Abstract
Previous theoretical research provides opposing arguments regarding the effect of environmental regulation on profitability. This study provides empirical evidence on this debated effect by testing the “strong” version of the Porter hypothesis. We employ panel data analysis to examine the effect of water regulation, as measured by permitted wastewater discharge limits, on the profitability of publicly held firms operating within the chemical manufacturing industries. We find that tighter water regulation meaningfully lowers profitability. By reinterpreting profitability in terms of sales and costs, the results demonstrate that tighter water regulation increases costs conditioned on a given level of sales. (JEL K23, Q52)
I. Introduction
Recent economic research posits that properly designed environmental regulation motivates firms to innovate, which ultimately increases profits (Porter 1990, 1991; Porter and van der Linde 1995; van der Linde 1993). This claim represents the Porter hypothesis, which many economists criticize, arguing that environmental regulation undermines competitive firms’ abilities to voluntarily pursue profit-maximizing opportunities (Palmer, Oates, and Portney 1995).
The controversy described above proves important for our economy. As a share of U.S. gross domestic product, total expenditures for pollution abatement and control were approximately 1.8%fromthe mid-1970s to the mid-1990s (Vogan 1996).1 A proper evaluation of the ultimate effect of abatement and control expenditures on the U.S. economy requires an understanding of whether the efforts associated with the expenditures facilitate or undermine profitability.
The Porter hypothesis asserts that environmental regulation eventually leads to greater profitability. As suggested in the following quotation, Porter and van der Linde (1995, 97–98) introduce their argument by indirectly attributing competitiveness to profitability in the context of lower costs and higher revenues: “Competitiveness at the industry level arises from superior productivity, either in terms of lower costs than rivals or the ability to offer products with superior value that justify a premium price.” Jaffe and Palmer (1997, 610) describe the focus on profitability as the “strong” version of the Porter hypothesis: “The shock of a new regulation may therefore induce [firms] to broaden their thinking and to find new products or processes that both comply with the regulation and increase profits.” Our study empirically tests this strong version of the Porter hypothesis: tighter environmental regulation improves firms’ profitability.
Ideally, an empirical test of the Porter hypothesis would assess all of the relevant dimensions: (1) whether regulations lead to some forms of innovation, (2) whether those forms of innovation produce measurable cost savings or revenue enhancements (“innovation offsets”), and (3) whether those innovation offsets increase profits. However, data limitations and other considerations prevent implementation of this ideal empirical test.
By testing the strong version of the Porter Hypothesis, our study contributes to the economic literature that examines various consequences of environmental regulation. In order to deliver our contribution, we study the effect of Clean Water Act regulation (hereafter “clean water regulation”) on profitability, as measured by the return on sales (i.e., profits divided by sales) of publicly held firms in the chemical manufacturing industries. As our measure of regulation, we use permitted wastewater discharge limits, which are imposed on individual facilities. To strengthen our analysis, our study draws upon a panel data set. Thus, we are able to control more completely for heterogeneity across firms and exploit both interfirm and intrafirm variation.
Our empirical results indicate that a negative relationship exists between stringent clean water regulation and publicly held firms’ profitability in the chemical manufacturing industries. In particular, based on annual profitability, our estimates imply that a 10% tightening of the permitted discharge limit decreases return on sales by 1.7%. With 90% confidence, this impact lies between 0.8% and 2.7%. By examining both quarterly data and annual data, the analysis connects clean water regulation to both short-term profitability fluctuations and long-term profitability levels. The results are strongly robust to the type of connection examined.
II. Relationship between Environmental Regulation and Profitability
Economic research addressing the effect of environmental regulation on profitability is largely theoretical. Two conflicting arguments exist. Porter and van der Linde (1995) argue that properly designed and implemented environmental regulation removes organizational inertia and ultimately improves profitability.2 Innovation and improved resource productivity are the mechanisms through which this relationship unfolds. As long as firms perceive their production processes and products as elements in a dynamic setting rather than a static setting, firms seize regulation as an opportunity to invest in technologies and techniques that not only reduce strains on the environment but also increase the efficiency of production processes and/or improve the quality of products. The results are decreased production costs and/or increased revenues.
According to Porter and van der Linde (1995, 101), regulation can lead to greater profitability because it induces “innovation offsets [that] can exceed the costs of compliance.” Properly designed regulation yields a spirit of innovation in a firm to focus on proactive rather than reactive approaches to compliance. Innovation, in turn, produces offsets to the costs of compliance that reduce the net cost of compliance and may yield a net benefit.3 Porter and van der Linde (1995) support their argument with a collection of case studies in which stringent environmental regulation improves polluting firms’ profits. This argument has become known as the Porter hypothesis.
From an alternative perspective, Mohr (2002) employs a general equilibrium framework with external economies of scale in production and discrete changes in technology to demonstrate that endogenous technical change supports the feasibility of the Porter hypothesis.
