Journal of Business Ethics

, Volume 109, Issue 3, pp 367–388

Corporate Social and Financial Performance Re-Examined: Industry Effects in a Linear Mixed Model Analysis

Authors

  • Philip L. Baird
    • Duquesne University, Palumbo–Donahue Schools of Business
  • Pinar Celikkol Geylani
    • Duquesne University, Palumbo–Donahue Schools of Business
    • Duquesne University, Palumbo–Donahue Schools of Business
Article

DOI: 10.1007/s10551-011-1135-z

Cite this article as:
Baird, P.L., Geylani, P.C. & Roberts, J.A. J Bus Ethics (2012) 109: 367. doi:10.1007/s10551-011-1135-z

Abstract

In this research, we shed new light on the empirical link between corporate social performance (CSP) and corporate financial performance (CFP) via the application of empirical models and methods new to the CSP–CFP literature. Applying advanced financial models to a uniquely constructed panel dataset, we demonstrate that a significant overall CSP–CFP relationship exists and that this relationship is, in part, conditioned on firms’ industry-specific context. To accommodate the estimation of time-invariant industry and industry-interaction effects, we estimate linear mixed models in our test of the CSP–CFP relationship. Our results show both a significant overall CSP effect as well as significant industry effects between CSP and CFP. In conflict with expectations, the unweighted average effect of CSP on CFP is negative. Our industry analysis, however, shows that in over 17% of the industries in our sample, the effect of CSP on CFP for socially responsible firms is positive. We also examine the multidimensional nature of the CSP construct in an industry context by exploring the CSP dimension–industry nexus and identify dimensions of social performance that are associated with either better or worse financial performance. Our results confirm the existence of disparate CSP dimension–industry effects on CFP, thus our results provide important and actionable information to decision makers considering whether and how to commit corporate resources to social performance.

Keywords

Corporate social performanceCorporate financial performancePanel data analysisLinear mixed modelResidual income model

Introduction

In their literature review of the empirical relation between corporate social and financial performance, van Beurden and Gossling (2008, p. 407) conclude that “Good ethics is good business,” by which they mean that good corporate social performance (CSP) leads to good financial performance (CFP).1 While the evidence suggests that, on balance, this is the case, it is not true for all firms under all conditions. Social performance is multidimensional, and there is no consensus on what should be considered part of an organization’s social responsibility (van Beurden and Gossling 2008). Moreover, whether social performance leads to good financial performance seems to depend on how social performance is measured. Orlitzky et al. (2003) found that reputation indices are more highly correlated with financial performance than are other indicators of social performance and a substantial number of studies found a negative relationship between corporate social and financial performance (Griffin and Mahon 1997). Consequently, for an individual firm, determining which social performance initiatives might improve (or erode) financial performance may not be readily apparent. Recognizing that corporate resources are limited, how should a firm prioritize socially responsible expenditures related to diversity, human rights, the environment, product quality, and corporate governance? Stakeholder Theory, which provides the theoretical foundation for corporate social responsibility, offers little guidance to decision makers on these questions (Jensen 2002). Hence, these questions must be addressed on empirical grounds.

This research into the empirical CSP–CFP link is conducted with an eye toward results that are useful to decision makers considering whether and how to commit corporate resources to social performance initiatives. Our view is that whether CSP leads to better or worse financial performance depends on the interactions of a number of factors: managerial motives, industry environment, the firm’s capabilities and limitations, and the particular modes of social performance the firm chooses to pursue. This view is motivated by previous research which shows that a firm’s response to stakeholder demands on social responsibility evolves from the firm’s industry context and is conditioned by its unique capabilities (Griffin and Mahon 1997; Waddock and Graves 1997; Hillman and Keim 2001; Rowley and Berman 2000; Simpson and Kohers 2002).

Our research differs from previous studies in three important ways. First, much of the literature indicates that stakeholder expectations and firms’ motives for CSP vary by industry. Hence, it seems reasonable that the CSP–CFP relationship varies by industry. We explicitly consider variation in the CSP–CFP relationship across more than 50 industries represented in our sample. To accommodate time-invariant industry effects, we estimate a linear mixed model (LMM) in lieu of a fixed effects model, and we statistically establish the existence of industry effects in the empirical CSP–CFP relationship. Second, we examine the multidimensional nature of the CSP construct in an industry context. In specific, we explore the CSP dimension–industry nexus, which allows us to identify the particular dimensions of social performance that are associated with either better or worse financial performance in those industries exhibiting a significant CSP–CFP relationship. Third, we measure financial performance in terms of stock price—a financial metric well suited to capture the long-run impact of social performance on stakeholder relationships. To more clearly isolate the relationship between social and financial performance, our models control for stock price intrinsic value—a measure shown to possess very high explanatory power in stock price studies (Courteau et al. 2001). Finally, our results on the CSP dimension–industry nexus provide insights that should prove useful to decision makers contemplating whether and how to commit corporate resources to social performance initiatives.

The remainder of the article is organized as follows. In §2, we review the CSP–CFP relationship literature. In §3, we outline the theoretical and empirical support for three hypotheses concerning the nature of industry–CSP–CFP relationships. §4 details the research environment of our work including data acquisition, description of measures, empirical models, estimation methods, and statistical results. Lastly, in §5 we summarize our results, discuss the implications of our findings and offer insights for practitioners interested in pursuing socially responsible initiatives.

Prior Literature

The empirical link between CSP and CFP has been studied extensively. Recent reviews suggest that, on balance, the relationship is positive (e.g., van Beurden and Gossling 2008; Orlitzky et al. 2003; Margolis and Walsh 2003; Roman et al. 1999; Griffin and Mahon 1997). The results of individual studies, however, vary widely. Although a majority of studies find a positive relation, many find no relation and some find it to be negative. One possible explanation for these mixed results is that the multidimensional nature of CSP and the absence of a consensus on what should comprise it make it difficult to measure. Marom’s (2006) unified theory of the CSP–CFP link suggests that aggregate indices of CSP may yield confounded results in empirical studies. Hillman and Keim (2001) distinguish between stakeholder management and social issue participation. These authors suggest that whereas improved relations with primary stakeholders can enhance competitive advantage and lead to better financial performance the allocation of corporate resources to social issues that do not increase competitive advantage can reduce financial performance. Hence, CSP will sometimes increase CFP and sometimes decrease it.

