Does It Pay to Be Sustainable? Corporate Sustainability and Corporate Financial Performance


Material Information

Does It Pay to Be Sustainable? Corporate Sustainability and Corporate Financial Performance a Study Based on the Dow Jones Sustainability Index (Djsi)
Physical Description:
1 online resource (60 p.)
Al Abri, Ibtisam H
University of Florida
Place of Publication:
Gainesville, Fla.
Publication Date:

Thesis/Dissertation Information

Master's ( M.S.)
Degree Grantor:
University of Florida
Degree Disciplines:
Food and Resource Economics
Committee Chair:
Committee Co-Chair:
Committee Members:


Subjects / Keywords:
corporate-financial-performance -- corporate-sustainability
Food and Resource Economics -- Dissertations, Academic -- UF
Food and Resource Economics thesis, M.S.
bibliography   ( marcgt )
theses   ( marcgt )
government publication (state, provincial, terriorial, dependent)   ( marcgt )
born-digital   ( sobekcm )
Electronic Thesis or Dissertation


Several attempts have been made to analyze the benefits of creating long term value for shareholders and stakeholders through corporate sustainability (CS), however empirical evidence has been mixed. This thesis has considered some core issues like covering a relatively longer time frame and accounting for the recessionary economic conditions that dominated the selected period. The Propensity Score Matching method was used to select comparable firms for analysis. Then, standardized multi-period panel data on financial performance and First Difference and Difference in Difference methods were applied to address the objectives of the study. This study proved the existence of a positive association between financial performance and CS, and found that a period of eight years was sufficient to reap the benefits of CS investment. Second, the financial performance of firms was sensitive to the level of CS applied. Third, the recession of 2008-09 had an insignificant influence on the relationship between CS and corporate financial performance (CFP). Fourth, on average, firms continuously practicing CS had an 11.3% higher Tobin's q and a 7.1% higher return on assets (ROA) compared to firms that never practiced CS. In addition, on average, the firms that were persistently involved in CS had a higher Tobin's q by 6% and a higher ROA by 3.1% compared to those firms occasionally investing in such practices. The significant differences among the three groups of firms persisted since the beginning of the study period in 2000, which suggests that corporate managers can quickly capture the benefits of CS. Fifth, empirical analysis revealed that the Utilities, Retail Trade, Information and Services industries were more efficient in capturing the benefits of CS investment during the study period of 8 years. Finally, only the Retail Trade and Information industries were found to be sensitive to the level of CS applied by firms.
General Note:
In the series University of Florida Digital Collections.
General Note:
Includes vita.
Includes bibliographical references.
Source of Description:
Description based on online resource; title from PDF title page.
Source of Description:
This bibliographic record is available under the Creative Commons CC0 public domain dedication. The University of Florida Libraries, as creator of this bibliographic record, has waived all rights to it worldwide under copyright law, including all related and neighboring rights, to the extent allowed by law.
Statement of Responsibility:
by Ibtisam H Al Abri.
Thesis (M.S.)--University of Florida, 2014.
Co-adviser: GROGAN,KELLY A.

Record Information

Source Institution:
Rights Management:
Applicable rights reserved.
lcc - LD1780 2014
System ID:

This item is only available as the following downloads:

Full Text




2014 Ibtisam Al Abri


3 ACKNOWLEDGMENTS First of all, I would like to extend my appreciation and gratitude to the Department of Food and Resource Economics at the University of Florida for offering me the great opportunity for completion of my master 's degree. Similarly, my great thanks and sincere grati tude go to the Government of my country, Oman, for fully sponsoring my study and stay in U.S.A. Likewise, a special thankfulness goes to RobecoSAM Company and the Dow Jones Sustainability Indices for their effective cooperation toward the success of this work by providing us the required data. The compilation of this thesis could not have been possible without the continuous encouragement, support and valuable guidance of my committee chair Profe ssor Alan Hodges. My dear chair I wish to thank him sincerely for his faith, help and support. Additionally, my great thanks go to my committee member Dr. Zhifeng Gao. He has offered remarkable guidance and continuous cooperation in various aspects especi ally in the empirical analysis part of it. He has appreciated my ability of understanding the other half of a number. I would like also to express the deepest appreciation to my committee member Dr. Kelly Grogan who has shown the attitude and the substance of a genius. She continually and persuasively conveyed a spirit of adventure in regard to research, and an excitement in regard to teaching. Most gratefully, I would love to give my appreciation to my committee member Dr. Xiang Bi for her tremendous suppo rt, help and appreciable contribution throughout the period of conducting this research. Her patience and kindness have impressed me so deeply. In addition, I would like to thank Dr. Jaclyn Kropp, who has contributed significantly toward the improvement of the analysis on this thesis.


4 Last but by no mean the least, I would like to give my most grateful appreciation and gratitude to the dearest people intimately related to my soul; my family and my friends far too many to be mentioned individually for their belie f i n me and their support during my life in Gainesville.


5 TABLE OF CONTENTS page ACKNOWLEDGMENTS ................................ ................................ ................................ ............... 3 LIST OF TABLES ................................ ................................ ................................ ........................... 7 LIST OF FIGURES ................................ ................................ ................................ ......................... 8 LIST OF ABBREV IATIONS ................................ ................................ ................................ .......... 9 ABSTRACT ................................ ................................ ................................ ................................ ... 10 CHAPTER INTRODUCTION ................................ ................................ ................................ .................. 12 Preamble ................................ ................................ ................................ ................................ 12 Justification of the Research Problem and the Expected Contribution ................................ .. 13 Objectives and Hypotheses of the Study ................................ ................................ ................ 15 The Dow Jones Sustainability Index ................................ ................................ ...................... 16 The Recession of 2008 2009 ................................ ................................ ................................ .. 17 LITERATURE REVIEW ................................ ................................ ................................ ....... 19 Definition of Sustainability ................................ ................................ ................................ ..... 19 Corporate Financial Performance and Corporate Sustainability ................................ ............ 19 Correlation between CS and CFP: Three perspectives ................................ ........................... 21 METHODS ................................ ................................ ................................ ............................. 27 Sample Selection ................................ ................................ ................................ .................... 27 Variables an d Measures ................................ ................................ ................................ .......... 28 Dependent Variables ................................ ................................ ................................ ....... 28 Independent Variables ................................ ................................ ................................ ..... 29 Control Variables ................................ ................................ ................................ ............. 29 METHODOLOGY ................................ ................................ ................................ ................. 31 Propensity Score Matching Method (PSM) ................................ ................................ ............ 31 Analysis of the Relation between CS and CFP ................................ ................................ ...... 32 Application of the Multi Period Panel Data FD and DID Model: Model 1 .................... 34 Application of the Multi Period Panel Data FD and DID Model: Model 2 .................... 36 DATA CONSTRUCTION ................................ ................................ ................................ ..... 38 Analysis of Residuals and Exposing Outliers ................................ ................................ ......... 38 Non Linear Relationships and Multicolinearity ................................ ................................ ..... 39


6 Detecting the Problem of Autocorrelation and Heteroskedasticity ................................ ........ 40 EMPIRICAL ANALYSIS AND RESULTS ................................ ................................ .......... 44 Statistical Description ................................ ................................ ................................ ............. 44 Resu lts ................................ ................................ ................................ ................................ ..... 46 FINDINGS SUMMARY AND CONCLUSION ................................ ................................ .... 52 LIST OF REFERENCES ................................ ................................ ................................ ............... 56 BIOGRAPHIC AL SKETCH ................................ ................................ ................................ ......... 60


7 LIST OF TABLES Table page 2 1 Summary of previous studies ................................ ................................ ............................. 24 4 1 Multi period panel data FD and DID estimator ................................ ................................ 35 5 1 Pearson correlation coefficients and probability values for model variables .................... 40 6 1 Descriptive statistics of the sampled firms ................................ ................................ ........ 45 6 2 Standardized parameter results of model 1 an d model 2 ................................ ................... 51


8 LIST OF FIGURES Figure page 4 1 Histogram of residuals ................................ ................................ ................................ ....... 34 5 1 Probability plot of residuals from model1 ................................ ................................ ......... 38 5 2 Fitted value of squared residuals for (Tobin's q) ................................ ............................... 41 5 3 Fitted value of squared residuals for (1/Tobin's q) ................................ ............................ 42


9 LIST OF ABBREVIATIONS CFP Corporate Financial Performance CS Corporate Sustainability CSR Corporate Social Responsibility DID Difference In Difference DJGI Dow Jones Global Indexes DJSI Dow Jones Sustainability Index FD First Difference KLD Kinder, Lydenberg, Domini index of social performance MVA Market Value Added NASDAQ National Association of Securities Dealers Automated Quotations (American Stock Market) NYSE New York Stock Exchange OLS Ordinary Least Squares PSM Propensity Score Matching ROA Return On Assets ROE Return On Equity ROS Return On Sales SAM Sustainable Asset Management S&P 500 Standards and Poor's 500


10 Abstract of Thesis Presented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Master of Science DOES IT PAY TO BE SUSTAINABLE? CORPORATE SUSTAINABILITY AND CORPORATE FINANCIAL PERFORMANCE: A ST UDY BASED ON THE DOW JONES SUSTAINABILITY INDEX (DJSI) By Ibtisam Al Abri May 2014 Chair: Alan Hodges Major: Food and Resource Economics Several attempts have been made to analyze the benefits of creating long term value for shareholders and stakeholders through corporate sustainability ( CS ), however empirical evidence has been mixed. This thesis has considered some core issues like coverin g a relatively longer time frame and accounting for the recessionary economic conditions that dominated the selected period. The Propensity Score Matching method was used to select comparable firms for analysis. Then, standardized multi period panel data o n financial performance and First Difference and Difference in Difference methods were applied to address the objectives of the study. This study proved the existence of a positive association between financial performance and CS, and found that a period of eight years was suffic ient to reap the benefits of CS investment. Second the financial performance of firms was sensitive to the level of CS applied Third, the recession of 2008 09 had an insignificant influence on the relationship between CS a nd corporate financial performance (CFP) Fourth on average, firms continuously practicing CS had a n 11.3% higher Tobin's q and a 7.1% higher return on assets ( ROA ) compar ed to firms that never practic ed CS In addition, on average, the firms that were persistently involve d in CS ha d a higher Tobin's q by 6% and a higher ROA by 3.1% compar ed to those firms occasionally


11 invest ing in such practices The significant differences among the three groups of firms persist ed since the beginning of the study perio d in 2000, which suggests that corporate managers can quickly capture the benefits of CS. Fifth empirical analysis reveal ed that the Utilities Retail Trade, Information and Services industries we re more efficient in capturing the benefit s of CS investmen t during the study period of 8 years. Finally only the Retail Trade and Information industries were found to be sensitive to the level of CS applied by firms


12 CHAPTER 1 INTRODUCTION Preamble Firms that operate in a complex global environment constantly search for competitive advantage s to ensure they are capable of creating value in the long term ( L pez et al. 200 7). These firms are incentivized by their internal and external stakeholders to initiate and implement a variety of sustainable practices into their operations (Searcy and Elkhawas 2012). The concept of s ustainability is generally perceived as the potential for long term maintenance of well being of all stakeholders It integrates the consideration s of economic growth, social equity and environmental protection. When firm s adopt sustainable practices, it is referred to as Corporate Sustainability (CS). Corporate sustainability is a business approach that considers all social, cultural, and economic dimensions to create l ong term value that is not limited to shareholders only, but towards the natural environment as well. It is an investment strategy that ideally seeks to balance the needs of present and prospect ive stakeholders (Report of the United Nations World Commissio n on Environment and Development 1987). This presupposes that gaining competitive advantage while maintaining a balance between investors needs and resource availability in the future is a complicated o bjective. Therefore, CS measures the firm 's capabilit y to adopt economic, environmental and social dimensions into its operations, and how such adoption will be effectively reflected on the firm itself and the society (Artiach et al. 201 0). Adopting sustainable activities that cont ribute to sustainable development is professed as engaging in corporate social practice s as well (Lacy et al. 201 0). Although, CS is the most commonly used concept to address such goals there are several researchers who conceptualize the affiliation between corporations and society as Corporate Social R esponsibility (CSR) (Lourenc et al. 2012). Both CS and CSR are widely acknowledged


13 and related to the concept of sustainability (Holme and Watts 200 0). In this study, we focus ed more on CS since it is the most broadly used concept although some authors still argue that these two concepts are distinct (Cheung 2011; Lo and Sheu 2007; et al. 2007). Research ers are interested in studying the impact of adopting sustainable practices, which has led to the emergence of sustainability indexes. This study focuse d on the Dow Jones Sustainability Index (DJSI) specifically the Dow Jones Sustainability Index North Am erica (DJSI NA). The (DJSI) was established in 1999 and is the first ever family of global sustainability stock market investment benchm arks and is the largest global resource for index based concepts, data and research; it h as become a reference point in sustainability i nvesting. In addition, the study utilized the Standard and Poor's 500 (S&P 500 ) which is a stock market index based on the market capitalization of 500 large companies having common stock listed on the NYSE or NASDAQ Researchers have investigated the relationship between CS and corporate financial performance (CFP); however, findings have been inconsistent. Although, the association is still debatable among researchers, they have agreed that, over a longer time period, sustainable practices can be managed to produce new strategic opportunities and control the accompanying risks Justification of the Research Problem and the Expected Contribution Since it could be hard to detect the true relationship in a short time hor izon, researchers suggest considering a longer time frame for those firms that adopt sustainable practices which in turn could strengthen the detected relationship. Although firms could improve their profitability during the beginning years of their invol vement in sustainable activities, this benefit could be offset later by incurring greater cost s or a reallocation of resources. Therefore, the general performance of firms may not reflect any improvement, which can be misinterpreted as no


14 association betwe en CS and CFP. Researchers have reported diverse finding s Factors like the length of the study period, the length of time since firms first started investing in CS, and the general economic conditions during the selected period could significantly alter t he results. Accordingly, this research has considered these issues by covering a relatively longer time frame and observing firms for up to 4 years after starting to invest in CS. In addition, this study accounted for the recessionary economic conditions t hat dominated the selected period. Since the time frame of this study is from 1999 to 2012, there is a need to account for the financial crisis that affected the U.S. and global economy during the time frame of the study I t is commonly acknowledged that recession s influence corporate performance which in turn may confound our main objective of clearly detecting the relationship between CS and CFP. Specifically, we are interested in whether the recession could either enforce or mitigate the association be tween CS and CFP. This is the first attempt that analyzes the impact of a recession on the relationship between CS and CFP Moreover, to ensure high quality findings, we constructed the data carefully in order to detect any phenomena that may lower the pre cision of the coefficient estimates. Such phenomena are expected to be one of the causes of the inconsistency in the results of previous studies in this field. Data construction is one of the strengths of this study, which in turn gives our results a highe r level of confidence. Furthermore, this study evaluated the effect of CS among industries to determine which industries are faster in absorbing the benefits of adopting sustainable practices. Additionally the methodology followed in this study is differ ent than previous studies in two ways. First, we applied the Propensity Score Matching (PSM) m ethod to ensure that the selection of companies is balanced, and thus, the comparisons are better than previous work This method is advantageous in formulating t he distribution of observed baseline covariates in


15 order to equalize these measures between treated and untreated subject firms (Austin 2011). Second, we used the Difference in Difference (DID) method for panel data using the First Difference (FD) to analy ze the relationship. The method is effective for this research since outcomes are observed for two time periods and two groups, one of which is the trea tment group. The well known two period panel data FD and DID model is applied here, but because several periods were analyzed, we refer to it as multi period panel data FD and DID. The multi period panel data FD and DID method is also re formulated to account for the recession and annual growth of firms continuously listed in the index Empirical analysis is expected to detect a clear relationship between CS and CFP and expose differences in performance between firms that always engage in CS and those firms that never used such practices. Objective s and Hypotheses of the Study The first objective of this study is to test whether there was a significant difference in financial performance between firms that continuously practice d sustainability activities and those firms that never invest ed in such practices while accounting for the persistence of sustainability effects during the global recession of 2008 09. The second objective is to determine whether corporate performance is sensitive to the level of corporate sustainability (CS) activities utilized by firms and if such sensiti vity persisted during the 2008 09 recession The third objective has two parts. The first part aims to find out by how much these persistently listed firms gain ed compared to the firms that have never been listed. The second part is to determine by how muc h these persistently listed firms gain ed compared to firms that were dropped from the listing at certain points. The fourth objective is to analyze th e effects of CS among industries by determining which industries we re faster in absorbing the benefits of investing in CS, and secondly, by testing the sensitivity of industries to the level of CS applied by firms.


