Social Norms and CSR Performance: An

Social Norms and CSR Performance: An Examination of Positive Screening and Activism
by Norm-Constrained Institutional Investors
Steven F. Cahan†, Chen Chen, Li Chen
University of Auckland Business School, Private Bag 92019, Auckland 1142, New Zealand
September 2013
† Corresponding author
[email protected], +64 9 373 7599 x87175
Social Norms and CSR Performance: An Examination of Positive Screening and Activism
by Norm-Constrained Institutional Investors
Abstract
We examine the investment portfolio forming decisions and shareholder activism of
institutions that are exposed to social norms. We find strong evidence suggesting that the
shareholdings of these norm-constrained institutional investors are positively associated with
firms’ performance in the corporate social responsibility (CSR) area, supporting positive
portfolio screening. In addition, we document a strong relation between the lagged and current
changes in the shareholdings of norm-constrained institutional investors and the current changes
in investees’ CSR performance, supporting shareholder activism. Our results indicate that the
effect of social norms on norm-constrained institutions is more pervasive than prior research
suggests.
1
I.
Introduction
Hong and Kacperczyk (2009) find that institutional investors who are more constrained
by societal norms are more likely to avoid investing in ‘sin’ stocks, i.e., alcohol, tobacco, and
gaming. These institutions are ‘norm-constrained’ because their stock positions may be known
publicly, because they have diverse constituents, or because they are exposed to public scrutiny.
Examples of norm-constrained institutions include pension funds, universities, and religious
organizations.
However, the negative screening that Hong and Kacperczyk (2009) examine is only one
part of socially responsible investing. Hylton (1992) identifies three approaches to socially
responsible investing – negative screening, positive screening, and activism. While negative
screening focuses on avoiding questionable corporate behavior, positive screening focuses on
identifying investments that are attractive in terms of social responsibility, e.g., companies that
have progressive employment practices or that are environmentally sound. Activism reflects the
institution’s desire to change the investee’s behavior once the investment is made. Activist
investors use their ownership to promote positive improvements in the social and environmental
areas.
While Hong and Kacpersczyk (2009) provide evidence on negative screening, whether
social norms are strong enough to cause norm-constrained institutions to go beyond negative
screening and engage in positive screening and activism is an open question. This question is
important because negative screening differs fundamentally from positive screening and activism
– negative screening focuses on avoiding ‘bad’ practices while positive screening and activism
focus on promoting ‘good’ practices. Also, negative screening is a narrower concept as only
certain industries or practices are singled out for avoidance, while theoretically positive
2
screening and activism can be applied to any industry or any practice. Thus, further investigation
of the investment strategies of norm-constrained institutions is clearly warranted.
In this study, we examine whether norm-constrained institutional investors engage in
positive screening and activism with regard to corporate social responsibility (CSR). We define
norm-constrained investors in two ways. First, we use a modified version of Hong and
Kacperczyk’s (2009) definition and define pensions, universities, and religious, charitable, and
not-for-profit institutions as norm-constrained institutions. We exclude banks and insurance
companies from Hong and Kacperczyk’s (2009) definition because Brickley, Lease, and Smith
(1988) and Chen, Harford, and Li (2007) argue that banks and insurance companies may have
existing or potential business relationships with investee firms which reduce their incentive to
challenge management or agitate for change.
Second, we use a measure for norm-constrained institutions that incorporates the location
of the institutional investor. Recently, Hilary and Hui (2009) show that the local social
environment can influence the decision-making of managers in that locale. They focus on
religiosity and find that firms in counties with more religious adherents have lower risk
exposure, lower investment rates, and lower growth. However, religion is just one aspect of the
social environment. A community will have views on a variety of social issues. Further, because
of self-selection and acclimation, within a locality, views on a particular issue would be expected
to converge or coalesce (e.g., Halek and Eisenhauer, 2001), leading to local norms. While this
does not suggest complete uniformity across individuals in a community, in terms of the
community’s expectations for corporate behavior, on average, some areas will be more proenvironmental and more socially progressive than other areas. Thus, in the spirit of Hillary and
Hui (2009), our second measure considers whether the investing decisions of pensions,
3
universities, and religious, charitable, and not-for-profit institutions will be further constrained
by local norms.
Using data from 1991-2011 and ratings of CSR performance from the KLD STATS
database, we start by examining whether the current level of shareholdings of norm-constrained
institutional investors is related to the current CSR performance of the investee firm. We
emphasize that our sample includes all types of institutional investors, not just those funds that
follow a CSR-oriented investment strategy (e.g., the Calvert Social Index funds). That is, we are
interested in how social norms affect investment decisions of institutions at a general level.
Unlike CSR-oriented funds, in the absence of social norms, it is difficult to predict why some
institutions would prefer firms with good CSR performance more than other institutions.
We find that norm-constrained institutions hold more shares in firms with high CSR
performance when compared to institutions that are not constrained by social norms. This result
is consistent with norm-constrained institutions selecting and investing in firms with high CSR
performance (i.e., positive screening). When we split CSR performance into strengths and
concerns, we find that norm-constrained investors hold more (less) shares in firms with more
strengths (concerns). These results suggest we are capturing an effect that is distinct from the
one documented by Hong and Kacperczyk (2009). Because negative screening is based on a noncompensatory decision rule based on negative actions, if our results are driven by negative
screening, only concerns would matter.
To shed more light on the activist role, we examine the effect of lagged changes in
institutional holdings on the current change in CSR performance of the invested firms. We find
that the one-year lagged and two-year lagged changes in shareholdings of norm-constrained
institutions are related to the current changes in CSR performance. This suggests that norm-
4
constrained institutions promote or agitate for CSR improvements in firms that they hold shares
in, a view consistent with institutional activism. Further, we show that norm-constrained
institutions with long-term and substantial holdings, who have the most incentive to take an
active interest in the firm, exhibit a stronger influence on firms’ CSR performance than other
norm-constrained institutions, or other long-term and substantial institutions.
Next, we consider whether norm-constrained institutions follow a CSR-oriented investing
strategy consistently over time. Hong, Kubik, and Scheinkman (2012) argue that financial
constraints limit the ability of firms to be a good corporate citizen. As a result, it will be more
difficult for firms to increase their CSR performance during recessions than in expansions. We
find that the sensitivity of norm-constrained institutions’ shareholdings on investee’s CSR
performance reduce significantly in recession periods than in expansion periods, consistent with
the benefits of activism being lower during an economic downturn.
Our paper contributes to the literature in the following ways. First, we complement Hong
and Kacperczyk (2009) by examining the role of positive screening and activism by institutional
investors in a CSR setting. Hong and Kacperczyk (2009) only examine negative screening.
Second, we contribute to the finance literature which studies the monitoring role of institutional
investors in influencing firms’ daily management. Brickley, Lease, and Smith . (1988), Agrawal
and Mendelker (1990), Bushee (1998), Hartzell and Starks (2003), Almazan, Hartzell and Starks
(2005), Borokhovich, Brunarski, Harman, and Parrino (2006), and Chen, Harford, and Li (2007)
have shown that certain types of, but not all, institutional investors exert influence on antitakeover amendments, R&D investment decisions, CEO compensation, and M&A performance.
Our study contributes to this stream of literature by showing that firms improve their nonfinancial performance – specifically, their CSR performance – in response to increased
5
ownerships of norm-constrained institutional investors. To our knowledge, no prior study has
examined the activist role of norm-constrained institutions. Third, we consider whether the
investing patterns of norm-constrained institutions vary over time. Fourth, we consider the effect
of local attitudes toward CSR as an additional set of norm-based constraints. Thus, we contribute
to a growing line of research that examines the effects of location (e.g., Malloy 2005, Pirinsky
and Wang 2006, Kang and Kim 2008, Hilary and Hui 2009). We show that local norms act as a
further constraint on the investment decisions of norm-constrained fund managers in that locale.
Finally, our findings have practical importance as ‘responsible’ investing becomes more
visible. For example, the United Nation’s ‘Principles of Responsible Investment’ (PRI) project
encourages responsible investing by institutional investors where responsible investing is defined
as “an approach to investment that explicitly acknowledges the relevance to the investor of
environmental, social and governance (ESG) factors, and the long-term health and stability of the
market as a whole”.1 Currently, over 1,160 institutional investors worldwide have agreed to
support the PRI initiative. While the efficacy of such an initiative is a separate question, our
results suggest that, even in the absence of engineered agreements, social norms can cause
institutions to take a more proactive stance toward social responsibility.
The remainder of the paper is divided as follows. Section 2 reviews the literature and
develops our hypothesis. Section 3 describes the research design. Section 4 provides the results,
and Section 5 is the conclusion.
II.
Literature and Hypotheses Development
A condition for an association between institutional ownership and CSR performance is
that the institutional investors make socially responsible investments. Said differently, socially
1
See http://www.unpri.org/about-ri/introducing-responsible-investment/.
6
responsible investments are the link between institutional investors and CSR performance.
Hylton (1992) summarizes two distinct mechanisms of socially responsible investing: portfolio
screening and shareholder activism. We use this as a framework to review prior literature on
institutional investors and CSR.
A.
Portfolio Screening
Portfolio screening in socially responsible investing includes negative and positive
screening (Hylton, 1992). The negative screening, or investor boycott, is an approach of
investment marked by an emphasis on separation and avoidance of ethically questionable
corporate behaviors. A common example is that socially responsible investors may refuse to
invest in firms that engage in sin industries such as alcohol, tobacco, and gaming. Hong and
Kacperczyk (2009) focus on negative screening and investigate certain types of institutions and
their ownerships of the sin stocks. They argue that pension funds, universities, religious
organizations, banks and insurance companies are subject to social norms, and find that these
institutions hold less sin stocks than other types of institutions. Hong and Kacperczyk’s (2009)
findings provide evidence suggesting that norm-constrained institutional investors employ a
negative portfolio screening method to practice socially responsible investing.
Positive screening, on the other hand, is to identify desirable firms. For example, the
forum for Sustainable and Responsible Investment in US states that positive screening involves
including strong CSR performers – profitable companies that contribute positively to society – in
an investment portfolio. The organization notes that ‘buy’ lists may include “enterprises with, for
example, good employer-employee relations, strong environmental practices, products that are
safe and useful, and operations that respect human rights around the world.”2 The positive
screening approach creates shareholders’ demand for good CSR performance. Harjoto, Jo, and
2
See http://ussif.org/resources/sriguide/srifacts.cfm.
7
Kim (2012) note that institutions may demand shares in good performing CSR companies
because their constituents prefer socially responsible investing even if it is costly. Fernando et al.
(2010) and Chava (2010) focus on institutional demand for firms with better environmental
performance.
