Sustainability development for supply chain management in U.S. petroleum industry by DEA environmental assessment

Energy Economics 46 (2014) 360–374

Contents lists available at ScienceDirect

Energy Economics journal homepage: www.elsevier.com/locate/eneco

Sustainability development for supply chain management in U.S. petroleum industry by DEA environmental assessment Toshiyuki Sueyoshi a,⁎, Derek Wang b a b

New Mexico Institute of Mining & Technology, Department of Management, 801 Leroy Place, Socorro, NM, 87801, USA McGill University, Desautels Faculty of Management, 1001 Sherbrooke Street West, Montréal, QC H3A 1G5, Canada

a r t i c l e

i n f o

Article history: Received 12 July 2014 Received in revised form 18 September 2014 Accepted 30 September 2014 Available online 16 October 2014 JEL classification: C60 C68 M52 Keywords: Environment Supply chain DEA Corporate sustainability

a b s t r a c t Environmental assessment and protection are important concerns in modern business. Consumers are interested in environmental protection. They avoid purchasing products from dirt-imaged companies even if their prices are much less than the ones produced by green-imaged companies. The green image is very important for business survivability in our global market. A business concern associated with the green image is how to attain corporate sustainability where companies can attain both economic success and pollution prevention in their operations. There is tradeoff between economic prosperity and environmental concern. To attain a high level of corporate sustainability, companies need to measure the current performance in terms of their operational and environmental achievements. This study proposes a use of Data Envelopment Analysis (DEA) for such assessment. The proposed DEA assessment provides corporate leaders and managers with not only the measure of corporate sustainability but also information regarding how to invest for technology innovation for abatement of undesirable outputs (e.g., CO2). As an application, this study utilizes the proposed approach to measure the corporate sustainability of petroleum firms in the United States. The petroleum industry is functionally separated into integrated and independent companies. The integrated companies, referred to as “Major”, have their large supply chains for both “upstream” (e.g., exploration, development and production of crude oil or natural gas) and “downstream” (e.g., oil tankers, refineries, storages and retails). Meanwhile, the independent companies focus upon the upstream function, but not the downstream, in their business operations. The empirical comparison between the two groups of petroleum firms identifies that the integrated companies outperform the independent companies because a large supply chain incorporated into the former group provides them with both a scale merit in their operations and an opportunity to obtain consumer's opinions on their business operations. Thus, the large supply chain system, covering business functions for upstream and downstream, enhances corporate sustainability in the U.S. petroleum industry. It is easily envisioned that the empirical findings discussed in this study are useful in preparing business strategy and industrial policy for the petroleum industry of not only the United States but also other nations involved in oil and gas production. © 2014 Elsevier B.V. All rights reserved.

1. Introduction Recently, the Intergovernmental Panel on Climate Change (IPCC, 2014), established by United Nations environmental program, has reported the policy suggestion that it is necessary for us to reduce an amount of greenhouse gas (GHG) emissions, in particular CO2, by 40– 70% (compared with 2010) until 2050 and to the level of almost zero by the end of the 21st century via shifting our current systems to energy efficient ones. Otherwise, we will have to bear severe consequences, such as heat waves, droughts, floods, food crisis as well as other damages to human, social and economic systems. The challenge on climate change makes the conventional profit-driven business logic ⁎ Corresponding author at: New Mexico Institute of Mining & Technology, Department of Management, 801 Leroy Place, Socorro, NM 87801, USA. E-mail addresses: [email protected] (T. Sueyoshi), [email protected] (D. Wang).

http://dx.doi.org/10.1016/j.eneco.2014.09.022 0140-9883/© 2014 Elsevier B.V. All rights reserved.

and practice inappropriate and incompatible with a world-wide trend toward a sustainable society. Firms need to change their business operations to adapt to various regulations for GHG emission abatement. They also face greater market pressures because an increasing number of consumers reject products and services from dirty-imaged companies even if their prices are much lower than those of green-imaged companies. The purpose of this study is to examine the corporate sustainability of U.S. petroleum companies, thereby obtaining empirical implications on their supply chain operations and carbon footprints. Among all industry sectors, the petroleum industry is of particular interest to us because the industry is the second largest contributor of GHG emission in U.S., which has emitted at least 217 million metric tons (MMT) of carbon dioxide equivalent (CO2e) in 2012, being the second only to the electric power industry (Environmental Protection Agency, 2013a). See also Environmental Protection Agency (2013b) for detailed

T. Sueyoshi, D. Wang / Energy Economics 46 (2014) 360–374

emission statistics of all industrial sectors. It is often reported that the climate change may exert negative impacts on the petroleum industry's ability to access, yield and distribute oil and gas products (Depart of Energy, 2013). Thus, limiting and reducing GHG emission in the petroleum industry have recently received considerable attentions among corporate leaders, policy makers and individuals who are interested in environment protection, as evidenced by the release of Oil and Natural Gas Air Pollution Standards from EPA (2012a). The typical functions, as found in the petroleum supply chain, include exploration, production, processing, distribution and retail of oil and natural gas products (American Petroleum Institute, 2013). In terms of the extent of involvement in the supply chain, petroleum companies can be functionally classified as integrated and independent companies. The integrated companies participate in the operation of the entire supply chain from upstream exploration to downstream retailing. Meanwhile, the independent companies typically engage in only upstream functions such as exploration and development. In this study, the proposed analysis focuses upon an amount of emission from onshore exploration and production segment. In the segment, GHG is generated directly through drilling processes and fossil fuel combustion and indirectly through well leaks and venting. Exploration and production is the first step where a significant amount of GHG is generated. In EPA's Greenhouse Gas Reporting Program, the segment has the largest reported GHG emission, or 88 MMT of CO2, in the petroleum industry (EPA, 2013a). Furthermore, this research focuses upon the segment that enables us to include as many companies into this study as possible. Almost all companies engage in onshore exploration and production, while only a relatively small fraction of firms participate in other activities such as offshore production, refining and distribution. As discussed previously, the petroleum industry may have a potential to reduce an amount of GHG emission by adapting green technologies and practices. For instance, a typical technology called “green completion” can capture the natural gas that would otherwise escape into the air during a well-completion period. According to estimate by the American Petroleum Institute, the green completion technology would cost $180,000 per well. Fortunately, companies have started to adapt green completion voluntarily even before EPA released the proposal to require the mandatory use of the technology (EPA, 2012b). In this study, it can be easily presumed that integrated and independent companies differ in their corporate efforts to reduce an amount of industrial pollution, so enhancing environmental performance. Integrated companies, with involvement in the entire supply chain, are more visible to consumers and regulators. They may face greater consumer pressure on environmental performance because of direct contact with consumers as their retail brands. Consumers' dissatisfaction on the environmental performance of a company would directly translate into lost sales. The large scale of the supply chain also makes them frequent targets of local governments and regulators. Moreover, the supply chain of integrated companies has global presence and the operations in multiple countries must obey international environmental regulation. In contrast, independent companies operate only in the United States where emission regulation is less stringent than some other countries. Even though environmental regulations in other countries do not directly affect U.S. operations, it can be conjectured that they would spur investment in technology innovation, which would benefit the U.S. operations through knowledge transfer. In addition, the most stringent environmental standards in the countries, where the company is present, may shape its overall environmental strategies. See, for example, Sueyoshi and Goto (2012a) that have compared the performance between international (Major) oil firms and national oil companies that are operating as a member of Organization of the Petroleum Exporting Countries (OPEC). Different from their study, this research examines the operational and environmental performance of U.S. petroleum firms from a supply chain perspective. This study refers the

361

readers to Gonzáles-Benito and Gonzáles-Benito (2006) and references therein for further discussion on the impacts of companies' supply chain on environmental performance. The goal to establish a sustainable society calls for a new methodology that can integrate the companies' environmental measures with their financial and operational measures to provide a holistic assessment for overall performance. This study proposes a usage of Data Envelopment Analysis1 (DEA) as a practical approach to evaluate the unified (operational and environmental) performance of petroleum companies from the perspective of corporate sustainability. The proposed DEA approach has several advantages over existing methods in production economics. For example, it incorporates four disposability concepts such as natural disposability and managerial disposability as well as these conceptual combinations. Outputs and inputs, characterizing companies' operational and environmental performance, are separated under these disposability concepts. Moreover, the proposed approach can identify effective investment opportunities for GHG abatement by utilizing the concept of congestion on undesirable outputs. This study summarizes both business and policy implications obtained from the proposed DEA application. The remainder of this study is organized as follows. Section 2 provides the literature review on corporate sustainability in supply chain management. Section 3 discusses underlying concepts incorporated into the proposed DEA environmental assessment. Section 4 describes non-radial models under natural and managerial disposability as well as these methodological combinations. Section 5 applies the proposed DEA approach to evaluate the operational and environmental performance of U.S. petroleum firms2 and summarizes empirical results obtained in this study. Section 6 concludes this research along with future extensions. 2. Literature review According to Sarkis et al. (2010), followed by Sueyoshi and Wang (2014), green supply chain management is classified into eight research groups. Although this study is concerned with corporate sustainability on the petroleum industry, their classification is useful in specifying the position of this research by comparing it with the other previous research efforts on supply chain management. The eight groups include (a) business complexity, (b) ecological modernization, (c) information utilization, (d) institutional externality, (e) resource-based view, (f) resource dependency, (g) social network and (h) stakeholder involvement. This study follows the previous research effort of Wang et 1 DEA has been extensively used for environmental assessment. See Glover and Sueyoshi (2009) and Ijiri and Sueyoshi (2010) for a historical description on DEA developments from the contribution of Professor William W. Cooper. DEA applications on environmental assessment can be found in a series of studies (e.g., Sueyoshi and Goto, 2009, 2010a,b, 2011a,b,c; Sueyoshi et al., 2009, 2010). An important feature of these studies is that they have developed a computational framework, but lacking a conceptual framework for DEA environmental assessment. The first article, which has discussed the conceptual framework such as natural and managerial disposability, can be found in Sueyoshi and Goto (2012a). Then, the concept of natural and managerial disposability has served a concept basis for proceeding research efforts. See, for example, Sueyoshi and Goto (2012b,d,e, f,g,h,i,j,k, 2013a,c,d, 2014a,b,c) and Sueyoshi and Goto (2013a,b). The disposability concepts are linked to an occurrence of desirable congestions, or technology innovation by their research efforts. In these articles, all inputs are separated into two categories under the two disposability concepts. Moreover, a conventional use of DEA was applied to renewable energy assessment (Sueyoshi and Goto, 2014c) and its combination with another methodology (e.g., Sueyoshi and Goto, 2012c). 2 Ross and Droge (2002) used DEA to measure the efficiency scores of a company's petroleum distribution network, but it did not consider the environmental performance. Ross and Droge (2004) extended their analysis for operation efficiency measurement in large-scale petroleum distribution systems, but not considering their environmental performance. Singhal and Singhal (2012) mentioned about a usage of DEA for supply chain management. They commented that DEA could identify DMUs on an efficiency frontier. This analytical feature of DEA was different from conventional regression that looked for average behavior. They suggested that researchers should use DEA as a starting point for analysis to understand why some DMUs were far from the efficient frontier. The three research efforts clearly indicate the importance of DEA in operations management.

