Energy 124 (2017) 46e58

Contents lists available at ScienceDirect

Energy
journal homepage: www.elsevier.com/locate/energy

Embodied energy analysis of building materials: An improved

IO-based hybrid method using sectoral disaggregation
Manish K. Dixit
Department of Construction Science, Texas A&M University, 3137 TAMU, College Station, TX, 77843, USA

a r t i c l e i n f o a b s t r a c t

Article history: Buildings consume approximately half of the annual energy supply of the United States in their con-
Received 22 November 2016 struction and operation. To effectively decrease this extensive energy footprint, both embodied and
Received in revised form operating energy must be quantified and optimized. Although validated standard methods are available
11 January 2017
to compute operating energy, quantifying embodied energy is still complicated and inconsistent. Among
Accepted 8 February 2017
the available embodied energy calculation methods, an input-output-based hybrid (IOH) method has the
potential to offer a more complete calculation. However, its calculation lacks specificity and reliability,
which can be improved using suggestions provided by literature. Studies across the globe have proposed
Keywords:
Embodied energy
techniques such as sectoral disaggregation to enhance not only the specificity and reliability, but also the
Life cycle energy completeness, of an IOH method.
Embodied carbon This study investigated and improved an IOH method of embodied energy calculation. Using the
Building materials improved method, the embodied energy of commonly used building materials was calculated and
Life cycle assessment evaluated. The results demonstrate a significant difference after disaggregating the relevant industry
sectors. The study concludes that using embodied energy values without industry sectors being dis-
aggregated can cause significant errors in a building’s embodied energy calculation.
© 2017 Elsevier Ltd. All rights reserved.

1. Introduction consumes recurrent embodied energy (REE) directly and indirectly

The building sector consumes nearly two-fifths of the annual
global energy supply in building operation alone, adding significant
carbon emissions to the atmosphere [1]. In the United States,
approximately 48% of annual energy is consumed in building
construction and operation [2]. Most of this energy originates from
fossil fuel sources and consequently contributes to over 39% of the
nation’s annual carbon emissions [1e3]. The total energy use of a
building includes embodied and operating energy; this study fo-
cuses on embodied energy. During a building’s initial construction,
embodied energy is used directly in onsite and offsite construction,
fabrication, transportation, administration, and related services,
and indirectly through the use of building materials, assemblies,
and equipment [3e5]. Each product installed in the building con-
sumes energy during its raw material extraction, manufacturing,
and delivery to the construction site. The sum of all direct and in-
direct energy consumed during building construction is called
initial embodied energy (IEE) [6,7]. While occupied, a building
E-mail address: mandix72@hotmail.com.
[7e9] throughout the maintenance, repair, replacement, and
renovation phases. To effectively optimize the total energy foot-
print of the building sector, reducing both embodied and operating
energy is recommended [10e12]. Due to building energy research
targeting mostly operating energy, advanced and energy efficient
materials, appliances, and assemblies are being installed in build-
ings that gradually decrease their operating energy [13e15]. Also,
more standardized methods and tools are now available to
consistently compute operating energy [6]. Quantifying embodied
energy, however, is still complicated and resource-intensive and
requires extensive quality data, which usually is not available
[16e19]. The lack of a standard method to completely and accu-
rately calculate embodied energy further complicates the
embodied energy analysis of buildings [6,17,20].
The available embodied energy calculation methods are prom-
ising but also have flaws that must be addressed [8,21e23]. For
instance, process-based methods provide material-specific and
reliable results but are grossly incomplete due to a system
boundary truncation [8,16,24]. The IO-based methods are relatively
complete but do not provide material-specific results [1,24]. A
method’s completeness refers to the extent to which all major and

