Strategic Management Journal

Strat. Mgmt. J., 27: 1101–1122 (2006)

Published online 13 September 2006 in Wiley InterScience (www.interscience.wiley.com) DOI: 10.1002/smj.557
Received 21 November 2003; Final revision received 3 March 2006

BEYOND DICHOTOMY: THE CURVILINEAR

RELATIONSHIP BETWEEN SOCIAL RESPONSIBILITY
AND FINANCIAL PERFORMANCE
MICHAEL L. BARNETT1 * and ROBERT M. SALOMON2
1
College of Business Administration, University of South Florida, Tampa, Florida,
U.S.A.
2
Stern School of Business, New York University, New York, New York, U.S.A.

A central and contentious debate in many literatures concerns the relationship between financial
and social performance. We advance this debate by measuring the financial–social performance
link within mutual funds that practice socially responsible investing (SRI). SRI fund managers
have an array of social screening strategies from which to choose. Prior studies have not
addressed this heterogeneity within SRI funds. Combining modern portfolio and stakeholder
theories, we hypothesize that the financial loss borne by an SRI fund due to poor diversification is
offset as social screening intensifies because better-managed and more stable firms are selected
into its portfolio. We find support for this hypothesis through an empirical test on a panel of
61 SRI funds from 1972 to 2000. The results show that as the number of social screens used
by an SRI fund increases, financial returns decline at first, but then rebound as the number of
screens reaches a maximum. That is, we find a curvilinear relationship, suggesting that two long-
competing viewpoints may be complementary. Furthermore, we find that financial performance
varies with the types of social screens used. Community relations screening increased financial
performance, but environmental and labor relations screening decreased financial performance.
Based on our results, we suggest that literatures addressing the link between financial and
social performance move toward in-depth examination of the merits of different social screening
strategies, and away from the continuing debate on the financial merits of either being socially
responsible or not. Copyright  2006 John Wiley & Sons, Ltd.

Are financial and social performance negatively study of socially responsible investing (SRI). SRI
or positively associated? Extant theoretical and is the practice of choosing financial investments on
empirical research has supported both contradic- the basis of social responsibility criteria. By some
tory positions (Margolis and Walsh, 2003; Or- accounts more than $1 trillion, or about 10 percent
litzky, Schmidt, and Rynes, 2003; Rowley and of all U.S. assets under management, including
Berman, 2000; Mahon and Griffin, 1999; Roman, about 160 mutual funds, can be categorized as SRI
Hayibor, and Agle, 1999; Griffin and Mahon, (Glassman, 1999; Hutton, D’Antonio, and Johnsen,
1997; Ullmann, 1985). In this paper, we recon- 1998; LaRose, 1998).
cile these divergent views through an empirical Many scholars have compared the financial per-
formance of SRI funds to those of funds that do not
Keywords: stakeholder theory; modern portfolio the- screen their holdings based on social criteria (Stat-
ory; corporate social responsibility; socially responsible man, 2000; Guerard, 1997; Sauer, 1997; Kurtz and
investing DiBartolomeo, 1996; Diltz, 1995; Hamilton, Jo,
∗
Correspondence to: Michael L. Barnett, College of Business and Statman, 1993; Luck and Pilotte, 1993; Teper,
Administration, University of South Florida, 4202 E. Fowler
Ave., BSN 3527, Tampa, FL 33620-5500, U.S.A. 1992; Mueller, 1991; Rudd, 1979). As with the
E-mail: [email protected] broader debate on the link between financial and

Copyright  2006 John Wiley & Sons, Ltd.

1102 M. L. Barnett and R. M. Salomon

social performance, the results of SRI and non-SRI moderate levels of social responsibility. In addi-
fund performance comparisons have been mixed. tion, we find that some types of social responsibil-
Many studies have shown that SRI funds can per- ity are linked to higher financial performance than
form as well as (e.g., Guerard, 1997; Diltz, 1995; others. We conclude by discussing the implications
Hamilton et al., 1993), and even better than (e.g., of these findings for the future of research on the
Statman, 2000; DiBartolomeo and Kurtz, 1999; link between social and financial performance.
Luck and Pilotte, 1993) unscreened funds. But
other studies have shown that SRI funds perform
THE LINK BETWEEN SOCIAL AND
worse than unscreened funds (Geczy, Stambaugh,
FINANCIAL PERFORMANCE
and Levin, 2003; Teper, 1992; Rudd, 1979). More-
over, SRI critics have suggested that the strong The nature of the relationship between the socially
financial performance of some SRI funds in recent beneficial behaviors of a corporation and its finan-
years may be the result of a decrease in the strin- cial performance has long been debated, yet it
gency of their social screening criteria (Glassman, remains unresolved (Margolis and Walsh, 2003).
1999; Goetz, 1997).1 That is, they contend that One group of scholars has argued, simply, that
SRI funds lowered their social performance in social responsibility detracts from a firm’s finan-
order to raise their financial performance. Thus, cial performance (Friedman, 1970; McWilliams
the stronger financial performance of SRI funds in and Siegel, 1997; Jensen, 2002). Any discretionary
recent years could actually serve as further evi- expenditures on social betterment unnecessarily
dence that financial and social performance are raise a firm’s costs, thereby putting it at an eco-
negatively, not positively, related. Indeed, SRI nomic disadvantage in a competitive market. In
funds vary greatly in the type and intensity of contrast, another group of scholars has argued that
social screens applied to their investments. Prior the better a firm’s social performance, the bet-
empirical studies have not addressed this hetero- ter it can attract resources (Cochran and Wood,
geneity, and so have confounded a range of social 1984; Waddock and Graves, 1997), obtain qual-
responsibility practices. ity employees (Greening and Turban, 2000; Tur-
In order to advance this long-standing and ban and Greening, 1996), market its products
contentious debate, rather than again comparing and services (Fombrun, 1996; Moskowitz, 1972),
socially screened to unscreened mutual funds, we and even create unforeseen opportunities (Fom-
address differences within SRI funds. In this paper, brun, Gardberg, and Barnett, 2000). Thus, social
we measure how variation in the intensity and type responsibility is a source of competitive advan-
of social screening employed by SRI funds affects tage (Porter, 1991; Porter and van der Linde,
their financial performance. We first review the 1995). Empirical tests of these opposing positions
debate over the link between financial and social have long produced mixed results, and so have
performance. We then develop a set of hypothe- not resolved this debate (Ullmann, 1985; Grif-
ses, grounded in modern portfolio theory (Camp- fin and Mahon, 1997; Margolis and Walsh, 2003;
bell et al., 2001; Fama, 1971; Markowitz, 1952; McWilliams and Siegel, 2000; Wood and Jones,
Sharpe, 1964) and stakeholder theory (Freeman, 1995).
1984; Donaldson and Preston, 1995; Jones, 1995), In this paper, we test the relationship between
that predict how variation in both the intensity and social and financial performance within mutual
type of social screening influences risk-adjusted funds. Mutual funds seek to maximize perfor-
financial performance. We test these hypotheses mance across a portfolio of firms, not within a
on a panel of 61 SRI funds. We find that the rela- single firm. As with the firm-level debate, the basic
tionship between financial and social performance issue concerns whether the costs of social respon-
sibility are offset or exceeded by financial returns
is neither strictly negative nor strictly positive.
over some period of time. However, mutual funds
Rather, it is curvilinear, with the strongest financial
are also concerned with diversification (Sharpe,
returns to low and high levels of social responsi-
1964; Black, Jensen, and Scholes, 1972; Campbell
bility, and significantly lower financial returns to
et al., 2001; Geczy et al., 2003).2 If a mutual fund

2
Throughout this paper, we use diversification to mean efforts
1
Refer to Figure 3 for consideration of this argument. of a mutual fund to invest in a broad set of firms so as to create a

Copyright  2006 John Wiley & Sons, Ltd. Strat. Mgmt. J., 27: 1101–1122 (2006)
DOI: 10.1002/smj
 Curvilinear Relationship between Social Responsibility and Financial Performance 1103

implements strict social performance criteria that costs and so, all else equal, below-average finan-
exclude firms, industries, or sectors from its portfo- cial performance. Thus, as one SRI critic bluntly
lio, that mutual fund may be unable to adequately concluded, ‘Socially conscious investing is a dumb
diversify. Without ample diversification, the fund idea, yielding sub-par returns, and screaming with
will be exposed to additional risk for a given level contradictions’ (Rothchild, 1996: 197).
of return and so by definition will incur a loss Moreover, screening may involve the exclusion
in risk-adjusted financial returns. SRI proponents of not merely certain firms, but entire industries
argue, though, that while there may be less poten- and even economic sectors from the portfolios
tial breadth in an SRI fund’s portfolio, those firms of SRI funds. For example, the tobacco indus-
that are chosen for the portfolio are substantively try is commonly screened out of SRI funds, and
better managed than the average firm and so tend to the entire defense sector is excluded from many
generate equal or higher financial returns, even on SRI funds (Social Investment Forum, 2002). The
a risk-adjusted basis. The remainder of this section exclusion of firms, industries, and economic sec-
reviews these opposing views on the link between tors has significant implications for the financial
social and financial performance and develops a set performance of an investment portfolio, regard-
of hypotheses that predict the financial outcomes of less of its social orientation. According to mod-
variability in intensity and type of social screens. ern portfolio theory, an investment portfolio bears
two types of risk: systematic and unsystematic, or
‘specific,’ risk (Markowitz, 1952; Sharpe, 1964;
The financial costs of social responsibility Fama, 1971). Systematic risk is the risk inher-
Critics of corporate social responsibility point ent in the volatility of the entire capital market,
out that it is costly and administratively burden- while specific risk is associated with the volatility
some for a firm to engage in socially responsi- of an individual security. Investors may assemble
ble practices such as doling out corporate philan- portfolios in such a way that the specific risk car-
thropy, providing employee day care, granting paid ried by any individual security within the portfolio
parental leave, and reducing environmental impact. is offset by the specific risk carried by another.
These additional costs and administrative burdens This is referred to as diversification. Efficient cap-
directly detract from the bottom line and so can ital markets reward investors for bearing system-
put socially responsible firms at a competitive dis- atic risk, but because diversification is possible,
advantage relative to rivals who do not engage in investors are not rewarded for bearing specific
such practices (Friedman, 1970; McWilliams and risk. That is, when a fund carries specific risk,
Siegel, 1997; Jensen, 2002). Through a process it fails to reach the efficient frontier, wherein the
termed ‘screening,’ SRI funds restrict their invest- risk/return trade-off is optimized. Because they
ments to those firms that engage in these costly exclude certain firms, industries, and sectors, SRI
and burdensome social practices: funds thus tend to bear a substantial degree of spe-
cific risk (Kurtz and DiBartolomeo, 1996; DiBar-
Screening describes the inclusion or exclusion of tolomeo and Kurtz, 1999), and so should experi-
corporate securities in investment portfolios based ence decreased risk-adjusted returns.
on social or environmental criteria. Socially con- However, a mutual fund can achieve diversifi-
cerned investors generally seek to own profitable cation ample to effectively eliminate most specific
companies with respectable employee relations, risk even if it does not select the entire universe
strong records of community involvement, excel-
lent environmental impact policies and practices, of securities. The traditional ‘rule of thumb’ in
respect for human rights around the world, and safe the finance literature is that a fund can closely
and useful products. Conversely, they often avoid approximate a well-diversified portfolio with as
investments in those firms that fall short in these few as 20 or 30 randomly selected stocks (Fisher
areas. (Social Investment Forum, 2002) and Lorie, 1970; Bloomfield, Leftwich, and Long,
1977). More recently, due to increasing volatility
As a result, SRI funds intentionally select firms in the stock market, researchers have concluded
that are likely to have above-average operating that the minimum number of randomly selected
stocks necessary to closely approximate a well-
portfolio of investments that eliminate unsystematic, or specific, diversified portfolio is at least 50 (Campbell et al.,
risk. 2001), and some have estimated this figure to be as
Copyright  2006 John Wiley & Sons, Ltd. Strat. Mgmt. J., 27: 1101–1122 (2006)
DOI: 10.1002/smj
1104 M. L. Barnett and R. M. Salomon

