Journal of Cleaner Production 57 (2013) 178e187

Contents lists available at SciVerse ScienceDirect

Journal of Cleaner Production

journal homepage: www.elsevier.com/locate/jclepro

A product-mix decision model using green manufacturing

technologies under activity-based costing
Wen-Hsien Tsai a, *, Hui-Chiao Chen a, Jun-Der Leu a, Yao-Chung Chang a, Thomas W. Lin b
a
b

National Central University, Jhongli, Taoyuan 32001, Taiwan

Leventhal School of Accounting, Marshall School of Business, University of Southern California, Los Angeles, CA 90089-0808, USA

a r t i c l e i n f o

a b s t r a c t

Article history:
Received 13 June 2012
Received in revised form
4 April 2013
Accepted 9 April 2013
Available online 15 June 2013

The purpose of this study is to assess how the integration of activity-based costing (ABC) and the theory
of constraints (TOC), as well as the application of a mixed-integer programming (MIP) model, can assist
in making decisions about product-mix using green manufacturing technologies (GMTs). This study
proposes a mathematical programming model to analyze the protability of a product-mix decision
based on the ABC and TOC, with the adoption of new GMTs. Using a numerical example from a metal
component parts manufacturer in the automotive industry, the ndings of this study provide insight into
the value of mathematical programming approaches for GMTs investment and product-mix decision
making based on ABC systems while simultaneously improving the value of green manufacturing
technology investments.

2013 Published by Elsevier Ltd.

Keywords:
Activity-based costing (ABC)
Green manufacturing technologies (GMTs)
Theory of constraints (TOC)
Product-mix decision
Mathematical programming approach

1. Introduction
Since production and consumption activities have been generating negative impact effects on the environment through their
production, use and disposal in recent years, environmental issues
have increasingly become primary concerns of corporate management. Meanwhile, stakeholders have also begun to put pressure on
organizations to be more environmentally responsible for both
their products and processes given regulatory requirements,
product stewardship, public image and potential competitive advantages. In such circumstances, many organizations are nding
that going beyond regulatory compliance can create value for their
customers and respond to pressure groups. For example, many facility operators are actively investigating the use of green
manufacturing technologies (GMTs), such as aqueous degreasers
and powder coatings, in an effort to reduce toxic air emissions and
control costs associated with the treatment of contaminated
efuent.
However, in practice, most companies continue to struggle to
invest their scarce resources to adopt new GMTs because of a lack of
proper justication tools that would both identify protable and
non-protable products and account for resource constraints.

* Corresponding author. Tel.: 886 3 4267247; fax: 886 3 4222891.

E-mail address: [email protected] (W.-H. Tsai).
0959-6526/$ e see front matter
2013 Published by Elsevier Ltd.
http://dx.doi.org/10.1016/j.jclepro.2013.04.011

Facility operators need accurate information from their management accounting systems in order to make protable choices
regarding environmental spending. The use of activity-based
costing (ABC) on protability analysis is called upon to help break
down and analyze the nature of environmental costs in products
that generate those costs, in order to lead to the best decisions.
Even though ABC provides a systematic approach to analyzing nonvalue added activities, processes and products, ABC may ignore
resource constraints in the production process (Yahya-Zadeh, 1998;
Kee and Schmidt, 2000). Under such conditions, the theory of
constraints (TOC) may provide a better solution when making decisions about reducing environmentally-damaging products from
the mix (Onwubolu and Mutingi, 2001; Lockhart and Taylor, 2007).
The purpose of this research paper is to propose a mathematical
programming model that analyzes the protability of a productmix decision based on the ABC and TOC, with the adoption of
new GMTs.
2. Background
Due to the growing awareness of environmental issues, new
GMTs have been widely considered for maintaining a competitive
advantage (Kong and White, 2010), enhancing production skills
(Puurunen and Vasara, 2007) and coping with pressure groups (Tsai
et al., 2011a). Nevertheless, facility managers still have difculty
analyzing the impact of GMTs on prots because of a lack of

W.-H. Tsai et al. / Journal of Cleaner Production 57 (2013) 178e187

appropriate performance measures (Azzone and Noci, 1998; Kuldip,

2006), systematic analysis of overhead and xed operating costs
(Kaplan and Cooper, 1998) and considerations of constrained resources in the production process (Wang et al., 2009; Tsai et al.,
2011b, 2012b). Recently, researchers have suggested a number of
methodologies and evaluation techniques that look promising insofar as the economic justication process for these advanced
manufacturing technologies is concerned. The solutions to these
shortages should be sought in the ABC and TOC.
2.1. Activity-based costing (ABC)
Currently, ABC has become a popular cost accounting system to
overcome the shortcomings of traditional cost accounting systems.
There are four key benets to the ABC system: (1) accurate identication of product cost, especially overhead; (2) more precise
information about value-added and non-value-added costs by the
identication of cost drivers; (3) direct allotment of costs to products or processes that consume resources; and (4) identication of
non-value-adding costs (Kaplan, 1989; Canada et al., 1996; Hilton,
2005). Various surveys have also indicated that the ABC approach
has been used to analyze different kinds of management decisions,
including: pricing, quoting, product-mix and joint products decisions (Tsai and Lai, 2007; Kee, 2007, 2008), outsourcing decisions
(Kee, 1998; Tsai and Lai, 2007; Tsai et al., 2007), quality improvement (Tsai, 1998), design and development (Qian and Ben-Arieh,
2008), nancial and physical ows analysis (Comelli et al., 2008)
and environmental management, among others (Tsai et al., 2007;
Tsai and Hung, 2009a, 2009b; Silva and Amaral, 2009).
The ABC uses a two-stage procedure to assign resource costs to
cost objects (Turney, 1991; Tsai, 2010). In the rst stage, resource
costs are assigned by resource drivers to activity cost pools that can
be classied by activity levels such as unit, batch, product and facility (Cooper, 1990). Each activity level can have several activity
cost pools. In the second stage, activity cost pools are assigned to
cost objects by activity drivers, which are activities that incur costs
(Hilton, 2005). Because the problem of assessing the benets and
cost drivers of the new GMTs are less well studied and understood,
in this study, we follow Kims (2009) approach and attempt to
identify the resource and activity drivers of production processes
using new GMTs. First, we estimate overall costs and possible resources of various production processes using new GMTs. Second,
we construct resource cost pools and develop resource cost drivers.
Third, we dene the main activities that consume resources of
production and calculate a total cost of each activity. Finally, we
identify the second stage activity drivers and allocate activity costs
to various products using the cost driver.
2.2. The theory of constraints (TOC)
In early ABC research, numerous studies demonstrated the differences between production costs using ABC and traditional
costing systems. Nevertheless, a method to reach the optimal
product-mix under ABC was seldom mentioned. Furthermore, ABC
has been criticized for its failure to incorporate constraints into
production-mix decisions (Kee and Schmidt, 2000). On account of
this, Kee (1995) began to integrate ABC with TOC in the productmix decision analysis. Several researchers proposed various
mathematical programming models in succession to conduct the
product-mix decision analysis under ABC (Kee, 1995; Malik and
Sullivan, 1995; Yahya-Zadeh, 1998; Kee and Schmidt, 2000; Tsai
et al., 2012c).
According to TOC, the performance of any production system is
determined by its slowest process (Goldratt et al., 1986; YahyaZadeh, 1998). Thus, managers should focus their attention on