http://dx.doi.org/10.1016/j.energy.2017.02.047
0360-5442/© 2017 Elsevier Ltd. All rights reserved.
 M.K. Dixit / Energy 124 (2017) 46e58
minor energy flows are covered in the calculation [8,22]. A hybrid
method combines the reliability of a process-based method with
the system boundary completeness of an IO-based method to
provide a complete and reliable embodied energy calculation [7,17].
However, if an IO-based hybrid framework is applied, the results
still are highly aggregated lacking specificity [8,25,26]. This
inability of an IO-based hybrid (IOH) method to provide material-
specific results is a major problem that may be addressed
through disaggregating an industry sector in an IO framework.
Studies (e.g. Refs. [16,25e29]) have demonstrated the process of
sectoral disaggregation using similar approaches. Various versions
of an IOH method have been proposed by Bullard et al. [30], Carter
et al. [31], Treloar [25], Dixit et al. [6], and Crawford [24], showing
gradual improvements.
This study builds upon a previous study [6], which quantified
and integrated human and capital energy into an IOH model. In this
study, some of the industry sectors of the IO model were dis-
aggregated to demonstrate the calculation of building material-
specific embodied energy. Using the improved IOH model, the
embodied energy of commonly used construction materials was
calculated. The results were also compared with the values ob-
tained from other embodied energy calculation methods to high-
light and discuss major differences.
2. Literature review
2.1. Embodied energy calculation methods
Among the widely used embodied energy calculation methods
are: (1) process-based, (2) IO-based, and (3) hybrid methods
[1,4,24,25]. Each method differs in the extent of its system
boundary coverage [8]. A system boundary defines the energy in-
puts included in a study.
2.1.1. Process-based analysis
While process-based analysis provides material-specific energy
values, its calculations are significantly incomplete [16,21,22]. It is a
bottom-up approach that starts with gathering data of actual en-
ergy use from manufacturers and works backward covering most of
the direct and some indirect energy inputs. Beyond a certain point
in the upstream, gathering energy use becomes difficult and in
some cases impractical due to data unavailability. Consequently,
some processes for which data is unavailable remain excluded from
the calculation causing a truncation of the system boundary
[32,33]. Lenzen [33] quantified the incompleteness and truncation
error due to boundary truncation as 50% and 10%, respectively. To
calculate the embodied energy of a building, all material quantities
are calculated and multiplied with respective process-based
embodied energy intensities [1,24]. Although such calculations
provide energy values specific to the building, the energy embodied
in building construction, administration, and related services (e.g.
banking) remains excluded from the calculation [24e26]. The ac-
tivities of construction management, financing, code compliance,
etc., often involve energy consumption through added labor,
equipment and vehicle use, and other non-energy material usage
such as office supplies, which may be quite difficult to quantify if a
conventional process-based approach is used [23e25]. Other en-
ergy inputs, such as the energy used in remediating the adverse
environmental impacts of building material production and con-
struction, are also excluded from process-based calculations [8,25].
For instance, if a material manufacturing plant treats its emission,
discharge, or waste before releasing it to the environment, such
processes also consume energy, which must be allocated to build-
ing material production. Construction sites are also required to
treat and divert any construction discharge from the sites, which
47
may involve energy use through equipment and vehicles [8,23].
2.1.2. Input-output (IO)-based analysis
An IO-based embodied energy calculation is a top-down
approach in which direct requirement coefficients are derived
from an economic input-output model [34]. A direct requirement
coefficient represents inputs (in $) required by an industry sector
from other sectors to produce a unit dollar output [35]. Using direct
requirement coefficients, direct energy inputs from energy
providing sectors can be quantified [6,25]. Since each industry
sector has a chain of suppliers, all direct requirements also cause
indirect requirements. For instance, when the cement industry
sector increases its production of cement by $1, all other industry
sectors supplying inputs, such as coke, limestone, gypsum etc., also
increase their production in order to meet the increased demand.
Such increased requirements are termed stage one indirect re-
quirements [6,35]. Each supply sector also has a chain of other
sectors supplying inputs, which also increase their output as stage 2
indirect requirements. Likewise, there are indirect requirements
associated with stages 3, 4, 5, and so on up to infinity. The increased
output of the cement sector actually causes an economy-wide in-
direct impact [8,23e26]. The total indirect requirement is the sum
of all indirect requirements spread over stage 1 to stage ∞ [6,8].
These stages of indirect requirements are known as indirect stages.
Fig. 1 illustrates the direct and indirect inputs associated with
various stages. To calculate indirect requirements, direct re-
quirements are subtracted from the total requirements. The total
requirements are calculated using either Leontief’s inverse matrix
or power series approximation (PSA) method [6,24e26]. More
details on these methods can be found in Miller and Blair [35].
Using appropriate energy tariffs, the direct and indirect energy
requirements can be translated from monetary to energy units
[24e26,36].
An IO-based analysis covers a comprehensive system boundary,
as it accounts for the economy-wide inflows and outflows
[31,36e38]. However, its results may be highly aggregated and not
product-specific [39e41]. For instance, in an IO analysis, the
embodied energy of structural steel is calculated by quantifying the
energy intensity of the manufacturing sector producing structural
steel as well as a wide range of other steel products. This approach
assumes that all of the sector’s products have the same embodied
energy, which may be inaccurate [6,8,23e26]. Also, since the input
and output of energy sources is tabulated in monetary units, energy
tariffs are used to convert them to energy units [6,8]. If energy
tariffs are miscalculated, the quantified values of embodied energy
may contain significant errors [6,8,26]. The energy intensity of a
manufacturing sector is calculated in energy units per unit of
monetary output. To convert energy intensity into energy units per
mass or volume, product prices are used. Like energy prices, any
fluctuation in product prices grossly affects the embodied energy
calculation [6,8,26]. IO tables are also prepared based on the as-
sumptions of homogeneity and proportionality [24,25]. According
to the homogeneity assumption, each product produced by a sector
has a homogeneous mix of inputs that may not be correct. In the
proportionality assumption, the cost of a product is directly pro-
portional to its input requirements, which may be inaccurate
[24,37]. According to Treloar [25], an IO-based embodied energy
calculation entails counting energy inputs multiple times. For
instance, if the electricity sector purchases large amounts of coal,
natural gas, and petroleum, the total energy embodied in elec-
tricity, according to the IO model, would include all energy pur-
chased, as well as the energy content of generated electricity [38].
2.1.3. Hybrid analysis
Hybrid analyses combine the benefits of process-based and IO-
48 M.K. Dixit / Energy 124 (2017) 46e58
Fig. 1. Direct and indirect embodied energy components.
based analyses to improve the reliability and specificity of calcu-
lations [6,8,23e26]. Depending on the calculation framework and
the type of data used, a hybrid analysis can be either process-based
or IO-based [6].
2.1.3.1. Process-based hybrid (PH) embodied energy calculation.
In a process-based hybrid (PH) calculation, IO data are integrated
into a process-based model to improve the system boundary
completeness of the calculation [25,26,31]. The direct energy
component is often calculated using actual energy use data,
whereas the IO data are used to cover indirect energy inputs. The
material quantities are then multiplied by the PH energy intensities
to quantify the total embodied energy [23,24]. Although the system
boundary completeness is improved slightly, certain energy inputs,
such as the embodied energy of services, transportation, labor, etc.,
are still not covered in the calculation [23e26]. In fact, the calcu-
lation may miss some direct energy inputs completely, particularly
in the case of a complex product [24]. For instance, if the quantities
of steel are used to compute the embodied energy of structural
steel products, such as open web joists, girders, trusses, and beams,
the direct energy consumed in their fabrication and transportation
may not be covered in the calculation. Using a PH method for a
more complex product such as a building may cause significant
errors due to its incompleteness [8,24,26]. Treloar [25] discussed an
incompleteness of 0e100% in a PH analysis, as some products and
processes may be covered only partially or missed completely by
the analysis.
2.1.3.2. IO-based hybrid (IOH) embodied energy calculation.
The IOH method inserts process-based or statistical energy use data
into an IO model to improve the calculation reliability [25,36]. The
term hybrid denotes either the use of both physical and monetary
units or a combination of process and IO data [42]. Numerous
methods have been proposed to integrate process or sector level
energy data into the IO model. In an approach proposed by Carter
et al. [31], the actual energy use data of industry sectors are inserted
in the IO model in energy units, while other flows of goods and
services remain in monetary units. This method actually combines
a statistical analysis and an IO analysis to improve the IO model’s
reliability [25]. This approach is very useful, particularly if the en-
ergy use data are available for each industry sector. Also, since the
calculated values of energy intensities are in energy units per unit
of industry output, the use of energy tariffs is avoided, which is one
of the major reliability issues with IO-based methods. A similar
approach was applied by Lissens and Rensbergen [43], who
developed a BIOSCOPE model by combining statistical data of in-
dustrial energy use and IO data to compute carbon dioxide in-
tensities of Flemish industries. Treloar [25] discussed the
integration of statistical energy use data into IO model as another
type of hybrid method. Aden et al. [44] discussed a statistical
analysis as a bottom-up approach quite similar to a process-based
method. Treloar [25] also proposed a unique approach to extract
direct energy paths from an IO model and replace them with
process-based energy data. This is an excellent method, particularly
when process-based data are not available for all energy sectors. To
avoid counting energy inputs multiple times, Treloar [25] has rec-
ommended keeping all energy and non-energy inputs of energy
providing sectors at zero in the IO model and using a set of primary
energy factors (PEFs) instead. The PEFs that are used to convert
delivered energy to primary energy represent all energy used and
lost in producing and distributing one unit of delivered energy such
as electricity [38]. Later, Crawford [24] suggested a correction to
Treloar’ method and proposed extracting and replacing total energy
paths instead. The technique proposed by Treloar has been applied
by numerous studies such as Crawford and Stephan [45], which
performed a life cycle energy demand assessment of a passive
house in Belgium. Acquaye [26] discussed another variation of an
IOH approach in which direct, rather than total, energy paths are
extracted. Another interesting version of an IOH analysis was dis-
cussed by Suh et al. [42,46] as an integrated hybrid analysis, which
through a technology matrix integrates detailed process data at the
level of unit processes into an IO model. This method avoids double
counting of inputs and offers a consistent approach to allocation. In
another interesting study by Lenzen and Crawford [47], a Path
Exchange Method was proposed that replaces comparable IO data
with process data at the very detailed structural path level. This
method is unique in that it avoids integrating process data and
modifies only a part of a structural path without disturbing the
whole system [47].
Efforts have also been made to incorporate human and capital
energy into an IO-based embodied energy calculation. As an
important improvement, Costanza [48] made household and gov-
ernment sectors endogenous to an intermediate transaction ma-
trix. In 2004, a study by Lenzen and Treloar [49] compared two
methods for endogenising capital inputs in an IO table: (1)
augmentation method and (2) flow matrix method. In the
augmentation method, capital is integrated into the intermediate
transaction matrix as a homogeneous commodity in additional
column and row vectors. In the flow matrix method, these addi-
tional column and row vectors of capital are further disaggregated
by the consuming and supplying sectors. They found that the
augmentation method results in an overestimation of multipliers
such as capital investments. The flow matrix method reduces the
error; it, however, is extremely data-intensive. Suh and Huppes [46]
compared various versions of conventional and hybrid methods in
terms of data requirements, uncertainty, system boundary, time
and labor-intensiveness, and available computational tools. They
discussed that no method is better than the other methods and the
suitability and choice of a method depends on the data availability
and quality, goal and scope of a study, and time and resources
available.
 M.K. Dixit / Energy 124 (2017) 46e58
2.2. Lack of specificity in an IOH model
Even after several improvements to the IOH method, its results
are still aggregated for an entire industry sector, which produces
not only the product under study but also a variety of other prod-
ucts [23]. Several studies have suggested disaggregating industry
sectors to quantify product-specific embodied energy. The idea of
sector disaggregation has been around for quite some time. Early
studies, such as Fei [52] and Wolsky [53], discussed disaggregating
industry sectors into two sub-sectors. Fei [46] suggested an easily
calculated proxy for a disaggregated matrix, whereas Wolsky [53]
offered an easy approach to derive the inverse of a disaggregated
matrix. According to Wolsky [53], the published IO tables are not
“detailed enough” for making a specific assessment. Another
approach by Treloar [25] actually involved disaggregation through
the extraction of an energy path and its replacement with com-
parable process data. This unique approach was later refined and
applied by a number of studies such as Crawford [24], Acquaye [26],
and Omar et al. [16]. To enhance the specificity of a calculation,
disaggregation in an IOH method can be accomplished at two
levels: (1) energy sector level to quantify fuel-specific embodied
energy, and (2) construction sector or manufacturing sector level to
compute product or building-specific embodied energy. Studies,
such as Duffy [54], Acquaye et al. [55], Acquaye [26], and Omar et al.
[16], disaggregated the energy sector as well as construction sector.
In another study, Joshi [27] proposed six models at several
disaggregation levels. The first model requires no modifications
because the product under study is the representative product of
the industry sector manufacturing it. In the second model, a hy-
pothetical industry sector is added to represent a product under
study that is not a representative product of the industry sector or
else it is a new product not produced by any of the industry sectors.
The third model, which has been applied to several studies, sug-
gests disaggregating an industry sector into two sectors: one rep-
resenting the product of interest and the other covering all other
products manufactured by the original industry sector. Using the
share of product of interest in the total sectoral output, the original
industry sector (n
n matrix) can be disaggregated into two sub
sectors (n þ 1
n þ 1 matrix) representing the original IO matrix.
The fourth model involves disaggregating the original sector pro-
ducing the study product as well as the sectors supplying inputs to
the upstream stages of the study product. As one might expect, this
is only possible if detailed information on upstream stage inputs is
available. The fifth and sixth models incorporate the environmental
impacts of the use and the end-of-life phases. Wiedmann et al. [50]
and Lenzen et al. [51] discussed the Australian Industrial Ecology
Virtual Laboratory (IELab) that facilitates the availability of IOH data
with spatial and sectoral disaggregation for more specific calcula-
tions. Cheng et al. [28] added a hypothetical sector representing
construction pollution to the original construction sector of Japan’s
13-sector IO table, resulting in a 14-sector matrix. They treated
construction pollution as a byproduct or a secondary product.
Acquaye et al. [55] disaggregated the Irish energy supply and
construction sectors to improve the specificity and reliability of
embodied energy results. Their results showed that not dis-
aggregating industry sectors may yield embodied energy values 3.5
times the values calculated after disaggregation. They found a
29e45% decrease in embodied energy values due to sectoral
disaggregation, which is significant. Later, in a similar study,
Acquaye [16] found that not using disaggregation may overestimate
embodied energy nearly 2.6 times the value calculated after
disaggregation. In a similar study, Suh and Huppes [56] proposed a
disaggregation model to disaggregate an industry sector into two
sectors based on Joshi [27]. Goggins et al. [57] applied the disag-
gregation constants calculated by Acquaye et al. [55] to
49
disaggregate Irish energy supply sectors. Recently, Omar et al. [16]
compared the impact of disaggregation on embodied energy at the
material level in Australia and Malaysia. In the Australian case, they
found very little change (<±1%) in embodied energy values after
disaggregation, whereas the change ranged between 5%
and 10% in the case of Malaysia possibly due to a higher aggre-
gation level of the Malaysian IO data than that in the Australian
data. In another recent study, Chang et al. [29] disaggregated the
Chinese construction sector into two sectors using the economic
output of disaggregated subsectors: (1) civil engineering projects
and (2) buildings. The buildings sub-sector was divided into resi-
dential and non-residential subgroups, which were further dis-
aggregated into thirteen subgroups representing various building
types. Their results suggested an overestimation of 15e225% in
product chain energy due to sectoral aggregation. They further
argued that the public buildings and urban residential buildings
embedded nearly 87e94% more product chain energy than rural
residential buildings. A number of IO databases called multi-region
IO (MRIO) accounts are also now being developed with better
resolution at regional and sectoral levels [50,51]. These databases
span across national and international boundaries and include an
extensive number of sectors. However, Lenzen et al. [51] have
argued that creating such accounts is arduous and extremely
resource and data-intensive. These accounts are also not aligned
properly with other environmental databases, making them diffi-
cult to use for energy and environmental assessment. Barret et al.
[58] discussed how using “incoherent” large datasets from different
regions to develop MRIO databases could add considerable uncer-
tainty to the model. Omar et al. [16] also warned about the multi-
plied uncertainties in MRIO data due to different parameters, such
as data sources, aggregation, proportionality and homogeneity as-
sumptions, data representativeness, multipliers, etc. These pa-
rameters may cause different levels of uncertainty in data across
countries, which may eventually be transferred to the MRIO data-
base. In spite of these issues, some studies (e.g. Ref. [59]) have
applied MRIO-based models to embodied energy analysis.
Although studies have emphasized a need to disaggregate IO
industry sectors to quantify product-specific embodied energy,
they also recognized the data-intensive nature of the disaggrega-
tion process. A consensus emerges on applying sectoral disaggre-
gation to IOH methods to improve their specificity so that complete
and product-specific embodied energy can be calculated in a reli-
able manner.
3. Research goal and objectives
Although an IOH hybrid method is considered the most com-
plete approach currently available, there remains the potential for
further improvement, particularly in specificity. The main goal of
this paper is to improve the specificity of an IOH method by using
the sectoral disaggregation approach, and calculate the life cycle
embodied energy of commonly used building materials. The
calculated values are compared with those calculated with IO-
based and process-based analyses. We also quantify and discuss
the difference in calculated embodied energy values before and
after disaggregation. This study builds upon a previous study [6],
which improved the completeness of an IOH method. We used the
IOH model from that study and applied sectoral disaggregation to
demonstrate quantifying product-specific embodied energy. The
term “improved IOH” denotes the IOH model with disaggregated
sectors. Note that the IOH model developed by Dixit et al. [6] did
not integrate process data of embodied energy sourced from
manufacturers or other databases such as Econinvent or Athena.
They collected, refined, and inserted the energy consumption data
of each industry sector (in energy units) into the IO model. In the
50 M.K. Dixit / Energy 124 (2017) 46e58

