Does being ethical pay in the banking industry?I II
Xuehan CuiIII Abstract In this study, we investigate the cost and benefit of being ethical compared to being unethical in the banking industry by discovering the market’s reaction upon corporate social responsibility (CSR) news. Specifically, we are interested in whether bank equities exhibit abnormal returns when exposed to CSR news. We find out that banks are mainly punished for negative CSR practices related to indirect economic impact, external initiatives and social impact (or compliance) of products. It is also found out that the comparative benefit of being ethical vs. being unethical increases during market downturns than during normal times, since the expected return of a market neutral hedge fund, with long positions in positive CSR news and short positions in negative CSR news, is much higher in the regime when the MSCI World Bank Index exhibits low mean return and high volatility. Keywords: Corporate Social Responsibility (CSR), Event Study, Market Regime
I DISCLAIMER: Covalence employs students and young graduates as intern ethical information analysts in partnership with various universities. During their 2 to 4 months’ internship analysts have the opportunity to conduct a research on a topic of their choice. They can present their findings during a staff meeting and write an article that may be published on Covalence website. These articles reflect the intern analysts’ own views, opinions and methodological choices, and are published under the responsibility of their individual author. II This is the second draft of the study, first written in December, 2012. III Student of MS in “Management” at Università Bocconi. Email: [email protected]
Table of Contents
1. Introduction ..................................................................................................................................... 1 2. Literature Review .......................................................................................................................... 2 2.1. Issues Related to Event Study ......................................................................................................... 2 2.2. Issues Related to Accounting Measure of Financial Performance ................................... 3
3. Data and Methodology ................................................................................................................. 3 4. Empirical Analysis: Bank-‐‑relevant Criteria ........................................................................ 5 4.1. Overview of Event Entries ................................................................................................................ 5 4.2. Standard Event Study .......................................................................................................................... 7 4.3. Relatively important CSR Criteria .................................................................................................. 8
5. Empirical Analysis: Cost & Benefit and Regime Dependence ................................... 12 5.1. Ethical and Evil Strategy ................................................................................................................. 12 5.2. Hedge Fund Returns’ Dependence on Market Regimes .................................................... 14
6. Summary and Conclusion ......................................................................................................... 17 Reference .............................................................................................................................................. 19 Appendix: Criteria and Definition ................................................................................................ I
1
1. Introduction Since the birth of “corporate social responsibility” in the 1970s, both firms and the society have experienced a complex and iterative learning path. Today most firms acknowledge the importance associated with CSR management and incorporate it into their decision making. Investors increasingly embrace socially responsible investing (SRI), where many non-‐‑financial aspects, such as environmental, social and corporate governance (ESG), are emphasized in the investment process. According to the Forum for Sustainable and Responsible Investment (US SIF), as of the beginning of 2014, $6.57 trillion are invested following the SRI strategy, representing 18% of the universe of assets under professional management. With the rising demand of investors, ethical indices are created, such as the FTSE4Good Index and the Dow Jones Sustainability World Index. While CSR practice recognized its importance in firms of many industries, it is particularly interesting for us to focus on the banking sector. Banks, being the most important player in the financial market, are expected to assume an inescapable responsibility in the healthiness and soundness of the growth of the world economy. Banks’ ethical or unethical behavior can thus indicate whether they are able to live up to this expectation. Everyday, investors are faced with all kinds of CSR-‐‑related news, upon which they (especially those following SRI strategy) may decide to increase or decrease exposures to some particular stocks, uplifting or dampening the (short-‐‑term) equity returns. With the aim of finding out whether market rewards or punishes banks with abnormal returns (AR) when confronted with positive and negative CSR news, this study makes focus on two research questions. First, what are the categories of CSR practices to which the banking sector is most susceptible? Since banks differ from other industrial firms, it can be expected that CSR information of some categories are more bank-‐‑relevant than others, and therefore can have a stronger impact on equity returns. For instance, banks that undertake initiatives detrimental to the environment or biodiversity probably suffers less than the revealing of corruptive behavior (such as Libor manipulation), because the latter reflects a flaw in the bank’s ethics and governance that may have continued for years and that will probably result in a severe fine; in the extreme case, when the situation is so acute as to have a strong negative social and economic impact, equities may suffer a great loss in the face of government intervention and court prosecution. Given this consideration, it is necessary to discover the bank-‐‑relevant CSR categories. This analysis can not only reveal the perspective of socially responsible investors (i.e. what are the types of CSR practices in the banking sector upon which their investment decisions depend mostly), but also shed light on the efficient CSR management for banks. Secondly, is there a comparative benefit of being ethical vs. being unethical? Many studies have analyzed the relation between firm’s CSR performance and the financial performance. Various methodologies and techniques have been employed, with result differing from study to study. In this analysis, we investigate the cost and benefit by formulating two trading strategies entirely based on CSR related information, with one strategy rebalancing the portfolio upon positive CSR information and the other upon negative information. The intuition is clear: if the market rewards banks for being ethical and punishes them for being unethical, we would expect the strategy trading on positive information outperforms the other. This analysis is made possible thanks to the dataset provided by Covalence EthicalQuote, a Geneva based CSR consulting firm who aggregates hundreds of thousands of documents extracted from diverse sources and classified into 50 sustainability criteria (categories) inspired by the Global Reporting Initiative. There are two main contributions of this study: first, it provides evidence of the CSR impact exclusively within the banking sector; and secondly, it draws attention on the CSR impact in different market regimes. As will become clear later, we find the cost and benefit between “being ethical” and “being unethical” is most evident during market downturns. This provides empirical evidence on the importance associated with ESG aspects in banks, and incentivizes them to take active and adequate initiatives when it comes to CSR management. This also justifies investment strategies following SRI, which is mostly likely to be successful in market downturns.
2
This study is organized as follows: in the next part, we will review empirical studies on the relationship between firms’ CSR performances and financial performances, a focus will be placed on the issues commonly encountered in such analysis and how we will treat them in this study; a brief discussion of our dataset and methodology is present in the part 3; part 4 and 5 will each deal with one of our research questions and the last part summarizes the main conclusions in this study.