Palmer, Oates, and Portney (1995) question the validity of the Porter hypothesis. In particular, they reject Porter and van der Linde’s (1995) assertion that regulation removes organizational inertia by providing firms with information and incentives that competitive markets somehow fail to provide. 4 Instead, they posit that firms voluntarily seek profit-maximizing opportunities regardless of regulation. Rather than a catalyst, environmental regulation serves to constrain firms’ abilities to pursue profit-increasing opportunities. As one specific consequence, firms facing more stringent regulation incur higher treatment costs. As the most general consequence, firms facing more stringent regulation are required to commit greater amounts of resources to uses that are neither productive nor profit increasing.5
In addition to these theoretical studies, other studies empirically examine relationships described by the Porter hypothesis. These studies fall into five sets. One set assesses the effect of environmental regulation on firms’ competitiveness, innovation activities, or productivity. Jaffe et al. (1995) examine the effects of environmental regulation on the competitiveness of U.S. manufacturing industries. They conclude that there is little evidence to support the hypothesis that regulation has had a large adverse effect on competitiveness. Jaffe and Palmer (1997) and Popp (2006) examine innovation; they both find a positive relationship between environmental regulation and innovation. Berman and Bui (2001a, 2001b) and Greenstone (2002) study the effect of air quality regulation on employment, productivity, or output. Berman and Bui (2001b) find no relationship between regulation and employment, while Greenstone (2002) finds a negative relationship. Berman and Bui (2001a) find that productivity rose sharply despite local air quality regulation, while Greenstone (2002) finds output losses due to regulation. Managi et al. (2005) explore the interactions among environmental regulation, technological innovation, and productivity growth in the offshore oil and gas industry, while testing two versions of the Porter hypothesis: (1) productivity measured in terms of market outputs only (e.g., oil and gas production) and (2) productivity measured in terms of both market outputs and nonmarket outputs (i.e., pollutant emissions). They find support for the latter version of the Porter hypothesis. Brown and Wilcoxen (2008) assess whether federally mandated pollution abatement investment leads to less-productive investment.
A second set of empirical studies examines the effect of environmental regulation on firms’ costs. Gray (1987), Hazilla and Kopp (1990), and Jorgenson and Wilcoxen (1990) present good examples. These studies generally find that regulation raises firms’ costs. While these studies contribute to our understanding, Porter and van der Linde (1995) argue that these studies’ results are subject to bias because the studies assume away innovation benefits.
A third set of empirical studies examines the effect of environmental regulation on firms’ location or investment decisions. In general, the empirical results are mixed. McConnell and Schwab (1990) and Levinson (1996) find little evidence that environmental regulation affects industry location decisions. Becker and Henderson (2000) investigate the effects of air quality regulation on firms’ decisions regarding plant birth, plant location, plant size, and investment patterns; they find fewer plant births for polluting industries in nonattainment areas and a shifting industrial structure toward less-regulated single-plant firms. List et al. (2003) and Dean, Brown, and Stango (2000) find further evidence of a negative relationship between environmental regulatory stringency and plant births.
A fourth set of empirical studies examines the effect of environmental performance, rather than environmental regulation, on firms’ financial performance. The studies improve our understanding of the Porter hypothesis to some extent even though they do not directly assess the effect of environmental regulation. Regardless, we use the studies to guide our empirical analysis. As the best examples, Russo and Fouts (1997), Khanna and Damon (1999), Konar and Cohen (2001), and Filbeck and Gorman (2004) provide conflicting evidence regarding the effect of environmental performance on financial performance. Russo and Fouts (1997) and Konar and Cohen (2001) find a positive relationship between environmental performance and financial performance. Filbeck and Gorman (2004) find a negative relationship between environmental performance and financial returns. Khanna and Damon (1999) find a negative relationship between environmental performance and current financial performance, but a positive relationship between environmental performance and long-run expected financial performance.
In contrast to the fourth set, the last set of empirical studies directly examines the effect of environmental regulation on financial performance. Brännlund, Färe, and Grosskopf (1995) use partially simulated data to study the effect of environmental regulation on profits among firms in the Swedish pulp and paper industry. Regulation is measured by the absolute amount of pollution a firmis permitted to discharge. The authors use a nonparametric, linear programming approach to simulate a ratio of short-run regulation-constrained profits to unconstrained short-run profits as a measure of the cost of regulation. Based on this simulation, the authors conclude that most firms in their sample are unaffected by regulation, although the profits of some are reduced. Alpay, Buccola, and Kerkvliet (2002) examine the effect of pollution regulations on the profitability and productivity growth of the Mexican and U.S. food industries. These authors find that U.S. pollution regulations have no impact on the profitability or productivity of U.S. food manufacturing, yet Mexico’s rising environmental standards enhance food processors’ productivity growth.
In the next section, we describe our use of the noted empirical studies to construct an empirical model that measures the effect of clean water regulation on firm-level profitability.
III. Measuring and Explaining Profitability
Our analysis draws upon the above empirical literature, especially that part of the literature linking environmental performance to financial performance, to examine the effect of clean water regulation on profitability. We focus on a single measure of profitability: return on sales. Return on sales is a good measure of profitability because it is widely used by firms to evaluate operational efficiency, reflecting how much profit is being earned per dollar of sales. By definition, return on sales, denoted ROS, is the ratio of a firm’s profits before interest and taxes, denoted Π, to the firm’s sales, denoted S:
[1]Since profits equal sales less costs, denoted C, we are able to reformulate return on sales as follows:
[2]Equation [2] says the return on sales captures a firm’s ability to contain costs, conditional on a particular level of sales. In other words, given a fixed level of sales, any change in costs signals the effectiveness with which a firm turns sales into profits.
