Download as pdf or txt
Download as pdf or txt
You are on page 1of 13

Resources, Conservation and Recycling 82 (2014) 8–20

Contents lists available at ScienceDirect

Resources, Conservation and Recycling


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

Supply chain sustainability assessment of the U.S. food manufacturing


sectors: A life cycle-based frontier approach
Gokhan Egilmez a,∗ , Murat Kucukvar b , Omer Tatari b , M. Khurrum S. Bhutta c
a
Industrial and Manufacturing Engineering, North Dakota State University, Civil and Industrial Engineering, 202K, 1410 14th Avenue North, Fargo, ND
58102, United States
b
Department of Civil, Environmental and Construction Engineering, University of Central Florida, United States
c
Management Department, College of Business, Ohio University, United States

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

Article history: Due to the fact that food manufacturing is one of the major drivers of the global environmental issues,
Received 27 April 2013 there is a strong need to focus on sustainable manufacturing toward achieving long-term sustainability
Received in revised form 2 September 2013 goals in food production of the United States. In this regard, current study assessed the direct and indi-
Accepted 13 October 2013
rect environmental footprint of 33 U.S. food manufacturing sectors by using the Economic Input-Output
Life Cycle Assessment (EIO-LCA) model. Then, a non-parametric mathematical optimization tool, namely
Keywords:
Data Envelopment Analysis (DEA), is utilized to benchmark the sustainability performance of food man-
Economic input-output analysis
ufacturing sectors by using the results of the EIO-LCA model. Next, sustainability performance indices
Life cycle assessment
Data envelopment analysis
(SPIs), rankings, target improvements, and sensitivity of environmental impact indicators are presented.
Food manufacturing The average SPI score of U.S. food manufacturing sectors is found as 0.76. In addition, 19 out of 33 food
Supply chain sectors are found as inefficient where an average of 45–71% reduction is indicated for various environ-
Sustainability mental impact categories. Analysis results also indicate that supply chains of food manufacturing sectors
are heavily responsible for the impacts with over 80% shares for energy, water and carbon footprint, fish-
ery and grazing categories. Especially, animal (except poultry) slaughtering, rendering and processing
sector is found as the most dominant sector in most of the impact categories (ranked as 2nd in fishery
and forest land). Sensitivity analysis indicated that forest land footprint is found to be the most sensitive
environmental indicator on the overall sustainability performance of food manufacturing sectors.
© 2013 Elsevier B.V. All rights reserved.

1. Introduction concerns throughout the world have encouraged stakeholders to


integrate sustainable manufacturing initiatives into the decision
Manufacturing of goods has changed the quality and speed making processes in several industrial and government projects
of daily life tremendously since the industrial revolution. Even (O’Connor and Spash, 1999). However, there is still more work to
though, 21st century humanity is benefiting from the advantages be done in terms of industry specific overview and policy making.
of living in a high-tech environment, this payoff comes with severe As the major contributor to the overall environmental impacts
environmental deterioration in many impact domains such as (e.g. greenhouse gas (GHG) emissions), food manufacturing could
air and water pollution and toxic and hazardous waste. Since, be performed in a more sustainable way to achieve a reduction of
the manufacturing activities are one of the major drivers of the the overall environmental impact. Indeed, food manufacturing is
entire situation; there is a direct need to focus on manufacturing an integral sector of U.S. economy with a 10% share of total GDP
initiatives that emphasize sustainability. In this regard, sustainable when the production, transportation and export are also taken into
manufacturing can be understood as “the creation of manufac- consideration (Antle, 2009). Despite contributing to U.S. economy
tured products that use processes that are non-polluting, conserve from several dimensions, food manufacturing has considerable
energy and natural resources, and are economically sound environmental impacts. For instance, in terms of water consump-
and safe for employees, communities and consumers” (Dept. tion, irrigation activities within food manufacturing industry
of Commerce, 2012). In recent decades, rising environmental account for 33.6% of nation’s water withdrawals (Blackhurst et al.,
2010). In terms of per capita greenhouse gas (GHG) emissions, food
manufacturing industry has slightly lower environmental impact
∗ Corresponding author. Tel.: +1 701 231 7286; fax: +1 701 231 7195. with a 8% share of total emissions (Hertwich and Peters, 2009),
E-mail addresses: gokhan.egilmez@ndsu.edu, gokhanegilmez@gmail.com which would increase the impact contribution after the population
(G. Egilmez). impact is also included compared to other sectors. In this regard,

0921-3449/$ – see front matter © 2013 Elsevier B.V. All rights reserved.
http://dx.doi.org/10.1016/j.resconrec.2013.10.008
G. Egilmez et al. / Resources, Conservation and Recycling 82 (2014) 8–20 9

food transportation has a considerable impact on air pollution and processes with details while considering the entire supply chain,
overall climate change. simultaneously (Acquaye et al., 2011). Furthermore, the hybrid LCA
In parallel with any life cycle based sustainability assess- is useful for minimizing the aggregation and uncertainty related
ment, food manufacturing sector’s sustainability assessment has to errors commonly encountered when both the P-LCA and EIO-LCA
include both onsite and supply chain impacts to be able to have the are used independently.
basic understanding about the complete picture of environmen- In terms of food manufacturing sustainability assessment, lit-
tal impacts associated with manufacturing sectors. According to a erature is abundant with works that consider certain processes or
recent survey conducted globally with 335 C-level executives from products. For example, meat (Calderón et al., 2010), non-wood fiber
world -including 95 from United States of America-, manufacturers production (González-García et al., 2010), canned food (Iribarren
believe that supply chains have to be at the center of their business et al., 2010) and (Hospido et al., 2006). While LCA models are uti-
strategies (KPMG, 2013). However, half of the executives (49% glob- lized for environmental impact assessment of industrial sectors,
ally and 54% USA) also admit that their visibility of supply chains is results usually provide insights about the selected environmen-
limited by onsite and only the 1st tier suppliers. In fact, to be able tal indicators. Even though such an assessment is crucial in terms
to cope with the environmental problems associated with manu- of quantifying the environmental burdens, it is also important
facturing practices, broader scoped assessments that consider the to compare industrial sectors from an overall sustainability per-
entire supply chain can only bring us the necessary information formance perspective. In this context, since LCA results provide
regarding the overall environmental impacts. Additionally, supply different environmental impact categories with different units of
chain of food unfortunately consists of 9000 t-km/year per house- measurement, comparing industrial sectors’ sustainability perfor-
hold indirect and 3000 t-km direct CO2 equivalent GHG emission mance in terms of environmental impacts and total production
impact (Weber and Matthews, 2008), which creates a necessity of output becomes a complicated task.
focusing on the supply chain of food from an environmental impact As a nonparametric linear programming based benchmarking
point of view along with its manufacturing impacts. While manu- model, Data Envelopment Analysis (DEA) has been successfully
facturing firms work to optimize the supply chains to minimize the applied to various environmental impact assessment studies to
travel times, the food travel in the United States raises concerns assess the overall sustainability performance (Zhou et al., 2008).
about the overall quality of available food. For instance, while the Unlike many other approaches, DEA does not require any subjec-
average travel for the final delivery of food was around 1020 miles, tive weighting procedure while benchmarking similar units (often
the average travel for the whole supply chain of food requires called “decision making units”) and generating an overall sustaina-
4200 miles of transportation, which puts a heavy burden on social bility performance score (Egilmez and McAvoy, 2013). In a typical
and environmental sustainability of nation (Weber and Matthews, DEA model, the motivation is to assess the productivity of a deci-
2008). Such results also point out the necessity of sustainable food sion making unit using a ration based on the amount of output
manufacturing and supply chain. per given input (Sarkis and Weinrach, 2001). In this case, the effi-
To be able to achieve sustainable manufacturing goals of energy ciency score can be defined as a measure of how well the inputs are
and natural resource conservation, waste and pollution mini- utilized toward producing the outputs for the preliminary defined
mization, the life cycle impacts of manufacturing processes and scope. In the context of sustainability performance assessment,
distribution activities have to be measured systematically. In this the inputs are considered as the environmental impact categories
context, Life Cycle Assessment (LCA) is a well-known and widely such as energy, carbon footprint, etc. and the output is the pro-
used method to assess the potential environmental impacts and duction of goods through manufacturing processes. Incorporating
resources used throughout the life cycle, including raw material environmental impacts and production outputs of industrial sec-
acquisition, production, use, and end-of-life phases (Finnveden tors into a single sustainability performance score is a very critical
et al., 2009). LCA was proposed in the early 1990s as practical due to increasing environmental concerns and global competition.
method to evaluate the potential environmental impacts of prod- A more detailed review about the usage of DEA in sustainability
ucts and processes, and has to be utilized to create polices for performance assessment is also given by Sarkis and Talluri (2011).
reducing the overall undesirable impacts of supply chain practices Literature is abundant with works that utilized a joint applica-
(Rebitzer et al., 2004). tion of LCA and DEA. For example, Barba-Gutiérrez et al. (2008)
Among the LCA models, process-based LCA (P-LCA), economic used a combined application of LCA and DEA to analyze the SPI
input-output LCA (EIO-LCA), and hybrid LCA are mainly used in of household electric appliances. In another study, Munksgaard
the environmental analysis of products or processes. In the P-LCA, et al. (2008) used the Economic Input-Output Life Cycle Assessment
every process that is included from the supply chain of the product (EIO-LCA) to analyze the environmental impacts of consumption,
analyzed needs to be properly inventoried. As the system bound- and developed a benchmarking DEA model for assessing the rela-
ary becomes broader, the life cycle results’ analysis becomes more tive efficiency of commodities and household types. Lozano et al.
complicated. However, with narrowly defined systems boundaries, (2009) applied DEA as a benchmarking tool to the assessment of
some important environmental impacts in the full production chain the operational efficiency of mussel cultivation in rafts along with
can be overlooked. The EIO-LCA model combines environmental LCA. Vázquez-Rowe et al. (2012) combined LCA and DEA in a hierar-
data with the economic input-output matrix of the U.S. economy chical methodology to quantify the operational efficiency of grape
to form a comprehensive system boundary. This model quanti- production. In another work, Tatari and Kucukvar (2012) used an
fies the environmental impacts of the products or processes of input-oriented DEA model to rank the SPI of different construction
direct and indirect suppliers at economy level (Hendrickson et al., materials. In their study, they used the BEES (Building for Envi-
2006). When compared to process-based LCA, EIO analysis consid- ronmental and Economic Sustainability) software to quantify the
ers interrelations between the sectors which form the structure of life cycle impacts and costs of exterior wall finishes, and used the
the nation’s economy. On the other hand, current EIO methodol- results in conjunction with the developed DEA model for SPI anal-
ogy does not allow for specific product comparisons which make ysis. As one of the recent works, Egilmez et al. (2013) used the
process assessments difficult. In order to take advantage of both hierarchical LCA and DEA approach to the sustainability assessment
the process-based LCA and EIO-LCA models and provide a more of the U.S. manufacturing sectors’ sustainability performance.
accurate and holistic LCA methodology, hybrid LCA models were The overarching goal of this paper is to develop an analytical
developed (Suh and Huppes, 2005). The combination of the EIO-LCA tool that can be used to analyze and compare the sustainability
and process-based LCA enabled the researchers to analyze specific performance of the U.S. food manufacturing sectors. In this case, a
10 G. Egilmez et al. / Resources, Conservation and Recycling 82 (2014) 8–20