16 The Dow Jones Sustainability Index The Dow Jones Sustainability Group Index is the most commonly used index in similar studies due to its excellent reputation and high quality assessment (Lo and Sheu 200 7; et al. 200 7; Consolandi et al. 200 9; Cheung 201 1; Robinson et al. 201 1; Ziegler and Schr der 201 0). Due to the increasing interest in sustainability and its long term value creation to corporations and shareholders, organizations have invested in developing such indexes. Corporate sustainability criteria are employed to generate economic, environmental and social benefits and control risks associated with them. A study conducted by SustAinability (2004) concluded that the requirements for firms to be assessed under DJSI are more far reaching than in other indexes that perform sustainability assessment. The DJSI defines sustainability as the follow s: changing global business environment. Companies that anticip ate and manage current and future economic, environmental and social opportunities and risks by focusing on quality, innovation and productivity will emerge as leaders that are more likely to create a competitive advantage an d long term stakeholder value. (DJSI guidebook 2012) Companies that are managed to respond efficiently to challenges and strategic opportunities offered by global and local markets are assessed by DJSI and identified as sustainable companies due to their ability to generate long term va lue to shareholders. The index also performs comparisons of these sustainable companies against their competitors within their industries to identify sustainability leaders. Firms included in the DJSI North America consist of the top 20% of the 600 largest firms from Canada and the United States in the S&P Global Broad Market IndexSM that lead the field in terms of sustainability (DJSI guidebook 2012) 1 The S ustainable A sset M anag ement (SAM) group evaluates the sustainab ility activities of 1 The DJSI Guidebook is available at


17 prospective firms The methodology of SA M captures both universal and specific to industry criteria for evaluat ing economic, environmental and so cial dimensions of sustainability (DJSI guidebook 2012). All companies are eligible to participate in the assessment. For non pa rticipating companies, the organization RobecoSAM reserves the right to assess them using the same methods and tools based on their disclos ed information for public purposes to ensure a best in class assortment C ompanies are assessed and assigned a score between 0 100, then ranked against other companies within the same industry. Lastly, the top 10% of firms from each industry are listed in the Dow Jones Sustainability World Index. Accordingly, firms continually enhance their sustainability activities to ensure their inclusion in the index. An increasing number of firms have a strategic goal to be continuously listed in the index. Consequently, the there is competition for index membership which in turn encourage s companies within the same industry to par ticipate in sustainability Researchers consider studies based on the DJSI to contribute to the research literature since there is a consensus that the index i s a good proxy for CS (Garcia Castro et al. 2010 ; Waddock and Graves 1 997; McWilliams and Siegel 2000 ; Becchetti et al. 2005 ). The R ecession of 2008 2009 The National Bureau of Economic Research (the official arbiter of U.S. recessions), defines the term recession as: A general slowdown in economic activity, a downturn in the business cycle, a reducti on in the amount of goods and services produced and sold; these are all c haracteristics of a recession. (BLS spotlight on statistics 2012) 2 2 BLS spotlight on statistics 2012 is available at ht/2012/recession/pdf/recession_bls spotlight.pdf


18 They declare d that the U S suffered ten recessions between 1948 and 2011. The latest one began in December 2007 an d ended in June 2009 a lthough many indicators suggest that the U.S. economy still has not achieved a full recovery from this latest recession. The duration of this recession is within the time frame of this study, which motivated an interest to analyze wh ether the 2008 2009 financial crisis ha d an impact on the relationship between CS and CFP (BLS spotlight on statistics 2012). Moreover, this recession represents an economic condition that may cause misleading findings if not account ed for. The sec ond chapter of this thesis provides a review of literature on the definition of sustainability and the analytical methods used to estimate the relationship between CS and CFP. The third chapter explains the method s followed to gather data and identify the variables for analysis of CS and CFP The fourth chapter presents the analytical methodology adopted in this study to estimate the relationship between CS and CFP. The fifth chapter describes the data management pr ocedures used in this study. The sixth chapter describes the statistical results of the analysis The se venth chapter summarizes the key findings and conclusion s


19 CHAPTER 2 LITERATURE REVIEW Definition of Sustainability The concept of s ustainability is one of the set of terms that has materialized in the field s of natural resources, management research and applied ecology. Although this concept is generally understood as a good thing it still conjures up different thoughts by environmental scientists and m anagers. Sustainability is a favorable and appealing mec hanism in the world of "green" investments. Researchers have tr ied to bring a common definition for emerging terms over the past decade, such as sustainability. Their main objective was to give a rati onale for a definition of sustainability that will provide a common ground for future communication and manage ment action (Allen and Hoekstra 1993). The concept of s ustainability has been used widely in several disciplines and a variety of contexts. The me aning of it depends heavily on the applied context; it could have a social, economic or ecological meaning. The definitio n can be narrow or broad depending on the issue being considered (Brown et al. 1987). Therefore, it is truly hard to find a single defi nition for the concept of sustainability and there is no commonly used method of measuring it ( et al. 200 7). However, since sustainability creates value, it is essential to agree on a method to define and measure it. The co ncept of s ustainability is generally perceived as the potential for long term maintenance of well being. It integrates consideration s of economic growth, social equity and environmental protection. Corporate Financial Performance and Corporate Sustainabili ty The incentive to gain competitive advantage encourages companies to engage in sustainability activities. These activities are acknowledged to provide internal and external


20 benefits to companies (Branco and Rodrigues 200 6; Orlitzky et al. 200 3). Internally, investment in current and future economic, environmental and social opportunities provides benefits by focusing on quality, innovation and productivity and helps companies in developing new re sources and capabilities which are related to improving their profitability. Also, CS can positively impact employees' productivity and performance by affecting their motivation and morale, toward being committed and loyal to the company (Brammer et al. 200 7) thereby enabling companies to save on expenses for recruitment and training of new employees (Vitaliano 201 0). Externally, engaging in CS has a positive effect on corporate reputation (Gallego Alv arez et al. 201 0; Hussainey and Salama 201 0; Orlitzky 200 8). Improved reputation has been recognized as an important invisible endowment that supplies sustainable advantage to a firm over its competitors (Roberts and Dowling 200 2). So, these companies would be able to establish better relations with customers, investors, bankers, suppliers, and c ompetitors as well as attract high qualified employees, which in turn improve financial performance For a company to maintain access to scarce resources, it needs to nurture relationship s with key stakeholders who control access to resources (Roberts 199 2). Although the findings of previous research have been inconsistent, it is agreed among researchers that, over a longer time period, sustainable practices can be managed to produce new strategic opportunities and control risks. CS requires firms to disclose more information than tho se is typically required for U.S. corporations ( et al. 200 7) and to invest in training, product quality and safely (Wad dock and Graves 1997). So, over the short term, expenses for implementation of CS practices could exceed the incremental revenue that such practic es generate (Simpson and Kohers 2002. p. 102). et al. (200 7) have indicate d in their study the following:


21 Another factor is that assigning resources to investments tha t take into account sustainability criteria depends on the availability of surplus funds, or on the allocation of resources that were destined a priori to another purpose. This may affect profit, since the availability of funds is limited. Only in the long term can the firm plan to obtain new funds to finance practices that require larger investments So, it is suggested that only in the long run can firms acquire the benefit s of their implemented sustainab ility activities. Since there is no consensus on what long term means and because of the period of record for the DJSI, the maximum possible period to identify the firms that continuously practice CS is from 2005 to 2012 1 We are interested to determine if it pays to be sustainable in about eight years of continuously practicing sustainability 2 Correlation between CS and CFP : T hree p erspectives Fairly few research papers have been published that analyze d the link between adopting sustainable practices and the effect on the firm's performance. These stu dies rep ort different and contradictor y results. The cause of such inconsistent results is explained by the fact that they follow ed different methodologies and used different measur es of sustainability (Griffin and Mahon 1997; Simpson and Kohers 2002). Some researchers have indicated no clear or neutral relationship between CS and CFP (Curran and Moran 200 7; Garcia Castro et al. 201 0; Surroca et al. 201 0; McWilliams and Siegel 2000) A majority of studies, however, have found a positive (increasing) or weak ly positive association (Waddock and Graves 1997; Berman et al. 1999; Graves and Waddock 2000; Hillman and Keim 2001; Margolis and Walsh 200 3; Doh et al. 201 0; Lo and Sheu 200 7; Consolandi et al. 200 9; Robin son et al. 201 1; Wagner 201 0; Artich et al. 2010; Cheung 2011; Lourence et al. 2012) A third group of researchers found a negative relationship between CS and CFP ( et al. 200 7). 1 This is the treatment period for this study, as discussed in the Methods Chapter. 2 The majority of firms in the sample have been practicing sustainability prior to 2005. The data showed they all started CS activities during the period of 2001 2004


22 Researchers, who found no clear and direct relationship between CS and CFP, construed from th eir findings that the association is complex, and there are unobserved intervening influences that can not be controlled and managed. For these reasons, Ullmann (1985) advocated that the existing theoretical presentations are insufficient to imply a direct clear relationship (Artiach et al. 201 0). For studies that found a positive associatio n between CS and CFP, the research can be divided into three groups in terms of interpreting the reason for this positive relationship. First, some researchers indicated that the financial payback from adopting sustainable practices exceed s the costs of ini tial investment (McGuire et al. 1988; Barnett 2005). Ano ther group of researchers based their interpretati ons on stakeholder theory which argues that investing in CS improves the financial performance by ideally managing stakeholders (Artiach et al. 201 0). A t hird group argues that firms that invest in CS have greater resources and t hat they are more capable to adopt sustainability into operations and management Having greater resources will ultimately be translated to higher financial perf ormance (Alexander and Buchholz 1978; Waddock and Graves 1997; Clarkson et al. 2006; Artiach e t al. 201 0). Finally, researchers who found out a negative relationship between CS and CFP argued that investing in corporate sustainability is costly (Alexander and Buchholz 1978; Becchetti et al. 2005) Those firms need to re allocate resources in order to meet sustainability standards such as adopting environmentally friendly practices, social and community development, employee training improving work ing conditions, conducting promotions and making corporate donations (Arti ach et al. 201 0). Researchers recommended that variation and ambiguity of previous studies in this area is likely due to application of diverse methodologies (Cochran and Wood 1984; Aupperle et al.


23 1985; Ullmann 1985; Pava and K rausz 19 96; Barnett 2005). In T able 2 1, which summarizes previous studies that are the closest in nature and purpose to this study, it can be noticed that it is hard to find common ground among them with different scope and methods. T here is noticeable variation in the selected measure s of CS and CFP, time periods examined and hypothesis tested. et al. (200 7) and Lourenc et al. (2012) are the most closely related studies to this study as they based their studies on the Dow Jones Sustainability Index (DJSI) as well. et al. (200 7) examined the association between corporate performance and the adoption of corporate social responsibility ( CSR ) as a proxy for sustainable practices for two groups of 55 European firms during the period 1998 2004. Corporate performance wa s measured by the growth of profit before tax. The effect of CSR on p rofit before tax was estimated by regression analysis. On the other hand, L ourenc et al. (2012) studied C F P and its effects on the market value of equity for a sample of 600 Canadian and American firms from 2007 to 2010 using r egression analysis


24 Table 2 1. Summary of previous s tudies Study Waddock and Graves (1997) 3 Berman et al. (1999) 3 Graves and Waddock (2000) 3 McWilliams and Siegel (2000) 3 Year 1989 1991 1991 1996 1991 1997 1991 1996 Data 469 American c ompanies belonging to Standard and Poor's 500 in 13 industries 81 American Fortune 500 companies in different industries 11 pairs of firms 524 companies Financial Performance ROE, ROA and ROS ROA ROE, ROA and ROS ROE and ROA Sample Cross sectional Longitudinal/ panel Longitudinal Cross sectional Using DJSI No No No No Account for Recession 4 Not applicable Not applicable Not applicable Not applicable Method OLS Pooled times series model and two step GLS Trend analysis, T tests OLS The detected Relationship Positive Positive Positive Neutral Findings Sustainab le performance leads to better f inancial performance Financial performance is positively affected with consumers and employees Increasing p ositive relationship between f inancial performance and s ustainable practices The effect on f inancial performance change s as specifications of the model change 3 (Garcia Castro et al. 2010) 4 Applicable when the study time frame covered 2008 2009


25 Table 2 1. Continued Study Hillman and Keim (2001) 3 et al. (200 7) Garcia Castro et al. (2010) 3 Artiach et al. (2010) Year 1994, 1995, 1995 1998 2004 1991 2005 2002 2006 Sample 308 A merican Fortune 1000 and Standard and Poor's 500 companies belonging to different industries Two groups of European firms : 55 firms included in the DJSI, and 55 European firms belong ing to the DJGI 658 companies in KLD and Datastream 26 firms from the S&P 500 are included in the index every year for the sample period, whilst 81 f irms are occasionally included Financial Performance MVA Profit before tax Tobin's Q, MVA, ROA,ROE CSR (Dummy variable) Data Cross sectional Panel Longitudinal/ Panel Panel Using DJSI No Yes No Yes Account for Recession Not applicable Not applicable Not applicable Not applicable Method OLS Regression analysis and hypotheses testing OLS, fixed effect and random effects estimations T test and Wilcoxon signed ranks tes t, the fixed effects model The detected Relationship Positive Negative Biased by unobserved firm specific variables Positive Findings The association be tween s takeholder management and shareholder value creation (MVA) is positive Differences in performance exist between firms that belong to the DJSI and to the DJGI and these differences are related to CSR practices. A short term negative impact on performance Positive relationship between social performance and f inancial performan ce but e stimations were non significant Leading firms that are significantly larger have higher levels of growth and a higher return on equity than conventio nal firms


2 6 Table 2 1. Continued Study Cheung (2011) Lourenc et al. (2012) This study (2014) Year 2002 2008 2007 2010 1999 2012 Sample 139 firms that were added to or deleted from the DJSI during the period of 2002 2008 A sample of 600 Canadian and American firms from 2007 until 2010 A sample of 493 firms in t he U.S. during the period 1999 to 2012 Financial Performance Stock return, risk and liquidity Liquidity: measured by trading volume and proportional bid ask spread Market value of Equity (MVE) Data Time series Panel Panel Using DJSI Yes Yes Yes Account for Recession No No Yes Method Event study methodology OLS and Breusch Pagan LM test Propensity Score Matching Method and multi period panel data FD and DID model The detected Relationship N o clear relationship Positive Positive Findings No significant impact on stock return and risk. Liquidity deteriorates There is association between Market value of Equity and CSP, this relationship is affected by the size and profitability of the firm P ositive association between CS and CFP. Firms' performance is sen sitive to the level of CS applied. Recession has no influence on such associations


27 CHAPTER 3 METHODS Sample Selection To keep in line with the main objectives of this study, which is investigating the relation between CS and CFP over a longer p eriod of time, and based on DJSI NA, we need to divide the sample into three groups of American firms. The first group is those firms that we re always included in the index during the period 2005 to 2012, however, they should not have been listed in the in dex during the period of 1999 2000, since this is the baseline period of this study and we need to ensure the absence of the treatment as we will discuss in more detail in the Methodology chapter The second are firms that we re occasionally listed during t he same period. In the second group, firms are added and removed at certain points since the index was established. These firms are referred to in this study as having a low level of corporate sustainability (CS). The last group represent s firms that were never listed in the index as they have never satisfied its requirements. We rely on the S&P 500 to identify this group. The data showed that 59 American firms the index tracks that lead the field in terms of sustainability by virtue of practicing sustainab ility for 8 years continuously. Based on the index, e leven companies out of the 59 have been investing in CS since 1999 2000, so these 11 firms cannot be included in this study due to their involvement in CS during the baseline period The remaining 48 fir ms represent the first group. Another 84 firms are included in the second group which represents non continuously sustainability practitioners. Corporate financial performance data for a total sample of 493 firms in The United States of America during th e period 1999 to 2012 is covered in this study. The period of 1999 2000 is the baseline as it is required by the applied methodology. The range of 1999 2000 cannot be expanded to cover more years since this would exclude more firms from the 48 firms that


28 c onstitute the first group. The lists of corporations were obtained from the DJSI which is the exclusive owner of such data and the financial data were retrieved from the COMPUSTAT Database. The COMPUSTAT Database belongs to Wharton Research Data Service d eveloped in 1993 by the Wharton School at the University of Pennsylvania. It has become a common tool for research by over 290 institutions around the world. The SAS and Minitab16 software we re used for analyses of the data. Variables and Measures Dependent V ariable s In order to measure the corporate financial performance, different researchers have used different indic ators as shown in Table 2 1. Particularly, et al. (200 7) focused on analyzing the growth of profit before tax and growth in revenue. However, King and Lenox (2001) analyzed the financial performance by using Tobin's q which is a measure of the market valuation of a firm relative to the replacement costs of tangible assets (Lindenberg and Ross 1981). It simply means the cash flow a firm will be able to generate by investing one more dollar in asset s (King and Lenox 2001). An increase in Tobin's q reflects better expectation s about future cash flows. Tobin's q can be calculated in various ways. We will be consistent with the dividing the sum of firm equity value, book value of long te rm debt, and net current liabilities by the book value of total assets (King and Lenox 2001). In order to double check the relationship under considerati on in this study, we used both return on asset s ( ROA ) and Tobin's q as the dependent variables These t wo variables are ones of the most commonly used in the literature.