high as 200 (Statman, 1987). Regardless, one need without social screening. The DSI, outperformed
not hold the entire universe of stocks to be suffi- the S&P 500 index from its inception in May 1990
ciently diversified. However, the subset selected through March 1999, earning a total of 470 per-
for the portfolio needs to be randomly chosen for cent as compared to 389 percent for the S&P 500
this rule of thumb to hold, and even then, some (DiBartolomeo and Kurtz, 1999). Even on a risk-
specific risk remains. For example, Campbell et al. adjusted basis, the DSI’s financial performance
(2001) found that a random portfolio of 50 securi- exceeded that of the unscreened S&P 500 (Luck
ties still bore a 5 percent excess standard deviation and Pilotte, 1993; Statman, 2000).
relative to the market portfolio. How can SRI funds possibly earn equal or higher
SRI portfolios, as with the holdings of many risk-adjusted returns than unconstrained funds?
other mutual funds, are not randomly chosen. They Though modern portfolio theory rightfully assesses
are intentionally selected based on a set of screen- the costs to limiting investment choices through
ing criteria. Thus, one can expect SRI funds, even social screening, it does not account for the bene-
those with large and relatively diverse holdings, fits that social screening may bring. Portfolio the-
to bear specific risk (Kurtz, 1997). For exam- ory assesses only the ability of a given stock to
ple, the Domini Social Index (DSI), which serves push a portfolio toward or away from the efficient
as the benchmark portfolio for socially respon- frontier, wherein risk-adjusted return is maximized
sible investing and is the SRI fund most often (Markowitz, 1952). However, it takes no account
criticized for having broad holdings (Glassman, of any variation in the ability of a firm upon which
1999; Goetz, 1997), holds some 400 non-randomly a stock’s value is based to create value. Rather,
selected stocks. Measures of both its beta and stan- under the assumption of perfectly efficient mar-
dard deviation of returns have shown that it is kets, each stock is treated as homogeneous in all
riskier than the S&P 500 (Statman, 2000). Other but its volatility relative to the market.3
researchers have also found that there are finan- SRI proponents counter that, while SRI portfolio
cial costs associated with the lack of diversifica- managers are constrained from choosing amongst
tion of SRI funds. Teper (1992) estimated that the entire universe of stocks, the pool of stocks
funds that chose their portfolios based on social from which they do choose is superior to that
criteria bore a one percent loss in returns rela- of the overall market and therein more likely
tive to diversified funds. Rudd (1979) measured to provide favorable financial returns over time.
the returns to portfolios that screened out firms Firms are embedded in a social environment (Gra-
with holdings in South Africa and found that they novetter, 1985; Scott, 1981). In order to main-
suffered a 4 percent loss in returns. More recently, tain legitimacy and effectively attract resources,
Geczy et al. (2003) found a range of losses to risk- firms must build favorable relations with those
adjusted return, from just a few basis points per groups that compose this environment. Strong
month, to more than 1500 basis points per month. social performance is an indicator that a firm pos-
Thus, a substantial body of both theoretical and sesses superior management talent (Alexander and
empirical research in modern portfolio theory indi- Bucholtz, 1978; Bowman and Haire, 1975) that
cates that SRI funds are bound to suffer a financial understands how to improve internal and external
loss of some magnitude due to inadequate diversi- relationships through socially responsible activities
fication. (Moskowitz, 1972). Thus, SRI proponents argue
that because social relationships matter to financial
The financial benefits of social responsibility
Despite the financial logic of modern portfolio the- 3
We recognize that some form of market failure is central to
ory, many researchers have found that SRI funds a strategy literature in which firms can achieve above-average
yield returns that equal or exceed those of mutual returns (see Barney, 1991; Dierickx and Cool, 1991; Werner-
funds that operate without the constraints of social felt, 1984; Williamson, 1975, 1985). This presents an obvious
dilemma relative to an efficient markets hypothesis (EMH),
responsibility. For example, Diltz (1995), Guer- the resolution of which is outside the scope of this paper.
ard (1997), and Hamilton et al. (1993) all found There remains considerable debate in the finance literature as
that there were no significant differences between to whether, and in what form, the EMH holds. In fact, this has
been one of most researched topics in the field of finance (see
the risk-adjusted returns of portfolios composed of Fama, 1991, for a review). The interested reader is encouraged
socially responsible firms and portfolios selected to consult this literature for more on this debate.

Copyright  2006 John Wiley & Sons, Ltd. Strat. Mgmt. J., 27: 1101–1122 (2006)
DOI: 10.1002/smj
 Curvilinear Relationship between Social Responsibility and Financial Performance 1105

performance, social responsibility is not merely a and Walsh, 2003: 277). Orlitzky et al. (2003)
cost, but a wise investment. recently conducted a meta-analysis that led to a
This basic rationale is supported by stakeholder similar conclusion: the prior literature, in aggre-
theory, which suggests that the better a firm man- gate, indicates that social and financial perfor-
ages its relationships with the myriad groups that mance are positively related. However, as Margolis
have some interest, or ‘stake,’ in the firm, the and Walsh (2003: 278) caution, such a conclu-
better its financial performance over time (Don- sion is illusory. A compilation of findings cannot
aldson and Preston, 1995; Freeman, 1984). For produce a definitive conclusion given the limita-
example, a firm with a favorable work environ- tions of the underlying studies. As many reviews
ment can decrease its hiring costs and increase of this body of literature note, these studies are
its employee retention rate, decrease community imperfect in a variety of ways (see Wood and
opposition and legal costs when opening a new Jones, 1995; Griffin and Mahon, 1997; Rowley and
factory, and more easily lobby for tax breaks from Berman, 2000). Thus, Margolis and Walsh (2003:
local governments (Freeman, 1984; Waddock and 278) argue that ‘[t]he CSP-CFP empirical liter-
Graves, 1997). A favorable social agenda builds ature reinforces, rather than relieves the tension
valuable goodwill that can buffer a firm from surrounding corporate responses to social misery.’
unforeseen problems and even provide valuable Overall, despite all the attention to the topic, the
new opportunities not available to less socially nature of the relationship remains contested.
responsible firms (Fombrun et al., 2000). All in Further fueling the debate, some critics of SRI
all, effective stakeholder management can create studies have argued that the strong financial perfor-
competitive advantage. Empirical results bear this mance of some SRI funds could be an indication
out. Graves and Waddock’s (2000) study of ‘built that the relationship between social responsibil-
to last’ companies suggested that the investments ity and financial performance is actually nega-
in stakeholder relations made by these firms led to tive. When SRI funds were first introduced, they
their above-average financial performance over an were ‘the butt of Wall Street jokes’ (Glassman,
8-year window, as measured by return on equity, 1999: 4) because their financial returns were often
return on assets, and return on sales. Hillman and quite poor. These critics suggest that many SRI
Keim’s (2001) study of the market value added of funds have become strong financial performers
308 firms within the S&P 500 found that effective only because their ‘once-strict screening criteria
stakeholder management was significantly corre- have turned porous’ (Goetz, 1997: 43). That is,
lated with, and preceded improved financial per- the improved financial performance of SRI funds
formance. Thus, even though SRI funds must draw is a result of gradually minimizing social perfor-
from a limited pool of firms, they draw from a mance standards for those firms to be included in
richer pool—one that is more likely to contain their portfolios. SRI funds have ‘opened the door
well-run, stable firms that outperform the broader to less-than-angelic companies whose high returns
market over the long run. The competitive advan- have helped SRI gain the upper hand in the long-
tage these individual firms possess aggregate into standing performance debate’ (Goetz, 1997: 43).
superior financial returns at the portfolio level. Therefore, social performance must indeed be sac-
rificed to gain financial returns.
This argument points to the need to account for
Variable financial returns to social
heterogeneity in the standards of social responsibil-
responsibility
ity employed by SRI funds. The standard approach
Many studies have been published supporting in research on SRI funds is to contrast the financial
a negative relationship, and many studies have performance of a set of screened funds with that
been published supporting a positive relationship. of a set of unscreened funds or the overall market
Which is it? If one sums up this conflicting empir- (e.g., Guerard, 1997; Hamilton et al., 1993). This
ical work, it appears that the relationship is pos- approach confounds a range of screening prac-
itive on the whole: ‘A simple compilation of the tices within SRI funds. Because some SRI funds
findings suggests there is a positive association, have more stringent social screening standards
and certainly very little evidence of a negative than others, the SRI literature must examine vari-
association, between a company’s social perfor- ances within screened funds to better determine
mance and its financial performance’ (Margolis the underlying nature of the relationship between
Copyright  2006 John Wiley & Sons, Ltd. Strat. Mgmt. J., 27: 1101–1122 (2006)
DOI: 10.1002/smj
1106 M. L. Barnett and R. M. Salomon

