Does it really pay to be green? Determinants and Consequences of Proactive Environmental Strategies *

Transcription

1 Does it really pay to be green? Determinants and Consequences of Proactive Environmental Strategies * Peter M. Clarkson UQ Business School The University of Queensland and Faculty of Business Administration Simon Fraser University p.clarkson@business.uq.edu.au Yue Li Joseph Rotman School of Management University of Toronto yueli@rotman.utoronto.ca Gordon D. Richardson Joseph Rotman School of Management University of Toronto gordon.richardson@rotman.utoronto.ca Florin P. Vasvari London Business School University of London fvasvari@london.edu Dated: September 8, 2006 * We acknowledge the financial support of the Canadian Academic Accounting Association and the Canadian Institute of Chartered Accountants. We thank the reviewer and participants at 2006 ASAC Conference for their comments. 1

2 Does it really pay to be green? Determinants and Consequences of Proactive Environmental Strategies Abstract Does it really pay to be green? Prior literature assumes that investments to protect the natural environment provide few financial benefits to firms. Recently some scholars argue that firms can benefit economically from pursuing a proactive environmental strategy, hence the argument it pays to be green. This argument is controversial since poor environmental performers have economic incentives to imitate good environmental performers and if that is true, we should not observe significant cross-sectional variations in corporate environmental performance in industries facing stringent environmental regulations over long periods. This study examines the relation between environmental performance and economic performance from a resource based view of the firm perspective. We hypothesize that firms pursuing a proactive environmental strategy: (1) possess financial resources and management capabilities that are not readily available to the other firms and (2) enjoy better financial performance in the subsequent periods. Using longitudinal data between 1990 and 2003 from four most polluting industries in the U.S., we show that firms experiencing a significant improvement in relative environmental performance over time ( repenters ) have higher Tobin s Qs, are more liquid, and spend more on R&D immediately prior to the improvement in the environmental performance. These firms also enjoy further improvement in Tobin s Qs, profitability, and liquidity, and sales growth in the subsequent periods. In contrast, firms experiencing a significant decline in relative environmental performance ( anti-repenters ) over time have, prior to the decline, lower Tobin s Qs, are less profitable, less liquid, more highly levered, have older assets and lower growth rates, and are less capital intensive. Overall, our results are consistent with the resource based view of the firm and indicate that although becoming green is associated with improvement in firm performance, such a strategy cannot be easily mimicked by all firms. 2

3 1. Introduction The costs of environmental protection for firms in the U.S. have increased substantially since the 1970s and are expected to increase even further in the future (Barbera and McConnell 1990; Chan- Fishel 2002). The magnitude of environmental protection costs is such that companies must integrate environmental protection into their business strategy (Christmann 2000). A growing body of the environmental management literature suggests that firms can gain sustainable competitive advantages by reducing the adverse impact of their operations on the natural environment. It is argued that a proactive environmental management strategy can enhance firm performance through process innovation and product differentiation (Porter and van der Linde 1995; Hart 1995). Although there has been some anecdotal evidence that links corporate environmental strategy and financial success, empirical studies have not established consistent evidence that a proactive environmental strategy enhances firm performance. Even in studies that find a positive association between environmental performance and financial success, it is not clear whether clean operations lead to financial success or financially healthy firms choose and can afford to operate clean (King and Lenox 2001; Bansal and Roth 2000; Griffin and Mahon 1997). Although the argument that it pays to be green is forceful, the specific mechanism under which industries and firms can improve their competitiveness through a proactive environmental strategy remains controversial. A recent survey of CFOs and top environmental officers from the 295 largest public companies in Canada indicates that less than half of the respondents believe environmental performance affects competitiveness or enhances shareholder value (Desjardins, Tabone, and Willis 2000). Conventional economics wisdom suggests that pollution abatement investments divert financial resources from productive investments that will not be recovered (Walley and Whitehead 1994). Furthermore, if a proactive environmental strategy leads to enhanced firm performance, it is difficult to explain why such a strategy is not widely duplicated by other firms in the industry. 1