context of this paper, the IOH model, therefore, should be inter-
preted as a hybrid of a statistical analysis and an IO analysis. The
research goal is accomplished by achieving the following
objectives:
 Investigate the United States’ economic input-output accounts
to develop a disaggregation framework.
 Improve the specificity of IOH-based embodied energy calcu-
lation by applying the process of sectoral disaggregation to the
IOH model proposed in Dixit et al. [6].
 Calculate and evaluate the IOH-based embodied energy values
of building materials under study using the improved IOH
model.
4. Research methods
This study utilizes the IOH model developed by Dixit et al. [6]
based on the approach proposed by Carter et al. [31]. The IOH
model was built by inserting energy use data directly into the IO
model, thus circumventing the use of unreliable energy prices, a
major methodological issue with IO-based methods. Additionally,
Dixit et al. [6] improved the system boundary completeness by
computing human and capital energy and integrating them into an
IOH model. More details on these improvements in reliability and
system completeness can be found in Dixit et al. [6]. In this
research, the IOH model proposed by Dixit et al. [6] is further
improved using sector disaggregation to compute material-specific
embodied energy. A total of six industry sectors representing 10
building materials under study were disaggregated to quantify
their embodied energy.
redefinition, the secondary products have been reallocated to the
industry sectors in which they were primary. This was done
through a process of redefinition. A detailed breakup of sectoral
output is regularly published by the United States Bureau of Eco-
nomic Analysis (USBEA) as a part of benchmark accounts. Table 2
illustrates an example of a detailed output of sector veneer and
plywood manufacturing provided by the USBEA. It can be seen that
each item produced as part of the sector’s total output was listed
along with its share in the output. Using this data, the share of
hardwood and softwood products can be extracted. Similar
breakup of total output was provided for most other sectors, which
also can be used to disaggregate industry sectors.
4.1.2. Detailed sectoral inputs
Monetary flows among various industry sectors included goods
and services. Any raw material consumed by an industry was
covered under the goods’ flow. For instance, glue and adhesives
consumed by sawmills was a flow of goods. To decompose the total
inputs used by an industry sector, the proportion of both goods and
services in total input was required. The United States Census Bu-
reau (USCB) publishes a series entitled the Materials Summary as a
part of its five-year Economic Census [60]. This summary provided
some detailed input data under the category of materials consumed
by kind. These data were specific to a subsector of the main
aggregated sector. Under the 2002 Economic Census, the USCB
published a data series entitled the Industry Series for each sector
and subsector [61]. Under Detailed Statistics by Industry, the in-
dustry series report listed various expenses such as material
consumed, fuel consumed, services used, etc. A set of disaggrega-
tion coefficients was developed for decomposing the total inputs of
energy and nonenergy goods and services.

4.1. Industry sector disaggregation 4.2. Applying disaggregation to economic data

For developing the improved IO model, the 2002 U.S. Bench-
mark IO accounts were used which included 426 industries pro-
ducing 430 commodities. Table 1 lists the industry sectors
disaggregated into two or more sectors to compute material-
specific embodied energy. Industry sectors such as iron and steel
mills and ferroalloy manufacturing were disaggregated into two
sectors. The first sector was iron and steel mills that produced one of
the study materials. The second sector represented all other
products manufactured by the aggregated sector. The following two
types of data were required for disaggregating an aggregated in-
dustry sector.
4.1.1. Detailed sectoral output
The output of an industry sector comprises primary and sec-
ondary products. Because we used input-output tables after
For calculating the embodied energy, a commodity-by-
commodity square matrix (430
430) of direct requirements
was used. However, the disaggregated inputs were given by in-
dustry and therefore could not be inserted directly into the direct
requirement matrix representing commodities. For the purpose of
obtaining a disaggregated commodity-by-commodity square ma-
trix, a few adjustments were made to both the Use Table and the
Make Table.
4.2.1. Use table modifications
The Use Table, a 430
426 commodity-by-industry matrix, lis-
ted the consumption of the commodities (goods and services) by
industry sectors. The row values represented the commodities used
by the industry at the top of the column. To disaggregate the in-
dustry column, a disaggregation coefficient for consumed materials

Table 1
List of disaggregated industry sectors.

