3000 Word Report Based on Two Related Papers

Journal of Financial Economics 122 (2016) 585–606

Contents lists available at ScienceDirect

Journal of Financial Economics

journal homepage: www.elsevier.com/locate/jfec

Socially responsible firms �

Allen Ferrell a , ∗, Hao Liang b , Luc Renneboog c

a Harvard Law School, Cambridge, MA 02138, USA b Lee Kong Chian School of Business, Singapore Management University, 50 Stamford Road, Singapore 178899, Singapore c CentER, Tilburg University, 50 0 0 LE Tilburg, The Netherlands

a r t i c l e i n f o

Article history:

Received 24 July 2014

Revised 3 December 2015

Accepted 14 December 2015

Available online 31 August 2016

JEL classifications:

G30

G31

G35

K22

L21

M14

Keywords:

Corporate social responsibility

Agency costs

Corporate governance

a b s t r a c t

In the corporate finance tradition, starting with Berle and Means (1932) , corporations

should generally be run to maximize shareholder value. The agency view of corporate so-

cial responsibility (CSR) considers CSR an agency problem and a waste of corporate re-

sources. Given our identification strategy by means of an instrumental variable approach,

we find that well-governed firms that suffer less from agency concerns (less cash abun-

dance, positive pay-for-performance, small control wedge, strong minority protection) en-

gage more in CSR. We also find that a positive relation exists between CSR and value and

that CSR attenuates the negative relation between managerial entrenchment and value.

© 2016 Published by Elsevier B.V.

1. Introduction

The desirability for corporations to engage in socially

responsible behavior has long been hotly debated among

� We are grateful to Kenneth Arrow, Martijn Cremers, Hans Degryse,

Marc Deloof, Elroy Dimson, Joost Driessen, Fabrizio Ferraro, Caroline

Flammer, Jessie Fred, Edward Freeman, Richard Friberg, William Goetz-

mann, Harald Hau, Rebecca Henderson, Oguzhan Karakas, Philipp Krueger,

Thomas Lambert, Alberto Manconi, Chris Marquis, Mark Roe, Amir Ru-

bin, Joaquim Schwalbach, Roy Shapira, Andrei Shleifer, Holger Spamann,

Alexander Wagner, as well as conference and seminar participants at

Harvard Business School, Harvard Law School, Paris Dauphine University,

Ècole Polytechnique Paris, Norwegian School of Economics, Stanford Law

School, Tilburg University, University of Antwerp, University of Ghent,

Second Geneva Summit on Sustainable Finance, Second Geneva-Harvard-

Renmin-Sydney Law Faculty Conference, and the China International Con-

ference in Finance (Shenzhen) for helpful comments. We also thank

Dennis de Buijzer for the excellent research assistant work. ∗ Corresponding author.

E-mail addresses: [email protected] (A. Ferrell), hliang@smu.

edu.sg (H. Liang), [email protected] (L. Renneboog).

http://dx.doi.org/10.1016/j.jfineco.2015.12.003

0304-405X/© 2016 Published by Elsevier B.V.

economists, lawyers, and business experts. Back in the

1930 s, two American lawyers, Adolf A. Berle, Jr., and E.

Merrick Dodd, Jr., had a famous public debate addressing

the question: To whom are corporations accountable?

Berle argued that the management of a corporation should

be held accountable only to shareholders for their actions,

and Dodd argued that corporations were accountable

to both the society in which they operated and their

shareholders ( Macintosh, 1999 ). The lasting interest in this

debate reflects the fact that the issues it raises touch on

the basic role and function of corporations in a capitalist

society.

Two general views, often reflecting the issues raised in

the Berle-Dodd debate, on corporate social responsibility

(CSR) prevail in the literature. The CSR good governance

view argues that socially responsible firms, such as firms

that promote effort s to help protect the environment, seek

social equality, and improve community relations, can and

often do adhere to value-maximizing corporate governance

practices. As such, well-governed firms are more likely to

586 A. Ferrell et al. / Journal of Financial Economics 122 (2016) 585–606

be socially responsible. In short, CSR can be consistent

with maximizing shareholder wealth as well as achieving

broader societal goals. Some proponents of the good gov-

ernance view further argue that firm value maximization

can incorporate stakeholder value, not merely shareholder

value (e.g., Edmans, 2011; Deng, Kang, and Low, 2013 ). The

opposite view on CSR begins with American economist

Milton Friedman’s well-known claim that “the only re-

sponsibility of corporations is to make profits” ( New York

Times Magazine, 1970 , p.122). Extending this view, several

researchers argue that CSR is often simply a manifestation

of managerial agency problems inside the firm ( Benabou

and Tirole, 2010; Cheng, Hong, and Shue, 2014; Masulis

and Reza, 2015 ) and, hence, problematic (the agency view).

That is to say, socially responsible firms tend to suffer from

agency problems, which are also manifested by managers

engaging in CSR that benefits themselves at the expense of

shareholders ( Krueger, 2015 ). Furthermore, managers en-

gaged in time-consuming CSR activities can lose focus on

their core managerial responsibilities ( Jensen, 2001 ). Over-

all, according to the agency view, CSR is generally not in

the interest of shareholders. Friedman even suggested that

to think that business should do anything other than make

a profit is to “harm the foundations of a free society” ( New

York Times Magazine, 1970 , p.122). Reality could lie some-

where between the good governance and agency views of

CSR. Some CSR-related corporate policies can be the re-

sult of good governance consistent with shareholder value,

while others can be driven by agency problems.

The empirical literature testing these two views is

mixed and thus has left the issues raised in the Berle-

Dodd debate largely unresolved. For instance, a number

of papers show that firm participation in certain social is-

sues, such as not engaging with sin industries, avoiding nu-

clear energy, and charity giving, is associated with higher

agency costs and lower shareholder value (e.g., Hillman

and Keim, 2001; Brown, Helland, and Smith, 2006; Di Giuli

and Kostovetsky, 2014; Masulis and Reza, 2015 ). In a recent

study based on the Kinder, Lydenberg, and Domini (KLD)

dataset, which provides CSR ratings for thousands of pub-

lic US companies, Cheng, Hong, and Shue (2014) find em-

pirical evidence supporting the argument that managers of

large US firms enjoy private benefits from investing in CSR.

Meanwhile, other papers, largely using the same KLD data

set, show that a higher CSR score is on average associated

with lower idiosyncratic risk and a lower probability of fi-

nancial distress ( Lee and Faff, 2009 ), a lower cost of capital

( Goss and Roberts, 2011; El Ghoul, Guedhami, Kwok, and

Mishra, 2011; Dhaliwal, Li, Tsang, and Yang, 2011; Albu-

querque, Durnev, and Koskinen, 2015 ), more positive sell-

side analysts’ recommendations ( Bushee, 20 0 0; Bushee and

Noe, 2001 ), and higher abnormal returns and long-term

post-acquisition returns ( Deng, Kang, and Low, 2013 ).

The CSR empirical literature to date has two major lim-

itations. First, much of the literature is largely focused only

on the ex post effects of CSR. That is, the principal research

focus is on measuring shareholder reactions to CSR as cap-

tured by abnormal stock returns (e.g., Dimson, Karakas,

and Li, 2015 ), the cost of capital (e.g., El Ghoul, Gued-

hami, Kwok, and Mishra, 2011 ), and ownership changes

(e.g., Cheng, Hong, and Shue, 2014 ) or on the financial con-

sequences of CSR spending (e.g., Lee and Faff, 2009 ). How-

ever, both the good governance and agency views are con-

cerned to a significant extent with managerial incentives,

which are ex ante in nature. In the agency view, the man-

agerial incentive to engage in CSR is a reflection of the gen-

erally poor incentives of managers at socially responsible

firms, i.e., these firms suffer from agency problems. These

agency problems then manifest themselves in the form of,

among others, CSR activities. In the good governance view,

well-run firms, meaning firms in which management is

generally properly incentivized, tend to have managers en-

gaging in appropriate CSR conduct. In this way, the debate

over CSR connects with the general corporate finance lit-

erature on agency problems and ex ante managerial in-

centives, a fact that we exploit in our empirical analy-

ses. Second, the objective function of a firm is often im-

plicitly assumed in the literature to be exclusively share-

holder wealth maximization, without any independent im-

portance being placed on third-party effects. In this regard,

it is worth noting that in many countries firms are re-

quired by law or social norms to be concerned not only

with shareholders, but also with other stakeholders, such

as employees. Given differing opinions concerning the ap-

propriate objective function within the literature, an im-

portant research question is whether well-governed firms

are more likely to be socially responsible.

In this paper, we take a comprehensive look at the

CSR agency and good governance views around the globe.

By means of a rich and partly proprietary CSR data set

with global coverage across a large number of countries

and composed of thousands of the largest companies,

we test these two views by examining whether tradi-

tional corporate finance proxies for firm agency problems,

such as capital spending cash flows, managerial compen-

sation arrangements, ownership structures, and country-

level investor protection laws, account for firms’ CSR ac-

tivities. While other studies using a within-country quasi-

experimental approach (e.g., Hong, Kubik, and Scheinkman,

2012; Cheng, Hong, and Shue, 2014 ) focus on the marginal

effect of variation in agency problems, our data and empir-

ical setting enable us to examine its average effect. Based

on this comprehensive analysis, we fail to find evidence

that CSR conduct in general is a function of firm agency

problems. Instead, consistent with the good governance

view, well-governed firms, as represented by lower cash

hoarding and capital spending, higher payout and lever-

age ratio and stronger pay-for-performance are more likely

to be socially responsible and have higher CSR ratings. In

addition, CSR is higher in countries with better legal pro-

tection of shareholder rights and in firms with smaller ex-

cess voting power held by controlling shareholders ( Liang

and Renneboog, 2016 ). Moreover, a higher CSR rating mod-

erates the negative association between a firm’s manage-

rial entrenchment and value. All these findings lend sup-

port to the good governance view and suggest that CSR

in general is not inconsistent with shareholder wealth

maximization.

The paper proceeds as follows. Section 2 identifies sev-

eral proxies drawn from the corporate finance literature for

firm agency problems and their possible relation to CSR.

Section 3 describes the samples and specifications used

A. Ferrell et al. / Journal of Financial Economics 122 (2016) 585–606 587

when testing the views on CSR. Section 4 reports and dis-

cusses the empirical results. Section 5 concludes.

2. Agency theory and CSR: hypotheses

Agency problems can manifest themselves through

non-value-maximizing investment choices ( Shleifer and

Vishny, 1989; La Porta , Lopez-de-Silanes, Shleifer, and

Vishny, 20 0 0 ) and managerial pay that is not tied to per-

formance ( Bebchuk and Fried, 2003 ). Economists have fo-

cused on possible mechanisms constraining these agency

problems, such as contract design, incentive systems, and

internal controls [see Holmstrom and Tirole (1989), Pren-

dergast (1999) , and Bebchuk and Weisbach (2010) for re-

views], as well as on external mechanisms such as la-

bor, capital, and product markets ( Fama, 1980; Fama and

Jensen, 1983 ) and institutional arrangements, including le-

gal rules ( La Porta, Lopez-de-Silanes, Shleifer, and Vishny,

1997, 1998, 20 0 0, 20 02 ).

To assess whether CSR should be regarded as a man-

ifestation of agency problems, we explore the underly-

ing mechanisms based on ex ante managerial incentives,

which connects the quality of corporate governance to CSR.

In better governed firms, managers are better incentivized

and their interests and behavior are more aligned with

those of shareholders. If CSR is consistent with or improves

firm performance, managers compensated for good perfor-

mance have a greater incentive to engage in CSR. That is,

good corporate governance induces more CSR activities. In

contrast, under the agency cost view, CSR is detrimental to

shareholder value but is favored by managers as they are

able to extract private benefits; i.e., bad corporate gover-

nance induces more CSR activities.

We examine the relations between CSR and two ex ante

incentive mechanisms of corporate governance, one on cor-

porate financial policies that manifest agency problems

and the other based on executive pay-for-performance.

First, for the corporate financial policies analysis, we ex-

plore the hypotheses based on agency theory at the firm

level in the spirit of Jensen and Meckling (1976) and Jensen

(1986) , which has played a seminal role in the corpo-

rate governance literature ( Morck and Yeung, 2005 ). Ac-

cording to this literature, agency problems can be partic-

ularly acute when the firm generates substantial free cash

flows in excess of those required to finance all positive net

present value projects, leading to serious agency problems

( Servaes and Tamayo, 2014 ). When liquid assets are abun-

dant, firms do not have to submit to the scrutiny of the

capital markets that occurs when new capital is needed,

and the managers have discretion to invest the funds as

they please. Because cash is the most liquid among all cor-

porate assets, it provides managers with the most latitude

as to how and when to spend it, and firms’ capital expen-

diture decisions could be a channel of spending the abun-

dant cash for empire building and private benefits extrac-

tion ( Masulis, Wang, and Xie, 2009 ). Dividends and debt,

given their demands on cash flow, can constrain managers

from diverting cash or committing cash to unprofitable

projects that generate private benefits to insiders ( La Porta,

Lopez-de-Silanes, Shleifer, and Vishny, 20 0 0; Morck and

Yeung, 2005; Jensen and Meckling, 1976; Jensen, 1986 ).

When cash is tight, managers are motivated to run the firm

efficiently, which can increase shareholder value ( La Porta,

Lopez-de-Silanes, Shleifer, and Vishny, 20 0 0 ).

This literature focusing on free cash flow creating an

agency problem suggests a causal effect running from cor-

porate liquidity and leverage to managerial incentives to

divert firm value. This leads to the following hypothesis

reflecting the CSR agency view: A higher level of CSR is

induced by higher cash holdings, free cash flows, and cap-

ital expenditure and lower leverage and dividend payout.

This hypothesis is consistent with the contention that CSR

usually requires long-term investments that do not neces-

sarily contribute to shareholder value maximization but do

contribute to managers’ private benefits of control ( Cheng,

Hong, and Shue, 2014 ). In contrast, the CSR good gover-

nance view suggests the opposite: CSR should be associ-

ated with fewer agency concerns and better managerial

decisions and, thus, with higher dividend payout and lever-

age and lower liquidity (cash and free cash flows) ( Krueger,

2015 ). This hypothesis is consistent with the agency the-

ory in that, when cash is tight, the firm tends to be bet-

ter governed as the manager is motivated or even forced

to run the firm efficiently. Both hypotheses are based on

the ex ante incentives of managers as identified in the cor-

porate finance literature, i.e., the abundance or scarcity of

cash can create bad or good managerial incentives.

Second, we consider the agency versus good gover-

nance view from a managerial incentive-performance per-

spective in the spirit of Jensen and Murphy (1990) and,

hence, investigate hypotheses concerning the relation be-

tween CSR and managerial pay-for-performance. In the

corporate finance literature, executive compensation is

among the central issues in the debate about the effects

of corporate governance as it helps align the interests

of managers and shareholders ( Masulis, Wang, and Xie,

2009 ), and a higher pay-performance sensitivity leads to

less severe agency problems (and thus shareholder value-

enhancement). Weak pay-for-performance sensitivity has

been widely regarded as a major form of incentive mis-

alignment and a symbol of bad governance ( Masulis, Wang,

and Xie, 2009 ), and it can be viewed as a proxy for agency

problems in the firm [“pay without performance” as in

Bebchuk and Fried (2003) ]. Accordingly, the CSR good gov-

ernance view would hypothesize that CSR is associated

with stronger pay-for-performance sensitivity or lower ex-

cess pay, whereas the agency view predicts the opposite.

CSR and agency problems can emerge simultaneously

as they are both corporate choices. This simultaneity (or

endogeneity) creates an empirical challenge for investigat-

ing the relation between CSR and firm agency problems.