179

capacity constraint resources (CCRs) and removing bottleneck

processes (Onwubolu and Mutingi, 2001). Since TOC treats overhead and operating expenses as given, TOC can also be viewed as a
short-run optimization procedure for managing resources and
opening bottlenecks, with the goal of maximizing throughput
(Holmen, 1995). The ABC and TOC approaches appear to complement one another in addressing short-run operational and longrun cost management problems (Kee, 1995; Lockhart and Taylor,
2007). Several researchers have suggested that TOC should be
used for short-term production-mix decisions where costs are
predominantly xed (Noreen et al., 1995) and that ABC should be
used to determine any increases or decreases in capacity and
products because all costs tend toward being variable in the long
term.
The following steps, called the TOC procedure, provide on-going
improvement to the throughput of a system (Goldratt, 1990):
1.
2.
3.
4.
5.

Identify the systems constraints.

Decide how to exploit the systems constraints.
Subordinate everything else to the above decision.
Elevate the systems constraints.
If in the previous steps, a constraint has been broken, go back to
step 1.

The above ve steps of the TOC can be applied to the productmix decision problem. Through the cycle of these ve steps, TOC
successively relieves bottlenecks and their associated constraints
by expanding the obtainable resources (capacities) or by improving
the companys operations (Tsai and Lin, 1990; Tsai et al., 2011a). In
addition, the product-mix decision problem that is solved using
TOC can also be formulated as a linear programming (LP) model
(Tsai and Lin, 1990; Luebbe and Finch, 1992; Plenert, 1993; Tsai
et al., 2011b, 2012a).
As compared to Kee (1995), the model proposed in this paper
has the following additional characteristics: (1) considering Environmental Regulatory Cost (xed cost), Volatile Organic Compounds (VOCs) emission cost (piecewise linear cost) and the
associated constraints to restrain the VOCs emission quantity, (2)
considering the facility-level activity cost increasing with a stepwise xed cost function, and (3) considering the capacity expansions for direct labor hours and machine hours.
In terms of point (3) mentioned above, regarding capacity
expansion, we consider capacity expansion for direct labor hours by
using overtime work with higher wage rates, which can be
formulated as the piecewise linear cost function. Herein, we also
consider capacity expansion for machine hours by renting machines or by buying machines, which can be formulated as a
stepwise linear cost function. Currently, this method is applied in
various research topics (Tsai and Kuo, 2004; Tsai and Lai, 2007; Tsai
et al., 2007, 2008, 2010).
2.3. Priming and top coating processes and activities in the
automotive industry
Many products and parts manufactured in the metal industries
are required to receive both priming and top coating processes.
Typical primer-top coat technologies are adopted for miscellaneous
metal work pieces, plastic part, and the automotive industry. The
typical sequence of operations performed by such a primer-topcoat
system in the automotive industry includes multiple stages, as
shown in Fig. 1.
In the beginning of a primer-topcoat process, the incoming
metal component parts or metalwork pieces for the car body are
often cleaned (e.g., degreased or steam-cleaned) and are passed
through a zinc phosphate stage before being immersed in a large

180

W.-H. Tsai et al. / Journal of Cleaner Production 57 (2013) 178e187

decision. The cost elements related to the product-mix decision

model are described below:
1. Total revenue:
The terms in Eq. (1) represent the total revenue of bid prices for
products.
n
X

Pi Xi

(1)

i1

2. Total direct labor cost:

Fig. 1. A typical process for applying a primer-topcoating system in the automotive

industry.

electro deposition tank in which a cathodic or anodic primer is

applied. After using dry-off ovens, the work pieces with electrodeposited primer are cured in a baking oven at temperatures
ranging from 300 to 400  F. In the sealers and sound deadening
phase, the metal work pieces receive a polyvinyl chloride (PVC)
coating that provides sound-deadening attributes. After undergoing light sanding, a second coat of primer is applied and then the
metal work pieces are again dried and cured in a baking oven. In the
top coating and/or clear coating phase, the metal work pieces enter
the top coating spray booth and a basecoat, a solid color top coat or
a wet-on-wet clear coat may be applied. Eventually, the metal work
pieces enter the nal baking oven in which the top coats are cured.
Given the importance of the primer-top coat process in the
automotive industry, new GMTs that prevent pollution are important issues for operations. Facilities typically use new GMTs such as
thermal oxidizers, catalytic incinerators, and carbon absorbers, or a
combination of these technologies, to control VOCs. Below, we will
describe the elements of product cost that will be used in developing the product-mix decision model under ABC systems.