Prior studies examine all institutional investors as a single group (e.g., Chava, 2010) or
rely on Bushee’s (1998) classification of institutions based on their investment horizon, i.e.,
dedicated, transient, or quasi-indexer (e.g., Dhaliwal, Li, Tsang, and Yang, 2011; Harjoto, Jo,
and Kim, 2012). Dedicated institutions are characterized by low turnover and concentrated
portfolios. Transient institutions have high turnover and diversified portfolios. Quasi-indexers
have low turnover and diversified portfolios. However, while Bushee’s (1998) categories are
widely accepted in the literature, their link to CSR performance is indirect and unclear. Harjoto,
Jo, and Kim (2012) argue that institutions with long horizons (i.e., low turnover) are “more likely
to appreciate social values from CSR activities whereas institutional investors with shorthorizons are sensitive to short-term changes in earnings.”
While investment horizon is likely to matter (and we consider it in later tests), our main
focus is on the role of social norms. Hong and Kacperczyk (2009) find that institutions that are
constrained by social norms are less likely to invest in alcohol, tobacco, or gaming stocks. Rather
than focusing on negative screening – a reactive strategy that only applies to a limited set of
stocks – we focus on positive screening. Positive screening is proactive in that it rewards or
encourages certain ‘good’ behaviors, and positive screening can be applied more broadly since
all firms have some responsibilities as corporate citizens (e.g., all firms have employees, all firms
impact the environment). However, positive screening is also more costly and difficult to apply.
Unlike negative screening which requires a binary decision in relation to a simple criterion (e.g.,
8
industry membership), positive screening requires that the institution identifies selection criteria,
determines suitable measures of the criteria, and establishes benchmarks to identify ‘good’ or
‘best’ practice.
As a result, it is not clear whether the social norms that lead institutions to engage in
negative screening (i.e., Hong and Kacperczyk, 2009) also will lead these institutions to partake
in positive screening. If norm-constrained institutional investors positively screen for good CSR
performance, we expect that the level of CSR performance will be positively associated with the
level of holdings by norm-constrained institutional investors relative to other institutional
investors.3 This leads to the first hypothesis (all hypotheses are stated in the alternative form):
H1:
Compared to other institutional investors, the shareholdings of norm-constrained
institutional investors are more positively associated with firms’ CSR performance.4
B.
Shareholder Activism
With respect to the shareholder activism mechanism, activist shareholders not only want
to avoid ‘bad’ firms and choose ‘good’ firms, they also want to promote improvement in firms
they have invested in. Institutional investors’ roles in influencing firms’ daily business are well
documented in the prior literature. For example, Brickley, Lease, and Smith (1988), Agrawal
and Mendelker (1990), Bushee (1998), Hartzell and Starks (2003), Almazan, Hartzell, and Starks
(2005), Borokhovich, Brunarski, Harman, and Parrino (2006), and Chen, Harford, and Li (2007)
3
Initially, our hypotheses and tests are based on comparisons between norm-constrained institutional investors and
other institutional investors. However, when we incorporate local norms, we compare norm-constrained institutions
located in areas with more positive CSR attitudes with norm-constrained institutions in areas with less positive CSR
attitudes.
4
In a somewhat related paper, Hong and Kostovetsky (2012) examine whether the political orientation of mutual
fund managers is associated with their propensity to invest in socially responsible companies. They classify fund
managers as Democrats or Republicans based on their political contributions. They find Democratic fund managers
are less likely to invest in ‘socially irresponsible’ companies such as firms involved in the tobacco, firearms, or
defense industries or firms with poor employee relations or hiring practices that lack diversity. One difficulty in
examining political donations is that the majority (66%) of fund managers are classified as ‘non-donors’ in their
tests. Also, political orientation is a broad concept that applies to one’s outlook on a variety of issues, e.g., ranging
from economic policy and national defense to specific issues like abortion and gun control. In our study, we
construct a measure of social norms that relates specifically to CSR.
9
have shown that certain types of, but not all, institutional investors exert influence on firms’
antitakeover amendments, R&D investment decisions, CEO compensation, and M&A
performance. Investments by norm-constrained institutions may promote subsequent
improvement in CSR performance. This could be because, once invested, the institutions are in a
better position to monitor and evaluate CSR performance. For instance, they can appoint their
representatives in the corporate board. Also, they may develop closer relationships with the
firm’s managers which gives them more influence (e.g., Useem, 1996).
Given that shareholder activism is documented in the normative CSR literature with realworld cases, e.g., Ralph Nader’s campaigns to improve General Motors’ vehicle safety and
reduce automobile pollution (Hylton, 1992), surprisingly there is lack of empirical research on
shareholder activism in the CSR context. If social norms create pressure for norm-constrained
institutions to become involved shareholder activism and if such activism is effective, we should
observe a relation between the current change in CSR performance and prior changes in the
shareholdings of norm-constrained institutions. This leads to a second hypothesis related to
shareholder activism:
H2:
Compared to other institutional investors, increased holdings of norm-constrained
institutional investors lead to firms’ CSR improvements in subsequent periods.
III.
Research Design
A.
Data and Sample
We obtain data on firms’ CSR performance from the KLD STATS database. KLD
evaluates the environmental, social, and governance performance for firms it covers along a
variety of dimensions. From 1991-2000, KLD covered approximately 650 firms every year. In
2001 and 2002, KLD included the largest 1,000 US firms by market capitalization. Since 2003, it
10
has expanded its coverage to the largest 3,000 US firms by market capitalization. Our data on
shareholdings come from the CDA Spectrum database of SEC Form13-F filings by institutional
investors, defined as those managing at least $100 million in assets. We focus on the end-of-year
holdings of a particular stock in terms of shares held by various types of institutional investors.
We extract data from the Compustat and CRSP databases to calculate the necessary control
variables in our models. Our final sample comprises firms that have requisite data from the KLD
STATS, CDA Spectrum, Compustat, and CRSP databases.
B.
Measurement of Variables
1.
CSR scores
KLD evaluates firms’ CSR performance in seven qualitative issue areas, i.e., corporate
governance, community, diversity, employee relations, environment, human rights, and product.
Within each area, KLD provides strength and concern ratings for multiple indicators. KLD also
reports controversial business issues such as alcohol, gambling, tobacco, firearms, military, and
nuclear power. We do not consider the nature of the industry in our analyses (although we
control for industry fixed effects) because we focus on the association between the institution
type and the CSR performance rather than the association between the institution type and ‘sin’
industries which Hong and Kacperczyk (2009) investigate.
Institutional investors may prefer firms with good corporate governance due to various
incentives documented in prior literature. Firms with good corporate governance (1) provide
strong internal monitoring mechanism that serve as a substitute for institutions’ costly
monitoring activities (e.g., Bushee and Noe, 2000), (2) exhibit higher firm value, better operating
performance, and potentially less wasteful investment (e.g., Gompers, Ishii, and Metrick, 2003;
Brown and Caylor, 2006; Larcker, Richardson, and Tuna, 2007), and (3) can reduce the
11
possibility of losses due to managerial fraud or negligence (e.g., Del Guercio, 1996). In order to
disentangle the effect of CSR and corporate governance on institutional holdings, we construct a
CSR score (CSR1), measured as total strengths minus total concerns in KLD’s six social rating
categories: community, diversity, employee relations, environment, human rights, and product.
That is, we exclude KLD’s corporate governance category in computing our CSR score. Instead,
consistent with Kim, Park, and Wier (2012), we include the net score for the corporate
governance category as a control variable in our regression models.5
In addition, we compute a second measure of CSR performance (CSR2) that excludes the
KLD product category. Product ratings include indicators such as quality, R&D/innovation, and
product safety, which may be correlated with firms’ financial performance. To disentangle the
effects of CSR performance and financial performance on institutional ownership, we use an
alternative CSR score calculated as total strengths minus total concerns from five of KLD’s
social rating categories: community, diversity, employee relations, environment, and human
rights.
2.
Institutional ownership and type of norm-constrained institutional investor
Institutional ownership (IO) is the percentage of a firm’s shares outstanding held by
institutions in the CDA Spectrum database at the end of year t. We calculate IO by aggregating
the shareholdings for different types of institutions at the end of the year and then dividing this
amount by total shares outstanding at the end of the year.
We construct an indicator variable to indicate whether an institution is norm-constrained
5
Kim, Park, and Wier (2012) excluded the human rights area in constructing CSR scores because this area was
added from 2002. However, we include this area for two reasons: (1) as stated in KLD’s ratings definitions, ratings
in this area were mostly taken from the former non-US operations and community category, which have been
embedded in the net score before 2002, and (2) to the extent that investors may view human rights as an important
aspect of firms’ CSR performance, a CSR score including this area will reflect a more complete CSR image of
firms.
12
(NC). We use two methods to define NC. First, similar to Hong and Kacpercyzk (2009), we use
the type of institution and define pensions, universities, and religious, charitable, and not-forprofit institutions as norm-constrained institutions. Because Brickley, Lease, and Smith (1988)
and Chen, Harford, and Li (2007) argue that banks and insurance companies may be less
independent due to existing or potential business relationships with investee firms, we do not
include banks and insurance companies in our definition of norm-constrained institutions.
Consistent with prior research, we view mutual funds and investment advisor as natural
arbitrageurs that are motivated purely by financial objectives.6 Thus, our first measure of normconstrained institutions, NC1, is equal to 1 for pensions, universities, and religious, charitable,
and not-for-profit institutions and 0 for all other institutional investors (i.e., banks, insurance
companies, mutual funds, investment advisors).
We based our classifications for NC1 on the CDA Spectrum database which contains
information from 13-F filings. The CDA Spectrum database classifies institutional investors into
five types: banks, insurance companies, mutual funds, investment advisors, and others. We
search the internet for detailed information on institutions classified as ‘others’ and find some
banks, insurance companies, mutual funds, and investment advisors are incorrectly classified.
Accordingly, we reassign these institutional investors into the appropriate group.7 After being
reclassified, the group ‘others’ includes pension funds, universities endowments, and religious,
charitable, and other non-for-profit organizations.
Our second measure for norm-constrained institutions, NC2, augments our first measure
by incorporating the location of each institution’s head office. We expect that a community’s
6
Similar to Hong and Kacperczyk (2009), we argue that while certain mutual funds and hedge funds may specially
practice socially responsible investing, other mutual and hedge funds still focus solely on achieving financial goals.
7
We removed a small number of fund managers for which we do not have sufficient information to make an
accurate classification.
13
norms and values will vary across different areas within US. These tests are motivated by Hilary
and Hui (2009) who find that the religiosity of geographic regions in the US is associated with
managerial decision making. Also, Rubin (2008) and Di Giuli and Kostovetsky (2012) provide
some evidence that the political views at the state-level are associated with CSR performance.
