Does Corporate Social Responsibility (CSR) Improve Credit Ratings?

Evidence from
Geographic Identification
Author(s): Pornsit Jiraporn, Napatsorn Jiraporn, Adisak Boeprasert and Kiyoung Chang
Source: Financial Management , FALL 2014, Vol. 43, No. 3 (FALL 2014), pp. 505-531
Published by: Wiley on behalf of the Financial Management Association International

Stable URL: https://www.jstor.org/stable/43280525

JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide
range of content in a trusted digital archive. We use information technology and tools to increase productivity and
facilitate new forms of scholarship. For more information about JSTOR, please contact [email protected].

Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at
https://about.jstor.org/terms

Wiley and Financial Management Association International are collaborating with JSTOR to
digitize, preserve and extend access to Financial Management

This content downloaded from

14.139.235.161 on Mon, 21 Nov 2022 09:28:56 UTC
All use subject to https://about.jstor.org/terms
 Does Corporate Social Responsibility
(CSR) Improve Credit Ratings?
Evidence from Geographic Identification
Pornsit Jiraporn, Napatsorn Jiraporn, Adisak Boeprasert,
and Kiyoung Chang*

We show that a firm 's CSR policy is significantly influenced by the CSR policies of firms in the
same three-digit zip code, an effect possibly due to investor clienteles, local competition, and/or
social interactions. We then exploit the variation in CSR across the zip codes to estimate the effect
of CSR on credit ratings under the assumption that zip code assignments are exogenous. We find
that more socially responsible firms enjoy more favorable credit ratings. In particular, an increase
in CSR by one standard deviation improves the firm 's credit rating by as much as 4.5%.

We find that a firm's corporate social responsibility (CSR) policy is significantly influenced
by the CSR policies of firms in the same three-digit zip code, an effect possibly due to investor
clientele, local competition, and/or social interactions. We then exploit the variation in CSR across
the zip codes to estimate the effect of CSR on credit ratings under the assumption that zip code
assignments are exogenous. We find that more socially responsible firms enjoy more favorable
credit ratings. In particular, an increase in CSR by one standard deviation improves the firm's
credit rating by as much as 4.5%.
Although a tremendous volume of research has been conducted regarding the issue of corporate
social responsibility (CSR), its effects on corporations are still not completely understood. One
difficulty is identifying exogenous shocks to CSR policies. Due to market segmentation, investor
clientele, social interactions, and/or local competition, we hypothesize that firms located in close
proximity to one another will have similar CSR policies. Exploiting the variation in CSR policies
across geographic locations, we then estimate the effect of CSR on credit ratings since ratings
significantly influence the cost of debt for the firm. Firms with favorable credit ratings have better
access to capital markets and can borrow at a much lower cost. Additionally, strong credit ratings
also inspire investor confidence.
Our proximity measure is based on zip codes. The US Postal Service (USPS) allocates zip
codes based on efficiency in mail delivery. Zip code changes are also rare and usually reflect
changes in macroeconomic factors such as demographics and urban development. Thus, zip
code assignments are unlikely related to corporate policies or outcomes and can be considered
exogenous.
The idea of corporate isomorphism and peer pressure is hardly new. Corporations that invest
in CSR have strong incentives to publicize their CSR activities and make them as visible as they

* Pornsit Jiraporn is an Associate Professor of Finance in the School of Graduate Professional Studies at Pennsylvania
State University in Malvern, PA and a Visiting Associate Professor of Finance at The National Institute of Development
Administration (NIDA) and SASIN Graduate Institute of Business Administration (GIBA), Chulalongkorn University in
Bangkok, Thailand. Napatsorn Jiraporn is an Assistant Professor of Marketing at State University of New York (SUNY) at
Oswego in Oswego, NY. Ardisak Boeprasert is a PhD candidate at the National Institute of Development Administration
(NIDA), Bangkok, Thailand and The Petroleum Authority of Thailand in Bangkok, Thailand. Kiyoung Chang is an
Associate Professor of Finance at the University of South Florida-Sarasota-Manatee in Sarasota, FL 34243.
Financial Management • Fall 2014 • pages 505 - 531

This content downloaded from

14.139.235.161 on Mon, 21 Nov 2022 09:28:56 UTC
All use subject to https://about.jstor.org/terms
 506

can asocial spending is "ak

1986; Navarro, 1988; We
activities are highly obser
that firms are influenced
Our results strongly conf
the degree of CSR of a gi
of the geographically prox
code exhibit similar CSR
significant even after cont
profitability, research an
as well as possible variatio
We explore the impact of
geographic similarity in C
Therefore, it comes from
in CSR across geographical
CSR level of the surround
squares (2SLS) analysis and
Therefore, CSR is recogni
impact of the endogenei
estimates and find that t
without properly account
underestimated by approx
We are also aware of a po
characteristics. Certain fir
credit ratings, possibly lea
that our results are not co
effects analysis, which co
bias. We obtain consistent
significant, but is also eco
deviation improves the fir
The literature on the effe
evidence appears to demon
2003; Saiia, Carroll, and Bu
and Davidson, 2010), a larg
(Aupperle, Carroll, and H
and Ferris, 1997; Teoh, W
documented a large numbe
default risk is found to be
and Reiter, 1992), whereas
and Urwitz, 1979; Boardm
are recent studies that de

1 The instrumental variable (IV

attenuation bias resulting from m
cient estimates toward zero. Sinc
approach helps mitigates the bias
2004).

This content downloaded from

14.139.235.161 on Mon, 21 Nov 2022 09:28:56 UTC
All use subject to https://about.jstor.org/terms
Jiraporn, et al. » Evidence from Geographic Identification

credit ratings (Ashbaugh-Skaife, Collins, an

of nonfinancial attributes, however, such as
The results of our study contribute to the lit
demonstrate that geography plays an impor
standard economic theory does not recogni
investor clientele. We argue that these facto
of geographically close firms. Finally, by in
paper, for the first time, points to a potent
of debt.
The remainder of this article is organized as follows. Section I reviews the prior literature
and develops the hypotheses. Section II discusses the sample formation and describes the data.
Section III presents the results regarding geography and CSR. Section IV estimates the impact
of CSR on credit ratings using geographic identification, while Section V offers our concluding
remarks.

I. Prior Research and Hypothesis Development

A. Corporate Social Responsibility (CSR)

Companies are usually considered socially responsible when they voluntarily take actions
that benefit not only their shareholders, but also broader groups of stakeholders as well as
society at large. Examples of CSR activities include philanthropy, environmental compliance
and improvements, community participation and enhancements, product safety, and promotion
of human rights. For instance, Tyson Foods donates large amounts of chicken to local food
banks as part of its hunger relief program. Haagen-Dazs promotes an environmental initiative to
preserve honeybees, which are disappearing at an alarming rate. The company supports research
that explores ways in which honeybees can be better protected in the environment. A growing
number of firms refuse to use parts manufactured in countries where human rights violations are
rampant. As awareness on CSR strengthens over the years, more and more companies engage in
more diverse CSR activities.
In the literature, a large number of recent studies explore the issue of CSR. For instance, Jiao
(2010) reports that more socially responsible companies exhibit higher firm value, as measured by
Tobin 's q. Benson and Davidson (2010) examine the relation between firm value, stakeholder re-
lations, and executive compensation. They find that stakeholder management is positively related
to firm value. However, firms do not compensate managers for having good relationships with
stakeholders. Atkas, Bodt, and Cousin (201 1) determine that acquisitions where the target firm is
more socially responsible elicit more favorable stock market reactions, suggesting that investors
value CSR. Finally, Deng, Kang, and Low (2013) report higher acquisition announcement returns
when the acquirer is more socially responsible. Their evidence supports the stakeholder value
maximization view. Firm that integrate various stakeholders' interests in their business operations
make better investments resulting in higher shareholder wealth and corporate value.

