Do Capital Markets Price Social
Responsibility?
An Analysis of the Relationship Between Market Risk and Corporate Social
Responsibility.
Abstract
This paper investigates the relationship between corporate social responsibility and market risk. In a sample of over 22.000 firm year observations I find that there are clear indications of a reduction in market risk for investing in stocks associated with the highest levels of corporate social responsibility. The risk reduction is strongest when using downside market risk as a measure of risk, but also
significant when using conventional measures of risk. Furthermore this paper uncovers that the risk reduction is strongest in periods of low volatility, limiting any wealth protective effects of social responsibility investing.
Bachelor Thesis Economie en Bedrijfskunde
Track: Finance & Organisation
Jakob Hijlkema 10099646
Supervisor: P.J.P.M. Versijp
Abstract
This paper investigates the relationship between corporate social responsibility and market risk. In a sample of over 22.000 firm year observations I find that there are clear indications of a reduction in market risk for investing in stocks associated with the highest levels of corporate social responsibility. The risk reduction is strongest when using downside market risk as a measure of risk, but also
significant when using conventional measures of risk. Furthermore this paper uncovers that the risk reduction is strongest in periods of low volatility, limiting any wealth protective effects of social responsibility investing.
I. Introduction
The idea of socially responsible investing is as old as humanity itself. There are numerous examples of ancient cultures and religions imposing some sort of ethical constraints on investments, but socially responsible investing as seen nowadays is a development of at most 20 years. The market for socially responsible investments has seen an explosive growth in recent years and is today a
multitrillion market. According to the Global Sustainable Investment Alliance (GSIA), socially responsible investments account for over 20% of the total assets under professional management worldwide. One of the largest and best documented sub-markets is the U.S.. The leading authority documenting the U.S. market for socially responsible investments, the United States Social
Investment Forum, estimates that there are $ 3.74 trillion dollars socially responsible invested. The amount of assets under socially responsible management has increased by 22% from 2009 to 2011 and keeps on growing.
But what exactly is socially responsible investing? It is the practice of incorporating some environmental, social or ethical values in the investment decision. The question is of course how to recognize these values in company policies. Company policies on the areas identified by socially responsible investing can be to some extent measured by corporate social responsibility or corporate social performance. Both terms correspond to the same values and are used interchangeably.
Some companies perform poorly in the light of the wider stakeholder environment by hazardous company policies that cause negative externalities. Examples are not containing any pollution of the environment, discrimination in hiring policies and extortion of employees. Other companies do not have harmful policies per se, but rather produce a very undesirable product from a social perspective, such as nuclear energy or weapons. On the other hand there are also companies that perform extremely well in the aforementioned areas, these companies are said to be socially responsible. The main goal of socially responsible investing is identifying any concerns or strengths regarding the areas concerning the wider stakeholder environment to exclude or include these investments in the investment portfolio. This process is called investment screening. Rather than excluding or including investments the same result can be achieved thru shareholder activism. Shareholder activism implies changing some socially undesirable policy thru the control of company shares. This paper focuses on the appliance of investment screens.
“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.” (Friedman, 1970)
According to the late Nobel laureate Milton Friedman, devoting resources to corporate social performance should be avoided. The idea of social responsibility being value destroying has been a basic assumption for the research conducted on the relationship between financial performance and corporate social responsibility.
Most of this research is focused on the relationship between investment returns and levels of corporate social responsibility. Often the hypothesized relationship is that stocks that are socially desirable earn a lower return as a result of the devotion of company resources to activities that do not directly increase profits. However the conclusions are ambiguous and researchers fail to clearly identify a relationship and its sign.
I argue in this paper that there is a more clear link between financial performance and corporate social responsibility and that is the relationship between the latter and market risk. I investigate this relationship in a sample of more than 22000 firm year observations over the period of 2006 to 2013. My theory for this phenomenon is that investors investing in stocks associated with high corporate social performance are less tempted to give in to market sentiment, since they do not only care about investment returns but also about the level of corporate social responsibility. This would then imply a more stable demand for highly socially desirable investments, resulting in lower market risk. The research question of this paper is therefore: “Stocks with higher levels of corporate
social responsibility are subject to less market risk than stocks with low levels of corporate social responsibility”.
I find in the researched sample that portfolios comprised of stocks with higher levels of corporate social responsibility are subject to less market risk than their equivalents comprised of stocks that are poor social performers. This relationship is even stronger when using downside deviations as a measure of risk. However an additional finding indicates that this reduction is greatest in times of low volatility, therefore restricting any wealth protective effects of corporate social performance.
The remainder of this paper is structured as follows: in part II the related literature is reviewed and I develop the research hypotheses. In part III I discuss the validity of the data and its descriptive statistics. Furthermore in Part IV the research methodology is discussed, including the defence of any assumptions made. In part V the results and main findings are presented, followed by a conclusion and discussion of these results in part VI.
II. Literature review and developping of hypotheses
Existing literature focusses on the relation between corporate social responsibility and investment returns, often using risk merely as an adjustment factor. There exist many studies investigating the excess returns of Socially Responsible Investment Funds or SRI-funds. The major part of the related literature attempts to show that these SRI-funds underperform the market. The rationale behind this is the following: applying ethical investment screens, either possitive or negative, reduce the investable universe, i.e. the portfolios of these funds can use less stocks to define their optimal portfolio which is thereby by definition suboptimal to the portfolio formed from the entire investable universe. The underperformance is then interpreted as the price capital
markets place on corporate social responsibility and in some cases as a lower bound on the altruistic utility investors may derive from investing in these funds. However there is no conclusive evidence in the existing literature indicating any statistical significant difference between the performance of
SRI-funds and conventional mutual SRI-funds. Since this paper focusses on the U.S.-market I will discuss some papers evaluating the relationship between corporate social responsibility and investment returns in the U.S..
One of the first studies investigating the relationship between corporate social responsibility and fund performance is the study by Hamilton, Jo and Statman (1993). This study shows that for a sample of 17 SRI-funds founded before 1986 and for 15 SRI-funds founded after 1986 there is no conclusive evidence of underperformance. In the first group 2 of the 17 funds showed excess performance as measured by the Jensen’s alpha statistically different from 0, however the both differing Jensen’s alpha showed different signs. Also the authors find no statistical difference in the performance of SRI-funds and conventional mutual funds. Although insignificant the study hints a negative relationship between corporate social responsibiility. The authors conclude that capital markets do not price corporate social responsibility.
Whereas the study of Hamilton, Jo and Statman (1993) consists of a rather small sample of funds, Goldreyer, Ahmed and Diltz (1999) perform a similar study of a larger sample. They compare the returns of 49 mutual funds that engage in ethical screening practices to the returns of
conventional mutual funds. The study finds both under- and overperformance of SRI-funds and concludes to find no unambigous advantage in performance for either of the groups. One interesting finding of this study is that SRI-funds employing positive screening methods outperform SRI-funds that do not.
Statman (2000) shows that the Domini Social Index, a socially screened version of the S&P 500, did no worse than the S&P 500 for the 1990-1998 period. Also the author finds, however insignificant, that socially responsible funds outperform conventional mutual funds. Furthermore the study shows that both conventional mutual funds and SRI-funds underperform the S&P 500 for this time period.
In more recent research Renneboog, Ter Horst and Zhang (2008) research a large sample of SRI-funds from all over the world. They find that most SRI-funds underperform domestic
benchmarkportfolios by 2,2% to 6,5% per annum. However their results are only significant in 4 out of the total of 15 investigated countries. Also they find that the apparant underperformance may be partly induced by poor stock picking skills in SRI-fund managers. For the US-market there is no significant underperformance.