2. Literature Review Empirical studies on the relationship between CSR and financial performance concluded different results. For example, Moskowitz (1972) selected 14 firms with good social responsibility credentials and compared their rates of return on the first half of 1972 with that of common stock index return and found they were appreciated; Waddock and Graves (1997) used social rating data of 469 firms and found a positive relation between social performance and financial performance, though the causation was unclear; Preston and O’Bannon (1997) used data for 67 large US stocks from 1982 to 1992, and showed there is no significant relationship between CSR and financial performance; Ainscough et. Al (2007) found the impact of CSR on abnormal returns was different among US, Europe and Asian, thus suggesting the difference was due to culture differences and should be converging in long term with globalization deepening. Empirical approaches can mainly be separated into two streams. The first is represented by the application of event study that aims to detect the short-‐‑term abnormal returns on the release of unanticipated information (see for example, Wright and Ferris (1997), Teoh et. Al (1999)). The second entails examining the relationship between CSR and long-‐‑term financial performance measured by accounting and financial numbers (see for example, Waddock and Graves (1997)). The inconsistency of results of both streams of studies have attracted researchers to examine the validity of assumptions and defects inherited in the research design. 2.1. Issues Related to Event Study The validity of semi-‐‑strong efficient-‐‑market hypothesis (EMH). This is the fundamental assumption underlying the event study approach, which implies that market prices reflect all publicly available information, so any new information coming to the market will be quickly reflected in the stock prices (McWilliam and Siegel, 1997). While in this study we do not prove the validity of EMH (when testing the EMH, one encounters the “joint hypothesis problem”, which states that it is never possible to prove or disprove the EMH, since inevitably one has to choose the asset pricing model), we do account for factors that may improve market efficiency. For instance, it is empirically proved that small “size” can lead to more anomalies and dampen the market efficiency (Brown and Warner, 1985). It is therefore not advisable to aggregate events occurred to firms of different market capitalization and liquidity. Since our dataset includes 30 of the largest banks/banking groups worldwide, we can be reasonably certain of the validity of EMH. Unanticipated events. This assumption requires that events are not anticipated or leaked prior to the formal announcement of the event, and implies that the event window can start after the event and not before it. Studies have shown that major corporate events such as earnings announcements and take-‐‑over announcements already have an impact on stock prices preceding the official date of release. The likelihood of information leakage and insider trading should depend on the financial relevance (profitability of taking advantage) of the events. Since CSR related news generally refer to non-‐‑financial aspects of banks, the likelihood of anticipation is at low level. Length of the event window and confounding Events. The choice of the event window length of empirical studies are rather arbitrary, ranging from several days to months. It is widely accepted that longer event window can reduce the test statistics, underestimating the power of the event has over the stock prices (Brown and Warner,1985). Longer event window also raises the
3
concern of eventual confounding effects included in the event window, making it difficult to claim that price change is all result of the event. The selection of event window length is a trade-‐‑off between being long enough to capture the significant effects and being short enough to exclude confounding effects. Common practices are either choosing a fixed event window (standard approach) or making it varying case by case (ad-‐‑hoc approach). As will be discussed later, we will choose the second approach, given the characteristics of the dataset and the nature of this study. 2.2. Issues Related to Accounting Measure of Financial Performance Reliability of accounting measure. Since financial reports are subject to possible manipulation by the reporting firm, the reliability of the accounting figure as measurement of financial performance is of high concern. Misspecifications of the true performance is a disadvantage compared to event study, which takes the price readily from market reflecting all expected cash flows from the future. Omitted variable bias. When regressing the financial performance, cautions should be made when selecting viable explanatory variables. There has been a long standing theoretical literature demonstrating that R&D has a strong impact on the future profitability (Lichtenberg and Siegel (1991), Hall (1999)). As a result, variables highly correlated with R&D (such as CSR) will be exposed to the omitted variable bias if R&D is excluded from the model. As is shown in McWilliams and Siegel (2000), the inclusion of R&D reduces the significance of the CSR coefficient, leading them to conclude that model misspecification is one source of mixed results in the CSR literature.
3. Data and Methodology Retrieved from the Covalence EthicalQuote databank, our dataset is a subset covering the banking sector, within the period from Jan 1, 2002 to Dec 20, 2012. The dataset contains news articles and documents extracted from 21 types of sources including press, trade unions and international organizations, etc. The news entries include the following information: date of release, news title, positive or negative, banks involved in the news, and a classified criterion that the news belongs to. A total number of 50 criteria are identified. Analysts are required to classify each news into 1, 2 or maximum 3 criteria to which the nature of the news directly belongs. Therefore, news is recorded multiple times if it is classified into multiple criteria, or if several banks are involved in the news. For instance, news regarding the incident of Libor manipulation is typically classified into 3 criteria: competition, corruption and social compliance. This is because Libor manipulation is a corruptive and unfair (anti-‐‑competition) practice violating social laws and regulations. A total number of 30487 entries are recorded in the dataset. As mentioned earlier, the first research question is to investigate which CSR criteria are closely related to the banking industry. Therefore, we will focus on the criteria rather than the news itself. In other words, a piece of news to our eyes is merely a vehicle of occurred CSR criteria, and it is the criteria that matters. Therefore, an event is uniquely identified by four attributes: (1) date, (2) positive or negative news, (3) bank involved, and (4) criterion. After deleting (from the 30487 news entries in the original dataset) duplicated entries1, we arrive at a total number of 26012 event entries. Our study will focus only on these unique event entries. This research design points out that even though news belonging to the same criterion can tell different stories, they entail some common traits that are characteristic of the criterion; thus, when a criterion is hypothetically influential in the banking sector, an event-‐‑study on the same criterion should conclude statistically significant ARs.
1 The duplicated entries are similar news articles released on the same date, therefore identical in the four attributes.
4
The 50-‐‑criterion classification represents an advantage to our analysis: the more classified, the more details we will know about exactly what kind of CSR information is bank-‐‑relevant. However, this can also impose two problems that complicate the event study. Firstly, when news is classified into a certain criterion, the content is lost. In other words, we are forced to ignore other attributes of the news, such as the degree of intensity (or severeness), the degree of surprise and the degree of pertinence to business, etc. For example, think of one piece of news saying that a particular bank is helping wealthy clients engaging in tax evasion and another news saying that a particular bank is being threatened to cease operation in the US due to money laundering with Iran. Both news may be classified as negative “Social Compliance” (criterion 43), however, market reaction is likely to be different, with the latter upsetting equity returns more severely. As a result, even news in the same criterion can be heterogeneous to some degree: some may lead to ARs whereas others may not. What makes even more complicated is the situation where the news is classified into multiple criteria, even though one criterion may be more relevant than another. The second problem resides in the fact that several banks may be involved in one piece of news. This means the event windows overlap for these observations, and the problem of clustering kicks in. The consequence is that ARs are likely to be correlated across securities and the test significance becomes a bit more problematic. The methodology applied in this study is that of the standard event study, which calculates the AR as the difference between the actual return and the expected return of that same day if no events were to occur.
𝐴𝑅#,% = 𝑟#,% − 𝐸(𝑅#,%|𝐹%) where 𝐴𝑅#,% is the abnormal return of firm i at time t, 𝑅#,% represents the expected return and 𝑟#,% is the actual return; 𝐹% is the information up to t. Commonly used approaches to estimate expected returns include constant mean return model and the market model. Although the two often give similar results (MacKinlay 1997), in this study we will use the latter, which excludes the portion of “surprise” coming from the market risk.
𝑅#,% = 𝛼# + 𝛽#𝑅2,% + 𝜀#,% where 𝛼# , 𝛽# are intercept and slope estimates, 𝑅2,% is the market index return on day t and 𝜀#,% is the noise following normal distribution with mean 0. The linear regression is performed in the estimation window, which spans from 180 days to 1 day before the event. We take MSCI World Bank Index (Bloomberg: MXWO0BK) as the market index. Abnormal return is calculated as:
𝐴𝑅4,% = 𝑟#,% − 𝛼4 − 𝛽4𝑅2,% Since the noise term is by assumption following a normal distribution, the abnormal return also follows normal distribution, with variable calculated as:
𝜎6(𝐴𝑅4,%) = 𝜎#6 + 1𝐿(1 +
𝑅2,%9𝜇26
𝜎26 )
where L is the length of the estimation window, 𝜇2 is the mean of the index return over the estimation window and 𝜎2
6 is its variance. The conditional variance has two terms, the first comes from the randomness of the noise and the second comes from the error of 𝛼4, 𝛽4
5
estimations. As the length of the estimation window increases, the second term vanishes. Therefore,
𝜎6(𝐴𝑅4,%) = 𝜎#6 and it can be tested whether the AR is statistically significant given a certain confidence level2. In order to draw overall inference, the cumulative abnormal returns (CAR) are calculated,
𝐶𝐴𝑅4 1, 𝑇 = (𝐴𝑅4,%)=
%>?
𝜎#6 𝐶𝐴𝑅4 1, 𝑇 = 𝑇𝜎#6 which is approximately true if zero correlation is assumed. The significance can then be tested.
4. Empirical Analysis: Bank-‐relevant Criteria 4.1. Overview of Event Entries Chart 1 shows the percentage of event entries of each criterion in the total sample. Positive news is termed as “true” in nature and are colored in green, whereas negative news is termed as “false” in nature and is colored in red. It can be concluded that banks are frequently generating positive CSR news for social sponsorship, local communities, awards, reports and comments, emissions; and are frequently criticized for their negative CSR practices related to social compliance, corruption, employment and fiscal contribution. The 8 criteria account for more than half of the CSR news in our sample.