joint application of two robust approaches is considered as the basis 2.2. Mathematical framework of EIO-LCA
of methodology. First, the sustainability assessment of food man-
ufacturing sectors is performed with the EIO-LCA considering the The EIO framework is employed to analyze the
entire supply chain of each sector. Then, DEA models are developed sustainability impacts of the U.S. food manufacturing
to compare the sustainability performance of U.S. food manufac- sectors from a holistic perspective. The EIO analysis
turing sectors. It is believed that the hybrid usage of EIO-LCA and is a well-known and widely used sustainability impact assessment
DEA will enable us to make sound interpretations about each sec- method, which was theorized and developed by Wassily Leontief
tor’s environmental impact contribution and later on the proposed in the 1970s, based on his earlier works of late 1930s, which were
synergistic hierarchical application will also provide significant recognized with the Nobel Prize (Leontief, 1970). The EIO-LCA
insights about individual sector’s performance with benchmarking model enables to calculate the environmental impact data with
all food manufacturing sectors among each other based on the SPI the EIO tables of the nation’s economy by using environmental
score, which is a normalized ratio of total production output to the impact multipliers and the theory of linear algebra. EIO method-
overall environmental impact. The rest of the paper is organized as ology has been used to analyze a wide range of policy issues in
follows. Section 2 introduces the mathematical structure of the EIO- environmental, economic and social areas, and several researchers
LCA model, data collection, and input-oriented DEA approach. The utilized the EIO model for analyzing the sustainability impacts
results are presented in Section 3. Finally, conclusions and future of infrastructures, energy technologies, sectors, international
work are explained in Section 4. trade, and household demand (Egilmez et al., 2013; Huang et al.,
2009; Huppes et al., 2006; Kucukvar and Tatari, 2011; Weber and
Matthews, 2007; Wiedmann et al., 2011).
2. Methodology In this paper, industry-by-industry EIO model has been utilized
(Kucukvar and Tatari, 2013). Then, a vector of total sustainability
2.1. Joint EIO-LCA and DEA approach impacts is formulated as follows:

In this paper, a two-phase hierarchical methodology is devel- r = Edir [(I − DB)−1 ] f (1)
oped to assess the environmental impacts of nation’s food In Eq. (1), r is the total impacts vector that represents total sus-
manufacturing sectors and compare them based on the per- tainability impacts per unit of final demand, and Edir represents
formance metric, “sustainability performance index (SPI)”. The a diagonal matrix, which accounts for the direct environmental
methodology basically consists of a joint application of the EIO-LCA impact values per dollar of output, I refers to the identity matrix,
and DEA. The first part quantifies the environmental footprints of and f is the total final demand vector for industries. In addition,
selected sectors and the second part compares by determining the B is the input requirements for products per unit of output of
efficiency value of each sector, which is considered as SPI. In this an industry matrix, and D is the market-share matrix. Also, the
regard, the SPI is defined as the ratio of total production output to term [(I-DB)−1 ] represents the total requirement matrix, which is
the overall environmental impact. The SPI indicates how efficient also known as the Leontief inverse and DB is the direct require-
the production of food is with regard to the environmental impacts. ment matrix, which is denoted as A matrix in the Leontief’s model
For a sector with higher SPI, higher production output is expected (Leontief, 1970). For more explanation about the integration of
to be provided with proportionally lower environmental impact. the supply and use tables into industry-by-industry input-output
To determine the SPI, DEA is employed due to its robust applicabil- model and calculation of the cumulative environmental impacts,
ity to multiple input and output data. DEA measures the efficiency please see the research paper published by Kucukvar and Tatari
by utilizing the normalized input(s) and the normalized output(s) (2013).
as a single efficiency score {output(s)/input(s)} without a need of
subjective weighting for inputs and outputs. 2.3. Data collection and mean normalization for DEA
The first phase of the hierarchical framework, LCA, consists of
two steps. First, the scope and goals are defined. The scope includes The aforementioned EIO model was used for calculating direct
the cradle-to-gate life cycle assessment of 33 major U.S. food man- and indirect environmental impacts of 33 food manufacturing sec-
ufacturing sectors based on their total production output (tons). tors, including carbon, energy footprint and water withdrawals and
In addition, the goals are evaluating each sector’s environmen- land footprint categories as fishery, grazing, forest land and crop-
tal impacts based on the selected impact categories, and finding land are then calculated by using the EIO model presented before.
the top contributing sectors based on the highest shares in the Economic input-output table supplied by the U.S. Bureau of Eco-
total environmental impacts. Then, EIO-LCA model is utilized to nomic Analysis (BEA, 2002) has used to calculate total production
determine environmental impacts of 33 U.S. food manufacturing output of each food sector. In addition, the EIO-LCA model has been
sectors including four land footprint categories: fishery (gha), graz- utilized to obtain energy, water and carbon footprint per dollar
ing (gha), forest land (gha) and cropland (gha) and three mid-point output of each U.S. sector. Global Footprint Network’s database is
indicators: water withdrawal, energy consumption and carbon also used to quantify the direct and indirect ecological land use of
footprint based on total production output of each sector. food manufacturing sectors in terms of fishery, grazing, cropland,
After developing LCA model to analyze the environmental forest land and CO2 uptake land for carbon sequestration (GFN,
impacts of each manufacturing sector based on aforementioned 2010). These environmental sustainability indicators are defined
impact categories, the results are used as input in the second phase, as follows:
“the DEA model”. And later on, the selected environmental impacts
form the inputs of DEA model. On the other hand, the output is • The carbon footprint is a measure of the total amount of car-
the amount of food products produced by each food sector. After bon dioxide (CO2 ), nitrogen oxides (N2 O), methane (CH4 ), and
integrating the results of the EIO-LCA model into DEA model and hydrofluorocarbon (HFC) emissions from fossil fuel combustion.
solving the associated linear programming models, SPI scores and Also, direct CO2 and CH4 emissions related to animal feed-
rankings, target, and percent improvement potential values of each stock production and direct CO2 emissions related to fertilizer
manufacturing sector are presented and in the final part potential applications are included with in the scope of carbon footprint
policy implications are discussed. The hierarchical methodology of calculations (CMU, 2002). In this analysis, all possible direct
the EIO-LCA + DEA is summarized in Fig. 1. and indirect emissions from food manufacturing sectors are
G. Egilmez et al. / Resources, Conservation and Recycling 82 (2014) 8–20 11