Study material Aggregated sector (n) Disaggregated sector (n1)

NAICS Sector NAICS Sector

Virgin steel 331110 Iron and steel mills and ferroalloy manufacturing 331111 Iron and steel mills
Primary aluminum 33131A Alumina refining and primary aluminum production 331312 Primary aluminum production
Bricks & other clay products 32712A Brick, tile, and other structural clay product manufacturing 327121 Brick and structural clay tile
Ceramic wall & floor tiles 32712A Brick, tile, and other structural clay product manufacturing 327122 Ceramic wall and floor tile
Vitrified Sewer Pipes 32712A Brick, tile, and other structural clay product manufacturing 327123 Vitrified clay sewer pipe and fittings
Gypsum 3274A0 Lime and gypsum product manufacturing 327420 Gypsum product manufacturing
Lime 3274A0 Lime and gypsum product manufacturing 327410 Lime manufacturing
Wood, lumber 321100 Sawmills and wood preservation 321113 Sawmills
Plywood & veneer, hardwood 32121A Veneer and plywood manufacturing 321211 Hardwood veneer and plywood manufacturing
Plywood & veneer, softwood 32121A Veneer and plywood manufacturing 321212 Softwood veneer and plywood manufacturing
 M.K. Dixit / Energy 124 (2017) 46e58 51

Table 2
A sample detailed item output for veneer and plywood manufacturing.

2002 Commodity code Commodity Item code Item description Item output

32121A Veneer and plywood manufacturing 3212111 Hardwood veneer 1121.2

32121A Veneer and plywood manufacturing 3212113 Hardwood plywood, except prefinished hardwood plywood 1049.9
32121A Veneer and plywood manufacturing 3212115 Prefinished hardwood plywood made from purchased hardwood plywood 132.0
32121A Veneer and plywood manufacturing 3212117 Hardwood plywood type products 476.0
32121A Veneer and plywood manufacturing 321211AO Hardwood veneer and plywood manufacturing, other miscellaneous receipts 9.4
32121A Veneer and plywood manufacturing 321211CW Hardwood veneer and plywood manufacturing, contract work 25.0
32121A Veneer and plywood manufacturing 321211IC Hardwood veneer and plywood manufacturing, inventory change 18.2
32121A Veneer and plywood manufacturing 321211RSL Hardwood veneer and plywood manufacturing, value of resale 0.1
32121A Veneer and plywood manufacturing 321211W Hardwood veneer and plywood manufacturing, nska 272.1
32121A Veneer and plywood manufacturing 3212121 Softwood veneer 867.2
32121A Veneer and plywood manufacturing 321212AO Softwood veneer and plywood manufacturing, other miscellaneous receipts 26.3
32121A Veneer and plywood manufacturing 321212CW Softwood veneer and plywood manufacturing, contract work 7.6
32121A Veneer and plywood manufacturing 321212IC Softwood veneer and plywood manufacturing, inventory change 12.7
32121A Veneer and plywood manufacturing 321212RSL Softwood veneer and plywood manufacturing, value of resale 0.2
32121A Veneer and plywood manufacturing 321212W Softwood veneer and plywood manufacturing, nska 64.4
32121A Veneer and plywood manufacturing 321212X Softwood plywood products: rough, sanded, and specialties 3581.3
a
nsk indicates not specified by kind.

and services was used. An existing industry sector (p in column)
was broken down into two columns (p1 and p2). The first column
vector (p1) depicted the disaggregated sector producing the study
material, whereas the second column (p2) represented all other
materials. The first column showed the inputs required to produce
the study material, whereas the second column contained the
remaining inputs consumed in manufacturing all other products. To
disaggregate commodity inputs in the Use Table, we calculated and
used a disaggregation coefficient (k) as per Equations (1) and (2).
The coefficient k represented the proportion of inputs to study
material in the total inputs of the original industry sector.
Two types of input data were sourced to compute the values of k
as mentioned in Detailed Sectoral Inputs section earlier: (1) specific
inputs from an industry sector and (2) aggregated inputs under the
categories of materials, electricity, fuel, etc. In some cases, specific
inputs of materials, services, and fuels to the production of a spe-
cific commodity were available, which were used to compute the
value of k. For instance, 38% and 62% of inputs from Stone Mining
and Quarrying sector went to the production of lime and gypsum
commodities, respectively. These percentages became the values of
k for the inputs from the Stone Mining and Quarrying sector.
However, such specific input data was not available for other input
supplying sectors. To calculate the values of k for such industry
sectors, we sourced and used aggregated proportions of inputs
under various categories as shown in Table 3. For instance, the
values of k for disaggregating electricity consumption were 0.35
and 0.65 for lime and gypsum product, respectively. Similarly, all
other material inputs for which no specific data were available were
disaggregated using k values of 0.18 and 0.82 for lime and gypsum
production, respectively. Note that there are more categories of
inputs that were used in the calculation but not presented here in
Table 3 due to space limitation. The row entries in the Use Table
contained the value of the row commodity used by the industry at
the top of the column. As the purpose was to create a decomposed
commodity-by-commodity square matrix of direct requirements, it
was important to create a row commodity to represent the dis-
aggregated industry sectors p1 and p2. In this study, we assumed
that the use of the disaggregated commodity by the column in-
dustries can be depicted by the disaggregated commodity’s share
(w) in the total output of the aggregated commodity. To disaggre-
gate commodities in the Use Table rows, we used the proportion of
the output of the study material in the total commodity output.
Si p1 ¼ Si p
k (1)
Si p2 ¼ Si p
ð1  kÞ (2)
Where Si p1 and Si p2 are the input of commodities required by in-
dustry sectors p1 and p2, respectively, and k represents the share of
p1’s inputs in the total inputs Si p of original sector p. The cells Sp1 p1,
Sp1 p2, Sp2 p1, and Sp2 p2 in the disaggregated column vectors were
calculated in the same manner using the coefficient k.
Sp1 i ¼ Sp i
w (3)

Table 3
Aggregated proportions of total inputs required from an industry sector by categories.

Commodities Proportion in total inputs required

Material Other inputs (e.g. services) Electricity Fuel

Wood Preservation 0.20 0.12 0.05 0.11

Sawmills 0.80 0.88 0.95 0.89
Hardwood Plywood & Veneer 0.36 0.72 0.21 0.25
Softwood Plywood & Veneer 0.64 0.28 0.79 0.75
Iron & Steel Mills 0.98 0.97 0.95 0.98
Ferroalloy 0.02 0.03 0.05 0.02
Lime 0.18 0.25 0.35 0.36
Gypsum Building Materials 0.82 0.75 0.65 0.64
Brick & Structural Clay Tiles 0.57 0.71 0.64 0.77
Ceramic Floor & Wall Tiles 0.36 0.29 0.31 0.17
Other Structural Clay Products 0.07 0.13 0.04 0.06
Primary Aluminum 0.90 0.64 0.99 0.62
Alumina Production 0.10 0.36 0.01 0.38
52 M.K. Dixit / Energy 124 (2017) 46e58
Sp2 i ¼ Sp i
ð1  wÞ
Where Sp1 i and Sp2 i are the purchases of commodities p1 and p2
required by industry sector i, respectively, and w represents the
share of p1’s output in the total outputs of the original sector p.
Fig. 2 demonstrates the changes made to the Use and Make Tables.
4.2.2. Make table modifications
The Make Table is a 426
430 industry-by-commodity matrix.
The row entries in the Make Table represented industries producing
commodities listed in the columns. The row representing the
aggregated industry sector p was divided into two rows. The first
row defined the disaggregated industry and its make of column
commodities. It was assumed that the share (w) of the dis-
aggregated industry in the output of the aggregated industry would
represent its proportion in making the column commodities.
Therefore, the amount of a commodity produced by the aggregated
sector was multiplied by w to compute the commodities made by
the disaggregated industry sectors p1 and p2. The column vectors of
the Make Table defined the amount of a commodity manufactured
by each row industry. The sum of a column is always 1.0, as each cell
in the column is a fraction of commodity produced by each row. It
was considered that the disaggregated commodity was produced in
the same proportion as the original aggregated commodity. Fig. 2
illustrates the Make Table modifications. The following assump-
tion were made:
Mi p1 ¼ Mi p
Mi p2 ¼ Mi p
Where Mi p1 and Mi p2 are the proportions of disaggregated com-
modities p1 and p2 produced by industry sector i, respectively. The
cells Mp1 p1, and Mp2 p2 are assumed to be equal to Mp p. In other
words, the proportion of the original commodity p produced by the
original industry sector p is assumed to represent the production of
disaggregated commodities p1 and p2 produced by the dis-
aggregated industry sectors p1 and p2. The cells Mp2 p1, and Mp1 p2 in
the disaggregated Make Table were kept at zero, meaning that the
disaggregated industry sectors p1 and p2 manufactured none of the
disaggregated commodities p2 and p1, respectively.
4.3. IO-based and IO-based hybrid (IOH) energy calculations
The embodied energy of construction materials was calculated
using both the IO-based and the improved IOH approaches to
compare, identify, and discuss gaps in embodied energy values.
First, using the direct and total requirement coefficients of the in-
dustry sectors under study, the IO-based embodied energy was
Fig. 2. Make and Use Table modifications for sectoral disaggregation.
calculated. Because energy prices vary across industrial establish-
(4) ments, energy tariffs were calculated by dividing the energy com-
modity output in monetary units by the output in physical units.
Using the calculated energy tariffs, the energy requirements were
converted from monetary to energy units in the IO-based analysis.
In the case of improved IOH analysis, actual energy use values were
inserted into the Use Table as energy source commodity in energy
units. More details about the basic IOH model can be found in Dixit
et al. [6]. The calculated direct requirements of energy sources were
in energy units per unit of output (e.g. MJ/$). The total requirements
were computed for 12 upstream stages using the Power Series
Approximation (PSA). The PSA method was used in order to
quantify and analyze indirect energy for each upstream stage. The
indirect energy was also computed using the Leontief’s inverse
matrix to verify the fact that calculations up to stage 12 covered
most of the total indirect energy. Stage 0 represented the direct
energy intensity of the industry sector. The indirect energy distri-
bution was calculated for each stage using the approach suggested
by Treloar [25]. The indirect energy intensity of sector “n” for stage
“k” was computed as (based on [6]):
X
N
εkn ¼ Xin
εiðk1Þ (7)
i¼1
Where, Xin and εi (k-1) form the direct requirement matrix for the
sector “n” for all inputs from a sector “i” and direct energy intensity
of sector “i” for one stage earlier. The value of the number of stages
(5) “k” is between 1 and ∞. Direct energy intensity was calculated
using the PEFs as (based on [6]):
(6) X
J
εi ¼ Dji
Pj (8)
j¼1
Where “j” represents an energy providing sector whose value is
between 1 and 5 because there are five energy providing sectors in
the U.S. economy. Terms Dji and Pj are the requirements of direct
energy and the PEF for energy source produced by sector “j,”
respectively. As described in detail in Dixit et al. [6], the calculation
of PEF was performed at the national level without considering the
state-level differences of fuel mix used. The PEF values, therefore,
represent the national average factors for delivered-to-primary
energy conversion.
5. Limitations of the study
The developed IOH model and the process of industry sector
disaggregation are based on a set of assumptions, which must be
discussed for transparency. As discussed in the Research Methods
section, the detailed and specific data of energy and nonenergy
inputs to industry sectors were not available for all disaggregated
sectors. For the sectors without detailed input data, all energy and
nonenergy inputs were assumed to be represented by the aggre-
gated inputs under various material categories as shown in Table 3.
For instance, all material inputs to Sawmills and Wood Preservation
sectors are assumed to be 80% and 20% of the total material inputs,
respectively. These inputs, however, may be consumed in different
proportions for different raw materials in reality. In order to modify
the Make Tables, the proportion of an aggregated commodity in the
total output of the industry sector was assumed to remain the same
for the disaggregated commodities, which may not be the case in
reality. The availability and quality of these input data are also
worth discussing. The quality of detailed and aggregated input data
used in this study was fairly robust as the data was sourced from
the United States’ Census Bureau. However, such data may not be
 M.K. Dixit / Energy 124 (2017) 46e58 53