2 In this study, we use a confidence level of 95%.
6
It is also useful to understand the amount of event entries over time, which is illustrated in Chart 2. Historically, the amount of event entries displays an increasing pattern. This is a reflection of the market’s gradual recognition of the importance of CSR in the banking industry. As investors become more and more demanding of the CSR related news, press and media are releasing reports and making comments much more frequently nowadays then 10 years ago. A sudden peak of negative event entries is noticed during July and August of 2012. These include major events such as Barclays’ Libor manipulation scandal, HSBC money laundering in the US, Wells Fargo fined for racial bias in lending, Deutsche Bank’s “hunger trade”, employment discrimination lawsuit against Bank of America, UBS fiscal fraud in France, Standard Chartered hiding Iranian transactions. The Barclays Libor manipulation has a particularly prolonged aftermath, with various other banks got involved into the scandal and press and media stepping into questioning the effectiveness of governance and anti-‐‑corruption procedures in the banking system. In the wave of incessant scandals during the second half of 2012, banks’ reputation is “at an all-‐‑time low”, as admitted Stephen Hester the former CEO of RBS group, who believes that banks have become “detached from society” and need to change their culture.
Intellectual Property RightsSecurity Practices
Local SourcingCustomer Privacy
Product SafetyHealth and Safety
Forced LaborChild Labor
Water ManagementCompetition
Product LabelingPricing / NeedsPublic FundingInfrastructures
MaterialsIndigenous Rights
Environmental Impact of TransportEmployee BenefitsLobbying Practices
Local HiringEnvironmental Compliance
Contributions to Political PartiesPollution
Humanitarian ActionBiodiversity
Training and EducationStakeholder Engagement
DiscriminationWaste Management
Trade UnionsMarketing Communications
United Nations PolicyProduct Compliance
Indirect Economic ImpactsDiversity and Equal Opportunity
Social Impacts of ProductsEnergyWages
Human Rights PolicyEnvironmental Impacts of ProductsCommitments to External Initiatives
GovernanceFiscal Contributions
CorruptionEmissions
EmploymentSocial ComplianceLocal Communities
Awards, Reports and CommentsSocial Sponsorship
0.00 0.04 0.08 0.12criterion
perc
enta
ge in
tota
l
natureFALSETRUE
Chart 1: News Counts by Criteria
7
4.2. Standard Event Study In a first treatment, we perform the standard event study on each criterion: we select event entries within the same criterion; for each entry we calculate the AR and the standard deviation from day 1 (the event day) up to a predefined length of window, say day 10. Then CARs and standard deviations are calculated. The same computation is done for each entry in the criterion. At last CARs are aggregated and tested against the null hypothesis of zero mean normal distribution3. Since we are not sure whether positive news and negative news have the same degree of impact on equity returns (if returns are suffered more at negative news than are appreciated at positive news in the same criterion, not differentiating the positive or negative ones will compromise the significance of aggregated abnormal returns), the analysis should be done separately. The following table shows the result obtained for criterion 1 (Governance): no aggregated CAR is resulted to be statistically significant. When considering the correlations across securities and the clustering effect, standard error should be even higher due to positive covariance, making the null hypothesis even more difficult to reject. Thus, we have to conclude that the standard event study does not recognize “Governance” as a bank relevant criterion.
Table 1: CARs for Criterion 1 (Governance) with Standard Event Study Day 1 Day 2 Day 3 Day 4 Day 5 Day 6 Day 7 Day 8 Day 9 Day 10 Positive Events
CAR(%) 0.14 0.16 -‐‑0.16 -‐‑0.14 -‐‑0.14 -‐‑0.08 -‐‑0.02 0.01 -‐‑0.14 -‐‑0.25 Std Err 0.5238 0.7408 0.9072 1.0476 1.1712 1.2830 1.3858 1.4815 1.5714 1.6564
Negative Events
CAR(%) 0.24 0.25 0.06 0.23 0.29 0.1 0.18 0.11 0.49 0.49 Std Err 0.7794 1.1022 1.3500 1.5588 1.7428 1.9091 2.0621 2.2045 2.3382 2.4647
The same analysis is repeated for all 50 criteria, and unfortunately, none of the them shows any significant CAR. The standard event study therefore does not recognize any criterion to be bank relevant and thus would take a neutral standpoint on the importance of CSR practices in the banking industry.
3 At this moment, we assume that returns are uncorrelated and that there is no clustering effect.
0
200
400
600
2002 2004 2006 2008 2010 2012time
new
s co
unts
natureFALSETRUE
Chart 2: News Counts by Time
8
4.3. Relatively important CSR Criteria The fact that the standard analysis of an event study fails to detect any abnormal return is in fact not surprising: news of CSR content does not play a role as nearly influential on stock returns as other content such as stock split, earnings or take-‐‑over announcement, which are more pertinent to the main business activity and more closely followed by general investors; even though sometimes CSR news can also relate to or can be a reflection of the main business, such as a severe fine due to violation of regulations or a significant employee layoff due to poor performance, they may not account for a large part and the degree of news intensity and surprise can differ significantly among themselves . Given the characteristics of CSR news, we feel the need of making some alterations to the standard event study. First, given that event entries belonging to the same criterion differ in the degree of intensity, surprise and business pertinence, we do not aggregate the abnormal returns. Instead, we calculate the percentage of entries that causes an abnormal return (they will be called hereafter “effective entries”). This percentage can be thought of as a representation of the empirical probability that the stock return be abnormal upon news of such criterion4. The comparison of empirical probabilities among the 50 criteria can give us some insights of which criterion is relatively more pertinent to banks. Secondly, since we acknowledge that news differs in degree of intensity and other characteristics, we have to assume the time needed for investors to react, or the event window, also differs with news. Chart 3 illustrates the number of effective entries (all criteria considered) changes with respect to the length of event window.
For instance, if the event window is set to one day (the event day), around 480 entries of negative events are effective and 800 entries of positive events are effective; when the window spans to two days, the numbers are approximately halved: around 230 of negative and 420 of positive events are effective. A decreasing pattern of effective entries can be observed with the event
4 If we assume our dataset is a good representation of all publicly available CSR news and that each news is classified into criteria with reasonably good judgment.
FALSE TRUE
0
200
400
600
800
0 5 10 15 0 5 10 15length of event window
coun
t
1020304050
criterion.id
Chart 3: Effective News Counts and Length of Event Window
9
window extending longer. Event entries that are effective in the 1-‐‑day window can be cases where traders intensively trade on information, with holding period of several hours, and net their positions by the end of the day; effective entries for longer window can be (1) cases where traders (such as portfolio managers that rebalance their portfolios weekly or biweekly) who need some time to make decisions of including or excluding a security from their portfolios, or (2) caused by confounding events included in the event window. Effective event entries are of high levels at the first few days after the event, when market adjusts stock prices to reflect the new information; as the market fully incorporates the information in prices, stock returns gradually go back to normal. During this period, the level of effective event entries should exhibit an obvious decreasing pattern. When the event window extends long enough for any event to be effective on stock returns, what is left are only confounding events. As a results, the event window should be long enough to include the true effective event entries so as to discover bank relevant criterion, and be short enough so that the likelihood of confounding events is at a tolerable level. As a trade-‐‑off, we set the longest event window as 6 days (or 1 trading week after the event). This means for a particular event entry, as long as one of the first 6 cumulative abnormal returns is tested to be statistically significant, we will conclude the entry is effective; otherwise, not effective. For each effective entry, we record the length of event window for which the CAR is abnormal, and the z-‐‑score (or the standardized CAR) under the assumption of normal distribution. We use z-‐‑score because it can be aggregated across event entries with different lengths of event window. The chart also shows the decomposition of effective entries by criterion: positive abnormal returns are generally generated by news related to “Governance”, “Economic” and “Environmental”, while negative abnormal returns are typically induced by news belonging to “Labor”, “Human Rights” and “Society” (bars of positive news are darker than those of negative). Table 2 shows the average percentage of effective news entries accounts for 12-‐‑13%. It is then not surprising the standard event study would fail to detect any aggregated abnormal returns. For the statistics to be reasonably reliable, we require the number of event entries should be at least 100. When excluding small sample criteria (in italic), an average across all other criteria is computed. Since compared to positive news, negative news on average have both a higher empirical probability of abnormal returns and a higher z-‐‑score5, we can conclude investors penalize negative CSR news more than they appreciate positive ones.
