Institutional Ownership Horizon Corporate Social Res 2019 Journal of Busine

Journal of Business Research 105 (2019) 61–79
Contents lists available at ScienceDirect
Journal of Business Research

journal homepage: www.elsevier.com/locate/jbusres
Institutional ownership horizon, corporate social responsibility and T

shareholder value☆
Otgontsetseg Erhemjamtsa,*, Kershen Huangb
a
McCallum School of Business, Bentley University, Waltham, MA 02452, United States of America
b
Huizenga College of Business and Entrepreneurship, Nova Southeastern University, Fort Lauderdale, FL 33314, United States of America
ARTICLE INFO ABSTRACT
Keywords: A widely held view among policymakers, corporate executives and the media is that short-termism among
Corporate social responsibility institutional investors is increasingly prevalent. However, some institutional investors are increasingly vocal
Institutional investors about taking a long-term approach, and these investors care about environmental, social and governance (ESG)
Investment horizon issues. The reality is that investors are a diverse set of stakeholders with various objectives and time horizons. In
Short-termism
the academic literature, empirical evidence on the relationship between institutional ownership horizon and
Long-term investing
corporate social responsibility (CSR) has been mixed. In this paper, we show that institutions with longer
(shorter) investment horizons promote (discourage) CSR at the firm level. In addition, the higher the proportion
JEL classification:
A13 of long-term (short-term) investors, the higher (lower) the effect of CSR on long-term (short-term) buy-and-hold
G23 returns. These findings are consistent with the view that short-termism on the part of institutional investors
M14 places short-term pressure on companies, and therefore discourages long-term investments that create value.
1. Introduction 2017. With a 100% annual turnover rate equal to a holding period of
one year, a 60% rate implies that investors are holding the average
Short-termism in financial markets has been the subject of ongoing NYSE stock for 1.67 years (20 months) at a time.2 A similar trend has
debate among leaders in business, government and academia for more been observed for most OECD stock markets, where average holding
than 30 years, with much of the discussion focusing on whether it de- period has fallen between one and three years in selected OECD ex-
stroys value. Critics of short-termism point to the record-short periods changes over the last 20 years. While this reflects investment transac-
of time stocks are being held, rise of high-frequency trading (HFT), tions driven by both individuals and institutional investors, institu-
shortening average CEO tenures and disproportionate Wall Street focus tional investors now account for the largest share of investment
on quarterly earnings.1 According to the New York Stock Exchange activity. Recent surveys of C-suite executives conducted by McKinsey &
(NYSE) Factbook, the year-to-date annualized turnover for NYSE stocks Company suggest that pressure to deliver short-term results has in-
- the rate at which stocks are bought and sold - is 60% as of December creased since 2013. From 2013 to 2016, the share of respondents who
☆
We are grateful to Naveen Donthu (Editor-in-Chief), Claire Crutchley Lending (Associate Editor), two anonymous referees, Senay Agca, Katya Salavei Bardos, and
the seminar participants at Florida Atlantic University, the 2019 Eastern Finance Association Annual Meeting in Miami, FL, the 2017 Financial Management
Association Annual Meeting in Boston, MA, and the 2016 IESE International Symposium on Ethics, Business and Society in Barcelona, Spain for their helpful
comments and suggestions. Any omissions or errors are the authors' alone. An earlier version of this paper has been circulated under the title “Examination of the
Relationship Between Institutional Ownership Horizon and Corporate Social Responsibility.”
*
Corresponding author.
E-mail addresses: [email protected] (O. Erhemjamts), [email protected] (K. Huang).
1
Advancements in trading technologies made very rapid placements of buy and sell orders possible. Trades can now be executed in micro-seconds, faster than a
blink of an eye. As a result, HFT has become widespread, especially in equity markets. According to the TABB Group (www.tabbgroup.com), a securities market
research firm, HFT was estimated to have accounted for 21% of all U.S. equity market volume in 2005. HFT market share increased to an all-time high of 61% in 2009
with aggregate revenues of $7.2 billion. However, following the financial crisis, the rise of HFT came to a halt and its market share started to recede. In 2017, HFT
accounted for little over 50% of daily trading volume with aggregate revenues of below $1 billion.
2
The NYSE Factbook also reveals that the average holding period for stocks in 1960 was 8.33 years. By 1970 it had dropped to 5.26 years. By 1980 it had dropped
further to 2.78 years, by 1990 to 2.17 years, by 2000 to just 1.14 years, and during 2008–2009 period, the average holding period was just nine months. While the
2017 figure shows improvement over the recent recession numbers, it is still far shorter than the average holding period of 3.43 years in the 20th century.
https://doi.org/10.1016/j.jbusres.2019.05.037
Received 4 January 2018; Received in revised form 31 May 2019; Accepted 31 May 2019
Available online 08 August 2019
0148-2963/ © 2019 Elsevier Inc. All rights reserved.
O. Erhemjamts and K. Huang Journal of Business Research 105 (2019) 61–79
reported feeling the most pressure to demonstrate strong financial institutions following stocks being added to major stock market indices
performance within two years or less rose from 79% to 87%.3 (Aghion, Van Reenen, & Zingales, 2013). We also use industry median
The topic of short-termism has gained the attention of global forums levels of ownerships in the first stage, as prior literature has shown that
such as the OECD, the APEC, the G20 and the World Economic Forum, institutions tend to herd within industries (Choi & Sias, 2009; Grinstein
evidenced by the numerous task forces, round tables, conferences, and & Michaely, 2005). Second, we use pseudo-Russell 1000 and 2000
research reports by them.4 Generally, these initiatives call for more memberships as instruments instead of S&P 500 membership, based on
“responsible” and long-term investment among institutional investors the empirical findings that S&P 500 additions and deletions do not
and a transformational change in investor behavior. However, the case exhibit as symmetric price effects as the annual Russell reconstitutions
for the long-term investments (e.g., environmentally and socially re- do (Chang, Hong, & Liskovich, 2015). Third, we use lagged CSR scores
sponsible investments) among institutional investors is far from settled and lagged short-term and long-term ownerships in the first-stage es-
in the academic community. On the one hand, the classic agency per- timations to allow institutions set preferences based on observable CSR
spective on corporate social responsibility (CSR) argues that good social scores at the time of their investments. The use of lagged levels of
performance comes at the expense of good financial performance be- ownerships as instruments is based on the empirical finding that in-
cause valuable resources are misused instead of being spent on value- stitutional holdings are highly persistent (Gompers & Metrick, 2001).
added projects or returned to shareholders. Therefore, firms' primary We find in each of the above specifications that the positive relation
goal should be to maximize shareholder wealth (Friedman, 1970). On between investor horizon and firm CSR persists.
the other hand, the stakeholder perspective argues that effective sta- Also, to directly address findings in prior literature that CSR scores
keholder management can enhance firms' ability to achieve competitive are not one-size-fits-all (e.g., Walls, Berrone, & Phan, 2012; Strike, Gao,
advantage and long-term value creation, and therefore, firms should & Bansal, 2006a), we utilize several alternative measures for CSR: the
invest in CSR (Freeman, 1984). Just as these theories offer opposite list of “100 Best Companies To Work For In America (BCW; Edmans,
views, empirical evidence on the relationship between institutional 2011, 2012),” industry-adjusted CSR scores and CSR scores that only
ownership and CSR has been mixed, making it difficult to advance the consist of “material” items, as defined by the Sustainability Accounting
conversation. For example, Graves and Waddock (1994) find no re- Standards Board (SASB). In all of these analyses, we find similar results
lationship between the percentage of shares institutionally owned and to the ones reported in our base models.
CSR. Johnson and Greening (1999) find that a proportion of a company As for the measurement of investor horizon, we implement Yan and
owned by pension funds is positively related to CSR, but equity own- Zhang (2009) classification of institutional investors, which relies on
ership by investment management funds exhibits no direct relationship the institutions' portfolio turnover rather than their legal type (e.g.,
with CSR. However, more recent studies (e.g., Neubaum & Zahra, 2006; mutual funds, pension funds). According to this classification, a higher
T. Chen, Dong, & Lin, 2017; Dyck, Lins, Roth, & Wagner, 2019) find a average churn rate implies a shorter investment horizon. For each year,
positive link between institutional holdings and CSR. we classify an institution as short-term (long-term) if it has an average
In addition, the link between long-term investments and long-term churn rate that is above (below) the sample median for that year. This
value creation needs to be explored further as the empirical evidence is approach allows us to capture heterogeneity in investment horizon
mixed here as well (e.g., Margolis & Walsh, 2003; Orlitzky, Schmidt, & within each legal type in that some mutual funds will be classified as
Rynes, 2003; van Beurden & Goessling, 2008). In 2017, McKinsey long-term (or short-term) and some pension funds will be classified as
Global Institute has released a study showing that companies they short-term (or long-term) based on their trading behavior. It also allows
classify as “long-term” outperform their short-term peers on a range of institutional investors to fall into different categories over time as their
key economic and financial metrics.5 While calling it an “important portfolio managers and portfolio turnover might change.7 Further, we
topic”, Lawrence Summers responded to the report in the Harvard ensure that our results are not driven by our choice of ownership
Business Review, arguing that the issue is still unresolved and that their proxies by employing several different measures of ownership, such as
findings deserve much discussion, debate, and attempts at replication.6 using annual averages instead of end-of-year calculations, adopting a
Given the growing interest in addressing the problem of short-ter- tercile cut instead of median cuts, employing Bushee (1998) classifi-
mism, the perceived lack of consensus in theoretical perspectives and cations (transient, quasi-indexer, and dedicated) and Gaspar, Massa,
the lack of clear and convincing empirical evidence, we examine the and Matos (2005) ownership classifications.
relationship between institutional ownership horizon, CSR and long- In addition to our analysis of buy-and-hold returns, we also examine
term value creation in this paper by using investor churn rates, KLD buy-and-hold abnormal returns (BHARs) and are able to find similar
scores and buy-and-hold returns. Our main results are twofold: First, we results that the short-term effect of CSR with SIO on prices is negative
show that institutions with longer (shorter) investment horizons pro- and the long-term effect of CSR with LIO on prices is positive. We
mote (discourage) CSR at the firm level; second, we provide evidence employ two different benchmarks in calculating BHARs: the industry
that the effect of CSR on long-term (short-term) buy-and-hold returns is median returns and the value-weighted CRSP returns. In each of these
higher (lower) for firms with higher proportion of long-term (short- checks, we are able to obtain qualitatively similar results.
term) investors. Our study contributes to the current literature in several ways. First,
In an effort to account for endogenous investor choices and reverse we address the perceived lack of consensus in theoretical perspectives
causality, we incorporate three sets of two-stage estimations: First, we and discuss how recent studies (e.g., Jensen, 2002; Barnett, 2007;
endogenize institutional investment preferences for CSR by using S&P Benabou & Tirole, 2010) help reconcile the conflicting views. In par-
500 index membership as an instrument to predict institutional own- ticular, Jensen (2002) argues that specifying long-term value max-
ership in the first stage, utilizing the variation of ownership by imization as objective of the firm solves the problems that arise from
3
www.fcltglobal.org/docs/default-source/default-document-library/fclt-
7
global-rising-to-the-challenge.pdf Capturing the heterogeneity in investment horizon is important in that there
4
www.oecd.org/finance/private-pensions/institutionalinvestorsandlong- is a disagreement on what “long-term” means even among pension funds, in-
terminvestment.htm stitutions typically regarded as long-term. According to a 2014 survey by IPE
5
www.mckinsey.com/~/media/McKinsey/Global Themes/Long term magazine, over three-quarters of respondents considered their fund to be a
Capitalism/Where companies with a long term view outperform their peers/ long-term investor. When asked to pin down their assumptions, a quarter de-
MGI-Measuring-the-economic-impact-of-short-termism.ashx fined long-term investing as taking a three to five-year view. A third saw it as a
6
https://hbr.org/2017/02/is-corporate-short-termism-really-a-problem-the- seven to ten-year view. www.ipe.com/news/esg/pension-funds-split-over-
jurys-still-out meaning-duration-of-long-termism/10004395.fullarticle
62
having to meet interest of various stakeholders. We provide supporting decreased their holdings of larger stocks. As institutions' ownership has
evidence for this argument by examining the link between long-term increased, their role as shareholders has also evolved. Some institu-
investments and long-term value creation. tional investors, particularly public pension funds and union pension
Second, we are able to offer new evidence on how strengths and funds, began to abandon their traditional passive shareholder role and
concerns components of the firms' CSR score are affected by institu- become more active participants in the governance of their corporate
tional ownership horizon. Most earlier studies calculate overall CSR holdings. Gillan and Starks (2000) argue that the constraints on selling
score for each firm by summing up scores from five dimensions in KLD under-performers imposed by the indexing strategy have provided an
database: employee relations, environment, community relations, pro- important motivation for shareholder activism by public pension funds.
duct, and diversity. The overall CSR score in these studies in effect is the Proponents of the increased activism argue that a number of positive
sum of all strengths minus the sum of all concerns, which is equivalent influences arise from such behavior. For example, Bebchuk, Brav, and
to our raw CSR Net Score. Focusing on CSR Net Score alone results in Jiang (2015) find that hedge fund activism leads to long-term im-
missing on heterogeneity among firms, reducing variation in the CSR provements in operating performance.
measure. More importantly, recent research shows that the KLD As the institutional ownership in equities becomes increasingly
strengths and concerns measures are theoretically and empirically dis- prominent, society's interest in corporate social responsibility (CSR)
tinct and represent two independent constructs even though they may and socially responsible investing (SRI) is on the rise as well.10 Ac-
correlate with each other (e.g., Walls et al., 2012; Mishina, Dykes, cording to the US SIF Foundation, in the 23 years between its first
Block, & Pollock, 2010; Strike et al., 2006a).8 By looking at the Trends Report in 1995 and its 12th in 2018, responsibly managed asset
strengths and concerns components of CSR scores separately, we are pools have grown more than 18-fold, from $639 billion to over $12.0
able to document channels through which institutional owners affect trillion.11 Common strategies for SRI funds include incorporating var-
CSR: short-term investors discourage investments that lead to im- ious environmental, social, and governance (ESG) criteria into their
provements in CSR strengths, and long-term investors encourage in- investment analysis and portfolio selection, filing shareholder resolu-
vestments that lead to reductions in CSR concerns. tions, or both.
Third, we acknowledge that CSR is a multidimensional construct
and that companies may treat environmental and social issues differ- 2.1. Theoretical perspectives
ently in practice. Aggregating all dimensions of CSR into a single
composite score fails to account for cases where firms are responsible in There has been a long-standing debate on whether firms should
some dimensions such as employee relations, and not in others, such as pursue socially responsible investments. On the one hand, the classic
environment (e.g., Walls et al., 2012; Strike et al., 2006a).9 Failure to agency perspective on CSR argues that good social performance comes
disaggregate responsible actions from irresponsible ones may partly at the expense of good financial performance because valuable re-
explain the inconclusive findings in prior research between institutional sources are misused instead of being spent on value-added projects or
ownership and CSR. Therefore, we examine how each of the CSR ca- returned to shareholders. Therefore, firms' primary goal should be to
tegories (environment, community relations, diversity, employee rela- maximize shareholder wealth (Friedman, 1970; Jensen & Meckling,
tions, human rights, product and controversial business involvement) is 1976). On the other hand, the stakeholder perspective argues that
affected by institutional ownership horizon. We find that the negative managers should make decisions so as to take account of the interests of
effect of short-term ownership on CSR is driven by employee relations all the stakeholders in a firm − not only the shareholders, but also
category, consistent with Benabou and Tirole (2010) who argue that the employees, customers, communities, the environment, etc. (Freeman,
short-termism implies both an intertemporal loss of profit and an ex- 1984). While the two theories seem to highlight competing interests of
ternality on stakeholders. In other words, managers make decisions that various stakeholders, resulting in multiple objectives, Jensen (2002)
increase short-term profit, but reduce shareholder value and hurt em- argues that having multiple objectives is equivalent to having no ob-
ployees or suppliers by reneging on implicit contracts with them to jective at all. According to Jensen (2002), firms cannot maximize long-
reduce costs. Alternatively, a firm could skimp on safety or pollution term market value while ignoring or mistreating any important con-
control to increase short-term profits, exposing the firm to liabilities stituency. He proposes the idea of “enlightened value maximization”
down the road such as future lawsuits, consumer boycotts and en- which utilizes much of the stakeholder theory but accepts maximization
vironmental clean-up costs. Finally, we find that the positive effect of of the long-run value of the firm as the criterion for making the re-
long-term ownership on CSR is present in community relations, di- quisite tradeoffs among its stakeholders and specifies long-term value
versity, human rights and product categories. maximization as the firm's objective. This proposal therefore solves the
problems that arise from having to meet multiple objectives. Based on
this framework, we hypothesize that firms with a higher proportion of
2. Prior research on institutional ownership and CSR long-term investors will be able to make more long-term investments
(such as socially responsible investments) to maximize long-term value.
Institutional investors have become increasingly important as equity Firms with a higher proportion of short-term investors will not be able
holders in the U.S. financial markets. Gompers and Metrick (2001) find to make long-term investments because long-term value maximization
that institutional ownership in equities nearly doubled from 1980 to will be at odds with the objectives of the short-term investors.
1996 to reach more than 50%. Blume and Keim (2012) report that Since Freeman published his seminal piece (Freeman, 1984), sta-
institutional ownership in equities reached 67% by the end of 2010. In keholder theory development has centered around classifying
contrast to earlier research that found institutional investors preferred
larger, more liquid stocks, they find that institutions, hedge funds in
10
particular, have increased their holdings of smaller stocks and Modern SRI gained global traction in the 1980s with the movement to
divest investments from South Africa in protest to its system of racial segre-
gation known as Apartheid. Then, with the Bhopal, Chernobyl, Exxon Valdez oil
8
Strike et al. (2006a) point out that aggregating concerns and strengths into a spills, as well as the more recent Deepwater Horizon spill in the Gulf of Mexico
net strengths score fails to recognize irresponsible actions for which there are and Fukushima nuclear disaster in Japan, the environment became the top
no responsible analogs. For example, violence against employees is irrespon- concern for socially conscious investors. There has also been a resurgence of
sible, bit the absence of violence is not necessarily responsible. policies restricting investments in firearms due to mass shootings in Sandy
9
Walls et al. (2012) mention that environmental practices tend to differ from Hook Elementary School in Newton, CT, Virginia Tech University in
other social practices since they require specific capabilities and significant Blacksburg, VA, a Century 16 movie theater in Aurora, CO, etc.
11
capital investment, and are guided by regulation. www.ussif.org/trends
63
stakeholders into categories that provide an understanding of how Millington (2004) find that the proportion of company owned by long-
stakeholders influence firms' operations, and how firms respond to term institutional investors (pension funds, life assurance funds, and
those influences. For example, Mitchell, Agle, and Wood (1997)’s charitable funds) is positively related to CSP providing further support
theory of stakeholder salience sorts stakeholders according to the pre- for Johnson and Greening (1999). Cox and Wicks (2011) use a larger
sence of the attributes of power, legitimacy and urgency. Based on this sample of UK firms, and wider variations in types of institutional in-
framework, we hypothesize that institutional investors with longer investors compared to Cox et al. (2004). In particular, they classify mu-
vestment horizon (such as public pension funds and private multi- tual funds, life insurance funds, externally managed pension funds as
employer funds) have higher salience relative to ones with shorter in- transient (short-term) institutions, and in-house-managed public and
vestment horizon (such as mutual funds, banks, and insurance private sector pension funds as dedicated (long-term) institutions. Cox
companies). Therefore, firms with a higher proportion of long-term and Wicks find that CSP positively and significantly influences the de-
(short-term) investors would make more (less) long-term investments mand for shares by dedicated institutional investors.12
such as CSP, to meet the needs of their more salient investors. While studies we mentioned so far mainly looked at what attracts
Moreover, Barnett (2007) argues that firms should view CSP as a institutional investors, there is a growing body of literature that looks at
long-term investment in creating the capacity to influence stakeholders. the effect of institutional ownership on firm behavior (Bushee, 1998;
Stakeholder influence capacity (SIC) is defined as “the ability of a firm Derrien, Kecskés, & Thesmar, 2013; Harford, Kecskés, & Mansi,
to identify, act on, and profit from opportunities to improve stakeholder 2018).13 Because different stakeholders often have competing ex-
relationships through CSR”. Firms with a weak history of social re- pectations, executives have to be attentive to the demands of their most
sponsibility have little or no SIC and are not credible with stakeholders. important stakeholders (Johnson & Greening, 1999; Hillman & Keim,
Therefore, firms with higher proportion of long-term investors will ei- 2001). Stakeholders who have more power and more actively express
ther passively comply with their needs and make more long-term in- themselves are likely to have a greater say in the firm's strategic deci-
vestments, or try to improve their relationship with these investors sions. According to Holderness and Sheehan (1988), “block-share-
through investing more in CSR. In contrast, firms with a higher pro- holders do not merely monitor management teams, they lead them.”
portion of short-term investors will make less long-term investments. Relying on the theory of stakeholder salience, Neubaum and Zahra
Benabou and Tirole (2010) discuss three alternative visions of CSR, (2006) suggest that institutional owners' investment horizons, as well as
where the first vision has a long-term perspective, and the remaining the frequency and coordination of institutional owners' activism,
two build on individual social responsibility. In the vision with a long- moderate the institutional ownership-CSP relationship. In particular,
term perspective, which refers to “doing well by doing good”, CSR is the authors suggest that demands of short-term investors may not be-
about taking a long-term perspective to maximizing (intertemporal) come salient to senior executives who may pay greater attention to
profits. This suggests that socially responsible investors should position those institutional owners with longer and more established ties to the
themselves as long-term investors who monitor management and exert firm. Using a sample of Fortune 500 firms, Neubaum and Zahra (2006)
voice to correct short-termism that results from poorly designed man- find evidence consistent with this argument. In particular, they find
agerial incentives or career concerns. The other two visions build on that the holdings by long-term institutional owners are significantly and
individual social responsibility. Benabou and Tirole (2010) argue that positively associated with future CSP, and this positive relationship
these three motives are mutually interdependent, and both policy-ma- grows stronger as institutional owners' activism and coordination in-
kers and social activists must have a good understanding of these increases.
teractions. Other studies also show that institutional investors influence firms'
CSR commitments through shareholder activism (e.g., Gillan & Starks,
2.2. Empirical research 2000; David, Bloom, & Hillman, 2007; Dimson, Karakas, & Li, 2015; T.
Chen et al., 2017; Dyck et al., 2019). For example, using ISS Risk Me-
Earlier studies on institutional ownership and CSR include Graves trics Shareholder Proposal and Vote Results database, T. Chen et al.
and Waddock (1994), and Johnson and Greening (1999). These studies (2017) show that there are increased amount and probability of SRI
use institutional ownership as a dependent variable, and corporate so- shareholder proposals for firms in the top of the Russell 2000 index
cial performance (CSP) as an explanatory variable. Graves and than firms in the bottom of Russell 1000. Also, the probability of SRI
Waddock (1994) find that the relationship between the percentage of proposals is higher for firms just included in Russell 2000. Using ex-
shares institutionally owned and CSP is positive but insignificant for S& tensive proprietary database of CSR engagements by an asset manager,
P 500 firms. Their composite CSP measure was constructed from the Dimson et al. (2015) show that shareholder engagements address ESG
KLD database. Johnson and Greening (1999) take a different approach concerns. Collaboration among activists is instrumental in increasing
and classify institutional investors into investment management funds the success rate of environmental/social engagements. After successful
(mutual funds and investment banks), and pension funds. They argue engagements, particularly on environmental/social issues, companies
that investment management funds tend to pursue short-term gains due
to the fact that investment managers' rewards are based on quarterly
results. In contrast, pension fund managers may monitor management 12
The notion of dedicated vs. transient investors in Cox and Wicks (2011) is
closely and press for needed changes because they cannot exit by selling from Bushee (1998), who classifies institutions based on their trading behavior.
large blocks of stock without driving the price down and suffering Transient institutions are characterized as having high portfolio turnover and
further losses. As a result, mutual funds may have much higher port- highly diversified portfolio holdings. These traits reflect the fact that transient
folio turnover, whereas public pension funds may hold some stocks for institutions tend to be short-term investors whose interest in the firm's stock is
decades. For their sample of 250 US firms, Johnson and Greening report based on short-term trading profits. Dedicated institutions are characterized by
that pension fund equity (i.e., a proportion of a company owned by large average investments in portfolio firms and extremely low turnover, con-
pension funds) is positively related to CSP, but equity ownership by sistent with a relationship-investing role and a commitment to provide long-
term patient capital.
investment management funds exhibits no direct relationship with CSP. 13
Bushee (1998) reports that firms with shorter investor horizons reduce
The strong positive relationship between institutional ownership and
research and development expenditures to increase short-term earnings.
CSP for pension funds was driven mostly by the product quality di- Derrien et al. (2013) find that when a firm is undervalued, greater long-term
mension (environment and product quality ratings) of CSP, not so much investor ownership is associated with more investment and, more equity fi-
by the people dimension (community, employee relations, and diversity nancing, and less payout to shareholders. Harford et al. (2018) find that long-
ratings). term investors strengthen corporate governance and restrain managerial mis-
In a sample of over 500 UK companies, Cox, Brammer, and behavior such as earnings management and financial fraud.
64
experience improved accounting performance and governance and in- 3. Sample

