Political Views and Corporate Decision-Making: The Case of

Corporate Social Responsibility

Amir Rubin∗

Abstract

This paper conducts an empirical analysis of the relationship between corporate social
responsibility (CSR) and political beliefs in the United States. By analyzing the 2004
presidential election results of communities in which corporate headquarters are located,
we establish a correlation between the political beliefs of corporate stakeholders and the
CSR rating of their firms. Companies with a high CSR rating tend to be located in
Democratic, or "blue" states and counties, while companies with a low CSR rating tend to
be located in Republican, or "red" states and counties.

∗
Amir Rubin, Finance Area, Faculty of Business Administration, Simon Fraser University, Burnaby, BC,
Canada. email: [email protected]. I thank Chris Perignon, Jacob Sagi, and Chris Veld for valuable suggestions
and advice. I gratefully acknowledge the research support of the Social Sciences and Humanities Research
Council of Canada.
 Political Views and Corporate Decision-Making: The
Case of Corporate Social Responsibility

September 12, 2005

Abstract

This paper conducts an empirical analysis of the relationship between corporate

social responsibility (CSR) and political beliefs in the United States. By analyzing the
2004 presidential election results of communities in which corporate headquarters are
located, we establish a correlation between the political beliefs of corporate stakeholders
and the CSR rating of their firms. Companies with a high CSR rating tend to be located
in Democratic, or “blue” states and counties, while companies with a low CSR rating
tend to be located in Republican, or “red” states and counties.

One of the most important business trends in the last decade has been the growing desire
of investors to place their money in socially responsible companies that go above and beyond
what is legally required to serve their communities and protect the environment. As a result,
a Corporate Social Responsibility (CSR) rating is now being applied to firms to help people
make their investment decisions. At the same time, there is some evidence that political
beliefs in the United States have become more and more polarized. Stories appear in the
media almost daily regarding culture wars and the chasm between Republican, or “red”
states and Democratic, or “blue” states.1 In this battle of ideas, the values inherent in
socially responsible investing appear to be more closely aligned with the Democratic Party
than with the Republican Party. In fact, Democratic leaders have incorporated many socially
responsible issues into their party platform.
1
See for example, Washington Post, “Faithful Standing More Firm, Poll Says”, January 23, 2005.

1
 It is reasonable to assume that CSR policy, an inherently political phenomenon, is af-
fected by the political beliefs of a firm’s stakeholders. To determine the validity of this
assumption, this paper analyzes the relationship between the CSR ratings of American firms
and political beliefs within the United States. We hypothesize that a firm’s CSR rating is
lower when the firm’s headquarters is located in a county or state where the election results
favor Republican candidates and higher in counties and states where election results favor
Democratic candidates. In particular, we focus our analysis on the relationship between
CSR and the 2004 presidential election. Similar to the electorate, John Kerry, a liberal, and
George Bush, a conservative, possess starkly different views regarding the role of government
and corporations. Therefore, the results from this election should provide especially relevant
data with which to analyze whether or not a correlation exists between the political views
of individual communities and the CSR ratings of firms located within those communities.
We believe a relationship exists for several reasons. First, financial economists show
that the location of investors is important for trading and portfolio decisions. There is
evidence that distance between investors and firms’ headquarter is important. Coval and
Moskowitz (1999, 2001), Grinblatt and Keloharju (2000) and Feng and Seasholes (2004)
find that investors who live near a firm’s headquarters are biased in their portfolio holdings
towards the local company. This bias suggests that one can proxy for shareholders’ political
views by headquarter location election results. Second, corporate executives tend to reside
near a firm’s headquarters; therefore, election results can also serve as reasonable proxies
for the political beliefs of the typical corporate executive. Third, due to the clustering of
other stakeholders such as employees, customers, and suppliers around a firm’s headquarter
location (Porter 1998, 2000), we hypothesize that election results reflect the political views
of other company stakeholders. It seems logical that corporate decision-makers would align
their policies with the views of their stakeholders in order to reduce conflict and create value
for the firm. Finally, even if most stakeholders do not reside near a firm’s headquarters–
and election results do not prove to be good proxies for stakeholders’ political views–we
hypothesize that the community where corporate executives reside will still exert an influence

2
on their political values. If we assume that some degree of social interaction occurs between
corporate executives and the people living in the state and county in which the company is
headquartered, election results should matter. This assumption is supported by Akerlof and
Kranton (2000) and Murphy and Shleifer (2004), who show that identity and social networks
tend to feed on each other.
For the purposes of our study, we conduct an empirical analysis to see whether or not
firms located in counties and states with Republican majorities have lower CSR ratings
than do firms located in counties and states with Democratic majorities. Throughout the
analysis, we are mindful of the relationship that may exist between CSR rating and industry
classification. Since it is conceivable that firms located in states with Republican majorities
belong to industries that are less socially responsible, such as energy, we pay special attention
to industry classifications in order to verify that our results are not driven by such factors.
Controlling for industry and other firm specific variables, we find that firms with high CSR
ratings do tend to be located in states with Democratic majorities, while firms with low
CSR ratings tend to be located in states with Republican majorities. The results of county
elections are similar in nature but tend to be somewhat less significant.
To conduct this study, we perform several robustness checks. For example, we analyze
a variety of sub-samples and different types of elections results. We also use an instrumen-
tal variable (IV) approach to rule out possible problems with the endogeneity of political
preference and CSR. While we do find that other elections that place people on the liberal-
conservative scale tend to be related to CSR rating, we find that the Bush-Kerry election
results are more significant in their ability to explain CSR ratings.
We wish to note that our paper is somewhat related to the literature that shows election
results lead to economic outcomes. For example, see Herbst and Slinkman (1984), Huang
(1985), Hensel and Ziemba (1995), Siegel (1998), Chittenden, Jensen, and Johnson (1999),
and Santa-Clara and Valkanov (2003). However, we do not infer a direct causality between
election results and economic outcome, but rather use election results as proxies for political
preferences that in turn affect corporate decision-making.

3
 The remainder of this paper proceeds as follows. In Section II we present our hypothesis.
In Section III we describe the data and the different variables that we use in our empirical
analysis. In section IV we conduct the empirical analysis. Finally, in Section V we summarize
our findings.

I Political Preference and CSR

The World Business Council for Sustainable Development defines CSR as “the continuing
commitment by business to behave ethically and contribute to economic development while
improving the quality of life of the workforce and their families as well as of the local
community and society at large.”2 The concept includes such issues as capacity building
for sustainable livelihoods, respect for cultural differences, social care for employees, and
giving back to the community. While different firms and organizations have defined the
term in different ways, considerable common ground exists between them.
One principle underlying the CSR movement is that blind pursuit of profit is likely to
prove socially harmful.3 According to the Republican platform, the role of government is
to provide only those critical functions that cannot be performed by individuals or private
organizations.4 Most Republicans believe that good government is based on a system of
limited taxes and limited spending. The view of low tax rates can be regarded as analogous
to the corporate view that profits should be distributed to shareholders and not spent on
social causes. Therefore, it would appear that the principles of CSR are not in alignment
with the values of most Republicans. On the other hand, Democratic Party policies in-
clude many of the precepts embraced by the CSR movement. For example, the Democratic
Party’s platform contains a commitment to support American workers and consumers by
creating internationally recognized labor and environmental standards; it also supports the
advancement of minorities and the right of workers to organize a union.5
2
“Making Good Business Sense” report of the World Business Council for Sustainable Development.
3
The Economist, “A Skeptical Look at Corporate Social Responsibility”, January 22, 2005.
4
Taken from the “Republican Oath” at the Republican National Committee website.
5
Taken from the 2004 Democratic Party Platform.