Fig. 1. Hierarchical framework of the proposed EIO-LCA + DEA methodology.

considered. All carbon footprint results are presented in terms • The forestland footprint is calculated based on the amount of lum-
of metric tons of CO2 equivalents (t CO2 -eqv.). ber, pulp, timber products, and fuel wood consumed by a country
• The water withdrawals impact category is a measure of direct on a yearly basis (GFN, 2010). The total ecological footprint of
and indirect water withdrawal by each food sector. The EIO- forest use (1.03 gha per capita) is allocated to the U.S. forestry
LCA model uses the United States Geological Survey (USGS) nurseries, forest products, and timber tracks sector.
data to estimate direct water withdrawals for each consump- • The fishery land footprint, in other words, fishing grounds foot-
tion category such as power generation, irrigation, industrial, print is calculated using estimates of the maximum sustainable
livestock and aquaculture, mining, public supply, and domes- catch for a variety of fish species. Marine areas outside continen-
tic water use. Some of these USGS categories are then allocated tal shelves are currently excluded from the ecological footprint
to different industrial sectors that are in the U.S. input-output accounts. Global footprint network used the catch data from the
table (Blackhurst et al., 2010; CMU, 2002). The polluted water UN’s Food and Agriculture Organization (FAO) which is used to
is not included within the scope of water withdrawal calcula- estimate demand on fishing grounds. Current accounts track both
tions. Power generation and supply sectors’ water withdrawal is fish catch for direct human consumption and catch for fish meal
primarily for cooling water and it is included in the water mul- (Kitzes et al., 2007). The EIO results calculate the direct and indi-
tiplier. All water withdrawals results are presented in terms of rect consumption of fishing ground based on final demand of food
kilo-gallons (kgal). manufacturing sectors in U.S. (GFN, 2010). Assigned completely
• The energy footprint of each sector is calculated by summing to the U.S. fishing sector is the total ecological footprint of fishing
the energy content of different fuel types such as coal, nat- ground (0.10 gha per capita). See Wackernagel (1995) for more
ural gas, petroleum, biomass, waste and non-fossil electricity information about the sustainability footprint indicators.
(Hendrickson et al., 2005). The consumption values of major fuels
by industrial sectors are obtained from the U.S. input-output Table 1 presents the results of LCA model for 33 food manufac-
tables (Joshi, 1999; CMU, 2002). The quantities of fuel consump- turing sectors. Since there is an imbalance in the data magnitude
tions are based on the average producer price of individual fuels due to multiple units, the mean normalization procedure has been
and are presented in terms of tera-joules (TJ). applied for the dataset obtained from the LCA model prior to
• The cropland footprint represents the most bio-productive of benchmarking. Similar normalization method was also applied in
all the land use types and includes areas used to produce food previous DEA research conducted by Talluri and Paul Yoon (2000).
and fiber for human consumption, feed for livestock, crops, and The mean normalization is simply conducted by calculating the
rubber (GFN, 2010). The National Footprint Accounts calculate mean for each input and output category and dividing each input
the cropland footprint according to the production quantities or output data element by the respective mean.
of 164 different crop categories. The total ecological footprint
of cropland use (1.08 gha per capita) is obtained from GFN 2.4. Explanation of DEA model
and proportionally allocated to the agricultural sectors of the
U.S. economy using the U.S. Census Database which shows the In this paper, DEA is utilized as the linear programming based
total harvested cropland for each agriculture sector (U.S. Census, mathematical optimization method to determine the SPI of U.S.
2012). food manufacturing sectors. Merging LCA results including six envi-
• The grazing land footprint is calculated by comparing the amount ronmental impact categories into a single efficiency measure (SPI)
of livestock feed available in a country with the amount of is the main reason of the complexity for performance assessment.
feed required for the livestock produced in that year, with To combine multiple impact categories into a single value, SPI mea-
the remainder of feed demand assumed to come from graz- surement requires assigning weights to each category to obtain
ing land (GFN, 2010). The total ecological footprint of grazing an overall score. To do so, prior studies have mostly used arbi-
use (0.14 gha per capita) is allocated to the U.S. agricultural trary equal weighting schemes or weights based on subjective
sectors. valuations (Huppes and Ishikawa, 2005). On the contrary, DEA
12
Table 1
EIO-LCA Results.

Food manufacturing sector Energy FP (TJ) Water withdrawals Carbon Fishery Grazing Forestry Cropland Total production
(kgal) FP (t) (gha) (gha) (gha) (gha) output (tons)

Dog and cat food manufacturing 1.31E + 05 1.74E + 09 1.48E + 07 7.72E + 04 4.12E + 05 3.78E + 04 8.16E + 06 9.59E + 03
Other animal food manufacturing 2.85E + 05 5.77E + 09 3.68E + 07 2.04E + 05 4.67E + 05 4.15E + 04 2.76E + 07 1.70E + 04
Flour milling and malt manufacturing 1.59E + 05 4.10E + 09 2.02E + 07 3.09E + 03 3.36E + 04 2.33E + 04 1.50E + 07 8.57E + 03
Wet corn milling 3.65E + 05 3.66E + 09 3.36E + 07 2.98E + 03 2.02E + 04 2.53E + 04 1.38E + 07 8.16E + 03

G. Egilmez et al. / Resources, Conservation and Recycling 82 (2014) 8–20


Soybean and other oilseed processing 2.61E + 05 1.20E + 09 3.38E + 07 6.50E + 03 4.06E + 04 3.72E + 04 4.21E + 07 1.33E + 04
Fats and oils refining and blending 1.33E + 05 6.78E + 08 1.53E + 07 4.81E + 03 1.10E + 05 2.28E + 04 1.50E + 07 7.02E + 03
Breakfast cereal manufacturing 9.70E + 04 9.19E + 08 8.16E + 06 1.37E + 04 1.86E + 04 5.02E + 04 5.12E + 06 8.67E + 03
Sugar cane mills and refining 1.53E + 05 1.19E + 09 9.52E + 06 1.17E + 03 7.78E + 03 1.76E + 04 6.40E + 06 4.18E + 03
Beet sugar manufacturing 8.13E + 04 8.54E + 08 5.92E + 06 5.79E + 02 5.26E + 03 1.17E + 04 4.64E + 06 2.24E + 03
Chocolate and confectionery manufacturing from cacao beans 6.77E + 04 3.47E + 08 4.40E + 06 1.22E + 03 1.06E + 04 2.51E + 04 3.22E + 06 3.86E + 03
Confectionery manufacturing from purchased chocolate 1.14E + 05 5.08E + 08 7.65E + 06 3.53E + 03 3.80E + 04 9.76E + 04 4.07E + 06 8.37E + 03
Nonchocolate confectionery manufacturing 8.40E + 04 3.21E + 08 5.80E + 06 2.63E + 03 1.15E + 04 6.71E + 04 2.35E + 06 5.70E + 03
Frozen food manufacturing 2.92E + 05 3.09E + 09 2.97E + 07 2.04E + 05 8.91E + 05 1.06E + 05 1.84E + 07 2.17E + 04
Fruit and vegetable canning, pickling, and drying 3.99E + 05 2.28E + 09 3.09E + 07 5.05E + 05 5.37E + 05 1.66E + 05 1.65E + 07 3.11E + 04
Fluid milk and butter manufacturing 3.54E + 05 1.64E + 09 5.51E + 07 2.84E + 04 4.13E + 05 9.55E + 04 1.09E + 07 2.43E + 04
Cheese manufacturing 3.23E + 05 1.71E + 09 5.29E + 07 6.13E + 04 4.44E + 05 7.69E + 04 1.06E + 07 2.10E + 04
Dry, condensed, and evaporated dairy product manufacturing 1.40E + 05 5.95E + 08 1.93E + 07 1.04E + 04 1.41E + 05 3.94E + 04 3.94E + 06 9.10E + 03
Ice cream and frozen dessert manufacturing 9.87E + 04 3.92E + 08 9.55E + 06 4.79E + 03 4.86E + 04 7.16E + 04 2.25E + 06 7.69E + 03
Animal (except poultry) slaughtering, rendering, and processing 1.28E + 06 8.86E + 09 3.36E + 08 6.37E + 05 4.83E + 07 1.91E + 05 6.08E + 07 8.22E + 04
Poultry processing 5.39E + 05 4.92E + 09 5.46E + 07 4.35E + 05 1.12E + 06 1.32E + 05 2.02E + 07 3.70E + 04
Seafood product preparation and packaging 1.19E + 05 1.56E + 08 9.96E + 06 5.91E + 06 9.79E + 04 2.63E + 04 6.89E + 05 7.98E + 03
Bread and bakery product manufacturing 3.96E + 05 2.79E + 09 3.22E + 07 1.96E + 04 1.02E + 05 1.23E + 05 1.26E + 07 3.67E + 04
Cookie, cracker, and pasta manufacturing 1.95E + 05 1.81E + 09 1.63E + 07 1.42E + 04 5.51E + 04 9.67E + 04 8.48E + 06 1.56E + 04
Tortilla manufacturing 1.87E + 04 2.05E + 08 1.68E + 06 7.02E + 02 4.22E + 03 5.63E + 03 8.23E + 05 1.46E + 03
Snack food manufacturing 2.11E + 05 1.06E + 09 1.70E + 07 1.01E + 04 7.34E + 04 1.01E + 05 1.48E + 07 1.71E + 04
Coffee and tea manufacturing 6.31E + 04 6.42E + 08 4.70E + 06 1.43E + 03 8.90E + 03 2.82E + 04 6.09E + 06 5.24E + 03
Flavoring syrup and concentrate manufacturing 4.69E + 04 2.02E + 08 3.15E + 06 6.16E + 03 9.45E + 03 9.58E + 04 1.44E + 06 8.01E + 03
Seasoning and dressing manufacturing 1.33E + 05 5.69E + 08 1.13E + 07 4.61E + 04 7.00E + 04 8.01E + 04 7.64E + 06 1.08E + 04
All other food manufacturing 1.82E + 05 1.40E + 09 1.64E + 07 1.93E + 04 1.65E + 05 8.34E + 04 1.02E + 07 1.44E + 04
Soft drink and ice manufacturing 4.57E + 05 1.08E + 09 2.99E + 07 2.45E + 04 3.57E + 04 2.94E + 05 4.26E + 06 3.22E + 04
Breweries 2.61E + 05 1.41E + 09 1.83E + 07 2.93E + 03 8.97E + 03 9.86E + 04 4.58E + 06 2.14E + 04
Wineries 8.63E + 04 9.71E + 08 5.76E + 06 1.61E + 03 6.74E + 03 2.90E + 04 4.75E + 06 9.61E + 03
Distilleries 4.47E + 04 3.03E + 08 2.95E + 06 4.47E + 02 1.27E + 03 3.01E + 04 6.50E + 05 7.58E + 03