Accounting Measures of Finance Performance

Like CSP, CFP is a multidimensional construct that has been measured in a variety of ways. Several authors find that the empirical CSP–CFP link depends on the operational definition of financial performance and that “CSP appears to be more highly correlated with accounting-based measures of CFP than with market-based indicators” (Orlitzky et al. 2003, p. 403; McGuire et al. 1988; Wu 2006; Margolis et al. 2007). Importantly, however, accounting-based measures may not provide meaningful measures of financial performance in these studies. Marom (2006) argues that while CSP-related expenditures occur in the short term, stakeholder reactions to them unfold over the long term. Consequently, the effects of CSP initiatives on CFP cannot be observed unless a long-term perspective is taken. Accounting measures, however, provide inherently short-term, historical measures of performance.2 Accounting performance does not adequately capture the intangible, long-term value created or destroyed through stakeholder relationships (Hillman and Keim 2001) and reflects managers’ discretionary choices made under generally accepted accounting principles. Financial ratios such as return on assets, return on sales, return on equity, growth rates, etc. capture a single aspect of financial performance that may or may not be correlated with the value of the firm (Lubatkin and Shrieves 1986). Finally, accounting performance varies by industry.

Market Returns and Financial Performance

Another line of inquiry investigates the relation between social performance and rates of return from common stocks, mutual funds, and stock indexes. These studies typically find no statistically significant relationship between returns and social performance. Hamilton et al. (1993) found that socially responsible mutual funds did not earn statistically significant excess returns and that the performance of such mutual funds was not statistically different from the performance of conventional mutual funds. Goldreyer and Diltz (1999) investigated mutual fund returns and concluded that social screening does not affect the investment performance of mutual funds in any systematic, predictable way. Statman (2000) found that the Domini Social Index (DSI) did as well as the Standard & Poor’s (S&P) 500 Index from 1990 to 1998 and that socially responsible mutual funds underperformed both the S&P 500 and the DSI but did no worse than conventional mutual funds. Statman (2006) analyzed the monthly alphas of four socially responsible indexes and could not reject the hypothesis that returns of socially responsible companies are equal to those of conventional firms. Bauer et al. (2005) examined the performance of 103 German, UK, and US ethical mutual funds and found no evidence of significant differences in risk-adjusted returns between ethical and conventional funds from 1990 to 2001. Brammer et al. (2006) found that the common stocks of companies with higher social performance scores had lower returns than the stocks of firms with lower social performance scores. They also found that higher scores on environmental and community criteria were associated with lower stock returns and that higher scores on employment criteria were associated with higher stock returns. Statman and Glushkov (2009) analyzed the excess monthly returns to long–short portfolios formed on KLD ratings on community, employee relations, diversity, environment, products, human rights and governance. From January 1992 through September 2007, the excess returns of high-rated stocks on two of these dimensions (community and employee relations) exceeded the returns of low-rated stocks on these dimensions by a statistically and economically significant margin. Their results, however, were not robust either to portfolio formation method or to time period. In a similar study, Kempf and Osthoff (2007) found that from 1992 to 2004 a simple trading strategy of going long on stocks rated high on KLD dimensions and going short on stocks rated low would have generated positive and statistically significant returns.

Kempf and Osthoff (2007) asked whether abnormal performance results from temporary mispricing or whether it compensates for an additional risk factor. If abnormal performance is the result of a correction of mispricing, then the light it sheds on the CSP–CFP relationship depends on the reason for the mispricing—which is not known. If abnormal performance is the result of risk compensation, then differential returns from investing in socially responsible companies indicate that such investments must involve greater or lesser risk than investments in conventional firms. Unless one assumes that differential returns compensate for risk, it is hard to draw firm conclusions on the nature of the CSP–CFP relationship from studies that measures financial performance in terms of rates of return.

Conceptual Background and Hypothesis Development

Over the past several years, scholars have made used of several conceptual and theoretical bases to develop and test hypotheses regarding the nature of the CSP–CFP relationship. Depending on the theoretical orientation, arguments have been made for neutral, negative and positive CSP–CFP relationships.

In support of a neutral CSP–CFP relationship, McWilliams and Siegel (2001) argue that in equilibrium there should be no significant CSP–CFP relationship. Using the supply and demand theory of the firm, McWilliams and Siegel contend that managers’ main objective is to maximize shareholder wealth. Accordingly, managers will choose the level of the attributes (including CSP attributes) that maximize firm value given the demand for various attributes and the costs of supplying them. Thus, in equilibrium, there is a neutral CSP–CFP relationship. Relatedly, Waddock and Graves (1997) argue that the environments in which firms and societies exist are so complex that a simple and direct link between CSP and CFP seems unlikely.

One of the bases for predicting a negative CSP–CFP relationship derives from the neoclassical theory of the firm. According to this view, the opportunity cost of expenditures for social performance exceeds the profitability of such investment, so that a tradeoff exists between CSP and CFP. Consequently, where stakeholders and others exert effective pressure for social performance, we should observe diminished financial performance and firm value. In addition, Preston and O’Bannon (1997) argue that, in spite of the tradeoff between CSP and CFP, managers may undertake socially responsible investment for their own private benefit (e.g., public acclaim) at shareholders’ expense. By their “managerial opportunism” argument, managers may seek to divert attention from poor financial performance by promoting their firms’ social performance. In contrast, managers may reduce expenditures on CSP programs in order to boost short-term profitability and, hence, their personal compensation. In either case, CSP would provide a signal to investors that management is prone to acting for its own private benefit. If so, we should see lower stock prices as investors come to expect that managers of socially responsible firms will act in a variety of ways detrimental to shareholders.

Lastly, there are several theoretical and conceptual bases leading to hypotheses for a positive CSP–CFP relationship. Waddock and Graves (1997), for example, explain a positive CSP–CFP relationship using a costs and potential benefits argument. They argue that the actual costs of CSP are minimal when they are compared to the potential benefits to the firm. Another explanation for the positive relationship between CSP and CFP is a “slack resources” argument where the firms that are financially superior and successful than their counterparts have slack resources that can be allocated to social performance (Waddock and Graves 1997; Preston and O’Bannon 1997).

To date, however, the most widely acknowledged theoretical support for a positive CSP–CFP relationship derives from Stakeholder Theory, which asserts that an organization’s success depends on its ability to meet the expectations and demands of all groups that have a stake in its activities (Freeman 1984, 1994; van Beurden and Gossling 2008). These stakeholders typically include customers, employees, and suppliers, as well as stockholders, and may also extend to the communities, societies, and natural environments which are impacted by the organization. In Stakeholder Theory firms that strive merely to satisfy stockholders while neglecting the interests of its broader set of stakeholder groups will ultimately fail; only by actively managing all key stakeholder relationships will firms achieve financial performance acceptable to stockholders. Consequently, Stakeholder Theory implies that corporations that seek better social performance will also enjoy better financial performance.