4
 Thus we hypothesize that corporate policies regarding social responsibility are related
to political preference. In addition, we hypothesize that companies located in communities
with a Republican majority will earn lower CSR ratings than will companies located in
Democratic communities. In particular, we expect to see a relationship between a firm’s
CSR rating and the political preferences of the community where the firm’s headquarters
are located. We focus on the location of a firm’s headquarters because this is where corporate
decision-making takes place, and corporate executives and other stakeholders tend to reside
in the area.

A Political Views of Shareholders

Brennan and Cao (1997) show that capital flow depends on asymetry of information between
foreign and domestic investors. Kang and Stulz (1997) show that foreign investors hold
disproportionately more shares of firms in manufacturing industries, large firms, and firms
with good accounting performance, low unsystematic risk, and low leverage. Coval and
Moskowitz (1999, 2001) and others identify a geographic bias toward home. For example,
they find that shareholders who reside near a firm’s headquarters tend to have a relatively
larger share of ownership compared to shareholders who live far away. In fact, there is some
evidence to support that firms hold their annual meetings away from their headquarters when
they wish to deter a large portion of shareholders from attending. For example, Disney moved
its annual meeting from Burbank, California, to Hartford, Connecticut, after a tumultuous
1997 meeting. Similarly, GM shifted its meetings from Detroit, Michigan, to Wilmington,
Delaware, after an annual meeting that was unfriendly to management.6

B Political Views in the Community

Since executives tend to reside near their firm’s headquarters, the political views of their
community would appear to make a relatively good proxy for their political beliefs. Even
if executives hold somewhat different political views from those held by individuals within
6
Business Week, Jan 7, 2002, page 14.

5
their community, it seems reasonable that social interactions between the firm and the com-
munity would influence executive decisions to some extent. This assumption is supported
by Akerlof and Kranton (2000), who argue that social identity affects people’s behavior and
that individuals tend to conform to their respective social groups.
Murphy and Shleifer (2004) show that conformity also plays a role in the attitudes of
people towards political issues. Indeed, Hong, Kubik and Stein (2004) demonstrate that
social interaction of individuals affects portfolio decisions. This would mean that executives’
beliefs and values should be influenced by their personal social network consisting of families,
friends and colleagues; the media to which they are exposed; and the local civic, educational,
charity and artistic institutions with which they associate. It is highly unlikely that corporate
executives would simply disregard the influence of their community and act in a way that
does not at least to some extent conform to it.

C Political Views of Customers, Workers and Suppliers

The initial corporate decision regarding where to site company headquarters tends to be
based on the need to attract and retain workers with the right combination of skills for the
company’s line of business. Companies also strive to be near their customers and suppliers.
As a result, they move their headquarters infrequently; when such moves do occur, the
impetus is often the desire to be closer to the company’s stakeholders.
Literature on real estate provides a theoretical basis for firms’ tendency to cluster around
customers and a large pool of potential employees. Porter (1998, 2000) and Glasmeier (1998)
show that proximity to consumers is particularly beneficial to firms that depend upon a
rapid differentiation of product to meet consumer demand because it enables them to beat
the competition with new products and a faster reaction time to the market. Such firms also
benefit by being close to a well-educated labor market that understands new technology. All
of this suggests that the location of corporate headquarters correlates with a concentration
of stakeholders. It also explains why companies would exhibit particular sensitivity to the
political preferences of their communities.

6
II Data Sources
In this section, we describe the variables used in our study. For clarity of exposition, we cat-
egorize the data into CSR ratings, political variables, and control variables. We obtained the
information on corporate social responsibility from Kinder, Lydenberg and Domini Research
& Analytics, Inc. (KLD), a leading research firm that rates corporate social performance
for investors. The KLD database screens close to 3,000 firms and categorizes them as either
socially responsible (SR) or socially irresponsible (SI).7 The firms in our sample account for
98 percent of the total market value of U.S. public equities. We gathered most of the politi-
cal variables from either USA Today or CNN. To provide control variables, we used the 13F
Schedule from institutional investors’ holdings, proxy statements, CRSP, and Compustat.
Our database is cross-sectional and consists of the most up-to-date information available at
the time of the 2004 national election (Q3 2003). Table I provides a description of the main
political and control variables used in the study.

[ Insert Table I about here ]

A The CSR Measure

In 2001 KLD launched the Broad Market Social Index (BMSI), a subset of close to 3,000 firms
that compose the Russell 3000 index. To generate this index, KLD submits each firm to a
two-stage screening process that results in three categories: 1) SR, 2) SI due to exclusionary
reasons, and 3) SI due to qualitative reasons. Only SR firms are included in the BMSI. In the
first stage of the screening process, KLD analyzes each company’s sources of revenue. Any
company whose revenues come from alcohol, tobacco, or gambling–or that derives more
than 2 percent of gross revenue from the production of military weapons–receives an SI
rating. Electric utilities that own interests in nuclear power plants or that derive electricity
from nuclear power plants in which they have an interest also receive an SI rating. It is
7
Aggarwal and Nanda (2004) use this data to study the impact of the size of a firm’s boards on managerial
incentives. Barnea and Rubin (2005) use the same data to study whether or not CSR policy is subject to a
conflict between shareholders

7
important to note that the exclusionary screening that KLD applies is a per-se criterion.
Firms that fail in this screening stage will never receive an SR rating unless they shut down
the “unethical” side of their business. For example, as long as Philip Morris continues to
manufacture cigarettes, it will receive an SI rating. Out of the 2,837 firms that KLD has
considered to date, 187 have received the SI rating for exclusionary reasons.
In the second stage of the process, KLD screens the remaining firms for quality of life
issues such as community relationships, workforce diversity, employee relationships, envi-
ronment, non-US operations, and product safety and use. Its goal is to determine a firm’s
strengths and weaknesses in each of these areas. It may find, for example, that a company
has paid a penalty for violating an environmental law. Conversely, it may also find that a
company is strong in employee relationships because its policies are particularly beneficial
to the family. Where possible, KLD uses quantitative criteria to determine the rating (e.g.,
dollar amounts paid in fines or the percentage of employees receiving certain kinds of bene-
fits). Some subjective judgments are necessary, of course, when determining the cutoff point
for a negative rating or deciding how to handle borderline cases. In our sample, 2,278 firms
passed the qualitative social screening and received an SR rating, while 372 firms failed the
qualitative screening and received an SI rating.
The dependent variable throughout our analysis is the CSR rating of each firm. This
binary variable equals “one” if a firm passes the screening conducted by KLD and “zero” if
it fails. Our underlying assumption is that corporate executives and other stakeholders have
an effect on the CSR policy of the firm. This means that we have not included SI due to
exclusionary screening firms in our analysis because they cannot change their CSR rating no
matter what policies they may adapt. After eliminating these firms from our sample, 2650
firms remain.