We assume that direct labor resources can be expanded using

overtime work or additional night shifts, or by hiring workers at a
higher wage rate. The total direct labor cost function will be a
piecewise linear function, as shown in Fig. 2. The available normal
direct labor hours are LH1 and the direct labor hours can be
expanded to LH2 with the total direct labor cost being LC1 and LC2 at
LH1 and LH2, respectively.
Thus, the total direct labor cost and associated constraints are as
follows (Tsai and Lin, 1990; Tsai et al., 2011b):

Total direct labor cost LC1 m1 LC2 m2

(2)

Constraints:

TL LH1 m1 LH2 m2

(3)

m0  h1  0

(4)

m1  h1  h2  0

(5)

m2  h2  0

(6)

m0 m1 m2 1

(7)

2.4. Product-mix decision model with capacity expansions

h1 h2 1

(8)

In this section, we develop a product-mix decision model using

the primer-top coat system as an example. In addition to the
assumption about the primer-top coat process, the product-mix
decision model presented includes the following assumptions:

(h1, h2) is an SOS1 set of 0e1 variables, within which exactly one
variable must be non-zero; (m0, m1, m2) is an SOS2 set of nonnegative variables, within which at most two adjacent variables,
in the order given to the set, can be non-zero (Beale and Tomlin,
1970; Williams, 1985); TL is the total direct labor hours we need,
and its function depends on the case under study.

1. The activities in this green, multi-product primer-topcoat system have been classied as unit-level, batch-level, productlevel, and facility-level activities. The related resource drivers
and activity drivers have also been chosen by the companys
ABC team through an ABC study.
2. The unit selling prices are constant within the relevant range.
3. The specic process is regarded as a stepwise xed cost that
varies with machine hours.
4. Machine hour resources can be expanded by renting or purchasing additional machines.
5. Direct labor resources can be expanded by using overtime work
or additional night shifts with higher wage rates.
6. VOC emissions are taxed at different rates, which are dependent on emission quantities, and the cost of VOC emissions is
regarded as a piecewise variable cost.

3. Stepwise machine cost:

The total machine cost is regarded as a common xed cost, and
its cost function is assumed to be a stepwise function (as shown in

3. A mathematical programming model

The following is a discussion on capacity expansions, capacity
constraints and VOC emission costs, which are incorporated into
the mathematical programming model for determining the optimal

Fig. 2. Piecewise direct labor cost.

W.-H. Tsai et al. / Journal of Cleaner Production 57 (2013) 178e187

Fig. 3) that varies with machine hours, as observed from prior cost
behavior analysis. The total machine cost is dened as FC0 under
the current capacity of MH0 machine hours. If the capacity is successively expanded to MH1, MH2,., MHt machine hours, then the
total machine cost increases to FC1, FC2,.,FCt, respectively. Let lih be
the requirement of machine hours for one unit of product i. Thus,
the total machine cost and associated machine hour constraints are
as follows (Tsai and Lin, 1990):

Total machine cost

t
X

FCk qk

181

Cost

(9)

VOC emissions
quantity (metric ton)

k0

Constraints:
n
X

lih Xi 

i1
t
X

t
X

MHk qk

Fig. 4. VOCs emission costs.

(10)

k0

qk 1

TVOC VOQ1 a1 VOQ2 a2 VOQ3 a3

(13)

a0  u1  0

(14)

a1  u1  u2  0

(15)

a2  u2  u3  0

(16)

a3  u3  0

(17)

a0 a1 a2 a3 1

(18)

u1 u2 u3 1

(19)

(11)

k0

(q0, q1,.,qt) is an SOS1 set of 0e1 variables, within which exactly

one variable must be non-zero (Beale and Tomlin, 1970; Williams,
1985). When qm 1 (m s 0), the capacity needs to be expanded
to the mth level, i.e. MHm machine hour.
4. Environmental and social cost e VOC emission costs:
Even though new GMTs can reduce emissions, there are still
small amounts of VOC emissions from the topcoat spray processes.
The emission costs are measured by the life cycle assessment (LCA)
method (Tsai et al., 2011a). LCA is a method of comparison of
environmental impacts of products, technologies or services with a
view of their whole life cycle. The VOC emissions to all areas of the
environment during product production, use and disposal are
considered. Following Ward and Chapmans (1995) approaches, Eq.
(12) is used to quantize the VOC emissions. VOC emissions are also
assumed to be taxed at different rates depending on the quantity of
these. Hence, the total VOCs emission cost function will be a
piecewise linear function (see Fig. 4). With increasing VOC emissions, taxation will be increased. The quantity of VOC emissions can
be increased from VOQ1 to VOQ2 and VOQ3. Therefore, the total VOC
emission cost is VOP1, VOP2, and VOP3 at VOQ1, VOQ2 and VOQ3,
respectively (Tsai et al., 2011b).
The total VOC emission costs and associated constraints are as
follows:

Total VOC emission costs VOP1 a1 VOP2 a2 VOP3 a3

Similarly, (u1, u2, u3) is also an SOS1 set of 0e1 variables,

within which exactly one variable must be non-zero; (a0, a1, a2,
a3) is an SOS2 set of non-negative variables, within which at
most two adjacent variables, in the order given to the set, can be
non-zero. TVOC in Eq. (13) is the total quantity of VOC emissions;
the function of TVOC will depend on the case problem. [Note that
the function of TVOC in Section 4. Numerical example is
(2X1 X2 X3), which means that the quantity of VOC emissions
for one unit of Products 1, 2 and 3 are 2, 1, and 1, respectively.]