While it is true that organizations must manage well their relationships with all key stakeholder groups, the expectations and demands of customers, suppliers, employees, stockholders, and creditors and community, environmental and other groups are frequently in conflict. In order to reconcile tradeoffs among stakeholders’ competing claims, Jensen (2002) argues that managers must have a mechanism by which they can judge what is better and what is worse for the organization and that Stakeholder Theory alone does not provide a principled basis by which managers can make such judgments. Moreover, Stakeholder Theory provides no basis for holding managers accountable for their stewardship of corporate resources, enabling them to fund projects that may yield private benefits for managers (e.g., public recognition) but that impose costs on the corporation. Therefore, Jensen (2002) proposes augmenting Stakeholder Theory with the provision that corporate expenditures toward improved social performance should increase the firm’s long-term market value. In what he calls Enlightened Stakeholder Theory, managers find a needed guide for weighing stakeholders’ competing claims that is independent of personal preferences, while stockholders and others external to the firm have a foundation for evaluating the expenditures of corporate resources.

In summary, there are numerous competing theoretical bases regarding the nature of the CSP–CFP relationship. On balance, we find ourselves aligned with the theoretical arguments of Stakeholder and Enlightened Stakeholder theorist. This view is strengthened by the considerable empirical evidence of a positive CSP–CFP relationship (e.g., see van Beurden and Gossling 2008). Thus, our first hypothesis is:

Hypothesis 1

There is a positive relationship between corporate social performance and corporate financial performance.

Companies face varying industry contexts and stakeholder expectations for social performance. Research has shown that studies that fail to account for industry effects are likely to produce confounded results (Griffin and Mahon 1997; Waddock and Graves 1997; Hillman and Keim 2001). In their recent review article, van Beurden and Gossling (2008: Table I) report that in 10 of 34 studies either industry or an industry-specific characteristic is a confounding variable. Griffin and Mahon’s (1997) criticism of studies which investigate the CSP–CFP link is that many of them consider samples comprised of firms from multiple industries. The mode of social performance a firm chooses to pursue is a unique response to the particular nature of its stakeholder demands conditioned by its inherited capabilities and industry environment. Because different industries face different stakeholder portfolios with different activity levels in different areas, the CSP–CFP relation must be considered in specific industry settings (Griffin and Mahon 1997; Rowley and Berman 2000; Simpson and Kohers 2002). Thus, the particular avenues to social performance that a firm chooses are a unique response to stakeholder demands that are conditioned by the firm’s capabilities and industry environment. This leads to our second hypothesis:

Hypothesis 2

The relationship between corporate social performance and corporate financial performance varies across industries.

As noted earlier, CSP is a multidimensional construct. Each dimension represents a unique opportunity for social performance. If these dimension ratings are reflected in the investment decisions of stock market investors, they should be systematically related to stock prices. Because it is difficult, if not impossible, to observe motivations of managers, it is not possible to predict a priori the industries for which dimension ratings are associated with higher or lower CFP. However, given the multidimensional nature of CSP and the importance of industry differences, we can and do expect a measurable overall dimension–industry effect to emerge in the CSP–CFP relationship. Thus, our third and final hypothesis is:

Hypothesis 3

The relationship between the dimensions of corporate social performance and corporate financial performance varies across industries.

Data and Empirical Methods

Data

We examine our research questions and hypotheses by means of existing and well established empirical models estimated using a specially constructed cross-sectional time-series (panel) dataset. In short, we empirically test the extent to which the systematic variation in a market-based measure of firm value (i.e., stock price) is associated with an indicator of CSP. As discussed in §2, unlike accounting-based measures of CFP, common stock price incorporates all relevant information about a firm’s future prospects—including the long-run impact of CSP on stakeholder relationships. By accounting for the fundamental or intrinsic value of each firm’s common stock price, we assess how much additional variation in observed stock price can be attributed to CSP. In the remainder of this section, we detail the sources of data and computation of measures used in this research.

Our panel dataset was assembled using secondary data from three market information providers—Value Line, Inc., S&P and KLD Research & Analytics, Inc. Central to our empirical models is the calculation of the intrinsic value of a firm’s common stock price (Penman and Sougiannis 1998). Simply put, for each time period we calculate each firm’s common stock intrinsic value as a function of common equity book value plus the present value of anticipated future earnings in excess of the firm’s equity capital opportunity cost. We calculate intrinsic value using forecasts of earnings and stock price obtained from Value Line, Inc. Value Line has produced independent research on public firms for over 75 years. Their research staff consists of approximately 80 experienced professional security analysts who possess deep knowledge of and experience with the firms they monitor. Because analysts typically cover a relatively small number of firms, they develop deep knowledge and experience with these firms and their industries. Consequently, they are aware of planned spending for research and development, advertising, and capital investment and of corporate initiatives on the environment, product quality, corporate governance, and other aspects of social responsibility. This is important for our research because Value Line analysts incorporate these factors in their forecasts, which we use to compute intrinsic value. A compilation of annual Value Line research data is available as an intellectual property product known as Value Line Earnings and Projections (hereafter, Value Line data).

Equally important to the development of our models is an indicator firms’ CSP. We use a firm’s inclusion in the Domini 400 Social Index (DS400) as recognition of the firm’s positive social performance.3 The DS400 was created in 1990 by KLD Research and Analytics (KLD) as a social investment benchmark. The DS400 is the first common stock index for which membership is based on environmental, social and governance criteria, and it is designed to reflect the general market performance of the average socially responsible investor. KLD’s research analysts specialize in industry-specific issues involving environment, community relations, employee programs and diversity, product safety and accessibility, labor relations, human rights, and governance. All firms in the KLD universe undergo annual review which involves an analysis of public documents—both financial and non-financial—that may provide information or insights regarding a firm’s socially responsibility behaviors. A compilation of annual KLD research data is available as an intellectual property product known as KLD Statistical Tool for Analyzing Trends in Social & Environmental Performance (hereafter, KLD data).

Lastly, we obtain the balance of the financial measures used in this research from a secondary source of market data available from S&P Compustat (Compustat). Since 1962, Compustat has been a leading provider in financial, statistical and market information products. Compustat’s information products are targeted toward investors and analysts involved in all aspects of the capital markets. The compilation of standardized market data used in this research is an intellectual property product known as Compustat North American Data (hereafter, Compustat data).