Other studies take a differing approach to identifying the nature of the relationship between corporate social responsibility and investment returns than solely comparing excess returns for SRI- and non-SRI-funds. Geczy, Stambaugh and Levin (2000) use modern portfolio theory to construct portolios of funds at least partly engaged in SRI . The comparison of these portfolios and portfolios constructed from funds that do not maintain any investment constraint shows that the cost of this constraint depends on the investors beliefs in managerial skills. If an investor completely rules out any stock picking skill and is a so called market investor, the cost of the constraint is on average a loss in returns of 0,76% per annum. However when the investor places some faith in pricing models that relate returns to size, market-to-book value or momentum factors, such as the Fama French three-factor model, the constaint results in an average loss of returns of 3,6% per annum.
Brammer, Brooks and Pavelin (2006) show that any significant underperformance in SRI-funds might be mitigated by the use of an aggregate measure. The rationale behind this is that whereas some areas of corporate social responsibility might be negatively related to investment returns, others show a positive relation. Therefore the both effects could even out, resulting in no
significant relation between levels of CSR and investment returns. This study uses disaggregate measures for the three most important areas of corporate social responsibility: Environment, Community and Employee relations. By using disaggregate measures the authors unravel a positive and significant relationship between policies regarding Employee Relations and excess investment returns. However the authors find an also signigficant and quite large negative relationship between Environment factors and excess investment return, alongside with a negative smaller, but significant, relation for Community factors and excess investment returns. Acknowledging these results I
consider their findings in the weighing of the various CSR areas in the portfolio construction. This will be explained more precisely in the methodology section.
The majority of aforementioned studies hints a negative relationship between investment returns and corporate social responsibility. This might be of any economical significance, but the results of the various studies are ambiguous and lack statistical significance. So is there no price that capital markets place on corporate social responsibility? Not entirely. There is another determinant of stock performance which is priced by capital markets and has been surpassed by the larger part of the literature: risk. Even though risk has been used as an adjustment factor in various studies, the amount of literature on the relation between risk as a standalone variable and corporate social performance is in the least scarce.
But why would there be any direct relation between risk and corporate social responsibility? As evident from the aforementioned studies the existent literature hints a negative relation between strong corporate social performance and investment returns. But then why is there still demand for these underperforming stocks? I find a rationale for this in the fact that investors have multivariate utility functions that are to some extent determined by social principles (Bollen, 2007). If certain investors derive altruistic value from investing in more socially responsible company stocks, this explains why investors still demand the underperforming stocks. If an investor would derive no such utility clearly investing in an underperforming stock would be unwise.
However Graves and Waddock (1994) show that the degree of institutional investor ownership does not vary with corporate social performance. Most institutional investors such as pensionfunds have no altruistic mission, so this is not in line with the expectations. One explanation may be found in the fact that corporate social responsibility benefits accrue in the long run and institutional investors tend to have a longer investment horizon (Cox, Brammer and Millington, 2004).
Though a much more simple explanation may be found in the fact that there are
government regulations and other external pressures forcing institutional investors to invest to some degree in socially responsible stocks. Although U.S. institutional investors are not yet obliged by regulation to invest socially responsible, institutional investors in other countries are. For example pension funds in Belgium, Germany and France must invest some of their funds taking socially responsible considerations in to account. Also U.K. charities must to some level invest funds in socially responsible investments. These regulations not only increase but also stabilize demand for all socially responsible assets, presumably with spill-over effects to the U.S..
This leads me to develop the following research hypothesis:
Hypothesis 1: A portfolio of stocks with higher corporate social performance is subject to less market risk than a portfolio of stocks with lower corporate social performance.
Downside risk as a measure of financial performance is an alternative for using variance as a measure of risk. Where variance measures deviations both on the up and downside of the mean, downside risk concerns only deviations on the bottom side of some measure. This characteristic corresponds well to theories of risk aversion, since the real risk that concerns investors is the risk of loss. Roy (1952) is among the first to acknowledge risk aversion in individuals and employs the socalled “Principle of safety first”. The rationale behind this principle of risk aversion is given by a metaphore comparing economic risk to a survivorship situation: if an individual risks losing everything, this person might be less concerned with getting a little bit more.
Further also Ahneman and Tversky (1979) show that expected utility theory exhibits some flaws and might not correspond to rational behaviour. In their experiments the researchers find that individuals avoid especially the risk of loss and rather choose certain payoffs, even if the expected payoffs of the risky choice are higher.
Acknowledging that individuals are risk averse, I now review some studies incorporating downside risk to evaluate stock performance. Post and Van Vliet (2004) use a sample over a period of almost 80 years to incorporate the what they call “bad states of the world” which correspond to financial crises and crashes over the period of time. The researchers find that the relationship between excess investment returns and risk is much better represented by models employing a measure of downside risk as opposed to the conventional models which use variance as a measure of risk. More specifically the authors find that during bad states of the world measures of downside risk restore the relationship between risk and return, which could only be weakly estimated by
conventional measures of risk in these states.
Also Ang, Chen and Xing (2006) find that when using downside market risk as a measure of risk that stocks with higher β figures command a premium on average returns for the increased downside risk. The researchers find also that downside β figures are more reliable predictors of future returns as compared to conventional measures of market risk.
These studies show that it makes sense to use downside market risk as a measure of risk as opposed to variance. Since I use a measure of downside market risk to model my research, I discuss the econometric characteristics of a measure for downside (market) risk in the methodology section. For now let me conclude that there are enough indicators for downside risk to be an appropriate measure of risk for any investment.
Besides downside risk being a superior measure of risk compared to the conventional variance. I also find reasons to investigate the relation between downside market risk and corporate social performance in the aforementioned multivariate utility theory and regulations imposed on institutional investors.
Both the forced demand by institutional investors and the hypothesized altruistic investor demand for socially responsible investments I would not expect to vary with stock performance. To clarify: I expect both the regulated demand by institutional investors and socially responsible investors to be fairly stable. For example when stock prices go up, investors that have no altruistic considerations do not all of a sudden demand socially responsible stocks. Also regulations do not
change overnight so institutional investors demands are quite stable. The stocks will be demanded by other investors since there are profits to be made. So on the upside of the market risk I would expect no difference resulting from corporate social performance. When stocks go down however,
institutional investors are still forced by regulations to invest in socially responsible investments and altruistic investors still derive non-financial utility, causing the demand of these groups to act as a buffer that stabilizes demand when the market goes down. Therefore I expect companies with higher levels of corporate social performance to be subject to less downside-market risk.
The arguments given above yield the following research hypothesis:
Hypothesis 2: A portfolio of stocks with higher corporate social performance is specifically subject to less downside market risk as compared to a portfolio of stocks with lower corporate social performance
There are studies investigating the relationship between idiosyncratic risk and corporate social performance such as Luo and Bhattacharya (2008). They find a negative relationship between coroporate social performance and firm specific risk. However the risk investors are primarily interested in is the non-diversifiable systematic risk.
Some early evidence on the relation between systematic risk and corporate social performance is given by Spicer (1978). Using Controls for Environmental Polution to measure the level of corporate social responsility, Spicer finds some strong negative correlations between both total and systematic risk and corporate social responsibility.
Furthermore Mcguire, Sungren and Schneeweiss (1988) investigate among other the influence of corporate social performance on market risk. The study finds an insignificant but negative relation between the two. However the measure for corporate social performance employed, Fortune magazine’s ratings of corporate reputations, is poor since it only contains some information on environmental and employee policies.
Sharfman and Fernando (2008) show in a sample of 276 U.S. firms that environmental risk management lowers the cost of capital. So essentially companies are rewarded by capital markets for their superior environmental performance. As the authors conclude much of the lower cost of capital results from a lower cost of equity capital which in turn is mainly caused by the reduction in
systematic risk. This article also suggests that more individuals purchase the stocks of firms that perform better environmentally. A limitation of this study is however that the authors focus only on the environmental performance, therefore not covering the wider range of corporate social
responsibility aspects.
