Business

ESSAYS IN EMPIRICAL ANALYSIS OF CORPORATE STRATEGY
AND CORPORATE RESPONSIBILITY
Jon Jungbien Moon
A DISSERTATION
in
Business and Public Policy
For the Graduate Group in Managerial Science and Applied Economics
Presented to the Faculties o f the University o f Pennsylvania
in Partial Fulfillment for the Requirements for the Degree of
Doctor o f Philosophy
2007
Supervisor of Dissertation
R(?rWx.—
Graduate Group Chairperson
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
UMI Number: 3271854
Copyright 2007 by
Moon, Jon Jungbien
All rights reserved.
INFORMATION TO USERS
The quality of this reproduction is dependent upon the quality of the copy
submitted. Broken or indistinct print, colored or poor quality illustrations and
photographs, print bleed-through, substandard margins, and improper
alignment can adversely affect reproduction.
In the unlikely event that the author did not send a complete manuscript
and there are missing pages, these will be noted. Also, if unauthorized
copyright material had to be removed, a note will indicate the deletion.
UMI
UMI Microform 3271854
Copyright 2007 by ProQuest Information and Learning Company.
All rights reserved. This microform edition is protected against
unauthorized copying under Title 17, United States Code.
ProQuest Information and Learning Company

300 North Zeeb Road
P.O. Box 1346
Ann Arbor, Ml 48106-1346
COPYRIGHT
Jon Jungbien Moon
2007
DEDICATION
To my maternal grandmother, Ms. Chung, Jung-Lim
ACKNOWLEDGEMENTS
My advisors, Elizabeth E. Bailey and Joel Waldfogel, have been my guides
through the doctoral program. I am very much grateful for their guidance over the course
of my nascent academic career. It has been a great fortune to study under their tutelage.
I am also indebted to Harbir Singh for many thoughtful comments, stimulating
discussions, and continued guidance. I am thankful to Thomas Dunfee, Justin Wolfers,
Matthew Kotchen, participants of 6th Annual Strategy and the Business Environment
Conference, and seminar participants at Wharton Applied Economics Workshop,
Shanghai Jiao Tong University, UC Santa Barbara, Korea Development Institute, and
Shanghai University of Finance and Economics for their insightful comments.
Over the years at the Wharton School, many other professors have given me great
support for my research. I learned a lot from them, and I would like to give my greatest
gratitude to them (in no particular order): Janet Pack, Dennis Yao, Sidney Winter, Gerald
Faulhaber, Paul Kleindorfer, Felix Oberholzer-Gee, Bob Inman, Howard Pack, Yuichi
Kitamura, Rafael Robb, Heather Berry, Jonathan Stroud, and Matthew White. I also
learned a lot from working for Bruce Allen, Keith Weigelt, Neil Doherty, and David
Crawford in Wharton’s managerial economics course for MBA’s and Executive MBA’s.
I would also like give my special thanks to Julia Kang and the staff of the Wharton
Business and Public Policy Department for their support during my years in the
iv
department. Bryan Elliott, in particular, gave me much help in gathering the data for my
second essay.
Financial support from the Zell-Lurie Real Estate Center, Zicklin Center for
Business Ethics Research, and Hower Chair Research Grant are gratefully acknowledged.
In the course of writing this dissertation, I had many helpful discussions with my
friends and fellow doctoral students. First of all, I would like to thank my best friend
Jayoung Yoon, who happens to be my wife. I would also like to thank (again, in no
particular order): Haitao Yin (U Michigan), Kevin Kyungchul Song (Penn), Karl Russo
(Congressional Joint Committee on Taxation), Sangyoung Song (Baruch College,
CUNY), Brian Wu (Ross School, U Michigan), Jay Hwa Hong (U Rochester), Leslie
Schaffer (PWC), Taeyoung Doh (FRB Kansas City), Yong Jin Kim (USC), Brett
Danaher, Carol Sojung Park, Walter Theseira, Jihae Wee, Leandra DeSilva, Hong Chong
Cho, Erica Johnson, Jingoo Kang, Ben Shiller, Jaewon Kim, Adam Isen, and Hailey
Hayeon Joo.
Finally, I would like to thank my parents Dr. Ok Ryun Moon and Dr. Uhn Cho,
and my brother Euibien, for their moral support, and I would like to thank my son Justin
for always brightening my day.
ABSTRACT
ESSAYS IN EMPIRICAL ANALYSIS OF CORPORATE STRATEGY AND
CORPORATE RESPONSIBILITY
Jon Jungbien Moon
Elizabeth E. Bailey and Joel Waldfogel
In this dissertation, I investigate three issues in the area of corporate strategy and
corporate social responsibility.
In my first essay, I investigate how corporate social performance (CSP) affects
corporate financial performance (CFP), using a panel of the KLD Social Ratings
Database. I find that the positive association between CSP and CFP is mainly due to the
unobserved heterogeneity specific to firms. I try to control for endogeneity issue by using
the enactment of Sarbanes-Oxley Act as an instrumental variable. I argue that spending
firm resources to increase CSP affects CFP negatively in the short term. I also find that
the positive side of CSP and the negative side of CSP have asymmetric association with
CFP.
vi
In the second essay that I jointly wrote with Elizabeth E. Bailey, we investigate
how the integration of market and non-market strategies via specialized Board-level
committees is associated with CSP. We analyze companies who have engaged their
boards of directors in integrated strategy, by establishing voluntary committees to handle
public affairs or social responsibility at the Board level. Companies with such committees
have significantly higher ratings in both the KLD negative social ratings index and their
positive social ratings index. A look at the panel data over the six-year period between
2000 and 2005 suggests that companies that had recently experienced problems in social
performance established such committees for mitigation purposes.
In the third essay that I jointly wrote with Joel Waldfogel, we investigate whether
history has persistent effect on economic outcomes by looking into the history of chain
restaurants in the U.S. Different chain restaurants came to existence at different points in
time, penetrating into the new market areas where they faced less competition and
expected to reap higher profit. It turns out that the chains that settled down in those new
market areas were able to withstand the arrivals of later competitors. As a result, we
observe that different chain restaurants are available in areas where current market
conditions are very similar, suggesting that history has persisting effects on market
outcomes, rendering them path-dependent.
TABLE OF CONTENTS
ABSTRACT..................................................................................................................................vi
TABLE OF CONTENTS......................................................................................................... viii
LIST OF TABLES.......................................................................................................................xi
LIST OF ILLUSTRATIONS...................................................................................................xiii
ESSAY ONE. IN GOOD COMPANIES? A CRITICAL EVALUATION OF THE
CORPORATE SOCIAL PERFORMANCE - CORPORATE FINANCIAL
PERFORMANCE L IN K ............................................................................................................. 1
1. Introduction.............................................................................................................. 1
2. Literature and Hypotheses..................................................................................... 3
2.1. Two Views on the CSP-CFP Link................................................................ 3
2.2. Empirical Literature........................................................................................ 5
2.3. Hypotheses....................................................................................................... 9
3. Methods...................................................................................................................12
3.1. Data Sources.................................................................................................. 12
3.2. Social Performance Data...............................................................................13
3.3. V ariables......................................................................................................... 19
3.4. Industry Classification..................................................................................22
3.5. Empirical Specification.................................................................................24
4. Analysis and Results.............................................................................................27
4.1. Companies in Top and Bottom D eciles..................................................... 27
4.2. Temporal Variation in the KLD D ata.........................................................29
viii
4.3. Pooled OLS Regression R esults................................................................. 30
4.4. Industry Interaction....................................................................................... 31
4.5. Fixed Effect Regression Results................................................................. 35
4.6. Addressing the Endogeneity Concern Further - IV Estimation Results 36
4.7. Asymmetry between the Effect of Positive and Negative C S P .............. 40
4.8. Controllable vs. Uncontrollable Components of Negative Social Ratings
...............................................................................................................................................43
5. Conclusion.............................................................................................................46
ESSAY TWO: BOARD COMMITTEES AND INTEGRATED STRATEGY: HOW
FIRMS INTEGRATE MARKET AND NON-MARKET STRATEGIES FOR
MITIGATION............................................................................................................................. 48
1. Introduction............................................................................................................48
2. Related Literature..................................................................................................50
3. D ata.........................................................................................................................54
3.1. Committee D ata............................................................................................ 54
3.2. Social Performance Data.............................................................................. 59
4. Analysis & Results................................................................................................65
4.1. Descriptive A nalysis.....................................................................................65
4.2. Regression Analysis......................................................................................68
5. Diffusion of a PA/SR Comm ittee.......................................................................71
5.1. Summary Statistics........................................................................................ 71
5.2. Regression Analysis......................................................................................75
6. Conclusion.............................................................................................................78
ix
ESSAY THREE. THE TIPPING POINT: PATH-DEPENDENCE IN RESTAURANT
M ARK ETS..................................................................................................................................79
1. Introduction............................................................................................................79
2. Theoretical Background and Literature............................................................. 81
3. D ata......................................................................................................................... 87
3.1. Geographic Market Definition and Market A ge........................................87
3.2. Chain Location and Zipcode Demographics.............................................. 89
3.3. History of Chain Restaurants.......................................................................92
4. Analysis & Results................................................................................................95
4.1. Descriptive A nalysis..................................................................................... 95
4.2. How Do Chain Founding Dates Correlate with Zip Vintage?...............102
4.3. Individual Chain’s Prevalence in Different Zip Vintages...................... 106
5. Conclusion........................................................................................................... 112
APPENDIX................................................................................................................................114
Appendix 1. List of Strength andConcern Items in the KLD Social Ratings
Database, by Category (Essays 1 & 2)...............................................................................114
Appendix 2. Alphabetical List of Chains (Essay 3)............................................ 117
BIBLIOGRAPHY.....................................................................................................................122
LIST OF TABLES
Table 1-1] Summary Statistics of the KLD Index over the 8-year Period 1998 - 2005... 18
Table 1-2] Summary Statistics and Correlation among Variables....................................... 21
Table 1-3] Industry Classification in the S am ple.................................................................. 23
Table 1-4] List of Top and Bottom Decile Companies with Largest Market Capitalization
at the End of 2005..........................................................................................................28
Table 1-5] Cross-sectional and Time Series Variation in the KLD D ata............................29
Table 1-6] Pooled OLS Regression Results of Financial Performance on Social
Performance....................................................................................................................32
Table 1-7] Regression Results of CFP on CSP by Industry................................................. 34
Table 1-8] Fixed Effect Regression Results of ROA and Tobin’s q on the KLD Index.. 36
Table 1-9] 2SLS Regression Results with SOX as an Instrumental Variable for the KLD
In d ex ............................................................................................................................... 39
Table 1-10] Regression Results of Financial Performance on the Lagged KLD Positive
and Negative Ratings Index......................................................................................... 42
Table 1-11] Regression Results of Financial Performance on the KLD Positive Ratings
Index and the KLD Negative Ratings Index with Uncontrollable and Controllable
Parts Separated............................................................................................................... 45
Table 2-1] Percentage of Companies with a PA/SR Committee in 2005, by Industry 60
Table 2-2] KLD Positive Ratings, KLD Negative Ratings, and KLD Overall Ratings
Indices by the Presence of a PA or SR Committee (2005)...................................... 62
xi
Table 2-3] Summary Statistics for Firm Characteristics (2005)...........................................64
Table 2-4] Regression Results of the KLD Overall, Positive, and Negative Ratings Index
on the Presence of a PA/SR Committee (2005)....................................................... 70
Table 2-5] Industry Breakdown of the Presence of a PA/SR Committee in 2000 and in
2005..................................................................................................................................72
Table 2-6] Summary Statistics of V ariables........................................................................... 73
Table 2-7] Correlation among V ariables.................................................................................74
Table 2-8] Probit Regression Results of the Presence of a PA/SR Committee on the KLD
Negative Ratings Index................................................................................ 77
Table 3-1] Summary Statistics for Variables.......................................................................... 91
Table 3-2] List of Chain Restaurants by the Time They Started Operating as Chains......93
Table 3-3] Timeline of Chain Expansion - an Example of Denny’s vs. Applebee’s ........94
Table 3-4] Percent of Chain Restaurants in a Founding Cohort in Markets of Different
V intages........................................................................................................................ 103
Table 3-5] Regression Results of the Fraction of Restaurants by Vintage on Market Age
.........................................................................................................................................105
Table 3-6] Tobit Regression Results of Individual Chain’s Fraction in Zipcode Area on
Zip Vintage Period D um m ies.................................................................................... 109
x ii
LIST OF ILLUSTRATIONS
Figure 1-1] Relationship between 5-year Average KLD Index (2000-2004) and Fortune
Score (2005).................................................................................................................. 16
Figure 1-2] Histogram of the KLD Index by Y ear.................................................................19
Figure 2-1] Integrated Strategy Framework (Baron, 2006. p. 654)..................................... 51
Figure 2-2] Average KLD Negative and Positive Ratings Indices by Industry................ 66
Figure 2-3] Within-industry Differences in Average Negative and Positive Ratings
Indices............................................................................................................................. 67
Figure 2-4] Number of S&P 500 Firms with a PA/SR Committee by Year....................... 71
Figure 3-1] Distribution of Outlet Locations by Zip Vintage for Selected C hains........... 97
Figure 3-2] Mean and Median Year of Founding by Zip Vintage of Market Areas, All
C hains............................................................................................................................. 98
Figure 3-3] Mean and Median Year of Founding by Zip Vintage of Market Areas,
Sitdown Chains.............................................................................................................. 99
Figure 3-4] Fraction of Chains Originated in Different Periods Over All Chain
Restaurants in the Areas with Different Zip Vintages............................................ 100
Figure 3-5] Fraction of Sitdown Chains Originated in Different Periods Over All Chain
Restaurants in the Areas with Different Zip Vintages............................................ 101
Figure 3-6] The Coefficients of Zip Vintage Period Dummies from Regressions in Table
3-6 (Denny’s vs. Applebee’s ) .................................................................................... I l l
x iii
ESSAY ONE. IN GOOD COMPANIES? A CRITICAL
EVALUATION OF THE CORPORATE SOCIAL PERFORMANCE -
CORPORATE FINANCIAL PERFORMANCE LINK
1. Introduction
How does corporate social responsibility (CSR) affect corporate financial
performance? An increasing number of companies have devoted their resources and
managerial attention to CSR in recent years. As will be shown in the second essay of this
dissertation, the number of S&P 500 companies with a committee that specializes in
CSR-related operations in their boards rose from 66 in 2000 to 89 in 2005, an increase of
almost 35%. Some large pension and mutual fund accounts screen for socially
responsible companies and invest only in companies that meet certain social performance
criteria (Barnett and Salomon, 2006), such as a certain environmental standard, a
corporate governance oversight board or institutionalized fair labor practices1. These
funds have grown in strength and influence since their inception in the 1970s. Social
Investment Forum reports that socially screened mutual funds hold $179 billion dollars at
the end of 2005 (SIF, 2006). TIAA-CREF’s Social Choice Account alone manages about
1 Socially screened mutual funds were created to include negative screens on controversial businesses,
sucha s tobacco and nuclear energy. It is only recently that they started including firms with a certain
positive attribute.
1
$7.8 billion in assets; Domini Social Investments Fund manages $1.8 billion; Pax World
Balanced Fund manages $2.1 billion. If a company fails to meet the fund’s social
screening standard, these funds will divest the company’s stock, and its share price could
take a hit.
Many researchers have tried to investigate the relationship between corporate
social performance (CSP) and corporate financial performance (CFP). A growing body of
research supports a positive correlation between a firm’s social performance and financial
performance. However, the existing studies, which find positive associations between
CSP and CFP, are vulnerable to a concern that unobserved heterogeneity drives the result,
perhaps because financially healthy firms invest more in social performance. In this paper,
I try to evaluate critically the most recent studies in the field and propose an alternative
identification strategy. I claim that the positive association between CSP and CFP found
in recent studies may be due to endogeneity, part of which is the unobserved firm-level
differences in capabilities. After controlling for firm-level heterogeneity and endogeneity,
I find that there is no significantly positive effect of CSP on CFP and that the effect is
likely negative. Furthermore, I find asymmetry in the association between negative social
performance and financial performance and the association between positive social
performance and financial performance. In pooled OLS regressions, negative social
performance does seem to be negatively correlated with profitability, whereas positive
social performance shows no correlation. But after controlling for firm-level
heterogeneity, positive social performance, or doing good, does seem to be negatively
correlated with a company’s financial performance, whereas negative social performance
shows no correlation with financial performance.
The study’s contributions are threefold: first, it provides empirical evidence that
challenges the current majority view on the topic, that it pays to do good. Second, it
raises a methodological issue that endogeneity concerns including individual firm-level
unobserved heterogeneity may be an important factor in all empirical studies in the CSP-
CFP link debate, and that ignoring these effects may lead a researcher to a very different
conclusion, even to an opposite one. Finally, it points out the need to investigate the
positive and negative sides of CSP separately in studying the CSP-CFP link by providing
the evidence that positive and negative social performances affect corporate financial
performance in an asymmetric way.
The organization of this essay is as follows: in section 2, I review theoretical and
empirical literature on the topic and develop my hypotheses. In section 3, I explain my
data and present empirical methodology. Section 4 presents analysis and results, and
section 5 concludes.
2, Literature and Hypotheses
2.1. Two Views on the CSP-CFP Link
Baron (2006) defines CSR in two slightly different ways. Citing Vogel (2005), the
first definition extends beyond compliance with laws and regulations, identifying CSR as
“policies and programs of private firms that go beyond legal requirement as a response to
public pressures and societal expectations.” Baron also provides a stakeholder-oriented
definition, explaining that CSR requires “balancing the legitimate interests of all
stakeholders.” Tirole (2005) mentions a list of concerns that managers should keep in
mind, based on externalities and concomitant duties toward four sets of stakeholders. The
first is a set of duties toward employees. Firms should refrain from laying off workers
when they make sizeable profits; firms should protect minority employees; and firms
should provide a safe work environment, good training and benefits. The second set
consists of duties toward communities. Firms should refrain from closing plants in
distressed economic areas, except when strictly necessary, and they should contribute to
their communities by offering voluntary services and charitable giving in normal times.
The third set of duties is those toward the greater society, in terms of ethical
considerations. They include the protection of the environment, staying away from
countries with oppressive governments and avoiding bribery and other forms of
corruption. The fourth set of duties consists of those toward creditors, whose interests
might conflict with those of the shareholders.
Two of the most respected scholars in the fields of economics and business
present two opposite views of how CSR affect CFP. In a well-publicized article in The
New York Times Magazine, Friedman (1970) states that corporate resources spent on
social responsibility-related activities come out of shareholders’ pockets.2 If his claim is
true, CSR cannot be value-maximizing to the firm. His point of view was later supported
by agency theory, where “agency problems that allow managers to act as principals rather
than as agents of shareholders can result in investment in CSR that is not rewarded in the
2 However, he acknowledges the em ployees and the customers as legitimate stakeholders o f a firm.
4
marketplace.3” If the agency theory view is correct, one would expect to see a negative
relationship between CSP and CFP. Also, in a simple economic model of a profit-
maximizing firm, any attempt to satisfy various stakeholders imposes additional
constraints on the firm, which can only reduce profits, if it has any effect.
In contrast, Bowman and Haire (1975) argue that it is a myth that activities in
social responsibility are in fundamental conflict with the interests of the shareholders.
They provide empirical evidence supporting their claim by analyzing how much
companies emphasize CSR in their annual reports. Stakeholder theory provides
mechanisms that explain why managing the stakeholder interests can increase
profitability of a firm (Wood, 1991; Freeman, 1991; Donaldson and Preston, 1995;
Hillman and Keim, 2001).
2,2. Empirical Literature
The empirical relationship between CSP and CFP has been studied extensively,
and most existing studies attempt to draw inferences about the effect of CSP on CFP
from a cross-sectional relationship between measures of CSP and CFP. Among the
empirical studies, three are of particular interest here, partly because they try to establish
a causal link, and partly because their measures of CSP are drawn from the same source
as in this study, the KLD Social Ratings Database.
Waddock & Graves (1997a) use a lagged cross-sectional approach to examine the
link between CSP and CFP using the KLD Social Ratings data and financial performance
3 Baron (2006) p.670.

5
measures such as return on asset (ROA), return on equity (ROE) and return on sales
(ROS). To deal with the concern that CSP is an endogenous variable, they use the lagged
KLD index as an explanatory variable. This approach eliminates the issue of
contemporaneous endogeneity. Their analysis of S&P 500 companies from 1989 to 1991
finds that the previous year’s social performance is positively correlated with the current
year’s financial performance, supporting the hypothesis that good management is linked
to good social performance. On the other hand, the previous year’s financial performance
is positively correlated with the current year’s social performance, supporting the
hypothesis that the social performance is affected by the availability of slack resources. I
attempt to use a similar empirical model in analyzing the social and financial
performance data from a panel dataset of S&P 500 companies between 1998 and 2005.
McWilliams and Siegel (2000) argue that Waddock and Graves’ empirical model is
misspecified in that it omitted an important explanatory variable, R&D intensity, which is
strongly correlated with both CSP and CFP. Their claim is that the positive association
between CSP and CFP disappears once an R&D-to-sales ratio is included in the empirical
specification as a measure of the R&D intensity.
Berman et al. (1999) investigate how social responsibility activities, measured by
the KLD Social Ratings data, affect ROA, using panel data of the companies on the
Fortune 100 list from 1991 to 1996. They try to control for firm strategy and
environmental variables, using various measures such as selling intensity, capital
expenditure and capital intensity. They also acknowledge the possibility of
heteroscedasticity and autocorrelation in the error term. However, their study does not
include firm fixed effects, so its results are vulnerable to an alternative explanation:
unobserved firm level heterogeneity could explain the association between CFP and CSP.
The inclusion of variables to capture strategy and environment cannot fully capture all
the strategic considerations going on inside the firm. For example, how much attention
the CEO pays to CSR is not observable in their study, but it can surely affect both the
social performance and financial performance of the firm. Likewise, the quality of the top
management team is likely a crucial factor in both financial performance and social
performance, but it is hard to observe empirically. In this study, I treat these issues by
including firm fixed effects and by using an instrumental variable for CSP.
Hillman and Keim (2001) estimate the effect of CSP on the change in CFP using
the CSP data from 1994 and the CFP data from 1995 and 1996. They find that better
stakeholder management leads to improved shareholder value (measured in terms of
market value added), whereas participation in social issues reduces shareholder value.4
Their empirical specification is to regress the change in CFP on the level of explanatory
variables, such as CSP, industry, sales, net income and a measure of risk. They find that
Ruf et al. (2001) take a first-difference approach and look at how the change in CSP
between 1990 and 1991 affects changes in financial performance in one, two and three
years’ time. They find that annual changes in ROE and ROS between 1993 and 1994 are
positively correlated with changes in CSP between 1990 and 1991. Their first-difference
approach, if applied appropriately, can be quite similar to the fixed-effect estimation
method. However, their empirical specification is not entirely consistent with fixed-effect
4 Their stakeholder management category includes community relations, diversity, em ployee relations,
environment, and product categories. Their social issue participation category includes corporate
governance (“other” at the time o f their publication), human rights, and controversial business areas.
7
specification because their model also contains a lagged dependent variable and the level
variables, such as firm size and industry.5
Several articles document the existing empirical analysis in this field. Griffin and
Mahon (1997) survey the existing empirical literature of eighty papers, in which two
thirds of the results support a positive correlation between CSP and CFP, and one quarter
of the results support negative correlation between the two. Margolis and Walsh (2001)
look at 80 empirical studies that examine the effect of CSP on CFP. They find a positive
correlation in 53% of the studies, a negative correlation in 5% of the studies, and no
correlation or mixed results in 43% of the studies. Frooman (1997) surveys existing event
studies and documents the evidence that legal or regulatory setbacks have a significantly
negative effect on stock prices. Orlitzky et al. (2003) compile 52 empirical studies and
conduct the meta-analysis to find that overall, the evidence suggests a positive correlation
between CSP and CFP.
To sum up, the majority of recent literature on the relationship between CSP and
CFP suggests a positive correlation between the two, even though a sizable minority of
studies still find either a negative correlation or none at all. In the following sections, I
provide results that challenge the majority view in this field, using a different
identification strategy and more extensive data.
5 In a sense, these studies are estimating a cointegration model o f CSP on CFP, but there is little theoretical
justification for such a model.
2.3. Hypotheses
A firm can manage its stakeholder relations in a way that benefit its bottom line
(Donaldson & Preston, 1995; Porter & Kramer, 2006). Managing environmental concerns
proactively can lower the costs of complying with existing and future environmental
regulations, even though it can increase operating costs in the short term. Managing
employee relations proactively can lower worker turnover and absenteeism and enhance
productivity by improving commitment and effort. Increased diversity in the workforce
means that the firm is not constrained by race or gender in selecting the best people for its
jobs, thereby enabling the company to recruit better talent (Berman et al., 1999; Waddock
and Graves, 1997a). It is also noted that managing a good inter-organizational
relationship with the firm’s suppliers can be the source of a competitive advantage (Dyer
and Singh 1998). Maintaining a good relationship with the community in which the firm
operates can not only lessen the likelihood of a negative media event such as picketing,
but also it may have the positive effect of attracting desirable residents (Tiebout, 1956;
Waldfogel, 2003). Customers can take actions to reward or punish a firm’s policies in an
attempt to change or reinforce such activities, thereby creating a positive link between
CSP and CFP (Rowley and Berman, 2000; Schuler and Cording, 2006). Shareholders are
also influential stakeholders, and Gompers, Ishii, and Metrick (2003) show that
companies with a governance structure that protects shareholder rights exhibit better
financial performance.6 There are also areas of social performance that are more closely
related to the normative aspect of CSR, such as human rights issues and involvement in
controversial businesses, which belong to the category of broad responsibility toward
society (Wood, 1991). Following these arguments, I hypothesize that CFP and CSP have
a positive association.
Hypothesis 1. CSP is positively associated with CFP.
Scholars claim that CSR is a strategic action taken by a company. A firm may
choose an optimal level of CSR by equilibrating the demand for CSR and the supply of
CSR (McWilliams & Siegel, 2001), or a firm should “integrate social considerations
more effectively into core business operations and strategy” (Porter & Kramer 2006).
Russo & Fouts (1997) emphasizes the usefulness of the resource based view (Barney,
1991; Wemerfelt, 1984) in investigating the effect of CSR by noting that “the resource
based view explicitly recognizes the importance of intangible concepts such as know how
n
and corporate culture.” These scholars point out that there are unobserved firm-specific
factors that play an important role in the CSP-CFP link, and I hypothesize that the
positive association between CSP and CFP can be due to the unobserved firm-level
heterogeneity.
6 In som e areas, shareholder rights can be in conflict with the interests o f other stakeholders. On the other
hand, there is little conflict between shareholder rights and good social performance in issues pertaining to
agency problems benefiting top management, and issues o f transparency and accountability in corporate
governance. The KLD data captures issues related to the latter in corporate governance category.
7 Russo & Fouts, (1997), p.535.
10
Hypothesis 2. The positive association between CSP and CFP can change once
you control fo r unobserved firm level heterogeneity.
It is possible to break down CSP into two parts. Negative CSP items can be
inherent to the business, such as the product a firm offers, or they can arise despite the
best efforts from the firm in trying to prevent them, such as strikes or tax disputes with
federal or local governments. On the other hand, positive CSP items are the results of
conscious efforts by the firm. Frooman (1997) finds that firms deemed socially
irresponsible also perform poorly in terms of their bottom lines. It is possible that the
positive association between CSP and CFP is driven mainly by the fact that firms with
high negative CSP ratings perform poorly. On the other hand, a company that achieves
high marks in the positive side of CSP, such as charitable giving or generous retirement
benefits, is spending its resources more on CSR-related programs and activities. In doing
so, the company is directing resources away from its core operations, and that could put it
at a financial disadvantage, compared to its competitors with similar capabilities and
resources. I hypothesize that the association between CSP and CFP after controlling for
the firm-level heterogeneity, is driven by the fact that companies that achieve higher
positive CSP ratings than the within-firm average level are suffering from lower financial
performance than the within-firm average level of financial performance.
Hypothesis 3. The positive association between CSP and CFP is driven by the
negative CSP ratings.
11
Hypothesis 4. The negative association between CSP and CFP after controlling
fo r the firm-level heterogeneity is driven by the positive CSP ratings.
3. Methods
3.1. Data Sources
I gathered the financial and social performance data for the firms that belong to
the KLD Social Ratings database from 1998 through 2005.8 1 collected accounting data
from the Compustat North America database, stock market data from the CRSP database
and social performance data from the KLD Social Ratings database. From the data, I
calculated financial performance measures such as ROA and Tobin’s q to use as my
financial performance measures.9 A great deal of existing research on the CSP-CFP link
has used ROA as its financial measure (Waddock & Graves 1997a; Berman et al. 1999),
and using ROA makes it easier to compare my results with the existing literature. The
fact that ROA is an accounting measure of profitability and does not capture the expected
future profitability is addressed by using Tobin’s q as financial performance measure in
addition to ROA. Defined as the ratio of the market value of the company over the
8 The KLD database has expanded substantially since 2000. Since its creation in 1991, it has covered the
social ratings data o f 650 large companies. In 2001 it expanded to include 1,100 companies, including the
entire Russell 1000 Index companies. In 2003, it expanded once again to include a total o f 3,100 companies,
covering all o f the Russell 2000 Index and the Broad Market Social Index.
9 ROA is calculated by dividing the earnings by total assets. In calculating Tobin’s q, I follow Gompers,
Ishii, and Metrick (2003), and define Tobin’s q as the following:
Tobin’s q = (market value o f assets) / (book value o f assets)
= (book value o f assets + market value o f common stocks - book value o f comm on stocks - deferred taxes)
/ (book value o f assets).
12
replacement cost of its assets (Lindenberg & Ross 1981), Tobin’s q factors in the stock
market valuation in measuring the financial performance. It is also similar to the market
value added (MVA) measure that Hillman and Keim used in their study of the CSP-CFP
link (Hillman & Keim, 2001). I restrict my attention to the companies that belong to the
S&P 500 Index for at least two years during the 1998-2005 period. This restriction is
necessary for keeping the sample constant while using the fixed effect estimator.
3.2. Social Performance Data
The KLD Social Ratings database is published by KLD Research & Analytics,
Inc., a Boston-based consulting firm that specializes in measuring CSP. It also publishes
the Domini Social 400 Index, the stock index for 400 companies whose business is
deemed socially responsible. The KLD Social Ratings data is very influential measure of
CSP, and many investment managers refer to KLD’s recommendations when drawing on
social screening. For example, TIAA-CREF announced that it divested its 1.2 million
shares of Coca-Cola stock, estimated at $52.4 million, following a periodic review of the
Broad Market Social Index by KLD Research & Analytics Inc. on July 18, 2006. KLD
saw shortcomings on the part of Coca-Cola in several areas, including worker rights at
overseas bottling plants, marketing of soda products to children and environmental issues
related to water usage at overseas facilities.10 The KLD Social Ratings data are also the
most frequently cited source of CSP in academic research (Harrison and Freeman, 1999;
Berman et al., 1999).
10 Reuters, Jul. 18, 2006.