B. Geography and Corporate Policy

A distinct area of the literature examines the impact of geography on firm policy. For example,
Kedia and Rajgopal (2009) find that a firm is more likely to grant stock options to rank and file
employees if other firms in the same area do so. The authors argue that location matters because
of local labor market conditions and social interaction with neighboring firms. John, Knyazeva,

This content downloaded from

14.139.235.161 on Mon, 21 Nov 2022 09:28:56 UTC
All use subject to https://about.jstor.org/terms
 508

and Knyazeva (201 1) rep

likely to use dividends t
geography can be found
from local acquisitions
dividend clientele influe
also affects chief execut
Knyazeva, Knyazeva, and
equity financing and be
location also influences
2008; Gao, Ng, and Wan

C. Geographic Proxi
A
crucial hypothesis of
of a firm is influenced b
made in support this hy
investor clientele. A larg
and individual investors
Moskowitz, 1999; Grin
Massa and Simonov, 200
different preferences fo
severe, local investors
environmentally conscio
affected by the same loc
policies.
In addition, local competition is expected to be a critical factor. CSR is usually viewed positively
by investors. How investors view a firm's CSR policy may depend, in part, on the CSR policies of
neighboring firms. Investors may have a negative view of a firm if its CSR policy is much weaker
than those of the surrounding firms. For this reason, when formulating its CSR policy, a firm
must take into consideration the CSR policies of surrounding firms. Local competition to attract
investors forces the CSR policies of geographically proximate firms to be similar. Moreover,
social interaction and peer effects can be particularly important for corporate decision makers.
Managers who work in the same geographic area usually have opportunities to network and
build valuable relationships with their peers, exchanging ideas and learning from one another's
experience (Pirinsky and Wang, 2010). The social interactions and peer effects of the executives
in the same geographic area have the potential to make the CSR policies of neighboring firms
more similar.

D. Corporate Social Responsibility and Credit Ratings

Credit ratings provide information about default likelihood and the financial health of firms,
thereby reducing a duplication of effort in the financial markets. They enable investors to readily
assess broad risk properties of a large number of firms using a single and well known scale.
Additionally, credit ratings play a crucial role in regulation and private contracting, as a tool

2 Consistent with this notion, Pirinsky and Wang (2010) argue that the local preference of investors naturally creates a
clientele of investors from the same region, which could have an influence on major corporate policies. For example, a
company that is headquartered in Boston would have a disproportionately large number of (local) institutional investors
that could affect its corporate financial policies.

This content downloaded from

14.139.235.161 on Mon, 21 Nov 2022 09:28:56 UTC
All use subject to https://about.jstor.org/terms
Jiraporn, et al. » Evidence from Geographic Identification

for measuring and mitigating risk. We

firm, such a positive effect should be v
the benefit of CSR should be reflected i
sophisticated financial intermediaries w
better information than the average inv
rating agencies are more likely to discove
Prior literature suggests that CSR engag
insulates them from adverse events in th
exposed to a lower degreeof risk (Godfre
and Hansen, 2009). Debtholders and credit
This risk mitigation view suggests that
agency view argues that CSR investment
to overinvest in CSR to enhance their ow
wealth (Friedman and Friedman, 1962
investments as an agency problem that re
shareholders. Because the risk mitigation
the effect of CSR on credit ratings rema

II. Sample Construction and

A. Sample Formation

Our CSR sample is from Kinder, Lydenb

is the most widely recognized and reliabl
peer reviewed articles. We obtain data on
accounting characteristics from Compust
digit zip code. A three-digit zip code is
Geographic and economic data, such as lan
Census Bureau report. Our final sample i
from 1995 to 2007. Please note that we
the variables. As a robustness check, we
R&D and advertising expenditures are zer
observations remaining after each screen

B. Corporate Social Responsibilit

One challenge for research in CSR is the
alternative CSR measures. However, KLD

3 Credit ratings agencies are more likely to disco

agencies have more resources dedicated to the analy
economies of scale, where the expertise developed fo
rating agencies accumulate their expertise over tim
activities such as CSR.

4 Our sample includes 133 unique three-digit zip codes with at least five corporate headquarters. There are a total of
323 unique three-digit zip codes with at least one corporate headquarters in the United States (according to the data
from Compustat). Thus, our sample represents 41.20% of all of the three-digit zip codes with at least one corporate
headquarters. We also check the robustness of our results using two-digit zip codes and obtain similar results.
5 Although there are other alternative measures of CSR, KLD is the most widely accepted. Other CSR measures can be
found in Carroll (1991), Hansen and Wernerfelt (1989), and Waddock and Graves (1997).

This content downloaded from

14.139.235.161 on Mon, 21 Nov 2022 09:28:56 UTC
All use subject to https://about.jstor.org/terms
 510

KLD includes strength ratin

(3) corporate governance,
(8) alcohol, (9) gambling,
assigns strengths and con
just exclusionary screens
a company can receive cr
registered for its environ
total CSR score. The total
and Roberts, 201 1).

C. Firm Location

The modern firm has fuzzy boundaries as its operations and management could encompass
numerous countries around the globe. The academic literature usually defines a firm's location
the location of its corporate headquarters. Corporate headquarters are usually close to corporat
core business activities. More importantly, corporate headquarters are the place where corpora
decision makers reside and are the center of information exchange between the firm and it
suppliers, service providers, and investors (Davis and Henderson, 2008; Pirinsky and Wang,
2010).
In this study, we follow the literature and define a firm's location by the location of its headquar-
ters. We identify the zip code of the headquarters location and classify firms as "geographically
proximate" if they are located in the same three-digit zip code. This method of geographic identi-
fication has several advantages. First, zip codes are easily identifiable and have clear boundaries,
facilitating geographic identification. Additionally, the zip code is determined by the US Postal
Service to maximize efficient mail delivery. Thus, it is unlikely related to corporate financial
characteristics. Moreover, zip code changes are rare. The US Postal Service does not modify zip
codes based on corporate performance.6 For these reasons, a geographic location based on a zip
code is likely exogenous.

D. Credit Ratings

We follow Klock, Mansi, and Maxwell (2005) and compute credit ratings using a conversion
process in which AAA-rated bonds are assigned a value of 22 and D-rated bonds are assigned a
value of 1. For example, a firm with an A+ rating from S&P would receive a score of 18. We
focus on the S&P credit ratings because Litov (2005) and Ashbaugh-Skaife et al. (2006) find that
the S&P ratings reflect the overall creditworthiness of the firm. Table I provides our credit rating
conversion.

E. Summary Statistics

Table II displays the descriptive statistics. A few observations are worth noting. The average
credit rating is 13.43, corresponding to BBB- . On average, the sample firms are profitable as
indicated by the average earnings before interest and taxes (EBIT) ratio of 0.09 of total assets.
The average leverage, calculated as total debt divided by total assets, is 0.30. R&D expenditures
and advertising spending average 0.03 and 0.01, respectively. The fixed assets ratio, computed as

6 Zip codes are occasionally reassigned based on demographics and mail delivery efficiency. For instance, when a rural
area becomes suburban, a new zip code is introduced. It is unlikely that corporate financial policies influence how the U.S.
Postal Service reallocates zip codes. In this sense, geographic identification based on zip codes is probably exogenous.

This content downloaded from

14.139.235.161 on Mon, 21 Nov 2022 09:28:56 UTC
All use subject to https://about.jstor.org/terms
Jiraporn, et al. » Evidence from Geographic Identification

Table I. Credit Rating Con

Conversion Number S&P Ratings

22 AAA
21 AA+
20 AA
19 AA-
18 A+
17 A
16 A-
15 BBB+
14 BBB
13 BBB-
12 BB+
11 BB
10 BB-
9 B+
8 B
7 B-
6 CCC+
5 CCC
4 CCC-
3 CC
2 C
1 D

Table

Credit ratin
R&D expend
assets. The f
industry (S
dummy is e
based on the

~ Mean Median SD 25th 75th

Credit Ratings 13.43 13.83 3.53 10.00 16.00
Total assets 13,484 3889 46,122 1872 9701
EBIT/total assets 0.09 0.09 0.10 0.06 0.14
Leverage 0.30 0.27 0.18 0.17 0.41
R&D intensity 0.03 0.01 0.05 0.00 0.04
Advertising intensity 0.01 0.00 0.06 0.00 0.00
Fixed assets ratio 0.09 0.07 0.12 0.00 0.13
Regulated 0.10 - - - -
S&P 500 0.53 - -
CSR score 0.22 0.00 2.67 -1.00 1.00