Most of the studies reviewed until now are comprised of rather small samples. A study using a larger sample, however not in the U.S. market, is provided by Salama, Anderson and Toms (2009). The authors investigate the relation between Community and Environmental Responsibility or CER rankings and systematic risk in the U.K. from 1994-2006. Using 1,625 observations the authors find a negative relation between systematic risk and CER. However significant the potential risk reduction is very small. The authors find a coefficient for market risk of only -0,026%. The study also does not evaluate the full spectrum of corporate social responsibility.
All of the above studies evaluate the relation between risk and some measure of corporate social responsibility in a mean-variance framework, using variance as a measure of risk. However Oikonomou, Brooks and Pavilin also use alternative measures of risk. The authors investigate the
corporate social performance risk relationship on the firm level using data from 1992-2009. They use a sample of about 7000 firm year observations and find a weak but significant negative relation between corporate social performance and systematic risk. One of the advantages of this study is the sample size. Another is the use of measures of downside risk, but the study only entails these to strengthen the found results by conventional methods of risk. An interesting finding is that the relation between corporate social performance and risk is stronger in more volatile times. A limitation is that the researchers quantify corporate social performance.
In addition to the earlier developed hypotheses. I build on the result found by Oikonomou, Brooks and Brammer (2012) and test whether the found relationship holds on a portfolio level and for a longer period of volatility, from the third quarter of 2008 to 2012, as opposed to 2008/2009. This yields the following research hypothesis:
Hypothesis 3: The hypothesized reduction in market risk is larger for more volatile periods.
In summary is the existant literature scarce and of limited sample size, with the exception of the study performed by Oikonomou, Brooks and Pavelin. This study contributes to existing literature by investigating a much larger sample size of approximately 22,000 firm year observations from 2006-2013.
Another contribution is the use of downside market risk to evaluate the relationship between corporate social performance and risk on a portfolio level. By using portolios this study establishes a method to measure the relationship between corporate social responsibility and risk without quantifying corporate social performance, which is to some extent a heterogeneous and subjective variable.
III. Data
To obtain a measure for corporate social performance I have relied on the KLD Research & Analytics database provided by Wharton Research Data Services. KLD Research & Analytics inc. (KLD) was founded in 1988 and has been providing the financial services market with both research
products and services ever since. KLD has the largest research staff in the world concerning corporate social research which provides consistent and high quality research. Most SRI-funds rely on data provided by KLD for screening purposes. KLD stands for Kinder, Lydenberg and Domini. Domini was also the initiator of the Domini 400 SocialSM Index or DS 400 Index. Founded in 1990, this market capitalization-weighted index consists of 400 socially screened stocks. The DS 400 has come to be known as a benchmark for socially screened equity investments. The social ratings that KLD provides date back to 1991. The database started out with ratings just for the companies included in the S&P 500 and the DS 400. In 2001 the databases expanded to over a thousand US-companies and for the period from 2003 until the present the database covers social data and ratings on over 3000 US-companies. In this study I will use data from 2005 to 2012 from the KLD database, so my whole sample will exist of almost 3000 companies each year. The use of this timeframe enables me to use a bigger and more diverse sample than previous literature.
KLD assigns yearly updated social ratings on both social strengths and concerns. Social strengths are company policies which are socially desirable, whereas concerns are company policies which may be considered harmful from a social perspective. These strengths and concerns are divided into 7 broad categories which are each divided into several sub-categories. The seven broad
categories are Community, Corporate Governance, Diversity, Employee Relations, Environment, Human Rights and Product. These categories consist of 8, 5, 8, 7, 8, 4 and 4 sub-categories of social strengths respectively. An example of a social strength in the Environment category is pollution prevention, measured by the existence of strong pollution prevention policies whereas an example of a concern in the same Environment category may be hazardous waste, measured by a threshold for the amount of liabilities a company has regarding hazardous waste. Since this study only concerns social strengths, I will continue describing the process of establishing a rating for strengths only. A rationale for this is given in the methodology section. All of the sub-categories are binary variables, that is, if the company passes the threshold the rating consists of 1 for that sub-category. The
maximum points a firm can achieve is therefore 44. The descriptive statistics on the KLD disaggregate and aggregate measures can be found in the appendix.
The data on stock prices for the KLD-rated companies I obtained from the Center for Research in Security Prices (CRSP) database from the Wharton Research Data Services. CRSP is the most comprehensive database regarding stock prices and volumes for stocks traded in the US equity market. The acquired data was in a daily format with stock prices quoted unadjusted. Also CRSP uses a daily ask/bid average when there is no closing price known. The database uses a negative price to signal that it concerns an ask/bid average as opposed to a closing price. To translate these prices to positive values I use only an absolute version of this value. Also I created weekly data by selecting every 5th date in a range of five consecutive dates. By selecting the Friday closing prices to calculate
weekly prices, my data is compatible with data in other weekly datasets. Next to prices and
identifying information I also obtained information regarding delisting and a cumulative adjustment factor to adjust for stock splits and cash and stock dividends.
However the dataset contained some missing and discontinuous values without any clear information as to why these were missing or discontinuous values. For evaluating the data and calculating returns later on I needed some form of continuity. Therefore I deleted any company data that contained missing or discontinuous values without indication. To test whether this did not bias my dataset I ran a few paired t-tests on a sample of my sub-datasets. Also since I am mainly interest in variance rather than returns I ran F-tests for equality of variances. The results of these tests can be found in table 1. Here I test for any bias arising from deleting values for continuity benefits. I test whether the 51 weekly returns and variance of these returns of an equally weighted portfolio for a sample of the years investigated and for different levels of corporate social responsibility differ when I delete values for continuity practices. I took a sample of 3 different levels of CSR portfolios in three different years: the 2006 with zero corporate social responsibility portfolio, the 2008 with high corporate social responsibility portfolio and the 2009 with low corporate social responsibility portfolio. The missing and discontinuous values bias, if any ,would be expected during the years of recent financial turmoil, 2008 and 2009, since these would contain more bankruptcies. However the differences in returns for these samples are even less statistically significant with p-values of 0.49 and 0.52 . This shows that adjusting my dataset and the efficiency and continuity gained from this adjustment largely compensate for the loss of statistically insignificant information.
Table 1: t-test for sample means to test for any bias arising from excluding values
2006 2006 2008 2008 2009 2009
Zero
CSP Zero CSP High CSP High CSP Low CSP Low CSP
Excluded Including Excluded Including Excluded Including
Deleted Deleted Deleted Deleted Deleted Deleted
Values Values Values Values Values Values
Mean 0.00253 0.00265 -0.00737 -0.00728 0.00927 0.00939 Variance 0.00045 0.00044 0.00333 0.00328 0.00272 0.00276 Observations 51 51 51 51 51 51 Pearson Correlation 1.00 1.00 1.00 Hyp. Mean Difference 0 0 0 df 50 50 50 t Stat -1.64 -0.69 -0.64 P(T<=t) two-tail 0.11 0.49 0.52 t Critical two-tail 2.01 2.01 2.01
Before any adjustments made the sample contained 23360 firm-year observations and
corresponding weekly stock prices. After adjusting for missing and discontinuous values the final sample yields 22857 firm year-observations. The average amount of firm-year observations is for the zero, low and high CSP portfolios 1400, 1058 and 400 respectively.
Since there is no known portfolio that comprises all the stocks in the investible universe i.e. the market portfolio, I use a proxy to approximate the market portfolio. The proxy I use is the S&P 500 Composite Index or S&P 500. This is a value-weighted index that consists of the 500 largest US listed companies. Even though the companies used for constructing the S&P 500 also appear in my dataset, I create equally-weighted portfolios so these companies are spread out over three portfolios and appear with differing weights from the S&P 500. Therefore my portfolios are not equally risky to the market portfolio. Using the S&P 500 as a proxy for the market portfolio is consistent with existing literature.