B Political Variables

We derive our major explanatory variables from the 2004 presidential election. We focus on
this election in particular because it is the first one to be held since the launch of the BMSI.

8
In addition, it is particularly suitable for our study because it represents two dichotomous
viewpoints. President George W. Bush is a conservative Republican, while John Kerry is
a liberal Democrat. In fact, The National Journal, a politically neutral magazine focusing
on policymakers in Washington, ranked Kerry the most liberal senator in 2003. Thus, this
election provided a clearer choice for American voters than did preceding political races when
the differences between candidates’ values were less obvious.
For the purposes of this paper, we focus on the variables that quantify the percentage
of votes that Bush received in the state or county where a firm’s headquarters is located.
To check whether or not our results are valid only with this particular election, we also use
other measures as proxies for the political preferences of voters. In particular, we use political
variables that have been calculated over longer periods of time, including one that sums up
the number of national presidential elections won by a Republican candidate in the state
during the years 1972-2004 and one that captures the degree of ”Republican strength” in the
state according to the Brookings Institute. We also use state senators, state representatives,
and state governors election results to check the importance of using national election results
rather than state-officials election results.

C Control Variables

We include several variables in our analysis to control for industry and firm characteristics.
To capture industry effects, we include sixty-four dummy variables for each two-digit SIC
code. We measure a firm’s size using the natural log of the book value of total assets. As
a proxy for growth opportunities, we use the market-to-book ratio calculated as the market
value of assets divided by the book value of assets. As our proxy for firm risk, we use a
stock’s 60-month return volatility. To measure a firm’s age, we use the number of years
its share price has appeared on the CRSP tape. Finally, we define the firm’s leverage as
long-term debt divided by the total book value of assets.
Barnea and Rubin (2005) show that CSR is related to ownership and capital structure.
Similar to what they have done, we focus on two groups of investors: insiders and insti-

9
tutions. We use one measure for ownership by insiders and one measure for ownership by
institutional investors. We define insiders’ ownership as “the percent of common stock held
by all officers and directors of the company plus beneficial owners who own more than 5
percent of the company’s stock as disclosed in the most recent proxy statement.” For insti-
tutional ownership, we use the Herfindahl-Hirschman Index (HHI) of concentration of the
P
top 15 institutional owners gathered from Schedule 13F filings. This is defined as 15 2
i=1 hi ,

where hi is the percentage ownership of institution i. Shleifer and Vishney (1986) show
that institutions have a greater influence when they are large shareholders, and Black (1992)
shows they have a greater influence because they can form a coalition. Therefore, we use a
measure that illustrates the concentration of institutional ownership.

D Summary Statistics

[ Insert Table II about here ]

Table II summarizes statistics for each state. These include the percentage of the popu-
lation that voted for Bush, the number of firms that went through KLD’s qualitative screen-
ing, and the percentage of SI firms.8 As might be expected, large variations exist among the
states–both in election results and in percentage of SI firms. For example, note the low 9.3
percent vote for Bush and the 0 percent of SI firms in the District of Columbia versus the
high 68.5 percent vote for Bush and the 33 percent of SI firms in Idaho. The percentage of
SI firms is low in most states because they represent only 14 percent of our sample. The
only state with a majority of SI firms is Mississippi. The number of Russell 3000 firms in
each state also varies widely. California has the most public companies by far, followed by
New York and Texas.
Table III Panel A reports the results of a Chi—Square test of independence to check
whether or not the observed number of SI firms is independent of whether the location of
the firms’ headquarter is in either a Bush or a Kerry state. If there were no relationship
between red and blue states and the CSR rating of firms (SI vs. SR), we would expect the
8
For brevity in description, we consider the District of Columbia as a state.

10
ratio of SI firms to the total number of firms for both candidates to be similar at around 14
percent. However, the results clearly show that there are comparatively more SI firms in red
states (17.16 percent) than in blue states (12.4 percent.)
We must be aware, however, that the type of industry can have a strong effect on our
results. For example, it is relatively easy for companies in the high-tech sector to receive
an SR rating because they are not involved in polluting activities, and their workforce is
composed of white-collar employees. On the other hand, it is much harder for companies
in the energy and basic material sectors to achieve an SR rating because they deal with
activities that almost by definition are somewhat harmful to the environment, and their
workforce is composed of blue-collar employees. Therefore, when correlating CSR rating to
election results, we must eliminate any effects stemming from type of industry. In Panel B
of Table III, we conduct a two-dimensional Chi-Square test in which the expected number
of SI firms has been calculated conditional on firms’ sector classification code. The results
show that we cannot reject the null that voting for Bush is independent of classification to
either SI or SR firms. Thus, after taking into account industry effects, it might well be that
there is no relationship between the classification of a state as red or blue and the probability
that a firm will be designated as SI.

[ Insert Table III about here ]

It is important to note a few important points regarding these results. First, even though
the two-dimensional Chi-Square test does not reject the null, it is still interesting that the
classification of a state as either red or blue picks up most of the sector classification effect. A-
priori, one might assume there would be no relationship between the distribution of industries
in a state and the classification of a state as either red or blue. However, these results show
that the types of industry within a state can act as proxies for the Bush-Kerry classification
and vice versa.
Second, it seems that in a multivariate analysis, the election variables would explain much
of the variation. For example, the four largest industries (i.e., energy, financial, services,

11
technology) represent more than half of our sample. Conditional on being classified to these
four industries, companies located in Bush voting states tend to more socially irresponsible
than what would be expected by the independence test. The independence test has low
power when the sectors are unevenly distributed, so it might not be able to detect this
relationship.
Finally, since we have so far only looked at a binary explanatory variable, i.e., Bush
states versus Kerry states, it is hard to know the effect of using the continuous variable of
the percentage of votes cast for Bush. This later variable may capture more of the variation
because it allows us to differentiate between different Bush and Kerry states.
In Table IV we present a difference of means test for important political variables. Since
we know that much of the political difference is related to industry, we also provide an
analysis that is mean-adjusted to the industry, where each observation is adjusted to the
2-digit SIC code industry average. The table provides t-statistics and indicates significance
at the five percent level. It is clear that political variables at the state level are significant
in explaining social responsibility–even after the adjustment for industry average.
The most significant variable is the percentage of people in each state who voted for
Bush, which is significant at the 1 percent level. The long-term political variables, that is
Republican state and Republican state (Brookings), perform almost as well. However, the
county election variables seem to be insignificant once we adjust to industry mean. It seems
that since elections at the county level are frequently tend to vary more from election to
election, they are less able to capture the CSR ratings of firms, which tend to be very stable
over time.