This content downloaded from

14.139.235.161 on Mon, 21 Nov 2022 09:28:56 UTC
All use subject to https://about.jstor.org/terms
 512 Financial Management • Fall 201 4
Table III. Most and Least Social
Most Socially Responsible Least Socially Responsible

3-DigitZip Area Avg. CSR 3-Digit Zip Area AVG. CSR

105 West Chester, New 2.35 727 Fayetteville, Arkansas -2.35
York
061 Hartford, Connecticut 2.27 443 Columbus, Ohio -2.31
551 St. Paul, Minnesota 1.95 662 Shawnee Mission, -2.10
Kansas
102 New York, New York 1.73 151 Pittsburgh, -1.60
Pennsylvania
941 San Francisco, 1.54 336 Tampa, Florida -1.56
California
971 Portland, Oregon 1.42 841 Salt Lake City, Utah -1.46
970 Portland, Oregon 1.42 153 Pittsburgh, -1.40
Pennsylvania
462 Indianapolis, Indiana 1.39 220 Northern Area, -1.26
Virginia
943 Palo Alto, California 1 .20 274 Greensboro, North - 1 .20
Carolina
146 Rochester, New York 1.13 731 Oklahoma City, -1.11
Oklahoma

plant, property, and equipment divided by total assets, averages 0.09. About 10% of the sample
firms are regulated (i.e., they belong to either the financial or utility industries). Approximately
53% of the sample firms are included in the S&P 500 Index. The average CSR score is 0.22. This
slightly positive average CSR score indicates that, on average, there are more CSR strengths than
CSR concerns.

F. Most and Least Socially Responsible Areas

Table III reports the 10 most socially responsible areas and the 10 least socially responsible
areas. The area where firms have the highest average CSR score is the three-digit zip code 105
in West Chester, New York. The next most socially responsible areas are Hartford, Connecticut,
St. Paul, Minnesota, and New York, New York.
The rest of the top 10 most socially responsible areas are predominantly in the West (California
and Oregon) and the Midwest (Indiana). Alternatively, the area where firms are given the lowest
average CSR score is the three-digit zip code 727 for Fayetteville, Arkansas. The rest of the top
10 least socially responsible areas scatter across the country, covering states in different regions
including the East (Pennsylvania and Virginia), the Southeast (Florida and North Carolina), and
the Midwest (Kansas and Ohio).

G. Control Variables

Based on the literature, we control for a number of firm characteristics that have been
to influence CSR. We control for firm size by including the log of total assets. Large firm
to attract more attention and pressure to respond to stakeholders' demands (Burke et al.,
Large firms are expected to exhibit more social responsibility. We control for profitabili

This content downloaded from

14.139.235.161 on Mon, 21 Nov 2022 09:28:56 UTC
All use subject to https://about.jstor.org/terms
Jiraporn, et al. « Evidence from Geographic Identification

including the ratio of EBIT to total assets.

more socially responsible. We also include th
leverage. Mc Williams and Siegel (2001) arg
Accordingly, we include the ratio of R&D e
Furthermore, we control for advertising
penditures to total assets. Since CSR promo
advertising. We also control for corporate i
tures to total assets. Firms that make heavy i
included in the S&P 500 enjoy more visibili
ble. Therefore, we include a dummy variable
500 Index. The awareness of CSR has incre
over time in CSR by including year dummie
effects. We by control for industry effects
two digits of the standard industrial class
econometrically inadvisable to include indu
control for any possible industry effects.

III. Geographic Proximity and

A. Geographic Proximity and CSR

Following the method in Bouwman (201 1),
exploring how in a particular firm's CSR is
proximate firms, while controlling for ot
reported in Table IV. The dependent variab
error due to clustering at the firm level. Ea
We compute the average CSR score of all o
In Model 1, the coefficient is positive and
geographically close firms. We control for i
score (base on the two-digit SIC). Year dum
variation over time. Thus, after controlling f
and time effects, the result indicates that
levels of the surrounding firms.7 It is impo
beyond the industry effect documented in
scoreof the neighboring firms is approximat
industry average, suggesting that the geog
effect.
To ensure that our result is robust, we execute additional regressions. Financial and utility firms
are unique as they are subject to regulation. The geographic effect of CSR may be different for

7 We use contemporaneous variables in Table IV. To alleviate concerns regarding possible reverse causality, we also run
a regression where the average CSR score of the surrounding firms is lagged by one year. The result remains similar.
Furthermore, we investigate the impact of geography on CSR using the same method as Gao et al. (201 1). In particular,
we construct a location dummy variable for each three-digit zip code. We regress the CSR score on all of the control
variables, including the industry dummies and the location dummies. Then, we examine whether the location dummies
are jointly significant. The F-statistics are significant suggesting that geographic location does matter to CSR. The large
F-statistics allow us to reject the null hypothesis that geographic location bears no effect on CSR. It is important to note
that the industry dummies are also included and are jointly significant. Thus, the location effects are above and beyond
those explained by industry.

This content downloaded from

14.139.235.161 on Mon, 21 Nov 2022 09:28:56 UTC
All use subject to https://about.jstor.org/terms
51 4 Financial Management • Fall 201 4
Table IV. The Impact of Geography

Credit ratings are the S&P credit ratings. Levera

R&D expenditures divided by total assets. Advertis
assets. The fixed assets ratio is plant, property, an
industry (SIC 6000 to 6999) and utility industry
500 dummy is equal to one if the firm is inclu
score is based on the KLD ratings and represents th
dummies are based on the first two digits of the
located in the same three-digit zip code as firm i.
from the Census Bureau report.

Model 1 Model 2 Model 3

(i-statistic) (i-statistic) (f-statistic)
Full Excluding Regulated With Additional
Sample Firms Controls
Constant -6.65*** -6.88*** -7.62
(-7.98) (-7.76) (-1.03)
Average CSR score of 0.25*** 0.29*** 0.25**
geographically close (2.43) (2.59) (2.37)
firms
Average CSR score 1.35*** 1.38*** 1.34***
of industry peers (7.85) (7.82) (7.85)
Ln(total assets) 0.67*** 0.70*** 0.69***
(7.61) (7.49) (7.69)
EBIT/total assets 2.72*** 2.56*** 2.66***
(2.82) (2.64) (2.72)
Leverage 0.02 -0.13 -0.01
(0.05) (-0.24) (-0.02)
R&D intensity 9.41*** 9.76*** 9.40***
(4.68) (4.61) (4.60)
Advertising intensity 0.31 1.02 0.41
(0.49) (1.59) (0.63)
Capital expenditures -0.02 -0.34** -0.02
ratio (-0.16) (-2.09) (-0.15)
S&P 500 0.06 0.10 0.06
(0.31) (0.45) (0.30)
No. of firms in the same -0.01
3-digit ZIP (-1.19)
Ln(land area) -0. 1 4
(-0.82)
Ln(population) -0. 1 4
(-0.56)
Ln(household income) 0.49
(0.67)
Year dummies Yes Yes Yes
Adjusted R2 26.72% 27.45% 26.98%
N 2,516 2,241 2,516

*** Significant at the 0.01 level.

"Significant at the 0.05 level.
* Significant at the 0.10 level.