3.1. Integrated cost models

The model for product-mix decision making with capacity expansions under activity-based costing is as follows:
Maximize:

(12)

p Total Revenue  Total Direct Material Cost

VOC emission constraints:

 Total Direct Labor Cost

 Total Unit  Level Activity Cost

Cost

 Total Batch  Level Activity Cost

 Total Product  Level Activity Cost
 Total Facility  Level Activity Cost
 Total VOCs Emission Cost

n
X

Pi Xi 

i1

X
jU

Machine Hour
Fig. 3. Stepwise machine costs.

n X
s
X

Cm aim Xi  LC1 m1 LC2 m2 

i1 m1

dj lij Xi 

n X
X
i 1 jB

n
X
i1

dj dij Bij 

n X
X
i 1 jP

dj rij Ri 

t
X

FCk qk

k0

 VOP1 a1 VOP2 a2 VOP3 a3

(20)

182

W.-H. Tsai et al. / Journal of Cleaner Production 57 (2013) 178e187

(h1, h2), qk (k 0, 1,.,t), (u1, u2, u3): SOS1 set of 0e1 variables.
h1, h2, Ri (i 1, 2,.,n), qk (k 0, 1,.,t), u1, u2, u3: 0e1 variables.

Subject to:
Direct material constraints:
n
X

where

aim Xi  Qm ;

m 1; 2; .; s

(21)

i1

Direct labor constraints:

TL LH1 m1 LH2 m2

(22)

m0  h1  0

(23)

m1  h1  h2  0

(24)

m2  h2  0

(25)

m0 m1 m2 1

(26)

h1 h2 1

(27)

Batch-level activity constraints:

Xi  sij Bij ;
n
X

i 1; 2; .; n;

dij Bij  Tj ;

jB

jB

(28)

(29)

i1

Product-level constraints:

Xi  Di Ri ;
n
X

i 1; 2; .; n

ri Ri  Vj

(30)

(31)

i1

Machine hour constraints:

n
X

t
X

lih Xi 

i1
t
X

MHk qk  0

(32)

k0

qk 1

(33)

k0

VOC emissions constraints:

TVOC VOQ1 a1 VOQ2 a2 VOQ3 a3

(34)

a0  u1  0

(35)

a1  u1  u2  0

(36)

a2  u2  u3  0

(37)

a3  u3  0

(38)

a0 a1 a2 a3 1

(39)

u1 u2 u3 1

(40)

Xi  0;

i 1; 2; .; n

Bij  0;

i 1; 2; .; n

jB

(m0, m1, m2), (a0, a1, a2, a3): SOS2 set of non-negative variables.

p e the companys prot;

Xi e the production quantity of product i;
Pi e the unit selling price of product i;
Cm e the unit cost of the mth material;
aim e the requirement of the mth material for one unit of
product i;
LH1 e the available normal direct labor hours;
LC1 e total direct labor cost in LH1;
LC2 e total direct labor cost in LH2;
dj e the actual running activity cost per activity driver for activity j;
lij e the requirement of the activity driver of unit-level activity j
(j U) for one unit of product i;
dij e the requirement of the activity driver of batch-level activity
j (j B) for product i;
Bij e the number of batches of batch-level activity j (j B) for
product i;
rij e the requirement of the activity driver of product-level activity j (j P) for product i;
sij e the number of units per batch of batch-level activity j (j B)
for product i;
Ri e the indicator for producing product i (Ri 1) or not producing product i (Ri 0);
TL e total direct labor hours;
MHk e the available direct machine hours;
VOP1 e total VOC emissions cost in VOQ1;
VOP2 e total VOC emissions cost in VOQ2;
VOP3 e total VOC emissions cost in VOQ3;
TVOC e the total quantity of VOC emissions;
VOQ1 e the total quantity of VOC emissions from product
execution (must increase the carbon tax rate);
VOQ2 e the total quantity of VOC emissions from product
execution (must increase the carbon tax rate);
VOQ3 e the total quantity of VOC emissions from product
execution (must increase the carbon tax rate);
qk e an SOS1 (special ordered set of type 1) set of 0e1 variables,
within which exactly one variable must be non-zero (Beale and
Tomlin, 1970; Williams, 1985), qk 1 (k s 0) means that the
capacity needs to be expanded to the kth level, i.e., MHk machine
hours;
h1, h2; u1, u2, u3 e an SOS1 (special ordered set of type 1) set of
0e1 variables, within which exactly one variable must be nonzero (Beale and Tomlin, 1970; Williams, 1985);
m0, m1, m2; a0, a1, a2, a3 e an SOS2 (special ordered set of type 2)
set of non-negative variables, within which at most two adjacent variables, in the order given to the set, can be non-zero
(Beale and Tomlin, 1970; Williams, 1985);
Di e the maximum demand of product i;
Vj e the capacity limit of the activity driver of product-level
activity j (j P);
Tj e the capacity limit of the activity driver of batch-level activity
j (j B);
Qm e the available quantity of the material.
Eq. (20) is the prot function (p) to be maximized in the model. The
P
rst term ni 1 Pi Xi of Eq. (20) is Total Revenue, and is described in
P
P
Eq. (1). The second term ni 1 sm 1 cm aim Xi of Eq. (20) is Total
Direct Material Cost and the associated constraints are in Eq. (21),
where the available quantity of the mth material is Qm. The third term
(LC1m1 LC2m2) of Eq. (20) is Total Direct Labor Cost and the associated
constraints are in Eqs. (22)e(27), which come from Eqs. (3)e(8).

W.-H. Tsai et al. / Journal of Cleaner Production 57 (2013) 178e187

P
P
The fourth term ni 1
dj lij Xi of Eq. (20) is Total Unit-level
jU

Activity Cost, which may have specic associated constraints (not

shown in the model), for a specic case problem. Strictly speaking,
direct material and direct labor costs are unit-level costs which
increase with the production quantity. The fth term
P
P
dj dij Bij of Eq. (20) is Total Batch-Level Activity Cost and
ni 1
jB

the associated constraints are in Eqs. (28) and (29), where Tj is the
capacity limit of the activity driver of batch-level activity j (j B) in
Eq. (29) and sijBij is the upper limit of the ith products quantity
from the perspective of the jth batch-level activity in Eq. (28). The
P
P
dj rij Ri of Eq. (20) is Total Product-Level Acsixth term ni 1
jp