Our second measure, NC2, is based on the notion that the investing strategy of fund managers in
norm-constrained institutions will be further constrained by local norms regarding CSR, so a
manager in a pro-social responsibility area will invest more heavily in good CSR performing
firms.8
We use the definitions of Metropolitan and Micropolitan Statistical Areas (MSA) under
the 2006 standards established by the Office of Management and Budget to separate the
geographical areas.9 We collect information on zip codes for firms and institutional investors
who filed 13-F, and match them with the corresponding MSAs. For each MSA, we calculate an
average CSR score of all firms with CSR ratings within that area.10 We rank all MSAs on the
area-based CSR scores, and define the top 30% as areas with strong local norms and the bottom
30% as areas weak local norms. Since these tests include only those institutions where NC1
equals 1, if the headquarters of a NC1 institution is located in an area with strong local norms, we
consider it to be norm-constrained since it faces further constraints relative to NC1 institutions in
other locations. In calculating the ownership of these institutions on the firms, we exclude
ownership by institutions in the same MSA area to eliminate the possibility that our results are
driven by a local or home bias where institutional shareholders hold more shares in the firms that
8
This does not mean that the fund manager cannot share the same view as the community. In fact, because of selfselection and acclimation, the fund manager’s and community’s views are likely to be similar in a general sense.
This is no different than a fund manager for a pension fund or university having a pro-CSR attitude.
9
Information on MSA is from the Office of Management and Budget website at http://www.whitehouse.gov/OMB
under “Bulletins” or “Statistical Programs and Standards”.
10
We require that each MSA area should have at least 10 firms with CSR score to calculate its average score.
14
are proximate to them (e.g., Coval and Moskowitz 1999, Campbell 2006).
Thus, we define NC2 equals 1 if the institution’s headquarter is located in an areas with
strong pro-CSR attitudes and the institution is a pension fund, university, or religious, charitable
or non-profit organization, and 0 if the institution’s headquarter is located in an area with weak
CSR attitudes and the institution is a pension fund, university, or religious, charitable or nonprofit organization.
C.
Regression Models
1.
Positive screening (H1)
To test H1, i.e., the association between CSR performance and ownership by norm-
constrained institutions versus other institutions, we estimate the following models:
(
)
∑
(1)
where
IO = institutional ownership by norm-constrained or other institutions at the end of year t,
calculated as a percentage of a firm’s shares outstanding;
CSR = the total CSR score for a firm at the end of year t. There are two measures under this
variable. CSR1 is the net score (strengths minus concerns) of community, diversity,
employee relations, environment, human rights, and product. CSR2 is similar except
it excludes the net score of product ratings;
NC = an indicator variable that equals 1 if the institution is classified as norm-constrained.
There are two types of classification. NC1 equals 1 if the institution is a pension fund,
university, or religious, charitable or non-profit organization, and 0 if the institution
is a bank, insurance company, mutual fund or investment advisor. NC2 equals 1 if
the institution’s headquarter is located in a norm-constrained area and the institution
15
is a pension fund, university, or religious, charitable or non-profit organization, and 0
if the institution’s headquarter is located in a non-norm-constrained area and the
institution is a pension fund, university, or religious, charitable or non-profit
organization.
Based on Del Guercio (1996), Bushee (2001), and Dhaliwal, Li, Tsang, and Yang (2011),
we control for variables that capture firms’ characteristics that can affect the investment
decisions of institutional investors, including corporate governance (CG), size (LMV), reputation
(SP500, Rating), financial and stock performance (ROE, AnnRet), liquidity (TV), risk (Lev, Beta,
IRisk), and fundamentals (EP, DP, BP, SaleChg). We note that financial and stock performance,
and reputation can control for good management which is relevant since good management and
CSR performance could be related. However, we emphasize that, in our context, good
management is not likely to be an omitted correlated variable because it is difficult to envision a
scenario where only norm-constrained institutions are interested in good management. See the
Appendix for the variable definitions.
Our empirical objective is to determine whether better CSR performance is associated
with more shareholdings by norm-constrained institutions. In equation (1), the estimate of the
association between CSR performance and holdings is
(NC = 0), and
for non-norm-constrained institutions
for norm-constrained institutions (NC = 1). If norm constrained
institutions invest more in firms with better CSR performance relative to other institutions, we
expect
2.
, the coefficient of interest, to be positive.
Shareholder activism (H2)
H2 hypothesizes that norm-constrained institutional investors can improve the CSR
performance of firms they have already invested in through activism. We estimate the following
16
model that examines the effect of current and lagged changes in holdings of norm-constrained
institutions versus other institutions on current changes in CSR performance:
(
(
)
)
(
(
)
)
(2)
∑
where
= a change from year t-1 to year t.
IOV = the institutional ownership with voting rights held by norm-constrained institutions or
other institutions at the end of year t, calculated as a percentage of a firm’s shares
outstanding;
All the other variables are defined as in equation (1).
We use institutional ownership with voting rights (IOV) in order to concentrate on firms
where institutions have the voting power to influence the decision-making process. We include
the concurrent and previous two periods’ changes in the test variable IOV (i.e., year t, t-1, and t2). Any significant results on changes in IOV, particular the lagged changes, will imply changes
in voting shareholdings leading to changes in CSR performance. Since the change in voting
shares is affected by the level of voting shares (e.g., institutions holding a low level of shares
have more room to increase their shareholdings), and since firms may continue to improve their
CSR performance based on historic – rather than current – holdings of norm-constrained
institutions, we also include the IOV variable at the beginning of the first change period (i.e., the
end of year t-3). We use a set of control variables that may affect firms’ CSR performance,
including firm-specific changes in S&P 500 status, leverage, book to market value of equity ratio,
ROE, and sales growth.
17
IV.
Results
Table 1 summarizes the distribution of relevant variables for the 1991-2011 period. We
trim all continuous variables at 1 and 99 percentiles of that variable’s distribution. For the
institutional ownership variable (IO) and ownership with voting rights variable (IOV), we also
present total IO (IOV), IO (IOV) for norm-constrained institutions and other institutions, using
three different measures, respectively. For the tests involving NC1 institutions, the mean for total
institutional shareholdings (holdings with voting rights) is 66% (41%).11 NC2 generates a lower
mean IO (IOV) which is expected since we omit investments by local institutions. The mean for
CSR scores using the first CSR measure (CSR1) is -0.206. This indicates that concerns outweigh
strengths on average, which is consistent with prior studies (e.g., the mean CSR score is negative
in Kim, Park, and Weir, 2011). The mean for our alternative CSR measure (CSR2), excluding the
product rating, is 0.075, higher than the first measure. Control variables are all consistent with
prior studies.
Table 2 reports the correlation matrix for the main and control variables with Pearson
correlations below the diagonal and Spearman correlations above the diagonal. The correlations
between total IO (IOV) and CSR scores are not conclusive. Almost all control variables are
significantly correlated with the total institutional shareholdings as well as voting shareholdings.
A.
Results for H1
Table 3 reports the multivariate regression results from estimating equation (1). We
present results for four combinations of two measures for CSR performance and two measures
11
The mean institutional holdings in our sample is higher compared with the mean holdings in the full dataset from
the CDA Spectrum database, which is 28% for shareholdings and 12% for shareholdings with voting rights. This
implies that firms covered by KLD attract more institutional ownership than other firms, on average.
18
for norm-constrained institutions.12 The overall results show that our models are well specified
with reasonably large R2 values. The explanatory power is strong with a value of 88% in the
regressions with the NC1 measure (based on the kind of institution), and less strong but still
reasonably good (38%) when the location-based NC2 measure is used. The coefficient on the
institution type indicator (NC) for the first measure (NC1) is negative and strongly significant
because the holdings of institutions when NC1=1 (e.g., pensions, etc.) are much lower relative to
institutions when NC1=0 (e.g., banks, mutual funds, etc.). Specifically, Table 1 shows that the
mean holdings when NC1=1 (IO_NC1) is 0.0375 whereas the mean holdings when NC1=0
(IO_Non-NC1) is 0.6306. Obviously, it is important to control for this effect by including
institution type in the regression as we have done. For the second type measure NC2, the mean
holdings when NC2=1 (IO_NC2) is 0.0478, close to IO_Non-NC2 (0.0482). As a result, the
coefficient on NC for the second measure NC2 is not as highly significant.
The coefficient of interest,
for CSR*NC, measures whether better CSR performance is
associated with more shareholdings by norm-constrained institutions relative to other institutions.
The results show that
is significantly positive in all four regressions in Table 3. For example,
when the CSR1 and institution NC1 measures are used,
is 0.0054 with a p-value less than
0.001. In terms of economic magnitude, this suggests that a one standard deviation increase in
the CSR score (s.d. for CSR1 = 1.88 from Table 1, panel A) would increase the shareholdings by
norm-constrained institutions by 1%. Since the average shareholdings by norm-constrained
institutions is 3.75% (from Table 1, panel A), this equates to a 27% increase in the portfolio
holdings of the norm-constrained institutions. Thus, the results indicate that norm-constrained
institutions, relative to other institutions, have more holdings in firms with better CSR
12
In subsequent tables, we only show the first CSR measure for brevity. The alternative CSR score measure
generates similar results for all regressions.
19
performance, and the effect is statistically and economically significant.
Most of the control variables are significant except EP and ROE. For corporate
governance (CG), norm-constrained institutions prefer firms with better governance mechanism
relative to other institutions, as implied by the significant and positive coefficient on CG*NC.13
For other control variables, institutional ownership is positively associated with size (LMV),
reputation (SP500, Rating), recent stock performance (AnnRet), liquidity (TV), leverage (Lev),
market model beta (Beta), book to market value of equity (BP), and negatively associated with
dividends to market value of equity (DP), percentage of sales change (SaleChg), and the standard
deviation of the market-model residuals (IRisk).
To address the concerns that the significant differential coefficient on CSR*NC may be
caused by the omission of differences in control variables between the two types of institutions,
we interact all control variables with the indictor variable NC in untabulated analyses. We find
the results still hold.
Overall, the results in Table 3 support H1, suggesting that norm-constrained institutions
seek out high performing CSR firms, which is consistent with positive screening. This adds
evidence to the literature that norm-constrained institutions not only exclude sin stocks as Hong
and Kacperczyk (2009) document, but that they also engage in positive screening when forming
investment portfolios. However, we acknowledge that these results can also be viewed as being
consistent with institutional activism, i.e., norm-constrained institutions that bought large stakes
in poor performing CSR firms in the past could have influenced the investee to raise its CSR
performance so that these firms have high CSR performance in the current period. In reality,
positive screening and activism are not mutually exclusive – and anecdotal evidence suggests
13
We include the CG*NC interaction as a control variable to ensure that our variable of interest, CSR*NC, is not
capturing a general preference among norm-constrained institutions for firms with good governance.
20
that both are likely to occur together.14 We provide a more direct test of institutional activism
(i.e., H2) below. Nonetheless, the results demonstrate a positive association between normconstrained institutional holdings and CSR performance, consistent with these institutions taking
an active interest in CSR performance.
A more serious alternative interpretation is that our results are driven by negative
screening, not positive screening. That is, since we measure CSR performance using a total score
of strengths minus concerns, it is possible that differences in CSR are driven by differences in
concerns. For example, if two firms have the same number of strengths but firm A has more
concerns than firm B, the variation in CSR will be due to concerns. If so, our results may be
driven by norm-constrained institutions seeking to avoid investing in firms with poor CSR
performance. In other words, we might just be detecting another form of negative screening.