Several studies resort to policy and market-wide shocks

as quasi-experiments to help identify a causal relation

between CSR and agency proxies (e.g., Hong, Kubik, and

Scheinkman, 2012; Cheng, Hong, and Shue, 2014; Flammer,

2013 ), but this approach is hard to apply in a multicoun-

try context. Instead, we use an instrumental variable (IV)

approach by employing several variables exogenous to the

focal firm’s financial policies and agency problems as IVs

for firm-level agency indicators.

588 A. Ferrell et al. / Journal of Financial Economics 122 (2016) 585–606

3. Data and methodology

We will now describe our data for CSR and our empiri-

cal strategy based on this data.

3.1. CSR data

Our primary data on CSR are from MSCI’s Intangible

Value Assessment (IVA) database and the Vigeo Corporate

environmental, social and governance (ESG) database. Both

databases are built by means of different proprietary data

sources and employ different rating metrics, which enables

us to cross-validate our results. The IVA indices measure

a corporation’s environmental and social risks and oppor-

tunities, which refer to issues in which companies gener-

ate large environmental and social externalities and can

be forced to internalize (future) unanticipated costs asso-

ciated with those externalities. The ratings are compiled

using company profiles, ratings, scores, and industry re-

ports, and they are available from 1999 to 2011. 1 Covered

are more than 25 hundred companies included in the ma-

jor equity indices around the world: the top 15 hundred

companies of the MSCI World Index (expanding to the full

MSCI World Index over the course of the sample period),

the top 25 companies of the MSCI Emerging Markets In-

dex, the top 275 companies by market capitalization of the

FTSE 100 and the FTSE 250 (excluding investment trusts),

the ASX 200, etc. For this large sample with global cov-

erage, MSCI constructs a series of 29 ESG scores cover-

ing the following categories: strategic governance, which

relates to traditional corporate governance concerns and

whether the firm adopts or has the ability to adopt cer-

tain strategic governance strategies; human capital, which

concerns labor relations as well as employees’ motivation

and health safety; stakeholder capital, which concerns re-

lations with customers, suppliers, and local communities;

products and services that relate to product safety and in-

tellectual capital product development; emerging markets,

which focus on issues related to human rights, child and

forced labor, and oppressive regimes arising from firms’

trade and operations in emerging markets; environmen-

tal risk factors, which include environmental-based liabil-

ities due to operating risks, industry-specific carbon risks,

and performance in leading sustainability risk indicators;

environmental management capacity, which includes en-

vironmental audit, accounting, reporting, training, certifi-

cation, and product materials; and environmental oppor-

tunity factors such as the firm’s competence in embed-

ding certain environmental opportunities in its strategies. 2

1 The information on which the IVA ratings are based is extracted from

corporate documents (annual reports, environmental and social reports,

securities filings, websites, and Carbon Disclosure Project responses), gov-

ernment data (central bank data, US Toxic Release Inventory, Comprehen-

sive Environmental Response and Liability Information System (CERCLIS),

Resource Conservation and Recovery Act (RCRA), etc. (for European com-

panies, the information is expanded by means of many other information

sources), trade and academic journals included in Factiva and Nexis, and

professional organizations and experts (reports from and interviews with

trade groups, industry experts, and nongovernmental organizations famil-

iar with the companies’ operations). 2 A key ESG issue is defined as an environmental or social externality

that has the potential to become internalized by the industry or the com-

The rating then takes into account the extent to which a

company has developed robust CSR strategies and demon-

strated a strong track record in managing these specific

risks and opportunities. A higher rating is assigned if the

company has done better than its peers in one or several

of the above-mentioned dimensions in its initiatives and

risk management. Furthermore, we cross-validate the re-

sults utilizing the IVA ratings with analyses using the Risk-

Metrics EcoValue21 Rating and the RiskMetrics Social Rating

scores, which are provided by RiskMetrics Group (now part

of MSCI) and capture the environmental and social aspects

of CSR, respectively. 3 Companies in the sample are rated

from CCC to AAA, which we then transform into numeric

ratings from 0 to 6. The whole IVA sample (including the

RiskMetrics ratings) covers 91,373 firm-time observations

from 59 countries.

The Vigeo Corporate ESG data set focuses more on CSR

compliance, as it applies a check-the-box approach to rate

how a firm and the country where it operates comply with

the conventions, guidelines, and declarations by interna-

tional organizations such as the United Nations (UN), In-

ternational Labor Organization (ILO), and Organization for

Economic Co-operation and Development (OECD). The Vi-

geo ratings cover six evaluation categories: (1) environ-

ment, (2) human rights, (3) human resources, (4) busi-

ness behavior (which concerns the relations with suppli-

ers and customers), (5) community involvement, and (6)

traditional corporate governance. These six domains are

further broken down into 38 ESG criteria (sustainability

drivers and risk factors) based on universally defined social

responsibility objectives and managerial action principles.

The indices used by Vigeo are Euronext Vigeo World 120,

Euronext Vigeo Europe 120, Euronext Vigeo Eurozone 120,

Euronext Vigeo US 50, Euronext Vigeo France 20, Euronext

Vigeo United Kingdom 20, and Euronext Vigeo Benelux 20,

and they are updated every six months. The whole Vigeo

sample covers 7,048 firm-time observations from 28 coun-

tries and 36 sectors. Both the MSCI sample and the Vi-

geo sample cover the well-established equity indices of the

largest companies across the world, not just a specific sam-

ple of firms that engage in CSR. Unlike the MSCI IVA rat-

ings, which are based on a scale of 0 to 6 (CCC to AAA), the

Vigeo ESG ratings are given on a scale of 0 to 100. As all

of the sample firms are listed and included in the major

global equity indices, our empirical results mostly speak

of the relations between CSR and corporate governance (or

agency problems) in the world’s largest corporations.

pany through one or more of the following triggers: pending or proposed

regulation; a potential supply constraint; a notable shift in demand; a

major strategic response by an established competitor; growing public

awareness or concerns. Once up to five key issues have been selected, an-

alysts work with sector team leaders to make any necessary adjustments

to the weightings in the model. The weightings take into account the im-

pact of companies, their supply chains, and their products and the finan-

cial implications of these impacts, illustrated in Online Appendix Table

OA1. On each key ESG issue, a wide range of data is collected to address

the question: To what extent is risk management commensurate with risk

exposure? 3 The two ratings from RiskMetrics use similar methodologies as the

IVA rating. Thus we combine them with the IVA ratings (and its constitu-

tive sub-dimensions) and refer to this combined sample as the “IVA sam-

ple,” which includes the overall IVA ratings.

A. Ferrell et al. / Journal of Financial Economics 122 (2016) 585–606 589

For both the MSCI and Vigeo samples, firms are rated

relative to their industry peers from both domestic and

international markets. Thus, the ratings do not depend

on the cross-country differences in jurisdiction, regulation,

and the local CSR situation. This makes our cross-country

data more credible and helps guarantee that our CSR rat-

ings are not biased by country-specific characteristics. In

addition, we triangulate the results using our public and

proprietary CSR data by using the ASSET4 data from Thom-

son Reuters, which also have global coverage and a similar

rating method (by adopting a global industry peer compar-

ison). The detailed descriptions of the MSCI IVA and the

Vigeo ESG samples are shown in the Online Appendix Ta-

bles OA1 and OA2, and their country distributions (as well

as that of ASSET4) are shown in Online Appendix Tables

OA3–OA5.

3.2. Empirical strategy

Our empirical strategy is to test the effects of prox-

ies for agency problems on CSR. Bae, Kang, and Wang

(2011) state, in line with Jensen (1986) , that firms with

significant free cash flow but few investment opportuni-

ties are more likely to invest beyond the optimal level and

that dividends and debt serve as disciplinary mechanisms

to prevent managers from wasting free cash. Based on our

earlier discussion of the academic literature, we utilize five

agency proxies (putting aside for the moment managerial

compensation): (1) capital expenditure (CapEx); (2) cash

holdings; (3) free cash flow measured as earnings before

interest and taxes (EBIT) after the change in net assets

(CapEx, minus depreciation and amortization, plus or mi-

nus the change in net working capital); (4) dividend pay-

out ratio; and (5) leverage, measured as the ratio of total

debt over total equity. Higher values of the first three vari-

ables (1-3) can be an indication of agency costs, especially

for large and mature multinational firms as in our sample,

and higher values of the last two (4 and 5) relate to mech-

anisms that can curb managerial agency problems.

The issue of endogeneity is, as always, important, which

is why we take an instrumental variable approach. In the

spirit of Lin, Malatesta, and Xuan (2011) and Laeven and

Levine (2009) , we use the industry peers’ average finan-

cial policies as IVs for firm-level financial policies. We take

the within-sample arithmetic means of each of our five

financial policies variables (agency indicators) by country,

by industry, and by year (country-industry-year average).

The use of industry peers’ average financial policies as IVs

for a focal firm’s financial policies can be justified in sev-

eral ways. First, ample evidence exists that a firm’s finan-

cial policies are affected by the policies of its peer firms.

According to Leary and Roberts (2014) , peer effects are

more important for capital structure determination than

most previously identified determinants. Such peer effects

are also found in corporate precautionary cash holdings

(e.g., Hoberg, Phillips, and Prabhala 2014 ), corporate invest-

ment decisions ( Cheng, 2011; Foucalt and Fresard, 2014 ),

and earnings fraud and other types of financial miscon-

duct ( Parsons, Sulaeman, and Titman, 2014 ). Second, lit-

tle reason exists to believe that a firm’s CSR practice is

affected by its peer firms’ financial policies through chan-

nels other than influencing its own financial policies. Even

if there were channels other than the focal firm’s own fi-

nancial policies, we attempt to take that into account by

controlling for firm fixed effects and time fixed effects.

We use the within-sample industry peers’ average fi-

nancial policies, instead of the financial policies of country-

wide industry peers, as our IVs, because our sample firms

are mostly large and mature companies that are listed in

major global equity indices such as the MSCI World Index

and, therefore, are comparable in size and tend to bench-

mark their financial policies to other large companies such

as those in our sample. It is less likely the case that an in-

dustry leader’s cash holdings and capital structure are de-

termined by the average cash holdings and capital struc-

ture of all the firms in its country and industry (as most

of these firms would be smaller and less internationally

active). On average, each firm has 49 peers (firms with-

out peers are dropped from the IV analysis). We also check

the robustness of our IV results by using alternative depen-

dent variables, alternative CSR samples, and an alternative

IV that captures other aspects of the agency problems.

Higher cash holdings, free cash flows, and capital ex-

penditures do not necessarily reflect higher agency costs as

long as there are sufficient growth and investment oppor-

tunities. The argument Jensen (1986) makes is that firms

with larger free cash flow but with limited investment op-

portunities can suffer from the agency problem of misus-

ing corporate funds. Therefore, we control for investment

opportunities proxied by Tobin’s q (market-to-book ratio of

assets) in all our regressions. Also, although each agency

indicator has some non-agency dimensions – for example,

higher cash holdings can signify higher profitability result-

ing from better past investments—our key argument is that

they can induce managerial incentives in relation to seek-

ing private benefits by engaging in costly CSR activities. We

therefore lag all independent variables (including the IVs)

by one year. Moreover, it is important to interpret our em-

pirical results collectively. That is, although a cash holdings

variable or leverage ratio in isolation can represent differ-

ent aspects of corporate policy and profitability other than

the degree of agency problems, the consistency of the signs

of all five potential agency indicators can provide a good

indication of whether CSR in general is induced by agency

problems or by good governance.

Turning to managerial compensation, we test the rela-

tion between CSR and managerial pay-for-performance by

regressing CSR on a pay-performance sensitivity variable,

along with other firm-level and country-level covariates.

In the literature, executive compensation is usually mea-

sured by the sum of cash-based pay (salaries and bonuses)

and equity-based pay (stock options, restricted stocks, and

payouts from long term incentive plans), and gauging ex-

ecutive pay-for-performance based on a single compensa-

tion dimension is difficult as both types of compensation

are benchmarked to different types of firm performance.

Furthermore, the relative use of these types of compen-

sation has changed considerably over time ( Frydman and

Jenter, 2010 ), which is why we focus on total compensation

and relate it to the total share performance benchmark.

A standard way of measuring a firm’s pay-for-performance

is by estimating the sensitivity of the change of executive

590 A. Ferrell et al. / Journal of Financial Economics 122 (2016) 585–606

compensation to the change of firm performance (e.g., re-

turn on assets (ROA) or Tobin’s q) over a long time series,

which is captured by the beta coefficient on the compensa-

tion variable for each individual firm ( Adams and Ferreira,

2008 ). However, this approach is not feasible in our setting

with companies from around the world, as we do not have

long-enough time series data on their executive compensa-

tion. In addition, the validity of measuring executive incen-

tives by pay-performance sensitivity is debatable, because

fractional equity ownership [the dollar change in chief ex-

ecutive officer (CEO) wealth for one dollar change in firm

value; Jensen and Murphy (1990) ] and equity-at-stake [the

dollar change for 1% change in firm value; Hall and Lieb-

man (1998) ] usually yield conflicting results for US samples

( Frydman and Jenter, 2010 ), let alone for international sam-

ples. Moreover, using sensitivity measures estimated from

regressing pay on performance (or vice versa) can lead

to spurious correlations because the performance variable

could already incorporate the effect of CSR. As these con-

cerns restrain us from using the traditional measures of

pay-for-performance as a proxy for managerial incentives,

we use two alternative pay-for-performance measures.

First, we use Thomson Reuter’s ASSET4 variable CEO

Compensation Link to Total Shareholder Return , which is a

dummy variable indicating whether managerial compensa-

tion is linked to total shareholder return (TSR) and is based

on the combination of textual analysis of a company’s an-

nual report and media coverage of executive compensa-

tion issues. In the annual report, a company usually in-

cludes a remuneration report or a compensation discussion

and analysis, which contains information through which

Thomson Reuters tracks whether specific return-based per-

formance benchmarks are set and whether the CEO com-

pensation is linked to TSR. If Thomson Reuters finds in

the above sources evidence that the executive top man-

agement has TSR as a performance criterion (performance

target), then a one is assigned to the pay-for-performance

variable and zero otherwise. We acknowledge that this

dummy variable is a crude proxy for pay-for-performance,

but it is a relatively objective indicator that captures the

ex ante managerial incentives of engaging in CSR (a per-

formance benchmark is clearly an ex ante mechanism) and

can be applied to international samples without long time

series.

The typical endogeneity concern that emerges is that

stronger pay-for-performance is a result of higher levels of

CSR or that CSR and pay-for-performance are jointly de-

termined by other firm-level factors. To address this endo-

geneity concern, we again employ an instrumental variable

approach, by using IVs for our pay-for-performance mea-

sure. The IVs we select are mostly related to board struc-

tures (collected from Datastream) and have been shown in

the literature to be key determinants of executive pay-for-

performance: the percentage of independent board mem-

bers as reported by the company [ Percent independent

board members ]; the percentage of independent compen-

sation committee members [ Compensation committee inde-

pendence ]; and a dummy variable indicating whether the

CEO simultaneously chairs the board [ CEO-chairman dual-

ity ]. Here, independent board members are directors who

are not employed by the company, have not served as an

executive board member in the firm for at least ten years,

are not or do not represent a reference shareholder with

more than 5% of the equity, do not hold cross-board mem-

berships, have no recent or immediate family ties to ex-

ecutive and nonexecutive directors of the company, and do

not accept any compensation other than fees for board ser-

vice. We cross-validate the information (e.g., board mem-

bers’ identities and status such as executive versus non-

executive) by manually checking the director reports of the

BoardEx database in each year of our sample and correct

mistakes.