[ Insert Table IV about here ]

12
III Multivariate Analysis
A Multivariate Analysis of CSR and Republican Preference

In this section we investigate the relationship between CSR and political variables. Our
measure of the social performance of firms is CSR, a dummy variable that equals “one” if a
firm passed the qualitative screening conducted by KLD and “zero” if it failed. The model
that we test in this subsection is the following:

CSRi = γ 0 + γ 1 (Degree of Republican)i + γ 2 (Stability)i

+γ 3−8 (Control variables)i + γ 8−71 (T wo − digit SIC code)i + εi (1)

On the right-hand side, we interchangeably use different political variables that measure
the degree of preference for the Republican Party in the region (state or county) of the firm’s
headquarters. The control variables are size, market to book, return volatility, age, leverage,
insiders’ ownership, and institutional HHI. To control for industry, we also use sixty-four
2-digit SIC code dummy variables.

[ Insert Table V about here ]

We present the results with robust standard errors in Table V. The most striking result
in our analysis is that the coefficients that represent Republican preference are negative and
significant in 4 out of 6 specifications. Similar to what was found in the univariate analysis,
the county election results are unable to explain either CSR or state-level variables. However,
while county election results are noisier, they still confirm our hypothesis at the 10 percent
significance level (see specification 4). All other control variables provide similar results to
those shown in Barnea and Rubin (2005), which are that SR firms tend to be smaller in size,
have a smaller percentage of insider ownership, and tend to have lower leverage.
Our results broadly confirm our hypothesis and show that in areas where Republicans
constitute a large portion of the population, firms tend to pursue a less-friendly CSR policy.
Also note that out of all the different specifications, the 2004 national election results are
the most significant in explaining CSR.

13
B Incorporating both State and County Results

One of the limitations in Table V is that we do not incorporate state and county results in
a single regression. This is because the correlation between the two continuous variables is
relatively high (0.56) and results in loss of power due to the increased variance of coefficients.
However, from a theoretical point of view, it seems that both variables should have an effect
on the firm’s decision-making. Firm decision-making should be affected not only by the
general preference of voters within a state, but also by the political attitudes of a specific
community. To eliminate the semi-multicollinearity problem and still preserve the valuable
county preference, we define the following variable:

Ma rginal countyi =
P op. Density Statei
[Bush vote county i − Bush vote state i ] (2)
P op. Density Countyi
The variable Marginal County captures the extra support for Bush that is present within a
county over and above the support found in corresponding election results for the state. P op.
Density State represents the average population per square mile in the state, while P op.
Density County represents the average population per square mile in the county. According
to Vercammen and Murray (1990), rural areas exhibit a stronger communal effect; therefore,
we normalize the differences in the Bush vote by the density ratio. This means that the more
rural the county, the more influence we would expect it to have on a firm’s CSR rating.

[ Insert Table VI about here ]

Table VI reports regression results for four specifications. In all specifications, the
Marginal County is negative and significant at the 10 percent level. It appears that
these specifications have improved the explanatory power because the R-square is some-
what higher. We conclude that both state and county election results show that the more a
community supports the Republican Party, the less likely it is that a firm in that community
will receive an SR rating.

14
C Other Election Results

In the previous subsections, we have concentrated on the results of the 2004 presidential
election. We intentionally focused on these elections because they fit our hypothesis well,
which suggests a difference in CSR attitude between liberals and conservatives. We also want
to see, however, if our results hold when measuring political beliefs in a different way. In
other words, we would like to test whether or not our hypothesis holds when the differences
between Republican and Democratic perspectives are less apparent. For example, Arnold
Schwarzenegger, the Republican governor of California, takes a position normally held by
Democrats when he strongly opposes giving companies permission to drill for oil off the coast
of California.9
In the following, we investigate how well local (state) political choices correlate with CSR
rating. To do so, we use four different variables: 1) the percentage of Republicans in the
Senate as of January 2005; 2) the percentage of Republicans in the House as of January
10
2005; 3) the percentage of the population that voted for the Republican candidate in the
last gubernatorial election;11 and 4) a variable that is a multiplication of the three former
variables.

[ Insert Table VII about here ]

Table VII reports the regression results. In all regressions, the coefficient that is associ-
ated with the political variable is negative; however, it is only significant in two out of four
specifications. House and Senate variables are more consistent with our hypothesis than are
gubernatorial election results. We conclude, therefore, that a relationship does exist between
the results of state elections and CSR policy. However, this relationship is less significant
than that between CSR policy and the 2004 national election results.
9
“Schwarzenegger on abortion, gays, environment”, August 28, 2003, CNN.com.
10
For state and House statistics, see http://www.stateside.com/thefiftystates/governors.shtml
11
This means that the data corresponds to the last election in the state, which could be in any of the years
2001-2004.

15
D Instrumental Variable (IV) Approach

Potentially, our analysis could suffer from a problem of endogeneity. Specifically, it is possible
to argue that voters in the county and state do not influence the CSR rating of a firm, but
that a firm’s CSR rating influences the voters. For example, if the firm engages in harmful
environmental activities, it may favor Republican candidates. Because of the influence of
the firm on the local community, the residents may be persuaded by the firm to prefer
Republican candidates. In general, this would seem like a reasonable occurrence of events
and it is important to rule out this endogeneity problem.
Throughout our multivariate analysis, we focus on two types of variables: those that
quantify a preference for the Republican Party at the state level and those that quantify a
preference for the Republican Party at the Marginal County level. Many studies, such as
those by Kahn (2002), demonstrate a strong relationship between demographics and election
results. As instruments, we use three variables: 1) GDP growth rate, 2) percentage of elderly
population, and 3) percentage of home ownership. Obtaining the demographics for the state
is relatively easy; however, demographics at the county level are unavailable. As a result, we
cannot produce a predicted variable for Marginal County; therefore, we report regression
results with and without this variable.
Table VIII presents the results of our instrumental variable regression analysis. Consis-
tent with earlier results, we find that Republican Party preference is significant and negatively
related to a firm’s CSR rating.

[ Insert Table VIII about here ]

E Robustness Analysis

In this subsection, we conduct an analysis to verify that our results regarding the relationship
between CSR rating and Republican Party preference hold when the sample is partitioned in
different ways. As stated above, it is apparent that SR and SI ratings are related to industry.
Almost by definition, some industries have a workforce that is well-paid and produce products

16
or services with limited (or no) negative effects on the environment. These two aspects are
probably the most important attributes affecting a firm’s CSR rating.
Throughout the multivariate analysis, we have so far classified firms according to the
2-digit SIC code, which includes 64 industries. To check the robustness of our results, we
run our main regressions with other types of industry classifications. Our classifications are
of a different kind and range from the broad Sector classification comprised of 12 sectors to
the very fine 4-digit NAICS code for 245 industries.