To estimate return on sales, we use a linear specification because return on sales can be negative, preventing the use of either a semilog or log-linear specification. In a linear specification, the absolute level of the return on sales of firm f at time t, denoted ROSft, is a function of the following factors: (1) clean water regulation, denoted Rft; (2) variables other than clean water regulation, denoted Xft; and (3) a stochastic error term, denoted εft. The following equation depicts the functional relationship:
[3]To estimate equation [3], we draw upon a panel data set. To control for firm-specific effects, we use standard panel estimators: pooled ordinary least squares (OLS), fixed effects, and random effects. We use standard tests to assess the three estimators. The test statistics are shown in Table 3. The Ftest for fixed effects indicates significant firm-specific effects, so the fixed-effects estimates always dominate the pooled OLS estimates. The Breusch-Pagan Lagrange multiplier test statistics reveal that the random-effects estimator is always more appropriate than the pooled OLS estimator. Consequently, we do not report the pooled OLS estimates. If the random-effects estimates are consistent, then the randomeffects estimator is the preferred estimator since it is more efficient, implying that this estimator is better able to identify a statistically significant effect. If the random-effects estimates are not consistent, then the fixed-effects estimator is the preferred estimator since it is at least consistent. The Hausman test for random effects indicates that the random-effects estimates do not differ significantly from the fixed-effects estimates, that is, the random-effects estimates appear consistent. Consequently, we focus on the randomeffects estimates. Since the fixed-effects estimates are at least consistent, we also comment on these estimates briefly.
IV. Data
To assess the effect of environmental regulation on profitability, we utilize environmental data that provide an effective measure of clean water regulation and can be matched to firm-level financial performance data. Specifically, as our measure of clean water regulation, we identify the permitted wastewater discharge limits that are imposed on major chemical manufacturing facilities regulated under the National Pollutant Discharge Elimination System (NPDES) Program.6 The data on permitted discharge limits come from the U.S. Environmental Protection Agency’s (EPA’s) Permit Compliance System (PCS) database. The NPDES Program is authorized by the Clean Water Act to control water pollution by regulating point sources of pollutant discharges. Under the program, the EPA issues permits to facilities that discharge pollutants from any point source into waters of the United States. A permit specifies a limit for the maximum amount of pollution that can be discharged for each regulated pollutant.
Permit writers consider two standards when determining a permitted discharge limit: (1) the state water quality–based standard and (2) the Effluent Limitation Guideline standard. After a limit is determined under each standard, the more stringent limit is written into the permit. The state water quality–based standard is designed to ensure that the ambient water quality of the receiving water body meets state-based ambient quality standards. In other words, the discharge limit is set so that the facility’s discharges do not cause the water body’s ambient water quality to fall below the acceptable level.7 Effluent Limitation Guideline standards are designed to require a minimum level of wastewater treatment in a given industry (i.e., they establish a uniform upper bound on limits across the entire United States in a given industry).8 Regardless of the standard used to determine a limit, a facility is allowed to use any available technology to comply with the limit.
Permitted discharge limits may vary from month to month for three reasons. First, permits are generally issued on a five-year cycle. However, these cycles are not tied to calendar years. Thus, the switch from one permit to another may occur midyear. Second, permits impose three types of limits: initial, interim, and final. The limit levels frequently vary across these three limit types. Again, the transition from one limit type to the next may occur midyear since they do not apply to particular calendar years. Third and most important, permits may impose seasonal limits, which vary within a calendar year.
As noted above, the permitted discharge limits are written into facility-specific permits that are generally issued on a five-year basis. The permit application process generally begins several months, if not years, before their permits become operative as well as after they are operative. Consequently, facilities are aware of their future discharge limits well before their permits begin. Thus, permitted discharge limits most likely affect a firm’s profitability without a lag. As demonstrated in equation [3], our estimation uses the contemporaneous discharge limit as the measure of environmental regulation when estimating the firm’s current rate of profitability.
Our study considers data on both an annual basis and a quarterly basis. Each type of data basis has strengths. The annual data are less sensitive to potentially irrelevant fluctuations. In this regard, use of the annual data may better capture a meaningful relationship between financial performance and environmental regulation if most of the fluctuations serve only to generate noise in the estimation process. Then again, the fluctuations captured by the quarterly data may be important for linking to seasonal variation in permitted discharge limits. Thus, the annual data may better capture overall financial conditions, while the quarterly data may better capture fluctuating financial conditions. A priori, it is not obvious which dimension is more important.
To measure clean water regulation, we use permitted discharge limit information on two regulated pollutants: biochemical oxygen demand (BOD) and total suspended solids (TSS). BOD and TSS provide a good generalization of other regulated pollutants for three reasons. First, they are both conventional pollutants, which receive the bulk of regulatory scrutiny. Second, all previous studies of wastewater discharges focus exclusively on either BOD or both BOD and TSS (e.g., Earnhart 2004). Third, TSS is the most prevalent pollutant in our sample, and BOD is the second most prevalent.