Table 2 Positive Event Entry Negative Event Entry
Group Criterion ID Num Prc(%) ADay AZsc Num Prc(%) ADay AZsc
Governance
Governance 1 458 12.45 2.23 2.30 532 10.90 2.45 -‐‑2.32 United Nations Policy 2 301 13.95 2.31 2.11 65 12.31 2.50 -‐‑2.04
Commitments to External Initiatives 3 673 11.29 2.43 2.42 229 10.48 2.50 -‐‑3.20 Stakeholder Engagement 4 177 12.43 2.27 2.56 88 13.64 3.33 -‐‑2.02
Economic
Fiscal Contributions 5 303 13.86 2.60 2.15 834 11.51 2.64 -‐‑2.43 Social Sponsorship 6 3057 12.33 2.34 2.22 189 11.11 2.19 -‐‑2.16
Public Funding 7 95 12.63 2.50 2.34 31 3.23 1.00 -‐‑2.23 Wages 8 230 13.04 2.53 2.41 442 8.82 2.38 -‐‑2.44
Local Sourcing 9 38 15.79 2.83 2.01 0 NA NA NA Local Hiring 10 145 9.66 1.93 2.27 60 8.33 1.80 -‐‑1.92
Infrastructures 11 129 14.73 2.95 2.17 16 6.25 1.00 -‐‑2.54 Indirect Economic Impacts 12 312 10.26 3.28 1.99 100 23.00 1.48 -‐‑3.34
Pricing/Needs 13 16 12.50 3.00 1.74 95 13.68 2.38 -‐‑2.15 Intellectual Property Rights 14 4 0 NA NA 0 NA NA NA
Environmental
Materials 15 127 13.39 2.76 2.19 24 4.17 1.00 -‐‑1.73 Energy 16 561 13.01 2.42 2.28 41 12.20 2.40 -‐‑1.76
Water Management 17 95 12.63 2.50 2.23 6 0 NA NA Biodiversity 18 214 10.75 2.43 1.99 43 6.98 3.00 -‐‑2.42
5 The empirical probability can potentially be subject to subjective judgment when classifying a new into criteria; the average z-‐‑score is less subject to this, it can therefore be more reliable than the empirical probability.
10
Emissions 19 1151 13.03 2.19 2.28 235 14.89 2.40 -‐‑2.11 Waste Management 20 265 15.09 2.43 2.45 38 7.89 1.33 -‐‑1.84
Pollution 21 136 10.29 2.00 2.23 118 13.56 2.69 -‐‑2.23 Environmental Impacts of Products 22 576 11.63 2.09 2.31 199 11.06 1.86 -‐‑2.15
Environmental Compliance 23 136 11.03 2.47 2.33 78 8.97 1.86 -‐‑2.27 Environmental Impact of Transport 24 121 14.05 1.82 2.47 42 23.81 3.10 -‐‑2.10
Labor
Employment 25 700 12.29 2.43 2.21 948 14.03 2.26 -‐‑2.42 Employee Benefits 26 135 15.56 2.43 2.25 43 6.98 1.00 -‐‑1.80
Trade Unions 27 93 15.05 2.50 2.54 211 9.00 2.37 -‐‑2.31 Health and Safety 28 37 5.41 1.50 2.00 20 10.00 3.50 -‐‑1.89
Training and Education 29 247 8.50 2.43 2.74 12 16.67 2.50 -‐‑1.80 Diversity and Equal Opportunity 30 395 9.62 2.18 2.22 19 5.26 3.00 -‐‑1.65
Human Rights
Human Rights Policy 31 260 15.77 2.71 2.17 496 13.31 2.85 -‐‑2.52 Discrimination 32 128 5.47 3.00 2.68 138 7.97 2.09 -‐‑2.26
Child Labor 33 58 12.07 3.00 2.89 13 15.38 3.50 -‐‑2.26 Forced Labor 34 24 4.17 2.00 1.91 41 7.32 3.67 -‐‑1.83
Security Practices 35 1 0 NA NA 4 0 NA NA Indigenous Rights 36 82 9.76 3.75 1.89 76 19.74 2.00 -‐‑2.12
Society
Local Communities 37 1573 12.33 2.40 2.30 266 11.28 2.13 -‐‑2.20 Humanitarian Action 38 231 15.15 2.29 2.13 24 16.67 2.75 -‐‑1.91
Corruption 39 160 10.63 2.94 2.14 1034 11.51 2.76 -‐‑2.48 Lobbying Practices 40 48 16.67 2.25 2.25 146 6.16 3.56 -‐‑2.43
Contributions to Political Parties 41 27 11.11 2.67 1.85 188 14.89 2.57 -‐‑2.24 Competition 42 14 7.14 2.00 2.20 93 15.05 2.86 -‐‑2.08
Social compliance 43 483 10.97 2.34 2.13 1328 11.97 2.67 -‐‑2.34 Awards, Reports and Comments 44 1410 12.34 2.43 2.20 441 13.38 2.61 -‐‑2.23
Product
Product Safety 45 16 25.00 1.75 2.20 37 5.41 1.50 -‐‑2.69 Product Labeling 46 85 10.59 2.11 2.26 25 16.00 1.75 -‐‑2.15
Marketing Communications 47 208 11.06 2.39 2.44 124 12.90 2.56 -‐‑2.10 Customer Privacy 48 17 0 NA NA 31 9.68 2.67 -‐‑2.91
Product Compliance 49 169 11.24 2.95 2.19 236 12.71 2.37 -‐‑2.67 Social Impact of Products 50 471 13.16 2.63 2.15 119 19.33 2.74 -‐‑2.54
Average 474 12.13 2.46 2.27 389 12.44 2.46 -‐‑2.41
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●●
●
●●
●
●●●
●
●
●●
1
2
3
4
56
8
10
11
12
15
16
18
19
20
212223
24
2526
29
3031
37
38394344
47
4950
2.0
2.5
3.0
3.5
0.10 0.15 0.20empirical probability
aver
age.
zsco
re
criterion.group●
●
●
●
●
●
●
EconomicEnvironmentalGovernanceHuman.RightsLaborProductSociety
Chart 4.1: Abnormal Returns (Positive)
11
The two dimensions (empirical probability and average z-‐‑score) together form up a coordinate system separated by the means of the two dimensions, shown on Chart 4. The further we move towards the first quadrant, the more relevant is the criterion to banks; the further we move towards the third quadrant, the less relevant is the criterion to banks. The fourth quadrant contains criteria that can result in AR more frequently, yet with limited magnitude; the second quadrant represents criteria that though less frequently cause AR, the magnitude of the AR is quite significant should it occurs. It is interesting to notice that positive criteria are alike among themselves, since they are much concentrated, sharing similar probability of AR and average z-‐‑score (with one possible outlier “Training and Education”). This finding implies that a CSR manager could be neutral in the types of positive news the bank generates: as long as they are positive CSR practices that reflect the bank as “being ethical”, they should have similar effect on uplifting the short-‐‑term stock return, regardless of the criterion they are classified into. The negative criteria are much more scattered, implying that they should not be treated indifferently: banks should avoid being exposed to negative news, especially those can easily generate highly negative abnormal returns. Four criteria deserve to be more carefully studied. “Indirect Economic Impacts” (Criterion 12) is placed in the first quadrant to the furthest for negative news while in the third quadrant to the furthest for positive news. In other words, investors take it for granted when banks have positive indirect economic impacts on the environment in which they operate, but put them under severe critics if they have negative impacts. Almost one out of four negative news will result in AR, which is on average 3.34 standard deviations away to the left!
