝑒𝑖𝑡,𝑖𝑡−1𝛽4+ Δ𝐺𝑟𝑜𝑤𝑡ℎ𝑖𝑡,𝑖𝑡−1𝛽5+ 𝐵𝑒𝑡𝑎𝑖𝑡𝛽3 + 𝑠𝑖𝑧𝑒𝑖𝛽5+ 𝑔𝑟𝑜𝑤𝑡ℎ𝑖𝛽6+ 𝐷𝐶,𝑦𝛽6,7+ 𝜀𝑖,𝑡 Model 3b: Δ𝑁𝐼 𝐴𝑠𝑠𝑒𝑡𝑠= (𝐶𝑆𝑅𝑖𝑡)𝛽1+ Δ𝐶𝑆𝑅𝑖𝑡,𝑖𝑡−1𝛽1+ Δ𝐵𝑒𝑡𝑎𝑖𝑡,𝑖𝑡−1𝛽3 + Δ𝑆𝑖𝑧𝑒𝑖𝑡,𝑖𝑡−1𝛽4+ Δ𝐺𝑟𝑜𝑤𝑡ℎ𝑖𝑡,𝑖𝑡−1𝛽5 + 𝐵𝑒𝑡𝑎𝑖𝑡𝛽3 + 𝑠𝑖𝑧𝑒𝑖𝛽5+ 𝑔𝑟𝑜𝑤𝑡ℎ𝑖𝛽6+ 𝐷𝐶,𝑦𝛽6,7+ 𝜀𝑖,𝑡
With 𝐷𝑦as a vector of dummies for the diffierent countries that takes the value of 1 if a firm is from the correlating countries. Similar, 𝐷𝑦is a vector of dummies for the different years that become 1 if an observation is made in the corresponding year.
Appendix 5 presents the results of the OLS regression results for all three models. The independent variables have equal signs and are also significant at the 1% level. Except for the coefficient of size all control variables have the same signs and significance levels. For model 2 the coefficients are no longer significant. So the relation between the CSR and the beta becomes questionable since the coefficient was only significant at the 10% level in the main analysis to begin with. Delta CSR becomes positive significant with model 3. Indicating that the change in CSR triggers an increase in sales. However, this result is not supported by the main analysis, so the evidence for this statement is limited.
As mentioned in section 4 the data suffers from heteroscedasticity and might show autocorrelation. Results of the White test and Breusch–Godfrey-Bertolo test indicate respectively that the data is indeed heteroscedastic and autocorrelated. This violates the assumptions related to the OLS model and might cause inefficient standard error, which makes the results misleading (although the model remains unbiased).
2008). According to Petersen (2009) the this approach still results in a bias, but the bias is less severe than for the other approach. Appendix 6 presents the results.
The results of model 1 correspond with the results of the main analysis and both hypotheses 1 and 4 can be accepted using OLS with West standard error as well. For model 2 the results differ, the coefficient of CSR with beta is now significant at the 1% level. However, CSR delta becomes positively significant. The signs of the other coefficients all correspond with the ones found in the main analysis, although the significance levels differ. Also for model 3 the coefficient of CSR becomes significant at the .01 level. Indicating that CSR might be related to lower returns in the long run, and the market does not forecast this effect correctly.
6
Conclusion
This paper attempts to answer the question whether investors are willing to pay a premium to comply with their morality. Hereto it determines whether investors accept a lower return to gain non-financial utility related to a CSR investment. Because the relation between CSR and market returns might also be attributed to other links unrelated to the morality of the investor (e.g. increase in sales, decrease/increase in costs, lower estimated risk) theories regarding these links are evaluated. These effects are classified as net operating returns, financial considerations of the investor and non-financial considerations of the investor. The latter is considered to be a moral consideration, and thus the topic of interest. By adding several control variables, delta CSR and employing a firm and year fixed effect model the study determines whether investors are willing to sacrifice return for to gain non-financial utility related to a CSR investment. The study uses Thomson Reuters’ Asset4 ESG data, over the period 01/2005 to 12/2013 covering over 4000 firm and extends the work of Dhaliwal (2011 and 2014) by uncovering that the lower discount rate can also be observed on the market in the form of a lower market return and by analysing the mechanism behind the lower market return with an emphasizes for morality(beta v.s. morality premium).
The main finding of this paper suggest that the market is indeed willing to sacrifice return for an investment in a CSR firm. In addition the positive, significant and robust coefficient for delta CSR indicates that the net effect of increasing CSR is positive. This positive effect might be caused by higher expected cash flows or a change in the discount rate. Therefore I examine whether this net positive effect is related to estimated risk, higher expected operating income or a premium the market is willing to pay for a CSR firm. In contrast to suggestions made by Ghoul et al. (2010) and Hughes et al. (2007) the estimated risk of a firm as measured by the beta is not affected by CSR; both delta CSR and CSR and beta are not significantly related with beta.5 In addition I evaluate the effect of CSR on operating performance. Again no significant association between these factors can be found. This indicates that the effect of CSR on market performance is primarily induced by non financial considerations of the investors, which supports the premise that stock market morality.
Although the results do confirm a negative relation between CSR and a firm’s market performance, which can be considered as proof of the premise that an investor is willing to sacrifice return to comply with morality, the economic significance is limited. There are several possible explanations for the lack of ecnomic significance. First, investors are indeed not willing to sacrifice return to
5
comply with their morality. Second, these scores do not represent the aspects important for an investors when they consider the CSR performance of a firm. Third, due to agencies problems the preferences of investors are not translated to investment decisions. According Merton’s framework and the survey conducted by Deloitte (2003), it would be logical that a premium should be paid for CSR firms. This is also illustrated by the fact that people, in the capacity of a consumer, are interested in CSR firms, but as investors they are not. As most people outsource investment decision to agents, these agents might not incorporate the preference of people for CSR firms in the investment decision. Thus this might be a point of improvement.
7
Sources
Aguinis, H., & Glavas, A. (2012). What we know and don’t know about corporate social responsibility a review and research agenda. Journal of Management, 38(4), 932-968.
Ambec, S., and Lanoie, P. (2008). Does it pay to be green? A systematic overview. The Academy
of Management Perspectives, 22(4), 45-62
Ang, A., Liu, J., and Schwarz, K. (2010). Using stocks or portfolios in tests of factor models.
Unpublished Working Paper, Columbia University and University of Pennsylvania.
von Arx, U., & Ziegler, A. (2014). The effect of corporate social responsibility on stock performance: new evidence for the USA and Europe. Quantitative Finance, 14(6), 977-991.
Berry, T. C., and Junkus, J. C. (2013). Socially Responsible Investing: An Investor Perspective. Journal of business ethics, 112(4), 707-720.
Bollen, N. P. (2007). Mutual fund attributes and investor behavior. Journal of Financial and
Quantitative Analysis, 42(03), 683-708.
Brammer, S., Brooks, C., & Pavelin, S. (2006). Corporate social performance and stock returns: UK evidence from disaggregate measures. Financial Management, 35(3), 97-116..
Brooks, C. (2008). Introductory econometrics for finance. Cambridge university press.
Cortez, M. C., Silva, F., & Areal, N. (2012). Socially responsible investing in the global market: the performance of US and European funds. International Journal of Finance & Economics, 17(3), 254-271.
Deloitte, Euronext, CSR Europe [2003], “Investing in responsible business – the 2003 survey of European fund managers, financial analysts and investor relations officers.”
Daves, P.R., M.E. Ehrhardt, R.A. Kunkel(2000), Estimating systemic risk: the choice of return interval and estimation period, Journal of Financial and Strategic Decisions Volume 13 number 1, 2000.