[ Insert Table IX about here ]

Table IX provides regression results for different industry classifications. The variables
Bush state vote and Marginal County are negatively correlated with CSR ratings in all 6
specifications. Bush state vote is very significant under all of the specifications. Marginal
county is significant in 3 out of 6 specifications. These results confirm that our findings are
not related to the type of industry classification applied.
Our analysis shows that the size of a firm and its ownership structure are the most
important factors in explaining CSR. These results are similar to those reported by Barnea
and Rubin (2005). Table X shows the regression results of the sub-sample regressions.
Specification 1 includes only small firms that are not part of the S&P 1500 index. Sub-sample
2 includes only the firms that are part of the S&P 1500 index. While our political variables
correlate negatively to CSR rating in both regressions, the results are more significant for the
S&P 1500 firms. Similarly, when we partition the sample of firms to those with small insider
holdings and those with large insider holdings, we find that the results are more significant
for the small insider holdings. Thus, a political effect is present in all firms, but it is much
more significant in large firms.

[ Insert Table IX about here ]

17
IV Conclusion
We take advantage of the results of the recent U.S. presidential elections to study whether or
not political beliefs play a role in corporate social responsibility. We find that the location of
a firm’s headquarters is a significant factor in explaining its CSR rating. There is a negative
correlation between the percentage of votes that President Bush received in the state or
county where a firm’s headquarters is located and the firm’s CSR ratings. The results hold
after controlling for industry, size and other firm characteristics. They also persist under
different specifications and robustness checks.
From the perspective of financial economics, we show that political views play a role in
corporate decision-making. We cannot discount the possibility that even though political
views affect corporate decisions, they represent an important part of the drive to maximize
firm value. It may well be that a firm’s CSR rating is driven by its stakeholders’ political
preferences and that incorporating their views into the firm’s policies is optimal.

Bibliography
Aggarwal, Rajesh K. and Dhananjay Nanda 2004, Access, common agency, and board size,
working paper, The University of Virginia.
Akerlof, George and Rachel Kranton, 2000, Economics and identity, Quarterly Journal of
Economics 115, 715-773.
Barclay-McLaughlin, Gina, 2000, Communal isolation: Narrowing the pathways to goal at-
tainment and work, in Coping with poverty: The social contexts of neighborhood, work,
and family in the African-American community, Ann Arbor: University of Michigan
Press, 52-76.
Barnea, Amir and Amir Rubin, 2005, Corporate social responsibility as a conflict between
shareholders, working paper.
Brennan, Michael, and Henry Cao, 1997, International portfolio investment flows, Journal
of Finance 52, 1851-1880.

18
Chittenden, William, Gerald R. Jensen, and Robert R. Johnson, 1999, Presidential politics,
stocks, bonds, bills, and inflation, Journal of Portfolio Management, 63-79.
Coval, Joshua, and Tobias Moskowitz, 1999, Home bias at home: Local equity preferences
in domestic portfolios, Journal of Finance 54, 2045-2073.
Coval, Joshua, and Tobias Moskowitz, 2001, The geography of investment: Informed trading
and asset prices, Journal of Political Economy 109, 811-841.
Feng, Lei and Mark S. Seasholes, 2004, Correlated trading and location, Journal of Finance
59, 2117-2143.
Friedman, Milton, 1972, An economist’s protest: Columns in political economy. Glen Ridge:
Thomas Horton.
Hensel, Chris. R, and William T. Ziemba, 1995, United States investment returns during
Democratic and Republican administrations, 1928-1993, Financial Analysts Journal,
61-69.
Herbst, Anthony, and Craig. W. Slinkman, 1984, Political economic cycles in the U.S. stock
market, Financial Analysts Journal, 35-39.
Hong Harrison, Jeffrey D. Kubik, and Jeremy C. Stein, 2004, Social interaction and stock-
market participation, Journal of Finance 59, 137-163.
Huang, Roger, 1985, Common stock returns and presidential elections, Financial Analysts
Journal, 24-39.
Glasmeier, Amy, 1988, Factors governing the development of high tech industry agglomer-
ations: A tale of three cities. Regional Studies 22, 287-301.
Grinblatt, Mark and Matti Keloharju, 2000, The investment behavior and performance
of various investor types: A study of Finland’s unique data set, Journal of Financial
Economics 25, 43-67.
Kang, Jun-Koo, and René Stulz, 1997, Why is there a home bias? An analysis of foreign
portfolio equity ownership in Japan, Journal of Financial Economics 46, 3-28.
Murphy, Kevin and Andrei Shleifer, 2004, Persuasion in politics, NBER working paper
series 10248.

19
Porter, M., 1998, Location, clusters, and the new ‘micro’ economics of competition, Business
Economics 33, 7-13.
Porter, M., 2000, Location, competition, and economic development: Local clusters in the
global economy, Economic Development Quarterly 14, 15-34.
Santa-Clara, Pedro and Rossen Valkanov, 2003, The presidential puzzle: Political cycles
and the stock market, Journal of Finance 58, 1841-1872.
Siegel, Jeremy, 1998, Stocks For the Long Run (Norton, New York).
Vercammen, James and Murray Fulton, 1990, The economic implications of farm interest
groups’ beliefs; American Journal of Agricultural Economics, November 72, 851-63.

20
 Table I
Definition and Source of Major Variables

Description Source
Election results variables
Bush state Dummy variable that equals 1 if the firm’s headquarter is USA Today
located in a state where Bush won.
Bush county Dummy variable that equals 1 if the firm’s headquarter is USA Today
located in a county where Bush won the popular vote.
Bush vote state (%) The percentage of votes that Bush received in the state USA Today
where the firm’s headquarter is located.
Bush vote county (%) The percentage of votes that Bush received in the county USA Today
where the firm’s headquarter is located.
Republican state The number of federal elections won by a republican US census
candidate, in the state where the firm’s headquarter is
located, during the years 1972-2004.
Republican state (Brookings) The degree of “republican strength” in the state where the Brookings Inst.*
firm’s headquarter is located according to the Brookings
Institute.
State governor The percentage of votes that the republican candidate won cnn.com
in the last gubernatorial election (years 2001-2004) in the
state where the firm’s headquarter is located.
State senate The percentage of republicans in the senate of the state cnn.com
where the firm’s headquarter is located. (Jan 2004)
State house The percentage of republicans in the house of cnn.com
representatives of the state where the firm’s headquarter is
located. (Jan 2004)
State Republican The product of State governor, State senate, and State cnn.com
house.
Control variables
Size Natural log of book value of total assets (data item #6) Compustat
Market to book The ratio of the market value of assets (book value of Compustat
assets (data item #6) plus the difference between the
market value of equity (data item #24 ×data item #25) and
the book value of equity (data item #60)) to the book value
of assets (data item #6).
Return volatility The standard deviation of share returns during the previous CRSP
60 months.
Firm’s age The year in which the firm’s share price (data item PRC) CRSP
first appeared on CRSP.
Leverage The book value of long term debt (data item #9) divided by Compustat
the book value of assets (data item #6)
Insiders’ ownership Percent of common stock held by all the officers and Proxy statements
directors of the company plus beneficial owners who own
more than 5 percent of the subject company's stock as
disclosed in the most recent proxy statement.
Institutional HHI The Herfindahl-Hirschman Index (HHI) of concentration 13F Schedule
15
of the top 15 institutional owners. It is defined as ∑ hi2 ,
i =1

where hi is the percentage ownership of institution i.