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
1
3
5
6
8
12
19
21
22
25
27
31
3237
39
41
43
44
47
49
50
2.0
2.5
3.0
3.5
0.10 0.15 0.20empirical probability
abs(
aver
age.
zsco
re)
criterion.group●
●
●
●
●
●
●
EconomicEnvironmentalGovernanceHuman.RightsLaborProductSociety
Chart 4.2: Abnormal Returns (Negative)
12
“Commitments to External Initiatives” (Criterion 3) has a significant difference in average z-‐‑scores between positive and negative news. Similar to banks’ indirect economic impact, investors do not embrace with exuberance banks’ active commitments to external initiatives such as social, environmental aspects; banks only need to fulfill their minimum responsibilities so as not to be considered detrimental to the society. Upon negative behaviors, 1 out of 10 cases will banks undergo a significant loss (3.2 standard deviations to the left), similar to tail risk. “Product Compliance” (Criterion 49) and “Social Impact of Products” (Criterion 50) are alike to some degree, emphasizing that the products and services provided to customers should be beneficial to the society or at least be compliant to laws and regulations. The negative criteria all have higher levels in probability and in average z-‐‑score than their positive counterparts. Our analysis so far has revealed 4 criteria that are most bank-‐‑relevant compared to others. These criteria can be summarized into 3 levels: impact of (1) banks’ products and services, (2) external (social, environmental) initiatives, and (3) economy at large. At all levels, we find that investors appreciate the benefits to a less degree than they punish the harms done by banks. From this perspective, banks are indeed punished for “being unethical”. Therefore, a bank CSR manager’s job should be more of avoiding unethical behaviors than of actively creating positive news. Unfortunately, as shown in Chart 1, banks make great efforts in giving social sponsorship and getting an award, while having poor reputation in social compliance. Banks’ predilection for social sponsorship may due to their goal of creating a positive imagine, a long-‐‑term benefit that can not be captured in short-‐‑term AR.
5. Empirical Analysis: Cost & Benefit and Regime Dependence 5.1. Ethical and Evil Strategy The previous analysis suggests that the percentage of effective event entries is so low that the standard event study would fail to detect any statistically significant aggregated CARs for any criterion. However, what if the analysis is done in relative terms, i.e. the cost and benefit (on returns) resulted from positive news compared with negative news. An easy way to visualize the result is to formulate two trading strategies, namely Ethical and Evil Strategy: the Ethical Strategy consistently invests in equities of banks exposed to positive news, and the Evil Strategy consistently invests in equities of banks exposed to negative news6 7. The two strategies are designed as follows. First, a benchmark is selected. We will use MSCI Word Bank Index (Bloomberg code: MXWO0BK)8. Then we choose a holding period. Given the information in Chart 3 and Table 2, 3 days should be an appropriate choice since it is long enough to incorporate most of the ARs and short enough to keep the potential confounding effects at low level. It is worth noting that in reality we can never earn the first day (the event day) return entirely. Since daily return is computed using closing price, in order to earn the return on the event day, we would have to buy the stock at the closing price of the previous day – before the event actually happens! Thus it is more reasonable to assume the returns can be earned are only
6 Here the comparison analysis is performed to the entire sample, without differentiating criterion; we could also construct a strategy referring to only one criterion, however this is subject to more inaccuracy since the subjective judgment of classification is necessary. 7 We could also perform the analysis with standard event-‐‑study such as in 4.2, and test whether returns generated by positive news are significantly higher than by negative news, however we would only get some statistics in the end without understanding the characteristics of returns over time. 8 Alternatively, we could also construct an index (for instance, an equally weighted index) using the 30 banks in our sample.
13
those of the second and third day. The Ethical Strategy will invest in the benchmark in case there is no news; in case on a particular day 𝑛 banks are exposed to positive news, the index will divest 𝜔𝑛 weights of its value and buy the stocks of 𝑛 banks (each with weights 𝜔) at the closing price9. The stocks are held in the portfolio for 2 days and are sold on the third day at the closing price. Proceeds are instantaneously invested in the benchmark. The Evil Strategy invests in the same way, except that it selects stocks of banks subject to negative news. The Ethical Strategy can be thought of as an index tracking a hypothetical group of banks that systemically generate positive CSR news; whereas the Evil mimics the return of a hypothetical group of banks consistently generate negative news. Certain assumptions should apply, such as no transaction costs and that stocks are fractionally divisible. Undoubtedly, the returns of index depend on 𝜔, which reflects the degree of bias. Ethical Strategy with a large 𝜔 implies a strategy strongly biased towards banks with good CSR practices. Generally, 𝜔 should be larger than 1/30 to reflect a reasonable degree of bias10. In case 𝜔𝑛 is larger than 1, this means Ethical Strategy would have to short the benchmark to finance its long positions in the 𝑛 banks. Chart 5 (upper) shows the benchmark and the two strategies’ (with 𝜔 equal to 2/30) performance through time with an initial value of 100 on Jan 1, 2002. Ethical Strategy is able to outperform benchmark, which in turn outperforms again the Evil Strategy. Had we chosen a higher 𝜔, the differences in performances would have got even larger11. Chart 5 (lower) illustrates the daily return difference between Ethical and Evil Strategy on a cumulative basis starting from Jan 1, 2002. Initially Ethical slightly outperforms Evil but then the two become commensurate until 2008; the cumulative return difference sees a steep rise from 2008 to the second half of 2009, corresponding almost perfectly with the recession as documented by NBER12, and this is the period when Ethical far beats Evil; after the recession, Ethical slightly underperforms Evil Strategy. The chart implies that daily return difference is somehow linked to market regimes. We can thus use a Hidden Markov Model (HMM) with two states13 to fit this time series; we can also think the time series as hedge fund return, which would lead to the same model specification.
9 This strategy formulation ensures that each event entry be treated equally: if we do not pre-‐‑define a weight 𝜔 and invests the entire portfolio equally among the 𝑛 banks, the weights will depend on the number of event entries on each day and will not be treated equally. 10 Had we constructed the benchmark with the 30 banks in our database with equal weights, each bank will represent exactly 1/30. 11 𝜔 can only affect the return differences in absolute terms, but not the direction: if on a particular day the return of Ethical Strategy is higher than that of the benchmark, then no matter what value 𝜔 takes (as long as it is positive), the return of Ethical Strategy is always higher than that of the benchmark. The same applies for the Evil Strategy. 12 NBER documents the recession started in December 2007 and lasted for 18 months; a trough is observed in June 2009 and the economy began to recover. 13 Two-‐‑state is proved to be the most appropriate choice compared to 3 or more states for our data, according to BIC information criteria.
14
5.2. Hedge Fund Returns’ Dependence on Market Regimes The daily return difference between Ethical and Evil can be viewed as the daily return of a hedge fund with a long position in Ethical and a short position in Evil launched on Jan 1, 2002. By modeling the hedge fund return as regime-‐‑dependent, we will be able to tell how the cost and benefits of “being ethical” vs. “being unethical” is linked with market regimes. There exists abundant research on hedge fund return modeling in the context of analyzing investment styles and measuring performances, where the excess return is explained by a linear combination of some risk factors or payoffs mimicking a certain trading strategy (see Fama and French (1992,1993); Carhart (1997); Fung and Hsieh (2004)). Recently, regime-‐‑dependent models such as Markov Switching models are also applied by hedge fund researchers (Agarwal and Naik (2004); Billio et. Al (2012)), where risk exposures are dependent on states governed by a Markov Chain.
𝑅% = 𝛼B + 𝛽C𝑀𝐾𝑇% + 𝜃G,C𝐹G,%
H
G>?
+ 𝜌B𝜀%
𝑀𝐾𝑇% = 𝜇J + 𝜎J𝜖%
The hedge fund return 𝑅% is explained by market index MKT and another K factors, with risk exposures all dependent on S, a state variable (governed by a Markov Chain) that characterizes the market index. 𝜀% and 𝜖% are noise terms following the standard normal distribution. This specification implies that the hedge fund changes risk exposures according to the market index.