Permitted discharge limits are a particularly useful measure of clean water regulation because they represent an output-based performance standard rather than an inputbased technology standard.9 The Porter hypothesis does not claim that any type of environmental regulation improves profitability. Instead, it maintains that a welldesigned regulation improves profitability. Porter and van der Linde (1995) offer performance-based standards as examples of well-designed regulation. Thus, our assessment of a performance-based standard represents a proper test of the Porter hypothesis. Any assessment of input-based technology standards would fail to test properly the Porter hypothesis because these types of standards constrain a firm’s ability to innovate in response to the environmental regulation.
This point notwithstanding, as argued by previous economic studies (Palmer, Oates, and Portney 1995), a performance-based standard may be less likely than an incentive-based policy instrument, such as an emissions tax or a system of marketable permits, to generate innovation offsets. In particular, use of an incentive-based policy instrument allows facilities to select the best level of performance, which is important if a performance-based standard represents a binding constraint. In this case, the incentive to innovate declines once performancebased compliance is achieved, but the incentive to lower emissions tax payments or marketable permit purchases remains at all discharge levels. While an incentivebased policy instrument may be more likely to generate innovation offsets, no such policy applies to wastewater discharges from the firms included in our sample. In the United States, few incentive-based policy instruments have ever been used. Our study instead focuses on the most prominently used policy instrument in the United States for controlling pollution: numeric effluent limits.
Permitted discharge limits are based on the quantity or the concentration of pollutants. Quantity-based limits identify the absolute amount of pollutant discharges allowed (e.g., pounds per day). Concentration-based limits identify the amount of pollution allowed relative to the amount of treated wastewater (e.g., milligrams per liter). A major chemical manufacturing facility may face one or both types of permitted discharge limits. Since we are not certain which permitted discharge limit type is more restrictive, we retain information on both types. Since permitted discharge limits may vary from month to month, as noted above, we gather data on a monthly basis. For each month, t, we assess potentially four different discharge limits: TSS quantity-based limit, denoted 
, TSS concentration-based limit, denoted 
, BOD quantity-based limit, denoted 
, and BOD concentration-based limit, denoted 
. First, we address each pollutant-specific limit separately, while transforming the two limit levels—quantity based and concentration based—into a single composite permitted discharge limit for a particular pollutant. Specifically, we divide each of the monthly permitted discharge limit levels by the relevant subsample means and then average the two scaled permitted discharge limit levels. Consider the case of TSS limits. We divide the TSS quantity-based permitted discharge limit, Lt TQ, by the average of all quantitybased permitted discharge limits across all years and facilities, denoted Lt TQS. The resulting scaled TSS quantity-based limit is denoted 
: 
. We perform similar calculations for each TSS concentration-based permitted discharge limit. Similarly, we divide each TSS concentration-based permitted discharge limit, 
, by the average of all concentration based permitted discharge limits across all years and facilities, denoted Lt TQ. The resulting scaled TSS quantity-based limit is denoted 
. Then we average the two TSS scaled limit levels in order to generate a comprehensive scaled discharge limit, denoted 
. Similarly, we generate a scaled composite discharge limit for BOD, denoted 
.
The resulting composite permitted discharge limits provide effective measures of regulation by comparing the stringency of permitted discharge limits at one facility relative to the stringency of permitted discharge limits at all other facilities in the sample. Thus, a lower “relative” permitted discharge limit indicates more stringent regulation. In order to assess both pollutants jointly, we generate a single comprehensive limit measure, denoted 
, by averaging the BOD and TSS relative permitted discharge limits: 
.
In addition to data on clean water regulation, we gather data on profitability and factors that explain profitability. This choice limits our analysis to firms with public financial data, which are available from Standard & Poor’s Compustat Research Insight© (1993–2003). Research Insight contains annual and quarterly data. To construct return on sales as defined in equation [1], we use profits before interest and taxes (i.e., operating profits) and sales less returns and allowances (i.e., net sales) from firms’ financial statements.
We incorporate into our empirical analysis the following control factors, which also explain profitability: (1) the firm’s sales growth, (2) the firm’s capital intensity, (3) age of the firm’s assets, (4) the firm’s size, (5) the firm’s market share, and (6) industry concentration. The inclusion of each factor and any stated a priori effect is supported by empirical evidence (Konar and Cohen 2001; Khanna and Damon 1999; Russo and Fouts 1997; Capon, Farley, and Hoenig 1990). In turn, we address each control factor.
Sales growth. While robust sales growth is generally indicative of a firm’s ability to compete and shield itself from cyclical market variations (Perez-Quiros and Timmerman 2000), there may be an optimal point beyond which further sales growth impairs a firm’s flexibility and adaptability, adversely affecting profits. Thus, the ultimate effect of sales growth on profits is ambiguous. We capture sales growth as the percent change in a firm’s sales from the previous three years.
Capital intensity. Capital intensity serves as a proxy of capacity utilization and is generally defined as the amount of fixed or real capital relative to other factors of production. We measure a firm’s capital intensity as the ratio of the firm’s capital investments to the firm’s sales. Sales depend on both the output price and the output level, which is a function of both capital and noncapital inputs. Thus, our measure varies with the output price whereas capacity utilization should not. Nevertheless, as the capital intensity ratio declines, the use of noncapital inputs must be rising, so that capital’s utilization must be increasing. Thus, capital intensity is representative of capacity utilization. In particular, higher capital intensity indicates lower capacity utilization. If a firm’s assets are idle, a firm’s profits are lower.