*
See http://www.thegreenpapers.com/G04/President-Strength.phtml
21
 Table II
The Distribution of SR (Socially Responsible) and SI (Socially Irresponsible) Firms by State

Percent Bush is the percentage of the population that voted for Bush in the respective state. Number of Firms is the
number of public firms that go through the qualitative screening of KLD. Percent of SI Firms is SI Firms divided by
the Number of Firms in the respective state.

State Percent Number Percent State Percent Number Percent

Bush of Firms SI Bush of firms SI
Alabama 62.5 24 16.67 Montana 59.2 3 0
Alaska 61.9 2 0 Nebraska 66.6 11 18.18
Arizona 55.2 27 14.81 Nevada 50.5 4 0
Arkansas 54.3 14 28.57 New Hampshire 49 9 0
California 44.3 491 8.55 New Jersey 46.5 103 16.50
Colorado 52.4 55 9.09 New Mexico 50.2 3 0
Connecticut 44 63 23.81 New York 40.5 215 14.42
Delaware 45.8 9 33.33 North Carolina 56.1 48 10.42
District of Columbia 9.3 13 0 North Dakota 62.9 3 33.33
Florida 52.2 99 17.17 Ohio 51 99 20.2
Georgia 58.6 79 10.13 Oklahoma 65.6 19 21.05
Hawaii 45.3 6 16.67 Oregon 47.5 31 16.13
Idaho 68.5 6 33.33 Pennsylvania 48.6 145 14.48
Illinois 44.7 121 15.70 Rhode Island 38.9 4 0
Indiana 60.1 40 15.00 South Carolina 58 14 0
Iowa 50.1 14 14.29 South Dakota 59.9 4 25
Kansas 62.2 11 18.18 Tennessee 56.8 47 17.02
Kentucky 59.6 20 25.00 Texas 61.2 213 19.72
Louisiana 56.8 23 8.70 Utah 71 14 21.43
Maine 45 4 25.00 Vermont 38.9 5 0
Maryland 43.2 47 8.51 Virginia 54 56 17.86
Massachusetts 37 130 8.46 Washington 45.7 52 11.54
Michigan 47.8 56 10.71 West Virginia 56.1 3 0
Minnesota 47.6 69 7.25 Wisconsin 49.4 38 18.42
Mississippi 59.7 10 60.00 Wyoming 69 0 -
Missouri 53.4 55 20.00

Total all states 51.02 2632 14.04

22
 Table III
Chi-Square Test

Panel A shows the result of a Chi–Square test of independence to check whether the observed number of SI firms in
Bush (Kerry) states is independent of the firms’ headquarters location. Expected number of SI firms is the number of
SI firms expected to be observed in Bush (Kerry) winning states given the overall probability of being a SI firm.
Observed SI firms are the actual number of observed SI firm in Bush (Kerry) winning states. P- Value shows the
degree of significance in rejecting the null of independence. Panel B show the results of a Chi-Square test when the
expected number of SI firms are adjusted to their corresponding industry. SI Probability is the probability of a firm to
be defined as SI given that it is part of the respective industry. Bush is the number of firms in that industry located in
Bush winning states. Expected SI Bush is the expected number of SI firms in Bush states given the probability of a
firm in the industry to be defined as SI. SI Bush is the observed number of SI firms in bush winning states in the
respective industry. P –Value shows the degree of significance in rejecting the null of independence.

Panel A
Number Percentage of total Expected number of Observed SI firms Chi-Square P Value
of firms firms SI firms (percentage)
Bush 1020 0.387 142.66 175 (17.16%) 7.33
0.001
Kerry 1611 0.612 225.33 194 (12.04%) 4.36

Table III - continued

Panel B
SI Expected SI Bush Chi –
Industry SI SR Total Probability Bush SI Bush (percentage) Square

Basic Materials 54 88 142 0.38 60 22.82 24 (40.0%) 0.06

Capital Goods 16 110 126 0.13 66 8.38 5 (7.6%) 1.36
Conglomerates 0 9 9 0 3 0.00 0 (33.3%) 0.00
Consumer Cyclical 13 109 122 0.11 37 3.94 2 (5.4%) 0.96
Consumer/Non-Cyclical 27 78 105 0.26 96 24.69 21 (21.9%) 0.55
Energy 68 390 458 0.15 181 26.87 29 (16.0%) 0.17
Financial 43 272 315 0.14 88 12.01 13 (14.8%) 0.08
Healthcare 11 73 84 0.13 36 4.71 7 (19.4%) 1.11
Services 88 534 622 0.14 271 38.34 52 (19.2%) 4.87
Technology 25 506 531 0.05 114 5.37 8 (7.0%) 1.29
Transportation 10 43 53 0.19 38 7.17 7 (18.4%) 0.00
Utilities 13 51 64 0.20 30 6.09 7 (23.3%) 0.13

Total 368 2263 1020 10.53

P Value 0.48

23
 Table IV
Difference of Means Tests

The dependent variable is CSR, a dummy variable which equals one if a firm passes the screening
conducted by KLD and zero if it fails. Bush state is a dummy variable that equals 1 if the firm’s
headquarters are located in a state where Bush won. Bush County is a dummy variable that equals 1
if the firm’s headquarters are located in a county where Bush won the most votes. Bush vote state is
the percentage of votes that Bush received in the state of the firm’s headquarters. Bush vote county
is the percentage of votes that Bush received in the county of the firm’s headquarters. Republican
state is the number of federal elections won since 1972 (post Vietnam era) by the republican
candidate in the state where the firm’s headquarter is located. Republican state (Brookings) is a
number that reflects the political view of the state according to the Brookings Institution criteria,
where 1 is strongly democratic and 5 is strongly republican. The table provides t-statistics and
indicates significance at the five percent (*) level.

