企業社會責任與財務積效—臺灣的實證研究

全文

(1)社會科學論叢 2009年4月第三卷第一期57-120頁. 企業社會責任與財務積效—臺灣的實證研究 Corporate Social Responsibility and Financial Performance: Empirical Evidence from Taiwan. 淡江大學財務金融系助理教授張元 Yuan Chang Assistant professor of the Department of Banking and Finance, Tamkang University.

(2) 58 社會科學論叢 2009 年 4 月第三卷第一期. 企業社會責任與財務績效—臺灣的實證研究張元淡江大學財務金融系助理教授. 中文摘要我們以遠見雜誌於2006年公佈的企業社會責任調查為基礎，將衡量企業社會責任績效的三個層面─社區參與、環境保護與財務透明，透過投資組合分析與迴歸分析來評估公司承擔社會責任與財務績效之間的關係。實證結果發現，首先，平均來說，公司的社會責任評分愈高，在評比結果公布後的股票報酬愈低，兩者之間呈現負向關係；第二，公司在不同層面的社會責任行為對於股價報酬的影響具有差異性。第三、透過公司的長期財務資料發現，社會責任評分高的公司其長期平均會計績效相對較佳，但長期平均的股票報酬相對較低，表示社會責任型公司不一定是一個好的投資標的。最後，我們找不到支持社會責任行為可以做為公司績效保險的證據。本文使用各別層面的評比指標可更廣泛地評估社會責任行為對財務績效的影響，改進既有文獻中僅使用單一指標的不足之處。. 關鍵字：企業社會責任、財務績效（JEL分類代號：G10、G14、M14、M20）.

(3) Corporate Social Responsibility and Financial Performance: Empirical Evidence from Taiwan 59. I. Introduction The concept of Corporate Social Responsibility (CSR)1 is gaining more and more attention in both the business and academic fields. Based on the concept of CSR, Socially Responsible Investment (SRI) describes an investment strategy which combines the intentions to maximize both financial return of stockholders and social benefits of stakeholders and is prevalent in major financial markets.2 According to Scholtens (2005) and Schroder (2004), it is estimated that SRI represent 2 trillion and 2.5 billion Dollars in assets in U.S. and Asia, respectively. A key question for investors considering firm's engaging in CSR activities is whether these firms forfeit financial returns for the sake of their concern for the environment or stakeholders of a society. More detail, firms with CSR outperform firms without considering CSR? Are good firms good investments? This topic has been the subject of numerous studies but their conclusion is controversial. Theoretically, there are two conflicting views regarding the impact of engaging in CSR activities on the financial performance of a firm. One is the social impact hypothesis, proposed by. 1. According to World Business Council for Sustainable Development (WBCSD), CSR is defined as "the continuing commitment by business to contribute to economic development while improving the quality of life of the workforce and their families as well as of the community and society at large". While there exists other definitions of CSR, they are similar. For example, Frooman (1997) define CSR as "An action by a firm, which the firm chooses to take, that substantially affects an identifiable social stakeholder’s welfare". McWilliams and Siegel (2001) describe CSR as "actions that appear to further some social good, beyond the interest of the firm and that which is required by law". 2. SRI is an investment process that considers social and environmental consequences of investments within the context of rigorous financial analysis in order to encourage corporate social responsibility such as environmentally responsible, support workplace diversity, and increase product safety and quality. Some (not all) also avoid businesses involved in alcohol, tobacco, gambling, weapons and other military industries, and/or abortion..

(4) 60 社會科學論叢 2009 年 4 月第三卷第一期 Cornell and Shapiro (1987) and Preston and O'Bannon (1997) and the other is the shift of focus hypothesis proposed by Becchetti, Ciciretti and Hasan (2007), where the former suggests that CSR a has positive relation with firm performance and the latter the opposite. For the social impact hypothesis, Moskowitz (1972), Parket and Eibert (1975), and Soloman and Hansen (1985) firstly advocate that CSR garners more benefits than the cost incurred. Other studies also proposed views that CSR has a positive effect on performance through several channels such as providing a better working place improves employee productivity (Turban and Greening, 1997); donating to public benefits increases social reputation, trust (Bowman and Haire, 1975; Alexander and Bucholtz, 1978), brand image and product competitiveness (Porter and van der Linde, 1995; Fombrun, Gardberg and Barnett, 2000); getting favors with investment institutions (Spicer, 1978; Moussavi and Evans, 1986); obtaining better insurance protection of brand image and financial performance during economic downturns or specific negative events (Tsoutsoura, 2004; Werther and Chandler, 2005; Peloza, 2006). The shift of focus hypothesis state that most of the CSR activities such as employee and community relationships, environmental protection and corporate governance are involved with a shift of focus from the maximization of stockholders' value to concern of interests of a wider set of stakeholders and are cost increasing. Previous studies also argue that corporations engaged in CSR activities are found to have lower market competitiveness and worse performance through inefficient use of resources (Friedman, 1970), product development limitation (Bragdon and Marlin, 1972) and maintaining cost-pushing and non-profitable social activities (Vance, 1975; Aupperle, Carroll and Hatfield, 1985; Ullmann, 1985)..

(5) Corporate Social Responsibility and Financial Performance: Empirical Evidence from Taiwan 61. Particularly, public responsiveness of philanthropic behavior is apathetic and the feedback effect on financial performance is insignificant (Walley and Whitehead, 1994; Henderson, 2002). A significant body of empirical studies also reached mixed results. We compress existing empirical studies about CSR and financial performance into three categories, and the main difference is about the measurement of firm performance. The performance measures of the first kind of study is accounting-based, like ROA, ROE, for examples, supporters of social impact hypothesis, Cochran and Wood (1984), adopted Moskowitz's (1972) reputation index, which rates firms into outstanding, honorable mention and worst companies, as the proxy of CSR measurement. They found CSR ratings positively affect firms' return on assets and return on sales. Waddock and Graves (1997) employed another similar reputation rating developed by Kinder, Lydenberg and Domini (KLD) as the proxy of CSR3 and also found that past and current KLD ratings are positively related to firm's ROA. Tsoutsourz (2004) also employed KLD rating confirming the results with positive relationship between CSR and ROA, ROE and ROS. Griffin and Mahon (1997), Orlitzky, Schmidt and Rynes (2003), Guenster, Derwall, Bauer and Koedijk (2005), Aigner (2006), Nelling and Webb (2009) and Dam (2006) provided summary of this strand of studies. While the accounting-based measures have only an indirect impact on stockholders' wealth, the second kind of studies use market-based measures like stock returns as performance measure. Earlier studies, such as Vance (1975), employed Moskowitz's (1972) ratings and found that firms with better ratings have inferior stock returns. Newgren, Rasher, LaRoe and Szabo (1985) 3. KLD assessed the performance of multi-dimension stakeholder's concerns among S & P 500 companies..

(6) 62 社會科學論叢 2009 年 4 月第三卷第一期 found that those firms with environmental assessment get inferior stock market returns. Lately, with development of concept of SRI, Brammer, Brooks and Pavelin (2005a) proposed a theoretical view about CSR and stock returns. They argue that from the perspective of a firm, given the efficient markets hypothesis satisfied, equilibrium should develop whereby engaging in expenditure on socially responsible activities takes place up to the point where its marginal profitability is zero. Thus, the returns to socially responsible and irresponsible firms should be the same, for given levels of risk and other firm characteristics. At the portfolio level, however, if this argument concerning the neutrality of corporate responsibility for returns holds, then investors must be made unambiguously worse off by the social responsible screening-out process. Removing some stocks, sectors, or even whole countries on ethical grounds from the investable universe of securities will reduce portfolio efficiency. From the viewpoint of portfolio analysis, some studies make comparison of stock returns between SRI funds and NonSRI funds. 4 Guerard (1997a) collect stock returns of 1,300 socially screened and 950 socially unscreened firms during 1987 and 1994 and find that there is little significant difference between the performances of socially screened versus unscreened investments. Kahn et al. (1997) show that divesture of tobacco stocks from S & P 500 firms would have made little difference to typical investors returns since allocations to such stocks is usually very small. Following his previous work, Guerard (1997b) shows that investment screens to exclude 4. Brammer, Brooks and Pavelin (2006) state that if the research make comparison stock performance between SRI and NonSRI funds will suffer from identification problem, that’ s the performance difference should attribute to good firms or good funds managers..

(7) Corporate Social Responsibility and Financial Performance: Empirical Evidence from Taiwan 63. environmental or alcohol/tobacco/gambling or nuclear stocks actually yield higher average returns than unscreened investments. Using more recent data for 1990-98, Statman (2000) use data during 1990 to 1998 to examine the performance of the Domini Social Index (DSI) and S & P 500 Index. They find that DSI slightly outperformed the S&P Index in pure returns, but showed slight underperformed on a risk-adjusted basis. Shefrin and Statman (2003) and Anginer, Fisher and Statman (2008) using CSR ratings of over 600 U.S firms by Fortune, and find that firms with better ratings get lower stock returns. Most of earlier studies of the performance of SRI- versus NonSRI-funds did not control for differential levels of risk between them. However, Hamilton, Jo and Statman (1993) use the CAPM framework to examine the performance of 32 socially responsible mutual funds during 1981 to 1991. Their empirical results show only two significant alphas with one positive and the other negative and they conclude that the market does not price socially responsible characteristics. Bauer, Koedijk and Otten (2002) use multi-factor model to show that both German and US ethical funds underperform their benchmark but outperform similar UK funds. Barnett and Salomon (2002) use 28 years of data on 67 socially responsible funds and find a U-shaped relationship between performance and the strength of the social responsible screen. They argue that funds which employ minimal screens are still able to diversify well, while funds that employ very strict screens are able to filter out poor quality firms effectively; funds with intermediate-level screens are found to provide the weakest performance. Brammer, Brooks and Pavelin (2005b) examine performance of America's 100 Best Corporate Citizens surveyed by Business Ethics and find evidence of negative abnormal returns of around 3%, but with positive abnormal returns after controlling firm.