362

T. Sueyoshi, D. Wang / Energy Economics 46 (2014) 360–374

al. (2014) and Sueyoshi and Wang (2014) for describing the literature review on green supply chain in operations management. Business complexity: Business complexity is defined through heterogeneity or diversity in environmental factors such as consumers, governmental regulations3 and technological advancements. Along with an increase in corporate size, firms have had a difficulty in planning and predicting their business actions, including product return, recycling, product inspection and quality check (Chakravarthy, 1997; Vachon and Klassen, 2006). This type of concern was also discussed by Sueyoshi and Goto (2010a). Utilizing a data set on 220 Japanese manufacturing firms from 2004 to 2007, so being the total of 853 observations, their study concluded that the Japanese firms tried to accumulate capital and then used it to prevent industrial pollution. As result, the size of firms was very important in understanding their corporate behaviors toward environmental protection. Acknowledging an influence of regulation, their study also discussed that firms could not make any strategic step toward corporate sustainability without capital accumulation even under regulation. Ecological modernization: As eco-innovators, firms were geared toward achieving industrial development and environmental protection through technology innovation. Regulation influenced corporate efforts on environmental innovation (Murphy and Gouldson, 2000). The study also suggested that manufacturing firms could overcome barriers to innovation and gain operational opportunities for performance improvement. However, as reported by Revell (2007), the corporate efforts did not bring financial benefits. Therefore, it was necessary to develop a diffusion mechanism in which core large firms motivated toward environmental innovation and then diffused innovation technology to small and medium ones (Hall, 2001). Information utilization: Firms needed to report their environmental performance to outside stakeholders (e.g., customers and equity holders), but they lacked full knowledge on products, processes and 3 The extraction of oil and natural gas produces various forms of pollutants, including solid wastes such as drill cuttings, liquid pollution such as well treatment fluids and oil spillage, and air toxics such as methane from well and process line leakage. Therefore, the extraction of oil and natural gas is governed by a number of environmental regulations at the federal and state levels. The federal regulations constitute the baseline requirements for all exploration and production activities, whereas the state governments can enact more stringent laws. At the federal level, the most prominent regulations include the National Environmental Policy Act (NEPA), Clean Air Act (CAA), Clean Water Act (CWA), Safe Drinking Water Act (SDWA), Endangered Species Act (ESA), Toxic Substance Control Act (TSCA), and Resource Conservation and Recovery Act (RCRA). All these federal statutes can impact oil and gas production directly and indirectly. EPA is the primary federal regulatory agency responsible for environmental protection. The operations on federal land are also governed by regulations from Bureau of Land Management. Specifically pertinent to this study are the regulations related to GHG emission in oil and gas industry. Among such regulations, the Greenhouse Gas Reporting Program (http://www.epa.gov/ ghgreporting/), the National Emissions Standards for Hazardous Air Pollutants (NESHAP: http://www.epa.gov/compliance/monitoring/programs/caa/neshaps.html) and the New Source Performance Standards (NSPS: http://www.epa.gov/compliance/monitoring/ programs/caa/newsource. html) are designed and supervised by EPA. The Greenhouse Gas Reporting Program, enacted as a law in 2010, requires the nation's large emitters of GHG to report the amount of GHG they emit annually. The NESHAP establishes air pollutants standards for 188 different hazardous air pollutants at new and existing sources. The NSPS targets at new, modified and reconstructed facilities. However, there are several noticeable loopholes in the implementation of the laws. For instance, under the rule of aggregation, NESHAP considers small polluting sources in close proximity to each other as one source of emissions. However, EPA exempts oil and gas wells from aggregation and hence the stricter standards of larger emitting sources are not applicable. Prior to 2012 there was no set of well management standards at the federal level. It was the state and local regulators who primarily regulated the drilling industry. For instance, Colorado requires the usage of green completion technology to minimize gas releases during well completion, while most other states do not require green completion. In order to reduce air pollution from the drilling process, especially the hydraulic fracturing for shale natural gas, in 2012 EPA revised NSPS and NESHAP to impose stricter emission rules on oil and gas production (http://www.epa.gov/airquality/oilandgas/actions.html). The focus of the new regulation is to limit the emission of volatile organic compounds (VOCs), SO2, methane and other hazardous air pollutants. The most noticeable feature of the new regulation is the mandatory usage of green completion for most wells. A transition period until January 2015 is granted to producers for them to make cost-effective adjustments. Finally, this study points out that all above regulations apply to integrated and independent petroleum companies alike regardless of their sizes and economic status.

material flows. Consequently, information asymmetry occurred between firms and outside stakeholders (Simpson, 2010). Information sharing is critical for coordinating between firms and stakeholders in terms of enhancing corporate image and satisfying regulation requirements (Wong et al., 2009). A difficulty associated with the information asymmetry and sharing is that all stakeholders need a brief summary which they can easily understand, not detailed and complicated information to describe the operational and environmental performance of firms. Institution externality: External pressures influenced a firm to adapt organizational practice (Lai et al., 2006). The concept could be used to examine how a firm addressed green issues because of external pressures (Jennings and Zandbergen, 1995). The concept of externality indicated a research direction to explain environmental practices. Public agencies were such an example of powerful institutions that influenced the business practice of an organization (Rivera, 2004). Moreover, according to Carter et al. (2000), 75% of U.S. consumers made their purchasing decisions with firm's environmental reputation taken into consideration, and 80% of them were willing to pay more for environmentally friendly products. Consumers increasingly heightened environmental awareness and were starting to purchase green products (Harris, 2006). Resource-based view: Competitive advantages of firms were sustained by harnessing resources (e.g., trust from customers) that were valuable, rare, imperfectly imitable and non-substitutable (Barney, 1991). Firm resources were often defined as assets, production and sale capabilities, firm attributes and history, all of which needed to utilized for improving their efficiency, effectiveness and competitiveness (Barney, 1991; Russo and Fouts, 1997). Resource dependency: Firms should seek high performance gains in a long term horizon instead of pursuing short term benefits at an expense of the others. Such a long term perspective is essential for the development of corporate sustainability (Paloviita and Luoma-aho, 2010). To attain the long term perspective, firms needed to manage internal and external coordination to attain such performance goals where partners' coordination and resource sharing are beneficial for operational and environmental improvements (Yang et al., 2008). Social network: The concept discusses how to develop corporate sustainability by establishing social relationships inside and outside a firm. A firm makes decisions according to information obtained through its social network (Wuyts et al., 2004). For example, informal human relationships consist of an internal social network and relationships with customers are an external social network. To enhance the social network development, it is important for firms to share information such as their new recyclable products, clean process and corporate efforts for environmental protection (Theyel, 2001; Walton et al., 1998). Stakeholders: They are a group of individuals who can affect or is affected by the achievement of firm's objectives (Freeman, 1984). All firms have externalities that influence a variety of stakeholders that are both internal and external to them. The scope of environmental issues is widely diverse because stakeholders consist of many different types of individuals. A group of important stakeholders are stock holders who pay attention to both a return to their investments in a short term horizon and the status of going concern in a long term horizon (Jacobs et al., 2010). All firms always face huge risk originated from an environmental issue because the violation makes them to pay large opportunity cost, often jeopardizing the status of going concern. Consumers are another type of stakeholders who recently think environmental protection as their first priority in their decision makings on purchases. It is almost impossible for modern corporations to survive if they cannot sell their products because of their dirty-images. The reality may be often different, but the image of a green company is essential for modern business. The last stakeholder group consists of employees who need a pride on the fact that they are working for a green company. An important research issue in the area is that it needs empirical evidences that confirm a positive relationship among stock holders in

T. Sueyoshi, D. Wang / Energy Economics 46 (2014) 360–374

terms of environmental protection (Sarkis et al., 2010; Tate et al., 2010; Vachon, 2007). Position of this study: Wang et al. (2014) and Sueyoshi and Wang (2014) investigated the corporate sustainability among seven U.S. industrial sectors by applying DEA environmental assessment. Their studies concluded that the energy sector provided an important investment opportunity to enhance corporate sustainability in the United States. As a research extension, this study focuses upon the petroleum industry and separates it into two groups of companies: integrated companies with a supply chain and independent companies without the supply chain. The integrated companies (e.g., BP, Chevron and Exxon), often referred to as “Major”, have a large supply chain system from drilling oil and gas wells to retailing at gas stations. On the other hand, the other firms do not have such a supply chain function, focusing upon the upstream business (e.g., drilling oil and gas). As mentioned previously, this study is interested in whether the supply chain system is useful in establishing corporate sustainability in the U.S. petroleum industry. No previous study has investigated the research concern. To attain the research objective, this study uses DEA environmental assessment as a methodology. This study knows that the previous studies have discussed the importance of corporate sustainability in modern business, but not providing any empirical evidence on environmental assessment and investment strategy. Such evidences are usually not obtained by applying statistical approaches (Jacobs et al., 2010; Sarkis et al., 2010). Thus, it is expected that the proposed use of DEA can deliver new empirical evidences on corporate sustainability on the U.S. petroleum industry, which cannot be found in any previous studies. The research concern to be explored in this study can be summarized in the following hypothesis4: H. The integrated petroleum companies outperform the independent companies in terms of their unified (operational and environmental) efficiency measures because the former group has a large supply chain capability for connecting downstream and upstream functions, but the latter group does not have the supply chain capability, focusing upon only the upstream. A rationale, supporting the hypothesis, is because the integrated firms have a large supply chain capability so that the green image is essential in attracting consumers and investors in their operational performance. The green image is also important for the independent companies to attract investors, but not consumers, because they do not have any retailing in their businesses. Furthermore, the operation of integrated companies is larger than that of independent companies so that the former group may take advantage of a scale merit on their operations. 3. Underlying concepts for DEA environmental assessment 3.1. Abbreviations and nomenclatures All abbreviations and nomenclatures used in this study are summarized as follows. DMU: Decision Making Unit, DEA: Data Envelopment Analysis, URS: Unrestricted, UE: Unified Efficiency, UEN: Unified Efficiency under Natural disposability, UEM: Unified Efficiency under Managerial disposability, UENM: Unified Efficiency under Natural & Managerial disposability, RTS: Returns to Scale, DTS: Damages to Scale, DTR: Damages to Return, DC: Desirable Congestion, UC: Undesirable Congestion, X: a column vector of m inputs, G: a column vector of s desirable outputs, B: a column vector of h undesirable outputs, dxi : an unknown slack variable of the i-th input, dgr : an unknown slack variable of the r-th desirable output, dbf : an unknown slack variable of the f-th 4 The null hypothesis is that the two groups of companies have same efficiency scores. The hypothesis documented in this section is an opposite hypothesis. If this study statistically rejects the null hypothesis, then we may accept the opposite hypothesis.

363

undesirable output, λ: an unknown column vector of intensity (or structural) variables, Rxi : a data range related to the i-th input, Rgr : a data range related to the r-th desirable output, and Rbf : a data range related to the f-th undesirable output. 3.2. Unification between natural and managerial disposability s Let us consider X ∈ Rm + as an input vector, G ∈ R+ as a desirable output vector and B ∈ Rh+ as an undesirable output vector. These vectors are referred to as “production factors” in this study. In addition to these vectors, the subscript ( j) is used to stand for the j-th DMU (Decision Making Unit: corresponding to an organization in private and public sectors) and λj indicates the j-th intensity variable ( j = 1, …, n) which is used for connecting production factors. Using an axiomatic expression, a unified (operational and environmental) production possibility set to express natural and managerial disposability is specified by the following two types of output vectors and an input vector, respectively:

N

P ðX Þ ¼

8 <

ðG; BÞ : G ≤

n X

G jλ j; B ≥

n X

B jλ j; X ≥

n X

X jλ j;

n X

λ j ¼ 1; λ j ≥ 0

9 =

: ; j¼1 j¼1 j¼1 j¼1 8 9 n n n n < = X X X X M P ðX Þ ¼ ðG; BÞ : G ≤ G jλ j; B ≥ B jλ j; X ≤ X jλ j; λ j ¼ 1; λ j ≥ 0 : : ; j¼1 j¼1 j¼1 j¼1

&

The difference between the two disposability concepts is that production technology under natural disposability, orPN(X), has X ≥ n

∑ X j λ j . Meanwhile, the managerial disposability, orPM(X), has X ≤ j¼1 n

∑ X j λ j. The two disposability concepts intuitively appeal to us because j¼1

an efficiency frontier for desirable outputs locates above or on all observations, while that of undesirable outputs locates below or on the observations. In the two axiomatic expressions, the operational performance of DMUs is the first priority and the environmental performance is the second one under natural disposability in assessing their unified efficiency measures. In contrast, the managerial disposability has an opposite priority order in the assessment. It is widely known that the managerial disposability has been long discussed by corporate strategists (e.g., Porter and van der Linde, 1995) in U.S. business schools. It is possible for us to consider the disposability concepts as two different criteria for DEA environmental assessment. Here, it is important to note that in previous research efforts on DEA, an input vector has been assumed to project toward a decreasing direction for performance enhancement. The assumption is often inconsistent with the reality of environmental protection in a private sector. For example, let us consider a manufacturing firm that can increase the input vector as long as average cost is less than average sale because the business condition produces profit to the firm. Thus, the conventional use of DEA is often unacceptable in a private sector because the previous DEA studies have implicitly assumed the minimization of total production cost. The cost concept may be acceptable for agencies in a public sector because the public sector does not need the concept of profit in the operation of agencies. Hence, it can be easily imagined that DEA environmental assessment in the private sector, discussed in this study, is different from the conventional DEA use for the public sector. In addition to the average cost or marginal cost for guiding companies, the opportunity cost, originated from business risk due to industrial pollutions, has a major role in modern business because the opportunity cost is much larger than any other cost components in modern business. Such cost concepts used for the private sector are implicitly incorporated into formulating the two disposability concepts discussed in this study.