This content downloaded from

14.139.235.161 on Mon, 21 Nov 2022 09:28:56 UTC
All use subject to https://about.jstor.org/terms
Jiraporn, et al. » Evidence from Geographic Identification

these firms. In Model 2, we run a regression

coefficient of the average CSR score stills re
excluding regulated firms, the result remain
specific factors may influence CSR. As a r
that are specific to each zip code. In particul
population, land area, and average househol
regression results with the additional contro
a positive and significant coefficient. Even
factors, the effect of the CSR of the surroun

B. Additional Robustness Checks

We execute additional robustness checks. First, we employ an alternative definiti

ographic area. Instead of the three-digit zip code, we use only the first two digits.
results remain consistent. Next, we hypothesize that firms that are included in the S&
are large and are expected to be less affected by local influences such as local comp
investor clientele. If our results are indeed driven by the effect of geographic proximity
the effect to be weaker for the S&P 500 firms, which operate nationally as well as int
To test this conjecture, we construct a dummy variable equal to one if the firm is
S&P 500 Index and zero otherwise. Then, we interact this variable with the average C
the surrounding firms. The coefficient of this interaction term should reveal the rel
of the average CSR score on S&P 500 vs. non-S&P 500 firms. The coefficient ap
negative and significant suggesting that the average CSR score of the neighboring
much weaker effect on the CSR of the S&P 500 firms. This is consistent with our ex
and with the notion that our results are driven by geographic proximity.9

C. Possible Reverse Causality

Thus far, we assume that the direction of causality runs from the average CSR score of the
neighboring firms to the CSR score of a given firm. Reverse causality would imply that the CSR
score of a given firm influences the average CSR score of the geographically proximate firms.
Reverse causality is unlikely as we require that each zip code contain at least five firms. Thus,
the influence of a single firm on the other four firms should not be so large (the average number
of firms in a zip code area is 22, the median 14). To further ensure robustness, we increase the
minimum number of firms in each zip code to 20 and 30. The results are presented in Table V.
The results hold even when the minimum number of firms is raised to 20 and 30. When one zip
code contains 30 firms, the influence of one firm on the other 29 firms in the same area should

8 We obtain the zip-code specific data from the US Census Bureau.

9 Because Compustat reports only the current address, it is possible that some sample firms relocated their headquarters
and, as such, did not have the same addresses throughout the sample period. This problem is not particularly serious as
headquarters relocations are quite rare (Alii, Ramirez, and Yung, 1991; Pirinsky and Wang, 2006). We execute additional
tests that should alleviate this problem. First, Pirinsky and Wang (2006) find that the vast majority of the headquarter
relocations in their sample come from small firms. Large firms rarely relocate their headquarters. Thus, we concentrate
on the S&P 500 firms. These firms are large and are highly unlikely to have their headquarters relocated. Consequently,
this subsample is considerably less vulnerable to the relocation bias. Using only the S & P 500 firms, we obtain similar
results. In addition, we focus on the subsample of firms in the last year of the sample, where we know with certainty
that the addresses are current. Evidently, the relocation bias cannot influence this subsample. We obtain similar results.
It appears that the relocation bias does not unduly affect our conclusion.

This content downloaded from

14.139.235.161 on Mon, 21 Nov 2022 09:28:56 UTC
All use subject to https://about.jstor.org/terms
 516

Table V. Regression

Model 1 (f-statistic) Model 2 (f-statistic)

Sample Restrictions At least 20 Firms At least 30 Firms
Average CSR score 0.57*" 0.84"*
of geographically close firms (2.41) (2.56)
Average CSR score 1.54*** 1.56***
of industry peers (4.76) (3.82)
Control variables Yes Yes
Year dummies Yes Yes
Adjusted R2 32.93% 36.44%
N 913 533

*** Significan
** Significant
* Significant

be negligib
weak.10
We also execute additional empirical tests to mitigate the concern for reverse causality. First,
we lag the average CSR score of the geographically close firms by one period (i.e., we regress
the CSR score of a given firm at time t on the average CSR score of the surrounding firms in
year t - 1). We continue to obtain consistent results. Moreover, we identify the first year when
each firm shows up in the sample. Then, we replace the average CSR score of the neighboring
firms in each year by the value in the earliest year. The logic is that the average CSR score in the
earliest year could not have resulted from the CSR score in any of the subsequent years, making
reverse causality unlikely. The regression result holds when we use the value in the earliest year.
These additional tests increase our confidence that an endogeneity bias due to reverse causality
is unlikely in our sample.

D. Fixed Effects Regression Analysis, Regression Based on Changes, and the

Possible Urban Effect

To gain further insight, we execute a fixed effects regression analysis. One critical adva
of the fixed effects analysis is that it only captures the variation over time. As such, it co
for the possible effects of unobservable firm characteristics that remain constant over
thereby minimizing the omitted variable bias. Table VI reports the results of the fixed ef
regressions. In Model 1, the average CSR score of the geographically close firms has a po
and significant coefficient. The result suggests that when the neighboring firms become m
socially responsible, firm i also increases its own CSR. In Model 3, we average the CSR scor
time for each firm. Thus, the regression in Model 3 captures only the cross-sectional varia
The coefficient of the average CSR score is positive and significant. When we only look

10 Another possible argument for reverse causality is that firms may choose a geographic location based on
other words, firms with high CSR choose to be located closer to other firms with high CSR. This is highly unli
Headquarter relocations are rare. It is implausible that firms change location in response to CSR. Furthermore, C
received much more attention recently. Thus, CSR has increased significantly over the years. Yet, firms still rarely
their headquarter locations. If reverse causality were valid, we would observe many more corporate relocations a
levels change over the years.

This content downloaded from

14.139.235.161 on Mon, 21 Nov 2022 09:28:56 UTC
All use subject to https://about.jstor.org/terms
Jiraporn, et al. • Evidence from Geographic Identification

Table VI. Within versus Between

and Random-effects Reg
With-in (Fixed Effects) Betw
Model 1 (f-statistic) Model
Average CSR score 0.51*** 0.88*** 0.87***
of geographically close (7.34) (7.63) (13.71)
firms
Average CSR score 0.08 0.93*** 0.80***
of industry Peers (1 07) (6.50) (9.90)
Control variables Yes Yes Yes
Year dummies Yes No Yes
Adjusted R2 78.68% 28.46% 34.39%
N 2,516 2,516 2.516

*** Significant at the 0.01 level.

** Significant at the 0.05 level.
•Significant at the 0.10 level.

cross-sectional variation
random-effects regressi
In addition, we perform
changes in the CSR score
firms. An analysis based
The results are displayed
CSR score is positive and
improve their CSR over
the evidence from the f
Furthermore, it is conce
metropolitan area may b
firmsin a rural area. We
located in one of the larg
are reported in Model 2
Apparently, the location
urban and rural areas.12

IV. Effect of Cor

The literature suggests

risk mitigation view arg
while the overinvestmen
hypotheses.

11 We obtain the population data from the US Census Bureau. According to the 2010 Census, the 10 largest cities are New
York, Los Angeles, Chicago, Houston, Philadelphia, Phoenix, San Antonio, San Diego, Dallas, and San Jose.
12 We thank an anonymous referee for this suggestion.

This content downloaded from

14.139.235.161 on Mon, 21 Nov 2022 09:28:56 UTC
All use subject to https://about.jstor.org/terms
518

Table VII. Regressio

Possible Urban Effect

Regression Based Model 1 Testing the Model 2

on Changes (f-statistic) Urban Effect (t-statistic)
Constant -0.32*** Constant -7.42***
(-11.01) (-6.67)
Urban Area ( 1 if Urban) -0. 1 7
(-0.56)
A Average CSR score 0.16** Average CSR score 0.18*
of geographically close firms (2.22) of geographically close firms (1 .76)
A Ln(total assets) -0. 1 7 Ln (total assets) 0.73***
(-0.99) (7.09)
A EBIT/total assets -1.09*** EBIT/total assets 2.59***
(-2.83) (2.58)
A Leverage 0.35 Leverage -0.5 1
(0.83) (-0.96)
A R&D Intensity -0.30 R&D intensity 1 1 .90***
( -0.21) (5.02)
A Advertising intensity 3.41* Advertising intensity 0.84
(1.71) (1.46)
A Capital expenditures ratio -0.23*** Capital expenditures ratio -0. 1 5
(-3.22) (-1.55)
S&P500 -0.91 S&P500 -0.07
(0.07) (-0.34)
Industry dummies Yes Industry dummies Yes
Adjusted R2 2.70% Adjusted R2 30.65%
N 1,978 N 2,516

*** Significant at the 0.01

** Significant at the 0.05
* Significant at the 0.10

A. The Risk Miti

One crucial argume

et al., 1 988; Stārks,
for the firm that f
2005; Gardberg and
the negative assessm
making firms less
and Zhu, 2010; Ye a
government and the
that would otherwi
Sharfman and Fer
number of prior st
Dufresne and Sava
Goss and Roberts, 2
rating agencies and
a firm because their
suggests that more

This content downloaded from

14.139.235.161 on Mon, 21 Nov 2022 09:28:56 UTC
All use subject to https://about.jstor.org/terms
Jiraporn, et al. « Evidence from Geographic Identification

B. The Overinvestment View

Based on agency theory, this perspective argues that CSR investments constitute costly dive
sions of firm resources. CSR engagement represents an agency conflict between managers an
shareholders. Managers over invest in CSR and gain private benefits at the expense of sharehol
ers. Bernea and Rubin (2010) contend that overinvestments in CSR occur because the credit for
CSR initiatives accrues principally to the manager, while the cost is borne by the shareholders
Friedman and Fridman (1962) argue that firms should not be involved in philanthropy as share
holders can better engage in philanthropic activities on their own. Porter and Kramer (2006) al
make a similar argument that philanthropy should be left in the hands of individual shareholde
Thus, CSR initiatives represent an inefficient deployment of corporate resources that do no
create value for shareholders. To the extent that CSR activities constitute an agency cost or
wasteful diversion of resources, firm value is expected to decline, making both shareholders an
debtholders worse off. The overinvestment view suggests that recognizing the agency conflict
engendered by CSR activities, credit rating agencies assign lower credit ratings to firms wit
higher CSR.