Age of assets. A firm with older assets may be less operationally efficient than a firm with new assets, which frequently embed updated technologies that lead to greater productivity. Alternatively, new equipment may be more expensive than old equipment, which yields higher depreciation. Thus, the a priori effect of the age of assets on profitability is ambiguous. We capture a firm’s age of assets as the ratio of net property, plant, and equipment to gross property, plant, and equipment, the difference between the two representing depreciation.
Size. Profitability is generally expected to vary with the scale of operations. Consistent with this expectation, empirical studies of financial performance often include this factor. Size can affect profitability either positively or negatively. We capture a firm’s size as the natural log of total assets.
Market share. Economic theory asserts that monopoly power is a positive factor behind profitability. To capture monopoly power, empirical studies generally include a measure of market share. We capture a firm’s market share as the ratio of the firm’s sales to total industry sales by four-digit Standard Industrial Classification (SIC) industry.
Industry concentration. Industry concentration refers to the extent to which a small number of firms account for a large proportion of economic activity in an industry. Accordingly, a firm in a high-concentration industry is expected to reap higher profits than a firm in a low-concentration industry. We capture industry concentration as the fourfirm sales ratio that is specific to the four-digit SIC industry in which a firm operates.10
In addition to these primary explanatory factors, we incorporate other factors. We control for variation in profitability across years by including six year dummies, with 1995 serving as the benchmark year. We also control for the average size of all facilities owned by a firm by including the average flow of wastewater that facilities are designed to manage (i.e., flow design capacity), which is calculated by considering the flow design capacity for all facilities owned by a particular firm.11 Similar to firm size, facility size may affect return on sales due to the presence of economies of scale or diseconomies of scale at the facility level, for example, larger facilities may tap into economies of scale, which lowers unit costs and increases profits.
We match the facility-level environmental data from the PCS database to the firmlevel financial data from the Research Insight database. In some cases, multiple facilities in the PCS database are owned by the same parent firm in Research Insight. In these cases, in order to examine firm-level profitability, we aggregate the facility-level environmental data to the firm level. In all cases, the environmental data are recorded on a monthly frequency, while the financial data are recorded on either an annual or quarterly frequency. To resolve this temporal difference, we aggregate the facility-level environmental data to the firm level by year and quarter.12
The sample period is January 1995 to June 2001. After we match the environmental and financial data and apply the necessary restrictions of nonmissing data for the regression variables, the samples available for analysis include two unbalanced panels. The sample panel of annual data contains 337 observations, consisting of 73 chemical manufacturing firms. The sample panel of quarterly data contains 926 observations, consisting of 59 chemical manufacturing firms.13 Table 1 contains descriptive statistics for each sample, and Table 2 contains correlation coefficients. In general, variable means and standard deviations are similar between the two samples, with slightly higher values in the quarterly sample.14 In addition, correlations among independent variables are generally similar between the two samples. The reasonably weak correlation coefficients of all independent variables should assuage any possible concerns about multicollinearity.
Descriptive Statistics: Annual and Quarterly Data
Correlations: Independent Variables in the Annual and Quarterly Data
Before assessing the empirical results in the next section, we address potential concerns. First, if permitted discharge limits depend on facilities’ financial conditions, our use of permitted discharge limits as a measure of regulation may generate erroneous conclusions regarding the effect of environmental regulation on financial performance. This possible dependence of discharge limits on financial conditions can be explained by using a political economy theoretical framework in which regulators attempt to maximize political support for their policies. If critics of the Porter hypothesis are correct that tighter limits reduce profitability, imposing tighter limits lowers political support, especially when tighter limits prompt firms to close some of their facilities, which are highly visible effects of tighter regulation. Given the desire to maximize political support, regulators may impose discharge limits that are inversely related to a firm’s financial performance, that is, decreased financial performance leads to higher discharge limits. In this case, the relationship between discharge limits and financial performance represents a reverse causation: financial performance causes discharge limits rather than discharge limits causing financial performance. However, the process of determining permitted discharge limits indicates that permitted discharge limits are highly unlikely to depend on financial performance. Permitted discharge limit levels are determined by Effluent Limitation Guidelines, which apply uniformly across all facilities within a particular industry, or by state water quality standards, which clearly are not based on financial performance. Thus, neither of these two dimensions relate to an individual facility’s financial conditions. More specifically, we are assured by conversations with EPA permit writers that financial performance is not considered when developing a permitted discharge limit.
Second, if permitted discharge limits depend on facility-level wastewater discharges or an individual facility’s ability to control discharges (i.e., cost of compliance), we would misinterpret the effect of environmental regulation on financial performance because limits would not be exogenously determined. As long as a facility’s ability to control pollution is correlated with financial performance to some degree, we would be incorrectly attributing influence to discharge limits rather than to a facility’s ability to control pollution. Since limits depend on industry-wide guidelines or ambient water quality concerns, facility-specific discharge limits cannot depend on facility-level wastewater discharges in the past, present, or future or facility-level costs of compliance.