(8) 64 社會科學論叢 2009 年 4 月第三卷第一期 characteristics and risk factors. At the firm level, Feldman, Soyka and Ameer (1997) use environmental aspect of CSR and find evidence that firms who improve their environmental performance can get lower CAPM betas and higher stock prices. Antunovich, Laster and Mitnick (2000) use CSR rating by the Fortune and get results of firms with better ratings get higher stock returns. Anderson and Smith (2006) get similar results. Derwall, Gunster, Bauer and Koedijk (2004) employ data from the Innovest rating database of "eco-efficiency" scores thus only cover environmental issues of CSR for the period 1995-2003. They form two portfolios which are composed of the highest and lowest eco-efficiency scores. Under CAPM framework and using a multifactor model and incorporating industry effects, their findings support high-scoring portfolio significantly outperforms the low-ranking one. Brammer, Brooks and Pavelin (2005a) examined the relationship between stock returns and CSR which is proxied by the composite indicator constructed from three aspects of CSR activities, environmental protection, employee interests and community relationship. They found that scores on a composite indicator are significantly negatively related to stock returns. This negative relationship is explained by the negative relation between disaggregate measure of employee interests and financial performance but not by the other two measures.5 One important concern is that CSR is multi-dimensional, such as public donation, taking account of employee and consumers' benefits, maintaining working opportunity of underprivileged minority, environmental protection of. 5. The third method is the event studies about impact on short-run stock returns from emersion of engagement or contravention of CSR activities, such as Posnikoff (1997), Wright and Ferris (1997), Teoh, Welch and Wazzan (1999), Brammer, Brooks and Pavelin (2005) and Becchetti, Ciciretti and Hasan (2007)..

(9) Corporate Social Responsibility and Financial Performance: Empirical Evidence from Taiwan 65. community and ecology, corporate governance and accounting transparency, etc. Various aspects of CSR activities bear different degree of importance among different industries of firms. Porter and Kramer (2006) suggested that managers should put a new premium on specific area of CSR issues that are most beneficial to the core business of the firm. For example, the issue of environmental protection is not the most important concern for pharmaceutical firm. Instead, they can make efforts in development and deterrence of HIV proliferation in Africa. Brammer and Pavelin (2005) provide evidence to show that a firm will get most improvement of its reputation and financial performance when they engage into the specific aspects of CSR activities that meet the greatest concern by stakeholders. The number of studies which using disaggregated measures of CSR activities to examine relation between CSR and financial performance is rare. The only exception is Brammer, Brooks and Pavelin (2005a). In Taiwan, the Global Views Monthly, a leading magazine related to management and finance, develops a similar framework to evaluate social responsibility of a firm from three dimensions, that is, a firm's performance on community participation, environmental protection and financial transparency. To be more specific, they refer to OEKOM, an independent research and rating agency of CSR in Germany, to design a questionnaire about engagement and effectuation of the above three aspects for listing companies on the Taiwan Stock Exchange. They then compute scores on each three dimensions of CSR activities based on respondents' reply and get composite score of 312 firms. The availability of disaggregate data on various aspects of CSR performance is likely to be important since CSR is mutli-faceted and these aspects may have differential impacts depending on the nature of the firm's business. Following Brammer, Brooks and Pavelin (2005a), we relate 's three.

(10) 66 社會科學論叢 2009 年 4 月第三卷第一期 disaggregate measures (interchangeable three individual scores thereafter) and composite measure (interchangeable with total score thereafter) to accounting and market performance of firms. Our analysis is threefold. First, as Global Views Monthly's ratings made public in 2006Q2, we examine its relationship with subsequent stock returns of firms by portfolio comparisons and regression analysis. Second, we relate long-term financial performance to these four CSR measures. Third, based on composite measure, we examine whether high-score (top 40% quantile) firms exhibit less aggravating decline of their financial performance than low-score (bottom 40% quantile) firms in stagnations and support the view that CSR plays a role of insurance of financial performance. Our basic findings are, first, score on composite social performance indicator is negatively related to subsequent stock returns and this relationship cannot be rationalized by multi-factor models for explaining the crosssectional variation in stock returns. The poor market reward offered by such firms is attributable to their good social performance on the financial transparency and to a lesser extent the community participation and environmental aspects. The impacts on different aspects of social performance on stock returns are diverse among industries. Second, composite social performance indicator is positively related to long-term accounting performance but negatively related to long-term market performance. Third, high-score firms exhibit a more aggravating decline of their financial performance than low-composite-score firms in stagnations, and thus inconsistent with view that CSR as insurance of financial performance. The organization of the paper is as follows. Section 2 describes the measures of corporate social responsibility and how Global Views Monthly's compile the CSR data bank. Section 3 discusses the statistical methods,.

(11) Corporate Social Responsibility and Financial Performance: Empirical Evidence from Taiwan 67. section 4 presents the empirical results and section 5 concludes the paper.. II. Measures of CSR and Financial Performance A. Measures of Corporate Social Responsibility The estimation of the effects of CSR activities on financial performance of firms often confronts with the problem of rating firm's contribution on CSR activities. In early stage of research, the CSR performance is often measured by the amount of expenditure on polluting control investments, spending on environmental recuperation and protection (Bragdon and Marlin, 1972; Folger and Nutt, 1975; Spicer, 1978). Other studies use prestige investigation from business school students, and social reputation ratings by leading business magazines, such as the Fortune, Times and Business Ethics (McGuire et al., 1988; Herremans, Akathaporn and McInnes, 1993 and Preston and O'Bannon, 1997).6 More recently, some research and financial institutions, like the above mentioned KLD and Financial Times Stock Exchange,7 develop some widely 6. For example, Fortune magazine has been publishing the results of an annual survey of company reputations since 1983 by asking thousands of senior executives, directors and securities analysts who responded to the survey to rate the ten largest companies in their industries on eight attributes of reputation, using a scale of zero (poor) to ten (excellent). The attributes were quality of management; quality of products or services; innovativeness; long-term investment value; financial soundness; ability to attract, develop, and keep talented people; responsibility to the community and the environment; and wise use of corporate assets. The score of a company is the mean of the ratings on the right attributes. Surveys were published in January during 1983-1990, February during 1991-1994 and March during 1995-2006. Anderson and Smith (2006) and Antunovich, Laster and Mitnick (2000) found that stocks of companies ranked high by Fortune had higher subsequent returns than stocks that ranked low. But Shefrin and Statman (2003) found conflicting results. 7. FTSE, a jointly owned company by London Stock Exchange and Financial Times, provides financial indices..

(12) 68 社會科學論叢 2009 年 4 月第三卷第一期 acknowledged social responsible criteria which gradually became an international standard. For example, KLD rate firms as a CSR firms on the basis of the eight criteria, i.e., community, corporate governance, diversity, employee relations, environment, human rights, product quality and controversial business issues. Also, the firms included in FTSE's Index (FTSE4GOOD) must meet criteria requirements in three areas, environmental, social and stakeholders, and human rights. Also, those companies whose business interests are involved in tobacco, nuclear weapons and power station, and uranium are also excluded from the index. These indices have been widely employed.8 In Taiwan, the Global Views Monthly, develops a framework to evaluate social responsibility of a firm from three dimensions, that is, a firm's performance on community participation (COM), environmental protection (ENV) and financial transparency (FIN). To be more specific, they refer to OEKOM, an independent research and rating agency of CSR in Germany, to design a questionnaire about engagement and effectuation of the above three aspects for listing companies on the Taiwan Stock Exchange. They then compute scores on each three dimensions of CSR activities based on respondents' reply, totally 312 firms. Finally, they ranked companies according to their total scores (TOT) of three aspects. The announcement of this ranking is on May 2006. One caveat is worth noting. Firms with the following infamies are eliminated from the rating: negative events challenged by government agencies like Environmental Protection Administration or Council of Labor 8. Chih, Shen and Kang (2007) apply FTSE4GOOD to study the relationship between the earnings management and CSR. Waddock and Graves (1997), Tsoutsoura (2004) use KLD ratings to study the performance between CSR and financial performance..

(13) Corporate Social Responsibility and Financial Performance: Empirical Evidence from Taiwan 69. Affairs; major controversy between the labor and capital, aggravation with consumers, litigation and departure restrictions of CEO; and losses for three years. Table 1 shows the name list of firms with the highest scores on various CSR measures. Firms with highest total score on aggregate CSR measure in order are Delta Electronics, TSMC, BenQ, CMC and Advantech Co. Table 2 shows scores of CSR measures for firms based on traditional industry (T.I), financial industry (F.I), electronic industry (E.I) and transportation, tourism, trade and merchandise industry (TTTM.I). On average, financial industry gets highest total scores and traditional industry is lowest. For community participation, performance of financial industry is best and traditional industry is worst. Table 1. Firms with Highest Scores on CSR Measures CSR Measures Scores on Community Participation(COM). Scores on Environmental Protection (ENV). Scores on Financial Transparency (FIN). Total Score (TOT). Delta Electronics. Delta Electronics. Delta Electronics. Delta Electronics. TSMC. TSMC. TSMC. TSMC. BenQ. Advantech Co.. BenQ. BenQ. E.SUN FHC. BenQ. CMC. CMC. CMC. CMC. Advantech Co.. Advantech Co.. Table 2. Scores of CSR Measures for Industries Industry. CSR Measure. COM. ENV. FIN. TOT. Mean St. Dev Mean St. Dev Mean St. Dev Mean St. Dev. T.I. 38.3. 9.02. 8.42. 2.92. 19.3. 4.41. 66.0. 13.5. F.I. 43.0. 9.14. 6.85. 2.31. 21.4. 4.19. 71.3. 13.1. E.I. 40.4. 8.02. 8.71. 2.82. 20.6. 4.17. 69.7. 11.7. TTTM.I. 39.8. 8.54. 8.26. 2.70. 19.7. 5.29. 67.7. 12.3. All samples of Firms. 39.7. 8.64. 8.43. 2.84. 20.1. 4.39. 68.2. 12.7.