4 Recent developments in management theory offer a theoretical framework to explain why a proactive environmental strategy may create a sustainable competitive advantage. The resource based view of the firm suggests that not all firms can benefit equally from a proactive environmental strategy (Hart 1995). In particular, firms pursuing a proactive environmental strategy are the ones with more financial resources and superior management capabilities (Christmann 2000). This framework predicts that proactive corporate environmental policies and financial success are endogenously determined. This study examines what factors affect a firm s decision to improve its relative environmental performance within the industry and whether pursuing a proactive environmental strategy enhances firm performance in the subsequent period. We explore the complex relation between corporate environmental performance and financial performance within the framework of the resource based view of the firm theory. We address the following two related empirical questions: What are the factors that affect the adoption of a proactive environmental strategy? Do changes in relative environmental performance enhance a firm s subsequent financial performance? We address these two empirical questions in the context of the four most polluting industries (Pulp & Paper, Chemicals, Oil & Gas, and Metals & Mining) in the U.S.. Since the decision to pursue a proactive environmental strategy is not directly observable, we characterize a firm s environmental strategy by changes in its relative environmental performance over time. Using longitudinal data between 1990 and 2003 from the four most polluting industries in the U.S., we show that firms experiencing a significant improvement in relative environmental performance over time ( repenters ) have better economic performance, greater cash flow, lower leverage, spend more on R&D, have greater growth, and higher Tobin s Q immediately prior to the improvement in the environmental performance. These firms also enjoy real economic benefits with improvements in profitability, cash flow, and Tobin s Q in the subsequent period. In contrast, firms experiencing a significant decline in relative environmental performance over time ( anti-repenters ) have lower profitability and cash flow, and higher leverage, spend less on R&D, experience lower 2

5 growth, and have a lower Tobin s Q, prior to the decline. They also experience a subsequent further decline in profitability, cash flow, and Tobin s Q. Thus, overall, our results are consistent with the resource based view of the firm and indicate that although becoming green is associated with an improvement in firm performance, such a strategy cannot be easily mimicked by all firms. The study contributes to the environmental accounting and management literature in several important ways. First, a main research issue in the environmental management literature is why some firms embrace ecologically responsive initiatives voluntarily, while others in seemingly similar circumstances merely attempt to comply with the environmental regulations with minimum spending and effort. By modeling the decision process and factors that affect the adoption of a proactive environmental strategy, we achieve a better understanding of the observed variation in corporate environmental performance. Second, this study provides a rigorous test of the resource based view of the firm by explicitly modeling the impact of financial resources and management capability on a firm s decision to improve its relative environmental performance within the industry. One potential reason previous studies in this area were inconclusive is that these studies do not consider the reciprocal impact of financial resources and the adoption of proactive environmental strategy on each other. Third, we develop a more reliable measure of proactive environmental strategies based on changes in relative environmental performance over time. Previous studies in this area rely mostly on management surveys or other subjective ratings to proxy for corporate environmental strategies. Using a broad sample of US firms from the four most polluting industries for the 1990 to 2003 period, we are able to measure changes in relative environmental performance over time as a proxy for corporate environmental strategy and to explore the subsequent changes in firm performance following these changes in relative environmental performance. The paper is organized as follows. Section 2 provides a literature review while Section 3 outlines our key hypotheses. We describe the sample and summary statistics of our data in section 4. Section 5 lays out the key research design and econometric models to test our hypotheses. Results are 3

6 presented and discussed in section 6. Section 7 concludes with a discussion of the implications for future research. 2. Literature Review Resource-based theory takes the perspective that valuable, costly to copy firm resources and capabilities provide the key sources of sustainable competitive advantage (Barney and Arikan 2005). The resource-based view of the firm posits that competitive advantage can be sustained only if the capabilities creating the competitive advantage are supported by resources that are not easily duplicated by competitors. Resources may include physical and financial assets as well as firm-specific assets such as employees skills and organizational processes. A firm s capabilities refer to its ability to accomplish specific value-added tasks with deployment of supporting resources. The resource-based view emphasizes resources that contribute to a firm s capabilities and are difficult to transfer or trade. Such resources can be difficult to replicate because they are invisible assets accumulated through past experience and learning-by-doing (Hart 1995). In addition, firms may also acquire socially complex resources that can support coordinated action engaged by a large numbers of employees or teams. Ultimately, firms that are able to accumulate resources to support unique capabilities will be able to enjoy a sustained competitive advantage (Russo and Fouts 1997). 1 Hart (1995) extends resource-based theory by considering the constraint of natural resources. He develops a conceptual framework, a natural-resource-based view of the firm, to characterize firms capabilities that facilitate three interrelated environmental strategies: pollution prevention, product stewardship, and sustainable development. Hart argues that the three environmental strategies can lead to competitive advantage in the form of improved manufacturing efficiency, enhanced reputation, and raising rival s costs by influencing future industry standards. Not all firms can realize these benefits in 1 We do not differentiate resource-based theory and the resource-based view of the firm in this paper. For a comprehensive review of resource-based theory and its origins, see Barney and Arikan (2005). 4