creased institutional ownership.
While Neubaum and Zahra (2006) are cautious to draw any causal 3.1. Data
inferences on whether institutional investors drive CSP, more recent
studies (T. Chen et al., 2017; Dyck et al., 2019) utilize quasi-natural We obtain corporate social responsibility (CSR) information from
experiments to establish causality. T. Chen et al. (2017) use Russell the KLD STATS database provided by MSCI ESG Research, Inc. The
index reconstitutions as exogenous shocks to institutional holdings. The database contains annual data on the strength and concern ratings
authors conduct their tests using 2SLS specifications, in which they first across seven CSR categories for 3000 publicly held companies in the US.
instrument total institutional ownership (TIO) with exogenous varia- The CSR categories include the environmental, social (i.e., community,
tions around 1000/2000 threshold and then test the effects of in- diversity, employee relations, human rights, and product), and gov-
strumented ownership on CSR engagements.14 T. Chen et al. (2017) ernance (ESG) sub-categories. It also provides information on firm in-
find that the exogenous increase in institutional ownership leads to volvements in controversial business areas including alcohol, firearms,
higher CSR scores. Similarly, Dyck et al. (2019) assess whether in- gambling, military, nuclear power, and tobacco.16
stitutional investors drive the environmental and social performance of Coverage of KLD has expanded over time. The database started with
firms in 41 countries using additions to the MSCI ACWI index as an roughly 650 unique firms during the 1990s and this number has in-
instrument for TIO. As a further test of the hypothesis that institutional creased to over 1000 by 2001. According to the MSCI ESG Methodology
investors cause changes in firms' environmental performance, they use Manual, the database started covering 3000 largest US companies by
the BP Deepwater Horizon Oil spill as a quasi-natural experiment. Dyck market capitalization starting from 2003. However, starting from 2014,
et al. (2019) find that firms with greater TIO at the time of the shock are the number of companies covered by KLD dropped to 2400. Therefore,
more reactive in improving environmental performance in the years for consistency in our sample size over time, we focus on a sample
following this shock. period of 2003 to 2013.17
Finally, Nguyen, Kecskés, and Mansi (forthcoming) study the effect Ownership information of institutions are extracted from the 13F
of CSR on shareholder value using KLD data. They regress Tobin's q filings of investment managers (per the Securities Exchange Act of
(market-to-book) on contemporaneous CSR score, lagged long-term 1934) provided by Thomson Reuters (TFN; formerly CDA/Spectrum).18
investor ownership (LIO), and their interaction. Based on 1991–2009 The TFN S34 Institutional Holdings File contains information on
data, the authors find that the coefficient on the CSR score is positive common stock holdings and transactions of managers with $100 million
and significant, the coefficient on the LIO is negative and significant, or more in assets under management. For firms in the KLD universe but
and the coefficient on the interaction between CSR and LIO is positive not in TFN, we set their institutional equity holdings to 0% because it is
significant. The interaction term is the focus of their analysis as they likely that their equity ownerships do not meet the filing requirements
acknowledge the endogeneity of the main effects (CSR and LIO). of the SEC (Grinstein & Michaely, 2005).
Therefore, based on the positive coefficient on the interaction term, the Financial statement items are from the Compustat Annual
authors argue that long-term investors are able to ensure that managers Fundamentals File and are as of the fiscal year during which the KLD
choose the amount of CSR that maximizes shareholder value. Later in scores are computed. Market capitalization, equity returns, and shares
their paper, Nguyen et al. (forthcoming) regress excess returns on CSR, outstanding data are from CRSP (Center for Research in Security
LIO and their interaction and find that the coefficient on the interaction Prices). Since the fundamentals of financial (SIC codes from 6000 to
term is negative and significant. 6999) and utility (SIC codes from 4900 to 4999) firms are subject to
Our paper complements and extends Neubaum and Zahra (2006) heavy regulatory supervision, and therefore do not necessarily reflect
and Nguyen et al. (forthcoming). Compared to Neubaum and Zahra the economic characteristics that we study, we exclude them from our
(2006), we use larger sample, use superior measures of investor hor- analysis.
izon, account for the multidimensionality of CSR and address reverse
causality. Compared to Nguyen et al. (forthcoming), which examine the 3.2. Measures of CSR
effect of lagged LIO on current Tobin's q, we examine the effect of both
long-term institutional ownership (LIO) and short-term institutional Our dependent variables are measures of CSR from MSCI ESG STATS
ownership (SIO) on long-term value creation using 1-year through 6- database. MSCI ESG STATS evaluates companies on over 60 environ-
year buy-and-hold returns. In addition, we follow Khan, Serafeim, and mental, social, and governance (ESG) indicators in seven categories:
Yoon (2016) to create a CSR score that only consists of “material” items. community, environment, diversity, employee relations, human rights,
The material items are defined by the Sustainability Accounting Stan- products, and governance. Each category consists of binary indicators in
dards Board (SASB).15 “strength” and “concern” dimensions. If a company meets the criteria
established for a rating (strength or concern), its score for that category is
equal to 1; the score takes the value of 0 otherwise. We construct vari-
ables “CSR Strengths” as the sum of ESG indicators on attributes that are
14
Russell 1000 and 2000 indices are constructed based on the end-of-May identified as strengths and “CSR Concerns” in an analogous manner.
market capitalization ranks each year. Since there are only very small differ- Following Hillman and Keim (2001) and other recent studies, we assign
ences in market values surrounding the 1000/2000 threshold and firms cannot equal importance to ESG categories and construct the variable CSR Net
control their rankings precisely, firms being assigned to the left or right of the Score (also known as the KLD index), our measure of overall CSR, by
cutoff is quasi-random. Because Russell indices are value-weighted, the random subtracting “CSR Concerns” from “CSR Strengths.”
assignment leads to significant differences in portfolio weights, and further in
institutional ownership, around the threshold.
15 16
To create the materiality score, we first map each of the SIC two-digit codes MSCI ESG Research Inc. is a subsidiary of MSCI Inc. The MSCI ESG KLD
of the firms in our sample to one of the SASB sectors based on their Sustainable STATS database was originally created by KLD Research & Analytics, Inc. (KLD)
Industry Classification System (SICS). These SASB SICS sectors include (i) in 1991. Founded in 1988, the firm was later acquired by RiskMetrics in 2009,
health care, (ii) financials, (iii) technology and communications, (iv) non-re- and then (together with RiskMetrics) by MSCI in 2010.
17
newable resources, (v) transportation, (vi) services, (vii) resource transforma- The number of companies covered is not the only change that is present in
tion, (viii) consumption, (ix) renewable resources and alternative energy, and KLD data post 2013. When we sum up all strength and concern ratings across
(x) infrastructure. Once mapped, we are then able to pin down individual KLD CSR categories, the average and median values change significantly in
items that are material to each of the two-digit SIC codes using the Materiality 2014–2016 period, compared to 2003–2013 period as well.
18
Map system of SASB. The US Congress passed Section 13(f) of the SEC Act in 1975.
65
In addition to these raw “CSR Strengths” and “CSR Concerns” variables 3.4. Control variables and final sample
(which are count variables), we calculate “Adjusted CSR Strengths” and
“Adjusted CSR Concerns” scores following Servaes and Tamayo (2013). The Based on earlier studies on CSR determinants, we control for firm
reason for adjusting the raw scores is that the number of possible strengths characteristics (firm performance, firm size, firm size squared, firm age,
and concerns varies from year to year. Every year, some new indicators are R&D intensity, firm risk, leverage, likelihood of financial distress and
introduced and/or other indicators get discontinued. Therefore, we scale the industry characteristics–new economy firms, industry concentration
score for each category (community, environment, etc.) by the maximum and change in industry sales). Following Erhemjamts, Li, and
value for that category in a given year. We then add up all adjusted Venkateswaran (2013), we use Tobin's q as a measure of firm perfor-
strengths and concerns scores across all categories to calculate Adjusted CSR mance, natural logarithm of sales as a measure of firm size, number of
Strengths and Adjusted CSR Concerns. In order to calculate Adjusted CSR years in CRSP as a proxy for firm age, R&D expenses scaled by total
Net Scores, we first find adjusted net score for each category, and then add assets as a measure of R&D intensity, standard deviation of monthly
up the adjusted net scores across all categories. returns in a fiscal year as a measure of firm risk, sum of long-term debt
Servaes and Tamayo (2013) exclude the corporate governance ca- and debt in current liabilities as a percentage of total assets as a mea-
tegory from their CSR measures by arguing: “corporate governance is sure of book leverage and modified Altman (1968) z-score as a proxy
about the mechanisms that allow the principals (shareholders) to re- for the likelihood of financial distress.19 New economy firms are de-
ward and exert control on agents (the managers)…CSR, on the other fined as companies competing in the business fields of computer, soft-
hand, deals with social objectives and stakeholders other than share- ware, internet, telecommunications, or networking (Murphy, 2003).
holders.” Accordingly, we also exclude corporate governance indicators Our final sample consists of 15,217 firm-year observations over the
from CSR Net Score, CSR Strengths, and CSR Concerns, as well as their 2003 to 2013 period, with 2860 unique firms. The characteristics of our
adjusted counterparts. sample firms are reported in Table 1. Panel A shows descriptive sta-
tistics of the main variables used, including the key dependent (CSR)
3.3. Measures of institutional ownership horizons and explanatory variables (short-term and long-term institutional
ownership).
To empirically proxy for the investment horizons of institutional Overall, we see figures that are consistent with prior literature. For
investors, we use investor churn rates calculated based on quarterly instance, the mean (median) net CSR score for our sample is −0.2832
portfolio turnovers (Gaspar et al., 2005; Yan & Zhang, 2009). Specifi- (− 1.0000), suggesting that the mean (median) firm has slightly more
cally, for each calendar quarter, we define the churn rate of investor k concerns than strengths in terms of CSR-related issues. The mean
at quarter t as (median) SIO and LIO are 0.3384 (0.3319) and 0.3911 (0.3987), re-
buy
spectively. It is worth noting that, due to our sample covering only
min Churn, Churnsell
k,t firms in the KLD universe (i.e., the 3000 largest firms every year), in-
stitutional ownerships of our sample firms appear larger than those
k, t
CR k, t 1
,
2 j J
(Nj, k, t Pj, t + Nj, k, t 1 Pj, t 1 ) (1) reported in earlier studies that examine the entire CRSP-Compustat
(CCM) Universe (e.g., Yan & Zhang, 2009). Our numbers are more in
where
line with those in studies that focus on, for instance, the markets for
Churnkbuy corporate bonds (e.g., Huang & Petkevich, 2016) or corporate control
,t = |Nj, k, t Pj, t Nj, k, t 1 Pj, t 1 Nj, k, t 1 Pj, t |
j J ; Nj, k, t > Nj, k, t 1 (2) (e.g., the bidders in Gaspar et al., 2005).
The mean and median net CSR scores can be more clearly observed
and in Panel B, where we show the number of firm-year observations in our
Churnsell
k,t = |Nj, k, t Pj, t Nj, k, t 1 Pj, t 1 Nj, k, t 1 Pj, t| sample by the number of CSR strengths and concerns. Based on our
j J ; Nj, k, t Nj, k , t 1 (3) data, 55.18% (25.48%) of the firm-years do not have CSR strengths
(concerns), meaning that 44.82% (74.52%) have at least one strength
are aggregate purchases and sales, respectively. Nj,k,t ≥ 0 denotes in- (concern), in any category. Although firms have more concerns than
vestor k’s shareholding of firm j ∈ J for quarter t, with J being the set of strengths on average, this relation flips as we examine cumulatively.
all sample firms. Pj,t presents the share price of firm j at the end of Specifically, at the count of four on either side, we see that 10.47% of
quarter t. The quarterly churn rate CRk,t above, specific to investor k at the sample firm-years have at least four strengths, while only 8.29%
time t, is therefore the minimum of purchase- and sale-generated have at least four concerns. This difference gradually widens as we
changes in number of shares valued using end-of-quarter prices at time move from the left to the right of Panel B. This highlights the im-
t, and further scaled by average portfolio size during the past quarter portance of examining strength and concern components separately.
from t − 1 to t.
For each December, institutional investor k’s horizon is determined
using its average churn rate over the past four quarters, i.e., 4. Analyses
3
1
avgCRk, t = CR k, t t . 4.1. Univariate analyses
4 t =0 (4)
Intuitively, the higher the average churn rate during a given year for an Table 2 presents the univariate relations between all variables used
institution, the higher its portfolio turnover, and therefore the more in this study. In Panel A, we present the correlation matrix. At first
short-term oriented it is likely to be. For each year, we classify an in- glance, we see from the bolded numbers that CSR is overall negatively
stitution as short-term (long-term) if it has an average churn rate that is
above (below) the sample median for that year. 19
The original Altman (1968) z-score consists of five components, which
Turning to the firm level, a firm j would have a portion of its equity
includes the ratio of market value of equity to book value of long-term debt.
ownership held by institutions during each year. This portion ranges from
Since we also control for a similar term, market-to-book, in our multivariate
0% to 100% and is referred to as the total institutional ownership (TIO) of models as a separate variable, we follow Graham, Li, and Qiu (2008) and em-
the firm. Using the above categorization of investment horizons (i.e., short- ploy a modified z-score that does not include the term. Compared to the original
term or long-term), we decompose this firm-level TIO into short-term in- z, while a higher modified z-score still indicates better financial health and thus
stitutional ownership (SIO) and long-term institutional ownership (LIO). lower default risk, the usual 1.81 and 2.99 cutoffs do not apply when using this
Therefore, the sum of SIO and LIO would equal to TIO. measure.
66
Table 1
Sample.
Percentiles
Variable N Mean Std. dev. 1st 5th 10th 25th Median 75th 90th 95th 99th
Panel A: Descriptive statistics