C. Empirical Results

A crucial challenge in the empirical literature in CSR is the endogeneity bias, which prevent
researchers from drawing casual inferences. Firms with better CSR may enjoy better firm per
formance. At the same time, better performing firms can afford to invest more in CSR. Thus,
is unclear whether the direction of causality runs from CSR to firm performance or vice vers
One possible solution to the endogeneity problem is the instrumental variable (IV) technique.
The idea is to identify a variable that is highly correlated with CSR, yet does not influence firm
performance except through CSR. It is difficult to find such a variable as financial variables ten
to be correlated. Geographic location has been used as an instrumental variable in many prio
studies as it is fixed and more likely to be exogenous.
Previously, we have demonstrated robust evidence that the degree of CSR of a given firm
in a particular area is influenced by the CSR of geographically proximate firms. We exploit th
variation in CSR across the geographic locations and estimate the impact of CSR on credit ratin
In particular, we employ as our instrumental variable the average CSR score of the surroundin
firms in the same three-digit zip code. This variable should be a legitimate instrument for tw
reasons. First, this variable is clearly related to the CSR score of a given firm, as demonstrate
earlier. As such, it meets the relevance requirement for an instrumental variable. In addition,
plausibly meets the exclusion requirement; that is, it is not correlated with credit ratings. The U
Postal Service allocates zip codes exclusively based on efficiency in postal delivery, not corpora
financial policies or outcomes. Thus, the variation in CSR across zip codes is likely exogenous
The two-stage least squares (2SLS) results are presented in Table VIII. Model 1 is the first stag
regression, where we estimate the CSR score using the average CSR score of the surrounding
firms. We include all of the control variables, as well as the year and industry dummies. As
expected, the average CSR score of the geographically close firms carries a positive and significa
coefficient.13 Model 2 is the second stage regression, where we take the instrumented value of

13 To ensure the strength of our instrumental variable, we compute the Anderson canonical correlation LM statistic
under-identification, as well as the Craig-Donald Wald F-statistic for weak identification. These two statistics examin
whether the relationship between the instrumental variable and the instruments is sufficiently strong to justify infere
from the results. We find that the two statistics are highly significant. Thus, our instrumental variable does not suff
from the weak instrument problem.

This content downloaded from

14.139.235.161 on Mon, 21 Nov 2022 09:28:56 UTC
All use subject to https://about.jstor.org/terms
520

Table VIII. Two-Sta

Credit Ratin

Credit ratings are the S&

R&D expenditures divided
assets. The fixed assets rat
industry (SIC 6000 to 699
500 dummy is equal to on
score is based on the KLD r
dummies are based on the
located in the same three-d

Modell Model 2 Model 3 Model 4

(f-statistic) (f-statistic) (f-statistic) (f-statistic)
First Stage Second Stage First Stage Second Stage
CSR Score Credit Rating CSR Score Credit Rating
Constant -6.70*** 6.31*** -7.20*** 6.27***
(-6.23) (5.85) (-6.76) (5.88)
Average CSR score 1.07*** - 1.03***
of geographically close firms (17.55) (1 7.09)
Average CSR score of industry - 0.95***
Peers (2-digit SIC) (7.51)
CSR score (instrumented) - 0.23*** - 0.22***
(4.28) (4.55)
Ln(total assets) 0.66*** 1.27*** 0.70*** 1.27***
(16.95) (23.98) (17.39) (25.11)
EBIT/total assets 2.13*** 8.97*** 1.99*** 8.99***
(4.19) (18.14) (3.94) (18.30)
Leverage -0.23 -4.54*** -0.22 -4.54***
(-0.77) (-16.60) (-0.77) (-16.64)
R&D intensity 9.54*** -4.12*** 9.61*** -4.06***
(8.08) (-3.24) (8.23) (-3.25)
Advertising intensity 0.61 0.09 0.57 -0.09
(0.83) (-0.13) (0.79) (-0.13)
Capital expenditures ratio -0.21* 0.35*** -0.19* 0.35***
(-1.94) (3.42) (-1.79) (3.41)
S&P 500 -0.06 -0.07 -0.06 -0.07
(-0.67) (-0.83) (-0.59) (-0.83)
Dividend/total assets 2.11* 8.63*** 1.93* 8.64***
(1.84) (7.99) (1.70) (8.02)
Fixed assets/total assets -0.31 2.86*** -0.19 2.86***
(-0.66) (6.58) (-0.42) (6.58)
Regulated 0.69 0.55 1.54 -1.91
(0.48) (-0.41) (0.87) (-1.14)
Year dummies Yes Yes Yes Yes
Industry dummies Yes Yes Yes Yes
Adjusted R2 38.27% 68.28% 39.66% 68.33%
Shea's (1997) partial R2 11.00% - 13.20%
F-statistics 305.32*** 75.63*** 185.78*** 75.78***
Sargan's (1958) Statistic - 0.07
N 2,516 2,516 2,516 2,516

*** Significant at the 0.01 level.

** Significant at the 0.05 level.
* Significant at the 0.10 level.

This content downloaded from

14.139.235.161 on Mon, 21 Nov 2022 09:28:56 UTC
All use subject to https://about.jstor.org/terms
Jiraporn, et al. » Evidence from Geographic Identification

the CSR score from the first stage and incl

regression. The coefficient of the instrumen
the evidence suggests that CSR improves c
significance, the coefficient of the CSR sco
an improvement in CSR by one standard dev
We execute further tests to ensure robust
(i.e., there is one endogenous variable and o
identified, it is not possible to calculate Sarg
ment is valid. As a consequence, we add one
enabling us to execute a test of overidentify
Several prior studies employ an industry ave
may be related to firm-level CSR. However,
that there are many firms in an industry,
exogenous (the average number of firms in
industry average CSR as an additional instrum
in Model 3, where we use as our instrument
industry-average CSR score. Both instrume
by their positive and highly significant coef
regression, we regress credit ratings on the
coefficient of the instrumented CSR score i
reveals that firms that engage in more CSR
to note that we can now conduct a test of o
in Model 4 is 0.067, which is statistically in
significant, our instrumental variables appea
Thus far, we have used the 2SLS estimation
nitude of the endogeneity bias, we run an O
and the CSR score that is not instrumente
coefficient of the CSR score is positive and
the result indicates that CSR has a favorabl
subject to the endogeneity bias and is not re
of CSR from the OLS regression is 0. 126, w
(Model 2 of Table VII) is 0.228. Without a pr
on credit rating would be underestimated
cant difference between the two estimates
negligible.
To further corroborate the result, we employ an alternative instrumental variable. Instead of
using the average CSR score of the surrounding firms in the same year, we now use the average
CSR score of neighboring firms from the earliest year in the sample. This instrumental variable is
even more unlikely related to credit ratings. Not only does it come from outside the firm (average

14 We also use the average CSR score of the surrounding firms in the same two-digit, instead of three-digit, zip code and
obtain consistent results.

15 To determine the economic significance, we multiply the standard deviation of the CSR score by the coefficient (2.672
X 0.228 = 0.609). Thus, an increase in CSR by one standard deviation improves the firm's credit rating by 0.609 or about
4.5% of the average credit rating (0.609/13.42 = 0.045).
16 Gao et al. (2011) report that leverage of firms located in close proximity exhibits similarity. This finding may be
relevant to our study as leverage and financial risk are taken into account by credit rating agencies. To explore this issue,
we run a regression similar to those in Table VII, but include, as an additional control variable, the average leverage of
firms located in the same three-digit zip code. The result remains consistent.