364

T. Sueyoshi, D. Wang / Energy Economics 46 (2014) 360–374

In applying the two disposability concepts for environmental assessment, it is possible for us to classify inputs into two categories under these disposability concepts. As an extension of the previous works (e.g., Sueyoshi and Goto, 2014b; Sueyoshi and Wang, 2014; Wang et al., 2014), this study can unify the two efficiency frontiers into a single one, based upon on which we can measure the level of unified efficiency after separating not only outputs but also inputs into two categories under the two disposability concepts. In unifying the disposability concepts, it is necessary for us to specify a same type of efficiency frontiers for desirable and undesirable outputs. Fig. 1 visually describes the type of efficiency frontiers in the horizontal axis (x) and the vertical axis (g and b). A production possibility set exists between the two lines in Fig. 1. For our descriptive convenience, all production factors are assumed to have a single component. As identified from the convex shape of the two functional forms, the efficiency frontier for an undesirable output is very similar to that of a desirable output. Such a frontier similarity occurs if technology innovation (more broadly speaking, including a managerial effort for pollution prevention) does not occur on an undesirable output, as discussed in the concept of natural disposability. Under limited technology innovation, an undesirable output can be considered as a “byproduct” of a desirable output. Therefore, the conceptual unification on disposability needs a similar type (but not same) of efficiency frontiers, as found in the left hand side of Fig. 1. However, the two frontier functions may have different shapes, as depicted in the right hand side, where an efficiency frontier for a desirable output increases along with an input increase, but the other frontier for an undesirable output decreases because of technology innovation for preventing industrial pollution. Such a case indicates the concept of managerial disposability. After describing the unification on two disposability concepts, it is necessary for us to discuss how to identify the shape of an efficiency frontier by DEA. It is impossible for us to identify the shape of the frontier by efficiency scores. However, a supporting hyperplane can provide information regarding the frontier shape so that it is possible for us to find an occurrence of desirable congestion (so, technology innovation). For example, the supporting hyperplane in Fig. 2, where the horizontal axis indicates a desirable output (g) and the vertical axis indicates an undesirable output (b), has the three types of supporting hyperplane

such as a–c, d–e and f–h, each of which indicates positive, zero and negative Damage to Return (DTR), respectively. The degree of DTR is measured by (db/dg)/(b/g) in the case of a single component of production factors. As indicated in Fig. 2, technology innovation is identified on negative DTR because an increase in the desirable output decreases an amount of the undesirable output and vice versa. The concept of DTR is different from conventional RTS (Returns to Scale) and DTS (Damages to Scale). See Sueyoshi and Goto (2014a) for a detailed description on their conceptual and methodological differences. 3.3. Conceptual comparisons with conventional disposability concepts To discuss differences among proposed and conventional disposability concepts, this study starts a description on a possible occurrence of undesirable congestion and RTS, both of which are well-known concepts in production economics. Fig. 3 depicts an occurrence of undesirable congestion in the conventional context of economics. The figure consists of an input (x) on the horizontal axis and a desirable output (g) on the vertical axis. A curve indicates a production frontier between the input and the desirable output. This study clearly understands that the curve is for our descriptive convenience. The production frontier is usually specified by a piecewise linear couture line in DEA. The undesirable congestion occurs in the right hand side of Fig. 3 in which an input increase leads to a decrease in the desirable output. Considering a possible occurrence of the undesirable congestion in Figs. 3, Fig. 4 classifies the type of RTS into five different categories: IRTS (Increasing RTS), CRTS (Constant RTS), DRTS (Decreasing RTS), no RTS and negative RTS. The first three types of RTS have been long discussed in any textbook on production economics. However, no RTS and negative RTS are relatively new. They are originated from an occurrence of undesirable congestion. See Sueyoshi and Goto (2012d,h,i,k, 2013a). An important feature of Fig. 4 is that the type of RTS is determined by a supporting hyperplane, as visually discussed in Fig. 2. The importance of a supporting hyperplace can be easily confirmed in Fig. 4, as well. Unfortunately, the role of a supporting hyperplane has not sufficiently explored in conventional production economics.

Desirable Output Function

e Desirable Congestion (DC: Technology Innovation) on Undesirable Output Function

Desirable Output (g) & Undesirable Output (b) c

d

a Right Hand

Left Hand

xk Input (x) Fig. 1. Desirable and undesirable output functions under desirable congestion. (a) A production possibility set locates between the desirable and undesirable output function (i.s., the two line). (b) The left hand side indicates natural disposability, while the right hand side indicates managerial disposability under the unification of the two disposability concepts. (c) If there is no technology innovation, then an undesirable output can be considered as a byproduct of a desirable output so that the two frontier lines behave in a similar manner. In contrast, the technology innovation, in particular for pollution prevention, drastically changes the situation. The two frontier lines behave differently as found in the right hand side. (d) The right hand side visually describes a possible occurrence of desirable congestion, or technology innovation on an undesirable output. (e) Source: Sueyoshi and Goto (2014b) and Sueyoshi and Wang (2014). See Sueyoshi and Goto (2014d), as well.

T. Sueyoshi, D. Wang / Energy Economics 46 (2014) 360–374

d

c

f

Desirable Congestion (Technology Innovation)

e

h

a

Undesirable Output (b)

365

Desirable Output (g) Fig. 2. A possible occurrence of desirable congestion (technology innovation). (a) The three supporting hyperplaces (a–c, d–e and f–h) indicate positive, zero and negative DTR, respectively. The technology innovation is found in the supporting hyperplane (f–h) that has a negative slope. The negative slope indicates that an increase in the desirable output leads to a decrease in the undesirable output. This study looks for firms with such negative DTR in environmental assessment. (b) Source: Sueyoshi and Goto (2014b) and Sueyoshi and Wang (2014). See also Sueyoshi and Goto (2014d).

A conceptual difficulty associated with DEA environmental assessment is that it must deal with three groups of production factors, i.e., inputs, desirable and undesirable outputs. The three different types of production factors consist of “three research triangle” in the environmental assessment, which cannot be found in the conventional use of DEA. Moreover, the input classification, as discussed in this study, makes the proposed assessment more complicate and difficult.

n

The inequality constraints

G≤∑ G j λ j

! are assigned to desirable

j¼1

outputs in the two disposability concepts. The inequality constraints ! n

B ≤∑ B j λ j

are also assigned to undesirable outputs in the strong

j¼1

n

disposability. In contrast, the equality constraints

An important break-through was developed by Färe et al. (1986, 1989). Their study (1989, p. 92) has specified the following strong (S) and weak (W) disposability concepts on the two output vectors and an input vector, respectively: S

8 <

ðG; BÞ : G≤

n X

G j λ j ; B≤

n X

n X

g

j¼1

X jλ j;

n X

j¼1

A

λ j ¼ 1; λ j ≥ 0 ð j ¼ 1; …; nÞ

9 =

& : ; j¼1 j¼1 j¼1 j¼1 9 8 = < n n n n X X X X W P ðX Þ ¼ ðG; BÞ : G≤ G jλ j; B ¼ B jλ j; X ≥ X jλ j; λ j ¼ 1; λ j ≥0 ð j ¼ 1; …; nÞ : ; : j¼1

B jλ j; X ≥

assigned to undesirable outputs in the weak disposability. The concept of strong and weak disposability has long dominated the literature (e.g., Watanabe and Tanaka, 2007; Yang and Pollitt, 2010 and many other articles) on DEA environmental assessment. The concept of strong and weak disposability has the two important features, both of which have been insufficiently discussed in the

g

Negave RTS

j¼1

Congeson

No RTS

CRTS

O a

b

E

DRTS

B

O

are

j¼1

3.4. Weak and strong disposability

P ðX Þ ¼

!

B ¼ ∑ B jλ j

x

Fig. 3. An occurrence of undesirable congestion. (a) An occurrence of undesirable congestion implies that an input increase leads to a decrease in a desirable output. This type of congestion belongs to its conventional definition in production economics. (b) Source: Sueyoshi and Goto (2012f).

A

F B

D C IRTS

a

b

x

Fig. 4. Classification of Returns to Scale (RTS). (a) IRTS: Increasing RTS, CRTS: Constant RTS, DRTS: Decreasing RTS, no RTS and negative RTS. (b) The supporting hyperplane determines the type of RTS. It is not impossible, but very difficult to determine the shape of a production frontier by comparing efficiency scores. Rather, as discussed in this study, we formulate a dual model from an original primal model and then determine the shape from dual variables. (c) Source: Sueyoshi and Goto (2012f).

366

T. Sueyoshi, D. Wang / Energy Economics 46 (2014) 360–374

previous research efforts on production economics. One of the two important features is that undesirable outputs are “byproducts” of desirable outputs. Hence, the two types of outputs have similar production relationships to a change of inputs. Consequently, it is possible for us to consider a similar type of production functions that can express the three production factors. A problem is that technology innovation on undesirable outputs, or an occurrence of desirable congestion, is not incorporated in the conventional disposability concepts. The other unique feature is that, as depicted in Fig. 5, the assignment of equality ! n

constraints on undesirable outputs

B ¼ ∑ B jλ j

makes it possible

j¼1

to identify a possible occurrence of undesirable congestion on desirable outputs. In the case, it is assumed, as visually discussed in Fig. 5, that an input is linked to a desirable output so that an increase in the undesirable output decreases an amount of the desirable output and vice versa. As depicted in the right hand side of Fig. 5, the supporting hyperplane, due to the equality constraints on undesirable outputs, indicates a possible occurrence of a negative slope like the line (a–c). This result implies a possible occurrence of undesirable congestion so that an increase in the undesirable output may occur with a decrease in the desirable output. It is difficult for us to accept the result because this study looks for an opposite result. That is, DEA environmental assessment needs a possible occurrence of desirable congestion, or technology innovation, as depicted in Fig. 2, in which an increase in the desirable output occurs with a decrease in the undesirable output, vice versa.

g2j, …, gsj)T N 0, and Bj = (b1j, b2j, …, bhj)T N 0 for j = 1, …, n. Here, the superscript “T” indicates a vector transpose. The inequality (N) implies that all components of the three column vectors are strictly positive. In addition, it is necessary for us to pay attention to the following slack variables related to inputs, desirable and undesirable outputs: dxi ≥ 0 for i = 1, …, m, dgr ≥ 0 for r = 1, …, s, and dbf ≥ 0 for f = 1, …, h, respectively. The proposed DEA approach needs the vector of λ = (λ1, …, λn)T to express unknown intensity or structural variables to connect all production factors. The following data ranges related to inputs, desirable and undesirable outputs are used in the proposed approach: n  o n  o−1   xi j  j ¼ 1; …; n −min xi j j ¼ 1; …; n for i ¼ 1; …; m;  n  o n  o−1   g −1 Rr ¼ ðm þ s þ hÞ max g r j  j ¼ 1; …; n −min g r j j ¼ 1; …; n for r ¼ 1; …; s; &  n  o n  o−1   b −1 R f ¼ ðm þ s þ hÞ max b f j j ¼ 1; …; n −min b f j j ¼ 1; …; n for f ¼ 1; …; h: x

Ri ¼ ðm þ s þ hÞ

−1



max

All the three data ranges are identified from an observed data set so that they are given to us before computing the proposed DEA assessment. Later, the inputs are further separated into two categories by the two disposability concepts. However, it is not necessary for us to change anything on the input range because it is determined by observations on each input. The previous research works (e.g., Sueyoshi and Goto, 2012a) have first specified the following non-radial model to measure the unified efficiency of the k-th DMU:

4. Formulations for DEA environmental assessment Maximize 4.1. Unified Efficiency (UE)

m X

s h   X X x xþ x− g g b b þ Ri di þ di Rr dr þ Rf df

The unified (operational and environmental) performance of DMUs is characterized by their production activities that utilize inputs to yield desirable and undesirable outputs as production factors. As mentioned previously, an important feature of DEA environmental assessment is that the achievement of each DMU is relatively compared with those of the remaining others. The performance level is referred to as “an efficiency measure”. Mathematical symbols to express such production factors in DEA are rewritten as follows: Xj = (x1j, x2j, …, xmj)T N 0, Gj = (g1j,

s:t:

j¼1 n X j¼1 n X

f ¼1

r¼1

i¼1

n X

xþ xi j λ j −di g

g r j λ j −dr

þ

x− di

¼ xik ði ¼ 1; …; mÞ; ¼ g rk ðr ¼ 1; …; sÞ;

b

b f jλ j þ d f

¼ b f k ð f ¼ 1; …; hÞ;

ð1Þ

j¼1

n X

λ j ¼ 1;

j¼1

λ j ≥ 0 ð j ¼ 1; …; nÞ; xþ x− di ≥ 0 ði ¼ 1; …; mÞ; di ≥ 0 ði ¼ 1; …; mÞ; g b dr ≥ 0 ðr ¼ 1; …; sÞ & d f ≥ 0 ð f ¼ 1; …; hÞ: After solving Model (1), the level of Unified Efficiency (UE) of the k-th DMU is determined by

a Undesirable Congestion

c

Desirable Output (g)

0 1 m s h   X X X x xþ x− g g b b UE ¼ 1−@ þ Ri di þ di Rr dr þ R f d f A; i¼1

Undesirable Output (b) Fig. 5. A possible occurrence of undesirable congestion. (a) The assignment of equality n

constraints on undesirable outputs (B ¼ ∑ B j λ j ) makes it possible to identify a possible j¼1

occurrence of undesirable congestion on desirable outputs. (b) The occurrence of undesirable congestion indicates that an increase in the undesirable output may occur with a decrease in the desirable output. It is difficult to accept because this study looks for an opposite result (a possible occurrence of desirable congestion or technology innovation), as depicted in Fig. 2.

r¼1

ð2Þ

f ¼1

where all the slacks within the parentheses are obtained from the optimality of Model (1). It is important to note that the two slacks related to the i-th input are mathematically defined as dxi + = (|dxi | + dxi )/2 and dxi − = (|dxi | − dxi )/2. They are mutually exclusive. Otherwise, Model (1) produces an “unbounded” solution because the dual formulation of Model (1) is infeasible. Therefore, a simultaneous occurrence of both dxi + N 0 and dxi − N 0 (i = 1, …, m) should be excluded from the optimal solution of Model (1). When such a case occurs on Model (1), it is necessary for Model (1) to incorporate the nonlinear conditions dxi +dxi − = 0 (i = 1, …, m), as additional constraints, that are solved by nonlinear or mixed integer programming. See Sueyoshi and Goto (2012a) for a detailed description on how to handle the computational difficulty by using nonlinear or mixed integer programming.

T. Sueyoshi, D. Wang / Energy Economics 46 (2014) 360–374

367

Like UEN, the degree of UEM on the k-th DMU is measured by Eq. (4), !

4.2. Unified Efficiency under Natural disposability (UEN)

m

The previous studies (e.g., Sueyoshi and Goto, 2012a) have proposed the following non-radial model to measure the level of UEN of the k-th DMU by maintaining only a slack vector of inputs (+dxi ): m X

Maximize

x x

Ri di þ

s:t:

g g

Rr dr þ

h X

b b

¼ xik

ði ¼ 1; …; mÞ;

g g r j λ j −dr

¼ g rk

ðr ¼ 1 ; …; sÞ;

b f jλ j þ d f ¼ b f k

ð f ¼ 1 ; …; hÞ;

j¼1

n X j¼1 n X

ð3Þ b

j¼1

n X

λ j ¼ 1;

j¼1 x

λ j ≥ 0 ð j ¼ 1; …; nÞ; di ≥ 0 ði ¼ 1; …; mÞ; g dr

b

≥ 0 ðr ¼ 1; …; sÞ & d f ≥ 0 ð f ¼ 1; …; hÞ:

Model (3) can measure the level of UEN. The model considers the input-related deviations + dxi (i = 1, …, m) to attain the status of natural disposability where all inputs decrease for improving the operational performance of the k-th DMU, while satisfying the requirements (i.e., the location of efficiency frontiers) on desirable and undesirable outputs. The UEN of the k-th DMU is determined by 0 1 m s h X X X x x g g b b A @ UEN ¼ 1− Ri di þ Rr dr þ Rf df ;

ð4Þ

f ¼1

r¼1

i¼1

h

g

b

, where all slack

f ¼1

r¼1

i¼1

variables are determined on the optimality of Model (5). The equation within the parentheses, obtained from Model (5), indicates the level of unified inefficiency. The UEM is obtained by subtracting the level of inefficiency from unity. 4.4. Unified Efficiency under Natural and Managerial disposability (UENM)

x di

xi j λ j þ

s

x

UEM ¼ 1− ∑ Rxi di þ ∑ Rgr dr þ ∑ Rbf d f

Rf df

f ¼1

r¼1

i¼1

n X

s X

or

where all slack variables are identified on the optimality of Model (3). The equation within the parentheses, obtained from Model (3), indicates the level of unified inefficiency. The UEN is obtained by subtracting the level of inefficiency from unity in Eq. (4).

All the models (1), (3) and (5) were discussed by the previous studies (e.g., Sueyoshi and Goto, 2012a). Hereafter, this study utilizes another type of non-radial model that simultaneously incorporates natural and managerial disposability (e.g., Goto et al., 2014). The proposed model can measure the level regarding UENM of the k-th DMU. Under the natural disposability, the DMU needs to decrease its input vector to improve the operational performance. In contrast, the DMU needs to increase an input vector to improve the environmental performance under managerial disposability. To satisfy the two conflicting requirements, the proposed non-radial model separates inputs into two categories according to the two disposability concepts. To unify the two disposability concepts, this study uses the following unified model:

−

Maximize s:t:

m X i¼1 n X j¼1 n X j¼1 n X j¼1 n X j¼1 n X

þ

x x−

R i di

m X

þ

x xþ

R q dq þ

−

x−

x i j λ j þ di þ

s X

g g

Rr dr þ

h X

b b

Rf df

f ¼1

r¼1

q¼1

¼ xik

−

ði ¼ 1; …; m Þ;

−

þ

þ

xq j λ j −dq

xþ

¼ xqk

ðq ¼ 1; …; m Þ;

g

¼ grk

ðr ¼ 1 ; …; sÞ;

gr j λ j −dr

b

b f jλ j þ d f ¼ b f k

ð6Þ

ð f ¼ 1 ; …; hÞ;

λ j ¼ 1;

j¼1

−

x−

λ j ≥0 ð j ¼ 1; …; nÞ; di ≥0 ði ¼ 1; …; m Þ; xþ þ g dq ≥0 ðq ¼ 1; …; m Þ; dr ≥0 ðr ¼ 1; …; sÞ & b

d f ≥0 ð f ¼ 1; …; hÞ:

4.3. Unified Efficiency under Managerial disposability (UEM) The previous studies (e.g., Sueyoshi and Goto, 2012a) have proposed the following non-radial model to measure the level of UEM of the k-th DMU by maintaining only a slack vector of inputs (− dxi ) of Model (1): Maximize

m X

x x

Ri d i þ

s:t:

j¼1 n X j¼1 n X

g g

Rr dr þ

x xi j λ j −di

h X

b b

Rf df

f ¼1

r¼1

i¼1

n X

s X

¼ xik

ði ¼ 1; …; mÞ;

gr j λ j −dr ¼ g rk

ðr ¼ 1 ; …; sÞ;

g

ð5Þ b

b f jλ j þ d f ¼ b f k

ð f ¼ 1 ; …; hÞ;

The number of original m inputs is newly separated into m− (under natural disposability) and m+ (under managerial disposability) in Model (6). So, the model maintains m = m− + m+. One of the two input categories uses inputs whose slacks (dx− for i = 1, …, m−) are fori mulated under the natural disposability. For example, the number of employees belongs to the input category. Meanwhile, the other category contains inputs whose slacks (dx+ for i = 1, …, m+) are formulated i under managerial disposability. For example, the amount of capital investment for technology innovation belongs to the input category. Another example is R&D expenditure. The amount of capital investment to attain technology innovation is important for enhancing the level of production activity and environmental protection. The UENM of the k-th DMU is determined by

j¼1

n X

0

λ j ¼ 1;

j¼1 x

λ j ≥ 0 ð j ¼ 1; …; nÞ; di ≥ 0 ði ¼ 1; …; mÞ; g

UENM ¼ 1−@

−

m X i¼1

þ

x x− Ri di

þ

m X q¼1

x xþ Rq dq

þ

s X r¼1

g g R r dr

þ

h X

1 b b R f d f A;

ð7Þ

f ¼1

b

dr ≥ 0 ðr ¼ 1; …; sÞ & d f ≥ 0 ð f ¼ 1; …; hÞ: Model (5) measures the level of UEM. The model considers the input-related deviations − dxi (i = 1, …, m) to attain the status of managerial disposability where all inputs increase for improving the environmental performance of the k-th DMU, while satisfying the requirement on desirable and undesirable outputs.

where all slack variables are determined on the optimality of Model (6). The equation within the parentheses, obtained from the optimality of Model (6), indicates the level of unified inefficiency. The UENM is obtained by subtracting the level of inefficiency from unity.

368

T. Sueyoshi, D. Wang / Energy Economics 46 (2014) 360–374

The Unified Efficiency under Natural and Managerial disposability, or UENM(DC), of the k-th DMU is measured like Eq. (4). More specifically, it becomes as follows:

Model (6) has the following dual formulation: −

Minimize

m X

þ

m s h X X X − þ vi xik − zq xqk − ur grk þ wf bfk þ σ q¼1

j¼1 −

s:t:

m X

s X

þ zq xq j −

q¼1

j¼1

f ¼1

r¼1

þ

m X

− v i xi j −

ur g r j þ

−

x

zq ≥ Rq

0

w f b f j þ σ ≥ 0 ð j ¼ 1; …; nÞ;

f ¼1

r¼1

x

vi ≥ Ri

h X

g

ur ≥ Rr

b

wf ≥Rf

σ : URS:

ði ¼ 1; …; m Þ; þ ðq ¼ 1; …; m Þ; ðr ¼ 1; …; sÞ; ð f ¼ 1; …; hÞ &

ð8Þ where vi (i = 1, …, m−), zq (q = 1, …, m+), ur (r = 1, …, s) and wf (f = 1, …, h) are all dual variables related to the first, second, third and fourth groups of constraints in Model (6). The dual variable (σ), which is unrestricted (URS), is obtained from the last equation of Model (6). At the end of this section, it is important to note that we can solve Model (6) by linear programming. Thus, the model has computational tractability, compared with Model (1). Moreover, Model (1) incorporates natural and managerial disposability in the formulation although inputs and outputs are based upon the two disposability concepts. Such unique features provide this study with an analytical capability to examine an effect of UENM where the two disposability concepts are equally treated in the evaluation.

UENMðDCÞ ¼ 1−@

−

m X

þ

x x Ri di

þ

m X

x x Rq dq

þ

1 b b R f d f A;

ð10Þ

f ¼1

q¼1

i¼1

h X

where all slack variables are determined on the optimality of Model (9). The equation within the parentheses, obtained from the optimality of Model (9), indicates the level of unified inefficiency. The level of UENM(DC) is obtained by subtracting the level of inefficiency from unity under a possible occurrence of DC, or technology innovation, on undesirable outputs. Model (9) has the following dual formulation: −

Minimize

m X

þ

m s h X X X − þ vi xik − zq xqk þ ur g rk − wf bfk þ σ q¼1

i¼1 −

s:t:

m X

r¼1

f ¼1

þ

− vi xi j

m s h X X X þ − zq xq j þ ur g r j − w f b f j þ σ ≥ 0 ð j ¼ 1; …; nÞ; r¼1

q¼1

i¼1 x

vi ≥Ri

f ¼1

−

x

zq ≥Rq

ur : URS

b

w f ≥R f

σ : URS;

ði ¼ 1; …; m Þ; þ ðq ¼ 1; …; m Þ; ðr ¼ 1; …; sÞ; ð f ¼ 1; …; hÞ &

4.5. Unified Efficiency under Natural & Managerial disposability: UENM(DC) with a possible occurrence of desirable congestion (technology innovation)

ð11Þ

As an alternative for unifying Models (3) and (5), Sueyoshi and Goto (2014b) and Wang et al. (2014) have proposed the following non-radial model in which the k-th DMU simultaneously attains the status of natural and managerial disposability along with a possible occurrence of desirable congestion (DC: technology innovation):

where vi (i = 1, …, m−), zq (q = 1, …, m+), ur (r = 1, …, s) and wf (f = 1, …, h) are all dual variables related to the first, second, third and fourth groups of constraints in Model (9). The dual variable (σ), which is unrestricted (URS), is obtained from the last equation of Model (9). An important feature of Model (11) is that it can be used for identifying investment strategy by the following rule.