(14) 70 社會科學論叢 2009 年 4 月第三卷第一期 B. Measures of Financial Performance Various measures of financial performance in this area of research are classified as market-based and accounting-based measures. According to McGuire et al. (1988), while the former are forward-looking but subjective and easily affected by investor sentiment, the latter are historical and objective but incurring accounting manipulation and earning management.9 Moore (2001) proposed that using accounting measures is more valid. In this study, we use return on asset (ROA), return on equity (ROE), earnings per share (EPS), stock returns (STK) and price earnings ratio (P/E Ratio). The first three measures are accounting-based, and the last two are market-based. It is also worth-noting that for factors inﬂuencing cross-section stock returns of firms, we consider three factors of Fama and French (1992), market risk (BETA), market to book value (MTBV), size measured by capital (CAPITAL). We also consider and momentum factor (MOM) of Carhart (1997). Because the Global Views Monthly announce CSR ranking on May 2006, we collect stock returns from June 2006 to Feb 2007 and annualized to yearly stock returns and defined as our subsequent stock returns. The reason of using CSR scores in May 2006 and stock returns from June 2006 to Feb 2007 is getting rid of reverse causation problem, that's we avoid the problem of reverse causation of CSR scores by firm's stock returns. Available data on firm characteristic and performance variables are collected from database of the Taiwan Economic Journal (TEJ).. 9. Using market measures as performance variables are Alexander and Buchholz (1978), Vance (1975), Brammer, Brooks and Pavalin (2005a,b). The studies used accounting measures are Cochran and Wood (1984), Waddock and Graves (1997), and Tsoutsoura (2004). Using both measures are Griffin and Mahon (1997) and Moore (2001)..

(15) Corporate Social Responsibility and Financial Performance: Empirical Evidence from Taiwan 71. Table 3 presents correlation coefficient matrix for variables of our analysis, and numbers within parentheses are p-values of thesis coefficients. From the fifth row, we observe that total score are highly positively and significantly correlated with score on community participation, environmental protection and financial transparency. This is a matter of course because total score is aggregation of three disaggregate measures. As we will mention, in our regression analysis, we avoid putting four measures into one regression equation to get rid of multi-collinearity problem. Table 3. Correlation Matrix of Variables COM. ENV. FIN. TOT. Subs. Returns. Market Risks. M/B Ratio. Capital. COM. 1.0000. ENV. 0.4897 (0.000). 1.0000. FIN. 0.3811 (0.000). 0.2229 (0.000). 1.0000. TOT. 0.9236 (0.000). 0.6353 (0.000). 0.6570 (0.000). 1.0000. Subs. Returns. -0.0995 (0.079). -0.0556 (0.327). -0.1434 (0.011). -0.1300 (0.022). 1.0000. Market Risks. 0.1261 (0.028). 0.1105 (0.054). 0.0402 (0.485). 0.1246 (0.030). -0.0161 (0.780). 1.0000. M/B Ratio. 0.0708 (0.216). 0.0126 (0.827). 0.1499 (0.009). 0.1029 (0.072). -0.0764 (0.182). 0.0833 (0.148). 1.0000. Capital. 0.3459 (0.000). 0.1997 (0.000). 0.1864 (0.001). 0.3450 (0.000). -0.0961 (0.093). 0.2191 (0.000). 0.0367 (0.522). 1.0000. Mom. -0.0145 (0.801). 0.0369 (0.520). 0.0097 (0.866). 0.0018 (0.976). 0.1253 (0.028). 0.1497 (0.009). 0.1555 (0.006). -0.0988 (0.084). Mom. 1.0000. From the sixth row, we observe that subsequent returns are slightly and negatively correlated with three of four CSR measures, for example, correlation coefficient between subsequent returns and total score is -0.13,.

(16) 72 社會科學論叢 2009 年 4 月第三卷第一期 means that firms with higher CSR ratings have lower subsequent stock returns. Size measure, CAPITAL is positively correlated with four CSR measures, implies that large firms tend to get higher CSR ratings. Some other characteristics of firms are also highly correlated with CSR measures.. III. Data and Statistical Methods As above mentioned, most of studies examine the subsequent stock performance and long-term future financial performance as CSR ratings made public. In this paper, we define stock return from June 2006 to Feb 2007 as subsequent return and can also seen as holding period returns for investors. Using returns after CSR ratings announcement, we can make causal inference for CSR ratings impact on stock performance. Our analysis is threefold. First, we relate CSR ratings to subsequent returns by using portfolio and regression analysis. For portfolio analysis, according to three disaggregate measures and one aggregate measure, we separate firms into five groups. The first group (Group I hereafter) composed of firms with top 20% quantile of the given measure. The second group (Group II hereafter) composed of firms with 20%-40% quantile. The final group (Group V hereafter) composed of firms with bottom 20% quantile of the given measure. We compute subsequent gross portfolio returns of each group. We also calculate subsequent portfolio returns excess of market index returns of each group. Portfolio returns are calculated by two methods. The one is weighted by market capitalization of each firm, and the other is equally weighted of returns in each firm for a given group. Because grouping firms is based on descending CSR ratings, comparing subsequent returns of each group of firms could inference the relationship between CSR ratings and subsequent stock returns. If the stock returns of the group which composed of.

(17) Corporate Social Responsibility and Financial Performance: Empirical Evidence from Taiwan 73. firms with higher ratings is systematically larger than group with lower CSR ratings, then it implies that firms with higher CSR ratings tends to get higher stock returns. For regression analysis, we regress three disaggregate measures and one aggregate measure on the subsequent stock returns. Thus, the explained variable of regression equation is subsequent stock return, and explanatory variables are scores on community participation, environmental protection and financial transparency and total score. If we consider including three factors of Fama and French (1993) and momentum factor of Carhart (1997) as control factors, our regression model is as follow: Ri, t = α0 + α1COMi, t-1 + α2ENVi, t-1 + α3FINi, t-1 + α4TOTi, t-1 + α5BETAi, t-1 + α6PTBVi, t-1 + α7CAPi, t-1 + α8Ri, t-1 + εi, t where Ri, t is subsequent stock return of firm i in time t, COM is score on community participation, ENV is score on environmental protection, FIN is score on financial transparency and TOT is total score, BETA is measure of market risk in CAPM, PTBV is market to book value, CAP is capital, Ri, t-1 is momentum factor, which is defined as one-year stock return prior to announcement of CSR rating. is the error term. As we consider that aggregate measure of CSR is sum of scores of three disaggregate measures, in order to avoid multi-collinearity problem, we do not include all CSR measures in one regression equation. Instead, we set our regression model for four model specifications, that using three disaggregate measures but not control four factors (Model I), using aggregate measure but not control four factors (Model II), using three disaggregate measures and control four factors (Model III), using aggregate measure and control four factors (Model IV). Following are regression equations for four models: Model I：Ri, t = α0 + α1COMi, t-1 + α2ENVi, t-1 + α3FINi, t-1 + εi, t. (1).

(18) 74 社會科學論叢 2009 年 4 月第三卷第一期 Model II：Ri, t = α0 + α4TOTi, t-1 + εi, t. (2). Model III：Ri, t = α0 + α1COMi, t-1 + α2ENVi, t-1 + α3FINi, t-1 + α5BETAi, t-1 + α6PTBVi, t-1 + α7CAPi, t-1 + α8Ri, t-1 + εi, t (3) Model IV：Ri, t = α0 + α4TOTi, t-1 + α5BETAi, t-1 + α6PTBVi, t-1 + α7CAPi, t-1 + α8Ri, t-1 + εi, t (4) The estimation of the regression models are not only based on a combined sample of all types of firms but also on specific industries, that are traditional industry, financial industry, electronic industry and transportation, tourism, trade and merchandise industry. In second part of our analysis, we relate CSR ratings to longer-term financial performance, where performance measures are returns on assets, returns on equity, earnings per share, stock return and price earnings ratio. The performance data of this part of analysis is ranged from 2002Q2 to 2006Q2. Although our CSR ratings are made public in May 2006, we assume that CSR ratings could be a roughly social performance proxy measures from 2002 to 2006. Thus, we could use these measures to analyze longer-term CSR activities and financial performance. We use correlation coefficient and portfolio analysis to relate CSR ratings and five year average financial performance of firms. Similar as above, according to three disaggregate measures and one aggregate measure, we separate firms into five groups and compute long-term average financial performance of each group. We also calculate the Sharpe ratio of each group to control for risk because the Sharpe ratio measure excess returns per unit of risk for given portfolio. To compute five-year average Sharpe ratio, we first subtract yearly gross return by risk-free interest rate and divided by standard deviation of returns for given year. Then, we average them into five-year average Sharpe ratios. Larger Sharpe ratio means higher returns after.