7 the same fashion. Pursuing these strategies implies both substantial investment and a long term commitment to the environment. Firms pursuing these strategies need sufficient financial resources and significant organizational capabilities in order to gain these competitive advantages. For instance, a pollution prevention strategy is people-intensive, focusing on continuous improvement through employee involvement to achieve well defined environmental objectives, rather than end of the pipe solutions. An environmental product stewardship strategy requires the integration of life-cycle-costing into a firm s product-development process. It implies an organizational ability not only to coordinate functional groups within the firm, but also to communicate across functions, departments, and organizational boundaries. Firms pursuing a sustainable development strategy must have an organizational wide dedication to a shared vision of the future. Creating such a shared vision deep into the management ranks requires strong moral leadership and an empowering social process inside the firm. Given the difficulty in generating such a consensus, shared vision becomes a firm-specific resource that cannot be easily duplicated by all firms. Klassen and McLaughlin (1996) try to disentangle the relationship between environmental performance and financial performance using event study methodology. They investigate market reactions to the announcements of positive environmental events such as environmental awards and negative environmental events such as oil spills or other environmental crises. The underlying assumption is that market reactions capture investors perception about the future financial performance as a result of these environmental events. Using a sample of 140 positive environmental events (involving 96 firms) and 22 negative environmental events (involving 16 firms), identified from the NEXIS newswire service database, they find that investors reward firms with positive environmental events and penalize firms with negative environmental events. Overall, their findings are consistent with positive environmental events leading to an improved perception about future financial performance. However, it should be noted that the incidental nature of environmental events in their study may not be a reliable proxy for corporate environmental strategies. 5

8 Hart and Ahuja (1996) examine whether emission reduction enhances a firms operating and financial performance in concurrent and subsequent periods. Using a sample of 127 S&P 500 firms, they find that the percentage change in emission between 1988 and 1989 is positively associated with three measures of firm performance (returns on sales, assets, and equity) in 1990, 1991 and 1992, but not in The authors interpret their findings as suggesting that it does pay to be green, although the economic benefits of becoming green appear to occur one to two years later. It should be noted that the study does not directly test the resource-based view of the firm as the empirical analyses do not control for resource related variables. In addition, environmental performance in this study is measured as the percentage change in emissions between 1988 and This limited time period does not capture continuous improvement in environmental performance over time nor does it reflect the environmental performance of each sample firm relative to its industry peers. 2 Dowell, Hart, and Yeung (2000) investigate whether multinational firms adopting a single stringent global environmental standard have higher Tobin s Q, as compared to those multinational firms defaulting to less stringent or poorly enforced host country standards. Their sample consists of 89 multinational S&P 500 firms with production operations in countries with GDP below $8,000 per capita. Their environmental performance indicator is based on a survey, conducted by the Investor Responsibility Research Center during the 1994 to 1997 period, which asks whether firms adopted internal environmental standards that exceed any national standards or only adhered to local standards in the host countries where environmental regulations are in general less stringent than the U.S. counterpart. The authors find that adopting more stringent environmental standards is positively associated with firms Tobin Q, after controlling for other variables known to affect Tobin s Q. However, such association tests do not permit causal assertions of the link between environmental strategy and Tobin s Q, something we seek to do in this study with longitudinal data. 2 Hart and Ahuja (1996, p.36) acknowledge the limitations caused by having only one year of emission reduction data and call for longitudinal measures of both environmental and firm performance measures, something we do in this study. 6

9 King and Lenox (2001) use a large sample of 652 manufacturing companies over the period 1987 to 1996 to further explore the relationship between environmental performance and financial performance. They examine whether relative environmental performance, measured as weighed average of toxic release normalized by firm size (number of employees) affects Tobin s Q. Consistent with Dowell et. al., they find that high relative environmental emission at the firm level is negatively associated with Tobin s Q. Again, these association findings do not permit causal assertions. 3 Christmann (2000) examines whether environmental strategies create competitive cost advantages and whether a firm s capability for process innovation and implementation affects its ability to benefit from implementing environmental strategies. Based on survey data from CEOs of 88 chemical companies, she concludes that a firm s capability in product and process innovation is a rare, valuable, non-substitutable, and imperfectly imitable asset and that the heterogeneity of this capability across firms explains why not all firms can obtain competitive cost advantage from proactive environmental strategies. In summary, although the resource-based view of the firm provides a promising conceptual framework that links firms environmental strategies to their financial performance, the findings in the existing empirical studies are largely mixed and inconclusive. In particular, there is limited evidence in studies of large cross-industry samples to support the causal relationship between proactive environmental strategies and improved financial performance. One problem in the existing studies is that environmental strategies are not directly observable. Using management survey and internal environmental policies to assess corporate environmental strategies can be unreliable due to their subjective nature. In addition, actual emission data in one single period can be difficult to interpret outside the industry context and it is well known that emission data evolve over time as industry wide environmental performance improves (Clarkson et al 2004). This study differs from existing studies in 3 It is interesting to note that when lagged Tobin s Q is included in the regression, environmental performance is no longer significant in explaining firms Tobin s Q. This finding further raises the question about the direction of causality in the relationship between environmental performance and financial performance. 7