CSR Score (Net) 15,217 −0.2832 2.2971 −5.0000 −3.0000 −2.0000 −2.0000 −1.0000 0.0000 2.0000 4.0000 8.0000
CSR Score (Adj.) 15,217 −0.2313 0.5991 −1.6405 −1.0000 −0.7500 −0.6333 −0.2500 0.0000 0.3810 0.7452 1.8810
SIO 15,217 0.3384 0.1554 0.0223 0.0931 0.1402 0.2263 0.3319 0.4412 0.5451 0.6110 0.7241
LIO 15,217 0.3911 0.1509 0.0564 0.1323 0.1829 0.2866 0.3987 0.4956 0.5804 0.6341 0.7345
Tobin's q 15,217 2.0318 1.2464 0.7687 0.9288 1.0249 1.2384 1.6290 2.3496 3.5099 4.6440 7.2428
Sales ($Bil.) 15,217 4.5700 17.9104 0.0050 0.0603 0.1148 0.3055 0.8605 2.6122 8.3655 18.3660 65.0300
Firm Age 15,217 20.1597 18.3268 1.0000 2.0000 3.0000 7.0000 14.0000 28.0000 42.0000 61.0000 83.0000
R&D/Assets 15,217 0.0419 0.0792 0.0000 0.0000 0.0000 0.0000 0.0033 0.0522 0.1275 0.1925 0.3981
Stk Volatility 15,217 0.1193 0.0651 0.0347 0.0473 0.0565 0.0754 0.1052 0.1460 0.1963 0.2389 0.3460
Leverage 15,217 0.1877 0.1749 0.0000 0.0000 0.0000 0.0098 0.1644 0.2994 0.4339 0.5257 0.6712
Modified z 15,217 1.5943 2.0159 −7.0966 −1.5428 −0.0349 0.9754 1.8329 2.6448 3.4291 4.0565 5.0796
HHI 15,217 0.2607 0.2015 0.0413 0.0584 0.0685 0.1185 0.2013 0.3353 0.5229 0.6844 1.0000
Panel B: Sample Size by CSR Strengths and Concerns

CSR Strengths 0 1 2 3 4 5 6 7 8 9 10 ≥11
Observations 8397 3105 1384 738 392 322 227 159 133 102 61 197
Pct. of Sample 55.18% 20.40% 9.10% 4.85% 2.58% 2.12% 1.49% 1.04% 0.87% 0.67% 0.40% 1.29%
Inverse Cumu. 100.00% 44.82% 24.41% 15.32% 10.47% 7.89% 5.78% 4.28% 3.24% 2.37% 1.70% 1.29%
CSR Concerns 0 1 2 3 4 5 6 7 8 9 10 ≥11

Observations 3878 5109 3795 1174 487 268 212 132 69 38 27 28
Pct. of Sample 25.48% 33.57% 24.94% 7.72% 3.20% 1.76% 1.39% 0.87% 0.45% 0.25% 0.18% 0.18%
Cumulative Pct. 100.00% 74.52% 40.94% 16.00% 8.29% 5.09% 3.33% 1.93% 1.06% 0.61% 0.36% 0.18%
This table presents information on the sample used in this study. Panel A shows descriptive statistics of the main variables used, including the key dependent
(corporate social responsibility; CSR) and explanatory variables (short-term and long-term institutional ownership). Adjusted CSR scores are computed according to
Servaes and Tamayo (2013). Panel B shows the numbers of firm-year observations with certain amounts of CSR strengths (top sub-panel) and concerns (bottom sub-
panel) according to the KLD database, as well as the associated percentages and cumulative percentages relative to the entire sample.
(positively) correlated with SIO (LIO). The correlation coefficient be- SIO/LIO and the net scores of individual categories within CSR, in-
tween raw (adjusted) CSR and SIO is −0.07 (−0.05), while that be- cluding environmental and social issues. Social issues are further nar-
tween raw (adjusted) CSR and LIO is 0.12 (0.11), all statistically sig- rowed down to issues related more specifically to community, diversity,
nificant at the 1% level. These correlations are consistent with our employee relations, human rights, and products. Consistent with the
argument that long-term institutions are more likely to encourage firm statistics reported earlier, we find that SIO (LIO) yields correlation
level CSR engagement, and that short-term institutions more likely coefficients with all components that are, at the very least, weakly
discourage it. While this can be valid, the reverse may hold true as well. negative (positive). In the right sub-panel of Panel C, we replace CSR
That is, different types of institutional investors are attracted to firms scores with various forms of the “100 Best Companies to Work For”
with higher CSR scores, and the correlations simply reflect such in- scorings (BCW) used in Edmans (2011, 2012) as an alternative mea-
stitutional preferences for securities. In Panel B, we make our initial sure.20 We continue to see similar results.
attempt to disentangle these two arguments, provide some argument
for the former, and more carefully address concerns regarding en- 4.2. Multivariate analyses
dogenous choices of institutions in the “Robustness Checks and Exten-
sions” section. 4.2.1. CSR estimations
The left sub-panel in Panel B shows the correlations between raw We now perform multivariate analyses in testing the relation be-
CSR and lagged, contemporary, and lead SIO/LIO (ordered from left to tween lead CSR scores and SIO/LIO levels of firms. Our baseline model
right), while the right sub-panel shows the same numbers derived using takes the following form:
adjusted CSR. For the ease of comparison, we present the contemporary
correlations in bold fonts (note that they are identical to the ones in CSR t + 1 = S SIOt + L LIOt + Xt BX + t , (5)
Panel A, also reported in bold fonts, but with only two digits after the
where CSR, our dependent variable, is one of the CSR scores (combi-
decimal). Overall, for both the raw and adjusted scores, the correlation
of CSR with lead SIO/LIO is larger than with either contemporary SIO/ nation of net/strength/concern and raw/adjusted). SIO and LIO are
vectors of institutional ownership categorized into short-term and long-
LIO or lagged SIO/LIO (with the only exception being SIO with lagged
CSR, which yields a relatively high correlation of −0.0817). term using Eq. (4). X is the matrix of all control variables and ϵ is the
error vector; β and B are the estimated coefficients for the ownership
In Panel C, we examine more closely the relation between institu-
tional investment horizons and individual CSR categories. The left sub-
panel shows separately the correlations between SIO/LIO and CSR 20
“BCW Score” is an annual ranking of firms, ranging from 0 to 100. For
strengths and concerns. For CSR strengths, either raw or adjusted, we instance, the No. 1 ranked firm on the list would have a BCW score of 100, the
continue to see patterns that are consistent with those for the net scores No. 2 ranked firm would have a score of 99, so on and so forth. Companies that
reported in Panel A. CSR concerns appears to be inconsistent at the are not on the list would have a score of 0. The “BCW Dummy” takes the value
univariate level. Judging from the magnitude of these correlations, it of 1 if a firm is one of the “100 Best” for a given year, and 0 otherwise. The
seems that the univariate evidence provided by net scores is driven by “BCW Accumulated Dummy” takes the value of 1 if a firm has been one of the
CSR strengths. The middle sub-panel shows the correlations between “100 Best” during or prior to the year of a particular observation, and 0
otherwise.
67
Table 2
Correlations.
CSR Raw CSR Adj. SIO LIO q Size Size2 Age R/A RDϕ SV Lev. z NE. HHI ΔS Ind.
Panel A: Correlations between key variables and controls (Obs: 15,217 Firm-years; 2860 Unique firms)
CSR Score (Raw) 1.00
CSR Score (Adj.) 0.94 1.00
SIO −0.07 −0.05 1.00
LIO 0.12 0.11 0.03 1.00
Tobin's q 0.09 0.09 0.07 −0.14 1.00
Size (Log of Sales) 0.22 0.16 −0.02 0.33 −0.26 1.00
Squared Size 0.24 0.17 −0.06 0.30 −0.22 0.98 1.00
Firm Age 0.14 0.10 −0.19 0.22 −0.15 0.45 0.47 1.00
R&D/Assets 0.05 0.05 0.01 −0.17 0.34 −0.46 −0.37 −0.17 1.00
Missing R&D −0.11 −0.10 −0.04 0.02 −0.16 0.12 0.09 −0.01 −0.39 1.00
Stk Volatility −0.14 −0.12 0.02 −0.26 −0.03 −0.34 −0.32 −0.21 0.19 −0.04 1.00
Leverage −0.03 −0.02 0.04 0.06 −0.29 0.26 0.24 0.07 −0.24 0.20 0.02 1.00
Modified z 0.07 0.06 0.04 0.24 −0.08 0.46 0.38 0.19 −0.55 0.12 −0.27 −0.09 1.00
New Econ. 0.12 0.08 0.04 −0.10 0.12 −0.14 −0.14 −0.19 0.24 −0.18 0.06 −0.20 −0.15 1.00
HHI 0.03 0.03 −0.07 0.08 −0.07 0.13 0.12 0.14 −0.17 0.00 −0.06 0.02 0.15 −0.15 1.00
Chg in Sales (Ind.) −0.04 −0.04 −0.02 0.02 0.07 −0.01 −0.01 −0.04 −0.01 0.04 −0.18 −0.02 0.03 0.01 −0.05 1.00
Panel B: Correlations between ownership and CSR of different timings

CSR Score (Raw) CSR Score (Adjusted)
Lag Cont. Lead Lag Cont. Lead

SIO −0.0817 −0.0712 −0.0797 −0.0597 −0.0546 −0.0641
p-val (0.00) (0.00) (0.00) (0.00) (0.00) (0.00)
LIO 0.0890 0.1150 0.1339 0.0883 0.1064 0.1326
p-val (0.00) (0.00) (0.00) (0.00) (0.00) (0.00)
Obs. 12,168 15,217 15,217 12,168 15,217 15,217
Panel C: Correlations between CSR components and SIO/LIO (Obs: 15,217)