29 Independent V ariables A dummy variable for corporate sustainability investment ( D i ) is introduced to the model, it represents the group to which a firm belongs ( continuously adopting CS no n continuously involv ed in CS or never invests in CS ). To account for the effect of the 2008 09 recession, we introduce d a dummy variable ( Rec i ), where Rec i equals 1 for 2008 and 2009 and 0 otherwise. In addition, the overall market performance influence s the detection of a possible relationship between CS and CFP. Good market conditions versus bad market condition s could strengthen or weaken the effect of sustainability on firms' performance. Consequently, the variable Year i is added to the model, where Year i ranges from 1999 to 2012. Control V ariables So as to use firms with similar characteristics and to ensure the homogeneity of the three groups analyzed, we include a number of measures commonly employed in the analysis of financial performance as controls (King and Lenox 2011). Additionally, c ontrolling for these variables guarantees that the change in the firm's corporate performance is explained only by being involved in sustainable practices. These measures include firm s ize calculated by takin g the log of total assets. The size of the firms is a vital factor that could positively affect financial performance. Larger corporations generally have greater access to resources which in turn may exaggerate their profitability in comparison to small si ze firms. Second, capital i ntensity which is presented in the model as capital expenditures divided by sales. Capital intensity is defined as the amount of current real and fixed capital relative to other available production factors, such as labor. It is acknowledged among researchers that the utiliz ation of machinery and equipment raises productivity of labor which in turn improves the overall performance ( Jorgenson and Vu 2005). Third, annual growth calculated as the percentage change in sales, noticeab ly impacts profitability. Fourth, l everage ratio is calculated as the ratio of debt to assets, and is used to


30 assess a firm's ability to meet its financial obligations when they become due 1 The mix of debt and equity used by the firm can seriously affect its performance. Fifth, research and development intensity ( R&D intensity ) is added to the model and is obtained by dividing in process research and development expenses by total assets. Researchers believe that investing in rese arch and development brings new technologies that lead to an increase in net return in the long term. Finally, we considered the industry sector which is determined by 4 digit Standard Industrial Classification ( SIC ) The assoc iation between CS and CFP naturally differ s among differ ent industries. Researchers suggests that some industries are faster to absorb the benefits of CS than others, which has motivated us to go further in the analysis and classify the firms by industry sector in order to determine which industries are faster in absorbing the benefits of investing in CS. The classification is done after applying the PSM method. Based on the DJSI data, we determined that the firms constituting our sample groups belong to 15 different industry sectors, which were then grouped in to 7 industry sectors based on the North American Industry Classification System 2 These seven groups are Services, Information, Utilities, Financial, Mining, Retail Trade and Manufacturing. Note that we avoided absolute values ; the entire data was scaled in an attempt to remove other characteristics of firm s or industr ies that could affect financial performance aside from involv ement in sustainab ility activities 1 Investop e dia website available at http://w 2 North American Industry Classification System is available at bin/sssd/naics/naicsrch?chart=2012


31 CHAPTER 4 METHODOLOGY Propensity Score Matching Method (PSM) Rosenbaum and Rubin (1983a) have defined the propensity score as the probability of treatment assignment conditional on observed baseline covariates, so the covariates distribution between treated and untreated subjects are alike. The propensity score matching method can be applied in randomized and non randomized studies. In non randomized studies, the true propensity score is unkn own but can be estimated using data from the study. The propensity scor e is most commonly obtained by applying a logistic regression model, in which the treatment group is regressed on characteristics of the baseline (Austin 2011). This method forms matched sets of treatment and non treatment groups who have similar chara cteristics, represented by obtaining similar values of the propensity score ( Rosenbaum and Rubin 1983a and 1985 ). One of the advantage s of using such an approach is that it ensures that the treatment group will not be confounded with either measured or unmeasured baseline characteristics. The second advantage is that once matched groups have been identi fied, the impact of the treatment can be directly analyzed by comparing the outcomes of treated and untreated subjects ( Greenland, Pearl and Robin s 1999 ). For the above mention ed reasons, PSM is employed in this study to ensure that the grouping procedure of companies is balanced, and thus, the comparisons are valid. In order to get the best matched groups of firms we need to run the following p ro bit (or l ogit) model for the p re treatment period 1999 2000: Pr {D i ( firm s ize+ capital i ntensity+ ann ual growth+ l everage ratio+ R&D intensity ) ( 4 1 )


32 w and D i represents the treatment group, so it equals 1 for firms that continuously practice CS. O ther variables are taken as covariates in normal linear terms. We r a n the model in equation (4 1) twice, first to get the best matched firms from the group of never invested in CS, and second to get a similar ma tched group from the firms occasionally investing in sustainability practices, in order to compare both of these groups to the treatment group. The results of the application of the PSM model assigns a probability score for each firm ranging from 0 to 1. T he firms that were close in terms of covariates to the continuously listed firms get a probability close to 1 and these were chosen for comparison purposes. Analysis of the R elation between CS and CFP We use the Difference in Difference (DID) method to analyze the relationship between CS and CFP The DID method is a technique used in econometrics that measures the effect of a treatment at a given period of time. It is often used to measure the change induced by a particular treatment or event ( Abadie 2005). Generally, the DID equation is commonly expressed as follows: Y i 0 1 D i 2 T i 3 D i T i + i ( 4 2 ) w here Di is the treatment, T i is the treatment period the interaction term represent s the treatment status 3 is the DID estimator, which reflects the difference between the treatment group and the control and i is the error term. However, equation (4 2) is not sufficient to analy ze our data, because we need to apply it along with the First Difference (FD) method. The FD method is an approach used to solve the problem of omitting relevant variables in panel data. It also removes time invariant omitted variables ( V i ) using repeated observations over tim e, which can be expressed as the following.


33 Y it 0 1 X it + V i + it ( 4 3 ) Y it 1 0 1 X it 1 + V i + it 1 t= 2000 ( 4 4 ) w here, X it represents independent variables. Differencing both equations, gives: Y it = Y it Y it 1 1 X it i t ( 4 5 ) This latter equation cancel s out the invariant unobserved effect V i (Wooldridge and Jeffrey 2001) This method is advantageous to guarantee that the change in performance of firms is only a result of being involved in CS. By combining the DID and FD methods, we get the tw o period panel data FD and DID, but we re formulated the lat er since we have multi ple periods and we need to add several interaction terms in a way that satisfies our technical objectives. The multi period panel data FD and DID model is discussed in detail in the following sections of this chapter. The DID method has the same assumptio ns as Ordinary Least Squares (OLS) regression : Simple random s ample ~ satisfied Quantitative response, Y i ~ satisfied since both Tobin's q and ROA are quantitative values error terms are independent and normal ly distributed with mean 0 and constant standard deviation for every combination of explanatory variable s ~ satisfied. This assumption is tested and F igure 4 1 shows the approximate normality of the residuals since it takes the shape of a standard bell curve. where changes with every combination of explanatory variables, but


34 Figure 4 1. Histogram of r esiduals A pplication of the M ulti P eriod P anel D ata FD and DID M odel: Model 1 The m ulti period panel data FD and DID model can be expressed as the following. Y it = 0 1 D it 2 T it i X it ) + V i it t= ..,2012 ( 4 6 ) Y it 1 = 0 1 D it 1 2 T it 1 i X it 1 ) + V i it 1 ..,2012 ( 4 7 ) Differencing both equations, gives: Y it = Y it Y it 1 = 2 + 1 D it + i X it ) + t ( 4 8 ) The latter equation cancel s out the invariant unobserved effect V i The term T it is a dummy for the time period that equals 1 for the treatment period 2005 12 and 0 otherwise. It does not vary across firms, and by applying the FD this term always gets the value of 0, except for year 2005 where it equals 1 (2005 2004 = 1 0 =1); which can be understood as (treatment period control period = 1 0 = 1). So, we will deal with it as with the case of a two period model, where the intercept is replaced by a period effect 2 ( T it = 1 for all units), as shown in equation (4 8) above. The term D it varies across firms and over time, as either 0 or 1. If we designate A as the control group and B as the treatment group, then we can summarize it as D i A = 0 and D i B = 1. We notice in equation (4 8), the interaction term between the treatment and the treatment period ( D it T it ) is dropped since T it equals 1, so it is the same as


35 D it However, in this study, there is a need to include the interaction term since we have a gap of 4 years (2001 2004) between the control period (1999 2000) and the treatment period (2005 2012). Therefore, we should re emphasize the treatment period after ta king the first difference. Then, we can determine the DID estimator from equation (4 9) as shown in Table 4 1 The term i X it ) represents all control variables after taking the first difference. The equation in (4 9) is our reference model in for a ll the analysis Y it 2 1 D it 3 D it *T it i X it t ( 4 9 ) Table 4 1. Multi p eriod p anel d ata FD and DID e stimator Post Treatment Period Pre Treatment Period Difference D 1 (treatment) 2 + 1 + 3 + i 1 + i 2 + 3 D 3 (control) 2 + i i 2 Difference 3 } The objective of Model 1 is to detect the association between CS and CFP by comparing firms that invest continuously in CS and those which have never invested in CS, with the la t t er taken a s the baseline. In addition, we will test the impact of the recession 2008 09 on this relationship. The dummy variable Rec i and the interaction term of Rec i and D i are added to the reference equation in (4 9), as shown in equation (4 10) below. The i nteraction term is needed because these variables may be interacting and the effect of Rec i on the d ependent variables will rely on whether the firm has CS or not. Additionally, we are interested in finding out by how much these always listed firms gain annual ly com pared to th o se firms which have never been listed. Therefore, we add the variable Year i and the interaction term Year i D i to the reference model to address this question, as shown in equation (4 10) below Finally, the interaction terms between each industry sector and D i are attached to the reference equation as well to achieve our fourth objective of determining which industries are faster in absorbing the benefits of CS.


36 Then, we apply standardized regr ession to equation in ( 4 10 ) in order to put all our coefficient estimators on an equal basis and therefor e can compare them directly. In other words, we can use the b eta coefficients as a measure of relative strength of the regressor variables Standardization is attained by taking the difference of each variable from its mean and divid ing by the standard deviation. Therefore, we can compare directly the performance of our three groups of firms ( 7 ) and the performance across different industries ( 13 to 18 ) Y it = 2 1 D it + 3 D it *T it + 4 Rec it 5 Rec it it 6 Year it + 7 Year it it + 8 firm s ize it + 9 capital i ntensity it + 10 a nnual growth it + 11 l everage ratio it + 12 R&D intensity it + 13 Services i D it + 14 Information i D it + 15 Utilities i D it + 16 Financial i D it + 17 Mining i D it + 18 Retails i D it + t ( 4 10 ) To answer our first concern in this study, we test the following hypothesis: H1: There is a difference in financial performance between firms that continuously practice sustainability activities and those firms that have never invested in such practices and 3 5 0. The interpretation of 7 answers the first part of our third objective, by how much these always listed firms gain annually compar ed to th o se firms which have never been listed The significance o 13 18 indicate which industries are more efficient in absorbing the benefits from CS which addresses the first part of our fourth objective Application of the Multi Period Panel Data FD and DID Model : Model 2 The objective of Model 2 is to test if there are differences in performance among corporations that i nvest in different levels of CS, which reflect s how corporate performance is sensitive to the level of CS utilized by firms. In this model, we focus on firms that continuously invest in CS and firms that non continuously invest in CS, with the latt er taken a s th e baseline. Keeping all conditions the same as equation (4 10). To answer our second objective the following hypothesis will be tested :


37 H1: Corporate performance is sensitive to the level of CS invested by firms and such sensitivity persists during the re 3 5 0. The interpretation of 7 answers the second part of our third objective, i.e. by how much these always listed firms gain compar ed to th o se which are occasionally listed Similarly, t he 1 3 1 8 answer the second part of our fourth objective, regarding which industries are more sensitive to the level of CS applied by firms.


38 CHAPTER 5 DATA CONSTRUCTION Analysis of Residuals and Exposing O utliers It is required to check the existence of outlier s in the data, which may mislead final results. An outlying observation is an observation that is much different in relation to the observations in the sample. In addition, outl iers are one of the sou rces of h etero ske dasticit y, a problem that is discussed lat er in this chapter. Probability p lot of residuals is one of the statistical tools used to expose outliers. Figure 5 1 shows that the data has six outliers, which are located beyond 2 and 2 standar d deviation s in the figure. Since the causes of these outliers in the data are unknown, it is not a wise decision to discard them immediately. We need to investigate their impacts on the study results, and then carefully deal with this issue Figure 5 1 Probability plot of r esiduals from m odel1 1 First, we r a n the model s with no action regarding the outliers as shown in T able 6 2 then, re r a n it after excluding the six ou tl ying observations. The results confirm that the exclusion of outliers has no effe ct on the significance of estimators but it improves the goodness of fit of 1 Prob ability plot of r esiduals from Model1 when ROA is the response exposed the same outliers since both Tob in's q and ROA are measures of return on asset. Model2 showed no outliers.


39 model 1 by increasing the adjusted R square to 23.29% in the case of Tob in's q and to 68.03% in the case of ROA So, it wa s concluded that exclusion of outliers was not necessary since they did not le ad to biased results Non Linear R elationship s and Multicolinearity A scatterplot of each predictor variable against the dependent variables demonstrate that there is no special kind of relationship. From the Pearson correlation m atr ix in T able 5 1, it shows some indications of collinearity. Not surprisingly, the correlation between Tobin's q and ROA is statistically significant since both are means of measuring the return on asset s Furthermore, Tobin's q and ROA are correlated to fi rm size, leverage ratio and R&D intensity Although the data suggest some level of collinearity, there were no pairwise correlations that exceed ed 61% except for capital i ntensity and annual growth where it reaches 95%. The correlation between capital i nte nsity and annual grow th is consistent with accounting literature. As we mentioned before, it is recognized that the utilization of machine ry and equipment raises the productivity of labor which in turn stimulates the growth of the firm (Jorgenson and Vu 20 05). To ensure reliability of the study results, we test ed whether the existence of collinearity may cause bias. The results of the models that include either capital i ntensity or annual growth are completely identical to the model s that include both as shown in Table 6 2 This indicates that the threat of multicolinearity is limited and we should not omit any variables


40 Table 5 1. Pearson correlation c oefficients and probability values for model variables Variable Tobin s q ROA F irm s ize C apital i ntensity A nnual G rowth L everage r atio Tobin s q 1 ROA 0.475 (0.000) 1 F irm s ize 0.608 (0.000) 0.455 (0.000) 1 C apital i ntensity 0.013 (0.285) 0.019 (0.121) 0.013 (0.281) 1 A nnual g rowth 0.012 (0.311) 0.016 (0.192) 0.012 (0.008) 0.991 (0.000) 1 L everage r atio 0.158 (0.000) 0.178 (0.000) 0.033 (0.008) 0.019 (0.117) 0.020 (0.103) 1 R & D i ntensity 0.049 (0.000) 0.045 (0.000) 0.031 (0.012) 0.001 (0.924) 0.001 (0.945) 0.011 (0.364) Probability ( p ) values are given in parentheses Detecting the P roblem of Autocorrelation and Heteroskedasticity The assumptions of OLS and DID methods require the absence of a utocorrelation and heteroskedasticity in th e study data. Autocorrelation means the error from one observation depends on the error from other observation s which may create pattern s among error terms. On the other hand, heteroskedasticity refers to the case where the variance of the disturbance term is not constant, which violate s the equal variance assumption of methods used in this study. W e r a n the Durbin Watson d test t o detect autocorrelation The results for the case of Tobin's q shows that neither positive nor negative autocorrelation are significant in the d ata (p value = 0.50 and 0.49, respectively). Similarly, the model s with ROA as dependent variable show the absence of positive and negative auto correlation with p value= 0.49 and 0.48, respectively as shown in Table 6 2


41 The presence of heteroskedasticity may lower the precision of the coefficient estimates. However, despite non constant variance, estimators are still linear unbiased and asymptotically normally distributed. By using White's General heteroskedasticity test as shown in Table 6 2, the null hypothesis that there is a constant variance is rejected when Tobin's q is the dependent variable (p value = 0.0003 for Model 1 and 0.0004 for Model 2 ). Figure 5 2 shows the absence of heteroskedasticity (constant variance). We can se e that at smaller fitted values of dependent variables the residuals have more positive values and as the fitted values get large r, the spread of the residuals slopes down and become negative values. The variability of the residuals can be understood as th ose firms with lower level s of return on asset s have predicted values lower than the actual ones, and the predicted values of firms with higher level s of return on asset s are much higher than the actual ones. This creates a down sloping pattern of residual s. Figure 5 2. Fitted v alue of squared r esiduals for (Tobin's q) To improve our analyses, we applied weighted r egression which is one of the methods used to correct for non constant va riance. Since the data indicate that the dependent variable ( Tobin's q) 2 changes with the variance of the residuals, we nee d to weight it in the r egression. 2 H eteroskedasticity test for ROA models wa s not significant at 10% level of significance as shown in table 6 2.