FP: Footprint.
G. Egilmez et al. / Resources, Conservation and Recycling 82 (2014) 8–20 13

compares alternatives solely based on mathematical optimization Subject to


theory, which does not require any prior subjective intervention εj
and weight assignment. The weighting of each impact category m for j = 1...M (7)
v
i=1 i
∗ xij
is assigned by DEA through rigorous mathematical programming
that aims to compare the sustainability performance of each food
manufacturing sector with others. vi ≥0 (8)
As a result, this study integrates DEA as a benchmarking method
where εj represents the total production output of DMU (food
by aggregating undesirable environmental impacts of food man-
manufacturing sector) j , that is being analyzed. Since total pro-
ufacturing sectors into a single sustainability performance score.
duction output is the only output, output multipliers are not
In this regard, the literature is abundant with similar approaches
needed for the proposed model. Due to the fact that corre-
which utilized DEA models for sustainability performance mea-
sponding DEA model does not force any weight restrictions
surement such as Korhonen and Luptacik (2004), Kuosmanen and
on environmental impact categories, the optimal weights (vi )
Kortelainen (2005). The notation for the general input-oriented
for the impact categories are enabled to maximize the relative
DEA multiplier model proposed by Charnes et al. (1978) is as fol-
efficiency of the DMU with respect to other compared DMUs
lows.
(Kortelainen, 2008). To solve this model as a linear program,
Objective function:
it is linearized by taking the inverse of the efficiency ratio as

k follows:
max q = r yro (2) Objective function:
r=1 1 1 m
max = ∗ vi ∗ xij for j = 1...M (9)
Subject to: z εj i=1

Subject to:

m

vi xio = 1 (3)
1 
m
i=1 ∗ vi ∗ xij for j = 1...M (10)
εj
i=1


k 
m
This mathematical model is solved through linear programming
r yrj − vi xij ≤ 0 j = 1, ..., n (4) and the efficiency ratio is derived by taking the inverse of z. The lin-
r=1 i=1 earized model is run 33 times to determine the optimal efficiency
(SPI) for all food manufacturing sectors. After analyzing the SPI of
r , vi ≥0 (5) each sector, the sensitivity analysis is also conducted to under-
stand the how each input category affects the SPI score in terms of
where r is the output multiplier, vi is the input multiplier, o is magnitude of change in SPI score, which is explained by the vari-
the DMU which is being evaluated, k represents the number of ations in specific DEA input variables. For sensitivity analysis, the
outputs, m represents the number of inputs, j is the number of super-efficiency DEA model was used to conduct sensitivity analy-
decision making units, yrj is the amount of output r produced by sis (Zhu, 2001).
DMU j, and xij is the amount of input i used by DMU j. The objec-
tive function q is the weighted sum of outputs for the DMU under
3. Results and discussion
evaluation.
The proposed DEA model consists of multiple inputs and out-
3.1. Results of life cycle assessment
put, which seeks to minimize the inputs to produce the desired
output. If the output cannot be produced by the combination of
LCA results consist of 33 food manufacturing sectors’ environ-
the input of all the other DMUs (food manufacturing sectors, the
mental impact assessment results. Since the LCA results include a
DMU in consideration is deemed on the efficient frontier and the
large amount of data, graphical representation is preferred instead
efficiency value is measured as 100%. An efficiency value of 100%
of the conventional tabulated format to enhance the comparability
indicates that the corresponding sector’s SPI score is 1. In cases
of the LCA results. LCA results are shown in Fig. 2 for top ten food
where the inputs of the other DMUs produce the output of DMU
manufacturing sectors based on the highest contribution to total
in consideration, that DMU is considered as “inefficient”, which
environmental impact. In Fig. 2, the orange colored bars represent
indicates an efficiency measurement that is less than 100% so that
the individual shares as percentages and the blue lines illustrate
an SPI score of less than 1. An efficient sector, in other words a
the cumulative impact share of the top ten sectors.
sector with an SPI score of 1, indicates that its overall environ-
mental impact is proportionally less according to its production
3.1.1. Energy footprint
output compared to other sectors input-output proportionality.
In terms of energy footprint (see Fig. 2a), animal (except poul-
Similar application of sustainability scoring can also be found
try) slaughtering, rendering and processing sector is the dominant
in Korhonen and Luptacik (2004), Kuosmanen and Kortelainen
sector with 16.9% share in total energy footprint. The remaining
(2005).
sectors’ energy footprint ranges between 0.1% and 8%. And the top
In this study, an input oriented DEA multiplier model was uti-
ten sectors account for 80% of the total energy footprint.
lized, since the primary focus is to minimize the aforementioned
environmental impacts while obtaining the same level of total pro-
duction output. Following is the formulation of the DEA model 3.1.2. Water withdrawals
developed: Similar trend observed in energy footprint is also observed in
Objective function: water withdrawals category. Animal (except poultry) slaughter-
ing, rendering and processing sector are the dominant sectors with
εj  15.4% share in total energy footprint. Remaining sectors have less
max z = m for j = 1...M (6)
v
i=1 i
∗ xijj than 10% footprint and the top ten sectors account for 85% of the
total water withdrawals.
14 G. Egilmez et al. / Resources, Conservation and Recycling 82 (2014) 8–20

a b

c d

e f

Fig. 2. LCI Results of top ten sectors for the selected environmental impact categories.