10 the following ways. First, we employ a large sample from multiple industries to explore the relationship between environmental performance and financial performance over an extended time period. Thus, the results from this study apply to a broader industry context. Second, the longitudinal data allow us to develop a more reliable proxy for proactive environmental strategies. Specifically, we can identify firms that experience either significant improvement or decline in environmental performance over our sample period of 1990 to Focusing on these two subsets of firms (repenters and anti-repenters) may increase the power of our statistical tests and provide sharper insights into why some firms chose to pursue proactive environmental strategies while others didn t during the sample period. Finally, our empirical model is more consistent with the resource-based view of the firm because we examine both the determinants and consequences of adopting a proactive environmental strategy. In addition, in order to explain a firm s decision to adopt a proactive environmental strategy, we consider both financial resources and proxies for management capabilities in our empirical model. 3. Hypotheses Development and Variable Selection 3.1 Hypotheses Pollution and environmental degradation are externalities because their effects are not typically reflected in the measurement of firm profits. Pollution abatement beyond the existing environmental regulation requires significant resources and managerial effort. Economic theory suggests that in order to derive competitive advantages from proactive environmental policies, firms need to create value and then capture it from customers, suppliers, or other economic agents (Reinhardt 1999). Proactive environmental strategies are sensible only if they lead to enhanced competitiveness and increase the expected value of the firm. Porter and van der Linde (1995) argue that environmentally proactive firms may enjoy the benefits of process innovation which can reduce both input costs and pollution. Reinhardt (1998) outlines the conditions under which companies can deliver shareholder value with improved environmental performance by producing environmentally preferable products. Other 8

11 economic benefits associated with superior environmental performance include raising rivals costs and risk management due to uncertainty of future environmental regulations (Reinhardt 1999). Thus, a firm s discretionary effort to pursue a proactive environmental strategy, as evidenced by improvements in its relative environmental performance over time, should be associated with enhanced competitiveness and financial performance in future periods. Our first hypothesis follows: H 1 : A sustained improvement in relative environmental performance over time leads to a subsequent improvement in a firm s financial performance. Although the environmental economics literature suggests that it is possible for firms to benefit economically from proactive environmental strategies, these benefits may not be equally accessible to all firms in an industry. If firms can simultaneously improve environmental performance and enhance competitiveness, we would not observe the wide variations in corporate environmental performance manifested in many existing studies (e.g., Li, Richardson and Thornton 1997; King and Lenox 2001; Clarkson et al 2004). A firm s ability to achieve such economic benefits depends on a number of factors such as internal resource constraints, management capabilities, industrial structure, the government regulatory framework, and the nature of the competition. The resource based view of the firm theory (Hart 1995) argues that firms must have critical financial resources and superior management capabilities to implement proactive environmental business strategies. Thus, corporate environmental performance and financial success are interrelated and the best managed firms with the most financial resources are more likely to pursue proactive environmental strategies. Our second hypothesis formalizes the prediction of the resource-based view of the firm and tests the interrelation between corporate environmental performance and financial resources and management capabilities: H 2 : Firms with a sustained improvement in relative environmental performance over time possess critical financial resources and management capabilities in the prior periods. 9

12 3.2 Variable Selection To examine the determinants of the decision to adopt a proactive environmental strategy (H 2 ), we require proxies for both a firm s financial resources and its management capabilities in the prior period. In order to pursue a proactive environmental strategy, firms must have the necessary financial resources. Herein, we argue that the level of financial resources available to a firm increases with profitability and cash flow, and decreases with leverage. In each instance, we use lagged measures of these firm characteristics in our empirical analysis based on the notion that resources available at the beginning of a fiscal period dictate spending for that period. According to the resource based view of the firm, firms must also have superior management commitment and capabilities in order to pursue proactive environmental strategies. 4 Firms with weaker management are more likely to be minimum compliance firms that spend on the environment in a reactive fashion as dictated by new regulation. While management talent is not directly observable, we argue that an innovative management team is more likely to be pursuing knowledge investments even if unrelated to the environment. We use R & D intensity as a surrogate for the propensity for innovation, and sales growth and Tobin s Q as manifestations of that proactive investment in intangibles. Thus, we posit that R & D intensity, sales growth, and Tobin s Q serve as proxies for unobservable management talent or capability with management capability varying directly with each proxy. Again, we use lagged measures of these proxies to capture the endowment of management talent going into the period in question. Turning to the consequences of adopting a proactive environmental strategy (H 1 ), we would expect that a sustained improvement in environmental performance will lead to enhanced firm profitability and cash flow in the future if indeed it pays to be green. Indirectly, enhanced firm profitability should also manifest itself in a higher Tobin s Q in future periods. 4 As an example of how management capability and commitment may affect corporate environmental strategies, see a recent article on Wal-Mart CEO Lee Scott in the Wall Street Journal: Can Wal-Mart Sustain a softer Edge? (Feb. 8, 2006 by Alan Murray). 10

