Process network modularity, commonality, and greenhouse gas emissions

• Author: Thomas J. Kull,Zhaohui Wu,Kevin J. Dooley,Surya D. Pathak,Jon Johnson,Elliot Rabinovich
• Published date: 01 March 2019
• DOI: http://doi.org/10.1002/joom.1007
• Date: 01 March 2019

ORIGINAL ARTICLE

Process network modularity, commonality, and greenhouse gas

emissions

Kevin J. Dooley

1

| Surya D. Pathak

2

| Thomas J. Kull

1

| Zhaohui Wu

3

| Jon Johnson

4

|

Elliot Rabinovich

1

1

Department of Supply Chain Management,

WP Carey School of Business, Arizona

State University, Tempe, Arizona

2

School of Business, University of

Washington, Bothell, Washington

3

College of Business, Oregon State

University, Corvallis, Oregon

4

Department of Management, Sam

M. Walton College of Business, University

of Arkansas, Fayetteville, Arkansas

Correspondence

Kevin J. Dooley, Department of Supply

Chain Management, WP Carey School of

Business, Arizona State University, Tempe,

AZ 85287-4706.

Email: kevin.dooley@asu.edu

Handling Editors: Anand Nair and Felix

Reed-Tsochas

Funding information

U.S. National Science Foundation, Grant/

Award Number: 1024752

Abstract

A process network is a complex system of linked unit processes that constitute the

life cycle of a product. In this article, we consider how the structural and functional

characteristics of a product's process network impact the network's collective green-

house gas (GHG) emissions. At a unit process level, GHG emissions are primarily

related to process efficiency. We hypothesize that a process network's GHG emis-

sions will be less when the process network has a modular structure and when its

constituent unit processes are more functionally similar. A modular process network

architecture promotes autonomous innovation and improvements in knowledge man-

agement and problem-solving capabilities, leading to more efficient processes. Func-

tional commonality in a process network enables economies of scale and knowledge

spillover and also leads to process efficiencies, thus reducing GHG emissions. We

test these two hypotheses using a sample of 4,189 process networks extracted from

an environmental lifecycle inventory database. Empirical results support our hypoth-

eses, and we discuss the implications of our findings for product development and

supply network design.

KEYWORDS

carbon footprint, commonality, environmental performance, greenhouse gas, life cycle, modular,

nearly decomposable, network, process, sustainability

1|INTRODUCTION

The observed and predicted impacts of climate change have

attracted unprecedented attention to reduce greenhouse gas

(GHG) emissions. In 2015, 155 countries agreed to the targets

of GHG emissions as part of the Paris Climate Conference.

Most of the participating countries havedeveloped regulations

and incentives to reduce GHG emissions and moved to

renewable energy sources that are less carbon intensive

(Williams et al., 2012). Companies and industry sectors have

in turn made GHG reduction commitments and renewable

energy commitments, and invested in processes and systems

that are less energy intensive or capture GHG emissions

(Hoffman, Corbett,Joglekar, & Wells, 2014; Obama, 2017).

Business organizations have concentrated on reducing

GHG emissions within their own operations as part of

their efforts to curb their direct operating costs (Corbett &

Klassen, 2006; Russo & Fouts, 1997). To reach global,

national, and corporate targets, however, GHG emissions

across the whole life cycle of a product or service need to be

addressed (Dooley, 2014). According to CDP (2017), GHG

emissions related to a company's own operations account, on

average, for less than 20% of a product or service's GHG

emissions; the majority of emissions occur in the company's

supply chain. Pressure from customers, competitors, regula-

tors, and civil society around life cycle or supply chain GHG

emissions is becoming common (Kleindorfer, Singhal, &

Van Wassenhove, 2005; Meinrenken, Sauerhaft, Garvan, &

DOI: 10.1002/joom.1007

J Oper Manag. 2019;65:93–113. wileyonlinelibrary.com/journal/joom © 2019 Association for Supply Chain Management, Inc. 93

Lackner, 2014; Piore, 2012). Because of both supply and

market reputation risks, manufacturers are held accountable

to not only their own actions but also the actions and

impacts of their upstream suppliers (Awaysheh & Klassen,

2010; Carter, 2005; Dooley & Johnson, 2015).

Walmart's recent commitment to remove 1 gigaton of

GHG emissions collectively from their supply chain is an

example of how companies are actively considering the risk

and opportunity associated with GHG emissions in their sup-

ply chain (Ragland, 2017). Reducing supply chain GHG

emissions typically leads to cost reductions because of effi-

ciency improvements, which aligns with a corporate strategy

of managing costs. Additionally, it reduces the cost risks asso-

ciated with any future carbon tax or energy price volatility.