We use the above-mentioned variables as our IVs be-

cause, on the one hand, stronger pay-for-performance sen-

sitivity has been shown to be steered by board indepen-

dence ( Chhaochharia and Grinstein, 2009; Bebchuk, Cre-

mers, and Peyer, 2011 ) and, on the other hand, no rea-

son exists to believe that a firm’s CSR is directly related

to board independence (especially the measure related

to compensation committee independence) through chan-

nels other than managerial pay-for-performance. Even if

one is still concerned that CSR is related to board struc-

ture and independence, which can create an omitted vari-

able bias, those channels should be taken into account by

controlling for firm fixed effects, as board structures are

mostly stable over time and are mainly changed by regula-

tions ( Linck, Netter, and Yang, 2012 ), and the identification

mostly comes from cross-sectional (instead of time series)

variations of board structures. Hence, time-invariant mech-

anisms induced by board structures should be captured by

firm fixed effects. We do not use industry peers’ pay-for-

performance measures as IVs, which is different from what

we have done for financial policies as agency indicators.

The reason is that, in the literature, the overall evidence for

the effective use of relative performance evaluation (rela-

tive to an industry benchmark) is weak ( Frydman and Jen-

ter, 2010 ).

Second, we estimate excess pay (or abnormal pay) from

a typical CEO compensation regression, which captures the

degree to which CEO compensation is not explained by

firm performance and general firm characteristics in a first

stage and then relate this predicted excess pay to the com-

pany’s CSR level in a second stage. Excess pay can re-

flect the occurrence of agency problems (or the quality

of governance) within a company as it regards the lack

of performance-driven incentives. We adopt this approach

but are mindful of the fact that several firm and CEO char-

acteristics can explain both excess pay of the CEO and

the firm’s CSR level. Therefore, we control not only for a

set of CEO and firm characteristics, but also for firm and

time fixed effects in both stages. We collect the yearly CEO

compensation from Datastream (ASSET4’s corporate gover-

nance pillar) and cross-validate the data with BoardEx’s di-

rector reports. We add the data on the CEO and boards

(CEO-chairman duality, independence of the compensation

committee, board size, independence of board members,

whether or not say-on-pay is sought) from Datastream and

BoardEx, blockholder ownership data from Orbis, and other

firm characteristics from Compustat and Datastream. In the

first stage, we estimate a typical pay-performance model

by regressing the logarithm of the CEO’s total compensa-

tion on the previous year’s Tobin’s q, along with the above

A. Ferrell et al. / Journal of Financial Economics 122 (2016) 585–606 591

CEO, board, governance, and firm characteristics variables

[following Bebchuk, Cremers, and Peyer (2011) and Peters

and Wagner (2014) ], as well as firm and time fixed ef-

fects. We call the residual of this regression “excess pay”

as it captures the component of CEO’s total compensation

that is predicted neither by firm performance measures

nor by other well-documented CEO and firm attributes (in-

cluding governance). Excess pay thus represents that part

of CEO pay that is not tied to performance. We use this

predicted pay residual as a proxy for the degree of devi-

ation from the expected pay-for-performance, that is, as

a proxy for poor governance. The larger this residual, the

greater the incentive misalignment for the CEO. In the

second stage, we regress the CSR rating on the one-year-

lagged excess pay and other control variables [firm size,

the largest shareholder’s cash flow rights, ROA , Tobin’s q,

interest coverage, current ratio , the Ln(GDP per capita) , and a

country-level globalization index], as well as firm and time

fixed effects. The agency cost view on CSR predicts a pos-

itive coefficient on Excess Pay in the second stage, because

the larger the excess pay, the more severe the agency prob-

lems are expected to be for the CEO because he could then

engage in non-performance-enhancing CSR activities. The

good governance view predicts the opposite. That is, when

the excess pay is zero or small, a CEO is more likely to

carry out value-consistent or value-enhancing CSR activi-

ties.

In robustness tests, we replace firm fixed effects with

industry fixed effects in all regressions to validate our re-

sults. The descriptive statistics of the variables mentioned

above for different sam ples (the MSCI IVA sample, the Vi-

geo ESG sample, and the ASSET4 ESG sample) are provided

in Table 1.

4. Results

We now present the results of our analyses.

4.1. Results on agency indicators

For our results on agency indicators, we will first

present our baseline IV results followed by results based

on analyses using alternative IVs, alternative dependent

variables and, finally, an alternative CSR sample.

4.1.1. Baseline IV results

In Table 2 , we examine the relation between CSR and

our five agency proxies: cash holdings, free cash flow,

CapEx, dividend payout ratio, and leverage. The agency

view predicts a positive relation between CSR and the first

three proxies and a negative one for the last two. The good

governance view on CSR predicts the opposite.

The five proxies are instrumented by the correspond-

ing within-sample firm-level industry peers’ financial poli-

cies. One important note is that the correlations between

the five industry-peer proxies are not high, ranging from

−0.8% to 23% for both the MSCI IVA and the Vigeo ESG sample companies, which suggests that the five financial

policies variables capture agency problems from different

angles and that we are thus not measuring the same re-

lation in each of the five models. In the second stage, CSR

ratings are regressed on the five predicted agency proxies

as estimated from the first stage and on the other control

variables with bootstrapping-adjusted standard errors. We

report regression results for both stages; all independent

variables are lagged.

In the first stage, industry peers’ cash holdings, free

cash flow, capital expenditures, dividend ratio, and lever-

age are all positively and significantly correlated with the

focal firm’s cash holdings, free cash flow, capital expendi-

tures, dividend ratio, and leverage, respectively, as is evi-

dent in Columns 1, 3, 5, 7, and 9. The Cragg-Donald f-test

statistics are all higher than the critical value of 16.38, with

P-values smaller than 0.01 in all specifications. This is es-

sentially a test for the weak instrument hypothesis (testing

for the relevance of the IV in the first stage), and a higher

test statistic (low P-value) indicates a rejection of the null

of a weak instrument. These positive and significant cor-

relations as well as the high F-statistics suggest that our

IVs are strongly correlated with our endogenous variables,

supporting the relevance criterion of our IVs.

In the second stage, the coefficients of the three

liquidity-focused agency proxies (cash holdings, free cash

flows, and capital expenditures) are all negative and sta-

tistically significant above the 95% confidence level, while

the coefficients on the financial constraint-focused agency

proxies (dividend payouts and leverage) are both signif-

icantly positive. The point estimate for cash holdings as

a percentage of total assets is −0.0127, indicating that a one standard deviation increase in cash holdings as a per-

centage of total assets (8.6%) is associated with a 0.11

( −0.0127 × 8.6) decrease in the overall IVA rating or 4% lower at the mean score (0.11/2.8). The point estimates

for the other agency indicators have similar magnitudes,

and, given that we use an IV approach, they can be inter-

preted as local average treatment effects (LATE). All these

findings therefore do not support the CSR agency view. In

addition, financial slack (as measured by the current ra-

tio) and, to some extent, the financial constraint proxy (in-

terest coverage) are mostly positively associated with the

ESG ratings, which likewise do not provide support for the

CSR agency costs perspective. We confirm that firms with

higher CSR ratings are larger firms with more concentrated

ownership, located in countries that are richer (in terms of

GDP per capita) and more globalized (as captured by the

globalization index). In terms of causation, the interpre-

tation of these results ought to be done with care. Given

our identification strategy, we tend to interpret them as

follows. Well-governed firms suffer less from agency con-

cerns. When cash is tight (lower cash reserves, free cash

flows and capital spending, and higher dividend payouts

and interest payouts), managers are motivated to run the

firm more efficiently and care more about the long run.

This is consistent with engaging in CSR activities.

4.1.2. Alternative IVs

To check the robustness of our previous results relat-

ing the level of CSR to various agency indicators, we con-

duct several additional tests using alternative IVs and CSR

indicators. First, we replace the industry peers’ average fi-

nancial policies, which we used as IVs, by a combination

of the existence of a poison pill and a classified board

592 A. Ferrell et al. / Journal of Financial Economics 122 (2016) 585–606

Table 1

Descriptive statistics of key variables.

Panel A. MSCI Intangible Value Assessment (IVA) sample and Vigeo environmental, social, and governance (ESG) sample

Variable MSCI IVA sample Vigeo ESG sample

Number of Mean Median Standard Minimum Maximum Number of Mean Median Standard Minimum Maximum

observations deviation observations Deviation

Overall IVA score 47,775 2.850 3 1.753 0 6

EcoValue score 90,496 2.926 3 1.833 0 6

Social score 61,119 2.857 3 1.725 0 6

Vigeo Environment score 7,048 33.867 34.0 0 0 18.534 0 87

Vigeo Human Resources score 7,048 32.378 31.0 0 0 17.939 0 84

Vigeo Customers & Suppliers

score

7,048 40.981 42.0 0 0 13.473 4 82

Cash holdings (scaled by

assets)

77,061 0.075 0.045 0.086 0 0.994 5,995 0.076 0.051 0.081 0 0.787

Free cash flows (scaled by

assets)

65,728 0.060 0.057 0.049 −0.034 0.159 4,804 0.104 0.094 0.055 0.020 0.227

Capital expenditure (scaled by

assets)

67,091 0.052 0.042 0.046 0 1.037 4,984 0.049 0.040 0.043 0 0.498

Dividend payout

(dividend-sales ratio)

(percent)

56,116 2.912 1.856 3.068 0 11.386 4,649 4.256 2.855 4.282 0 16.246

Leverage ratio (winsorized) 78,004 0.615 0.613 0.208 0.228 0.955 6,038 0.646 0.094 0.194 0.288 0.961

ROA (winsorized) 74,993 0.049 0.043 0.038 −0.001 0.149 5,876 0.047 0.040 0.038 0.012 0.117 Tobin’s q (winsorized) 82,269 1.730 1.427 0.825 0.970 3.977 6,766 2.751 1.935 2.911 0.620 8.020

Financial constraints

(winsorized)

62,076 0.264 0.006 0.495 0 1.832 4,738 0.296 0.035 0.500 0 1.784

Interest coverage (winsorized) 73,948 11.988 5.975 13.807 1.115 45.310 5,821 9.799 5.388 10.317 1.118 33.80

Financial slack (current ratio) 63,342 1.721 1.365 1.572 0.038 184.984 4,852 0.850 0.774 0.472 0 6.527

Blockholder ownership 54,746 35.57% 23.12% 33.92% 0% 100% 6,755 35.31% 23.56% 34.27% 0% 100%

Largest shareholder’s

ownership

25,558 21.22% 13.21% 18.56% 0% 100% 4,787 18.79% 11.05% 17.38% 0.06% 100%

Board size 55,990 11.518 11 4.165 1 45 5,513 11.942 11 4.031 2 44

Independence of board

members

47,701 59.15% 41.67% 30.09% 0% 100% 4,702 60.56% 50% 27.50% 0% 100%

Compensation committee 56,482 0.835 1 0.371 0 1 5,533 0.890 1 0.313 0 1

Compensation committee

independence

45,165 86.42% 100% 27.51% 0% 100% 4,691 81.11% 100% 30.17% 0% 100%

Say on pay 56,558 0.210 0 0.407 0 1 5,533 0.278 0 0.473 0 1

CEO-chairman duality 56,431 0.369 0 0.483 0 1 5,533 0.278 0 0.448 0 1

CEO compensation link to TSR 56,482 0.361 0 0.480 0 1 5,533 0.417 0 0.493 0 1

Adjusted anti-director rights

index

89,765 3.371 4 1.184 2 5 7,006 3.757 4 1.098 2 5

Anti-self-dealing index 89,947 0.617 0.650 0.212 0.170 1 7,047 0.546 0.500 0.240 0.2 1

Public enforcement of

anti-self-dealing

89,947 0.197 0 0.339 0 1 7,047 0.331 0 0.403 0 1

Panel B. ASSET4 Sample

Variable Number of Mean Median Standard Minimum Maximum

observations deviation

Wedge1 (voting minus cash flow rights) 20,573 1.165% 0 7.245% −89.84% 99.99% Wedge2 (voting over cash flow rights) 20,562 4.039 1 170.790 0 10 0 0 0

Largest shareholder’s ownership 23,797 22.029% 13.6% 19.578% 0 100%

Largest shareholder’s voting rights 20,716 23.590% 14.3% 20.881% 0 100%

Equity book-to-market (winsorized) 46,583 2.359 1.800 1.757 0.500 7.280

Firm size (Total assets) 31,133 3612965 6123 2.15 ×10 8 0 3.06 ×10 10 Firm age 23,374 34.740 23 31.655 0 185

Annual sales growth rate (winsorized) 46,799 12.627% 8.16% 21,157% −19.070% 69.830% CapEx to sales ratio (winsorized) 29,015 0.017 0.001 0.044 2.54 ×10 −6 0.185 Leverage ratio (winsorized) 31,061 21.081% 15.932% 382.758% −0.034% 67392% Dividend per share (winsorized) 47,541 4.014 0.345 9.940 0 41

ROE (winsorized) 31,082 0.121 0.118 0.143 −0.209 0.427 ROA (winsorized) 31,084 0.051 0.045 0.060 −0.073 0.179 Entrenchment Index 1 53,472 0.690 0 1.037 0 5

Entrenchment Index 2 53,472 0.889 0 1.239 0 5

A . F

e rre

ll e

t a

l. / Jo

u rn

a l o

f F

in a

n c ia

l E

c o

n o

m ic

s 1

2 2 (2

0 1 6

) 5

8 5

– 6

0 6

5 9

3

Table 2

Corporate social responsibility and agency indicators with firm-level instrumental variables (IVs): two-stage least-square (2SLS) estimations with industry-peer average financial policies as IVs.

The table shows the results from 2SLS estimations using an instrumental variable approach. The dependent variables (DVs) in the first stage (columns 1, 3, 5, 7, and 9), are cash holdings ( Cash , scaled by total

assets), free cash flows ( FCF , scaled by total assets), capital expenditure ( CapEx , scaled by total assets), dividend payout (dividend/sales ratio), and leverage (debt-to-asset ratio), respectively. These five agency proxies

are then instrumented by the corresponding within-sample firm-level industry peers’ financial policies in the second stage, calculated as the arithmetic means of each of the five financial policies variables

for a firm’s industry peers (the industry classification is based on Worldscope) by year and by countries. The dependent variables in the second stage (Columns 2, 4, 6, 8, and 10), are the (IVA) rating from

MSCI’s Intangible Value Assessment database. All independent variables are lagged by one year. The Cragg-Donald F-test statistics (the weak instruments’ test) and the P-values are reported for the first stage.

All regressions control for firm and time fixed effects. Bootstrapped standard errors are clustered at the firm level and are reported in parentheses. ∗∗∗, ∗∗ , and ∗ indicate significance at the 1%, 5%, and 10% level, respectively. Stock-Yogo weak identification value: 10% maximal IV size—16.38; 15% maximal IV size—8.96; 20% maximal IV size—6.66; 25% maximal IV size 5.53.