financial and social performance. The uncertainty with poor stakeholder relations are more risky and
of extant empirical results, despite decades of susceptible to crises (Cornell and Shapiro, 1987;
study, may be a function of treating social respon- Fombrun et al., 2000), and so more likely to be
sibility as a dichotomous variable. in the left tail of the distribution. As a result of
If the heterogeneity in the intensity of social using social screens that exclude firms with poor
screens applied by SRI funds is accounted for, the stakeholder relations and funnel in firms with good
combination of modern portfolio and stakeholder stakeholder relations, SRI fund managers become
theories points toward neither a strictly positive nor more likely to select firms that will achieve above-
negative relationship, but a curvilinear relationship average returns, and less likely to select firms
between social and financial performance. Based that will earn below-average returns.4 Figure 1
on the efficient market assumption that underlies illustrates this relationship.
modern portfolio theory, consider the entire uni- The combination of modern portfolio theory and
verse of stocks to have a uniform distribution of stakeholder theory, as well as a long history of
returns. Those stocks in the center of the distribu- mixed empirical findings, then suggests that the
tion earn the market return, while those in the left relationship between social and financial perfor-
tail earn less, and those in the right tail earn above- mance may be curvilinear, not strictly monotonic.
average returns. A fund manager taking random SRI funds that have relatively weak social respon-
draws from this universe can expect to assem- sibility standards will be able to choose from a
ble a portfolio that will earn the market return, larger universe of potential investments, thereby
so long as the resulting portfolio is diversified. increasing their odds of achieving ample diversi-
If the stocks picked do not sum to a diversified fication and hence improving risk-adjusted finan-
portfolio, the fund carries unsystematic risk and cial performance. As an SRI fund’s social stan-
so can expect to have a risk-adjusted return that dards increase, its pool of investment opportunities
underperforms the market. Since social screening shrinks, and so it will have a decreased likeli-
systematically constrains the ability to diversify, hood of establishing a well-diversified portfolio.
an SRI fund is thus expected to underperform the However, this negative effect is offset as the strin-
market. gency of social screening intensifies. Those funds
However, a fund manager using social screens that greatly restrict potential investments bene-
may have better odds of avoiding stocks in the fit from improved selection of investment targets
left tail of the distribution and picking stocks in
the right tail. Based on stakeholder theory, we
4
expect that firms engaging in socially responsible This argument does not address skill differences across fund
managers. Rather, it is premised on averages; given that the pool
practices are more likely to achieve superior long- of firms from which SRI fund managers select is richer, better
run performance (Freeman, 1984; Jones, 1995; returns are more likely, regardless of individual fund manager
Wicks, Berman and Jones, 1999). Thus, socially characteristics. Nonetheless, it seems reasonable to conclude,
and some have argued, that the actions inherent in screening
responsible firms are more likely to be in the can provide useful information to fund managers about a target
right tail of the distribution. In contrast, firms firm’s relationships with its stakeholders (e.g., Lowry, 1991).

Unscreened
Increasing Diversification

Increasing Selectivity

Low Screening

Moderate Screening

Strict Screening

Figure 1. The effects of social screening on the universe of stock choices

Copyright  2006 John Wiley & Sons, Ltd. Strat. Mgmt. J., 27: 1101–1122 (2006)
DOI: 10.1002/smj
 Curvilinear Relationship between Social Responsibility and Financial Performance 1107

(Lowry, 1991). Though an SRI fund may bear from a broad set of stakeholders and a lim-
more and more specific risk by choosing from ited budget, firms must decide where to allocate
an increasingly smaller pool of stocks, the pool their resources. Which investments in stakeholder
from which it does choose becomes richer. As relationship building are likely to generate the
a fund manager dips into this increasingly rich greatest financial returns?
pool, he/she is more likely to pick a stock that Instrumental stakeholder theory provides a the-
will provide above-average financial returns. SRI oretical basis for predicting the varying financial
funds that are ‘stuck in the middle’ may bear all the implications of differing types of social perfor-
costs of either pure strategy without gaining any of mance. It points out that some stakeholder relation-
the benefits. That is, an SRI fund with a moderate ships are more instrumental to a firm’s success than
level of social screening may bear specific risk yet others. The more a firm relies on a particular stake-
not consistently exclude underperforming firms or holder group, the more that firm stands to gain by
consistently select those firms with above-average investing in the creation and maintenance of trust-
financial performance. Thus, we hypothesize: ing relations with that group (Preston and Post,
1975; Waddock and Graves, 1997; Wicks et al.,
1999). Employees, in particular, are instrumental
Hypothesis 1: The relationship between the in-
to a firm’s financial performance. Employees con-
tensity of social screening and financial perfor-
stitute the ‘front line’ of the firm, and are responsi-
mance for SRI funds is curvilinear (U-shaped).
ble for transforming the firm’s inputs into outputs.
As we have moved further and further away from
The financial returns to different types of the industrial age and begun to rely more and
social responsibility more on the knowledge and creativity of employ-
ees to create value, labor relations have become
SRI funds vary not only in the intensity of their increasingly important (Florida, 2002). Positive
social screening, but also in the types of social labor relations can facilitate increased productivity,
screens they employ. SRI fund investors can decreased turnover, and decreased strife (Freeman,
choose from a variety of funds tailored to a specific 1984). For example, during a period when most
social issue or group of social issues. For exam- major airlines suffered one or more strikes, South-
ple, the Aquinas line of SRI funds pursue invest- west Airlines, a firm with much-heralded labor
ment in only those firms that they deem to amply relations, avoided such disruptions and maintained
reflect Catholic religious values, and the Sierra profitability. Several studies have shown that the
Club mutual funds invest in only those firms that better a firm’s labor relations, the better its finan-
they believe have acceptable environmental per- cial performance (Berman et al., 1999; Greening
formance. Whereas previous studies, in comparing and Turban, 2000; Jones and Murrell, 2001; Tur-
SRI to non-SRI funds, have largely ignored the ban and Greening, 1996; Waddock and Graves,
rich heterogeneity within SRI funds, we expect to 1997; Wright et al., 1995). Thus, we hypothesize
find significant financial performance differences the following:
not only across varying levels of screening inten-
sity, but also across the varying types of social Hypothesis 2: SRI funds that select firms for
screens that SRI funds use. their portfolios based on labor relations screen-
As Jones (1995: 430) noted: ‘Certain types ing criteria will earn higher financial returns
of corporate social performance are manifesta- than those that do not.
tions of attempts to establish trusting, coopera-
tive firm/stakeholder relationships and should be Firms must physically locate their operations
positively linked to a company’s financial perfor- within the boundaries of specific communities.
mance.’ However, engaging in the socially respon- Poor relations with these host communities can
sible behaviors that build these relationships is create a variety of costly problems. For example,
costly. For example, firms that dole out corporate poor community relations can increase the diffi-
philanthropy, provide day care centers and paid culty and cost of expansion and thereby limit a
parental leave, or engage in other such socially firm’s growth. A firm with poor community rela-
responsible behaviors incur significant expenses. tions may face ‘not-in-my-back-yard’ (NIMBY)
Faced with a large set of often-conflicting demands protests when attempting to open new plants. Such
Copyright  2006 John Wiley & Sons, Ltd. Strat. Mgmt. J., 27: 1101–1122 (2006)
DOI: 10.1002/smj
1108 M. L. Barnett and R. M. Salomon

protests decrease the ability of a firm to obtain it can be quite costly for a firm to substantially
essential building and zoning permits from local improve its environmental performance. Do the
governments and force the firm to bear significant benefits outweigh the costs? A growing body of
litigation expenses (Dear, 1992; Sellers, 1993). literature indicates that it does, indeed, ‘pay to be
On the other hand, favorable community relations green.’ Studies have shown that strong environ-
bring a number of benefits. Favorable commu- mental performance is associated with increased
nity relations can not only decrease the likelihood operational efficiency, improved learning and inno-
and intensity of NIMBY protests, but can also vation, decreased insurance costs, improved rela-
decrease the likelihood of attacks by shareholder tionships with stakeholders, differentiation of prod-
activists (Rehbein, Waddock, and Graves, 2004: ucts and services, and other such benefits that, indi-
239). Moreover, favorable community relations vidually and in combination, can more than offset
increase the likelihood of successful bargaining the costs of implementing environmental improve-
with local government officials for favorable taxa- ments (Hart and Ahuja, 1996; King and Lenox,
tion and regulation (Waddock and Graves, 1997). 2002; Klassen and McLaughlin, 1996; Klassen and
More generally, ‘Companies that treat local com- Whybark, 1999; Konar and Cohen, 2001; Russo
munities well reap many returns, including bet- and Fouts, 1997; Porter and van der Linde, 1995).
ter schools, fewer local restrictions, and a better We therefore hypothesize:
infrastructure to support the firm. In the long term,
these decrease corporate operating costs’ (Wad- Hypothesis 4: SRI funds that select firms for
dock and Smith, 2000: 79). This better infrastruc- their portfolios based on environmental screen-
ture includes access to more highly skilled employ- ing criteria will earn higher financial returns
ees (Greening and Turban, 2000). A variety of than those that do not.
empirical studies have indeed found that a firm’s
efforts to improve relations with their host com-
munities through activities such as philanthropy METHODOLOGY
and support of employee volunteering can lead
The data we employ come from several sources.
to improved financial performance (Hillman and
Our initial sample consisted of the socially respon-
Keim, 2001; Preston and O’Bannon, 1997; Simp-
sible mutual funds tracked by the Social Invest-
son and Kohers, 2002; Waddock and Graves 2000).
ment Forum. The Social Investment Forum is a
Therefore, we expect the following:
national non-profit organization that encourages
and promotes the growth of socially responsible
Hypothesis 3: SRI funds that select firms for
investing. Data from this source provide informa-
their portfolios based on community relations
tion about the social screening strategies (number
screening criteria will earn higher financial
and type of social screens used) of 67 socially
returns than those that do not.
responsible funds.5 After identifying our initial
sample of 67 socially responsible mutual funds, we
Finally, responsible environmental practices have used CRSP data to track each fund’s financial per-
become critical to a firm’s relationships with a formance. We compiled monthly financial perfor-
variety of stakeholder groups. Over the last few mance data from 1972 to 2000, which encompass
decades, stakeholders have increased their expec- the entire substantive existence of SRI funds. We
tations about the degree to which firms should supplemented this data with mutual fund informa-
assume responsibility for protection of the natu- tion from Weisenberger and ICDI. Weisenberger
ral environment (Hoffman, 1997, 1999). Today,
firms with poor environmental performance risk 5
Although screening data were only available from 1997 on, in
consumer disfavor, protest by activist groups, neg- the results presented herein, we use performance data from as far
ative media coverage, and general degradation of back as was available to get as complete a picture as possible.
their reputation (Fombrun et al., 2000; King and We acknowledge that the inherent assumption in using such data
is that the screening strategy of the mutual fund did not change
Lenox, 2000). Poor environmental practices can prior to 1997. To check the implications of this assumption, we
also place firms at increased risk for serious indus- performed two sensitivity analyses. First, we limited the sample
trial accidents that may result in large regulatory to those funds that were founded in 1997 or later. Second, we
eliminated the data for all funds prior to 1997 to check for
fines, costly lawsuits, and even the shutdown of potential biases. In both cases, the quantitative and qualitative
operations (Perrow, 1984; Rees, 1994). However, results were consistent with those shown.