available in the future due to the discontinuation of some of the
data series. The timings of data reporting also put some limitations.
For instance, the United States’ Department of Energy reports
manufacturing energy survey data in a 4-year cycle (1998, 2002,
2006, 2010, etc.), which does not match with the 5-year frequency
of publishing the benchmark IO data (1997, 2002, 2007, 2012).
Fortunately, the reporting cycles for this study matched for both the
types of data, which could be a serious issue for the earlier and later
reporting cycles. In such cases, interpolating energy consumption
data between reporting cycles would necessitate more assump-
tions. The IO data, energy usage data, and primary energy factors
referred in this study are the national averages, which may be
different in different states. For instance, each state has a different
energy mix for electricity generation, which may differ from the
national average energy mix. Similarly, IO data may also be
different at the state and national levels, which may introduce
some level of uncertainty in the model. Because most of the as-
sumptions are made due to data unavailability, quantifying any
resulting uncertainty in the IOH model becomes difficult. Note that
the results of the improved IOH method may still carry some of the
limitations of an IO framework due to the assumptions of propor-
tionality and homogeneity. In addition, energy or non-energy ma-
terial inputs for which no monetary transaction occurred may not
show up in IO accounts and thus, remain excluded from the
calculation.
6. Findings
Table 4 provides the values of embodied energy calculated for
each of the study materials. The values are listed for both the IO-
based and improved IOH calculation methods. Note that the
improved IOH method denotes an enhancement in the specificity of
the IOH model proposed by Dixit et al. [6] by disaggregating some
of its industry sectors (shaded cells in Table 4). The values shown in
the second column are the results of a conventional IO-based
approach. Column 3 lists embodied energy values calculated us-
ing the IOH model integrated with human and capital energy but
without sectoral disaggregation. Column 4 shows the final results
of this study (in bold) after disaggregating the six industry sectors.
The values in column 4, therefore, come from a hybrid method with
improved completeness and specificity. Only the six sectors pro-
ducing materials such as wood, plywood, brick, clay tiles, vitrified
clay pipes, gypsum building products, lime, steel, and primary
aluminum were disaggregated. Industry sectors producing carpet,
glass, cement, stone, and adhesive did not need disaggregation due
to their production focusing mainly on the products under study.
Column 5 contains the % difference between the values quan-
tified using the improved IOH and a conventional IO-based method.
Column 6 reports the % difference obtained from sectoral disag-
gregation from the IOH model of Dixit et al. [6]. There is negligible
difference between the IO-based and improved IOH values of glass,
virgin steel, and carpet (1e3%). An underestimation of 3%e181%
and an overestimation of 1%e59% is found in embodied energy of
study materials if calculated with the conventional IO-based
method. Column 6 compares IOH embodied energy values before
and after disaggregation. Although the embodied energy of mate-
rials such as lime, gypsum, hardwood plywood and veneer, and clay
wall and floor tiles changed significantly after sector disaggrega-
tion, there is little difference (<5%) in the cases of virgin steel,
primary aluminum, and wood lumber. Over 25% overestimation is
found in the embodied energy of hardwood plywood and veneer,
clay wall and floor tiles, and gypsum building products. The
embodied energy of materials such as lime may remain grossly
underestimated (68%) if the relevant industry sector is not
disaggregated.
Table 5 lists improved IOH embodied energy values broken

Table 4
Embodied energy values from the IO model and improved IOH model after disaggregation.

Study material A: IO-based B: IOH analysis without sector C: Improved IOH analysis with sector Change from IO-based Change due to sector
analysis (MJ/kg) disaggregation (MJ/kg) disaggregation (MJ/kg) analysis (C-A)/A disaggregation (C-B)/B

Carpet (3/800 Thk.), 547.2 563.1 563.1 3% Underestimation N/A

Level Loop
Wood Lumber 5.1 6.3 6.5 28% Underestimation 4% Underestimation
Hardwood Plywood 26.8 35.3 24.6 8% Overestimation 30% Overestimation
& Veneer
Softwood Plywood & 7.0 9.2 9.2 31% Underestimation 1% Overestimation
Veneer
Paints & Coatings 67.4 56.1 56.1 17% Overestimation N/A
Adhesives 130.6 53.5 53.5 59% Overestimation N/A
Plastic Pipes & 98.2 113.4 113.4 15% Underestimation N/A
Fittings
Polystyrene Foam 243.9 256.1 256.1 5% Underestimation N/A
Insulation
Bricks 4.8 3.9 4.5 7% Overestimation 16% Underestimation
Clay Wall & Floor 44.2 35.4 26.4 40% Overestimation 25% Overestimation
Tiles (1/400 Thk.)
Vitrified Clay Sewer 19.5 15.6 14.3 27% Overestimation 9% Overestimation
Pipes
Flat Glass 24.7 24.7 24.7 Negligible difference N/A
Cement 4.4 7.5 7.5 69% Underestimation N/A
Concrete 1.1 1.3 1.3 26% Underestimation N/A
Gypsum, Bldg. 21.1 24.1 17.4 17% Overestimation 28% Overestimation
Products
Lime 3.9 4.5 7.4 90% Underestimation 65% Underestimation
Stone 3.0 3.3 3.3 9% Underestimation N/A
Mineral Wool 27.5 29.3 29.3 7% Underestimation N/A
Insulation
Virgin Steel 24.2 24.2 24.0 1% Overestimation 1% Overestimation
Primary Aluminum 67.9 186.5 190.6 181% Underestimation 2% Underestimation
Copper 43.6 59.9 59.9 37% Underestimation N/A
54 M.K. Dixit / Energy 124 (2017) 46e58

Table 5
Break-down of total embodied energy by energy sources based on the improved IOH method.