13
The KLD data cover approximately 80 indicators in seven major qualitative issue
areas over the sample period: community, corporate governance, diversity, employee
relations, environment, human rights and product quality and safety.11 In addition, KLD
data provide information for involvement in the following “controversial business
issues:” alcohol, gambling, firearms, military, nuclear power and tobacco. Business
12
involvement in any of these sectors results in a negative indicator. Each issue area has a
number of strength and concern items, where a binary measure indicates the presence or
absence of that particular strength or concern. For example, the community category
contains seven strength items (charitable giving, innovative giving, non-U.S. charitable
giving, support for housing, support for education, volunteer programs and other strength)
and four concern items (investment controversies, negative economic impact, tax disputes
and other concerns). Each year, KLD evaluates the companies in the database on each
item through various sources, such as public records and media reports, monitoring of
corporate advertising, surveys and on-site evaluations. The database provides substantial
cross-section and time series variation in CSP, as I will explain in detail later in section
4.2. The complete list of strength and concern items by category is provided in the
appendix.
I define the KLD index as the sum of all strength items minus the sum of all
concern items. This simple manipulation has advantage over more complex methods that
bestow different weights on different categories in the sense that it is not affected by
11 The exact number o f items varies slightly year to year, both in total and within each category.
12 As o f 2005, the KLD data did not include ratings explicitly related to animal rights or biotechnology
issues.
14
some indicators’ switching between categories. Over the 8 year sample period, some
indicators do change categories.13
To test the validity of the KLD index, I apply two tests of consistency: one
external, and the other internal. For an external validity check, I look at the correlation of
the index with Fortune’s reputation score (Wood, 1995; Fortune, 2005). For the 514
companies listed in Fortune Magazine’s 2006 list of America’s Most Respected
Companies, the correlation between the KLD index in 2005 and the Fortune Score in
2006 is positive and highly significant. The correlation coefficient is 0.2556, and is
significant at the 0.1% level. I also look at the relationship between the Fortune score in
2006 and the 5-year average KLD index from 2000-2004 because it may take some time
to build a reputation in social performance. Figure 1-1 shows this relationship, with the
straight line indicating the linear fitted value and the shaded area indicating the 95%
confidence interval. The correlation coefficient is 0.2643, and it is significant at the 0.1%
level. These results confirm that the KLD index is consistent with one of the most widely
used measures of reputation, rendering it credible.
13 I also used a modified index, calculated in the follow ing way: seven qualitative issues areas (i.e.
corporate governance, community, diversity, em ployee relations, environment, product quality and safety,
and human rights categories) receive a weight o f 1, while involvement in alcohol, firearms, gambling,
military, nuclear, and tobacco categories receive a weight o f 0.5. This weighting scheme is closer to what
W addock & Graves (1997a) use after consulting an expert panel. This modified index puts more weight on
the effort and attitude o f the company toward social performance, rather than the products they provide.
Since simple and modified indices lead to very similar estimation results, I only report the results from the
regressions using the simple index.
15
-10 -5 0 5 10
5-year A verage KLD Index
9 5 % Cl Fitted va lu es
Fortune S co re 2 0 0 5 ________
Figure 1-1] Relationship between 5-year Average KLD Index (2000-2004) and Fortune
Score (2005)
For an internal validity check, I calculate the summary statistics and
autocorrelation of the KLD index. The KLD index for my sample displays little variation
in its standard deviation over time, and it also shows strong autocorrelation. During the 8-
year sample period, the mean of the index stays between 0.0812 (1998) and -0.6401
(2004), with the within-year standard deviation remaining close to 2.9. The
autocorrelation in the index between year t and year (t-1) is about 0.9, and the
autocorrelation falls down smoothly as the years between two observations become
farther away. Table 1-1 presents the summary statistics and autocorrelation in the KLD
16
index.14 This pattern is consistent with the idea that it takes time to build the capabilities
in social performance but that those capabilities will be lasting once established. Figure
1-2 shows the histograms of the KLD index by year, with the histogram over all years at
the end. All histograms are overlaid with a normal distribution. The histogram of the
KLD index over all years shows no skewness, but it is leptokurtic, with more acute peak
and thicker tails than a normal distribution. The histogram of the KLD index by year
shows the same pattern, with skewness ranging from -0.21 to 0.41, and with kurtosis
ranging from 3.29 to 4.02.15
In addition to the KLD index, I also define the KLD positive ratings index as the
sum of all strength items in the KLD database, and the KLD negative ratings index as the
sum of all concern items in the KLD database. Therefore, the following relationship
holds among the KLD index, the KLD positive ratings index and the KLD negative
ratings index:
KLD Index = KLD Positive Ratings Index - KLD Negative Ratings Index
14 I also examine the autocorrelation in other financial variables. Those variables that are related to a firm’s
size (such as assets and number o f em ployees) tend to exhibit strong autocorrelation, whereas the variables
related to the firm’s stock market valuation tend to show much smaller autocorrelation.
15 No year shows skew ness significantly different from 0 at the 5% level. In contrast, kurtosis is
significantly different from 3, the kurtosis o f normal distribution, in all but one year at the 5% level.
17
Reproduced
with permission
of the copyright owner. Further reproduction
Year Mean SD N 1998 1999 2000 2001 2002 2003 2004

1998 0.0812 2.8621 394 1
1999 0 .0 2 9 6 2 .9 0 3 6 439 0.9175*** 1
2000 0 .0199 2 .8 3 6 6 453 0.8276*** 0.8903*** 1
2001 -0.2627 2 .8 7 9 0 510 0.7926*** 0.8461*** 0.9089*** 1
2002 -0.2386 2 .9 1 5 6 503 0.7526*** 0.8105*** 0.8536*** 0.9198*** 1
2003 -0.5253 2 .9 1 2 0 514 0.6912*** 0.7538*** 0.7772*** 0.8189*** 0.8711*** 1
2004 -0.6401 2 .9 9 7 9 514 0.6393*** 0.6952*** 0.6927*** 0.7455*** 0.8026*** 0.8901*** 1
2005 -0.2797 3 .2 8 9 8 497 0.5873*** 0.6503*** 0.6502*** 0.6814*** 0.7317*** 0.7748*** 0.8325***
*** significant at 0.1% level.
Table 1-1] Summary Statistics of the KLD Index over the 8-year Period 1998 - 2005
prohibited without perm ission.
18
KLD Index
Density
------------ normal kldind
Graphs by year
Figure 1-2] Histogram of the KLD Index by Year
3.3. Variables
Table 1-2 presents the summary statistics and correlation among the variables
used in the regression analysis: the KLD index, the KLD positive ratings index, the KLD
negative ratings index, ROA, Tobin’s q, natural log of assets, natural log of sales, natural
log of number of employees and debt ratio. Financial data tend to have large outliers. I
take the natural log of variables such assets, sales and the number of employees to
minimize the skewness problem in these variables. For accounting performance measures
19
with large outliers, I drop those observations that are outside of the I 5,1percentile and the
99th percentile of the population distribution. One of the most notable correlation
coefficients appears to be the one between the KLD positive ratings index and the KLD
negative ratings index. The two indices’ overall correlation coefficient is 0.3052, and it is
significant at the 5% level. This suggests that those companies in controversial businesses,
as well as companies that have troubles in their stakeholder relationships, might invest
more in social performance to mitigate the negative impact on their reputation (Kotchen,
2006). This also raises the possibility that one may benefit from disentangling the
positive and negative sides of the social performance separately in studying the CSP-CFP
link.
20
Reproduced
with permission
KLD
KLD KLD Pos. Log Log Log #
N Mean St. Dev. Neg. ROA Tobin’s q
Index Ratings Assets Sales Emp.
Ratings
KLD Index (t-1) 2771 -0.2288 2.9620 1

KLD Pos.Ratings (t-1) 2771 2.6572 2.5228 0.5939* 1
KLD Neg.Ratings (t-1) 2771 2.8860 2.4829 -0.5849* 0.3052* 1
ROA 2721 0.0445 0.0886 0.0843* 0.0106 -0.0890* 1
Tobin's q 2221 2.1547 1.7239 0.1772* 0.0095 -0.1914* 0.4095* 1
Log Assets 2721 9.2270 1.4195 0.011 0.4283* 0.4187* -0.1235* -0.2857* 1
Log Sales 2721 8.7496 1.1834 -0.0252 0.4358* 0.4692* 0.0437* -0.1826* 0.7500* 1
Log # Emp 2688 3.1402 1.1704 -0.0063 0.3157* 0.3239* 0.0344 -0.1310* 0.4948* 0.8040* 1
Debt Ratio
2717 0.6357 0.2160 -0.0292 0.1305* 0.1662* -0.2043* -0.3706* 0.4891* 0.3088* 0.1746*
* significant at 5%
Table 1-2] Summary Statistics and Correlation among Variables
21
3.4. Industry Classification
I classify the industry following Waddock & Graves’ industry classification
(Waddock & Graves, 1997).16 This classification has two advantages over a finer
classification, for example, using the two-digit SIC, which results in 55 different
industries. Using an industry classification that is close to Waddock & Graves’ enables a
close comparison of my estimation results to theirs. In addition, the number of industries
is manageable, and I can tease out an interesting estimation result by industry, which I
will discuss in table 1-7.
CSP appears to vary a great deal by industry. The industry with the lowest
average KLD index is mining and construction at -2.4857. The industry with the highest
average KLD index is bank and financial services at 0.7226. Paper and publishing,
computers and precision products, and hotel and entertainment industries have favorable
CSP ratings on average, whereas refining, rubber and plastic, auto and aerospace, and
hospital management industries have unfavorable CSP ratings on average. Table 1-3
details my industry classification.
16 One exception is that I break the computer, auto and aerospace category into two separate categories:
computers and precision products, and auto and aerospace. In my view, these are different industries and
should not be lumped into a single category.
22
Reproduced
with permission
A verage
Number SIC Category N
KLD Index
1 1000 ~ 1799 Mining & Construction 105 -2 .4 8 5 7

2 2000 ~ 2399 Food, Textiles, Apparel 145 -0 .2 8 9 7
3 24 0 0 ~ 2799 Paper & Publishing 142 0.3 0 9 9
4 28 0 0 ~ 2899 C hem icals & Pharmaceuticals 216 -0.4120
5 29 0 0 ~ 3199 Refining, Rubber, Plastic 78 -1 .2 9 4 9
6 3200 ~ 3569 Heavy Manufacturing 173 -0.5723
7 35 7 0 ~ 3699 Computers & Precision Products 447 0.6 9 1 3
8 3700 ~ 3799 Auto & A erospace 104 -2 .3 4 6 2
9 4000 ~ 4789 Transportation S ervices 51 -0.7451
10 4800 ~ 4991 T eleph one & Utilities 247 -1.2713
11 5000 ~ 5999 W holesale & Retail 306 -0 .0 4 5 8

12 6000 ~ 6799 Bank & Financial S ervices 447 0 .7 2 2 6
13 7000 ~ 7999 Hotel & Entertainment 243 -0.0370
14 8000 ~ 8999 Hospital M anagem ent 57 -1 .2 1 0 5
Table 1-3] Industry Classification in the Sample
23
3.5. Empirical Specification
Shareholders are known to be highly responsive to news items related to CSP,
especially when the news is negative (Frooman, 1997). In the following analysis, I
assume that other stakeholders (customers, employees and people who reside near the
company’s plants or laboratories) also pay close attention to the operation of the
company in which they have a stake, and the activities the company initiates are noticed
by those stakeholders. However, because of the time lag associated with compiling the
KLD database, I assume that the stakeholders observe last year’s social performance in
determining the course of actions toward the com pany.17 These actions affect the
company’s bottom line either directly or indirectly. This behavioral model necessitates an
empirical specification in which a company’s financial performance is affected by its
social performance with a one-year time lag, which is consistent with the existing
literature (Waddock and Graves, 1997a; Hillman and Keim, 2001). This formulation also
takes care of the contemporaneous endogeneity issue in the following regression analysis.
My sample of firms includes the firms that belonged to the S&P 500 Index for at
least two years between 1998 and 2005. After dropping those companies with incomplete
social performance data, I have 527 companies in my sample.181 estimate the following
model to get the pooled OLS regression results in Table 1-7:
17 The KLD Social Ratings Database is published once annually, in mid to late January.
18 The results reported in this section are robust to the selection o f the sample. Regressions on the sample o f
firms in the entire KLD database, as well as those on the sample o f firms that stayed in the S&P 500 index
for all eight years, yield similar results.
24
yi, =Xi,'-iP+zi,r+Dils+£u (i)
ya is the financial performance measure, X u./ is the lagged KLD index, Zi( is the
vector of firm-level characteristics that affect financial performance, D it is the vector of
industry and year dummies, and eit is the error term. This empirical model postulates that
the financial performance is affected by the contemporaneous control variables, such as
debt ratio, sales and the size of the workforce, industry, and business cycle, along with
the lagged social performance variable. The error term is assumed to be independently
distributed for each firm and is uncorrelated with right-hand side variables. I allow
heteroscedasticity in the error term, and calculate Huber-White robust standard errors
clustered at each firm level in the estimation. As measures of CFP, I use ROA and
Tobin’s q.19 As firm-level characteristics that may affect financial performance, I include
the debt ratio of a firm to capture the interest cost and leverage risk and natural log of
20
sales, and the number of employees to capture the size of a company. Industry is
measured in 14 industry categories, similar to those in Waddock and Graves (1997a). 21
Because financial performance is heavily affected by the business cycle, I also include
the year dummy to control for the business cycle effect.
However, I argue that this approach suffers from the existence of unobserved
firm-level heterogeneity, and as a result, the coefficient estimates are not consistent.
19 Regressions using other accounting measures such as ROE and ROS show similar results to the
regression results using ROA as the dependent variable.
20 Gompers and Metrick (2003) include the state o f incorporation in their regression analysis. I do not
include the variable because a fixed effect model cannot estimate the effect o f the variable.
21 The industry classification using the 2-digit SIC yields basically the same results.
25
There is an endogeneity problem caused by the unobserved firm-specific characteristics.
If I break down the error term into two parts, the empirical specification (1) becomes:
y» = + z i,r + c <s + a + v.„ (2)
Vectors X ir,.i and Z„ are defined as before, and the vector C, contains the dummy
variables for years. The error term is assumed to contain a time-invariant firm level
characteristic /u„ as well as the idiosyncratic error v it. Depending on the correlation
between the explanatory variables and the firm-level unobserved heterogeneity,
estimating equation (2) can yield inconsistent coefficient estimates. In other words, the
consistency of the estimator hinges on the assumption that the firm-specific fixed effect is
uncorrelated with the explanatory variables:
E(Vi I X it_r Z lt) = 0 (3)
If assumption (3) is violated, a fixed-effect transformation is necessary to obtain
consistent estimates. For each pair of fixed effect and random effect regressions, I
perform the Hausman specification test to determine whether assumption (3) is violated.
My fixed effect specification is the following, where the variables with upper bar denote
the within-firm average values over the sample period:
(y„ - y , ) = (X„_, - X , ) / ? + ( Z „ - Z J r H C ' - O S H V H - v , ) (4)
26
4. Analysis and Results
4.1. Companies in Top and Bottom Deciles
Table 1-4 shows the list of the 15 largest companies, measured by their market
capitalization at the end of 2005, in the top and bottom deciles of the KLD index in 2005.
Technology firms dominate the top decile, as eight technology companies (Microsoft,
Intel, IBM, Cisco, HP, Dell, Motorola and Texas Instruments) have their spots in top 15.
There are major financial institutions (Citi, Wells Fargo and American Express) and
major consumer product producers (Johnson & Johnson, P&G and 3M), as well. Due to
their consumer-oriented nature of business, these financial corporations and consumer-
product manufacturers seem to invest heavily in their reputations for being good
corporate citizens. On the other hand, the bottom decile is dominated by energy
conglomerates. ExxonMobil, Chevron, ConocoPhillips and Occidental Petroleum are
ranked at the bottom, mainly due to their environmental concerns. There are also
companies who are in a controversial line of business, such as defense contractors
Halliburton and Lockheed Martin and the tobacco company Altria. Companies whose
business practices are particularly controversial, such as Wal-Mart and News Corp., also
hold spots in the bottom decile.22 Table 1-4 also displays non-trivial temporal variation in
the KLD index within firms. This implies that there may be enough temporal variation in
the KLD index to exploit, despite its high autocorrelation.
22 Among the other respected companies, Starbucks is in the top decile; G oogle is in the second decile; and
GE has a relatively poor KLD index due to its substantial revenue from the defense sector.
27
2 0 0 5 Market
2 0 0 2 KLD 2 0 0 4 KLD 2005 KLD
Top Decile Capitalization
Index Index Index
($billion)
Microsoft Corporation 1 5 5 2 8 0 .0 7
Citigroup Inc. 2 2 6 2 4 1 .6 9
Johnson & Johnson 1 3 8 178.77
Intel Corporation 8 10 10 147.74
Procter & G am ble Com pany 8 5 7 143.13
IBM 6 4 9 129.38
C isco S y ste m s, Inc. 5 6 7 1 08.39
W ells Fargo & Com pany 3 6 7 105.40
United Parcel S ervice, Inc. 0 0 5 8 2.44
Hewlett-Packard Com pany 7 9 11 81.23
Dell Inc. - 2 6 6 9.78
American E xpress Company 2 2 7 63.86
3M C om pany 5 4 6 5 8 .4 8
Motorola, Inc. 6 4 11 56.50
T exas Instruments Incorporated 5 5 9 5 1.20
2005 Market
2002 KLD 2 0 0 4 KLD 2 0 0 5 KLD
Bottom Decile Capitalization
Index Index Index
($billion)
Exxon Mobil Corporation -6 -7 -9 3 4 4 .4 9
Wal-Mart Stores, Inc. -6 -6 -7 194.92
Altria Group, Inc. - 0 -4 1 55.74
Chevron Corporation -4 -3 -4 126.75
ConocoPhillips -8 -9 -8 8 0.16
Tyco International Ltd. -5 -7 -6 5 8.15
Schlum berger N.V. -2 -1 -4 5 7.20
N ew s Corporation - -2 -5 5 0.80
Carnival Corporation, Inc. -6 -6 -5 4 3 .2 6
Caterpillar Inc. -6 -4 -5 3 8 .7 6
Occidental Petroleum Corporation -10 -5 -4 3 2.12
Halliburton C om pany -3 -7 -5 3 1.85
Lockheed Martin Corporation -8 -4 -4 2 7.49
Dominion R esou rces, Inc. -2 -3 -6 2 6 .7 9
Burlington Northern S anta Fe Corp. -4 -5 -7 26.31
Table 1-4] List of Top and Bottom Decile Companies with Largest Market Capitalization
at the End of 2005
4.2. Temporal Variation in the KLD Data
Table 1-5 shows the breakdown of cross-sectional and temporal variation in KLD
indices compared with that of Gompers-Metrick governance index (Gompers, Ishii,
Metrick, 2003).23 The two datasets display similar overall ranges and dispersions. The
within-firm standard deviation in the Gompers-Metrick governance index is about a
quarter of the size of the between-firm standard deviation. In contrast, the within-firm
standard deviation is about one half the size of between-firm standard deviation in the
KLD index and about 42% of the between-firm standard deviation in the KLD positive or
negative ratings index. This variation over time helps to control for firm-specific
heterogeneity and to identify the effects of social performance on financial performance.
Variable Mean Std. Dev. Min Max O bservations