tivity Cost and the associated constraints are in Eqs. (30) and (31). If
Ri 1, this means that the ith product will be produced in the
period, Eq. (30) is the constraint related to the demand limit of the
ith product and Eq. (31) is the constraint related to the capacity
limit of the activity driver of product-level activity j (j P) (Vj). The
P

t
seventh term
k 0 FCk qk of Eq. (20) is Total Facility-level Activity Cost and the associated constraints are in Eqs. (32) and (33)
which correspond to Eqs. (10) and (11). The machine hours
required increase proportionally with the production quantity, but
the cost of this is a stepwise xed cost function. This is the typical
example of theory of constraints.
The eighth term (VOP1a1 VOP2a2 VOP3a3) of Eq. (20) is Total
VOC Emissions Cost and the associated constraints are in Eqs. (34)e
(40) as described earlier in this section for Eqs. (12)e(19). TVOC in
Eq. (34) is the total quantity of VOC emissions, and the function of
TVOC will depend on the case problem.
As for the constraints in the model, some constraints are used
for constructing the specic cost functions like piecewise linear
functions and stepwise linear functions. However, the labor
resource constraint is hidden in the formulations. Some constraints
are the real constraints for resources, like Eqs. (21), (29), (31) and
(32) for setting up the resource limits of materials, of the activity
drivers of batch-level activities and product-level activities, and of
machine hours, respectively.
As mentioned in Section 2.2 the theory of constraints (TOC), we
can relax the bottleneck resources by using the methods of capacity
expansions for the specic resources. This will be done in the postoptimal analyses. However, the model proposed in this paper
incorporating the possible levels of capacity expansions for the
various resources. Then it can, simultaneously, consider more than
one resource constraints.

183

Despite using electrostatic turbo bells, the transfer efciency is

still too low for such paint spray guns. Thus, VOC emissions already
come near the permitted cap and are almost double what was
originally estimated when the permits were applied for.
Assume that Company A is considering producing products 1,
2 and 3 (i 1, 2, 3) by using new GMTs and that these products
need three different kinds of direct materials (m 1, 2, 3).
Company A needs to calculate the following essential costs in
producing these products: (1) unit-level costs and activities:
manufacturing costs include machine costs, labor costs and material costs; (2) batch-level costs: inventory handling costs, carbon absorption costs and setup costs; (3) product-level costs:
design costs; (4) facility-level costs: environment regulatory costs
and VOC emissions costs. The related data for this example are
presented in Table 1. Environmental Regulatory Cost, which refers
to the costs associated with handling regular inspections, discharging waste and ensuring that processes are in compliance
with the Environmental standard value specication according to
the laws and regulations of local government, is a xed cost and
so can be expressed with a constant (12,000 in this example).
Company A has to decide the optimal quantity of products with
its current capacity. By using Eqs. (20)e(40) and the example data
in Table 1, the green product-mix decision model for the example
is formulated as follows:
Maximize

p Total Revenue  Total Direct Material Cost

 Total Direct Labor Cost
 Total Unit  Level Activity Cost
 Total Batch  Level Activity Cost
 Total Product  Level Activity Cost
 Total Facility  Level Activity Cost
 Total VOCs Emission Cost

n
X

Pi Xi 

i1

X
jU

n
s
X
X

Cm aim Xi  LC1 m1 LC2 m2 

i1 m1

dj lij Xi 

n X
X

n
X
i1

dj dij Bij 

i 1 jB

n X
X
i 1 jP

dj rij Ri 

t
X

FCk qk

k0

 VOP1 a1 VOP2 a2 VOP3 a3

200X1 230X2 250X3  10*6 2*5 1*2 3*1X1
10*7 2*9 1*2 3*1X2 10*8 2*10 1*2
3*2X3   120; 000m1 220; 000m2  2*5 4*6X1

4. Numerical example
A numerical example is presented to illustrate the application of
a mathematical programming model to determine the optimal
product-mix under activity-based costing. Company A manufactures metal component parts for the automotive industry.
Numerous environmental problems are preventing the Company
from getting its nished components to the market. Company A
now has applied GMTs to produce their products and has overcome
the following environmental problems:
(1) Solid sludge from the metal pretreatment process requires the
disposal of relatively large volumes of poisonous waste.
(2) Large volumes of rinse water from the topcoating spray booth
for topcoating and/or clear coating need more special treatment due to the local water treatment plants requirements.
(3) The nished metal component parts do not have the same gloss
and the quality of the coating is unstable.

2*5 4*7X2 2*5 4*8X3   100*2B13

100*2B23 100*2B33   2*10B14 2*10B24
2*20B34  15*2B15 15*2B25 15*4B35 
 100*20R1 100*10R2 100*30R3   40; 000q0
75; 000q1 120; 000q2  100; 000a1 150; 000a2
195; 000a3 12; 000
91X1 99X2 100X3  120; 000m1  220; 000m2
 200B13  200B23  200B33  20B14  20B24  40B34
 30B15  30B25  60B35  2000R1  1000R2  3000R3
 40; 000q0  75; 000q1  120; 000q2  100; 000a1
 150; 000a2  195; 000a3  12; 000:
Subject to:
Direct material constraints:

184

W.-H. Tsai et al. / Journal of Cleaner Production 57 (2013) 178e187

Table 1
Example data.