N SR SI Firms t - statistics
Firms
Number of firms 2632 2278 372

Election variables
Bush state 2632 37.40 47.68 -3.76*
Bush county 2632 36.51 37.87 -0.50
Bush vote state 2632 48.77 50.60 -4.17*
Bush vote county 2628 45.22 46.51 -1.82
Republican state 2632 5.24 5.67 -3.80*
Republican state (Brookings) 2632 2.70 2.97 -3.90*

Election variables mean adjusted to 2 digit SIC

code industry average
Bush state 2632 -0.64 3.96 -1.80
Bush county 2632 0.07 -0.48 0.21
Bush vote state 2632 -0.15 0.95 -2.69*
Bush vote county 2628 -0.13 0.78 -1.32
Republican state 2632 -0.04 0.22 -2.39*
Republican state (Brookings) 2632 -0.02 0.12 -2.08*

24
 Table V
The Relation between CSR and 2004 Federal Election Results - Probit Regressions

The dependent variable is CSR, a dummy variable which equals one if a firm passes the screening
conducted by KLD and zero if it fails. Bush state is a dummy variable that equals 1 if the firm’s
headquarters are located in a state where Bush won. Bush County is a dummy variable that equals 1
if the firm’s headquarters are located in a county where Bush won the most votes. Bush vote state is
the percentage of votes that Bush received in the state of the firm’s headquarters. Bush vote county
is the percentage of votes that Bush received in the county of the firm’s headquarters. Republican
state is the number of federal elections won since 1972 (post Vietnam era) by the republican
candidate in the state where the firm’s headquarter is located. Republican state (Brookings) is a
number that reflects the political view of the state according to the Brookings Institution criteria,
where 1 is strongly democratic and 5 is strongly republican. The Control variables are defined in
Table I. All specifications include 2-digit SIC code indicators. The table provides z-statistics
calculated with robust standard errors.

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

Intercept 2.47 2.36 2.66 2.60 2.66 2.61
(3.31) (3.16) (3.29) (3.41) (3.51) (3.44)

Bush state -0.1535

(-1.59)
Bush county 0.0020
(0.03)
Bush vote state -0.0128
(-2.85)
Bush vote county -0.4572
(-1.68)
Republican state -0.0378
(-2.18)
Republican state (Brookings) -0.0499
(-1.82)

Control variables
Size -0.2115 -0.2095 -0.2138 -0.2127 -0.2118 -0.2112
(-8.10) (-7.99) (-8.21) (-8.13) (-8.13) (-8.11)
Market to book 0.0500 0.0524 0.0501 0.0531 0.508 0.0511
(1.29) (1.35) (1.31) (1.38) (1.32) (1.33)
Return volatility 0.2210 0.2624 0.2087 0.2327 0.2581 0.2294
(0.41) (0.50) (0.39) (0.44) (0.48) (0.43)
Firm’s age -0.0010 -0.0010 -.0011 -0.0011 -0.0011 -0.0011
(-0.40) (-0.41) (-0.43) (-0.43) (-0.46) (-0.45)
Leverage -0.5758 -0.5850 -0.5746 -0.5661 -0.5769 -0.5759
(-3.00) (-3.05) (-2.99) (-2.94) (-3.00) (-3.00)
Insiders’ ownership -0.0102 -0.0102 -0.0101 -0.0101 -0.0101 -0.0101
(-6.04) (-6.08) (-6.01) (-6.00) (-6.00) (-6.01)
Institutional HHI 0.0684 0.0906 0.0436 0.0789 0.0629 0.0401
(0.08) (0.10) (0.05) (0.09) (0.07) (0.04)

N 2530 2526 2530 2526 2530 2530

“Pseudo R2” 0.145 0.145 0.148 0.146 0.146 0.146

25
 Table VI
The Relation between CSR and 2004 Federal Election Results - Joint State and County
Regressions
The dependent variable is CSR, a dummy variable which equals one if a firm passes the screening
conducted by KLD and zero if it fails. Bush vote state is the percentage of votes that Bush received
in the state of the firm’s headquarters. Republican state is the number of federal elections won since
1972 (post Vietnam era) by the republican candidate in the state where the firm’s headquarter is
located. Republican state (Brookings) is a number that reflects the political view of the state
according to the Brookings Institution criteria, where 1 is strongly democratic and 5 is strongly
republican. Marginal county is the difference (Bush vote county -Bush vote state) times the
population density of the state divided by the population density of the county. The Control
variables are defined in Table I. All specifications include 2-digit SIC code indicators. The table
provides z-statistics calculated with robust standard errors.

(1) (2) (3)

Intercept 2.64 2.66
(3.26) (3.49)

Bush vote state -0.0125

(-2.80)
Republican state -0.0370
(-2.12)
Republican state (Brookings) -0.0488
(-1.78)
Marginal county -0.0064 -0.0065 -0.0066
(-1.80) (-1.84) (-1.88)
Control variables
Size -0.2112 -0.2092 -0.2086
(-8.09) (-8.01) (-7.98)
Market to book 0.0488 0.0495 0.0498
(1.28) (1.29) (1.29)
Return volatility 0.2167 0.2662 0.2376
(0.41) (0.50) (0.45)
Firm’s age -0.0014 -0.0015 -0.0014
(-0.56) (-0.59) (-0.58)
Leverage -0.5793 -0.5820 -0.5817
(-3.01) (-3.02) (-3.02)
Insiders’ ownership -0.0103 -0.0102 -0.0103
(-6.08) (-6.06) (-6.08)
Institutional HHI 0.0440 0.0634 0.0412
(0.05) (0.07) (0.04)

N 2526 2526 2526

“Pseudo R2” 0.150 0.148 0.147

26
 Table VII
The Relation between CSR and State Election Results

The dependent variable is CSR, a dummy variable which equals one if a firm passes the screening
conducted by KLD and zero if it fails. State governor is the percentage of votes that the republican
candidate won in the last gubernatorial election (years 2001-2004) in the state where the firm’s
headquarter is located. State senate is the percentage of republicans in the senate of the state where
the firm’s headquarter is located (Jan 2004). State house is the percentage of republicans in the
house of representatives of the state where the firm’s headquarter is located (Jan 2004). State
republican is the product of State governor, State senate, and State house. The Control variables are
defined in Table I. All specifications include 2-digit SIC code indicators. The table provides z-
statistics calculated with robust standard errors.

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

Intercept 2.11 2.55 2.64 2.13
(2.49) (3.42) (3.51) (2.71)

State governor -0.0030

(-0.47)
State senate -0.0041
(-1.55)
State house -0.0051
(-1.86)
State republican -0.0106
(-1.99)
Control variables
Size -0.2110 -0.2103 -0.2130 -0.2125
(-8.07) (-8.07) (-8.21) (-8.16)
Market to book 0.0497 0.0516 0.0504 0.0498
(1.29) (1.35) (1.32) (1.30)
Return volatility 0.3159 0.2004 0.2107 0.2050
(0.59) (0.37) (0.39) (0.38)
Firm’s age -0.0010 -0.0010 -0.0010 -0.0010
(-0.42) (-0.40) (-0.41) (-0.42)
Leverage -0.5865 -0.5722 -0.5687 -0.5694
(-3.06) (-2.98) (-2.97) (-2.98)
Insiders’ ownership -0.0102 -0.0103 -0.0104 -0.0104
(-6.03) (-6.08) (-6.16) (-6.12)
Institutional HHI 0.1484 0.1325 0.1258 0.1600
(0.16) (0.14) (0.14) (0.18)

N 2517 2506 2506 2506

“Pseudo R2” 0.145 0.145 0.146 0.146

27
 Table VIII
The Relation between CSR and Election Results – IV Regression

The dependent variable is CSR, a dummy variable which equals one if a firm passes the screening
conducted by KLD and zero if it fails. Instrumental-variable, two-stage probit regressions of CSR,
where GDP growth rate from 1980-2003, percentage of state population with home ownership
(2003), and percentage of state population above 55 (2003), are used as instruments for all Predicted
variables of the election results. Bush state is a dummy variable that equals 1 if the firm’s
headquarters are located in a state where Bush won. Bush vote state is the percentage of votes that
Bush received in the state of the firm’s headquarters. Marginal county is the difference (Bush vote
county -Bush vote state) times the population density of the state divided by the population density
of the county. The Control variables are defined in Table I. All specifications include 2-digit SIC
code indicators. The table provides z-statistics calculated with robust standard errors.