KLD Index overall -0.2453 2 .8 9 0 9 5 5 -11 11 N =3198
betw een 2.575611 -8.5 8 .3 7 5 n = 527
within 1 .281818 -6.43724 7 .5 6 2 7 6 2
KLD KLD overall 2 .7 2 1 4 9 6 2 .5 9 4 4 0 6 0 18 N =3198
(1998- Positive betw een 2 .3 4 1 6 2 2 0 13.25 n = 527
2005) within 0 .9 9 0 9 8 4 -4.5285 1 0.0965
KLD overall 2 .9 6 6 7 8 9 2 .4 9 6 4 9 4 0 16 N =3198
N egative betw een 2 .2 8 5 1 7 8 0 14.6 6 6 6 7 n = 527
within 1.045665 -2.40821 8 .3 4 1 7 8 9
G om pers G-index overall 9.164561 2 .8 4 2 1 1 6 2 18 N = 5779
et al. betw een 2 .7 7 1 5 6 4 2 17.25 n = 2281
(1990-98) within 0.7 3 2 6 9 3.664561 14 .1 6 4 5 6
Table 1-5] Cross-sectional and Time Series Variation in the KLD Data
23 One o f the concerns in using the fixed-effect estimator is whether there is enough variation in the data
over time. Especially, index-type data do not vary too much over time. Gompers, Ishii, and Metrick (2003)
explain that the reason why they do not consider firm fixed effect regression is because there is not much
temporal variation in the governance index they have developed.
29
4.3. Pooled OLS Regression Results
Table 1-6 displays the pooled OLS regression results of financial performance
measures on the lagged KLD index and other explanatory variables. In all regressions,
the coefficient estimate of the lagged KLD index is positive and significant, confirming
my first hypothesis. Column (1) reports the regression coefficients of ROA on the lagged
KLD index without industry dummies, and column (2) reports the regression coefficients
of ROA on the lagged KLD index with industry dummies. The inclusion of industry
dummies increases the R-squared from 0.09 to 0.15, but it does not affect the coefficients
substantially. Columns (3) and (4) repeat the same estimation with Tobin’s q as the
dependent variable. Here, the inclusion of industry dummies reduces the size of the
coefficient of the KLD index, even though the sign and the level of significance remain
unchanged. The coefficient of the lagged KLD index in column (2) is of particular
interest, as the specification in that regression emulates Waddock and Graves’ (1997a)
specification. It is interesting to note that the size of the coefficient on the KLD index is
similar to what Waddock and Graves obtained, after accounting for the fact that the KLD
index I use has ten times the range of Waddock and Graves’.24 In all regressions, the
coefficients of the KLD index are positive and highly significant, suggesting that CSP is
positively associated with CFP, thereby confirming my first hypothesis. The effect is
24 Due to different weights, my KLD index ranges between -11 and 11, whereas W addock and Graves’
ranges between -1.17 and 1.06. M y coefficient is about one tenth o f the size o f theirs. It should also be
noted that the definition o f industry is slightly different. They use a thirteen-industry categorization using
the 4-digit SIC code, as discussed in section 3.4.
30
economically meaningful, as well. Taking the coefficient of the KLD index in column (2),
a 10-point swing in the lagged KLD index is associated with a change in a firm’s ROA
by 2.9 percentage points.
4.4. Industry Interaction
To examine the differential association between CSP and CFP by industry, I
estimate the following model with an interaction term between the KLD index and
industry classification (Elsayed & Paton, 2004):
yit=KLDitj3 + Indltr+ K LD itx In d HS + X it9 + Sit (5)
KLDit is the lagged KLD index, lndlt is a vector of dummy variables indicating the
industry category, Xit is the vector of control variables and year dummies, and eit is the
error term. The coefficient estimates 8 are reported. Then, I run the F-test that tests the
linear relation y + 8s = 0, where the superscript j refers to the j th industry. Table 1-7
reports the regression results of specification (5).
31
Reproduced
with permission
(1) (2) (3) (4)

Dep. Variable: ROA ROA Tobin's q Tobin's q
KLD Index (t-1) 0.0 0 2 5 2 0 0.002943 0.084214 0 .0 6 9 4 4 2
(0.000934)“ (0.000976)** (0.020016)** (0.018755)**
Debt Ratio -0.102234 -0.104302 -3.066251 -2 .4 9 6 0 8 5
(0.018288)“ (0.022410)** (0.482028)** (0.492439)**
Log Net S ales 0.0 1 4 9 0 6 0.0 1 7 3 8 4 0.015155 0 .2 0 6 9 4 6
(0.004470)“ (0.004168)** (0.082594) (0.102076)*
Log # Emp -0.006387 -0.010880 -0.075979 -0 .2 8 1 9 9 4
(0.003111)* (0.004332)* (0.070388) (0.109871)*
Constant 0.0 0 9 3 4 2 0.002076 4.555915 3 .2 0 4 4 9 9
(0.032762) (0.030817) (0.472723)** (0 .6 0 3 9 5 1 )“
Industry Dummies No Y es No Y es
Year Dummies Y es Y es Y es Y es
Observations 2684 2684 2201 2201
R-squared 0.10 0.16 0.19 0.27
Robust standard errors clustered by firm in parentheses

* significant at 5 % level; ** significant at 1% level
Table 1-6] Pooled OLS Regression Results of Financial Performance on Social Performance
32
The paper and publishing, chemicals and pharmaceuticals, and hospital
management industries display a positive and significant association between the KLD
index and the measure of financial performance, measured either with ROA or Tobin’s q.
The association between CSP and CFP is positive and significant at the 5% level for the
computers and precision products industry when the dependent variable is ROA, but the
association is not significant at the 5% level when the dependent variable is Tobin’s q.
The richness of my data enables me to parse out the differential association between CSP
and CFP across industries, and the reasons behind these cross-industry differences is an
interesting subject for further research. I also ran the regressions with the firm fixed
effects, but none of the associations between CSP and CFP turn out to be significant.
33
(1) F-test (2) F-test
D ependent Variable ROA Tobin's q
KLD Index (t-1) -0 .0 0 2 7 3 0 -0 .0 4 5 3 1 2
(0.001749) (0.048740)
KLD Index (t-1 )x 0.002721 0 .0 4 6 5 9 9
Food, Textile, Apparel (0.005222) (0.105152)
KLD Index (t-1 )x 0 .0 1 5 4 5 0 F = 6 .04 0 .2 6 2 7 4 8 F = 6 .68
Paper & Publishing (0.005478)** (P=0.014) (0.097189)** (P=0.010)
KLD Index (t-1)x 0 .0 1 1 6 0 9 F = 2 0.10 0 .2 0 0 1 5 6 F = 12.83
C hem icals & Pharmaceutical (0.002685)** (P<0.001) (0.066630)** (P<0.001)
KLD Index (t-1 )x 0 .0 0 4 9 5 7 0 .0 7 7 5 5 4
Refining, Rubber, Plastic (0.002341)* (0.085498)
KLD Index (t-1 )x 0 .0 0 2 9 8 2 0 .0 6 3 6 0 7
Heavy Manufacturing (0.002483) (0.071035)
KLD Index (t-1 )x 0 .0 0 8 7 4 3 F = 5.1 0 0 .1 2 1 1 3 7
Computers (0.003182)** (P<0.024) (0.065213)
KLD Index (t-1 )x -0.0 0 0 8 8 7 0 .0 4 6 8 3 2
Auto & A erospace (0.003055) (0.056556)
KLD Index (t-1 )x 0 .0 0 3 3 1 0 0 .1 0 1 8 5 6
Transportation S ervices (0.002066) (0.059596)
KLD Index (t-1 )x 0 .0 0 0 2 2 8 0 .0 3 8 2 0 3
T elephone & Utilities (0.004022) (0.052357)
KLD Index (t-1 )x 0 .0 0 4 4 4 7 0 .2 0 0 7 7 3
W holesale & Retail (0.002479) (0.095736)*
KLD Index (t-1 )x 0 .0 0 1 7 1 3 0 .0 3 4 2 1 2
Bank & Financial S ervices (0.001949) (0.053503)
KLD Index (t-1 )x 0 .0 1 1 6 4 0 0 .2 7 0 7 5 6
Hotel & Entertainment (0.006325) (0.158383)
KLD Index (t-1 )x 0 .0 2 8 0 5 3 F = 6.59 0 .4 1 6 4 1 5 F = 7 .384
Hospital M anagem ent (0.009972)** (P<0.011) (0.144884)** (P=0.007)
Debt Ratio -0 .1 0 1 4 8 2 -2.389431
(0.022374)** (0.504965)**
Log Net S a les 0 .0 1 5 1 2 8 0 .1 8 6 9 9 6
(0.004224)** (0.103874)
Log # E m ployees -0 .0 0 9 2 4 0 -0.2 6 3 2 7 0
(0.004027)* (0.101507)**
Constant 0 .0 0 2 1 2 2 2 .3 0 5 0 9 7
(0.030901) (0.632437)**
Industry Dum m ies Y es Y es
Year Dum m ies Y es Y es
O bservations 2684 2201
R-squared 0 .18 0.29
Robust standard errors clustered by firm in parentheses; * significant at 5% level; ** significant at 1% level.
Omitted industry category is Mining and Construction. Only those F-test results that are significant at 5%
level are reported.
Table 1-7] Regression Results of CFP on CSP by Industry
4.5. Fixed Effect Regression Results
Table 1-8 reports the regression results of CFP measures such as ROA and
Tobin’s q on the KLD index. In both pairs of regressions reported in Table 1-8, the
Hausman test statistics are highly significant, and I reject the null hypothesis that the
fixed effect specification and the random effect specification lead to the same coefficient
estimates. In this case, only the fixed effect model yields consistent coefficient estimates
(Wooldridge, 2002). In column (1), the coefficient estimate of the lagged KLD index on
ROA in the fixed effect specification is negative, but it is not significant at the 5% level.
In column (2), the coefficient estimate of the lagged KLD index on Tobin’s q in the fixed
effect estimation is negative and significant at the 5% level. These results imply that after
controlling for unobserved firm-level heterogeneity, the positive relationship between the
CSP and CFP has disappeared, and possibly has been replaced by a negative relationship,
confirming my second hypothesis. The fixed effect specification implies that achieving a
better-than-firm-average level of social performance is associated with a worse-than-
firm-average level of financial performance, which presents a stark contrast to the
existing literature. I infer from these results that the positive association between CSP and
CFP is the result of unobserved differences in resources and capabilities among firms. In
other words, it is the resources and capabilities within a firm that affect both social and
financial performances.
35
(1) (2)
Dep. Variable: ROA Tobin's q
KLD Index (t-1) -0.001553 -0.054445
(0.001172) (0.025677)*
Debt Ratio -0.177316 0.0 7 9 4 1 9
(0.040942)** (0.430486)
Log Net S a le s 0 .0 7 5 8 0 2 0 .8 5 3 9 1 4
(0.013451)** (0.232670)**
Log # Emp -0.0 7 9 2 0 9 -1.437219
(0.014513)** (0.310682)**
Constant -0.227261 -1.209502
(0.086175)** (1.649159)
Firm Dum m ies Y es Y es
Year Dum m ies Y es Y es
Observations 2684 2201
Number of Firms 509 440
R-squared 0 .65 0 .75
Hausm an Test 77.51 (P<.001) 2 5 .2 3 (P c.001)
* significant at 5% level; ** significant at 1% level
Table 1-8] Fixed Effect Regression Results of ROA and Tobin’s q on the KLD Index
4.6. Addressing the Endogeneity Concern Further - IV Estimation Results
The fixed effect specification takes care of the endogeneity concern that is caused
by the fixed firm-specific portion of the error term. However, there still remains the
concern that the firm- and year-specific unobserved heterogeneity is present, thereby
making my estimates inconsistent. An instrumental variable approach can help resolve
this concern (Wooldridge, 2002). If there exists a variable that is highly correlated with
CSP but not directly affected by an individual firm’s CFP, that variable can be used in the
estimation as an instrument for CSP.
36
I claim that the enactment of the Sarbanes-Oxley Act (SOX) can serve as such an
instrument. The Sarbanes-Oxley Act took effect in July 2002, as a result of a bipartisan
effort by the U.S. Congress to increase the transparency and accountability of corporate
accounting and governance practices in the aftermath of large corporate scandals, such as
those at Enron, Tyco and WorldCom (Romano, 2005). I coded an indicator variable as 1
if the observation is from the period after the SOX legislation. Therefore, the
observations from 2002 to 2005 are assigned the SOX value of 1, and the observations
from 1998 to 2001 are assigned the SOX value of 0. The enactment of SOX shows a
negative correlation with CSP measures for the S&P 500 companies. The raw correlation
between SOX legislation and the KLD index is -0.0604, and the correlation is significant
at the 1% level. After breaking down this correlation into the positive and negative part of
the KLD index, I find that SOX legislation has a correlation coefficient of 0.0850 with
the KLD positive ratings index and a correlation coefficient of 0.1608 with the KLD
negative ratings index. Both correlation coefficients are significant at the 0.01% level.
The fact that SOX legislation is so highly correlated with the KLD negative ratings index
can be interpreted as evidence of heightened public awareness on corporate actions, and
ensuing media coverage. In particular, the enactment of SOX shows a correlation
coefficient of 0.1619 with the number of corporate governance concerns in the KLD
index (which is significant at the 0.01% level).
Table 1-9 reports the two-stage least squares estimation results with the enactment
of SOX as an instrument for the KLD index.25 Columns (1) and (2) show the results of
25 Because contemporaneous endogeneity is not a concern under this instrumental variable specification,
the KLD index variable is not lagged in these results.
37
first-stage regressions, with column (1) reporting the regression results without year
dummies or firm dummies, and column (2) reporting the regression results with both year
dummies and firm dummies. In column (1), the regression results show that SOX
legislation is negatively correlated with the KLD index after controlling for industry, debt
ratio, log of sales, and log of number of employees. In column (2), SOX legislation is
negatively correlated with the KLD index after controlling for firm fixed effects, year
dummies, debt ratio, log of sales, and log of number of employees, even though the
coefficient is not significant at the 5% level.26 Column (3) and (5) reports the second
stage regression results from pooled OLS specification with ROA and Tobin’s q as
dependent variable, respectively. In both regressions, the coefficient estimates of the
instrumented KLD index are positive and significant. In columns (4) and (6), second
stage regression results from firm fixed effect specification are reported, with ROA and
Tobin’s q as dependent variable, respectively. When ROA is dependent variable, the
coefficient estimate of the instrumented KLD index is negative, albeit not significant at
the 5% level. When Tobin’s q is dependent variable, the coefficient estimate of the
instrumented KLD index is negative and significant at the 5% level. These results are
consistent with the fixed effect regression results reported in table 1-8, and I claim that
these results resolve the endogeneity concern that the error term after removing the firm
fixed effect might be correlated with the KLD index.
26 Because the variable SOX is collinear with the entire set o f year dummies, year dummy for 2002 was
dropped in estimation.
38
Reproduced
with permission
(1) (2) (3) (4) (5) (6)

Dep. Variable: KLD Index KLD Index ROA ROA Tobin's q Tobin's q
SOX -0 .4 1 7 2 6 0 -0 .1 8 9 6 9 2
(0.098093)** (0.129355)
KLD Index 0 .0 2 1 7 8 4 -0.0 2 0 7 6 0 1.3 7 5 5 6 0 -0.240101
(0.007242)** (0.012652) (0.244088)** (0.070048)**
Debt Ratio -0 .6 1 3 9 3 4 -0.9 3 6 1 0 0 -0.0 9 0 3 7 8 -0.188081 -1 .6 3 7 7 7 2 0.340681
(0.640126) (0.509121) (0.023449)** (0.047658)** (0.540204)** (0.605838)
Log Net S a les 0 .0 1 8 0 3 0 -0 .3 1 3 2 9 3 0 .0 1 5 3 9 5 0.056411 0 .2 0 4 2 8 9 0 .6 1 2 9 9 6
(0.182006) (0.279782) (0.004046)** (0.013841)** (0.104696) (0.219516)**
Log # Emp 0 .0 7 3 0 9 3 0 .3 5 2 3 7 7 -0 .0 1 0 8 7 5 -0.055111 -0 .4 5 0 5 3 9 -1.2 9 6 9 0 5
(0.180802) (0.308743) (0.004447)* (0.016146)** (0.126936)** (0.329502)**
Constant -2 .2 7 7 5 9 5 1.8 2 8 7 2 3 0 .0 4 8 4 9 6 -0 .0 8 9 5 2 7 5.356781 -0.6 4 9 6 6 8
(1.261755) (1.948948) (0.031332) (0.171876) (0.732881)** (1.992931)
Ind. Dum m ies Y es N/A Y es N/A Y es N/A
Firm Dum m ies No Y es No Y es No Y es
Year Dum m ies No Y es No Y es No Y es

Observations 3198 3198 3198 3198 2621 2621
No. of Firms 526 526 4 54
R-squared 0.10 0.82 0.11 0.52 0 .25 0.71
Hausm an T est 83.01 (P c.001) 8 5 .9 2 (P c.001)
Table 1-9] 2SLS Regression Results with SOX as an Instrumental Variable for the KLD Index
39
4.7. Asymmetry between the Effect of Positive and Negative CSP
So far in my regression analysis, I have used the KLD index, which is the
combination of all positive CSP items and all negative CSP items. However, given that
positive CSR activities are often the conscious effort by the management in trying to
mitigate the impact of negative CSR incidents, positive KLD ratings and negative KLD
ratings may be associated with a firm’s financial performance in an asymmetric way. One
feature of the KLD data that is of particular interest is the fact that positive CSP (strength)
items and negative CSP (concern) items are separately recorded.
In Table 1-10,1 run pooled OLS and firm fixed effect regressions of the financial
performance measures on the lagged KLD positive and negative ratings indices. In
column (1), with ROA as the dependent variable, the coefficient on the lagged KLD
positive ratings index is very close to zero and not significant at all. On the contrary, the
coefficient on the lagged KLD negative ratings index is negative and highly significant at
the 1% level. This result suggests that the positive association between CSP and CFP is
driven by the negative side of the CSP. The companies that suffer less from negative
social performance tend to have higher ROA than the companies that suffer more from
negative social performance, while the companies boasting high positive social
performance do not have any higher ROA than companies with low positive social
performance. In column (3), with Tobin’s q as the dependent variable, the coefficient on
the lagged KLD positive ratings index is positive and significant at the 5% level, and the
coefficient on the lagged KLD negative ratings index is negative and significant at the
40
1% level. Even though higher KLD positive ratings index is associated with a higher
Tobin’s q in the next period, the effect of preventing negative social performance seems
to be much more strongly correlated with a higher Tobin’s q values, as the size of the
coefficient is larger and the significance level is higher. These results confirm my third
hypothesis that the positive association between CSP and CFP is driven mainly by the
negative CSP ratings.
Once again, though, these results change with the fixed effect regression. In
column (2), with ROA as the dependent variable, the coefficient of the lagged KLD
positive ratings index is negative and significant at the 5% level, while the coefficient of
the lagged KLD negative ratings index is not significant at all. I obtain the same result
when I use Tobin’s q as the dependent variable. In column (4), the coefficient of the
lagged KLD positive ratings index is negative and significant at the 1% level, while the
coefficient of the lagged KLD negative ratings index is not significant at the 5% level.
These results confirm my fourth hypothesis that the negative association between CSP
and CFP, after controlling for the firm level heterogeneity, is driven by the positive CSP
ratings. The Hausman test results suggest that a fixed effect model is the right
specification in both column (2) and column (4). Using an F-test, I also test the
hypothesis that the coefficients of the KLD positive ratings and those of the KLD
negative ratings sum to zero, which would imply no asymmetry. The F-test statistics are
significant at the 5% level except for column (3), when I run the pooled OLS regression
with Tobin’s q as the dependent variable. I reject the null hypothesis of no asymmetry in
other three regressions.
41
Reproduced
with permission
(1) (2) (3) (4)

Dep. Variable ROA ROA Tobin’s q Tobin’s q
KLD Positive 0 .0 0 0 6 8 7 -0 .0 0 4 2 6 3 0.0 5 6 8 7 5 -0 .1 3 0 9 9 6
Ratings (t-1) (0.001321) (0.001725)* (0.026840)* (0.049001)**
KLD N egative -0.00 5 6 1 3 -0 .0 0 0 9 7 2 -0.083162 -0.017041
Ratings (t-1) (0.001353)** (0.001431) (0.022292)** (0.033599)
Debt Ratio -0.10 2 1 3 6 -0 .1 7 8 8 5 5 -2 .4 8 3 0 4 7 0 .0 7 3 4 4 2
(0.022350)** (0.040335)** (0.490677)** (0.419370)
Log Net S a les 0.0 2 2 2 2 4 0 .0 7 4 9 9 3 0.2 3 2 9 8 8 0 .8 2 1 3 8 8
(0.005108)** (0.013289)** (0.111363)* (0.225595)**
Log # Emp -0 .0 1 0 5 3 7 -0 .0 7 8 5 0 4 -0.281131 -1.380601
(0.004427)* (0.014076)** (0.110477)* (0.293364)**
C onstant -0.0 3 0 4 5 3 -0 .2 1 0 0 1 9 3 .0 3 0 9 8 4 -0 .6 1 2 6 9 4
(0.037341) (0.086803)* (0.666893)** (1.656961)
Ind. Dummies Y es N/A Y es N/A
Firm Dummies No Y es No Y es
Observations 2684 2684 2201 2201
# of Firms 509 440

R-squared 0.16 0 .65 0 .27 0.75
F-test 7.31 (P=0.007) 5 .7 5 (P=0.017) 0 .66 (P=0.417) 5 .0 6 (P=0.025)
H ausm an Test 1 0 4 .6 4 (P c.001) 142.51 (P c.001)
* significant at 5 % level; ** significant at 1% level
Table 1-10] Regression Results of Financial Performance on the Lagged KLD Positive and Negative Ratings Index
42
4.8. Controllable vs. Uncontrollable Components of Negative Social Ratings
I separate the KLD index items into those that the firm can directly control and
those that are out of the firm’s direct control. The seventeen items that I categorize to be
not directly under the control of the firm are: investment controversies, negative
economic impact (contamination, etc.), tax disputes, other community concern
(mobilized opposition from the community), other corporate governance concern,
diversity controversies, other diversity concern, violation of health and safety standards,
other employee relations controversies, hazardous waste, regulatory problems, other
environmental controversies, supply chain labor standards controversies, other human
rights controversies, product safety concern, marketing violations, and other product-
related controversies. These items share the common feature that an outside force - be it
a regulatory body or private citizens - is more influential than the decision of the
management. The tax disputes item, for example, is recorded to have a value of one only
when a federal or local government initiates an investigation. Similarly, the negative
economic impact item can result from a probabilistic event such as a leak from an
underground storage tank (Yin, 2006). I define two sub-indices of the KLD negative
ratings index. For the lack of better terms, I define the KLD negative ratings index
(uncontrollable) as the sum of aforementioned seventeen items, and I define the KLD
27 Under this line o f logic, all the items in the KLD positive ratings index are deemed controllable. There is
little chance that the positive items are recorded unless the company is making a conscious effort on the
matter and is trying to get public recognition.
43
negative ratings index (controllable) as the sum of the rest of the items in the KLD
negative ratings index.
When the KLD negative ratings index is broken into uncontrollable and
controllable subindices, the basic results from earlier regressions stand. Table 1-11
reports that Hausman test statistics suggest that the fixed effect specification is the correct
one to estimate, and that negative ratings indices tend to have negative and significant
coefficients only under the pooled OLS specification, and that the KLD positive ratings
index has negative and significant coefficient only under the fixed effect specification.
There is also an interesting contrast between the specification with ROA as dependent
variable and the specification with Tobin’s q as dependent variable. When ROA is used
as dependent variable, the KLD controllable negative ratings index displays a stronger
and significant association with the dependent variable (column 1). When Tobin’s q is
used as dependent variable, the KLD negative ratings index (uncontrollable) has a
stronger and significant association with the dependent variable (column 3). This contrast
seems to make sense. Because Tobin’s q reflects the volatility in the stock market, the
uncontrollable social performance items should have much greater possibility of affecting
Tobin’s q than they can affect ROA.
44
Reproduced
with permission
(1) (2) (3) (4)

Dep. Var.: ROA ROA Tobin's q Tobin's q
KLD Positive 0 .0 0 0 4 8 0 -0 .0 0 4 4 3 3 0 .0 5 8 6 0 2 -0 .1 2 6 7 4 2
Ratings Index (0.001314) (0.001796)* (0.026686)* (0.050657)*
KLD N egative -0 .0 0 2 4 0 6 0.0 0 0 5 5 9 -0 .1 1 0 4 9 6 -0 .0 6 1 8 3 7
(Uncontrollable) (0.001499) (0.002152) (0.026968)** (0.040593)
KLD N egative -0 .0 1 0 4 9 2 -0.0 0 2 4 8 3 -0 .0 4 2 8 7 3 0 .0 2 4 2 9 2
(Controllable) (0.002409)** (0.002192) (0.045295) (0.047320)
Debt Ratio -0 .1 0 5 3 5 9 -0 .1 8 0 0 1 7 -2 .4 5 4 6 4 0 0 .1 1 2 2 9 4
(0.022246)** (0.040453)** (0.489363)** (0.422926)
Log Net S ales 0 .0 2 2 2 6 0 0.0 7 5 2 0 3 0 .2 3 3 1 2 5 0 .8 0 5 6 4 0
(0.005076)** (0.013187)** (0.111223)* (0.227460)**
Log # Emp -0 .0 1 1 4 1 4 -0 .0 7 9 4 3 4 -0 .2 7 4 9 1 5 -1 .3 3 7 0 7 7
(0.004439)* (0.013919)** (0.111375)* (0.300787)**
Constant -0 .0 2 2 7 1 8 -0.2 0 7 3 5 6 2 .9 6 5 7 3 7 -0.6 4 7 0 7 0
(0.036604) (0.087430)* (0.662457)** (1.660804)
Industry Dummies Y es N/A Y es N/A