3.1.3. Carbon footprint 3.1.5. Grazing


In carbon footprint category, animal (except poultry) slaugh- In grazing category, the overall ecological implication is dom-
tering, rendering and processing sector has the greatest and most inated by animal (except poultry) slaughtering, rendering and
dominant effect on the overall footprint compared to previous foot- processing sector with a share of 89.9%, which was the dominant
print categories as shown in Fig. 2c. The share of this sector is 35.2% sector in all three footprint categories (See Fig. 2e). The total share
while the total share of top ten sectors is 87.3%. It is important of first ten sectors is observed as almost 100%.
to note that in carbon footprint category, CO2 emissions (fossil and
process related), methane (CH4 ), nitrous oxide (N2 O), and hydroflu- 3.1.6. Forest land
orocarbons (HFCs) are also included. Soft drink and ice manufacturing sector has the highest share
with 12.1% in forest category, as shown in Fig. 2f. The first ten sec-
tors’ shares range between 4% and 13%, which accounts for a total
3.1.4. Fishery share of 81.4%. It is also important to note that, in forest land cate-
As one of the ecological indicators, expected results are observed gory, individual sectors’ impacts are evenly distributed in terms of
in fishery (see Fig. 2d). The seafood product preparation and pack- % share values.
aging is the dominant sector with 71.5% share. Remaining sectors
have less than 10% share. Animal (except poultry) slaughtering, ren- 3.1.7. Cropland
dering and processing sector follows the dominant sector with 7.7% As shown in Fig. 2g, the results associated with cropland cate-
contribution. And, the first ten sectors account for almost 100% of gory indicate that animal (except poultry) slaughtering, rendering
the total fishery. and processing sector is responsible for 16.5% of the overall impact
G. Egilmez et al. / Resources, Conservation and Recycling 82 (2014) 8–20 15

1.00 0.81 Wet corn milling


1.00 0.78
1.00 Beet sugar manufacturing 0.76 U.S. Average
1.00 Breweries 0.76
1.00 0.70
1.00 0.69
1.00 Wineries 0.64
1.00 Distilleries 0.59
1.00 0.56
1.00 0.54
1.00 0.53
1.00 0.50 Snack food manufacturing
1.00 Tortilla manufacturing 0.47
1.00 0.35 Frozen food manufacturing
0.89 0.33
0.87 0.28 Cheese manufacturing
0.83 0.13 Poultry processing

0.00 0.20 0.40 0.60 0.80 1.00 0.00 0.20 0.40 0.60 0.80 1.00

Fig. 3. SPI scores and ranks.

in cropland category, which is the highest share. Remaining sectors are measured with least SPI scores. For instance, animal (except
impacts range between 0.1% and 12%. The first ten sectors account poultry) slaughtering, rendering and processing is the major driver
for 85% of the total ecological impact. of carbon, energy footprint and water withdrawals, grazing and
cropland categories and its SPI is found to be 0.47.
3.2. Results of data envelopment analysis
3.2.2. Improvement potentials
3.2.1. SPI results DEA also provides quantitative guidance to inefficient sectors
DEA approach provides significant insights into the sustaina- in terms of percent improvements to reduce the environmen-
bility performance of different sectors. The SPI results provide an tal impacts while the production output is kept the same. The
understanding about how food manufacturing sectors affect the projected percent improvements for inefficient sectors are sum-
environment while producing food. In this regard, the total produc- marized in Table 2 based on each environmental and ecological
tion output (K-ton) is determined as a positive indicator, whereas impact indicator.
the selected environmental impact categories are negative indica- For example, based on DEA results, for “Poultry Processing” to
tors. In this context, “how the environment is affected?” or “how become 100% efficient, reductions of 88.7% in energy footprint,
environmentally efficient are the manufacturing sectors?” are the 89.0% in water withdrawals, 86.8% in carbon footprint, 96.4% in
main subjects of interest in this research with the focus on the U.S. fishery, 91.1% in grazing, 86.8% in each of forest land and cropland
food manufacturing sectors. Indeed, such a sustainability perfor- have to be achieved. The projected improvements for the 19 inef-
mance measurement can provide vital guidance to stakeholders ficient food manufacturing sectors are given in Table 2. Analysis
for a more effective policy making. This section presents the SPI results also indicate that food manufacturing sectors, in average,
analysis results (See Fig. 3). need to reach reduction of 54.8% in energy footprint, 54.9% in water
First of all, U.S. food manufacturing sectors are ranked based on withdrawals, 53.2% in carbon footprint, 52.3% in fishery, 48.1% in
the SPI scores as shown in Fig. 3. The results indicate that the SPI grazing, 45.9% forest land and 70.3% in cropland, which dramati-
scores range between 0.13 and 1.0. Among the sectors, 14 food man- cally points out that current food manufacturing processes cause
ufacturing sectors are found to be efficient whereas the remaining significant environmental impacts.
19 food manufacturing sectors’ sustainability performance is classi-
fied as “inefficient”, which indicates that half of the food industry’s
environmental impact is more severe compared to its benefit to the 3.2.3. Sensitivity analysis
economy. It is important to note that more than 50% of total U.S. This section provides insights about the average sensitivity
food manufacturing sectors are found to be inefficient, and require of each environmental impact category on the SPI of food man-
significant improvements in their life cycles. The average SPI of 33 ufacturing sectors. The sensitivity analysis provides an overall
food manufacturing sectors is found to be 0.76 and over 40% of the understanding about the magnitude of change in the SPI, which
sectors’ sustainability performances are found to be below the U.S. is explained by the variation in specific environmental impact
average. The most environmentally damaging sector is found to be category. The sensitivity analysis is conducted with Zhu’s super
“poultry processing” compared to its total production output with efficiency DEA model (Zhu, 2001). Higher sensitivity indicates that
an SPI of 0.13. a relatively small change in an environmental input category will
It is important to note that there is a parallelism between have a relatively higher impact on the SPI score. The analysis results
EIO-LCA and DEA results in terms of life cycle impacts and SPI indicated that the average sensitivity ratios ranged from 20.1%
scores. The sectors with the significant environmental impacts to 49.2% (See Fig. 4). Forest land has the highest (49.2%) average
16 G. Egilmez et al. / Resources, Conservation and Recycling 82 (2014) 8–20

Table 2
Improvement Potentials.

Industrial sector Energy FP Water Carbon FP Fishery Grazing Forest land Cropland
withdrawals

Poultry processing −88.7% −89.0% −86.8% −96.4% −91.1% −86.8% −86.8%


Cheese manufacturing −75.6% −71.8% −80.4% −85.7% −83.2% −71.8% −74.4%
Animal (except poultry) slaughtering, −75.2% −52.8% −90.0% −52.8% −99.0% −54.8% −63.7%
rendering, and processing
Fluid milk and butter manufacturing −68.6% −67.4% −72.5% −67.3% −73.5% −67.3% −69.5%
Frozen food manufacturing −68.3% −65.3% −65.3% −78.5% −73.3% −65.3% −72.2%
Fats and oils refining and blending −68.2% −58.2% −66.6% −46.4% −72.2% −46.4% −90.4%
Wet corn milling −60.7% −69.3% −71.2% −38.8% −19.1% −19.1% −56.4%
Seasoning and dressing manufacturing −55.8% −44.0% −44.0% −88.8% −44.0% −44.0% −77.2%
Snack food manufacturing −55.6% −60.4% −49.7% −55.5% −49.7% −49.7% −88.6%
Ice cream and frozen dessert −54.5% −30.1% −47.0% −30.1% −39.5% −62.5% −34.0%
manufacturing
Confectionery manufacturing from −51.2% −36.0% −36.0% −36.0% −50.5% −61.1% −72.2%
purchased chocolate
Nonchocolate confectionery −49.7% −16.9% −42.7% −16.9% −16.9% −42.5% −57.2%
manufacturing
Chocolate and confectionery −49.1% −24.3% −35.1% −24.3% −28.4% −29.3% −73.5%
manufacturing from cacao beans
Cookie, cracker, and pasta −48.1% −73.0% −47.4% −74.3% −47.4% −47.4% −79.2%
manufacturing
All other food manufacturing −42.9% −60.6% −22.2% −22.2% −29.6% −27.4% −66.7%
Breakfast cereal manufacturing −41.4% −52.5% −46.2% −90.7% −41.4% −41.4% −68.9%
Bread and bakery product −40.4% −68.7% −33.4% −38.8% −11.5% −11.5% −67.1%
manufacturing
Flour milling and malt manufacturing −34.6% −80.9% −60.2% −30.6% −30.6% −30.6% −71.7%
Coffee and tea manufacturing −13.4% −22.1% −14.4% −18.7% −13.4% −13.4% −65.8%