Str. Str. Con. Con. CSR Components BCW Edmans
(Raw) (Adj.) (Raw) (Adj.) Envir. Soc. Comm. Div. Emp. Hum. Prod. Score Dum. Accum.
SIO −0.14 −0.14 −0.10 −0.09 −0.06 −0.04 −0.02 −0.02 −0.06 0.00 −0.01 −0.04 −0.05 −0.07
LIO 0.15 0.15 0.05 0.03 0.09 0.09 0.03 0.09 0.05 0.01 0.02 0.03 0.05 0.07
This table presents Pearson correlations between the variables used in this study. Panel A presents the correlation matrix of all key (CSR scores and SIO/LIO) and
control variables (e.g., Tobin's q, firm size age, etc.). Panel B shows the univariate relation between institutional ownership and CSR scores of various timings (i.e.,
one-year lagged, contemporary, and one-year lead). Panel C presents the correlations between short-term/long-term institutional ownership and (i) separated
strength and concern scores, for both raw and adjusted measures, (ii) detailed components of the KLD CSR raw scores (environmental and social, where the social
component is further sub-categorized into community, diversity, employee relations, human rights, and product scores), as well as (iii) the BCW (i.e., best companies
to work for) index from Edmans (2011, 2012, see also Fortune Magazine and the Great Place To Work Institute).
vectors and controls matrix, respectively. All models include year and 1.99% and 3.09%, respectively.21 For Model 2, an increase in SIO from
industry (based on the two-digit SIC code) fixed effects. Further, all 0% to 100% lowers the adjusted CSR score by 0.1533, and an increase
estimations are reported using two-way robust standard errors (clus- in LIO from 0% to 100% increases the adjusted CSR score by 0.2516.
tered by firm and year) to simultaneously control for cross-sectional The two institutional holdings of our sample both have standard de-
and time-series dependencies (Cameron, Gelbach, & Miller, 2011; Gow, viations of more than 15%, and as such, an increase from 0% to 100%
Ormazabal, & Taylor, 2010; Petersen, 2009). ownership presents roughly an increase of six standard deviations.
The results are reported in Table 3. Models 1, 3, and 5 (2, 4, and 6) These estimated effects of institutional holdings on CSR scores are
show the estimations of lead raw (adjusted) CSR net score, strength, economically significant, considering that the 1st and 99th percentiles
and concern, respectively. Given the characteristics of the dependent of the scores are −1.6405 and 1.8810, respectively.
variables, Models 1, 3, and 5 are estimated using ordered logit models, Models 3 through 6 yield quite interesting results that are overall
and Models 2, 4, and 6 are estimated using OLS. consistent with, yet not completely identical to, those in Models 1 and
Models 1 and 2 estimate lead raw and adjusted CSR net scores, re- 2. As shown in Model 3, the estimated coefficients for SIO and LIO in
spectively. Consistent with the univariate results earlier, we find that the estimation of CSR raw strengths are −0.6259 and 0.9784, respec-
SIO (LIO) is associated with lower (higher) one-year-ahead CSR scores. tively. Both are statistically significant in the way we expect. However,
The estimated coefficients for SIO and LIO are −0.4508 and 0.9197 when turning to the estimation of adjusted strengths in Model 4, the SIO
(−0.1533 and 0.2516), respectively, for the raw (adjusted) score. For and LIO coefficients are estimated to be −0.1877 and 0.0570,
Model 1, given a cutpoint estimate of 1.5135 at the net score of −1 (not
reported, but available upon request), the probability of a 100% LIO
firm to have a non-negative CSR score is 35.58%. The same numbers for 21
CSR net scores, strengths, and concerns range from −9 to 18, 0 to 21, and 0
a 100% SIO firm and a non-IO firm (i.e., 0% for both SIO and LIO) are to 15, respectively, for our sample. Therefore, the estimated cutpoints of all
12.30% and 18.04%, respectively. In addition, given that CSR net ordered logit estimations in this study are omitted from reporting due to space
scores have a mean of −0.2832 and a standard deviation of 2.2971 (see concerns, but are available upon request. In the examples in the text above, the
Table 1), the probability of a 100% LIO firm to have a score that is at three probabilities for non-negative scores are calculated as
least one standard deviation above the mean (roughly a net score of 2), 1 1 + exp{ 1.5135 0.4508} = 12.30% for SIO, 1 1 + exp{ 1.5135 + 0.9197} = 35.58% for
1 1
based on a cutpoint estimate of 3.4442 at the score of 1, is 7.42%. Once LIO, and 1 1 + exp{ 1.5135} = 18.04% for non-IO. Likewise, the three probabilities
1
again, the same numbers for a 100% SIO firm and a non-IO firm are for scores at least 2 are calculated following the same logic, with a cutpoint of
3.4442.
68
Table 3
Lead CSR Net Score, Strength, and Concern estimations.
CSR Net Score CSR Strengths CSR Concerns
Model 1 Model 2 Model 3 Model 4 Model 5 Model 6

Raw Adjusted Raw Adjusted Raw Adjusted
Coef. p-val Coef. p-val Coef. p-val Coef. p-val Coef. p-val Coef. p-val
SIO −0.4508 (0.034) −0.1533 (0.017) −0.6259 (0.005) −0.1877 (0.000) −0.2170 (0.318) −0.0343 (0.441)
LIO 0.9197 (0.000) 0.2516 (0.000) 0.9784 (0.000) 0.0570 (0.137) −0.7570 (0.002) −0.1946 (0.000)
Tobin's q 0.1448 (0.000) 0.0452 (0.000) 0.2196 (0.000) 0.0390 (0.000) −0.0281 (0.327) −0.0061 (0.282)
Size (Log of Sales) −0.3532 (0.008) −0.0746 (0.150) −0.5302 (0.000) −0.2866 (0.000) −0.7889 (0.000) −0.2119 (0.000)
Squared Size 0.0523 (0.000) 0.0121 (0.020) 0.1023 (0.000) 0.0343 (0.000) 0.0897 (0.000) 0.0222 (0.000)
Firm Age 0.0005 (0.866) 0.0007 (0.419) 0.0051 (0.014) 0.0027 (0.000) 0.0081 (0.004) 0.0019 (0.004)
R&D/Assets 1.4271 (0.010) 0.5290 (0.001) 1.0240 (0.060) 0.0230 (0.801) −2.3551 (0.000) −0.5060 (0.000)
Missing R&D −0.1732 (0.036) −0.0619 (0.016) −0.3902 (0.000) −0.0858 (0.000) −0.1133 (0.137) −0.0239 (0.172)
Stk Volatility −1.3839 (0.004) −0.4804 (0.001) −0.7326 (0.112) −0.2111 (0.002) 1.3289 (0.093) 0.2693 (0.068)
Leverage −0.0818 (0.603) 0.0445 (0.393) −0.4033 (0.123) −0.0473 (0.214) −0.3471 (0.038) −0.0918 (0.013)
Modified z 0.0337 (0.137) 0.0144 (0.018) −0.0549 (0.009) 0.0002 (0.963) −0.0901 (0.005) −0.0142 (0.008)
New Econ. 0.3907 (0.000) 0.1100 (0.006) 0.6364 (0.000) 0.1509 (0.000) 0.1104 (0.354) 0.0410 (0.096)
HHI 0.1118 (0.464) 0.0613 (0.216) −0.0450 (0.803) 0.0493 (0.233) −0.1180 (0.425) −0.0120 (0.738)
Chg in Sales (Ind.) 0.0100 (0.952) 0.0026 (0.959) −0.5316 (0.045) −0.0777 (0.336) −0.3840 (0.133) −0.0802 (0.199)
Intercept Omit −0.8040 (0.000) Omit 0.8360 (0.000) Omit 1.6400 (0.000)
(Pseudo-)R2 0.0537 0.1819 0.1590 0.4712 0.108 0.3052
N 15,217 15,217 15,217 15,217 15,217 15,217
This table presents results from estimations of lead CSR (corporate social responsibility) using short-term and long-term institutional ownerships. The models take the
following functional form:
CSRt + 1 = S SIOt + L LIOt + Xt BX + t ,
where CSR is one of the CSR scores (combination of net/strength/concern and raw/adjusted). SIO and LIO are vectors of institutional ownership categorized into
short-term and long-term. X is the matrix of all control variables and ϵ is the error vector; β and B are the estimated coefficients for the ownership vectors and controls
matrix, respectively. All models include year and industry (based on the two-digit SIC code) fixed effects. All estimations are reported using robust standard errors
clustered by firm and year to simultaneously control for cross-sectional and time-series dependencies (Petersen, 2009; Cameron et al., 2011). Models 1, 3, and 5 (2, 4,
and 6) show the estimations of lead raw (adjusted) CSR net score, strengths, and concerns, respectively. Models 1, 3, and 5 are estimated using ordered logit models,
and Models 2, 4, and 6 are estimated using OLS.
respectively, with the latter being barely significant (p-value = 0.137). behind low strengths and LIO is more effective in driving down con-
Therefore, there is some evidence suggesting that, once we address the cerns.
potential effects from the year-to-year changes in CSR strength cate- In Table 5, we report estimations of lead scores for each of the five
gories (Servaes & Tamayo, 2013), SIO retains its adverse effect on one- social categories. From the estimated SIO coefficients across all models,
year-ahead CSR strength, but LIO is not as effective in increasing it. For it appears that the key category where short-term investors are exerting
the estimation of CSR raw concerns, we see in Model 5 that the SIO and adverse effects is employee relations. LIO covers a wider scope, with
LIO coefficients are −0.2170 and −0.7570, respectively, with the estimated coefficients being positively significant for all categories ex-
former being insignificantly different from zero. Qualitatively similar cept employee relations (p-value = 0.112). Notably, despite statistical
results are obtained when using adjusted concerns instead. That is, significance, the point estimates of SIO (LIO) coefficients are overall
while LIO shows a beneficial effect in lowering the one-year-ahead CSR negative (positive).
concerns in all cases, SIO does not have a significant effect. In sum, the Given that LIO shows a stronger effect in mitigating CSR concerns,
negative (positive) effect of SIO (LIO) on lead CSR net scores appears to we further examine how firm level engagement in controversial busi-
be driven primarily by SIO (LIO) lowering CSR strengths (concerns). ness issues can vary in institutional investment horizons. The KLD da-
tabase provides information on business involvement in alcohol, fire-
4.2.2. Components of CSR arms, gambling, military, nuclear, and tobacco products. Involvement
Next, we examine more closely the individual components of CSR. in these business areas can include the licensing, manufacturing (either
Essentially, we are asking the empirical question: Where is our base part or whole), operations or revenue generation (to a certain percent of
finding coming from? The construction of the CSR net scores in KLD total revenue), and large ownership by or of another involved firm,
allows us to break the aggregated score down into its environmental among many others.
and social components. Within the broad social category, we can fur- We report these results in Table 6. In Panel A, we estimate an all-in-
ther differentiate between the strengths and concerns attributed to the one model where the dependent variable is the lead indicator for in-
community, diversity, employee relations, human rights, and product volvement in any of the five controversial business areas. From our
aspects of firms. finding, we see that the estimated coefficient for SIO is insignificant,
In Table 4, we report results from estimating the adjusted net scores, while that for LIO is negatively significant. Once again, LIO is effective
strengths, and concerns for the environmental (in the left panel) and in mitigating firm level engagement in negative CSR issues. In Panel B,
social (in the right panel) categories separately. For both scores, we we run six models, one for each of the six sins. All controls are included
continue to see negative estimated coefficients for SIO and positive ones in each estimation, but we omit reporting their estimated coefficients
for LIO, indicating that larger long-term ownerships encourage CSR due to space concerns (available upon request). Looking across the
engagement in both categories. The effect from SIO is relatively weaker Panel, we observe that only the LIO coefficients in the firearms (Model
in both models, but overall still significantly negative (only at the 10% 2) and military (Model 4) regressions are negatively significant. All
level for the environmental category). Breaking down net scores into other estimated coefficients, including those for SIO, are not sig-
strengths and concerns, we see results that are consistent with those nificantly different from zero.
derived from the total CSR net scores. Specifically, SIO is the main force
69
Table 4
Environmental and social components of CSR.
Environmental components of CSR (Adj.) Social components of CSR (Adj.)
Model 1 Model 2 Model 3 Model 4 Model 5 Model 6

Net Strengths Concerns Net Strengths Concerns
Coef. p-val Coef. p-val Coef. p-val Coef. p-val Coef. p-val Coef. p-val
SIO −0.0235 (0.100) −0.0263 (0.056) −0.0028 (0.736) −0.1298 (0.018) −0.1614 (0.000) −0.0315 (0.445)
LIO 0.0345 (0.012) −0.0023 (0.844) −0.0368 (0.002) 0.2171 (0.000) 0.0593 (0.066) −0.1578 (0.001)
Tobin's q 0.0069 (0.000) 0.0061 (0.006) −0.0008 (0.513) 0.0383 (0.000) 0.0329 (0.000) −0.0054 (0.313)
Size (Log of Sales) −0.0350 (0.000) −0.0638 (0.000) −0.0288 (0.000) −0.0397 (0.389) −0.2228 (0.000) −0.1831 (0.000)
Squared Size 0.0038 (0.002) 0.0075 (0.000) 0.0038 (0.000) 0.0083 (0.060) 0.0268 (0.000) 0.0184 (0.000)
Firm Age 0.0000 (0.941) 0.0009 (0.000) 0.0009 (0.000) 0.0007 (0.323) 0.0018 (0.000) 0.0010 (0.072)
R&D/Assets 0.0512 (0.145) −0.0882 (0.005) −0.1394 (0.000) 0.4778 (0.000) 0.1112 (0.134) −0.3666 (0.000)
Missing R&D −0.0197 (0.003) −0.0309 (0.000) −0.0112 (0.014) −0.0422 (0.054) −0.0549 (0.000) −0.0127 (0.419)
Stk Volatility −0.1073 (0.000) −0.0702 (0.049) 0.0371 (0.257) −0.3732 (0.007) −0.1409 (0.010) 0.2322 (0.070)
Leverage 0.0301 (0.022) 0.0067 (0.528) −0.0233 (0.008) 0.0144 (0.756) −0.0541 (0.083) −0.0685 (0.046)
Modified z 0.0018 (0.091) −0.0014 (0.127) −0.0032 (0.000) 0.0126 (0.022) 0.0016 (0.614) −0.0110 (0.030)
New Econ. 0.0342 (0.001) 0.0319 (0.000) −0.0024 (0.574) 0.0757 (0.023) 0.1191 (0.000) 0.0434 (0.060)
HHI 0.0343 (0.020) 0.0309 (0.027) −0.0034 (0.762) 0.0269 (0.531) 0.0184 (0.581) −0.0086 (0.783)
Chg in Sales (Ind.) 0.0033 (0.902) −0.0172 (0.361) −0.0206 (0.036) −0.0008 (0.989) −0.0604 (0.513) −0.0597 (0.337)
Intercept −0.0546 (0.675) 0.1384 (0.008) 0.1930 (0.198) −0.7494 (0.000) 0.6976 (0.000) 1.4470 (0.000)
R2 0.2074 0.3531 0.4062 0.1562 0.4330 0.2278
N 15,217 15,217 15,217 15,217 15,217 15,217
This table presents results from estimations of lead environmental (left panel) and social (right panel) components of CSR (corporate social responsibility) using
short-term and long-term institutional ownerships. The models take the following functional form:
where CSR is one of the adjusted CSR scores (net/strength/concern) in the corresponding category. SIO and LIO are vectors of institutional ownership categorized
into short-term and long-term. X is the matrix of all control variables and ϵ is the error vector; β and B are the estimated coefficients for the ownership vectors and
controls matrix, respectively. All models include year and industry (based on the two-digit SIC code) fixed effects. All estimations are reported using robust standard
errors clustered by firm and year to simultaneously control for cross-sectional and time-series dependencies (Petersen, 2009; Cameron et al., 2011). Models 1, 2, and
3 (4, 5, and 6) show the estimations of lead adjusted environmental (social) net score, strengths, and concerns, respectively.
Table 5
Details of the social component.
Model 1 Model 2 Model 3 Model 4 Model 5
Community Diversity Emp. Relations Human Rights Product
Coef. p-val Coef. p-val Coef. p-val Coef. p-val Coef. p-val
SIO −0.0175 (0.157) −0.0311 (0.368) −0.0697 (0.000) 0.0055 (0.484) −0.0170 (0.201)
LIO 0.0314 (0.015) 0.1020 (0.007) 0.0295 (0.112) 0.0243 (0.006) 0.0299 (0.087)
Tobin's q 0.0044 (0.005) 0.0119 (0.013) 0.0120 (0.000) 0.0014 (0.193) 0.0086 (0.000)
Size (Log of Sales) −0.0203 (0.066) −0.0589 (0.003) 0.0148 (0.289) 0.0071 (0.349) 0.0177 (0.071)
Squared Size 0.0020 (0.070) 0.0101 (0.000) −0.0008 (0.524) −0.0007 (0.340) −0.0022 (0.023)
Firm Age −0.0004 (0.079) 0.0014 (0.000) 0.0001 (0.806) −0.0001 (0.519) −0.0002 (0.291)
R&D/Assets 0.1042 (0.001) 0.1488 (0.065) 0.1403 (0.003) 0.0471 (0.013) 0.0374 (0.264)
Missing R&D −0.0048 (0.396) −0.0361 (0.008) −0.0091 (0.227) 0.0100 (0.085) −0.0021 (0.746)
Stk Volatility −0.0848 (0.000) −0.1703 (0.022) −0.1001 (0.226) 0.0272 (0.239) −0.0451 (0.083)
Leverage 0.0065 (0.571) −0.0091 (0.726) 0.0048 (0.770) −0.0122 (0.139) 0.0244 (0.128)
Modified z 0.0037 (0.000) −0.0032 (0.260) 0.0090 (0.001) 0.0010 (0.230) 0.0021 (0.066)
New Econ. 0.0160 (0.011) −0.0140 (0.427) 0.0598 (0.000) −0.0033 (0.651) 0.0173 (0.034)
HHI 0.0236 (0.072) 0.0166 (0.509) −0.0141 (0.355) −0.0060 (0.588) 0.0068 (0.588)
Chg in Sales (Ind.) 0.0291 (0.239) −0.0128 (0.758) 0.0315 (0.294) −0.0938 (0.250) 0.0452 (0.014)
Intercept −0.1025 (0.254) −0.2029 (0.186) −0.1786 (0.017) −0.0988 (0.053) −0.1666 (0.191)
R2 0.0860 0.2841 0.1556 0.0594 0.0951
N 15,217 15,217 15,217 15,217 15,217
This table presents results from estimations of lead social components of CSR (corporate social responsibility) using short-term and long-term institutional own-
erships. The models take the following functional form:
where CSR is one of the adjusted social component net scores (community/diversity/employee relations/human rights/product). SIO and LIO are vectors of in-
stitutional ownership categorized into short-term and long-term. X is the matrix of all control variables and ϵ is the error vector; β and B are the estimated coefficients
for the ownership vectors and controls matrix, respectively. All models include year and industry (based on the two-digit SIC code) fixed effects. All estimations are
reported using robust standard errors clustered by firm and year to simultaneously control for cross-sectional and time-series dependencies (Petersen, 2009; Cameron
et al., 2011).
70
Table 6
Controversial business issues (Sins).
Panel A: Aggregate Sin Panel B: Individual Sins
Any “Sin” Model 1: Alcohol Model 2: Firearms