Study material Embodied energy from improved IOH method (% of total)

Oil & Gas Coal Electricity Natural gas Petroleum Human energy Capital energy
00
Carpet (3/8 Thk), Level Loop 1.6 4.2 36.3 23.1 29.0 3.6 2.2
Wood Lumber 1.4 0.8 28.4 10.1 49.4 5.7 4.3
Hardwood Plywood & Veneer 1.7 1.0 32.9 14.6 39.2 6.9 3.8
Softwood Plywood & Veneer 0.7 0.8 40.7 16.7 32.5 5.4 3.3
Paints & Coatings 2.8 4.5 23.9 21.7 41.9 2.7 2.6
Adhesives 2.5 4.7 26.0 21.5 39.4 3.1 2.8
Plastic Pipes & Fittings 3.3 2.1 23.3 20.1 47.3 1.8 2.0
Polystyrene Foam Insulation 2.6 3.1 24.3 24.6 40.5 2.6 2.3
Bricks 0.2 2.0 20.4 54.9 17.3 2.9 2.2
Clay Wall & Floor Tiles (1/400 Thk) 0.5 2.0 31.4 40.3 17.2 5.0 3.6
Vitrified Clay Sewer Pipes 0.4 2.2 18.3 52.3 19.2 4.6 3.2
Flat Glass 0.3 1.2 28.4 55.9 11.1 1.5 1.7
Cement 0.3 38.3 28.1 6.0 24.4 0.7 2.3
Concrete 0.0 21.3 24.3 14.7 33.1 2.9 3.7
Gypsum, Bldg. Products 0.3 23.7 20.2 28.1 25.4 1.3 1.0
Lime 0.3 23.9 17.1 27.3 27.5 1.9 2.0
Stone 0.6 3.9 33.6 17.1 30.3 9.3 5.1
Mineral Wool Insulation 0.5 3.8 39.0 35.5 15.5 2.5 3.1
Virgin Steel 0.2 26.6 37.7 24.9 7.9 1.5 1.2
Primary Aluminum 2.6 0.3 72.3 7.5 16.1 0.7 0.5
Copper 0.1 6.3 51.1 27.0 11.2 2.2 2.1

down by each energy source after sector disaggregation (shaded
cells represent disaggregation). Among the most electricity-
intensive materials are aluminum and copper whose electricity
use alone contributes over 72% and 51% of their total embodied
energy, respectively. A major portion of the embodied energy of
cement (38%) and virgin steel (27%) is attributed to coal con-
sumption. Due to higher heating requirements, all ceramic prod-
ucts such as glass, bricks, clay tiles, and pipes show higher natural
gas consumption (40e56% of the total embodied energy). Likewise,
due to a higher use of petroleum products, both as energy sources
and raw materials, most plastic, paint, adhesives, and polystyrene
foam products consume maximum petroleum-based energy sour-
ces (42e51% of total embodied energy). Surprisingly, nearly 50% of
the total embodied energy of wood lumbers comes from petroleum
products. The most human energy (9.3%) and capital energy (5.2%)
intensive construction material is stone in terms of human and
capital energy’s proportion of total embodied energy. Table 6 shows
the difference in fuel-specific embodied energy of study materials
after disaggregation. As a result of disaggregation, the fuel usage of
study materials increased or decreased by different magnitudes. In
the case of lime, the use of different energy sources increased by
38%e168%, whereas it decreased by 13%e45% for gypsum. Note
that both lime and gypsum are manufactured by the same industry
sector. When the industry sector producing bricks, clay tiles, and
vitrified sewer pipes was disaggregated, the energy use in all en-
ergy source categories increased in the case of bricks.
Fig. 3 illustrates the change in the direct and indirect energy
components of the ten study materials due to sectoral disaggre-
gation. The change differs for the direct and indirect energy com-
ponents. For instance, the direct embodied energy of clay wall and
floor tiles decreased by 34% after disaggregation, whereas their
indirect energy increased by 20%. Similarly, the direct energy
component of hardwood plywood and veneer increased by 9% after
disaggregation, but its indirect component dropped slightly by 2%.
In some cases, such as gypsum, both direct (20%) and indirect en-
ergy (47%) components dropped significantly after disaggregation.
On the contrary, materials such as brick and lime demonstrated an
increase in both direct and indirect embodied energy as a result of
sectoral disaggregation. As also discussed in Dixit et al. [6], mate-
rials such as concrete (87%) and carpet (83%) have considerably
higher indirect energy components than direct energy because
they are manufactured using energy-intensive raw materials such
as cement and plastic products. Likewise, materials such as paints
and coatings, adhesives, and plastic products containing petroleum
products as raw materials also show a higher indirect energy
component (82e92%). All ceramic products (clay products and
glass) manufactured with less energy-intensive natural materials,
such as clay and sand, have significantly lower indirect energy
proportions (14e21%). Similarly, materials such as cement and
lime, requiring higher direct energy consumption and less energy-
intensive materials such as limestone, demonstrate lower indirect
energy components (11e19%).

Table 6
Change in fuel-specific embodied energy after disaggregation (- indicates an increase after disaggregation).

Oil & Gas Coal Electricity Natural gas Petroleum Human energy Capital energy

Wood, Lumber 4.57% 1.71% 7.69% 0.29% 3.06% 0.56% 0.68%

HW Plywood 4.34% 40.18% 30.51% 27.28% 35.94% 6.36% 12.25%
SW Plywood 49.72% 32.74% 23.11% 18.73% 23.89% 5.05% 9.43%
Bricks 3.37% 11.42% 1.76% 22.69% 17.23% 15.70% 4.09%
Clay Tiles 8.41% 30.42% 1.03% 42.07% 24.69% 26.34% 10.57%
Vitr. Sewer Pipes 8.32% 6.90% 28.18% 8.01% 2.85% 41.14% 16.24%
Gypsum 13.00% 23.75% 21.08% 23.49% 38.04% 25.83% 44.75%
Lime 38.09% 75.54% 53.13% 69.74% 53.78% 168.00% 158.31%
Steel 0.06% 0.11% 1.78% 0.11% 0.23% 0.49% 0.69%
Aluminum 24.18% 36.85% 14.30% 16.41% 21.98% 17.41% 7.34%
 M.K. Dixit / Energy 124 (2017) 46e58 55

Fig. 3. Direct and indirect energy components after sectoral disaggregation.

7. Result evaluation and discussion
The results of the improved IOH method demonstrate a
considerable gap from conventional IO-based and process-based
analyses. When compared with the conventional IO-based re-
sults, the improved IOH results show gaps ranging from
approximately 59% to þ181%, indicating the under and over-
estimation of embodied energy. The gaps between the embodied
energy values calculated using the two methods ranged
from 77 MJ/kg to þ123 MJ/kg with average and standard devia-
tion values of 4 MJ/kg and 32 MJ/kg, respectively. The difference
was significant in the case of primary aluminum, lime, cement, and
adhesives. In the case of lime, a major difference occurred due to
sectoral disaggregation because the process of lime manufacturing
may consume more energy than gypsum manufacturing. After
disaggregation, the gap increased considerably, particularly in the
case of clay products, lime, and gypsum.
Process-based analysis may cover a wide range of system
boundaries based on data availability. For instance, it may cover
only direct inputs or it may also include some indirect inputs
[24e26]. To compare process-based results to improved IOH re-
sults, three hypothetical situations of embodied energy values were
created to represent the process data: (1) direct inputs only; (2)
direct inputs and stage 1 indirect inputs; and (3) direct inputs and
stage 2 indirect inputs. However, it is important to note that many
stage 1 and 2 energy inputs are not covered by a process-based
analysis. The comparison reveals that, even if the process-based
data covered all of the direct inputs, a 10%e99% gap remains due
to the exclusion of indirect energy. Moreover, if the process analysis
were extensive enough to include all stage 1 indirect energy inputs,

Table 7
Comparison with process-based scenarios.

Study materials Process analysis up to Stage 1 Process analysis up to Stage 2

% Indirect energy % Total energy Incompleteness (MJ/ % Indirect energy % Total energy Incompleteness (MJ/
excluded excluded kg) excluded excluded kg)