To determine whether this is the case, we separate the strengths and concerns into two
separate measures. If negative screening is the prominent effect, we expect that the results will
be significant for the interaction between concerns and NC, but not strengths and NC. This is
because negative screening is a non-compensatory decision rule, i.e., if a firm is performing
badly in some CSR areas, it does not matter if the firm is performing well in other CSR areas.
For instance, if a firm is involved in alcohol production, using a negative screening rule, a normconstrained institution would not invest in the company, regardless of the positive contributions
that firm might make to the community. On the other hand, positive screening focuses on the
overall CSR performance, selecting firms that follow best practice. Thus, a positive screen takes
in account a firm’s CSR strengths as well as its CSR weaknesses.
Consequently, we replace the total CSR score with separate variables for strengths and
For example, the UN’s PRI website encourages signatories to consider principles of responsible investing both
before they invest and later as on-going, active shareholders. See http://www.unpri.org/signatories/become-asignatory/.
14
21
concerns, and examine whether institutional investors care about both positive and negative
aspects of firms’ CSR performance, i.e.:
(
)
(
)
∑
(3)
where Str (Con) is the total CSR strengths (concerns) of a firm’s ratings in six areas: community,
diversity, employee relations, environment, human rights, and product. All other variables are
defined as in equation (1). As strengths and concerns in the KLD database are both count
numbers, if negative screening predominates, we expect that
and
expect
on Str*Type will be insignificant
on Con*Type will be negative and significant. If positive screening predominates, we
to be positive and significant, and
to be negative and significant.
Table 4 reports empirical results from estimating equation (3). The coefficient for
Str*Type is significantly positive (e.g.,
= 0.0052, p-value < 0.001 for NC1) and the coefficient
for Con*Type is significantly negative (e.g.,
= -0.0055, p-value < 0.001 for NC1) for all
regressions with different definitions of norm-constrained institutions. In sensitivity tests, we
also define Str and Con to exclude the product dimension and find similar results.
In short, we find evidence suggesting that both strengths and concerns in firms’ CSR
performance matter to norm-constrained institutional investors. Thus, we are not merely
capturing a different aspect of the negative screening that Hong and Kacpercyzk (2009) identify
in their study. Instead, our results extend Hong and Kacpercyzk (2009) by showing that social
norms can lead norm-constrained institutions to engage in positive screening.
B.
Results for H2
Table 5 reports results from estimating equation (2) in which the change in CSR is the
dependent variable. The coefficients of interest include
22
,
, and
that measure the
incremental effect of changes in voting shares by norm-constrained institutions versus other
institutions on changes in firms’ CSR performance. For NC1, the results show that all three
coefficients on the interaction between ∆IOV and NC1 are strongly significant (p-value < 0.001)
and positive (4.59, 5.08, and 5.59 for
,
, and
, respectively). In particular, the findings for
– the change in voting shares in previous two years – suggest that an increase in voting
and
shares held by norm-constrained institutions, when compared with other institutions, leads to
greater improvement in the investee firm’s CSR performance one and two years down the track.
This evidence is consistent with the investor activism by norm-constrained institutions, i.e., the
institutions that are subject to social norms take an active interest in improving the CSR
performance of firms they have invested in.
For the NC2 measure, the results on the coefficients of interest are similar to those for
NC1.
*Type)
,
, and
for the current and lagged changes (∆IOVt *Type, ∆IOVt-1 *Type and ∆IOVt-2
are all positive and significant at the 0.05 level at least. This finding indicates that among
norm-constrained institutions (i.e., pensions, universities, and religious, charitable, and nonprofit organizations), those that face additional pressure from local norms have a greater
inclination to actively monitor the invested firm, take a more active interest in improving firms’
CSR performance when compared to similar institutions from areas that have less intense CSR
attitudes. Thus, our results for NC2 support the local social environment can act as an additional
constraint on the investment decisions of some fund managers.
While the main results are qualitatively similar between the NC1 and the NC2 measures,
it is worth noting that coefficients on the current and lagged changes in shareholdings for the
benchmark group are different. Coefficients on ∆IOVt, ∆IOVt-1 , and ∆IOVt-2 are statistically
insignificant for NC1 because the benchmark group in this instance is non-norm-constrained
23
institutions that have fewer incentives to be activists. As such, we do not expect shareholdings
from these institutions would have any impact on firms’ CSR performance. In contrast,
coefficients on ∆IOVt, ∆IOVt-1 , and ∆IOVt-2 are positive and statistically significant for NC2.
This is because the benchmark group is pensions, universities, and religious, charitable, and notfor-profit organizations that are located in areas with weaker local norms in regard to CSR. Even
though these institutions exhibit less shareholder activism compared to the similar institutions
from areas with strong CSR norms, these institutions still care about social values. Therefore, the
results for the benchmark institutions are consistent with our variable constructions.
Overall, Table 5 provides evidence consistent with H2, suggesting that norm-constrained
institutions use their influence to improve the CSR performance of their invested firms in
subsequent periods.
C.
Additional Analyses
1.
Long-term and substantial holdings in tests of shareholder activism
Chen, Harford, and Li (2007) document that institutional investors with long-term (at
least one year) and substantial holdings in a firm specialize in monitoring activities rather than
gaining profit from short-term trades. We now focus on the NC1 measure and consider the size
of the stake and the length of investment when estimating equation (2) for the investor activism
hypothesis (i.e., H2). Specifically, we identify institutions which are among the five largest
institutional investors of a firm and at the same time hold the shares in that firm for more than
four years.15, 16 Consequently, we have a matrix of institutions by ‘norm-constrained’ and ‘long-
15
We also use the alternative requirement for substantial holdings such as the single largest institutional investor, or
holdings with at least 5% of the shares, as in Chen, Harford, and Li . (2007). The results are weaker due to the small
sample size.
16
We define institutions as long-term shareholders if they hold the stocks of one particular firm for at least four
years as our shareholder activism model in equation (2) requires shareholdings for consecutive four years. Our
results remain similar if we relax our criteria into at least two or three years.
24
term and substantial’. Then, we conduct two different comparison analyses. First, we focus on
norm-constrained investors only, and compare long-term and substantial institutions with other
institutions. The underlying rationale is that, while all norm-constrained institutions may care
about firms’ CSR performance, those with long-term and substantial holdings will have stronger
power in influencing firms’ operations, and hence, we expect that these institutions will show
higher level of shareholder activism than other norm-constrained institutions. Second, we focus
on long-term and substantial institutions only, and compare norm-constrained institutions with
other institutions. We reason that while long-term and substantial institutions prefer to exert
monitoring effort, norm-constrained institutions will be more willing to put effort in improving
firms’ CSR performance than other institutions. Finally, we use the long-term and substantial
and norm-constrained institutions only to re-examine equation (2) to provide further evidence of
their role in monitoring firms’ CSR performances.
Table 6 reports results from three different analyses. In the first comparison analysis
shown in column (1), we only look at the concurrent change in IOV on changes in CSR, because
it is not sensible to examine lagged changes when some institutions hold their shares for less than
four years. The coefficient of the interaction term, ∆IOVt *NC, is significantly positively
associated with the current change of CSR performance, indicating that compared to other normconstrained investors, long-term and substantial norm-constrained investors have larger
influences on firms’ CSR performance. With respect to the results of the second comparison
analysis show in column (2), the coefficients of three interaction terms, ∆IOVt *NC, ∆IOVt-1 *NC
and ∆IOVt-2 *NC are all significantly positively associated with the current change of CSR
performance, suggesting that compared to other long-term and substantial institutions, the longterm and substantial and norm-constrained institutions are more likely to affect investees’ CSR
25
performance. Column 3 presents results for current and lagged changes for norm-constrained
institutions with long-term and substantial holdings. Results show that these institutions exhibit
influence on firms’ CSR performance with all the coefficients on the change in the current and
lagged shareholdings significantly positively correlated with the change in firms’ CSR
performance.
Overall, our findings suggest that norm-constrained institutional investors with long-term
and substantial holdings have stronger influence on improving investees’ CSR performance,
when compared to other norm-constrained institutional investors or other long-term and
substantial institutional investors. This provides additional insight about the activist role of
norm-constrained institutions.
2.
Macroeconomic conditions in tests of shareholder activism
We investigate whether the effectiveness of shareholder activism on CSR performance
vary with the economic conditions. Hong, Kubik, and Scheinkman (2012) finds firms do good
only when they have sufficient financial slack. During economic recessions, firms’ resources are
restrained. Even with the same level of monitoring inputs from pro-CSR activist investors, one
would expect that the improvement of CSR performance may be less compared to economic
expansions, simply because firms do not have enough resources to invest in CSR-related projects.
To test this hypothesis, i.e., whether macroeconomic conditions affect the relationship between
CSR performance and ownership by norm-constrained institutions versus other institutions, we
estimate the following models:
26
(
(
(
(
(
)
)
)
)
)
(
(
(
)
(
)
)
)
(
)
(
(
)
)
(
)
(4)
∑
where Rec is an indicator variable equal to 1 if the period is identified as a recessionary period by
the National Bureau of Economic Research (NBER), and 0 otherwise. All the other variables are
defined as in equation (2).
In equation (4), we consider the changing macroeconomic conditions. If the effectiveness
of shareholder activism on CSR performance is weaker during recessionary periods than during
expansionary periods, we expect the positive association between CSR performance and
holdings of norm-constrained institutions relative to other institutions to be weaker during
recessionary periods (Rec = 1) than during expansionary periods (Rec = 0). That is,
for the current and lagged changes
,
, and
should be negative.
Table 7 reports results from estimation equation (4). For the NC1 measure, coefficients
,
, and
for the interaction between the current and both lagged changes in shareholdings
and the institution type are all significantly positive (
= 5.25,
= 6.14,
= 7.21) during
expansionary periods. This is consistent with the results in Table 5 with all sample periods
pooled together, indicating that CSR performance is positive associated with shareholdings from
norm-constrained institutions. During recessionary periods, however, the differential coefficients
for the interaction between the current and both lagged changes in shareholdings and the
institution type are all significantly negative (
= -4.51,
= -5.84,
= -6.72). While these
differential coefficients do not swamp the corresponding coefficients for expansions, i.e., the
association between shareholdings changes and CSR changes is still positive during recessions,
the decremental effect of recessions on the association is significant in magnitude. For example,
27
for two-year-lagged changes, the magnitude of the association is reduced from 7.21 (
expansions to 0.49 (
-
) for
) for recessions. NC2 produces similar results.
Overall, the results show that the stronger positive association between changes in CSR
performance and current and previous changes in ownership by norm-constrained institutions is
less pronounced in recessionary periods than expansionary periods. This is consistent with the
interpretation that the effectiveness of shareholder activism is affected by economic conditions.
3.