Coef. p-val Coef. p-val Coef. p-val
SIO −0.0943 (0.835) SIO −1.4212 (0.217) SIO 1.7495 (0.380)

LIO −1.4007 (0.003) LIO −0.9842 (0.349) LIO −4.3930 (0.013)
Tobin's q −0.0617 (0.383) All Ctrl/FE Yes All Ctrl/FE Yes
Size (Log of Sales) −0.0356 (0.900)
Squared Size 0.0166 (0.368)
Firm Age 0.0213 (0.000) Model 3: Gambling Model 4: Military
R&D/Assets −4.6017 (0.089) Coef. p-val Coef. p-val
Missing R&D −0.4612 (0.033) SIO −1.1086 (0.295) SIO 0.5032 0.4270
Stk Volatility −2.4862 (0.003) LIO −1.4131 (0.293) LIO −1.3017 0.0340
Leverage 0.2757 (0.597) All Ctrl/FE Yes All Ctrl/FE Yes
Modified z 0.0108 (0.842)
New Econ. −0.4191 (0.192)
HHI 0.3669 (0.277) Model 5: Nuclear Model 6: Tobacco
Chg in Sales (Ind.) 0.2958 (0.522) Coef. p-val Coef. p-val
Intercept 4.0974 (0.010) SIO 0.9989 0.3060 SIO −0.0954 0.9320
Pseudo-R2 0.2396 LIO −1.1764 0.2330 LIO 0.0430 0.9820
N 15,217 All Ctrl/FE Yes All Ctrl/FE Yes
This table presents results from logit estimations of controversial business involvement of firms (sins) using short-term and long-term institutional ownerships. The
models take the following functional form:
CSRt + 1 = S SIOt + L LIOt + Xt BX + t .
In Panel A, CSR is the indicator variable for firm involvement in any one of the controversial business areas including alcohol, firearms, gambling, military, nuclear,
and tobacco; in Panel B, CSR is the indicator variable for firm involvement in a particular one of the controversial business areas. SIO and LIO are vectors of
institutional ownership categorized into short-term and long-term. X is the matrix of all control variables and ϵ is the error vector; β and B are vectors of estimated
coefficients for the ownership vectors and controls matrix, respectively. All models include year and industry (based on the two-digit SIC code) fixed effects. All
estimations are reported using robust standard errors clustered by firm and year to simultaneously control for cross-sectional and time-series dependencies (Petersen,
2009; Cameron et al., 2011).
5. Robustness checks and extensions where Z and X are the matrices of the IVs and all other control vari-
ables, respectively. Λ and Θ are the vectors of estimated coefficients,
We run a battery of robustness checks to ensure the empirical va- and u in both equations are the vectors of first-stage errors. From the
lidity of our results. These include addressing endogeneity, reverse first-stage estimation, we retrieve the predicted ownership values and
causality, and autocorrelation concerns, as well as using alternative use them in place of the actual ownership variables for our second-stage
measures for our key variables and employing alternative econometric estimation:
specifications.
CSR t + 1 = 0 + S SIOt + L LIOt + Xt BX + t + 1, (7)
5.1. Endogeneity and reverse causality where CSRt+1, as before, denotes the lead CSR adjusted net score. SIO
and LIO present the predicted values of institutional ownership from
5.1.1. Two-stage estimations Eq. (6). X, as in the previous stage, is the matrix of all control variables.
From our base results of estimating one-year-ahead CSR scores using β and B are the estimated coefficients for the ownership vectors and
proxies for institutional investment horizons, we argue that the ex- controls matrix, respectively. As in our base models, we employ year
istence of certain types of ownership can have an impact on CSR en- and industry fixed effects and robust standard errors clustered by firm
gagement at the firm level. However, our results may suffer from en- and year. We present three sets of two-stage estimations outlined above
dogeneity, in that a certain set of firm characteristics may attract in Table 7. All panels present the second-stage lead CSR estimation in
particular types of investors, and that given that set of firm character- Model 1 and the first-stage SIO and LIO estimations in Models 1a and
istics, the firm simply more likely take CSR engagement to the direction 1b, respectively.
that we find. Further, long-term oriented institutions may simply be In Panel A, we use S&P 500 index membership as an instrument in
attracted to firms that are consistently improving their CSR (i.e., reverse predicting ownership (Aghion et al., 2013). This follows several studies,
causality). It is also possible that some unobserved idiosyncratic shocks including Appel, Gormley, and Keim (2016), Bena, Ferreira, Matos, and
simultaneously affect both the ownership and CSR variables in our Pires (2017), and Dyck et al. (2019), that utilize the variation of
sample, leading to spurious correlation. We now more formally address ownership by institutions following stocks being added to major stock
these potential concerns using instrumental variables (IVs) for owner- market indices. Further, we include industry median levels of short-
ship in a two-stage approach. term and long-term institutional ownerships as additional instruments
In all our two-stage models, we first estimate SIO and LIO as follows: (Choi & Sias, 2009; Michaely & Vincent, 2012). Prior studies have
S shown that institutions tend to herd within industries and, more im-
SIOt = + Zt S + Xt S + utS
0
portantly, the median ownership level of an industry is not likely re-
L
LIOt = 0 + Zt L + Xt L + utL , (6) lated to the CSR investments of particular firms within that industry.
71
Table 7 CSR estimations using the predicted SIO and LIO levels, we obtain re-
Endogenous investment choices and reverse causality. sults that are qualitatively similar to those reported earlier.
Model 1 Model 1a Model 1b In Panel B, we replace S&P 500 membership with pseudo-Russell
Stage 2 Lead CSR Stage 1 SIO Stage 1 LIO 1000 and 2000 memberships as instruments (Appel et al., 2016; Bird &
Karolyi, 2016; Crane, Michenaud, & Weston, 2016). This in part follows
Panel A: S&P 500 Coef. p-val Coef. p-val Coef. p-val
the empirical phenomenon that S&P 500 additions and deletions do not
Membership and
Ind Med IOs
exhibit as symmetric price effects as the annual Russell reconstitutions
SIO −0.5309 (0.000) do (e.g., Chang et al., 2015). To the extent that S&P 500 index mem-
LIO 0.7562 (0.000) bership may also be asymmetrically associated with ownership (e.g.,
S&P 500 −0.0058 (0.501) 0.0338 (0.000) see Pruitt & Wei, 1989), the use of Russell indices as an alternative set
Ind Med SIO 0.5615 (0.000) 0.0432 (0.098)
of instruments may offer an empirically cleaner capture of ownership.22
Ind Med LIO 0.0138 (0.613) 0.5233 (0.000)
R2 0.1206 Specifically, we mimic the Russell constituents as closely as we can
N 12,218 by computing the end-of-May market cap for the universe of CRSP/
Compustat firms to infer “pseudo-Russell” rankings each year. We then
Panel B: R1000/2000 Coef. p-val Coef. p-val Coef. p-val
take the first 1000 as firms that most likely be in the Russell 1000 and
Memberships and
Ind Med IOs
take the next 2000 as those that most likely fall in the Russell 2000.
SIO −0.5549 (0.000) These rankings will not be the exact constituents, but rather a closely
LIO 0.7881 (0.000) mimicked version of the Russell index memberships. As expected, we
Russell 1000 0.0575 (0.000) 0.0785 (0.000) show in Models 1a and 1b that the first-stage estimations exhibit
(mimicked)
stronger overall effects in determining ownership. These larger effects
Russell 2000 0.0740 (0.000) 0.0633 (0.000)
(mimicked) can be at least partially attributed to how sizes impact institutional
Ind Med SIO 0.5427 (0.000) 0.0308 (0.230) investment preferences (e.g., Bennett, Sias, & Starks, 2003; Falkenstein,
Ind Med LIO 0.0096 (0.737) 0.5207 (0.000) 1996; Yan & Zhang, 2009). In the second-stage lead CSR estimations,
R2 0.1146 we continue to obtain consistent results.
N 12,218
In Panel C, we use lagged short-term and long-term ownerships in
Panel C: Lagged IOs Coef. p-val Coef. p-val Coef. p-val our first-stage estimations of ownership. The use of lagged levels of
SIO −0.3000 (0.001) ownerships as instruments is based on the empirical fact that institu-
LIO 0.3788 (0.000) tional holdings are highly persistent (Gompers & Metrick, 2001). When
Lagged SIO 0.7397 (0.000) 0.1276 (0.000)
we employ these predicted ownerships in the second-stage lead CSR
Lagged LIO 0.0765 (0.000) 0.7406 (0.000)
R2 0.2704 estimation, we find that the estimated coefficients for SIO and LIO are
N 12,218 −0.3000 and 0.3788, respectively; both are significant at the 1% level.
That is, having controlled for investor persistence, we continue to see
This table presents results from two-stage estimations of Lead CSR. In all that our base results hold.
models, short-term and long-term ownerships are first estimated as In all IV regressions, we are able to validate that our models are
SIOt = S
+ Zt S + Xt S + utS ; LIOt = L
+ Zt L + Xt L + utL, adequately identified using the rk-statistic of Kleibergen and Paap
0 0
(2006). That is, the null of the LM test for underidentification is rejected
where Z and X are the matrices of the IVs and all other control variables, re- at conventional levels in all cases. Further, we are able to reject weak
spectively. Λ and Θ are the vectors of estimated coefficients, and u in both identification either using the rule of thumb suggested by Staiger and
equations are the vectors of first-stage errors. The first stage predictions of SIO Stock (1997, i.e., an F-stat of at least 10) or applying the critical values
and LIO are then used in the second stage lead CSR estimation: of Stock and Yogo (2005) to address a 5–10% largest tolerable relative/
estimator bias. The lowest Kleibergen-Paap rk Wald F from our two-
CSRt + 1 = + S SIOt + L LIOt + Xt BX + t + 1,
0
stage estimations is, for instance, 70.89 from Panel A. Finally, the
where SIO and LIO present the predicted values of institutional ownership. X is
Hansen (1982) J-statistics are 0.585 and 0.480 for Panels A and B, re-
the matrix of all control variables. β and B are the estimated coefficients for the spectively, both far from rejection of non-overidentification. Alto-
ownership vectors and controls matrix, respectively. Year and industry fixed gether, we are statistically confident about the appropriateness of our
effects and robust standard errors clustered by firm and year are employed instrument sets.
(Petersen, 2009; Cameron et al., 2011). Coefficient estimates of control vari- To further ensure robustness in our first-stage modeling of institu-
ables are omitted due to space concerns. Panel A uses S&P 500 index mem- tional preferences, we also use (i) contemporary CSR instead of the
bership (Aghion et al., 2013) and industry median short-term and long-term lagged and (ii) a decomposition of the contemporary CSR into a trend
institutional ownerships (Choi & Sias, 2009; Michaely & Vincent, 2012) as in- and a lagged CSR component and use them as our instruments. The first
struments in predicting firm-level ownership; Panel B replaces S&P 500 mem- alternative alleviates the concern that investors may be able to observe
bership with the mimicked Russell 1000 and Russell 2000 memberships (Chang
the contemporary CSR levels when making investment decisions. The
et al., 2015); Panel C endogenizes investor preferences for existing firm CSR
second alternative addresses the logic that investors may pay more
scores and address the empirical persistence of institutional holdings (Gompers
& Metrick, 2001). All panels present the second-stage lead CSR estimation in attention to the most recent trends in firm CSR engagement (e.g.,
Model 1 and the first-stage SIO and LIO estimations in Models 1a and 1b, re- whether it is increasing or decreasing) than to lagged scores. Our results
spectively. remain qualitatively unchanged, therefore not reported. With these
results, we continue to show that the channel through which investor
heterogeneity affects firm CSR engagement remains at work, even after
From the first-stage estimations of ownership, we see that S&P 500