(19) Corporate Social Responsibility and Financial Performance: Empirical Evidence from Taiwan 75. considering the risk of returns. As far as our third part of analysis, although the most of researches related corporate social performance to its financial performance, some examine influence of firm's CSR activities on consumer satisfaction and purchase intentions (Sen and Bhattacharya, 2001); sales growth (Brown and Dacin, 1997); business image (Fombrun and Shanley, 1990) and employee loyalty, the number of study which examining CSR as insurance protection of firms is rare. Refer to Peloza (2006), an important yet underemphasized benefit from CSR is insurance against negative events that would otherwise harm financial performance. Although previous researchers conceptualized CSR as a form of "operating license" or simply the actions of the firm that conform to social norms, the potential for CSR to act as an insurance policy that can mitigate the effects of negative events. For this analysis, we extend the data from 1991 to 2006 and examine whether high-CSR-ratings firms suffer less severe performance decreases during recessions or negative specific events, such as Blacconiere and Patten (1994). We use Monitoring Indicator constructed by the Council for Economic Planning and Development (CEPD) and compare financial performance of Group I and II (top 40% quantile of CSR ratings of firms) versus group IV and V (bottom 40% quantile of CSR ratings of firms). Insurance protection of performance is valid if we get smaller performance decay of Group I and II than Group IV and V..

(20) 76 社會科學論叢 2009 年 4 月第三卷第一期. IV. Empirical Results A. CSR and Subsequent Stock Returns (A) Analysis of Portfolio Returns Table 4 shows subsequent stock returns of five groups of firms from five samples of firms, that are samples of all 307 firms, 126 firms of traditional industry, 24 financial industries, 132 electronics industries and 25 transportation, tourism, trade and merchandise industries, which presents in Panel A to Panel E. Each group is composed of firms with different CSR ratings at descending order, that's the group I composed of firms with top 20% quantile of CSR ratings, group II composed of firms with 20%-40% quantile of CSR ratings and the portfolio return of each group is calculated as weighted average by market capitalization of each firm. Interesting findings are presented first. Group I and II get negative excess market returns regardless of using three disaggregate measures or aggregate measure. Second, using measure of community participation for ranking, portfolio return of Group I is the lowest of the five groups, 14.24% (market excess returns is-6.16%). Similar results occur when financial transparency and aggregate measure are used for ranking, and the subsequent returns of group I are 16.04% and 14.48%, respectively. When measure of environmental protection for rankings basis, stock performance of Group II is worst and it's subsequent portfolio returns is 7.96% and market excess returns is -12.44%. Third, higher subsequent portfolio returns occur in groups with lower CSR ratings. For example, as environmental protection as the CSR measure, group V gets the highest returns, as financial transparency as the.

(21) Corporate Social Responsibility and Financial Performance: Empirical Evidence from Taiwan 77. CSR measure, group IV got the highest returns. Fourth, the most interesting, the portfolio stock performance difference between group I and group V is negative regardless of CSR measures, implying there exists a trend that groups with higher CSR ratings get lower subsequent stock returns. Thus, we find that better performance in CSR does not necessarily get better stock performance. Instead, there are valuation discounts to firms with better CSR performance. Panel B shows similar arrangement as Panel A but using traditional industry as samples. We observe that, first, market excess returns of five groups are positive, means that stock performance of traditional industry is good during this period. Second, under the rankings of four CSR measures (community participation, environmental protection, financial transparency and aggregate measure in order), the groups with highest stock performance are group V, group III, group IV and group II, the groups with worst stock performance are group IV, group II, group II and group I. Third, performance difference between group I and group V are all negative. Thus, we get a trend that the groups with better CSR ratings will get lower subsequent stock performance. CSR discount still exists as we use only firms belonging to traditional industry. Table 4. Weighted Average Stock Returns of Five Groups of Firms Panel A. All Firms (Samples: 307) CSR Measure. COM. ENV. FIN. TOT. Portfolio. STK E.STK STK E.STK STK E.STK STK E.STK No.Firms. Group I. 14.24. -6.16. 17.04. -3.36. 16.04. -4.36. 14.48. -5.92. 62. Group II. 16.25. -4.15. 7.960. -12.4. 16.49. -3.91. 18.31. -2.09. 62. Group III. 28.55. 8.15. 22.65. 2.25. 17.13. -3.27. 19.00. -1.40. 61. Group IV. 25.71. 5.31. 15.08. -5.32. 23.96. 3.56. 26.24. 5.84. 61. Group V. 19.66. -0.74. 27.60. 7.20. 20.98. 0.58. 19.33. -1.07. 61. Diff (I and V). -5.42. -10.56. -4.94. -4.85.

(22) 78 社會科學論叢 2009 年 4 月第三卷第一期 Panel B. T.I (Samples: 126) CSR Measure. COM. Portfolio. STK. ENV. FIN. TOT. E.STK STK E.STK STK E.STK STK E.STK No.Firms. Group I. 30.57. 10.17. 29.75. 9.35. 33.96. 13.56. 28.89. 8.49. 25. Group II. 27.01. 6.61. 25.13. 4.73. 24.09. 3.69. 34.45. 14.05. 25. Group III. 33.04. 12.64. 35.61. 15.21. 28.96. 8.56. 29.52. 9.12. 25. Group IV. 24.83. 4.43. 27.15. 6.75. 31.71. 11.31. 31.73. 11.33. 25. Group V. 41.73. 21.33. 32.47. 12.07. 34.08. 13.68. 32.88. 12.48. 26. Diff (I and V). -11.16. -2.72. -0.12. -3.99. ENV. FIN. TOT. Panel C. F.I (Samples: 24) CSR Measure. COM. Portfolio. STK E.STK STK E.STK STK E.STK STK E.STK No.Firms. Group I. 13.71. -6.69. 12.48. -7.92. 2.19. -18.2. 12.63. -7.77. 5. Group II. 11.59. -8.81. 10.78. -9.62. 11.86. -8.54. 11.22. -9.18. 5. Group III. -5.40. -25.80 11.58. -8.82. 11.21. -9.19. 13.11. -7.29. 5. Group IV. 27.28. 12.59. -7.81. 13.75. -6.65. 13.36. -7.04. 5. Group V. 7.33. -13.07 13.97. -6.43. 11.71. -8.69. 7.00. -13.40. 4. Diff (I and V). 6.88. 6.38. -1.49. -9.52. 5.63. ENV. FIN. TOT. Panel D. E.I (Samples:132) CSR Measure Portfolio. COM STK. E.STK. STK. E.STK. STK. E.STK. STK. E.STK No.Firms. Group I. 8.50. -11.90 13.57. -6.83. 11.39. -9.01. 9.60. -10.80. 26. Group II. 19.81. -0.59. 1.34. -19.06 23.41. 3.01. 20.16. -0.24. 26. Group III. 21.93. 1.53. 10.97. -9.43. 7.56. -12.84 12.08. -8.32. 26. Group IV. 12.91. -7.49. 24.06. 3.66. 8.14. -12.26 17.11. -3.29. 27. Group V. 18.43. -1.97. 25.52. 5.12. 11.39. -9.01. -5.88. 27. Diff (I and V). -9.93. -11.95. 0.00. 14.52. -4.92.

(23) Corporate Social Responsibility and Financial Performance: Empirical Evidence from Taiwan 79 Panel E. TTTM.I (Samples:25) CSR Measure. COM. ENV. FIN. TOT. Portfolio. STK E.STK STK E.STK STK E.STK STK E.STK No.Firms. Group I. 13.09. -7.31. 13.98. -6.42. 15.81. -4.59. 13.63. -6.77. 5. Group II. 20.16. -0.24. 21.39. 0.99. 21.42. 1.02. 11.35. -9.05. 5. Group III. 42.79. 22.39. 20.11. -0.29. 27.03. 6.63. 49.69. 29.29. 5. Group IV. 51.12. 30.72. 70.40. 50.00. 45.17. 24.77. 52.12. 31.72. 5. Group V. 35.24. 14.83. 2.96. -17.44 48.08. 27.68. 37.25. 16.85. 5. Diff (I and V). -22.15. 11.02. -32.27. -23.62. Panel C shows a similar arrangement as Panel B but using financial industry as samples. We observe that, the first, market excess returns of most of the groups are negative, meaning that stock performance of this industry is weak during this period. Second, under the rankings of four CSR measures (community participation, environmental protection, financial transparency and aggregate measure in order), the groups with the highest stock performance are group II, group V, group IV and group IV, the groups with the worst stock performance are group III, group II, group I and group V. Third, the situation of performance difference between group I and group V are negative occurs when environmental protection and financial transparency are used as the CSR rankings. Thus, the former trend of higher CSR group get lower stock returns deteriorate. Panel D and Panel E show similar arrangement as before. We observe that, the first, under the rankings of four CSR measures (community participation, environmental protection, financial transparency and aggregate measure in order), the groups with highest stock performance are group I, group II, group III and group I, the groups with worst stock performance are group III, group IV, group II and group II. Second, the situation of.