This content downloaded from

14.139.235.161 on Mon, 21 Nov 2022 09:28:56 UTC
All use subject to https://about.jstor.org/terms
522

Table IX. Two-Stag

Credit Ratin

Credit ratings are the S&P cre

divided by total assets. Adver
is plant, property, and equipm
industry (SIC 4900 to 4999)
in the S&P 500 Index and zer
strengths minus the concerns.
proximate firms are those loc

Model 1 Model 2 Model 3 Model 4 Model 5

OLS First Stage First Stage Second

(Earliest year Second (Fixed Stage
as instrument) Stage Effects) (Fixed Effects)
Credit Rating CSR Score Credit Rating CSR Score Credit Rating
Constant -3.57*** -4.74*** 2.76***
(-3.35) (-6.90) (1.24)
CSR score 0.13*** - - - -
(3.41)
CSR score (instrumented) - - 1.38*** - 0.31***
(3.28) (5.51)

Average CSR score of - - 0.73***

geographically close firms (11.10)
Average CSR score of 0.20*** - - -
geographically close firms (3.59)
(earliest year)
Ln(total assets) 1.18*** 0.61*** 0.38 0.03 0.76***
(13.83) (15.12) (1.38) (0.28) (11.14)
EBIT/total assets 9.64*** 3.46*** 4.75*** -0.30 1.44***
(5.21) (6.19) (2.75) (-0.51) (3.91)
Leverage -3.93*** -0.20 -3.68*** 0.32 -1.53***
(-6.91) (-0.67) (-7.83) (0.78) (-6.06)
R&D intensity 0.08 14.06*** -18.64*** 1.93 1.78
(0.03) (12.28) (-2.90) (0.96) (1.43)
Advertising intensity 0.08 0.35 -0.49 5.14** -1.21
(0.88) (0.44) (-0.40) (2.43) (-0.90)
Capital expenditures ratio 0.08 -0.10 0.18 -0.27 0.25
(0.13) (-0.97) (1.14) (-1.08) (1.58)
S&P 500 -0.05 0.06 -0.12 -0.44* 0.38***
(-0.27) (0.56) (-0.73) (-2.84) (3.82)
Fixed assets/Total assets 2.37*** -0.18 2.47*** -5.31*** 1.88
(3.12) (-0.43) (3.88) (-4.42) (2.29)
Regulated 1.98*** 0.12 1.15***
(6.90) (0.61) (5.15)
Industry-average 0.56*** -0.04 0.59*** -0.24*** 0.36***
Credit rating (9.72) (-1.21) (10.94) (-7.28) (18.10)
Year dummies Yes Yes Yes Yes Yes
Adjusted /?2 /pseudo R2 66.04% 17.88% - 4.59% 53.61%
Shea's Partial R2 - 0.51% - - -
F-statistics 58.51*** 12.87% 63.86*** 15.96*** 24.36***
N 2,516 2,516 2,516 2,516 2,516

*** Significant at the 0.01 level.

** Significant at the 0.05 level.
* Significant at the 0.10 level.

This content downloaded from

14.139.235.161 on Mon, 21 Nov 2022 09:28:56 UTC
All use subject to https://about.jstor.org/terms
Jiraporn, et al. » Evidence from Geographic Identification

CSR of other firms in the same zip code), it a

instead of year i for each firm). Therefore, it
both dimensions (both in terms of time as w
is probably exogenous. In Table IX, Model 2
the earliest year and the control variables. N
score in the earliest year is positive and sig
CSR score instrumented from the first stag
significantly positive, again corroborating t
ratings.
The above tests are meant to alleviate concerns about reverse causality. Yet, there is another
type of endogeneity that could produce a spurious relationship. It is possible that both firm value
and CSR are related to certain unobservable firm characteristics that are omitted in the model.
We address this potential problem by running a two-stage fixed effects analysis. Like the regular
two-stage model, this approach helps alleviate endogeneity due to reverse causality. In addition,
this method also controls for unobservable firm attributes that remain constant through time,
thus mitigating the omitted variable bias. In Model 4, we run a fixed effects regression of the
CSR score on the average CSR score of the surrounding firms. The coefficient of the average
CSR score is positive and significant. In Model 5, we regress credit ratings on the CSR score
instrumented from the first stage. The coefficient of the instrumented CSR score is positive and
significant. So, even after controlling for both possible reverse causality and unobservable firm
attributes, we continue to find that more socially responsible firms enjoy better credit ratings.

D. Further Analysis for the Risk Mitigation View

The risk mitigation view argues that firms engage in CSR activities to build up moral capital
thereby mitigating the adverse effects of negative events. There are two implications of the risk
mitigation view. First, if CSR activities are associated with lower risk, then the effect of CSR
on credit ratings should be positive. Our results support this notion indicating that more socially
responsible firms enjoy more favorable credit ratings. In addition, if CSR reduces risk, then the
effect should be attributed more to CSR concerns, rather than CSR strengths. To examine this
second implication, we execute additional analysis as follows. We separate the CSR score into
strengths and concerns, instead of combining them into one aggregate CSR score. Then, we
perform a regression analysis for the strengths and the concerns separately. The regression results
are found in Table X. The evidence corroborates our hypothesis. In particular, the strengths carry
an insignificant coefficient, while the concerns exhibit a significantly negative coefficient. The
effects of the strengths and the concerns are indeed asymmetric. This is consistent with the notion
that companies invest in CSR primarily to reduce their risk exposure. The risk mitigation view is
corroborated.

E. Additional Robustness Tests

We execute several additional robustness checks including using an expanded sample

the minimum number of firms in each zip code, using the exclusionary screens, and dist
between investment grade and noninvestment grade credit ratings. Due to space limi
present and discuss the results of these additional tests in the appendix.

This content downloaded from

14.139.235.161 on Mon, 21 Nov 2022 09:28:56 UTC
All use subject to https://about.jstor.org/terms
 524

Table X. The Effect

Total CSR Strengths equal t

total number of CSR conce
Table IX.

Credit Rating Credit Rating

Constant 5.25*** 4.75***
(1.04) (8.59)
Total CSR strengths 0.03
(1.24)
Total CSR concerns -0. 1 1 ***
(-3.98)
Control variables Yes Yes
Year dummies Yes Yes
Industry dummies Yes Yes
Adjusted R2 66.95% 67. 1 9%
N 2,516 2,516

*** Significant at the

** Significant at the 0
* Significant at the 0

VI. Conclud

Due
to market s
firm is likely to
own CSR policy.
firms located in
even after contro
R&D spending, a
firm' CSR is inf
exploit the varia
two-stage least s
credit ratings. Ou
that more sociall

Appendix
Additional Robustness Tests

A. Expanded Sample

We execute further robustness checks. First, in our sample selection process, we lost a su
number of firms due to missing data regarding R&D and advertising expenditures. It is we
in the literature in empirical finance that these two variables are missing for a large
firms in Compustat. One common solution to this problem is to assume that R&D and a
expenditures are zero when the data are missing. This assumption has been used in ma
studies that significantly deal with these two variables. When we apply this assumptio

This content downloaded from

14.139.235.161 on Mon, 21 Nov 2022 09:28:56 UTC
All use subject to https://about.jstor.org/terms
Jiraporn, et al. » Evidence from Geographic Identification

Table A1. Sample Construction an

No.

1 . KLD data, credit ratings, and firm characteristics available 7,085

on Compustat (except for R&D and Advertising
expenditures)
2. There are at least five corporate headquarters in each zip code 4,726
3. R&D and Advertising expenditures available on Compustat 2,5 1 6

study, the sample size increases from 2,5 1 6 to 4,726. The results based on the expanded sample are
shown in Table A2. Model 1 is the first-stage regression where the CSR score is the dependent
variable. The average CSR score of the neighboring firms exhibits a positive and significant
coefficient, confirming the results based on the smaller original sample. Model 2 is the second-
stage model. The dependent variable is credit ratings. The CSR score instrumented from the first
stage shows a significantly positive coefficient. Thus, the results based on the expanded sample
reinforce the conclusion that firms with higher CSR enjoy better credit ratings.
Please note that we do not use the expanded sample as our primary sample as the assumption of
zero R&D and advertising expenditures when the data are missing is made simply to address the
sample size problem in empirical research and, as such, is not based on an economic rationale.
Just because a firm does not report its R&D and advertising expenditures does not necessarily
mean that it does not spend anything on R&D and advertising. As a result, we do not believe that
the results based on this assumption are as reliable as those from our original example (which,
although smaller, contains actual data on R&D and advertising). In any event, the results are
remarkably consistent, regardless of which sample is used.