membership is positively related to LIO and not statistically associated 22
Prior literature suggests that additions to and deletions from the S&P 500
with SIO. This is consistent with Boone and White (2015), who examine index do not exhibit symmetric effects. H. Chen, Noronha, and Singal (2004),
the ownerships of institutional investors surrounding the Russell index for instance, find permanent price increases for stocks of firms added to the S&P
reconstitutions and find that the increase in ownership is primarily 500 index, but no permanent decline for deleted firms. In contrast, studies that
driven by long-term and diversified investors. In the second-stage lead use the Russell reconstitution, such as Chang et al. (2015), observe symmetric
price effects.
72
empirically controlling for reverse causality as much as we can.23 serial correlation through Newey-West errors with three lags (Fama &
In untabulated results, we use a dividend paying dummy as an in- MacBeth, 1973; Newey & West, 1987).24 We report this estimation in
strument for ownership (Aggarwal, Erel, Ferreira, & Matos, 2011). The Model 3 of Panel C. As shown, the estimated coefficients for SIO and
first-stage results are consistent with findings presented in prior lit- LIO are −0.2456 and 0.2749, respectively, both remain economically
erature that domestic institutional ownership is positively associated and statistically significant.
with dividend-paying stocks and that firms with shorter institutional Finally, in untabulated results, we re-estimate our base model
investment horizons more likely replace dividends with repurchases in (Model 2 in Table 3) and two-stage least squares models (Table 7) by
structuring payout policies (Ferreira & Matos, 2008; Gaspar, Massa, employing HAC standard errors that are robust to arbitrary hetero-
Matos, Patgiri, & Rehman, 2012). Importantly, we continue to find that skedasticity and within panel autocorrelated disturbances (clustering
lead CSR scores are negatively associated with SIO and positively as- on firm) and cross-panel autocorrelated disturbances that disappear
sociated with LIO. after two lags (clustering on time combined with kernel-based HAC)
(Cameron et al., 2011; Thompson, 2011). We obtain standard errors for
5.2. Autocorrelation and cross-sectional robustness SIO and LIO that are even lower than those reported earlier.
To sum up, in all attempts described above, we continue to find a
Since CSR scores are highly persistent, it is worth noting that our significantly negative (positive) association between firm SIO (LIO) and
estimations employ two-way clustered errors that simultaneously con- CSR scores. Thus, our key finding remains qualitatively unchanged. The
trol for cross-sectional and time-series dependencies (Petersen, 2009; results from our Fama-MacBeth estimation also ensure the cross-sec-
Cameron et al., 2011). Further, the year fixed effects that we in- tional robustness of our study. In untabulated results, we find that SIO
corporate in all models also alleviate problems arising from time-series (LIO) yields a negative estimated coefficient in all ten out of the ten
dependence. In this section, we continue to exert effort in addressing cross sections in our sample, in which six (nine) are statistically sig-
autocorrelation concerns in several ways. nificant at the 5% level.
First, when we include a lagged CSR term into our CSR estimation,
as shown in Model 1 of Table 8, our results remain qualitatively un- 5.3. Alternative measures for key variables
changed. Second, as an attempt to address both reverse causality and
autocorrelation at the same time, we employ a simultaneous estima- 5.3.1. Institutional ownership
tions system, where CSR scores and the two types of institutional When computing our main explanatory variables, SIO and LIO, we
ownerships are jointly determined (Zellner, 1962). The specification is apply three rules that factor into our presentation of results. First, we
as follows: categorize total institutional ownership into short-term and long-term
CSRt = C
+ S C C C based on the annual investor level churn medians. Second, the investor
Ct 1 CSRt 1 + S SIOt + L LIOt + X t BC + t
level churn rates are based on Yan and Zhang (2009). Third, to combine
0
S S S S
SIOt = + Ct CSR t + Ct 1 CSRt 1 + X t BS +
0 t
quarterly data from TFN with annual data from KLD, we use only
L L S L
LIOt = 0 + Ct CSR t + Ct 1 CSRt 1 + X t BL + t, (8) ownership data from December of each calendar year. To ensure that
our results are not purely driven by these methodological choices, we
where CSR is the vector of adjusted CSR score and SIO and LIO are
apply different calculations, cuts, and filters to each of these aspects
vectors of institutional ownership categorized into short-term and long-
when computing our empirical proxy for institutional investment hor-
term. X is the matrix of all control variables and ϵ is the error vector; β
izons. We also employ theoretically different measures of ownership.
and B are the estimated coefficients for the key variables and controls
These results are reported in Table 9.
matrix, respectively. We report the results in Panel B of Table 8. Model
In Model 1, instead of using ownerships from December, we try
2 estimates CSR scores; Model 2a estimates SIO; and Model 2b estimates
using the March, June, and September figures. We also use the mean of
LIO. In Model 2, the estimated coefficient for lagged CSR is 0.7951,
all four ownership levels during the year. Doing so allows us to alleviate
indicating that average firm-level engagement in corporate social re-
any seasonality-related concerns about our results. We report only re-
sponsibility is highly persistent. Importantly, the estimated coefficients
sults from using the mean ownerships in Model 1. As shown, the esti-
for SIO and LIO in the same model are −0.0398 and 0.0633, which are
mated coefficient for SIO (LIO) remains significantly negative (posi-
both significant at conventional levels.
tive). Results from using the ownerships from the other three months
Third, we also estimate lead CSR using a Fama-MacBeth estimation,
are qualitatively similar. We omit reporting them due to space con-
where each cross section is modeled separately while correcting for
cerns.
In Model 2, instead of cutting the TFN investor universe based on
23
As an additional attempt to address reverse causality, we also employ a churn medians, we do so based on terciles. As a result, there will be an
simultaneous equations system, where CSR scores and the two types of in- additional medium-term institutional ownership (MIO) at the firm level
stitutional ownerships are jointly determined in the following specification: in our model, presenting those between the short-term and the long-
C C C C
term. While doing so seems to bias towards finding results if the CSR-
CSRt = + SIOt + LIOt + Xt BC +
0 S L t
horizon relation is empirically monotonic, we will be able to observe
S S S
SIOt = 0
+ C
CSRt + X t BS + t such monotonicity if there indeed is one. If, oppositely, we observe any
strictly non-monotonic relation (e.g., U- or inverted U-shape), our ar-
L L L
LIOt = 0 + C CSR t + Xt B L + t
gument that longer investment horizons of institutions promote CSR
where CSR is the vector of adjusted CSR score and SIO and LIO are vectors of
cannot be generalized. From the estimation, not only do we find results
institutional ownership categorized into short-term and long-term. X is the
that are consistent with those reported earlier, we also confirm that the
matrix of all control variables and ϵ is the error vector; β and B are the esti-
mated coefficients for the key variables and controls matrix, respectively. The CSR-horizon relation is at least weakly monotonic. Although the point
estimated coefficients for CSR in the SIO and LIO equations are −0.0096 and estimate of the MIO coefficient is slightly lower then that of the LIO
0.0141, respectively. While both are strongly significant, the estimated coeffi- coefficient, a test of equality between the two yields an F-value of only
cients for SIO and LIO in the CSR equation remains statistically and econom- 0.0007, suggesting that they are not statistically different.
ically significant at −0.1568 and 0.2552, respectively. Given that these results In Model 3, we replace Yan and Zhang’s (2009) investment horizon
are qualitatively similar to those in the simultaneous equations system with measure, which is based on the minimum of aggregate buy and sell in
lagged CSR scores incorporated as a part of our effort to address autocorrelation
in the dependent variable (see Panel B of Table 8), we do not tabulate them to
24
conserve space. Our results are robust to other choices of lags.
73
Table 8
Autocorrelation and cross-sectional robustness.
Panel A: CSR Est. Panel B: Simultaneous Est. of CSR and SIO/LIO Panel C: FM Reg
Model 1 Model 2 Model 2a Model 2b Model 3
CSR CSR SIO LIO CSR
Coef. p-val Coef. p-val Coef. p-val Coef. p-val Coef. p-val
SIO −0.0435 (0.042) −0.0398 (0.044) −0.2456 (0.000)

LIO 0.1133 (0.002) 0.0633 (0.003) 0.2749 (0.000)
CSR −0.0057 (0.066) 0.0080 (0.005)
Lagged CSR 0.6297 (0.000) 0.7951 (0.000) −0.0071 (0.028) 0.0070 (0.018)
Tobin's q 0.0263 (0.000) 0.0131 (0.000) 0.0106 (0.000) −0.0046 (0.016) 0.0493 (0.000)
Size (Log of Sales) −0.0675 (0.000) −0.0496 (0.000) 0.0823 (0.000) 0.0647 (0.000) −0.0301 (0.548)
Squared Size 0.0098 (0.000) 0.0060 (0.000) −0.0054 (0.000) −0.0037 (0.000) 0.0088 (0.187)
Firm Age 0.0005 (0.258) 0.0002 (0.334) −0.0013 (0.000) 0.0006 (0.000) 0.0007 (0.144)
R\&D / Assets 0.2192 (0.010) 0.0838 (0.073) 0.1601 (0.000) 0.0795 (0.029) 0.6595 (0.001)
Missing R\&D −0.0258 (0.059) −0.0080 (0.295) −0.0208 (0.002) −0.0030 (0.649) −0.0538 (0.031)
Stk Volatility −0.1926 (0.036) −0.0106 (0.820) −0.0386 (0.245) −0.3743 (0.000) −0.6756 (0.000)
Leverage 0.0291 (0.364) 0.0083 (0.646) 0.1044 (0.000) 0.0072 (0.622) 0.0224 (0.508)
Modified z 0.0013 (0.678) 0.0036 (0.026) 0.0021 (0.169) 0.0069 (0.000) 0.0093 (0.143)
New Econ. 0.0646 (0.000) 0.0345 (0.001) −0.0014 (0.875) −0.0226 (0.007) 0.1042 (0.005)
HHI 0.0240 (0.388) 0.0116 (0.463) −0.0353 (0.009) −0.0043 (0.757) 0.0669 (0.001)
Chg in Sales (Ind.) −0.0923 (0.154) −0.1137 (0.017) 0.0275 (0.051) −0.0113 (0.403) Omit
Intercept −0.0951 (0.681) −0.1008 (0.296) 0.0510 (0.446) 0.1234 (0.000) −0.9302 (0.000)
R2 0.5045 0.6670 0.1924 0.2340 0.1978
N 12,168 14,029 14,029 14,029 15,217
This table further addresses serial correlation concerns. Panel A reports the results from the base CSR model when including a lagged CSR term (analogous to Model 2
in Table 3); Panel B includes a lagged CSR term in a simultaneous estimations system, where CSR scores and short-term and long-term ownership levels are jointly
determined (Zellner, 1962):
C S C C C
CSRt = 0
+ Ct 1
CSRt 1 + S
SIOt + L
LIOt + Xt BC + t
S S S S
SIOt = 0
+ Ct
CSRt + Ct 1
CSRt 1 +X t BS + t
L L S L
LIOt = 0
+ Ct
CSRt + Ct 1
CSRt 1 + X t BL + t,
where CSR is the vector of adjusted CSR score and SIO and LIO are vectors of institutional ownership categorized into short-term and long-term. X is the matrix of all
control variables and ϵ is the error vector; β and B are the estimated coefficients for the key variables and controls matrix, respectively. All models include year and
industry (based on the two-digit SIC code) fixed effects and clustered errors. In Panel B, Model 2 estimates CSR scores, Model 2a estimates SIO, and Model 2b
estimates LIO. Panel C presents results obtained under the Fama-MacBeth setting, where each cross section is modeled separately while correcting for serial
correlation (Fama & MacBeth, 1973; Newey & West, 1987)
calculating churn rates, with that of Gaspar et al. (2005, 2012), which indexers and dedicated investors be positively related to CSR. Further,
uses the sum of aggregate buys and sells. In short, Eq. (1) is replaced by if CSR is related to institutional investment horizons through price re-
the following calculation: sponses, we should also observe that any effect be the weakest with
dedicated investors since they are the least sensitive to changes in
|Nj, k, t Pj, t Nj, k, t 1 Pj, t 1 Nj, k, t 1 Pj, t | earnings. We find what we expect: The estimated coefficients for BTIO,
CRGMM
k,t .
1
j J
(Nj, k, t Pj, t + Nj, k, t 1 Pj, t 1 ) (9) BQIO, and BDIO are −0.2789, 0.1795, and 0.0449 in estimating the
2
lead adjusted net CSR scores, with only the former two being statisti-
The results are qualitatively similar to those reported earlier, and thus cally significant at conventional levels.
continues to provide support to our argument.
Finally, in Model 4, we replace short-term and long-term owner- 5.3.2. Corporate social responsibility
ships with transient, quasi-indexer, and dedicated institutional owner- Our main analyses surround the use of aggregated CSR scores from
ships (BTIO, BQIO, and BDIO, respectively) derived from the k-means the KLD database. To ensure robustness, we re-estimate our main re-
clustering approach of Bushee (1998). By definition, transient investors sults using variations of and alternatives to the aggregated scores in this
hold highly diversified portfolios with high turnover and pursue mo- section. We report these results in Table 10.
mentum strategies following earnings news; quasi-indexers hold highly In Model 1, we use the list of “100 Best Companies To Work For In
diversified portfolios with low turnover and use long-term buy-and- America (BCW)” as an alternative measure for CSR (Edmans, 2011,
hold strategies; and dedicated investors have high concentration and 2012). The list of firms in the BCW is published every January in the
low turnover in equity ownership, with very little trading sensitivity to Fortune Magazine by the Great Place To Work (GPTW) Institute, which
earnings. is headquartered in San Francisco, CA. To make the interpretations of
Bushee (2001) finds that, when decomposing firm value into ex- coefficients consistent between the BCW and the CSR, we calculate a
pected near-term earnings and expected long-term terminal value, score that is based on the annual rankings of the 100 Best. Specifically,
short-term oriented (i.e., transient) investors are positively associated the firm ranked No. 1 on the list would have a score of 100, the No. 2
with the former and negatively associated with the latter. This indicates firm would have a score of 99, so on and so forth. Firms not in the list
that, consistent with our key hypothesis, a short-term-focused institu- have BCW scores of zero. We present results from this ordered logit in
tional ownership base may pressure managers into a short-term focus Model 1. Once again, we find results that are consistent with our ar-
through institutions myopically pricing firms by overweighting short- guments.
term earnings potential and underweighting long-term earnings po- In untabulated results, we employ two variations of the BCW score.
tential. Directly capitalizing on this earlier finding, we would expect One is a simple BCW Dummy that takes the value of 1 if a firm is one of
that transient ownership be negatively related to CSR and that quasi- the 100 Best for a given year, and 0 otherwise. The other is a BCW
74
Table 9
Alternative ownership measures.
Model 1 Model 2 Model 3 Model 4
Mean IO Tercile Cuts Gaspar Hor. Bushee
Coef. p-val Coef. p-val Coef. p-val Coef. p-val
SIO Mean −0.1744 (0.007)

LIO Mean 0.2692 (0.000)
SIO (Tercile) −0.2963 (0.000)
MIO (Tercile) 0.1374 (0.021)
LIO (Tercile) 0.1590 (0.038)
SIO (GMM) −0.1639 (0.014)
LIO (GMM) 0.1592 (0.010)
BTIO −0.2789 (0.000)
BQIO 0.1795 (0.000)
BDIO 0.0449 (0.724)
Tobin's q 0.0453 (0.000) 0.0439 (0.000) 0.0432 (0.000) 0.0452 (0.000)
Size (Log of Sales) −0.0754 (0.141) −0.0768 (0.138) −0.0775 (0.132) −0.0776 (0.130)
Squared Size 0.0121 (0.019) 0.0121 (0.021) 0.0123 (0.018) 0.0123 (0.017)
Firm Age 0.0006 (0.488) 0.0009 (0.353) 0.0009 (0.340) 0.0009 (0.323)
R&D/Assets 0.5222 (0.001) 0.5125 (0.001) 0.5121 (0.001) 0.5118 (0.001)
Missing R&D −0.0620 (0.016) −0.0620 (0.015) −0.0618 (0.015) −0.0623 (0.015)
Stk Volatility −0.4750 (0.000) −0.4613 (0.001) −0.4810 (0.001) −0.4608 (0.001)
Leverage 0.0442 (0.396) 0.0370 (0.474) 0.0414 (0.422) 0.0387 (0.458)
Modified z 0.0143 (0.018) 0.0141 (0.018) 0.0143 (0.016) 0.0138 (0.021)
New Econ. 0.1105 (0.006) 0.1095 (0.006) 0.1086 (0.006) 0.1099 (0.006)
HHI 0.0613 (0.214) 0.0621 (0.209) 0.0623 (0.208) 0.0598 (0.229)
Chg in Sales (Ind.) 0.0047 (0.925) 0.0065 (0.899) 0.0000 (1.000) 0.0061 (0.902)
Intercept −0.8047 (0.000) −0.7891 (0.000) −0.7926 (0.000) −0.8096 (0.000)
R2 0.1821 0.1812 0.1804 0.1813
N 15,217 15,217 15,217 15,217
This table presents results from lead CSR (corporate social responsibility) net adjusted score estimations using variations of ownership measures. All models include
year and industry (based on the two-digit SIC code) fixed effects and are reported using robust standard errors clustered by firm and year to simultaneously control
for cross-sectional and time-series dependencies (Petersen, 2009; Cameron et al., 2011). Model 1 employs mean ownership for each calendar year in place of the
December figures; Model 2 uses a tercile cut for ownership in place of a median cut; Model 3 uses Gaspar et al.’s (2005) horizon measures instead of Yan and Zhang’s
(2009); Model 4 uses Bushee’s (1998) classification of transient, quasi-indexer, and dedicated institutional investors (BTIO, BQIO, and BDIO, respectively).
Accumulated Dummy, which takes the value of 1 if a firm has been one actual score (and therefore the exact rankings). For the BCW
of the “100 Best” during or prior to the year of a particular observation, Accumulated Dummy, we continue to see consistent results.
and 0 otherwise. For the BCW Dummy, we find almost identical results Recent CSR literature stresses the importance of using more focused
with those produced by Model 1. Further estimation of a Heckman measures of the social aspect of firms, rather than simply aggregating
selection model confirms that, while SIO and LIO are associated with the counts of all strengths and all concerns at a given firm-year, im-
the likelihood of being in the 100 Best, they do not have an effect on the plicitly treating all items equal (Strike, Gao, & Bansal, 2006b; Walls
Table 10
Alternative CSR measures.
Model 1 Model 2 Model 3 Model 4
BCW Ind-Adj. CSR Material CSR SIC4
Coef. p-val Coef. p-val Coef. p-val Coef. p-val