Third, we claim that firms are most likely able to anticipate future permitted discharge limits with sufficient lead time so that the contemporaneous discharge limit level affects current-period profitability. However, firms may not be able to fully anticipate future limit levels. If true, the connection from limit stringency to profitability then involves a lag. Alternatively, changes in regulation may be expected to have benefits that are more likely to improve future profitability rather than current-period profitability because these benefits require time to develop (e.g., improved reputation, goodwill, etc.). To assess either of these possibilities, we replace the contemporaneous permitted discharge limit with a lagged permitted discharge limit in the specification of equation [3]. Results support our claim that permitted discharge limits most likely affect a firm’s profitability without a lag.
Fourth, the profit-increasing innovation described by Porter and van der Linde (1995) is more likely to be undertaken by a firm that faces a permanent or long-term increase in regulatory stringency because the firm would consider such an innovation as more worthwhile. In contrast, a temporary change in regulatory stringency, such as a tighter seasonal limit, may not send a sufficiently strong signal of innovation. Therefore, since our empirical analysis includes both permanent changes (at least for the five-year permit period in general) and temporary changes, it is not as well positioned to identify the Porter effect as it would be if all the studied changes in discharge limits were permanent. For this reason, our apparent rejection of the Porter hypothesis need not generalize to the context of permanent increases in regulatory stringency.
Fifth, the empirical analysis combines the BOD and TSS discharge limits into a single regressor. This construction appears justified since Wald tests reveal no statistically significant differences between the effect of the BOD permitted discharge limit on profitability and the effect of the TSS permitted discharge limit on profitability when both factors are included as separate regressors in the estimation of profitability.
Sixth, flow design capacity, which proxies for facility size, and quantity-based limits are most likely correlated. Inclusion of firm-average flow design capacity as a regressor in our estimation of profitability avoids any potential omitted variable bias associated with this correlation. Without this inclusion, we might incorrectly attribute the influence of facility size on profitability as part of the effect of the permitted discharge limit on profitability.
Seventh, the seasonal variability in discharge limits may be correlated with seasonable variability in demand, which could produce a spurious correlation between environmental stringency and profitability. The possible correlation between seasonal limits and demand variability is unlikely since seasonal limits are not connected to the standard four seasons. Instead, seasonal limits are connected to the flow conditions of the receiving water body. In general, limits are tighter when stream flow is low. This concern applies only to the estimation of quarterly financial data because the annual financial data do not contain any seasonal variation. Since our empirical results are robust to the chosen frequency of the financial data reporting, this concern does not appear to drive our results.
V. Empirical Results
Using the described data, our analysis seeks to estimate the effect of clean water regulation, as measured by permitted discharge limits, on profitability, as measured by return on sales. Table 3 displays results from the estimation of equation [3] using a random effects estimator and drawing upon the annual and quarterly data samples.
Random Effects Estimates of Return on Sales: Based on Annual and Quarterly Data
Based on both samples, the coefficient estimates of the traditional control variables are generally consistent with our expectations. Return on sales is positively related to sales growth and firm size in both the annual and quarterly samples. Industry concentration and the age of a firm’s assets have negative effects on return on sales in both samples. The coefficient estimates are not significant for capital intensity and a firm’s market share in either sample. We neither report nor interpret the results for the year indicators or the average size of the chemical facilities owned by a firm because most of the estimates prove insignificant and because the results are robust to the exclusion of these factors. In this regard, our estimation results are robust to the selection of a regressor set.
The overall R-squared is higher in the annual sample, which appears to indicate that the random effects model is better able to estimate long-term overall levels of profitability than short-term fluctuations in profitability.
Most importantly, the permitted discharge limits’ coefficient estimates are positive and significant at the 1% level in both the annual and the quarterly samples. Thus, we draw the same conclusion whether we link clean water regulation to short-term profitability fluctuations or to long-term profitability levels.15
Even though the fixed-effects estimates are dominated by the random-effects estimates, the fixed-effects estimates are consistent, so their interpretation may provide insight. In contrast to the random-effects estimates, the coefficient on permitted discharge limits proves statistically insignificant, regardless of the chosen sample— quarterly or annual. This result need not be surprising since it is more difficult to identify a significant relationship based exclusively on intrafirm variation. For this reason and because the random-effects estimates are more efficient, we base our conclusions exclusively on the random-effects estimates.
Recall that the permitted discharge limit is a relative measure, with lower values indicating more stringent clean water regulation. Accordingly, a positive coefficient estimate for the permitted discharge limit indicates that more stringent clean water regulation hinders profitability. These results appear to reject the strong version of the Porter hypothesis. Indeed, they seem consistent with critics’ arguments against the Porter hypothesis: stringent clean water regulation has a negative effect on profitability. Tighter clean water regulation appears to constrain firms’ abilities to turn sales into profits. As an explanation, recall the arguments provided by critics of the Porter hypothesis. Firms facing more stringent limits may incur higher treatment costs, may experience a more constrained ability to pursue profit-increasing activities, and/or may be required to commit more resources to nonproductive uses.
As an alternative understanding, recall the reformulation of return on sales, as shown in equation [2]. Given this reformulation, return on sales captures a firm’s costs relative to sales. The estimated effect of the relative permitted discharge limit on return on sales indicates that tighter limits raise costs relative to sales, thereby reducing return on sales. In other words, clean water regulation constrains firms’ abilities to contain costs, conditional upon a given level of sales.