Average of Inefficient Sectors −54.8% −54.9% −53.2% −52.3% −48.1% −45.9% −70.3%

sensitivity ratio in comparison with other input categories, contrary category, which has the second largest sensitivity, should definitely
cropland is found to be the least sensitive (20.1%) input category. be considered to improve the relative SPI of this food manufacturing
Fig. 4 reveals crucial information about the relationship between sector.
average improvement potential and sensitivity. According to Fig. 4
and conducted correlation analysis results shown in Table 3, 3.3. Supply chain decomposition analysis
environmental impact categories with a lower sensitivity score
generally require higher percentage improvements to become When dealing with the policy making toward reducing envi-
100% efficient and the indicators with higher sensitivity needs rel- ronmental impacts of food manufacturing sectors, it is crucial
atively less% reduction for the inefficient sectors to become 100% to consider both onsite and supply chain shares and significant
efficient. In fact, according to the correlation results, there is a sig- contributor sectors with the average percent shares on the overall
nificant negative correlation (p = 0.014 < 0.05, Pearson’s r = −0.858) sustainability footprint for each indicator category. To have an
between improvement potential and average sensitivity of envi- overall understanding about the insight dynamics of life cycle
ronmental impact indicators (See Table 3). For example, forest land assessment, supply chain decomposition analysis is used as a
has the highest average sensitivity ratio (49.2%) compared to other robust method. Studying EIO based life cycle impacts of food man-
inputs, whereas it requires relatively lower average improvement ufacturing sectors is important from the fact that supply-chain
potential (45.9%) compared to other impact categories. and onsite impacts are both considered. Therefore, it is also crucial
Among the environmental impact categories, forest land, water to take a more detailed look into the supply chains of 33 food
and energy footprint categories are found to be responsible for the manufacturing sectors. This section reveals the results of supply
highest sensitivity. These three footprint categories indicate a high chain decomposition analysis (See Table 4) for the top ten sectors
impact on SPI with sensitivity values over 37%. In addition, ani- whose impacts are given as percentages in Fig. 2. The supply chain
mal (except poultry) slaughtering, rendering, and processing used decomposition analysis is performed considering the average per-
a high amount of water for production process. Hence, water use centage shares of the top ten sectors onsite and their supply chain

Forest Land -45.9% 49.2%


Water With drawals -54.9% 41.2%
Carbon Foot print -53.2% 37.0%
Grazing -48.1% 35.9%
Energy Foot print -54.8% 35.7%
Fishery -52.3% 32.7%
Cropland -70.3% 20.1%

-80% -60% -40% -20% 0% 20% 40% 60%


Avg. Performance Improvement Sensitivity

Fig. 4. Sensitivity analysis and average improvement potential.


G. Egilmez et al. / Resources, Conservation and Recycling 82 (2014) 8–20 17

Table 3 – In fishery category, onsite shares of top ten contributing sectors


Correlation analysis.
accounted for an average of 67.6%. The top three contributing
Correlations sectors in the supply chains are seafood product preparation and
packaging; fruit and vegetable canning, pickling, and drying; and
Sensitivity Improvement
potential poultry processing with percent shares as 39.3%, 26.6% and 22.6%,
respectively.
Sensitivity Pearson correlation 1 −.858*
Sig. (2-tailed) 0.014 – The ecological impact of grazing is also observed as heavily on
N 7 7 the supply chains of the food manufacturing sectors. The aver-
age percent share of supply chains is 93.4%. The most significant
Improvement Pearson correlation −.858* 1
potential contributing sector in the supply chains is found to be animal
Sig. (2-tailed) 0.014 (except poultry) slaughtering, rendering, and processing with an
N 7 7 average percent share of 46%. Cattle ranching and farming sector
*
Correlation is significant at the 0.05 level (2-tailed). dairy cattle and milk production are following with 33.5% and
27.3% shares, respectively.
– With regard to results of forest land category, supply chains of
food sectors significantly dominated the overall forest land with
sectors environmental impact shares. A more detailed analysis is
98.8% share. The top three sectors with the greatest average share
also performed to show the top three contributor sectors in the
in the supply chains of the food manufacturing sectors are lodging
supply chains of top ten sectors for each environmental impact
(32.2%), sawmills and wood preservation (17.7%) and paperboard
category. The analysis results are highlighted as follows.
mills (15.9%).
– In terms of cropland category, the total cropland usage is shared
– In energy footprint category, the average shares of the sectors between onsite and supply chains of the ten sectors with approx-
that are in supply chains accounted for a significant amount with imate shares of 60% and 40%, respectively. Soybean and other
82.1% share. On the other hand, onsite manufacturing energy oilseed processing sector is observed as the only significant con-
footprints are found relatively insignificant with an average of tributor in the supply chains with an average share of 22.9%.
17.9% share. The top contributing sector is found as electric power
generation, transmission and distribution (23.2% average share). The overall decomposition of LCA results by top ten sectors are
Cattle ranching and farming and wet corn milling sectors follow also illustrated in Fig. 5. The sectors in supply chains indicated over
with 19.6% and 17.2% shares. 80% share in all footprint, forest land and grazing impact categories.
– In water withdrawals category, supply chains significantly dom- On the other hand, in fishery and cropland categories, over 60% of
inate the water withdrawals with an average share of 96.8%. impacts are attributed to the industries at the end of the supply
Further analysis results indicate that grain farming accounts for chain.
an average of 76.2% share for the top ten contributing sectors sup-
ply chains. Remaining two sectors have relatively smaller percent 3.4. Discussion
shares that range between 5% and 9%.
– Similar to water and energy footprint categories, supply chains The findings of this paper reveal significant insights to pol-
are observed as the dominating part with 90.8% share compared icy makers toward improving the overall supply chain-linked
to onsite carbon footprint impacts. In specific, dairy cattle and sustainability performance of food manufacturing sectors, which
milk production sector has the greatest share with an average of will require more detailed analysis of processes and consumption
53.6% in the top ten contributing sectors supply chains. behaviors. In fact, food consumption is estimated to have 20–30%

Table 4
Supply Chain Decomposition Analysis.

Avg. Onsite Energy Footprint 17.9% Avg. Onsite Carbon Footprint 9.2%

Avg. Supply-Chain Energy Footprint 82.1% Avg. Supply-Chain Carbon Footprint 90.8%

Top Three Sectors in Supply-chains Top Three Sectors in Supply-chains


1 Electric power generation, transmission, and distribution 23.2% 1 Dairy cattle and milk production 53.6%
2 Cattle ranching and farming 19.6% 2 Oilseed farming 22.8%
3 Wet corn milling 17.2% 3 Electric power generation, transmission, and distribution 16.9%

Avg. Onsite Water Withdrawals 3.2% Avg. Onsite Fishery 67.6%


Avg. Supply-Chain Water Withdrawals 96.8% Avg. Supply-Chain Fishery 32.4%
Top Three Sectors in Supply-chains Top Sectors in Supply-chain Fishery Impact
1 Grain farming 76.2% 1 Seafood product preparation and packaging 39.3%
2 Cattle ranching and farming 8.2% 2 Fruit and vegetable canning, pickling, and drying 26.6%
3 Electric power generation, transmission, and distribution 5.4% 3 Poultry processing 22.6%

Avg. Onsite Grazing 6.6% Avg. Onsite Forest Land 1.2%


Avg. Supply-Chain Grazing 93.4% Avg. Supply-Chain – Forest Land 98.8%
Top Three Sectors in Supply-chains Top Three Sectors in Supply-chains
1 Animal (except poultry) slaughtering, rendering, and processing 46.0% 1 Logging 32.2%
2 Cattle ranching and farming 33.5% 2 Sawmills and wood preservation 17.7%
3 Dairy cattle and milk production 27.3% 3 Paperboard Mills 15.9%

Avg. Onsite Cropland 59.6%


Avg. Supply-Chain Cropland 40.4%
Top Three Sectors in Supply-chains
1 Soybean and other oilseed processing 22.9%
2 Flour milling and malt manufacturing 9.5%
3 Wet corn milling 8.9%
18 G. Egilmez et al. / Resources, Conservation and Recycling 82 (2014) 8–20