Firm Dummies No Y es No Y es
Observations 2684 2684 2201 2201
R-squared 0.17 0.65 0 .2 7 0.75
Hausm an Test 104.11 (<.001) 4 7 .8 9 (<.001)
Table 1-11] Regression Results of Financial Performance on the KLD Positive Ratings Index and the KLD Negative Ratings
Index with Uncontrollable and Controllable Parts Separated
45
5. Conclusion
In today’s corporate world, there are good companies, which have built better
capabilities than other companies. These companies can outperform their peers not only
in terms of financial performance, but also in terms of social performance. They have
better bottom lines and enjoy superior reputations. In short, the corporate world is not fair.
In this paper, I use the panel structure of the Social Ratings database that the KLD
Research Inc. has accumulated over eight years, and the fact that the KLD Social Ratings
database has recorded considerably more social concerns since the Sarbanes-Oxley Act
took effect, to identify the effect of CSP on CFP.
After controlling for firm-level unobserved heterogeneity in capabilities by
introducing firm fixed effects in the regression analysis, and using the enactment of the
Sarbanes-Oxley Act as an instrument for CSP, I find that CSP does not have a positive
effect on CFP, when CSP is measured by the KLD index, defined by the net sum of all
items in the KLD Social Ratings database. The effect is found to be negative.
In addition, I find asymmetry in how negative CSP and positive CSP are
associated with CFP. In pooled OLS regressions, negative social performance does seem
to be negatively associated with the profitability, whereas positive social performance
shows no significant correlation. This implies that preventing an event that will impact
the firm’s social performance ratings negatively is associated with better financial
performance.
46
When unobserved firm-level heterogeneity is controlled, positive social
performance is negatively associated with financial performance, whereas negative social
performance has no significant association with financial performance. This result
implies that, after controlling for the unobserved firm-level heterogeneity in capabilities,
engaging in activities that would qualify as positive social performance is associated with
lower financial performance of the firm.
In terms of methodological contribution, the results in this study suggest that
close attention must be paid to the endogeneity concerns, including the unobserved firm-
level heterogeneity in capabilities, when studying the relationship between CSP and CFP.
47
ESSAY TWO: BOARD COMMITTEES AND INTEGRATED
STRATEGY: HOW FIRMS INTEGRATE MARKET AND NON-
MARKET STRATEGIES FOR MITIGATION
(Joint with Elizabeth E. Bailey)
1. Introduction
Over the past decade, more firms have begun to consider the benefits of adopting
a systematic approach to managing issues that arise in the public sector. Basically, the
view is that firms should establish an ongoing process that identifies, measures and
mitigates public sector risks and opportunities that could result in material financial
impacts on the company (Baron, 2006). Such risks and opportunities are managed by
creating strategies about when to avoid, reduce, transfer or retain risk, and when to
promote public sector alliances and initiatives (constructive engagements and societal
alignments). Once firms have decided to adopt such polices, they must assign responsible
parties to manage the integration of public sector strategies with other market-oriented
strategies of the firm.
This research will attempt to identify empirically the circumstances in which such
an integrated strategy process is called for, and the extent to which such a process helps
improve a firm’s reputation and social performance. We will use the establishment of
public affairs (PA) committees or social responsibility (SR) committees on boards of
48
directors as a proxy for the routinization of non-market strategy (Nelson and Winter,
1982). The PA/SR committees are not mandated by the Sarbanes-Oxley Act or other
governance laws; rather, they are established voluntarily by firms. In contrast, all boards
are virtually mandated by the Act to have audit, nominating and compensation
committees and staff their directors from outside the company.
For our sample of firms in the S&P 500 Index as of December 30, 2005, we
observe more than a 34% increase over the prior six years in firms that have a PA/SR
committee. However, the majority of S&P 500 companies still do not have such
committees. It is also worth noting that the presence of a PA/SR committee varies widely
by industry.
By examining the adoption of these committees, we find that companies establish
such committees as defensive measures against negative social ratings. We also find that
the existence of such committees is positively associated with a firm’s positive social
performance rating, as well as the firm’s negative social ratings. Combined, the presence
of a PA/SR committee has no association with the overall social performance ratings. It
appears that these committees are effective in neutralizing the effect of the negative
social performance, so that their overall social ratings are not adversely affected in the
face of negative events.
The organization of this paper is as follows. In section 2, we briefly review the
existing literature of the field. In section 3, we describe our two main datasets, the board
committee data and the social performance data, in detail. In section 4, we describe how
the degree to which firms integrate their market and non-market strategies through a
board level committee is associated with their social performance by looking at data from
49
the most recent year of our sample. In section 5, we analyze how the adoption of a PA/SR
committee is related to the social performance of a firm, using panel data of S&P 500
firms from 2000 to 2005. In section 6, we conclude and offer directions for future
research.
2. Related Literature
Baron (1995) proposes the necessity of integrating market and non-market
strategies in formulating successful business strategies tailored to individual business
environments. In a further step, Baron (2006) states that formulating integrated strategies
requires consideration of the market and non-market environments, as well as attention to
moral concerns and social responsibilities (figure 2-1). He argues that companies,
especially large companies that may be targeted by social activists, need to build social
issues into strategy in a manner that reflects their potential business importance.
The structure of a business organization is reflective of its strategy (Chandler,
1969). Therefore, we postulate that the committee structure on the board can reflect the
integrated strategy framework. Some firms with serious needs for non-market strategies
may have established an extra committee that specializes in the non-market aspect of the
business environment. Firms that pay particular attention to ethical concerns and social
responsibility may have a committee that focuses on corporate social responsibility. At
the board level, these committees help integrate non-market and ethical issues with
market environment issues. From the resource dependence perspective, having a
50
specialized board-level committee can better provide advice and counsel related to the
non-market environment and ethical considerations (Hillman & Dalziel, 2003).
Integrated Strategy Framework

S o cial R esp on sib ility
E thics P rinciples
Issues
M a rk e t S tructure Interests
C o m p e tito rs Institutions
B randing Inform ation
M a rk e t positioning N o n m a rk e t
positioning
Figure 2-1] Integrated Strategy Framework (Baron, 2006. p. 654)
Although we find no empirical literature that measures the impact of integrated
strategy on firm performance, we would like to build on the body of work that
investigates the effect of the composition of the board on corporate decision-making and
performance. Dalton, Daly et al. (1998) conduct a meta-analysis of the relationship
between the composition and leadership structure of corporate boards and financial
performance. They review 54 existing empirical studies of board composition and 31
51
studies of board leadership structure. They conclude that the existing studies provide little
evidence of a systematic relationship between a board’s composition and leadership
structure and the firm’s financial performance. They also deduce that one potentially
promising avenue for future research might be the relationship between board
subcommittee composition and financial performance, since many critical decisions are
made at the subcommittee level.
In their study of golden parachutes, Singh and Harianto (1989) also emphasized
the importance of committees on a board. They state that:
“One o f the most important means fo r facilitating the decision-making process o f
the board is the creation o f various committees such as executive, audit, executive
28
compensation and other ad-hoc committees. ”
For the purpose of our paper, the relevant board-level committees are public
affairs and/or social responsibility committees, which deal with formulation of integrated
strategy and CSR issues.
Hillman, Keim et al. (2001) examine the link between board composition and
stakeholder relations among 250 of S&P 500 companies. Their measure of board
composition is the number of stakeholder directors on corporate boards, representing
such stakeholders as community, employee, customers, and suppliers. For five categories
of stakeholder relations including community relations, employee relations, customer
relations, environment, and diversity, their measure of stakeholder relations is social
28 Singh and Harianto, 1989, p. 147.

52
performance ratings compiled by KLD. They find that overall, the number of stakeholder
directors is not significantly correlated with stakeholder performance. When they look
into the four different stakeholder directors separately, only the number of community
directors shows some significant associations: it is positively correlated with stakeholder
performance in diversity, and it is negatively correlated with stakeholder performance in
environment. They conclude that the presence of stakeholder directors does not make
much difference in terms of stakeholder performance.
Klein (1998) studies the composition of boards of directors and concludes that
firms with a higher percentage of inside directors on their finance and investment
committees achieve better financial performance. Adams (2003) finds that, while
corporate boards appear to spend most of their effort on monitoring, boards of older,
larger and growing firms devote more effort to shareholder interests.
Other relevant literature covers the effect of political and non-market strategies of
the firm on corporate performance. In an event study, Hillman, Zardkoohi et al. (1999)
finds that firms can gain firm-specific benefits from political strategies. In their study,
firms enjoy positive abnormal returns when a link is established between a firm and the
government through a cabinet-level appointment or Congressional election. Shaffer,
Quasney et al. (2000) studies the airline industry and suggests that non-market actions
have a positive and significant impact on performance, measured by profits, market share
and capacity utilization.
The association between the committee presence, as a proxy for integrated
strategy, and corporate financial performance is an important question. We do not have
sufficient evidence to report that the practice of integrated strategy is significantly
53
associated with corporate financial performance. Instead, we focus on corporate social
performance. In particular, we look at positive and negative social performance ratings
separately, as the first chapter of this dissertation documents evidence of an asymmetry in
how positive social performance ratings and negative social performance ratings are
associated with corporate financial performance.
3. Data
3.1. Committee Data
Our sample of firms includes the companies in the S&P 500 Index as of
December 30, 2005. For these companies, we collect the committee information for the
years 2000 through 2005 by reading the proxy statements of the companies. We gather
the social performance information from the KLD Social Ratings database and gather the
financial information of these companies, which we use as control variables in our
regression analysis, from the merged CRSP/Compustat database.
We categorize policy, public policy, public affairs, public issues, and public
interest committees as public affairs committees. There are environment, safety and
public policy, and public responsibility committees, which we also classify as public
affairs committees. Sometimes, firms assign the functions of a public affairs committee
within the compliance, audit or corporate governance committees. When in doubt, we
54
determine whether a committee qualifies as a public affairs committee based on the
company’s proxy statement.
We categorize the following committees as “social responsibility committees:”
corporate responsibility, social responsibility, public responsibility, corporate issues and
governance, and ethics committees.
To illustrate what these committees do, we present the following three examples
of such committees from the proxy statements that companies filed with the Securities
and Exchange Commission (SEC): CSX, McDonald’s and Merck. Even though each
company faces distinctive public policy issues, such as public safety (CSX), obesity and
health concerns (McDonald’s) and product safety issues (Merck), they are the most likely
targets of activists for two main reasons. First, as industry leaders, they have clear name
recognition and therefore can attract greater attention to activist causes; and second, their
products and services can be substituted by rival companies’ offerings within the same
industry.
In the proxy statement of the CSX Corporation filed on Mar. 30, 2006, the
company reports that it has six committees on its board: executive, audit, compensation,
finance, governance and public affairs. The proxy statement defines the responsibilities
of the Public Affairs Committee as follows:
“The Public Affairs Committee reviews and makes recommendations concerning
the Com pany’s practices and program s designed to address important public policy
issues that may impact the Company, its shareholders, and the general public.”
55
In the proxy statement of the McDonald’s Corporation filed on Apr. 7, 2006, the
company reports that it has six committees on its board: audit, compensation, governance,
corporate responsibility, finance and executive. The proxy statement defines the role of
the Corporate Responsibility Committee as follows:
“The Corporate Responsibility Committee acts in an advisory capacity to the
Com pany’s management with respect to policies and strategies that affect the Com pany’s
role as a socially responsible organization, including issues pertaining to health and
safety, the environment, employee opportunities, consumers and the communities in
which the Company does business. ’’
In addition, in the proxy statement of Merck & Co. Inc. filed on Mar. 9, 2006, the
company reports that it has seven committees on its board: audit, compensation and
benefits, corporate governance, public policy and social responsibility, executive, finance,
and research. The proxy statement defines the role of the Public Policy and Social
Responsibility Committee as follows:
“The Committee on Public Policy and Social Responsibility, which is comprised
o f independent directors, advises the Board o f Directors and m anagem ent on Company
policies and practices that pertain to the C om pany’s responsibilities as a global
corporate citizen, its obligations as a pharm aceutical company whose products and
services affect health and quality o f life around the world, and its commitment to high
standards o f ethics and integrity. It reviews social, political and economic trends that
56
affect the Com pany’s business; reviews the positions and strategies that the Company
pursues to influence public policy; monitors and evaluates the Com pany’s corporate
citizenship program s and activities; and reviews legislative, regulatory, privacy and
other matters that could impact the Com pany’s stockholders, customers, employees and
communities in which it operates. ”
From these statements, we can infer that many boards of directors consider the
impact of issues relating to public policy and social responsibility in their strategic
planning, and that such thinking is consistent with good corporate governance and
fiduciary duty.
CSX’s experience illustrates one mechanism that may lead to the establishment of
a PA/SR committee, as well as the committee’s usefulness. CSX Corporation suffered a
jury verdict in excess of $2 billion including punitive damages after a toxic release
incident took place near New Orleans in 1997, in which there was de minimus harm or
injury. An analysis of the lessons learned from the experience led the company to
routinize the handling of non-market issues, which led to the establishment of a Public
Affairs Committee in July 2001. The lessons learned also helped the company in
handling a Baltimore tunnel fire in 2001, in which the company’s handling of the matter
earned public praise.
One of the empirical findings of this paper is that no company in our sample has
two separate committees, one on public affairs and the other on social responsibility, on
their board. Out of 89 companies that had either committee by the end of 2005, 44 had a
public affairs committee, 23 had a social responsibility committee, and 22 had a
57
combined “public responsibility” committee. This implies that the distinction between the
roles of two committees, theorized in Baron’s framework, is not clearly established in the
business practices of S&P 500 companies. We conjecture that the reason that no company
has actually separate PA and SR committees is the difficulty associated with having too
many board-level committees, given that three are already virtually mandated. Even the
largest companies rarely have more than five or six board-level committees. Within the
combined committee, board members are aware of the difference between the two and
adopt appropriate strategies for each.
The presence of a PA/SR committee shows substantial variation across industries.
Among the railroad companies, for example, CSX has a Public Affairs Committee, while
Norfolk Southern does not have such a committee. Among the tobacco companies, Altria
has its Public Affairs and Social Responsibility Committee, while Reynolds American
Inc. does not have an analogous one. We categorize the industry broadly using the two-
digit Global Industry Classification Standard (GICS) used in classifying companies in the
S&P Indices. According to our industry classification, Telecommunication Services,
Industrials, and Consumer Staples industries have a relatively higher percentage of firms
with a PA/SR committee, whereas Information Technology, Financials, and Consumer
Discretionary industries have a lower percentage. Table 2-1 displays the breakdown of
the prevalence of companies with a PA committee, with a SR committee and with either
of the two committees by industry.
58
3.2. Social Performance Data
We use the social performance data from the KLD Social Ratings database, which
is published by KLD Research & Analytics, Inc. As discussed in the first essay, the KLD
Social Ratings data are widely regarded as the most influential measure of CSP (Harrison
and Freeman, 1999; Waddock and Graves, 1997a; Berman et al., 1999). Compared with
other frequently used social performance measures, such as reputation scores, the KLD
Social Ratings data offer two advantages. First, the database consists of many categories,
which remain relatively stable over time. Therefore, an index based on the KLD
database is more analytical in nature. Second, the KLD database distinguishes between
positive social performance and negative social performance. Since the process that
generates positive social performance can be quite different from the process that
generates negative social performance, the ability to investigate the two separately
presents a researcher with an opportunity to analyze the relationship between committee
presence and CSP in greater detail.
29 Over the 6-year sample period, some, though not many, indicators do switch categories.
59
Reproduced
with permission
No. of
Public Affairs Social Responsibility Any O ne of the Two
GICS Industry C om panies
N Percentage N P ercentage N P ercentage
10 Energy 29 5 17.24% 1 3.45% 6 20.69%

15 Materials 31 8 25.81% 2 6.45% 8 25.81%
20 Industrials 53 8 15.09% 7 13.21% 14 26.42%
25 Consum er Discretionary 90 5 5.56% 12 13.33% 15 16.67%
30 C onsum er S tap les 38 8 21.05% 7 18.42% 10 26.32%
35 Health Care 57 8 14.04% 5 8.77% 11 19.30%
40 Financials 84 12 14.29% 8 9.52% 13 15.48%
45 Information T echnology 78 2 2.56% 1 1.28% 2 2.56%
50 Telecom m unications S ervices 8 3 37.50% 0 0.00% 3 37.50%
55 Utilities 32 7 21.88% 2 6.25% 7 21.88%
Total 500 66 13.20% 45 9.00% 89 17.80%
Table 2-1] Percentage of Companies with a PA/SR Committee in 2005, by Industry
60
The KLD data cover approximately 80 indicators in seven major qualitative issue
areas: community, corporate governance, diversity, employee relations, environment,
human rights, and product quality and safety. In addition, KLD data provide information
for involvement in the following “controversial business issues:” alcohol, gambling,
firearms, military, nuclear power, and tobacco. Business involvement in any of these
industries results in a negative indicator. Each area has a number of strength and concern
items, where a binary measure indicates the presence or absence of that particular
strength or concern. For example, the employee relations category contains six strength
items (union relations strength, cash profit sharing, employee involvement, strong
retirement benefits, health and safety strength, and an item for other strength) and five
concern items (union relations concerns, safety controversies, workforce reductions,

30
pension/benefits concerns, and an item for other concerns).
We define the KLD positive ratings index as the sum of all strength items in the
KLD Social Ratings Database. Likewise, we define the KLD negative ratings index as
the sum of all concern items in the database. We then define the KLD overall ratings
index as the sum of all strength items minus the sum of all concern items. This simple
manipulation has an advantage over more complex methods that bestow different weights
on different categories: it is not affected by some indicators’ switching between
categories.
Table 2-2 presents the summary statistics for the three KLD indices of the
companies that had a PA/SR committee and the summary statistics for the companies that
did not in 2005. After we merge the KLD data with our S&P 500 board committee data,
30 The complete list o f social ratings items in the KLD database can be found in the appendix.
61
we are left with 486 firms in our sample from 2005. At first glance, the KLD overall
ratings index is no different between the two groups of companies. The mean difference
is small and statistically insignificant. If we consider the KLD overall ratings index as the
sole measure of corporate social performance, we might conclude that the presence of
these committees is not associated with the companies’ social performance.
PA/SR N Mean SD Min Max Difference
KLD No 398 -0.2261 3 .1526 -10 11 0.1261

Overall Y es 88 -0.3523 3.9481 -9 9 (P=0.7464)
Ratings
Total 486 -0.2490 3 .3 0 6 6 -10 11
KLD No 398 2 .9 8 2 4 2 .8 5 6 3 0 18 -2.6994

Positive Y es 88 5.6 8 1 8 3 .7 4 0 3 0 14 (P c.001)
Ratings
Total 486 3 .4 7 1 2 3 .2 0 4 7 0 18
KLD No 398 3.2 0 8 5 2 .2 2 7 4 0 11 -2.8255

N egative Y es 88 6.0341 3 .5 9 2 6 0 16 (P c.001)
Ratings
Total 486 3 .7 2 0 2 2 .7 5 0 0 0 16
Table 2-2] KLD Positive Ratings, KLD Negative Ratings, and KLD Overall Ratings
Indices by the Presence of a PA or SR Committee (2005)
However, a deeper look at the data tells a different story. Companies with a
PA/SR committee have significantly higher KLD positive ratings index scores, as well as
significantly higher KLD negative ratings index scores. The companies with a PA/SR
committee on average have a KLD positive ratings index score that is 2.70 higher than
that of their counterparts with no such committee, and they have on average a KLD
negative ratings index score that is 2.83 higher than their counterparts with no such
62
committee. The differences are significant at the 0.1% levels. The size of the differences
is substantial, given that the sizes are almost as large as the standard deviations of the
respective ratings indices. This result implies that the companies with a PA/SR
committee may be subject to a more heavily scrutinized business environment, due either
to the nature of their product or the nature of their business processes. They may need to
integrate their market and non-market strategies better to navigate their environment. In
the next section, we analyze the relationship between the presence of a PA/SR committee
and corporate social performance in more detail. However, we caution readers that those
companies with a PA/SR committee and those without differ along some important
dimensions as well. Companies with such committees seem to possess more assets, have
larger sales revenue, employ more workers, have higher debt ratios and are larger in
terms of market capitalization.
Firms with a PA/SR committee have on average 68% debt ratio compared to 60%
debt ratio of the firms without such a committee. They have almost three times the asset
and sales, almost twice the number of employees, and 2.4 times the market capitalization.
These differences are all significant at the 1% level. We therefore control for these
differences in our regression analysis.31 The only characteristic in table 2-3 that does not
show a significant difference between the two groups is ROA. Even though the firms
with a PA/SR committee have a ROA that is about 10% lower (0.0609, compared to
0.0675), the difference is not statistically significant at the 5% level. Table 2-3 present
the summary statistics.
31 Because o f the large skew ness in accounting and financial market data, we use the natural logarithm o f
asset, sales, market capitalization, and employment in our regressions.
63
Reproduced
with permission
No PA / SR Comm ittee Firms with PA / SR Comm ittee All Firms

Variable Mean SD N Mean SD N Mean SD N
Debt Ratio 0.60 0.22 380 0.68 0.1 7 84 0.61 0.22 464
Total A sset ($M) 3 5 9 5 5 .1 2 1106 6 7 .2 0 381 109699.20 2 4 9 5 3 5 .6 0 84 492 7 6 .6 3 1 4 8 2 4 0 .2 0 465
S a le s ($M) 122 4 5 .1 6 2 1 5 4 0 .3 4 381 3 5 9 4 7 .5 2 5 3 1 4 7 .7 3 84 16526.88 3 1 1 2 2 .3 2 465
Market Cap. ($M) 188 0 7 .7 0 2 8 9 1 3 .8 6 378 4 4 8 0 1 .0 8 6 6 2 2 7 .7 6 83 2 3 4 8 7 .6 4 3 9 5 8 5 .3 6 461
Employment ('000) 4 0 .9 8 10 6 .0 4 374 8 0 .9 8 9 3.29 83 48.25 104.88 457
ROA 0.0 6 7 5 0 .0 7 4 2 381 0 .0609 0.0 5 0 6 84 0.0663 0 .0 7 0 6 465
A ll variables except for ROA are significantly larger for firms with a PA/SR committee at the 1% levels.
Table 2-3] Summary Statistics for Firm Characteristics (2005)

64
4. Analysis & Results
4.1. Descriptive Analysis
Depending on the magnitude of their social performance ratings, we classify the
Standard & Poor’s ten industry sectors into four distinct categories: industries that exhibit
low positive social performance ratings and low negative social performance ratings
(Low-Low); industries that exhibit low positive social performance ratings but high
negative social performance ratings (Low-High); industries that exhibit high positive
social performance ratings but low negative social performance ratings (High-Low); and
industries that exhibit high ratings in both positive and negative social performance
(High-High). In 2005, the Low-High category includes the Energy and Utilities sectors,
and the High-Low category includes the Information Technology, Financials, Consumer
Discretionary and Health Care sectors. Finally, the High-High category includes the
Materials, Consumer Staples, Telecom and Industrials sectors (figure 2-2). No sector
appears to fall into the Low-Low category.32
When we look at the percentage of corporations that have a PA/SR committee on
their boards, we find an interesting pattern. The fraction of companies with those
committees is low for the sectors in the High-Low category. Each of four sectors in the
category shows the fraction below 20%. The Information Technology sector has the
32 In 2005, the cutoff for high negative ratings index is 4, and the cutoff for high positive ratings is 3, as
they are defined as the mean o f the indices rounded up to an integer.
65
lowest fraction: only two firms (Coming Inc. and Lexmark International Inc.) out of 78
have a PA/SR committee.
CO
05 • M aterials (25.8%)
c
• IT (2.6%)
CO
CC • C onsum er Staples (26.3%)
0)
>
• Telecom (37.5%)
'<75 • Financials (15.9% ) e Industrials (26.4%)

0
Q. • C onsum er D iscretionary (16.7%)
d)
ci*
U. c'o • Health Care (19.3% )
CD • U tilities (21.9%)
1
• Energy (20.7%)
CM -
—r~
3 4 5
A verage N egative Ratings
Numbers in parentheses indicate the fraction o f companies with a PA/SR committee on the board.
Figure 2-2] Average KLD Negative and Positive Ratings Indices by Industry
The fraction of companies with those committees is high for the industries in the
High-High category. Each of four sectors in the category shows the fraction above 25%.
The industries in the Low-High category show an intermediate frequency of those
committees, with 21.9% of Utilities sector firms and 20.7% of Energy sector firms having
one of those committees.
This result could be interpreted to suggest that a corporation establishes a PA/SR
committee when it faces the business environment that generates high ratings in negative
66
social performance, and once established, such a committee helps improve the firm’s
positive social performance ratings.
00
♦ C onsum er Discretionary
♦ Telecom
(0
o> ♦ Financials
-E CD ♦ Industrials
as ♦ M aterials
CC ♦ IT
a>> ♦ C onsum er Staples
'■M
'(/>
o ♦ Health Care
Q.
0>
O )
CO r f ♦ Energy
O IT
<
>
D ♦ Utilities
O Materials
< O C onsum er Staples
o Financials
© Health Care © Utilities
O C onsum er Di;
rs-ona
CM - © Energy
© Telecom
6 8 10 12
A verage N egative Ratings
Solid diamonds represent the average among companies with a PA/SR committee, and hollow diamonds
represent the average among companies without a PA/SR committee.
Figure 2-3] Within-industry Differences in Average Negative and Positive Ratings

Indices
Figure 2-3 looks at the difference in KLD positive ratings and KLD negative
ratings within each industry in 2005. Within virtually all industries, the companies with a
PA/SR committee have on average a higher KLD positive ratings index and a higher
KLD negative ratings index. While the majority of the differences are statistically
significant, the differences in the energy and the consumer discretionary sectors are
67
particularly large. In the following subsection, we will show that these differences are
statistically significant even after controlling for firm characteristics.
4.2. Regression Analysis
Table 2-4 presents the regression results of the KLD overall, positive, negative
ratings index on the presence of a PA/SR committee for 2005. The specification of our
empirical model is as follows: For each company i,
(Social Performance), =
a + J3* (Committee), + y* (Firm Characteristics), + <5*(lndustry Dummies), + £,
£■, satisfies standard OLS assumptions. We calculate the Huber-White robust
standard errors in order to account for the possibility that companies belonging to
different industries have different variances for the error term.
The results in columns (1) and (2) of table 2-4 show that a company’s KLD
overall ratings index is not significantly correlated with the presence of a PA/SR
committee. The coefficient of the committee variable is not significant in either of the
specifications. However, columns (3) through (6) show that companies with a PA/SR
committee on their boards show both higher positive KLD ratings and higher negative
KLD ratings, after controlling for size, leverage and industry-fixed effects. In other words,
companies with a PA/SR committee have higher positive social performance ratings, as
well as higher negative social performance ratings, within each industry.
68
Columns (3) and (4) show the association between a PA/SR committee and the
KLD positive ratings index. Column (3) includes debt ratio, log of size of employment,
log of sales, log of total assets and ROA as firm characteristics, but it does not include
industry fixed effects. Column (4) repeats the estimation with industry-fixed effects, and
the results show that controlling for industry does not change either the size of the
coefficient or its significance level. Columns (5) and (6) report the regression results
using the KLD negative ratings index as the dependent variable, and the coefficients of
the committee variable are positive and significant at the 1% level in both regressions.
Do these results mean that having a specialized committee to integrate market
strategy, non-market strategy and ethical concerns does not affect a company’s social
performance at all? We do not believe so. Rather, we interpret this result as evidence that
companies learn to mitigate negative social outcomes by establishing a PA/SR committee
and integrating market and non-market strategies. We turn our attention to the adoption
of a PA/SR committee in the next section.
69
(1) (2) (3) (4) (5) (6)