(24) 80 社會科學論叢 2009 年 4 月第三卷第一期 performance difference between group I and group V are negative only occurs when financial transparency as CSR rankings. Third, in TTTM industry, the situation of performance difference between group I and group V are all negative except for environmental protection measure are used. Thus, CSR discount still occurs in TTTM industry. As we compare last row of Panel B to Panel E of Table 4, we can understand the CSR discount among four industries. For community participation as CSR measure, the largest CSR discount occurs in TTTM industry, -22.15%, the smallest CSR discount occurs in financial industry, 6.38%. For environmental protection as CSR measure, the largest CSR discount occurs in electronic industry, -11.95%, the smallest CSR discount occurs in financial industry, 11.02%. For financial transparency as CSR measure, the largest CSR discount occurs in TTTM industry, -32.37%, the smallest CSR discount occurs in electronic industry, 0.00%. For aggregate CSR measure is used, the largest CSR discount occurs in TTTM industry, -23.62%, the smallest CSR discount occurs in financial industry, 5.63%. From the above analysis, we know that the public responses of CSR activities are different according to different aspects of CSR and industries. For example, the public give positive valuation of financial industry on efforts on activities on community participation, but negative valuation to TTTM industry. The public give large negative valuation of electronic industry on efforts in environmental protection, but positive valuation to TTTM industry. The public give neutral valuation of electronic industry on efforts in financial transparency, but negative valuation to TTTM industry. Thus, this result implies that the public give different valuation to the efforts of different aspects of CSR activities. This conform to the proposition by Porter and Kramer (2006) that business should not blindly engage themselves in any.

(25) Corporate Social Responsibility and Financial Performance: Empirical Evidence from Taiwan 81. scope of CSR activities, instead, wise managers have to put a new premium on specific area of CSR issues that most to be beneficial to core business of firm and topics which the public most concerned. Speaking in more detail, a firm should devote to a CSR issue which not only have benefits to social but also establish competitive advantage, achieve a positive-sum game rather than zero-sum game. Table 5 shows subsequent stock returns of five groups of firms from five samples of firms, that are samples of all 307 firms, 126 in traditional industry, 24 in financial industry, 132 in electronics industry and 25 in the transportation, tourism, trade and Merchant industry, which are present in Panels A to E. While the portfolio returns of each group are calculated as equally weighted average by market of each firm's return, each group are composed of firms with different CSR ratings at descending order, that's the group I composed of firms with top 20% quantile of CSR ratings, group II composed of firms with 20%-40% quantile of CSR ratings. Because portfolio returns of groups are equally weighted, we can perform t-test for group comparisons. For Panel A, when all samples are used, under the rankings of four CSR measures (community participation, environmental protection, financial transparency and aggregate measure in order), the groups with highest stock performance are group III, group V, group V and group IV, while the groups with worst stock performance are group I, group III, group I and group I. Performance difference between group I and group V are all negative, for example, when community participation are used as CSR measure, difference of stock return between group I and group V are -7.14, but not significant; when financial transparency are used as CSR measure, difference of stock return between group I and group V are negatively significant, -7.14; when.

(26) 82 社會科學論叢 2009 年 4 月第三卷第一期 aggregate CSR measure is used as CSR measure, difference of stock return between group I and group V are -11.23 and are significant. We got a similar trend as before that the groups with better CSR ratings will get lower subsequent stock performance, thus CSR discount exists. From Panel B, the trend with higher CSR group get lower stock returns is not so evident, and so is the CSR discount. Similar results are gained from panel C to D. From Panel E we find that when financial transparency is the CSR measure, significant negative CSR discount exists. Significant CSR discount exist with environmental protection as the CSR measure. Thus, as with the MMMT industry, efforts on CSR activities should concentrate on financial but not environmental aspects. Thus, as before, firms should engage CSR in most-favored aspects by investors. Table 5. Simple Average Stock Returns of Five Groups of Firms Panel A. All Firms (Samples: 307) CSR Measure. COM. ENV. FIN. TOT. Portfolio. STK. E.STK. STK. E.STK. STK. E.STK. STK. Group I. 20.46. 0.06. 23.16. 2.76. 18.96. -1.44. 18.06. E.STK No.Firms -2.34. 62. Group II. 21.91. 1.51. 26.30. 5.90. 23.99. 3.59. 23.38. 2.98. 62. Group III. 30.83. 10.43. 21.50. 1.10. 23.66. 3.26. 28.94. 8.54. 61. Group IV. 28.88. 8.48. 28.67. 8.27. 30.30. 9.90. 30.01. 9.61. 61. Group V. 27.60. 7.20. 29.92. 9.52. 32.74. 12.34. 29.29. 8.89. 61. Diff (I and V). Statistics p-value Statistics p-value Statistics p-value Statistics p-value -7.14. 0.1138. -6.76. 0.1328. -13.78. 0.0151. -11.23. 0.0227.

(27) Corporate Social Responsibility and Financial Performance: Empirical Evidence from Taiwan 83 Panel B. T.I (Samples: 126) CSR Measure. COM. ENV. FIN. TOT. Portfolio. STK. E.STK. STK. E.STK. STK. E.STK. STK. Group I. 30.06. 9.66. 26.52. 6.12. 29.75. 9.35. 23.34. 2.94. 25. Group II. 27.02. 6.62. 32.30. 11.90. 25.01. 4.61. 39.32. 18.92. 25. Group III. 32.88. 12.48. 31.10. 10.70. 34.22. 13.82. 25.76. 5.36. 25. Group IV. 27.46. 7.06. 37.23. 16.83. 32.79. 12.39. 30.84. 10.44. 25. 39.14. 18.74. 29.79. 9.39. 34.96. 14.56. 37.38. 16.98. 26. Group V Diff (I and V). E.STK No.Firms. Statistics p-value Statistics p-value Statistics p-value Statistics p-value -9.08. 0.2128. -3.27. 0.3737. -5.21. 0.3031. -14.04. 0.0946. Panel C. F.I (Samples: 24) CSR Measure Portfolio. COM. ENV. FIN. STK. E.STK. STK. E.STK. TOT. STK. E.STK. STK. E.STK No.Firms. Group I. 8.80. -11.60. 10.19. -10.21. 9.78. -10.62. 12.17. -8.23. 5. Group II. 24.05. 3.65. 8.75. -11.65. 11.38. -9.02. 7.60. -12.80. 5. Group III. 7.64. -12.76. 8.28. -12.12. 13.40. -7.00. 15.11. -5.29. 5. Group IV. 10.11. -10.29. 20.91. 0.51. 16.74. -3.66. 19.59. -0.81. 5. Group V. 12.62. -7.78. 15.70. -4.70. 11.75. -8.65. 7.78. -12.62. 4. Diff (I and V). Statistics p-value Statistics p-value Statistics p-value Statistics p-value -3.82. 0.2959. -5.51. 0.1398. -1.97. 0.4183. 4.39. 0.3147. Panel D. E.I (Samples:132) CSR Measure. COM. ENV. FIN. TOT. Portfolio. STK. E.STK. STK. E.STK. STK. E.STK. STK. Group I. 14.28. -6.12. 20.01. -0.39. 19.25. -1.15. 15.78. E.STK No.Firms -4.62. 26. Group II. 19.04. -1.36. 25.05. 4.65. 20.54. 0.14. 13.88. -6.52. 26. Group III. 33.77. 13.37. 9.82. -10.58. 16.25. -4.15. 31.93. 11.53. 26. Group IV. 17.90. -2.50. 26.30. 5.90. 29.15. 8.75. 26.71. 6.31. 27. 23.44. 3.04. 26.98. 6.58. 23.25. 2.85. 20.37. -0.03. 27. Group V Diff (I and V). Statistics p-value Statistics p-value Statistics p-value Statistics p-value -9.16. 0.1686. -6.97. 0.2464. -4.00. 0.3384. -4.59. 0.3089.

(28) 84 社會科學論叢 2009 年 4 月第三卷第一期 Panel E. TTTM.I (Samples:25) CSR Measure. COM. ENV. FIN. TOT. Portfolio. STK. E.STK. STK. E.STK. STK. E.STK. STK. Group I. 17.82. -2.58. 28.35. 7.95. 28.78. 8.38. 35.65. 15.25. 5. Group II. 26.38. 5.98. 35.09. 14.69. 20.57. 0.17. 50.06. 29.66. 5. Group III. 38.75. 18.35. 36.12. 15.72. 36.95. 16.55. 47.97. 27.57. 5. Group IV. 46.13. 25.73. 62.57. 42.17. 29.98. 9.58. 13.25. -7.15. 5. 36.79. 16.39. 3.73. -16.67. 49.59. 29.19. 18.95. -1.45. 5. Group V Diff (I and V). E.STK No.Firms. Statistics p-value Statistics p-value Statistics p-value Statistics p-value -18.97. 0.1348. 24.62. 0.0598. -20.81. 0.0981. 16.7. 0.1870. (B) Regression Analysis Results We use regression analysis to examine the relationship between measures of CSR and subsequent returns for firms. Our samples contain all previous firm types; traditional industry; financial industry; electronic industry; transportation, tourism, trade and merchant industry. For each group of samples, as we consider multi-collinearity among CSR measures and four factors of Fama and French (1992) and Carhart (1997), we have four kinds of estimated models. Table 6 reports the OLS estimation results. We get several striking results. First, for all samples, for model I, the estimated coefficients for three disaggregate measures of CSR are negative but are significant for measures of financial transparency. It means that firms with more efforts in financial transparency have lower subsequent stock returns. If we control four factors, as in model III, similar results are obtained. Second, for model II, the estimated coefficients for aggregate measure of CSR is negative and significant (-0.369) and thus mean that firms with more efforts in aggregate CSR activities will have lower subsequent stock returns. We get similar results if we control four factors, as in model IV. Thus we draw the conclusion that firms with higher performance in CSR will get worse stock.