42 Therefore, we calculate d the reciprocal of Tobin's q (1/ Tobin's q ) for the entire set of observations. Weighted r egression works by weighting each observation based on the variability of its fitted value. In our case, we want ed to give observations with a different level of Tobin's q different weights in order to shrink their squared residuals. With the proper weight, this procedure minimizes the sum of weighted squared residuals to produce residuals with a constant variance ( heteroskedasticity ). In Table 6 2, we see that the non constant variance in our data was resolved with p value = 0.054 for Model 1 and Model 2 Figure 5 3 demonstrates that the heteroskedasticity pr oblem wa s clearl y minimized after applying the weighted r egression. Fig ure 5 3. Fitted v alue of squared r esiduals for (1/ Tobin's q ) It is important to emphasize the impact of violating the assumption of constant variance on the findings of this study. I gnoring the presence of heteroskedasticity led to a major misinterpretation of a negative relationship between CS and CFP and also turning one variable to be statistically significant. It is clear then that such phenomenon can not only lower precision of the coefficient estimates, but completely shift the result s For this reason, we would


43 recommend that non constant variance is expected to be one of the causes that have led to the inconsistency in the results of previous studies in this field


44 CHAPTER 6 EMPIRICAL ANALYSIS AND RESULTS Statistical Description Table 6 1 presents the descriptive statistics for three sub samples. The first section displays the statistics of DJSI continuously listed firms, the middle section shows the statistics of DJSI occasionally included firms, and the last section shows the fi rms that were never listed in DJSI When comparing these three groups, we see that the mean and the median values for all variables are slightly greater for DJSI continuously listed firms compar ed to occasionally included firms and both continuously and oc casionally listed have higher mean and median values compar ed to the never listed firms These findings are consistent with Lourenc et al. (2012) who studied C F P and its effects on the market value of equity and Artiach et al. (2010) who analyzed the deter minants of C F P. Both studies concluded that continuously listed corporation s are significantly larger and have a higher return on equity ( ROE ) than non continuously listed firms In addition, all three sub sample s show that the distributions of capital i nt ensity and annual growth are highly skewed toward the right In the case of continuously listed firms the s kewness values were 79.23 and 80.93 for capital i ntensity and annual growth respectively. The same two variables ha d a higher and sharper distribution peak, which is pr esented as kurtosis values of 6367.98 and 6557.77, respectively, for continuously listed firms. At the same time, both capital i ntensity and annual growth have higher standard deviations for all sample gr oups, compar ed to other control variables


45 Table 6 1. Descriptive s tatistics of the sampled firms Variable Mean Median Std Dev Min Max Kurtosis Skewness Continuously listed f irms Tobin s q 0.79 0 0.967 0.231 0.608 1 .000 2.738 1.883 ROA 0.297 0.259 0.234 0.669 1.63 0 1.602 1.129 Firms ize 4.015 3.981 0.663 1.458 6.38 0 0.837 0.27 0 C apital i ntensity 1.027 0.039 64.216 0.0001 5169.18 6382.81 79.233 Annual G rowth 0.742 0.074 45.678 1.000 3701.47 6557.77 80.931 Leverage R atio 0.204 0.179 0.167 0 1.5108 5.158 1.447 R & D I ntensity 0.001 0.001 0.012 0.58 6 0.006 1102.5 29.151 Non continuously listed f irms Tobin s q 0.788 0.883 0.232 0.699 1.000 2.775 1.823 ROA 0.292 0.252 0.229 0.669 1.534 1.665 1.184 Firms ize 4. 017 3.502 0.669 1.333 6.383 0.681 0.275 Capital i ntensity 1.020 0.029 64.582 0 5144.18 6367.98 79.21 3 Annual G rowth 0.738 0.074 47.027 1.001 3701.31 6557.17 81.003 Leverage R atio 0.164 0.166 0.177 0 1.173 5.132 1.623 R & D I ntensity 0.0009 0.0001 0.012 0.666 0.005 1114.50 29.374 Never listed f irms Tobin s q 0.780 0.867 0.292 0.409 1 .000 2.723 1.782 ROA 0.291 0.24 0 0.237 0.658 1.501 1.274 1.027 Firms ize 4.016 3.518 0.703 1.444 6 .000 0.379 0.298 Capital i ntensity 1.019 0.03 0 64.002 0 5160.05 6321.29 79.2 0 Annual G rowth 0.731 0.071 45.002 1 .000 3600.21 6722.81 80.902 Leverage R atio 0.128 0.129 0.204 0 1.444 5.364 1.426 R & D I ntensity 0.0002 0 0.013 0.825 0.001 1101.17 29.151


46 Results The application of the Propensity Score Matching m ethod (PSM) results in two sets of two groups each that are similar to the firms that are persistently listed in the index in terms of firm s iz e, capital intensity, ann ual growth, leverage ratio and R&D int ensity The first is a subset of 48 firms from the group of firms that never practice d CS and the second is a subset of 48 firms from the group of occasionally listed firms. We will rely on these two subsets in the empirical analysis. Consequently, we expe ct most of the control variables to be statistically insignificant. We can answer our first objective based on multiple regression analysis of model 1 with ( 1/Tobin's q ) as the dependent variable and after the heteroskedasticity correcting procedure with results as shown in the second and third columns of T able 6 2 The result of our first hypothesis, H 0 3 5 =0 is significantly rejected (p value= 0.012) where we can conclude that there is a significant difference in financial performance between fi rms that persistently practice CS for 8 years or more and those that have never do ne it Importantly, we notice that the individual t tests of these coefficients indicate 3 is statistically significant (p value= 0.009), but 5 is statistically insign ificant (p value= 0.819). Therefore, the total difference between the two groups 3 only and the effect of recession on the actual difference is negl igible We can conclude that the presence of recession neither enforce d nor moderated the t otal difference of the effect of CS, and such difference persist ed in the same magnitude during the recession of 2008 2009. The negative 1 3 indicates that sustainable companies outperform the non sustainable companies and this difference persists in the same magnitude during recession times. It is important to understand from this study that managers can reap the avails of their investment 1 Signs in the model reflect the relationship between the indicators and ( 1/Tobin's q ), so negative sign means a positive association between the indicator and Tobin's q


47 on CS within a period of 8 years 2 Moreover, the coefficient of the term D it reflect s the difference in performance between the two groups throughout the study period; it is not conditional to the treatment period for individual years ( T it ). All findings are consistent when analyzing the variation on ROA instead of 1/Tobin's q as illustra ted in the fourth and fifth columns of T able 6 2. The second objective tests if there are differences in performance among corporations that invest in different levels of CS, which may reflect how corporate performance is sensitive to the level of CS inves ted by focusing on firms continuo usly invest in CS and those that sometimes invest with the latter treated as the baseline. The result of the second hypothesis 3 5 =0 supports our claim that the fin ancial performance of firms is sensitive to the level of CS invested (p value = 0.027, see sixth and seventh column s of T able 2 6 ) The individual t tests of these coefficients indicate that 3 is statistically significant (p value= 0.031 ), however 5 is statistically insignificant (p value= 0.721 ). Similarly we can conclude here that existence of the 2008 09 recession The negative 3 sign of 3 indicates that continuously listed companies also outperform ed the occasionally listed companies and this difference persist ed in the same magnitude during recession times. In other words, the financial performance of firms is sensitive to the level of CS applied. This also supports our previous finding that managers could be motivated by such findings to pr ioritize the investment in CS. The regression analysis of model 2 with ROA as the dependent variable gave similar results as shown in the eighth and ninth columns of T able 6 2 2 Eight years: from 2005 to 2012. 3 Signs in the model reflect the relationship betwee n the indicators and ( 1/Tobin's q ), so negative sign means a positive association be tween the indicator and Tobin's q


48 The first part of our third objective aims to find out by how much the persistently listed firms gain ed compar ed to the firms that have never been listed Based on the standardized coefficient estimator of the D it *Year it interaction variable as shown in the second and fourth columns of T able 6 2, on average, continuously p racticing CS firms realized a higher Tobin's q by 11.3% and a higher ROA by 7.1% compar ed to firms that never practic ed CS, holding all other variables constant. The second part of the third objective examines by how much these persistently listed firms gain ed annually compar ed to th o se which were dropped off at from time to time. We can see in the sixth and eighth column s of Table 6 2 on average, the firms that were persistently involve d in CS ha d a higher To bin's q by 6% and a higher ROA by 3.1% compar ed to those firms occasionally invest ing in such practices, holding all other variables constant. The significant differences among the three groups of firms persist ed across all time s studied since 2000, which offers a promise for managers to capture the benefits of CS by efficiently utilizing available resources. In addition, the coefficient of the Year it variable reveals that the overall market performance wa s slightly increasing during the study period. Specifically, the general performance of the corporations constituting our sample groups gr ew annually by 3% to 6.4% on average since 2000. The fourth objective of this study is to analyz e the effect of CS among industries. First, our goal wa s to figure ou t which industries we re faster in absorbing the benefits of investing in CS. The results of the empirical analysis in T able 6 2 (column 2 and 3) reveals that the Utilities Retail Trade, Information and Services industries more greatly reflect ed the benef it s of CS investment during the study period of 8 years. Firms in the Utilities industry can gain more than 12.4% on average as a result o f persistently being involv ed in sustainable practices, compar ed to Manufacturing industry firms that invest at the same intenseness in such practices, holding all


49 other variable constant. Similarly, continuously practicing CS firms in Retail Trade, Information and Services industries have a higher gain by 9.2%, 7.8% and 3%, respectively, comparing to continuously practicing CS firms in Manufacturing The analysis with ROA as the dependent variable supports all the above mentioned findings (see T able 6 2, column s 4 and 5). Secondly, we can conclude that only Retail Trade and Information industries are sensitive to the level of CS applied by firms as shown in Table 6 2 column s 6 and 7. In these two industries, it matters how intensive is the investment in CS by firms. In the case of comparing the Retail Trade industry to Manufacturing firms that have CS over a period of 8 years can gain more than 8.8% on average compar ed to occasionally listed firms, holding all other variable s constant. The results are consistent when analyzing ROA as shown in T able 6 2 column s 8 and 9. In addition, the variables firm s ize, capital i ntensity and annual g rowth ha d no influence in all models. These three indicators met our expectations since firms groups are of similar level s for all control variables as a result of applying the PSM method. Whilst R& D i ntensity is si gnificant at the 10% probability level for the case of compari ng continuously listed and never listed firms it is not significant in the case of comparing continuously listed with non continuously firms This means that continuously listed firms invest sig nificantly more in Research and Development practices compar ed to firms that never invest in CS and such difference s d id n't exist between continuously and non continuously listed firms, as we can see in T able 6 2. Notably, the l everage r atio explain s some of the variation in the models, which means that there is a wide variance in the data such that the application of the PSM method did not completely correct for this predictor. Th is finding is consistent with previous literature; firms that are highly lev eraged have a higher Tobin's q and ROA holding all other variables constant.


50 Notably, the adjusted R Squared of the ROA models we re much higher than the adjusted R Squared of Tobin's q models, which indicate that our independent variables are more efficient in explaining the variation in ROA during the study period. Additionally, the intercepts of the models represent two components, the first is the period effect 2 as equation 4 10 symbolizes and the second is the average time invariant unobserved effect (average fixed effect) as the software used in this study always report s All intercepts are statistically significant which implies the existence of period effect s and average fixed effect in this study. It definitely reflects that the application of standardized m ulti period panel data FD and DID methods is a preferred approach to analy ze such data and handle the objectives of this study since the period effect is suggested by the DID method application and the fixed effect indicated by applying the FD method.


51 Table 6 2. Standardized p arameter results of model 1 and m odel 2 Model 1 Model 2 Dep. Var. (1/Tobin'sq ) (ROA) (1/Tobin'sq ) (ROA) Predictor Coeff. P value Coeff. P value Coeff. P value Coeff. P value Intercept + 2 16 .000*** 7 13 .047** + 1 94 .000*** 3 61 .091* it .0 019** +.0 18 .042* .0 78 .044** + .025 021** D i t *T it .0 87 009 ** +.0 99 .002 *** .0 62 .031 ** + .031 004*** Rec it +.0001 .8 07 .00 1 396 +.0 17 9 43 .00 2 .9 01 it *Rec it +.051 .819 .00 4 85 5 +.001 .7 21 .00 3 .980 Year it .064 .000*** +.030 .017** .061 .000** +.037 .034** it *Year i t 113 .000*** +.07 1 .021 ** .06 .000** +.031 .0 27 ** Firms ize i t 1 50 .371 + 1 18 .557 1 43 .844 + 8 63 .284 Capital i ntensity it .001 .758 +.0001 .881 .001 .710 + 7 12 .787 Annual g rowth i t .001 .765 + .0002 .415 .002 .719 + .569 441 Leverage r atio i t .301 .000*** +.241 .0 29* .277 .000*** +.9 38 .044** R & D i ntensity i t .402 .091* +.518 089* .285 .562 + .616 .888 Services i t it .003 .0 77 + 007 .0 27 .000 1 432 + .00 06 499 Information it it .0 78 .0 41 ** + .031 .042** .011 .0 57 + .002 .07 1* Utilities it it 124 .003 *** +.112 .0 39 ** 095 .151 + .0 27 .601 Financial i t it .002 119 + .033 .4 42 .001 788 + .0 27 699 Mining i t it .0002 121 + .0001 14 2 .013 367 + .002 109 RetailTrade it it 092 .000** + 087 .01 2** .0 88 .032 + .072 .0 1 1 3 5 =0 .0 12 ** .0 11** .0 27 ** .0 0 1* * Adj R Sq 2 3 07 % 67 .2 1 % 22 88 % 83 16 % # of Obs. 24,960 24,951 24,960 24,051 D W Test .50* .49* .49 .48 WhiteTest1 .0 0 0 *** .058* .0 0 0 *** .058* WhiteTest2 .054* .058* .054* .058* ***, ** and s ignificance at the 0.01, 0.05 and 0.10 levels, respectively. Dep. Var. stands for Dependent Variable; Coeff. stands for Coefficient; # of Obs. is Number of Observation; D W Test is Durbin Watson Statistic o f Autocorrelation; White's Test1 is the first heteroskedasticity test ; White's Test2 is the heteroskedasticity test after correcting for heteroskedasticity Dependent variables: is calculated by dividing the sum of firm equity value, book value of long term debt, and net current liabilities by the book value of total assets; ROA is Return on Asset, calculated as Net Income/Total Asset s Independent variables: D it is an indicator that equals 1 if the firm have been continuously listed in the DJSI during the sample period 2005 2012 and 0 if a firm has never been listed (baseli ne of model 1) or a firm is non continuously listed (baseline of model 2) ; T it is the treat ment period (2005 12); Rec it is the recession dummy variable it equals 1 for 2008 2009; Year it ranges from 2000 to 2012; Firm s ize is the log of total assets ; Capital i ntensity equals capital expenditures / sales ; annual growth is the percentage change in sales; Leverage ratio is the ratio of debt to assets ; R&D intensity equals in process research and development expenses / total assets ; Industry dummy variables: Services, Information Utilities Financial Mining Retail Trade Manufacturing ( baseline).


52 CHAPTER 7 FINDINGS SUMMARY AND CONCLUSION Although, there have been several attempts to analyze the benefits of creating long term value for shareholders and stakeholders through CS activities empirical evidence has been mixed. In fact, results could be influenced by factors like the study period length, time since firms started investing in CS, and the overall e conomic performance during the period of study For these reasons, this research h as considered these issues by covering a relatively longer time frame and accounting for defined economic conditions that dominated the selected period. The first objective of this study wa s to test whether there is a significant difference in financial pe rformance between firms that continuously practice sustainability activities and those firms that never invest in such practices while accounting for the persistence of sustainability effects during the global recession of 2008 09. The second objective wa s to examine whether corporate performance is sensitive to the level of corporate sustainability (CS) activities by firms and if such sensitivity persisted during the 2008 09 recession The third objective to find out by how much these persistently listed firms annually gain more comparing to the firms that have never been listed and measure and compared to firms which were dropped off at certain points. The fourth objective was to analyze th e effect of CS among industries including which industries we re fa ster in absorbing the benefits of investing in CS and testing the sensitivity of industries to the level of CS applied. Importantly, we have constructed the dataset to detect any phenomena that may lower the precision of the coefficient estimates. Such phe nomena are normally expected to be one of the causes le a d ing to inconsistency in the results of previous studies in this field. Based on our sample, the existence of heteroskedasticity changed the sign of the DID estimator and turned the significance of on e con trol variable. If this problem uncorrected, clearly it w ould have shift ed the


53 findings. Data construction is one of the strength s of this study, which gives our results a higher level of confidence. This study contributes to the literature by providing further support to the group of researchers who have reported a significant positive relationship between CS and CFP. We have proved the existence of a positive association by analyzing the variation in ROA and the reciprocal of Tobin's q stat istic. Second, a period of eight years was found to be sufficient to reap the benefits of CS and completely cover the initial cost of investment in CS This finding supports the explanations offered by McGuire et al. (1988) and Barnett (2005) regarding the positive relationship between CS and CFP. They believe d that the association between CS and CFP becomes positive when the financial payback from adopting sustainable practices exceed s the costs of initial investment. Third the financial performance of firms was sensitive to the level of CS applied; Tobin's Q and ROA were significantly greater for those firms always in DJSI since 2005. Fourth, the recession of 2008 09 didn't shift the total differen ce between CS and CFP It is acknowledged that in recession times larger and smaller companies may suffer more relative to medium and large companies, and the majority of our sample is companies of medium size. Fifth, on average, continuously practicing CS firms realize a higher Tobin's q by 11.3% and a higher ROA by 7.1% compar ed to those that have never been practicing CS In addition, on average, the firms that are persistently involve d in CS have a higher Tobin's q by 6% and a higher ROA by 3.1% compar e d to those firms occasionally invest ing in such practices The significant differences among the three groups of firms persist ed across the time since 2000, which offers promise for managers to capture the benefits of CS based on their utilization of attainable resources. Sixth, empirical analysis reveal ed that the Utilities Retail Trade, Information and Services industries realized greater benefit s of CS investment during the study


54 period of 8 years. Firms belong ing to the Utilities industry can gain about 12.4% greater returns on average as a result o f persistent involv e ment in sustainable practices compar ed to firms in Manufacturing that invest at the same intens ity in such practices. Moreover, we conclude that the Retail Trade and Information industries we re most sensitive to the level of CS applied by firms. In the case of comparing Retail Trade to Manufacturing firms with CS over a period of 8 years can gain up to 8.8% greater on average compar ed to occasionally listed firms. To summarize, this study found out that it really pays to be sustainable and it pays more for those persistently investing in CS. Such practice is proven to improve firms' financial performance. The increase in financial per formance is in the favor of the firms themselves as well as investors, stakeholders and society. For example, one face of CS is establishing better work conditions for employees, which can positively impact their productivity. Companies can save on their e xpenses by cutting down the costs of recruitment and training of new employees as well. All of these factors result in high quality products and better prices, which works toward gaining committed and loyal consumers, and this is what stakeholder theory ca lls for. Some researchers like Artiach et al. (201 0) based their interpretations of a positive relationship between CS and CFP on this theory which argues that investing in CS improves the financial performance by ideally manag ing stakeholders. It is worth mention ing that, although it is reported by The National Bureau of Economic Research that the duration of the recession was from 2008 to 2009, financial indicators still suggest that the U.S. economy is not fully recovered, wh ich may be a limitation of this study. Future research efforts are expected to conduct more studies o n this issue and considering even longer time frame s where then U.S. economy is expected to be fully recovered. In addition, research could broaden the sc ope and compare the effects of CS investment between developing


55 and developed countries. Questions like when does it pay to be sustainable and under what conditions do corporate sustainability (CS) efficiently work better could be interesting for other res earchers.