An initial panel for years 2001 through 2008 was obtained by merging the Value Line, KLD, and Compustat data for firms with December fiscal years. Our final reconciled panel consists of 1,153 unique firms with 5,073 firm-year observations. Table 1 shows the sample distribution by year and by economic sector. Our sample is roughly evenly distributed over the years 2001 through 2008, a period which covers a range of stock market conditions. Table 1 also shows the sample distribution for firms classified according to S&P Global Industrial Classification (GIC) Table 2.
Table 1

Industry composition of sample

 

DS400

Non-DS400

Total

 

DS400

Non-DS400

Total

Panel A—Distribution by Industry Classification

Panel B—Distribution by Year

Energy

234

82

316

2001

253

181

434

Materials

274

114

388

2002

267

189

456

Industrials

533

302

835

2003

433

179

612

Consumer Discretionary

779

326

1,105

2004

489

200

689

Consumer staples

125

105

230

2005

481

206

687

Health care

442

203

645

2006

476

204

680

Financials

405

270

675

2007

526

206

732

Information Technology

610

145

755

2008

574

209

783

Telecom services

28

14

42

    

Utilities

69

13

82

    

Totals

3,499

1,574

5,073

 

3,499

1,574

5,073

Industry classification is based on Compustat GIC at the two-digit level. The sample consists of 5,073 firm-year observations from years 2001 to 2008. There are 1,153 unique firms

Table 2

Summary statistics

 

Mean

Median

Std Dev

Max

Min

Overall sample (N = 5,073)

     

 Current stock price

35.71

29.12

42.89

906.00

1.80

 Intrinsic value

35.56

28.59

43.94

1,064.07

1.10

 Market value ($ millions)

11,147

2,978

29,740

486,720

81

 Sales ($ millions)

7,559

2,039

21,014

358,600

8

 Return on invested capital

17%

15%

16%

507%

−41%

 Total debt to total capital

39%

33%

41%

476%

0%

Non-DS400 observations (N = 1,1574)

 Current stock price

40.33

32.96

54.40

906.00

1.80

 Intrinsic value

41.09

32.92

58.51

1,064.07

1.10

 Market value ($ millions)

15,728

6,424

27,830

199,525

81

 Sales ($ millions)

8,478

4,121

12,805

116,353

91

 Return on invested capital

20%

17%

20%

507%

−31%

 Total debt to total capital

43%

35%

46%

476%

0%

DS400 observations (N = 3,499)

 Current stock price

33.64

27.41

36.37

820.50

4.55

 Intrinsic value

33.07

26.81

35.20

868.35

1.76

 Market value ($ millions)

9,085

2,235

30,340

486,720

125

 Sales ($ millions)

7,147

1,511

23,780

358,600

8

 Return on invested capital

16%

14%

14%

342%

−41%

 Total debt to total capital

38%

32%

39%

474%

0%

Predictor Variables

Corporate Social Performance

The primary focus of this research is to examine the relation between CSP and a market-based measure of firm performance—stock price. In the test of H1, we use a firm’s inclusion in the DS400 as a proxy for an overall favorable indicator of CSP. The DS400 was created in 1990 by KLD Research and Analytics “as a social investment benchmark—a basket of firms that defines the market for social investors” and is designed to reflect the general market behavior of stocks of the average socially responsible investor.4 The DS400 is a capitalization-weighted index representing 400 common stocks selected from the 3,000 largest U.S. corporations, and it is the first index for which membership is based on both financial and environmental, social, and governance criteria. Firms passing financial screens for market capitalization, earnings, liquidity, stock price, and debt-to-equity ratio, are selected for inclusion in the DS400 based on their social and environmental records. KLD evaluates firms’ social performance along seven dimensions: community relations, diversity, employee relations, human rights, product quality and safety, environmental governance, and corporate governance.

In a follow-up analysis, we explore more deeply the relationship between the seven dimensions of CSP and CFP. The seven CSP dimensions outlined above are further decomposed by KLD into between 8 and 12 sub-dimensions with each sub-dimension being classified as either a “strength” or a “concern”. In their annual review of a firm, KLD analysts assess each firm’s performance for each sub-dimension (be it strength or concern) using a binary scoring scheme. A rating of 1 indicates the analysts’ recognition of the existence of the behavior identified with the sub-dimension. A rating of 0 indicates the analysts’ recognition of the absence of the behavior identified with the sub-dimension. To assess a firm’s overall performance for a CSP dimension, we calculate a Z-score for the difference between the sum of strengths and the sum of concerns.

Intrinsic Value

Penman and Sougiannis (1998) demonstrate the equivalence between the intrinsic value of a firm’s common stock and the current book value of its common equity plus the present value of future earnings in excess of the opportunity cost of equity capital. This excess income is known as residual income and is the basis for the Residual Income Valuation Model (RIM). Empirical research has shown the RIM to have exceptionally good explanatory power for stock price with R2s exceeding .90 (e.g., see Abarbanell and Bernard 2000; Courteau et al. 2001). Moreover, a firm’s intrinsic value (as derived from the market-based RIM) is largely free of the previously discussed issues confounding accounting data. We compute intrinsic value according to the following equation:
$$ V_{t} = B_{t} + \sum\limits_{\tau = 1}^{T} {R^{ - \tau } E_{t} \left[ {x_{t + \tau }^{r} } \right] + R^{ - T} E_{t} \left[ {V_{t + T} - B_{t + T} } \right]} $$
where Vt denotes the intrinsic value of a share of common stock at time t; Bt denotes the book value of common equity; R denotes one plus the required return on equity (r); Et is an expectation operator conditioned on information available at time t, xt denotes net income. Residual income equals income in excess of the opportunity cost of beginning-of-period equity capital: \( x_{t}^{r} = x_{t} - rB_{t - 1} \). In this research, we compute intrinsic value with T = 4-year horizon for each firm in our sample.
The required return on equity (ri) is estimated from the Fama–French (1992) 3-factor model in which required return depends on market beta, firm size and the ratio of the book value of common equity to the market value of common equity. Market beta captures a stock’s sensitivity to movements in the market portfolio, and size and book-to-market equity are thought to capture additional sources of non-diversifiable risk.5 Required return is calculated according to the following:
$$ r_{\rm{i}}\,{=}\,r_{\rm{f}} { + }\left( {r_{\rm{M}}{ - }r_{\rm{f}} } \right){\beta }_{{\rm{i,M}}} {\rm{ + SMB \times \beta }}_{{\rm{i,SMB}}} \rm{ + HML \times \beta }_{{\rm{i,HML}}} $$
where rf denotes the risk-free return, operationalized as the 1-year Treasury bill return. The market return rM is the return of the value-weighted portfolio of all NYSE, AMEX, and NASDAQ stocks. The market risk premium, rM − rf, is the expected excess return of stocks over T-bills and reflects the expected compensation for bearing average market risk. SMB denotes the expected return of a portfolio that is long small stocks and short large stocks. It reflects the expected compensation for bearing non-diversifiable risk associated with firm size. HML denotes the expected return of a portfolio that is long high book-to-market equity stocks and short low book-to-market equity stocks, and it reflects the expected compensation for bearing non-diversifiable risk associated with book-to-market equity. Monthly T-bill, market and factor portfolio returns from July 1926 to the present are available from Professor Kenneth French’s website.6 These data are used to estimate factor risk premia and betas for each stock in the sample. Annual T-bill yields are obtained from the US Federal Reserve Weekly Statistical Release H15. Refer to the Appendix for the complete derivation of the Intrinsic Value measure.