−

Maximize s:t:

m X i¼1 n X j¼1 n X j¼1 n X j¼1 n X j¼1 n X

þ

x x

Ri di þ

m X

h X

x x

Rq dq þ

−

x

x i j λ j þ di þ

x

b b

4.6. Investment strategy

Rf df

f ¼1

q¼1 −

−

þ

þ

xq j λ j −dq

¼ xqk

ðq ¼ 1; …; m Þ;

gr j λ j

¼ g rk

ðr ¼ 1 ; …; sÞ;

b

b f jλ j− d f

After solving Model (11), an occurrence of DC, or technology innovation, is identified by the following rule with the assumption on a unique optimal solution (Sueyoshi and Goto, 2014b; Sueyoshi and Wang, 2014; Wang et al., 2014):

ði ¼ 1; …; m Þ;

¼ xik

¼ bfk

ð9Þ

ð f ¼ 1 ; …; hÞ;

λ j ¼ 1;

j¼1

−

x

λ j ≥ 0 ð j ¼ 1; …; nÞ; di ≥0 ði ¼ 1; …; m Þ; x dq

þ

≥0 ðq ¼ 1; …; m Þ &

b d f ≥0

ð f ¼ 1; …; hÞ;

where the number of original m inputs is newly separated into m− (under natural disposability) and m+ (under managerial disposability), as mentioned previously. The model maintains the relationship (m = m− + m+). There are two differences between UENM and UENM(DC). One of the two differences is that UENM(DC) considers a possible occurrence of DC, or technology innovation to reduce an amount of undesirable outputs, as depicted in Figs. 1 and 2, in Model (9). In contrast, UENM does not consider such an important occurrence in Model (6). The other difference is that the third group of constraints related to desirable outputs ! n

does not have any slacks ∑ g r j λ j ¼ grk so that Model (9) can analytj¼1

ically duplicate Figs. 1 and 2 under natural and managerial disposability concepts. Consequently, Model (9) can measure an occurrence of positive, zero and negative DTR, as visually specified in Fig. 2.

(a) if u⁎r = 0 for some (at least one) r, then “zero DTR” occurs on the k-th DMU, (b) if u⁎r b 0 for some (at least one) r, then “negative DTR” occurs on the k-th DMU and (c) if u⁎r N 0 for all r, then “positive DTR” occurs on the k-th DMU. Note that if u⁎r b 0 occurs on some r and ur⁎' = 0 occurs on another r′, then this study considers that the negative DTR occurs on the k-th DMU, indicating a status of DC, or technology innovation, on undesirable outputs. It is indeed true that u⁎r b 0 for all r are the best case because an increase in any desirable output always decreases an amount of undesirable outputs. Meanwhile, if u⁎r b 0 is identified for some r, then it indicates that there is a chance to reduce an amount of undesirable output(s). Therefore, this study includes the second case for considering an investment opportunity. Under an occurrence of negative DTR (i.e., u⁎r b 0 for at least one r), the effect of investment on undesirable outputs is determined by the following rule: (a) if zq⁎ N Rxq for q, then the q-th input for investment under managerial disposability can “effectively” decrease an amount of undesirable outputs and (b) if zq⁎ = Rxq for q, then the q-th input for investment has a “limited” effect on decreasing an amount of undesirable outputs.

T. Sueyoshi, D. Wang / Energy Economics 46 (2014) 360–374

It is important to note that the investment on inputs under managerial disposability is not recommended in the other two cases (i.e., positive and zero DTR) as depicted in Fig. 2. Furthermore, this study uses “a limited effect” in the second case. The term implies that if zq⁎ = Rxq is found on the q-th input, then there is a high likelihood that zq⁎ may become a very small positive number. So, the investment has only a limited effect in reducing an amount of undesirable outputs. Moreover, zq⁎ N Rxq is required for some q, but not necessary for all q. See Sueyoshi and Goto (2014d) in the case of radial measurement.

5. Empirical results 5.1. Data The study utilizes a data set on oil and natural gas production companies, all of which are listed in NYSE Energy Index (2013). The index covers 82 independent producers and 20 integrated companies. This study restricts the proposed analysis to the petroleum companies with operations in the United States. The amount of their greenhouse gas emissions is obtained from EPA's GHG Reporting Program (2013b). The program has requested facilities emitting more than 25,000 metric tons of CO2 equivalents per year to submit mandatory reporting on GHG emissions from 2010. The sources covered by the program account for 85– 90% of the total U.S. GHG emission. The emission data set is reported at the facility level. Each company may operate multiple oil and gas projects in different regions. This study extracts the emission data from all onshore oil/gas production sites and then aggregates them at a firm-level. A company's emission is closely related to its drilling activity. Therefore, this study extracts the number of wells drilled by a company from its annual report. In alignment with the emission data, the drilled wells used in this study count only those drilled onshore in the United States. We also obtain the three-year average during 2010–2012 on Acquisition, Finding and Development (AFD) expenses from Young (2013). The AFD expenses reflect companies' abilities to access and obtain petroleum reserves, which indirectly affect the GHG emission. Higher acquisition cost is generally paid for easy accessibility to resources which require less effort to drill, hence incurring less emission. Higher finding and development cost may imply more intensive exploration, hence producing more emission from exploration. Finally, this study collects the companies' financial and operational data from COMPUSTAT database. The data set used in this study is summarized below in detail.

369

(a) GHG emission: This measures an amount of emissions from onshore/offshore production of a company, including not only CO2 but also methane (CH4) and nitrous oxide (N2O). The cost of adapting pollution prevention practices and the effectiveness of pollution prevention as a strategy for reducing emissions may vary with a scale of current emission. (b) Number of employees: This is regarded as a proxy for a firm size. Larger firms may have more resources to adapt GHG mitigation practices. (c) Capital expenditure: This is included to indicate the operating liquidity of a firm. Firms with higher capital expenditure may invest more in GHG mitigation. (d) R&D expense: This is a measure of a firm's technology capacity and serves a proxy for technology innovation. It is expected that firms with higher R&D expense are more likely to acquire and implement efficient emission control technology. (e) Total assets: This includes current assets, property, plant and equipment, all of which are used as another proxy for a corporate size. (f) Number of net wells drilled: This gives the number of wells drilled by a company for the calendar year, accounting for fractional working interest owned by the company in each well. For instance, a 50% interest in a well is counted as 0.5 well. (g) Revenue: This is income received from sale of oil and gas, and indicates an operational size of the business. (h) Acquisition, Finding and Development (AFD) expense: This is calculated as the sum of proved reserve acquisition cost as well as finding and development cost. Specifically, the proved reserve acquisition cost is calculated as reserve purchasing cost divided by the reserve purchased. The finding and development cost is calculated as unproved reserve acquisition cost, exploration and development expenditures relative to the added reserve. Overall, this study computes the sum of acquisition, finding and development costs as an aggregate measure of each company's ability to obtain access to per barrel of oil equivalent (BOE) resources. This study utilizes one desirable output (s = 1): total revenue, one undesirable output (h = 1): an amount of GHG emission, five inputs under natural disposability (m− = 5): the number of employees, the amount of capital expenditure, the total assets, the number of net wells drilled, and the amount of total AFD expense, one input under managerial disposability (m+ = 1): the amount of R&D expenditure.

Table 1 Descriptive statistics. Outputs & inputs

Desirable output

Undesirable output

Variables

Revenue

GHG emission

AFD expense

R&D expense

Total assets

Inputs Capital expenditure

Employees

Net wells drilled

Unit

Million $

1,000 tons

$ per BOE

Million $

Million $

Million $

1000

Well

Independent companies Avg. 6650.2026 S.D. 14103.9993 Min. 356.1330 Max. 72556.0000

775.4225 657.2929 110.0000 3521.8010

39.1163 25.1597 7.0000 135.0000

271.5116 492.0496 2.0000 1946.0000

18534.2787 28230.3581 1381.7880 129273.0000

3527.6308 4429.4995 485.4790 20837.0000

5.06 11.49 0.12 60.24

344.38 337.93 26.20 1411.20

Integrated companies Avg. S.D. Min. Max.

232695.4759 189253.3820 28616.3310 467153.0000

6849.6230 8385.8624 558.7897 24519.1410

57.7143 16.3678 26.0000 77.0000

1188.5714 625.6942 60.0000 1840.0000

187925.9490 153499.0642 17522.6430 360325.0000

19100.2316 14421.3860 1369.0000 34271.0000

51.65 34.10 9.19 87.00

468.00 328.60 107.50 951.00

Overall Avg. S.D. Min. Max.

38296.5409 104085.5747 356.1330 467153.0000

1625.8106 3676.1641 110.0000 24519.1410

41.7200 24.8572 7.0000 135.0000

399.9000 598.9874 2.0000 1946.0000

42249.1126 84222.6042 1381.7880 360325.0000

5707.7949 8489.8539 485.4790 34271.0000

11.58 22.85 0.12 87.00

361.68 336.12 26.20 1411.20

370

T. Sueyoshi, D. Wang / Energy Economics 46 (2014) 360–374

Table 2 Unified efficiency measures. Company name Independent companies Anadarko Petroleum Corporation Antero Resources LLC Apache Corporation Berry Petroleum Company BHP Billiton Group Bill Barrett Corporation Cabot Oil & Gas Corporation Chesapeake Energy Corporation Cimarex Energy Co. Concho Resources Conoco Phillips CONSOL Energy Inc. Continental Resources, Inc. Denbury Resources Inc. Devon Energy Corporation EnCana Corporation Energen Corporation EOG Resources, Inc. EP Energy LLC EQT Corporation EXCO Resources, Inc. Forest Oil Corporation Laredo Petroleum Holdings, Inc. Linn Energy, LLC National Fuel Gas Company Newfield Exploration Company Noble Energy, Inc. Occidental Petroleum Corporation PDC Energy, Inc. Pioneer Natural Resources Company Plains E & P Company QEP Resources, Inc. Quicksilver Resources, Inc. Range Resources Corporation Rosetta Resources Inc. SandRidge Energy, Inc. SM Energy Company Southwestern Energy Company Swift Energy Inc. Talisman Energy Inc. Ultra Petroleum Corporation Whiting Petroleum Corporation WPX Energy, Inc. Avg. Max. Min. S.D. Integrated companies BP PLC Chevron Corporation Exxon Mobil Corporation Hess Corporation Marathon Oil Corporation Murphy Oil Corporation Royal Dutch Shell PLC Avg. Max. Min. S.D. P-value (Welch's t-test)

UE

UEN

UEM

UENM

UENM(DC)

1.0000 1.0000 1.0000 1.0000 1.0000 0.9999 1.0000 1.0000 0.9584 1.0000 1.0000 0.9991 1.0000 0.9998 1.0000 1.0000 0.9999 1.0000 1.0000 0.9967 1.0000 0.9277 0.9997 1.0000 0.9824 0.9552 1.0000 1.0000

0.7495 1.0000 0.7484 1.0000 1.0000 0.9627 1.0000 0.7335 0.9683 0.9156 0.7910 1.0000 0.9438 1.0000 0.8070 1.0000 0.9467 0.8622 0.9725 0.9639 0.9723 0.9134 1.0000 1.0000 0.9088 0.9266 0.8978 0.6949

1.0000 0.7837 0.8878 0.6758 1.0000 0.7163 0.6473 0.8467 0.7488 0.8873 0.7811 0.8400 0.7977 0.8991 0.6644 1.0000 0.8391 0.7984 0.7805 0.8401 0.8253 0.8418 0.8886 0.7266 1.0000 0.8115 0.7896 1.0000