56 LIST OF REFERENCES Abadie, A. (2005). Semiparametric difference in differences estimators. Review of Economic Studies 72 (1): 1 19. Doi : 10.1111/0034 6527.00321 Allen, T. F. H., & Hoekstra, T. W. (1993). Toward a definition of sustainability. Sustainable ecological systems: implementing an ecological approach to land management. Rocky Mountain Forest and Range Experiment Station, Fort Collins, Colorado 98 107. Artiach, T., Lee, D., Nelson, D., & Walker, J. (2010). The deter minants of corporate sustainability performance. Accounting and Finance 50, 31 51. Austin P. (2011). An introduction to propensity score methods for reducing the effects of confounding in observational studies. Multivariate Behav Res 15(3):399 424. doi: 10.1080/00273171.2011.568786. Berman, S. L., Wicks A. C., Kotha S. & Jones T. M. (1999). Does Stakeholder Orientation Matter? The Relationship between Stakeholder Management Models and the Firm Financial Performance. Academy of Management Journal 42(5 ), 488 506. Brammer, S., Millington, A., & Rayton, B. (2007). The contribution of corporate social responsibility to organizational commitment. The International Journal of Human Resource Management 18(10), 1701 1719. Branco, M. C., & Rodrigues, L. L. ( 2006). Social responsibility disclosure: A study of proxies for the public visibility of Portuguese banks. The British Accounting Review 40(2), 161 181. Brown, B., Hanson, M., Liverman, D., & Merideth, R. (1987). Global sustainability: Toward definition Environmental Management 11, Number 6, Page 713. Cheung, A. (2011). Do stock investors value corporate sustainability? Evidence from an event study. Journal of Business Ethics 99(2), pp 145 165. Compustat. (Novermber 2013). Financial Data, 1999 2012. [Data]. Retrieved from Wharton Reasearch Data Service. Consolandi, C., Jaiswal Dale, A., Poggiani, E., & Vercelli, A. (2009). Global standards and ethical stock indexes: The case of the Dow Jones sustainability stock ind ex. Journal of Business Ethics 87, 185 197. Curran, M. M., & Moran, D. (2007). Impact of the FTSE4Good Index on firm price: An event study. Journal of Environmental Management 82, 529 537. Doh, J. P., & Guay, T. R. (2006). Corporate social responsibili ty, public policy, and NGO activism in Europe and the United States: An institutional stakeholder perspective. Journal of Management Studies 43(1), 47 73.


57 Gallego lvarez, I., P rado Lorenzo, J. M., Rodrguez Domnguez, L., & Garca S nchez, I. M. (2010). Are social and environmental practices a marketing tool? Empirical evidence for the biggest European companies. Management Decision 48(10), 1440 1455. Garcia Castro, R., Ar i o, M. A., & Canela, M. A. (2010). Does social performance really lead to financi al performance? Accounting for endogeneity. Journal of Business Ethics 92, 107 126. Graves, S. B. & to Last Companies. Business and Society Review 105, 393 418. Greenland S., Pea rl J., Robins J.M. (1999). Causal diagrams for epidemiologic research. Epidemiology. 10:37 48. Griffin, J. J. & J. F. Mahon. (1997). The Corporate Social Performance and Corporate Financial Performance Debate: Twenty Five Years of Incomparable Research. B usiness and Society 36(1), 5 31. Hillman, A. J. & G. D. Keim. (2001). Shareholder Value, Stakeholder Management, and Social Strategic Management Journal 22(2), 125 139. Holme, R., & Watts, P. (2000). Corporate Social Responsibility: making good business sense, World Business Council for Sustainable Development Hussainey, K., & Salama, A. (2010). The importance of corporate environmental reputation to investors. Journal of Applied A ccounting Research 11(3), 229 241. Jorgenso n, Dale W.; Vu, Khuong (2005). Information T echnology and the World Economy Scandinavian Jou rnal of Economics 107 (4): 631 650. doi : 10.1111/j.1467 9442.2005.00430. King A. A. & Lenox M. J. (2001). Does It Really Pay to Be Green? An Empirical Study of Firm Environmental and Financial Performance. Journal of Industrial Ecology 5,1. Lacy, P., Cooper, T., Hayward, R., & Neuberger, L. (2010). A new era of sustainability UN Global Compact Accenture CEO Study 2010, Accenture Lindenberg, B., & Stephen A. (1981). Tobin's q Ratio and Industrial Organization. Journal of Business 54:1 32. M. V., Garcia, A., & Rodriguez L. (2007). Sustainable development and corporate performance: A study based on the Dow Jones sustainability index. Journal of Business Ethics 75, 285 300. Lo, S., & Sheu, H. (2007). Is corporate sustainability a value increasing strategy for business? Corporate Governance 15(2), 345 358.


58 Lourenco I. C., Br anco M. C., Curto J. D. & Eug nio T. (2012). How Does the Market Value Corporate Sustainability Performance? J Bus Ethics 108:417 428. Margolis, J. D., & Walsh, J. P. (2003). Misery loves compani es: Rethinking social initiatives by business. Administrative Science Quarterly 48, 268 305. McWilliam s, A. & D. Siegel. (2000). Corporate Social Responsibility and Financial Performance: Correlation or Misspecification? Strategic Management Journal 21, 603 609. North American Industry Classification System. (March 2014). 2012 NAICS, 2012. Retrieved from United States Census, U.S. Department of Commerce. Orlitzky, M. (2008). Corporate social performance and financial performance: A research synthesis In A Orlitzky, M., Schmidt, F. L., & Rynes, S. L. (2003). Corporate social and financial performance: A meta analysis. Organization Studies 24(3), 403 441. Roberts, R. W. (1992). Determinants of corporate social responsibility disclosure: An applicati on of stakeholder theory. Accounting, Organizations and Society 17(6), 595 612. Roberts, P. W., & Dowling, G. R. (2002). Corporate reputation and sustained superior financial performance Strategic Management Journal 23, 1077 1093. Robinson, M., Kleffn er, A., & Bertels, S. (2011). Signaling sustainability leadership: Empirical evidence of the value of DJSI membership. Journal of Business Ethics 101, 493 505. Rosenbaum P.R., Rubin D.B. (1983a). The central role of the propensity score in observational studies for causal effects. Biometrika 70:41 55. Rosenbaum P.R., Rubin D.B. (1985). Constructing a control group using multivariate matched sampling methods that incorporate the propensity score. The American Statistician 39:33 38. Searcy C. & Elkhawas D. (2012). Corporate sustainability ratings: an investigation into how corporations use the Dow Jones Sustainability Index. Journal of Cleaner Production 35 79e92. Simpson, W. G. & T. Kohers (2002). The Link between Corporate Social and Financial Perfor mance: Evidence from the Banking Industry. Journal of Business Ethics 35(2), 97 109. Surroca, J., Trib J. A., & Waddock, S. (2010). Corporate responsibility and financial performance: The role of intangible resources. Strategic Management Journal 31(5 ),


59 463 490. Vitaliano, D. F. (2010). Corporate social responsibility and laborturnover. Corporate Governance 10(5), 563 573. Waddock, S. A., & Graves, S. B. (1997). The corporate social performance financial performance link. Strategic Management Journal 18(4), 303 319. Waddock, S. A. & Graves, S. B. (1997). Quality of Management and Quality of Stakeholder Relations: Are They Synonymous? Business & Society 36(3), 250 279. Wagner, M. (2010). The role of corporate sustainability performance for ec onomic performance: A firm level analysis of moderation effects. Ecological Economics 69, 1553 1560. Wooldridge, Jeffrey M. (2001). Econometric Analysis of Cross Section and Panel Dat a MIT Press. 279 291. 978 0 262 23219 7 Ziegler, A., & Schr der, M. (2010). What determines the inclusion in a sustainability s tock index? A panel data analysis for European firms. Ecological Economics 69, 848 856.


60 BIOGRAPHICAL SKETCH Ibtisam Al Abri is a student in the Master of Science program in the Food and Resource Economics Department at the University of Florida (UF). Sh e received h er MS from the University o f Florida in the spring of 2014. Ibtisam is from the Sultanate of Oman. She attended Sultan Qaboos University (SQU) in Oman and gained in economics in 2010. She is working as a fac ulty member in the Department of Natural Resource Economics at SQU. After graduation, she plans to continue her education and seek a PhD degree in applied economics or n atural resource economics


HowDoestheMarketValueCorporateSustainability Performance?IsabelCostaLourenc o€ManuelCasteloBranco€Jose DiasCurto€TeresaEuge nioReceived:2May2011/Accepted:26October2011/Publishedonline:9November2011 SpringerScience+BusinessMediaB.V.2011Abstract Thisstudyprovidesempiricalevidenceonhow corporatesustainabilityperformance(CSP),asproxiedby membershipoftheDowJonessustainabilityindex,is reectedinthemarketvalueofequity.Usingatheoretical frameworkcombininginstitutionalperspectives,stakeholdertheory,andresource-basedperspectives,wedevelop asetofhypothesesthatrelatethemarketvalueofequityto CSP.ForasampleofNorthAmericanrms,ourpreliminaryresultsshowthatCSPhassignicantexplanatory powerforstockpricesoverthetraditionalsummary accountingmeasuressuchasearningsandbookvalueof equity.However,furtheranalysessuggestthatweshould notfocusoncorporatesustainabilityitself.Ourndings suggestthatwhatinvestorsreallydoistopenalizelarge protablermswithlowlevelofCSP.FirmswithincentivestodevelopahighlevelofCSPnotengagingonsuch strategyare,thus,penalizedbythemarket. Keywords Corporatesustainability Valuerelevance Canada USA Introduction Theconceptofsustainabledevelopmentintegratesthe considerationofeconomicgrowth,environmentalprotection,andsocialequity,simultaneouslyandonamacrolevel(FiggeandHahn 2004 ).Whenincorporatedbythe rm,itiscalledcorporatesustainability(CS)(ibid.). Althoughotherconceptshavebeenproposedovertheyears toconceptualizebusinessandsocietyrelations,suchas corporatesocialresponsibility(CSR),CShasbecomethe conceptusedmostwidelytoaddresstheserelationships. Eventhoughsomeauthorsproposedistinctionsbetween CSRandCS(Cheung 2011 ;LoandSheu 2007 ;Lo pez etal. 2007 ),widelyacknowledgeddenitionsandanalyses ofCSRrelateitwithsustainabledevelopment(Holmeand Watts 2000 ;EuropeanCommission 2002 ).Thus,inthis article,theseconceptsareconsideredtoaddressthesame basicissues,inthesensethattheyallareaboutcompanies' impactson,relationshipswith,andresponsibilitiesto, society. Engaginginactivitiestocontributetosustainable developmenthasemergedasanimportantdimensionof corporatevoluntarypractice(Lacyetal. 2010 ).Corporate sustainabilityperformance(CSP)measurestheextentto whicharmembraceseconomic,environmental,social, andgovernancefactorsintoitsoperations,andultimately theimpacttheyexertonthermandsociety(Artiachetal. 2010 ).Engagementinactivitiespromotingsustainable developmentisincreasinglyanalyzedasasourceofcompetitiveadvantagefortherm(PorterandKramer 2006 ). Animportantstreamofresearchanalyseswhetherrms whichareperceivedassustainableout-performorunderperformrmswhicharenotperceivedinthesameway. Somemixedresultscanbefound.Surveysofthenumerous studiesabouttherelationshipbetweenCSandcorporate I.C.Lourenc o( & ) J.D.Curto UNIDE,LisbonUniversityInstitute(ISCTE-IUL), AvenidaForc asArmadas,1649-026Lisbon,Portugal M.C.Branco FacultyofEconomics,UniversityofPorto,Porto,Portugal M.C.Branco OBEGEF(ObservatoryinEconomicsandManagement ofFraud),RuaDr.RobertoFrias,4200-464Porto,Portugal T.Euge nio SchoolofTechnologyandManagement, PolytechnicInstituteofLeiria,Leiria,Portugal123JBusEthics(2012)108:417428 DOI10.1007/s10551-011-1102-8


nancialperformance(CFP)thathavebeenundertaken abound.Findingsofthemajorityofthemindicatenoclear tendency(Ullman 1985 ;Aupperleetal. 1985 ;Pavaand Krausz 1996 ;WoodandJones 1995 )orapositivebutweak correlationbetweenthetwo(MargolisandWalsh 2003 ; Orlitzkyetal. 2003 ;Romanetal. 1999 ).Recentresearch stillprovidesmixedresults:thereisevidencebothofa negativerelation(Lo pezetal. 2007 ),norelation(Curran andMoran 2007 ;Garcia-Castroetal. 2010 ;Surrocaetal. 2010 ),andapositiverelation(Dohetal. 2010 ;Loand Sheu 2007 ;Consolandietal. 2009 ;Cheung 2011 ;Robinsonetal. 2011 ;Wagner 2010 )betweenthetwo. Inspiteofmixedresultsofindividualstudies,aconsistentconclusionemergeswhenwetaketheminaggregate:marketforcesgenerallydonotpenalizeandare morelikelytorewardcompanieswithhighlevelsofCSP (Dohetal. 2010 ).Departingfromthisconclusion,the purposeofthisstudyistoextendresearchbyanalyzing whetherthemarketpenalizescompanieswithlowerlevels ofCSPandwhetherrm'scharacteristicslikesizeand protabilityinteractwithsuchpenalizations. Thisstudycontributestotheextantliteratureonthis issuebyinvestigatinghowthemarketviewsCSP,as proxiedbymembershipoftheDowJonessustainability index.Usingamulti-theoreticalframeworkwhichcombinesinstitutionalperspectives,stakeholdertheory,and resource-basedperspectives(RBP),asetofhypothesesare developedthatrelatethemarketvalueofequitywithCSP, consideringtheinteractionofsizeandprotabilitywith CSP.Inthisstudy,companiesareconsideredtoengagein CSactivitiestoconformtostakeholdernormsandexpectationsbecausetheyexpectthathavinggoodrelationswith themmayleadtoincreasednancialreturnsbyassistingin developingandmaintainingvaluableintangibleassets. Theempiricalanalysisreliesonthelargest600rms fromCanadaandtheUnitedStatesofAmericaintheDow Jonesglobaltotalstockmarketindex(DJGTSM),which includestwosetsofrms,thosethatbelongtotheDow JonessustainabilityUnitedStatesindex(DJSI)North America(higherlevelofCSP)andthosethatbelongtothe DJGTSMbutarenotincludedintheDJSINorthAmerica (lowerlevelofCSP). Giventhatnationalinstitutionalcontextsarerelevant whenassessingthestockmarketvaluerelevanceof nancialandnon-nancialperformancemeasures(Cormier andMagnan 2007 ),ourstudyfocusesoncompaniesfrom CanadaandtheUSAmainlyinordertoobtainalarge samplethatishomogeneousastotheinstitutionalsetting. OurpreliminaryresultsindicatethatCSPhassignicant explanatorypowerforstockpricesoverthetraditional summaryaccountingmeasuressuchasearningsandbook valueofequity.However,furtheranalysessuggestthatwe shouldnotfocusontheCSitself.Ourndingsshowthat whatinvestorsreallydoistopenalizelargeprotablerms withlowlevelofCSP,whichfacegreaterpublicscrutiny andpressuresfromstakeholders. Thisstudycontributestotheliteratureinseveralways. First,webringadditionalevidenceonthevaluerelevance ofnon-nancialinformation.Somepreviousstudieshave alreadyfoundasignicantrelationbetweenthemarket valueofequityandnon-nancialinformation,likenetwork advantages(Rajgopaletal. 2003 ),environmentalperformance(Hasseletal. 2005 ),eco-efciency(Sinkinetal. 2008 ),ortechnologicalconditions(MatolcsyandWyatt 2008 ).WeextendtheseconclusionstotheissueofCSP. Second,weprovideadditionalevidenceontherelationship betweenCSPandrms'nancialperformance.Thisarticle reportsevidenceofapositiverelationbetweenCSPandrmperformance.Finally,wecontributewithnewempiricalevidencesupportingthatrmswithincentivesto developahighlevelofCSPnotengagingonsuchstrategy arepenalizedbythemarket.Artiachetal ( 2010 )have alreadydemonstratedthatsizeandprotabilityareincentivestoinvestinsustainability.Wealsondthatsizeand protabilityareissuesthatmatterintermsofCSP.In addition,ourresultssuggestthattheinformationonthe relationbetweensize,protability,andlevelofCSPis relevantforinvestors. Theremainderofthearticleisorganizedasfollows. Section 2 developsthetheoreticalframeworkofthisstudy. Section 3 describestheresearchdesignandSect. 4 presents theempiricalresults.Finally,Sect. 5 discussesthendings andoffersconclusionsandimplicationsforfutureresearch. TheoryandHypothesesDevelopment Thetheoreticalframeworkadoptedinthisstudycombines institutionalperspectives,stakeholdertheory,andRBP. Someauthorsalreadyprovidedimportantstudiesinwhich similarcombinationswereattempted(see,forexample, Bansal 2005 ;HillmanandKeim 2001 ;Rufetal. 2001 ; Surrocaetal. 2010 ). Institutionalperspectiveshavebeenusedasalens throughwhichtoexploreCS(DohandGuay 2006 ;Doh etal. 2010 ;Campbell 2007 ).Institutionaltheorypredicts thatrmsadoptspecicbehaviorstoobtainaccessto resourcesandsupportbycriticalstakeholders(Dohetal. 2010 ).Theanalyticalfocusofinstitutionalperspectivesis onsociallegitimacy,whichreferstotheacceptanceofthe rmbyitssocialenvironment,byitsexternalconstituents. Failuretoconformtocritical,institutionalizednormsof acceptabilitycanthreatenitslegitimacy,resources,and, ultimatelysurvival.Thisperspectivesuggeststhatrms willrespondstrategicallytoinstitutionalnormsandto changesintheirsocialenvironmenttogainormaintain 418 I.C.Lourenc oetal.123