Likewise, the positive momentum of the product environmen-

tal footprinting efforts in Europe (Finkbeiner, 2014) suggests

that life cycle accounting for GHG emissions will be increas-

ingly used for purchasing decisions, despite the moderate suc-

cess so far of product carbon footprinting efforts.

There is a paucity of research that links characteristics of

supply chains to GHG emissions. In this article, models of

the activities that constitute the product life cycle as a pro-

cess network are presented (Ruddell & Kumar, 2009). A

process network is a directed network of unit processes that

yields a final product, where each node represents a transfor-

mation process and connections represent physical flow of

process inputs and outputs. There is little understanding of

what drives the level of a network's GHG emissions, other

than having unit processes with low or high emissions.

Process networks are complex in that they involve many

parts acting in a nonsimple way (Simon, 1962) and adapt to

the fact that the organizations (i.e., agents) who design or

operate unit processes within the process network make

decisions over time that impact unit process performance

(Pathak, Day, Nair, Sawaya, & Kristal, 2007). Most real-

world process networks involve hundreds or thousands of

unit processes, thus the network is not under the control of

any single organization, but rather its structure and function

are emergent from the actions and interactions of many orga-

nizational decision makers. If process networks are similar

to other complex systems, then it is possible that structural

and functional characteristics of the network itself, beyond

the node-level effects, may also have an impact on the net-

work's performance (Simon, 1962). Thus, our research ques-

tion is: How do the structural and functional characteristics

of a process network impact its collective GHG emissions?

To answer this question, we consider theory and empirical

studies related to the benefits of modularity and commonality

in product, process, organization, and supply networks

(Barabasi, 2007; Bellamy, Ghosh, & Hora, 2014; Cheng,

2011; Danese & Filippini, 2013; Fixson, 2007; Jacobs,

Vickery, & Droge, 2007;Lau, Yam, & Tang, 2010; Randall &

Ulrich, 2001; Thatte, 2013; Ulrich, 1995; Worren, Moore, &

Cardona, 2002). As a process's GHG emissions stem mostly

from the energy and fuel it uses in production, GHG emis-

sions are reduced as processes become more productive and

efficient. Our theory suggests that a modular process network

architecture promotes more autonomous process innovation

and improvements in knowledge management and problem-

solving capabilities, leading to more efficientprocesses. Addi-

tionally, functional commonality of processes in a process

network enables economies of scale and knowledge sharing

and spillover and also leads to process efficiencies, thus

reducing network GHG emissions.

We test these two hypotheses using a sample of 4,189 pro-

cess networks extracted from an environmental life cycle

inventory database. We use recent graph theory methods to

operationalizemodularity (Leicht & Newman, 2008, Dugué&

Perez, 2015, Fortunato and Hric 2016) and process common-

ality of using an entropy-based measure. We test the hypothe-

ses using a hierarchal linear model involving the two main

explanatory variables and appropriate firm-level, industry-

level, and network-level controls. Empirical results support

our hypotheses, namely, that more modular process networks

and networks with morecommonality among their constituent

unit processes have lowerGHG emissions, ceteris paribus.

Our study makes several unique contributions. First, the

study addresses modularity and commonality across the entire

life cycle, which contrasts with previous studies that only focus

on the relationship between a manufacturer and their first-tier

suppliers (e.g., Jacobs et al., 2007). Likewise, in contrast to

studies of product architecture which focus only on compo-

nents and materials present in final pr oducts (e.g., Danese &

Filippini, 2013), our model and data are the only known ones

that include intermediaryprocesses used at multiple tiers of the

supply network. Second, most studies of modularity have used

perceptual measures of modularity (e.g., Jacobs et al., 2 007);

in contrast, our study uses secondary data, enabling a more

objective measure of modularity. Third, our study is the first in

this area to use network community detection algorithms

(Blondel, Guillaume, Lambiotte, & Lefebvre, 2008; Fortunato

and Hric 2016) to operationalize network modularity. Fourth,

our study is the first to examine the linkage between modular-

ity, commonality, and GHG emissions (or energy intensity).

Finally, our study uses a sample roughlyan order of magnitude

larger than any previous study of modularity and commonality,

encompassing the whole manufacturing sector.

We first review the relevant literature that provides some

general insight into our research question. We present our

theoretical proposition and argument, linking process net-

work modularity, commonality, and GHG emissions. This

article then discusses the study's methodology and empirical

findings and concludes with implications for theory and

future research.

94 DOOLEY ET AL.