Copyright  2006 John Wiley & Sons, Ltd. Strat. Mgmt. J., 27: 1101–1122 (2006)
DOI: 10.1002/smj
 Curvilinear Relationship between Social Responsibility and Financial Performance 1109

70

60

50
Number of Funds

40

30

20

10

0
71

73

75

77

79

81

83

85

87

89

91

93

95

97

99
19

19

19

19

19

19

19

19

19

19

19

19

19

19

19
Year

Figure 2. Population of SRI funds

and ICDI are mutual fund tracking services that Dependent variable
provide a standard directory of information on
We test for the effects of social screening on
mutual funds and their holdings such as mutual
financial performance. Thus, our dependent vari-
fund total assets and general investment strategy able is the risk-adjusted financial performance of a
(e.g., growth, income, capital appreciation). given SRI fund in a given month. Risk-adjusted
The resultant sample from these sources is an performance (RAP) is defined as the average
unbalanced panel of 61 funds and 4,821 fund- monthly return, measured as the percentage change
month observations. The total available sample in a fund’s market value from the beginning to the
could have reached 22,512 fund-month obser- end of a given month, adjusted by the fund’s spe-
vations (67 funds × 28 years × 12 months of cific beta (see Sharpe, 1964). For details on how
performance data). However, six funds failed to we used the CAPM model to calculate RAP, refer
report their social screening strategies to the Social to the Appendix.
Investment Forum. Most data were lost, though,
because SRI boomed only in recent years. Only
one fund, the PAX World Fund, existed at the Independent variables
beginning of the sample (1972). No other fund Previous research has taken a largely dichotomous
entered the sample until 1982. As illustrated in approach to categorizing SRI funds: either a fund
Figure 2, most SRI funds did not exist prior to screens for social responsibility, or it does not.
1996, with a 218 percent growth in the number of Most empirical work then compares the perfor-
funds since then. Fortunately, none of the funds mance of SRI funds to non-SRI funds. However,
in our sample exited during the event window, SRI funds are not homogeneous. Some have more
therein eliminating concern of a survival bias that stringent social screening standards than others.
has plagued other studies of this type (see Elton, Gains to highly diversified but weakly screened
Gruber and Blake, 1996b). SRI funds may have offset performance losses in
Copyright  2006 John Wiley & Sons, Ltd. Strat. Mgmt. J., 27: 1101–1122 (2006)
DOI: 10.1002/smj
1110 M. L. Barnett and R. M. Salomon

less intensively screened SRI funds. Then again, fund. As illustrated in Figure 1, the greater a
intensively screened SRI funds may have selected fund’s screening intensity, the smaller its universe
a stronger portfolio, thereby subsidizing the poor of potential investment targets. In contrast, lesser
choices of the weakly screened funds. Thus, lump- screening intensity implies a larger universe from
ing all SRI funds into a single category may have which a fund manager might select, and so a
contributed to the largely contradictory findings closer resemblance to a broadly diversified port-
of prior research. Moreover, such studies shed no folio. Thus, a large value for screening intensity
light on the argument of SRI opponents that ‘the indicates an increasing tendency toward a narrower
key to the recent success of many SR[I] funds SRI portfolio, while a small value for screening
could be that they own what conventional funds intensity reflects a more diversified SRI portfolio.7
own’ (Glassman, 1999: 4). To advance the debate, In order to test Hypotheses 2, 3, and 4, we must
we examine the performance implications of vary- tease out the variance in performance associated
ing the stringency of screening strategies used by with employment, community, and environmental
different SRI funds. screens. To do this, we created a dichotomous vari-
The Social Investment Forum lists 12 types of able for each of the screening strategies related to
screens that SRI funds may use to filter firms from workforce, community, and environmental issues.
their investment portfolios. Potential screening cri- For instance, in order to test Hypothesis 4, we
teria include excluding firms based upon their affil- assigned a value of 1 to the variable environment
iation with the following 12 industries or issues: if a fund screened out firms based on environ-
alcohol, tobacco, gambling, defense/weapons, ani- mental performance, zero otherwise. We similarly
mal testing, product/service quality, environment, defined two variables related to workforce issues
human rights, labor relations, employment equal- in order to test Hypothesis 2: labor relations and
ity, community investment, and community rela- equal employment. Finally, as a test of Hypothe-
tions. We refer to an SRI fund’s choice of number sis 3, if a fund screened out firms based on what
of screens to apply to its investment portfolio as its it deemed as a poor community record, then the
screening intensity. Screening intensity varies from variables community investment and community
1 to 12. If a fund’s screening intensity is given a services, respectively, received a value of 1, zero
value of 12, this indicates that the fund employs otherwise. Table 1 provides general definitions of
all 12 of the above-listed screens, whereas a value how each of the above screens is employed by the
of 1 indicates that the fund uses only 1 of the various socially responsible mutual funds.
12 available screens.6 Each SRI fund determines
how many and which of these screens it wishes to
Control variables
use. For example, the Ariel Socially Responsible
Fund uses only two screens (tobacco and nuclear Because our dependent variable captures the finan-
power) while the Calvert Social Equity Fund uses cial performance of a fund, we must control for
all twelve screens to select its investment portfolio.
The number of screens employed by the fund 7
The inherent assumption in using such an additive screening
proxies for the extent of diversification of the intensity variable is that any one screen is the same as another
inasmuch as it decreases the opportunity for diversification. We
acknowledge that this is a coarse proxy, as each screen may
6
In sensitivity analyses we dropped those funds that reported not contribute equally to the ability of a mutual fund manager
extreme screening intensity values (e.g., fewer than 3 and greater to diversify. We therefore checked the robustness of the results
than 10 screens) to test for outliers or influential points. Results by using different weightings of the screening intensity variable.
remained qualitatively unchanged. Results were largely consistent with those presented herein.

Table 1. Definitions of social screens

Type of social screen Firms affected by screen

1. Environment Excludes firms with a record of poor environmental performance

2. Labor relations Excludes firms with a record of poor labor relations practices
3. Employment/equality Excludes firms that violate norms of equal employment and diversity at work
4. Community investment Excludes firms that do not invest in and/or develop economically depressed communities
5. Community relations Excludes firms that have a poor record of accountability to local community stakeholders

Copyright  2006 John Wiley & Sons, Ltd. Strat. Mgmt. J., 27: 1101–1122 (2006)
DOI: 10.1002/smj
 Curvilinear Relationship between Social Responsibility and Financial Performance 1111

factors that could systematically affect financial Even after adjusting for a fund’s risk profile,
performance. We therefore include variables pre- there may still remain some difference in the risk-
viously identified as likely to influence the finan- adjusted performance between bond funds and
cial performance of mutual funds, while control- stock funds. Bonds may be a better investment
ling for unobservables using a combination of vehicle than stocks, or vice versa. We therefore
fixed and random effects. We discuss the details include two control measures to capture the effects
of the econometric specification in the subsequent of pursuing different general investment strategies
section. (see Elton, Gruber, and Blake, 1996a). ICDI gath-
A fund’s age is a potential factor in its finan- ers information on the percentage of total assets
cial performance. Older funds may have different each fund invests in stocks vs. bonds. We label
cost structures from new or young funds. More- these measures as percent stock and percent bonds,
over, the collective experience of a particular fund respectively.
accumulates over time, and that learning may be a Finally, we include yearly dummy variables to
valuable asset in choosing and managing its port- control for any residual macro-economic factors
folio (Argote, 1999). We control for any age effects that affect all funds similarly and to control for the
with the variable fund age, a count of the number potential for simultaneity bias and residual serial
of months since the fund’s inception. correlation of the error (Greene, 2000).
Larger funds may outperform smaller funds
because of economies of scale in fund man-
Statistical methods
agement. For instance, larger funds may spread
costs of information gathering, investor solicitation In selecting an appropriate multivariate statistical
and communication, and other fund management method, we begin with an ordinary least squares
expenses across a greater asset base. On the other (OLS) specification. As shown in Equation 1, we
hand, larger funds may face liabilities of size. As first specify a fund’s risk-adjusted performance
they seek ways in which to invest cash, they may (RAPit ) as a linear function of the vector X of
find it increasingly difficult to uncover bargains. independent variables for fund i at time t that we
Moreover, when larger funds make sizable mar- wish to examine and can measure, in addition to
ket purchases, the size of the trades may move an error term, which we label uit :
the market, making it difficult for these funds to
purchase undervalued stocks without the act of RAPit = Xit β + uit (1)
purchase raising the value of the stock. Therefore,
larger funds may be constrained to make smaller Given the panel data structure, with several obser-
purchases than desired, spread across a larger num- vations per fund, the possibility arises that uit in
ber of stocks than desired. In order to control for Equation 1 will not be independent across time
any potential size effect, we include a measure of (Greene, 2000). Thus, any systematic effect on
overall fund assets (measured in millions of U.S. risk-adjusted performance that is not included in
dollars). We label this variable total assets. X will be captured in the error term. Previous
Global economic cycles and the risks (or poten- research has identified many macro-economic fac-
tial rewards), over and above those inherent in tors associated with performance, including gov-
the underlying U.S. market, may affect financial ernment policy or systemic shocks that influence
performance. Thus, funds with holdings only in industries and sectors. Should we be unable to
the United States may perform differently from identify and measure all of these effects, there
funds with international holdings. In order to con- exists the potential for a systematic component to
trol for performance differentials across funds with be embedded in uit . This systematic component
national and international holdings, we include the will lead to correlated errors across observations,
dummy variable global fund. This variable takes which violates an assumption of OLS (Kmenta,
the value of 1 for funds with international hold- 1997; Kennedy, 1998).
ings, zero otherwise.8 Conceptually, we can decompose uit into a vec-
tor of systematic (fixed) effects, which we label
8
Zt , plus a truly random error component, which
In sensitivity analyses we eliminated those funds with inter-
national holdings. Results were largely consistent with those we label eit . In this case, Zt represents the yearly
presented herein. dummy variables. After we extract Zt from uit ,
Copyright  2006 John Wiley & Sons, Ltd. Strat. Mgmt. J., 27: 1101–1122 (2006)
DOI: 10.1002/smj
1112 M. L. Barnett and R. M. Salomon

we can more confidently assume that eit is i.i.d. In this case, Fi represents the individual mutual
normal with zero mean. Equation 2 represents this fund disturbance. The efficient estimator employed
decomposition of uit : is generalized least squares, and nested models can
be compared by the chi-square test.
RAPit = Xit β + Zt + eit (2) In summary, we include a fixed year effect and
a random fund effect to control for characteris-
tics not directly measured by our other variables
Finally, because there are several observations for but that might correlate with risk-adjusted mutual
each mutual fund, the possibility still exists that fund performance. The advantage of the fixed- and
eit in Equation 2 will not be independent within random-effects specifications are that they control
a common fund. This would occur, for instance, for unobserved heterogeneity without having to
if some funds performed, on a risk-adjusted basis, precisely specify the source of that heterogene-
differently from others over time owing to sys- ity. Therefore, they provide robust estimates that
tematically better fund management, or owing to eliminate bias in statistical results. The disadvan-
each fund’s idiosyncratic application of particu- tage, however, is that we cannot precisely isolate
lar screens. In theory, either a fund fixed- or or identify every individual factor that influences
random-effects model may be used to correct for the dependent variable. Because our goal is to con-
this (Greene, 2000). However, in our data, some trol for and not investigate or test these effects, we
funds exhibit very little variance in total assets and accept this trade-off.
screening strategies across time. Under this condi-
tion, and because we have few observations per
fund on average, a random-effects model is pre- RESULTS
ferred (Kennedy, 1998). We therefore arrive at our
final econometric specification as listed below in Table 2 presents descriptive statistics and product
Equation 3: moment correlations for the variables we used to
test the hypotheses. For the most part, the descrip-
RAPit = Xit β + Zt + Fi + vit (3) tive statistics and correlations are as expected. The

Table 2. Descriptive statistics and product moment correlations

1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12.