1.0000 1.0000 0.9390 0.9312 1.0000 0.9591 1.0000 0.8112 0.9713 0.9196 1.0000 0.9760 0.9457 0.9758 0.9479 1.0000 0.9474 0.8836 1.0000 0.9760 0.9724 0.9136 1.0000 0.9363 0.9008 0.9300 0.9520 0.8167

1.0000 1.0000 1.0000 1.0000 1.0000 0.9904 1.0000 1.0000 0.9773 0.9218 1.0000 0.9766 0.9495 0.9758 1.0000 1.0000 0.9503 0.8899 1.0000 0.9781 0.9999 0.9281 1.0000 0.9444 0.9015 0.9350 1.0000 0.8192

1.0000 0.9878 0.7750 1.0000 1.0000 0.8931 0.7334 0.9244

1.0000 0.9398

1.0000 0.9809 0.9998 0.9999 1.0000 1.0000 1.0000 1.0000 0.9998 0.9767 1.0000 0.9485 0.9797 0.9931 1.0000 0.9277 0.0163

1.0000 0.9410 1.0000 0.9464 1.0000 0.8932 0.9551 1.0000 0.9750 0.9146 1.0000 0.9295 0.9311 0.9313 1.0000 0.6949 0.0825

1.0000 0.7821 0.8719 0.9061 0.8031 0.6766 0.6941 0.9359 0.9152 0.8075 1.0000 0.8398 0.7222 0.8321 1.0000 0.6473 0.1022

1.0000 0.9410 1.0000 1.0000 1.0000 0.8917 0.9619 1.0000 0.9785 0.9230 1.0000 1.0000 0.9402 0.9574 1.0000 0.8112 0.0473

0.9142 0.9462 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 0.9269 1.0000 1.0000 0.9488 0.9724 1.0000 0.8192 0.0407

1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 0.0000 0.0085

1.0000 1.0000 1.0000 0.8498 1.0000 1.0000 1.0000 0.9785 1.0000 0.8498 0.0568 0.0849

1.0000 1.0000 1.0000 0.7055 0.6733 0.8192 1.0000 0.8854 1.0000 0.6733 0.1496 0.3937

1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 0.0000 0.0000

1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 0.0000 0.0001

In collecting the data set, this study removes companies with missing data in any of the relevant fields. Eventually, we have obtained a set of fifty companies (n = 50), including forty-three independent companies and seven integrated companies. The data set consists of roughly half of all production companies in NYSE Energy Index. Meanwhile, the total emission from these 50 companies is 82.3 MMT emissions from the entire onshore production segment in the United States.

Table 1 summarizes descriptive statistics of the data set used in this study where Avg., S.D., Min. and Max. stand for average, standard deviation, minimum and maximum, respectively. On average, integrated companies are larger than independent companies in all data fields. The GHG emissions by integrated firms have average 6849.6230 thousand tons, being 8.8 times of the average emission of 775.4225 thousand tons produced by independent companies. The largest emission of 24,529 thousand tons comes from Exxon Mobil, whereas EnCana Corporation has the smallest emission of 110 thousand tons. The integrated companies drill 468 wells on average, being 36% higher than the independent companies. However, the most active driller, Occidental Petroleum Corporation with 1411.2 net wells drilled, is an independent company. See the end of this study that provides the data set used for this empirical study. 5.2. Unified efficiency measures This research programmed the proposed DEA models for UE, UEN, UEM, UENM and UENM(DC) by Matlab (Version 2012). The computer used for this study is a Dell workstation with Intel i5 CPU of 3.10 GHz and RAM of 8.0 GB. See the end of this study that lists a computer code. Table 2 documents the five efficiency measures, along with the Pvalue of the t-test at the bottom, to confirm the hypothesis on independent and integrated oil companies that is summarized in Section 2. The worst performers among independent companies are Forest Oil in UE, Occidental Petroleum in UEN, Cabot Oil & Gas Corporation in UEM, Chesapeake Energy Corporation in UENM, Occidental Petroleum in UENM(DC) and the average efficiency. Among integrated companies, the worst performers are Hess Corporation in UEN and Marathon Oil in UEM. All seven integrated companies have efficiency of 1.0000 in UE, UENM and UENM(DC). Moreover, there are seven companies with an efficiency score of 1.0000 in all five measures, including three independent companies (i.e., BHP Billiton, EnCana, and Ultra Petroleum) and four integrated companies (i.e., BP, Chevron, Exxon Mobil, and Royal Dutch Shell). On average, Occidental Petroleum and Chesapeake Energy have lower efficiency scores. The relatively low performance of both companies can be partly attributed to their significant stakes in shale plays. The development of shale wells releases more GHG especially methane than conventional wells due to hydraulic fracturing (Howarth et al., 2011). EPA issued new rules to reduce air emissions from hydraulic fracturing (EPA, 2012c). In addition, the low price on natural gas in recent years also undermined their revenues. Overall, the efficiency scores of independent petroleum companies are 0.9931 in UE, 0.9313 in UEN, 0.8321 in UEM, 0.9574 in UENM and 0.9724 in UENM(DC) on average, while these scores of integrated companies are 1.0000, 0.9785, 0.8854, 1.0000 and 1.0000, on average. Thus, the integrated companies outperform the independent companies on average in terms of all five efficiency measures. The largest difference occurs in UEM and the smallest difference occurs in UE. This study has applied the t-test to the integrated and independent companies' efficiency scores to statistically examine the hypothesis summarized in Section 2. The results in the bottom row in the table indicate that integrated companies have higher UE, UENM and UENM(DC) than independent companies at 1% significance level, and higher UEN at 10% significance level. Therefore, this study accepts the hypothesis that firms with a supply chain outperform those without the supply chain. An exception can be found in the t-test on UEM between the two groups because U.S. environmental regulation is equally effective on the two types of petroleum firms. The results in Table 2 also imply that the unified (operational and environmental) performance of U.S. petroleum firms is positively affected by the size of their supply chains. In fact, the four consistently efficient integrated companies have more branded retail outlets than the integrated companies (National Association of Convenience Stores, 2013). A larger retail network is associated with higher visibility to consumers, hence resulting in higher market pressure on the unified

T. Sueyoshi, D. Wang / Energy Economics 46 (2014) 360–374

371

Table 3 Dual variables, DTR and investment opportunity. Company name

Dual Variables

DTR

Revenue

GHG emission

AFD expense

R&D expense

Total assets

Capital expenditure

Employees

Net wells drilled

Independent companies Anadarko Petroleum Corporation Antero Resources LLC Apache Corporation Berry Petroleum Company BHP Billiton Group Bill Barrett Corporation Cabot Oil & Gas Corporation Chesapeake Energy Corporation Cimarex Energy Co. Concho Resources Conoco Phillips CONSOL Energy Inc. Continental Resources, Inc. Denbury Resources Inc. Devon Energy Corporation EnCana Corporation Energen Corporation EOG Resources, Inc. EP Energy LLC EQT Corporation EXCO Resources, Inc. Forest Oil Corporation Laredo Petroleum Holdings, Inc. Linn Energy, LLC National Fuel Gas Company Newfield Exploration Company Noble Energy, Inc. Occidental Petroleum Corporation PDC Energy, Inc. Pioneer Natural Resources Company Plains E & P Company QEP Resources, Inc. Quicksilver Resources, Inc. Range Resources Corporation Rosetta Resources Inc. SandRidge Energy, Inc. SM Energy Company Southwestern Energy Company Swift Energy Inc. Talisman Energy Inc. Ultra Petroleum Corporation Whiting Petroleum Corporation WPX Energy, Inc.

0.0000 −0.0011 0.0018 0.0001 0.0002 0.0000 −0.0010 0.0025 0.0000 0.0000 −0.0001 0.0000 0.0000 0.0000 0.0022 −0.0009 0.0000 0.0000 −0.0046 0.0000 −0.0001 0.0000 0.0000 0.0000 0.0000 0.0000 0.0019 0.0000 0.0001 0.0000 0.0000 0.0000 −0.0009 −0.0057 −0.0056 0.0038 −0.0051 0.0000 0.0000 0.0000 −0.0076 −0.0063 0.0000

0.0000 0.0000 0.0001 0.0001 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0001 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000

0.5936 3.3567 0.0142 0.1036 1.1654 0.0010 3.5498 0.1458 0.0010 0.0010 0.7317 0.0013 0.0010 0.0010 0.6130 2.6179 0.0010 0.0010 0.0409 0.0010 0.0070 0.0010 0.0010 0.0010 0.0010 0.0010 0.1567 0.0010 0.4467 0.0010 0.0010 0.0010 2.1018 0.0132 1.7752 1.1917 0.0108 0.0010 0.0010 0.0010 0.0047 0.0129 0.0010

0.0950 0.0311 0.1942 0.0555 0.2224 0.0001 0.0482 0.0330 0.0001 0.0001 0.1747 0.0001 0.0001 0.0001 0.1067 0.1984 0.0001 0.0001 0.0534 0.0001 0.0076 0.0001 0.0002 0.0001 0.0001 0.0001 0.1867 0.0001 0.0931 0.0001 0.0001 0.0001 0.1040 0.2210 0.0832 0.0091 0.2333 0.0001 0.0002 0.0001 0.1951 0.1517 0.0001

0.0004 0.0006 0.0000 0.0026 0.0002 0.0000 0.0007 0.0001 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0002 0.0015 0.0000 0.0000 0.0000 0.0000 0.0001 0.0000 0.0000 0.0000 0.0000 0.0000 0.0001 0.0000 0.0003 0.0000 0.0000 0.0000 0.0005 0.0081 0.0050 0.0004 0.0079 0.0000 0.0000 0.0000 0.0053 0.0012 0.0000

0.0044 0.0022 0.0001 0.0186 0.0005 0.0000 0.0100 0.0009 0.0000 0.0000 0.0007 0.0000 0.0000 0.0000 0.0015 0.0013 0.0000 0.0000 0.0006 0.0000 0.0488 0.0000 0.0000 0.0000 0.0000 0.0000 0.0467 0.0000 0.0057 0.0000 0.0000 0.0000 0.0062 0.0005 0.0052 0.0068 0.0002 0.0000 0.0000 0.0000 0.0008 0.0376 0.0000

0.7119 1.6896 7.2572 0.4610 0.2067 0.0728 2.1386 0.1857 0.0014 0.0014 2.5243 0.0014 0.0092 0.0014 0.6841 1.4290 0.0014 0.0014 32.3743 0.0014 0.1416 0.0014 0.0061 0.0014 0.0014 0.0014 0.1438 0.0014 1.8094 0.0014 0.0014 0.0014 2.1241 15.5337 13.3544 0.5981 16.0648 0.0014 0.0097 0.0014 23.8990 21.3112 0.0014

0.0147 0.0878 0.2511 0.0050 0.4496 0.0001 0.0829 0.0061 0.0001 0.0001 0.2273 0.0001 0.0001 0.0001 0.0658 0.2933 0.0001 0.0001 0.0196 0.0001 0.0588 0.0002 0.0001 0.0001 0.0001 0.0001 0.0057 0.0001 0.3986 0.0001 0.0001 0.0001 0.3420 0.0023 0.0253 0.0021 0.0420 0.0001 0.0001 0.0001 0.1390 0.0046 0.0001

Z N P P P Z N P Z Z N Z Z Z P N Z Z N Z N Z Z Z Z Z P Z P Z Z Z N N N P N Z Z Z N N Z

Integrated companies BP PLC Chevron Corporation Exxon Mobil Corporation Hess Corporation Marathon Oil Corporation Murphy Oil Corporation Royal Dutch Shell plc

−0.0016 0.0000 0.0000 −0.0003 −0.0016 0.0000 −0.0011

0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000

0.2816 0.0002 0.1146 0.1206 0.3081 0.0010 0.5164

0.0696 0.0000 0.0022 0.1538 0.0817 0.0001 0.0247

0.0011 0.0000 0.0001 0.0005 0.0020 0.0000 0.0004

0.0092 0.0000 0.0002 0.0045 0.0047 0.0000 0.0024

0.3624 0.0014 0.1695 0.4568 0.4492 0.0014 2.1316

0.1491 0.0000 0.0065 0.1886 0.0266 0.0002 0.1379

N Z Z N N Z N

performance. It can be conjectured that consumer pressure in the downstream spreads over the supply chain and gives an impact on the performance of exploration and production in the upstream. 5.3. Investment strategy Table 3 summarizes dual variables, the type of DTR and investment strategy, all of which are obtained from the optimal solution of Model (11). In the table, P, N and Z stand for positive, negative and zero DTR, respectively. Overall, eight firms are rated as P, sixteen firms are rated as N and twenty-six firms are rated as Z. An occurrence of N indicates a potential investment opportunity, which is further classified based on the rule in Section 4.5. Specifically, the symbol (E) stands for an effective investment opportunity. The symbol (L), implying limited (but still