Cash FCF CapEx Dividend ratio Leverage

(1) (2) (3) (4) (5) (6) (7) (8) (9) (10)

First stage Second stage First stage Second stage First stage Second stage First stage Second stage First stage Second stage

Agency indicator DV = cash DV = CSR DV = FCF DV = CSR DV = CapEx DV = CSR DV = dividend DV = CSR DV = leverage DV = CSR

Cash (scaled by assets) −0.0127 ∗∗∗ (0.0037)

Industry peer cash (IV) 0.749 ∗∗∗ (0.010)

FCF (scaled by assets) −0.0131 ∗∗ (0.0055)

Industry peer FCF (IV) 0.657 ∗∗∗ (0.022)

CapEx (scaled by assets) −0.0345 ∗∗∗ (0.007)

Industry peer CapEx (IV) 0.819 ∗∗∗ (0.024)

Dividend/sales ratio 0.0035 ∗∗∗ (0.0012)

Industry peer dividend (IV) 0.090 ∗∗∗ (0.011)

Leverage ratio 0.0555 ∗∗∗ (0.007)

Industry peer leverage (IV) 0.254 ∗∗∗ (0.015)

Control variable

Ln(Assets) −2.215 ∗∗∗ 0.415 ∗∗∗ 1.416 ∗∗∗ 0.510 ∗∗∗ −0.278 0.4 4 4 ∗∗∗ −1.144 0.536 ∗∗∗ 2.182 ∗∗∗ 0.371 ∗∗∗ (0.220) (0.0573) (0.364) (0.0594) (0.232) (0.0570) (7.538) (0.0713) (0.561) (0.0668)

Largest shareholder’s ownership −0.018 ∗∗∗ 0.0043 ∗∗∗ −0.042 ∗∗∗ 0.00316 ∗∗ −0.028 ∗∗∗ 0.00367 ∗∗ −0.177 0.0140 ∗∗∗ 0.005 0.00435 ∗∗ (0.006) (0.00155) (0.010) (0.00157) (0.006) (0.00155) (0.205) (0.00196) (0.015) (0.00174)

Tobin’s q −0.055 0.0356 −0.491 ∗ 0.0699 −1.122 ∗∗∗ 0.0235 −2.610 0.0484 −0.258 0.121 ∗∗ (0.171) (0.0447) (0.276) (0.0447) (0.182) (0.0456) (6.231) (0.0591) (0.427) (0.0504)

ROA −0.0013 0.00251 −0.032 0.00175 −0.010 0.00264 −0.715 −0.00278 0.404 ∗∗∗ −0.0192 ∗∗∗ (0.015) (0.00403) (0.024) (0.00398) (0.016) (0.00398) (0.542) (0.00520) (0.039) (0.00528)

Interest coverage 0.011 ∗∗∗ 0.0 0 0305 0.041 ∗∗∗ 0.0 0 0767 0.033 ∗∗∗ 0.00129 0.138 0.00201 ∗∗ −0.094 ∗∗∗ 0.00559 ∗∗∗ (0.003) (0.0 0 0799) (0.005) (0.0 0 0844) (0.003) (0.0 0 0847) (0.104) (0.0 010 0) (0.008) (0.00115)

Current ratio 0.424 ∗∗∗ 0.00586 0.037 0.0359 ∗∗ 0.146 ∗∗ 0.0337 ∗ 6.952 ∗∗∗ −0.0292 ∗∗ −0.367 ∗∗∗ 0.0259 ∗∗ (0.041) (0.0109) (0.112) (0.0181) (0.073) (0.0180) (1.284) (0.0143) (0.101) (0.0123)

Ln(GDP per capita) −2.758 ∗∗∗ 0.838 ∗∗∗ 4.966 ∗∗∗ 0.957 ∗∗∗ 1.200 0.918 ∗∗∗ −16.210 0.131 7.084 ∗∗∗ 0.334 (0.947) (0.247) (1.619) (0.262) (1.069) (0.262) (31.762) (0.300) (2.400) (0.290)

Globalization −0.142 ∗∗∗ 0.0866 ∗∗∗ 0.045 0.0868 ∗∗∗ −0.055 0.0870 ∗∗∗ −1.822 0.0505 ∗∗∗ 0.224 ∗∗ 0.0655 ∗∗∗ (0.044) (0.0115) (0.072) (0.0117) (0.048) (0.0117) (1.442) (0.0137) (0.112) (0.0133)

Firm fixed effects Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes

Time fixed effects Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes

Observations 8,624 8,458 8,468 6,537 8,431

R -squared 0.8633 0.8650 0.8653 0.8662 0.8492

First stage Cragg-Donald F-test statistics 4985.916 (P = 0.00) 857.34 (P = 0.00) 1175.45 (P = 0.00) 66.18 (P = 0.00) 281.46 (P = 0.00)

594 A. Ferrell et al. / Journal of Financial Economics 122 (2016) 585–606

Table 3

CSR and agency indicators with firm-level instrumental variables (IVs):

two-stage least-square (2SLS) estimations with poison pill and classified

board as the IV.

The table shows the results from 2SLS estimations using an IV ap-

proach. Only the second–stage results are shown for conciseness (the

setup is similar to Table 2 ). The dependent variables in the first stages

are cash holdings ( Cash , scaled by total assets), free cash flows ( FCF , scaled

by total assets), capital expenditure ( CapEx , scaled by total assets), dividend

payout (dividend/sales ratio), and leverage (debt-to-asset ratio). These five

agency proxies are then instrumented by the ordinal variable poison pill

in combination with classified board that takes the value zero if the firm

neither adopted a poison pill nor has a classified board , one if the firm

adopted either of the two, and two if the firm has both. The depen-

dent variables in the second stage (columns 1–5) are the IVA ratings from

MSCI’s Intangible Value Assessment database. All independent variables

are lagged by one year. All regressions include the control variables re-

ported in Table 2 and control for firm and time fixed effects. Bootstrapped

standard errors are clustered at the firm level and are reported in paren-

theses. ∗∗∗, ∗∗ , and ∗ indicate significance at the 1%, 5%, and 10% level, re- spectively.

Variable (1) (2) (3) (4) (5)

Cash (scaled −0.0749 ∗ by assets) (0.0454)

FCF (scaled by −0.132 assets) (0.234)

CapEx (scaled −0.247 ∗∗∗ by assets) (0.0825)

Dividend payout 0.886 ∗∗

(0.425)

Leverage ratio 0.151 ∗

(0.0871)

Control variables Yes Yes Yes Yes Yes

Time fixed effects Yes Yes Yes Yes Yes

Firm fixed effects Yes Yes Yes Yes Yes

Number of 9,544 8,790 8,959 7,371 8,873

observations

within a firm, and conduct similar two stage least squares

(2SLS) tests. The reasons that we combine these anti-

takeover mechanisms are twofold. First, whenever the use

of a poison pill is legally allowed, the combination of the

poison pill and a classified board is the most effective de-

fense available to a target company and, hence, most likely

to entrench its management ( Bebchuk, Coates, and Subra-

manian, 2002 ). Second, antitakeover provisions are usually

regulated by laws, and their prevalence varies across coun-

tries. For example, while the poison pill is widely used in

US companies and can be triggered by the directors, the

mandatory business neutrality rule in the United Kingdom

does not allow directors to activate a poison pill but re-

quires that the shareholders decide on accepting or reject-

ing a takeover bid ( Davies, Schuster, and Van de Walle de

Ghelcke, 2010; Enriques, Gilson, and Pacces, 2014 ). In con-

nection with the use of this variable as a robustness check,

a significant fraction of our firm-year observations are

from the United States and the United Kingdom (see On-

line Appendix and Tables OA3 and OA4). Hence, the com-

bining of these two mechanisms into one broader mecha-

nism is more suitable for our cross-country setting. Man-

agerial entrenchment enhanced by antitakeover provisions

can amplify the agency problems in a firm, which can be

directly manifested by the firm’s cash flows and payout

policies. There is no reason to believe, however, that the

decision to engage in CSR is influenced by the likelihood

of a firm being taken over, except via the channel of its

financial policies induced by agency problems.

Table 3 shows the results using the existence of a poi-

son pill and a classified board as alternative IVs. We report

the coefficients on the five agency indicators only in the

second stage for conciseness, but the model specification is

essentially the same as that in Table 2 . The results are con-

sistent with our previous findings. The (instrumented) in-

dicators of cash holdings, free cash flows, and CapEx are all

negatively correlated with the aggregate CSR score, and the

(instrumented) indicators of dividend payout and leverage

ratio are positively correlated with CSR. The coefficients are

significant within the 5% level for CapEx and dividend pay-

out and within the 10% for cash holdings and leverage.

4.1.3. Alternative dependent variables

As an additional robustness check, we replace the gen-

eral CSR variable in Table 2 by the RiskMetrics EcoValue Rat-

ing (CSR focusing on ecological efficiencies) and the Risk-

Metrics Social Rating (CSR focusing on social issues), and

we conduct similar tests as in the baseline results by us-

ing industry peers’ average financial policies as IVs. These

two sub-dimensional ratings use similar metrics as the IVA

rating, and they measure two important, but different, as-

pects of CSR: a firm’s environmental impact and social im-

pact (the IVA rating gauges a firm’s overall CSR engage-

ment, which also includes other dimensions). The results

of using these two alternative dependent variables are re-

ported in Table 4 , with Panel A showing results for the

EcoValue Rating and Panel B for the Social Rating . We re-

port the coefficients of the five (predicted) agency indica-

tors only in the second stage for reasons of conciseness,

but all first-stage f-tests satisfy the relevance criteria of IV.

In Panel A, three out of the five agency indicators ( FCF,

CapEx , and dividend payout ) are significant at the 1% level

and have signs consistent with those in Tables 2 and 3 .

In Panel B, four out of five agency indicators ( cash, FCF,

CapEx , and the leverage ratio ) are significant at the 99%

confidence level with consistent signs. The economic mag-

nitudes are also similar to those in Table 2 . In unreported

analysis, we also replace the dependent variables by the

three sub-indices that receive the highest weights in the

general IVA index ( Labor Relations, Industry-specific Carbon

Risks , and Environmental Opportunities ) and three aggregate

sub-scores [ Strategic Governance (including traditional gov-

ernance), Human Capital , and Stakeholder Capital ], and sim-

ilar results are obtained. Again, using these alternative de-

pendent variables yields results consistent with the good

governance view, not the agency view, on CSR.

4.1.4. Alternative CSR sample

We also turn to an alternative CSR sample, the Vigeo

Corporate ESG sample, to further cross-validate our results.

The Vigeo Corporate ESG data set focuses more on a firm’s

compliance (rather than engagement) to CSR standards and

enables us to test another aspect of CSR to triangulate

our previous approach. Different from the IVA ratings, the

Vigeo ratings are given on a scale of 0-100. The results

from this alternative CSR sample are shown in Table 5 , in

which the dependent variables are the Environment Score

(Panel A), Customers & Suppliers Score (Panel B), and Human

A. Ferrell et al. / Journal of Financial Economics 122 (2016) 585–606 595

Table 4

Corporate social responsibility and agency indicators: robustness with alternative dependent variables.

The table shows the results from two-stage least squares estimations using an instrumental variable approach. Only the second-stage results are shown

for conciseness (the setup is similar to Table 2 ). The dependent variables in the first stage are cash holdings ( Cash , scaled by total assets), free cash flows

( FCF , scaled by total assets), capital expenditure ( CapEx , scaled by total assets), dividend payout (dividend/sales ratio), and leverage (debt-to-asset ratio).

These five agency proxies are then instrumented by the corresponding within-sample firm-level industry peers’ financial policies and are calculated as the

arithmetic means of each of the five financial policies variables for a firm’s industry peers (the industry classification is based on Worldscope) by year and

across countries. The dependent variable in the second stage is the EcoValue rating (environmental rating) as in Panel A and is the Riskmetrics Social rating

as in Panel B. All independent variables are one-year lagged. All regressions include the control variables reported in Table 2 and control for firm- and

time-fixed effects. Bootstrapped standard errors are clustered at the firm level and are reported in parentheses. ∗∗∗, ∗∗ , and ∗ indicate significance at the 1%, 5%, and 10% level, respectively.

Panel A. Dependent variable is EcoValue Rating (environmental rating)

Variable (1) (2) (3) (4) (5)

Cash (scaled by assets) 0.0297

(0.0219)

FCF (scaled by assets) −0.0244 ∗∗∗ (0.0045)

CapEx (scaled by assets) −0.0290 ∗∗∗ (0.0071)

Dividend / sales ratio 0.0 0 013 ∗∗∗

(0.0 0 0 03)

Leverage ratio 0.00197

(0.00167)

Control variables Yes Yes Yes Yes Yes

Time and firm fixed effects Yes Yes Yes Yes Yes

Number of observations 10,530 10,513 10,716 7,819 10,727

Panel B. Dependent variable is RiskMetrics Social Rating

Cash (scaled by assets) −0.0115 ∗∗∗ (0.00345)

FCF (scaled by assets) −0.0320 ∗∗∗ (0.00506)

CapEx (scaled by assets) −0.0202 ∗∗∗ (0.00745)

Dividend / sales ratio 0.0115

(0.0218)

Leverage ratio 0.0203 ∗∗∗

(0.00510)

Control variables Yes Yes Yes Yes Yes

Time and firm fixed effects Yes Yes Yes Yes Yes

Number of observations 9,961 9,776 9,786 7,332 9,715

results.

Resources Score (Panel C), which most represent the inter-

ests of a company’s direct stakeholders. In line with our

baseline tests on the MSCI IVA sample ( Table 2 ), we use

the industry peer average financial policies as our IVs for

our five financial policies variables. 4 As is evident in Table

5 , we show similar results. Variables related to cash and

liquidity mostly have a negative sign, and dividend payout

and leverage ratio are positively correlated with all three

ESG ratings from this alternative sample. Take cash hold-

ings as an example. Economically, a one standard deviation

increase in cash holdings as a percentage of total assets

(8.1%) is associated with 4.9% decrease in the Environment

Score , 7.1% decrease in the Customers & Suppliers Score , and

7.9% decrease in the Human Resources Score . 5

4 Again, the coefficients of the five agency indicators in the second

stage only are shown for conciseness. Tables with more detailed results

are available upon request. 5 In unreported analyses, we also use several country-level shareholder

protection law indices as IVs for firm-level agency concerns, and simi-

lar results are found. These country-level shareholder protection indices

include the antidirect rights index first developed by La Porta, Lopez-de-

Silanes, and Shleifer (1998) and further adjusted by Spamann (2010) , anti-

In sum, the CSR agency view predicts a positive and sig-

nificant correlation between liquidity focused agency prox-

ies (e.g., the abundance of cash) and CSR. As long as the

coefficients are not positive and significant, the agency

view is unsubstantiated, which is the case as we find con-

sistent significantly negative relations. Our results yield

that CSR is adopted by firms characterized by good gov-

ernance. Also, in all our models above we control for firm

and time fixed effects, which mitigate concerns that unob-

servable firm characteristics or time trends could drive our

self-dealing index (ASDI) as in Djankov, La Porta, Lopez-de-Silanes, and

Shleifer (2008) , the private enforcement of securities law index as in La

Porta, Lopez-de-Silanes, and Shleifer (2006) , the revised one-share one-

vote rule (mandatory proportionality of voting and cash flow) index as in

Spamann (2010) , and the revised mandatory waivable dividend index as

in Spamann (2010) .

596 A. Ferrell et al. / Journal of Financial Economics 122 (2016) 585–606

Table 5

Corporate social responsibility (CSR) and agency indicators: robustness with alternative CSR sample.

The table shows the results from the two–stage least squares (2SLS) estimations using an instrumental variable approach. Only the second–stage results

for the main explanatory variables are shown for conciseness. The dependent variables in the first stage are cash holdings ( Cash , scaled by total assets), free

cash flows ( FCF , scaled by total assets), capital expenditure ( CapEx , scaled by total assets), dividend payout (dividend/sales ratio), and leverage (debt-to-asset

ratio). These five agency proxies are then instrumented by the corresponding within-sample firm-level industry peers’ average financial policies (country-

industry-year average) and are the arithmetic means of each of the five financial policies variables for a firm’s industry peers (the industry classification is

based on Worldscope) by year and by countries. The dependent variables in the second stage are the Environmental Score (Panel A), Customers & Suppliers

Score (Panel B), and Human Resources Score (Panel C), all are from Vigeo’s Corporate environmental, social, and governance database. The control variables

are the same as those reported in Table 2 . The regressions include firm and time fixed effects. Bootstrapped standard errors are clustered at the firm level

and are reported in parentheses. ∗∗∗, ∗∗ , and ∗ indicate significance at the 1%, 5%, and 10% level, respectively.