Also there is no completely risk free interest rate. I use the annual interest rate on three-month treasury bills as a proxy. The three three-month treasury bill interest rate is essentially risk-free, however small the remote possibility of a U.S.-government default in the time period of three months still exists. Acknowledging this flaw, the interest rate on U.S.-government short-term treasury bills is sufficiently risk-free to serve as an appropriate proxy for a risk-free interest rate. I converted the annual interest rate on treasury bills to a weekly return.
The data on the performance of the S&P500 is also obtained from CRSP. The weekly returns on the S&P 500 are constructed identically to the method employed to obtain weekly stock returns, as mentioned above. Data on the 3-month treasury bill rates are obtained from Datastream.
Table 2: Descriptive statistics remaining dataset
Return Zero CSP Low CSP High CSP S&P 500 Rf
Mean 0,002056 0,001234 0,001736 0,001006 0,000257 Standard Error 0,001846 0,001821 0,001667 0,001343 0,000018 Median 0,003796 0,00201 0,00278 0,001708 0,000029 Standard Deviation 0,037289 0,036784 0,033679 0,027136 0,00036 Sample Variance 0,001391 0,001353 0,001134 0,000736 0 Kurtosis 5,431897 4,973454 6,438605 6,862817 -0,71807 Skewness -0,06572 -0,06881 -0,14307 -0,61643 1,041221 Minimum -0,18343 -0,17713 -0,1905 -0,18196 0,000002 Maximum 0,216721 0,196692 0,188614 0,120258 0,000945 Amount 408 408 408 408 408
After constructing portfolios based on corporate social performance I am left with 3 portfolios for the full sample period. The summary statistics of which are given in table 2.
IV. Methodology
In the following part I will describe which research methods and techniques I have used to both analyze and conform my raw data and acquire research results. First I will explain on which criterion I sorted my data and why this criterion was established. Second I will show which research model I used and why I have used this particular model to evaluate the variables researched.
The data I obtained from the KLD, the social ratings, are divided into three portfolios for each year of data. In the creation of these portfolios I distinguished between firms obtaining respectively a 0 score for corporate social responsibility, a non-zero score which is no larger than 3 and a score of higher or equal to 4. However since KLD provides disaggregate measures rather than one aggregate measure for total corporate social responsibility, I had to create one. But why would one want to use an aggregate measure?
In the first place it enables me to divide the data in three portfolios, all based on one statistic. This simplifies the process of portfolio selection greatly and enables me to evaluate far more data in a much greater timeframe, improving the statistical validity of my research. Also creating an
aggregate measure for corporate social responsibility makes it possible to evaluate firms uniformly on the wide range of different aspects that comprise corporate social responsibility (Bollen 2007) and is consistent with existing literature (Graves & Waddock, 1994). Furthermore even though some studies show that the use of disaggregate measures is of importance for researching the impact of corporate social responsibility on stock returns (Pavelin, Brammer, Brooks, 2006), Oikonomou, Brooks and Pavelin (2012) find that when investigating the impact of corporate social responsibility on risk rather than returns the most significant measure is an aggregate measure.
However creating an aggregate measure does involve some loss of information. A solution to confine this information loss to a minimum is to weigh the aggregate measure. For continuity I will still refer to the weighted aggregate measure as the aggregate measure. One can easily imagine that performing well on the different corporate social responsibility areas is not evenly costly for a firm to
undertake. For example doing well in the employment area may, next to being socially desirable, also be very well a sound strategy to attract and retain key personnel . This opposes for example the impact a former human rights policy in South-Africa during Apartheid has on increasing current or future profits. Since this study is on the effect of social performance on market risk and not merely on the impact of the probability of future cash-flow announcements on market risk, I doubled the scores in areas for which it can be shown that those are more costly to perform well in. But how do I decide which of the seven areas are the most costly?
Most of the decisions regarding which areas are costly can follow from logical thinking, however I have found evidence on whether these areas are costly for almost all of the seven areas in the existing literature. First of all Brammer, Brooks and Pavelin (2008) show in their study the effect on stock returns for the most important of the disaggregate measures, Environment, Community and Emloyee relations. They find a very significant negative relation between scoring well in the
Environment area and stock returns, implying that Environment scores are costly. Second they find a negative but less significant relation between Community scores and stock returns, deeming also the Community area costly. Also they find a very significant positive relation for the Employee relations area, this area is value adding and therefore not costly. Diversity is also an area that is not perceived as costly (Niederle,Segal and Vesterlund, 2013). Furthermore Corporate Governance improves transparency and reduces the principal agent problem between shareholders and management. These qualities result in good Corporate Governance policies being value adding (Berk and DeMarzo, 2007) and are therefore not costly. Regarding the Human Rights area there are very little non-zero values and even though there is no scientific proof in existing literature, Human Rights scores I do perceive as costly. There are no direct financial benefits to be obtained by allocating firm assets to pursue Human Rights goals. This is consistent with for example the weighing of the Environment area. Lastly the Product area I do not perceive as costly. One could easily follow the rationale that improving for example product safety would be value adding. To summarize: all scores in the Environment, Community and Human Rights areas are doubled, the scores in the remaining areas remain unchanged.
Since I do not investigate the direct relation between a social score and its impact on firm returns or risk, the weighing factor is only of ordinal importance. Strong policies in the Community area may not be 2 times as costly as strong policies in the Corporate Governance area, but firms that perform well in more costly areas may end up in a portfolio of more socially desirable stocks. This is consistent with the hypothesized shareholder altruism.
To determine the aggregate measure I use only social strengths and not concerns. As mentioned before I do not attempt to estimate the effect of potential future cash-flows on risk. Corporate social responsibility concerns are much more closely linked to potential future cash-flows and the expected values of these cash-flows could easily be calculated and transferred into current prices. The concerns may very well be an indicator of impending lawsuits or fines, which can be roughly estimated and are potential negative cash-flows. Corporate social responsibility strengths on the other hand do not have such obvious links with future cash-flows, since the benefits of strong social performance accrue mostly in the long run and affect the wider stakeholder environment (Cox, Brammer and Millington, 2004).
The following part evaluates the models and research methods involved in estimating market risk. The first model that I use to estimate market risk is the Capital Asset Pricing Model (CAPM) as
set out by Markowitz. In this framework the variance is the measure of risk. The variance together with the co-variance are the building blocks of the CAPM based β. Since I will use another β figure later on I will refer to the CAPM based β as the βCAPM.
𝐸𝐸(𝑟𝑟𝑖𝑖) = 𝑟𝑟𝑓𝑓+ 𝛽𝛽(𝐸𝐸(𝑟𝑟𝑀𝑀) − 𝑟𝑟𝑓𝑓)
Where: rf = risk-free interest rate E(ri)= Expected return on a stock or portfolio
E(rM ) = Expected return on the market portfolio
And β is defined as:
𝛽𝛽𝐶𝐶𝐶𝐶𝐶𝐶𝑀𝑀= 𝐶𝐶𝐶𝐶𝐶𝐶(𝑟𝑟𝑖𝑖, 𝑟𝑟𝑀𝑀) 𝑉𝑉𝑉𝑉𝑟𝑟(𝑟𝑟𝑀𝑀)
Where: Cov(ri,rM) = co-variance of a stock with the market portfolio
Var(rM) = variance of the market portfolio
The CAPM is a widely used model and by using the βCAPM the results of this study can be easily
compared with results found by similar existing or future studies.
