Stakeholder Information and Shareholder Value

Text-only Preview

Stakeholder Information and
Shareholder Value
Philipp Kr¨
This Draft: August 31, 2009
This paper examines the relationship between incidents of social re-
sponsibility (positive and negative) and shareholder value in an event
study setting. I find significant negative abnormal returns when third
parties such as newspapers, non-governmental organizations or regu-
latory authorities publicly report about negative social responsibility.
Negative returns are particularly strong for information concerning
product safety and the firm’s impact on communities and the wellbe-
ing of employees. In contrast, positive news do not generate a system-
atic reaction. Analyzing positive events in greater detail, I nonetheless
identify two causal mechanisms relating positive social responsibility
and shareholder value. On the one hand, events containing signals
about past and likely future economic performance bring about pos-
itive abnormal returns. As such, charitable donations to communi-
ties, profit distribution to employees and the appointment of women
or members of minority groups to senior executive positions or the
board are rewarded by investors. On the other hand, positive events
indicative of higher company expenses due to the use of alternative en-
ergy or (non)-monetary benefits to employees generate a significantly
negative reaction.
JEL-Classification: M14, G14, G24, D21, L21
Keywords: Social Responsibility, Corporate Governance, Corporate Culture,
Event Study, Shareholder Value

The question of whether managers should attend to the interests of share-
holders, stakeholders or both is longstanding in the economics, finance and
management literature. As Kacperczyk (2009) notes, two opposing theo-
retical views have generated separate streams of research concerned with
the purpose of the firm. On the one hand, shareholder value theory (see
Jensen and Meckling (1976), Fama and Jensen (1983), Jensen (2001) or Ti-
role (2001)) is based on the argument that the only objective of the firm is to
attend to the interests of those owning the claims to the firm’s residual profit
or assets. On the other hand, stakeholder theory advances the idea that man-
agers running a firm should balance the interests of all stakeholders, because
not one group of stakeholders has a priority over other groups (see Freeman
et al. (2007)). In a similar spirit Allen et al. (2009) show that under certain
circumstances, stakeholder oriented firms can be more valuable.
In order to answer the unresolved question of whether attending to stake-
holders’ interests increases shareholder value, a myriad of studies has exam-
ined the empirical relationship between profitability and responsible business
practices (see Orlitzky et al. (2003)).
However, the majority of these studies have not been able to account for
the problem of reverse causality: It is still not clear whether attending to
stakeholders, that is acting in a more (socially) responsible way, increases
profitability, or whether causality runs from profitability to responsible firm
In the present study, I revisit this salient question in an event study set-
ting. I analyze how the public release of information concerning a broad set
of different stakeholders influences shareholder value. The examination of
changes in shareholder value due to exogenous changes in a firm’s consider-
ation of non-shareholding stakeholders provides insights into the question of
whether companies should only pursue their fiduciary duty vis `
a vis share-
holders, or explicitly account for interests of non-shareholding stakeholders.
The issue
More precisely, I study whether the stock market systematically reacts when-
ever third parties such as newspapers, regulatory authorities or non govern-
mental organizations report about incidents of positive or negative (social)
responsibility. In the business, finance and economics literature, there re-
mains a high level of disagreement as to what constitutes (socially) respon-

sible firm behavior. While some argue that a company is acting in a socially
responsible way whenever it is sacrificing profits for the general public good
(see Friedman (1970) or more recently Reinhardt et al. (2008)), others (see
Campbell (2007)) think of a firm as being socially responsible whenever its
actions (1) do not knowingly harm stakeholders and (2) indemnify harmed
stakeholders once irresponsible behavior is brought to the firm’s attention.
The first definition is conceptually rooted in shareholder value oriented
theories and builds on the idea that only stakeholder benefiting actions which
go beyond what is legally required from the firm should be considered socially
responsible. In contrast, the second definition explicitly includes illegal firm
actions within the set of socially irresponsible firm behavior.
In this study, I rely on data and definitions from social investment re-
search firm Kinder, Lydenberg and Domini Research and Analytics2 (KLD).
KLD’s data are widely considered to be the gold standard in empirical re-
search concerned with corporate social responsibility and as such, their data
definitions have been employed extensively in previous empirical studies (see
Waddock and Graves (1997), Deckop et al. (2006) or Kempf and Osthoff
(2007)). At present, KLD focuses on seven different stakeholder issue ar-
eas (e.g. employee relations, community relations, environment, etc.). Since
some of the issues make explicit reference to the extent to which a company
is breaching the law, KLD’s interpretation of social (ir)responsibility is closer
to one including illegal company behavior (see Campbell (2007)). As an ex-
ample, KLD records whether a company has been at odds with regulatory
or other legal authorities.
Under a scheme known as @KLD, the social investment research firm
sends out irregularly spaced newsletters which contain references to publi-
cized incidents of social responsibility. Broadly speaking, the disseminated
events relate to a company’s relationship with its various stakeholders. Nega-
tive events (concerns) are reflective of matters such as product safety, environ-
mental accidents or workplace safety violations. Positive events (strengths)
make reference to corporate philanthropy, the use of alternative energy or
family benefits offered to employees. I study how the stock market reacts to
the events relevant to KLD’s analysis.
Main results
A descriptive analysis reveals that KLD included more negative than posi-
tive events in the newsletters. Furthermore, the relative importance of each
issue area (e.g. communities, environment, human rights, etc.)3 differs sub-
stantially for the set of negative and positive events. As an example, issues
related to a company’s diversity, the environment or the relationship with