Alternative explanation for results in Table 5
In Table 5, we find evidence that the changes of the shareholdings of such institutions
lead to the changes of firms’ CSR performance, suggesting that norm-constrained institutions
improve the CSR performance of firms they have already invested in. While this is consistent
with the shareholder activism interpretation, an alternative explanation for this finding is that
norm-constrained institutions can somehow predict firms’ future CSR performance, and
accordingly, invest in firms that will have an improved CSR performance in future. Following
this thought, if an institution forecasts a firm’s CSR performance based on the firm’s historic
CSR performance, and subsequently, if the institution forms its investment portfolio based on the
forecast, then we would observe a relation between change in the institutions’ current
shareholdings and changes in the investees’ previous CSR performance. In this subsection, we
more closely examine this alternative explanation.
Similar to equation (2), we specify the following model based on changes:
(
(
)
)
)
)
(5)
∑
where
(
(
denotes a change from year t-1 to year t. All variables are defined as in equations (1)
and (2). We include the concurrent and previous two periods’ changes in CSR performance (i.e.,
28
year t, t-1, and t-2). If improved CSR performance attracts more shareholdings by normconstrained institutions than other institutions, then the coefficients
,
and
on interaction
terms between the ∆CSR and the indicator NC should be positive. Since the change in CSR is
affected by the level of CSR (e.g., firms with poor CSR have more room for improvement) and
since norm-constrained institutions may continue to increase their holdings in firms based on
their historic CSR performance (e.g., buys shares of firms that had good CSR performance in the
past), we also include the CSR level variable at the beginning of the first change period (i.e., the
end of year t-3).
Table 8 reports results from estimating equation (5). The predictive power for the
regressions ranges between 14% and 15%, which is reasonably good considering variables are
measured as changes. The coefficients of interest include
,
, and
which measure the
incremental effect of changes in CSR score in the concurrent and most recent two periods,
respectively, on changes in shares held by norm-constrained institutions compared to other
institutions. The results in Table 8 do not support a relation between the changes of institutions’
current shareholdings and the investees’ pervious change of CSR performance (i.e., insignificant
coefficients on ∆CSRt-1*NC and ∆CSRt-2*NC). Rather, the results show that, after the level of
CSR performance in t-3 and other changes in a firms’ characteristics are controlled for, the
coefficient on the concurrent change ∆CSRt*Type is significant and positive for NC1 (
=
0.0007, p-value = 0.06) and NC2 ( =0.0074, p-value < 0.001). The coefficients on the lagged
changes are all insignificant.
In short, the findings from Table 8 help us rule out the alternative explanations for the
findings in Table 5 by showing evidence of a positive association between the contemporaneous
changes in CSR performance and holdings by norm-constrained institutions, but an insignificant
29
effect of the lagged changes in CSR on changes in ownership by norm-constrained institutions.17
4.
Alternative measure for institutional ownership in changes analyses
In the shareholder activism analyses, we use voting share ownership IOV in equation (2)
because voting shares serve our purpose of testing institutions’ monitoring role. On the other
hand, we use common shareholdings IO in equation (5) to examine whether institutional
shareholdings change in response to previous changes in CSR performance. While different
empirical goals require IO or IOV in different estimating models, it may raise concerns that the
different results on the lagged changes variables from Table 5 and Table 8 are possibly caused by
using different institutional ownership variables.
To address this issue, we replace IOV with IO in equation (2) and replace IO with IOV in
equation (5). Untabulated results show our findings still hold. For example, when CSR1 and NC1
are used in equation (2), while all three coefficients on the interaction terms between changes in
shareholdings and NC are smaller and less significant than those reported in Table 5, they still
remain positive and statistically significant. Using IOV in equation (5) only produces a
significant and positive coefficient on the concurrent change ∆CSRt*NC. The coefficients on the
lagged changes are both insignificant, similar to the results reported in Table 8 when IO is used.
Thus, our findings from the changes analyses are robust whether IO or IOV is used.
5.
Non-zero changes in testing H2
To rule out possibilities that results in Table 5 may be affected by observations with zero
changes in CSR performance, we estimate equation (2) again while requiring the dependent
variable
to be non-zero. The untabulated results show no statistically significant
difference from those reported in Table 5. We also require non-zero changes in institutional
17
We cannot completely rule out the possibility that norm-constrained investors quickly identifying and investing in
firms that have improved their CSR performance.
30
ownership (
) when re-estimating equation (5). The results are qualitatively similar to those
presented in Table 8.
6.
Industry-adjusted CSR measures
We replace our CSR measures with industry-adjusted CSR scores in the regression tests.
Untabulated results show that findings from estimating equation (1) are weaker. However, results
from equation (2) are qualitatively similar to those reported above. Thus, the evidence of positive
screening and shareholder activism from norm-constrained institutional investors is robust
whether CSR scores are industry adjusted or not.
V.
Conclusion
In this study, we examine how the concept of social norms affects institutional
shareholders’ investment behavior in relation to CSR performance. We use two definitions of
norm-constrained institutional investors – one based on a refined measure derived from Hong
and Kacperczyk (2009) and another that further incorporates the CSR attitudes that prevail in the
geographic location of the institution. Using levels tests, we find that both definitions produce
evidence of a positive association between norm-constrained institutions and their holdings in
firms with higher CSR performances relative to other institutions. This is consistent with positive
screening, and we show that this effect is distinct from the negative screening that Hong and
Kacperczyk (2009) identify.
We further explore norm-constrained institutions’ monitoring role in driving their
invested firms’ CSR performance. In particular, we regress the current change in CSR
performance on the one-year lagged and two-year lagged changes in holdings of normconstrained institutions. We find that the current change in CSR performance is more positively
related to both lagged and current changes of norm-constrained institutions’ ownership. This
31
evidence suggests that firms improve their CSR performance because of encouragement or
pressure from their norm-constrained institutional shareholders, consistent with these institutions
taking an activist role.
In further analyses, we show that norm-constrained institutional investors with long-term
and substantial holdings exhibit stronger influence on firms’ CSR performance than other normconstrained institutions, or other long-term and substantial institutional investors. In addition, we
document that the activist role of norm-constrained institutions is less effective in recession
periods rather than expansion periods. Lastly, we rule out alternative explanation of our
empirical findings that norm-constrained institutions can predict future improvements in CSR by
documenting that past changes in CSR performance are unrelated to current changes in the
shareholdings of norm-constrained institutions.
Our study complements Hong and Kacperczyk (2009) as they focus on negative
screening while we focus on positive screening and institutional activism. Our results suggest
that social norms can lead norm-constrained institutions to screen their investments for good
CSR performance and to take an activist role to promote improvements in CSR. Thus, our results
show that social norms have a much wider influence on norm-constrained institutions than the
negative screening role documented by Hong and Kacperczyk (2009).
32
References
Agrawal, A., and Mandelker, G. 1990. Large shareholders and the monitoring of managers: the
case of antitakeover charter amendments. Journal of Financial and Quantitative Analysis 25,
143-161.
Almazan, A., Hartzell, J., and Starks, L. 2005. Active institutional shareholders and costs of
monitoring: evidence from executive compensation. Financial Management 34, 5-34.
Borokhovich, K., Brunarski, k., Harman, Y., and Parrino, R. 2006. Variation in the monitoring
incentives of outside stockholders. Journal of Law and Economics 49, 651-680.
Brickley, J., Lease, R., and Smith, C. 1988. Ownership structure and voting on antitakeover
amendments. Journal of Financial Economics 20: 267-291.
Brown, L., and Caylor, M. 2006. Corporate governance and firm valuation. Journal of
Accounting and Public Policy 25, 409-434.
Bushee, B. 1998. The influence of institutional investors in myopic R&D investment behavior.
The Accounting Review 73, 305-333.
Bushee, B. 2001. Do institutional investors prefer near-term earnings over long-run value?
Contemporary Accounting Research 18, 207 - 246
Bushee, B., and Noe, C. 2000. Corporate disclosure practices, institutional investors, and stock
return volatility. Journal of Accounting Research 38, 171-202.
Campbell, J. 2006. Household finance, The Journal of Finance 61, 1553-1604.
Chava, S. 2010. Socially responsible investing and expected stock returns. Working paper,
Georgia Institute of Technology.
Chen, X., Harford, J., and Li, K. 2007. Monitoring: Which institutions matter? Journal of
Financial Economics 86, 279-305.
33
Coval, J., and Moskowitz, T. 1999. Home bias at home: local equity preference in domestic
portfolios. The Journal of Finance 54, 1249-1290
Del Guercio, D. 1996. The distorting effect of the prudent-man laws on institutional equity
investments. Journal of Financial Economics 40, 31-62.
Dhaliwal, D., Li, O., Tsang, A., and Yang, Y. 2011. Voluntary nonfinancial disclosure and the
cost of equity capital: the initiation of corporate social responsibility reporting. The
Accounting Review 86, 59-100.
Di Giuli, A., and Kostovetsky, L. 2012. Are red or blue companies more likely to go green?
Politics and corporate social responsibility. Working paper, University of Rochester.
Fernando, C., Sharfman. M, and Vahap, U. 2010. Does greenness matter? The effect of corporate
environmental performance on ownership structure, analyst coverage, and firm value.
Working paper, University of Oklahoma.
Gompers, P., Ishii, J., and Metrick, A. 2003. Corporate governance and equity prices. The
Quarterly Journal of Economics 118, 107-156.
Halek, M., and Eisenhauer, J. 2001. Demography of risk aversion. Journal of Risk and Insurance
68, 1-24.
Harjoto, M., Jo, H., and Kim, Y. 2012. Is institutional ownership related to corporate social
responsibility? The nonlinear relation and its implication for stock return volatility. Working
paper, Pepperdine University.
Hartzell, J., and Starks, L. 2003. Institutional investors and executive compensation. The Journal
of Finance 58, 2351-2374.
Hilary, G., and Hui, K. 2009. Does religion matter in corporate decision making in America?
Journal of Financial Economics 93, 455-473.
34
Hong, H., and Kacperczyk, M. 2009. The price of sin: the effects of social norms on markets.
Journal of Financial Economics 93, 15-36.
Hong, H., and Kostovetsky, L. 2012. Red and blue investing: values and finance. Journal of
Financial Economics 103, 1-19.
Hong, H., Kubik, J., and Scheinkman, J. 2012. Financial constraints on corporate goodness.
Working paper, Princeton University.
Hylton, M. 1992. "Socially responsible" investing: doing good versus doing well in an inefficient
market. The American University Law Review 42, 1-52.
Kang, J.-K., and Kim, J.-M. 2008. The geography of block acquisitions. Journal of Finance 63,
2817-2858.
Kim, Y., Park, M., and Wier, B. 2012. Is earnings quality associated with corporate social
responsibility? The Accounting Review 87, 761-796.
Larcker, D., Richardson, S., and Tuna, I. 2007. Corporate governance, accounting outcomes, and
organizational performance. The Accounting Review 82, 963-1008.
Malloy, C. 2005. The geography of equity analysis. Journal of Finance 60, 719-755.