Panel A. Dependent variable is Vigeo Environment Score

Variable (1) (2) (3) (4) (5)

Cash (scaled by assets) −0.604 ∗ (0.344)

FCF (scaled by assets) −0.198 (0.267)

CapEx (scaled by assets) −0.455 (0.400)

Dividend/sales ratio 1.270 ∗∗

(0.549)

Leverage ratio 1.335 ∗∗∗

(0.333)

Control variables Yes Yes Yes Yes Yes

Time and firm fixed effects Yes Yes Yes Yes Yes

Panel B. Dependent variable is Vigeo Customers & Suppliers Score

Cash (scaled by assets) −0.939 ∗ (0.507)

FCF (scaled by assets) −6.083 (14.03)

CapEx (scaled by assets) 0.176

(0.350)

Dividend/sales ratio 1.655 ∗∗∗

(0.460)

Leverage ratio 1.035 ∗

(0.541)

Control variables Yes Yes Yes Yes Yes

Time and firm fixed effects Yes Yes Yes Yes Yes

Panel C. Dependent variable is Vigeo Human Resources Score

Cash (scaled by assets) −1.035 ∗∗ (0.523)

FCF (scaled by assets) −0.314 (0.263)

CapEx (scaled by assets) −0.436 (0.396)

Dividend / sales ratio 1.056 ∗

(0.560)

Leverage ratio 0.460 ∗

(0.257)

Control variables Yes Yes Yes Yes Yes

Time and firm fixed effects Yes Yes Yes Yes Yes

Number of observations 3,414 3,383 3,487 3,295 3,434

4.2. Results on pay-for-performance and excess pay

We now present the results of our pay-for-performance

and excess pay analyses starting with our pay-for-

performance baseline IV results.

4.2.1. Pay-for-performance: baseline IV results

We next investigate how CSR is related to an-

other incentive-based mechanism of corporate governance,

namely, executive pay-for-performance. The question we

try to answer is whether firms with stronger pay-for-

performance (a proxy for better governance) have higher

CSR ratings. To deal with the potential endogeneity issue,

we again apply an instrumental variable approach by in-

strumenting pay-for-performance with Percent independent

board members, Compensation committee independence , and

CEO-chairman duality . We report the results from these

pay-for-performance analyses in Table 6 , showing both the

first-stage and the second-stage results, as well as the

Cragg-Donald f-test statistics. In the first stage, the dummy

variable pay-for-performance [measured by the ASSET4 en-

try CEO Compensation Link to Total Shareholder Return ] is

A. Ferrell et al. / Journal of Financial Economics 122 (2016) 585–606 597

Table 6

Corporate social responsibility (CSR) and executive pay-for-performance: two-stage least squares (2SLS) estimations.

The table shows the results of CSR and executive pay-for-performance from 2SLS estimations using the instrumental variable (IV) approach. The depen-

dent variable (DV) in the first stage is a dummy variable pay-for-performance , which indicates whether executive pay is linked to total shareholder return

(TSR) and is measured using Thomson Reuter’s ASSET4 data category “CEO Compensation Link to Total Shareholder Return”. This variable is based on a

textual analysis from companies’ annual reports. The IVs for pay-for-performance are the percentage of independent board members as reported by the

company ( percent independent board members , as in Columns 1 and 2), the percentage of independent compensation committee members as stipulated by

the company ( Compensation committee independence , as in Columns 3 and 4), and a dummy variable indicating whether the chief executive officer (CEO)

simultaneously chairs the board ( CEO-chairman duality , as in Columns 5 and 6). The dependent variable in the second stage is the IVA rating from MSCI’s

Intangible Value Assessment database. All independent variables are one-year lagged. The Cragg-Donald f-test statistics (weak instrument test) are reported

for the first stage. All regressions control for firm and time fixed effect. Bootstrapped standard errors are clustered at the firm level and are reported in

parentheses. ∗∗∗, ∗∗ , and ∗ indicate significance at the 1%, 5%, and 10% level, respectively. Stock-Yogo weak identification test critical values: 10% maximal IV size—16.38; 15% maximal IV size—8.96; 20% maximal IV size—6.66; 25% maximal IV size—5.53.

IV = Percent independent IV = Compensation IV = CEO-chairman board members committee independence duality

(1) (2) (3) (4) (5) (6)

Variable DV = Pay–for- DV = CSR DV = Pay–for- DV = CSR DV = Pay-for- DV = CSR performance performance performance

Instrumental variable 0.0011 ∗∗∗ 0.0011 ∗∗∗ −0.043 ∗∗∗ (0.0 0 02) (0.0 0 03) (0.010)

Pay-for-performance 4.145 ∗∗∗ 3.355 ∗∗∗ 3.368 ∗∗∗

(predicted from first stage) (1.012) (1.099) (1.199)

Ln(Assets) 0.031 ∗ 0.321 ∗∗∗ −0.030 ∗ 0.648 ∗∗∗ 0.0256 ∗ 0.301 ∗∗∗ (0.016) (0.0910) (0.016) (0.0848) (0.0145) (0.0802)

Largest owner shares 0.0 0 03 0.00176 −0.0 0 08 0.00193 −0.0 0 04 0.00555 ∗∗∗ (0.0 0 05) (0.00265) (0.0 0 06) (0.00329) (0.0 0 04) (0.00213)

Tobin’s q −0.035 ∗∗∗ 0.198 ∗∗ −0.026 ∗∗ 0.188 ∗∗∗ −0.026 ∗∗ 0.170 ∗∗ (0.013) (0.0786) (0.013) (0.0706) (0.011) (0.0662)

ROA 0.005 ∗∗∗ −0.0206 ∗∗ 0.004 ∗∗∗ −0.00344 0.004 ∗∗∗ −0.0134 ∗ (0.001) (0.00807) (0.0012) (0.00754) (0.001) (0.00761)

Interest coverage 0.0 0 01 −0.0 0 02 −0.0 0 05 ∗ 0.00222 0.0 0 01 0.0 0 01 (0.0 0 03) (0.00141) (0.0 0 03) (0.00147) (0.0 0 02) (0.0011)

Current ratio −0.007 ∗∗ 0.0219 −0.011 ∗∗∗ 0.0278 −0.006 ∗ 0.0182 (0.003) (0.0175) (0.003) (0.0204) (0.003) (0.0160)

Ln(GDP per capita) 0.340 ∗∗∗ 0.0575 0.165 ∗∗ 0.971 ∗∗ 0.251 ∗∗∗ 0.310 (0.067) (0.493) (0.071) (0.419) (0.059) (0.426)

Globalization 0.018 ∗∗∗ 0.0246 −0.020 ∗∗∗ 0.147 ∗∗∗ 0.011 ∗∗∗ 0.0637 ∗∗∗ (0.003) (0.0248) (0.005) (0.0335) (0.003) (0.0195)

Time fixed effects Yes Yes Yes Yes Yes Yes

Firm fixed effects Yes Yes Yes Yes Yes Yes

Number of observations 7,871 6,797 8,967

R -squared 0.747 0.772 0.772

Cragg-Donald f-test statistics 29.12 20.61 17.87

regressed on the three above IVs (respectively shown in

Columns 1, 3 and 5). In the second stage, the dependent

variable is the overall IVA rating (Columns 2, 4, and 6) and

the key explanatory variable is Pay-for-performance pre-

dicted from the first stage.

The results on pay-for-performance are again not com-

patible with the agency view, but instead support the

good governance view. In the first stage, Percent indepen-

dent board members and Compensation committee indepen-

dence are both positively and significantly correlated with

Pay-for-performance , and CEO-chairman duality is negatively

and significantly correlated with Pay-for-performance . This

is consistent with the notion that board independence, es-

pecially the independence of the compensation committee,

induces executive pay-for-performance and with the fact

that a CEO who is also chairing the board makes himself

more entrenched and more likely to self-grant unjustified

high salaries, which can diminish the pay-for-performance

relation. The first-stage Cragg-Donald f-test statistics are all

above the critical value of 16.38, supporting the relevance

of these IVs. In the second stage, the coefficients on the

pay-for-performance variable are positive and significant.

The point estimates suggest that the relations are eco-

nomically meaningful. Firms that have an explicit pay-for-

performance benchmark on average have more than three-

grade higher CSR ratings.

4.2.2. Pay-for-performance with alternative dependent

variables and CSR sample

To verify the robustness of our pay-for-performance re-

sults, we use alternative dependent variables and an alter-

native CSR sample, similar to what we have also done for

financial policies as agency indicators ( Sections 4.1.3 and

4.1.4 ). The alternative dependent variables are the RiskMet-

rics EcoValue Rating (environmental rating) and Social Rat-

ing from the MSCI sample, and the alternative CSR sample

is the Vigeo ESG data from which we use the Environment

Score , the Human Resources Score , and the Customers & Sup-

pliers Score (see Table 7 ). For reasons of conciseness, we

show the coefficients of the predicted Pay-for-performance

only in the second stage of 2SLS estimations. We find

that firms with explicit and clear pay-for-performance

598 A. Ferrell et al. / Journal of Financial Economics 122 (2016) 585–606

Table 7

Corporate social responsibility (CSR) and executive pay-for-performance: robustness with alternative dependent variables (DVs) and CSR Sample.

The table shows the results of CSR and executive pay-for-performance from two–stage least squares estimations using the independent variable IV

approach. The setup is similar to Table 6 , but only the second stage results for the main explanatory variable are shown for conciseness. The dependent

variable in the first stage is a dummy variable pay-for-performance which indicates whether executive pay is linked to total shareholder return (TSR) and is

measured using Thomson Reuter’s ASSET4 data category “CEO Compensation Link to Total Shareholder Value”. This variable is based on a textual analysis

from companies’ annual reports. The IV for pay-for-performance is the percentage of independent board members ( Percent independent board members ).

The control variables are Ln(Assets), Largest shareholder’s ownership, Tobin’s q, ROA, Interest coverage, Current ratio, Ln(GDP per capita) , and Globalization . All

regressions control for firm and time fixed effects. Bootstrapped standard errors are clustered at the firm level and are reported in parentheses. ∗∗∗, ∗∗ , and ∗ indicate significance at the 1%, 5%, and 10% level, respectively.

Panel A. IVA – RiskMetrics sample Panel B. Vigeo ESG sample

DV = EcoValue rating DV = Social DV = Environment DV = Human DV = Customers & (environment) rating score resources score suppliers score

(1) (2) (3) (4) (5)

Pay–for–performance 2.840 ∗∗∗ 1.431 ∗∗ 31.99 ∗∗ 32.25 ∗ 40.45 ∗

(predicted from first stage) (0.922) (0.691) (9.616) (18.16) (24.27)

Control variables Yes Yes Yes Yes Yes

Time fixed effects Yes Yes Yes Yes Yes

Firm fixed effects Yes Yes Yes Yes Yes

Number of observations 9,276 12,487 2,631 2,631 2,631

R -squared 0.789 0.798 0.983 0.787 0.618

benchmarks have on average 1.4–2.8 grades higher (on a

scale of 7) on the EcoValue Rating and on the Social Rat-

ing or 32 - 40 grades higher (on a scale of 100) on the

Vigeo ESG ratings, relative to firms without explicit total

shareholder return benchmarks. These results are consis-

tent with those in Table 6 and provide additional support

for the good governance view on CSR (and not for the

agency view). Firms with better governance mechanisms to

incentivizing their top executives also engage in and com-

ply more with CSR.

4.2.3. Excess pay

We next use another empirical approach, which con-

sists of estimating Excess pay (the residual from a typical

CEO compensation regression) that captures the degree to

which CEO compensation is not explained by firm perfor-

mance and firm - CEO characteristics in a first stage, and

then regress a company’s CSR rating on this predicted ex-

cess pay in a second stage. Excess pay can be interpreted as

a deviation from pay-for-performance and should be pos-

itively correlated with CSR in the second stage under the

agency costs view, but negatively correlated with CSR un-

der the good governance view.

We report results of the CSR - excess pay estimation

in Table 8 , with Columns 1, 3, 5, and 7 reporting the

first-stage results of regressing the logarithm of CEO’s to-

tal compensation (in US dollars) on different sets of vari-

ables capturing firm performance, firm - CEO characteris-

tics, and corporate governance. These variables include To-

bin’s q (winsorized at 5%), return on assets ( ROA ) (win-

sorized at 5%), firm size (measured by the logarithm of to-

tal assets), CEO-chairman duality (a dummy variable cap-

turing whether the CEO simultaneously chairs the board),

and a self-constructed entrenchment index 6 , as in Column

6 More detailed description of the entrenchment index can be found in

Section 4.5 and in the Online Appendix. The entrenchment index aims at

capturing managerial entrenchment as described in Bebchuk, Cohen, and

Ferrell (2009) and is the sum of dummies that capture the presence of

1. Column 3 includes all the variables of Column 1 but

adds a dummy Say-on-pay which captures whether the

company’s shareholders have the right to vote on execu-

tive compensation. Column 5 includes all the variables as

in Column 1 and adds the logarithm of board size. Col-

umn 7 includes all the above mentioned variables and also

the percentage of independent directors on the board ( Per-

cent independent board members ), a compensation commit-

tee dummy, and blockholder ownership. 7 Columns 2, 4, 6,

and 8 report the second-stage results of regressing the ag-

gregate IVA rating on the predicted residual ( Excess pay )

from the first stage along with the control variables used

in, for example, Table 2 . Inevitably, lagging independent

variables and adding many control variables reduces our

sample size. Even within the sample for which interna-

tional data on CEO compensation and characteristics, board

structures, and other firm variables are available, the pat-

terns are clear. As is evident in Table 8 , Excess pay is nega-

tively correlated with the aggregate CSR rating. This means

that CEOs with high pay not related to performance invest

less in CSR, which supports the good governance view in-

stead of the agency cost perspective. Robustness tests with

the IVA environmental and social ratings as well as CSR

ratings from the Vigeo database confirm the relation. 8

4.3. Investor protection laws and CSR

As the main purpose of this paper is to evaluate

whether CSR results from agency problems or is, on the

contrary, present in well-governed firms, we also turn to

the regulatory context of agency and governance at the

country level. Agency problems can also be shaped by the

a poison pill, a golden parachute, a classified board, other antitakeover

devices, and supermajority requirements for amending the charter and

bylaws. 7 More detailed descriptions of these variables are in the Online Ap-

pendix. 8 Tables with robustness tests are available upon request.

A. Ferrell et al. / Journal of Financial Economics 122 (2016) 585–606 599

Table 8

Corporate social responsibility (CSR) and excess pay: two-stage least squares (2SLS) estimations.

The table shows the results of CSR and excess pay from 2SLS estimations. The dependent variable in the first stage is the CEO’s total compensation,

which is regressed on Tobin’s q, return on assets (ROA), firm size, CEO–chairman duality , the degree of entrenchment ( E-index ), a dummy variable capturing

whether say-on-pay is sought, board size , the Percent independent board members , Compensation committee independence , and blockholder ownership (the

percent of equity held by all blockholders who own more than 5% of the firm’s shares). The residual of the first stage is called excess pay and is an

explanatory variable of CSR (the overall IVA rating from MSCI’s Intangible Value Assessment database) in stage two, which also includes the following

variables: Tobin’s q (winsorized at 5%), ROA (winsorized at 5%), firm size, largest shareholder ownership stake (percent), interest coverage, current ratio, ln(GDP

per capita) , and a country level globalization index . All independent variables are one-year lagged. All regressions control for firm and time fixed effects.

Bootstrapped standard errors are clustered at the firm level and are reported in parentheses. ∗∗∗, ∗∗ , and ∗ indicate significance at the 1%, 5%, and 10% level, respectively.