D ependent KLD Overall KLD Overall KLD Positive KLD Positive KLD N egative KLD N egative
Variable: Ratings Ratings Ratings Ratings Ratings Ratings
PA/SR -0 .4 1 5 2 5 5 -0.085103 1 .3 9 9 0 5 7 1 .406566 1.814312 1 .4 9 1 6 6 8
Committee (0.471079) (0.431770) (0.420283)** (0.413886)** (0.353876)** (0.332069)**
Log Total 0 .7 9 2 4 4 9 0 .6 9 0 4 8 4 0 .7 4 0 7 4 9 1.309144 -0 .0 5 1 7 0 0 0 .6 1 8 6 6 0
A sse ts (0.175164)** (0.221986)** (0.157991)** (0.201078)** (0.123754) (0.171545)**
Log Net S a les -1 .0 9 1 9 2 3 -0.382582 0 .2 8 8 8 3 6 0 .0 9 3 3 0 2 1 .380759 0 .4 7 5 8 8 4
(0.281204)** (0.267176) (0.230472) (0.237861) (0.229369)** (0.205869)*
Log # Emp 0 .6 8 3 1 6 0 0.081489 0 .4 2 3 1 5 7 0 .1 7 6 2 6 2 -0 .2 6 0 0 0 3 0 .0 9 4 7 7 4
(0.220343)** (0.236455) (0.173747)* (0.201404) (0.159742) (0.160221)
Debt Ratio -1 .9 8 5 7 0 6 -0.3 9 5 9 9 4 -1 .5 4 2 2 3 3 0 .0 3 0 4 0 4 0 .4 4 3 4 7 3 0 .4 2 6 3 9 8
(0.912826)* (0.891791) (0.688003)* (0.677725) (0.559672) (0.619795)
ROA 5 .1 1 3 2 3 5 7.017931 4 .0 0 5 4 0 9 6 .3 9 4 1 1 7 -1 .1 0 7 8 2 6 -0 .6 2 3 8 1 5
(3.114155) (2.812661)* (2.221520) (2.081077)** (1.751818) (1.634223)
Constant 0 .7 4 9 5 1 4 -3.920443 -7 .0 5 7 3 2 0 -1 1 .0 5 2 0 7 5 -7 .8 0 6 8 3 3 -7 .1 3 1 6 3 3
(1.629251) (1.594633)* (1.327691)** (1.446478)** (1.104396)** (1.103798)**
Industry FE No Y es No Y es No Y es
Observations 459 459 459 459 459 459
R-squared 0.06 0.22 0.30 0.39 0 .39 0 .50
Robust standard errors in parentheses
Table 2-4] Regression Results of the KLD Overall, Positive, and Negative Ratings Index on the Presence of a PA/SR
Committee (2005)
70
5. Diffusion of a PA/SR Committee
5.1. Summary Statistics
In this section, we try to show the diffusion of the presence of a PA/SR committee
over time and how it relates to the social performance of a firm. Figure 2-4 and table 2-5
show the diffusion of a PA/SR committee over the sample period. The number of firms
with a PA/SR committee has increased more than 34% during the six-year period
between 2000 and 2005. In our sample of S&P 500 companies, there were 66 such
committees in 2000, compared to 89 in 2005.
o
o -
2000 2001 2002 2003 2004 2005
Figure 2-4] Number of S&P 500 Firms with a PA/SR Committee by Year
71
Many notable companies such as ConocoPhillips, Alcoa, CSX, GE, Lockheed
Martin, Eastman Kodak, Gap, Merck and AT&T established a PA/SR committee during
this period. Table 2-5 presents a summary of the number of companies that established a
PA/SR committee between 2000 and 2005 by industry. The Industrials, Consumer
Discretionary, and Financials sectors each include at least four companies that added a
PA/SR committee during that period, whereas the Consumer Staples industry has no
company that added such a committee.
Year 2000 2005

Number .. ...
... „ % with a “ f % with a
GICS S ector with a PA/SR with a PA/SR
PA/SR „ KA/b.t? PA SR
„ ... Committee /-v ... Comm ittee
Com m ittee Comm ittee
10 Energy 5 17.24% 6 20.69%

15 Materials 6 19.35% 8 25.81%
20 Industrials 9 16.98% 14 26.42%
25 C onsum er Discretionary 11 12.22% 15 16.67%
30 C onsum er S taples 10 25.64% 10 25.64%
35 Health Care 8 14.04% 11 19.30%
40 Financials 9 10.71% 13 15.48%
45 Information T echnology 1 1.30% 2 2.60%
50 Telecom m unications S ervices 1 12.50% 3 37.50%
55 Utilities 6 18.75% 7 21.88%
Total 66 13.20% 89 17.80%
Table 2-5] Industry Breakdown of the Presence of a PA/SR Committee in 2000 and in
2005
Tables 2-6 and 2-7 show summary statistics and correlations among variables.
During the six-year sample period, the presence of a PA/SR committee is positively and
significantly correlated with both the KLD positive ratings index and the KLD negative
ratings index. Larger firms tend to have a PA/SR committee, as all the size variables (log
of assets, log of sales, log of number of employees) show significant positive correlation
with the presence of a PA/SR committee. Debt ratio is also positively correlated with the
committee presence at the 5% significance level. The KLD overall ratings index is
negatively correlated at the 5% significance level with the presence of a PA/SR
committee, even though the correlation is relatively small at -0.0474.
Variable N Mean Std. Dev. Min Max
PA/SR Committee 2994 0.158651 0.365411 0 1

KLD Overall Ratings Index 27 4 9 -0.32521 2 .9 9 1 5 7 4 -11 11
KLD Positive Ratings Index 27 4 9 2 .8 2 3 2 0 8 2.7 1 7 4 0 8 0 18
KLD N egative Ratings Index 27 4 9 3 .1 4 8 4 1 8 2 .5 8 8 2 8 4 0 16
ROA 2728 0 .0 5 0 1 0 7 0.0 8 2 3 5 -0.75678 0 .4 5 7 9 9
Log Total A sse ts 2728 9.2 7 8 5 5 1.418927 5 .5 5 5 7 1 2 14.2 1 6 9 9
Log S a le s 2728 8 .7 6 1 4 6 4 1 .184459 4 .6 2 0 9 5 5 12.70142
Log # of E m ployees 2689 3.098031 1 .190872 0 .3 0 3 0 6 3 7 .4 9 6 0 9 8
Debt Ratio 27 2 0 0 .6 1 9 3 4 4 0 .2 2 1 2 9 5 0 .0 4 2 7 6 6 1.8 7 9 3 6 4
Table 2-6] Summary Statistics of Variables
73
Reproduced
with permission
00
Variable 1 2 3 4 5 6
1 PA /SR Comm ittee 1

2 KLD Overall Ratings Index -0.0474* 1
3 KLD Positive Ratings Index 0.2810* 0.5926* 1
4 KLD N egative Ratings Index 0.3497* -0.5337* 0.3650* 1
5 ROA 0 .0 0 8 5 0.1026* -0.0048 -0.1228* 1
6 Log Total A sse ts 0.2628* 0.0 0 5 3 0.4524* 0.4657* -0.2050* 1
7 Log S a les 0.3290* -0.0534* 0.4663* 0.5477* -0.013 0.7423* 1
8 Log # of Em ployees 0.2727* -0 .0 1 0 9 0.3648* 0.3931* -0 .00 4 7 0.4827* 0.7958* 1
9 Debt Ratio 0.1784* -0 .0 3 7 3 0.1457* 0.1948* -0.2241* 0.5190* 0.3439* 0.1919*
* significant at the 5% level
Table 2-7] Correlation among Variables
74
5.2. Regression Analysis
I follow Podolny (1994) in using a binary choice model to study the adoption of a
PA/SR committee.33 Table 2-8 reports regression results of Probit regression of the
following specification:
P r ( y , = l l X „ ) = <D(X„/?)
y l = x u P + £m y» = K y i > 0)
P r ( y , = l l A , ) = Pr(y*>0IA,)
= Pr(£, > - X j \ X ) = \ - < t > ( - X j ) = ® ( X J )
, where <I>(z) = J (2;r)_1/2 exp(-w 2 / 2)du
The dependent variable y,, is 1 if a PA/SR committee is present for firm i in year t.
Otherwise, it is 0. Xit is a vector of explanatory variables that includes the KLD negative
ratings index for firm i in year t. y*„ is a latent dependent variable that gives rise to the
observable y„ through an indicator function. If y*,, is above 0, you observe the value of y lt
as 1. Bi, is an error term that is assumed to be normally distributed. O is the cumulative
distribution function of a normal random variable.
Table 2-8 reports the marginal effects calculated at the means of the explanatory
variables. In columns (1) through (3), we start with a single explanatory variable of
interest, the KLD negative ratings index, and add more explanatory variables. When there
33 The other approach would be to use a hazard model, as in D avis & Greve (1997), but time-varying
independent variables in our data do not suit the typical hazard model formulation.
75
is no control variable for firm level characteristics, a 1-point increase in the KLD
negative ratings index in a given year for a company with average level explanatory
variable values is associated with a 4.1% increase in the probability of a PA/SR
committee being present. The association is significant at the 1% level. When we add
time-varying firm level characteristics such as debt ratio, log of assets, log of sales, log of
number of employees and ROA, the probability becomes 2.5%, and it is significant at the
1% level. When we add the industry dummies and year dummies, the probability
becomes 1.9%, and it is significant at the 1% level. This means that in a given year, a
firm with average-level explanatory variable values is 1.9% more likely to have a PA/SR
committee in its board when its KLD negative ratings index is 1 point higher than its
industry peers. We interpret these results as evidence that firms establish a PA/SR
committee in response to recent events that could drive their negative social performance
ratings higher, thereby affecting the overall social performance ratings negatively.
76
Reproduced with permission
(1) (2) (3)

Dep. Variable PA/SR Comm ittee PA/SR Comm ittee PA/SR Com m ittee
KLD N egative 0.040951 0 .0 2 5 4 1 8 0.0 1 9 3 7 5
Ratings (0.002504)** (0.002825)** (0.003131)**
Debt Ratio 0 .2 5 1 4 3 0 0 .233523
(0.037172)** (0.039805)**
Log # Emp 0 .0 2 4 1 7 5 0 .0 1 6 2 6 7
(0.008911)** (0.010448)
Log Net S a les 0.038831 0.0 1 6 3 3 8
(0.012829)** (0.013738)
Log Total A sse ts -0 .0 0 7 1 0 5 0.0 2 7 5 2 6
(0.007807) (0.011799)*
ROA 0 .3 7 2 3 4 7 0.340731
(0.106023)** (0.113674)**
Industry Dum m ies No No Y es
Year D um m ies No No Y es
prohibited without p erm ission .
Firm Dum m ies No No No

O bservations 2749 2686 2686
Pseudo-R * 0.12 0.18 0.20
Reported coefficients are marginal effects; standard errors in parentheses
Table 2-8] Probit Regression Results of the Presence of a PA/SR Committee on the KLD Negative Ratings Index
77
6. Conclusion
In this paper, we analyze companies that have engaged their boards of directors in
integrated strategy by establishing voluntary public affairs or social responsibility
committees at the board level. Among the companies in the S&P 500 at the end of 2005,
17.8% of the companies have such a committee, up from 13.2% in 2000.
We find that companies with such a committee have significantly higher scores in
the KLD negative ratings index. We also find that companies with such a committee have
significantly higher scores in the KLD positive ratings index. However, when we look at
the KLD overall ratings index, we find no association between the presence of such a
committee and the KLD overall ratings index. These results suggest that companies learn
to balance overall social performance by forming a specialized board-level committee
that seeks to integrate market and non-market strategies.
A panel data analysis over the six-year period between 2000 and 2005 also
reveals that the presence of such a committee is closely associated with high negative
social performance ratings. We interpret this result in the following way: companies that
have recently experienced problems in their social performance ratings tend to establish
these committees in order to mitigate the impact of the negative event.
Given that the existing research on the relationship between corporate governance
schemes such as board composition and board leadership structure and corporate
performances display few systematic results, we believe that our findings make an
important contribution to the literature.
78
ESSAY THREE. THE TIPPING POINT: PATH-DEPENDENCE IN
RESTAURANT MARKETS
(Joint with Joel Waldfogel)
1. Introduction
Do markets tend toward unique and efficient equilibrium? Or are markets path-
dependent, leading to different, and potentially inefficient, outcomes depending on initial
conditions? In short, does history matter? Few questions have generated as much
interesting scholarship - and heated invective - as this one. The topic traces its recent
intellectual history to works by Paul David and W. Brian Arthur arguing for path-
dependence.34 David offers the vivid example of the QWERTY typewriter keyboard
which, he argues, was an inferior design adopted to slow typists to within the capacity of
early machines. Once adopted, the argument goes, this standard has proven hard to shake
even though better keyboard layouts are available. Liebowitz and Margolis (1990)
dispute David’s evidence on typing speed and go on to argue at length that it is difficult
for markets to follow paths to inefficient equilibria.
The dispute over path-dependence persists in part due to the difficulty of finding
convincing evidence for one side or the other. The reason for this is that it is generally
34 See David (1985) and Arthur (1989).

79
very difficult to observe multiple markets with different initial conditions, asking whether
they arrive at the same equilibrium. Instead, one must look to the historical development
of markets for comparisons sharing some features of this ideal. To this end, this study
makes use of different restaurant markets separated by geography. The idea is simple.
Residential areas in the U.S. developed at different times. In each period since 1950,
different chain restaurants stood ready to enter the geographic areas as residential
development created demand for restaurants. For instance, the set of restaurants that
might initially have entered an area that developed in 1960 is not the same as the set of
restaurants that were ready to enter areas developed in 1970. How are both of these
markets configured in 2004? If the products that became available later replaced the
incumbents who entered earlier, then initial conditions will have no effect on available
products, and history does not matter. On the other hand, if the restaurants that are
currently operating in two otherwise similar areas differ based on initial conditions, then
history matters.
Unlike most contexts where scholars seek evidence of path-dependence or lock-in,
this one has low switching costs. The usual case has complementary “platforms” and
“applications,” such as trains and track, or hardware and software. Here, the only
switching cost is the cost of trying - and coming to prefer - a new restaurant over an old
one. Hence, our approach offers a “one-sided test”: a failure to detect effects of history is
uninformative, while detected effects of history - in a context without explicit switching
costs - will provide evidence that a wide range of markets can be path-dependent.
We use detailed Census data on the nature of demand (demographic
characteristics), as well as information on the location of each restaurant from
80
169 chains. We know when each individual restaurant outlet opened, if it opened after
1984. And we know the date that each chain was founded along with fragmentary data on
the timing of each chain’s growth prior to 1985. These chain restaurants make up roughly
thirty percent of all restaurants in the U.S. We also have data on the total number of
restaurants in each zipcode from the Economic Census. Finally, the Census of Population
and Housing provides information on the median age of housing in each zipcode, which
we use as a measure of the time when demand arrived in the local market. Using these
data, we can ask whether older areas have older chains, and so on.
The paper is organized as follows. Section 2 presents theoretical background and
reviews the existing empirical literature on path-dependence. Section 3 describes the
data used in the study and presents brief histories of the chain restaurants in the dataset.
Section 4 describes the empirical strategy and results. Section 5 concludes.
2. Theoretical Background and Literature
This paper draws upon various streams of existing literature. The first is the
literature debating over the existence of lock-in and path dependence, following David
(1985) and Liebowitz and Margolis (1990). David (1985) argues that the technically
superior Dvorak keyboard was defeated by the QWERTY keyboard in the battle for the
standard, for no other reason than those events that could be characterized as ‘historical
accidents’. Arthur (1989) presents a theoretical model where agents choose between
competing technologies which improve as they gain in adoption. He shows that over time,
81
due to random historical events, the economy can become locked in to a technological
path that is not necessarily efficient.
The debate on path dependence caught the public’s attention when the US
Department of Justice filed a lawsuit against Microsoft for its anti-competitive behaviors.
Believers of lock-in and path dependence sided with the DOJ and accused Microsoft of
being an inefficient monopoly that had happened to dominate the market. And once it
dominated the market, they argued, it was trying to use its bottleneck operating system
product in order to suppress all the other competitors that had the potential of developing
technically superior products to its Windows-based applications. Liebowitz and Margolis
(1999) objected to the view, and argued extensively for the technical superiority of
Microsoft’s products - the operating system, the word processor, and the Internet browser.
They also provided an argument that the existence of switching costs could not possibly
lead to an inefficient outcome. In so doing, they categorized path dependence into three
classes. First-degree path dependence arises because some economic decisions involve
dynamic optimization, which may not be optimal from a static point of view. In their
example, you may purchase a house that maximizes you utility over the length of your
stay in that house, but that may be unnecessarily large at the moment. Second-degree path
dependence arises due to limited foresight. A decision which is suboptimal ex post may
well have been optimal ex ante - this type of path dependence is not remediable. Third-
degree path dependence is the kind of remediable inefficiency that arises from a
coordination failure - one individual has no incentive to unilaterally change her behavior,
but the entire group would benefit if everyone changed. Liebowitz and Margolis argue
that only this type of path dependence causes inefficiency in the economic
82
sense. Puffert (1999) criticizes Liebowitz and Margolis’ categorization based on the fact
that both foresight and remediability are matter of degree, therefore the distinction
between second- and third-degree path dependence is not so clear after all.
Crucial in this debate on path dependence is the existence of multiple equilibria.
In his attempts to build a model explaining the agglomeration in manufacturing,
Krugman (1991a) theorizes that in a situation with multiple equilibria one of two factors
determines which equilibrium gets established: history or expectation. In his model,
whether history or expectation plays the decisive role depends on three factors: the
interest rate, the magnitude of the external economy, and the speed of adjustment. History
matters more when people discount the future more heavily, when the size of the external
economy is smaller, and when the speed of adjustment is slower. The restaurant industry
- which has typically fairly low setup costs, can be characterized as low external
economy / fast adjustment industry, and the only factor that sides with history over
expectation in determining the outcome in this industry is geography.
Krugman (1991b) argues that whether geography is path-dependent depends on a
combination of three factors: large setup costs that yield strong economies of scale,
sufficiently small costs of transportation, and a sufficiently large share of ‘footloose’
production not tied by natural resources. Restaurant industry has small setup costs and the
transportation costs are relatively large compared to the production costs. And even
though the restaurant industry is not bound by natural resources, it is heavily dependent
upon proximity to consumers. As Waldfogel (2004) argues, the restaurant industry is a
highly local industry, just like agriculture. As such, the geography of restaurant industry
83
does not seem likely to be path dependent. Therefore this paper can be thought of as
providing a one-sided test of whether history matters.
Even without explicit lock-in, history can matter for other reasons. First, new
firms will prefer to avoid competition with established firms. Given a choice between
entry in markets with and without established competition, a new firm would prefer to
enter in a place that lacks competition. Given the localized geographic markets for
restaurants, areas of new housing development constitute such new markets. Thus, we
would expect new firms to enter in new markets and, until they seek direct competition
with established firms, to be more likely to operate in new markets. That is, we expect a
positive relationship between the prevalence of new chains and the time that markets are
established.
As a firm gets established - perhaps developing brand recognition that puts it on
equal footing with older firms - it can “defy history.” That is, it can enter in proximity of
established firms. Eventually it stands on equal footing with existing chains, and -
conditional on demand factors - the probability of entry is equal across markets of
different vintages. Starbucks, for example, is as likely to be found in older zipcode areas
as in new zipcode areas. Because a Starbucks store needs smaller space for its operation -
therefore can be characterized as having a small footprint (Baldwin & Clark, 2006) - and
because its main clientele is the pedestrians in downtown districts, Starbucks has
penetrated into the old market areas successfully, despite being one of the youngest of
chains - thereby defying history.
There is a second, related sense in which history can matter. Sunk costs of entry
place a wedge between the auspicious conditions needed to prompt entry and
84
the much-less-favorable conditions - failure to cover ongoing costs - needed to prompt
exit. The chains entering in the past when facing good conditions in then-new local
markets continue to operate their locations in places where they would not currently
choose to enter. This is the hysterisis explored by Dixit (1989).35 As a result, we would
expect to see old firms more prevalent than new ones in older markets. That is, for the
oldest chains, we expect a negative relationship between prevalence and the time when
their markets are established. The sunk costs of restaurant entry include both costs of
physical assets as well as costs of developing brand recognition, as in Klein & Leffler
(1981).
According to Puffert (2002), a path dependent economic process is “one in which
specific contingent events have a persistent effect on the subsequent course o f
allocation.” He addresses the question of whether the path dependence does take place
and whether it is a source of inefficiencies by investigating the history of railway track
gauges in Britain, Continental Europe, North America, and Australia. He argues that
random events reinforced by positive feedback determined both particular standards and
the geographic extent of standardization. His cellular automata Monte Carlo simulation
shows that the construction and conversion of railway track gauge is a symmetry-
breaking process, in the sense that eventually more than 90% of the time the
configuration winds up with a single standard, whether it be a narrow gauge or a broad
one. In his model, it is also notable that stochastic events, in both the order of
35 In D ixit’s setup the hysterisis-inducing effect o f sunk costs is magnified by uncertainty about future
demand.
85
construction and in the gauges favored by locals, make the process path-dependent,
thereby making it impossible to predict the outcome at the outset.
There are also other documented examples of path dependence in various
circumstances, including ready-made wet soups (Sutton, 1991), sports leagues (Mueller,
1997), Swedish manufacturing (Carlson, 1997), and pest control system (Cowan &
Gunby, 1996).
This paper also complements the management strategy literature that studies the
internal workings of franchise-type enterprises. Winter and Szulanski (2001) propose
replication as a strategy, especially in service industries - they claim that companies
operate a regime of exploration in the beginning in order to establish a business model
that suits them the best, and then turn to a phase of exploitation when they have enough
confidence in that business model, applying the model to various places through large-
scale replication. This study provides evidence that such replication strategy is taking
place en masse. Raff (2000) investigates the recent history of the bookselling industry.
From the emergence of two bookselling superstores, Borders and Barnes & Noble, he
argues that the market only imposes loose constraints on the strategies of firms, and
profitable firms in an industry can differ substantially in ways that affect their
performances. He also states that part of those differences is attributable to the history
that formed the core capabilities of the firms.
This paper also contributes to a growing industrial organization literature that
studies franchise enterprises. Holmes (2005) studies the expansion of Wal-Mart franchise
and conjectures that Wal-Mart’s huge success can be attributed to realizing and
exploiting the economies of density. Wal-Mart’s inside-out expansion strategy
86
cannot be applied directly to chain restaurants where the transportation costs are much
smaller and logistics is not at the core of its production process, but this Wal-Mart
example may be another incidence where history clearly matters. With a different entry
sequence, Wal-Mart may not have been as successful.
Basker (2005) looks into the price effect of a low-cost entrant who is a franchise
operator (Wal-Mart). She finds that the prices of drugstore items fall significantly with
the entry of Wal-Mart, while the prices of convenience store items may not always
decline. The prices of clothing are not affected significantly with Wal-Mart’s entry. She
also shows that the impact on prices by Wal-Mart’s entry is less pronounced in large
cities where there are more stores that carry the same products.
In the following sections, we seek to address the following questions. First, does
the vintage of available restaurants in a market area vary with market age? Second, are
restaurants of more recent vintage more prevalent in newer markets areas? Finally, are
older-vintage restaurants more prevalent in older markets (and less prevalent in new
ones)? Our undertaking is a rare attempt to provide evidence of path dependence using a
large set of data.
3. Data
3.1. Geographic Market Definition and Market Age
Before we move on to the question of whether history matters in the restaurant
87
industry, two methodological concerns need to be resolved. The first is what the right
choice of geographic market is in answering the question, and the second is how to
measure the age of a market area.
For the right choice of geographic market, we propose the five-digit zipcode area.
Although necessary demographic and economic variables about the geographic market
areas are available at various levels from Census blocks to states, the more aggregated
levels such as county or MSA are problematic because they are clearly too wide for a
geographic market for restaurants. Smaller levels such as Census tract or block groups are
too narrow, and the business pattern information is not available at that level of
disaggregation. Waldfogel (2004) also suggests that a 5-digit zipcode area is a good
proxy for a geographic market for restaurants. He finds that an average 5-digit zipcode
area is 2.95 miles wide in radius, which is reasonable distance to travel for a meal, and
that the composition of chain restaurants does not seem to change as one aggregates to
higher levels of geography. Even though we do not claim that each zipcode provides the
perfect boundary for the restaurants within, the use of a 5-digit zipcode area does not
seem to create a systematic bias as a geographic market for restaurants.
The next question is how to measure the age of a market area, and here we use the
concept of “zip vintage.” A zip vintage corresponds to the variable “median year housing
was built within the zipcode area”, as defined in the 2000 Census of Population and
Housing. The Census has reported this variable since 1990. Even though the zip vintage
is a limited measure of the age of the zipcode area, it is the best one available, and it is
88
certain that half of the residential structures in the zipcode were built after the zip
vintage36.
3.2. Chain Location and Zipcode Demographics
The basic data for this study is a zipcode level cross section containing
information on the number of restaurants, overall and in each of 169 restaurant chains
operating in each zipcode. The data also include zipcode level demographic information.
The data are drawn from three basic sources. Data on the individual restaurants’
locations, sales, and years of entry are collected from the Reference USA Database. The
number of NAICS722 (eating and drinking places) establishments are from the Zip
Business Pattern Dataset for the year 2002. Zipcode level demographic data is collected
from the 2000 Census of Population and Housing. In addition, company history data -
including founding dates - were collected through individual chains’ websites and
various other sources, including the books that document the lives of popular chains’
founders. Chain founding dates provide our basic measure of chain vintage.
The 169 restaurant chains in our sample are selected from Scarborough Research
survey data and Consumer Reports July 2003 issue. Chains are classified as “fast food”
or “sit-down,” and further classified by cuisine types. Fast food chains specialize in cheap
and fast services and usually have smaller store size. The data cover 59 fast food chains,
including McDonald’s, Subway, Burger King, and Starbucks. Sit-down chains provide
full service meals and range from economy to upscale. A detailed list of chains is
36 The variable in the Census is left censored at the year 1939.