Pirinsky, C., and Wang, Q. 2006. Does corporate headquarters location matter for stock returns?
Journal of Finance 61, 1991-2015.
Rubin, A. 2008. Political views and corporate decision making: The case of corporate social
responsibility. The Financial Review 43, 337-360.
Useem, M. 1996. Investor capitalism: How money managers are changing the face of corporate
America. New York: Basic Books.
35
Appendix Variable Definitions
Main variables
IO = Institutional ownership by norm-constrained and other institutions at the end of year t,
calculated as a percentage of a firm’s shares outstanding.
IOV = Institutional ownership with voting rights held by socially responsible or non-normconstrained institutions at the end of year t, calculated as a percentage of a firm’s
shares outstanding.
CSR = Total CSR score for a firm at the end of year t. There are two measures under this
variable. CSR1 is the net score (strengths minus concerns) of community, diversity,
employee relations, environment, human rights, and product. CSR2 is similar except
excludes the net score of product ratings.
NC = Indicator variable that equals 1 if the institution is classified as norm-constrained.
There are three types of classification. NC1 equals 1 if the institution is a pension
fund, university, or religious, charitable or non-profit organization, and 0 if the
institution is a bank, insurance company, mutual fund or investment advisor. NC2
equals 1 if the institution’s headquarter is located in a norm-constrained area and the
institution is a pension fund, university, or religious, charitable or non-profit
organization, and 0 if the institution’s headquarter is located in a non-normconstrained area and the institution is a pension fund, university, or religious,
charitable or non-profit organization.
Rec = Indicator variable equal to 1 if the period is identified as a recessionary period by
NBER, and 0 otherwise.
Controls variables
CG = Net score of corporate governance ratings for a firm at the end of year t.
36
LMV = Natural log of the market value of equity.
SP500 = Indicator variable being 1 if a firm is in the S&P500 index, and 0 otherwise.
Lev = Ratio of debt to total assets.
EP = Ratio of income before extraordinary items to the market value of equity.
ROE = Ratio of income before extraordinary items to the book value of equity.
DP = Ratio of dividends to the market value of equity.
BP = Ratio of the book value of equity to the market value of equity.
TV = Average monthly trading volume relative to total shares outstanding measured over
the year.
SaleChg = Percentage change in sales.
AnnRet = Market adjusted buy-and-hold stock returns measured over the year.
Rating = Measure of the S&P stock rating (9 = A+, …, 1 = not rated).
Beta = Market model beta calculated from daily stock returns measured over the year.
IRisk = Standard deviations of the market-model residuals of daily stock returns measured
over the year.
37
Table 1 Descriptive statistics for key variables
Variable
Mean
STD
Q1
Median
Q3
Institutional ownership variables: NC1
IO_total
0.6577
0.2183
0.5091
0.6935
0.8338
IO_NC1
0.0375
0.0226
0.0206
0.0367
0.0483
-0.0001
0.0149
-0.0063
0.0002
0.0060
IO_Non-NC1
0.6306
0.1919
0.5005
0.6600
0.7860
∆IO_Non-NC1
0.0201
0.0682
-0.0241
0.0167
0.0621
IOV_total
0.4060
0.2213
0.2642
0.4565
0.5774
IOV_NC1
0.0243
0.0180
0.0115
0.0228
0.0346
∆IOV_NC1
0.0012
0.0125
-0.0044
0.0000
0.0065
IOV_Non-NC1
0.3691
0.2121
0.2338
0.4269
0.5357
∆IOV_Non-NC1
0.0165
0.0677
-0.0195
0.0022
0.0515
∆IO_NC1
Institutional ownership variables: NC2
IO_total
0.4619
0.2241
0.3426
0.4706
0.6230
IO_NC2
0.0478
0.0923
0.0124
0.0415
0.0632
∆IO_NC2
0.0093
0.0537
-0.0076
0.0078
0.0119
IO_Non-NC2
0.0482
0.0955
0.0140
0.0446
0.0615
∆IO_Non-NC2
0.0071
0.0572
-0.0106
0.0049
0.0134
IOV_total
0.3164
0.1897
0.1782
0.3254
0.4493
IOV_NC2
0.0449
0.0879
0.0119
0.0411
0.0619
∆IOV_NC2
0.0091
0.0476
-0.0069
0.0073
0.0115
IOV_Non-NC2
0.0455
0.0836
0.0119
0.0397
0.0604
∆IOV_Non-NC2
0.0055
0.0488
-0.0097
0.0037
0.0127
CSR1
-0.2063
1.8809
-1.0000
0.0000
1.0000
∆CSR1
-0.0161
1.0747
0.0000
0.0000
0.0000
Str1
1.1508
1.7157
0.0000
0.0000
2.0000
Con1
1.3571
1.3990
0.0000
1.0000
2.0000
CSR2
0.0754
1.8129
-1.0000
0.0000
1.0000
-0.0126
1.0111
0.0000
0.0000
0.0000
Str2
1.0812
1.6282
0.0000
0.0000
2.0000
Con2
1.1566
1.1899
0.0000
1.0000
2.0000
CSR performance variables
∆CSR2
Control variables
CG
-0.2417
0.6899
-1.0000
0.0000
0.0000
∆CG
-0.1123
0.5727
0.0000
0.0000
0.0000
38
Variable
Mean
LMV
7.2294
1.4474
6.1244
7.1046
8.1761
∆LMV
0.0474
0.4137
-0.1647
0.0690
0.2834
SP500
0.2681
0.4430
0.0000
0.0000
1.0000
∆SP500
0.0054
0.1086
0.0000
0.0000
0.0000
Lev
0.2111
0.1878
0.0422
0.1836
0.3261
-0.0001
0.0742
-0.0257
-0.0005
0.0157
EP
0.0238
0.1138
0.0201
0.0461
0.0665
∆EP
0.0070
0.2557
-0.0161
0.0018
0.0204
ROE
0.0744
0.2592
0.0385
0.1041
0.1633
∆ROE
0.0577
5.9158
-0.0421
0.0002
0.0351
DP
0.0147
0.0207
0.0000
0.0062
0.0229
-0.0012
0.0369
-0.0007
0.0000
0.0015
BP
0.5352
0.3479
0.2932
0.4677
0.6954
∆BP
0.0191
0.5262
-0.0730
0.0074
0.1011
TV
0.1665
0.1310
0.0719
0.1293
0.2197
∆TV
0.0051
0.0821
-0.0204
0.0047
0.0333
SaleChg
0.1094
0.2264
0.0000
0.0674
0.1774
-0.2026
25.3191
-0.0966
0.0000
0.0751
0.0490
0.3580
-0.1778
0.0002
0.2128
-0.0523
0.6681
-0.3075
-0.0246
0.2403
Beta
1.2870
0.5397
0.9016
1.2374
1.6125
∆Beta
0.0044
0.4702
-0.2749
0.0107
0.2859
IRisk
0.0224
0.0100
0.0151
0.0202
0.0274
-0.0010
0.0094
-0.0050
-0.0008
0.0029
∆Lev
∆DP
∆SaleChg
AnnRet
∆AnnRet
∆IRisk
STD
Q1
Median
Q3
This table provides descriptive statistics for the institutional ownership, CSR performance, and control
variables. ‘_NC’ indicates the subsample of norm-constrained institutions. ‘_Non-NC’ indicates the subsample
of other institutions. See Appendix for variable definitions.
39
Table 2 Correlation matrix of main and control variables
Variable
IO_ total
IOV_ total
CSR1
CSR2
CG
LMV
SP500
Lev
EP
ROE
DP
BP
TV
SaleChg
AnnRet
Beta
IRisk
IO_ total
1.000
.
0.692
0.000
0.005
0.514
0.024
0.000
-0.242
0.000
0.230
0.000
0.085
0.000
0.067
0.000
0.029
0.000
0.057
0.000
-0.183
0.000
-0.042
0.000
0.401
0.000
0.007
0.280
0.042
0.000
0.051
0.000
-0.079
0.000
0.788
0.000
1.000
.
-0.124
0.000
-0.105
0.000
-0.147
0.000
-0.037
0.000
-0.258
0.000
0.010
0.145
-0.027
0.000
-0.026
0.000
-0.173
0.000
0.053
0.000
0.404
0.000
0.030
0.000
0.052
0.000
0.162
0.000
0.054
0.000
-0.014
0.041
-0.109
0.000
1.000
.
0.956
0.000
0.006
0.360
0.250
0.000
0.259
0.000
-0.030
0.000
0.037
0.000
0.072
0.000
0.029
0.000
-0.094
0.000
-0.007
0.288
-0.041
0.000
-0.023
0.001
-0.101
0.000
-0.069
0.000
0.004
0.613
-0.096
0.000
0.955
0.000
1.000
.
-0.033
0.000
0.339
0.000
0.331
0.000
-0.004
0.526
0.048
0.000
0.086
0.000
0.050
0.000
-0.095
0.000
-0.006
0.407
-0.061
0.000
-0.027
0.000
-0.135
0.000
-0.100
0.000
-0.234
0.000
-0.162
0.000
0.022
0.001
-0.013
0.054
1.000
.
-0.310
0.000
-0.230
0.000
-0.058
0.000
0.010
0.147
-0.021
0.002
0.058
0.000
0.021
0.002
-0.183
0.000
0.039
0.000
0.031
0.000
0.002
0.797
0.066
0.000
0.227
0.000
0.001
0.834
0.226
0.000
0.297
0.000
-0.312
0.000
1.000
.
0.700
0.000
0.141
0.000
0.214
0.000
0.265
0.000
0.069
0.000
-0.329
0.000
0.096
0.000
-0.001
0.862
0.097
0.000
-0.216
0.000
-0.412
0.000
0.047
0.000
-0.210
0.000
0.236
0.000
0.300
0.000
-0.229
0.000
0.677
0.000
1.000
.
0.095
0.000
0.093
0.000
0.148
0.000
0.102
0.000
-0.153
0.000
-0.009
0.186
-0.098
0.000
-0.041
0.000
-0.218
0.000
-0.219
0.000
0.054
0.000
-0.016
0.021
-0.012
0.089
0.017
0.013
-0.058
0.000
0.199
0.000
0.149
0.000
1.000
.
-0.029
0.000
-0.032
0.000
0.300
0.000
0.028
0.000
-0.039
0.000
-0.041
0.000
-0.021
0.002
-0.111
0.000
-0.133
0.000
-0.020
0.004
-0.052
0.000
0.018
0.008
0.037
0.000
-0.006
0.417
0.169
0.000
0.116
0.000
0.082
0.000
1.000
.
0.651
0.000
0.087
0.000
-0.169
0.000
-0.092
0.000
0.073
0.000
0.122
0.000
-0.183
0.000
-0.358
0.000
0.029
0.000
-0.064
0.000
0.105
0.000
0.126
0.000
-0.044
0.000
0.378
0.000
0.238
0.000
0.000
0.998
0.664
0.000
1.000
.