1. RAP 1
2. Screening −0.001 1
intensity
3. Fund age 0.019 0.066 1
4. Total assets 0.025 0.035 0.604 1
5. Percent 0.072 −0.039 0.000 0.113 1
stocks
6. Percent −0.107 −0.082 0.118 0.019 −0.630 1
bonds
7. Global fund −0.033 0.134 −0.127 −0.027 0.199 −0.153 1
8. Labor −0.019 0.588 −0.163 0.065 0.050 0.048 0.238 1
relations
9. Equal −0.019 0.609 0.151 0.074 −0.046 −0.157 0.167 0.201 1
employment
10. Community −0.018 0.442 0.195 −0.104 −0.211 0.092 0.054 0.039 0.409 1
investment
11. Community −0.002 0.735 −0.216 −0.075 0.030 −0.066 0.133 0.612 0.570 0.316 1
relations
12. Environment −0.044 0.405 0.045 −0.016 −0.016 0.070 0.124 0.260 0.235 −0.067 0.214 1
Mean 0.13 7.79 68.38 93.00 67.30 16.02 0.07 0.46 0.73 0.50 0.47 0.83
S.D. 0.03 3.02 58.62 173.20 37.51 29.11 0.26 0.49 0.44 0.50 0.50 0.37
Minimum −3.02 1 1 0.19 0 0 0 0 0 0 0 0
Maximum 3.08 12 354 1483.92 106.7 100 1 1 1 1 1 1

Copyright  2006 John Wiley & Sons, Ltd. Strat. Mgmt. J., 27: 1101–1122 (2006)
DOI: 10.1002/smj
 Curvilinear Relationship between Social Responsibility and Financial Performance 1113

average fund held about 67 percent of its portfolio Figure 3(b) provides modest support for this posi-
in equity and 16 percent in bonds, thus explaining tion, as it shows that average screening inten-
the relatively low fund beta.9 The greater the per- sity has declined over time, though most of the
centage of assets allocated to stocks, the greater the decline occurred in the early 1990s, with only
mutual fund financial performance. Likewise, the small declines from the late 1990s until the end of
more a mutual fund invested in bonds, the lower the study in 2000. Of course, these graphs do not
the risk-adjusted performance of the fund. control for intervening factors that have the poten-
Table 2 suggests that many of the screens tend to tial to influence the relationship between RAP and
be used in conjunction with others. For instance, screening intensity. We address these additional
mutual funds that screened on the basis of com- influences in the analyses below.
munity relations also tended to screen out firms In Table 3 we present results of regression mod-
with poor labor relations (ρ = 0.612). It is also els testing the first hypothesis. While we included
interesting to note the relative popularity of par- fixed year effects in the empirical specification, we
ticular types of social screens. Whereas many SRI did not report them here. Briefly though, the results
funds screened out firms based on their environ- suggested that while mutual fund performance was
mental performance (83%) and equal employment best in 1997 and 1998 in raw terms, the funds
records (73%), less than half of SRI funds con- in this sample assumed more risk during those
cerned themselves with labor relations (46%) or years, which raised their risk profile and actually
community relations (47%). This presents some decreased their risk-adjusted performance. Consis-
econometric challenges. The main concern is that
tent with the correlation tables, funds with a greater
such high correlations may infuse multicollinearity
percentage of equity investments achieved the best
into the regression. Given this concern, we per-
overall risk-adjusted performance, while funds that
formed various sensitivity analyses to ensure the
invested heavily in bonds were the worst over-
robustness of the results. First, we added each
all performers in our sample. However, we should
type of social screening strategy separately into
the regressions; the results were largely consistent point out that the economic impact of such findings
across specifications. Second, we entered the inde- is negligible. For example, the marginal effects in
pendent variables in different orders. The results Model 1 indicate that a 10 percent increase in
did not change substantively. Finally, we explored stock holdings amounts to about a 0.06 percent
the variance inflation contribution of each of the increase in risk-adjusted performance per month,
independent variables. All were well within the or about 0.72 percent per year. Not surprisingly,
acceptable range (Kennedy, 1998; Belsley, Kuh, global mutual funds performed consistently worse
and Welsch, 1980). As a result, interpretations of than funds with a purely domestic investment ori-
the findings do not change. entation. This result may be due to the additional
Figure 3 presents two graphs that display (a) risk involved with investing in foreign locations
average monthly RAP and (b) average screening (Hymer, 1976) over and above the risk imposed
intensity for all funds in the study over time. by the domestic market.
Figure 3(a) illustrates that there was a dip in per- In Model 1, we posit risk-adjusted performance
formance in the early 1990s, and then the finan- to be a linear function of screening intensity. In
cial performance of SRI funds tended to increase this specification we test whether including more
over time. As discussed previously, some have social screens is positively or negatively related to
argued that this increase in performance has come fund financial performance. A negative relation-
at the expense of a decrease in the stringency of ship would support those who contend that social
social screening (Glassman, 1999; Goetz, 1997). screening has a detrimental affect on financial per-
formance by limiting a fund’s ability to diversify,
9
The maximum percentage of equity reached 106.7 percent while a positive relationship would support the
in this sample. This suggests that a fund borrowed to invest stakeholder argument that well-screened socially
more than its total assets in stocks. Mutual funds are generally
restricted from leveraging assets by the Investment Company Act
responsible funds outperform more broadly diver-
of 1940; however, at times, they may borrow to invest more than sified funds because they are able to select from
their total assets. This is not a common practice. In our sample, a better subset of firms. Interestingly, we find no
only two of the funds were ever levered (invested greater than
100% of their assets) at any point in time. When we eliminated linear association between the number of screens
these levered funds from the analysis, the results did not change. and fund performance.
Copyright  2006 John Wiley & Sons, Ltd. Strat. Mgmt. J., 27: 1101–1122 (2006)
DOI: 10.1002/smj
1114 M. L. Barnett and R. M. Salomon

0.7

0.6

0.5
Monthy RAP

0.4

0.3

0.2

0.1

0
1992 1993 1994 1995 1996 1997 1998 1999 2000
(a) Year

8.6

8.4

8.2
Screening Intensity

7.8

7.6

7.4

7.2

6.8
1991 1992 1993 1994 1995 1996 1997 1998 1999 2000
(b) Year

Figure 3. Fund performance and screening intensity over time. (a) Average monthly RAP (all funds). (b) Average
screening intensity (all funds)

Model 2 adds a squared screening intensity maximum social screening intensity of 12 screens.
term. Because the models are nested, we can We note that even at the maximum of 12 screens,
directly compare the χ 2 statistic across the two however, performance does not recover to reach
to determine which dominates. The χ 2 (1) increase the levels achieved by those funds with 1 screen.
of 7.12 from Model 1 to Model 2 is significant In fact, the results suggest that we should expect
at the p < 0.01 level, suggesting that Model 2 funds with 12 screens to suffer performance decre-
better fits the data. Consistent with our expecta- ments of about 0.2 percent per month (about 2.4%
tions, we find a negative and significant coeffi- per year) vs. more broadly diversified funds. We
cient for screening intensity and a positive and therefore cannot conclude that screening comes
significant coefficient for its quadratic. This result without costs. Figure 4 depicts this relationship
implies a curvilinear, non-monotonic relationship graphically.
between screening intensity and fund performance, In Table 4 we explore the variance across the
thus supporting Hypothesis 1. Risk-adjusted per- screening strategies of interest. Not surprisingly,
formance declines at first as screening intensity we find that some screening strategies significantly
increases, reaching a minimum at 7 screens, but influence mutual fund performance. However,
then increases continuously until it reaches the some of the relationships are not as hypothesized.
Copyright  2006 John Wiley & Sons, Ltd. Strat. Mgmt. J., 27: 1101–1122 (2006)
DOI: 10.1002/smj
 Curvilinear Relationship between Social Responsibility and Financial Performance 1115
Table 3. Regression results for screening intensity Table 4. Regression results for screen types