13 4. Sample and Descriptive Statistics The sample data consist of all firms in the Pulp & Paper (SIC = 26), Chemicals (SIC = 28), Oil & Gas (SIC = 29), and Metals & Mining (SIC = 33) industry sectors for which environmental performance, stock price, and financial data are available during the period The study period ends in 2003 because environmental data are published by the EPA with a two-year lag. Table 1 presents a frequency distribution of the sample observations. As revealed in Panel A, the sample is comprised of observations on 242 distinct firms, with the greatest number of sample firms being in the Chemicals and Metals & Mining industry sectors. Panel B presents a frequency distribution of firm-year observations by year and industry. As can be seen, there are slightly fewer observations in the last five years of the study period ( ) but otherwise the sample size is relatively consistent over time. In sum, the final sample consists of 2,376 firm-years of data on 242 distinct firms with 412 firm-year observations from the Pulp & Paper sector, 995 from the Chemicals sector, 261 from the Oil & Gas sector, and 708 from the Metals & Mining sector. Table 2 presents distributional statistics for the sample firms environmental performance, the measures of the firm s financial resources and management capabilities discussed in Section 3.2, and for the following three additional control variables that we include in our econometric models, firm size, equipment newness, and capital intensity. We include firm size as a control in keeping with prior environmental studies (Chen and Metcalf 1980). We include equipment newness as a control since newer equipment is expected to employ newer and less polluting technologies and as such, it may potentially show up in our measures of relative environmental performance in a particular period, although its impact on a firm s ability to change its environmental performance over time (our focus) is less clear. For similar reasons, firms with higher sustaining capital expenditures are expected to have 5 We did not include utilities, another major polluting industry in the U.S., in this study for two reasons. First, utilities are heavily regulated, which means their financial performance may be difficult to interpret in a cross-sectional study. Second, our readings of 10Ks indicate that R&D spending data, a key variable in our analysis, for this industry is minimal, and for most firms non-existent. 11

14 newer equipment and thus we also control for capital intensity. Panel A presents statistics for the pooled sample of 2,376 firm-years whereas Panel B presents statistics for the sample partitioned by environmental performance. In terms of a firm s environmental performance, we initially document its pollution propensity (PP), measured as toxic release inventory (TRI) in pounds per thousand dollar cost of goods sold (TRI/COGS). For ease of interpretation, for much of the subsequent empirical analysis, we then focus on a firm s environmental performance (EP), defined as the inverse of its PP. 6 In this regard, firms with higher measures of EP (and thereby lower relative pollution propensity measures) are viewed as better environmental performers. The way in which we measure each of the remaining variables is as follows: TA RNOA CF LEV RDIN GRTH Q NEW CAPIN = total assets ($ millions); = return on net operating assets, measured as net operating income divided by beginning period net operating assets; = liquidity, measured as net cash flow from operations deflated by sales; = leverage, measured as total debt divided by total assets; = research and development intensity, measured as R&D expenses divided by sales; = growth, measured by change in sales divided by beginning of period sales; = Tobin s Q, measured as the sum of market value of equity plus the book value of total debt and preferred shares all divided by total assets; = equipment newness, measured as net property, plant, and equipment divided by gross property, plant, and equipment; and = capital intensity, measured as capital expenditures divided by sales. Panel A of Table 2 indicates that there is considerable diversity in the characteristics of the sample. To begin, the pollution propensity measure (PP) has an interquartile range (Q1 to Q3) of to 3.582, with a mean (median) value of (1.211).Thus, on average, our sample firms have pounds of toxic release per $1,000 of cost of goods sold. Firms also vary considerably in terms of size, 6 For observations where PP is equal to zero, we arbitrarily set EP equal to the maximum EP for firms in the same year and industry. Since it is possible that PP observations with a value of zero for a given year represent missing data rather than zero pollution output, we alternatively treat these observations as missing, finding results based on this alternative treatment (not reported) to be qualitatively identical to those presented in the Tables. 12

