Use of interval data envelopment analysis, goal programming and dynamic eco-efficiency assessment for sustainable supplier management

Highlights

• Comprehensive method for the management of supplier sustainability performance.

• Data Envelopment Analyses (DEA) used to estimate supplier eco-efficiency.

• Dynamic monitoring of supplier eco-efficiency through a Malmquist approach.

• Bicriteria optimization model integrating order quantity discounts.

• Methodology demonstrated with real data from an auto parts manufacturer.

Abstract

Two big challenges for sustainable supplier management consist of (1) properly integrating the multiple factors necessary to measure the sustainability performance of suppliers and (2) modeling the dynamic nature of those factors. In this study, we demonstrate the aggregation of environmental and economic supplier performance criteria into a single (eco-efficiency) index using Data Envelopment Analysis (DEA) for interval data. Suppliers are pre-qualified according to this index and are allocated orders using multiple criteria models for single and multiple sourcing. In multiple sourcing, preemptive, non-preemptive and fuzzy goal programming models are used with purchase price and eco-efficiency as conflicting objectives. Finally, eco-efficiency change is assessed using an approach based on the Malmquist productivity index, which monitors suppliers over time. Thus, our methodology can be used for supplier evaluation, selection and monitoring. Data obtained from an auto parts manufacturer is used to illustrate the proposed method.

Introduction

Beyond a purely responsive attitude, organizations are finding that ensuring that their products meet or exceed environmental and social expectations can generate significant new opportunities in terms of operational efficiencies and as new markets are created for lower carbon, more sustainable, goods and services. Some estimations suggest that a typical company, using best-practice sustainability approaches, could improve its profit by at least 51–81% within three to five years, while avoiding a potential 16–36% erosion of profits, if it did nothing (Willard, 2012). Other studies (Cruz and Wakolbinger, 2008, Cruz, 2009) have shown that performing well in the social and environmental domains can positively impact transaction costs, risk and environmental impact. The Index, the tool introduced by Wal-Mart in 2009 to measure the environmental performance of its then 60,000 global supplier network, is an example of how most big companies are now requesting their suppliers to commit to high environmental and social standards (Humes, 2011). By virtue of initiatives like this, detailed management and reporting will require a better approximation to actual supplier sustainability performance.

Part of the complexity of introducing sustainability expectations, when allocating orders to suppliers, lies in the proper integration of the many factors which impact the economic, social and environmental performance of companies. Sustainability criteria possess certain properties which make particularly challenging those problems involving this type of criteria (Munda, 2005):

• Incommensurability refers to the presence of multiple criteria associated with different units of measure, which create value conflict when deciding what common term of comparison should be used to represent a real-world system.

• Mixed criterion scores. Although economic performance is usually measured with quantitative criteria, social and environmental well-being require qualitative criteria in most cases.

• Non-compensability. In composite indicators there exists a theoretical inconsistency between the way weights are actually used and what their real theoretical meaning is. For instance, some sorts of natural capital cannot be readily substituted (compensated) by man-made capital. An example is trees. We may say that one tree is used to make a set of chairs, however, this does not imply that we can transform that set of chairs into the original tree (i.e. waste needs to be accounted for).

• Uncertainty. We rarely find information that is precise, certain, exhaustive and unequivocal when dealing with environmental problems. In consequence, we face the uncertainty of the stochastic and/or fuzzy nature of data.

• Risk. There is always the potential that performance under certain criteria will deviate from what is considered a normal behavior. In these situations, the undesirable outcome is usually unavoidable and must be assumed by the decision maker.

One more challenge of implementing the sustainability concept lies in the fact that it is also difficult to specify the long-term damage to an ecosystem and allocate it to a particular company. Because of this, companies seek measures that can easily be translated into practice, while helping address the multicriteria and dynamic dimensions of sustainability. The current article presents an approach in that direction.