SIO −3.5821 (0.007) −0.1566 (0.012) −0.0111 (0.010) −0.1478 (0.015)
LIO 2.1560 (0.018) 0.2562 (0.000) 0.0095 (0.026) 0.1916 (0.000)
Tobin's q 0.5732 (0.000) 0.0459 (0.000) 0.0010 (0.051) 0.0362 (0.000)
Size (Log of Sales) 5.3102 (0.000) −0.0757 (0.139) −0.0225 (0.000) −0.0874 (0.084)
Squared Size −0.2618 (0.000) 0.0120 (0.019) 0.0025 (0.000) 0.0137 (0.009)
Firm Age −0.0300 (0.003) 0.0007 (0.407) 0.0003 (0.000) 0.0017 (0.060)
R&D/Assets 3.8454 (0.146) 0.4987 (0.001) −0.0710 (0.000) 0.2248 (0.126)
Missing R&D −0.9650 (0.005) −0.0602 (0.018) −0.0072 (0.000) −0.0639 (0.008)
Stk Volatility −4.8922 (0.019) −0.5724 (0.000) −0.0140 (0.085) −0.4508 (0.001)
Leverage −2.3938 (0.042) 0.0339 (0.527) 0.0096 (0.018) 0.0775 (0.132)
Modified z 0.0473 (0.759) 0.0120 (0.057) 0.0000 (0.901) 0.0208 (0.001)
New Econ. 1.5301 (0.031) 0.1084 (0.007) 0.0049 (0.041) 1.1490 (0.098)
HHI 1.4449 (0.086) 0.0632 (0.201) 0.0136 (0.004) 0.0074 (0.902)
Chg in Sales (Ind.) −1.7594 (0.007) 0.0486 (0.151) −0.0046 (0.318) 0.0014 (0.981)
Intercept Omit −0.0885 (0.646) 0.0927 (0.025) −0.8047 (0.000)
R2 0.2202 0.1346 0.4398 0.2855
N 15,217 15,217 15,217 15,217
This table presents results from estimations of certain transformations of lead CSR (corporate social responsibility) using short-term and long-term institutional
ownerships (SIO and LIO, respectively). Model 1 employs BCW scores calculated using the list of “100 Best Companies To Work For In America” published by the
Fortune Magazine as an alternative measure for CSR (Edmans, 2011, 2012); Model 2 estimates lead industry-adjusted CSR net adjusted scores; Model 3 estimates
material CSR net scores calculated based on Khan et al.’s (2016) materiality mapping; Model 4 estimates lead CSR net adjusted scores while incorporating 4-digit SIC
industry fixed effects. All models include year and industry (based on the two-digit SIC code, other than Model 4) fixed effects. All estimations are reported using
robust standard errors clustered by firm and year to simultaneously control for cross-sectional and time-series dependencies (Petersen, 2009; Cameron et al., 2011).
75
et al., 2012). We address this in two ways: First, we use industry-ad- coefficient estimate for the interaction between LIO and CSR becomes
justed CSR scores, based on the two-digit SIC code, instead of the un- positively significant, while the interaction between SIO and CSR re-
adjusted scores to estimate lead CSR using SIO and LIO. As shown in mains insignificant.
Model 2 of Table 10, the estimated coefficients are qualitatively similar The sample size differences across our BHR models warrant further
to those reported in the main analyses. discussion. Since we require firms to have full 1–6 years of returns in
Second, we follow Khan et al. (2016) to create a CSR score that only order to calculate the BHRs, the sample sizes in our returns regressions
consists of “material” items. The material items are defined by the vary from model to model. Due to the survivorship, however, we expect
Sustainability Accounting Standards Board (SASB). To create the ma- that estimating cumulative returns using the maximum possible number
teriality score, we first map each of the SIC two-digit codes of the firms of observations in each regression would bias us against finding results,
in our sample to one of the SASB sectors based on their Sustainable if any; since by doing so we are including in our shorter-window returns
Industry Classification System (SICS). These SASB SICS sectors include estimations firms that do not survive long enough to make it into the
(i) health care, (ii) financials, (iii) technology and communications, (iv) longer-window returns estimations. One may view these firms as worse-
non-renewable resources, (v) transportation, (vi) services, (vii) resource performing ones in the sample.
transformation, (viii) consumption, (ix) renewable resources and al- To confirm this, we re-estimate the same buy-and-hold returns re-
ternative energy, and (x) infrastructure. Once mapped, we are then able gressions for the 1- through 6-year windows using a fixed sample of
to pin down individual KLD items that are material to each of the two- 7423 firm-year observations (1407 unique firms), which is the sample
digit SIC codes using the Materiality Map system of SASB.25 The results that makes it into the original cumulative 6-year BHR regression. We
from estimating the material CSR scores are reported in Model 3, where find that our results not only qualitatively hold, but also stronger: The
we continue to find similar results. adverse effect from the interaction between SIO and CSR on returns still
In Model 4, we further address the importance of differences be- dies out relatively quickly, while the beneficial effect from the inter-
tween industries by employing fixed effects based on the four-digit SIC action between LIO and CSR kicks in much earlier and remains sig-
code. Our results, once again, remain qualitatively unchanged. Using nificant for at least six years cumulatively. These results using the fixed
the three-digit SIC code does not alter our estimations. sample are reported in Panel B of Table 11.
In untabulated results, we analyze buy-and-hold abnormal returns
(BHAR) instead of buy-and-hold raw returns. We employ two different
5.4. Additional discussions benchmarks, the first is industry median returns and the second is the
value-weighted CRSP returns. Both sets of regressions produce overall
5.4.1. Buy-and-hold returns similar results: The interaction between SIO and CSR is negatively as-
Following our results presented thus far, a natural question to ask is: sociated with BHARs during the first few years and then fades away; the
Do investments in CSR create value for shareholders? Specifically, how interaction between LIO and CSR is positively associated with BHARs
do stocks perform in the presence of higher CSR scores, coupled with a during the later years during our estimation window. In sum, we find
larger existing SIO or a larger existing LIO? For short-term institutions what we expect: The short-term effect of CSR with SIO on prices is
to discourage CSR endeavors, the short-term effect of CSR on prices negative; the long-term effect of CSR with LIO on prices is positive.
should be negative for firms with high proportions of short-term in- A caveat of our inference is that it is difficult to tell whether the
vestors; for long-term institutions to encourage CSR, the long-term ef- long-term effects have anything to do with the CSR points of such a long
fect should be positive for firms with high proportions of long-term time ago. We acknowledge this potential shortcoming: despite our best
investors. We examine this conjecture in our sample of firms. efforts, other factors such as the empirical persistence of institutional
In Panel A of Table 11, we perform buy-and-hold return regressions ownership may be at least partially responsible for the results.
for up to six years following an institutional ownership-CSR combina- However, if merely persistence or something else other than the own-
tion. In each model, we regress individual stock level buy-and-hold ership variables themselves were to drive the results, we should also
returns on the interactions of short-term and long-term institutional observe that SIO exhibit a long-term effect, which we do not.
ownership (SIO and LIO as of time t) with lead CSR net adjusted score
(CSR as of time t + 1), the lead CSR score itself and firm level de-
5.4.2. Does investment horizon matter in bad times ?
terminants of stock returns. The firm level determinants of returns in-
As stated in the quote from CalPERS Beliefs at the very beginning of
clude firm size (market capitalization), Tobin's q, dividend yield, share
this paper, some institutional investors adapt a long-term view, toler-
price, and lagged returns (Brennan, Chordia, & Subrahmanyam, 1998).
ating a near-term volatility. Our analysis so far suggests that LIO (SIO)
Due to space concerns, we only report coefficient estimates for the key
is associated with higher (lower) CSR. However, a related question is
variables. The dependent variables in Models 1 through 6 are logs of
whether the effect of investor horizon still persists in bad times, i.e.,
buy-and-hold returns from time t to t + τ, τ ∈{1,2,3,4,5,6}, respec-
does LIO (SIO) still encourage (discourage) investments in CSR during
tively. All models are reported using robust standard errors clustered by
bad times?
firm and year to simultaneously control for cross-sectional and time-
To answer this question, we estimate the following regression
series dependencies (Petersen, 2009; Cameron et al., 2011) and include
model:
industry fixed effects (based on the two-digit SIC code).
In Model 1, we see that the interaction between SIO and CSR carries CSR t + 1 = 0 + d( d
S SIOt + d
L LIOt ) + Xt B + t ,
a significantly negative coefficient in explaining the one-year buy-and- d {dgood, dbad}
hold return, while the interaction between LIO and CSR, as well as CSR
on its own, do not show up significantly. In Models 2 albeit exhibiting where dgood and dbad are dummy variables that indicate good and bad
the expected directions, we show that none of the estimated coefficients times, respectively.
are significant. This insignificance seems to hold up to year t + 4. For We use two empirical proxies for bad times. The first is at the
the buy-and-hold returns from t to t + 5 and from t to t + 6, the macro-level, i.e., the business cycle contraction periods published by
the National Bureau of Economic Research (NBER).26 The second is at
the micro-level, where we determine whether a firm, at a given year,
25
The detailed descriptions for each of the SASB SICS sectors can be found has a higher stock volatility than that of its industry median. Under
online at www.sasb.org/approach/sics; the SASB Materiality Map can be found both proxies, the LIO coefficient remains positive and significant during
at www.sasb.org/materiality/sasb-materiality-map. Khan et al. (2016) provide
the materiality mapping for six of the ten SASB sectors in their study. We hand-
26
collect the mapping of the remaining four. www.nber.org/cycles.html
76
Table 11
Buy-and-Hold returns.
Panel A: Varying samples Panel B: Fixed sample
Model 1: Year t+1 Coef. p-val Coef. p-val

SIO t times CSR t+1 −0.1942 (0.043) Obs 12,229 −0.2075 (0.018) Obs 7423
LIO t times CSR t+1 0.0512 (0.360) Firms 2279 0.1201 (0.053) Firms 1407
All CTRL/FE Yes R2 0.1334 Yes R2 0.1390
Model 2: Years t+1 to t+2 Coef. p-val Coef. p-val