communities make up more than 60 per cent of the positive events. By con-
trast, negative ones are largely dominated by matters related to economic
aspects of the firm (e.g. product safety, governance or the extent to which
fraudulent marketing and contracting practices are being employed). The
latter result sheds doubts on whether events of negative social responsibility
recorded by KLD are solely measuring the extent to which a company ig-
nores the interests of stakeholders, or more generally inferior and/or illegal
corporate behavior. Since negative news of economic nature is of interest to
KLD’s social analysts, one could be inclined to think that the same set of
corporate behavior adversely affects both shareholder value and a company’s
attendance to stakeholders. Put differently, companies ignoring potentially
adverse effects on stakeholders by cutting corners might not be maximizing
shareholder value either. This idea is at odds with the common perception
in shareholder value oriented theories that attending to stakeholders must
necessarily be related to sacrificing profits.
Relying on content analysis, a technique from linguistics employed for
the systematic analysis of textual data, I add to the result that negative
events are reflective of illegal firm behavior. Statistical evidence shows that
negative events are often concerned with legal actions taken by stakehold-
ers (employees, customers, regulatory authorities). I also find that negative
event descriptions contain significantly more quantitative information than
positive ones, suggesting that negative events are often related to fines by
legal authorities and/or indemnity payments to employees and other stake-
In a second step, I study the implications of the events for shareholder
wealth. I take the day on which third parties release a piece of information
relevant to KLD’s analysts as the event date. Since I rely on KLD’s clas-
sification which is known with certainty only three weeks after the event, I
have to address a problem of endogeneity: A priori, it is not clear whether
abnormal share price movements draw the attention of KLD’s analysts to
record an event.
Stated differently, if KLD’s analysts are responding to
observed changes in stock prices in order to identify events of social respon-
sibility, event study methodology is no longer valid (see Viswanathan and
Wei (2008)). For this reason, I test whether changes in shareholder value
are affecting the way KLD’s analysts classify events by comparing ordinary
least squares (OLS) and two stage least squares (2SLS) estimates. Regres-
sion based Hausman tests reject the hypothesis that KLD’s analysts respond
to observed changes in shareholder value. Since the tests reject the hypoth-
esis of endogenously sampled events, I subsequently identify what kind of
stakeholder information has a significant impact on shareholder value.
I find a strong and significantly negative stock market reaction whenever

negative stakeholder information is released. In contrast, positive news does
not affect stock prices in a systematic way.
Analyzing the stock price reaction conditional on KLD’s seven issue areas,
I find that negative news related product safety and a company’s relations
with communities reduces shareholder value substantially. Furthermore, neg-
ative news concerning governance or employee related issues also generate
statistically significant negative abnormal returns. I also provide statisti-
cal evidence (marginally significant) that negative news related to a firm’s
diversity bring about negative abnormal returns.
Turning to positive events, I find some weak evidence that investors do
respond positively to human rights and negatively to environmental issues.
This paper is related to the large literature examining the relationship be-
tween social responsible business practices and financial performance. Re-
lying on meta-analysis, Orlitzky et al. (2003) find a positive association be-
tween financial and corporate social performance. Kempf and Osthoff (2007)
and Statman and Glushkov (2008) study the impact of KLD screens on the
performance of investment portfolios. Both studies find significant excess re-
turns for portfolios long socially responsible, and short socially irresponsible
companies. In contrast, Schr¨
oder (2007) finds no differences in risk adjusted
returns between social and conventional benchmarks. My paper is different
since I do not study long term investment performance of socially responsible
trading strategies. In contrast, I focus on the short term implications of so-
cially (ir)responsible firm behavior for shareholder value. My results suggest
that excess returns of trading strategies based on KLD’s indicators like those
documented by Kempf and Osthoff (2007); Statman and Glushkov (2008))
are largely driven by the short component of the strategy.
Bechetti et al. (2007) analyze market reaction to entry (exit) announce-
ments to (from) KLD’s Domini 400 Social Index Index4 in an event study
setting. Their results suggest that the exclusion from the index causes signifi-
cantly negative abnormal returns. The negative reaction following a firm’s ex-
clusion from the index is most likely explained to passive social fund mangers
tracking the index. My paper differs from their study, since I rely on events
of social responsibility other than the exclusion or inclusion of companies in
KLD’s index.
On the theoretical side, Heinkel et al. (2001) show that with the presence
of ethical investors, prices of stocks of irresponsible companies are lower, driv-
ing up their expected returns. This is consistent with evidence in Hong and
Kacperczyk (2009) showing that returns on sin stocks (alcohol, gambling,