important research gap by providing a hierarchical approach that


consists of the EIO and SPI analyses.
Based on LCA findings, animal (except poultry) slaughtering,
rendering and processing sector is found to have the greatest
impact on the energy, carbon, water withdrawals and grazing
and cropland categories. The same sector is also responsible for
the second greatest impact share on the forest land and fishery
indicators. To improve the overall sustainability performance, the
processes performed in this sector has to be analyzed with a more
detailed process-based LCA to determine the potential improve-
ment areas. Additionally, a supply chain decomposition analysis is
also conducted to analyze the driver sectors in each impact category
considering onsite and supply chain components. Results indicated
Fig. 5. The overall analysis of supply chain decomposition. that supply chains of food sectors have over 90% shares in energy,
carbon and water withdrawals, forest land and grazing categories,
impact share on the overall environmental impacts of private con- whereas a more balanced pair of shares is observed in fishery and
sumption (Weidema and Wesnæs, 2008); (Tukker et al., 2009). cropland. To enable a successful policy making, the top three con-
In this regard, grain farming and animal production-related sec- tributing sectors with the greatest impact shares are also provided
tors are found to have the greatest shares in the supply chains for each environmental impact category. For instance, grain farm-
in various impact categories including carbon, water and land ing drives the water withdrawal with average percent share of
footprint. Similar findings were also provided and more detailed 73% on supply chains. According to the results of DEA model, the
assessments are made for livestock and dairy products in several average SPI score is found as 0.76, where 19 out of 33 sectors are
works such as Lesschen et al. (2011), Leip et al. (2010), Weidema found as inefficient. The sensitivity analysis indicated that forest
and Wesnæs (2008). To improve the sustainability performance land is the most sensitive input to the sustainability scores of food
of in these sectors, the food waste reduction, mobile combustion manufacturing sectors. In fact, forests are the integral parts of the
impacts associated with food shopping, agricultural improvements ecosystem’s balance and millions of people depend on forests for
such as copper and methane reducing diets for dairy cattle, biogasi- their livelihoods – directly through the consumption and sale foods
fication for water, land and carbon footprint reduction and power harvested in forests, and indirectly through forest-related employ-
saving improvements can be highlighted for specific improvement ment and income generation. Role of ecosystem services provided
areas to focus on. In this context, U.S. household consumption- by forests and trees include the integral support to agricultural pro-
focused input-output LCA studies are also required for a more duction, protection of water and soil resources, contribution to soil
holistic decision making (Kucukvar et al., 2013). development processes, including increasing soil fertility, regulat-
ing climate and providing habitat for animals. From these aspects,
4. Conclusion and recommendations it is crucial to focus on forest land sustainability as one of the key
policy areas.
In this paper, the U.S. food manufacturing sectors’ sustaina- This research provides an overall understanding about the sup-
bility performance is analyzed considering seven environmental ply chain sustainability assessment of 33 U.S. food manufacturing
impact categories and the production amounts as the output cat- sectors. The authors believe that the findings of this research will
egory. A hierarchical methodology is developed to quantify the provide an accurate guidance to stakeholders, scientists, and pol-
environmental impacts of 33 food manufacturing sectors and icy makers to improve the overall sustainability performance of
benchmark them based on the SPI score. This study introduces a food manufacturing. If the limited resources and projects budgets
new sustainability benchmarking methodology by taking various are considered, it is very important to prioritize the most prob-
environmental sustainability indicators into account and combin- lematic areas in terms of sustainability performance based on
ing them into a single sustainability performance score with a the environmental impact areas prior to taking action. However,
linear programming based benchmarking model (DEA). In this con- this research needs to be extended with process-based LCA for a
text, using DEA to compare the sustainability performance of food more detailed impact assessment and potential improvement sug-
manufacturing sectors by considering environmental impacts and gestions for the observed top contributor sectors. Especially, the
production outputs is also vital from sustainability assessment environmental implications of U.S. food manufacturing sectors are
perspective, since environment, ecosystem and economy are inte- not limited to the findings of this research. Especially, excessive use
gral elements of the sustainable development. of fertilizer for crop cultivation and animal production can have
The overreaching goal of this study is to contribute to the body substantial adverse effects on environmental quality in terms of
of knowledge on the sustainability performance analysis, repre- eutrophication and acidification of terrestrial and aquatic ecosys-
sented with SPI of 33 U.S. food manufacturing sectors. The proposed tems (Leip et al., 2011). In addition, imported products related to
sustainability performance benchmarking model provides signif- food manufacturing are assumed to be produced with domestic
icant insights for the holistic sustainability assessment of U.S. technologies. As economic input-output tables at national level
food manufacturing and offers vital guidance for decision makers, assume domestic production of imports, bringing input-output
regarding the relative SPI. LCA approach uses a set of useful sus- tables in global dimension eliminates those errors. Thus, multi-
tainability assessment metrics such as carbon, water and energy regional EIO models can be developed to link international trades
footprint, which enables us to see the overall picture about the into single region EIO models. Studies pointing out the importance
current scheme from an environmental perspective. On the other of applying multi-regional EIO models can be found in literature
hand, it would be interesting to create a sustainability perfor- (Weber and Matthews, 2008; Hetwich and Peter, 2008).
mance index (SPI) as a combination of all sustainability assessment For future research, a supply chain network focused sustaina-
metrics, which enables an overall comparison across the food man- bility performance assessment can extend the findings (especially
ufacturing sectors. In doing so, the multidimensionality can be the improvement paths) to a more detailed level. To add the time
summarized and total sustainability assessment picture can be pre- dimension to sustainability policy making framework, a dynamic
sented for all food manufacturing sectors. Hence, this paper fills this modeling approach to food manufacturing sector’s sustainability
G. Egilmez et al. / Resources, Conservation and Recycling 82 (2014) 8–20 19