Model 1 (RAP) Model 2 (RAP) Model 1 Model 2 Model 3

(RAP) (RAP) (RAP)
Constant 0.533∗∗ 1.090∗∗∗
(1.84) (2.56) Constant 0.605∗∗∗ 0.711∗∗∗ 0.950∗∗∗
Screening intensity −0.005 −0.202∗∗ (2.59) (2.98) (3.58)
(−0.22) (−1.78) Labor relations 0.110 −0.152 −0.099
Screening intensity2 0.014∗∗ (0.68) (−0.77) (−0.51)
(1.77) Equal −0.287∗ −0.571∗∗∗ −0.471∗∗
Fund age 0.001 0.001 employment (−1.61) (−2.50) (−2.07)
(0.87) (0.88) Community −0.091 −0.138
Total assets 0.000 0.000 investment (−0.54) (−0.84)
(0.33) (0.46) Community 0.550∗∗ 0.535∗∗
Percent stocks 0.006∗∗∗ 0.006∗∗∗ relations (2.20) (2.22)
(2.35) (2.53) Environment −0.381∗∗
Percent bonds −0.009∗∗∗ −0.009∗∗∗ (−1.91)
(−2.69) (−2.68) Fund age 0.002 0.002∗ 0.002∗
Global fund −0.609∗∗ −0.698∗∗∗ (1.01) (1.44) (1.35)
(−2.19) (−2.46) Total assets 0.000 0.000 0.000
Mutual fund effects Included Included (0.17) (0.15) (0.36)
No. of observations 4821 4821 Percent stocks 0.004∗∗ 0.004∗∗ 0.004∗∗
No. of mutual funds 61 61 (1.92) (1.72) (1.81)
χ 2 (d.f.) 94.23∗∗∗ 101.35∗∗∗ Percent bonds −0.011∗∗∗ −0.011∗∗∗ −0.010∗∗∗
(16) (17) (−3.39) (−3.42) (−3.24)
Global fund −0.570∗∗ −0.430∗ −0.396∗
∗ ∗∗ ∗∗∗
(−2.03) (−1.51) (−1.44)
p-value < 0.10; p-value < 0.05; p-value < 0.01 (one-
Mutual fund Included Included Included
tailed tests)
effects
No. of 4821 4821 4821
observations
Although there is no significant relationship No. of mutual 61 61 61
between labor relations screening strategies and funds
risk-adjusted performance, mutual funds that χ 2 (d.f.) 100.57∗∗∗ 110.63∗∗∗ 134.24∗∗∗
actively screened out firms based on their (16) (18) (19)
equal employment records suffered performance ∗ ∗∗ ∗∗∗
p-value < 0.10; p-value < 0.05; p-value < 0.01 (one-
decrements compared to the baseline fund. The tailed tests)
latter result, as shown in Model 1, runs counter

1
0.9
Risk-adjusted performance

0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
1 2 3 4 5 6 7 8 9 10 11 12
Screening Intensity

Figure 4. Non-monotonic effects of screening

Copyright  2006 John Wiley & Sons, Ltd. Strat. Mgmt. J., 27: 1101–1122 (2006)
DOI: 10.1002/smj
1116 M. L. Barnett and R. M. Salomon

to Hypothesis 2. Funds that excluded firms that slightly weaker in statistical significance given a
violated norms of equal employment actually reduction in sample size, but the marginal effects,
suffered a financial penalty of about 0.29 percent and so inferences, do not change.
per month. This result implies at the firm level Second, because we are using data both pooled
that the costs of implementing such programs across time, and within funds across time, there
may outweigh their benefits, and/or that the is the potential for residual serial correlation of
diversification penalty incurred by mutual funds the error. In order to assess whether our results
that use such screens exceed financial gains. were biased by serial correlation, we ran two
Model 2 adds our proxies for community screen- specific tests. We pooled the data by year and
ing strategies. Although there was no systematic tested whether AR(1) processes were at work after
relationship between the screening of those firms controlling for year effects. The Durbin–Watson
that invested in their local community (community statistic did not suggest that serial correlation of the
investment) and performance, mutual funds that AR(1) type was present in any of the models. We
included firms that fostered positive relationships then arranged the data by fund-year and assessed
with their local community (community relations) the within-fund correlations of the residuals over
performed better. Therefore, Hypothesis 3 receives time. We found a within-fund average correlation
some support. This implies, consistent with instru- of around 0.007 across all models; further, none of
mental stakeholder theory, that firms that foster the individual fund correlations was greater than
positive relationships within their communities are 0.30 or less than −0.30. We therefore conclude
financially rewarded. that, after controlling for fund and year effects,
Finally, contrary to Hypothesis 4, screening on the results are not biased by serial correlation.
the basis of environmental criteria was negatively Third, the inherent assumption underlying our
related with risk-adjusted financial performance. screening intensity measure is that any one screen
All else equal, SRI funds that culled environmen- is the same as another inasmuch as it decreases
tally poor performers out of their holdings per- the opportunity for diversification and increases
formed about 0.38 percent worse per month than the opportunity for selectivity. However, some of
the baseline SRI fund. The costs of implementing the screens represent ‘positive’ screens in that they
environmentally sound policies appear to outweigh select certain firms into the portfolio (e.g., environ-
their potential benefits, and they do not compen- ment, labor relations, community relations) while
sate mutual funds for the loss in diversification some screens are negative screens in that they
that they impose. These findings stand in contrast eliminate entire industries and sectors from a port-
to a growing body of literature that suggests that folio. To the extent that negative screens impose
firms that perform better on environmental crite- more diversification costs upon a fund and positive
ria will experience better overall financial perfor- screens disproportionately increase a fund’s selec-
mance (Dowell, Hart, and Yeung, 2000; Porter and tivity, we might plausibly expect a negative perfor-
van der Linde, 1995; Russo and Fouts, 1997). mance effect for ‘negative’ screens and a positive
performance effect for ‘positive’ screens. In order
to test for this possibility, we grouped screens into
Sensitivity analyses
negative or positive and reran the results presented
In order to assess the robustness and sensitivity of in Model 1 of Table 3 using counts of positive
the results, we tested several variants of the models and negative screens instead of screening intensity.
presented herein.10 First, because equity returns Both variables were economically and statistically
are sensitive to the time frame of study, the time insignificant, suggesting that positive and negative
window used may impact the findings. Although screens contribute to both decreased diversification
we include time dummies to control for bias that and increased selectivity.
may be induced from pooling data across time, Fourth, we sought to verify that the results
we reran the results using various time windows. presented in Table 4 are not just an artifact of the
Regardless of the time frame selected, the results screening intensity data. If the results are driven
did not change. In some instances the results were by low screening intensity funds employing all of
the screens that appear positive and significant,
10
We thank the anonymous reviewers for motivating and sug- mid-range intensity funds employing all of the
gesting several of the sensitivity analyses. screens with negative and significant coefficients,
Copyright  2006 John Wiley & Sons, Ltd. Strat. Mgmt. J., 27: 1101–1122 (2006)
DOI: 10.1002/smj
 Curvilinear Relationship between Social Responsibility and Financial Performance 1117

and high screening intensity funds employing all described. Following prior literature, therefore, we
of those with neutral coefficients, we could be do not include a bond term to help quantify RAP in
mistakenly attributing our results to differences our CAPM equation. Instead, we incorporate per-
in screen usage though the results are actually centage bonds as a control variable in our regres-
driven by variation in the degree of diversification. sion models. To determine whether the absence of
In order to assess this possibility, we split the a bond term introduces any bias, we reran our anal-
sample into low (1–4 screens in use), medium yses excluding the 11 funds in our sample that
(5–8 screens in use), and high (9–12 screens hold bonds at any given point in time. This left a
in use) screening intensity subsets to conduct a set of 50 pure equity funds and 3026 fund-month
more nuanced analysis of the screen usage patterns observations. The results did not change.11
of the funds. We found no patterns to suggest
that low intensity funds were disproportionately
using community relations screens or that medium DISCUSSION
intensity funds were disproportionately favoring
equal opportunity and environmental screens. We If we assume that investors are rewarded only for
therefore conclude that we are capturing variance bearing systematic risk and all firms are homoge-
that is unique to the individual screen in question. nous in all but their risk premium, as modern
Fifth, there is some disagreement among finan- portfolio theory contends, then SRI can only harm
cial economists as to which asset-pricing model is financial performance. However, as stakeholder
most accurate (for a review see Fama and French, theory argues, some firms may be consistently bet-
1992, 1996; Kothari, Shanken, and Sloan, 1995). ter financial performers than others because of their
Some support the CAPM advanced by Sharpe socially oriented characteristics. The fundamental
(1964) and Lintner (1965), while others advo- market logic of SRI is that social screening can
cate more elaborate models such as the 3-factor help in selecting these firms. Our findings support
model proposed by Fama and French (1993) or both portfolio and stakeholder theories to varying
the 4-factor model proposed by Carhart (1997). degrees. It appears that even though social screen-
Although we use CAPM to calculate our mea- ing forces a narrowing of investment choices,
sure of risk-adjusted performance (RAP) in this if adequately implemented, social screening can
study, those who believe in a Fama and French 3- lead to an increase in financial returns. That is,
factor or Carhart 4-factor world might suggest that the financial performance of those limited firms
using CAPM will bias our results because betas chosen through intensive social screening offsets
alone do not adequately explain average return. costs from loss of portfolio diversification to some
We therefore reran our results using RAP calcu- degree.
lated from the Fama and French (1993) 3-factor These trade-offs help explain how two long
model and then the Carhart (1997) 4-factor model. competing viewpoints in the SRI literature may
In the case of the Fama and French (1993) 3-factor actually be complementary. Funds that employ
model, results were similar in statistical signif- many social screens may effectively eliminate
icance but slightly weaker in magnitude. Using underperforming firms from their portfolio in order
the Carhart (1997) 4-factor model, results were to improve financial performance. On the other
both statistically and economically stronger than hand, SRI funds that employ few social screens
those reported in this paper. Because CAPM is improve financial performance through benefits
the standard asset-pricing model applied in strat- received from increased diversification.12 Those
egy research, we report CAPM results herein.
Finally, asset-pricing models generally do not 11
All results discussed in this section are available from the
incorporate a bond term in addition to other mar- authors upon request.
ket factors to explain returns. Fama and French 12
It could be argued that since mutual funds are required to per-
(1993) found that using a bond term explains little form active analysis of social performance, one would expect
operating costs (and the funds’ associated expense ratios) to be
variation in the return of funds comprised mostly higher as screening intensity increases. Increased costs reduce
of stocks. Because bond holdings comprise only 16 performance, and would propose an alternative interpretation of
percent of the total invested assets of the funds in the negative portion of our curvilinear effect, potentially calling
our diversification inferences into question. However, returns
our sample, the funds are likely representative of compiled by CRSP and used in our analyses do not factor in
the type of portfolios that Fama and French (1993) expense ratios. This allows us to specifically isolate the effects