15 profitability, and cash flow. The total assets of our sample firms (TA) has an interquartile range of $ million to $5.432 billion with a mean (median) value of $5.194 billion ($1.336 billion), the return on net operating assets (RNOA) has an interquartile range of 8.9 percent to 23.8 percent with a mean (median) value of 18.7 percent (14.8 percent), and the ratio of cash flow to sales (CF) has an interquartile range of 5.4 percent to 15.6 percent with a mean (median) value of 9.9 percent (9.0 percent). On average, firms are somewhat levered with a mean (median) debt-to-assets ratio of (0.289). Firms also vary in terms of the ratio of R&D to sales with RDIN having an interquartile range of to and a mean (median) value of (0.015). Realized growth (GRTH) has an interquartile range of to with a mean (median) value of (0.056). The firms Tobin s Q (Q) has an interquartile range of to with a mean (median) of (1.133). Finally, the equipment of newness (NEW) and capital intensity (CAPIN) measures have interquartile ranges of to and to 0.107, respectively. Panel B of Table 2 presents descriptive statistics for the sample partitioned on the basis of pollution propensity as measures by PP. Herein, the sample has been ranked on PP within year and industry, and the descriptive statistics presented are for the firm-years ranked in the two extreme quartiles (the so-called best and worst PP categories). As revealed, there are several dimensions along which these two PP categories differ. Consistent with prior studies (Hart and Ahuja, 1996; Dowell, Hart, and Yeung, 2000; King and Lennox, 2001), the best environmental performers (lowest PP) are more profitable (RNOA) (p = 0.010) and have a higher Tobin s Q (p < 0.001). The results confirm a joint positive association between environmental performance and financial performance. In addition, the best environmental performers are larger (TA) (p < 0.001), have greater cash flow (CF) (p = 0.069), and experience higher growth (GRTH) (p = 0.082), but have a lower capital intensity measure (CAPIN) (p = 0.029). There appears to be no difference between the two sub-samples in terms of the remaining attributes, leverage, R&D intensity, and equipment newness. However, caution must be exercised when interpreting these results because they are simply based on the aggregate of all firm- 13

16 years within the two extreme quartiles and do not discriminate between firms with stable environmental performance patterns and those for which their relative performance changes over the study period. Table 3 presents a pair-wise correlation matrix for the pooled sample where all variables have been ranked within industry and year. As noted above, for ease of interpretation, we invert our pollution propensity measure, PP, and base this and all subsequent analyses on the inverted measure, environmental performance (EP). Thus, the highest ranked firms are those with the best environmental performance (EP) (lowest values of TRI / COGS). Consistent with Panel B of Table 2, from a univariate perspective there is a positive and significant association between EP and financial performance as measured by RNOA and Q. In addition, there are also positive and significant associations between EP and each of lnta, CF, LEV, and NEW. For the various explanatory variables in the econometric models, none of the correlations exceed 0.5. Thus, the threat of multicollinearity appears limited (Gujarati 1995) Experimental Design In this study we seek insights into both the determinants and consequences of a firm s decision to change its environmental strategy by exploiting the longitudinal nature of our data. The experimental design we employ is a matched pair design wherein we compare characteristics and attributes of firms for which their environmental performance changed appreciably during the study period with those of industry-matched firms which exhibited stable environmental performance throughout the entire study period. We argue that significant changes in relative environmental performance over time within industry must result from a change in a firm s discretionary efforts in environmental protection, hence a change in its environmental strategy. By focusing on firms for which there has been an appreciable 7 For example, Gujarati (1995) argues that multicollinearity will not be a problem, even with high pairwise correlations (in excess of 0.85), if there is sufficient variability in the values taken by the regressors. As further support for the view that multicollinearity is not a threat in any of our regressions, we consider the Variance Inflation Factor (VIF). The VIFs (not reported) also support the view that we should not be concerned about multicollinearity. 14

17 change in relative environmental performance and comparing them with firms with stable environmental performance, we expect to capture the impact of a change in environmental strategies on firm performance in the subsequent period. Since this methodology allows us to infer approximately when a change in environmental strategy takes place, the power of both determinants and consequences analyses is enhanced. In a traditional association study relating EP to financial performance, the researcher is not sure when an environmental strategy change occurred (i.e., it could even have occurred a decade or more earlier) but can only observe current EP category membership. This lowers the validity of causal inferences concerning the determinants and consequences of EP choices. Since we pool firms from across four industry sectors, we rank all variables within industry and year, and base our analyses on these rank values. Finally, as noted above, the environmental performance measure (EP) has been defined so that the highest ranked firms are those with the best environmental performance (lowest values of TRI / COGS). 5.1 The Matching Procedure We begin by creating five environmental performance change partitions, stable good environmental performers (SG), stable poor environmental performers (SP), repenters (R), antirepenters (AR), and variable environmental performers (VP). Specifically, we rank our sample firms based on their TRI/COGS within year and industry and then classify the firms as: stable good performers (SG) = firms ranked in the best two quartiles during the entire study period; stable poor performers (SP) = firms ranked in the worst two quartiles during the entire study period; repenters (R) = firms whose ranking improved at least two quartiles over a three year span at some point during the study period and then stabilized within the best two quartiles for remainder of the study period; 8 anti-repenters (AR) = firms whose ranking worsened at least two quartiles over a three year span at some point during the study period and then stabilized within the worst two quartiles for the remainder of the study period; 8 For the repenter classification, firms moved from the third or fourth quartile to the first quartile, or from the fourth quartile to the second. For the anti-repenter classification, firms moved from the first to the third or fourth quartile, or from the second quartile to the fourth. 15