DV = Ln(CEO pay) (in US dollars) (1) (2) (3) (4) (5) (6) (7) (8) First stage Second stage First stage Second stage First stage Second stage First stage Second stage

Explanatory variable (lagged) DV = Ln(Pay) DV = CSR DV = Ln(Pay) DV = CSR DV = Ln(Pay) DV = CSR DV = Ln(Pay) DV = CSR

Excess pay −0.0451 ∗ −0.0430 ∗ −0.0357 −0.0574 ∗∗ (0.0241) (0.0241) (0.0242) (0.0257)

Tobin’s q 0.212 ∗∗∗ 0.238 ∗∗∗ 0.212 ∗∗∗ 0.238 ∗∗∗ 0.210 ∗∗∗ 0.234 ∗∗∗ 0.245 ∗∗∗ 0.178 ∗∗

(0.0242) (0.0688) (0.0242) (0.0689) (0.0242) (0.0690) (0.0255) (0.0819)

ROA −0.0025 0.0143 ∗∗ −0.0023 0.0144 ∗∗ −0.00235 0.0144 ∗∗ −0.0027 0.0104 (0.0020) (0.0064) (0.0020) (0.0064) (0.0019) (0.0064) (0.0020) (0.0067)

Ln(Assets) 0.180 ∗∗∗ 0.620 ∗∗∗ 0.181 ∗∗∗ 0.621 ∗∗∗ 0.174 ∗∗∗ 0.620 ∗∗∗ 0.239 ∗∗∗ 0.613 ∗∗∗

(0.0265) (0.0847) (0.0265) (0.0847) (0.0266) (0.0847) (0.0271) (0.0906)

CEO–chairman duality 0.165 ∗∗∗ 0.165 ∗∗∗ 0.170 ∗∗∗ 0.121 ∗∗∗

(0.0266) (0.0266) (0.0268) (0.0279)

Entrenchment index 0.0154 0.0169 0.0176 0.00429

(0.0118) (0.0118) (0.0119) (0.0123)

Say–on–pay −0.0877 ∗∗∗ −0.0524 ∗ (0.0300) (0.0301)

Ln(Board size) 0.0848 ∗ 0.0767 (0.0468) (0.0496)

Percent independent board members −0.234 ×10 −3 (0.419 ×10 −3 )

Compensation committee independence −0.179 ∗∗∗ (0.0447)

Blockholder ownership 0.0165

(0.0170)

Largest shareholder’s ownership −0.0 0 07 −0.0 0 08 −0.0 0 02 0.0031 (0.0031) (0.0031) (0.0031) (0.0036)

Interest coverage −0.0017 −0.0017 −0.0016 −0.0013 (0.0016) (0.0016) (0.0017) (0.0018)

Current ratio −0.0048 −0.0047 −0.0048 −0.0038 (0.0125) (0.0125) (0.0125) (0.0140)

Ln(GDP per capita) 1.630 ∗∗∗ 1.633 ∗∗∗ 1.626 ∗∗∗ 2.055 ∗∗∗

(0.415) (0.415) (0.415) (0.446)

Globalization index 0.0297 0.0299 0.0280 0.0828 ∗∗

(0.0318) (0.0318) (0.0318) (0.0337)

Constant 10.08 ∗∗∗ −29.54 ∗∗∗ 10.06 ∗∗∗ −29.60 ∗∗∗ 9.982 ∗∗∗ −29.34 ∗∗∗ 8.666 ∗∗∗ −38.48 ∗∗∗ (0.459) (5.738) (0.459) (5.741) (0.466) (5.746) (0.770) (6.152)

Number of observations 14,153 3,599 14,153 3,599 14,106 3,598 12,187 3,060

Firm fixed effects Yes Yes Yes Yes Yes Yes Yes Yes

Time fixed effects Yes Yes Yes Yes Yes Yes Yes Yes

R -squared 70.32% 89.27% 70.30% 89.27% 70.23% 89.27% 73.13% 90.04%

legal protection of investor rights at the country level. In

countries with stronger legal protection, agency problems

are likely to be lower, which can entail that CSR activities

are lower if one assumes that they are due to agency

problems. We therefore explore the relation between

country-level investor protection laws and firm-level CSR.

The relevant country-level investor protection laws are

those that provide legal protection of shareholder rights

( La Porta, Lopez-de-Silanes, Shleifer, and Vishny, 20 0 0 ).

Broadly speaking, the laws that aim at addressing agency

problems and investor expropriation concern corporate

decision making and voting (corporate law), information

disclosure in securities transactions (securities law), and

regulation of related parties transactions (anti-self-dealing

law), as well as the effectiveness of their enforcement ( La

Porta, Lopez-de-Silanes, and Shleifer, 2006; Djankov, La

Porta, Lopez-de-Silanes, and Shleifer, 2008 ). We therefore

use country-level investor protection laws as a proxy for

firm-level agency problems in exploring the CSR agency

and good governance views. In the agency view, stronger

legal protection of shareholder rights reduces the incentive

and ability of corporate insiders (directors and officers) to

extract private benefits through CSR-related spending. In

contrast, the CSR good governance view predicts that CSR

spending is positively related to shareholder protection. To

test the relations between firm-level CSR and country-level

600 A. Ferrell et al. / Journal of Financial Economics 122 (2016) 585–606

shareholder protection laws, we use several country-level

investor protection indices, which all come from well-

established sources. We use the anti-director rights index

(ADRI) which was first developed by La Porta, Lopez-de-

Silanes, Shleifer, and Vishny (1998) and revised in Djankov,

La Porta, Lopez-de-Silanes, and Shleifer (2008) and

Spamann (2010) . 9 We stick to the most recent version of

ADRI as amended by Spamann (2010) . For comparison,

we use the anti-self-dealing index (ASDI) developed by

Djankov, La Porta, Lopez-de-Silanes, and Shleifer (2008) ,

which is not directly related to corporate decision making

but more related to the regulation on insider expropri-

ation, and contains ex ante control of self-dealing and

ex post control of self-dealing variables. We include the

variable public enforcement of anti-self-dealing index also

developed by Djankov, La Porta, Lopez-de-Silanes, and

Shleifer (2008) .

We regress CSR ratings on various shareholder protec-

tion laws variables ( Table 9 ) and show that company law

on shareholder protection (ADRI) is strongly and positively

correlated with CSR. 10 The positive correlations between

corporate law and CSR suggest that when legal rules are

stronger in disciplining corporate behavior toward good

conduct to investors, especially to minority shareholders,

firms are more likely to engage in social responsibilities.

The effects are also economically meaningful. For exam-

ple, a one standard deviation in ADRI is associated with

0.35 (0.297 × 1.184) increases in the overall IVA rating, or a more than 12% (0.35 / 2.85) increase from the mean IVA.

In Panel B, in which the dependent variables are the Vi-

geo ESG ratings that focus more on CSR compliance (in-

stead of on the CSR practice or engagement of Panel A),

company law on shareholder protection (ADRI) still plays

a positive role. However, we do not find consistent re-

sults for the anti-self-dealing index and the public enforce-

ment of self-dealing in Panels A and B. The insignificance

of the latter two indices can be explained by the fact that,

as CSR represents a shareholder-stakeholder trade-off, the

most relevant regulation is the one related to corporate de-

cision making and voting. In contrast, ASDI and the public

enforcement index mostly capture a constraint on insider

trading transactions and are not directly related to how

companies make decisions regarding stakeholders’ welfare.

9 Both the original La Porta, Lopez-de-Silanes, Shleifer, and Vishny

(1998) ADRI and the Spamann (2010) revised ADRI consist of six key com-

ponents: (1) proxy by mail allowed, (2) shares not blocked before share-

holder meeting, (3) cumulative voting or proportional representation, (4)

oppressed minority protection, (5) preemptive rights to new share issues,

and (6) percentage of share capital to call an extraordinary shareholder

meeting. 10 To save space, we do not report the parameter estimates of the con-

trol variables: cash holdings (scaled by total assets), leverage ratio, return

on assets (ROA), Tobin’s q, interest coverage, current ratio, ownership dis-

persion (the Bureau van Dijck’s independence indicator), as well as in-

dustry and time fixed effects. Full tables are available upon request. As

a robustness test, we set up specifications that include the original ADRI

from La Porta, Lopez-de-Silanes, Shleifer, and Vishny (1998) and the re-

vised ADRI from Djankov, La Porta, Lopez-de-Silanes, and Shleifer (2008) ,

and find similar results.

4.4. Large shareholders and CSR

Besides firm-level liquidity ( Section 4.1 ), pay-for-

performance ( Section 4.2 ), and country-level shareholder

protection laws ( Section 4.3 ), another important gover-

nance mechanism affecting agency issues is the firm’s

ownership structure; in particular, the degree to which

control is concentrated in the hands of large sharehold-

ers and whether large shareholders can use excessive vot-

ing power to entrench themselves ( La Porta, Lopez-de-

Silanes, and Shleifer, 1999; Claessens, Djankov, Fan, and

Lang, 2002 ). Therefore, we further evaluate the agency ver-

sus good governance views on CSR from the perspective

of ownership and control. In countries other than the US,

the UK, and Australia, large firms typically have share-

holders that own a significant proportion of the equity

( Claessens, Djankov, Fan, and Lang, 2002 ). Ownership pat-

terns are rather stable in general, especially outside the

US, and are largely shaped by the companies’ histories and

their founding or controlling families ( La Porta, Lopez-de-

Silanes, Shleifer, and Vishny (2002) . Therefore, large share-

holders’ ownership concentration could be considered as

relatively exogenous to CSR decisions of a firm, and our

empirical analyses on the effects of large shareholders are

conducted in a setting without IVs.

The relation between the level of concentrated owner-

ship and firm-level agency problems is theoretically un-

clear. On the one hand, ownership in the hands of one or

a few large shareholders could create agency problems be-

tween controlling and minority shareholders ( Bebchuk and

Weisbach, 2010 ). The concern is diversion of firm value

from the minority to the controlling shareholder. The pos-

sibility of this expropriation, and hence this type of large

shareholder agency problem, can be augmented when the

firm’s free cash flow increases and leverage and dividend

payouts decrease (as there is now more to divert). On

the other hand, the controlling shareholders can effec-

tively steer managerial decision making and, hence, also

function as a mechanism to curb the managerial agency

problem (which Sections 4.1 –4.3 were mainly about). Ei-

ther way, large shareholders’ ownership and control can

shape the degree to which agency problems are present

within the firm and can also be used as a proxy for

firm-level agency problems. However, the two mechanisms

mentioned above—one capturing the major shareholder’s

incentive to monitor the manager to maximize firm value

(incentive effect) and the other capturing the large share-

holder’s expropriation of the rights of minority sharehold-

ers (expropriation effect)—lead to opposite predictions on

the relation between large shareholder ownership and CSR,

which creates an empirical challenge to directly test such

relation. One way to circumvent this problem is to dis-

entangle the incentive and expropriation effects of large

shareholders’ ownership, which can be achieved through

separating control rights from cash flow rights. Control-

ling shareholders can establish control over firms with only

minimal cash flow rights (ownership) when a deviation

from the one share, one vote rule applies ( La Porta, Lopez-

de-Silanes, and Shleifer 1999; Bebchuk, Kraakman, and Tri-

antis, 20 0 0; Claessens, Djankov, and Lang, 20 0 0; Faccio

and Lang, 2002; Lins, 2003 ). This wedge between large

A . F

e rre

ll e

t a

l. / Jo

u rn

a l o

f F

in a

n c ia

l E

c o

n o

m ic

s 1

2 2 (2

0 1 6

) 5

8 5

– 6

0 6

6 0

1

Table 9

Direct effects of legal protection of shareholder rights on CSR.

The dependent variables are various environmental, social, and governance (ESG) indices, and the key explanatory variables are the adjusted anti-director rights index (ADRI), anti-self-dealing index (ASDI), and

the public enforcement of the anti-self-dealing regulation. Control variables are legal origins (French, German, and Scandinavian; the English origin is taken as benchmark and omitted from regressions), Ln(GDP

per capita) per capita, return on assets (ROA) (winsorized at 5%), Tobin’s q (winsorized at 5%), financial constraints, interest coverage, current ratio , an ownership dispersion indicator, investment opportunities, and

year and industry fixed effects. Standard errors are clustered at the country level and reported in the parentheses. ∗∗∗, ∗∗ , and ∗ indicate significance at the 1%, 5%, and 10% level, respectively.

Panel A. Dependent variables are ESG ratings (overall ratings and subdimensional ratings) from the MSCI IVA sample

IVA rating EcoValue rating Social rating Labor relations Industry-specific carbon risks Environmental opportunities

Adjusted ADRI 0.297 ∗∗∗ 0.333 ∗∗∗ 0.269 ∗∗∗ 0.243 ∗∗∗ 0.221 ∗∗∗ 0.151 ∗∗∗

(0.110) (0.060) (0.055) (0.070) (0.053) (0.046)

ASDI 1.329 1.966 ∗∗∗ 1.184 1.003 1.302 ∗∗ 0.967 ∗∗∗

(1.325) (0.676) (1.174) (0.940) (0.489) (0.307)

Public enforcement 0.753 ∗∗∗ 0.158 0.725 ∗∗∗ 0.523 ∗∗∗ 0.004 −0.018 (0.229) (0.211) (0.208) (0.169) (0.202) (0.128)

Number of observations 25,449 25,549 25,549 48,858 48,958 48,958 32,495 32,483 32,483 32,504 32,604 32,604 40,508 40,606 40,606 47,976 48,075 48,075

Controls Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes

Year fixed effects Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes

Industry fixed effects Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes

R -squared 13.5% 12.2% 12.9% 18.3% 17.5% 16.3% 10.7% 9.5% 10.4% 14.0% 13.2% 13.5% 41.3% 41.6% 41.2% 27.3% 27.2% 27.0%

Panel B. Dependent variables are ESG ratings (overall and subdimensional ratings) from the Vigeo corporate ESG sample

Overall ESG Environment Human resources Customers and suppliers Human rights Community involvement

Adjusted ADRI 1.969 ∗∗∗ 2.789 ∗∗∗ 3.363 ∗∗∗ 0.980 2.558 ∗∗∗ 2.622 ∗∗∗

(0.585) (0.520) (1.123) (0.674) (0.811) (0.762)

ASDI −5.395 7.104 0.665 −3.116 −4.828 −7.227 (9.169) (10.904) (11.472) (9.148) (9.046) (10.608)

Public enforcement −0.323 −2.337 0.698 −1.623 0.908 1.325 (1.516) (1.711) (2.255) (1.376) (1.688) (1.384)

Number of observations 3,586 3,610 3,610 3,586 3,610 3,610 3,586 3,610 3,610 3,586 3,610 3,610 3,586 3,610 3,610 3,586 3,610 3610

Controls Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes

Year fixed effects Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes

Industry fixed effects Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes

R -squared 33.8% 32.2% 32.2% 28.5% 27.3% 27.4% 41.7% 39.7% 39.8% 18.7% 18.2% 18.3% 24.5% 23.0% 23.0% 27.7% 26.7% 26.7%

602 A. Ferrell et al. / Journal of Financial Economics 122 (2016) 585–606

shareholder voting and cash flow rights can single out the

expropriation effects of large shareholders and can help

capture the large shareholder agency problem. Claessens,

Djankov, Fan, and Lang (2002) separate the largest

shareholder’s voting rights and cash flow rights and find

that firm value increases with the cash flow ownership of

the largest shareholder, consistent with a positive incen-

tive effect, but that firm value falls when the control rights

of the largest shareholder exceed its cash flow ownership,

consistent with an expropriation effect.

We therefore test the relation between the largest

shareholder’s voting rights in excess of its cash flow

rights (wedge) and CSR using the ASSET4 sample, which

is composed of standardized data on the largest share-

holder’s voting and cash flow rights for a sample of com-

panies around the world. 11 Our model specifications follow

those of Claessens, Djankov, Fan, and Lang (2002), Morck,

Shleifer, and Vishny (1988) , and Bebchuk, Cohen, and Fer-

rell (2009) in that we also capture the non-monotonic ef-

fects of large shareholders’ cash flow rights by including

both the Largest shareholder’s ownership and its square. We

also control for country, industry, and year fixed effects

(the above studies controlled only for industry). Our main

explanatory variables are Wedge1 , which is the difference

between the largest shareholder’s voting and cash flow

rights, and Wedge2 , which is the ratio of voting rights to

cash flow rights. To control for doing good by doing well,

we include the equity market-to-book ratio and add the

standard control variables [used by Claessens, Djankov, Fan,

and Lang (2002) and Bebchuk, Cohen, and Ferrell (2009) ].