Copyright  2006 John Wiley & Sons, Ltd. Strat. Mgmt. J., 27: 1101–1122 (2006)
DOI: 10.1002/smj
1118 M. L. Barnett and R. M. Salomon

funds that are ‘stuck in the middle’ may not be engage in reactive environmental initiatives, the
able to either effectively diversify away unsys- findings could thus be skewed. Ultimately, finer-
tematic risk, or eliminate enough underperforming grained measures may be needed to adequately
firms from their portfolios to improve financial tease out such variance in performance.
performance. Second, in our study, we used market measures
The results of our analyses of different types of financial performance. Orlitzky et al. (2003:
of social screens prove more vexing. We found 403) concluded that ‘CSP [Corporate Social Per-
support for Hypothesis 3 that, indeed, funds that formance] appears to be more highly correlated
screened on the basis of community relations had with accounting-based measures of CFP [Cor-
relatively stronger financial performance. Thus, porate Financial Performance] than with market-
our findings concur with those of others who have based indicators . . ..’ A fruitful area of future
shown that the costs incurred by a firm to improve research on the nature of the relationship between
its relations with its local communities are more social and financial performance therefore might
than offset by financial gains (Hillman and Keim, be to relate how differences in research measures
2001; Simpson and Kohers, 2002; Waddock and and methods might contribute to different empir-
Graves, 2000). However, our findings ran counter ical findings. Prior studies have used a variety
to Hypotheses 2 and 4.13 These results suggest of measures, to include both market-based indica-
that the financial costs of increasing equal employ- tors (e.g., Klassen and McLaughlin, 1996; Shane
ment opportunity and diversity, as well as envi- and Spicer, 1983; Belkaoui, 1976) and accounting-
ronmental performance to levels adequate to pass based measures (e.g., King and Lenox, 2002;
the screening standards of SRI funds (above and Russo and Fouts, 1997; Hart and Ahuja, 1996).
beyond what is mandated by law) may outweigh A recent study by DellaVigna and Pollet (2005)
their financial benefits. Considering recent find- suggests that investors rarely attend to information
ings suggesting that it ‘pays to be green’ (e.g., that extends beyond a 5-year horizon and there-
Klassen and Whybark, 1999; Konar and Cohen, fore do not accurately price for the long-term. For
2001; Russo and Fouts, 1997; Porter and van der SRI funds, this implies that the market might not
Linde, 1995), our result concerning environmental accurately value social initiatives with extended
screening seems particularly difficult to reconcile. payback periods.
There are several possible reasons why we found Finally, as with any study of financial perfor-
that it did not pay to screen out firms with bet- mance, our analysis is retrospective and so must
ter labor relations or to ‘screen for green.’ First, be interpreted with caution. By necessity, we ana-
our measures of CSP are admittedly coarse. For lyzed the appropriateness of screening strategies
example, the environmental performance measure based upon the historical performance of mutual
cannot distinguish between proactive and reactive funds. Some of the financial benefits of certain
environmental initiatives. King and Lenox (2002) social screens may not be discernible until far-
found that waste prevention, not ‘end of pipe’ ther in the future. For example, though we found
cleanup drives financial gains. If the SRI funds in that screening out firms with poor environmen-
this sample disproportionately include firms that tal records actually harmed an SRI fund’s perfor-
mance, a firm’s up-front investment in environ-
of diversification so as to ensure that they are not confounded mental improvements may pay off in future years.
with the effects of screening costs on mutual fund perfor- Since the majority of the funds in our study were
mance—thereby reducing the plausibility of such an alternative.
Moreover, in sensitivity analyses conducted using expense ratios, less than 5 years old, our study may have missed
we did not find any relationship between screening intensity and these gains. The market’s preference for certain
expense ratios, and our results did not change.
13
types of social screens may also change over time.
It is important to note that the negative results for the labor
relations and environmental screens do not invalidate the stake- For example, the market may at first be slow to
holder theory arguments used to justify the positive portion of reward firms for particular social actions, then pro-
the curvilinear relationship suggested (and found) in Hypothe- vide increasing financial incentives as the issue
sis 1. Rather, these negative results only imply that the loss of
diversification associated with adopting either of these two par- becomes salient and popular, only to later simply
ticular screens in isolation outweighed any gains. In contrast, expect firms to take such actions without financial
the financial gains from superior stakeholder relations possessed reward. For all of these reasons, we are hesitant
by firms with positive community relations added more to the
RAP of SRI funds than the associated decrease in diversification to draw strong conclusions from our findings on
necessary to select such firms took away from their RAP. labor and environmental screening.
Copyright  2006 John Wiley & Sons, Ltd. Strat. Mgmt. J., 27: 1101–1122 (2006)
DOI: 10.1002/smj
 Curvilinear Relationship between Social Responsibility and Financial Performance 1119