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

Intercept 2.54 2.91 2.05 2.87
(3.36) (3.57) (2.62) (3.51)

Predicted Bush state -0.1900 -0.1807

(-1.87) (-1.78)
Predicted Bush vote state -0.0092 -0.0062 -0.0086
(-1.68) (-1.81) (-1.57)
Marginal county -0.0062
(-1.80)
Control variables
Size -0.2134 -0.2128 -0.2107 -0.2102
(-8.20) (-8.16) (-8.07) (-8.03)
Market to book 0.0498 0.0501 0.0488 0.0490
(1.29) (1.30) (1.26) (1.27)
Return volatility 0.1984 0.2157 0.2091 0.2261
(0.37) (0.41) (0.39) (0.43)
Firm’s age -0.0009 -0.0009 -0.0012 -0.0012
(-0.36) (-0.37) (-0.48) (-0.49)
Leverage -0.5768 -0.5807 -0.5820 -0.5856
(-3.01) (-3.02) (-3.03) (-3.05)
Insiders’ ownership -0.0102 -0.0102 -0.0104 -0.0104
(-6.07) (-6.08) (-6.13) (-6.14)
Institutional HHI 0.0519 0.0652 0.0536 0.0665
(0.06) (0.07) (0.06) (0.07)

N 2530 2530 2526 2526

“Pseudo R2” 0.146 0.146 0.148 0.147

28
 Table IX
Industry Robustness

The dependent variable is CSR, a dummy variable which equals one if a firm passes the screening
conducted by KLD and zero if it fails. Classification type is the industry classification type. Number
of industries is the number of industries in the respective classification. Bush vote state (%) is the
percentage of votes that Bush received in the state of the firm’s headquarters. Marginal county is
the difference (Bush vote county -Bush vote state) times the population density of the state divided
by the population density of the county. The Control variables are defined in Table I. The table
provides z-statistics calculated with robust standard errors.

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

Industry 2 digit 3 digit 2 digit 3 digit 4 digit
Classification type Sector SIC SIC NAICS NAICS NAICS

Number of industries 12 64 240 27 94 245

Intercept 3.55 2.64 2.97 3.43 2.74 2.71

(8.98) (3.26) (3.55) (4.17) (2.86) (2.78)

Bush vote state -0.0117 -0.0125 -0.0102 -0.0119 -0.0111 -0.0091

(-2.63) (-2.80) (-2.04) (-2.71) (-2.40) (-1.83)
Marginal county -0.0040 -0.0064 -0.0058 -0.0060 -0.0070 -0.0044
(-1.05) (-1.80) (-1.50) (-1.65) (-1.89) (-1.26)
Control variables
Size -0.2123 -0.2112 -0.2520 -0.2051 -0.2110 -0.2227
(-8.58) (-8.09) (-8.46) (-8.56) (-7.90) (-7.55)
Market to book 0.0140 0.0488 0.0102 0.0364 0.0477 -0.0092
(0.42) (1.28) (0.25) (1.00) (1.24) (-0.23)
Return volatility -0.5107 0.2167 -0.6625 0.1579 0.1399 -0.8769
(-1.06) (0.41) (-1.13) (0.32) (0.25) (-1.43)
Firm’s age -0.0023 -0.0014 0.0012 -0.0028 -0.0010 -0.0006
(-0.99) (-0.56) (0.41) (-1.15) (-0.40) (-0.22)
Leverage -0.2119 -0.5793 -0.5989 -0.3411 -0.6199 -0.6348
(-1.14) (-3.01) (-2.81) (-2.02) (-3.22) (-3.05)
Insiders’ ownership -0.0099 -0.0103 -0.0107 -0.0010 -0.0111 -0.0114
(-5.99) (-6.08) (-5.67) (-6.11) (-6.48) (-6.12)
Institutional HHI -0.5333 0.0440 -0.4086 -0.0265 0.0145 0.3163
(-0.55) (0.05) (-0.40) (-0.03) (0.02) (0.32)

N 2555 2526 2104 2553 2485 2063

“Pseudo R2” 0.119 0.150 0.191 0.119 0.157 0.191

29
 Table X
Sub-sample Regressions

The dependent variable is CSR, a dummy variable which equals one if a firm passes the screening
conducted by KLD and zero if it fails. Sub-sample type is the defining characteristic of the sub-
sample. Small (S&P 1500) firms are those firms that are not part of (part of) the S&P 1500 firms,
Low (High) holdings are those firms whose insiders’ holdings is less (more) than the median firm
insiders holdings. Bush vote state is the percentage of votes that Bush received in the state of the
firm’s headquarters. Marginal county is the difference (Bush vote county -Bush vote state) times the
population density of the state divided by the population density of the county. The Control variables
are defined in Table I. All specifications include 2-digit SIC code indicators. The table provides z-
statistics calculated with robust standard errors.

Size Insiders
(1) (2) (3) (4)
Sub-sample type Small firms S&P 1500 Low holdings High holdings

Intercept 4.20 1.90 1.98 3.6492

(4.32) (1.82) (1.93) (4.53)

Bush vote state -0.0244 -0.0101 -0.0115 -0.0158

(-2.70) (-1.60) (-1.77) (-2.40)
Marginal county -0.0002 -0.0182 -0.0083 -0.0066
(-0.05) (-2.33) (-1.87) (-1.03)
Control variables
Size -0.3325 -0.1491 -0.2039 -0.2551
(-6.28) (-3.79) (-5.48) (-6.17)
Market to book -0.0694 0.1882 0.0343 0.0699
(-1.79) (2.33) (0.50) (1.37)
Return volatility -0.3388 2.1631 0.5182 -0.3531
(-0.54) (1.84) (0.61) (-0.51)
Firm’s age -0.0008 -0.0033 -0.0012 -0.0018
(0.15) (-1.05) (-0.35) (-0.39)
Leverage -0.4320 -0.5237 -0.6384 -0.4699
(-1.77) (-1.31) (-2.08) (-1.89)
Insiders’ ownership -0.0092 -0.0090 -0.0169 -0.0119
(-4.31) (-2.35) (-0.99) (-4.83)
Institutional HHI 0.1687 2.1669 1.1155 0.3433
(0.18) (0.62) (0.30) (0.36)

N 1208 1183 1210 1204

“Pseudo R2” 0.173 0.162 0.163 0.175

30