Investment

E

E

E

E

E E

E E E E

E E

E

E E E

effective) investment, does not occur in the data set used in this study. A blank in the last column corresponds to a positive or zero DTR. The black implies that the investment is not effective in enhancing the unified efficiency of each petroleum company. There are twelve independent companies and four integrated companies that have effective investment opportunities. The percentage (57.14% = 4/7) of effective investment opportunities among the integrated companies is much higher than that (27.91% = 12/43) of the independent companies. In other words, the integrated companies present better green investment opportunities than the independent companies. The empirical result implies that the integrated companies provide attractive investment opportunities for operational enhancement and emission mitigation, so attaining their corporate sustainability. This is because they are much larger emitters than the independent

372

T. Sueyoshi, D. Wang / Energy Economics 46 (2014) 360–374

Undesirable Output (b)

F

G

Efficiency Frontier under Constant DTR Desirable Congestion (Technology D Innovation)

C B

E Efficiency Frontier under Variable DTR

A

0

Desirable Output (g)

Fig. 6. Scale efficiency measurement. (a) The degree of SE(DC) is measured by the distance between the two DMUs {F and G} divided by the one between the two DMUs (F and C}. The degree of SE(DC) measures how each DMU effectively manages its operational size with a possible occurrence of technology innovation to enhance the level of corporate sustainability. (b) This study clearly understands that we need to measure the degree of RTS, DTS and DTR in order to conclude whether a DMU has a scale merit regarding corporate sustainability. The scale efficiency discussed in this study is for our methodological convenience, but being not an exact methodology.

companies. Moreover, they need to pay attention to environmentconscious consumers in their supply chain systems.

5.4. Scale efficiency As mentioned previously, the petroleum industry can be considered as a very large process industry from exploration of oil and gas to retail services for end users. Chakravarthy (1997) and Vachon and Klassen (2006) have discussed that the petroleum firms may have increasing difficulties in planning and predicting their businesses, when their corporate sizes grow. Thus, it is necessary for us to examine how they manage the unified performance from the perspective of their business sizes. In other words, it is expected from the previous studies that the integrated companies are much larger than the independent companies so that the operation of their businesses is more difficult than that of the independent companies. To investigate the research concern, this study measures a degree of UENM(DC) under two cases: variable and constant DTR. The degree of UENM(DC) under variable DTR is measured by Model (9). The efficiency, or UENM(DC)*, under constant DTR is measured by Model (9) after n

dropping ∑ λ j ¼ 1 in the formulation. An important feature of the two j¼1

efficiency measures is that the former incorporates a possible occurrence of desirable congestion, or technology innovation. The difference can be found in the type of DTR in their formulations. After obtaining the two measures on each DMU, this study measures SE(DC) = UENM(DC)* /UENM(DC) as the level of scale efficiency. The scale efficiency measures how each DMU effectively operates by considering the operational size. Thus, the magnitude of SE(DC) indicates the level of size utilization on the DMU operation. Assuming that two production functions for desirable and undesirable outputs have similar behaviors so that they are expressed by part of a convex function on a change of inputs, Fig. 6 visually describes a difference between efficiency frontiers under constant and variable DTR. A contour line, connecting A–B–C–D–E, stands for an efficiency frontier under variable DTR, while the straight line, passing on DMU{G}, from the origin indicates the efficiency frontier under constant DTR. Since an observed DMU {F} is inefficient, it needs to be projected on to the two efficient frontiers. Such projections can be found on two efficient

Table 4 Scale efficiency measures. Company name

UENM(DC)

UENM(DC)*

Scale efficiency (DC)

Independent companies Anadarko Petroleum Corporation Antero Resources LLC Apache Corporation Berry Petroleum Company BHP Billiton Group Bill Barrett Corporation Cabot Oil & Gas Corporation Chesapeake Energy Corporation Cimarex Energy Co. Concho Resources Conoco Phillips CONSOL Energy Inc. Continental Resources, Inc. Denbury Resources Inc. Devon Energy Corporation EnCana Corporation Energen Corporation EOG Resources, Inc. EP Energy LLC EQT Corporation EXCO Resources, Inc. Forest Oil Corporation Laredo Petroleum Holdings, Inc. Linn Energy, LLC National Fuel Gas Company Newfield Exploration Company Noble Energy, Inc. Occidental Petroleum Corporation PDC Energy, Inc. Pioneer Natural Resources Company Plains E & P Company QEP Resources, Inc. Quicksilver Resources, Inc. Range Resources Corporation Rosetta Resources Inc. SandRidge Energy, Inc. SM Energy Company Southwestern Energy Company Swift Energy Inc. Talisman Energy Inc. Ultra Petroleum Corporation Whiting Petroleum Corporation WPX Energy, Inc. Avg. Max. Min. S.D.

1.0000 1.0000 1.0000 1.0000 1.0000 0.9904 1.0000 1.0000 0.9773 0.9218 1.0000 0.9766 0.9495 0.9758 1.0000 1.0000 0.9503 0.8899 1.0000 0.9781 0.9999 0.9281 1.0000 0.9444 0.9015 0.9350 1.0000 0.8192 1.0000 0.9398 0.9142 0.9462 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 0.9269 1.0000 1.0000 0.9488 0.9724 1.0000 0.8192 0.0407

1.0000 1.0000 1.0000 1.0000 1.0000 0.9875 1.0000 0.8884 0.9693 0.9094 1.0000 0.9601 0.9395 0.9576 1.0000 1.0000 0.9410 0.8817 1.0000 0.9665 0.9693 0.9279 0.9666 0.9404 0.8868 0.9276 1.0000 0.8102 1.0000 0.9386 0.8965 0.9378 0.9847 1.0000 1.0000 0.9168 0.9899 0.9462 0.9387 0.9154 1.0000 1.0000 0.9457 0.9591 1.0000 0.8102 0.0444

1.0000 1.0000 1.0000 1.0000 1.0000 0.9970 1.0000 0.8884 0.9917 0.9865 1.0000 0.9830 0.9894 0.9814 1.0000 1.0000 0.9902 0.9907 1.0000 0.9881 0.9694 0.9999 0.9666 0.9958 0.9836 0.9921 1.0000 0.9890 1.0000 0.9988 0.9806 0.9912 0.9847 1.0000 1.0000 0.9168 0.9899 0.9462 0.9387 0.9876 1.0000 1.0000 0.9967 0.9864 1.0000 0.8884 0.0233

Integrated companies BP PLC Chevron Corporation Exxon Mobil Corporation Hess Corporation Marathon Oil Corporation Murphy Oil Corporation Royal Dutch Shell PLC Avg. Max. Min. S.D. P-value (Welch's t-test)

1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 0.0000 0.0001

1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 0.0000 0.0000

1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 0.0000 0.0004

DMUs {G and C} under constant and variable RTS, respectively. The degree of SE(DC) is specified by the distance between the two DMUs {F} and {G} divided by the one between the two DMUs {F} and {C}. Thus, the degree of SE(DC) measures how each DMU effectively manages its operational size with a possible occurrence of technology innovation to enhance corporate sustainability. Table 4 summarizes UENM(DC), UENM(DC)* and SE(DC) of all independent and integrated petroleum companies. The SE measure of

T. Sueyoshi, D. Wang / Energy Economics 46 (2014) 360–374

independent firms is 0.9864 and that of integrated companies is 1.0000 on average. The mean test confirms at the 1% significance that the integrated companies outperform the independent companies in term of their size utilizations. The result implies that the integrated companies may have more managerial difficulty, along with an increase in their operational sizes, than the independent companies. However, by carefully controlling their large supply chain systems from exploration to retail services, the integrated companies can effectively increase their unified (operational and environmental) performance, so corporate sustainability.

6. Conclusion and future extensions To measure the level of corporate sustainability regarding companies, this study discussed a use of DEA for environmental assessment. The proposed environmental assessment provided corporate leaders and policy makers with not only the level of corporate sustainability but also information regarding how to invest for technology innovation for abatement of undesirable outputs. As an application, this study used the proposed DEA approach to examine the corporate sustainability of petroleum firms in the United States. The integrated companies have large supply chain systems for spanning both upstream and downstream functions. Meanwhile, the independent companies have only the upstream function in their business operations. This empirical study found that the integrated companies outperformed the independent companies, because a large supply chain incorporated into the former group provided them with a scale merit in their operations and gave a business opportunity to directly contact consumers (Wuyts et al., 2004). Consequently, it can be considered that the large supply chain can enhance corporate sustainability in the U.S. petroleum industry. Moreover, the vertical integration in upstream and downstream may be a promising business trend toward better environmental performance in the industry. A problem of this empirical study was that it did not examine whether governmental regulation influenced the performance of petroleum firms. It is easily imagined that such regulation policy influence exists on the operation of all integrated and independent firms. Unfortunately, this study did not directly examine the policy issue because the research concern of this study was whether a supply chain influenced the performance of petroleum firms. The business concern was the gist of recent studies on operations management in U.S. business schools. An inquiry, on whether the regulation influences the petroleum industry, will be explored as a future research task. The proposed DEA approach has four drawbacks as a methodology for environmental assessment. First, the proposed DEA approach assumes that all unified efficiency measures are uniquely determined on optimality. If the assumption is dropped, the proposed nonradial models need to incorporate strong complementary slackness conditions into their formulations to obtain a unique optimal solution. Second, this study can incorporate a time horizon into the computational framework of DEA environmental assessment (Sueyoshi and Goto, 2013a,b,c,d; Sueyoshi et al., 2013a,b). Consequently, the extension makes it possible to handle a time series data set on environmental assessment. Third, this study does not discuss analytical features on RTS, DTS and DTR. The mathematical issue will be discussed in another article as an extension of this study. Finally, the empirical study is related to only the U.S. petroleum industry. It is important for this study to expand the research horizon from the United States to other OPEC nations in order to confirm the validity of empirical evidences obtained in this study. See Sueyoshi and Goto (2014d) that discussed the radial approach to examine corporate governance and technology investment. In conclusion, it is hoped that this study makes a contribution in DEA environmental assessment on the petroleum industry. We look forward to seeing future extensions as discussed in this study.

373

Appendix A. Supplementary data Supplementary data to this article can be found online at http://dx. doi.org/10.1016/j.eneco.2014.09.022.