Carpet (3/800 Thk), Level Loop 62.6% 52.0% 292.7 32.3% 26.9% 151.3
Wood Lumber 56.1% 27.2% 1.8 33.4% 16.2% 1.0
Hardwood Plywood & Veneer 45.4% 27.9% 6.9 24.5% 15.0% 3.7
Softwood Plywood & Veneer 57.4% 26.4% 2.4 33.6% 15.4% 1.4
Paints & Coatings 44.6% 40.8% 22.9 18.0% 16.5% 9.3
Adhesives 45.1% 36.9% 19.7 18.3% 15.0% 8.0
Plastic Pipes & Fittings 42.0% 36.4% 41.3 15.9% 13.8% 15.7
Polystyrene Foam Insulation 45.8% 33.3% 85.3 17.2% 12.5% 32.1
Bricks 52.7% 7.5% 0.3 24.9% 3.5% 0.2
Clay Wall & Floor Tiles (1/400 52.6% 13.5% 3.6 24.6% 6.3% 1.7
Thk)
Vitrified Clay Sewer Pipes 51.4% 11.8% 1.7 24.5% 5.6% 0.8
Glass 27.4% 5.7% 1.4 11.9% 2.5% 0.6
Cement 31.2% 3.3% 0.3 12.9% 1.4% 0.1
Concrete 17.5% 15.3% 0.2 6.4% 5.6% 0.1
Gypsum, Bldg. Products 44.5% 8.8% 1.5 19.5% 3.9% 0.7
Lime 22.5% 4.3% 0.3 8.8% 1.7% 0.1
Stone 36.7% 24.4% 0.8 16.2% 10.7% 0.3
Mineral Wool Insulation 39.0% 14.3% 4.2 16.3% 6.0% 1.7
Virgin Steel 33.2% 3.3% 2.7 12.6% 1.2% 1.0
Primary Aluminum 29.1% 7.2% 13.6 9.9% 2.4% 4.6
Copper 50.3% 32.0% 19.2 24.9% 15.9% 9.5
56 M.K. Dixit / Energy 124 (2017) 46e58
the gap is between 3% and 52% (third column of Table 7). Although
impractical to assume, if all of the stage 2 inputs are covered by a
process analysis, the incompleteness still remains in the range of
1e27%. Table 7 presents the amount of incompleteness if process-
based analyses are used for embodied energy calculation (shaded
cells represent disaggregation). As also discussed in Ref. [6], the
share of human and capital energy in total embodied energy after
human and capital energy integration and sectoral disaggregation
was in the range of 1e14% across study materials.
The results of the disaggregation of some industry sectors reveal
the importance of improving the specificity of hybrid methods. For
construction materials, such as lime, gypsum, and other clay
products, larger differences surfaced after the sectoral disaggrega-
tion of relevant industry sectors. Using embodied energy values
without disaggregation may be inaccurate and misleading. For
instance, if the energy embodied in gypsum and lime is calculated
without sector disaggregation, the calculation would overestimate
the embodied energy of gypsum (by over 38%) because after
disaggregation the embodied energy of gypsum decreased from
24.2 to 17.4 MJ/kg. Similarly, the energy intensity of hardwood
plywood and veneer, and clay tiles, would be overestimated by 44%
and 34%, respectively, without disaggregation. For materials such as
brick and lime, the energy intensity would be underestimated by
14% and 40%, respectively, if their industry sectors were not dis-
aggregated. When energy intensities by energy source type were
examined, some materials showed considerable deviations. For
instance, the electricity intensity of aluminum went up by 14% after
disaggregation, while its natural gas and coal intensities decreased
by 16% and 37%, respectively. The natural gas intensities of brick
and lime increased by 23% and 70% as a result of disaggregated
sectors. Interestingly, the electricity and natural gas use increased
for softwood plywood and veneer products, but petroleum and coal
intensities went down. The human energy and capital energy of
lime production increased over 2.5 times as a result of disaggre-
gation. Note that these differences in energy intensities are per $
output, which can be significant at the building level. For instance, a
recent survey of home builders by the National Association of
Home Builders (NAHB) reported that the average cost of drywall in
an average single-family home was roughly $12,000 [62]. Consid-
ering this data and 45% construction labor (based on [63]), if the
industry sector is not disaggregated, then the energy embodied in
just gypsum drywall may be overestimated by over 127 GJ of en-
ergy per single-family home (nearly 0.5 GJ/m2). Considering other
extensively used materials such as steel, aluminum, brick, hard-
wood and softwood, and ceramic tiles, the embodied energy
calculation at the building level may contain significant errors if no
sectoral disaggregation is applied.
The results of the improved IOH model also substantiated the
assertions put forth by Treloar [25] and Miller and Blair [35] that
the indirect energy inputs calculated up to stage 12 can cover most
of the total energy inputs. Fig. 4 illustrates the 12-stage breakdown
of embodied energy by each upstream stage. The results of this
study show that stage 12 consisted of only 0.002e0.02% of total
embodied energy, indicating that over 99% of total embodied en-
ergy can be traced by covering stages 1 through 12. In fact, the
energy intensities of most industry sectors became nearly zero after
stage 8. However, it is important to compare with the total re-
quirements calculated using the Leontief’s inverse matrix before
concluding that over 99% of indirect inputs are covered up to stage
12. Since indirect energy intensities depend mainly on direct
requirement coefficients, the 12-stage breakdown differs across
industry sectors. Materials with higher indirect energy, such as
carpet and plastic products, hold considerable amounts of indirect
energy in the later upstream stages (e.g. up to stage 6). Because the
values of embodied energy are computed with a breakdown of
Fig. 4. Indirect energy associated with indirect upstream stages.
fossil fuels, these values can provide a more accurate estimate of
carbon emissions. One should be cautious, however, when using
only embodied energy values for computing the amount of carbon
emission. Materials such as cement may involve non-fossil fuel-
based carbon dioxide emissions coming from some manufacturing
processes such as calcination.
8. Conclusions
According to the results of this study, it is possible to improve
the completeness, specificity, and reliability of current IOH
embodied energy calculation methods. The impacts of these im-
provements on embodied energy values could be significant. For
instance, the results of Dixit et al. [6] showed that the energy
embodied in labor and capital inputs can be significant for some
construction materials. Similarly, differences in embodied energy
occurring after sectoral disaggregation are substantial in some
cases, a fact which cannot be overlooked. The results of this study
suggest significant under and over estimation of embodied energy
of study materials. Although the reliability of an IOH method can be
improved by circumventing the use of unreliable energy prices as
shown by Ref. [6], the use of material prices may still be required to
convert embodied energy per $ to per unit of weight or volume.
Using material prices can be avoided if reliable data of actual ma-
terial production is available in physical units. The findings also
reflect a considerable amount of incompleteness in process-based
methods, thus demonstrating that an improved IOH method
could provide complete, accurate, and material-specific embodied
energy calculations. However, the results must be used or inter-
preted with caution since the IO data used were from 2002. The
energy intensity of the manufacturing industry is decreasing with
the advent of more advanced and efficient technology. Hence, these
results may not represent current practices. However, the improved
IOH method can be applied to compute embodied energy using the
latest data. Future research should focus on further improving the
reliability of IOH methods by avoiding the use of product prices,
which are still required to compute embodied energy per unit of
mass or volume. In addition, any accuracy relating to proportion-
ality and homogeneity assumptions inherent in an IO model must
be addressed in future research. Note that the embodied energy as
calculated in this paper may only represent the environmental
impacts associated with energy consumption such as carbon
emission, pollution, global warming potential, etc. The future
 M.K. Dixit / Energy 124 (2017) 46e58
research should focus on developing IOH models to include not
only energy-related but also non-energy related environmental
impacts of construction.
References
[1] Dixit MK, Fern
andez-Solís J, Lavy S, Culp CH. Need for an embodied energy
measurement protocol for buildings: a review paper. Renew Sustain Energy
Rev 2012;16(6):3730e43.
[2] Baum M. Green building research funding: an assessment of current activity in
the United States. Washington, DC: US Green Building Council; 2007.
[3] Dixit MK, Culp CH, Ferna
ndez-Solís JL, Lavy S. Reducing carbon footprint of
facilities using a facility management approach. Facilities 2016;34(3/4):
247e59.
[4] Stephan A, Stephan L. Life cycle energy and cost analysis of embodied, oper-
ational and user-transport energy reduction measures for residential build-
ings. Appl Energy 2016;161:445e64.
[5] Crowther P. Design for disassembly to recover embodied energy. In: The 16th
annual conference on passive and low energy architecture, Melbourne-
Brisbane-Cairns, Australia; 1999.
[6] Dixit MK, Culp CH, Ferna
ndez-Solís JL. Embodied energy of construction ma-
terials: integrating human and capital energy into an IO-based hybrid model.
Environ Sci Technol 2015;49(3):1936e45.
[7] Rauf A, Crawford RH. Building service life and its effect on the life cycle