SIO t times CSR t+1 −0.1002 (0.447) Obs 12,229 −0.1108 (0.315) Obs 7423

SIO t times CSR t+1 0.0824 (0.553) Obs 10,939 0.0271 (0.845) Obs 7423

SIO t times CSR t+1 0.0640 (0.652) Obs 9877 0.1017 (0.510) Obs 7423

SIO t times CSR t+1 −0.0681 (0.639) Obs 8981 −0.0641 (0.658) Obs 7423

SIO t times CSR t+1 −0.0291 (0.843) Obs 7423 −0.0291 (0.843) Obs 7423
This table presents results from regressions of individual stock returns on the interactions of short-term and long-term institutional ownership (SIO and LIO as of time
t) with lead CSR net adjusted score (CSR as of time t + 1), the score itself, and firm level determinants of stock returns including size (market capitalization), Tobin's
q, dividend yield, share price, and past one-year stock returns (Brennan et al., 1998). Panel A reports results from varying samples based on data availability for each
respective regression; Panel B reports results from a fixed sample of observations that have the full six years of lead returns (i.e., the 7423 observations in Model 6).
Only coefficient estimates for the key variables are reported due to space concerns. The dependent variables in Models 1 through 6 are logs of buy-and-hold returns
from time t to t + τ, τ ∈{1, 2, 3, 4, 5, 6}, respectively. All models are reported using robust standard errors clustered by firm and year to simultaneously control for
cross-sectional and time-series dependencies (Petersen, 2009; Cameron et al., 2011) and includes industry fixed effects (based on the two-digit SIC code).
both good and bad times. In particular, when we use recessions to de- maximizing the short-term financial performance of the organization. It
fine bad times, the good time LIO coefficient is 0.2516 and the bad time is well documented that firms often suffer from a short-term bias due to
LIO coefficient is 0.2534 (both are statistically significant at the 1- poorly structured managerial incentives, recency bias, career concerns,
percent level). When we use high stock volatility to define bad times, etc. As a result, short-termism implies maximization of short-term
the estimated good and bad time coefficients are 0.3324 and 0.1642, profits, often at the expense of other stakeholders. Another explanation
respectively (both are statistically significant at the 1-percent level). for the lack of consensus is the temptation to interpret the stakeholder
Therefore, LIO has a positive and significant effect on CSR in both good theory as firms facing conflicting and competing interests of various
and bad times. stakeholders.
As for SIO, its' negative effect on CSR is not consistently significant. Jensen (2002) clarifies the proper relation between value max-
For example, when we use recessions to define bad times, the effect of imization and stakeholder theory by proposing a new corporate ob-
SIO on CSR is negative but not significant during bad times (−0.1227), jective function that he calls “enlightened value maximization”. En-
and negative and significant during good times (−0.1685). Similarly, lightened value maximization adds the simple specification that the
when we use high stock volatility to define bad times, the effect of SIO objective function of the firm is to maximize the total long-term market
is negative but not significant during bad times (−0.0684), and nega- value of the firm. In this way, stakeholder theorists can see that al-
tive and significant during good times (−0.2387). On balance, the though shareholders are not some special constituency that ranks above
negative effect of SIO on CSR is driven by the good times. These results all others, long-term stock value is an important determinant (along
are not tabulated, but are available upon request. with the value of debt and other instruments) of total long-term firm
value. They would see that long-term value creation gives management
6. Conclusion a way to assess the tradeoffs that must be made among competing
constituencies, and that it allows for principled decision making in-
Given the growing interest in addressing the problem of short-ter- dependent of the personal preferences of managers and directors.27
mism, the perceived lack of consensus in theoretical perspectives and
the lack of clear empirical evidence, we examine the relationship be-
tween institutional ownership horizon, CSR and shareholder value in 27
Benabou and Tirole (2010) also point out that long-term perspective and
this paper. We first address the perceived lack of consensus in theore-
delegated philanthropy notions of CSR predict a positive correlation between
tical perspectives (i.e., agency vs. stakeholder perspectives), and discuss CSR and profits, while insider-initiated corporate philanthropy predicts the
how recent studies (e.g., Jensen, 2002; Benabou & Tirole, 2010) help reverse. In this latter interpretation of CSR, corporate prosocial behavior is not
reconcile the conflicting views. In particular, the perceived lack of motivated by stakeholders demands of willingness to sacrifice money for a good
consensus between value maximization and stakeholder theory arises cause, but rather reflects management's or the board members' own desires to
due to the temptation to consider value simply as a matter of engage in philanthropy.
77
Thus, the enlightened value maximization theory solves the problems more than 1100 investment management professionals conducted by
arising from the multiple objectives that accompany traditional stake- the CFA Institute revealed that nearly 80% of investment managers
holder theory by giving managers a clear way to think about and make have less than half of their compensation based on long-term perfor-
the tradeoffs among corporate stakeholders. mance measures.28 There is also an implication for monitoring invest-
Next, we empirically examine the link between institutional own- ment managers in that they need to be held accountable for deviating
ership horizon and firm policies on CSR, given the recent evidence on too far from their stated objectives in regards to portfolio turnover.
how institutional investors have the ability to impact corporate policies. Recent studies show that investment managers with low portfolio
While there are some existing studies on this topic, they have produced turnover and concentrated positions outperformed managers without
fragmented and contradictory results, making it difficult to advance the these two combined characteristics by a statistically significant
conversation in the academic community. For instance, Graves and 2.3%–3.5 % per year over 20-plus year observation periods (e.g.,
Waddock (1994) find no relationship between the percentage of shares Harford et al., 2018; Cremers & Pareek, 2016). Pension funds in par-
institutionally owned and CSR. In contrast, Johnson and Greening ticular, with their high political profile, need to rethink their invest-
(1999) and Neubaum and Zahra (2006) find a strong, positive link. We ment strategies and governance. If implementation is an issue despite
address this situation by adopting a comprehensive empirical approach having good policy intentions and long-term goals, the increasing
that improves on the previous studies. In particular, we improve on the availability of ESG data from a large number of data providers (e.g.,
previous studies by employing a superior horizon measure based on Bloomberg, MSCI, Thomson Reuters, Sustainalytics, CDP) makes im-
trading behavior rather than legal type, accounting for multi- provements in performance monitoring and measurement systems
dimensionality of CSR (strengths vs. concerns dimension, as well as CSR possible.
categories such as community relations, employee relations, human
rights, diversity, environment and product), and accounting for en- References
dogeneity, reverse causality and autocorrelation.
We find that firms with higher percentage of long-term (short-term) Aggarwal, R., Erel, I., Ferreira, M., & Matos, P. (2011). Does governance travel around the
investors have higher (lower) CSR net scores. However, because our world? Evidence from institutional investors. Journal of Financial Economics, 100,
154— 181.
comprehensive approach accounted for multidimensionality of CSR, we Aghion, P., Van Reenen, J., & Zingales, L. (2013). Innovation and institutional ownership.
have made some new discoveries. For example, the positive and sig- American Economic Review, 103, 277–304.
nificant effect of long-term ownership is driven by its negative and Altman, E. I. (1968). Financial ratios, discriminant analysis and the prediction of cor-
porate bankruptcy. Journal of Finance, 23, 589–609.
significant effect on the concerns dimension of CSR, whereas the short- Appel, I. R., Gormley, T. A., & Keim, D. B. (2016). Passive investors, not passive owners.
term ownership has a negative and significant effect on the strengths Journal of Financial Economics, 121, 111— 141.
dimension of CSR. When we examine the effect of institutional own- Barnett, M. L. (2007). Stakeholder influence capacity and the variability of financial re-
turns to corporate social responsibility. Academy of Management Review, 32, 794–816.
ership horizon on the components of CSR, a similar picture emerges.
Bebchuk, L. A., Brav, A., & Jiang, W. (2015). The long-term effects of hedge fund activism.
Short-term institutional ownership has a negative and significant effect Columbia Law Review, 115, 1085—11 56.
on both environmental strengths and social strengths scores. In con- Bena, J., Ferreira, M. A., Matos, P., & Pires, P. (2017). Are foreign investors locusts? The
long-term effects of foreign institutional ownership. Journal of Financial Economics,
trast, long-term institutional ownership has a significant and negative
126, 122— 146.
effect on both environmental and social concerns scores. When we look Benabou, R., & Tirole, J. (2010). Individual and corporate social responsibility.
deeper within social categories of CSR, the negative effect of short-term Economica, 77, 1–19.
ownership on CSR seems to be driven by employee relations category, Bennett, J. A., Sias, R. W., & Starks, L. T. (2003). Greener pastures and the impact of
dynamic institutional preferences. Review of Financial Studies, 16, 1203—12 38.
whereas the positive effect of long-term ownership on CSR is present in Bird, A., & Karolyi, S. A. (2016). Do institutional investors demand public disclosure?
community relations, diversity, human rights and product categories. Review of Financial Studies, 29, 3245—32 77.
Similarly, our examination of controversial business issues reveals that Blume, M. E., & Keim, D. B. (2012). Institutional investors and stock market liquidity:
Trends and relationships. University of pennsylvania working paper..
long-term institutional ownership has a significant negative effect on Boone, A. L., & White, J. T. (2015). The effect of institutional ownership on firm trans-
the likelihood of firm getting involved with any controversy (e.g., al- parency and information production. Journal of Financial Economics, 117, 508— 533.
cohol, firearms, gambling, military, nuclear and tobacco-related issues). Brennan, M., Chordia, T., & Subrahmanyam, A. (1998). Alternative factor specifications,
security characteristics, and the cross-section of expected stock returns. Journal of
This negative effect seems to be mostly driven by the lower likelihood Financial Economics, 49, 345—3 73.
of controversies regarding firearms and military. Bushee, B. J. (1998). The influence of institutional investors on myopic R&D investment
To test Jensen's “enlightened value maximization” perspective for behavior. Accounting Review, 73, 305–333.
Bushee, B. J. (2001). Do institutional investors prefers near-term earnings over long-run
CSR, we examine the effect of CSR along with long-term institutional value? Contemporary Accounting Research, 18, 207—24 6.
ownership (LIO) and short-term institutional ownership (SIO) on near- Cameron, A. C., Gelbach, J. B., & Miller, D. (2011). Robust inference with multiway
term and long-term buy-and-hold returns. For short-term institutions to clustering. Journal of Business & Economic Statistics, 29, 238— 249.
Chang, Y.-C., Hong, H., & Liskovich, I. (2015). Regression discontinuity and the price
discourage CSR endeavors, the short-term effect of CSR on prices should
effects of stock market indexing. Review of Financial Studies, 28, 212–246.
be negative for firms with high proportions of short-term investors; for Chen, H., Noronha, G., & Singal, V. (2004). The price response to S&P 500 index additions
long-term institutions to encourage CSR, the long-term effect should be and deletions: Evidence of asymmetry and new explanation. Journal of Finance, 59,
positive for firms with high proportions of long-term investors. 1901—19 40.
Chen, T., Dong, H., & Lin, C. (2017). Institutional shareholders and corporate social re-
Consistent with this conjecture, we find that higher CSR alongside sponsibility: Evidence from two quasi-natural experiments. Working paper..
higher short-term institutional ownership is negatively associated with Choi, N., & Sias, R. (2009). Institutional industry herding. Journal of Financial Economics,
near-term buy-and-hold returns, while higher CSR alongside higher 94, 469—4 91.
Cox, P., Brammer, S., & Millington, A. (2004). An empirical examination of institutional
long-term institutional ownership is positively associated with long- investor preferences for corporate social performance. Journal of Business Ethics, 52,
term buy-and-hold returns. 27–43.
Our findings support the view that short-termism on the part of Cox, P., & Wicks, P. G. (2011). Institutional interest in corporate responsibility: Portfolio
evidence and ethical explanation. Journal of Business Ethics, 103, 143–165.
institutional investors places short-term pressure on companies, and Crane, A. D., Michenaud, S., & Weston, J. P. (2016). The effect of institutional ownership
therefore discourages long-term investments, such as environmentally on payout policy: Evidence from index thresholds. Review of Financial Studies, 29,
and socially responsible investments. In addition, firms with long-term 1377—14 08.
Cremers, M., & Pareek, A. (2016). Patient capital outperformance: The investment skill of
investors seem to be able to make the necessary investments to reduce high active share managers who trade infrequently. Journal of Financial Economics,
environmental and social concerns, as well as controversial business 122, 288–306.
practices. These findings highlight the importance of structuring the
compensation incentives for investment managers so that they are in-
centivized to have a long-term view. A 2008 compensation study of 28
www.cfainstitute.org/Survey/short_termism_survey_report_may_2008.pdf
78
David, P., Bloom, M., & Hillman, A. J. (2007). Investor activism, managerial respon- Johnson, R. A., & Greening, D. W. (1999). The effects of corporate governance and in-
siveness, and corporate social performance. Strategic Managament Journal, 28, stitutional ownership types on corporate social performance. Academy of Management
91–100. Journal, 42, 564–576.
Derrien, F., Kecskés, A., & Thesmar, D. (2013). Investor horizons and corporate policies. Khan, M., Serafeim, G., & Yoon, A. (2016). Corporate sustainability: First evidence on
Journal of Financial and Quantitative Analysis, 48(06), 1755–1780. materiality. Accounting Review, 91, 1697—172 4.
Dimson, E., Karakas, O., & Li, X. (2015). Active ownership. Review of Financial Studies, 28, Kleibergen, F., & Paap, R. (2006). Generalized reduced rank tests using the singular value
3225—32 68. decomposition. Journal of Econometrics, 127, 97 —126.
Dyck, A., Lins, K. V., Roth, L., & Wagner, H. F. (2019). Do institutional investors drive Margolis, J. D., & Walsh, J. P. (2003). Misery loves companies: Rethinking social in-
corporate social responsibility: International evidence. Journal of Financial Economics, itiatives by business. Administrative Science Quarterly, 48, 268—3 05.
131, 693–714. Michaely, R., & Vincent, C. (2012). Do institutional investors influence capital structure
Edmans, A. (2011). Does the stock market fully value intangibles? Employee satisfaction decisions? Cornell university working paper..
and equity prices. Journal of Financial Economics, 101, 621— 640. Mishina, Y., Dykes, B. J., Block, E. S., & Pollock, T. G. (2010). Why “good” firms do bad
Edmans, A. (2012). The link between job satisfaction and firm value, with implications things: The effects of high aspirations, high expectations, and prominence on the
for corporate social responsibility. Academy of Management Perspectives, 26, 1— 19. incidence of corporate illegality. Academy of Management Journal, 53, 701–722.
Erhemjamts, O., Li, Q., & Venkateswaran, A. (2013). Corporate social responsibility and Mitchell, R. K., Agle, B. R., & Wood, D. J. (1997). Toward a theory of stakeholder iden-
its impact on firms' investment policy, organizational structure, and performance. tification and salience: Defining the principle of who and what really counts.
Journal of Business Ethics, 118, 395–412. Academy of Management Review, 22, 853–886.
Falkenstein, E. G. (1996). Preferences for stock characteristics as revealed by mutual fund Murphy, K. J. (2003). Stock-based pay in new economy firms. Journal of Accounting and
portfolio holdings. Journal of Finance, 51, 111–135. Economics, 34, 129–147.
Fama, E. F., & MacBeth, J. D. (1973). Risk, return, and equilibrium: Empirical tests. Neubaum, D. O., & Zahra, S. A. (2006). Institutional ownership and corporate social
Journal of Political Economy, 81, 607–636. performance: The moderating effects of investment horizon, activism, and co-
Ferreira, M. A., & Matos, P. (2008). The colors of investors' money: The role of institu- ordination. Journal of Management, 32, 108— 131.
tional investors around the world. Journal of Financial Economics, 88, 499—5 33. Newey, W. K., & West, K. D. (1987). A simple positive definite, heteroscedasticity and
Freeman, R. E. (1984). Strategic management: A stakeholder approach. Boston: Pitman. autocorrelation consistent covariance matrix. Econometrica, 55, 703–705.
Friedman, M. (1970). The social responsibility of business is to increase its profits. New Nguyen, P.-A., Kecskés, A., & Mansi, S. (2018). Does corporate social responsibility create
York Times Magazine, September 13. shareholder value: The importance of long-term investors. Journal of Banking and
Gaspar, J.-M., Massa, M., & Matos, P. (2005). Shareholder investment horizons and the Finance. https://doi.org/10.1016/j.jbankfin.2017.09.013 forthcoming.
market for corporate control. Journal of Financial Economics, 76, 135—1 65. Orlitzky, M., Schmidt, F. L., & Rynes, S. L. (2003). Corporate social and financial per-
Gaspar, J.-M., Massa, M., Matos, P., Patgiri, R., & Rehman, Z. (2012). Payout policy formance: A meta-analysis. Organization Studies, 24, 403–441.
choices and shareholder investment horizons. Review of Finance, 17, 261–320. Petersen, M. A. (2009). Estimating standard errors in finance panel data sets: Comparing
Gillan, S. L., & Starks, L. T. (2000). Corporate governance proposals and shareholder approaches. Review of Financial Studies, 22, 435–480.
activism: The role of institutional investors. Journal of Financial Economics, 57, Pruitt, S. W., & Wei, K. C. J. (1989). Institutional ownership and changes in the S&P 500.
275—3 05. Journal of Finance, 44, 509–513.
Gompers, P. A., & Metrick, A. (2001). Institutional investors and equity prices. Quarterly Servaes, H., & Tamayo, A. (2013). The impact of corporate social responsibility on firm
Journal of Economics, 116, 229— 259. value: The role of customer awareness. Management Science, 59, 1045—10 61.
Gow, I. D., Ormazabal, G., & Taylor, D. J. (2010). Correcting for cross-sectional and time- Staiger, D., & Stock, J. H. (1997). Instrumental variables regression with weak instru-
series dependence in accounting research. Accounting Review, 85, 483–512. ments. Econometrica, 65, 557–586.
Graham, J. R., Li, S., & Qiu, J. (2008). Corporate misreporting and bank loan contracting. Stock, J. H., & Yogo, M. (2005). Testing for weak instruments in linear IV regression. In
Journal of Financial Economics, 89, 44–61. Identification and Inference for Econometric Models: Essays in Honor of Thomas
Graves, S. B., & Waddock, S. A. (1994). Institutional owners and corporate social per- Rothenberg, Cambridge University Press, 80–108.
formance. Academy of Management Journal, 37. Strike, V. M., Gao, J., & Bansal, P. (2006a). Being good while being bad: Social respon-
Grinstein, Y., & Michaely, R. (2005). Institutional holdings and payout policy. Journal of sibility and the international diversification of US firms. Journal of International
Finance, 60, 1389—14 26. Business Studies, 37, 850–862.
Hansen, L. (1982). Large sample properties of generalized method of moments estimators. Strike, V. M., Gao, J., & Bansal, P. (2006b). Being good while being bad: Social respon-
Econometrica, 50, 1029–1054. sibility and the international diversification of US firms. Journal of International
Harford, J., Kecskés, A., & Mansi, S. (2018). Do long-term investors improve corporate Business Studies, 37, 850–862.
decision making? Journal of Corporate Finance, 50, 424–452. Thompson, S. B. (2011). Simple formulas for standard errors that cluster by both firm and
Hillman, A. J., & Keim, G. D. (2001). Shareholder value, stakeholder management, and time. Journal of Financial Economics, 99, 1–10.
social issues: What's the bottom line? Strategic Management Journal, 22, 125–139. van Beurden, P., & Goessling, T. (2008). The worth of values − A literature review on the
Holderness, C. G., & Sheehan, D. P. (1988). The role of majority shareholders in publicly relation between corporate social and financial performance. Journal of Business
held corporations: An exploratory analysis. Journal of Financial Economics, 20, Ethics, 82, 407–424.
317—3 46. Walls, J. L., Berrone, P., & Phan, P. H. (2012). Corporate governance and environmental
Huang, K., & Petkevich, A. (2016). Corporate bond pricing and ownership heterogeneity. performance: Is there really a link? Strategic Management Journal, 33, 885–913.
Journal of Corporate Finance, 36, 54–74. Yan, X. S., & Zhang, Z. (2009). Institutional investors and equity returns: Are short-term
Jensen, M. C. (2002). Value maximization, stakeholder theory, and the corporate objec- institutions better informed? Review of Financial Studies, 22, 893–924.
tive function. Business Ethics Quarterly, 12, 235—2 56. Zellner, A. (1962). An efficient method of estimating seemingly unrelated regressions and
Jensen, M. C., & Meckling, W. H. (1976). Theory of the firm: Managerial behavior, agency tests for aggregation bias. Journal of the American Statistical Association, 57, 348–368.
costs and ownership structure. Journal of Financial Economics, 3, 305—36 0.
79

Institutional Ownership Horizon Corporate Social Res 2019 Journal of Busine

Uploaded by

Copyright:

Available Formats

Institutional Ownership Horizon Corporate Social Res 2019 Journal of Busine

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Institutional Ownership Horizon Corporate Social Res 2019 Journal of Busine

Uploaded by

Copyright:

Available Formats

Journal of Business Research 105 (2019) 61–79

Contents lists available at ScienceDirect

Journal of Business Research

Institutional ownership horizon, corporate social responsibility and T

ARTICLE INFO ABSTRACT

experience improved accounting performance and governance and in- 3. Sample

Panel A: Descriptive statistics

Panel B: Sample Size by CSR Strengths and Concerns

CSR Concerns 0 1 2 3 4 5 6 7 8 9 10 ≥11

Panel B: Correlations between ownership and CSR of different timings

Lag Cont. Lead Lag Cont. Lead

Panel C: Correlations between CSR components and SIO/LIO (Obs: 15,217)

Model 1 Model 2 Model 3 Model 4 Model 5 Model 6

CSRt + 1 = S SIOt + L LIOt + Xt BX + t ,

Model 1 Model 2 Model 3 Model 4 Model 5 Model 6

CSRt + 1 = S SIOt + L LIOt + Xt BX + t ,

CSRt + 1 = S SIOt + L LIOt + Xt BX + t ,

Any “Sin” Model 1: Alcohol Model 2: Firearms

SIO −0.0943 (0.835) SIO −1.4212 (0.217) SIO 1.7495 (0.380)

CSRt + 1 = S SIOt + L LIOt + Xt BX + t .

From the first-stage estimations of ownership, we see that S&P 500

SIO −0.0435 (0.042) −0.0398 (0.044) −0.2456 (0.000)

SIO Mean −0.1744 (0.007)

Coef. p-val Coef. p-val Coef. p-val Coef. p-val

Model 1: Year t+1 Coef. p-val Coef. p-val

Model 2: Years t+1 to t+2 Coef. p-val Coef. p-val

Model 3: Years t+1 to t+3 Coef. p-val Coef. p-val

Model 4: Years t+1 to t+4 Coef. p-val Coef. p-val

Model 5: Years t+1 to t+5 Coef. p-val Coef. p-val

Model 6: Years t+1 to t+6 Coef. p-val Coef. p-val

You might also like