(29) Corporate Social Responsibility and Financial Performance: Empirical Evidence from Taiwan 85. returns and this relationship is mostly explained by the negative relationship between financial transparency and stock returns. For samples of traditional industry, as shown in panel B, estimated coefficients of all measures of CSR are negative but insignificant regardless of what model specification. Although we cannot make a conclusion on the negative relationship between CSR and stock returns, we get clear results that engagement in CSR at least does not improve stock returns of firms. Similar results are obtained for financial and electronic industry, we shown in panel C and panel D. From model I in panel E, we observe that the estimated coefficient for financial transparency is negative and significant (-2.174); in model III, the estimated coefficient for community participation is negative and significant (-1.378). Table 6. OLS Results of CSR Ratings and Subsequent Stock Returns Panel A. All Firms (Samples: 307) Estimated. Intercept. COM. ENV. FIN. TOT. Coefficient. α0. α1. α2. α3. α4. 54.4. -0.212. -0.041. -1.006**. (4.77). (-0.75). (-0.050). (-2.021). Model I Model II Model III Model IV. 50.6. Four Beta Loadings α5. α6. α7. α8. -0.369**. (4.57). (-2.31). 49.1. 0.010. -0.467. -0.890*. (4.01). (0.03). (-0.60). (-1.88). -0.294. -1.408 -5.98×10 - 8 0.095. (-0.04). (-0.04). 45.7. -0.268*. 0.101. -1.542 -5.48×10 - 8 0.094. (-1.08). (3.84). (-1.67). (0.01). (-1.51). (-0.99). (1.89) (1.87).

(30) 86 社會科學論叢 2009 年 4 月第三卷第一期 Panel B. T.I (Samples: 126) Estimated. Intercept. COM. ENV. FIN. TOT. Coefficient. α0. α1. α2. α3. α4. Model I Model II Model III Model IV. 53.60. -0.140. -0.064. -0.843. (3.44). (-0.34). (-0.05). (-1.11). 50.34. -0.287. (3.39). (-1.30). Four Beta Loadings α5. α6. α7. α8. 34.70. -0.066. -1.014. -0.414. 23.64. -2.887. -1.79×10 - 8. 0.073. (2.00). (-0.17). (-0.84). (-0.57). (2.24). (-1.59). (-0.14). (0.78). 35.44. -0.291. 22.68. -2.878. -1.76×10 - 8. 0.078. (2.14). (-1.36). (2.19). (-1.60). (-0.14). (0.84). Panel C. F.I (Samples: 24) Estimated Intercept Coefficient Model I Model II Model III Model IV. α0. COM. ENV. FIN. TOT. α1. α2. α3. α4. 20.14. 0.351. -1.598. -0.541. (1.10). (0.77). (-1.01). (-0.65). 17.91. -0.074. (1.09). (-0.33). -5.875. 0.296. -0.058. -0.598. (-0.20). (0.52). (-0.03). (-0.62). Four Beta Loadings α5. α6. 30.35. -8.391. α7. α8. -1.87×10 - 8 0.083. (1.67). (-0.74). (-0.02). -12.42. 0.049. 31.33. -7.971. 1.62×10 - 8. (0.30) 0.024. (-0.47). (0.17). (1.93). (-0.74). (0.21). (0.11). Panel D. E.I (Samples: 132) Estimated Intercept Coefficient Model I Model II Model III Model IV. α0. COM. ENV. FIN. TOT. α1. α2. α3. α4. 46.12. -0.094. -0.953. -0.593. (2.02). (-0.17). (-0.63). (-0.64). 44.98. -0.333. (2.06). (-1.08). 55.64. 0.377. -0.901. -0.794. (2.34). (0.74). (-0.66). (-0.94). Four Beta Loadings α5. α6. -25.94. 0.014. (-1.99). (0.01). 51.73. -0.079. -25.05. -0.077. (2.26). (-0.26). (-1.95). (-0.05). α7. α8. -3.68×10 - 8 0.077 (-0.44). (0.98). -3.43×10 - 8 0.067 (-0.41). (0.86).

(31) Corporate Social Responsibility and Financial Performance: Empirical Evidence from Taiwan 87 Panel E. TTTM.I (Samples: 25) Estimated. Intercept. COM. ENV. FIN. TOT. Coefficient. α0. α1. α2. α3. α4. Model I Model II Model III Model IV. 82.87. -0.192. 0.077. -2.174**. (2.57). (-0.25). (0.03). (-1.96). 78.38. -0.668. (2.43). (-1.42). Four Beta Loadings α5. α6. α7. α8. 16.24. -1.378*. 3.169. -1.104. 76.93. 4.788. -3.00×10 - 7. 0.321. (0.41). (-1.77). (1.46). (-0.91). (3.34). (0.75). (-0.66). (3.05). 23.44. -0.659. 65.03. 2.330. -2.82×10 - 7. 0.279. (0.63). (-1.10). (3.15). (0.36). (-0.62). (2.70). From the above results we know that firms with larger efforts in engaging CSR activities are not necessarily positively evaluated by investors. Instead, our above analysis presents evidence that negative response occurred more often. Investors do not support firms with philanthropic ties to the public, because these behavior of firms are not thought to be beneficial to financial healthy of firms, at least from the view point of typical investor. As we use disaggregate measures, we get the results of negative relationship between financial transparency and financial returns. This can be explained by the inefficiency of stock market in Taiwan. Firms with less limpid financial information may probably boom the stock price by bluffing sales.10 Instead, more well-behaved firms with more transparent financial information do not go in this direction and demonstrate less striking performance in stock returns. The above phenomenon cannot be rationalized by multi-factor models 10. According to Shen (2002), when the products of a company are not sold out, the company sells the products to its subsidiaries, which are not listed in the stock market and hence are not responsible to the public. The fake increase in sale of the mother company stimulates its stock price. The company then uses the high price of stock as collateral to borrow more money from banks. Revenue from the product sale is recorded as "accounts receivable" in the company's balance sheet, but the cash will never come in. The growing sales generate no profit..

(32) 88 社會科學論叢 2009 年 4 月第三卷第一期 for explaining the cross-sectional variation in stock returns.. B. CSR and Long-term Financial Performance We examine the relationship between CSR ratings and long-term financial performance by collecting financial performance data from 2002 to 2006 for all samples of firms. Measures of financial performance are return on asset, return on equity, earnings per share, stock returns and price earning ratio. The first three are accounting-based measures and the latter two are market-based. Table 7 presents correlation coefficients between four CSR measures and five-year average financial performance of firms. From the left part of Panel A which all samples are used, we observe a striking result which shows that the correlation coefficients between CSR ratings and ROA, ROE and EPS are positive, but correlation coefficients between CSR and STOCKRET and P/E ratio are negative regardless what CSR measure is used. It means that firms with higher CSR ratings are accompanied by better average long-tern return on asset, return on equity and earnings per share, but worse long-term average stock returns and price earnings ratio. That's firms with better CSR performance have better accounting-based performance but lower marketbased performance. Good firms are good in books, but not good investments for investors. From right part of Panel A which firms of traditional industry are used, similar results are obtained. Firms with higher CSR ratings have better long-term accounting-based performance but worse long-term marketbased performance..

(33) Corporate Social Responsibility and Financial Performance: Empirical Evidence from Taiwan 89. Table 7. Correlation between CSR and Long-term Financial Performance Panel A. Samples-All Firms and T.I Performance Measure. All Firms. T.I. COM. ENV. FIN. TOT. COM. ENV. FIN. TOT. ROA. 0.165. 0.157. 0.252. 0.237. 0.327. 0.165. 0.245. 0.338. ROE. 0.118. 0.116. 0.235. 0.190. 0.272. 0.082. 0.196. 0.266. EPS. 0.196. 0.123. 0.235. 0.245. 0.333. 0.111. 0.225. 0.323. STOCKRET. -0.083. -0.016. -0.088. -0.092. -0.134. 0.014. -0.163. -0.142. P/E Ratio. -0.066. -0.130. -0.110. -0.114. -0.140. -0.186. -0.117. -0.175. Panel B. Samples-F.I, E.I and TTTM.I Performance F.I Measure COM ENV FIN. E.I TOT COM ENV. TTTM.I FIN. TOT COM ENV. FIN. TOT. ROA. 0.226 0.264 0.326 0.309 0.095 0.140 0.238 0.183 0.115 -0.055 0.474 0.277. ROE. -0.016 0.045 0.457 0.137 0.094 0.148 0.234 0.183 0.106 -0.087 0.343 0.206. EPS. 0.317 0.183 0.125 0.296 0.137 0.106 0.235 0.202 0.244 -0.037 0.527 0.395. STOCKRET -0.063 -0.172 -0.055 -0.093 0.036 0.034 0.119 0.075 0.070 -0.242 -0.012 -0.008 P/E Ratio. -0.283 0.076 0.070 -0.165 -0.022 -0.120 -0.059 -0.065 0.050 -0.233 -0.379 -0.182. Panel B of table 7 presents correlation coefficients between four CSR measures and five-year average financial performance of firms of financial industry, electronics industry and the industry of transportation, tourism, trade and merchant. First, as we observe the left part of panel B of table 7, the correlation coefficients between measures of community participation are positive for ROA and ROE but negative for EPS, STOCKRET and P/E ratio. The correlation coefficients between measures of environmental protection are positive for all accounting-based performance but negative for marketbased performance. Similar results are obtained if measure of financial transparency and aggregate CSR measure are used. Second, from the middle part of panel B which firms of electronic industry are used, we observe that.