embodied energy of buildings. Energy 2015;79:140e8.
[8] M.K. Dixit, An input-output-based hybrid method for embodied energy
calculation, in: Proceedings of ASC 51st Annual International Conference held
in conjunction with CIB Workgroup 89, April 22-25, 2015, College Station, TX.
[9] Dixit MK, Culp CH, Ferna
ndez-Solís JL, Lavy S. A facility management approach
to reducing energy and carbon footprint of built facilities. In: Proceedings of
joint CIB W070, W111 and W118 international conference, Copenhagen, may
21-23; 2014.
[10] NSTC. Federal research and development agenda for net-zero energy, high-
performance green buildings. Report of the subcommittee on buildings
technology research and development. National Science and Technology
Council; 2008.
[11] M. Paleari, reportLife cycle assessment and zero energy residential buildings:
environmental assessment, economics and social remarks, [PhD dissertation,
Italy].
[12] Karimpour M, Belusko M, Xing K, Bruno F. Minimising the life cycle energy of
buildings: review and analysis. Build Environ 2014;73:106e14.
[13] Plank R. The principles of sustainable construction. IES J Part A Civ Struct Eng
2008;1(4):301e7.
[14] Davies PJ, Emmitt S, Firth SK. Challenges for capturing and assessing initial
embodied energy: a contractor’s perspective. Constr Manag Econ 2014;32(3):
290e308.
[15] Copiello S. Economic implications of the energy issue: evidence for a positive
non-linear relation between embodied energy and construction cost. Energy
Build 2016;123:59e70.
[16] Omar WW, Doh JH, Panuwatwanich K. Variations in embodied energy and
carbon emission intensities of construction materials. Environ Impact Assess
Rev 2014;49:31e48.
[17] Henry AF, Elambo NG, Tah JHM, Fabrice OE, Blanche MM. Embodied energy
and CO 2 analyses of mud-brick and cement-block houses. Aims Energy
2014;2(1):18e40.
[18] Lützkendorf T, Foliente G, Balouktsi M, Wiberg AH. Net-zero buildings:
incorporating embodied impacts. Build Res Inf 2015;43(1):62e81.
[19] Praseeda KI, Reddy BV, Mani M. Embodied and operational energy of urban
residential buildings in India. Energy Build 2016;110:211e9.
[20] Stephan A, Crawford RH, De Myttenaere K. A comprehensive assessment of
the life cycle energy demand of passive houses. Appl Energy 2013;112:23e34.
[21] Robertson AB, Lam FC, Cole RJ. A comparative cradle-to-gate life cycle
assessment of mid-rise office building construction alternatives: laminated
timber or reinforced concrete. Buildings 2012;2(3):245e70.
[22] Ristima €ki M, S€
ayna€joki A, Heinonen J, Junnila S. Combining life cycle costing
and life cycle assessment for an analysis of a new residential district energy
system design. Energy 2013;63:168e79.
[23] Dixit MK. Embodied energy and cost of building materials: correlation anal-
ysis. Build Res Inf 2016:1e16.
[24] Crawford RH. Using input-output data in life cycle inventory analysis, Ph.D.
Thesis. Victoria, Australia: Deakin University; 2004.
[25] Treloar GJ. A comprehensive embodied energy analysis framework, Ph.D.
Thesis. Victoria, Australia: Deakin University; 1998.
[26] Acquaye A. A stochastic hybrid embodied energy and CO 2_eq intensity
analysis of building and construction processes in Ireland. Ph.D. Thesis.
Dublin: Dublin Institute of Technology; 2010.
[27] Joshi S. Product environmental life-cycle assessment using input-output
techniques. J Indus Ecol 1999;3(2e3):95e120.
[28] Cheng EW, Chiang YH, Tang BS. Exploring the economic impact of construc-
tion pollution by disaggregating the construction sector of the inputeoutput
table. Build Environ 2006;41(12):1940e51.
[29] Chang Y, Ries RJ, Man Q, Wang Y. Disaggregated IO LCA model for building
product chain energy quantification: a case from China. Energy Build 2014;72:
57
212e21.
[30] Bullard CW, Penner PS, Pilati DA. Net energy analysis: handbook for
combining process and input-output analysis. Resour Energy 1978;1(3):
267e313.
[31] Carter AJ, Peet NJ, Baines JT. Direct and indirect energy requirements of the
New Zealand economy. New Zealand: New Zealand Energy Research and
Development Committee; 1981.
[32] Dixit MK, Culp CH, Ferna
ndez-Solís JL. System boundary for embodied energy
in buildings: a conceptual model for definition. Renew Sustain Energy Rev
2013;21:153e64.
[33] Lenzen M. Errors in conventional and input output base life cycle inventories.
J Indus Ecol 2000;4(4):128e48.
[34] Horowitz KJ, Planting MA. Concepts and methods of the input-output ac-
counts. Washington, D.C: United States Bureau of Economic Analysis; 2009.
[35] Miller RE, Blair PD. Input-output analysis, foundations and extensions. New
York: Cambridge University Press; 2009.
[36] Dixit MK, Fern
andez-Solís J, Lavy S, Culp CH. Identification of parameters for
embodied energy measurement: a literature review. Energy Build 2010;42(8):
1238e47.
[37] Langston YL. Embodied energy modeling of individual buildings in Mel-
bourne, the inherent energy-cost relationship. Ph.D. Thesis. Victoria,
Australia: Deakin University; 2006.
[38] Dixit MK, Culp CH, Ferna
ndez-Solís JL. Calculating primary energy and carbon
emission factors for the United States’ energy sectors. RSC Adv 2014;97:
54200e16.
[39] Joshi SV. Comprehensive product life-cycle analysis using input output tech-
niques. Ph.D. Thesis. Pittsburgh, USA: Carnegie Mellon University; 1998.
[40] Khasreen MM, Banfill PFG, Menzies GF. Life cycle assessment and the envi-
ronment impact of buildings: a review. Sustainability 2009;1(3):674e701.
[41] M.K. Dixit, An input-output-based hybrid recurrent embodied energy calcu-
lation model for commercial facilities, in: CIB world building Congress 2016,
may 31-june 3 2016, Tampere.
[42] Suh S, Lenzen M, Treloar GJ, Hondo H, Horvath A, Huppes G, et al. System
boundary selection in life-cycle inventories using hybrid approaches. Environ
Sci Technol 2004;38(3):657e64.
[43] Lissens G, Rensbergen JV. BIOSCOPE: a model combining Markov-chain theory
and inputeoutput table data to estimate the CO2 content of final demand.
Environmetrics 1996;7(1):97e107.
[44] Aden N. Lifecycle assessment of Beijing-area building energy use and emis-
sions: summary findings and policy applications. Lawrence Berkeley National
Laboratory; 2010.
[45] Crawford RH, Stephan A. The significance of embodied energy in certified
passive houses. In: Proceedings of world academy of science, engineering and
technology (No. 78, p. 453). World Academy of Science, Engineering and
Technology (WASET); January 2013.
[46] Suh S, Huppes G. Methods for life cycle inventory of a product. J Clean Pro-
duction 2005;13(7):687e97.
[47] Lenzen M, Crawford R. The path exchange method for hybrid LCA. Environ Sci
Technol 2009;43(21):8251e6.
[48] Costanza R. Embodied energy and economic valuation. Science
1980;210(4475):1219e24.
[49] Lenzen M, Treloar G. Endogenising capital: a comparison of two methods.
J Appl Input-Output Anal 2004;10(12):1e11.
[50] Wiedmann T, Crawford R, Seo S, Giesekam J. The industrial ecology virtual
laboratory and its application to sustainability and environmental
engineering-the case of low carbon living. In: Sustainable Engineering Society
(SENG) 2013 conference: looking back... looking forward (p. 102). Engineers
Australia; 2013.
[51] Lenzen M, Geschke A, Wiedmann T, Lane J, Anderson N, Baynes T, et al.
Compiling and using inputeoutput frameworks through collaborative virtual
laboratories. Sci Total Environ 2014;485:241e51.
[52] Fei JCH. A fundamental theorem for the aggregation problem of input-output
analysis. Econ J Econ Soc 1956:400e12.
[53] Wolsky AM. Disaggregating input-output models. Rev Econ Stat 1984:
283e91.
[54] Duffy A. Land use planning in Irelandda life cycle energy analysis of recent
residential development in the Greater Dublin Area. Int J Life Cycle Assess
2009;14(3):268e77.
[55] Acquaye AC, Duffy AP, Basu B. Development of a construction sub-sector
embodied energy hybrid analysis. Dublin: Dublin Institute of Technology;
2008.
[56] Suh S, Huppes G. Methods in the life cycle inventory of a product. In: Hand-
book of input-output economics in industrial ecology (pp. 263e282).
Netherlands: Springer; 2009.
[57] Goggins J, Keane T, Kelly A. The assessment of embodied energy in typical
reinforced concrete building structures in Ireland. Energy Build 2010;42(5):
735e44.
[58] Barrett J, Peters G, Wiedmann T, Scott K, Lenzen M, Roelich K, et al. Con-
sumption-based GHG emission accounting: a UK case study. Clim Policy
2013;13(4):451e70.
[59] Hong J, Shen GQ, Guo S, Xue F, Zheng W. Energy use embodied in China
construction industry: a multi-regional inputeoutput analysis. Renew Sustain
Energy Rev 2016;53:1303e12.
[60] USCB. Materials summary: 2002, 2002 economic Census, manufacturing
subject series. EC02-31SM-1. Washington, D.C: United States Census Bureau;
58 M.K. Dixit / Energy 124 (2017) 46e58

2006. Retrieved from, http://www.census.gov/prod/ec02/ec0231sm1.pdf. [62] Taylor H. Cost of constructing a home. Housing Economics. National Associ-
[61] USCB. American community survey data. Washington, D.C: United States ation of Home Builders; 2015.
Census Bureau; 2002. Retrieved from, http://www2.census.gov/acs/ [63] Carliner M. New home cost components, Housing Economics. March 2003.
downloads/Core_Tables/2002/. p. 7e11.