The second measure of risk I will use is the Bawa and Lindenberg β figure. This is a measure of systematic downward risk that follows from the Mean-Lower Partial Moment framework that was developed by Bawa. This Mean-Lower Partial Moment or M-LPM framework was in the first place developed to solve the portfolio optimization problem. The M-V framework received critiques from financial theorists for being optimal only when there is a quadratic utility function which implies the highly implausible Increasing Absolute Risk Aversion or IARA leading to the framework only being optimal in 2-parameter distributions such as the normal distribution. Bawa (1975) shows that the M-LPM framework is a good approximation of the Third order Stochastic Dominance rule consistent with Decreasing Absolute Risk Aversion or DARA which is more plausible than IARA for all return distributions.
Besides being superior for solving the portfolio optimization problem the measure of risk in the M-LPM framework possesses a quality of far more importance for this study . This measure is the LPM2 or second order Lower Partial Moment. This LPM2 is given by Bawa as:
𝐿𝐿𝐿𝐿𝐿𝐿2�𝑟𝑟𝑓𝑓, 𝑋𝑋� = � �𝑟𝑟𝑓𝑓− 𝑦𝑦�2𝑑𝑑𝐹𝐹𝑥𝑥(𝑦𝑦) 𝑟𝑟𝑓𝑓
𝑎𝑎
Where: y =return on a stock or portfolio
rf =risk-free interest rate
From this equation specifying the LPM2 it is evident that the LPM2 considers downside variance. The
regular variance measure, measures all deviations from the mean of a distribution, but the LPM2
measures only those deviations that fall below a pre-specified target rate or threshold. In the Bawa Lindenberg model this threshold is the risk-free rate.
Bawa and Lindenberg (1977) develop a model for asset pricing similar to CAPM, but in a M-LPM framework. The great advantage of this model is that it is easily compared to the popular CAPM model. In this specific case the measure of LPM2 used is the LPM2(rf, rM), the Lower Partial Moment
of the second order of the return on the market portfolio.
𝐸𝐸(𝑟𝑟𝑖𝑖) = 𝑟𝑟𝑓𝑓+ 𝛽𝛽𝐿𝐿𝐶𝐶𝑀𝑀(𝐸𝐸(𝑟𝑟𝑀𝑀) − 𝑟𝑟𝑓𝑓) Where: 𝛽𝛽𝐿𝐿𝐶𝐶𝑀𝑀 =𝐶𝐶𝐿𝐿𝐿𝐿𝐿𝐿(𝑟𝑟𝑓𝑓, 𝑟𝑟𝑖𝑖, 𝑟𝑟𝑀𝑀) 𝐿𝐿𝐿𝐿𝐿𝐿2(𝑟𝑟𝑓𝑓, 𝑟𝑟𝑀𝑀) And: 𝐶𝐶𝐿𝐿𝐿𝐿𝐿𝐿2�𝑟𝑟𝑓𝑓; 𝑟𝑟𝑀𝑀; 𝑟𝑟𝑖𝑖� = � � �𝑟𝑟𝑓𝑓− 𝑟𝑟𝑀𝑀�2�𝑟𝑟𝑓𝑓− 𝑟𝑟𝑖𝑖�𝑑𝑑𝐹𝐹(𝑟𝑟𝑀𝑀, 𝑟𝑟𝑖𝑖) ∞ 𝑟𝑟𝑖𝑖=−∞ 𝑟𝑟𝑓𝑓 𝑟𝑟𝑀𝑀=−∞
However the integrals used in the definition of the LPM2 and CLPM2 measures are not very practical
therefore I use the more widely used specifications of these measures as is consistent with existing literature (Oikonomou, Brooks, and Brammer, 2012). These are:
𝐿𝐿𝐿𝐿𝐿𝐿2(𝑟𝑟𝑓𝑓, 𝑟𝑟𝑀𝑀) = 𝑇𝑇 �[min�𝑟𝑟1 𝑀𝑀− 𝑟𝑟𝑓𝑓, 0�]2 𝑇𝑇 𝑖𝑖=1 And: 𝐶𝐶𝐿𝐿𝐿𝐿𝐿𝐿2�𝑟𝑟𝑓𝑓; 𝑟𝑟𝑀𝑀; 𝑟𝑟𝑖𝑖� = 1𝑇𝑇 �(𝑟𝑟𝑖𝑖− 𝑟𝑟𝑓𝑓)min [( 𝑇𝑇 𝑖𝑖=1 𝑟𝑟𝑀𝑀− 𝑟𝑟𝑓𝑓),0]
These are the models used for assessing systematic or market risk in this study. Please note that I do not use any control variables so only a limited part of the variation in portfolio returns may be explained by the models used.
After I construct the portfolios I calculate the corresponding βLPM and the βCAPM of the portfolios for
the entire 2006-2013 period. Although this method might give some useful insight, the amount of figures this method produces is so small, that any statistical test is unverified.
Therefore I employ the following model :
𝛽𝛽𝑖𝑖 = 𝛼𝛼0+ 𝛼𝛼1𝐿𝐿𝐶𝐶𝐿𝐿 + 𝛼𝛼2𝐻𝐻𝐻𝐻𝐻𝐻ℎ + 𝜀𝜀𝑖𝑖
Where: Low = Dummy variable equal to 1 if a stock is in the low-CSP portfolio
High = Dummy variable equal to 1 if a stock is in the high-CSP portfolio βi = The βLPM or βCAPM of firm i, dependent on the hypothesized
relationship. ε = Error term
Earlier on in this paper I developed several research hypotheses.
Hypothesis 1: A portfolio of stocks with higher corporate social performance is subject to less market risk than a portfolio of stocks with lower corporate social performance.
The hypothesized relationship in hypothesis 1 translates to the following statistical hypotheses: Hypothesis 1a: 𝐻𝐻0: 𝛽𝛽0𝐶𝐶𝐶𝐶𝐶𝐶𝑀𝑀− 𝛽𝛽𝐿𝐿𝐿𝐿𝐿𝐿𝐶𝐶𝐶𝐶𝐶𝐶𝑀𝑀= 0 𝐻𝐻1: 𝛽𝛽0𝐶𝐶𝐶𝐶𝐶𝐶𝑀𝑀− 𝛽𝛽𝐿𝐿𝐿𝐿𝐿𝐿𝐶𝐶𝐶𝐶𝐶𝐶𝑀𝑀≠ 0 Hypothesis 1b: 𝐻𝐻0: 𝛽𝛽𝐿𝐿𝐿𝐿𝐿𝐿𝐶𝐶𝐶𝐶𝐶𝐶𝑀𝑀− 𝛽𝛽𝐻𝐻𝑖𝑖𝐻𝐻ℎ𝐶𝐶𝐶𝐶𝐶𝐶𝑀𝑀= 0 𝐻𝐻1: 𝛽𝛽𝐿𝐿𝐿𝐿𝐿𝐿𝐶𝐶𝐶𝐶𝐶𝐶𝑀𝑀− 𝛽𝛽𝐻𝐻𝑖𝑖𝐻𝐻ℎ𝐶𝐶𝐶𝐶𝐶𝐶𝑀𝑀≠ 0 Hypothesis 1c: 𝐻𝐻0: 𝛽𝛽0𝐶𝐶𝐶𝐶𝐶𝐶𝑀𝑀− 𝛽𝛽𝐻𝐻𝑖𝑖𝐻𝐻ℎ𝐶𝐶𝐶𝐶𝐶𝐶𝑀𝑀= 0 𝐻𝐻1: 𝛽𝛽0𝐶𝐶𝐶𝐶𝐶𝐶𝑀𝑀− 𝛽𝛽 𝐻𝐻𝑖𝑖𝐻𝐻ℎ𝐶𝐶𝐶𝐶𝐶𝐶𝑀𝑀≠ 0 The next hypothesis developed is hypothesis 2.