policy making (similar to the system dynamics modeling applica- Journal of Industrial Ecology 2006;10(3):129–46, Retrieved from
tion to holistic transportation sustainability policy making, Egilmez http://doi.wiley.com/10.1162/jiec.2006.10.3.129
Iribarren D, Moreira MT, Feijoo G. Life Cycle Assessment of fresh and canned mussel
and Tatari, 2011) can be useful to simulate the current results processing and consumption in Galicia (NW Spain). Resources, Conservation and
and apply quantitative policy making alternatives for the long run Recycling 2010;55(2):106–17.
to stabilize the impacts. Additionally, more advanced DEA mod- Joshi S. Product environmental life-cycle assessment using input-output tech-
niques. Journal of Industrial Ecology 1999;3(2–3):95–120, Retrieved from
els such as cone ratio DEA and multi criteria decision making http://doi.wiley.com/10.1162/108819899569449
approaches can be utilized to reflect predetermined weights on the Kitzes J, Peller A, Goldfinger S, Wackernagel M. Current methods for calculating
inputs and can be compared with the results of current research. national ecological footprint accounts. Science for Environment and Sustainable
Society 2007;4(1).
Korhonen PJ, Luptacik M. Eco-efficiency analysis of power plants: An
extension of data envelopment analysis. European Journal of Opera-
Appendix A. Supplementary data tional Research 2004;154(2):437–46, Retrieved from http://ideas.repec.
org/a/eee/ejores/v154y2004i2p437-446.html
Supplementary data associated with this article can be found, Kortelainen M. Dynamic environmental performance analysis: a Malmquist
index approach. Ecological Economics 2008;64(4):701–15, Retrieved from
in the online version, at http://dx.doi.org/10.1016/j.resconrec. http://dx.doi.org/10.1016/j.ecolecon.2007.08.001
2013.10.008. KPMG. Survey: global manufacturers lack supply chain visibility beyond
tier 1; 2013, Retrieved from http://www.manufacturing.net/news/2013/
05/survey-global-manufacturers-lack-supply-chain-visibility-beyond-tier-1
References Kucukvar M, Tatari O, Egilmez G. Sustainability impact assessment of U.S. final con-
sumption and investments: a triple-bottom-line approach (Working Paper);
2013.
Acquaye AA, Wiedmann T, Feng K, Crawford RH, Barrett J, Kuylenstierna J, et al. Kucukvar M, Tatari O. A comprehensive life cycle analysis of cofiring algae in
Identification of “carbon hot-spots” and quantification of GHG intensities in the a coal power plant as a solution for achieving sustainable energy. Energy
biodiesel supply chain using hybrid LCA and structural path analysis. Environ- 2011;36(11):6352–7, http://dx.doi.org/10.1016/j.energy.2011.09.039.
mental Science and Technology 2011;45(6):2471–8.
Kucukvar M, Tatari O. Towards a triple bottom-line sustainability assessment of the
Antle JM. Agriculture and the food system adaptation to climate change; 2009. p. 24,
U.S. construction industry. International Journal of Life Cycle Assessment 2013,
Retrieved from http://www.rff.org/rff/documents/rff-rpt-adaptation-antle.pdf
Retrieved from http://link.springer.com/10.1007/s11367-013-0545-9
Barba-Gutiérrez Y, Adenso-Díaz B, Lozano S. Eco-efficiency of electric and
Kuosmanen T, Kortelainen M. Measuring eco-efficiency of production with data
electronic appliances: a Data Envelopment Analysis (DEA). Environ-
envelopment analysis. Journal of Industrial Ecology 2005;9(4):59–72, Retrieved
mental Modeling and Assessment 2008;14(4):439–47, Retrieved from
from http://doi.wiley.com/10.1162/108819805775247846
http://www.springerlink.com/content/118r003x6711v05g/
Leip A, Weiss F, Wassenaar T, Perez I, Fellmann T, Loudjani P, et al. Evaluation
Blackhurst BM, Hendrickson C, Vidal JSI. Direct and indirect water with-
of the livestock sector’s contribution to the EU greenhouse gas emis-
drawals for U.S. industrial sectors. Environmental Science and Technology
sions (GGELS) 2010, Retrieved from http://ec.europa.eu/agriculture/analysis/
2010;44(6):2126–30, http://dx.doi.org/10.1021/es903147k.
external/livestock-gas/full text en.pdf
Bureau of Economic Analysis. Benchmark input-output data; 2002, Retrieved from Leip, Adrian, Britz W, Weiss F, de Vries W. Farm, land, and soil nitrogen budgets for
http://www.bea.gov/industry/io benchmark.html agriculture in Europe calculated with CAPRI. Environmental pollution (Barking,
Calderón LA, Iglesias L, Laca A, Herrero M, Díaz M. The utility of life cycle assess- Essex: 1987) 2011;159(11):3243–53.
ment in the ready meal food industry. Resources, Conservation and Recycling Leontief W. Environmental repercussions and the economic structure: an input-
2010;54(12):1196–207. output approach. REStat 1970;52(3):262–71, Retrieved from http://www.jstor.
Carnegie Mellon University. EIO-LCA (Economic Input-Output Life Cycle Assess- org/discover/10.2307/1926294?uid=3739840&uid=2129&uid=2&uid=70&uid=
ment); 2002, Retrieved from http://www.eiolca.net/ 4&uid=3739256&sid=21101262763267
Charnes A, Cooper WW, Rhodes E. Measuring the efficiency of decision-making units. Lesschen JP, van den Berg M, Westhoek HJ, Witzke HP, Oenema O. Green-
European Journal of Operational Research 1978;3(4):338–9. house gas emission profiles of European livestock sectors. Animal
Dept. of Commerce. How does commerce define sustainable manufac- Feed Science and Technology 2011;166–167:16–28, Retrieved from
turing?; 2012, Retrieved from http://www.trade.gov/competitiveness/ http://dx.doi.org/10.1016/j.anifeedsci.2011.04.058
sustainablemanufacturing/how doc defines SM.asp Lozano S, Iribarren D, Moreira MT, Feijoo G. The link between operational
Egilmez G, Kucukvar M, Tatari O. Sustainability assessment of U.S. manufacturing efficiency and environmental impacts. A joint application of Life Cycle Assess-
sectors: an economic input output-based frontier approach. Journal of Cleaner ment and Data Envelopment Analysis. Science of The Total Environment
Production 2013;53:91–102, http://dx.doi.org/10.1016/j.jclepro.2013.03.037. 2009;407(5):1744–54, http://dx.doi.org/10.1016/j.scitotenv.2008.10.062.
Egilmez G, McAvoy D. Benchmarking road safety of U.S. states: a DEA-based Munksgaard J, Wier M, Lenzen M, Dey C. Using input-output analysis
Malmquist productivity index approach. Accident Analysis and |Prevention to measure the environmental pressure of consumption at differ-
2013;53(1):55–64. ent spatial levels. Journal of Industrial Ecology 2008;9(1–2):169–85,
Egilmez G, Tatari O. A dynamic modeling approach to highway sustaina- http://dx.doi.org/10.1162/1088198054084699.
bility: strategies to reduce overall impact. Transportation Research, Part A
O’Connor M, Spash CL. Valuation and the environment: theory, method, and practice
2011;46(7):1086–96.
(Advances in Ecological Economics). Northampton, MA: Edward Elgar Pub; 1999.
Finnveden G, Hauschild MZ, Ekvall T, Guinée J, Heijungs R, Hellweg
p. 339.
S, et al. Recent developments in life cycle assessment. Journal
Rebitzer G, Ekvall T, Frischknecht R, Hunkeler D, Norris G, Rydberg T, et al.
of Environmental Economics and Management 2009;91(1):1–21,
Life cycle assessment part 1: framework, goal and scope definition, inven-
http://dx.doi.org/10.1016/j.jenvman.2009.06.018.
tory analysis, and applications. Environment International 2004;30(5):701–20,
Global Footprint Network. GFN national footprint accounts: ecological foot- http://dx.doi.org/10.1016/j.envint.2003.11.005.
print and bio-capacity; 2010, Retrieved from http://www.footprintnetwork.
Sarkis J, Weinrach J. Using data envelopment analysis to evaluate environmen-
org/en/index.php/GFN/page/footprint for nations/ [pp. accessed 15.06.12].
tally conscious waste treatment technology. Journal of Cleaner Production
González-García S, Hospido A, Feijoo G, Moreira MT. Life cycle assessment of raw
2001;9(5):417–27, http://dx.doi.org/10.1016/S0959-6526(00)00084-6.
materials for non-wood pulp mills: Hemp and flax. Resources, Conservation and
Recycling 2010;54(11):923–30. Suh S, Huppes G. Methods for life cycle inventory of a product. Journal of Cleaner
Hendrickson CT, Lave LB, Matthews HS. Environmental life cycle assessment of goods Production 2005;13(7):687–97.
and services: an input-output approach; 2005. p. 261, Resources for the Future. Talluri S, Paul Yoon K. A cone-ratio DEA approach for AMT justification. Inter-
Hendrickson CT, Lave LB, Matthews HS. Environmental life cycle assessment of goods national Journal of Production Economics 2000;66(2):119–29, Retrieved from
and services: an input-output approach; 2006, Resources for the Future. http://ideas.repec.org/a/eee/proeco/v66y2000i2p119-129.html
Hertwich EG, Peters GP. Carbon Footprint of nations: a global, trade-linked analysis. Tatari O, Kucukvar M. Eco-efficiency of construction materials: a data envel-
Environmental Science & Technology 2009;43(16):6414–20. opment analysis. Journal of Construction Engineering and Management
Hospido A, Vazquez ME, Cuevas A, Feijoo G, Moreira MT. Environmental assessment 2012;138(6):733–41.
of canned tuna manufacture with a life-cycle perspective. Resources, Conserva- Tukker A, Bausch-Goldbohm S, Verheijden M, Koning A, Kleijn R, Wolf O, et al. Envi-
tion and Recycling 2006;47(1):56–72. ronmental impacts of diet changes in the EU Annex Report; 2009. p. 88, Retrieved
Huang YA, Lenzen M, Weber CL, Murray J, Matthews HS. The role of input–output from http://ftp.jrc.es/EURdoc/EURdoc/JRC50544 Annex1.pdf
analysis for the screening of corporate carbon footprints. Economic Systems U.S. Census. Statistical Abstract of the United States: Agriculture; 2012, Retrieved
Research 2009;21(3):217–42, http://dx.doi.org/10.1080/09535310903541348. from http://www.census.gov/prod/2011pubs/12statab/agricult.pdf
Vázquez-Rowe I, Villanueva-Rey P, Iribarren D, Teresa Moreira M, Feijoo G. Joint life
Huppes G, Ishikawa M. A framework for quantified eco-efficiency anal-
cycle assessment and data envelopment analysis of grape production for vinifi-
ysis. Journal of Industrial Ecology 2005;9(4):25–41, Retrieved from
cation in the Rías Baixas appellation (NW Spain). Journal of Cleaner Production
http://doi.wiley.com/10.1162/108819805775247882
2012;27:92–102.
Huppes G, Koning A, Suh S, Heijungs R, Oers L, Nielsen P, et al.
Wackernagel M. Our ecological footprint: reducing human impact on the Earth.
Environmental impacts of consumption in the european union:high-
Gabriola Island, B.C. Canada: New Society Publishers; 1995. p. 158.
resolution input-output tables with detailed environmental extensions.
20 G. Egilmez et al. / Resources, Conservation and Recycling 82 (2014) 8–20

Weber CL, Matthews HS. Embodied environmental emissions in U.S. Wiedmann TO, Suh S, Feng K, Lenzen M, Acquaye A, Scott K, et al. Application of
international trade, 1997–2004. Environmental Science and Technol- hybrid life cycle approaches to emerging energy technologies – the case of wind
ogy 2007;41(14):4875–81, Retrieved from http://www.ncbi.nlm.nih. power in the UK. Environmental Science and Technology 2011;45(13):5900–7,
gov/pubmed/17711196 Retrieved from http://dx.doi.org/10.1021/es2007287
Weber CL, Matthews HS. Food-miles and the relative climate impacts of Zhou P, Ang BW, Poh KL. A survey of data envelopment analysis in energy
food choices in the United States. Environmental Science and Technology and environmental studies. European Journal of Operational Research
2008;42(10):3508–13, Retrieved from http://dx.doi.org/10.1021/es702969f 2008;189(1):1–18, http://dx.doi.org/10.1016/j.ejor.2007.04.042.
Weidema BP, Wesnæs M. Environmental improvement potentials of meat and Zhu J. Super-efficiency and DEA sensitivity analysis. European Journal of Operational
dairy products, vol. 217; 2008, Retrieved from http://www.saiplatform.org/ Research 2001;129(2):443–55.
uploads/Library/EnvironmentalImprovementsPotentialsofMeatandDairy
Products.pdf

You might also like