89
available in Appendix 2. Sample chains include 134,646 outlets, 110,464 of which are
operating in one of 331 Metropolitan Statistical Areas (MSA’s). Among 31,656
residential zipcodes in the United States, 14,762 of them are located within an MSA.
Demographic variables vary substantially by zipcode area. Between 5th percentile
and 75th percentile, population varies from 295 to 24,055, the fraction of African
American population varies from 0% to 9.33%, median age in the zipcode area varies
from 27.5 to 39.3, unemployment rate varies from 1.1% to 6.7%, median household
income varies from $22,831 to $55,365, and the median year housing was built varies
from 1939 to 1979. The number of restaurants and the number of chain restaurants,
ancestry of population, and the percentage of households with children also show large
variation. Table 3-1 presents the summary statistics of these demographic variables for
the MSA zipcodes.
90
Reproduced
with permission
Statistics N Mean S.D . 5th 25th 50th 75th 99th
Population 14762 15319 157 8 5 295 2716 9890 24055 65868

% Black 14762 9 .57 17.62 0.00 0.48 2 .07 9 .33 87.81
% Hispanic 14762 8.95 15.81 0.30 1.07 2 .6 4 8 .52 8 2 .5 0
Median Household Incom e ($) 14729 4 6 7 5 2 .3 4 19 1 6 4 .2 9 22831 346 6 7 43364 55365 109771
Median A ge 14762 36.6 5.7 27.5 3 3.7 36.6 3 9.3 55.6
% Unemployed 14729 5.65 5.41 1.10 2.90 4 .40 6 .70 2 5 .7 0
Median Year Housing Built (Zip Vintage) 14696 1968 14 1939 1958 1970 1979 1995
# Restaurants 14762 2 6.53 3 3 .9 4 0 3 13 40 139
# Chain Restaurants 14762 7.40 10.21 0 0 3 11 43
# Sitdown Chain R estaurants 14762 1.59 3 .08 0 0 0 2 15
% English Ancestry 14741 9 .78 15.58 0.84 5.66 9 .1 7 12.90 2 8 .3 8
% German Ancestry 14741 17.99 2 1 .4 0 1.59 8.53 15.72 2 4 .2 3 57.11
% Irish Ancestry 14741 12.41 16.78 1.43 7.90 11.67 15.48 3 4 .0 4
% Italian Ancestry 14741 6.26 7 .87 0.00 1.56 3.65 8.0 3 3 3 .0 0
% American Ancestry 14741 8.30 5 7 .2 4 1.11 3.29 5 .66 1 0.17 3 0 .5 5

% M iscellaneous Ancestry 14741 2 2.97 26.41 2.71 7.34 15.59 31.51 8 3 .6 4
% Household with Children 14687 3 4 .2 8 10.70 16.95 2 8 .8 7 3 4 .3 8 3 9 .6 5 6 5 .6 3
Table 3-1] Summary Statistics for Variables
91
3.3. History of Chain Restaurants
Table 3-2 provides capsule historical information about 138 of our sample chains
for which the founding dates are available. Chain restaurants date back almost a century.
The oldest chain in the data, White Castle, was established in 1921. However, it was in
the 1950s that chain restaurants became commonplace in American daily lives. Twenty-
four sample chains, including such major chains as McDonald’s, Burger King, KFC, and
Pizza Hut, all came into existence during the 1950s. The 1960s continued to see the trend
of growing fast food chains. This trend, however, started to change in the early 1980s.
Perhaps because the fast food market was saturated, new entrants developed innovative
concepts and established themselves at niches. Some chains, like Papa John’s, went into
downtown areas where no upscale competitors could afford high rents. Some others, such
as Starbucks, successfully targeted increasing number of urbanites who did not mind
spending money for a quality break. The 1990s saw the upscale differentiation, where
chains like II Fomaio and Morton’s of Chicago differentiated themselves from other
chains by offering high quality meals and services. The 10 largest sitdown chains in our
data, measured by the number of outlets, are Pizza Hut, Denny’s, Applebee’s, Waffle
House, International House of Pancakes, Chili’s, Outback, Ruby Tuesday, Red Lobster,
and Godfather’s Chicken. The 10 largest fast food chains are Subway, McDonald’s,
Burger King, Wendy’s, Taco Bell, KFC, Starbucks, Domino’s, Dairy Queen, and Arby’s.
92
Number of
Founded S elected Chains
Chains Started
Before 1940 White C astle, Friendly's, Dairy Q ueen 6
1940s Baskin Robbins, Dunkin Donuts 8
KFC, D enny’s, Burger King, S onic Drive-in,

1950s 22
McDonald's, IHOP, P izza Hut
Hardee's, Donato's, Arby's, Taco Bell, Subway, TGI

1960s Friday’s, Domino's, Legal S e a Foods, Red Lobster, 33
Cracker Barrel
Bennigan's, C h e e se c a k e Factory, Houlihan's, P opeye's,

1970s 27
Ruby T uesday, W endy's
Chili's, Quizno's, Hard Rock Cafe, A pplebee’s, Papa

John's, Boston Market, California Pizza Kitchen, Cici's,
1980s 34
Starbucks, Outback S teakh ouse, Caribou C offee, Olive
Garden
J. Alexander's, M aggiano's Little Italy, Buca di Beppo,

1990s 9
PF C hang's China Bistro, II Fomaio
Table 3-2] List of Chain Restaurants by the Time They Started Operating as Chains
Table 3-3 shows a detailed timeline of two major sitdown chains, Denny’s and
Applebee’s. Commanding similar numbers of outlets by 2004, they started and expanded
over very different times. Denny’s started in 1953 and expanded aggressively throughout
1970s. Applebee’s came into existence after Denny’s had already more than a thousand
restaurants in operation, but its successful expansion during the 1990s eventually made it
into one of the largest sitdown chains.
93
Reproduced
with permission
Year Denny’s A p p leb ee’s Event

1953 1 Harold Butler founded Danny’s Donuts in Lakewood, CA.
1959 20 The chain w a s renam ed D enny’s Restaurants.
1963 78
1968 192
1980 1 Bill and T.J. Palm er founded A p p leb ee’s in Atlanta, GA.
1981 1000
1982 2
1983 A pplebee’s sold to W R Grace, a chain operator
1988 54 A pplebee's purchased by franchisees, b eco m e s A pp lebee’s International
1995 1529
1996 1596 819

1997 1652 960
1998 1000
2001 1400
2002 1676
2003
2004 1638 1521
Table 3-3] Timeline of Chain Expansion - an Example of Denny’s vs. Applebee’s
94
4. Analysis & Results
In order to answer the question posed in the beginning, we propose the following
empirical strategy. From the discussion in the previous section, it is clear that different
chains entered different markets in different periods. From today’s viewpoint, these are
historical facts, not affected by the demand and supply parameters currently at work. If
history does matter, one might expect to witness the outcomes that are not explained
solely by current economic conditions. Specifically, we look at the age of the market area.
Different market areas were developed in different times. If two markets that are identical
in all aspects other than when they were developed have two very different sets of chain
restaurants, it should be viewed as evidence that history has persisting effects on market
outcomes.
We first offer evidence that older chains are more prevalent in old zipcode areas,
and younger chains are more prevalent in younger zipcode areas, preserving the order of
existence. Regression analysis controlling for key demographic variables is presented to
confirm the hypothesis that history has lasting impacts. Then we look into prominent
individual chains to see if their current locations are systematically correlated with when
they started operating as chain.
4.1. Descriptive Analysis
Different chains enter different zipcode areas seeking the type of patrons that are
95
the most attractive to them. Figure 3-1 depicts the fraction of the number of outlets by zip
vintage, both cumulatively and non-cumulatively. The upper graphs show that Pizza Hut
and Outback Steakhouse are located in the areas with distinctive zip vintages. Pizza Hut,
being the older chain, generally has stores in much older zipcode areas. To borrow the
language of finance, the upper left diagram in figure 3-1 shows the relationship analogous
to first order stochastic dominance. What it means is that for any zip vintage, the fraction
of Pizza Hut outlets located in the market area of the particular zip vintage or earlier zip
vintages is always higher that that of Outback’s. For a market with 1960 zip vintage,
about 20% of Pizza Hut outlets are located in older market areas, whereas only about
10% of Outback outlets are located in older market areas. The median zip vintage for
Pizza Hut is 1972, whereas the median zip vintage for Outback is 1976.
The lower graphs compare McDonald’s and Subway, the two chains with the
largest number of outlets in the U.S. The Subway chain has the most outlets with more
than 15,000, while the McDonald’s is a close second with more than 13,000 outlets. The
picture shows that they operate in very similar zipcode areas in terms of zip vintage, in
spite of the fact that Subway came into existence a decade later than McDonald’s. Burger
King, not shown in the diagrams, is the third largest restaurant chain with about 7,000
outlets. It also has a very similar profile to those of McDonald’s and Subway. This
implies that the largest fast food chains have basically saturated the entire U.S. market, in
which case their differences in the number of restaurants come mainly from their
differences in minimum efficient scale, of which Subway possesses the smallest. It is no
surprise that these chains’ main expansion came from international locations during the
1990s.
96
Reproduced
with permission
in
00
oo
CO
o
CM
o
o
CM
o 1940 1960 1980 2000

M e d ia n Y e a r Z ip D eve lo p ed
1940 1960 1980 2000
M e d ia n Y e a r Z ip D e v e lo p e d
• P iz z a Hut M e dian bands
P iz z a Hut * O u tb a c k * O u tb a c k M e d ian bands
CO CO
o
o
o
CM
o
o 1940 1960 1980 2000
M e d ia n Y e a r Z ip D eve lo p ed
1940 1960 1980 2000
M e d ia n Y e a r Z ip D e v e lo p e d
• M c D o n a ld 's — M e d ian bands
• M c D o n a ld 's * S u b w ay * S u b w ay M e d ian bands
The upper graphs show that the older chain, Pizza Hut, is generally operating in older zipcodes than the younger chain, Outback Steakhouse. The lower graphs
show that the tw o most popular chains, M cD onald’s and Subway, share very similar pattern in zip vintages that they operate in.
Figure 3-1] Distribution of Outlet Locations by Zip Vintage for Selected Chains
97
CD
CD
o>
CD
o> • • • ••
a
A
CM
CD
O)
A
O
CD
a>
oo
in
o> • •• •••
1940 1960 1980 2000

Median Year Zip D eveloped
• Med Yr Chain Founded Median bands

* Mean Yr Chain Founded — — Median b ands
Figure 3-2] Mean and Median Year of Founding by Zip Vintage of Market Areas, All
Chains
Before turning to more formal analysis, some simple tabulation is instructive.
The basic question is whether older markets have older restaurants. Figures 3-2 and 3-3
show the mean and median vintage of sample chain restaurant outlets by the median year
that housing was built, overall and for sit-down restaurants. A clear pattern from both
figures is that newer market areas ~ zipcodes developed more recently - have chain
restaurants of more recent vintage operating.
98
1940 1960 1980 2000
Median Year Zip D eveloped
• Med Yr Sitdown Chain Founded Median bands

* Mean Yr Sitdown Chain Founded Median bands
Figure 3-3] Mean and Median Year of Founding by Zip Vintage of Market Areas,
Sitdown Chains
A second simple way to ask the question is to divide the chain restaurants into
four groups, according to founding dates: before 1950, during the 1950s, during the
1960s and 1970s, and 1980 and later. For each of these categories, we can ask how the
share of a zipcode’s chain restaurants in this vintage category varies across zipcodes with
their housing vintage. Figures 3-4 and 3-5 show these results for all chains and for sit-
down chains, respectively.
99
Zip Vintage -10 Year Period
_
CO
CM
o
1940zip s 1950zip s 1960ztps 1970zip s 198 0 zip s 1990zip s
before 1950 ■ ■ ■ in1950s

in 1 9 6 0 s7 0 s sin c e l 9 8 0
Figure 3-4] Fraction of Chains Originated in Different Periods Over All Chain
Restaurants in the Areas with Different Zip Vintages
Figure 3-4 shows that, among all the chain restaurants located in the zipcode areas
with zip vintages in the 1940s, the outlets from the chains originating before 1950
account for about 15%, those from the sitdown chains originating in the 1950s account
for about 36%, those from the sitdown chains originating during the 1960s and 1970s
account for about 35%, and those from the sitdown chains originating since 1980 account
for about 14%. The most striking feature of the graph is where the peaks lie. Chains
originating before 1950 reach the highest fraction in the 1950s zipcodes, while the
fraction of chains originating in the 1950s remains flat between zipcodes with 1940s
100
vintages and zipcodes with 1970s vintages and then declines. The pattern continues to
hold for the chains originating afterwards: chains originating during the 1960s and 1970s
have managed the highest fraction in zipcodes with 1980s vintages, and those chains
originating since 1980 have the highest fraction in zipcodes with 1990s vintages. Taking
the time of expansion into consideration, this picture suggests that the peak presence of
sitdown chains tends to coincide with the chains’ period of expansion. Figure 3-5
presents the fraction of sitdown restaurants that originated in different periods in time
over all chain restaurants, showing a very similar pattern to that found in figure 3-4.
Zip Vintage -10 Year Period
1940zip s 1950zip s 1960zip s 197 0 zip s 1980zip s 199 0 zip s
before 1950 in 1 9 5 0 s
in 1 9 6 0 s7 0 s sin c e l 980
Figure 3-5] Fraction of Sitdown Chains Originated in Different Periods Over All Chain
Restaurants in the Areas with Different Zip Vintages
101
Similar results can also be seen in table 3-4. The results are striking; the older
restaurants that were founded before 1950 are more prevalent in older zipcodes and less
prevalent in newer ones. Outlets of the newest restaurant chains, by contrast, have the
opposite pattern: they are more prevalent in newer zipcodes and less prevalent in older
zipcodes. These figures and tables provide suggestive evidence that history matters, in
both of the senses we outline. However, they leave open many alternative explanations.
For example, new and old zipcodes may differ systematically in their consumer
characteristics, in ways that are correlated with preferences for older or newer restaurants.
For example, areas with older housing may be occupied by older consumers who have
been long time residents of the area and prefer older-vintage restaurants. To see whether
history matters, we must determine whether the relationships that appear to hold in
figures 3-2 through 3-5 survive the inclusion of controls for consumer preferences. We
now turn to this in the next subsection.
4.2. How Do Chain Founding Dates Correlate with Zip Vintage?
We begin, in row 1 of table 3-5, with regressions of the share of a zipcode’s
restaurants in each of the four age groups on market age, with no controls. The purpose
of these regressions is simply to reproduce the substantive results of figures 3-4 and 3-5
in a regression context before adding controls. The results reflect the figures.
102
Reproduced
with permission
Zip Built <’50 ‘5 0 -‘60 '6 0 -7 0 ‘7 0 -‘80 ‘8 0 -‘90 ‘90+

All
Founded <50 10.9 9.2 7.4 4 .4 3.9 3 .4
Founded 80+ 18.4 20.1 22.0 24.1 27.3 3 0.3
Sitdown
Founded <50 11.3 9.2 7.5 4.5 3.9 3.3
Founded 80+ 18.9 20.1 22.2 24 .3 2 7.4 3 1.3
Table 3-4] Percent of Chain Restaurants in a Founding Cohort in Markets of Different Vintages
103
The coefficient of the median year housing is built is negative when the
dependent variable is the share of chain outlets founded before 1950. The coefficients
turn positive and increasing when the dependent variable is the share of chain outlets
founded in the 1950s, in the 1960s and 70s, and after 1980. This relationship is even
stronger when we restrict our attention to the sitdown restaurants in columns (5) through
(8). Older restaurants decline as a share of chain restaurants as zipcodes are newer. The
opposite is true for newer restaurants.
The second row adds a long list of controls, including median age, median
household income, population, the fraction of African American and Hispanic population,
as well as the shares in each of the ancestry groups in the Census.37 The coefficients of
interest are unchanged by the inclusion of the full set of controls.
The third row includes MSA fixed effects, and the coefficients of interest still do
not change. This implies that the positive correlation between the founding dates and the
zip vintage of where the outlets are located is not due to some regional variation. Within
a given MSA, older chain restaurants decline as a share of all chain restaurants as
zipcodes are newer, and newer chain restaurants increase as a share of all chain
38
restaurants as zipcodes are newer.
These results establish that older markets have older restaurant chains, while
newer markets have newer chains. These relationships survive an extensive battery of
statistical controls for heterogeneous preferences and MSA fixed effects.
37 The Census o f Population and Housing has data on 109 ancestry categories. In terms o f fraction o f
population, major ancestry groups in the Census are English, Irish, German, Italian, and American.
38 W e have repeated this exercise using a Tobit regression with unconditional M SA fixed effects, and the
coefficients stay basically unchanged.
104
Reproduced
with permission
(1) (2) (3) (4) (5) (6) (7) (8)

Sitdown Sitdown
D ependent Share Pre- Founded Founded Founded Sitdown Sitdown
Share Pre- S hare sin ce
Variable: '50 '50's '60,'70's >'80 Share ’5 0 s Share '60-80
■50 '80
No controls -0.0002 0 .0 0 0 7 0 .0006 0 .0 0 0 9 -0.0018 -0.0011 0 .0006 0 .0 0 2 3
(0.0000)** (0.0000)** (0.0000)** (0.0000)** (0.0000)** (0.0001)** (0.0001)** (0.0001)**
Controls -0.0002 0 .0 0 0 7 0.0 0 0 6 0 .0009 -0.0017 -0.0011 0 .0 0 0 5 0 .0 0 2 3
(0.0000)** (0.0000)** (0.0000)** (0.0000)** (0.0000)** (0.0001)** (0.0001)** (0.0001)**
+ MSA FE -0.0002 0 .0 0 0 8 0 .0 0 0 6 0.0009 -0.0018 -0.0009 0 .0005 0 .0 0 2 2
(0.0000)** (0.0000)** (0.0000)** (0.0000)** (0.0001)** (0.0001)** (0.0001)** (0.0001)**
Note: Each cell is a coefficient on the median year housing is built. The dependent variable is the zipcode’s share o f chain restaurants founded, say, prior to
1950. The first four columns use all sample chain restaurants. Columns (5)-(8) use only sample chain sitdown restaurants.
Table 3-5] Regression Results of the Fraction of Restaurants by Vintage on Market Age
105
4.3. Individual Chain’s Prevalence in Different Zip Vintages
Now we delve into a more detailed picture to investigate how the fraction of each
chain restaurant is correlated with the zip vintage after controlling for heterogeneous
tastes. We break down the zip vintage into thirteen 5-year periods and create semi-decade
dummies, in order to incorporate the nonlinear relationship between the prevalence of a
chain and the age of the market area. And we include the demographic variables such as
population, median household income, fraction of households with children, fraction of
African American and Hispanic population, and ancestry.
Our empirical specification is as follows: Let C be the set of all chains. The
dependent variable y\m is the fraction of chain c e C over all restaurants in a zipcode i
39
within an MSA m. X im is the vector of demographic and economic variables for a
zipcode i within an MSA m, Djm is the vector of dummy variables for five-year period
zip vintages for a zipcode i within an MSA m, and the error term e cim consists of two parts.
One part, / l cm, captures the unobserved MSA specific characteristics that is pertinent to
the fraction of chain c in all zipcode areas belonging to the MSA, and the other part, V-m ,
captures the unobserved characteristics of zipcode area i pertinent to the fraction of chain
c in that zipcode area. The idiosyncratic error term v.m is assumed to be distributed
normally, and its variance is assumed to be the same across different MSA’s. We clearly
39 The dependent variable is conditional upon the fact that the chain is present in the MSA.
106
have a problem estimating this equation using OLS, because in a large number of zipcode
areas a particular chain restaurant is not present at all. For example, even the chain with
the most number of outlets, Subway, is present in only 6,916 of the 14,653 zipcode areas
located within M SA’s. In other words, there are many observations in our data with a
comer solution outcome (Wooldridge, 2002). We use the following Tobit model with
unconditional MSA fixed effect to address this issue.40
Then for each chain c e C ,
y*icm - X imr'Bc + D imf• ' + £^im