Maximum demand
Selling price
Direct material (Unit-level)

Product 1

Product 2

Product 3

Available capacity

Di
Pi
ai1
ai2
ai3
ai4

3000
200
6
5
2
1

2500
230
7
9
2
1

5000
250
8
10
2
2

W1
W2
W3
W4

m1
m2
m3
m 4 (Disposal of
hazardous material)
Machine hours
Labor hours

$10/unit
$2/unit
$1/unit
$3/unit
$2
$4

1
2

li1
li2

5
6

5
7

5
8

Batch-level activity
Inventory handling

Handling hours

$100

Carbon adsorption

Machine hours

$2

Setup

Setup hours

$15

Product- level activity e Design

Facility-level cost
(Excluding regulatory costs)
Machine hours
Environment regulatory cost
Direct labor constraint-cost
Labor hours
Wage rate
VOCs emission constraint
Cost
Emission quantities
Tax rate

Drawings
FC0 $40,000

$100
FC1 $75,000

6
FC2 $120,000

di3
si3
di4
si4
di5
si5
ri6

2
10
10
4
2
5
20

2
10
10
4
2
5
10

2
30
20
8
4
10
20

MH0 20,000
Total cost
LC1 $120,000
LH1 30,000
r1 $4/h

MH1 30,000
$12,000
LC2 $ 220,000
LH2 50,000
r2 $5/h

MH2 40,000

VOP1 $100,000
VOQ1 10,000
T1 $10/ton

VOP2 $150,000
VOQ2 12,500
T2 $20/ton

VOP3 $195,000
VOQ3 14,000
T3 $30/ton

vi

Unit-level activity

6X1 7X2 8X3  120; 000

a1  u1  u2  0

5X1 9X2 10X3  100; 000

a2  u2  u3  0

2X1 2X2 2X3  50; 000

a3  u3  0

X1 X2 2X3  40; 000

a0 a1 a2 a3 1

Stepwise facility-level machine hour constraints:

5X1 5X2 5X3  20; 000q0  30; 000q1  40; 000q2  0

q0 q1 q2 1
Direct labor constraints:

6X1 7X2 8X3  30; 000m1  50; 000m2 0

m0  h1  0
m1  h1  h2  0
m2  h2  0

V6 55

X2  10B23  0
X3  30B33  0
2B13 2B23 2B33  900;
Batch-level carbon adsorption activity constraints:

h1 h2 1

X3  8B34  0

a0  u1  0

T5 3000

X1  10B13  0

X2  4B24  0

2X1 X2 X3  10; 000a1  12; 500a2  14; 000a3 0

T4 18,000

Batch-level inventory handling activity constraints:

X1  4B14  0

10B14 10B24 20B34  18; 000;

Batch-level setup activity constraints:

X1  5B15  0

120,000
100,000
50,000
40,000

T3 900

u1 u2 u3 1

m0 m1  m2  1

VOC emissions constraints:

W.-H. Tsai et al. / Journal of Cleaner Production 57 (2013) 178e187

X2  5B25  0
X3  10B35  0
2B15 2B25 4B35  3000;
Product-level constraints:

X1  3000R1  0
X2  2500R2  0
X3  5000R3  0
20R1 10R2 20R3  55
Xi  0;

i 1; 2; 3

Bij  0;

i 1; 2; 3;

j 3; 4; 5;

(h1, h2), (q0, q1, q2), (u1, u2, u3): SOS1 set of 0e1 variables.
h1, h2, R1, R2, R3, q0, q1, q2, u1, u2, u3: 0e1 variables.
This is a mixed-integer programming (MIP) model. We solve this
problem by utilizing the software, LINGO 13.0, and obtain the
following optimal solutions.

X1

q0
m0
h1
R1

a1
u1
B13
B14
B15

3000
0
0
0
1
1
1
300
750
600

X2

q1
m1
h2
R2

a2
u2
B23
B24
B25

0
0
0
1
0
0
0
0
0
0

X3

q2
m2
R0
R3

a3
u3
B33
B34
B35

4000
1
1
0
1
0
0
134
500
400

According to the results, the optimal quantity of green productmix is (X1, X2, X3) (3000, 0, 4000), which requires 50,000 units
(6  3000 7  0 8  4000) of the rst kind of material, 55,000
units (5  3000 9  0 10  4000) of the second kind of
material, 14,000 units (2  3000 2  0 2  4000) of the third
kind of material, 35,000 (5  3000 5  0 5  4000) machine
hours, 50,000 (6  3000 7  0 8  4000) direct labor hours,
and 10,000 (2  3000 1  0 1  4000) VOC emissions
quantities. The total prot p is $52,200. This means that the machine capacity does not exceed 40,000 h or that the company has to
rent machines, that the direct labor capacity is equal to 50,000 labor hours and that the VOC emissions are equal to 10,000 tons of
emission quantity which is in the upper limit of the rst range of
VOC emission cost function.
Company A has adopted ABC. Therefore, some related ABC data
can be obtained directly from the accounting department of Company A. We can also ask about the related business units to provide
related data required in the product-mix decision models. For
example, we can ask the production department to provide information concerning maximum demand, the sales department to
provide information concerning selling price, the purchasing
department to provide cost data for direct materials, and the accounting department to provide information on unit-level, batchlevel, product-level and facility-level activity costs obtained from
the ABC study. As for the Environmental Regulatory cost and VOC