In the agency view of CSR, the controlling shareholders

can use their majority voting rights to expropriate minority

shareholders by approving CSR projects that benefit only

themselves. Therefore, a positive association between CSR

and the control wedge is expected under the agency view.

The results from our general least squares (GLS) regres-

sions are shown in Table 10 . Some interesting observations

can be made. First, throughout all specifications, the coeffi-

cients on both Wedge1 and Wedge2 are negative and signif-

icant. This negative sign does not support the agency view

that considers CSR spending as a corollary of controlling

shareholders’ entrenchment and possible expropriation of

minority shareholders. Second, the effect of the largest

shareholder’s ownership seems to be non-monotonic on

different aspects of CSR, as the coefficients of largest share-

holder’s ownership are all negative and significant, while

that of the square of ownership are all positive. This is con-

sistent with the previous literature in that both incentive

and entrenchment mechanisms of controlling sharehold-

ers affect corporate outcomes. The simplified specifications

(controlling only for performance) and the more complex

ones (also including other traditional financial controls)

11 The reason that we resort to the ASSET4 sample is that its CSR ratings

can be directly matched to the data on largest shareholders’ voting rights

and cash flow rights, and we can preserve the numbers of observations

to the largest extent compared with using other CSR samples. Neverthe-

less, when we conduct the same analysis with the MSCI IVA sample and

the Vigeo sample (but with smaller numbers of observations due to more

missing data on largest shareholders’ ownership and control), similar re-

sults are obtained. Tables with alternative CSR sample tests are available

upon request.

yield both qualitatively and quantitatively similar results,

although the sample size for the latter shrinks. These re-

sults also hold for various ESG sub-indices that we do not

report for reasons of conciseness. In terms of control vari-

ables, the positive coefficients on the equity market-to-book

mostly support the doing good by doing well conjecture.

Firm size and year since incorporation also have positive

loadings on CSR, indicating that larger and more estab-

lished companies are more likely to engage in CSR. Overall,

the direct effects of controlling shareholder ownership and

control (the wedge between voting and cash flow rights)

imply that CSR is not likely to be used as a self-serving

tool for controlling shareholders to extract private benefits,

shirk, or build empires, though large shareholders do not

seem to overspend on CSR (due to the internalization of

its costs). This reflects that a CSR policy is expensive but

does not by itself provide support for the agency view.

4.5. CSR, agency problems, and shareholder value

As a final extension, we consider the relations between

CSR, agency problems, and shareholder value together in a

cross-country setting, which has not been explored in the

literature. If CSR is not incompatible with good governance,

this should have value implications. We therefore further

explore the role of CSR in facilitating value enhancement

and also test whether CSR counterbalances the negative ef-

fects of agency problems and poor corporate governance

on firm value. To do so, we use the ASSET4 sample and uti-

lize data on several governance provisions under its corpo-

rate governance pillar to construct a global entrenchment

index (global E-index) as a proxy for poor governance. Our

global E-index follows the structure of the original US E-

index by Bebchuk, Cohen, and Ferrell. (2009) . We have

tried our best to mimic accurately the original E-index by

applying the same governance provisions across countries.

Only slight differences relative to the original US index oc-

cur due to data availability in Datastream. The provisions

in our global E-index include the presence of a poison pill,

a golden parachute, a classified board, other antitakeover

devices, and supermajority requirements for amending the

charter and bylaws. We term this E-index as Entrenchment

Index 1 . We also create Entrenchment Index 2 by replacing

the classified board in Entrenchment Index 1 by a staggered

board. 12

We conduct our test on a panel data set of more than

47 hundred of the largest public firms from 60 countries

in the ASSET4 sample from 2002 to 2013. Again, the main

reason for using the ASSET4 sample is that its CSR rat-

ings can be directly matched to the data on the E-index,

and we can thus preserve the number of observations to

the largest extent (compared with using other CSR sam-

ples). The dependent variable for all specifications is To-

bin’s q, defined as the ratio of market value of equity to

the book value of equity, winsorized at the 5% level. The

12 A classified board is a general term that refers to the situation in

which the terms of board directors can be different from each other (and

end in different years), while a staggered board refers to the situation

in which the terms of board directors are uniform (but end in different

years).

A . F

e rre

ll e

t a

l. / Jo

u rn

a l o

f F

in a

n c ia

l E

c o

n o

m ic

s 1

2 2 (2

0 1 6

) 5

8 5

– 6

0 6

6 0

3

Table 10

Direct effects of large shareholders’ ownership and control on corporate social responsibility (CSR).

The dependent variables are various environmental, social, and governance (ESG) indices from the ASSET4 sample, and the key explanatory variables are the l argest shareholder’s ownership (cash flow rights)

and its square and the wedge between the largest shareholder’s voting rights and cash flow rights. Wedge1 stands for voting rights minus cash flow rights, and Wedge2 stands for the ratio of voting rights to

cash flow rights. Control variables are Equity market-to-book (winsorized at 5%), the logarithm of total assets (size), the logarithm of firm age, annual sales growth rate (winsorized at 1%), and CapEx – to – sales ratio

(winsorized at 1%). All regressions control for country, industry, and time fixed effects. Standard errors are clustered at the firm level and reported in parentheses. ∗∗∗, ∗∗ , and ∗ indicate significance at the 1%, 5%, and 10% level, respectively.

Dependent variables are ESG ratings from the ASSET4 sample

Overall CSR rating Environmental rating Social rating

Ownership and control

Wedge1 (Voting - Cash flow rights) −0.118 ∗∗∗ −0.089 ∗∗ −0.072 ∗∗ −0.066 ∗ −0.088 ∗∗∗ −0.079 ∗∗ (0.032) (0.036) (0.031) (0.036) (0.031) (0.035)

Wedge2 (Voting / Cash flow rights) −0.002 ∗∗∗ −0.001 ∗∗∗ −0.002 ∗∗∗ −0.002 ∗∗∗ −0.001 ∗∗∗ −0.001 ∗∗ (0.0 0 02) (0.0 0 04) (0.0 0 02) (0.0 0 03) (0.0 0 02) (0.0 0 04)

Largest shareholder’s ownership −0.274 ∗∗∗ −0.278 ∗∗∗ −0.310 ∗∗∗ −0.315 ∗∗∗ −0.223 ∗∗∗ −0.215 ∗∗∗ −0.234 ∗∗∗ −0.232 ∗∗∗ −0.175 ∗∗∗ −0.181 ∗∗∗ −0.223 ∗∗∗ −0.226 ∗∗∗ (0.054) (0.054) (0.073) (0.073) (0.053) (0.054) (0.079) (0.078) (0.054) (0.054) (0.076) (0.076)

Largest shareholder’s ownership square 0.002 ∗∗∗ 0.002 ∗∗∗ 0.002 ∗∗∗ 0.003 ∗∗∗ 0.002 ∗∗∗ 0.002 ∗∗∗ 0.002 ∗∗ 0.002 ∗∗∗ 0.001 ∗∗ 0.001 ∗∗ 0.002 ∗∗ 0.002 ∗∗

(0.001) (0.001) (0.001) (0.001) (0.001) (0.001) (0.001) (0.001) (0.0 0 06) (0.0 0 06) (0.001) ()0.001

Control variables

Equity market-to-book 0.129 0.121 0.375 ∗∗ 0.376 ∗∗ −0.046 −0.052 0.352 ∗ 0.350 ∗ 0.168 0.162 0.470 ∗∗ 0.472 ∗∗ (0.134) (0.135) (0.189) (0.189) (0.132) (0.132) (0.181) (0.182) (0.135) (0.136) (0.197) (0.198)

Log(Size) 7.261 ∗∗∗ 7.265 ∗∗∗ 7.689 ∗∗∗ 7.691 ∗∗∗ 7.195 ∗∗∗ 7.199 ∗∗∗

(0.486) (0.486) (0.462) (0.461) (0.474) (0.473)

Log(Age) 3.940 ∗∗∗ 3.962 ∗∗∗ 2.647 ∗∗∗ 2.657 ∗∗∗ 2.919 ∗∗∗ 2.945 ∗∗∗

(0.614) (0.615) (0.607) (0.607) (0.617) (0.617)

Annual sales growth rate 0.002 0.002 −0.015 ∗∗∗ −0.015 ∗∗∗ −0.013 ∗∗ −0.013 ∗∗ (0.005) (0.005) (0.005) (0.005) (0.006) (0.006)

CapEx – to – sales ratio −0.077 ∗∗ −0.077 ∗∗ 0.012 0.012 −0.048 −0.048 (0.034) (0.033) (0.040) (0.040) (0.038) (0.037)

Constant −64.214 ∗∗∗ −64.822 ∗∗∗ −44.976 ∗∗∗ −45.233 ∗∗∗ −39.148 ∗∗∗ −39.790 ∗∗∗ (7.664) (7.665) (8.071) (8.046) (7.384) (7.372)

Number of Observations 18,905 18,894 9,064 9,060 19,467 19,456 9,193 9,189 19,467 19,456 9,193 9,189

Country, industry, year fixed effects Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes

R -squared 20.5% 20.4% 42.0% 41.8% 28.3% 28.3% 45.1% 45.0% 24.2% 24.2% 41.9% 41.8%

604 A. Ferrell et al. / Journal of Financial Economics 122 (2016) 585–606

Table 11

Corporate social responsibility (CSR), entrenchment, and firm value: ASSET4 sample.

The dependent variable is Tobin’s q (the ratio of equity market capitalization to equity book value) winsorized at 5% level for all regressions . Entrenchment

Index 1 is the sum of the following dummy variables from Datastream: the presence of a poison pill, a golden parachute, a supermajority requirement for

amending bylaws and charter, a classified board, and other anti-takeover provisions and treats non-available values as zeros. Entrenchment Index 2 has the

same composition as Entrenchment Index 1 , except that classified board (directors’ terms can be different) is replaced by staggered board (directors’ terms

are uniform). CSR is measured by ASSET4’s Overall CSR Rating for Columns 1 and 2, ASSET4’s aggregate Environmental Rating for Columns 3 and 4, and

ASSET4’s aggregate Social Rating for Columns 5 and 6. All specifications include country, industry, and year fixed effects. Standard errors are clustered at

the firm level and reported in parentheses. ∗∗∗, ∗∗ , and ∗ indicate significance at the 1%, 5%, and 10% level, respectively.

The world sample: Dependent variable = Tobin’s q Overall CSR Rating Environmental Rating Social Rating

Variable (1) (2) (3) (4) (5) (6)

Entrenchment Index 1 −0.0767 ∗∗ −0.0707 ∗∗∗ −0.0780 ∗∗∗ (0.0318) (0.0274) (0.0299)

Entrenchment Index 2 −0.0689 ∗∗ −0.0618 ∗∗ −0.0805 ∗∗∗ (0.0296) (0.0254) (0.0275)

CSR 0.0021 ∗∗ 0.0022 ∗∗ 0.0 0 05 0.0 0 07 0.0016 ∗ 0.0014 (0.0010) (0.0011) (0.0010) (0.001) (0.0010) (0.0010)

CSR × Entrenchment Index 0.0011 ∗∗ 0.0 0 08 ∗ 0.0012 ∗∗∗ 0.0 0 09 ∗∗ 0.0013 ∗∗∗ 0.0011 ∗∗∗ (0.0 0 05) (0.0 0 04) (0.0 0 04) (0.0 0 04) (0.0 0 04) (0.0 0 04)

Log(Assets) −0.2775 ∗∗∗ −0.2772 ∗∗∗ −0.2694 ∗∗∗ −0.2692 ∗∗∗ −0.2784 ∗∗∗ −0.2784 ∗∗∗ (0.0284) (0.0283) (0.0275) (0.0275) (0.0280) (0.0280)

Largest shareholder’s ownership 0.0017 0.0015 0.0 0 07 0.0 0 05 0.0 0 09 0.0 0 08

(0.0042) (0.0042) (0.0042) (0.0042) (0.0042) (0.0042)

Largest shareholder’s ownership square −0.0 0 0 0 −0.0 0 0 0 0.0 0 0 0 0.0 0 0 0 0.0 0 0 0 0.0 0 0 0 (0.0 0 01) (0.0 0 01) (0.0 0 01) (0.0 0 01) (0.0 0 01) (0.0 0 01)

Leverage ratio 0.0 0 08 0.0 0 08 0.0 0 05 0.0 0 05 0.0 0 05 0.0 0 05

(0.0029) (0.0029) (0.0029) (0.0029) (0.0029) (0.0029)

Dividend per share −0.0 0 0 0 −0.0 0 0 0 0.0 0 0 0 0.0 0 0 0 −0.0 0 0 0 −0.0 0 0 0 (0.0 0 01) (0.0 0 01) (0.0 0 01) (0.0 0 01) (0.0 0 01) (0.0 0 01)

ROE 0.0227 0.0226 0.0230 0.0229 0.0229 0.0229

(0.0150) (0.0150) (0.0150) (0.0150) (0.0151) (0.0151)

Year fixed effects Yes Yes Yes Yes Yes Yes

Country fixed effects Yes Yes Yes Yes Yes Yes

Industry fixed effects Yes Yes Yes Yes Yes Yes

Number of observations 16,077 16,077 16,278 16,278 16,278 16,278

R -squared 25.4% 25.4% 25.0% 25.0% 25.3% 25.3%

key explanatory variables are the global E-index, the CSR

rating (measured by ASSET4’s overall CSR score, environ-

mental score, and social score), and the interaction term

between the E-index and CSR. If CSR is induced by good

governance (at least those governance mechanisms related

to the efficient use of cash, the disbursement of earnings

to shareholders, and pay-for-performance), this can coun-

terbalance the negative impact of managerial agency prob-

lems otherwise induced by entrenchment (as proxied by

the E-index). We use standard financial controls, such as

firm size [measured as Log(Assets) ], the largest shareholder’s

ownership and its square, return on equity (ROE), leverage

ratio and dividends per share , as well as year, country, and

industry fixed effects .

The coefficients on the two measures of our global E-

index are mostly negatively related to Tobin’s q which

signifies that entrenchment reduces value ( Table 11 ). This

finding is consistent with the US results based on the orig-

inal E-index as in Bebchuk, Cohen, and Ferrell (2009) . The

effects of CSR on a non-entrenched firm (the noninteracted

CSR rating) are positively related to firm value for the over-

all and social CSR ratings or insignificantly so for the envi-

ronmental ratings specifications. We also show that CSR af-

fects firms with strong entrenchment. The interaction term

between CSR and the global E-index is positive and sig-

nificant for almost all CSR ratings (the environmental, so-

cial, and overall indices). This reinforces our earlier find-

ings supporting the good governance view and suggests

that CSR and the firm governance that induces CSR, in-

stead of being an agency problem, attenuates the nega-

tive effects of some types of agency problems (e.g., those

related to managerial entrenchment as proxied by the E-

index) on firm value. Potential endogeneity issues could

still exist, and unfortunately no single instrumental vari-

able is readily available for the interaction that captures all

aspects of CSR as well as of entrenchment. Moreover, the

entrenchment index is directly relevant for companies with

dispersed ownership, not those with controllers. Therefore,

the interaction results should be interpreted with caution.

Nevertheless, corporate charters and bylaws as well as the

use of the other antitakeover provisions are very stable

over time ( Bebchuk, Cohen, and Ferrell, 2009 ), which could

partly eliminate endogeneity concerns.

5. Conclusions

In most Anglo-American countries, consensus exists

that corporate governance is about “how investors get the

managers to give them back their money” ( Shleifer and

Vishny, 1997 , p.738). Corporate social responsibility, be-

cause of its focus on stakeholders in addition to share-

holders, is often considered as a form of cash diversion

A. Ferrell et al. / Journal of Financial Economics 122 (2016) 585–606 605

and an agency problem. In contrast to this agency perspec-

tive on CSR stands the good governance view, which states

that CSR activities are often adopted by firms character-

ized by good governance. In this debate, legal rules and

ownership structures are very different outside the Anglo-

American world, which significantly influences the execu-

tives’ incentives, the fiduciary duties of the management

and the board of directors, and the decision-making pro-

cess. The debate on the role of corporate social responsibil-

ity therefore often reflects the varieties of capitalism across

countries and the boundaries of the firm.