2002 2004 2006 2008 2010 2012
5010
020
0
Chart 5.1: Ethical vs. Evil Strategy
time
valu
e
BenckmarkEthicalEvil
2002 2004 2006 2008 2010 2012
0.0
0.2
0.4
0.6
Chart 5.2: Cumulative Return Difference
time
valu
e
Ethical−EvilNBER recession
15
The alpha and idiosyncratic risk is dependent on Z, a state variable governed by another Markov Chain designed to capture additional non-‐‑linearity. When this model specification is applied to our hedge fund return, it can be simplified due to the characteristics of the trading strategy. First, since the hedge fund invests only in equity, it has only 4 relevant systemic risk factors, namely market risk, size factor (smb), value factor (hml) and momentum factor (wml), as demonstrated by lots of empirical research on equity hedge funds or mutual funds (Fama and French (1993); Carhart (1997); Fung and Hsieh (2004)). Secondly, since the hedge fund has zero net position (long 100% in Ethical Strategy and short 100% in Evil Strategy), it does not bear any market risk; moreover, since it temporarily deviates its components weights’ from the benchmark in face of positive or negative CSR news, it does not follow any size, value or momentum strategy. The long/short strategy implies that the risk exposures to these factors are also close to zero. As a result, the model reduces to:
𝑅% = 𝛼B + 𝜌B𝜀% indicating that the hedge fund performance is all attributed to alpha (management skills), which is embedded in the successfulness of the trading strategy entirely based on CSR information. The state variable Z is governed by a Markov Chain, and 𝜀% is the random term following the standard normal distribution. More specifically, if 2-‐‑states are imposed on the return series,
𝑅% = 𝛼L + 𝜌L𝜀% , 𝑖𝑓 𝑍 = 0 𝑅% = 𝛼? + 𝜌?𝜀% , 𝑖𝑓 𝑍 = 1
and the distribution of Z is determined by the transition probability matrix P,
𝑃 =𝑝LL 𝑝L?𝑝?L 𝑝??
where 𝑝LL denotes the probability that Z remains 0 in the next period given that it equals 0 in the current period; 𝑝L? denotes the probability that Z transit to 1 in the next period given that it equals 0 in the current period, etc. It follows that 𝑝LL + 𝑝L? = 1 and that 𝑝?L + 𝑝?? = 1. EM algorithm (see for example, Kim and Nelson (1999)) can be used for likelihood maximization and parameters can thus be estimated. Here we will use weekly returns instead of daily returns, this choice is due to 2 reasons: first, out of 2865 daily returns, 292 are equal to 0, corresponding to the 292 days when no news occurred. These observations are not related to the successfulness of the strategy and if were included, the estimation result would be erroneous. When returns are aggregated into weekly frequency, out of 573 weekly observations only 6 are equal to 0, which will unlikely to have any impact on the result. Secondly, if daily returns were used, states would constantly switch among each other with short lengths. This is not really desirable in our analysis, since we are also interested in understanding the timing of the states.
Table 3: Estimation Result 𝛼 𝜌 Transition Matrix
Z=0 0.038% 0.0036 0.945 0.055 Z=1 0.142% 0.0147 0.141 0.859
The market neutral hedge fund trading on CSR news seems to be quite successful, since in both states alpha reveals to be positive. This result implies that banks being ethical are expected to have a higher return than being unethical, regardless of the market regimes. Moreover, the fact
16
that alpha is much higher in state 1 than in state 0 (though higher volatility) indicates that the cost and benefit of being ethical vs. unethical is expected to be highest in states 1. We might be interested in understanding what the State 0 and State 1 represent by observing the timing of the state sequence. The most-‐‑likely sequence of states can be obtained from the Viterbi algorithm (see Forney (1973)), which maximizes the joint probability of the entire sequence of states. Chart 6 (upper) shows the timing of state 1. We then model the MSCI World Index (the benchmark) with a 2-‐‑state Markov Switching model, resulting one regime (normal times) characterized with higher mean returns and lower volatility and another regime (recession) characterized with lower mean returns and higher volatility. Again, we plot the most-‐‑likely sequence of regimes, shown on lower part of the chart. Interestingly, the state sequence resulting from the hedge fund return and the regime sequence from the index series share similar pattern. In fact, the odds ratio test strongly indicates that the two sequence share similar property. Odds ratio is a commonly used method to quantify the strengths of association of one property (hedge fund return being in state 1) to another property (index return being in recession). Table 4 shows there are 331 weeks being in state 0 and in normal market condition, accounting for 57.77% of the total sample; there are 108 weeks being in state 1 and in recession, accounting for 18.85%, etc. Odds ratio results to be 11.4714, leading to a Z value equal to 4.3415, which strongly rejects the null hypothesis that Z follows a standard normal distribution. Therefore, we conclude that state 1 and recession are strongly associated, in the sense that the index being in recession raises the odds that the hedge fund in state 1. Or loosely speaking, state 1 corresponds to periods of market downturn.
14 computed as ST.TT∗?W.WS
S.6X∗?W.?S
15 computed as YZ[ (??.\T)]
^_.__`]
]a.a^`]
^.bc`]
]a.]^
17
Table 4: Odds Ratio Test Regimes of Benchmark
Normal Recession
States of Hedge Fund S=0 331 (57.77%) 104 (18.15%) S=1 30 (5.23%) 108 (18.85%)
6. Summary and Conclusion Banks are frequently generating positive CSR news for social sponsorship, local communities, awards/reports/comments, emissions; and are frequently criticized for their negative CSR practices related to social compliance, corruption, employment and fiscal contribution. The 8 criteria account for more than half of the CSR news. The amount of events displays an increasing trend over time, which reflects that market is placing more and more importance on the CSR aspects in the banking industry and that investors are becoming more and more demanding of the CSR related news. The standard event study does not conclude statistically significant ARs for any criterion. This is because news belonging to the same criterion are also heterogeneous in the contents, varying in degrees of intensity, surprise and pertinence to business, etc. In fact, on average the percentage of news in a criterion resulting in a significant AR is less than 13%.
2002 2004 2006 2008 2010 2012
0.0
0.2
0.4
0.6
Chart 6: Hedge Fund States and Benchmark Regimes
time
valu
e
Hedge fund weekly returnMost likely state sequence
2002 2004 2006 2008 2010 2012
5010
015
0
time
valu
e
Benchmark (MXWO0BK)Most likely regimes
18
While the standard event study does not give any informative conclusions, it is still possible to identify criteria that are relatively more bank-‐‑relevant. 4 criteria are revealed: (1) “product compliance” and “social impact of products” (criterion 49, 50): the products and services provided by banks to their customers should be beneficial to the society or at least be compliant to social laws and regulations. (2) “commitments to external initiatives” (criterion 3): while banks do not need to actively engage in initiatives of external entities, they shall still fulfill the minimum responsibilities so as not to be considered detrimental to the environment and the society. (3) “indirect economic impacts” (criterion 12): investors take it for granted when banks have positive indirect economic impacts on the environment in which they operate, but put them under severe critics if they have negative impacts. The two-‐‑dimension (empirical probability, average z-‐‑score) analysis reveals (1) compared to positive ones, negative news on average results in higher empirical probability and higher z-‐‑score, (2) criteria of positive news are alike themselves, because they are concentrated in the center of the two dimensions; however, criteria of negative news are much more scattered. These findings imply that the priority job of a CSR manager should be more of minimizing negative CSR news rather than actively engaging in generating positive news, and that the treatment of different categories of negatives CSR news should not be indifferent. To investigate the comparative benefit of being ethical vs. being unethical, we form up two strategies, Ethical and Evil, that temporarily deviate from the benchmark to gain exposure of equities exposed to positive and negative CSR news, respectively. It is found that over the sample period considered, Ethical outperforms the benchmark which again outperforms Evil. Moreover, the cumulative return difference between Ethical and Evil exhibits a steep increase during the 2008 financial crisis. In fact, the Markov Switching model indicates that a market neutral hedge fund with the only strategy of longing in positive news and shorting in negative news would gain a much higher return in market downturns than in normal times. This is a strong empirical evidence that favors the ethical CSR practices in the banking industry: although the benefit of being ethical (compared to being unethical) may be hidden in normal market conditions, it becomes evident in market recession. So being ethical pays. The result also justifies SRI where ESG aspects are incorporated into the investment process, which should be expected to be more successful during market downturns.