0.058
0.000
-0.191
0.000
-0.058
0.000
0.047
0.000
0.111
0.000
-0.185
0.000
-0.305
0.000
-0.217
0.000
-0.259
0.000
0.086
0.000
0.108
0.000
0.061
0.000
0.192
0.000
0.228
0.000
0.285
0.000
0.274
0.000
0.118
0.000
1.000
.
0.216
0.000
-0.199
0.000
-0.133
0.000
-0.097
0.000
-0.219
0.000
-0.246
0.000
-0.044
0.000
0.038
0.000
-0.099
0.000
-0.105
0.000
0.040
0.000
-0.327
0.000
-0.167
0.000
0.078
0.000
0.126
0.000
-0.483
0.000
0.186
0.000
1.000
.
-0.033
0.000
-0.171
0.000
-0.258
0.000
0.058
0.000
0.153
0.000
0.519
0.000
0.531
0.000
-0.032
0.000
-0.022
0.001
-0.211
0.000
0.139
0.000
-0.024
0.000
-0.055
0.000
-0.099
0.000
-0.037
0.000
-0.339
0.000
-0.076
0.000
1.000
.
0.094
0.000
0.027
0.000
0.340
0.000
0.335
0.000
0.042
0.000
0.041
0.000
-0.034
0.000
-0.051
0.000
0.027
0.000
0.034
0.000
-0.095
0.000
-0.081
0.000
0.061
0.000
0.185
0.000
-0.201
0.000
-0.227
0.000
0.084
0.000
1.000
.
0.101
0.000
0.094
0.000
0.034
0.000
0.056
0.000
0.048
0.000
-0.016
0.023
-0.015
0.030
0.019
0.006
0.146
0.000
-0.018
0.009
-0.007
0.303
0.024
0.001
0.176
0.000
-0.049
0.000
-0.264
0.000
0.000
0.987
0.116
0.000
1.000
.
0.031
0.000
0.054
0.000
0.068
0.000
0.158
0.000
-0.113
0.000
-0.146
0.000
-0.002
0.774
-0.238
0.000
-0.237
0.000
-0.150
0.000
-0.206
0.000
-0.210
0.000
-0.286
0.000
0.045
0.000
0.336
0.000
0.080
0.000
-0.022
0.001
1.000
.
0.435
0.000
-0.032
0.000
0.057
0.000
-0.079
0.000
-0.112
0.000
0.070
0.000
-0.442
0.000
-0.242
0.000
-0.206
0.000
-0.312
0.000
-0.295
0.000
-0.426
0.000
0.053
0.000
0.333
0.000
-0.006
0.368
-0.028
0.000
0.452
0.000
1.000
.
IOV_total
CSR1
CSR2
CG
LMV
SP500
Lev
EP
ROE
DP
BP
TV
SaleChg
AnnRet
Beta
IRisk
This table reports the correlation matrix for the main and control variables with Pearson correlations below the diagonal and Spearman correlations above the diagonal.
Correlation coefficients in bold are significant at the 5% level. See Appendix for variable definitions.
40
Table 3 Level analysis of association between institutional ownership and CSR performance
(
Coefficient
(t-statistic)
Exp.
Sign
)
∑
(1)
Institution = NC1
Institution = NC2
CSR1
CSR2
0.5231***
0.5237***
CSR1
CSR2
Variable
Intercept
CSR
NC
(17.21)
-0.0030**
(17.28)
-0.0032**
(-2.39)
(-2.48)
-0.5816***
(-174.65)
CSR*NC
CG
+
SP500
Lev
EP
ROE
(11.25)
0.0012***
(12.44)
0.0009***
(3.02)
-0.5822***
(-175.30)
0.0106***
(9.12)
(2.89)
0.0133***
(7.69)
0.0057***
0.0014***
0.0017***
(3.85)
-0.0424***
(3.96)
-0.0428***
(3.26)
0.0024***
(4.45)
0.0028***
(5.26)
(6.16)
0.0587***
(17.75)
LMV
0.0307***
0.0054***
(-14.29)
CG*NC
0.0313***
0.0022*
(-14.46)
0.0593***
(18.02)
0.0056***
(6.09)
0.0022
0.0042***
(7.33)
0.0002*
0.0003*
(1.66)
0.0135***
(1.64)
0.0135***
(1.81)
0.0015**
(1.93)
0.0013**
(3.66)
(3.66)
(2.33)
(2.09)
0.0400***
0.0399***
0.0009
(0.66)
0.0007
(6.26)
(6.26)
-0.0007
-0.0007
0.0004
0.0006
(-0.09)
0.0010
(-0.09)
0.0010
(0.31)
0.0011
(0.29)
0.0016
(0.25)
(0.24)
(1.33)
(1.18)
41
(0.48)
Coefficient
(t-statistic)
DP
Exp.
Sign
Institution = NC1
CSR1
-0.6828***
(-11.39)
BP
0.0243***
(7.42)
TV
SaleChg
AnnRet
0.1937***
Beta
IRisk
CSR2
-0.6832***
(-11.39)
0.0243***
(7.38)
0.1936***
CSR1
CSR2
-0.1124**
-0.1285**
(-2.15)
0.0010**
(2.45)
0.0244***
(21.38)
-0.0055*
(21.35)
-0.0054*
(9.39)
0.0016
(-1.78)
(-1.77)
(1.23)
0.0106***
(6.30)
Rating
Institution = NC2
0.0105***
(6.29)
0.0029***
(3.25)
(-2.05)
0.0014**
(2.36)
0.0265***
(10.22)
0.0013
(1.08)
0.0033***
(4.28)
0.0014**
0.0014**
0.0006
0.0005
(2.43)
0.0038**
(2.42)
0.0037**
(0.78)
0.0019
(0.63)
0.0021
(2.10)
(2.08)
(1.19)
(1.25)
-2.2531***
-2.2519***
-0.6752***
-0.7305***
(-16.66)
(-16.67)
(-11.19)
(-8.26)
Industry Fixed
Effects
Included
Included
Included
Included
Year Fixed Effects
Included
Included
Included
Included
Adj_R-Sqr
0.884
0.884
0.375
0.375
No_Obs
38170
38170
24119
24119
This table reports coefficients and t-statistics (in parentheses) from firm cluster, industry and year fixed effect, and heteroscedasticity adjusted regressions of institutional
ownership on CSR performance, institutional investor type, two-way interaction terms between CSR performance and investor type, and control variables. Results regarding
industry and year intercepts are not reported. See Appendix for variable definitions. ***, **, and * denote significance at the 1%, 5%, and 10% levels in a two-tailed test,
respectively.
42
Table 4 Level analysis of association between institutional ownership and CSR performance
separating CSR strengths and concerns
(
)
(
)
(3)
∑
Coefficient
(t-statistic)
Exp.
Sign
Variable
Intercept
Institution = NC1
0.5083***
(16.76)
Str
-0.0046***
(-3.29)
Con
0.0005
(0.29)
-0.5811***
NC
(-135.26)
Str*NC
Con*NC
CG
+
-
0.0052***
SP500
Lev
ROE
DP
(5.32)
-0.0025***
(-4.37)
0.0021***
(3.06)
0.0018**
-0.0055***
-0.0021**
(-2.94)
-0.0421***
(-2.07)
0.0012**
0.0587***
(2.41)
0.0028***
(3.59)
0.0039***
0.0041***
(2.88)
0.0164***
(2.89)
0.0056***
(4.36)
(2.95)
0.0410***
(6.47)
EP
0.0029***
(2.26)
(17.71)
LMV
0.0321***
(15.48)
(3.30)
(-14.11)
CG*NC
Institution = NC2
0.0119
(1.25)
-0.0003
0.0009
(-0.04)
0.0005
(0.16)
0.0017
(0.12)
(0.78)
-0.6642***
-0.4169***
(-11.06)
(-6.42)
BP
0.0259***
0.0088***
TV
(7.93)
0.1917***
(4.26)
0.1538***
(21.19)
SaleChg
-0.0066**
(-2.15)
AnnRet
0.0101***
(6.08)
43
(8.48)
0.0011
(0.44)
0.0075***
(6.16)
Coefficient
(t-statistic)
Rating
Exp.
Sign
Institution = NC1
0.0014**
(2.43)
Beta
0.0035**
IRisk
(1.97)
-2.1975***
(-16.29)
Industry Fixed
Effects
Year Fixed Effects
Adj_R-Sqr
Included
Included
0.884
Institution = NC2
0.0007*
(1.68)
-0.0026*
(-1.79)
-0.8926***
(-9.24)
Included
Included
0.377
No_Obs
38170
24119
This table reports coefficients and t-statistics (in parentheses) from firm cluster, industry and year fixed effect, and
heteroscedasticity adjusted regressions of institutional ownership on CSR strengths and concerns, institutional
investor type, two-way interaction terms between CSR strengths/concerns and investor type, and control variables.
Results regarding control variables, industry and year intercepts are not reported. See Appendix for variable
definitions. ***, **, and * denote significance at the 1%, 5%, and 10% levels in a two-tailed test, respectively.
44
Table 5 Change analysis of increased institutional ownership with voting rights leading to
improved CSR performance
(
(
)
)
(
(
)
)
(2)
∑
Coefficient
(t-statistic)
Exp.
Sign
Institution = NC1
Institution = NC2
Variable
Intercept
-0.1929**
(-2.00)
∆IOVt
0.1620
NC
∆IOVt *NC
+
(2.19)
0.9019***
(5.64)
0.0942*
(1.89)
4.5949***
-0.0463
(-0.22)
∆IOVt-1 *NC
+
∆IOVt-2
5.0842***
(5.22)
0.2465
(1.30)
∆IOVt-2 *NC
IOVt-3
IOVt-3 *NC
CSRt-1
+
5.5900***
∆CGt
(2.26)
0.8891**
(1.99)
1.7175***
(2.81)
1.0651**
(2.43)
1.8816**
(2.07)
0.0062
(0.05)
1.5542***
(3.39)
6.0269***
1.0716*
(8.31)
(1.76)
-0.0593***
-0.1176**
0.0026
(0.20)
0.1818***
(8.12)
∆SP500
2.1154**
(5.99)
(-10.85)
CGt-1
1.7544**
(0.73)
(4.72)
∆IOVt-1
0.3325
(0.41)
-0.0485
(-1.34)
(-2.49)
0.0541
(0.80)
0.1517***
(7.36)
0.0645
(1.08)
∆Lev
0.0103
0.0991
∆BP
(0.12)
0.1617
(0.30)
0.0017*
(1.03)
(1.80)
∆ROE
-0.0358
(-0.71)
45
0.0014*
(1.67)
Coefficient
(t-statistic)
∆SaleChg
Exp.