In this paper, we utilize public and proprietary data on

corporate compliance and engagement in stakeholder is-

sues to comprehensively assess the agency and good gover-

nance views of CSR. Our empirical set-up is well grounded

in fundamental economic theory: incentives, information

asymmetry, and control. We do not find empirical evi-

dence that CSR is associated with ex ante agency concerns,

such as abundance of cash (as proxied by cash holdings,

free cash flow, capital expenditures, dividend payout, and

leverage), or a weak connection between managerial pay

and corporate performance (as proxied by a total share-

holder return benchmark and excess CEO pay). Instead,

higher CSR performance is closely related to tighter cash

constraints—usually a proxy for better disciplined manage-

rial practice in the traditional corporate finance literature

( Jensen, 1986 )— and higher pay-for-performance sensitiv-

ity. In addition, CSR is positively related to legal protec-

tion of shareholder rights and negatively related to con-

trolling shareholders’ expropriation of minority sharehold-

ers. Whereas the vast majority of the literature has em-

phasized the agency costs of managerial entrenchment and

large shareholders’ control, as well as their economic con-

sequences such as distorting resource allocation and im-

peding economic growth, our empirical findings show that

these costs are at least not incurred through CSR activi-

ties. Moreover, we find evidence that a positive correla-

tion exists between CSR and Tobin’s q in firms with few

agency problems and that CSR and the firm governance

that induces CSR counterbalances the negative association

between firm value (proxied by Tobin’s q) and manage-

rial entrenchment (captured by the global entrenchment

index). Our empirical results (based on an instrumental

variables estimation) suggest that good governance causes

high CSR and that a firm’s CSR practice is not inconsis-

tent with shareholder wealth maximization, which induces

a positive stance on CSR, also found in Dimson, Karakas,

and Li (2015) , and Deng, Kang, and Low (2013) .

None of this is to say that more CSR is always better.

Undertaking some CSR activities can be driven by manage-

rial utility considerations, such as the satisfaction of some

personal or moral imperative of the manager, instead of

the enhancement of shareholder wealth ( Moser and Mar-

tin, 2012 ). Moreover, shareholders always internalize the

costs of CSR expenditures, and as their ownership stakes

increase, they can reduce spending on CSR. Our main ar-

gument is that, in general, corporate social responsibility

need not to be inevitably induced by agency problems but

can be consistent with a core value of capitalism, gener-

ating more returns to investors, through enhancing firm

value and shareholder wealth.

Taking the evidence in this paper at face value, several

policy implications emerge for the improvement of corpo-

rate governance, particularly in the area of corporate social

responsibility. Undoubtedly, governments have a responsi-

bility for dealing with market failures and externalities, but

government might not always be incentivized and effec-

tive in achieving this goal. Governments can be corrupt,

inefficient, and even predatory towards the private sector

( Shleifer and Vishny, 1998 ), in which case they fail to pro-

vide public goods. Therefore, corporate social responsibility

in the private sector—the private provision of public goods

( Kitzmueller and Shimshack, 2012 )—can be important for

preserving social welfare. While many researchers believe

that such private provision of public goods can be associ-

ated with agency problems that divert shareholder wealth

and even undermine the foundations of capitalism, we cast

doubt on such belief. Corporate governance reforms should

take into account such positive externalities.

Supplementary materials

Supplementary material associated with this article can

be found, in the online version, at doi:10.1016/j.jfineco.

2015.12.003 .

References

Adams, R. , Ferreira, D. , 2008. Do directors perform for pay? Journal of Ac-

counting and Economics 46 (1), 154–171 .

Albuquerque, R., Durnev A., Koskinen, T., 2015. Corporate social respon- sibility and firm risk: theory and empirical evidence. Unpublished

working paper, http://ssrn.com/abstract=1961971 . Bae, K. , Kang, J. , Wang, J. , 2011. Employee treatment and firm leverage: a

test of the stakeholder theory of capital structure. Journal of Financial Economics 100, 130–153 .

Bebchuk, L. , Coates IV, J. , Subramanian, G. , 2002. The powerful anti- takeover force of staggered boards: theory, evidence, and policy. Stan-

ford Law Review 54 (5), 887–951 .

Bebchuk, L. , Cohen, A. , Ferrell, A. , 2009. What matters in corporate gover- nance? Review of Financial Studies 22 (2), 783–827 .

Bebchuk, L. , Cremers, M. , Peyer, U. , 2011. The CEO pay slice. Journal of Financial Economics 102 (1), 199–221 .

Bebchuk, L. , Fried, J. , 2003. Executive compensation as an agency problem. Journal of Economic Perspectives 17, 71–92 .

Bebchuk, L. , Kraakman, R. , Triantis, G. , 20 0 0. Stock pyramids, cross-own-

ership, and dual class equity: the creation and agency costs of sepa- rating control from cash flow rights. In: Morck, R. (Ed.), Concentrated

Corporate Governance National Bureau of Economic Research. Cam- bridge, MA, pp. 445–460 .

Bebchuk, L. , Weisbach, M. , 2010. The state of corporate governance re- search. Review of Financial Studies 23 (3), 939–961 .

Benabou, R. , Tirole, J. , 2010. Individual and corporate social responsibility.

Econometrica 77, 1–19 . Berle, A. , Means, G. , 1932. The Modern Corporation and Private Property.

Harcourt, Brace & World, New York, NY . Brown, W. , Helland, E. , Smith, K. , 2006. Corporate philanthropic practices.

Journal of Corporate Finance 12, 855–877 . Bushee, B. , 20 0 0. Do institutional investors prefer near-term earnings over

long-term value? Contemporary Accounting Research 18 (2), 207–246 .

Bushee, B. , Noe, C. , 2001. Corporate disclosure practices, institutional in- vestors, and stock return volatility. Journal of Accounting Research 38,

171–202 . Cheng, I., 2011. Corporate governance spillovers, Unpublished working pa-

per, http://ssrn.com/abstract=1299652 . Cheng, I.-H. , Hong, H. , Shue, K. , 2014. Do managers do good with other

people’s money?. University of Chicago Booth School of Business,

Chicago, Ill Unpublished working paper, 12-47 . Chhaochharia, V. , Grinstein, Y. , 2009. CEO compensation and board struc-

ture. Journal of Finance 64, 231–261 . Claessens, S. , Djankov, S. , Fan, J. , Lang, L. , 2002. Disentangling the incen-

tive and entrenchment effects of large shareholders. Journal of Fi- nance 57 (6), 2741–2771 .

606 A. Ferrell et al. / Journal of Financial Economics 122 (2016) 585–606

Claessens, S. , Djankov, S. , Lang, L. , 20 0 0. The separation of ownership and control in Eastern Asian corporations. Journal of Financial Economics

58, 81–112 . Davies, P. , Schuster, E.-P. , Van de Walle de Ghelcke, E. , 2010. The takeover

directive as a protectionist tool. In: Ringe, W.G., Bernitz, U. (Eds.), Company Law and Economic Protectionism. Oxford University Press,

New York, NY, pp. 105–161 .

Deng, X. , Kang, J.-K. , Low, B. , 2013. Corporate social responsibility and stakeholder value maximization: evidence from mergers. Journal of

Financial Economics 110, 87–109 . Dhaliwal, D. , Li, O. , Tsang, A. , Yang, Y. , 2011. Voluntary nonfinancial disclo-

sure and the cost of equity capital: the initiation of corporate social responsibility reporting. Accounting Review 86 (1), 59–100 .

Di Giuli, A. , Kostovetsky, L. , 2014. Are red or blue companies more likely to go green? Politics and corporate social responsibility. Journal of Fi-

nancial Economics 111 (1), 158–180 .

Dimson, E. , Karakas, O. , Li, X. , 2015. Active ownership. Review of Financial Studies 28 (12), 3225–3268 .

Djankov, A. , La Porta, R. , Lopez-de-Silanes, F. , Shleifer, A. , 2008. The law and economics of self-dealing. Journal of Financial Economics 88,

430–465 . Edmans, A. , 2011. Does the stock market fully value intangibles? Employee

satisfaction and equity prices. Journal of Financial Economics 101 (3),

621–640 . El Ghoul, S. , Guedhami, O. , Kwok, C. , Mishra, D. , 2011. Does corporate so-

cial responsibility affect the cost of capital? Journal of Banking and Finance 35 (9), 2388–2406 .

Enriques, L. , Gilson, R. , Pacces, A. , 2014. The case for an unbiased takeover law. Harvard Business Law Review 41 (1), 85–127 .

Faccio, M. , Lang, L. , 2002. The ultimate ownership of western European

corporations. Journal of Financial Economics 65 (3), 365–395 . Fama, E. , 1980. Agency problems and the theory of the firm. Journal of

Political Economy 88 (2), 288–307 . Fama, E. , Jensen, M. , 1983. Separation of ownership and control. Journal

of Law and Economics 26, 301–325 . Flammer, C. , 2013. Corporate social responsibility and shareholder reac-

tion: the environmental awareness of investors. Academy of Manage-

ment Journal 56 (3), 758–781 . Foucalt, T. , Fresard, L. , 2014. Learning from peers’ stock prices and corpo-

rate investment. Journal of Financial Economics 111, 554–577 . Frydman, C. , Jenter, D. , 2010. CEO compensation. Annual Review of Finan-

cial Economics 2 (1), 75–102 . Goss, A. , Roberts, G. , 2011. The impact of corporate social responsibility

on the costs of bank loans. Journal of Banking and Finance 35 (7),

1794–1810 . Hall, B. , Liebman, J. , 1998. Are CEOs really paid like bureaucrats? Quarterly

Journal of Economics 113 (3), 653–691 . Hillman, A. , Keim, G. , 2001. Shareholder value, stakeholder management,

and social issues: what’s the bottom line? Strategic Management Jour- nal 22 (2), 125–139 .

Hoberg, G. , Phillips, G. , Prabhala, N. , 2014. Product market threats, pay-

outs, and financial flexibility. Journal of Finance 69 (1), 293–324 . Holmstrom, B. , Tirole, J. , 1989. The theory of the firm. In: Schmalensee, R.,

Willig, R. (Eds.), Handbook of Industrial Economics, Part 1. Elsevier Publishing B.V., Amsterdam Chapter 2 .

Hong, H., Kubik, J., Scheinkman, J., 2012. Financial constraints on corpo- rate goodness. Unpublished working paper, http://ssrn.com/abstract=

1734164 . Jensen, M. , 1986. Agency costs of free cash flow, corporate finance, and

takeovers. American Economic Review 76 (2), 323–329 .

Jensen, M. , 2001. Value maximization, stakeholder theory and the cor- porate objective function. Journal of Applied Corporate Finance (Fall)

8–21 . Jensen, M. , Meckling, W. , 1976. Theory of the firm: managerial behavior,

agency costs, and ownership structure. Journal of Financial Economics 3 (4), 305–360 .

Jensen, M. , Murphy, K. , 1990. Performance pay and top-management in-

centives. Journal of Political Economy 98 (2), 225–264 .

Kitzmueller, M. , Shimshack, J. , 2012. Economic perspectives on corporate social responsibility. Journal of Economic Literature 50 (1), 51–84 .

Krueger, P. , 2015. Corporate goodness and shareholder wealth. Journal of Financial Economics 115 (2), 304–325 .

La Porta, R. , Lopez-de-Silanes, F. , Shleifer, A. , 1999. Corporate ownership around the world. Journal of Finance 54 (2), 471–517 .

La Porta, R. , Lopez-de-Silanes, F. , Shleifer, A. , 2006. What works in securi-

ties laws? Journal of Finance 61 (1), 1–32 . La Porta, R. , Lopez-de-Silanes, F. , Shleifer, A. , Vishny, R. , 1997. Legal deter-

minants of external finance. Journal of Finance 52 (3), 1131–1150 . La Porta, R. , Lopez-de-Silanes, F. , Shleifer, A. , Vishny, R. , 1998. Law and

finance. Journal of Political Economy 106 (6), 1113–1155 . La Porta, R. , Lopez-de-Silanes, F. , Shleifer, A. , Vishny, R. , 20 0 0. Agency

problems and dividend policies around the world. Journal of Finance 55 (1), 1–33 .

La Porta, R. , Lopez-de-Silanes, F. , Shleifer, A. , Vishny, R. , 2002. In-

vestor protection and corporate valuation. Journal of Finance 57 (3), 1147–1170 .

Laeven, L. , Levine, R. , 2009. Bank governance, regulation, and risk taking. Journal of Financial Economics 93, 259–275 .

Leary, M. , Roberts, M. , 2014. Do peer firms affect corporate financial pol- icy? Journal of Finance 69 (1), 139–178 .

Lee, D. , Faff, R. , 2009. Corporate sustainability performance and idiosyn-

cratic risk: a global perspective. Financial Review 44 (2), 213–237 . Liang, H. , Renneboog, L. , 2016. On the foundations of corporate social re-

sponsibility. Journal of Finance forthcoming . Lin, C. , Malatesta, P. , Xuan, Y. , 2011. Ownership structure and the cost of

corporate borrowing. Journal of Financial Economics 100, 1–23 . Linck, J. , Netter, J. , Yang, T. , 2012. The determinants of board structure.

Journal of Financial Economics 87 (2), 308–328 .

Lins, K. , 2003. Equity ownership and firm value in emerging markets. Journal of Financial and Quantitative Analysis 38, 159–184 .

Macintosh, J. , 1999. The issues, effects, and consequences of the Berle– Dodd debate, 1931-1932. Accounting, Organizations and Society 24,

139–153 . Masulis, R. , Reza, S.W , 2015. Agency problem of corporate philanthropy.

Review of Financial Studies 28 (2), 592–636 .

Masulis, R. , Wang, C. , Xie, F. , 2009. Agency problems at dual-class compa- nies. Journal of Finance 64, 1697–1727 .

Morck, R. , Shleifer, A. , Vishny, R. , 1988. Management ownership and mar- ket valuation: an empirical analysis. Journal of Financial Economics

20, 293–315 . Morck, R. , Yeung, B. , 2005. Dividend taxation and corporate governance.

Journal of Economic Perspectives 19 (3), 163–180 .

Moser, D. , Martin, P. , 2012. A broad perspective on corporate social responsibility research in accounting. Accounting Review 87 (3),

797–806 . New York Times Magazine , 1970. The Social Responsibility of Business is

to Increase Profits. M. Friedman, pp. 122–126. September 13 . Parsons, C., Sulaeman, J., Titman, S., 2014. The geography of financial mis-

conduct. National Bureau of Economic Research, Cambridge, MA. Un-

published working paper, 20347. Peters, F. , Wagner, A. , 2014. The executive turnover risk premium. Journal

of Finance 69, 1529–1563 . Prendergast, C. , 1999. The provision of incentives in firms. Journal of Eco-

nomic Literature 37 (1), 7–63 . Servaes, H. , Tamayo, A. , 2014. How do industrial peers respond to control

threat? Management Science 60 (2), 380–399 . Shleifer, A. , Vishny, R. , 1989. Management entrenchment: the case of man-

ager-specific investment. Journal of Financial Economics 25, 123–139 .

Shleifer, A. , Vishny, R. , 1997. A survey of corporate governance. Journal of Finance 52 (2), 737–783 .

Shleifer, A. , Vishny, R. , 1998. The Grabbing Hand: Government Pathologies and Their Cures. Harvard University Press, Cambridge, MA .

Spamann, H. , 2010. The ‘antidirector rights index’ revisited. Review of Fi- nancial Studies 23 (2), 467–486 .