19
Reference Agarwal, V. & Naik, N. Y. ‘Risk and portfolio decisions involving hedge funds’, The Review of Financial Studies, Vol. 17, No. 1, 2004, pp. 63–98. Ainscough, T.S., Hill, P.R. & Mnullang, D., ‘Corporate social responsibility and social responsible investing: a global perspective’, Journal of Business Ethics, Vol. 70, 2007, pp. 165-‐‑ 174 Billio, M., Getmansky, M. & Pelizzon, L., ‘Dynamic risk exposures in hedge funds’, Computational Statistics & Data Analysis, Vol. 56, No. 11, 2012, pp. 3517-‐‑3532. Brown, S. & Warner, J.B., ‘Using daily stock returns’, Journal of Financial Economics, Vol. 14, 1985: pp.3-‐‑31. Carhart, M., ‘On persistence in mutual fund performance’, Journal of Finance, Vol. 52, No. 1, 1997, pp. 57–82. Fama, E.F. & French, K.R., ‘The Cross-‐‑Section of Expected Stock Returns’, Journal of finance, Vol. 47, No. 2, 1992, pp. 427-‐‑465. Fama, E.F. & French, K.R., ‘Common risk factors in the returns on stocks and bonds’, Journal of Financial Economics, Vol. 33, No. 1, 1993, pp. 3-‐‑56. Fung, W. & Hsieh, D. A., ‘Hedge fund benchmarks: a risk based approach’, Financial Analyst Journal, Vol. 60, No. 5, 2004, pp. 65–80. Hall, B. H., ‘Innovation and market value’, National Bureau of Economic Research, 1999. Hamilton, J. D., ‘A new approach to the economic analysis of nonstationary time series and the business cycle’, Econometrica, Vol. 57, No. 2, pp. 357-‐‑384. Kim, C. J. & Nelson, C. R., ‘State-‐‑space models with regime switching: classical and Gibbs-‐‑sampling approaches with applications, Massachusetts Institute of Techonology, 1999, The MIT Press. Lichtenberg, F. & Siegel, D. ‘The impact of R&D investment on productivity: new evidence using linked R&D-‐‑LRD data’, Economic Inquiry, Vol. 29, 1991, pp. 203–228. MacKinlay, A.C., ‘Event study in Economics and Finance’, Journal of Economic Literature, Vol. 35, No. 1, 1997, pp. 13-‐‑39. McWilliams, A. & Siegel, D., ‘Event study in management research: theoretical and empirical issues’, Academy of Management Journal, Vol. 40, No. 3, 1997, pp. 626-‐‑657. Moskowitz, M. ‘Choosing social responsible stocks’, Business and Society Review, Issue 1, 1972, pp. 71-‐‑75. Preston, L.E. & O’Bannon, D.P., ‘The corporate social-‐‑financial performance relationship: a typology and analysis’, Business and Society, Vol. 36, No. 4, 1997, pp. 419-‐‑429. Teoh, S. H., Welch, I. & Wazzan, C.P., ‘The effect of socially activist investment policies on the financial markets: Evidence from the South African boycott’, Journal of Business, Vol. 72, No. 1, 1999, pp. 35–89. Waddock, S. & Graves, S., ‘The corporate social performance -‐‑ financial performance link’, Strategy Management Journal, Vol. 18, No. 4, 1997, pp. 303-‐‑319. Wright, P. & Ferris, S., ‘Agency conflict and corporate strategy: The effect of divestment on corporate value’, Strategic Management Journal, Vol. 18, No. 1,1997, pp. 77–83.
I
Appendix: Criteria and Definition (Source: Covalence EthicalQuote Company Website) Id Criterion
Name Definition Criterion
Group 1
Governance
"Corporate governance is "the system by which companies are directed and controlled". It involves regulatory and market mechanisms, and the roles and relationships between a company’s management, its board, its shareholders and other stakeholders, and the goals for which the corporation is governed" (Wikipedia). It covers topics such as: Governance structure, Cumulative functions, Board independence, Shareholders expression rights, Sustainable compensation (link between compensation and the company's social and environmental performance), Conflicts of interest, Board diversity, Mission statements and codes of conduct (GRI G3.1 4. Governance, Commitments, and Engagement)
Governance
2 United Nations Policy
Dialogue, partnerships, or controversies between a company and the United Nations (programmes, agencies, or UN-supported projects, such as the Global Compact, UNEP, UNDP, the Global Reporting Initiative, etc.); adressing the precautionary approach or principle as stated in Article 15 of the Rio Principles
Governance
3 Commitments to External Initiatives
Participation in economic, environmental, and social charters, principles, platforms, partnerships or other initiatives that haven't been principally created by the company itself, but by external organisations. (GRI G3.1 Part 2.4).
Governance
4 Stakeholder Engagement
Engagement, consultation, dialogue of a company with its stakeholders regarding its impact on sustainable development and on stakeholders, such as civil society, NGOs, customers, employees, other workers, local communities, shareholders and providers of capital, suppliers. (GRI G3.1 Part 2.4)
Governance
5 Fiscal Contributions
Payment of taxes by the company, globally and in individual countries; fiscal policy; transparency about the payment of taxes; impact of fiscal contributions on local economic and social development
Economic
6 Social Sponsorship Donation of money or goods by a company to an external organization in the pursuit of social or environmental objectives; cause-related marketing: when the support to social / environmental projects is linked to the selling of a product.
Economic
7 Public Funding
Financial assistance received from government by a company: "subsidies; investment grants, research and development grants, and other relevant types of grants; Awards; Royalty holidays; Financial assistance from Export Credit Agencies (ECAs); Financial incentives; Other financial benefits received or receivable from any government for any operation." (GRI G3.1 EC4)
Economic
8 Wages Wages paid to employees and executives within the company; comparisons with local minimum wage; Range of ratios of standard entry level wage by gender compared to local minimum wage at significant locations of operation. (GRI G3.1 EC5)
Economic
9 Local Sourcing "Use of locally-based suppliers at significant locations of operation"; "Supporting local business in the supply chain" (GRI G3.1 EC6)
Economic
10 Local Hiring Hiring of employees and managers from the local community at
locations of significant operation. (GRI G3.1 EC7) Economic
11 Infrastructures "Development and impact of infrastructure investments and services
provided primarily for public benefit", such as water supply facility, road, Economic
II
hospital, and other public services (GRI 3.1 EC8)
12 Indirect Economic Impacts
Indirect economic impacts: "additional impacts generated as money circulates through the economy" (Direct economic impacts: "immediate consequences of monetary flows to stakeholders"). Examples of Indirect economic impacts: "Economic development in areas of high poverty (e.g., number of dependents supported through income from one job)"; Wages paid by suppliers and sub- contractors; "Enhancing skills and knowledge amongst a professional community or in a geographical region"; "Jobs supported in the supply chain or distribution chain"; "Economic impact of change in location of operations or activities (e.g., outsourcing of jobs to an overseas location)"; "Economic impact of the use of products and services (e.g., linkage between economic growth patterns and use of particular products and services)." (GRI G3.1 EC9)
Economic
13 Pricing / Needs Price of products and services considering their social utility and capacity to respond to essential human needs, such as life-saving drugs, electricity or water supply; "Availability of products and services for those on low incomes (e.g., preferential pricing of pharmaceuticals contributes to a healthier population that can participate more fully in the economy"; "pricing structures that exceed the economic capacity of those on low incomes" (GRI G3.1 EC9)
Economic
14 Intellectual Property Rights
Social and environmental impacts of a company's intellectual property rights on other companies and countries. In relation to the management of intellectual property rights and patents, measures - or lack of measures - that promote human and economic development, the protection of biodiversity, respect of traditional knowledge and local natural resources, for example through research & development, voluntary licenses, agreements, cooperation with research institutes and local communities. (GRI G3.1 EC9)
Economic
15 Materials Environmental impact of use of materials by the company; "contribution to the conservation of the global resource base and efforts to reduce the material intensity and increase the efficiency of the economy"; use of recycled input materials (GRI G3.1 EN1, EN2)
Environmental
16 Energy Direct energy consumption; Indirect energy consumption; "Energy saved due to conservation and efficiency improvements"; "Initiatives to provide energy-efficient or renewable energy based products and services" (GRI G3.1 EN3, EN4, EN5, EN6, EN7)
Environmental
17 Water Management
Management of water used by the company; water withdrawal from any kind of source; environmental impact of the use of water; recycling and reuse of water. (GRI G3.1 EN8, EN9, EN10)
Environmental
18 Biodiversity Impacts of activities, products, and services on biodiversity; Habitats protected or restored; IUCN Red List species and national conservation list species with habitats in areas affected (GRI G3.1 EN11, EN12)