c
K , \ X im, Dim~ N ( 0 , a l )
y ,cm = max(0, y ‘m)
Table 3-6 displays the Tobit regression results with unconditional MSA fixed
effect for ten of the most popular chains. For the zip vintage period dummies, the omitted
category is the 5-year period that each chain started operating as chain. One can observe a
clear pattern: the fraction of each chain is lower in the zip vintage periods that precede
the chain’s entry; it is higher in the zip vintage periods that come after the chain’s entry.
To highlight the findings in table 3-6, figure 3-6 displays the regression
coefficients of Denny’s and Applebee’s with their 95% confidence intervals. These are
two prominent sitdown chains with very different dates of origin. Denny’s started
operating as a chain in 1953. The three zip vintage periods preceding this have
40 W e acknowledge that the unconditional fixed effect model in Tobit produces biased estimates. However,
we believe that our sample size is sufficiently large to obtain consistent estimates.
107
significantly negative coefficients, implying that the fraction of Denny’s restaurant is
lower in those zipcode areas, whereas all of the following zip vintage periods have
significantly positive coefficients, implying that the fraction of Denny’s restaurant is
higher in those zipcode areas.
Similarly, the graph shows that any zip vintage periods before the first half of the
1980s when Applebee’s started to expand as a chain have significantly negative
coefficients, implying a lower fraction of Applebee’s compared to the first half of the
1980s. In contrast, the zipcodes with 1990s vintages have higher fraction of Applebee’s
restaurants.
These results complement the aggregate results discussed in the earlier subsection
and illustrate the mechanism with which older chains are more prevalent in older market
areas and younger chains are more prevalent in younger market areas: In its early stage of
expansion, every chain tends to enter market areas that are relatively new. Coupling this
with the fact that different chains came into existence at different times, we observe the
pattern that older chains are more prevalent in older market areas and younger chains are
more prevalent in younger market areas. Behind this phenomenon, there is path-
dependence, without which the later entrants can quickly drive the incumbents out, and
markets should reflect current economic conditions only. We argue that these findings are
evidence that prominent chain restaurants exhibit path dependence in terms of the vintage
of locations where they are operating.
108
Reproduced
with permission
Table 3-6] Tobit Regression Results of Individual Chain’s Fraction in Zipcode Area on Zip Vintage Period Dummies
Dependent APPLE- CHILI'S PIZZA HUT DENNY'S BURGER McDonald's STAR KFC TACO BELL SUBWAY
Variable: BEE'S KING BUCKS
Older than 1939 -0.074915 -0.077966 -0.036234 -0.034182 -0.016378 -0.050196 -0.031839 -0.022804 -0.038575 -0.052755
(0.010056)** (0.014263)** (0.006257)** (0.010104)** (0.006918)* (0.006066)** (0.007796)** (0.006042)** (0.005816)** (0.007274)**
1940's, -0.071689 -0.070101 -0.018112 -0.013916 -0.025476 -0.035866 -0.032792 -0.008368 -0.025495 -0.047682
First Half (0.012128)** (0.016598)** (0.007063)* (0.010396) (0.008908)** (0.007280)** (0.010272)** (0.006762) (0.006515)** (0.008827)**
1940's, -0.066515 -0.072734 -0.015726 -0.011204 -0.026906 -0.030520 -0.041349 -0.013787 -0.025087 -0.030749
Second Half (0.010614)** (0.014933)** (0.006154)* (0.009440) (0.008022)** (0.009528)** (0.006152)* (0.005749)** (0.007703)**
(0.006400)**
1950's, -0.039118 -0.052805 -0.007321 -0.013130 -0.029807 -0.011422 -0.011268
First Half (0.007340)** (0.009717)** (0.004971) (0.005102)* (0.007768)** (0.004447)* (0.006142)
1950's, -0.030287 -0.033645 0.019322 0.007393 -0.028542 0.001301 -0.003324 -0.006498
Second Half (0.006285)** (0.007229)** (0.006409)** (0.005805) (0.007029)** (0.004715) (0.003868) (0.005513)
1960's, -0.025436 -0.035616 0.003220 0.027188 0.005852 -0.002353 -0.024506 0.001483 -0.001597
First Half (0.006026)** (0.007065)** (0.004355) (0.006349)** (0.005991) (0.004547) (0.006846)** (0.004736) (0.005368)
1960's, -0.017913 -0.026259 0.009809 0.024111 0.012258 0.000840 -0.029011 0.003648 0.002570
Second Half (0.005521)** (0.006318)** (0.004247)* (0.006345)** (0.005895)* (0.004487) (0.006577)** (0.004720) (0.003706)
1970's, -0.014725 -0.008965 0.011070 0.028769 0.008053 0.004804 -0.025786 0.004445 0.004963 0.009572
First Half (0.004906)** (0.005333) (0.004088)** (0.006207)** (0.005732) (0.004256) (0.006134)** (0.004620) (0.003517) (0.004840)*
1970's, -0.004425 -0.009650 0.010985 0.031500 0.010162 0.015084 -0.025280 0.006606 0.006716 0.017075
Second Half (0.004564) (0.005035) (0.004184)** (0.004719) (0.003598) (0.004945)**
(0.006366)** (0.005968) (0.004280)** (0.005887)**
1980's, 0.006080 0.033933 0.011607 0.016941 -0.015239 0.000907 0.007583 0.026538
First Half (0.004524) (0.006660)** (0.006292) (0.004582)** (0.005817)** (0.005131) (0.003847)* (0.005323)**
1980's, 0.007029 0.005061 0.009742 0.035026 0.010793 0.033818 0.008325 0.011924 0.047778
Second Half (0.005427) (0.005685) (0.005106) (0.007398)** (0.007296) (0.005631) (0.004501)** (0.006305)**
(0.005523)**
1990's, 0.004070 0.012501 0.016558 0.033259 0.036971 0.051748 0.024998 0.019215 0.023590 0.069967
First Half (0.007194) (0.006953) (0.006288)** (0.008640)** (0.007836)** (0.006857)** (0.007377)** (0.006966)** (0.005473)** (0.007751)**
1990's, 0.043889 0.028427 0.055928 0.034429 0.031922 0.073030 0.096445 0.050947 0.042898 0.119855
Second Half (0.009368)** (0.009998)** (0.007940)** (0.012100)** (0.012366)** (0.009927)** (0.008126)** (0.008876)** (0.007493)** (0.011341)**
109
Reproduced
with permission
Dependent APPLE- CHILI'S PIZZA HUT DENNY'S BURGER McDonald's STAR KFC TACO BELL SUBWAY
Variable: BEE'S KING BUCKS
Total Population 0.000002 0.000002 0.000003 0.000002 0.000003 0.000003 0.000002 0.000002 0.000002 0.000003
(0.000000)** (0.000000)** (0.000000)** (0.000000)** (0.000000)** (0.000000)** (0.000000)** (0.000000)** (0.000000)** (0.000000)**
% Black -0.000535 -0.000545 -0.000245 -0.000236 -0.000075 0.000476 -0.001099 0.000517 -0.000144 0.000062
(0.000189)** (0.000213)* (0.000102)* (0.000126) (0.000149) (0.000109)** (0.000146)** (0.000095)** (0.000094) (0.000156)
% Hispanic -0.000797 -0.000358 -0.000179 0.000182 -0.000165 0.000386 -0.001101 -0.000039 -0.000255 -0.000125
(0.000249)** (0.000242) (0.000126) (0.000138) (0.000187) (0.000145)** (0.000175)** (0.000122) (0.000118)* (0.000197)
Median 0.000233 0.000323 -0.000051 -0.000208 -0.000665 -0.000079 0.001698 -0.000359 -0.000312 -0.000100
Household (0.000126) (0.000131)* (0.000097) (0.000117) (0.000121)** (0.000095) (0.000084)** (0.000100)** (0.000088)** (0.000113)
Income (‘000)
Median Age -0.001077 -0.001329 -0.000651 -0.001236 -0.002061 -0.001672 -0.003015 -0.000778 -0.001417 -0.001714
(0.000360)** (0.000391)** (0.000266)* (0.000292)** (0.000337)** (0.000283)** (0.000335)** (0.000272)** (0.000238)** (0.000335)**
% Unemployed -0.003064 -0.001928 -0.001029 -0.001018 -0.000709 -0.000229 -0.000069 -0.000711 -0.000354 -0.000284
(0.000647)** (0.000680)** (0.000317)** (0.000377)** (0.000348)* (0.000285) (0.000334) (0.000307)* (0.000250) (0.000333)
% Households -0.168288 -0.151922 -0.059647 -0.132216 0.010673 -0.073571 -0.288024 -0.039154 -0.060300 -0.107869
With Children (0.019957)** (0.021433)** (0.013005)** (0.016740)** (0.015088) (0.013653)** (0.017433)** (0.012346)** (0.011641)** (0.016266)**
Constant -0.016568 0.067275 -0.029750 -0.011615 -0.000381 -0.032963 0.113091 -0.100744 0.030251 0.064462
(0.057004) (0.046617) (0.032158) (0.038654) (0.044343) (0.039929) (0.056005)* (0.046020)* (0.027584) (0.043093)
Ancestry Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes
Variables
MSA FE Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes
# Observations 13987 12491 14633 12083 14645 14653 12825 14625 14553 14653
# Uncensored 1189 731 3841 1234 4312 6421 2623 3506 3517 6916
Standard errors in parentheses; * significant at 5%, ** significant at 1%
Table 3-6 (continued)] Tobit Regression Results of Individual Chain’s Fraction in Zipcode Area on Zip Vintage Period Dummies
110
— APPLEBEE'S » 95% Cl (lower)-A 95% Cl (upper)-A
DENNY'S -*-9 5 % Cl (lower)-D —-9 5 % Cl (upper)-D
0.08
0.06
0.04
Coefficients
0.02
0
- 0 . 02|
-0.04
-0.06
-0.08
0.1
-
Zip Vintage - 5 Year Period
For each graph, the break corresponds to the omitted category, which is the 5-year period that the chain started operating.
Figure 3-6] The Coefficients of Zip Vintage Period Dummies from Regressions in Table 3-6 (Denny’s vs. Applebee’s)
5. Conclusion
In this paper, we investigate whether history has a persistent effect in shaping the
geographic market in the restaurant industry. The question of whether history matters is a
long-standing one, and there have been both theoretical debates and anecdotal evidence
on why history should or should not matter, but this is a rare attempt to empirically test
the proposition using a large set of data.
This paper addresses the question of whether history matters or not by utilizing
the fact that different chain restaurants came to existence at different points in time,
penetrating into the new market areas of the time where they faced less competition and
expected to reap higher profit. It turns out that those chains that settled down to then-new
market areas are quite resilient to the arrivals of later competitors, and we observe that
different restaurants are available in areas where current market conditions are very
similar, depending on who was available to enter at the early stage of the geographic
market’s development. In particular, older chain restaurants are located in older market
areas and younger chain restaurants are located in younger market areas, after taking the
heterogeneity in current market conditions into account.
The evidence discussed in the paper suggests that history matters in determining
the identity of the chain restaurants currently available. One may argue that, as long as
those different chains available in similar areas are close substitutes to one another, the
welfare implication of history may not be too large. We do not have evidence to disprove
112
that claim yet. Still, the size of the profit associated with the control of a certain market is
certainly not negligible.
The potential welfare effect of path-dependence in restaurant markets is clearly an
important issue. If the path-dependence that we document in this paper is systematically
correlated with the type of cuisines available, there can be a welfare effect. Besides the
welfare implication of path-dependence, we are interested in the following two questions
that we hope to address in the near future.
The first one is how long would history matter if it matters. It is possible to
imagine that the effect of history also fades away as time passes by. Does the persistence
of history have the same effect for chains founded half a century ago as it does for chains
founded a decade ago? If not, what determines how long history might matter? We plan
to provide an answer to this question using our data in the near future.
The second question is about the possibility of “defying history,” as we briefly
touched upon in section 2. Although the discussion in this paper establishes that initial
conditions have rather persistent effects on products available decades later, it is not
necessarily true that all new restaurants have trouble penetrating into old market areas. It
would be interesting to see if we can identify what are the characteristics of businesses
that enable them to defy history.
113
Reproduced
with permission
APPENDIX
Appendix 1. List of Strength and Concern Items in the KLD Social Ratings Database, by Category (Essays 1 & 2)
Category Strength Concerns

Charitable Giving Investment Controversies
Innovative Giving Negative Economic Impact
Non-US Charitable Giving Tax Disputes
Community
Support for Housing Other Concern
Support for Education
Volunteer Programs
Other Strengths
Limited Compensation to Top Management High Compensation to Top Management
Ownership Strength Ownership Concern
Corporate
Transparency Strength Accounting Concern
Governance
Political Accountability Strength Transparency Concern
Other Strength Political Accountability Concern
Other Concern
CEO Controversies
Promotion Non-representation
Board of Directors Other Concern
Diversity Work/Life Benefits
Women & Minority Contracting
Employment of the Disabled
Gay & Lesbian Policies
Other Strength
114
Reproduced
with permission

Union Relations Union Relations Concern
No-layoff Policy Health and Safety Concern

Employee Cash Profit Sharing Workforce Reductions
Relations Employee Involvement Retirement Benefits Concern
Retirement Benefits Strength Other Concern
Health and Safety Strength
Other Strength
Beneficial Products and Services Hazardous Waste
Pollution Prevention Regulatory Problems
Recycling Ozone Depleting Chemicals
Environment Clean Energy Substantial Emissions
Communications of Environmental Agricultural Chemicals
Performance Climate Change
Property, Plant and Equipment Other Concern
Other Strength
Indigenous Peoples Relations Strength Controversies in Mexico
Human Rights Labor Rights Strength Indigenous Peoples Relations Concern

Other Strength Labor Rights Concern
Other Concern
Quality Superiority Product Safety Concern
Product R&D / Innovation Marketing / Contracting Concern
Benefits to Economically Disadvantaged Antitrust
Other Strength Other Concern
115
Reproduced
with permission

Licensing, manufacturing, retaihng of alcohol
Alcohol and ownership relation to an alcohol