Hypothesis 2: A portfolio of stocks with higher corporate social performance is specifically subject to less downside market risk as compared to a portfolio of stocks with lower corporate social performance.
Hypothesis 2a: 𝐻𝐻0: 𝛽𝛽0𝐿𝐿𝐶𝐶𝑀𝑀− 𝛽𝛽𝐿𝐿𝐿𝐿𝐿𝐿𝐿𝐿𝐶𝐶𝑀𝑀= 0 𝐻𝐻1: 𝛽𝛽0𝐿𝐿𝐶𝐶𝑀𝑀− 𝛽𝛽𝐿𝐿𝐿𝐿𝐿𝐿𝐿𝐿𝐶𝐶𝑀𝑀 ≠ 0 Hypothesis 2b: 𝐻𝐻0: 𝛽𝛽𝐿𝐿𝐿𝐿𝐿𝐿𝐿𝐿𝐶𝐶𝑀𝑀− 𝛽𝛽𝐻𝐻𝑖𝑖𝐻𝐻ℎ𝐿𝐿𝐶𝐶𝑀𝑀 = 0 𝐻𝐻1: 𝛽𝛽𝐿𝐿𝐿𝐿𝐿𝐿𝑀𝑀𝐿𝐿𝐶𝐶𝑀𝑀− 𝛽𝛽𝐻𝐻𝑖𝑖𝐻𝐻ℎ𝑀𝑀𝐿𝐿𝐶𝐶𝑀𝑀≠ 0 Hypothesis 2c: 𝐻𝐻0: 𝛽𝛽0𝐿𝐿𝐶𝐶𝑀𝑀− 𝛽𝛽𝐻𝐻𝑖𝑖𝐻𝐻ℎ𝐿𝐿𝐶𝐶𝑀𝑀 = 0 𝐻𝐻1: 𝛽𝛽0𝑀𝑀𝐿𝐿𝐶𝐶𝑀𝑀− 𝛽𝛽𝐻𝐻𝑖𝑖𝐻𝐻ℎ𝐿𝐿𝐶𝐶𝑀𝑀 ≠ 0
To examine the third hypothesis proposed, I need some adjustment to the model described above to discriminate between observations in more and less volatile periods. The adjusted model is:
𝛽𝛽𝑖𝑖 = 𝛼𝛼0+ 𝛼𝛼1𝐶𝐶𝑟𝑟𝐻𝐻𝐶𝐶𝐻𝐻𝐶𝐶 + 𝜀𝜀𝑖𝑖
Where: Crisis = Dummy variable that equals 1 when the observation is in a more volatile period.
When the sample consists of either βCAPM or βLPM measures that belong to a different level of
corporate social performance, thhis adjustment enables me to investigate the hypothesized relationship in hypothesis 3:
Hypthesis 3: The hypothesized reduction in market risk is larger for more volatile periods.
This translates to the following statistical hypotheses for all 3 levels of corporate social performance: Hypothesis 3a: 𝐻𝐻0: 𝛽𝛽𝑁𝑁𝐿𝐿𝐶𝐶𝑟𝑟𝑖𝑖𝑁𝑁𝑖𝑖𝑁𝑁𝐿𝐿𝐶𝐶𝑀𝑀 − 𝛽𝛽 𝐶𝐶𝑟𝑟𝑖𝑖𝑁𝑁𝑖𝑖𝑁𝑁𝐿𝐿𝐶𝐶𝑀𝑀 = 0 𝐻𝐻1: 𝛽𝛽𝑁𝑁𝐿𝐿𝐶𝐶𝑟𝑟𝑖𝑖𝑁𝑁𝑖𝑖𝑁𝑁𝑀𝑀𝐿𝐿𝐶𝐶𝑀𝑀 − 𝛽𝛽 𝐶𝐶𝑟𝑟𝑖𝑖𝑁𝑁𝑖𝑖𝑁𝑁𝐿𝐿𝐶𝐶𝑀𝑀 ≠ 0 Hypothesis 3b: 𝐻𝐻0: 𝛽𝛽𝑁𝑁𝐿𝐿𝐶𝐶𝑟𝑟𝑖𝑖𝑁𝑁𝑖𝑖𝑁𝑁𝐶𝐶𝐶𝐶𝐶𝐶𝑀𝑀 − 𝛽𝛽𝐶𝐶𝑟𝑟𝑖𝑖𝑁𝑁𝑖𝑖𝑁𝑁𝐶𝐶𝐶𝐶𝐶𝐶𝑀𝑀= 0 𝐻𝐻1: 𝛽𝛽𝑁𝑁𝐿𝐿𝐶𝐶𝑟𝑟𝑖𝑖𝑁𝑁𝑖𝑖𝑁𝑁𝐶𝐶𝐶𝐶𝐶𝐶𝑀𝑀 − 𝛽𝛽𝐶𝐶𝑟𝑟𝑖𝑖𝑁𝑁𝑖𝑖𝑁𝑁𝐶𝐶𝐶𝐶𝐶𝐶𝑀𝑀≠ 0
Since the remaining four hypotheses are identical in all but the sample they are tested on, I see no need to state these as well. These hypotheses will be labelled alphabetically as the ones stated above. By testing all ten aforementioned statistical hypotheses, I am able to test the research hypotheses and in turn the research question of this paper.
V. Results
The first results to discuss are the βCAPM and βLPM figures calculated for the equally weighted
portfolios over the whole length of the dataset. Although these figures can not be tested for differences, they do present a more general insight in the relationship between market risk and corporate social performance. As evident from table 3 the figures calculated hint a risk reduction for the portfolios formed from stocks with high corporate social performance scores. But whereas portfolios with high scoring social performers seem to obtain the hypothesized risk reduction, there seems to be no difference between some low level of CSR and no CSR at all. This would be, if statistically verified, a peculiar finding.
Table 3: β figures for the equally weighted portfolios constructed of different levels of CSR
βCAPM βLPM
Zero 1,2841*** 1,2427
Low 1,2807*** 1,2523
High 1,2096*** 1,1741
Table 4 models the results from testing hypothesis 1. To recall:
Hypothesis 1: A portfolio of stocks with higher corporate social performance is subject to less market risk than a portfolio of stocks with lower corporate social performance.
To statistically verify this hypothesis I calculate all of the individual β figures for the stocks that comprise the different portfolios. The regression yields the following results. The βCAPM figures for
stocks that have zero CSR are on average over the whole period 1,2756. The βCAPM figures for stocks
with low but non-zero CSP scores are only very slightly higher and this result is insignificant. The βCAPM figures for the highest levels of CSR differ significantly from the figures for zero and low
performers. The realized reduction in market risk for high corporate social performance stocks is 0,1079, meaning that on average, high corporate social performance results in a level of market risk nearly 8,5% lower than that of stocks with zero CSR. The results confirm the hypothesized
relationship of hypothesis 1.
Table 4 : regression of βCAPM on different levels of CSR
βCAPM Coef. Rob. Std. Err. t P>|t| [95% Conf. Interval]
Low 0,0021 0,0097 0,22 0,826 -0,0169 0,0211
High -0,1079*** 0,0118 -9,16 0,000 -0,1310 -0,0848
constant (Zero) 1,2756*** 0,0062 207,17 0,000 1,2635 1,2877
R-squared 0.0033
The second hypothesis yields the results presented in table 5 . The hypothesized relationship is the following:
Hypothesis 2: A portfolio of stocks with higher corporate social performance is specifically subject to less downside market risk as compared to a portfolio of stocks with lower corporate social performance.