etc.) are higher than those of other companies. Relying on KLD data, Hong
and Kostovetsky (2009) examine whether political values influence invest-
ing and show that Democrat fund managers overweight firms with positive
social characteristics (good employee relations and superior environmental
Finally, there is a recent literature examining the validity and impact of
social responsibility ratings. Chatterji et al. (2009) investigate the predictive
power of KLD’s environmental ratings and find that KLD’s negative environ-
mental ratings are fairly good summaries of past environmental performance,
whereas positive ratings do not significantly forecast future levels of pollution
or compliance violations.
The rest of the paper is organized as follows. The next section presents the
sources of the data. Section 3 briefly reviews the methodology. In section 4,
I present and discuss the results before concluding in section 5.
Social analysts at KLD monitor current affairs via customized press searches.
In this process, the analysts single out information relevant to socially re-
sponsible investors. If an event (e.g. a newspaper article about a socially
controversial topic) concerning a company contained in the Russel30005 in-
dex is considered sufficiently important, the analysts will record, classify and
privately disseminate it in the form of a social rating. The present study uses
events between 2001 and 2007 concerning 836 companies different companies.
At present, KLD classifies (social) events according to two different sets
of criteria. On the one hand, it records incidents relevant for stakeholders in
its Environmental, Social and Governance (ESG) catalog. Table 1 provides
an overview of relevant issues.
[Table 1 (see page 26) about here.]
On the other hand, the Controversial Business Involvement (CBI) catalog
documents a company’s level of involvement in industries such as gambling,
tobacco or pornography. I focus on events belonging to the first catalog,
which covers the following stakeholder issue areas:
1. Community
2. Diversity
3. Employee Relations

4. Human Rights
5. Product
6. Corporate Governance
7. Environment
Some representative examples of the analyzed events can be found in
appendix A.
Concerns and Strengths
For each category, KLD has identified mutually exclusive positive and neg-
ative social indicators. Positive indicators are referred to as Strengths,
and negative ones are known as Concerns.
For instance, on July 17th
2003, Raleigh (North Carolina) based newspaper The News & Observer and
other information service providers6 made reference to a press release issued
by West Pharmaceutical Services, Inc. In the press release the company
announced that it had settled a citation issued by the North Carolina De-
partment of Labor, Occupational Safety and Health Division (NCOSHA)
following an investigation of an accident at one of the company’s facilities in
January 2003. The citation was issued by the NCOSHA on July 16th, 2003.
Subsequent to this event, KLD generated a negative Employee Relations rat-
ing7, which it worded as follows:
Following the January 2003 explosion at its manufacturing plant near
Raleigh, North Carolina, in July 2003, West Pharmaceutical Services
agreed to pay a penalty of $100,000 to the North Carolina Department
of Labor to settle allegations that the company had violated North
Carolina’s Occupational Safety and Health Act. The company
also agreed to pay $300,000 to local charities and groups that had
assisted with rescue work following explosion.
For an overview of the positive and negative social indicators under which
the analyzed events where filed, please refer to tables 2 and 3.
[Tables 2 and 3 (see pages 27 and 29) about here.]
@KLD Newsletters
In addition to maintaining a database with company ratings of the abovemen-
tioned form, KLD sends out irregularly spaced newsletters. The newsletters
contain the most noteworthy additions and modifications of social ratings
and inform about recent incidents of corporate social responsibility. I an-
alyze the information content of 92 such newsletters disseminated between

August 2001 and April 2007 concerning 836 different companies. During this
time frame, any two newsletters were separated by three weeks on average.
Figure 1 (see page 30) provides an example of the format and layout of a
typical @KLD newsletter.
[Figure 1 (see page 30) about here.]
From each newsletter I extract all social ratings and the date on which
the newsletter was privately disseminated, which I refer to as the Sent date.
[Figure 2 (see page 30) about here.]
Figure 2 shows the typical format of the ratings, which consist of the
following information:
• Company Name (Allied Capital Corporation)
• Company Ticker (ALD)
• KLD Category (Corporate Governance)
• KLD Social Indicator (Other Concern)
• Rating Text
• Source Date (11/12/2007)
The last item deserves further explanation. The Source Date refers to
the point in time when socially relevant information leading to a subsequent
KLD rating is publicly released. For example, this date might coincide with
the date a newspaper article on a socially controversial topic is published, a
regulatory filing is made or irresponsible corporate behavior is reported by
a non-governmental organization. In the example given above, the Source
Date was the day The News & Observer reported about the settlement.
Analysts at KLD systematically collect this kind of information and file it
under the respective reference date (Source Date).
Event selection
Whenever a positive and a negative event for a company fall on the same
Source Date, I drop both events from the analysis. Furthermore, I ran-
domly drop one of any pair of two events which are separated by less than 15
calendar days. I also remove events for which the three day average closing
price around the Source Date is less than $3. Ignoring penny stocks reduces
possible biases due to stocks with extreme return movements. Finally, some
of the references KLD disseminates do not make reference to a specific event.
For instance, in March 2002, KLD’s analysts sent out a newsletter containing
the following rating:

”As of December 2001, five of Honeywell’s facilities had received
a Star certification under the federal OSHA’s Voluntary Protec-
tion Program [VPP]. The VPP recognizes a plant for excellence
in safety and health management. The Star certification is the
highest award under the VPP. (12/31/2001)”
While the analysts filed this event with the Source Date being December
31st, 2001, a manual check on Dow Jones Factiva8 database reveals that
on and around that day no information concerning Honeywell’s health and
safety management was released in the press. I check and remove all events
with ambiguous time stamps. For a randomly chosen subset of 200 events, I
manually check on Factiva (1) for confounding events and (2) whether KLD’s
Source Date coincides with the date the information is publicly released. I
remove events with confounding events and in most of the cases the event date
in Factiva coincides with the Source Date. This procedure yields a final
sample of 2142 events. Table 4 provides an overview of events conditional on
stakeholder issue area.
[Table 4 (see page 32) about here.]
The table allows for two interesting observations. First, the newsletters
referenced substantially more negative than positive events. Second, the rel-
ative importance of each issue area differs for positive and negative events.
While traditional stakeholder issues related to the community, the diversity
and the environment make up more than 60 % of the positive events, these
categories only account for about 30 % of the negative ones. In contrast,
the set of negative events is dominated by issue areas of substantial eco-
nomic relevance: The product, employee relations and corporate governance
categories account for more than 70 % of all negative ratings. Overall, the
product category is the most important in absolute and relative terms: Ap-
proximately 34 % (733) of all disseminated events pertain to this category.
The second most important category with approximately 22 % (490) of all
disseminated events are employee relations. Purely social issues such as com-
munity, human rights and diversity only make up for approximately 25 % of
all ratings. Similarly, only about 10 % (203) of all events are concerned with
the company’s impact on the environment.
Content analysis
Besides categories and indicators, I also extract the event descriptions from
the newsletters. Analyzing these might deliver additional insights into what
differentiates KLD’s perception of negative and positive social responsibility.

Recent work in accounting and finance has focussed on the systematic analy-
sis of qualitative information in the form of textual data. Tetlock (2007) uses
a simple word counting algorithm to construct a market sentiment measure
from systematically analyzing Wall Street Journal’s ”Abreast of the Market”
column. Other studies have analyzed the information content of earnings
announcements (see Demers et al. (2008)) or the extent to which 10-Q and
10-K forms contain incremental information beyond financial measures such
as earnings surprises (see Feldman et al. (2008)). Here, I examine whether
the language used to describe the events differs systematically between pos-
itive and negative events. This form of content analysis seeks to extract
characteristics contained in text data in a systematic way. As Ole R. Holsti
puts it, content analysis is defined as ”any technique for making inferences
by objectively and systematically identifying specified characteristics of mes-
sages” (see Holsti (1969)). In mathematical terms, the technique boils down
to creating a quantitative profile of a text by mapping the words of the mes-
sage to predefined word categories. I focus on the following three predefined
word categories based on the Harvard IV-4 dictionary9:
[email protected] 510 words of an economic, commercial, industrial, or business
orientation, including roles, collectivities, acts, abstract ideas, and sym-
bols, including references to money. Includes names of common com-
modities in business. (e.g. ANTI-TRUST, AUCTION, BANKRUPT)
• Legal 192 words relating to legal, judicial, or police matters. (e.g.
• Quan 314 words indicating the assessment of quantity, including the
First, I examine to what extent event descriptions contain language be-
longing to a certain institution (e.g. economic or legal language). Second, I
test whether the presence of words used for the assessment of quantities or in-
dicating the presence of numbers differs systematically between positive and
negative event descriptions and across social categories. I start by compar-
ing negative and positive event descriptions independently from their social
category. Let wτ denote the total number of words and nk,τ the number of
words pertaining to word category k describing event τ . I am interested in
the statistical properties of r
k,τ =
, that is the relative importance of word

category k in the description of event τ . Since the fraction of words belonging
to a certain word category is non-negative and thus not normally distributed,
I rely on non parametric statistics and rank the data. I apply Kruskal-Wallis