B. Minimum Number of Firms in Each Zip Code

Our requirement that each zip code has at least five headquarters may seem a bit arbitrary.
We impose this minimum requirement to avoid any mechanical relationship when the number of
firms in each zip code is too low. For robustness, we change the minimum requirement to three
and seven firms. The alternative tests produce consistent results. The minimum number of firms
in each zip code does not materially affect the conclusion.

C. Exclusionary Screens

Please recall that there are six other CSR categories that KLD uses only as exclusionary screens.
These CSR components are not included in the calculation of the CSR score. These categories
are alcohol, firearms, gambling, military, nuclear power, and tobacco. For these CSR categories,
KLD reports only the concerns and not the strengths, while for the other seven categories included
in the CSR score, both the strengths and the concerns are reported. Thus, it is quite evident that
the six exclusionary screens are not meant to be interpreted the same way as the other seven
categories. For the sake of completeness, we construct an index based on these six exclusionary
screens by summing the concerns in these six categories. Then, we run a regression with this index
as an independent variable. The coefficient of the total exclusionary concerns is not significant. It
does not appear that credit ratings agencies take these exclusionary CSR categories into account
when assigning credit ratings. The result confirms our assumption that these six exclusionary
screens should not be interpreted the same way as the other seven CSR categories.

This content downloaded from

14.139.235.161 on Mon, 21 Nov 2022 09:28:56 UTC
All use subject to https://about.jstor.org/terms
526

Table A2. The Effect

Credit ratings are the S&P

R&D expenditures divided b
assets. The fixed assets ratio
industry (SIC 6000 to 6999
500 dummy is equal to on
score is based on the KLD ra
dummies are based on the f
located in the same three-d

First Stage Second Stage

Constant -3.93*** 4M***

(-4.40) (6.04)
CSR score (instrumented) - 0.22***
(5.87)
Average CSR score of 1.1 2*** -
geographically close firms (23.86)

Ln(total assets) 0.4 1 *** 1 .09***

(14.34) (35.07)
EBIT/total assets 3.37*** 9.14***
(7.58) (21.90)
Leverage -0.58 -4.78***
(-2.78) (-25.66)
R&D intensity 12.49*** -4.79***
(10.28) (-3.94)
Advertising intensity 7.56*** -1.54
(5.63) (-1.23)
Capital expenditures ratio 1 .46 -2.29***
(1.47) (-2.59)
S&P 500 0.23 1.02***
(1.09) (5.49)
Fixed assets/total assets 0.82*** 1 . 1 6***
(2.84) (4.55)
Regulated -0.01 0.01
(-0.01) (0.01)
Year dummies Yes Yes
Industry dummies Yes Yes
Adjusted R2 31 .96% 64. 1 2%
Shea's partial R2 10.93% -
F-statistics 25.34*** 95.70***
N 4,726 4,726

*** Significant at the

"Significant at the 0
* Significant at the 0.

This content downloaded from

14.139.235.161 on Mon, 21 Nov 2022 09:28:56 UTC
All use subject to https://about.jstor.org/terms
Jiraporn, et al. • Evidence from Geographic Identification 527

Table A3. The Effects of CSR Components on Cred

Model

CSR Corporate Employee Human

Components Governance Community Divers
Constant 4.73*** 5.31*** 5.19*** 4.87*** 5.09*** 4.83*** 4.84***
(4.62) (5.14) (5.01) (4.79) (4.94) (4.73) (4.73)
Corporate 0.11*
governance (1.65)
Community 0.20***
(3.07)
Diversity 0.08**
(2.05)
Employee 0.21***
relations (4.55)
Environment 0. 1 1 ***
(2.06)
Human rights -0.30***
(-2.15)
Products 0.00
(0.02)

Control Yes Yes Yes Yes Yes Yes Yes

variables
Year dummies Yes Yes Yes Yes Yes Yes Yes
Industry Yes Yes Yes Yes Yes Yes Yes
dummies
Adjusted R2 67.08 67.17% 67.10% 67.32% 67.10% 67.10% 67.94%
N 2,516 2,516 2,516 2,516 2,516 2,516 2,516

*** Significant at the 0.01 level.

** Significant at the 0.05 level.
* Significant at the 0.10 level.

D. CSR Components

The CSR score is constructed based on seven CSR

community, (3) diversity, (4) employee relations,
products. The evidence thus far suggests that the
credit ratings. To gain further insight, we conduct a
of each individual CSR category on credit ratings. T
To avoid collinearity, we include only one CSR com
of the seven components, five of them are positiv
apparent that the positive impact of CSR on credi
two CSR categories. Rather, it is based on a broad
study is on the overall impact of CSR, our investi
nature. Nevertheless, we encourage future research

E. Investment-Grade versus Noninvestment-Grade

Additionally, certain financial institutions are only allowed to hold investment-grade secur
Thus, the difference between investment-grade and noninvestment-grade ratings is much

This content downloaded from

14.139.235.161 on Mon, 21 Nov 2022 09:28:56 UTC
All use subject to https://about.jstor.org/terms
 528

crucial than the differen

dummy variable equal to
above BBB- is usually co
predicting the likelihood
use the CSR score instru
and significant. The res
being classified as invest

References

Atkas, N., E.D. Bodt, and J.G. Cousin, 2011, "Do Financial Markets Care About SRI? Evidenc
Mergers and Acquisitions," Journal of Banking and Finance 35, 1753-1 761 .

Alii, K., G.G. Ramirez, and K. Yung, 1991, "Corporate Headquarters Relocation: Evidence from the C
Markets," AREUEA Journal 19, 583-599.

Ashbaugh-Skaife, H., D. Collins, and R. LaFond, 2006, "The Effects of Corporate Governance on
Credit Ratings," Journal of Accounting and Economics 42, 203-243.

Aupperle, K., A.B. Caroli, and J. Hatfield, 1985, "An Empirical Examination of the Relationship
Corporate Social Responsibility and Profitability," Academy of Management Journal 28, 446-463.

Benson, B.W. and W.N. Davidson, 2010, "The Relation Between Stakeholder Management, Firm Va
CEO Compensation: A Test of Enlightened Value Maximization," Financial Management 39, 9

Becker, B., Z. Ivkovic, and S. Weisbenner, 2011, "Local Dividend Clienteles," Journal of Fina
655-683.

Bernea, A. and A. Rubin, 2010, "Corporate Social Responsibility as a Conflict Between Shareholders,"
Journal of Business Ethics 97, 71-86.

Bhojraj, S. and P. Sengupta, 2003, "Effect of Corporate Governance on Bond Ratings and Yields: The Role
of Institutional Investors and the Outside Directors," Journal of Business 76, 455-475.

Boardman, C. and R. McEnally, 1981, "Factors Affecting Seasoned Corporate Bond Prices," Journal of
Financial and Quantitative Analysis 16, 207-216.

Boutin-Dufřesne, F. and P. Savaria, 2004, "Corporate Social Responsibility and Financial Risk," Journal of
Investing 13, 57-66.

Bouwman, C.H.S., 201 1, "The Geography of Executive Compensation," Case Western Reserve University
and Wharton Financial Institutions Center Working Paper.

Brammer, S. and A. Millington, 2008, "Does It Pay to Be Different? An Analysis of the Relationship
Between Corporate Social and Financial Performance," Strategic Management Journal 29, 1325-1343.

Burke, L., J.M. Logsdon, W. Mitchell, M. Reiner, and D. Vogel, 1986, "Corporate Community Involvement
in the San Francisco Bay Area," California Management Review 28, 122-141.

Carroll, A.B., 1991, "The Pyramid of Corporate Social Responsibility: Toward the Moral Management of
Organizational Stakeholders," Business Horizons 34, 39-48.

Covai, J.D. and T.J. Moskowitz, 1999, "The Geography of Investment: Informed Trading and Asset Prices,"
Journal of Political Economy 109, 81 1-841.