185

emissions, we can invite the HSE management department to

provide information, and can also ask the human resources
department to provide related direct labor cost information.
However, the most direct, and fastest, method is to invite the decision maker to request business units to submit related data to the
accounting department for collection, and then provide this information for the study.
In addition, VOCs can also be obtained from the emission cost
analysis of the government unit or the company. Furthermore,
additional overtime situations and related operation activities,
caused by production capacity expansion, can also be provided by
the production department of Company A.
Using the TOC procedure, the constraints are identied and used
so that new GMTs are used most protably. Company A is able to
choose products that produce the least amount of VOC emissions
since the TOC approach fosters the selection of optimal productmix, which emits fewer VOC emissions. In effect, when the
optimal product-mix is chosen, the output volumes of product 2 are
zero. This means that there is an unsatised demand for product 2
in the market, and Company A may increase the price of product 2
to reect its environmental costs. Although investments in new
GMTs can be very expensive, by adopting the TOC procedure, a
sound environmental investment by Company A can maximize
their prots.
5. Discussion
It seems that the model proposed in this paper will select a
product mix with higher pollution with the only objective of
maximizing the prot of a product mix based on the most constrained resources. Although the model does not explicitly select an
optimal product mix that emits fewer VOC emissions, we can use
the related constraints to restrain VOC emission quantity within
certain limits. As a matter of fact, it is impractical to aim to minimize VOC emission quantity or cost. Alternatively, we can set up a
goal programming model to maximize the operating prot, to
minimize deviation from the VOC emission quantity target, and
meet other objectives. In the numerical example presented in
Section 4, it was assumed that the total VOC emissions cost function
would be a piecewise linear function. The unit cost for VOC emissions (i.e. tax rate) will increase with increasing VOC emissions
from $10/ton, $20/ton to $30/ton within the different ranges of
emission quantity [0, 10,000], [10,000, 12,500] and [12,500, 14,000],
respectively. This constraint will restrain the product-mix solution
to emit the VOCs within the upper limit of the rst range. Otherwise, the tax rate will increase double. Managers may want to
search for new GMTs with fewer VOC emissions from the top coat
spray processes to simultaneously increase operating prots and to
maintain corporate social responsibility.
6. Conclusion
In this paper, we considered the capacity expansions for machine hours and direct labor hours. We also used piecewise and
stepwise linear functions to approximate the non-linear direct labor costs and machine costs. Specically, we further extended the
piecewise linear function to quantize the VOC emissions so that we
could depict the total VOC emissions cost function, which in turn
helps to build a product-mix model using the green, multi-product
primer-topcoat system under ABC. Limited by cost considerations,
the prots combination also undergoes certain changes along with
adjustment, such that a maximum prot combination is calculated
according to the conditions. Compared with early scholars and
early articles that only use gures for description, such as Kee
(1995), this study has explored a detailed mathematical

186

W.-H. Tsai et al. / Journal of Cleaner Production 57 (2013) 178e187

programming model. Furthermore, consideration is given to the

carbon factor and costs have been discussed.
The integration of ABC and TOC and the application of an MIP
model in our research can improve the efciency and effect of the
product-mix decision model for the green, multi-product primertop coat system. By integrating ABC and TOC approaches, traditional drawbacks and problems can be solved (e.g., ignoring
resource constraints, relying heavily on the assumption of proportional cost structures, short-term focus of TOC, and so on). In
addition, these techniques can be readily utilized to improve
product-mix decisions in the context of the current regulatory
environment, while producing a better mix of environmentallyfriendly products.
As indicated in the research background, facility operators need
accurate information from their management accounting systems
to make protable choices, especially with regard to environmental
spending. This MIP model identies and incorporates the costs of
VOC emissions. The optimal product-mix from ABC combined with
the TOC approach will be more accurate than traditional accounting
approaches.
This paper specically focused on the priming and topcoating
system and metal component parts for the automotive industry,
and so is limited by only considering single objectives when evaluating the product-mix decision. Future research may attempt to
extend these techniques to different industries and diverse activities that may produce certain kinds of emissions. Additionally,
incorporating various unit selling prices, unit direct material costs
and other kinds of emissions into the product-mix decision model
to reect the dynamics of the real world may better evaluate
environmental impact. Besides, we can consider multiple objectives to make the model more realistic.

Acknowledgment
We would like to thank the National Science Council of Taiwan
for nancially supporting this research under Grant No. NSC1002410-H-008-007-MY3.

References
Azzone, G., Noci, G., 1998. Identifying effective PMSs for the development of Green
manufacturing strategies. International Journal of Operations & Production
Management 18 (4), 308e335.
Beale, E.M.L., Tomlin, J.A., 1970. Special facilities in a general mathematical programming system for non-convex problems using ordered sets of variables. In:
Lawrence, J. (Ed.), Proceedings of the 5th International Conference on Operational Research. Tavistock, London, pp. 447e454.
Canada, J.R., Sullivan, W.G., White, J.A., 1996. Capital Investment Analysis for Engineering and Management, second ed. Prentice Hall, Upper Saddle River, NJ:
Comelli, M., Fenies, P., Tchernev, N., 2008. A combined nancial and physical ows
evaluation for logistic process and tactical production planning: application in a
company supply chain. International Journal of Production Economics 112 (1),
77e95.
Cooper, R., 1990. Cost classication in unit-based and activity-based manufacturing
cost systems. Journal of Cost Management 4 (3), 4e14.
Goldratt, E., Cox, J., 1986. The Goal. North River Press, Croton-on-Hudson, New York.
Goldratt, E.M., 1990. Theory of Constraints: What is This Thing Called the Theory of
Constraints and How Should it be Implemented? North River, Croton-onHudson, NY.
Hilton, R.W., 2005. Managerial Accounting. McGraw-Hill, New York.
Holmen, J.S., 1995. ABC vs. TOC: Its a Matter of Time, Management Accounting,
pp. 37e40. January.
Kaplan, R., Cooper, R., 1998. Cost & Effect: Using Integrated Cost Systems to Drive
Protability and Performance. Harvard Business School Press, Boston, MA,
pp. 13e180.
Kaplan, R.S., 1989. Management accounting for advanced technological environments. Science 245 (25), 819e823.
Kee, R., Schmidt, C., 2000. A comparative analysis of utilizing activity-based costing
and the theory of constraints for making product-mix decisions. International
Journal of Production Economics 63 (1), 1e17.