The results in table 5 show that there is also when employing measures of downside risk a reduction in market risk for companies which perform well in social responsibility areas. This reduction in downside market risk is nearly twice as big as the reduction in conventional market risk. The figures in table 5 indicate that stocks associated with high levels of corporate social performance suffer significantly less downward deviations than stocks which are associated with poor social
performance. The reduction as indicated by the coefficient is 0,2058, which implies that stocks screened for high levels of corporate social performance are not much riskier than a well diversified market portfolio. In addition the figures in table 5 indicate that stocks of companies that do not engage in any level of corporate social responsibility are on average 1,26 times as risky as a well diversified market portfolio.
Table 5 : regression of βLPM on different levels of CSR
βLPM Coef. Rob. Std. Err. t P>|t| [95% Conf. Interval]
Low 0,0355* 0,0191 1,86 0,063 -0,0019 0,0730
High -0,2058*** 0,0147 -14,00 0,000 -0,2347 -0,1770
constant (Zero) 1,2603*** 0,0085 148,80 0,000 1,2437 1,2769
R2 = 0.0045
*significant at the 0.1 level.
However insignificant at the 1% and the 5%, but now significant at the 10% level, the coefficient for stocks from companies that are engaged in some low leve of corporate social performance is again positive. So the results suggest that however small, there is some increased risk of losses when investing in companies that have very low but non-zero corporate social responsibility. The results support research hypothesis 2 in the sense that stocks associated with high levels of CSR are
specifically subject to less downside market risk. However as seen with the testing for differences in βCAPM for the various levels of CSR, if anything ,there is an increase in market risk for low but non-zero
levels of CSR.
To test the third research hypothesis I use the model described in the methodology section. In this model the only explanatory variable for any difference in β figures is the presence of highly volatile times, represented by the dummy variable “Crisis”. This means that the model lacks even more explanatory power as will be evident by the low R2 figures of the regressions. However the method
does uncover some indications for differences in β figures in periods of high and lower volatility. The results from the regressions testing research hypothesis 3 are presented in tables 6 to 11.
Hypothesis 3: The hypothesized reduction in market risk is larger for more volatile periods.
Table 6 regression of βCAPM for the Zero CSR Portfolio on Crisis
βCAPM Coef. Rob. Std. Err. t P>|t| [95% Conf. Interval]
Crisis 0,0648*** 0,01276 5,08 0 0,0398 0,0898
Constant (No Crisis) 1,2054*** 0,00984 122,48 0 1,1861 1,2247
R2 = 0.0023
In table 6 the results for the regression of the βCAPM of the portfolio comprised of stocks with zero
corporate social responsibility are presented. Both coefficients estimated are significant, but the increase in market risk, as measured by the coefficient for the Crisis variable, is rather small; only 0,06479.
Table 7 models the results of the regression for exactly the same portfolio as table 6 but the regression now measures the relation between more volatile times and its effect on market risk as measured by the βLPM. The coefficients that the regression yields are both significant, so there exists
a clear difference between highly volatile and more stable periods. This is a somewhat intuitive finding when using downside market risk as the measure of market risk. The coefficient for the Crisis dummy is 0,091384 in this regression. Implying that stocks with zero corporate social responsibility are subject to 7,8% more downside market risk under bearish conditions.
Table 7: regression of βLPM for the Zero CSR Portfolio on Crisis
βLPM Coef. Rob. Std. Err. t P>|t| [95% Conf. Interval]
Crisis 0,0914*** 0,01803 5,07 0 0,0560 0,1267
Constant (No Crisis) 1,1732*** 0,01531 76,62 0 1,1432 1,2032
R2 = 0.0024
Table 8: regression of βCAPM for the Low CSR Portfolio on Crisis
CAPM Coef. Rob. Std. Err. t P>|t| [95% Conf. Interval]
Crisis 0,0569*** 0,0151 3,76 0 0,0272 0,0866
Constant (No Crisis) 1,2600*** 0,0108 116,79 0 1,2389 1,2812
R2 =0.0016
Table 9: regression of βLPM for the Low CSR Portfolio on Crisis
BL Coef. Rob. Std. Err. t P>|t| [95% Conf. Interval]
Crisis 0,3268*** 0,0318 10,27 0 0,2644 0,3891
Constant (No Crisis) 0,9842*** 0,0296 33,2 0 0,9261 1,0423
R2 = 0.0100
Subsequently tables 8 and 9 model the results for the regression of βCAPM and the βLPM figures on the
Crisis dummy for stocks with low CSR scores. For the conventional βCAPM table 8 yields some
interesting results. In periods absent of high volatility the market risk for stocks that have some non zero level of corporate social responsibility appears to be higher than for stocks that have zero corporate social performance. The coefficient for the low stocks is 1,26 and is significant at the 1% level. For comparison: the coefficient for the same period of relatively low volatility for the zero CSR stocks is 1,2054. However the increase in market risk in the presence of financial turmoil is slightly smaller than for stocks with zero CSR. The results are consistent with results in table 4 that maps the performance of the stocks with low social performance over the entire period.
The regression coefficients in table 9 show that during years of relative stability the downside market risk of stocks associated with low but non-zero corporate social performance as measured by the constant of 0,9842 do not differ that much from a well diversified unscreened portfolio. More specifically, the 95% confidence interval includes the value 1,00; implying that there is no statistical difference at the 5% level. However in periods that are characterized by higher volatility there is a large increase in downside market risk. The coefficient of the crisis variable is 0,3268 and significant at the 1% level. These results insinuate that also when using downside risk, stocks linked to low but non-zero CSP, are riskier than stocks associated with no CSR at all in periods of higher volatility.
Table 10: regression of βCAPM for the High CSR Portfolio on Crisis
βCAPM Coef. Rob. Std. Err. t P>|t| [95% Conf. Interval]
Crisis 0,0712*** 0,0199 3,58 0 0,0322 0,1101
Constant (No Crisis) 1,1210*** 0,0138 81,15 0 1,0939 1,1481
Table 10 yields the results for the regression of the βCAPM for stocks with the highest levels of CSR.
The regression yields an increase in market risk of 0,0712 in periods of high volatility, significant at the 1% level. This finding is consistent with results in table 4 that estimates the coefficient for this portfolio over the entire period. In periods of relatively low volatility the market risk as measured by the βCAPM is lower than the coefficients found for the other two levels of CSR.
Finally table 11 yields the results for the same regression as table 10 but using the measure of downside risk as independent variable. Both coefficients are significant at the 1% level. Both coefficients are also very interesting results. Remember that hypothesis 3 builds on the research of Oikonomou, Brooks and Brammer (2012), in which the authors find that the relationship between risk and CSR is stronger in more volatile periods. The coefficients found in table 11 show that there is less downside market risk in a high CSR portfolio in periods of low volatility, as indicated by the coefficient of the constant of only 0,8682. This coefficient is significantly lower than 1, which is the coefficient of the market portfolio.
In times of financial crisis there is a large increase in downside market risk. This is somewhat logical since there are more downward deviations in periods of higher periods of economic
downturn. But the found coefficient of 0,3247 implies that the downside risk reduction that high levels of CSR may cause is mainly a result of very low downside risk in periods of lower volatility. This finding suggest that any wealth protective effects of high CSR might be mainly limited to more stable periods.
Table 11: regression of βLPM for the High CSR Portfolio on Crisis
ΒLPM Coef. Rob. Std. Err. t P>|t| [95% Conf. Interval]
Crisis 0,3247*** 0,02433 13,35 0 0,2770 0,3724
Constant (No Crisis) 0,8682*** 0,01916 45,33 0 0,8307 0,9058
R2 = 0.0542
In summary the hypothesized relationship between market risk and CSR in volatile periods is not supported by the empirical evidence. The results show a similar risk increase in periods of high volatility for differing levels of CSR . That is the results do not show any larger reduction in market risk in periods of high volatility for stocks with higher levels of CSR. If anything, the results indicate that the reduction in market risk associated with high levels of CSR is highest in periods of low volatility. The wealth protection effects of CSR seem to be mitigated in periods of financial crisis to some level of market risk only slightly lower than the worst social performers.