Eco-efficiency has been traditionally defined from a managerial perspective as the ratio of value added to environmental damage. Thus, eco-efficiency is measured in terms of the value generated to the firm. Value may come from fair price quotations, high-quality materials, reliable delivery and a company’s good reputation. Environmental damage may originate from different environmental concerns: hazardous waste, landfill waste, air pollution, etc. (Lahouel, 2016, Michelsen et al., 2006).

$$\mathit{Eco}-efficiency=\frac{\mathit{product}orservicevalue}{\mathit{environmental}damage}$$

According to this definition, the type of eco-efficiency measures proposed throughout the business literature include: simple ratio indicators related to specific consumption (e.g. kg of water consumed in 2016/kg of product produced in 2016), proportional values (e.g. drinking water input in kg in 2016/water input in kg in 2015) or index figures over the course of time (water input in 2016/water input in the base year 2010). In addition, indicators using Life Cycle Analysis (LCA) and Environmental Impact Assessment (EIA) have been widely used. Companies may also define their own environmental performance indicators. These are seldom disclosed in the literature and are more likely to be specifically oriented towards the company’s objectives (Tyteca, 1996), which prevents or, at least, makes more difficult comparison among firms. For a detailed review of the various types of environmental performance indicators and recommendations for their creation, refer to (Azzone and Manzini, 1994, Govindan et al., 2014, Tyteca, 1996).

Based on the previous considerations, a well-balanced evaluation of a firm’s eco-efficiency should not limit the number of sources considered, neither for the value criteria nor for the environmental damage criteria. This implies that aggregation mechanisms should be used to have a single or few plant-level indicators that allow interfirm comparison in an objective way. Data Envelopment Analysis (DEA) is a mathematical programming methodology proposed by Charnes, Cooper, and Rhodes (CCR) (Charnes, Cooper, & Rhodes, 1978) as benchmarking tool and classification system that allows the aggregation of multiple criteria into a single efficiency measure. It categorizes suppliers as “efficient” or “inefficient” so that, for a given set of suppliers, DEA forms an efficient frontier by joining the most efficient suppliers. The efficiency value or productivity index of each supplier is measured using the relative distance projection toward that frontier. Suppliers are evaluated on benefit criteria (output) and cost criteria (input), which can involve categorical and numerical values having different measurement units. A supplier's efficiency is given by the ratio of the weighted sum of its outputs to the weighted sum of its inputs.

Although DEA is a very practical benchmarking tool, from the perspective of supplier selection its use has been limited to provide assessments in static or single period settings. However, since firms change over time, it becomes fundamental to account for changes in their efficiencies. Particularly, it becomes important to capture how changes in eco-efficiency over time are caused by special initiatives on part of individual firms or by general technological progress affecting an entire industry or group of firms. Accounting for this type of changes would decrease the exposure to risk of a company by providing feedback about deviations from what is considered normal. In this regard, a useful addition to the traditional research on DEA and supplier selection is the introduction of the Malmquist Index.

The Malmquist Index (MI) considers the evolution of relative efficiency values over time and is the most popular approach to dynamic productivity evaluations. It was introduced theoretically by Caves, Christensen, and Diewert (1982) and empirically developed by Färe, Grosskopf, Lindgren, and Roos (1994). The latter author proposed the decomposition of total factor productivity to provide relevant insights with respect to the supplier’s progress achieved across time as result of the supplier’s efforts versus the progress achieved as a result of technological changes affecting an entire industry.

In this paper, we present a three-phase methodology (see Fig. 1) where phase one uses DEA to pre-qualify a large set of initial suppliers by the estimation of their relative eco-efficiency scores based on interval data. In phase two, for a single sourcing strategy, a multiple criteria ranking method is applied and, for a multiple sourcing strategy, a Bi-criteria Optimization Model (BOM) is used with eco-efficiency scores and total purchase price as conflicting objectives. The BOM incorporates all unit quantity discounts based on the total purchase value and is solved using Preemptive, Non-Preemptive and Fuzzy Goal Programming. Finally, in the third phase, we monitor supplier eco-efficiency changes for adjoining time periods by estimating the Environmental Productivity Index (EPI), which we decompose to distinguish between the eco-efficiency changes due to innovations in technology and those due to the actual behavioral change on part of the supplier. The methodology is applied to the evaluation, selection and monitoring of suppliers for an autoparts manufacturer in Mexico.

The paper is organized as follows: Section 2, presents the literature review; Section 3 introduces the proposed model; Section 4 presents the case study using the outlined approach and Section 5 presents concluding remarks and future research.