Control Variables

Previous research indicates that the control variables (e.g., industry, firm risk, size, etc.) used in multivariate analysis could play an important role in the assessment of the CSP–CFP relationship (Ullmann 1985; Waddock and Graves 1997; McWilliams and Siegel 2001). Further, the type of control variables that studies use could account for some of the variance in the CSP–CFP relation (Allouche and Laroche 2005). Thus, in order to more accurately assess the extent to which remaining variation in observed stock price can be attributed to firms’ social performance, we control for as much additional systematic variation in stock price as possible. Thus, we include three control variables in our model—firm size, operating profitability, and the use of debt in the capital structure. All control variables are calculated using Compustat data for the most recently reported fiscal year.

Researchers such as Fama and French (1992), find that after controlling for beta in the CAPM, common stock returns are decreasing in firm size. That is, small-firm equities have historically produced above average returns when compared with large-firm equities. Thus, firm size may be a proxy for dimensions of risk not adequately captured by the CAPM. If large stocks are less risky than small stocks, then expected return apart from that predicted by the CAPM should be decreasing in firm size. Thus, we include a commonly used proxy for size—annual sales—as a control variable for firm size in our model (Al-Khazali and Zoubi 2005; Ruf et al. 2001).

Previous research also shows that Value Line forecasts may tend to be optimistic (Klein 1990; Ramnath et al. 2005). These results suggest that Value Line forecasts tend to be optimistic for firms that are currently unprofitable vis-à-vis other firms. Thus, we include return on invested capital as a control for firm profitability in our model.

Lastly, because required equity returns have been shown to increase with financial leverage (Brav et al. 2005), we also include a control for the amount of debt in the capital structure which we measure as the debt ratio.7

Empirical Methods

To test our propositions regarding the impact of socially responsible behavior on stock price we estimate two LMMs8 using firm-level panel data covering years 2001 through 2008. In a LMM framework, both fixed and random effects are accommodated. The “fixed” effects of the LMM are analogous to linear predictors from standard OLS. The “random” effects are assumed to be distributed according to an empirically derived probability density function (Kreft and de Leweuw 1998).

In the present case, the principal benefit of LMM is its capacity to account for variance among firms’ initial conditions (intercepts) within the framework of the fixed parameter estimates (Wallace and Green 2002). Thus, LMM allows us to account for variation in firms’ initial conditions without requiring estimation of a large number firm-level fixed effects thereby allowing the inclusion of time-invariant variables of interest such as industry membership.9 Moreover, the LMM allows fine grain control over the structure of the model’s error variance–covariance matrix.10 Following error variance–covariance selection strategies advocated by several authors (e.g., see Singer and Willett 2002; Wallace and Green 2002; Littell et al. 2006), we estimate our model using an AR(1) within-subject error structure while allowing for panel-level heteroskedasticity at the between-subject level. In addition, we estimate standard errors for the model coefficients using Huber–White (robust) standard errors. Equation 1 provides the basis on which we investigate the overall CSP–CFP relationship (H1) as well as the presence of industry effects (H2):
$$ \begin{gathered} P_{it} = \pi_{0i} + \beta_{1} CSP_{it} +\beta_{2} IV_{it} + \beta_{3} ROIC_{it} + \beta_{4} DEBT_{it} +\beta_{5} SALES_{it} + \beta_{6 - 12} YEAR_{t} \hfill \\ +\beta_{{13 - 70}} GIC_{i} + \beta_{71 - 128} CSP_{it} \times GIC_{i} + \varepsilon_{it} \hfill \\ \end{gathered} $$
(1)
where \( \varepsilon_{it} \sim N\left( {0,\sigma^{2} } \right) \), \( \pi_{0i} = \beta_{0} + u_{0i} \) and \( u_{0i} \sim N\left( {0,\tau_{00} } \right) \), and i = 1..1,153 firms and t = the years 2001..2008.

The dependent variable (Pit) is the observed stock price for firm i in year t. The predictor variable CSPit is a dichotomous indicator of recognized positive CSP for firm i in year t. CSPit takes on the value one for those firms included in the DS400 during year t and zero otherwise. The predictor variable IVit is the calculated intrinsic value for the stock price of firm i in year t. The control variables ROICit, DEBTit, and SALESit represent return on invested capital, debt ratio, and annual sales, respectively. To control for shared market conditions experienced by all firms during the years covered by our panel, we include year-based fixed effects (YEARt). The set of control variables GICi represents the time-invariant polychotomous industry classification reflecting the S&P GIC of firm i. The variables CSPit × GICi capture the interaction of firm CSP and industry classification. The intercept in Eq. 1 (π0i) reflects the firm-level random effect, which is assumed normally distributed with mean zero and variance τ00. The remaining coefficients are population average estimates and are analogous to linear predictors from standard OLS.

We describe above the multidimensional aspects of CSP, and hypothesis H3 states that the importance of CSP dimensions for CFP varies by industry. To examine the effects of CSP dimensions on CFP by industry, we re-conceptualize our dichotomous measure of CSP as firm performance along the seven dimensions of CSP as held by KLD. These dimensions are community relations, diversity, employee relations, human rights, product quality and safety, and environmental and corporate governance. Equation 2 provides the basis for our test of H3:
$$ \begin{gathered} P_{it} = \pi_{0i} + \beta_{1 - 7} \left\{ {DIM_{dit} } \right\} + \beta_{8} IV_{it} + \beta_{9} ROIC_{it} + \beta_{10} DEBT_{it} + \beta_{11} SALES_{it} + \beta_{12 - 18} YEAR_{t} + \hfill \\ \beta_{19 - 35} GIC_{i} + \beta_{36 - 153} GIC_{i} \times \left\{ {DIM_{dit} } \right\} + \varepsilon_{it} \hfill \\ \end{gathered} $$
(2)
where \( \varepsilon_{it} \sim N\left( {0,\sigma^{2} } \right) \), \( \pi_{0i} = \beta_{0} + u_{0i} \) and \( u_{0i} \sim N\left( {0,\tau_{00} } \right) \), and i = 1..1153 firms, t = the years 2001..2008 and DIMd = {community relations, diversity, employee relations, human rights, product quality and safety, and environmental and corporate governance}.

The variable DIMdit represents CSP dimension Z-score (see §4) for dimension d of firm i in year t. Given the large number of industry–dimension interactions in Eq. 2, we impose two additional constraints on our data. First, we restrict our analysis of industry–dimension relationships to those industries exhibiting a significant CSP–industry interaction in the estimation of Eq. 1. Second, we restrict our analysis to GIC strata containing five or more firms. All industries not meeting both of the above criteria are absorbed into the intercept for the estimation of Eq. 2. Descriptions for all other variables remain unchanged from Eq. 1.