legitimacybecausetheyrecognizethatconformingwill resultinimprovedaccesstoresources(Suchman 1995 ; Bansal 2005 ). Stakeholdertheorycanbethoughasfocusingtheinstitutionalperspective.Becausethesocialenvironmentwithin whichrmsoperateisconstitutedbystakeholders,legitimacydependsonmeetingtheirexpectations.Armbuilds legitimacybyconformingtostakeholderexpectations (BansalandBogner 2002 ).Postetal.( 2002 ,p.8)dene thestakeholdersofacompanyastheindividualsand constituenciesthatcontribute,eithervoluntarilyorinvoluntarily,toitswealth-creatingcapacityandactivities,and whoarethereforeitspotentialbeneciariesand/orrisk bearers.''CSPcanbeassessedintermsofacompany meetingthedemandsofitsmultiplestakeholdergroups (Rufetal. 2001 ). Stakeholdertheorycanbealsocomplementedbythe RBPsincermsmayviewmeetingstakeholderdemandsas astrategicinvestment,requiringcommitmentsbeyondthe minimumnecessarytosatisfystakeholders(Rufetal. 2001 ).EngaginginCSactivitieswhentheseareexpected tobenetthecompanyisabehaviorthatcanbeexamined throughthelensoftheRBP(BrancoandRodrigues 2006 ; Gallego-A lvarezetal. 2010 ;HussaineyandSalama 2010 ; McWilliamsetal. 2006 ;Siegel 2009 ;Surrocaetal. 2010 ). TheRBPsuggestthatcompaniesgeneratesustainable competitiveadvantagesbyeffectivelycontrollingand manipulatingtheirresourcesthatarevaluable,rare,cannot beperfectlyimitated,andforwhichnoperfectsubstituteis available(see,forexample,Barney 1999 ;Bowmanand Ambrosini 2003 ;Kraaijenbrinketal. 2010 ;Pertusa-Ortega etal. 2010 ). CompaniesengageinCSbecauseitisacknowledged thatsomekindofcompetitiveadvantageaccruestothem. CSisseenasprovidinginternalorexternalbenets,orboth (BrancoandRodrigues 2006 ;Orlitzkyetal. 2003 ). Investmentsinsociallyandenvironmentallyresponsible activitieshaveinternalbenetsbyhelpingacompanyin developingnewresourcesandcapabilitieswhicharerelatedtoknow-howandcorporateculture.Theseinvestments haveimportantconsequencesonthecreationordepletion offundamentalintangibleresources,namelythoseassociatedwithemployees.CScanbedemonstratedtohave positiveeffectsonemployees'motivationandmorale,as wellasontheircommitmentandloyaltytothecompany (Brammeretal. 2007 ).Aswellasproductivitybenets, companiesalsosaveoncostsforrecruitmentandtraining ofnewemployees(Vitaliano 2010 ). TheexternalbenetsofCSarerelatedtoitseffecton corporatereputation(BrancoandRodrigues 2006 ;GallegoA lvarezetal. 2010 ;HussaineyandSalama 2010 ;Orlitzky etal. 2003 ;Orlitzky 2008 ).Corporatereputationhasbeen identiedasoneofthemostimportantintangibleresources thatprovidearmsustainablecompetitiveadvantage (RobertsandDowling 2002).CompanieswithagoodCS reputationareabletoimproverelationswithexternalactors suchascustomers,investors,bankers,suppliers,and competitors.Theyalsoattractbetteremployeesorincrease currentemployees'motivationandmoraleaswellastheir commitmentandloyaltytothecompany,whichinturn mayimprovenancialoutcomes.Stakeholdersultimately controlarm'saccesstoscarceresourcesandrmsmust managetheirrelationshipwithkeystakeholderstoinsure thatsuchaccesstoresourcesismaintained(Roberts 1992 ). Insum,CScanraisebenetsinthelongrunnamely throughimprovedrelationswithstakeholdersandreduced costofconictswiththem,reputationcreation,and employeeproductivity.Alltheseaspectsmakermsmore attractivetoinvestors.HigherlevelsofCSParesubjectto lowereconomicuncertainty,morepredictableearnings, andlowerriskforinvestors.Inaddition, t hedegreeof institutionalizationreachedbyCSpractices,suchasISO 14001certicationorsustainabilityreporting,hasmade thesepracticesnecessaryrequirementsforenteringthe markets.Companiesthatdonotconformtothesepractices arelikelytobepenalized.Thus,weexpectthat: H1 ThemarketpenalizesrmswithalowerlevelofCSP, whencomparedwithrmswithahigherlevelofCSP. Ascompaniesgrowlargertheirvisibilityincreasesand theybecomemoresusceptibletothescrutinyoftheir stakeholdersandhencemorevulnerabletothepotential adversereactionsofthesegroups.Largecompanies,on average,aremorediversiedacrossgeographicaland productmarketswhichmeansthattheyhavelargerand morediversestakeholdergroups(BrammerandPavelin 2004 ). Largerrmsaremorevisiblepoliticallyandsodraw greaterattentionfromthegeneralpublic,government,and otherstakeholders.Theyaremorelikelytocreatecorrespondinglargersocialproblemsbecauseofthesheerscale andprominenceoftheiractivities.Thus,apassiveoreven negativeresponsetostakeholder'sdemandsisunlikelyto beasuccessfulstrategyforbigrmswhichfacegreater publicscrutinyandexternalpressures(Artiachetal. 2010 ). Sizemayalsobeconsideredasanindicatorforthecapacity ofarmtoengageinenvironmentalandsocialactivities, whichleadtoxedcoststhatarelessimportantforlarger companies(ZieglerandSchro der 2010 ). Godfreyetal.( 2009 ,p.430)suggestthatrmswitha largermarketpresenceincurmoreriskthantheirsmaller counterparts.Theyarguethatalargermarketpresence translatesintomoretransactions,whichleadtoahigher probabilityofnegativeevents(therearesimplymore opportunitiesfornegativeoutcomes'')(ibid.).Theconsequenceisthatlargerrmsshouldbemorewillingto HowDoestheMarketValueCSP? 419123


engageinsociallyandenvironmentallyresponsibleactivitiestocoverthisincreasedriskthansmallerrms. TheliteratureonthedeterminantsofCSPprovide empiricalevidenceonapositiverelationshipbetween rm'ssizeandCSP(e.g.,Artiachetal. 2010 ;Zieglerand Schro der 2010 ;Chihetal. 2010 ).Thus,weexpectthat: H2 Themarketpenalizationofrmswithalowerlevelof CSPishigherforlargerrms,whencomparedwithsmaller rms. WaddockandGraves( 1997 )studiedthelinkbetween rms'socialandnancialperformance,hypothesizingthat socialperformanceisbothapredictorandconsequenceof nancialperformance.Theyconcludedthatcorporate socialperformancedependsonnancialperformanceand thatthesignoftherelationshipispositive.Thesendings wereinterpretedasmeaningthatrmswithslackresources potentiallyavailablefromstrongnancialperformance mayhavegreaterfreedomtoinvestinsociallyandenvironmentallyresponsibleactivities,andthatthoseinvestmentsmayresultinimprovedsocialperformance. Artiachetal.( 2010 )alsodemonstratethatprotable rmsaremorelikelytohaveahigherlevelofCSP.The managersofnon-protablermsareaskedtoreducecosts andmaximizeeconomicreturnstonancialstakeholders, insteadofmeetingsocialstakeholder'sdemandsthrough expenditureonsustainableactivities.Inperiodsoflow economicperformance,thecompanies'economicobjectiveswillbegivenmoreattentionthansocialconcerns (Ullman 1985 ). Ontheotherhand,companieswhichpresentabnormally highlevelsofprotsarejustasexposedtopressuresfrom stakeholdersasthoseofabnormallylargecompaniesor thosethatoperateinsociallysensitiveindustries(Branco andRodrigues 2008 ).Publicvisibilitymayberelatedto highprots,withthemoresuccessfulcompaniescoming undermoreintensestakeholderscrutiny(ibid.).Itfollows that: H3 Themarketpenalizationofrmswithalowerlevelof CSPisalsohigherforprotablermswhencomparedwith non-protablerms. ResearchDesign SampleandData Theempiricalanalysisreliesonthelargest600rmsfrom CanadaandtheUnitedStatesofAmericaintheDowJones GlobalTotalStockMarketIndex(DJGTSM)attheendof 2010.Westartedbylookingforallthermswithdata availableeveryyearforthefour-yearperiod20072010. Weexcludermswithnegativebookvalueatleastinone ofthe4years. Second,weclassiedthermsintotwogroups,those includedintheDJSIinallthe4yearsofthesampleand thosermsneverincludedintheDJSIduringtheentire periodofthesample,therebyrepresentinganongoinglack ofinvestmentinCSP.1Thisclassicationgivesrisetothe mostimportantindependentvariableforourstudy,aproxy forthelevelofCSP. FirmsincludedintheDJSINorthAmericaconsistofthe top20%ofthe600largestrmsfromCanadaandthe UnitedStatesintheDJGTSMthatleadtheeldintermsof sustainability(DJSIguidebook,2010).2Firm'ssustainabilityisevaluatedbythesustainableassetmanagement (SAM)group.TheSAM'smethodologyisbasedonthe applicationofcriteriatoassesstheopportunitiesandrisks derivingfromeconomic,environmentalandsocial dimensionsforeachoftheeligiblerms(DJSIguidebook, 2010).TheintegrityoftheDJSIasaproxyforCSPis highlightedbysomeauthors,whorecommendtheSAM GroupresearchasthebestpracticeinCSresearch(Artiach etal. 2010 ).AnincreasingnumberofstudiesontherelationbetweenCSPandrmperformanceconsidersDJSIas aproxyforCSP(LoandSheu 2007 ;Lo pezetal. 2007 ; Consolandietal. 2009 ;Cheung 2011 ;Robinsonetal. 2011 ;ZieglerandSchro der 2010 ). Theaccountingandmarketdatawerecollectedfromthe ThompsonWorldscopeDatabase.Toinsurethatregression resultsarenotinuencedbyoutlyingobservations,thetop andbottom1%ofeachmainvariable'sdistributionhave beenexcludedfromthesample.Thisapproachisin accordancewithsomeothervaluerelevancestudies.3The nalsampleisanunbalancedpanelcomposedby241 rms-yearobservationsforthe63rmsincludedinthe DJSIduringthe4yearsofthesampleand1,356rms-year observationsforthe355rmsneverincludedintheDJSI duringtheentireperiodofthesample. Table 1 presentsthesampledistributionacrossindustries.Whenalltheobservationsareconsideredtogether, theindustrialsectoristhemostrepresentativewith36%of thesample.Thesmallestrepresentations,witharound10%, arethemining,thecommercialandtheservicesindustries. Asexpected,bothDJSIandNon_DJSIrmsarefoundin eachindustryandthelatterdominatesinallcases.The proportionofDJSIrms-yearobservationsfromeach 1FirmswhicharepersistentlyincludedintheDJSIhaveamore substantialnancialandstrategicinvestmentsinCSPthanrmsthat areonlyoccasionallyincluded.FollowingArtiachetal ( 2010 ),the latestrmswerethusexcludedfromthesample.2TheDJSIGuidebookisavailableat .3SeeCurtoetal.( 2011 )wheretheimpactofinuentialobservations onregressionresultsisdiscussed. 420 I.C.Lourenc oetal.123


industryisbetween15and20%,exceptforthenancial industrywherethispercentageissomewhatlower. ResearchMethod TotestthehypothesesformulatedinSect. 3 ,weestimate severalregressionsbasedonthesamemodel,whichrelies ontheaccountingbasedvaluationmodeldevelopedin Ohlson( 1995 ),whoshowshowthermvaluerelatesto accountingdataandotherinformation.Thisapproachis currentlyusedinempiricalstudiesonthevaluerelevance ofnon-nancialinformation(e.g.,Rajgopaletal 2003 ; Hasseletal 2005 ;MatolcsyandWyatt 2008 ;Johnston etal 2008 ;Sinkinetal 2008 ;SchadewitzandNiskala 2010 ) .Ourprimarymodelshowsthatthemarketvalueof equityisalinearfunctionoftwosummarymeasuresof informationreectedinnancialstatements,namelythe bookvalueofequityandearnings,givenbytheEq. 1 MVit a0 a1BVit a2NIit eit 1 where MV isthemarketvalueofequity,4BV representsthe bookvalueofequity,and NI isthenetoperatingincome. Allthevariablesareonapersharebasis. TheAssociationofMarketValueofEquitywithCSP Inordertoaccesswhetherthemarketpenalizesrmswitha lowerlevelofCSP,whencomparedtormswithahigher levelofCSP,weuseanewregressionequation,Eq. 2 whichcomprisesthevariable Non_DJSI ,whichassumesthe value1ifthermisnotincludedintheDJSINorthAmerica and0otherwise.Ifthemarketpenalizesrmswithalower levelofCSP,wewouldexpecttheestimatedcoefcienton Non_DJSI a3,tobenegativeandstatisticallysignicant. MVit a0 a1BVit a2NIit a3Non DJSIit eit 2 TheAssociationofMarketValueofEquitywithCSP: EffectofSize Inordertoaccesswhetherthemarketpenalizationofrms withalowerlevelofCSPishigherforlargerrmswhen comparedwithsmallerrms,weuseanewregression equation,Eq. 3 ,whichcomprisestwobinaryvariables splittingthe Non_DJSI intwogroupsbasedontherm's size( Non_DJSI_Big and Non_DJSI_Small ).Thevariable Non_DJSI_Big assumesthevalue1ifthermhasalower levelofCSPandits SIZE isabovethemedian5and0 otherwise.Thevariable Non_DJSI_Small assumesthe value1ifthermhasalowerlevelofCSPandits SIZE is belowthemedianand0otherwise. Ifthemarketpenalizationoflargerrmsishigher,when comparedwithsmallerrms,wewouldexpectthe estimatedcoefcientson Non_DJSI_Big a3,andon Non_DJSI_Small a4,tobenegativeandstatisticallysignicantandtheabsolutevalueoftheformertobestatisticallyhigherthanthelatter.Ifontheotherhandthemarket doesnotdistinguishgroupsofrmswithalowerlevelof CSPbasedonsize,thenwewouldexpectthat a3= a4.An alternativesituationisalsopossiblewherebythemarket penalizesonlythosermswithincentivestopresenta higherlevelofCSPbutthatdonotengageonsuchstrategy,i.e.,thelargerrmsnotincludedintheDJSINorth America.Inthiscase,wewouldexpecttheestimated coefcienton Non_DJSI_Big a3,tobenegativeandstatisticallysignicantandtheestimatedcoefcienton Non_DJSI_Small a4,tobestatisticallyinsignicant. MVit a0 a1BVit a2NIit a3Non DJSI Bigit a4Non DJSI Smallit eit 3 Table1 Samplecompositionbyindustry IndustrySICcodeDJSIrms-yearobs.Non_DJSIrms-yearobs.Allrms-yearobs. n % n % n % MiningSIC123512491369 IndustrialSIC2and3117494523356936 UtilitiesSIC434141841421814 CommercialSIC53213128916010 FinancialSIC62193232434422 ServicesSIC7and825101451117011 24110013561001597100 DJSI rmsarethoseincludedintheDJSIeveryyearforthesampleperiod20072010; Non_DJSI rmsarethosewhohaveneverbeenincluded intheDJSIduringthesampleperiod20072010 4Weusethemarketvalueofequityasofscalyear-end.However, untabulatedndingsrevealthatourinferencesarenotsensitiveto usingpricesasofscalyear-endorasof3monthsafterscalyearend. 5Themedianiscomputedbasedonthemeanvalueofeachrmin thefourconsideredyears. HowDoestheMarketValueCSP? 421123