company
Licensing, manufacturing, supporting of
Gambling gambling products and services and
ownership relation to a gambling company
Licensing, manufacturing, retailing of tobacco
Tobacco products and ownership relation to a tobacco
company
Manufacturing and retailing of firearms and
Firearms
ownership relation to a firearms company
Manufacturing of weapons and weapon systems,
Military
ownership relation to a military company
Construction of nuclear power plants, supplying
Nuclear nuclear power fuel, parts, and services, and
ownership relation to a nuclear company
116
Appendix 2. Alphabetical List of Chains (Essay 3)
Founded/
Initial
First Number Initial
Nam e C uisine Type Location-
Franchis of Stores Location-City
State
ed
Fastfood
Arby's 1964 3218 Burgers Boardman OH
Backyard Burgers 1987 123 Burgers Cleveland MS
Baskin Robbins 1946 1785 Ice Cream Glendale CA
Burger King 1954 7004 Burgers Miami FL
Blimpie 1964 1464 Sub Hoboken NJ
Bojangle’s 1977 316 Burgers Charlotte NC
Boston Market 1985 640 Fastfood Newton MA
Braum’s 1968 278 Ice Cream OK
Bruegger's 1983 253 Bagel VT
Captain D's 1969 563 S eafood D onelson TN
Caribou 1990 117 C offee Minneapolis MN
Carl's Jr 1956 884 Fastfood Anaheim CA
Checker's 385 Fastfood
Chick-fil-a 1967 895 Chicken Atlanta GA
Church's 1952 1060 Chicken San Antonio TX
Cici's 1985 436 Pizza Plano TX
D'angelo's 1967 11 Fastfood Dedham MA
Del Taco 1964 392 Mexican Barstow CA
Domino's 1967 4890 Pizza Ypsilanti Ml
Donato's 1963 191 Pizza Columbus OH
Dairy Q ueen 1940 4503 Frozen Yogurt Joliet IL
Dunkin Donuts 1950 3131 Donut Quincy MA
Einstein Bros Bagel 1993 362 Bagel N ew York NY
El Polio Loco 1980 308 Chicken LA CA
Fazolis 1988 365 Italian Lexington KY
Hardee's 1961 1938 Burgers Rocky Mount NC
Hungry Howie 1983 446 Pizza Taylor Ml
In N Out 1948 162 Burgers Baldwin Park CA
Jack in the Box 1951 1901 Fastfood San Diego CA
Point
Jersey Mike's 1956 267 Sub NJ
Pleasant
KFC 1952 5309 Chicken Corbin KY
Founded/ Number Initial
Initial
N am e First of Cuisine Type Location-
Location-City
Franchised S tores State
Krystal's 1932 302 Fastfood C hattanooga TN
L ees Fam ous
151 Chicken
R ecipe Chicken
Little C aesar's 1959 1821 Fastfood Garden City Ml
Long John Silver 1969 1174 S eafood Louisville KY
McDonald's 1955 13276 Burgers D es P laines IL
Miami Sub 1983 81 Sub Key W est FL
Mr. Gatti's 150 Fastfood
Mrs. Winner 1977 118 Fastfood Nashville TN
Noah Bagel 46 Bagel
Papa John's 1984 2529 Pizza Jeffersonville IN
P apa Murphy's 1988 805 Fastfood Petalum a CA
Peter Piper 1975 114 Pizza G lendale AZ
P opeye's 1972 1368 Chicken N ew Orleans LA
Quizno's 1981 2436 Sub Denver CO
Rally's 377 Fastfood
Seattle's Best
1971 C offee Seattle WA
C offee 77
Schlotzsky’s 1971 521 Fastfood Austin TX
Sonic 1954 2690 Burgers S h a w n ee OK
Starbucks 1985 5023 C offee Seattle WA
Subway 1965 15148 Sub Bridgeport CT
Taco Bell 1964 5460 Mexican Dow ney CA
Taco Bueno 1967 128 Mexican Abilene TX
Taco C abana 1978 135 Mexican San Antonio TX
Taco Time 1962 2 17 Mexican T acom a WA
TCBY 1981 693 Frozen Yogurt Little Rock AR
W endy's 1972 5587 Burgers Columbus OH
Corpus
What a Burger 1950 625 Burgers TX
Christi
White C astle 1921 249 Fastfood Wichita KS
118
Founded/
Initial
First Initial
Nam e Nf u ™b e r Cuisine Type Location-
Franchis of Stores Location-City
State
ed
Sitdown
A pplebee's 1983 1521 Pub/ Grill Atlanta GA
A zteca Mexican 1974 40 Mexican Seattle WA
Bakers Square 1983 142 Family D es Moines IA
Bennigan's 1976 285 Pub/ Grill
Bickford's Family
1959 54 Family P eabody MA
R estaurants
Bob Evans 1962 517 Family Rio Grande OH
Bonanza 1969 70 Family
Buca di Beppo 1993 90 Italian Minneapolis MN
California Pizza
1985 146 Various LA CA
Kitchen
Carraba’s 1986 154 italian Houston TX
Carrows 1970 111 Family Santa Clara CA
Chart H ouse 1961 27 American A spen CO
C h e ese ca k e
1972 75 American Los A ngeles CA
Factory
Chi Chi's 1977 78 Mexican
Chili's 1983 823 Pub/ Grill
Claim Jumper 1977 32 American Los Alamitos CA
Orange
C oco's 1948 137 Family CA
County
Copeland's of New
1983 33 Various N ew Orleans LA
Orleans
Country Kitchen 1958 2 57 Family Cincinnati OH
Cracker Barrel 1969 504 Family Lebanon TN
Damon's 1979 89 Pub/ Grill Columbus OH
Denny's 1953 1538 Family Lakewood CA
Don Pablo's 1985 66 Mexican Lubbock TX
Eat'n Park 1949 66 Family Pittsburgh PA
Fam ous Dave's 1994 102 Various Hayward Wl
First W atch 1983 48 Family
Friendlys 1935 519 Family Springfield MA
Frisch's 1939 102 Family Cincinnati OH
Fuddrucker's 1980 186 Pub/ Grill Sari Antonio TX
Furr's 1947 59 Family
Godfather's
1973 582 Chicken O m aha NE
Chicken
Golden Corral 1973 458 Family
Ground Round 1969 89 Pub/ Grill
Hard Rock C afe 1982 17 Pub/ Grill London U.K.
119
Initial
N am e First of C uisine Type Location-
Location-City
Franchised Stores State
Hops 1988 50 Pub/ Grill Clearwater FL
Houlihan’s 1972 82 Pub/ Grill K ansas City MO
IHOP 1958 1152 Family Los A ngeles CA
II Fornaio 1997 25 Italian Burlingame CA
J. Alexander's 1991 25 American Nashville TN
Legal S e a Foods 1968 27 S eafood Cambridge MA
Logan's R oadhouse 1991 120 Pub/ Grill
Winston-
Lone Star 1989 255 Steak NC
Salem
Luby’s 1965 162 Family S an Antonio TX
M aggiano's Little
1991 28 Italian
Italy
Marie C alendar's 1964 142 Family LA area CA
Max and Erma's 1975 90 American Dayton OH
Mimi's Cafe 1978 91 American Anaheim CA
Ninety Nine
1952 84 American Woburn MA
Restaurant & Pub
O'Charley's 1985 210 American
Old County Buffet 1983 179 Buffet
Old Spaghetti
1969 38 Italian Portland OR
Factory
Olive Garden 1982 543 Italian
Original Pancake
1953 87 Family Portland OR
H ouse
Outback
1988 762 Steak Tam pa FL
S teakh ouse
Perkin's 1958 480 Family Cincinnati OH
PF Chang's China
1993 106 Various
Bistro
Pizza Hut 1958 6731 pizza
Pizzaria Uno 1943 141 Italian Chicago IL
P onderosa 1969 342 Family
Red Lobster 1968 6 62 S eafood Lakeland FL
Red Robin 1969 211 Pub/ Grill Seattle WA
Pem broke
R oadhouse Grill 1993 77 Pub/ Grill FL
P ines
R omano's Macaroni
1989 196 Italian
Grill
Round Table Pizza 1959 483 Family Menlo Park CA
Rubio's Fresh
1983 143 Mexican San Diego CA
Mexican Grill
Ruby T uesday 1972 696 Pub/ Grill Knoxville TN
Ruby's Diner 1982 19 Family Newport CA
Ruth's Chris Steak 1965 79 Steak N ew Orleans LA
Shari's 1978 101 Family Hermiston OR
Shell 1986 28 Family Tam pa FL
120
Initial
N am e First of Cuisine Type Location-
Location-City
Franchised Stores State
Shoney's 1953 358 Family Charleston WV
Sizzler's 1958 225 Family Culver City CA
Sonny's Real Pit
1968 67 Various G ainesville FL
BBQ
Spaghetti
1972 17 Family Dallas TX
W arehouse
Steak and Ale 1967 63 Steak
Steak n Shake 1934 395 Family Normal IL
Stuart Anderson's 1964 30 Steak Seattle WA
S w eet T om ato 1978 64 Family San Diego CA
TGIF 1965 505 Pub/ Grill N ew York NY
Tony Roma's 1972 132 Pub/ Grill Miami FL
Tum blew eed's 1975 54 Family N ew Albany IN
Village Inn 1958 313 Family Denver CO
Waffle H ouse 1955 1392 Family Atlanta GA
BIBLIOGRAPHY
Adams, Renee B. 2003. "What Do Boards Do? Evidence from Board Committee and
Director Compensation Data." Federal Reserve Bank o f New York Working Paper.
Adamson, Jim. 2000. The Denny's Story: How a Company in Crisis Resurrected Its Good
Time: John Wiley & Sons, Inc.
Agle, Bradley R., Ronald K. Mitchell, and Jeffrey A. Sonnenfeld. 1999. "Who Matters to
CEOs? An Investigation of Stakeholder Attributes and Salience, Corporate
Performance, and CEO Values." The Academy o f Management Journal, 42:5, pp.
507-25.
Antonelli, Cristiano. 1997. "The Economics of Path-dependence in Industrial
Organization." International Journal o f Industrial Organization, 15:643-675.
Arrow, Kenneth J. 2000. "Increasing Returns: Historiographic Issues and Path
Dependence." European Journal o f Economic Thought, 7:2, pp. 171-80.
Arthur, Brian W. 1989. "Competing Technologies, Increasing Returns, and Lock-in by
Historical Events." The Economic Journal, 99, pp. 116-31.
Bagnoli, Marc and Susan Watts. 2003. "Selling to Socially Responsible Consumers:
Competition and the Private Provision of Public Goods." Journal o f Economics &
122
Management Strategy, 12:3, pp. 419-45.
Bailey, Elizabeth E. and Constance E. Helfat. 2003. "External Management Succession,
Human Capital, and Firm Performance: An Integrative Analysis." Managerial
and Decision Economics, 24, pp. 347-69.
Bailey, Elizabeth E. and Constance E. Helfat. 2005. "External Succession and Disruptive
Change: Heirs-apparent, Forced Turnover and Firm Performance." Strategic
Organization, 3:1, pp. 47-83.
Baldwin, Carliss Y. and Kim B. Clark. 2006. "Architectural Innovation and Dynamic
Competition: The Smaller "Footprint" Strategy." HBS Working Paper 07-014.
Baltagi, Badi H. 2001. Econometric Analysis o f Panel Data'. John Wiley & Sons, Ltd.
Barnett, Michael L. and Robert M. Salomon. 2006. "Beyond Dichotomy: The Curvilinear
Relationship between Social Responsibility and Financial Performance." Strategic
Management Journal, 27:11, pp. 1101-22.
Barney, Jay. 1991. "Firm Resources and Sustained Competitive Advantage." Journal o f
Management, 17, pp. 771-92.
Baron, David P. 1995. "Integrated Strategy: Market and Nonmarket Components."
California Management Review, 37, pp. 73-85.
Baron, David P. 1999. "Integrated Market and Nonmarket Strategy in Client and Interest
Group Politics." Business and Politics, 1:April, pp. 3-34.
123
Baron, David P. 2001. "Private Politics, Corporate Social Responsibility, and Integrated
Strategy." Journal o f Economics & Management Strategy, 10:1, pp. 7-45.
Baron, David P. 2006. Business and Its Environment'. Pearson Prentice Hall.
Basker, Emek. 2005. "Job Creation or Destruction? Labor-Market Effects of Wal-Mart
Expansion." Review o f Economics and Statistics, 87:1, pp. 174-83.
Berman, Shawn L., Andrew C. Wicks, Suresh Kotha, and Thomas M. Jones. 1999. "Does
Stakeholder Orientation Matter? The Relationship between Stakeholder
Management Models and Firm Financial Performance." The Academy o f
Berry, Steven T. and Joel Waldfogel. 1999. "Free Entry and Social Inefficiency in Radio
Broadcasting." RAND Journal o f Economics, 30:3, pp. 397-420.
Berry, Steven T. and Joel Waldfogel. 2001. "Do Mergers Increase Product Variety?
Evidence from Radio Broadcasting." Quarterly Journal o f Economics, 116:3, pp.
1009-25.
Bertrand, Marianne and Antoinette Schoar. 2003. "Managing with Style: The Effect of
Managers on Firm Policies." The Quarterly Journal o f Economics, 118:4, pp.
1169-208.
Bilimoria, Diana and Sandy Kristin Piderit. 1994. "Board Committee Membership:
Effects of Sex-based Bias." The Academy o f Management Journal, 37:6, pp.
124
1453-77.
Bloom, Paul N., Steve Hoeffler, Kevin L. Keller, and Carlos E. B. Meza. 2006. "How
Social-Cause Marketing Affects Consumer Perceptions." M IT Sloan Management
Review, 47:2, pp. 49-55.
Bowman, Edward H. and Mason Haire. 1975. "A Strategic Posture Toward Corporate
Social Responsibility." California Management Review, 18:2, pp. 49-58.
Brammer, S. J. and S. Pavelin. 2006. "Corporate Reputation and Social Performance: The
Importance of Fit." Journal o f Management Studies, 43:3, pp. 435-55.
Brammer, S. and A. Millington. 2005. "Corporate Reputation and Philanthropy: An
Empirical Analysis.” Journal o f Business Ethics, 61:1, pp. 29-44.
Bresnahan, Timothy and Peter Reiss. 1991. "Entry and Competition in Concentrated
Markets." Journal o f Political Economy, 99, pp. 977-1009.
Brinker, Norman and Donald T. Phillips. 1996. On the Brink: The Life and Leadership o f
Norman Brinker. Arlington, TX: The Summit Publishing Group.
Cameron, A. Colin and Pravin K. Trivedi. 2005. Microeconometrics: Methods and
Applications. Cambridge: Cambridge University Press.
Carhart, Mark M. 1997. "On Persistence in Mutual Fund Performance." The Journal o f
Finance, 52:1, pp. 57-82.
Carlsson, Bo. 1997. "On and Off the Beaten Path: the Evolution of Four Technological
125
Systems in Sweden." International Journal o f Industrial Organization, 15, pp.
775-99.
Chandler, Alfred D. 1969. Strategy and Structure: Chapters in the History o f American
Industrial Enterprise. Cambridge, MA: The MIT Press.
Chhaochharia, Vidhi and Yaniv Grinstein. 2006. "CEO Compensation and Board
Structure." NBER Summer Institute working paper.
Coffey, Betty S. and Jia Wang. 1998. "Board Diversity and Managerial Control as
Predictors of Corporate Social Performance." Journal o f Business Ethics, 17, pp.
1595-603.
Cohen, W. M. and R. C. Levin. 1989. "Ch. 18 Empirical Studies of Innovation and
Market Structure," in Handbook o f Industrial Organization. Richard Schmalensee,
and Robert D. Willig ed. Amsterdam: Elsevier.
Cowan, Robin and Philip Gunby. 1996. "Sprayed to Death: Path Dependence, Lock-in
and Pest Control Strategies." The Economic Journal, 106:May, pp. 521-42.
Daily, Catherine M. 1996. "Governance Patterns in Bankruptcy Reorganizations."
Strategic Management Journal, 17:5, pp. 355-75.
Dalton, Dan R., Catherine M. Daily, Alan E. Ellstrand, and Jonathan L. Johnson. 1998.
"Meta-analytic Reviews of Board Composition, Leadership Structure, and
Financial Performance." Strategic Management Journal, 19, pp. 269-90.
126
David, Paul A. 1985. "Clio and the Economics of QWERTY." The American Economic
Review, 75:2, pp. 332-37.
David, Paul A. 1986. "Understanding of the Economics of QWERTY: The Necessity of
History," in Economic History and Modern Economist. W. N. Parker ed. New
York: Basil Blackwell.
Davis, Gerald F. and Henrich R. Greve. 1997. "Corporate Elite Networks and
Governance Changes in the 1980s." The American Journal o f Sociology, 103:1,
p p . 1-37.
Davis, Peter. 2006. "Spatial Competition in Retail Markets: Movie Theaters." RAND
Journal o f Economics ', forthcoming.
Deckop, John R., Kimberly K. Merriman, and Shruti Gupta. 2006. "The Effects of CEO
Pay Structure on Corporate Social Performance." Journal o f Management, 32:3,
pp. 329-42.
Dilnersoz, Emin M. 2004. "Firm Organization and the Structure of Retail Markets."
Journal o f Economics & Management Strategy, 13:2, pp. 207-40.
Dixit, A. 1992. "Investment and Hysterisis." Journal o f Economic Perspectives, 6:1, pp.
107-32.
Dixit, A. 1995. "Irreversible Investment with Uncertainty And Scale Economies."
Journal o f Economic Dynamics & Control, 19:1-2, pp. 327-50.
127
Donaldson, Thomas and Lee E. Preston. 1995. "The Stakeholder Theory of the
Corporation: Concepts, Evidence, and Implications." Academy o f Management
Review, 20:1, pp. 65-91.
Dyer, Jeffrey H and Harbir Singh. 1998. "The Relational View: Cooperative Strategy
and Sources of Interorganizational Competitive Advantage." Academy o f
Management Review, 23:4, pp. 660-79.
Eesley, Charles and Michael Lenox. 2006. "Firm Responses to Secondary Stakeholder
Action." Strategic Management Journal, 27:8, pp. 765-81.
Ellickson, Paul B. 2004. "Sunk Costs and Supermarkets." Duke University Economics
Department Working Paper.
Elsayed, Klaled and David Paton. 2005. "The Impact of Environmental Performance on
Financial Performance: Static and Dynamic Panel Data Evidence." Structural
Change and Economic Dynamics, 16:3, pp. 395-412.
Fama, Eugene F. and Kenneth R. French. 1993. "Common Risk Factors in the Returns on
Stocks and Bonds." Journal o f Financial Economics, 33, pp. 3-56.
Fama, Eugene F. and Kenneth R. French. 1997. "Industry Costs of Equity." Journal o f
Financial Economics, 43, pp. 153-93.
Fama, Eugene F. and Michael C. Jensen. 1983. "The Separation of Ownership and
Control." Journal O f Law & Economics, 26, pp. 301-25.
128
Fama, Eugene F. and James D. MacBeth. 1973. "Risk, Return, and Equilibrium:
Empirical Test." The Journal o f Political Economy, 81:3, pp. 607-36.
Fisher, Anne. 2006. "America's Most Admired Companies." Fortune, Vol. 153: 65-76.
Freeman, R. Edward. 1999. "Divergent Stakeholder Theory." Academy o f Management
Review, 24:2, pp. 233-36.
Freeman, R. Edward. 1991. Business Ethics: the State o f the Art: Oxford University Press.
Friedman, Milton. 1970. "The Social Responsibility of Business Is to Increase Its
Profits." The New York Times Magazine.
Frooman, Jeff. 1997. "Socially Irresponsible and Illegal Behavior and Shareholder
Wealth: A Meta-Analysis of Event Studies." Business & Society, 36:3, pp. 221-49.
Fudenberg, Drew and Jean Tirole. 1991. Game Theory. Cambridge, MA: MIT Press.
Fujita, Masahisa, Paul Krugman, and Anthony J. Venables. 1999. The Spatial Economy.
Cambridge, MA: MIT Press.
George, Lisa and Joel Waldfogel. 2002. "Does the New York Times Spread Ignorance
and Apathy?" Working Paper, The Wharton School, University o f Pennsylvania.
George, Lisa and Joel Waldfogel. 2003. "Who Affects Whom in Daily Newspaper
Markets?" Journal o f Political Economy, 111:4, pp. 765-84.
Goldberg, Jeffrey. 2007. "Selling Wal-Mart." The New Yorker.
129
Gompers, Paul, Joy Ishii, and Andrew Metrick. 2003. "Corporate Governance and Equity
Prices." The Quarterly Journal o f Economics, 118, pp. 107-55.
Goodpaster, Kenneth E. 1991. "Business Ethics and Stakeholder Analysis." Business
Ethics Quarterly, 1:1, pp. 53-73.
Griffin, Jennifer. J. and John F. Mahon. 1997. "The Corporate Social Performance and
Corporate Financial Performance Debate: Twenty-five Years of Incomparable
Research." Business & Society, 36:March, pp. 5-31.
Guner, A. Burak, Ulrike Malmendier, and Geoffrey Tate. 2006. "The Impact of Board
with Financial Expertise on Corporate Policies." NBER Working Paper.
Harrison, Jeffery S. and R. Edward Freeman. 1999. "Stakeholders, Social Responsibility,
and Performance: Empirical Evidence and Theoretical Perspectives." The
Academy o f Management Journal, 42:5, pp. 479-85.
Harvard Business Review. 2003. Harvard Business Review on Corporate Social
Responsibility. Cambridge, MA: HBS Press.
Hillman, Amy J. and Thomas Dalziel. 2003. "Boards of Directors and Firm Performance:
Integrating Agency and Resource Dependence Perspectives." Academy o f
Hillman, Amy J. and Gerald D. Keim. 2001. "Shareholder Value, Stakeholder
Management, and Social Issues: What's the Bottom Line?" Strategic Management
130
Journal, 22:2, pp. 125-39.
Hillman, Amy J., Gerald D. Keim, and Rebecca A. Luce. 2001. "Board Composition and
Stakeholder Performance: Do Stakeholder Directors Make a Difference?"
Business & Society, 40:3, pp. 295-314.
Hillman, Amy J., Asghar Zardkoohi, and Leonard Bierman. 1999. "Corporate Political
Strategies and Firm Performance: Indications of Firm-specific Benefits from
Personal Service in the U.S. Government." Strategic Management Journal, 20, pp.
67-81.
Hirschman, Albert O. 1970. Exit, Voice, and Loyalty: Responses to Decline in Firms,
Organizations, and States. Cambridge, MA: Harvard University Press.
Holmes, Thomas J. 2005. "The Diffusion of Wal-Mart and Economies of Density."
NBER Conference: Cambridge, Massachusetts.
Hong, Harrison and Marcin Kacperczyk. 2007. "The Price of Sin: The Effects of Social
Norms on Markets." Princeton University Working Paper.
Hotelling, Harold. 1929. "Stability in Competition." Economic Journal, 39, pp. 41-57.
Huang, Chi-fu and Robert H. Litzenberger. 1988. Foundations fo r Financial Economics:
Prentice Hall.
Husted, Bryan W. and Jose de Jesus Salazar. 2006. "Taking Friedman Seriously:
Maximizing Profits and Social Performance." Journal o f Management Studies,
131
43:1, pp. 75-91.
Ilinitch, Anne Y., Naomi S. Soderstrom, and Tom E. Thomas. 1998. "Measuring
Corporate Environmental Performance." Journal o f Accounting and Public Policy,
17, pp. 383-408.
Institutional Shareholder Services. 2006. "Corporate Governance: From Compliance
Obligation to Business Imperative." Global Institutional Investor Study.
Jia, Panle. 2006. "What Happens When Wal-mart Comes to Town: An Empirical
Analysis of the Discount Retailing Industry." Working paper, Department o f
Economics, Yale University.
Johnson, Richard A. and Daniel W. Greening. 1999. "The Effects of Corporate
Governance and Institutional Ownership Types on Corporate Social
Performance." The Academy o f Management Journal, 42:5, pp. 564-76.
Jones, Thomas M. 1995. "Instrumental Stakeholder Theory: A Synthesis of Ethics and
Economics." Academy o f Management Review, 20:2, pp. 404-37.
Jones, Thomas M. and Andrew C. Wicks. 1999. "Convergent Stakeholder Theory."
Academy o f Management Review, 24:2, pp. 206-21.
Kale, Prashant, Jeffrey H Dyer, and Harbir Singh. 2002. "Alliance Capability, Stock
Market Response, and Long-term Alliance Success: the Role of the Alliance
Function." Strategic Management Journal, 23, pp. 747-67.
132
King, Andrew and Michael Lenox. 2001. "Does It Really Pay to Be Green? An Empirical
Study of Firm Environmental and Financial Performance." Journal o f Industrial
Ecology, 5:1, pp. 105-16.
King, Andrew and Michael Lenox. 2002. "Exploring the Locus of Profitable Pollution
Reduction." Management Science, 48:2, pp. 289-99.
Klein, April. 1998. "Firm Performance and Board Committee Structure." Journal o f Law
and Economics, 41, pp. 275-303.
Klein, Benjamin and Keith B. Leffler. 1981. "The Role of Market Forces in Assuring
Contractual Performance." Journal o f Political Economy, 89:4, pp. 615-41.
Klepper, Steven. 2002a. "The Capabilities of New Firms and the Evolution of the US
Automobile Industry." Industrial and Corporate Change, 11:4, pp. 645-66.
Klepper, Steven. 2002b. "Firm Survival and the Evolution of Oligopoly." RAND Journal
o f Economics, 33:1, pp. 37-61.
Klepper, Steven and Elizabeth Graddy. 1990. "The Evolution of New Industries and the
Determinants of Market Structure." RAND Journal o f Economics, 21, pp. 27-44.
Klepper, Steven and Kenneth L. Simons. 2000. "Dominance by Birthright: Entry of Prior
Radio Producers and Competitive Ramifications in the U.S. Television Receiver
Industry." Strategic Management Journal, 21:997-1016.
Kogut, B. and N. Kulatilaka. 1994. "Operating Flexibility, Global Manufacturing, And
133
The Option Value O f A Multinational Network." Management Science, 40:1, pp.
123-39.
Kotchen, Matthew J. 2006. "Voluntary Provision of Public Goods for Bads: A Theory of
Environmental Offset." University o f California, Santa Barbara Working Paper.
Krugman, Paul. 1991a. "History and Industry Location: The Case of the Manufacturing
Belt." The American Economic Review, 81:2, pp. 80-83.
Krugman, Paul. 1991b. "History Versus Expectations." The Quarterly Journal o f
Economics, 106:2, pp. 651-67.
Krugman, Paul. 1991c. "Increasing Returns and Economic Geography." The Journal o f
Political Economy, 99:3, pp. 483-99.
Krugman, Paul. 1996. The Self-Organizing Economy. Oxford, UK: Blackwell.
Lafontaine, Francine. 1995. "Pricing Decisions in Franchised Chains: A Look at the
Restaurant and Fast-Food Industry." NBER Working Paper 5247.
Levinthal, Daniel A. and James G. March. 1993. "The Myopia of Learning." Strategic
Management Journal, 14, pp. 95-112.
Liebowitz, Stan J. and Stephen E. Margolis. 1990. "The Fable of the Keys." Journal o f
Law and Economics, 33:1, pp. 1-25.
Liebowitz, Stan J. and Stephen E. Margolis. 1999. Winners, Losers, and Microsoft:
134
Competition and Antitrust in High Technology: Independent Institute.
Lindenberg, Eric B. and Stephen A. Ross. 1981. "Tobin's q Ratio and Industrial
Organization." The Journal o f Business, 54:1, pp. 1-32.
Love, John F. 1995. McDonald's: Behind the Arches: Bantam Books.
Luo, Xueming and C. B. Bhattacharya. 2006. "Corporate Social Responsibility, Customer
Satisfaction, and Market Value." Journal o f Marketing, 70:October, pp. 1-18.
Luoma, Patrice and Jerry Goodstein. 1999. "Stakeholders and Corporate Boards:
Institutional Influence on Board Composition and Structure." The Academy o f
Mahon, John F. and Jennifer. J. Griffin. 1999. "Painting a Portrait: A Reply." Business &
Society, 38:March, pp. 126-33.
Mankiw, N. Gregory and Michael D. Whinston. 1986. "Free Entry and Social
Inefficiency." RAND Journal o f Economics, 17:1, pp. 48-58.
Margolis, Joshua D. and James P. Walsh. 2001. People and Profits? The Search fo r a
Link between a Company's Social and Financial Performance. Mahwah, NJ:
Lawrence Erlbaum.
Margolis, Joshua D. and James P. Walsh. 2003. "Misery Loves Companies: Rethinking
Social Initiatives by Business." Administrative Science Quarterly, 48, pp. 268-305.
Markovitch, Dmitri G., Joel H. Steckel, and Bernard Yeung. 2005. "Using Capital
135
Markets as Market Intelligence: Evidence from the Pharmaceutical Industry."
Management Science, 51:10, pp. 1467-80.
Mattingly, James E. and Shawn L. Berman. 2006. "Measurement of Corporate Social
Action." Business & Society, 45:1, pp. 20-46.
Mazzeo, Michael J. 2002. "Product Choice and Oligopoly Market Structure.” RAND
Journal o f Economics, 33:2, pp. 221-42.
McWilliams, Abagail and Donald Siegel. 2000. "Corporate Social Responsibility and
Financial Performance: Correlation or Misspecification?" Strategic Management
Journal, 21:5, pp. 603-09.
McWilliams, Abagail and Donald Siegel. 2001. "Corporate Social Responsibility: A
Theory of the Firm Perspective." Academy o f Management Review, 26, pp. 117-
27.
McWilliams, Abagail, Donald Siegel, and Patrick M. Wright. 2006. "Corporate Social
Responsibility: Strategic Implications." Journal o f Management Studies, 43:1, pp.
1-18.
Moore, Geoff. 2001. "Corporate Social and Financial Performance: An Investigation in
the U.K. Supermarket Industry." Journal o f Business Ethics, 34, pp. 299-315.
Mueller, Dennis C. 1997. "First-mover Advantages and Path Dependence." International
Journal o f Industrial Organization, 15, pp. 827-50.
136
Nelson, Richard R. 1995. "Co-evolution of Industry Structure, Technology and
Supporting Institutions, and the Making of Comparative Advantage."
International Journal o f the Economics o f Business, 2:2, pp. 171-84.
Nelson, Richard R. and Sidney G. Winter. 1982a. An Evolutionary Theory o f Economic
Change. Cambridge, MA: The Belknap Press of Harvard University Press.
Nelson, Richard R. and Sidney G. Winter. 1982b. "The Schumpeterian Tradeoff
Revisited." The American Economic Review, 72:1, pp. 114-32.
Ogden, Stuart and Robert Watson. 1999. "Corporate Performance and Stakeholder
Management: Balancing Shareholder and Customer Interests in the U.K.
Privatized Water Industry." The Academy o f Management Journal, 42:5, pp. 526-
38.
Orlitzky, Marc, Frank L. Schmidt, and Sara L. Rynes. 2003. "Corporate Social and
Financial Performance: A Meta-analysis." Organization Studies, 24.3, pp. 403-41.
Podolny, Joel M. 1994. "Market Uncertainty and the Social Character of Economic
Exchange." Administrative Science Quarterly, 39:3, pp. 458-83.
Porter, Michael E. and Mark R. Kramer. 2002. "The Competitive Advantage of Corporate
Philanthropy." Harvard Business Review, 80:12, pp. 56-68.
Porter, Michael E. and Mark R. Kramer. 2006 "Strategy & Society: The Link between
Competitive Advantage and Corporate Social Responsibility." Harvard Business
137
Review, 84:12, pp. 78-92.
Puffert, Douglas J. 1999. "Path Dependence in Economic History," in Handbook o f
Evolutionary Economics.
Puffert, Douglas J. 2000. "Path Dependence, Network Form, and Technological Change."
History Matters: Economic Growth, Technology, and Population: Stanford
University.
Puffert, Douglas J. 2002. "Path Dependence in Spatial Networks: The Standardization of
Railway Track Gauge." Explorations in Economic History, 39:3, pp. 282-314.
Raff, Daniel M. G. 2000. "Superstores and the Evolution of Firm Capabilities in
American Bookselling." Strategic Management Journal, 21, pp. 1043-59.
Roman, Ronald M., Sefa Hayibor, and Bradley R. Agle. 1999. "The Relationship
between Social and Financial Performance: Repainting a Portrait." Business &
Society, 38:March, pp. 109-25.
Romano, Roberta. 2005. "The Sarbanes-Oxley Act and the Making of Quack Corporate
Governance." The Yale Law Journal, 114:7, pp. 1521-611.
Rowley, Tim and Shawn L. Berman. 2000. "A Brand New Brand of Corporate Social
Performance." Business & Society, 39:4, pp. 397-418.
Ruf, Bernadette M., Krishnamurty Muralidhar, Robert M. Brown, Jay J. Janney, and
Karen Paul. 2001. "An Empirical Investigation of the Relationship between
138
Change in Corporate Social Performance and Financial Performance: A
Stakeholder Theory Perspective." Journal o f Business Ethics, 32:2, pp. 143-56.
Russo, Michael V. and Paul A. Fouts. 1997. "A Resource-based View Perspective on
Corporate Environmental Performance and Profitability." The Academy o f
Schlosser, Eric. 2002. Fast Food Nation'. Perennial.
Schuler, Douglas A. and Margaret Cording. 2006. "A Corporate Social Performance -
Corporate Financial Performance Behavioral Model for Consumers." Academy o f
Schultz, Howard and Dori Jones Yang. 1997. Pour Your Heart into It: How Starbucks
Built a Company One Cup at a Time. New York: Hyperion.
Seim, Katja. 2006. "An Empirical Model of Firm Entry with Endogenous Product-Type
Choices." Rand Journal o f Economics, forthcoming.
Shaffer, Brian, Thomas J. Quasney, and Curtis M. Grimm. 2000. "Firm Level
Performance Implications of Nonmarket Actions." Business & Society, 39:2, pp.
126-43.
Shane, Scott A. 2005. From Ice Cream to the Internet: Using Franchising to Drive the
Growth and Profits o f Your Company: Prentice Hall.
Siegel, Donald and Donald F. Vitaliano. forthcoming. "An Empirical Analysis of the
139
Strategic Use of Corporate Social Responsibility." Journal o f Economics &
Management Strategy.
Sinai, Todd and Joel Waldfogel. 2004. "Geography and the Internet: Is the Internet a
Substitute or a Complement for Cities?" Journal o f Urban Economics, 56:1, pp.
1-24.
Singh, Harbir and Farid Harianto. 1989. "Top Management Tenure, Corporate Ownership
Structure and Magnitude of Golden Parachutes." Strategic Management Journal,
10:Summer Special Issue, pp. 143-56.
Smith, Warren and Matthew Higgins. 2000. "Cause-Related Marketing: Ethics and the
Ecstatic." Business <&Society, 39:3, pp. 304-22.
Social Investment Forum. 2006. "2005 Report on Socially Responsible Investing Trends
in the United States - 10-YEAR REVIEW."
Spence, Michael. 1976. "Product Selection, Fixed Costs, and Monopolistic Competition."
The Review o f Economic Studies, 43:2, pp. 217-35.
Sutton, John. 1991. Sunk Cost and Market Structure. Cambridge, MA: MIT Press.
Sutton, John. 1994. "History Matters. So What?" Journal o f the Economics o f Business,
1:1, pp. 41-44.
Sutton, John. 1998. Technology and Market Structure. Cambridge, MA: MIT Press.
Swanson, Diane L. 1999. "Toward an Integrative Theory of Business and Society: A
140
Research Strategy for Corporate Social Performance." Academy o f Management
Review, 24:3, pp. 506-21.
Szulanski, Gabriel. 2000. "The Process of Knowledge Transfer: A Diachronic Analysis
of Stickiness." Organizational Behavior and Human Decision Processes, 82:1, pp.
9-27.
Tiebout, Charles M. 1956. "A Pure Theory of Local Expenditures." The Journal o f
Political Economy, 64:5, pp. 416-24.
Tirole, Jean. 2001. "Corporate Governance." Econometrica, 69:1, pp. 1-35.
Tirole, Jean. 2005. The Theory o f Corporate Finance: Princeton University Press.
Troske, K. R. 1996. "The Dynamic Adjustment Process of Firm Entry and Exit in
Manufacturing and Finance, Insurance, and Real Estate." Journal o f Law &
Economics, 39:2, pp. 705-35.
Vogel, David. 2005. The Market fo r Virtue: The Potential and Limits o f Corporate Social
Responsibility. Washington D.C.: Brookings.
Waddock, Sandra A. and Samuel B. Graves. 1997a. "The Corporate Social Performance -
Financial Performance Link." Strategic Management Journal, 18:4, pp. 303-19.
Waddock, Sandra A. and Samuel B. Graves. 1997b. "Quality of Management and Quality
of Stakeholder Relations." Business & Society, 36:3. pp. 250-79.
Waldfogel, Joel. 2003. "Preference Externalities: An Empirical Study of Who Benefits
141
Whom in Differentiated Product Markets." RAND Journal o f Economics, 34:3, pp.
557-68.
Waldfogel, Joel. 2004a. "The Median Voter and the Median Consumer: Local Private
Goods and Residential Sorting." Working paper, Wharton School, University o f
Pennsylvania.
Waldfogel, Joel. 2004b. "Who Benefits Whom in Local Television Markets?" Brookings-
Wharton Papers on Urban Affairs, 2004:5, pp. 257-305.
Waldfogel, Joel. 2007. The Tyranny o f the Market: Why You Can’t Always Get What You
Want. Cambridge, MA: Harvard University Press.
Weaver, Gary R., Linda K. Trevino, and Philip L. Cochran. 1999. "Integrated and
Decoupled Corporate Social Performance: Management Commitments, External
Pressures, and Corporate Ethics Practices." The Academy o f Management Journal,
42:5, pp. 539-52.
Werhane, P. H. and R. Edward Freeman. 2006. Business Ethics (The Blackwell
Encyclopedia o f Management)-. Blackwell Publishing Limited.
Wemerfelt, Birger. 1984. "A Resource-based View of the Firm." Strategic Management
Journal, 5, pp. 171-80.
Werther Jr., William B. and David Chandler. 2005. Strategic Corporate Social
Responsibility: Stakeholders in a Global Environment: Sage Publications, Inc.
142
White, Halbert. 1980. A Heteroskedasticity-consistent Covariance Matrix Estimator and a
Direct Test for Heteroskedasticity." Econometrica, 48, pp.817-838.
Windsor, Duane. 2006. "Corporate Social Responsibility: Three Key Approaches."
Journal o f Management Studies, 43:1, pp. 93-114.
Winter, Sidney G. and Gabriel Szulanski. 2001. "Replication as Strategy." Organization
Science, 12:6, pp. 730-43.
Witt, Ulrich. 1997. ""Lock-in" vs. "Critical Mass" - Industrial Change under Network
Externalities." International Journal o f Industrial Organization, 15, pp. 753-73.
Wood, Donna J. 1991a. "Corporate Social Performance Revisited." Academy o f
Management Review, 18, pp. 691-718.
Wood, Donna J. 1991b. "Social Issues in Management: Theory and Research in
Corporate Social Performance." Journal o f Management, 17:2, pp. 383-406.
Wood, Donna J. 1995. "The Fortune Database as a CSP Measure." Business & Society,
34:2, pp. 197-98.
Wood, Donna J. and Raymond E. Jones. 1995. "Stakeholder Mismatching: A Theoretical
Problem in Empirical Research on Corporate Social Performance." International
Journal o f Organizational Analysis, 3:3, pp. 229-67.
Wooldridge, Jeffrey M. 2002. Econometric Analysis o f Cross Section and Panel Data.
143
Cambridge, MA: MIT Press.
Yin, Haitao. 2006. "The Environmental and Economic Impacts of Environmental
Regulations: the Case of Underground Storage Tank Regulations." Ph. D.
dissertation, University o f Pennsylvania.

Business

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Business

Uploaded by

Copyright:

Available Formats

ESSAYS IN EMPIRICAL ANALYSIS OF CORPORATE STRATEGY

AND CORPORATE RESPONSIBILITY

Jon Jungbien Moon

Business and Public Policy

For the Graduate Group in Managerial Science and Applied Economics

Presented to the Faculties o f the University o f Pennsylvania

in Partial Fulfillment for the Requirements for the Degree of

All rights reserved.

ProQuest Information and Learning Company

Jon Jungbien Moon

To my maternal grandmother, Ms. Chung, Jung-Lim

My advisors, Elizabeth E. Bailey and Joel Waldfogel, have been my guides

I am also indebted to Harbir Singh for many thoughtful comments, stimulating

discussions, and continued guidance. I am thankful to Thomas Dunfee, Justin Wolfers,

Conference, and seminar participants at Wharton Applied Economics Workshop,

Shanghai University of Finance and Economics for their insightful comments.

(Congressional Joint Committee on Taxation), Sangyoung Song (Baruch College,

for always brightening my day.

ESSAYS IN EMPIRICAL ANALYSIS OF CORPORATE STRATEGY AND

Jon Jungbien Moon

Elizabeth E. Bailey and Joel Waldfogel

corporate social responsibility.

In my first essay, I investigate how corporate social performance (CSP) affects

the enactment of Sarbanes-Oxley Act as an instrumental variable. I argue that spending

boards of directors in integrated strategy, by establishing voluntary committees to handle

performance established such committees for mitigation purposes.

outcomes, rendering them path-dependent.

TABLE OF CONTENTS......................................................................................................... viii

ESSAY ONE. IN GOOD COMPANIES? A CRITICAL EVALUATION OF THE

CORPORATE SOCIAL PERFORMANCE - CORPORATE FINANCIAL

2. Literature and Hypotheses..................................................................................... 3

2.1. Two Views on the CSP-CFP Link................................................................ 3

2.2. Empirical Literature........................................................................................ 5

3.1. Data Sources.................................................................................................. 12

3.2. Social Performance Data...............................................................................13

3.4. Industry Classification..................................................................................22

3.5. Empirical Specification.................................................................................24

4. Analysis and Results.............................................................................................27

4.1. Companies in Top and Bottom D eciles..................................................... 27

4.2. Temporal Variation in the KLD D ata.........................................................29

4.4. Industry Interaction....................................................................................... 31

4.5. Fixed Effect Regression Results................................................................. 35

4.6. Addressing the Endogeneity Concern Further - IV Estimation Results 36

4.7. Asymmetry between the Effect of Positive and Negative C S P .............. 40

4.8. Controllable vs. Uncontrollable Components of Negative Social Ratings

ESSAY TWO: BOARD COMMITTEES AND INTEGRATED STRATEGY: HOW

FIRMS INTEGRATE MARKET AND NON-MARKET STRATEGIES FOR

3.1. Committee D ata............................................................................................ 54

3.2. Social Performance Data.............................................................................. 59

4. Analysis & Results................................................................................................65

4.1. Descriptive A nalysis.....................................................................................65

4.2. Regression Analysis......................................................................................68

5. Diffusion of a PA/SR Comm ittee.......................................................................71

5.1. Summary Statistics........................................................................................ 71

5.2. Regression Analysis......................................................................................75

2. Theoretical Background and Literature............................................................. 81

3.1. Geographic Market Definition and Market A ge........................................87

3.2. Chain Location and Zipcode Demographics.............................................. 89

3.3. History of Chain Restaurants.......................................................................92

4. Analysis & Results................................................................................................95

4.1. Descriptive A nalysis..................................................................................... 95

4.2. How Do Chain Founding Dates Correlate with Zip Vintage?...............102

4.3. Individual Chain’s Prevalence in Different Zip Vintages...................... 106