To facilitate interpretation of results, we implement several important variable transformations. Due to a moderate degree of skewness in the data, we apply a natural log transformation to continuous model variables P, IV, and SALES. We re-code all polychotomous categorical variables—YEAR, GIC, CSP, and CSP × GIC—using a dummy variable effects-coding scheme in lieu of customary reference coding scheme. For reference coded dummy variables, the βs represent deviations from the selected “reference group” effect. While this approach is tenable for ordinal data and for data with a natural reference group, in our data no such group exists. More importantly, we are interested in an overallaverage effect of CSP on CFP, not the effect relative to a referent group. For effects-coded variables, each \( \beta \) indicates the difference between the effect group and the unweighted average of all groups (Hardy 1993). Thus, effects-coding allows interpretation of dummy variable coefficients as deviations from the unweighted mean of the dependent variable. Second, we grand-mean center the previously transformed continuous variables. Lastly, for Eq. 1 our dichotomous indicator of social performance (CSP) is reference coded using zero as the reference group. It is important to note that the last two transformations play an important role in the interpretability of our results. By centering all continuous variables and reference coding, the indicator of social performance (CSP), the intercept represents the average log stock price for firms where CSP is zero (Sweeney and Ulveling 1972). Thus, it follows that the coefficient for the reference-coded CSP dummy is interpretable as the change in average log stock-price for socially responsible firms.

In Eq. 1, the primary variable of interest (CSP) is a dichotomous indicator of CSP and is operationalized as a firm’s inclusion in the Domini 400 Social Index; CSP equals 1 for firms in the DS400 and zero otherwise. Consistent with H1, we expect the coefficient on CSP to be positive and significant. Consistent with H2, we expect (1) the inclusion of CSPGIC interactions to significantly improve model performance and (2) the relative importance of CSP to vary among industries. IV denotes the intrinsic value of a share of common stock, which we expect to be a positive and highly significant predictor of stock price. We expect the coefficients on profitability (ROIC) and size (SALES) to be positive and significant, and we expect the coefficient on the debt ratio (DEBT) to be negative and significant. Consistent with H3, we expect (1) that the collective of CSP dimension–industry interactions featured in Eq. 2 will significantly improve model performance and (2) the relative importance of these interactions to vary across industries.

Estimation and Results

Estimation results for Eq. 1 are reported in Table 3.11 The coefficient on CSP (β1) captures the average effect of CSP on our measure of CFP. By hypothesis H1, we expect CSP to have a positive and statistically significant relationship with stock price. In conflict with H1, however, we find that CSP is negatively associated with stock price (β1 = −.0290, p < .01). This result suggests that good CSP (as reflected in membership in the DS400 index) reduces stock price by 2.85% (100 − [100 * e−.0290]) on average. Coefficient estimates for the continuous variables in our model are statistically significant with the expected signs. Intrinsic value (IV) is a highly significant predictor of stock price (β2 = .7110, p < .001). Both ROIC and SALES are positively and significantly associated with stock price (β3 = .2134, p < .01; β5 = .0380, p < .001), and DEBT is negatively and significantly associated with stock price (β4 = −.0454, p < .01).
Table 3

LMM estimation results

Variable(1)

Coef(1)

Intercept

3.3352***

CSP

−.0290*

IV

.7110***

ROIC

.2134***

DEBT

−.0454**

SALES

.0380***

CSP × Oil, Gas & Consmbl Fuels

.1730*

CSP × Chem

.0022

CSP × Containers & Pkg

−.1824*

CSP × Metals & Mining

−.1763**

CSP × Paper & Forest Prod

−.0388

CSP × Bldg Prods

−.1669+

CSP × Constr & Engr

−.0598

CSP × Electrical Equip

−.0803

CSP × Industl Conglomerates

.2919***

CSP × Machinery

−.0169

CSP × Trading Comps & Dist

.1568

CSP × Coml Serv & Supplies

.0567

CSP × Profnl Serv

−.2561***

CSP × Air Frt & Logistics

−.0176

CSP × Airlines

.0110

CSP × Rd & Rail

.1514**

Variable(2)

Coef(2)

CSP × Auto Components

.1368

CSP × Autos

−.1400

CSP × Household Durables

−.1073

CSP × Leisure Equip & Prods

−.0585

CSP × Textiles, Apprl & Lxry Gds

.1523**

CSP × Hotels, Rstrnt & Leisure

−.0247

CSP × Diversified Consmr Serv

−.2861*

CSP × Media

.0765

CSP × Distributors

.1751***

CSP × Multiline Retail

−.2072*

CSP × Specialty Retail

−.1085**

CSP × Food & Staples Retailing

−.0285

CSP × Beverages

.0316

CSP × Food Products

.0817

CSP × Household Prods

.1225

CSP × Personal Prods

.0774

CSP × Hlth Care Equip & Supls

.1005**

CSP × Hlth Care Providers & Serv

.0188

CSP x Hlth Care Tech

−0.0642*

CSP x Biotech

0.0243

CSP x Pharma

0.0477

CSP x Life Sciences Tools & Serv

0.1302

Variable(3)

Coef(3)

CSP × Coml Banks

−.0266

CSP × Thrifts & Mortg Financ

.2197**

CSP × Diversified Financ Serv

.1872+

CSP × Capital Mkts

.1171

CSP × Insurance

.0503

CSP × R/E Mgt & Development

−.3761**

CSP  × Internet S/W & Serv

−.0883+

CSP × IT Serv

−.1209

CSP × S/W

.2061+

CSP × Comm Equip

.0218

CSP × Cmptr & Periph

.0159

CSP × Electr Equip, Instr & Comps

.1725**

CSP × Office Electrs

−.4222***

CSP × Semicond & Semicond Equip

−.1098

CSP × Diversified Telecomm Serv

.4031*

CSP × Wireless Telecomm Serv

−.2872+

CSP × Electric Util

.1690***

CSP × Gas Util

−.1030*

CSP × Multi-Util

−.0568

CSP × Indep Power Prodcr & Enrgy Trdrs

.0660

−2 Residual log likelihood

−345.7

Akaike’s information criterion (AIC)

−81.7

Psuedo-R2

.8657

τ00 (random effect variance)

.0129***

p < .05, ** p < .01, *** p < .001. Number of observations = 5,073 from 1,153 unique firms. Dependent variable is the natural log of stock price in U.S. dollars. YEAR, GIC, and CSP–GIC interaction dummy variables are effects-coded. IV and SALES are natural log transformed and grand-mean centered. ROIC and DEBT are grand-mean centered. YEAR dummies and main effects for industry are omitted in the interest of conciseness. Full model results are shown the Appendix

Turning our attention to the random component of the model, not surprisingly, we note that our data exhibit significant variation across firms’ initial conditions. The estimated variance component (τ00) of the random effect (π0i) is significantly different from zero (p < .001). This result confirms our expectation that firms differ significantly in average initial stock prices.