(34) 90 社會科學論叢 2009 年 4 月第三卷第一期 correlation coefficients between CSR and long-term average ROA, ROE, EPS and STOCKRET are positive, but are negative between CSR and P/E ratio regardless what CSR measure is used. Third, from the right part of panel B where firms of industry of transportation, tourism, trade and merchant are used, we observe that if CSR measure is community participation, correlation coefficients between CSR and five long-term financial performance indicators are all positive, but are all negative if environmental protection is used as CSR measure. If financial transparency is used as the CSR measure, correlation coefficients between CSR and long-term accounting-based performance indicators are positive, but correlation coefficients between CSR and long-term market-based performance indicators are negative. Similar results are obtained if aggregate CSR measure is used. Although the results are slight mixed, we get general results that more philanthropic firms have better accounting performance but worse market performance. According to aggregate CSR measure, we divide firms into five groups on descending order of CSR ratings. As mentioned before, the first group composed of top 20% quantile of samples, the second group composed of 20%-40% quantile of samples. The fifth group composed of bottom 20% quantile of samples. Table 8 reports five year long-term simple average financial performance of five groups of firms. Panel A shows five-year average of return on asset of five groups of firms. From the second column which all firms are used, average ROAs of group I and II are 8.286% and 6.19%, respectively. The average ROA of group V is 1.585%. The difference between group I and group V is 6.7%. Thus, it seems to have descending trend in performance which presents that group of firms with better CSR ratings get higher average returns on asset. As we further do this analysis using different industries, we get similar results. For example, the third.

(35) Corporate Social Responsibility and Financial Performance: Empirical Evidence from Taiwan 91. column in panel A, only the firms of traditional firms are used, average ROAs of group I and II are 7.09% and 6.149%, respectively, and the average ROA of group V is 0.981%. The difference between group I and group V is 6.109%. It also show descending trend in performance which presents that group of firms with better CSR ratings get higher average returns on asset. Similar results obtained if only firms in financial Industry, firms in electronic industry and firms in transportation, trade, tourism and merchandise industry are used. The panel B shows long-term five year average returns on equity of five groups of firms. As all samples of firms are used, as shown in the second column, average ROEs of group I and II are 12.03% and 8.381%, respectively. The average ROE of group V is -3.139%. The difference between group I and group V is 15.17%. Thus, it also shows descending trend in performance which presents that group of firms with better CSR ratings get higher average returns on equity. Similar results are obtained if only firms in traditional, financial, electronics and transportation, trade, tourism and merchandise industries are used. Panel C show familiar results and presents a trend that groups of firms with better CSR ratings got higher long-term average earnings per share. From panel A to panel C we can make a conclusion that firms with better CSR ratings get better accounting-based financial performance. Panel D in table 8 shows long-term five year simple average stock returns of five groups of firms. As all samples of firms are used, as shown in second column, the trend of higher CSR-lower performance is not obvious. For example, average STOCKRETs of group I and II are 16.83% and 13.54%, respectively. The average STOCKRET of group V is 26.31%. The difference between group I and group V is -9.47%. Similar results are obtained if only firms in traditional, financial, electronics or transportation, trade, tourism and.

(36) 92 社會科學論叢 2009 年 4 月第三卷第一期 merchandise industries are used. Panel E is similar to panel D but presents weighted average stock returns of five groups of firms. When all samples of firms are used, average STOCKRETs of group I and II are 21.37% and 16.18%, respectively. The average STOCKRET of group V is 24.09%. The difference between group I and group V is -2.714%. As we use firms in traditional industry, firms in financial industry, firms in electronic industry and firms in transportation, trade, tourism and merchandise industry for analysis, we did not get a clear trend that higher CSR-lower performance is presented. Panel F show even contrary results relative to panel A to panel C. Average P/E ratio of group I and II are 44.79 and 31.58, respectively. The average STOCKRET of group V is 101.4. The difference between group I and group V is -56.58. Similar results are obtained if only firms in traditional, financial industry, electronics or transportation, trade, tourism and merchandise industries are analyzed. At the same time we can find a slightly different trend which is contrary to the trend obtained from panel A to panel C, that's the group with higher CSR ratings get lower P/E ratio. From now on, we can make a conclusion that the group of firms with higher CSR ratings tend to have higher returns on assets, returns on equity and earnings per share, but not on stock returns and price earnings ratio. Especially, firms with higher CSR ratings tend to have lower price earnings ratio than firms with lower CSR ratings. This is partially consistent with former results that higher CSR firms have better accounting-based performance but worse market-based performance than firms with lower CSR ratings. Good companies are good in books but not necessarily good investments for investors..

(37) Corporate Social Responsibility and Financial Performance: Empirical Evidence from Taiwan 93. Table 8. Long-term Financial Performance of Firms Panel A. Performance Measure: ROA Portfolio. All Firms. T.I. F.I. E.I. TTTM.I. Group I. 8.286. 7.090. 5.748. 9.193. 10.92. Group II. 6.190. 6.149. 1.911. 7.640. 6.002. Group III. 4.769. 4.798. 1.650. 6.392. 5.494. Group IV. 4.838. 3.937. 0.222. 5.186. 0.824. Group V. 1.585. 0.981. 1.524. 2.797. 2.347. Diff (I and V). 6.700. 6.109. 4.225. 6.396. 8.572. Panel B. Performance Measure: ROE Portfolio. All Firms. T.I. F.I. E.I. TTTM.I. Group I. 12.03. 9.292. 7.111. 14.11. 18.74. Group II. 8.381. 8.475. 0.652. 9.662. 10.35. Group III. 3.061. 8.053. 1.785. 9.109. -10.89. Group IV. 4.633. 5.059. -28.48. 3.438. 0.584. Group V. -3.139. -7.033. -3.332. 0.384. 1.420. Diff (I and V). 15.17. 16.33. 10.44. 13.73. 17.32. Panel C. Performance Measure: EPS Portfolio. All Firms. T.I. F.I. E.I. TTTM.I. Group I. 2.520. 2.077. 1.271. 3.165. 2.964. Group II. 1.793. 1.437. 0.099. 2.311. 1.708. Group III. 1.244. 1.222. 0.288. 1.756. 1.099. Group IV. 1.410. 1.097. 0.977. 1.678. 0.040. Group V. 0.500. 0.306. -0.176. 1.031. 0.322. Diff (I and V). 2.020. 1.771. 1.447. 2.134. 2.642.

(38) 94 社會科學論叢 2009 年 4 月第三卷第一期 Panel D. Performance Measure: Average Yearly Stock Returns (Weighted by Market Cap.) Portfolio. All Firms. T.I. F.I. E.I. TTTM.I. Group I. 16.83. 33.33. 13.49. 10.79. 33.10. Group II. 13.54. 32.20. 14.67. 3.647. 20.26. Group III. 19.28. 26.55. 10.84. 8.688. 47.23. Group IV. 19.39. 32.16. 10.20. 9.210. 24.61. Group V. 26.31. 40.79. 16.30. 12.44. 32.62. Diff (I and V). -9.479. -7.461. -2.815. -1.650. 0.475. Panel E. Performance Measure: Average Yearly Stock Returns (Equally Weighted) Portfolio. All Firms. T.I. F.I. E.I. TTTM.I. Group I. 21.37. 29.81. 15.85. 13.04. 30.20. Group II. 16.18. 22.60. 17.26. 10.49. 18.54. Group III. 19.59. 26.89. 21.61. 9.072. 41.89. Group IV. 18.18. 27.88. 17.07. 11.06. 23.86. Group V. 24.09. 37.99. 18.31. 5.789. 25.32. Diff (I and V). -2.714. -8.186. -2.463. 7.252. 4.874. Panel F. Performance Measure: Average P/E Ratio Portfolio. All Firms. T.I. F.I. E.I. TTTM.I. Group I. 44.79. 22.75. 96.76. 46.98. 27.65. Group II. 31.58. 31.23. 119.2. 32.73. 17.21. Group III. 36.78. 33.82. 17.87. 42.62. 34.35. Group IV. 104.4. 111.9. 32.40. 39.90. 576.3. Group V. 101.4. 47.70. 455.7. 63.87. 123.6. Diff (I and V). -56.58. -24.96. -359.0. -16.89. -95.98. Figure 1 presents a bar chart of five long-term average financial performance indicators of each group firms. As before, samples of firms are divided into five groups by their aggregate CSR ratings. From A we can find that there exits a descending trend in return on assets from group I to group V..

(39) Corporate Social Responsibility and Financial Performance: Empirical Evidence from Taiwan 95. Similar results are obtained for B and C. This means that the group of firms with higher CSR ratings gets better accounting-based performance indicators. But this trend disappeared as we observe diagram D, E and F, which shows stock returns (simple average, weighted average) and price earnings ratio of five groups of firms. Thus, the group of firms do not necessarily get better market-based performance indicators. This is consistent with our former results that Good Companies which engage more efforts in CSR are good in books, but not good investments, firms with corporate social responsibility is not necessarily responsible for investors. Figure 1. Bar Chart of Long-term Financial Performance of Firms A.ROA B.ROE. C.EPS D. Stock Return (Simple Average).