Implications Belkaoui A. 1976. The impact of the disclosure of the

environmental effects of organizational behavior on
The aforementioned caveats notwithstanding, our the market. Financial Management 5: 26–31.
study has several important implications for both Belsley DA, Kuh E, Welsch RE. 1980. Regression
practitioners and scholars. First, it implies that Diagnostics: Identifying Influential Data and Sources
of Collinearity. Wiley: New York.
fund managers need to more carefully consider Berman SL, Wicks AC, Kotha S, Jones TM. 1999.
the effects their chosen screening strategies are Does stakeholder orientation matter? The relationship
likely to have on the performance of their funds. between stakeholder management models and firm
The choice is not as simple as either being an financial performance. Academy of Management
SRI fund or not, but rather, just how socially Journal 42(5): 488–506.
Black F, Jensen M, Scholes M. 1972. The capital asset
responsible to be. The prescription is that managers
pricing model: Some empirical tests. In Studies in the
should either wholeheartedly commit to broadly Theory of Capital Markets, Jensen M (ed). Praeger:
screening socially irresponsible firms from their New York; 79–121.
funds, or exclude very few firms such that they Bloomfield T, Leftwich R, Long J Jr. 1977. Portfolio
do not interfere with their ability to diversify. strategies and performance. Journal of Financial
Additionally, SRI fund managers must be aware Economics 5: 201–218.
Bowman E, Haire M. 1975. A strategic posture towards
of which types of social screens are rewarded at CSR. California Management Review 18(2): 49–58.
any particular point in time. Campbell J, Lettau M, Malkiel B, Xu Y. 2001. Have
For scholars, this result suggests that both camps individual stocks become more volatile? An empirical
can peacefully coexist. Those who base their argu- exploration of idiosyncratic risk. Journal of Finance
ments on the financial logic of modern portfolio 56(1): 1–43.
Carhart M. 1997. On persistence in mutual fund
theory and those who support instrumental stake-
performance. Journal of Finance 52: 57–82.
holder theory may not be at odds. Funds that use Cochran PL, Wood RA. 1984. Corporate social responsi-
few screens gain the benefits of diversification, bility and financial performance. Academy of Manage-
and those that filter stocks and limit their universe ment Journal 27(1): 42–56.
of investments do not handicap their portfolio as Cornell B, Shapiro AC. 1987. Corporate stakeholders
much as some contend. The real danger lies in not and corporate finance. Financial Management 16(1):
5–14.
committing to one strategy or the other—in being Dear M. 1992. Understanding and overcoming the
‘stuck in the middle.’ Failure to control for screen- NIMBY syndrome. Journal of the American Planning
ing intensity amongst SRI funds may explain, in Association 58(3): 288–300.
part, the mixed findings that have allowed this DellaVigna S, Pollet JM. 2005. Attention, demographics,
debate to rage on for so long. and the stock market. NBER working paper #11211.
DiBartolomeo D, Kurtz L. 1999. Managing risk expo-
In conclusion, we have uncovered findings that
sures of socially screened portfolios. Northfield Infor-
appear to bring together long-divided factions in mation Services working paper.
the broad literature on the link between financial Dierickx I, Cool K. 1991. Asset stock accumulation and
and social performance. Given our findings, fur- sustainability of competitive advantage. Management
ther research that moves beyond the dichotomy Science 35(12): 1504–1511.
inherent in prior studies seems well warranted. We Diltz JD. 1995. The private cost of socially responsible
investing. Applied Financial Economics 5: 69–77.
encourage others to improve upon our contribu- Donaldson T, Preston LE. 1995. The stakeholder theory
tion by looking at improved data sets with more of the corporation: concepts, evidence, and implica-
fine-grained measures. tions. Academy of Management Review 20(1): 65–91.
Dowell G, Hart S, Yeung B. 2000. Do corporate global
environmental standards create or destroy market
value? Management Science 46: 1059–1074.
REFERENCES Elton E, Gruber MJ, Blake CR. 1996a. The persistence
of risk-adjusted mutual fund performance. Journal of
Alexander G, Bucholtz R. 1978. Corporate social respon- Business 69(2): 133–157.
sibility and stock market performance. Academy of Elton E, Gruber MJ, Blake CR. 1996b. Survivorship bias
Management Journal 21: 479–486. and mutual fund performance. Review of Financial
Argote L. 1999. Organizational Learning: Creating, Studies 9(4): 1097–1120.
Retaining and Transferring Knowledge. Kluwer: Fama EF. 1971. Risk, return, and equilibrium. Journal of
Boston, MA. Political Economy 79(1): 30–55.
Barney J. 1991. Firm resource and sustained competitive Fama EF. 1991. Efficient markets: II. Journal of Finance
advantage. Journal of Management 17: 99–120. 46: 1575–1617.
Copyright  2006 John Wiley & Sons, Ltd. Strat. Mgmt. J., 27: 1101–1122 (2006)
DOI: 10.1002/smj
1120 M. L. Barnett and R. M. Salomon
Fama EF, French KR. 1992. The cross-section of bottom line? Strategic Management Journal 22(2):
expected stock returns. Journal of Finance 47: 125–139.
427–465. Hoffman A. 1997. From Heresy to Dogma: An Institu-
Fama EF, French KR. 1993. Common risk factors in the tional History of Corporate Environmentalism. New
returns on stocks and bonds. Journal of Financial Lexington Press: San Francisco, CA.
Economics 33: 3–56. Hoffman A. 1999. Institutional evolution and change:
Fama EF, French KR. 1996. The CAPM is wanted dead environmentalism and the U.S. chemical industry.
or alive. Journal of Finance 51: 1947–1958. Academy of Management Journal 42: 351–371.
Fisher L, Lorie JH. 1970. Some studies of the variability Hutton RB, D’Antonio L, Johnsen T. 1998. Socially
of returns on investments in common stocks. Journal responsible investing: growing issues and new
of Business 43(2): 99–134. opportunities. Business and Society 37(3): 281–305.
Florida R. 2002. The Rise of the Creative Class. Basic Hymer SH. 1976. The International Operations of
Books: New York. National Firms: A Study of Direct Foreign Investment.
Fombrun CJ. 1996. Reputation: Realizing Value from the MIT Press: Cambridge, MA.
Corporate Image. Harvard Business School Press: Jensen M. 2002. Value maximization, stakeholder theory,
Boston, MA. and the corporate objective function. Business Ethics
Fombrun CJ, Gardberg NA, Barnett ML. 2000. Opportu- Quarterly 12: 235–256.
nity platforms and safety nets: corporate citizenship Jones R, Murrell A. 2001. Signaling positive corporate
and reputational risk. Business and Society Review social performance: an event study of family-friendly
105(1): 85–106. firms. Business and Society 40(1): 59–78.
Freeman RE. 1984. Strategic Management: A Stakeholder Jones T. 1995. Instrumental stakeholder theory: a
Approach. Pitman: Boston, MA. synthesis of ethics and economics. Academy of
Friedman M. 1970. The social responsibility of business Management Review 20(2): 404–437.
is to increase its profits. New York Times Magazine 13 Kennedy P. 1998. A Guide to Econometrics. MIT Press:
September: 122–126. Cambridge, MA.
Geczy C, Stambaugh R, Levin D. 2003. Investing in King A, Lenox M. 2000. Industry self-regulation without
socially responsible mutual funds. Wharton School of sanctions: the chemical industry’s Responsible Care
Business working paper. program. Academy of Management Journal 43(4):
Glassman JK. 1999. Letting your conscience be your 698–716.
investment guide. Houston Chronicle 8 February: 4. King A, Lenox M. 2002. Exploring the locus of profitable
Goetz T. 1997. Dealing with the devil. Village Voice 19 pollution reduction. Management Science 48(2):
August: 43–44. 289–299.
Granovetter M. 1985. Economic action and social Klassen R, McLaughlin C. 1996. The impact of
structure: the problem of embeddedness. American environmental management on firm performance.
Journal of Sociology 91(3): 481–510. Management Science 42(8): 1199–1214.
Graves S, Waddock S. 2000. Beyond built to last . . . Klassen R, Whybark D. 1999. The impact of environ-
stakeholder relations in ‘built-to-last’ companies. mental technologies on manufacturing performance.
Business and Society Review 105(4): 393–418. Academy of Management Journal 42(6): 599–615.
Greene WH. 2000. Econometric Analysis. Macmillan: Kmenta J. 1997. Elements of Econometrics. Macmillan:
New York. New York.
Greening DW, Turban DB. 2000. Corporate social Konar S, Cohen M. 2001. Does the market value
performance as a competitive advantage in attracting environmental performance? Review of Economics and
a quality workforce. Business and Society 39(3): Statistics 83(2): 281–289.
254–280. Kothari SP, Shanken J, Sloan RG. 1995. Another look at
Griffin JJ, Mahon JF. 1997. The corporate social the cross-section of expected stock returns. Journal of
performance and corporate financial performance Finance 50: 185–224.
debate: twenty-five years of incomparable research. Kurtz L. 1997. No effect, or no net effect? Studies on
Business and Society 36(1): 5–31. socially responsible investing. Journal of Investing
Guerard JB Jr. 1997. Is there a cost to being socially 6(4): 37–49.
responsible in investing? Journal of Investing 6(2): Kurtz L, DiBartolomeo D. 1996. Socially screened port-
11–18. folios: an attribution analysis of relative performance.
Hamilton S, Jo H, Statman M. 1993. Doing well while Journal of Investing 5(3): 35–41.
doing good? The investment performance of socially LaRose JA. 1998. More than profits: selected resources
responsible mutual funds. Financial Analysts Journal , in socially responsible investing. Reference and User
November–December: 62–66. Services Quarterly 38(1): 33–40.
Hart S, Ahuja G. 1996. Does it pay to be green? An Lintner J. 1965. The valuation of risk assets and the
empirical examination of the relationship between selection of risky investments in stock portfolios and
emissions reduction and firm performance. Business capital budgets. Review of Economics and Statistics
Strategy and the Environment 5: 30–37. 47: 13–37.
Hillman A, Keim G. 2001. Shareholder value, stake- Lowry R. 1991. Good Money: A Guide to Profitable
holder management, and social issues: what’s the Investing in the ‘90’s. W. W. Norton: New York.
Copyright  2006 John Wiley & Sons, Ltd. Strat. Mgmt. J., 27: 1101–1122 (2006)
DOI: 10.1002/smj
 Curvilinear Relationship between Social Responsibility and Financial Performance 1121
Luck C, Pilotte N. 1993. Domini Social Index perfor- Scott R. 1981. Organizations: Rational, Natural, and
mance. Journal of Investing 2: 60–62. Open Systems. Prentice-Hall: Englewood Cliffs, NJ.
Mahon JF, Griffin JJ. 1999. Painting a portrait. Business Sellers M. 1993. NIMBY: a case study in conflict politics.
and Society 38(1): 126–133. Public Administration Quarterly 16(4): 460–477.
Margolis J, Walsh J. 2003. Misery loves company: Shane P, Spicer B. 1983. Market responses to envi-
rethinking social initiatives by business. Administra- ronmental information produced outside the firm.
tive Science Quarterly 48: 268–305. Accounting Review 58: 521–538.
Markowitz H. 1952. Portfolio selection. Journal of Sharpe WF. 1964. Capital asset prices: a theory of
Finance 7(1): 77–91. market equilibrium under conditions of risk. Journal
McWilliams A, Siegel D. 1997. The role of money of Finance 19: 425–442.
managers in assessing corporate social responsibility Simpson WG, Kohers T. 2002. The link between
research. Journal of Investing 6(4): 98–107. corporate social and financial performance: evidence
McWilliams A, Siegel D. 2000. Corporate social respon- from the banking industry. Journal of Business Ethics
sibility and financial performance: correlation or mis- 35: 97–109.
specification? Strategic Management Journal 21(5): Social Investment Forum. 2002. http://www.socialinvest.
603–609. org/Areas/SRIGuide/Screening.htm [15 January 2002].
Moskowitz M. 1972. Choosing socially responsible Statman M. 1987. How many stocks make a diversified
stocks. Business and Society Review 10: 71–75. portfolio? Journal of Financial and Quantitative
Mueller S. 1991. The opportunity cost of discipleship: Analysis 22(3): 353–363.
ethical mutual funds and their returns. Sociological Statman M. 2000. Socially responsible mutual funds.
Analysis 52: 111–124. Financial Analysts Journal May/June: 30–39.
Orlitzky M, Schmidt F, Rynes S. 2003. Corporate Teper J. 1992. Evaluating the cost of socially responsible
social and financial performance: a meta-analysis. investing. In The Social Investment Almanac (1st edn),
Organization Studies 24: 403–441. Kinder P (ed.). Henry Holt: New York; 340–349.
Perrow C. 1984. Normal Accidents: Living with High- Turban DB, Greening DW. 1996. Corporate social
Risk Technologies. Basic Books: New York. performance and organizational attractiveness to
Porter M. 1991. Towards a dynamic theory of strategy. prospective employees. Academy of Management
Strategic Management Journal , Winter Special Issue Journal 40(3): 658–672.
12: 95–117. Ullmann A. 1985. Data in search of a theory: a
Porter M, van der Linde C. 1995. Green and competitive. critical examination of the relationship among
Harvard Business Review 73(5): 121–134. social performance, social disclosure, and economic
Preston LE, O’Bannon D. 1997. The corporate performance. Academy of Management Review 10:
social–financial performance relationship. Business 450–477.
and Society 36(4): 419–429. Waddock SA, Graves SB. 1997. The corporate social
Preston LE, Post JE. 1975. Private Management and performance–financial performance link. Strategic
Public Policy: The Principle of Public Responsibility. Management Journal 18(4): 303–319.
Prentice-Hall: Englewood Cliffs, NJ. Waddock SA, Graves SB. 2000. Performance character-
Rees J. 1994. Hostages of Each Other: The Transforma- istics of social and traditional investments. Journal of
tion of Nuclear Safety Since Three Mile Island . Uni- Investing 9(2): 27–38.
versity of Chicago Press: Chicago, IL. Waddock SA, Smith N. 2000. Corporate responsibility
Rehbein K, Waddock S, Graves S. 2004. Understanding audits: doing well by doing good. Sloan Management
shareholder activism: which corporations are targeted? Review 41(2): 75–83.
Business and Society 43(3): 239–267. Wernerfelt B. 1984. A resource-based view of the firm.
Roman RM, Hayibor S, Agle BR. 1999. The relationship Strategic Management Journal 5(2): 171–180.
between financial and social performance: repainting Wicks AC, Berman SL, Jones TM. 1999. The
a portrait. Business and Society 38(1): 109–125. structure of optimal trust: moral and strategic
Rothchild J. 1996. Why I invest with sinners. Fortune 13 implications. Academy of Management Review 24(1):
May: 197. 99–116.
Rowley T, Berman S. 2000. A brand new brand of Williamson OE. 1975. Markets and Hierarchies: Analysis
corporate social performance. Business and Society and Antitrust Implications. Free Press: New York.
39(4): 397–418. Williamson OE. 1985. The Economic Institutions of
Rudd A. 1979. Divestment of South African equities: how Capitalism. Free Press: New York.
risky? Journal of Portfolio Management 5(3): 5. Wood D, Jones R. 1995. Stakeholder mismatching:
Russo M, Fouts P. 1997. A resource-based perspective on a theoretical problem in empirical research on
corporate environmental performance and profitability. corporate social performance. International Journal of
Academy of Management Journal 40: 534–559. Organizational Analysis 3: 229–267.
Sauer D. 1997. The impact of social-responsibility Wright P, Ferris S, Hiller J, Kroll M. 1995. Competi-
screens on investment performance: evidence from the tiveness through management of diversity: effects on
Domini 400 Social Index and Domini Equity mutual stock price valuation. Academy of Management Jour-
funds. Review of Financial Economics 6(2): 137–149. nal 38(1): 272–287.

Copyright  2006 John Wiley & Sons, Ltd. Strat. Mgmt. J., 27: 1101–1122 (2006)
DOI: 10.1002/smj
1122 M. L. Barnett and R. M. Salomon

APPENDIX: CALCULATION OF on the market portfolio for month t, Rf t reflects

RISK-ADJUSTED PERFORMANCE the risk-free rate of return (e.g., the 30-day T-bill
(RAP) rate), Bi captures the fixed beta of fund i, and eit
represents random error.
A fund’s expected return is a linear function of The risk-adjusted return of the fund is the differ-
the market return. A market index represents the ence between the risk premium of that fund and the
market return (in our case, the S&P 500), and the fund’s expected return, given its beta and the mar-
coefficient beta represents the linear relationship of ket’s risk premium. The risk-adjusted performance
the fund’s return to the market return. As is con- (RAP) then for fund i in month t, RAPit , is:
ventional in CAPM (see Sharpe, 1964), this study
computes beta on the basis of monthly returns. RAPit = (Rit − Rf t ) − Bi ∗ (Rmt − Rf t ) (5)
Specifically:
In essence, RAPit , from Equation 5, captures the
Rit − Rf t = ai + Bi ∗ (Rmt − Rf t ) + eit (4) fund’s return over and above what is expected
based upon its beta.14
where Rit represents the return on fund i in month
t, ai is Jensen’s alpha, Rmt represents the return

14
When we model RAP as the dependent variable, we are
parameterizing the error term and Jensen’s alpha to uncover
residual systematic noise in a fund’s performance. We found
similar results when we parameterized the standard CAPM
equation.

Copyright  2006 John Wiley & Sons, Ltd. Strat. Mgmt. J., 27: 1101–1122 (2006)
DOI: 10.1002/smj