By hypothesis H2, we expect significant cross-industry variation in the CSP–CFP relationship. As can be seen in Table 3, we observe cross-industry variation on the CSP–industry interactions from the estimation of Eq. 1. As a formal test of H2; however, we expect the following two conditions to hold: (1) the combined explanatory power of the industry interactions will be statistically significant, (2) CSP–industry interaction coefficients will exhibit significant variability across industries. As a test of condition 1, we carry out a likelihood ratio test of the significance of CSP–industry interaction coefficients β71–128. Consistent with expectations, results confirm that this is the case (Χ2(58) = 80.2, p < .05). As a test of condition 2, we construct a linear contrast to test the null hypothesis that β71–128 = β, i.e., that the CSP–industry interaction coefficients are constant (Belsley 1973). Consistent with H2, test results support condition 2 as the contrast F-statistic rejects the null hypothesis (F(57,3850) = 12989; p < .001). Hence, we find strong support in favor of H2 that significant cross-industry variation exists in the CSP–CFP relationship.

Recall that hypothesis H3 states that the importance of CSP dimensions on CFP will vary by industry. Table 4 summarizes the significant CSP dimensions by industry. From Table 4, we observe the appearance of cross-industry variation for each of the seven CSP dimensions. As a formal test of H3, we expect the following three conditions to hold: (1) the combined explanatory power of the CSP dimension–industry interactions will be statistically significant; (2) the interaction coefficients will exhibit significant variability across all CSP dimensions and across all industries, and (3) within each CSP dimension, industry interaction coefficients will exhibit significant cross-industry variability.
Table 4

Summary of significant CSP dimensions by industry group

Industries

CSP dimensions

Corporate governance

Community

Diversity

Employee relations

Environment

Human rights

Products

101020:

Oil, Gas & Consumable Fuels

   

.1577*

.1432**

.1119**

 

151030:

Containers & Packaging

−.1455***

 

.1922**

    

201050:

Industrial Conglomerates

.0900+

.1831**

 

.3496*

   

202020:

Professional Srvcs

−.0888+

−.1824**

−.1954**

    

203040:

Road & Rail

 

.2749*

.2885*

.1477+

   

252030:

Textiles, Apparel & Luxury Goods

     

−.1485*

 

253020:

Diversified Consumer Srvcs

.1373**

.1604+

     

255030:

Multiline Retail

 

.1874+

−.2038*

−.1629+

  

−.1471+

351030:

Health Care Technology

−.0512+

 

.1326+

    

401020:

Thrifts & Mortgage Finance

 

−.1395+

−.2303**

   

.0783**

501010:

Diversified Telecommunication Srvcs

.3413***

−.7015**

−.0859**

 

−1.1255**

−1.2307**

−.1555**

551010:

Electric Utilities

     

.1874+

.1945*

551020:

Gas Utilities

−.1179**

−.3097**

−.4772**

 

.6098***

.4879***

.1441***

F-statistic for test of significant cross-industry variation by CSP dimension

2.99***

5.33***

8.57***

1.72*

4.02***

3.79***

3.58***

Values are coefficient estimates. + p < .10, * p < .05, ** p < .01, *** p < .001. Results shown for GIC strata containing 10 or more firms. Positive relationships are show in shaded cells. Negative relationships are show in un-shaded cells.  Degrees of freedom for all contrast F-statistics are (15, 3782)

As a test of condition 1, we carry out a likelihood ratio test of the significance of coefficients β36–153. Consistent with expectations, results confirm that this is the case (Χ2 (135) = 203.3, p < .01). As a test of condition 2, we construct a single linear contrast to test the null hypothesis that β36–153 = β, i.e., that the GICDIM interaction coefficients are constant. Again consistent with expectations, test results support condition 2 as the contrast F-statistic rejects the null hypothesis (F(104,3782) = 36.96; p < .001). Lastly, as a test of condition 3, we construct seven linear contrasts (one for each CSP dimension) to test the cross-industry variability within each of the CSP dimensions. Referring to the last row of Table 4, all test statistics reject the null hypothesis of constant GICDIM interaction coefficients in favor of the alternative hypothesis of variability. Overall, we interpret the above results as strong support of H3. Complete estimation results for Eqs. 1 and 2 are shown in the Appendix.

Discussion of Results

Applying innovative empirical models and methods to a uniquely constructed dataset, we show that a measurable and statistically significant relationship exists between corporate social and financial performance. Moreover, our results show that this relationship reflects firms’ unique capabilities within their specific industry settings Relying on Stakeholder Theory and substantial empirical evidence, we hypothesized a positive relationship between CSP and CFP. In this view, a firm’s stock price will rise if it successfully meets the expectations of its stakeholders. We therefore expected firms with acknowledged social performance would have stock prices greater than firms that did not. In conflict with H1, we find that CSP is associated with a stock price that is 2.85% lower than the unweighted average stock price.

One explanation for this contrary finding may be found in the context of Preston and O’Bannon (1997) managerial opportunism argument discussed in §3. Jensen (2002) argues that Stakeholder Theory provides no principled mechanism for weighing tradeoffs among stakeholders’ competing claims, which leaves managers free to justify social performance expenditures that satisfy their own personal agendas or for their private benefit. Consistent with this view is a study by Barnea and Rubin (2006) showed that managers of firms rated highly on social performance metrics held smaller portions of their firms’ shares and thus bore relatively little of the cost of CSP-related activities.

A different construal of how managerial opportunism might manifest as below average stock prices can be explained in terms of risk. Merton (1987) presents a theory of capital market equilibrium with incomplete information in which expected returns increase with idiosyncratic risk. If the managers of firms that made expenditures for social performance did so for self-serving or opportunistic reasons, then markets might reasonably expect these managers to continue to such behavior into the future to the detriment of shareholders. In this way, managerial opportunism can be conceptualized as firm-level idiosyncratic risk. Firms with such risk would be expected to experience higher expected returns and lower stock prices. Several recent empirical studies are consistent with this interpretation. Both Kempf and Osthoff (2007) and Statman and Glushkov (2009) found that companies rated highly on social performance metrics had higher stock returns. If these higher returns can be attributed to risk (as is normally assumed), then socially responsible companies must be riskier than their conventional counterparts. Thus, all else equal, it is expected that risker firms have higher expected returns and lower stock prices.

In support of hypothesis H2, we confirm the existence of significant industry effects in the CSP–CFP relationship via tests of overall significance and variation of the CSP–industry interaction effects in Eq. 1. Our results show that for 20 of 58 industries the effect of the CSP–industry interaction is statistically significant—10 positive and 10 negative. For example, the Oil, Gas and Consumable Fuels industry interaction (CSP*101020) has a coefficient estimate of .1730 and is statistically significant, indicating that in this industry CSP is associated with higher stock prices, and increased shareholder wealth, of approximately 17%. Similar interpretations can be made regarding the 19 remaining significant CSP–industry interaction coefficients.