VI. Conclusion and Discussion
After evaluating the results of testing research hypothesis 1 I can conclude that there is a clear reduction in market risk as measured by the βCAPM for stocks associated with the highest levels of
corporate social performance. The significant coefficient of -,1079 clearly shows that investing in stocks with higher levels of CSR is less risky than investing in stocks with no CSR at all.
Also when investigating the relationship between downside market risk and CSR I find results that support the hypothesized risk reduction. The coefficient of -0.2058 for stocks with the highest levels of CSR confirms that these stocks are very much less risky than stocks associated with either zero or low levels of CSR. These results support my theory regarding altruistic investor utility
functions and institutional investment in CSR stabilizing demand for investments in companies that achieve high levels of CSR.
However this relationship is not straightforward. I find that when evaluating levels of CSR and their effects on the riskiness of stocks it is better to have no CSR than to have a low level of CSR. The regressions yield a non-significant but positive relationship between market risk and a low level of CSR for conventional risk measures and a positive significant at the 10% level relationship between low levels of CSR and downside market risk. So how could this be?
There might be an explanation for this phenomenon. If low levels of CSR correspond to companies only trying to reach the bare minimum necessary to not be avoided by socially responsible investors, than the altruistic utility that such an investor may derive from the investment might be no higher than from investing in a company that does not entail any CSP at all. Metaphorically said: socially responsible investors might rather not share their lunch at all than giving away some crumbs. However investigating the exact nature of this relationship is confined to behavioural sciences and should be subject of future research.
When evaluating the effects of highly volatile periods on the hypothesized risk reduction by CSR I find mixed results. When the βCAPM is used as measure of market risk I find some differences across the
three levels of CSR investigated, but the increase in market risk as a result of a financial crisis is quite similar for the three levels of CSR. However in periods of low volatility there is a clear reduction in market risk for high levels of CSR.
When using βLPM as a measure of market risk very large reductions in market risk of over 30% can be
achieved by investing in stocks with high levels of CSR in periods of low volatility. However in a period of financial crisis only a small reduction of some 6% lower downside market risk remains from
investing in stocks with the highest levels
These results suggest that investors can achieve large reductions in market risk in more stable periods by investing in stocks with high CSR, but that there are much less risk reduction gains from this investment strategy in more unstable periods. Therefore I conclude that the risk reduction is, if anything, smaller in periods of higher volatility.
This in inconsistent with the results found in existing literature, but could be consistent with my theory of investor utility. Investors might for the sake of altruism be much more tolerant of downward deviations in prosperous economic times for investments which have high corporate social responsibility, but might very well dislike huge losses more than any utility gains from altruism. The same goes for institutional investors which must account for their social investment strategies, but even more so for big losses. This would explain the large increase in market risk in more volatile times and the risk reduction in more stable periods.
To answer the research question of this paper: stocks associated with higher levels of corporate social responsibility are indeed subject to less market risk than stocks with lower levels of corporate social responsibility.
A major limitation of this research is the exclusion of any control variables. Although it can be argued that incorporating some control variables would still not make the models employed perfect, my models lack explanatory power as evident from the low R2. Also for example including firm size in the
not only carry less market risk, but might also have higher levels of CSR than young fast growing firms. Future research could focus on isolating the corporate social responsibility component to model the exact contribution of high CSR to risk reduction.
Other research might also concentrate on the non-straightforward relationship between CSR and risk. Why do stocks with low levels of CSR seem to be subject to higher market risk? And would these companies be better off in equity markets by a different level of corporate social
References
Ang, A., J. Chen, and Y. Xing. “Downside Risk.” Review of Financial Studies, 19 (2006), 1191-1239. Bawa, V. “Optimal Rules for Ordering Uncertain Prospects.” Journal of Financial Economics, 2
(1975), 95–121.
Bawa, V., and E. B. Lindenberg. “Capital Market Equilibrium in a Mean-Lower Partial Moment Framework.” Journal of Financial Economics, 5 (1977), 189–200.
Berk, J., and P. DeMarzo. “Corporate Finance.” Pearson Education (2007), 917-921
Bollen, N. “Mutual Find attributes and investor behaviour.” Journal of Financial and Quantative
Analysis, 42 (2007), 683-708.
Brammer, S., C. Brooks, and S. Pavelin “Corporate Social Performance and Stock Returns: UK Evidence from Disaggregate measures.” Financial Management, 3 (2006), 97-116. Cox, P., S. Brammer,and A. Millington. “An empirical examination of institutional investor
preferences for corporate social performance.” Journal of Business Ethics, 1(2004), 27-43. Friedman, M. “The Social Responsibility of Business is to Increase Its Profits” The New York Times
Magazine, 32 (1970), 122-126.
Geczy, C. C., R. F. Stambaugh, and D. Levin. “Investing in Socially Responsible Mutual Funds.” Working paper (2005)Available at SSRN:
http://papers.ssrn.com/sol3/papers.cfm?abstract_id=416380
Goldreyer, E. F., P. Ahmed, and D. J. Diltz. “The Performance of Socially Responsible Mutual Funds: Incorporating Sociopolitical Information in Portfolio Selection.” Managerial Finance, 1 (1999), 23-36.
Graves, S. B., and S. A. Waddock. “Institutional Owners and Corporate Social Performance.” The
Academy of Management Journal, 4 (1994), 1034-1046.
Hamilton, S., H. Jo, and M. Statman. “Doing Well while Doing Good? The Investment Performance of Socially Responsible Mutual Funds.” Financial Analysts Journal, 6 (1993), 62-66.
Kahneman, D., and A. Tversky. “Prospect Theory: An Analysis of Decision Under Risk.”
Econometrica, 47 (1979), 263-291.
Lintner, J. “The Valuation of Risk Assets and the Selection of Risky Investments in Stock Portfolios and Capital Budgets.” Review of Economics and Statistics, 47 (1965), pp. 13–37.
Luo, X., and C.B. Bhattacharya. “The Debate over Doing Good: Corporate Social Performance, Strategic Marketing Levers and Firm-Idiosyncratic Risk.” Journal of Marketing, 73 (2009), 198-213
Markowitz, H. “Portfolio Selection.” Journal of Finance, 7 (1952), 77-99.
Performance.” Academy of Management Journal, 31 (1988), 854-872.
Oikonomou, I., C. Brooks, and S. Brammer. “The Impact of Corporate Social Performance on Financial Risk and Utility: A Longitudinal Analysis.” Financial Management, 2 (2012), 483-515.
Post, T., and P. van Vliet. “Conditional Downside Risk and the CAPM.” Working paper (2004) Available at SSRN: http://papers.ssrn.com/sol3/papers.cfm?abstract_id=797286
Renneboog, L., J. Ter Horst,and Zhang, C. “The price of ethics and stakeholder governance: the performance of socially responsible mutual funds.” Journal of Corporate Finance, 14 (2008), 302-322.
Roy, A. D., “Safety First and the Holding of Assets,” Econometrica, 20 (1952), 431-449.
Salama, A., K. P. Anderson, and S. Toms. “Does Community and Environmental Responsibility Affect Firm Risk? Evidence form UK Panel Data 1994-2006.” Business Ethics: A European Review, 20 (2009), 192-204
Sharfman, M. P., and C. S. Fernando. “Environmental Risk Management and the Cost of Capital.”
Strategic Management Journal, 29 (2008), 569-592.
Sharpe, W. F. “Capital Asset Prices: A Theory of Market Equilibrium Under Conditions of Risk.”
Journal of Finance, 19 (1964), 425-441.
Spicer, B. H. “Investors, Corporate Social Performance and Information Disclosure: An Empirical Study.” Accounting Review, 53 (1978), 94-111.