18 and variable performers (VP) = all remaining firms. 9 Table 4 presents a frequency distribution by industry and environmental performance change partition. As revealed, of the 242 sample firms, 70 are classified as stable good environmental performers (SG), 67 as stable poor environmental performers (SP), 43 as repenters (R), 24 as antirepenters (AR), and the remaining 38 as variable environmental performers (VP). Table 4 also reveals that the proportions of firms in the five categories are relatively stable across the industries. Based upon the reported Chi-square test statistic, independence between environmental performance change partition and industry sector cannot be rejected at conventional levels of significance (χ 2 = 5.642, p = 0.933). Based upon these classifications, we then consider two separate comparisons. First, we compare the characteristics and attributes of repenters with stable poor performers. By construction, both groups were ranked in the worst fifty percent of the sample firms within their industry sector in terms of environmental performance at the beginning of the study period (1990). However, at some subsequent point during the study period, the relative environmental performance of the R subsample firms improved significantly so that they moved into the best fifty percent whereas the SP subsample firms remained in the worst fifty percent. Following this, the second comparison we make is between anti-repenters and stable good environmental performers. These two groups started in the best fifty percent of sample firms within their industry sector but over the study period, the AR subsample firms dropped into the worst fifty percent while the SG subsample firms remained within the best fifty percent. Thus, in both instances, a comparison of the characteristics and attributes of the two related subsamples provides insights into how they differed both before and after relative environmental 9 One concern relating to this classification strategy is that changes in a firm s environmental performance may have arisen artificially because of merger and acquisition (M&A) activity. To ensure that our results are not being influenced by these firms, we searched the COMPUSTAT database finding that two sample firms classified as repenters and one firm classified as an anti-repenter had engaged in M&A activity during the study period, with the M&A activity roughly coinciding with the change in their environmental performance in each instance. We then repeated all analyses are dropping these observations, finding our results to be qualitatively unaffected. 16

19 performance changed for the R or AR firms, as appropriate. We argue that this use of the SP and SG firms as benchmarks against which to judge the R and AR firms, respectively, is reasonable because, all else held equal, these benchmark firms could also have changed their environmental performance but opted not to. Finally, we match and align firms in the following fashion. First, we define the event year (t 0 ) as the year in which the industry ranked environmental performance of a firm classified as either R or AR started to change. We then identify the R and AR firms (the treatment firms) for which complete data are available over the seven-year period, t -3 to t +3. Lastly, we match each treatment firm with a control firm drawn from the same industry and from within the appropriate stable classification (i.e., R with SP and AR with SG) after ensuring that the matched firm has complete data available over the same seven year interval. 10 For each matched pair, the selected control firm is the one which has the closest environmental performance (EP) measure to that of the treatment firm in year t -1. Our aim is to have no difference in EP for treatment relative to control firms in the year prior to a sustained improvement or deterioration. This enables us to infer when a change in environmental strategy takes place. 11 This procedure results in 41 matched repenter / stable poor environmental performance firm pairs and 23 matched anti-repenter / stable good environmental performance firm pairs for which complete data are available over the seven-year period, t -3 to t , 5.2 Determinants and Consequences of Environmental Performance We begin by comparing firm characteristics and attributes of the matched pairs on a year-byyear basis over the seven-year interval, t -3 to t +3. Herein, this univariate analysis should provide 10 If, for example, the relevant interval for a given R firm is 1992 to 1998 since its environmental performance started to change in 1995, the matched SP firm must also have complete data available for the period 1992 to Standard statistical tests (not reported) reveal that there is no statistical difference in mean EP t-1 rank or mean raw EP t-1 values for either the R / SP or the AR / SG pairs. 12 Clearly, the choice of the appropriate interval involves a tradeoff between sample size and having a sufficiently long time series of observations to draw inferences about how the characteristics and attributes of the various subsamples change over time. The cost associated with selecting a seven-year interval is the loss of two R / SP pairs and one AR / SG pair. If the interval is shortened to five years (t -2 to t +2 ), only one of the R / SP pairs is recovered and results are qualitatively identical to those reported based upon the seven-year interval. 17