References American Petroleum Institute, 2013. Understanding the oil and natural gas supply chains. Accessed June 23, 2014 http://www.api.org/policy-and-issues/policy-items/safety/ understanding-the-oil-natural-gas-supply-chains. Barney, J.B., 1991. Firm resources and sustained competitive advantage. J. Manag. 17 (1), 99–120. Carter, C.R., Kale, R., Grimm, C.M., 2000. Environmental purchasing and firm performance: an empirical investigation. Transp. Res. E 36, 219–228. Chakravarthy, B., 1997. A new strategy framework for coping with turbulence. Sloan Manag. Rev. 38, 69–82. Depart of Energy, 2013. U.S. energy sector vulnerabilities to climate change and extreme weatherAccessed June 23, 2014 http://energy.gov/sites/prod/files/2013/07/f2/ 20130710-Energy-Sector-Vulnerabilities-Report.pdf. Environmental Protection Agency, 2012a. http://www.epa.gov/climatechange/ EPAactivities/regulatory-initiatives.html. Environmental Protection Agency, 2012b. http://www.bloomberg.com/news/2012-0417/drillers-say-costs-manageable-from-pending-gas-emissions-rule.html. Environmental Protection Agency, 2012c. http://www.epa.gov/airquality/oilandgas/ actions.html. Environmental Protection Agency, 2013a. Greenhouse Gas Reporting Program: petroleum and natural gas systems 2012 data summaryAccessed June 23, 2014 http://www.epa. gov/ghgreporting/documents/pdf/2013/documents/SubpartW-2012-Data-Summary. pdf. Environmental Protection Agency, 2013b. Greenhouse Gas Reporting Program. http:// www.epa.gov/ghgreporting. Färe, R., Grosskopf, S., Pasurka, C., 1986. Effects on relative efficiency in electric power generation due to environmental controls. Resour. Energy 8, 167–184. Färe, R., Grosskopf, S., Lovel, C.A.K., Pasurka, C., 1989. Multilateral productivity comparison when some outputs are undesirable: a nonparametric approach. Rev. Econ. Stat. 71, 90–98. Freeman, R.E., 1984. Strategic Management: A Stakeholder Approach. Pittman, Marshfield. Glover, F., Sueyoshi, T., 2009. Contributions of Professor William W. Cooper in operations research and management science. Eur. J. Oper. Res. 197, 1–16. Gonzáles-Benito, J., Gonzáles-Benito, Ó., 2006. A review of determinant factors of environmental proactivity. Bus. Strateg. Environ. 15, 87–102. Goto, M., Ootsuka, A., Sueyoshi, T., 2014. DEA (data envelopment analysis) assessment of operational and environmental efficiencies on Japanese regional industries. Energy 66, 535–549. Hall, J., 2001. Environmental supply-chain innovations. Greener Manag. Int. 35, 105–119. Harris, P.G., 2006. Environmental perspectives and behavior in China: synopsis and bibliography. Environ. Behav. 38, 5–21. Howarth, R., Ingraffea, A., Engelder, T., 2011. Natural gas: should fracking stop. Nature 477, 271–275. Ijiri, Y., Sueyoshi, T., 2010. Accounting essays by Professor William W. Cooper: revisiting in commemoration of his 95th birthday. ABACUS, 46, 464–505. Intergovernmental Panel on Climate Change, 2014. http://ipcc.ch/index.htm. Jacobs, B.W., Singhal, V.R., Subramanian, R., 2010. An empirical investigation of environmental performance and the market value of the firm. J. Oper. Manag. 28, 430–441. Jennings, P.D., Zandbergen, P.A., 1995. Ecologically sustainable organizations: an institutional approach. Acad. Manag. Rev. 20, 1015–1052. Lai, K.H., Wong, C.W.Y., Cheng, T.C.E., 2006. Institutional isomorphism and the adoption of information technology for supply chain management. Comput. Ind. 57, 93–98. Murphy, J., Gouldson, A., 2000. Environmental policy and industrial innovation: integrating environment and economy through ecological modernization. Geoforum 31, 33–44. National Association of Convenience Stores, 2013. NACS retail fuels report. http://www. nacsonline.com/YourBusiness/FuelsReports/GasPrices_2012/Pages/WhoSellsGas. aspx. NYSE Energy Index, 2013. http://www.nyse.com/about/listed/nyeid_components.shtml. Paloviita, A., Luoma-aho, V., 2010. Recognizing definitive stakeholders in corporate environmental management. Manag. Res. Rev. 33, 306–316. Porter, M., van der Linde, C., 1995. Toward a new conception of the environmentcompetitiveness relationship. J. Econ. Perspect. 9, 97–118. Revell, A., 2007. The ecological modernisation of SMEs in the UK's construction industry. Geoforum 38, 114–126. Rivera, J., 2004. Institutional pressures and voluntary environmental behavior in developing countries: evidence from the Costa Rican hotel industry. Soc. Nat. Resour. 17, 779–797. Ross, A., Droge, C., 2002. An integrated benchmarking approach to distribution center performance using DEA modeling. J. Oper. Manag. 20, 19–32. Ross, A., Droge, C., 2004. An analysis of operations efficiency in large-scale distribution systems. J. Oper. Manag. 21, 673–688. Russo, M.V., Fouts, P.A., 1997. A Resource-based perspective on corporate environmental performance and profitability. Acad. Manag. J. 40, 534–559.

374

T. Sueyoshi, D. Wang / Energy Economics 46 (2014) 360–374

Sarkis, J., Gonzalez-Torre, P., Adenso-Diaz, B., 2010. Stakeholder pressure and the adoption of environmental practices: the mediating effect of training. J. Oper. Manag. 28, 163–176. Simpson, D., 2010. Use of supply relationships to recycle secondary materials. Int. J. Prod. Res. 48, 227–249. Singhal, K., Singhal, J., 2012. Opportunities for developing the science of operations and supply chain management. J. Oper. Manag. 30, 245–252. Sueyoshi, T., Goto, M., 2009. Can environmental investment and expenditure enhance financial performance of US electric utility firms under the clean air act amendment of 1990? Energy Policy 37, 4819–4826. Sueyoshi, T., Goto, M., 2010a. Measurement of a linkage among environmental, operational and financial performance in Japanese manufacturing firms: a use of data envelopment analysis with strong complementary slackness condition. Eur. J. Oper. Res. 207, 1742–1753. Sueyoshi, T., Goto, M., 2010b. Should the US clean air act include CO2 emission control?: examination by data envelopment analysis. Energy Policy 38, 5902–5911. Sueyoshi, T., Goto, M., 2011a. Operational synergy in the electric utility industry under the influence of US deregulation policy: a linkage to financial performance and corporate value. Energy Policy 39, 699–713. Sueyoshi, T., Goto, M., 2011b. DEA approach for unified efficiency measurement: assessment of Japanese fossil fuel power generation. Energy Econ. 33, 195208. Sueyoshi, T., Goto, M., 2011c. A combined use of DEA (data envelopment analysis) with strong complementary slackness condition and DEA-DA (discriminant analysis). Appl. Math. Lett. 24, 1051–1056. Sueyoshi, T., Goto, M., 2012a. Data envelopment analysis for environmental assessment: comparison between public and private ownership in petroleum industry. Eur. J. Oper. Res. 216, 668–678. Sueyoshi, T., Goto, M., 2012b. DEA radial measurement for environmental assessment and planning: desirable procedures to evaluate fossil fuel power plants. Energy Policy 41, 422–432. Sueyoshi, T., Goto, M., 2012c. Efficiency-based rank assessment for electric power industry: a combined use of data envelopment analysis (DEA) and DEA-discriminant analysis (DA). Energy Econ. 34, 634–644. Sueyoshi, T., Goto, M., 2012d. Returns to scale and damages to scale under natural and managerial disposability: strategy, efficiency and competitiveness of petroleum firms. Energy Econ. 34, 645–662. Sueyoshi, T., Goto, M., 2012e. Environmental assessment by DEA radial measurement: U.S. coal-fired power plants in ISO (independent system operator) and RTO (regional transmission organization). Energy Econ. 34, 663–676. Sueyoshi, T., Goto, M., 2012f. Weak and strong disposability vs. natural and managerial disposability in DEA environmental assessment: comparison between Japanese electric power industry and manufacturing industries. Energy Econ. 34, 686–699. Sueyoshi, T., Goto, M., 2012g. DEA radial and non-radial models for unified efficiency under natural and managerial disposability: theoretical extension by strong complementary slackness conditions. Energy Econ. 34, 700–713. Sueyoshi, T., Goto, M., 2012h. Returns to scale, damages of scale, marginal rate of transformation and rate of substitution in DEA environmental assessment. Energy Econ. 34, 905–917. Sueyoshi, T., Goto, M., 2012i. Returns to scale and damages to scale with strong complementary slackness conditions in DEA environmental assessment: Japanese corporate effort on environmental protection. Energy Econ. 34, 1422–1434. Sueyoshi, T., Goto, M., 2012j. DEA environmental assessment of coal fired power plants: methodological comparison between radial and non-radial models. Energy Econ. 34, 1854–1863. Sueyoshi, T., Goto, M., 2012k. Returns to scale and damages to scale on U.S. fossil fuel power plants: radial and non-radial approaches for DEA environment assessment. Energy Econ. 34, 2240–2259. Sueyoshi, T., Goto, M., 2013a. Returns to scale vs. damages to scale in data envelopment analysis: an impact of U.S. clean air act on coal-fired power plants. OMEGA 41, 164–175. Sueyoshi, T., Goto, M., 2013b. A use of DEA-DA to measure importance of R&D expenditure in Japanese information technology industry. Decis. Support. Syst. 54, 941–952.

Sueyoshi, T., Goto, M., 2013c. A comparative study among fossil fuel power plants in PJM and California ISO by DEA environmental assessment. Energy Econ. 40, 130–145. Sueyoshi, T., Goto, M., 2013d. DEA environmental assessment in a time horizon: Malmquist index on fuel mix, electricity and CO2 of industrial nations. Energy Econ. 40, 370–382. Sueyoshi, T., Goto, M., 2014a. DEA radial measurement for environmental assessment: a comparative study between Japanese chemical and pharmaceutical firms. Appl. Energy 115, 502–513. Sueyoshi, T., Goto, M., 2014b. Investment strategy for sustainable society by development of regional economies and prevention of industrial pollutions in Japanese manufacturing Sectors. Energy Econ. 42, 299–312. Sueyoshi, T., Goto, M., 2014c. Photovoltaic power stations in Germany and the United States: a comparative study by data envelopment analysis. Energy Econ. 42, 271–288. Sueyoshi, T., Goto, M., 2014d. Environmental assessment for corporate sustainability by resource utilization and technology innovation: DEA radial measurement on Japanese industrial sectors. Energy Econ. 46, 295–307. Sueyoshi, T., Wang, D., 2014. Radial and non-radial approaches for environmental assessment by data envelopment analysis: corporate sustainability and effective investment for technology innovation. Energy Econ. 45, 537–551. Sueyoshi, T., Goto, M., Shang, J., 2009. Core business concentration vs. corporate diversification in the US electric utility industry: synergy and deregulation effects. Energy Policy 37, 4583–4594. Sueyoshi, T., Goto, M., Ueno, T., 2010. Performance analysis of U.S. coal-fired power plants by measuring three DEA efficiencies. Energy Policy 38, 1675–1688. Sueyoshi, T., Goto, M., Snell, M.A., 2013a. DEA environmental assessment: measurement of damages to scale with unified efficiency under managerial disposability or environmental efficiency. Appl. Math. Model. 37, 7300–7314. Sueyoshi, T., Goto, M., Sugiyama, M., 2013b. DEA window analysis for environmental assessment in a dynamic time shift: performance assessment of U.S. coal-fired power plants. Energy Econ. 40, 845–857. Tate, W.L., Ellram, L.M., Kirchoff, J.F., 2010. Corporate social responsibility reports: a thematic analysis related to supply chain management. J. Supply Chain Manag. 46, 19–44. Theyel, G., 2001. Customer and supplier relations for environmental performance. Greener Manag. Int. 35, 61–69. Vachon, S., 2007. Green supply chain practices and the selection of environmental technologies. Int. J. Prod. Res. 45, 4357–4379. Vachon, S., Klassen, R.D., 2006. Extending green practices across the supply chain: the impact of upstream and downstream integration. Int. J. Oper. Prod. Manag. 26 (7), 795–821. Walton, S.V., Handfield, R.B., Melnyk, S.A., 1998. The green supply chain: integrating suppliers into environmental management processes. J. Supply Chain Manag. 34, 2–11. Wang, D., Li, S., Sueyoshi, T., 2014. DEA environmental assessment on U.S. industrial sectors: investment for improvement in operational and environmental performance to attain corporate sustainability. Energy Econ. 45, 254–267. Watanabe, M., Tanaka, K., 2007. Efficiency analysis of Chinese industry: a directional distance approach. Energy Policy 35, 6323–6331. Wong, C.W.Y., Lai, K.H., Cheng, T.C.E., 2009. Complementarities and alignment of information systems management and supply chain management. Int. J. Shipp. Transp. Logist. 1, 156–171. Wuyts, S., Stremersch, S., Van Den Bulte, C., Franses, P.H., 2004. Vertical marketing systems for complex products: a triadic perspective. J. Mark. Res. 41, 479–487. Yang, H., Pollitt, M., 2010. The necessity of distinguishing weak and strong disposability among undesirable outputs in DEA: environmental performance of Chinese coalfired power plants. Energy Policy 38, 4440–4444. Yang, J., Wang, J., Wong, C.W.Y., Lai, K.H., 2008. Relational stability and alliance performance in supply chain. Omega 36, 600–608. Young, Ernst, 2013. U.S. oil and gas reserves studyAccessed June 23, 2014 http://www.ey. com/Publication/vwLUAssets/US_oil_and_gas_reserves_study_2013/$FILE/US_oil_ and_gas_reserves_study_2013_DW0267.pdf.