Kee, R., 1995. Integrating activity-based costing with the theory of constraints
to enhance production-related decision making. Accounting Horizons 9 (4),
48e61.
Kee, R., 1998. Integrating ABC and the theory of constraints to evaluate outsourcing
decisions. Journal of Cost Management 12 (1), 24e36.
Kee, R., 2007. Cost-volume-prot analysis incorporating the cost of capital. Journal
of Managerial Issue 19 (4), 478e493.
Kee, R., 2008. The sufciency of product and variable costs for production-related
decisions when economies of scope are present. International Journal of Production Economics 114 (2), 682e696.
Kim, J., 2009. Activity-based framework for cost savings through the implementation of an ERP system. International Journal of Production Research 47 (7),
1913e1929.
Kong, G., White, R., 2010. Toward cleaner production of hot dip galvanizing industry
in China. Journal of Cleaner Production 18 (10e11), 1092e1099.
Kuldip, S.S., 2006. Performance value analysis for justication of green
manufacturing systems. Journal of Advanced Manufacturing Systems 5 (1),
59e73.
Lockhart, J., Taylor, A., 2007. Environmental considerations in product mix decisions
using ABC and TOC. Management Accounting Quarterly 9 (1), 13e31.
Luebbe, R., Finch, B., 1992. Theory of constraints and linear programming. International Journal of Production Research 30, 1471e1478.
Malik, S.A., Sullivan, W.G., 1995. Impact of ABC information on product mix and
costing decisions. IEEE Transactions on Engineering Management 42 (2),
171e176.
Noreen, E., Smith, D., Mackey, J.T., 1995. The Theory of Constraints and Its Implications for Management Accounting. North River Press, Great Barrington, Mass.
Onwubolu, G.C., Mutingi, M., 2001. Optimizing the multiple constrained resources
product mix problem using genetic algorithms. International Journal of Production Research 39 (9), 1897e1910.
Plenert, G., 1993. Optimizing theory of constraints when multiple constrained resources exist. European Journal of Operational Research 70
(1), 126e133.
Puurunen, K., Vasara, P., 2007. Opportunities for utilising nanotechnology in
reaching near-zero emissions in the paper industry. Journal of Cleaner Production 15 (13e14), 1287e1294.
Qian, L., Ben-Arieh, D., 2008. Parametric cost estimation based on activity-based
costing: a case study for design and development of rotational parts. International Journal of Production Economics 113 (2), 805e818.
Silva, P.R.S. da., Amaral, F.G., 2009. An integrated methodology for environmental
impacts and costs evaluation in industrial processes. Journal of Cleaner Production 17 (15), 1339e1350.
Tsai, W.-H., Hung, S.-J., 2009a. A fuzzy goal programming approach for green supply
chain optimisation under activity-based costing and performance evaluation
with a value-chain structure. International Journal of Production Research 47
(18), 4991e5017.
Tsai, W.-H., Hung, S.-J., 2009b. Treatment and recycling system optimisation with
activity based costing in WEEE reverse logistics management: an environmental supply chain perspective. International Journal of Production Research
47 (19), 5391e5420.
Tsai, W.-H., Kuo, L., 2004. Operating costs and capacity in the airline industry.
Journal of Air Transport Management 10 (4), 269e275.
Tsai, W.-H., Lai, C.-W., 2007. Outsourcing or capacity expansions: application of
activity based costing model on joint product decisions. Computers & Operations Research 34 (12), 3666e3681.
Tsai, W.-H., Lin, T.M.,1990. Nonlinear multiproduct CVP analysis with 0-1 mixed integer
programming. Engineering Costs and Production Economics 20 (1), 81e91.
Tsai, W.-H., Hsu, J.-L., Chen, C.-H., 2007. Integrating activity-based costing and
revenue management approaches to analyse their manufacturing outsourcing
decision with qualitative factors. International Journal of Revenue Management
1 (4), 367e387.
Tsai, W.-H., Lai, C.-W., Tseng, L.-J., Chou, W.-C., 2008. Embedding management
discretionary power into an ABC Model for a joint products mix decision. International Journal of Production Economics 115 (1), 210e220.
Tsai, W.-H., Kuo, L., Lin, T.W., Kuo, Y.-C., Shen, Y.-S., 2010. Price elasticity of demand
and capacity expansion features in an enhanced ABC product-mix decision
model. International Journal of Production Research 48 (21), 6387e6416.
Tsai, W.H., Chen, H.C., Liu, J.Y., Chen, S.P., Shen, Y.S., 2011a. Using activity-based
costing to evaluate capital investments for green manufacturing technologies.
International Journal of Production Research 49 (24), 7275e7292.
Tsai, W.H., Lin, S.J., Liu, J.Y., Lin, W.R., Lee, K.C., 2011b. Incorporating life cycle assessments into building project decision making: an energy consumption and
CO2 emission perspective. Energy 36 (5), 3022e3029.
Tsai, W.H., Lee, K.C., Lin, H.L., Liu, J.Y., Chou, Y.W., Lin, S.J., 2012a. A mixed activity
based costing decision model for green airline eet planning under the
constraints of the European Union Emissions Trading Scheme. Energy 39 (1),
218e226.
Tsai, W.H., Lin, W.R., Fan, Y.W., Lee, P.L., Lin, S.J., Hsu, J.L., 2012b. Applying a mathematical programming approach for a green product mix decision. International Journal of Production Research 50 (4), 1171e1184.
Tsai, W.H., Shen, Y.S., Lee, P.L., Chen, H.C., Kuo, L., Huang, C.C., 2012c. Integrating
information about the cost of carbon through activity-based costing. Journal of
Cleaner Production 36, 102e111.
Tsai, W.-H., 1998. Quality cost measurement under activity-based costing. International Journal of Quality and Reliability Management 15 (7), 719e752.

W.-H. Tsai et al. / Journal of Cleaner Production 57 (2013) 178e187

Tsai, W.-H., 2010. Project management accounting using activity-based costing
approach. In: Bidgoli, Hossein (Ed.), 2010. The Handbook of Technology Management, vol. 1. John Wiley & Sons, Inc., pp. 469e488.
Turney, P.B.B., 1991. Common Cents: the ABC Performance Break-throughdHow to
Succeed with Activity-based Costing. Cost Technology, Hillsboro.
Ward, S.C., Chapman, C.B., 1995. Risk-management perspective on the project lifecycle. International Journal of Project Management 13 (3), 145e149.

187

Wang, J.Q., Sun, S.D., Si, S.B., Yang, H.A., 2009. Theory of constraints product mix
optimization. International Journal of Production Research 47 (16), 4521e4543.
Williams, H.P., 1985. Model Building in Mathematical Programming, second ed. John
Wiley & Sons, Inc., NY, pp. 173e177.
Yahya-Zadeh, M., 1998. Product-mix decisions under activity-based costing with
resource constraints and non-proportional activity costs. Journal of Applied
Business Research 14 (4), 39.