Environmental
19 Emissions Direct and indirect greenhouse gas emissions; Initiatives to reduce greenhouse gas emissions; Emissions of ozone-depleting substances; NO, SO, and other significant air emissions; Initiatives to reduce emissions of ozone-depleting substances and air emissions. (GRI G3.1 EN16, EN17, EN18, EN19, EN20)
Environmental
20 Waste Management
Waste management and disposal method; Transport of hazardous waste; Water discharge; Impact of water discharge on biodiversity. (GRI G3.1 EN21, EN22, EN24, EN25)
Environmental
21 Pollution
Pollution; Significant spills of chemicals, oils, wastes, and fuels; Impact of pollution on the environment; Initiatives to avoid pollution and spills of hazardous materials. "Spill: accidental release of a hazardous substance that can affect human health, land, vegetation, water bodies, and ground water." (GRI G3.1 EN23)
Environmental
22 Environmental Impacts of Products
Impacts of products and services on the environment, nature, and animals; Initiatives to mitigate such impacts; Reuse and recycling of products and package; New products or services that are friendly to the environment, nature, animals. (GRI G3.1 EN26, EN27)
Environmental
III
23 Environmental Compliance
Compliance and noncompliance with environmental laws and regulations; "Monetary value of significant fines and total number of non-monetary sanctions for non- compliance with environmental laws and regulations." (GRI G3.1 EN28)
Environmental
24 Environmental Impact of Transport
"Environmental impacts of transporting products and other goods and materials used for the organization’s operations, and transporting members of the workforce" (GRI G3.1 EN29)
Environmental
25 Employment Employment; Creation of jobs; Job cuts (downsizing, restructuring); Rate of new employee hires and employee turnover; "workforce by employment type, employment contract, and region, broken down by gender"; "new employee hires and employee turnover by age group, gender, and region." (GRI G3.1 LA1, LA2)
Labor
26 Employee Benefits Benefits attributed to a company's employees in addition to wages: pension plan, retirement plan, health insurance, parental leave, maternity leave, paternity leave, return to work and retention rates after parental leave (GRI G3.1 LA3, LA15)
Labor
27 Trade Unions Relations of companies' management with trade unions (dialogue, partnership or confrontations); Freedom of Association and Collective Bargaining within the company as well as at major suppliers; Employees covered by collective bargaining agreements; Strikes; "Minimum notice period(s) regarding operational changes, including whether it is specified in collective agreements" (GRI G3.1 LA5)
Labor
28 Health and Safety Health and safety of employees with the company and in the supply chain (occupational health and safety); Injury, occupational diseases, lost days, absenteeism; Work-related fatalities (deaths); "Education, training, counseling, prevention, and risk-control programs in place to assist workforce members, their families, or community members regarding serious diseases"; "Health and safety topics covered in formal agreements with trade unions" (GRI G3.1 LA6, LA7, LA8, LA9)
Labor
29 Training and Education
Training and education offered by a company to its employees, "skills management and lifelong learning that support the continued employability of employees and assist them in managing career endings"; Performance and career development reviews (GRI G3.1 LA10, LA11, LA12)
Labor
30 Diversity and Equal Opportunity
Diversity and equal opportunities among employees and in governance bodies, "according to gender, age group, minority group membership, and other indicators of diversity." (GRI G3.1 LA13)
Labor
31 Human Rights Policy
Incorporation of human rights concerns when deciding major investments or contracts; relations of companies with governments regarding human rights issues; human rights screening; boycott of certain countries and governments because of the human rights situation (GRI G3.1 HR1, HR2, HR3, HR10, HR11)
Human Rights
32 Discrimination Discrimination: "the unjust or prejudicial treatment of different categories of people, especially on the grounds of race, age, or sex" (Oxford Dictionaries); Initiatives aiming at reducing or avoiding discrimination within the company, along the supply chain and in other sectors of society. (GRI G3.1 HR4, LA14)
Human Rights
33 Child Labor "Risk for incidents of child labor, and measures taken to contribute to the effective abolition of child labor" (GRI G3.1 HR6)
Human Rights
34 Forced Labor Operations and significant suppliers confronted to, or risking incidents of
forced or compulsory labor, "and measures to contribute to the elimination of all forms of forced or compulsory labor." (GRI G3.1 HR7)
Human Rights
35 Security Practices Impact of a company's security practices on the respect of human rights among its stakeholders; Training of security personnel on human rights. (GRI G3.1 HR8)
Human Rights
36 Indigenous Rights Initiatives in favor of indigenous people; "Incidents of violations involving rights of indigenous people and actions taken" (GRI G3.1 HR9)
Human Rights
IV
37 Local Communities
Positive or negative impacts on local communities; Prevention and mitigation of negative impacts; "Local community engagement, impact assessments, and development programs"; Community investments; Involvement in local communities related to topics such as education, health, the environment, or security. (GRI G3.1 SO1)
Society
38 Humanitarian Action
Behavior of a company within and about emergency situations such as wars, civil wars and natural disasters. (GRI G3.1 SO1)
Society
39 Corruption Cases of corruption and bribery of public or private actors by a company;
actions taken by the company to avoid the use of corruption; anti-corruption training; analysis of risks related to corruption. To bribe: "Dishonestly persuade (someone) to act in one’s favour by a gift of money or other inducement" (Oxford Dictionaries); (GRI G3.1 SO2, SO3, SO4)
Society
40 Lobbying Practices Lobbying activities of companies: activities aiming at influencing decisions taken by governments at the national and international levels; Social and environmental impacts of such lobbying activities; "Public policy positions and participation in public policy development" (GRI G3.1 SO5)
Society
41 Contributions to Political Parties
"Financial and in-kind contributions to political parties, politicians, and related institutions" (GRI G3.1 SO6)
Society
42 Competition "Anticompetitive behavior, anti-trust, and monopoly practices and their
outcomes"; Unfair business practices; Measures oriented towards fair competition. (GRI G3.1 SO7)
Society
43 Social Compliance Compliance and noncompliance with social laws and regulations; Monetary value of significant fines and total number of non-monetary sanctions for non-compliance with laws and regulations. (GRI G3.1 SO8)
Society
44 Awards, Reports and Comments
Award, prize and other marks of recognition received or given by a company in the field of sustainability, ethics, Corporate Social Responsibility (CSR); inclusion in, or exclusion from, Socially Responsible Investing (SRI) funds and indexes; publication of CSR and sustainability reports; general comments, positive or negative, about a company's behavior and international presence
Society
45 Product Safety Impacts of products and services on health and safety of consumers; Risks relating to the health and safety of consumers, and measures mitigating such risks; compliance and noncompliance "with regulations and voluntary codes concerning health and safety impacts of products and services during their life cycle" (GRI G3.1 PR2)
Product
46 Product Labeling Information about labeling of products and services; compliance and "noncompliance with regulations and voluntary codes concerning product and service information and labeling" (GRI G3.1 PR4); "Practices related to customer satisfaction, including results of surveys measuring customer satisfaction" (GRI G3.1 PR5)
Product
47 Marketing Communications
Compliance and noncompliance with "laws, standards, and voluntary codes related to marketing communications, including advertising, promotion, and sponsorship" (GRI G3.1 PR6, PR7)
Product
48 Customer Privacy Respect and "breaches of customer privacy and losses of customer data" (GRI G3.1 PR8)
Product
49 Product
Compliance Compliance and "noncompliance with laws and regulations concerning the provision and use of products and services" (GRI G3.1 PR9)
Product
50 Social Impacts of
Products Impacts of products and services on society and the people; human and social utility of products and services; socially innovative products and services such as life-saving drugs, education material or communications facilities; research & development (R&D) of products or services that present a particular interest for responding to human needs and contributing to economic and social development.
Product