Next, we assess the economic consequences of our results from the annual sample by assessing how changes in the relative permitted discharge limit affect return on sales (an assessment of the quarterly sample results yields similar conclusions). A oneunit decrease in the relative permitted discharge limit causes return on sales to decrease by a factor of 0.007. However, a one-unit decrease is dramatic since the average relative permitted discharge limit, 1.382, is small by construction (see Table 1). As an alternative approach, we assess a 10% reduction in the average relative permitted discharge limit, denoted 
, which causes the return on sales to decrease according to the following equation:
[4]We consider the point estimate of βR and its 90% confidence interval, which lies between 0.0032 and 0.0113. By considering the 90% confidence interval, we are able to assess more fully economic significance and plausibility of the estimated effects.
A 10% reduction in the average relative permitted discharge limit equals 0.1382. Based on the point estimate, the 10% reduction causes the return on sales to drop by a factor of 0.00097 (5 0.1382 * 0.007), which represents a 1.7% decline in return on sales relative to the sample mean of 0.057. Based on the 90% confidence interval, the drop in return on sales lies between 0.00044 and 0.00156, which lies between 0.8% and 2.7% relative to the sample mean. Thus, based on the point estimate and the 90% confidence interval associated with the coefficient on discharge limits, the impact of a 10% tightening in environmental regulation appears both economically meaningful and plausible.
VI. Conclusion
We study the effect of clean water regulation on the profitability of publicly held firms in the chemical manufacturing industries. We focus on return on sales as our measure of profitability. After controlling for other factors that are expected to affect profitability, our analysis identifies a negative relationship between clean water regulation and profitability, which is consistent with the critics’ arguments against the Porter hypothesis: more stringent clean water regulation undermines profitability. In other words, a more stringent wastewater permitted discharge limit prompts a decrease in return on sales. In particular, based on the estimation of annual financial data, a 10% reduction in the average relative permitted discharge limit causes the return on sales to decrease by as little as 0.8% and as much as 2.7% according to the 90% confidence interval of the estimated coefficient on the discharge limit. Thus, the relationship is both statistically and economically significant. Using quarterly and annual data, we are able to demonstrate a substantial effect on both short-term profitability fluctuations and long-term profitability levels.
We recognize that our results are specific to the regulation of two pollutants—BOD and TSS—of publicly held firms in the chemical manufacturing industries and may not generalize to the regulation of other 86(2) Rassier and Earnhart: Clean Water Regulation and Profitability 341 wastewater pollutants, pollution in other media (e.g., air), pollution discharged by privately held firms in the chemical manufacturing industries, or pollution discharged by all types of firms in other industries. Nevertheless, our study is an important step toward resolving the debate regarding the effect of environmental regulation on profitability.
Appendix
We describe here the match between the firmlevel data from the Research Insight database and the facility-level data from the EPA’s PCS database. In addition, we describe how we address temporal frequency differences between the two databases, as well as any potential differences between the firmlevel data and the facility-level data.
We match the facility-level environmental data to the firm-level financial data by year using the firmlevel name. The firm-level name is not initially available in the environmental data. Instead, we use the EPA’s Toxic Release Inventory (TRI) database to find the firm-level name by year for each facility in the environmental data. The TRI database contains information regarding a facility’s parent firm. For some facilities, the TRI database does not contain the parent firm for a given year. In these cases, we are usually able to identify the parent firm using additional information. As the most useful method, we use the parent firm from the preceding and succeeding years if the name remains the same. If we are unable to recover a parent firm from the TRI database, we search for the facility name in Research Insight or Thomson Gale’s Business and Company Resource Center© (2004). Business and Company Resource Center contains data on privately and publicly held firms, including ownership structure. If neither of the business databases indicates ownership by a publicly held firm, we assume the facility is owned by a privately held firm. For cases in which the Business and Company Resource Center does not provide data on ownership structure, we assume the parent firm is publicly held if found in Research Insight and privately held if not found in Research Insight.
After we match the environmental and financial data, we address potential level differences within firms between the two sets of data. The environmental data are recorded at the facility level with a monthly frequency for major chemical manufacturing facilities. The financial data are recorded at the firm level with either an annual or quarterly frequency for firms in all industries. To resolve these differences, we aggregate the environmental data across months and commonly owned facilities to annual and quarterly firm levels using a simple, nonweighted averaging protocol.
This approach resolves temporal frequency differences for all firms. It also resolves level differences for firms that own only chemical manufacturing facilities contained in the environmental data. However, this approach does not completely resolve potential level differences for firms that own facilities outside the chemical manufacturing industry. In these cases, the financial data reflect activity from all facilities, but the environmental data reflect activity only from major chemical manufacturing facilities. To address these potential level differences, we use the firm-level primary two-digit SIC code from Research Insight to divide the sample into two subsamples: one subsample includes only firms with a primary two-digit SIC code for the chemical manufacturing industry, while the other subsample includes firms with a primary two-digit SIC code outside the chemical manufacturing industry. Chow tests reject the null hypothesis that the coefficient estimates are the same between the chemical and nonchemical samples. Based on these test results, we proceed with a subsample of firms with a primary two-digit SIC code for the chemical manufacturing industry.16
Table A1 provides information on the sample selection process. In particular, the table assesses the numbers of facilities, firms, and observations in various subsamples. The table distinguishes between publicly held firms and privately held firms. Of the publicly held firms, the table distinguishes between those firms operating primarily in the chemical manufacturing sector and those firms operating primarily outside of the chemical manufacturing sector.