Davis, J. and J.V. Henderson, 2008, "The Agglomeration of Corporate Headquarters," Regional Science and
Urban Economics 63, 431-450.

This content downloaded from

14.139.235.161 on Mon, 21 Nov 2022 09:28:56 UTC
All use subject to https://about.jstor.org/terms
Jiraporn, et al. » Evidence from Geographic Identification

Deng, X., J.K Kang, and B.S. Low, 2013, "Corp

mization: Evidence from Mergers," Journal of

Dhaliwal, D., S. Eheitzman, and O.Z. Li, 2009, "T

of Accounting Research 44, 691-723.

Dorfman, R. and RO. Steiner, 1954, "Optimal A

Review 43, 153-193.

El Ghoul, S., O. Guedhami, C. Kwok, and D. Mis

the Cost of Capital?" Journal of Banking and

Francis, B., I. Hasan, K. John, and M. Waismann,

Institute, New York University, and Fordham U

Friedman, M. and R. Friedman, 1962, Capitalism

Gao, W, L. Ng, and Q. Wang, 201 1, "Does Corpor

Policies?" Financial Management 40, 1 13-138.

Gardberg, N.A. and C. Fombrun, 2006, "Corpor

tional Environment," Academy of Management

Godfrey, P.C., 2005, "The Relationship Between

Management Perspective," Academy of Manag

Godfrey, P.C., C.B. Merrill, and J.M. Hansen

sponsibility and Shareholder Value: An Empiri
Management Journal 30, 425-445.

Goss, A. and G.S. Roberts, 2011, "The Impact o

Loans," Journal of Banking and Finance 35,1 7

Grinblatt, M. and M. Kelohaiju, 2001, "How Dis

Trades," Journal of Finance 56, 1053-1073.

Hansen, G.S. and B. Wernerfelt, 1989, "Determ

Economic and Organizational Factors," Strateg

Horrigan, J., 1966, "The Determinants of Long

Accounting Research 4, 44-62.

Ivkovic, Z. and S. Weisbenner, 2005, "Local Doe

Individual Investors Common Stock Investmen

Jiao, Y., 2010, "Stakeholder Welfare and Firm V

John, K. and, A. Knyazeva, and D. Knyazeva, 2

Precommitment," New York University and Uni

Kang, J.-K. and J.-M. Kim, 2008, "The Geograp

2858.

Kaplan, R. and G. Urwitz, 1979, "Statistical Models of Bond Ratings: A Methodological Inquiry," Journal
of Business 52, 23 1 -26 1 .

Kedia, S., V Panchapagesan, and V Uysal, 2008, "Geography and Acquirer Returns," Journal of Financial
Intermediation 17, 256-275.

Kedia, S. and S. Rajgopal, 2009, "Neighborhood Matters: The Impact of Location on Broad Based Stock
Option Plans," Journal of Financial Economics 92, 109-127.

This content downloaded from

14.139.235.161 on Mon, 21 Nov 2022 09:28:56 UTC
All use subject to https://about.jstor.org/terms
 530

Klock, M., S. Mansi, and W.

of Financial and Quantitativ

Knyazeva, A., D. Knyazeva

University of Rochester and

John, K. and D. Kadyrzhanova

Paper.

Lamy, R. and R. Thompson, 1988, "Risk Premia and the Pricing of Primary Issue Bonds," Journal of
Banking and Finance 12, 585-601.

Lee, D.D. and R.W. Faff, 2009, "Corporate Sustainability Performance and Idiosyncratic Risk: A Global
Perspective," Financial Review 44, 213-237.

Litov, L.P., 2005, " Corporate Governance and Financing Policy : New Evidence Manuscript ," New York,
NY, Stern School of Business, New York University.

Liu, Y. and P. Jiraporn, 2010, "The Effect of CEO Power on Bond Ratings and Yields," Journal of Empirical
Finance 17, 744-762.

Loughran, T. and P. Schultz, 2006, "Asymmetric Information, Firm Location, and Equity Issuance," Uni-
versity of Notre Dame Working Paper.

Massa, M. and S. Simonov, 2006, "Hedging, Familiarity, and Portfolio Choice," Review of Financial Studies
19, 633-685.

McGuire, J., A. Sundgren, and T. Schneeweis, 1988, "Corporate Social Responsibility and Firm Financial
Performance," Academy of Management Journal 3 1 , 854-872.

Mc Williams, A. and D. Siegel, 2001, "Corporate Social Responsibility and Financial Performance: Corre-
lation or Misspecification," Strategic Management Journal 21, 603-609.

Miguel, E., S. Satyanath, and E. Sergenti, 2004, "Economic Shocks and Civil Conflicts: An Instrumental
Variables Approach," Journal of Political Economy 1 12, 725-753.

Milgrom, P. and J. Roberts, 1986, "Price and Advertising Signals of Product Quality," Journal of Political
Economy 94, 796-821.

Navarro, P., 1988, "Why Do Corporations Give to Charity?" Journal of Business 61, 66-75.

Orlitzky, M., F.L. Schmidt, and S.L. Rynes, 2003, "Corporate Social and Financial Performance: A Meta
Analysis," Organizational Studies 24, 403-441.

Peloza, J., 2006, "Using Corporate Social Responsibility as Insurance for Financial Performance: A Meta-
Analysis," Organizational Science 24, 403-441.

Porter, M. and M. Kramer, 2006, "Strategy and Society: The Link Between Competitive Advantage and
Corporate Social Responsibility," Harvard Business Review 84, 76-92.

Pirinsky, C. and Q. Wang, 2010, "Geographic Location and Corporate Finance: A Review," Handbook of
Emerging Issues in Corporate Governance , UK, World Scientific, Imperial College, Press.

Pirinsky, C. and Q. Wang, 2006, "Does Corporate Headquarters Location Matter for Stock Returns?" Journal
of Finance 61, 1991-2015.

Saiia, D.H., A.B. Carroll, and A.K. Buchholtz, 2003, "Philanthropy as Strategy: When Corporate Charity
"Begins at Home," Business and Society 42, 169-201.

Sargan, J.D., 1958, "The Estimation of Economic Relationships Using Instrumental Variables," Economet-
rica 26, 393-415.

This content downloaded from

14.139.235.161 on Mon, 21 Nov 2022 09:28:56 UTC
All use subject to https://about.jstor.org/terms
Jiraporn, et al. • Evidence from Geographic Identification 531
Sen, S. and C.B. Bhattacharya, 2001 , "Does Doing Good Always Lead to Doing Be
to Corporate Social Responsibility," Journal of Marketing Research 38, 225

Sharfman, M.R and C.S. Fernando, 2008, "Environmental Risk Management

Strategic Management Journal 29, 569-592.

Starks, L.T., 2009, "Corporate Governance and Corporate Social Responsibili

About? What Should Investors Care About? The Financial Review 44, 461-4

Teoh, S.H, I. Welch, and C.R Wazzan, 1999, "The Effect of Socially Activist
Financial Markets: Evidence from the South African Boycott," Journal of Bu

Waddock, S.A. and S.A Graves, 1997, "The Corporate Social Performance-Fin
Strategic Management Journal 18, 303-319.

Webb, N.J. and A. Farmer, 1996, "Corporate Goodwill," Annals of Public and
29-50.

Wright, P. and S.P. Ferris, 1997, "Agency Conflict and Corporate Strategy: The Effect of Divestment on
Corporate Value," Strategic Management Journal 18, 77-83.

Ye, K. and R. Zhang, 2011, "Do Lenders Value Corporate Social Responsibility? Evidence from China,"
Journal of Business Ethics 104, 197-206.

Zhang, R., J. Zhu, H. Yue, and C. Zhu, 2010, "Corporate Philanthropic Giving, Advertising Intensity, and
Industry Competition Level," Journal of Business Ethics 91, 51-63.

Zhu, N., 2002, "The Local Bias of Individual Investors," Yale University Working Paper.

Ziebart, D. and S. Reiter, 1992, "Bond Ratings, Bond Yields and Finance Information," Contemporary
Accounting Research 9, 252-282.

This content downloaded from

14.139.235.161 on Mon, 21 Nov 2022 09:28:56 UTC
All use subject to https://about.jstor.org/terms