TheAssociationofMarketValueofEquitywithCSP: EffectofSizeandProtability Inordertoaccesswhetherthemarketpenalizationofrms withalowerlevelofCSPisalsohigherforprotablerms whencomparedwithnon-protableones,weuseanew regressionequation,Eq. 4 ,whichcomprisestwobinary variablessplittingthe Non_DJSI_Big ( Non_DJSI_Small )in twogroupsbasedontherm'sprotability,namelythe Non_DJSI_Big_Prot andthe Non_DJSI_Big_Loss ( Non_DJSI_Small_Prot and Non_DJSI_Small_Loss ).The variable Non_DJSI_Big_Prot ( Non_DJSI_Small_Prot ) assumesthevalue1ifthermhasalowerlevelofCSP,its SIZE isabove(below)themediananditsROEispositive and0otherwise.Thevariable Non_DJSI_Big_Loss ( Non_DJSI_Small_Loss )assumesthevalue1iftherm hasalowerlevelofCSP,its SIZE isabove(below)the medianandits ROE isnegativeand0otherwise. Ifthemarketpenalizationofprotablermsishigher, whencomparedwithnon-protableones,wewouldexpect theestimatedcoefcientson Non_DJSI_Big_Prot ( Non_ DJSI_Small_Prot ), a3( a5),andon Non_DJSI_Big_Loss ( Non_DJSI_Small_Loss ), a4( a6),tobenegativeandstatisticallysignicantandtheabsolutevalueoftheformertobe statisticallyhigherthanthelatter.Ifontheotherhandthe marketdoesnotdistinguishgroupsofBig(Small)rmswith alowerlevelofCSPbasedonprotability,thenwewould expectthat a3= a4( a5= a6).Analternativesituationisalso possiblewherebythemarketpenalizesonlythosermswith economicincentivestopresentahigherlevelofCSPbut thatdonotengageonsuchstrategy,i.e.,thelargerand protablermsnotincludedintheDJSINorthAmerica. Inthiscase,wewouldexpecttheestimatedcoefcienton Non_DJSI_Big_Prot a3,tobenegativeandstatistically signicantandtheotherestimatedcoefcients, a4, a5, a4,and a6,tobestatisticallyinsignicant. MVit a0 a1BVit a2NIit a3Non DJSI Big Profitit a4Non DJSI Big Lossit a5Non DJSI Small Profitit a6Non DJSI Small Lossit eit 4 Finally,followingpreviousliteratureonthevalue relevanceofnancialandnon-nancialinformation, additionalvariablesareusedinthisstudytocontrolfor protability,leverage,size,cash-ows,marketrisk, internationallisting,andindustry.Thus,Eqs. 2 4 are estimatedincludingthefollowingcontrolvariables: ROE LEV SIZE CF RISK LIST ,andIndustry. ROE isthe returnonequity, LEV isend-of-yeartotaldebtdividedby end-of-yearmarketcapitalization, SIZE isthelogarithmof totalassetsasoftheendoftheyear, CF isnetcash-ow fromoperatingactivitiesscaledbyend-of-yeartotalassets, RISK isBetaasreportedbyWorldScope,and LIST isa dummyvariablethatassumesthevalue1ifthermis listedinaforeignstockexchangeand0otherwise.There aresixdummiesforindustry:the Mining dummywhich assumesthevalue1inthecaseofSIC1and0otherwise, the Industrial dummywhichassumesthevalue1incases ofSIC2or3and0otherwise,the Utilities dummywhich assumesthevalue1inthecaseofSIC4and0otherwise, the Commercial dummywhichassumesthevalue1incase ofSIC5and0otherwise,the Financial dummywhich assumesthevalue1incasesofSIC6and0otherwiseand, nally,the Services dummywhichassumes1incasesof SIC7or8and0otherwise. Asoursampledataisanunbalancedpanelwith418 rmsand4yearsofobservations,empiricalresearchis basedonstatisticaltechniquestoestimatepaneldata regressionmodels.Asseveralvariablesdonotvarywithin thermsinthefourconsideredyears,speciallythedummy variables,theyshouldbedroppedfromthemodelifxed effectsregressionwasconducted.However,asthesevariablesareveryimportantfortestingtheresearchhypotheses formulatedbefore,xedeffectsregressionhasbeendiscarded.Duetothis,andinordertocheckwhichone, pooled(wherenopaneleffectsexist)orrandomeffects regression,isstatisticallymoreappropriatetodescribethe relationshipbetweenthedependentandtheexplanatory variablesincludedintheregressionmodels,theBreusch Pagantestwascomputed. Results DescriptiveStatisticsandCorrelations Table 2 presentsthedescriptivestatisticsfortheentire sampleaswellasforthesub-samplesof241DJSIrmsyearobservationsand1,356Non_DJSIrms-year observations.Whencomparingthesetwogroupsof observations,wendthatforallthevariables,exceptfor LEV and RISK ,themeanandthemedianvaluesarehigher fortheDJSIrms.Untabulatedresultsfortheequalityof meansparametric t testshowthatthemeanvaluesare statisticallydifferentforthevariables MV NI ROE SIZE CF ,and RISK .ThesendingsareconsistentwiththoseofArtiachetal ( 2010 )intheirstudyonthedeterminantsof CSP.TheyfoundthatleadingCSPrmsaresignicantly largerandhaveahigherreturnonequitythannon-leading CSPrms. Table 3 showsPearsoncorrelationsforthecontinuous variablesincludedintheregressions.Consistentwith establishedresultsintheaccountingliterature,themarket valueofequityispositivelyandsignicantlyassociated with BV and NI .Notsurprisingly,thecorrelationbetween 422 I.C.Lourenc oetal.123


marketvalueand ROE LEV SIZE CF ,and RISK isalso statisticallysignicant.Thesignsofthecorrelationcoefcientsarelargelyconsistentwithndingsinprior research. RegressionResults BasedonBreuschPagantestresults(seeTable 4 ),the pooledregressionhypothesiswasalwaysrejectedinfavor oftherandomeffectsregression.Thus,werunrandom effectsregressions6toestablishtherelationshipbetween thedependentandtheexplanatoryvariables. TheAssociationofMarketValueofEquitywithCSP Table 5 presentssummarystatisticsresultingfromtheestimationoftheEq. 2 ,includingtheestimatedcoefcientsfor thecontrolvariables.TheregressionincolumnC1includes allthecovariates.ColumnsC2andC3dropfromC1the variable Non_DJSI andthecontrolvariables,respectively,in ordertocheckifthereareinteractioneffectswithindifferent setsofexplanatoryvariables.Theestimateforthecoefcient ofthevariable Non_DJSI isnegativeandstatisticallysignicant( 4.157; p value = 0.020),whichmeansthatrms notincludedintheDJSINorthAmericaareassociatedwitha loweraveragemarketprice,afterconsideringthecompeting variablesincludedintheregressions. Theestimatesfortheaccountinginformationarestatisticallysignicantandtheyhavetheexpectedsign.For Table2 Descriptivestatistics MeanMedianSDMinMaxSkewnessKurtosis Allrm-yearobs.( n = 1,597) MV 37.46034.54021.0241.290103.0500.7170.061 BV 17.14114.40211.1090.26555.6050.9840.543 NI 1.9541.9322.329 8.3179.900 0.4082.802 ROE 0.1320.1340.325 8.8242.545 13.154368.412 LEV 0.7700.2762.8960.00071.78016.201327.701 SIZE 16.57916.4521.30513.96721.5410.7150.651 CF 0.1040.0940.074 0.1220.4300.7011.060 RISK 1.2501.1000.7040.0904.1101.3002.198 DJSIrm-yearobs.( n = 241) MV 43.84241.45022.0622.98098.3700.392 0.686 BV 17.74615.33211.8781.81355.6051.2620.312 NI 2.7072.5802.259 8.2369.477 0.3263.168 ROE 0.1770.1710.226 1.7040.916 2.99225.301 LEV 0.5410.1991.5810.00018.6938.13379.935 SIZE 17.21017.1351.14715.34621.3420.8300.942 CF 0.1210.1160.070 0.0820.3010.1890.095 RISK 1.0920.9800.6210.3003.6101.5333.203 Non_DJSIrm-yearobs.( n = 1356) MV 36.32633.62520.6361.290103.0500.7810.321 BV 17.03414.18710.9680.26554.9320.9170.322 NI 1.8201.8512.316 8.3179.900 0.4322.854 ROE 0.1240.1260.339 8.8252.545 13.459364.575 LEV 0.8100.2973.0700.00071.78015.797303.844 SIZE 16.46716.3351.30013.96721.5410.7900.781 CF 0.1000.0900.075 0.1220.4300.8021.326 RISK 1.2781.1400.7140.0904.1101.2622.071 DJSI rmsarethoseincludedintheDJSIeveryyearforthesampleperiod20072010, Non_DJSI rmsarethosewhohaveneverbeenincluded intheDJSIduringthesampleperiod20072010, MVisthemarketpriceatthescalyear-end, BV isthebookvalueofequityasoftheendofthe year, NI isthenetincomeoftheyear, ROE isthereturnonequity, LEV isend-of-yeartotaldebtdividedbyend-of-yearmarketcapitalization, SIZE isthenaturallogarithmoftotalassetsasoftheendoftheyear, CF isnetcash-owfromoperatingactivitiesscaledbyend-of-yeartotal assets, RISK isBetaasreportedbyWorldScope 6STATA10hasbeenusedtocomputetheBreusch-Pagantestandto estimatetherandomeffectsmodels. HowDoestheMarketValueCSP? 423123


example,inthemainregression,the BV and NI coefcients are0.773and2.587,respectively,andthe p valueassociated totheindividual t testsis \ 0.01inbothcases.Themajority ofthecontrolvariablesarealsostatisticallysignicantand theirsignisinaccordancewiththeliterature.Forexample, rmswithlargercash-owsfromoperationsareassociated withahighermarketprice,whilehighleveragermsare associatedwithalowermarketprice.Contrarytotheliterature,theestimateforthevariable SIZE isstatistically signicantbutwithanegativesign.Furtheranalysisshows thattheestimatedcoefcientassociatedwiththisvariable isstatisticallysignicantbutonlywhenthevariable Non_DJSI isalsoincludedintheregression(inC2,the SIZE isnotstatisticallysignicant),whichmeansthattherelation between SIZE and MV isonlyobservableforoneofthe groupsofrmsbasedonthesustainabilitycriteria.Thenext sectionprovidesmoredetailedinformationonthisissue. TheAssociationofMarketValueofEquitywithCSP: EffectofSize Table 6 presentssummarystatisticsresultingfromthe estimationoftheEq. 3 .Thecoefcientestimateforthe variable Non_DJSI_Big isnegativeandstatisticallysignicant( 5.060; p value \ 0.01),whilethecoefcient estimatefor Non_DJSI_Small isnotstatisticallysignicant. Theseresultsshowthatthemarketdoesnotpenalizeallthe rmswithalowerlevelofCSP.Onthecontrary,the marketpenalizesonlythosermswithincentivestopresentahighlevelofCSP(largerms)butthatdonotengage onsuchstrategy,i.e.,thegroupofthelargerrmsnot includedintheDJSINorthAmerica. TheAssociationofMarketValueofEquitywithCSP: EffectofSizeandProtability Table 7 presentssummarystatisticsresultingfromthe estimationofEq. 4 .Thecoefcientestimateforthevariable Non_DJSI_Big_Prot isnegativeandstatisticallysignicant( 5.119; p value \ 0.01),whilethecoefcientestimatesfor Non_DJSI_Big_Loss Non_DJSI_Small_Prot and Non_DJSI_Small_Loss areallstatisticallynotsignicant.Theseresultsshowthatinaveragethemarketdoesnot penalizeallthelargerrmswithalowerlevelofCSP,but onlythosethatareprotable.Thus,themarketdistinguishes groupsofrmswithalowerlevelofCSPbasednotonlyon sizebutalsoonprotability. Overall,ourndingsseemtosuggestthatsizeand protabilitymatterintermsofCSP.Theinformationonthe relationbetweensizeandprotabilityandthelevelofCSP isvaluerelevantforthemarket. DiscussionandConcludingComments Thisstudyprovidesvaluablenewinsightsthathelptoclarify thendingsofrecentstudiesontherelationshipbetween CSPandCFP.Somestudies,suchasCurranandMoran ( 2007 ),Consolandietal.( 2009 ),Cheung( 2011 ),Dohetal. ( 2010),andRobinsonetal.( 2011 )testwhetherinclusionin, ordeletionfrom,sustainabilityindexes(suchasthe FTSE4GoodUK50Index,theDowJonesSustainability StoxxIndex,theDowJonessustainabilityworldindexand theCalvertsocialindex),resultsinapositive(negative) impact.ResultssuggestthatinvestorsdovalueCSP. Table3 Correlationmatrix MVBVNIROELEVSIZECFRISK MV 1 BV 0.470***1 NI 0.605***0.383***1 ROE 0.191*** 0.083***0.435*** 1 LEV 0.186*** 0.004 0.162*** 0.099***1 SIZE 0.062**0.395***0.116*** 0.052**0.187***1 CF 0.197*** 0.289***0.237***0.311*** 0.223*** 0.435***1 RISK 0.195***0.036 0.230*** 0.200***0.186***0.007 0.250***1 MV isthemarketpriceatthescalyear-end, BV isthebookvalueofequityasoftheendoftheyear, NI isthenetincomeoftheyear, ROE isthe returnonequity, LEV isend-of-yeartotaldebtdividedbyend-of-yearmarketcapitalization, SIZE isthelogarithmoftotalassetsasoftheendof theyear, CF isnetcash-owfromoperatingactivitiesscaledbyend-of-yeartotalassets, RISK isBetaasreportedbyWorldScope Table4 BreuschPaganLMtestresults TestvalueSig. Equation 2 400.000.000 Equation 3 394.950.000 Equation 4 382.900.000 TheBreuschPaganLagrangeMultipliertestisusedtotestthepooled regressionagainsttherandomeffectsregression.Ifthenullhypothesisisrejected,therandomeffectsregressionismoreappropriatethan thepooledregression(Baltagi 2001 ) 424 I.C.Lourenc oetal.123


Otherstudies,suchasLoandSheu( 2007 ),GarciaCastroetal.( 2010 ),andWagner( 2010 ),aremoresimilar toourowninthattheyusepaneldataandexaminewhether CSPhasanimpactonmarketvalue.Giventhattheirresults suggestthatKLDdoesnotimpactonnancialperformance,Garcia-Castroetal.( 2010 )arguethatthepositive relationshipfoundinmostofthepreviousresearchonthe linkbetweenCSPandFPbecomesanon-signicantor evenanegativerelationshipwhenendogeneityisproperly takenintoaccount. Thendingsoftheothertwostudiessuggestthatsustainablermsaremorelikelytoberewardedbyinvestors. ThendingsofLoandSheu( 2007 ),whoexaminewhether CShasanimpactonmarketvalueusinglargeUSnonnancialrmsfrom1999to2002,areespeciallyrelevant toourstudy.TheyusedlistingintheDJSGIUSAasthe proxyforCSandtheTobin'sqastheproxyforrmvalue. Theirkeyndingisthatsustainablermsarerewarded withhighervaluationsinthemarketplace. OurndingsareconsistentwiththendingsofLoand Sheu( 2007 )andthoseofotherstudiesthatndapositive relationbetweenCSPandCFP.Wendthatinvestorsdo valueCSP.However,whathappensisthattheypenalize largeprotablermswithlowlevelofCSP,whichface greaterpublicscrutinyandpressuresfromstakeholders. Thesecompaniesareexpectedtosignalsustainability Table5 Firsthypothesistest:theassociationofmarketvalueof equitywithCSP Exp.signC1C2C3 Intercept45.617***35.048***24.917*** Mainvariables BV ? 0.773***0.767***0.618*** NI ? 2.587***2.601***2.963*** Non_DJSI -4.157** 4.432*** Controlvariables ROE ?1.187 1.249 LEV -0.627*** 0.629*** SIZE ?1.321** 0.888 CF ? 28.248***29.545*** RISK -3.590*** 3.663*** LIST ? 2.8523.098 Mining 1.1070.555 Utilities 5.584*** 5.995*** Commercial 2.930 2.877 Financial 1.593 2.620 Services 1.007 1.046 Overall R20.4770.4740.422 Waldtest810.04***801.35***140.85*** Dependentvariable: MV marketpriceatthescalyear-end;Independentvariables: BV bookvalueofequityasoftheendoftheyear, NI netincomeoftheyear, Non_DJSI anindicatorthatequals1ifthe rmhaveneverbeenincludedintheDJSIduringthesampleperiod 20072010, ROE returnonequity, LEV end-of-yeartotaldebtdivided byend-of-yearmarketcapitalization, SIZE logarithmoftotalassetsas oftheendoftheyear, CF netcash-owfromoperatingactivities scaledbyend-of-yeartotalassets, RISK BetaasreportedbyWorldScope, LIST anindicatorthatequals1ifthermislistedinaforeign stockexchangeand0otherwise;Industryvariables: Mining (SIC1), Utilities (SIC4), Commercial (SIC5),Financial(SIC6),and Services (SIC7andSIC8) Duetothesamplepaneldata,andbasedonBreuschPagantest, randomeffectsregressionwasconducted ***,**,and*Signicanceatthe0.01,0.05,and0.10levels, respectively Table6 Secondhypothesistesttheassociationofmarketvalueof equitywithCSP:effectofsize Exp.sign Intercept31.612** Mainvariables BV ? 0.764*** NI ? 2.594*** Non_DJSI_Big -5.060*** Non_DJSI_Small -1.957 Controlvariables ROE ?1.231 LEV -0.626*** SIZE ?0.525 CF ? 27.758*** RISK -3.657*** LIST ? 3.316 Mining 1.250 Utilities 5.039** Commercial 2.792 Financial 1.629 Services 0.954 OverallR20.480 Waldtest815.81*** Dependentvariable: MV marketpriceatthescalyear-end;Independentvariables: BV bookvalueofequityasoftheendoftheyear, NI netincomeoftheyear, Non_DJSI_Big anindicatorthatequals1if thermhaveneverbeenincludedintheDJSIduringthesample period20072010andhissizeisabovethemedian, Non_DJSI_Small anindicatorthatequals1ifthermhaveneverbeenincludedinthe DJSIduringthesampleperiod20072010andhissizeisbelowthe median, ROE returnonequity; LEV end-of-yeartotaldebtdividedby end-of-yearmarketcapitalization, SIZE logarithmoftotalassetsasof theendoftheyear, CF netcash-owfromoperatingactivitiesscaled byend-of-yeartotalassets, RISK BetaasreportedbyWorldScope, LIST anindicatorthatequals1ifthermislistedinaforeignstock exchangeand0otherwise;Industryvariables: Mining (SIC1), Utilities (SIC4), Commercial (SIC5), Financial (SIC6),and Services (SIC7andSIC8) Duetothesamplepaneldata,andbasedonBreuschPagantest, randomeffectsregressionwasconducted ***,**,and*Signicanceatthe0.01,0.05,and0.10levels, respectively HowDoestheMarketValueCSP? 425123