Sign
Institution = NC1
-0.0172
(-0.40)
Industry Fixed Effects
Year Fixed Effects
Adj_R-Sqr
Institution = NC2
0.0000
(0.05)
Included
Included
Included
0.049
Included
0.033
No_Obs
20723
13119
This table reports coefficients and t-statistics (in parentheses) from firm cluster, industry and year fixed effect, and
heteroscedasticity adjusted regressions of change in CSR performance on change in institutional ownership with
voting rights, institutional investor type, two-way interaction terms between change in voting shares and investor
type, and control variables. Results regarding industry and year intercepts are not reported. See Appendix for variable
definitions. ***, **, and * denote significance at the 1%, 5%, and 10% levels in a two-tailed test, respectively.
46
Table 6 Change analysis of increased institutional ownership with voting rights leading to
improved CSR performance: Long-term and substantial holdings
(1)
Coefficient
(t-statistic)
Exp.
Sign
Intercept
(2)
Norm-constrained:
Long-Term & Substantial
versus Others
0.0714
0.4771
(0.91)
∆IOVt
(1.44)
3.7306***
0.778
(3.88)
NC
∆IOVt *NC
(1.29)
0.0695**
+
Long-Term & Substantial:
Norm-constrained
versus Others
(2.37)
6.8231**
-0.0043
(-0.05)
9.887***
(4.51)
(-0.56)
3.1775**
(2.01)
∆IOVt-1
0.5592
(0.65)
+
4.432**
(2.05)
2.5519**
(2.18)
-0.8854
∆IOVt-2
(-1.16)
∆IOVt-2 *NC
Norm-constrained
and Long-Term &
Substantial
-0.0259
(2.01)
∆IOVt-1 *NC
(3)
+
6.569***
(2.75)
3.2776***
(2.99)
IOVt-3
0.4779
(0.76)
4.2558***
IOVt-3 *NC
8.455***
(4.13))
(4.64)
IOVt-1
6.3055***
(10.27)
IOVt-1 *NC
CSRt-1
1.0205
(0.53)
-0.0748***
-0.0575**
(-14.69)
CGt-1
(-2.40)
-0.0013
0.0010
(-1.14)
∆CGt
∆SP500
∆Lev
(0.09)
0.1410***
0.1338**
∆ROE
(-4.64)
-0.0086
(-0.18)
0.2166***
(7.96)
0.0000
(2.15)
0.1518
(3.18)
0.1195
(0.00)
(0.50)
(0.56)
-0.6648
-0.4914
0.1488
(1.35)
∆BP
-0.0678***
(-1.17)
0.0538**
0.2779**
(-1.08)
0.4026**
(2.03)
0.0063
(2.09)
0.0038
(2.54)
0.1018
(0.29)
(0.31)
(1.42)
47
Coefficient
(t-statistic)
∆SaleChg
Exp.
Sign
(1)
(2)
(3)
Norm-constrained:
Long-Term & Substantial
versus Others
Long-Term & Substantial:
Norm-constrained
versus Others
Norm-constrained
and Long-Term &
Substantial
-0.0256
0.0889
(-0.89)
Industry Fixed
Effects
Year Fixed
Effects
Adj_R-Sqr
0.1720***
(0.94)
(1.45)
Included
Included
Included
Included
0.049
Included
0.074
Included
0.127
No_Obs
22031
15338
2225
This table reports coefficients and t-statistics (in parentheses) from firm cluster, industry and year fixed effect, and
heteroscedasticity adjusted. Results regarding industry and year intercepts are not reported. See Appendix for
definitions of other variable in equation (2) except for NC. In Column 1, NC is an indicator variable that equals 1 if
an institution is a pension fund, university, or religious, charitable or non-for-profit organizations and has longterm and substantial holdings in one particular firm, and 0 if an institution is a pension fund, university, or
religious, charitable or non-for-profit organizations but does not have long-term and substantial holdings. In
Column 2, NC is an indicator variable that equals 1 if an institution is a pension fund, university, or religious,
charitable or non-for-profit organizations and has long-term and substantial holdings in one particular firm, and 0 if
an institution is a bank, an insurance company, a mutual fund or an independent investment manager and has longterm and substantial holdings in one particular firm. In Column 3, the sample is restricted to pension funds,
universities, religious, charitable and non-for-profit organizations that have long-term and substantial holdings in
one particular firm. ***, **, and * denote significance at the 1%, 5%, and 10% levels in a two-tailed test,
respectively.
48
Table 7 Change analysis of increased institutional ownership with voting rights leading to
improved CSR performance: Macroeconomic conditions
(
(
(
(
Coefficient
(t-statistic)
)
)
)
)
)
Exp.
Sign
(
(
(
(
)
(
(
)
Institution = NC1
)
)
)
(
(
)
(
)
∑
Institution = NC2
Variable
Intercept
0.1355
(0.97)
∆IOVt
0.0936
NC
Rec
∆IOVt *NC
∆IOVt*Rec
(1.68)
-0.2546
-0.0877*
(4.56)
(2.40)
-4.5120**
(-0.23)
∆IOVt-1 *NC
+
6.1446***
(5.50)
∆IOVt-1*Rec
∆IOVt -1*NC*Rec
0.0911
-
(0.20)
-5.8431***
(-3.13)
∆IOVt-2
0.3042
(1.34)
∆IOVt-2 *NC
+
7.2145***
(7.05)
∆IOVt-2*Rec
∆IOVt-2 *NC*Rec
-0.2073
(-0.53)
-
0.0015
(1.06)
2.4643**
(-2.47)
-0.0581
∆IOVt-1
(-1.70)
(0.15)
5.2504***
0.2541
-
1.8089**
(4.93)
(0.54)
∆IOVt *NC*Rec
(0.28)
(2.07)
0.0940*
0.0248
+
0.2791
(0.33)
0.8484***
(-1.40)
NC*Rec
-6.7220***
(-3.91)
49
)
-0.1271
(-0.32)
-0.8895**
(-2.47)
0.9952*
(1.84)
2.0375**
(2.29)
-0.2092*
(-1.73)
-0.4877**
(-2.33)
1.2086**
(2.54)
2.2199**
(2.12)
-0.2506*
(-1.88)
-0.7245***
(-3.07)
(4)
Coefficient
(t-statistic)
Exp.
Sign
IOVt-3
Institution = NC1
0.0913
IOVt-3 *NC
∆IOVt-3*Rec
(4.24)
1.4962**
(9.88)
(2.47)
-0.3738
CSRt-1
-
-0.9845***
(-6.06)
-0.0593***
(-3.51)
-0.0721***
-0.0010
(-0.08)
∆CGt
∆SP500
∆Lev
0.1773***
∆ROE
Year Fixed Effects
Adj_R-Sqr
(-0.47)
0.1448***
(0.15)
(0.35)
0.1557
-0.0398
(-0.79)
0.0411
-0.0162
(-0.38)
Industry Fixed Effects
-0.0233
(6.69)
0.0308
(1.25)
∆SaleChg
(-8.42)
(7.93)
0.0130
(0.99)
∆BP
-0.4649**
(-2.04)
-7.4194***
(-10.81)
CGt-1
1.7690***
(0.60)
7.7397***
(-1.63)
∆IOVt-3 *NC*Rec
Institution = NC2
0.1229
(0.67)
0.0017*
(1.80)
0.0275
(0.73)
-0.0029
(-0.17)
Included
Included
Included
0.049
Included
0.034
No_Obs
20723
13119
This table reports coefficients and t-statistics (in parentheses) from firm cluster, industry and year fixed effect, and
heteroscedasticity adjusted regressions of change in CSR performance on change in institutional ownership with
voting rights, institutional investor type, two-way interaction terms between change in voting shares and investor
type, and control variables. Results regarding industry and year intercepts are not reported. See Appendix for variable
definitions. ***, **, and * denote significance at the 1%, 5%, and 10% levels in a two-tailed test, respectively.
50
Table 8 Change analysis of institutional shareholdings in response to lagged CSR
performance
(
(
)
)
(
(
)
)
(5)
∑
Coefficient
(t-statistic)
Exp.
Sign
Variable
Intercept
∆CSRt
NC
∆CSRt*NC
+
∆CSRt-1
∆CSRt-1*NC
+
∆CSRt-2
∆CSRt-2*NC
CSRt-3
CSRt-3*NC
∆CGt
∆CGt*NC
∆CGt-1
∆CGt-1*NC
∆CGt-2
∆CGt-2*NC
CGt-3
+
Institution = NC1
Institution = NC2
0.0054
(0.86)
-0.0011**
(-2.00)
-0.0186***
(-25.55)
0.0007
(1.72)
-0.0007
(-1.28)
0.0003
(0.57)
-0.0002
(-0.33)
-0.0002
(-0.26)
0.0006**
(1.97)
-0.0010***
(-3.36)
0.0003
(0.28)
0.0003
(0.24)
0.0028**
(2.19)
-0.0033**
(-2.48)
0.0038***
(2.61)
-0.0035**
(-2.32)
0.0053***
(5.49)
-0.0082
(-0.60)
0.0106***
(5.19)
0.0298***
(7.05)
0.0074***
(2.61)
0.0011
(1.14)
-0.0075
(-1.24)
0.0026
(1.07)
0.0031
(0.78)
0.0004*
(1.72)
-0.0001
(-0.42)
0.0022**
(2.43)
0.0008
(0.14)
0.0019*
(1.71)
-0.0006
(-0.80)
0.0011*
(1.88)
0.0000
(0.02)
0.0041***
(3.29)
51
Coefficient
(t-statistic)
CGt-3*NC
∆LMV
∆SP500
∆Lev
∆EP
∆ROE
∆DP
∆BP
∆TV
∆SaleChg
∆AnnRet
∆Rating
∆Beta
∆IRisk
Exp.
Sign
Institution = NC1
Institution = NC2
-0.0055***
(-5.60)
0.0069***
(6.15)
0.0026
(1.02)
0.0017
(0.42)
0.0026*
(1.75)
-0.0001
(-1.24)
-0.0087
(-0.58)
-0.0012
(-0.80)
0.0282***
(6.55)
-0.0000
(-0.74)
0.0009*
(1.67)
-0.0102**
(-2.35)
0.0001
(0.15)
-0.2800***
(-6.05)
-0.0007
(-0.04)
0.0055**
(2.28)
0.0008
(0.36)
-0.0027
(-0.69)
0.0019*
(1.80)
-0.0002
(-0.19)
-0.0092
(-0.37)
-0.0028
(-0.74)
-0.0031
(-0.42)
0.0007
(1.26)
0.0005*
(1.71)
0.0082**
(2.44)
-0.0022*
(-1.69)
-0.3185**
(-2.43)
Industry Fixed
Effects
Included
Included
Year Fixed
Effects
Included
Included
Adj_R-Sqr
0.147
0.137
No_Obs
22319
15774
This table reports coefficients and t-statistics (in parentheses) from firm cluster, industry and year fixed effect,
and heteroscedasticity adjusted regressions of change in institutional ownership on change in CSR performance,
institutional investor type, two-way interaction terms between change in CSR performance and investor type,
and control variables. Results regarding industry and year intercepts are not reported. See Appendix for variable
definitions. ***, **, and * denote significance at the 1%, 5%, and 10% levels in a two-tailed test, respectively.
52