20 preliminary insights into how the matched firms differ and also into whether there are differences in the way in which they change over time, and thereby into the determinants and consequences of our sample firms environmental choice. Following this, in order to examine the determinants of environmental performance choice, we consider the following logistic regression model: TREAT t = β 0 + β 1 RNOA t-k + β 2 CF t-k + β 3 LEV t-k + β 4 RDIN t-k + β 5 GRTH t-k + β 6 Q t-k + β 7 lnta t-k + β 8 NEW t-k + β 9 CAPIN t-k + β 10 EP t-k + ε (1) For our consideration of the decision to repent, we base the analysis on the sample of 41 R / SP matched pairs and set TREAT set equal to 1 if the sample firm is classified as a repenter and 0 if classified as a stable poor environmental performer. Thus, TREAT serves as a proxy for decision to adopt a proactive environmental strategy by the R/SP matched firms. Alternatively, for our consideration of the decision to anti-repent, we base the analysis on the sample of 23 AR / SG matched pairs and set TREAT set equal to 1 if the sample firm is classified as an anti-repenter and 0 if classified as a stable good environmental performer. For this group of firms, TREAT can be interpreted as a firm s decision to abandon its previous proactive environmental strategy. All remaining variables are as previously defined. In undertaking this analysis, lags of one and two years will be considered (k = 1, 2). 13 Notwithstanding the matching criterion we impose, we include the lagged measure of EP in the model as an additional control to ensure that results are not being driven by the remaining (residual) differences in lagged environmental performance between the matched pairs. Within the context of the R model, we expect that relative to the stable poor environmental performers, the repenter firms should have a higher endowment of financial resources going into the current period as proxied by the lagged values of RNOA, and CF, and lower leverage as proxied by LEV. Similary, repenter firms should have, relative to stable poor environmental performers, a higher 13 We also considered models with three-year lags (i.e., k = 3). The results for these models (not tabulated) are considerably weaker that those based on one- and two-year lags with only size and R&D intensity significant in the repenter model and only growth significant in the anti-repenter model. 18

21 endowment of management capability going into the current period as proxied by the lagged values of RDIN, GRTH, and Q. Thus, we would predict the following signs: RNOA (β 1 > 0), CF (β 2 > 0), LEV (β 3 < 0), RDIN (β 4 > 0), GRTH (β 5 > 0), and Q (β 6 > 0). Within the context of the AR model, we expect all relations to have the sign opposite to those of the repenter model. Finally, in order to examine the consequences of environmental performance choice, we compare mean values of one-year changes in rank for the economic measures (Q, RNOA, and LIQ) over the period t -1 to t +3. For the R model, we predict a positive sign of association between EP rank improvement and subsequent financial performance rank improvement. We argue that the difference in profitability between the repenter firms and the benchmark firms can be attributed to improvement in relative environmental performance after controlling for other profit or value drivers. We make analogous directional predictions for the AR model. If environmental performance choice has consequences for subsequent economic performance, we would expect improvements for the repenter firms relative to their matched stable poor counterparts and conversely, declines for the anti-repenter firms relative to their matched stable good counterparts. Finally, in conjunction with this analysis, we compare one-year changes in rank for CAPIN and RDIN over the same period since these also represent potential profit drivers and thereby potential confounds. 6. Empirical Results 6.1 Year-by-Year Univariate Comparisons Year-by-year univariate comparisons for firm performance as measured by RNOA and Q are presented in Table 5. Given H 1 and H 2, we argue that there is two-way causality relationship between environmental performance and financial performance. Specifically, we argue that financial performance creates the resources to spend on the environment with such spending in turn leading to enhanced financial performance if it is indeed expected positive NPV. For example, for the R / SP 19

22 matched pairs, we expect to observe a pattern where R firms are more profitable than SP firms prior to the year of pronounced rank change (i.e., they have the deep pockets to adopt proactive environmental strategies). Further, we expect the separation in profits to increase subsequent to the year when we observe a marked change in relative EP ranking (t 0 ) compared to before. By similar reasoning, we would expect the opposite for the AR / SG matched pairs. In Panel A of Table 5, we present the results for the R / SP matched pairs. Focusing on RNOA, the panel reveals that the mean intra-industry percentile of RNOA for the R firms in year t -3 is where year t 0 is the first year of a marked change in relative EP percentile. The corresponding mean intra-industry percentile for the SP firms in year t -3 is The difference in the mean percentiles for year t -3 is (p = 0.052). The corresponding differences for years t -2 and t -1 are (p = 0.186) and (p = 0.018), respectively. This pattern is consistent with the notion that profitability differences precede improvements in environmental performance. However, the mean differences in years t 0, t +!, t +2, and t +3 are (p = 0.001), (p = 0.012), (p = ), and (p = 0.006), respectively, differences that are all greater than those in the pre-event period and monotonically increasing over the post-event period. Thus, it would appear that a marked improvement in environmental performance precedes enhanced profitability differences relative to the no-change group. More formally, the F-statistic for the comparison of mean percentile differences over the period t -3 to t -1 with those for the period t +1 to t +3 of (p < 0.001) confirms that, for the R / SP matched pairs, the difference in mean percentiles has increased significantly from the pre- to the post-event period. Similar patterns are apparent for an alternative measure of financial performance, Tobin s Q. Here the difference metrics for the three years prior to t 0 are , , and , respectively. The difference increases to in year t 0, the year of pronounced EP rank change, consistent with an efficient market that rapidly impounds the benefits of green investments. The corresponding difference metrics for the three years subsequent to t 0 are , , and , respectively, consistent with a widening in the gap in Q between the two groups subsequent to a pronounced change 20