leadership.Iftheydonot,theyarepenalizedbythemarket. Thendingsinthisstudyareimportanttotheongoing debateaboutthenancialconsequencesofcorporate investmentinsustainabilityactivities. WeaddresssomeproblemsidentiedbyGarcia-Castro etal.( 2010 )asbeinglikelytoexplaintheheterogeneous resultsfoundinpreviousstudies,suchasusingaconsistent measureofCSP,includingalltherelevantcontrolvariables,distinguishingbetweenshort-andlongrunnancial effects,andtheendogeneityofstrategicdecisions.The lattertwoproblemsareaddressednamelybyusingofpanel datamethods. WealsoaddressthesuggestionofGarcia-Castroetal. ( 2010 ),whousedthemostcompleteKLDpaneldata availableatthetime(19912005),thatfutureresearch shouldlookatrm-speciccharacteristicsthatpushrms toadoptsustainabilitypracticesintherstplace.Recently, Artiachetal.( 2010 )examinedtheincentivesforUSrms toinvestinsustainability.Theyexaminedrm-specic factorsassociatedwithhighCSP,asproxiedbymembershipoftheDowJonessustainabilityindex.Theyfoundthat leadingCSPrmsaresignicantlylargerandprotable whencomparedwithconventionalrms.Ourndings suggestthatlargeandprotablermsarewelladvisedto investinCS,becausetheywouldnotdowellotherwise. OurpurposeistodescribethelinkbetweenCSPand CFP.Inanovelapproach,wedistinguishrmsbasedon sizeandprotability.OurndingssuggestthatCSPis positivelyassociatedwiththenancialperformanceof largeandprotablermswhichareabletosignaltheir sustainabilityperformance,andhasanegativeassociation withtheperformanceoflargeandprotablermsthatare notabletosignaltheirsustainabilityperformance.CS makeslargeandprotablermsthathaveareputationfor beingcommittedtosustainabilitybetterandlargeand protablermswithoutthatreputationworse. Itisprudenttoconcludethatourndings,obtainedin theNorthAmericaninstitutionalsetting,arenotsusceptible ofgeneralizationtoothercountries,especiallythosewith verydifferentcharacteristics.CormierandMagnan( 2007 ) suggestthatnationalinstitutionalcontextsarerelevant whenassessingthestockmarketvaluerelevanceof nancialandnon-nancialperformancemeasures.Thisis oneofthereasonsforlimitingoursampletoNorth Americancompaniesandalsoapromisingavenuefor futureresearch.Webelievethatstudyinginternationaldata forcross-countrycomparisonsandindustrycomparisons wouldbeaninterestingfutureresearcharea. ReferencesArtiach,T.,Lee,D.,Nelson,D.,&Walker,J.(2010).The determinantsofcorporatesustainabilityperformance. AccountingandFinance,50 ,3151. Aupperle,K.E.,Carroll,A.B.,&Hateld,J.D.(1985).Anempirical examinationoftherelationshipbetweencorporatesocial Table7 Thirdhypothesistesttheassociationofmarketvalueof equitywithCSP:effectofsizeandprotability Exp.sign Intercept31.430** Mainvariables BV ? 0.761*** NI ? 2.707*** Non_DJSI_Big_Prot -5.119*** Non_DJSI_Big_Loss -3.879 Non_DJSI_Small_Prot -2.040 Non_DJSI_Small_Loss -0.724 Controlvariables ROE ?1.077 LEV -0.635*** SIZE ?0.529 CF ? 28.048*** RISK -3.662*** LIST ? 3.227 Mining 1.239 Utilities 4.961** Commercial 2.692 Financial 1.604 Services 0.899 Overall R20.481 Waldtest823.05*** Dependentvariable: MV marketpriceatthescalyear-end;Independentvariables: BV bookvalueofequityasoftheendoftheyear, NI netincomeoftheyear, Non_DJSI_Big_Prot anindicatorthatequals 1ifthermhaveneverbeenincludedintheDJSIduringthesample period20072010andhissizeisabovethemedianandROEis positive, Non_DJSI_Big_Loss anindicatorthatequals1iftherm haveneverbeenincludedintheDJSIduringthesampleperiod 20072010andhissizeisabovethemedianandROEisnegative, Non_DJSI_Smal_Prot anindicatorthatequals1iftherm haveneverbeenincludedintheDJSIduringthesampleperiod 20072010,andhissizeisbelowthemedianandROEispositive, Non_DJSI_Smal_Loss anindicatorthatequals1ifthermhavenever beenincludedintheDJSIduringthesampleperiod20072010,and hissizeisbelowthemedianandROEisnegative, ROE returnon equity, LEV end-of-yeartotaldebtdividedbyend-of-yearmarket capitalization, SIZE logarithmoftotalassetsasoftheendoftheyear, CF netcash-owfromoperatingactivitiesscaledbyend-of-yeartotal assets, RISK BetaasreportedbyWorldScope, LIST anindicatorthat equals1ifthermislistedinaforeignstockexchangeand0otherwise;Industryvariables: Mining (SIC1), Utilities (SIC4), Commercial (SIC5), Financial (SIC6),and Services (SIC7andSIC8) Duetothesamplepaneldata,andbasedonBreuschPagantest, randomeffectsregressionwasconducted ***,**,and*Signicanceatthe0.01,0.05,and0.10levels, respectively 426 I.C.Lourenc oetal.123


responsibilityandprotability. AcademyofManagementJournal,28 (2),446463. Baltagi,B.H.(2001). Econometricanalysisofpaneldata (2nded.). Chichester:Wiley. Bansal,P.(2005).Evolvingsustainability:Alongitudinalstudyof corporatesustainabledevelopment. StrategicManagementJournal,26 (3),197218. Bansal,P.,&Bogner,W.C.(2002).DecidingonISO14001: Economics,institutions,andcontext. LongRangePlanning, 35 (3),269290. Barney,J.B.(1999).Howarm'scapabilitiesaffectboundary decisions. SloanManagementReview,40 (3),137145. Bowman,C.,&Ambrosini,V.(2003).Howtheresource-basedand thedynamiccapabilityviewsoftherminformcorporate-level strategy. BritishJournalofManagement,14 (4),289303. Brammer,S.,Millington,A.,&Rayton,B.(2007).Thecontribution ofcorporatesocialresponsibilitytoorganizationalcommitment. TheInternationalJournalofHumanResourceManagement, 18 (10),17011719. Brammer,S.,&Pavelin,S.(2004).Voluntarysocialdisclosuresby largeUKcompanies. BusinessEthics:AEuropeanReview,13 (2/ 3),8699. Branco,M.C.,&Rodrigues,L.L.(2006).Corporatesocial responsibilityandresourcebasedperspectives. Journalof BusinessEthics,69 (2),111132. Branco,M.C.,&Rodrigues,L.L.(2008).Socialresponsibility disclosure:AstudyofproxiesforthepublicvisibilityofPortuguese banks. TheBritishAccountingReview,40 (2),161181. Campbell,J.L.(2007).Whywouldcorporationsbehaveinsocially responsibleways?Aninstitutionaltheoryofcorporatesocial responsibility. AcademyofManagementReview,32 (3),946967. Cheung,A.(2011).Dostockinvestorsvaluecorporatesustainability? Evidencefromaneventstudy. JournalofBusinessEthics forthcoming. Chih,H.-L.,Chih,H.-H.,&Chen,T.-Y.(2010).Onthedeterminants ofcorporatesocialresponsibility:Internationalevidenceonthe nancialindustry. JournalofBusinessEthics,93 ,115135. Consolandi,C.,Jaiswal-Dale,A.,Poggiani,E.,&Vercelli,A.(2009). Globalstandardsandethicalstockindexes:ThecaseoftheDow Jonessustainabilitystoxxindex. JournalofBusinessEthics,87 185197. Cormier,D.,&Magnan,M.(2007).Therevisedcontributionof environmentalreportingtoinvestors'valuationofarm's earnings:Aninternationalperspective. EcologicalEconomics, 62 (3/4),613626. Curran,M.M.,&Moran,D.(2007).ImpactoftheFTSE4GoodIndex onrmprice:Aneventstudy. JournalofEnvironmental Management,82 ,529537. Curto,J.,Morais,A.,Lourenc o,I.,&Pinto,J.(2011).The heteroskedasticity-consistentcovarianceestimatorinaccounting. ReviewofQuantitativeFinanceandAccounting forthcoming. Doh,J.P.,&Guay,T.R.(2006).Corporatesocialresponsibility, publicpolicy,andNGOactivisminEuropeandtheUnited States:Aninstitutional-stakeholderperspective. Journalof ManagementStudies,43 (1),4773. Doh,J.P.,Howton,S.D.,Howton,S.W.,&Siegel,D.S.(2010). Doesthemarketrespondtoanendorsementofsocialresponsibility?Theroleofinstitutions,information,andlegitimacy. JournalofManagement,36 (6),14611485. EuropeanCommission(EC).(2002). CorporatesocialresponsibilityAbusinesscontributiontosustainabledevelopment .Luxemburg:OfceforOfcialPublicationsoftheEuropean Communities. Figge,F.,&Hahn,T.(2004).SustainablevalueaddedMeasuring corporatecontributionstosustainabilitybeyondeco-efciency. EcologicalEconomics,48 ,173187. Gallego-A lvarez,I.,Prado-Lorenzo,J.-M.,Rodr guez-Dom nguez,L., &Garc a-Sa nchez,I.-M.(2010).Aresocialandenvironmental practicesamarketingtool?Empiricalevidenceforthebiggest Europeancompanies. ManagementDecision,48 (10), 14401455. Garcia-Castro,R.,Arin o,M.A.,&Canela,M.A.(2010).Doessocial performancereallyleadtonancialperformance?Accounting forendogeneity. JournalofBusinessEthics,92 ,107126. Godfrey,P.C.,Merrill,C.G.,&Hansen,J.M.(2009).The relationshipbetweencorporatesocialresponsibilityandshareholdervalue:Anempiricaltestoftheriskmanagement hypothesis. StrategicManagementJournal,30 (4),425445. Hassel,L.,Nilsson,H.,&Nyquist,S.(2005).Thevaluerelevanceof environmentalperformance. EuropeanAccountingReview, 14 (1),4161. Hillman,A.J.,&Keim,G.D.(2001).Shareholdervalue,stakeholder management,andsocialissues:What'sthebottomline? StrategicManagementJournal,22 ,125139. Holme,R.,&Watts,P.(2000). CorporateSocialResponsibility: makinggoodbusinesssense ,WorldBusinessCouncilfor SustainableDevelopment. Hussainey,K.,&Salama,A.(2010).Theimportanceofcorporate environmentalreputationtoinvestors. JournalofApplied AccountingResearch,11 (3),229241. Johnston,D.,Sefcik,S.,&Soderstrom,N.(2008).Thevalue relevanceofgreenhousegasemissionsallowances:AnexploratorystudyintherelatedUnitedStatesCO2market. European AccountingReview,17 (4),747764. Kraaijenbrink,J.,Spender,J.-C.,&Groen,A.J.(2010).Theresourcebasedview:Areviewandassessmentofitscritiques. Journalof Management,36 ,349372. Lacy,P.,Cooper,T.,Hayward,R.,&Neuberger,L.(2010). Anewera ofsustainabilityUNGlobalCompact AccentureCEOStudy 2010 ,Accenture. Lo,S.,&Sheu,H.(2007).Iscorporatesustainabilityavalue-increasing strategyforbusiness? CorporateGovernance,15 (2),345358. Lo pez,M.V.,Garcia,A.,&Rodriguez,L.(2007).Sustainable developmentandcorporateperformance:Astudybasedonthe DowJonessustainabilityindex. JournalofBusinessEthics,75 285300. Margolis,J.D.,&Walsh,J.P.(2003).Miserylovescompanies: Rethinkingsocialinitiativesbybusiness. AdministrativeScience Quarterly,48 ,268305. Matolcsy,Z.,&Wyatt,A.(2008).Theassociationbetweentechnologicalconditionsandthemarketvalueofequity. TheAccountingReview,83 (2),479518. McWilliams,A.,Siegel,D.,&Wright,P.M.(2006).Corporatesocial responsibility:Strategicimplications. JournalofManagement Studies,43 (1),118. Ohlson,J.(1995).Earnings,bookvalueanddividendsinsecurity valuation. ContemporaryAccountingResearch,11 ,661687. Orlitzky,M.(2008).Corporatesocialperformanceandnancial performance:Aresearchsynthesis.InA.Crane,A.McWilliams, D.Matten,J.Moon,&D.S.Siegel(Eds.), TheOxfordhandbook ofcorporatesocialresponsibility (pp.113134).Oxford:Oxford UniversityPress. Orlitzky,M.,Schmidt,F.L.,&Rynes,S.L.(2003).Corporatesocial andnancialperformance:Ameta-analysis. OrganizationStudies,24 (3),403441. Pava,M.L.,&Krausz,J.(1996).Theassociationbetweencorporate social-responsibilityandnancialperformance:Theparadoxof socialcost. JournalofBusinessEthics,15 ,321357. Pertusa-Ortega,E.M.,Molina-Azor n,J.F.,&Claver-Corte s,E. (2010).Competitivestrategy,structureandrmperformance:A comparisonoftheresource-basedviewandthecontingency approach. ManagementDecision,48 (8),12821303. HowDoestheMarketValueCSP? 427123


Porter,M.E.,&Kramer,M.R.(2006).Strategyandsociety:Thelink betweencompetitiveadvantageandcorporatesocialresponsibility. HarvardBusinessReview,84 (12),7892. Post,J.E.,Preston,L.E.,&Sachs,S.(2002).Managingtheextended enterprise:Thenewstakeholderview. CaliforniaManagement Review,45 (1),628. Rajgopal,S.,Venkatachalam,M.,&Kotha,S.(2003).Thevalue relevanceofnetworkadvantages:thecaseofE-commercerms. JournalofAccountingResearch,41 (1),135162. Roberts,R.W.(1992).Determinantsofcorporatesocialresponsibility disclosure:Anapplicationofstakeholdertheory. Accounting, OrganizationsandSociety,17 (6),595612. Roberts,P.W.,&Dowling,G.R.(2002).Corporatereputationand sustainedsuperiornancialperformance. StrategicManagement Journal,23 ,10771093. Robinson,M.,Kleffner,A.,&Bertels,S.(2011).Signalingsustainabilityleadership:EmpiricalevidenceofthevalueofDJSI membership. JournalofBusinessEthics,101 ,493505. Roman,R.M.,Hayibor,S.,&Agle,B.R.(1999).Therelationship betweensocialandnancialperformance. BusinessandSociety, 38 (1),109125. Ruf,B.M.,Muralidhar,K.,Brown,R.M.,Jay,J.,&Paul,K.(2001). Anempiricalinvestigationoftherelationshipbetweenchangein corporatesocialperformanceandnancialperformance:A stakeholdertheoryperspective. JournalofBusinessEthics, 32 (2),143156. Schadewitz,H.J.,&Niskala,M.(2010).Communicationvia responsibilityreportinganditseffectonrmvalueinFinland. CorporateSocialResponsibilityandEnvironmentalManagement,17 (2),96106. Siegel,D.S.(2009).Greenmanagementmattersonlyifityieldsmore green:Aneconomic/strategicperspective. AcademyofManagementPerspectives,23 (3),516. Sinkin,C.,Wright,C.J.,&Burnett,R.D.(2008).Eco-efciencyandrm value. JournalofAccountingandPublicPolicy,27 ,167176. Suchman,M.C.(1995).Managinglegitimacy:Strategicandinstitutionalapproaches. AcademyofManagementReview,20 (3), 571610. Surroca,J.,Tribo ,J.A.,&Waddock,S.(2010).Corporate responsibilityandnancialperformance:Theroleofintangible resources. StrategicManagementJournal,31 (5),463490. Ullman,A.E.(1985).Datainsearchofatheory:Acritical examinationoftherelationshipsamongsocialperformance, socialdisclosureandeconomicperformanceofU.S.rms. AcademyofManagementReview,10 (3),540557. Vitaliano,D.F.(2010).Corporatesocialresponsibilityandlabor turnover. CorporateGovernance,10 (5),563573. Waddock,S.A.,&Graves,S.B.(1997).Thecorporatesocial performancenancialperformancelink. StrategicManagement Journal,18 (4),303319. Wagner,M.(2010).Theroleofcorporatesustainabilityperformance foreconomicperformance:Arm-levelanalysisofmoderation effects. EcologicalEconomics,69 ,15531560. Wood,D.J.,&Jones,R.E.(1995).Stakeholdermismatching:A theoreticalprobleminempiricalresearchoncorporatesocial performance. TheInternationalJournalofOrganizational Analysis,3 (3),229267. Ziegler,A.,&Schro der,M.(2010).Whatdeterminestheinclusionin asustainabilitystockindex?ApaneldataanalysisforEuropean rms. EcologicalEconomics,69 ,848856. 428 I.C.Lourenc oetal.123