Sample Selection
Footnotes
The authors are, respectively, research economist, U.S. Department of Commerce, Bureau of Economic Analysis; and professor, Department of Economics, University of Kansas. The views expressed in this manuscript are those of the authors and not necessarily those of the U.S. Department of Commerce, Bureau of Economic Analysis. This manuscript was developed under STAR Research Assistance Agreement No. R-82882801-0 awarded by the U.S. Environmental Protection Agency. It has not been formally reviewed by the EPA. The EPA does not endorse any products or commercial services mentioned in this manuscript.
↵1 This statistic represents the most recent pollution and abatement cost estimate published for the United States by the U.S. Bureau of Economic Analysis, which was published in September 1996.
↵2 Ambec and Barla (2002) construct a theoretical model that supports Porter and van der Linde’s (1995) suggestion that environmental regulation creates external pressure to overcome organizational inertia. The authors show that environmental regulation may reduce agency costs borne by firms when they attempt to induce truthful reporting about productivity by division managers. Regulation separately prompts investment in pollutionreducing technology and an improvement in firms’ expected profits.
↵3 Porter and van der Linde (1995) broadly divide innovation offsets into product offsets and process offsets. Product offsets occur when regulation yields higher-quality products, safer products, lower product costs, or lower costs of product disposal for users. Process offsets occur when regulation yields higher resource productivity, less process downtime, better utilization of by-products, lower energy consumption, reduced material storage and handling costs, reduced waste disposal costs, or safer workplace conditions.
↵4 The innovation espoused by Porter and van der Linde (1995) is operative in a variety of market settings as demonstrated by the firms they discuss as case studies, including chemical firms. However, the authors focus largely on the potential for environmental regulation to generate innovation in competitive markets.
↵5 Other theoretical studies address particular aspects of the organizational behavior underlying the Porter hypothesis (e.g., King 1999, 2000; King and Lenox 2002).
↵6 Major facilities are distinguished from minor facilities by the EPA as those facilities with a significant effect on the water quality of the receiving water body or a major rating code equal to or greater than 80. A major rating code is the total numeric score of ranking points assigned to a facility from the NPDES Permit Ranking Worksheet. Highly limited information on permitted discharge limits imposed on minor facilities is available from the EPA.
↵7 While the state water quality standards do not differ within a state, the discharge limits identified by state water quality–based standards will differ across facilities and time, since the background pollution of water bodies differs across space and time.
↵8 Since the passage of the 1972 Federal Water Pollution Control Act, which preceded the Clean Water Act, the EPA has developed industry-specific Effluent Limitation Guidelines. The guidelines have been developed by the EPA based on the degree of pollution reduction attainable by facilities of a given industry. If no industry-specific Effluent Limitation Guideline applies to a particular facility, a permit writer uses his or her “best professional judgment,” which draws upon all reasonably available and relevant data. In particular, the permit writer evaluates the effect of a permitted discharge limit on the environment. However, best professional judgment is relevant for none of the sampled firms, since all of the sampled facilities operate in industries covered by Effluent Limitation Guidelines.
↵9 Here, “output” refers to the pollutant generated from production rather than the economic good produced.
↵10 We use the Research Insight database to construct market share and industry concentration measures.
↵11 Flow design capacity measures the amount of water a facility is designed to handle, not the amount of pollution discharged by a facility on either a quantity basis or concentration basis. As an analogy, flow design capacity measures the size of the bathtub not the amount of dirt scrubbed from the bather’s body and discharged down the drain.
↵12 For a detailed explanation of how we match the PCS database to the Research Insight database, address any potential differences between the firm-level data and the facility-level data, and address temporal frequency differences, see the Appendix.
↵13 The Appendix and associated Table A1 provide more details on the sample selection process.
↵14 In each sample, the mean relative permitted discharge limit need not equal one for two reasons. First, the sample-wide mean values used for scaling are based on individual facilities, while the reported mean values are based on individual firms. As noted, some firms own multiple facilities. Second, the general restrictiveness of concentration-based limits may vary from the general restrictiveness of quantity-based limits. As noted, some facilities face only quantity-based limits, while other facilities face only concentration-based limits.
↵15 Mohr and Saha (2008) demonstrate that, in the presence of other market failures, environmental regulation may benefit firms without fostering innovation. If true, our empirical analysis would not be able to distinguish between the noted outcome and the Porter hypothesis outcome. Since our empirical analysis does not reveal that regulation benefits firms, this concern is not relevant to our conclusions.
↵16 The Chow test results indicate that the firms focusing on chemicals respond differently to environmental regulation than do firms producing a mix of products. Based on these tests, our results do not appear to generalize to all firms that produce at least chemicals. Of course, this difference may stem solely from our inability to examine the wastewater effluent limits for facilities that do not manufacture chemicals yet are owned by firms that do manufacture chemicals at other facilities. This difference need not surprise us. Firms that focus on chemical production may be better able to innovate in response to tighter limits because they focus on a smaller set of products. On the other hand, firms that produce a broad mix of products may be better able to innovate in response to tighter limits because their breadth of products facilitates a broader perspective on pollution control techniques (i.e., effective pollution control techniques for one product may transfer to the pollution control applied to another disparate product).