(40) 96 社會科學論叢 2009 年 4 月第三卷第一期 E. Stock Return (Weighted Average) F. P/E Ratio. Table 9 reports similar data as table 8 but with the difference of using three disaggregate measures as rankings to divide samples of firms into five groups. For example, the left part of panel A shows long-term five-year average returns on assets of five groups of firms. Firms are grouped according to descending order of community participation ratings. The samples of firms for analysis are five, all samples of firms, firms in traditional industry, firms in financial industry, firms in electronic industry and firms in transportation, tourism, trade and merchandise industry. We can get similar results as before that groups with better CSR ratings (in community participation) have higher five-year average returns on assets. When we observe the middle and right part of panel A, which CSR measures are environmental protection and financial transparency, a similar trend is obtained. For panel B and panel C, although it is not perfect, we still find a trend that group of firms with better CSR ratings get higher returns on equity and earnings per share no matter what CSR ratings are used. But as we observe panel D to panel F, this trend is non-existent. Thus, as before, we get a general result which state that firms with better CSR ratings have higher accountings-based performance indicators but not market-based indicators, and this results is invariant to different measures of CSR..

(41) Corporate Social Responsibility and Financial Performance: Empirical Evidence from Taiwan 97. Table 9. L ong-term Financial Performance of Firms (Grouping Firms Based on Three Disaggregated CSR Measures) Panel A. Performance Measure: Average ROA Portfolio. COM. ENV. FIN. All. T.I. F.I. E.I. TTTM. All. T.I. F.I. E.I. TTTM. All. T.I. F.I. E.I. TTTM. Group I. 7.60. 6.24. 5.75. 8.40. 10.3. 7.42. 7.15. 5.96. 9.04. 5.69. 8.54. 8.57. 3.41. 10.6. 10.9. Group II. 5.92. 6.40. -0.55. 7.85. 4.23. 5.53. 3.29. 1.54. 5.80. 2.30. 7.16. 4.28. 2.63. 7.29. 9.29. Group III 4.84. 5.36. 1.65. 5.42. -0.05. 6.34. 6.19. -0.64. 8.07. 4.90. 4.08. 4.27. 4.09. 5.49. 2.29. Group IV. 4.14. 4.89. 2.82. 5.60. 3.78. 2.95. 4.88. 2.59. 6.24. 11.3. 3.50. 4.65. 1.54. 4.80. 1.02. Group V. 3.19. 0.11. 1.83. 3.88. 7.37. 3.41. 1.43. 1.78. 2.05. 1.38. 2.37. 1.18. 2.35. 3.02. 2.06. Diff (I and V). 4.41. 6.13. 3.92. 4.53. 2.89. 4.01. 5.73. 4.18. 6.98. 4.32. 6.17. 7.39. 1.05. 7.59. 8.86. F.I. E.I. TTTM. All. T.I. F.I. E.I. TTTM. 18.7. Panel B. Performance Measure: Average ROE Portfolio. COM All. T.I. Group I. 10.7. Group II. 6.24. Group III. 5.04. Group IV Group V Diff (I and V). ENV. FIN. F.I. E.I. TTTM. All. T.I. 8.01. 7.11. 13.0. 17.4. 9.85. 9.32. 6.34. 12.9. 8.97. 12.4. 12.0. 3.57. 15.9. 9.45. -34.5. 11.3. 6.73. 6.40. 4.74. -1.48. 8.40. -12.7. 10.8. 6.42. 4.46. 10.6. 13.3. 8.37. 1.79. 5.73. -15.6. 9.26. 9.60. -31.5. 10.9. 6.51. 3.31. 5.11. 5.60. 4.33. -14.2. 2.42. 7.30. 6.55. 1.64. 1.61. -1.95. 2.53. 5.28. 7.48. 15.3. 3.13. 6.79. 0.59. 5.67. -0.41. 0.57. -9.19 -0.07. 4.93. 10.2. 1.41. -2.64. 0.86. -3.07. 2.14. -4.69 -6.59 -3.50. 0.12. 2.78. 10.2. 17.2. 8.07. 7.19. 8.44. 12.0. 5.49. 16.0. 6.84. 17.1. 18.6. 15.8. 16.0. 7.19. 7.07. Panel C. Performance Measure: Average EPS Portfolio. COM. ENV. FIN. All. T.I. F.I. E.I. TTTM. All. T.I. F.I. E.I. TTTM. All. T.I. F.I. E.I. TTTM. Group I. 2.36. 1.89. 1.27. 3.00. 2.87. 2.08. 1.96. 1.17. 2.70. 1.41. 2.61. 2.54. 0.66. 3.48. 2.96. Group II. 1.76. 1.61. 0.20. 2.41. 0.96. 1.61. 0.85. -0.06. 1.48. 0.62. 1.99. 0.96. 0.59. 2.26. 2.27. Group III. 1.23. 1.22. 0.29. 1.46. 0.19. 1.71. 1.53. 0.58. 2.92. 1.11. 1.04. 0.99. 0.95. 1.63. 0.28. Group IV. 1.36. 1.46. 0.88. 1.84. 0.63. 1.01. 1.21. 0.69. 1.78. 2.46. 1.00. 1.23. 0.18. 1.41. 0.19. Group V. 0.75. -0.03. 0.19. 1.22. 1.47. 1.04. 0.59. 0.30. 1.05. 0.54. 0.82. 0.42. 0.55. 1.17. 0.42. Diff (I and V). 1.61. 1.92. 1.08. 1.79. 1.40. 1.04. 1.37. 0.87. 1.65. 0.87. 1.78. 2.12. 0.11. 2.31. 2.54.

(42) 98 社會科學論叢 2009 年 4 月第三卷第一期 Panel D. Performance Measure: Average Yearly Stock Returns (Weighted by Market Cap.) Portfolio. COM. ENV. FIN. All. T.I. F.I. E.I. TTTM. All. T.I. F.I. E.I. TTTM. All. T.I. F.I. E.I. TTTM. Group I. 15.0. 30.4. 13.5. 9.96. 31.0. 17.7. 34.2. 12.2. 10.6. 17.7. 18.0. 34.6. 17.2. 10.5. 33.1. Group II. 16.2. 36.6. 12.2. 7.06. 27.1. 10.8. 23.8. 11.7. 3.38. 41.8. 14.7. 27.9. 12.1. 8.70. 23.3. Group III. 25.0. 29.0. 10.8. 12.6. 37.1. 21.1. 35.9. 16.9. 17.6. 18.6. 15.1. 24.7. 12.6. 10.5. 47.8. Group IV. 21.2. 35.3. 15.9. 11.0. 32.9. 15.9. 32.8. 12.4. 7.01. 46.0. 19.3. 30.5. 7.30. 8.29. 29.9. Group V. 23.4. 35.8. 17.4. 11.2. 31.9. 22.4. 27.6. 16.4. 15.3. 17.3. 24.4. 39.2. 11.7. 5.95. 18.4. -8.35 -5.42 -3.93 -1.23. -0.89. -4.75. 6.60. -4.22 -4.68. 0.40. -6.33 -4.64. 5.52. 4.58. 14.7. Diff (I and V). Panel E. Performance Measure: Average Yearly Stock Returns (Equally Weighted) Portfolio. COM. ENV. FIN. All. T.I. F.I. E.I. TTTM. All. T.I. F.I. E.I. TTTM. All. T.I. F.I. E.I. TTTM. Group I. 18.8. 25.7. 15.8. 12.8. 29.8. 22.1. 32.9. 14.7. 13.0. 17.2. 21.5. 30.4. 19.4. 13.4. 30.2. Group II. 19.6. 27.9. 14.1. 9.16. 29.4. 15.4. 22.9. 15.1. 7.25. 33.3. 17.7. 24.5. 17.6. 10.4. 29.1. Group III. 19.8. 26.8. 21.6. 12.3. 24.8. 20.9. 32.2. 21.6. 11.4. 25.7. 16.0. 23.2. 17.6. 7.66. 31.2. Group IV. 15.1. 32.1. 20.2. 5.77. 27.2. 19.3. 28.5. 18.1. 12.0. 48.2. 20.1. 29.2. 18.6. 13.8. 23.5. 26.0. 32.9. 18.7. 9.44. 28.7. 21.7. 29.1. 21.2. 5.78. 15.4. 24.0. 37.8. 18.0. 4.20. 25.9. -7.18 -7.27 -2.85. 3.36. 1.07. 0.34. 3.84. -6.51. 7.17. 1.80. -2.51 -7.46. 1.36. 9.22. 4.34. Group V Diff (I and V). Panel F. Performance Measure: Average P/E Ratio Portfolio. COM ENV. FIN. All. T.I. F.I. E.I. TTTM. All. T.I. F.I. E.I. TTTM. All. T.I. F.I. E.I. TTTM. Group I. 46.0. 32.3. 96.8. 43.3. 27.9. 33.6. 38.9. 177. 32.7. 27.1. 46.9. 22.4. 176. 36.9. 27.6. Group II. 63.6. 26.3. 112. 38.1. 402. 47.5. 36.4. 360. 55.3. 35.7. 30.4. 35.4. 50.5. 30.4. 15.0. Group III. 49.1. 27.6. 17.9. 44.7. 247. 45.2. 41.7. 13.8. 36.6. 443. 62.8. 30.2. 17.2. 50.6. 34.8. Group IV. 73.8. 111. 48.2. 50.8. 90.4. 108. 33.3. 38.5. 19.1. 68.9. 90.2. 98.2. 364. 67.7. 562. Group V. 85.2. 50.4. 366. 48.0. 47.9. 83.3. 104. 52.9. 79.7. 225. 90.0. 62.8. 134. 40.5. 227. -39.2 -18.1 -269 -4.63. -20.0. -49.7 -65.1. 124. -47.0. -198. -43.2 -40.4. 41.2. -3.61. -200. Diff (I and V). The idea of the above analysis is to use portfolio comparisons of longterm performance for various groups of firms. Under this framework, if performance indicator is stock returns, then comparing portfolio stock returns should consider risks associated with portfolio returns. Based on above.