Under hypothesis H3, we explored the inter-industry variability of CSP dimensions on CFP. In support of H3, we found that the industry–CSP dimension interactions significantly enter our model and that the impact of these interactions on CFP varies across all CSP dimensions. Our results confirm that while CSP has a measurable effect on CFP, the nature of the relationship is variable across industries. Moreover, these results support the view that empirical studies that fail to account for industry effects are likely to produce confounded results (cf. van Beurden and Gossling 2008).

Considering the Industry–CSP dimension interactions, we note that every industry has unique characteristics with respect to management and stakeholder actions. Moreover, considering the complex nature of the industry–CSP dimension relationship coupled with scant theoretical support in this area, developing precise ex-ante expectations regarding the nature of the relationships across all (or any) industries and all CSP dimensions would require extensive industry-specific knowledge and is beyond the scope of this research. That said, in several instances we are able to draw some comparisons between our results and the results of previous research that examines the CSP–CFP relationship along specific dimensions of CSP.

Several studies examine the community relations dimension of CSP on financial performance. Barnett and Salomon (2006), Hillman and Keim (2001), Simpson and Kohers (2002), and Preston and O’Bannon (1997) show that firms that put effort into improving relationships with their local communities also improve their financial performance. Consistent with these findings, our results show that in several industries such as industrial conglomerates, road and rail, diversified consumer services, and multiline retail, firms that exhibit strong community relations also observe higher stock prices. Similarly, studies by Berman et al. (1999), Turban and Greening (1996), and Waddock and Graves (1997) find a positive financial impact from improved employee relations. Again, consistent with these findings, our results show that firms that have better employee relations also have higher stock prices in the oil, gas and consumable fuels, industrial conglomerates, and road and rail industries.

Several authors have investigated CSP–CFP relationship focusing on the environmental dimension of CSP. Feldman et al. (1997) found that firms that improve their environmental performance can raise their stock prices by as much as 5%. In a portfolio analysis approach using corporate eco-efficiency scores, Derwall et al. (2005) show that a high-eco-efficiency portfolio substantially outperforms its low-eco-efficiency counterpart. Other studies such as Dowell et al. (2000), Russo and Fouts (1997) and Konar and Cohen (2001) also suggest firms exhibiting superior performance on environmental criteria also experience better overall financial performance. Finally, Brammer et al. (2006) contend that in industries with high environmental impact one would expect that superior performance on the environmental dimension of CSP would play an important role in the CSP–CFP relationship. Consistent with the above, our results show that in industries with high environmental impact such as oil, gas and consumable fuels and gas utilities, firms demonstrating positive environmental performance experience stock prices well above average.

In summary, our analysis confirms that there are variable returns to CSP dimensions that are conditioned on industry membership. Table 4 highlights the significant industry–CSP dimension interactions results from the estimation of Eq. 2. These results provide valuable information and actionable insights to decision makers considering whether and how to commit resources to social performance initiatives. For example, improved environmental performance in the oil, gas and consumable fuels industry is associated with a 14.4% increase in stock price. Making use of this knowledge, a firm in this industry might rank environmental performance improvement above other possible CSP initiatives when considering the allocation of scarce corporate resources. Likewise, similar tradeoff considerations can be made by firms in other industries regarding potentially advantageous CSP investments.

Conclusions

Our aim in this research is to shed new light on the empirical link between corporate social and financial performance via the application of empirical models and methods new to the CSP–CFP literature. Our research contributes to the empirical CSP–CSP literature in several important ways. First, we take the view that whether social performance is associated with better or worse financial performance depends on the motives driving an individual firm’s response to stakeholder demands. We admit the possibility that a firm’s response evolves from its specific industry context and is conditioned by its unique capabilities. Our research design therefore allows the CSP–CFP link to vary across the 58 industries in our sample. In order to quantify industry effects, we estimate LMMs which can accommodate both fixed and random effects thereby allowing for the estimation of time-invariant variables of interest such as industry and CSP–industry interactions.

Second, we measure financial performance in terms of stock price, which is the only financial metric that can capture the long-run impact of social performance on stakeholder relationships. For each firm in our sample, we compute the intrinsic value of its common stock based on residual income forecasts, which allows us to control a large proportion of the variation in stock price due to factors unrelated to social performance. We then examine the extent to which remaining variation in stock price can be attributed to measures of social performance.

Third, we operationalize a firm’s CSP in terms of its membership in the Domini Social 400 Stock Index. The DS400 was created by KLD Research and Analytics to serve as a social investment benchmark and index membership is determined via a rigorous evaluation of a firm’s activities on seven dimensions of social performance and thus an appropriate indicator of acknowledged social performance. In the initial phase of this research, we investigate the association between social performance and stock price. In the second phase, we investigate the association between seven dimensions of social performance and stock price. Lastly, because we are able to differentiate the effects of the CSP–CFP relationship among industry strata, we believe our results are useful to decision makers considering the allocation of corporate resources to social performance.

Footnotes
1

We acknowledge the distinction between the phrases “corporate social responsibility” and “corporate social performance,” however, for ease of exposition we will use CSP throughout this manuscript.

 
2

According to the FASB Statement of Financial Accounting Concepts Number 1, accounting information “largely reflects the financial effects of transactions and events that have already happened.”

 
3

The Domini 400 Social Index was renamed the FTSE KLD 400 Social Index in July 2009.

 
5

In addition to calculations of required return using the Fama–French 3-factor model, we also computed intrinsic value using required returns using Value Line-supplied betas in the Capital Asset Pricing Model (CAPM). We also estimated betas from regressions of industry portfolio returns on the value-weighted NYSE/AMEX/NASDAQ index and assigned these industry betas to individual stocks matching on SIC codes. All three computations of required return resulted in qualitatively equivalent results.

 
7

Debt ratio is the ratio of the sum of short- and long-term debt to the sum of short- and long-term debt, preferred stock, and common equity.

 
8

LMMs are also often referred to as random coefficient models.

 
9

One frequent lament of a traditional fixed effects model is the “differencing away” of time-invariant model parameters of interest (Wooldridge 2002). For example, firms rarely, if ever, are re-classified into a different industry, thus rendering estimation of the time-invariant industry effects impossible in a fixed effects model.

 
10

The ability to accurately depict the true nature of the error structure of one’s data is often cited as the raison d'être of the LMM (Singer 2002).

 
11

We estimate our LMMs using SAS/MIXED software, Version 9.1 of the SAS System for Windows. The SAS LMM estimation procedure is a generalization of the standard linear model in that the data are permitted to exhibit correlation and non-constant variability as random effects (Littell et al. 2006).

 
12

These returns are available from Professor Ken French’s website, for which we are grateful.

 

Copyright information

© Springer Science+Business Media B.V. 2011